LIN325 Introduction to Semitic Languages Chapter 3: Phonology 1 3.1.0 Semitic Consonant Inventories Twenty-nine consonant phonemes are typically reconstructed for Proto-Semitic, as follows: Table 1: Proto-Semitic Consonants Labial
Apical Dental
Stop
Fric
Lateral
Nasal
(Alveo-) Palatal
p
t
k
vd
b
d
g
emph
t2
q
vls
T
s
vd
D
z
emph
T2
s2
vls
Â
vd
l
emph
Â2 m
S
Gutteral Pharyngeal Glottal
vls
?
x
ð
Ä
À
h
n r
Rhotic Glide
Alveolar
Dorsal
w
j
3.1.1 Emphatics and Gutterals 3.1.1.1 Emphatics Most modern Semitic languages have a set of apical obstruents with a secondary articulation termed emphatic. The term emphatic is used because the actual articulation of these consonants varies from language to language. They may be velarised (with or without accompanying labialisation), pharyngealised, or glottalised (with or without a raised larynx, i.e. ejective, articulation). Velarised or pharyngealised articulations are most common for Arabic and modern Aramaic. Emphasis in modern South Arabian and in Ethiopic is usually described as ejective. In Arabic, the emphatic consonants have the effect of retracting or centralising an adjacent (particularly a tautosyllabic) vowel. Note that the uvular stop /q/ of many languages patterns with the emphatics; in some languages the corresponding consonant is just an emphatic velar stop. McCarthy (1991 and elsewhere -see below) groups the emphatics in a class with the gutterals, characterised by a feature [pharyngeal].
14/2/02
Draft Only: Do Not Quote Or Circulate Without Permission
LIN325 Introduction to Semitic Languages Chapter 3: Phonology 2 Five emphatic consonants are reconstructed for PSEM (see Table 1). Amongst modern Semitic languages, the distinction between emphatic and non-emphatic obstruents is retained, to my knowledge, in all languages except Modern Hebrew and some Arabic dialects (for example, Maltese, Cypriot, and some sub-Saharan dialects). In some languages and dialects, the contrast has in fact been extended. For example, in Amharic and Tigrinya it is a feature of all oral stops. In some modern eastern Aramaic dialects it has been extended to all consonants, largely as a result of emphasis spreading. At least in some modern Semitic languages, emphasis is a prosodic feature whose domain extends beyond a single segment. In many modern Aramaic languages, emphasis spreads over portions of the phonological word; in some eastern Aramaic languages, like the modern koiné (Hoberman 1997:333) emphasis is a feature of the phonological word. In Moroccan Arabic, all the alveolars in the same stem are either plain or emphatic, and all alveolars except /n/ have plain and emphatic forms. *?
**ð
*À
*h
*x
*Ä
Ugaritic
?
ð
À
h
x
Ä
Akkadian
?
?
?
?
x
?
Standard Arabic
?
ð
À
h
x
Ä
Iraqi Arabic
Proto-Semitic
?
ð
À
h
x
Ä
1
Maltese
Ø
ð
(Ä)
ð
ð
(Ä)
Tiberian Hebrew
?
ð
À
h
ð
À
?~Ø
x
?~Ø
h~Ø
-
-)
?
ð
À
h
ð
À
?
x
?
h
-
-
Ge’ez
?
ð
À
h
x
À
Tigrinya
?
ð
À
h
ð
-
Amharic
?~Ø
h
?~Ø
h
h
-
(Modern Hebrew Syraic Eastern Aramaic
2
Table 2: Some Semitic Gutteral Consonant Correspondences (Dead languages are in italics; standard Arabic has no native speakers, but is in other respects a living language.) 3.1.1.1 Gutterals Pharyngeal and glottal fricatives are found in most modern Semitic languages, and are reconstructible for Proto-Semitic. Modern pharyngeals Proto-Semitic pharyngeals are not always reflected directly as pharyngeals in modern languages, though. In some Modern 1
Maltese /ð/ is the result of * the merger of *ð, *x, and *h. Its phonetic realisation varies. *À is realised only as a lengthening of the formerly tautosyllabic vowel.
2
The pharyngeals survive in some words containing emphatics, and in borrowings from Arabic.
14/2/02
Draft Only: Do Not Quote Or Circulate Without Permission
LIN325 Introduction to Semitic Languages Chapter 3: Phonology 3 Eastern Aramaic languages, the historical pharyngeals have merged with velars/glottals, but pharyngeal articulations have been maintained through borrowings -- see section 3.1.3.3. Loss and merger of the pharyngeals and glottals appears to have been a feature of Semitic throughout its recorded history. In Akkadian, for example, it is usually assumed that the pharyngeals and glottals had merged on glottal stop. Some Arabic dialects preserve the Proto-Semitic contrasts, while others (to some degree at least) do not. (Koranic Arabic orthography suggests, for example, that Mohammed's own Meccan dialect had lost glottal stop.) In a series of papers3 , McCarthy proposed that the pharyngeals, laryngeals, and uvular (velar) fricatives of Semitic languages form a natural class he terms gutteral. He extends that class to include the uvular stop /q/, and other emphatic consonants (see below), and suggests that the class is characterised by articulation in the oral pharynx. McCarthy proposes the feature [pharyngeal] to capture the class. Historical mergers like those in Table 2 suggest an affinity amongst at least this set of six consonants. In addition, McCarthy points out a number of synchronic phenomena common to his gutterals in various languages. Amongst these are: a. restrictions on the gemination of gutterals in Tiberian Hebrew4 and many Ethiopic languages (and of glottals in some modern Eastern Aramaic languages -see Hoberman 1997:318). b. harmony constraints between vowels across a gutteral consonant in Tiberian Hebrew and Ge'ez c. restrictions on gutterals in internal syllable final position in Tiberian Hebrew and some Arabic dialects. (In TH, the ultra-short hatef vowels appear after gutterals, where other consonants would take (± vocalic) shwa -- see section 3.3.5.) d. both synchronic and diachronic vowel lowering in the environment of a gutteral; in Tigrinya, for example, /æ/ and /ö/ tend to lower to /a/ in the environment of a gutteral. 3.1.2 Conservative Semitic Consonant Inventories: Standard Arabic and Modern South Arabian Modern Standard Arabic has a relatively conservative consonant inventory, differing from that reconstructed for Proto-Semitic in two principal respects. First, it shows evidence of phonetic lenition of stops; PSEM *p has become a labio-dental fricative and PSEM *g, an alveopalatal affricate /dZ/. The latter becomes [Z] in some dialects, and returns to [g] in others. Affrication is extended in some dialects to reflexes of PSEM *k (realised as [tS] or 3
See particularly McCarthy (1991). A summary of the issue appears in Hoberman (1995:840-843).
4
In Tiberian Hebrew, /r/ does not occur geminate either.
14/2/02
Draft Only: Do Not Quote Or Circulate Without Permission
LIN325 Introduction to Semitic Languages Chapter 3: Phonology 4 [ts]). (Note that the dorsal emphatic stop is uvular in the standard language and in many dialects, but a velar, glottal, or alveopalatal in others.) Table 3: Modern Standard Arabic Consonants Labial
Apical Dental
vls Stop
vd
Fric
(Alveo-) Palatal
t b
emph vls
Alveolar
Dorsal
Pharyngeal Glottal
k
d
?
dZ
t2 d2 f
Gutteral
q
T
s
vd
D
z
emph
z2
s2
S
x
ð
Ä
À
h
vls Lateral
vd
l
emph Nasal
[l2] m
r [r2]
Rhotic Glide
n
w
j
Second, the inventory of emphatics is expanded for the apicals, in the fashion outlined in section 3.1.1.1 above. The standard orthography does not indicate these new emphatics, and not all are present in all dialects. Arabic /z2/ ([D2]~ [z2]) reflects PSEM *T. Arabic /d2/ reflects the Proto-Semitic emphatic lateral fricative *Â2. The only consonant merger that Standard Arabic has undergone is that of the plain lateral fricative *Â with *S to become Arabic /s/. The phonemic inventory of the Modern South Arabian is identical to that reconstructed for Proto-Semitic, except for the problematic additional phoneme, the emphatic alveopalatal fricative MSA /S2/. All South Arabian languages have this phoneme, but (as reported, for example, in Lonnet and Simeone-Senelle 1997:350) seldom agree on which etyma contain it.
14/2/02
Draft Only: Do Not Quote Or Circulate Without Permission
LIN325 Introduction to Semitic Languages Chapter 3: Phonology 5 Table 4: Modern South Arabian Consonants Labial
Apical Dental
vls Stop
vd
b
Fric
Lateral
d
g
t2
q s
vd
D
z
emph
T2
s2
vls
Â
vd
l
emph
Â2 m
S
?
x
ð
Ä
À
h
S2
n r
Rhotic Glide
Gutteral Pharyngeal Glottal
k
T
Nasal
f
(Alveo-) Palatal
t
emph vls
Alveolar
Dorsal
w
j
Though Proto-Semitic *Â/*s@ and *Â2/*d2 are reconstructible on the basis of correspondences between Syriac and Akkadian, for example, it is only the evidence of Modern South Arabian languages that suggests they were phonetic lateral fricatives. As Table 5 illustrates, in many languages, these they have fallen together with reflexes of PSEM sibilants: *s
*s2
*Â
*Â2
*S
Ugaritic
s
s2
S
s2
S
Akkadian
s
s2
S
s2
S
Standard Arabic
s
s2
S
d2
s
Tiberian Hebrew
s
s2
sÛ
s2
S
Syraic
s
s2
s
À
S
Ge’ez
s
s2
sÛ
d2
S
Epigraphic S. Arabian
s3
s2
s3
d2
s2
Modern S. Arabian
s
s2
Â
Â2
S/h
Proto-Semitic
Table 5 Some Alveolar and Lateral Fricative Correspondences 14/2/02
Draft Only: Do Not Quote Or Circulate Without Permission
LIN325 Introduction to Semitic Languages Chapter 3: Phonology 6 3.1.3 Aramaic Consonants This section considers the history of consonants in Aramaic. I choose Aramaic first because it is reasonably well-documented over a period of some three thousand years (c. 1000 BCE - present) and, second, because it exemplifies most of the types of change that have affected Semitic consonants. In particular: i. reduction in the PSEM fricative inventory ii. context-free and context-sensitive lenition of stops iii. emphasis spreading 3.1.3.1 Early Aramaic Fricatives Although Aramaic was written was with (a derivative of) the 22-letter Phoenician alphabet, changes in orthography from the Early Aramaic period (pre-7th century BCE), through the Imperial Aramaic period (7th-2nd centuries BCE) and beyond suggest that Early Aramaic distinguished at least 26 of the 29 Proto-Semitic consonants. In Early Aramaic materials, the grapheme ù shin is used in forms reflecting Proto-Semitic *T and *Â, as well as Proto-Semitic *S. By the Imperial Aramaic period, forms reflecting *T were written with ú taw, and those reflecting *Â with ñ semkat, leaving only those reflecting *S written with shin. Similarly, æ zayn was used for *z and *D, ö s2ade for *s2 and *T2, and ÷ qof for *q and *Â2. By Imperial Aramaic times , qof represents only *q, s2ade only *s2 and zayn only *z. These patterns suggest that the Proto-Semitic interdental and lateral fricatives persisted as distinct phonemes into Early Aramaic (though not written distinctly because of the nature of the Phoenician alphabet). At a later period, the interdental fricatives merged with the dental stops, and the voiced and voiceless lateral fricatives with the voiced pharyngeal and voiceless alveolar fricatives, respectively. There is no corresponding evidence for the survival of the Proto-Semitic velar fricatives into Aramaic, since their graphemic representation remains constant through the history of the language. They may have remained distinct for some time, or may have merged on the corresponding pharyngeals at an earlier period. 3.1.3.1.2 Early Aramaic Glides In all of Northwest Semitic, word-initial Proto-Semitic *w > y, with some exceptions, in Aramaic. 3.1.3.2 Lenition of Stops (Begad Kefat) At some point probably toward the end of the first millennium BCE, the non-emphatic stops {/b/,/g/, /d/; /k/, /p/, /t/} developed fricative allophones postvocalically. (In Hebrew and Aramaic grammar, this set is known by the acronym begad kefat). This pattern of allophony is reflected in the orthography of Tiberian Hebrew and of Classical Syriac-14/2/02
Draft Only: Do Not Quote Or Circulate Without Permission
LIN325 Introduction to Semitic Languages Chapter 3: Phonology 7 see section 2.9.2.1 and Morag 1962:52-53. In forms of Classical Syriac5 (and, it would appear, other forms of Middle and Late Aramaic), the fricative allophones were { [v], [Ä], [D], [x], [f], [T] }, respectively. In the Ashkenazi tradition of Tiberian Hebrew, all fricative allophones are marked orthographically, but not all are phonetically distinct from the corresponding stop. The fricative allophones are { [v], -, -, [x], [f], [s] }; that is, amongst the voiced stops, only /b/ has a distinct fricative allophone. In the Ashkenazi tradition of Tiberian Hebrew, the fricative allophones are { [v], -, -, [x], [f], [s] }. In Modern Hebrew and in Modern Aramaic, begad kefat allophony has become phonemicised. In the former, the phonemicisation is a consequence of the loss of gemination, giving rise to a contrast between fricatives and (formerly geminate) stops intervocalically. In the latter, reduction of geminates and deletion of the shwa position from lexical representations6 had the same effect. The difference between Modern Hebrew and Modern Aramaic is that, in Modern Aramaic (but not Modern Hebrew), most of the resulting stop/fricative morphophonemic alternations have been neutralised. Compare: NH MEA
ùÌáÞì ùÍaÞì"úÏä
/lavaS/ 'to wear something' /hitlabeS/'to get dressed' la:wiS 'that he wear' malwo:Se 'dressing'
where the Modern Hebrew forms show an alternation between /b/ and /v/ in the root √ lbS ~ lvS 'to dress', while the Modern Eastern Aramaic has an invariable root √ lwS (where /w/ reflects the fricative allophone of *b). 3.1.3.3 Consonant Shifts in Modern Western Aramaic In Modern Western Aramaic, the earlier begad kefat fricative allophones have become the phonemes {/b/, /Ä/, /D/, /x/, /f/, /T/}. Note that the fricative allophone [ b ] of earlier *b has become the voiced stop /b/, as part of a major shift in the phonetics of the nonemphatic stops in Modern Western dialects, by which Late Aramaic phonetic voiced stops become voiceless, and voiceless stops in most case are lenited. The changes (following Jastrow 1997:335) are as follows: 5
Depending on how one chooses to phonemicise these cases, there are exceptions to post-vocalic spirantisation in Syriac, the best known cases being the 1s and 2s suffix forms of the III weak verbs: for example, h2di t 'I am glad', h2dit 'you (ms) are glad'; orthographically t{YDX and tYDX}, respectively. There is some historical evidence for an analysis in which the latter has an underlying pre-suffix glide (/h2kiyt/), but the former does not (/h2dit/). Under that analysis, the exceptionality vanishes. Otherwise, one must assume that the phonemicisation of post-vocalic spirantisation had begun in Syriac. 6
It is possible to analyse Classical Syriac without these shwa positions; for example, k«tab vs. kTab. The latter analysis removes the environment for many fricatives, rendering them phonemic. 14/2/02
Draft Only: Do Not Quote Or Circulate Without Permission
LIN325 Introduction to Semitic Languages Chapter 3: Phonology 8 Late Aramaic
p
p#
b
Modern Western
b
p
Aramaic
f
t
t
d
k
t
tS~ts B
d
T
k
g
g#
k~tS x
D
Ä
Figure 1: Western Aramaic Consonant Shifts (The mergers that have affected the phonemicisation of earlier begad kefat fricative allophones in Modern Eastern Aramaic are shown in Figure 2.) Table 6: Modern Eastern Aramaic Consonants Labial
Dental Alveolar
Stop
(Alveo-) Palatal
p
t
(tS)
k
vd
b
d
(DZ)
g
t23
Gutteral Pharyngeal Glottal
vls
emph
Fric
Dorsal4
Apical
?
q
vls
f1
T
s
S
x
ð
vd
(v)2
D
z
(Z)
Ä
À
emph
h
s2
vls Lateral
vd
l
emph Nasal
m
r
Rhotic Glide
n
w
j
1. Mainly in borrowings, except in some dialects in the west of the ENA area in which earlier [f] does not go to [p]. 2. Phonemes in parentheses are found in borrowings. 3. Most non-gutterals have corresponding emphatics; only those inherited from Proto-Semitic are given here. 4. In some dialects, velar stops are fronted to palatals; /q/ is uvular.
14/2/02
Draft Only: Do Not Quote Or Circulate Without Permission
LIN325 Introduction to Semitic Languages Chapter 3: Phonology 9 ProtoSemitic
Early Aramaic
m *
m
p *
p
Imperial Aramaic
Modern E. Aramaic
/p/ [p] [f]
b *
Syriac
b
m
m
p
p~f
b
b~v
[f]
/b/ [b] [B]
Tiberian Hebrew
[B]
w *
w
w
w/y
*y
y
y~Ø
y
n *
n
n
n
l*
l
l
l
r*
r
r
r
t*
t
/t/ [t]
t
t~T
[T]
T
*T
S1
*s
s
Â*
S2
d *
d
D *
z1
z*
z
z
z
z
*s2
s2
s2
s2
s2
*T2
s2
t2
t2
s2
t* 2
t2
*s&
S
S
S
*?
?
?
?
Â* 2
q1
*Ä
À
S s
s
s• /d/ [d]
d
[D]
D
t2
À
À
k* *x
s2
À g
k ð
d~D z
*À g *
s
/g/ [Ä]
À
[[ Ä]
g~Ä
[g]
g
/k/ [k]
k
[x]
x ð
h * 5
k~x ð
ð
ð
q *
q
q
q
h *
h
h
h
1. orthographic representation only (see 3.1.3.1)
Figure 2: PSEM Consonants in Modern Eastern Aramaic 14/2/02
Draft Only: Do Not Quote Or Circulate Without Permission
LIN325 Introduction to Semitic Languages Chapter 3: Phonology 10 3.1.3.3 Effects of Borrowing on the Consonant Inventory Note that in most varieties of Modern Aramaic, many of the phonetic consonants lost through merger or allophonic change have been reintroduced through borrowings. In Eastern Aramaic those borrowings are largely from Kurdish and Arabic, and have given rise to an /f/ in those varieties in which the historical fricative allophone of /p/ has merged with the stop. Phonemic pharyngeals are also found in Eastern Aramaic, mostly in borrowings but in some native etyma as well. New contrastive consonants (for example, alveopalatal affricates) have also arisen through borrowing in Modern Eastern Aramaic. 3.2.0 Syllable Structure Unless the orthography marks vowel position and quantity and consonant gemination, it is difficult to glean information about syllable structure in older Semitic languages. In Phoenician, for example, there is no orthographic indication of either vocalism or consonant gemination, so comparative evidence is all we have. In the cuneiform orthography of Eblaite, the basic graphemic unit is the open syllable. Closed syllables (and geminate consonants) in Eblaite must again be inferred through comparative evidence. The Ugaritic cuneiform consonantary indicates vowels only in the environment of glottal stop. Amongst the older Semitic languages, syllable structure constraints are most transparent for Akkadian. For languages like Classical Arabic, Tiberian Hebrew, Middle Aramaic, Syriac, and Ge'ez, with (auxiliary) orthographic representation for vowels, syllable structure information is relatively transparent, at least for the received reading tradition(s). 3.2.1 Classical Arabic Syllable Structure Of the languages for which syllable structure data is readily available, Classical Arabic appears to be the most conservative, since it has not undergone any of the processes of short vowel loss or reduction that have affected most other West Semitic languages, and preserves most PSEM consonant distinctions. (The consonant mergers that have occurred in Classical Arabic do not affect syllable structure.) In Classical Arabic, syllables may be light (CV) or heavy (CVC, CV:). All syllables have an onset. Initial consonant clusters are disallowed per se, but there is evidence, both internal and comparative, for earlier *#CC clusters to which a prothetic [?i] or [?u] (termed elidible alif) has been added; for example: Vï:òJî7 /yaktub/ 'let him write', Vï:ò2ïÇ [?uktub] 'write (ms)!'; ðæUîM#ò ðÇ [iTna#ni] 'two (m)' (cf TH í›éžð" [Snajim] Ö 'two (m)'), but ðæUîM#ò ðÇ ðæîÇbî3îè [walada#ni 7 Tna#ni] 'two boys (nom)' . In general, superheavy syllables (CVCC or CV:C) are disallowed, except in two cases. The 7
A number of nouns in CA, most reflecting apparent PSEM biradical roots, also show elidable a l i f ; for
example: ìrò'ðÇ [?ismun] 'a name', ìs!ò ðÇ [?ibnun] 'a son', ìÄdò4ðÇ [?imru?un] 'a man', but ìrò?ð! [bismi] 'in the name of', ís!ò îs4 ð [mina bnin] 'from a son', íÄðdò4ð™ [limri?in] 'to a man'.
14/2/02
Draft Only: Do Not Quote Or Circulate Without Permission
LIN325 Introduction to Semitic Languages Chapter 3: Phonology 11 first are Ca:C syllables in active participles of geminate verbs like √DLL 'to guide'; for example: yäÇîÏ [da:llun] 'guiding (msn)'. The second case is in pausal (and citation) forms of nouns, in which is lost the short final vowel (with or without nunation8 ) that marks case. With nouns of the shape CVCC, a final cluster may occur in pause; for example, òVòKîJò3îÇ [?alkalb] for îVòKîJò3îÇ [?alkalba] 'the dog (acc)'. In such noun forms, the final vocalic case marker may also metathesise around the final consonant: VïKîJò3 [?alkalub] for ïVòKîJò3îÇ [?alkalbu] 'the dog (nom)'. In all other cases where a superheavy syllable would arise morphologically in CA, the vowel is shortened or the cluster broken; for example, in jussives and imperatives of hollow verbs9 (wð4uïI"î /taqu#miÜ/ 'you (fs) may get up', rïIî" /taqum/ 'you (ms) may get up', rï1 /qum/ 'arise (ms)!' or of doubled verbs (w=3ïbî" /tadulliÜ/ 'you (fs) may guide', qï3bî" /tadlul/ ~ ]äïbî" /tadulla/ ~ =äïbî" /tadulli/ 'you (ms) may guide'). The minimal word in CA (clitics aside) is bimoric: wð! [bi:] 'with me', sî4 [man] 'who?'. But bimoric nominal and verbal forms are in fact uncommon. 3.2.2 Syllable Structure in Akkadian and in Proto-Semitic Akkadian syllable structure constraints appear to have been similar to those of CA except that: i. phonologically empty onsets are permitted (where a consonant was lost)1 0 ii. CV:C superheavy syllables occur freely (again apparently as a consequence of consonant loss): OB is&a#l 's/he asked' cf. TH ìÌà"Ö›é/yiS?al/ 'he asks'. With the exception of a few syncategorematic items like u 'and', s&a 'that, which', and -ma 'and', the minimal word in Akkadian is bimoric, as in CA. And as in CA, bimoric forms are themselves uncommon, limited largely to a few prepositions like ana 'to, for, at' and pronouns like s&u# 'he' (but including some nouns like mû 'water'). Most major category 8
In Classical Arabic, nunation is a morphological phenomenon by which certain paradigm members (particularly indefinite nouns) acquire a final /n/ of uncertain provenance and function. 9
A h o l l o w verb is a triliteral root in which the medial consonant is a glide.
10
Glottals, laryngeals, the voiced velar fricative, pre-/post-consonantal glides and initial *y merge in Akkadian. Whether they merge on glottal stop or on Ø is a matter of debate. Loss is suggested by the fact that, in general, adjacent vowels across a glottal stop (from whatever source) elide: OB telqi^ 'you (fs) took' < *telqe+i# < *talqað+ i# Cases in which the expected vowel elision, or other effects of loss of glottal stop, fail to occur suggest that the glottal stop has been retained: s&ua#tu~s&a#tu 'he, that' . Such cases are relatively rare. (It has been argued that the pharyngeals were in fact phonemically distinct, though orthographically merged with the glottals etc., as evidenced by particular sound changes occurring only in the environment of PSEM pharyngeals. That analysis depends on the very unlikely claim that vowel allophony was represented orthographically in Akkadian. 14/2/02
Draft Only: Do Not Quote Or Circulate Without Permission
LIN325 Introduction to Semitic Languages Chapter 3: Phonology 12 words are disyllabic and/or trimoric, or longer. Are we justified in reconstructing superheavy (CV:C or CVCC) syllables or syllable initial consonant clusters for Proto-Semitic? Certainly Proto-Semitic final consonant clusters are unlikely; though morphemes might have that shape, it does not follow that syllables did. Where syllable-final clusters developed, they are typically a consequence of the loss of a following vowel; for example, in 2sf suffix verb forms in TH ("záÙ ’úÞ/ka# k tabt/1 1 'she wrote' (cf. CA ðX9ò î:î2 /katabti/ 'she wrote'), or CVCC nouns in some Modern Arabic dialects (Iraqi n3îÇ /?alf/, CA ìnò3îÇ /?alfun/ 'thousand'). Complex syllable onsets have likewise arisen in many Semitic languages, but again as a result of the loss of short vowels, as in Modern Aramaic or Maghrebian Arabic. A few Proto-Semitic morphemes, like *Tna# 'two' or *Sm- 'name' might be reconstructed with an initial consonant cluster, but those clusters are only rarely reflected in any of the older witnesses. It is difficult to know whether such clusters were actually permitted in PSEM, or whether even in the proto-language, they had to be resolved by prothesis or epenthesis. As already noted, underlying CV:C syllables in CA only surface in morphologically delimited cases. CV:C syllables are usually assumed for Akkadian, but it is, of course, impossible to know for sure how they might have been realised phonetically in that language. TH (and various forms of Aramaic) have such syllables word-finally, but most seem to have developped as a result of final vowel loss; for example TH øéÏòÞö /s2a#Ài#r/ CA ìdðFî• /s2aÀi#run/ 'small'. Of particular interest are the hollow verbs for which CV:C stems must be postulated for PSEM. We have already noted that in CA, the vowel of such stems shorten when in the word-final syllable, as in jussives or imperatives. Related TH forms are as follows: íÈ÷•é/ya#ku#m/ 'he rises', íÈ÷ /ku#m/ 'arise (ms)!', íS•é/ya#kom/ 'let him rise'. The TH jussive suggests the same short *u as the CA form, though the imperative suggests *u:. I would venture that the evidence from the jussive should carry more weight (since the imperative form may have been lengthened secondarily), and that the language antecedent to TH had the same constraint barring superheavy syllables as does CA. 3.3 Word Stress Except possibly for some Ethiopic and South Arabian languages, all Semitic languages have word stress. Blau (1979:50) points out that the stress pattern Classical Arabic has frequently been assumed for Akkadian and reconstructed for Proto-Semitic. The problem is in determining what the stress pattern of Classical Arabic is. The traditional Arabic grammarians were not concerned with word stress and provide no information about it. 11
I am not taking into consideration those analyses that view such forms as having a final syllable with an empty nucleus. 14/2/02
Draft Only: Do Not Quote Or Circulate Without Permission
LIN325 Introduction to Semitic Languages Chapter 3: Phonology 13 Classical Arabic reading traditions are influenced by the stress patterns of the modern spoken language, and those patterns vary considerably across the Arabic-speaking world. 3.3.1 Classical Arabic Two slightly different accounts of Classical Arabic stress appear in contemporary literature. Pausal forms aside, both accounts agree that the main stress falls on the rightmost non-final heavy syllable. When there is no heavy syllable, on one account the antepenult is stressed, and on the other, the initial syllable1 2. Thus: kataba waladun al-baytu Muh2ammadan al-madiinati
'he wrote' 'a boy (nom)' 'the house (nom)' 'Mohammed (acc)' 'the city (gen)'
The two account differ for forms like: katabahu katabahu
'he wrote it (m)'
In terms of Metrical Stress Theory, the pattern with default initial stress can be analysed as follows: Foot Construction: Left-headed, quantity sensitive, unbounded feet Word Level: Right Prominent Extrametricality: Final (heavy) syllable is extrametrical yielding: (* ) (* ) ka.ta.〈ba〉
(* ) (* ) wa.la.〈dun〉
( * ) ( * )( * ) ?al.bay.〈tu〉
(
(
(* ) (* ) ka.ta.ba.〈hu〉
* ) (∗ ) ma.dii.na.〈ti〉
* ) (* ) mu.ham.ma.〈dun〉
With default antepenultimate stress:
12
Angoujard (1990:1) appears to regard this variant as artificial, if not questionable.
14/2/02
Draft Only: Do Not Quote Or Circulate Without Permission
LIN325 Introduction to Semitic Languages Chapter 3: Phonology 14 Foot Construction: Moraic trochees, right-to-left Word Level: Right Prominent Extrametricality: Final syllable is extrametrical yielding: (* ) (* ) ka.ta.〈ba〉
(* ) (* ) wa.la.〈dun〉
( * ) ( * )( * ) ?al.bay.〈tu〉
(
(
(*
* ) (∗) ma.dii.na.〈ti〉
* ) (* ) mu.ham.ma.〈dun〉 )
(* ) ka.ta.ba.〈hu〉
According to Wright (1874:27) the proclitic prepositions and conjunctions of Classical eschew stress. When short final vowels (and any accompanying nunation) are lost in pause (sentence or verse-finally), the position of the stress is unaltered. 3.3.2 Modern Arabic There are a number of detailed descriptions of stress in modern Arabic dialects. For some references, see Angoujard (1990) and Hayes (1995). Eastern Arabic dialects, with minor variations, agree with the accounts of the classical language in giving priority to a rightmost heavy syllable, and do not show the default initial stress reported for the Classical language. Colloquial Cairene Arabic (Hayes 1995:67ff) has one of the less elaborate systems. The final syllable is stressed only if it is superheavy. Otherwise, if the penult is heavy, it is stressed. If neither of these conditions holds, then main stress falls either on the penult or the antepenult, depending on which is separated from the closest preceding heavy syllable (or the start of the word) by an even number of (light) syllables. In the following examples, Classical Arabic forms in Cairene pronunciation are marked CA: i. final stress katabt hajja:t ii. penult stress be:tak katabta fihim katabitu mudarrisit iii. antepenult stress
'I wrote' 'pilgrimages'
CA
your (ms) house' 'you (ms) wrote' CA 'he understood' 'she wrote it (m)' 'teacher (fc)'
14/2/02
Draft Only: Do Not Quote Or Circulate Without Permission
LIN325 Introduction to Semitic Languages Chapter 3: Phonology 15 ?adwiyatuhuma: 'their (d) drugs' kataba 'he wrote'
CA CA
Hayes (ibid.:69) observes that the Cairene pattern can be captured by the following constraints: Foot Construction: Moraic trochees, left-to-right1 3 Word Level: Right Prominent Extrametricality: Final consonant is extrametrical Stress in Maghrebian (western) dialects stress is complicated by the radical reduction of the vowel system in most of those dialects, in which earlier long vowels and diphthongs survive as 'full' vowels, but earlier short vowels are reduced, in some places to a single colourless central vowel. The reduced vowels are subject to syncope and the resulting consonant clusters to prothesis and epenthesis. At this point, I will leave the investigation of Maghrebian Arabic stress as an exercise for the reader. 3.3.3 Modern South Arabian and Ethiopic Materials available to me on stress in Ethiopic languages suggest that there is no obvious word stress in these languages. A typical remark is that of Raz (1997:447), regarding Tigré, that "…stress in Tigré is non-distinctive and shifts easily from one syllable to the other." (My own experience with Oceanic and Eastern Indonesian languages leads me to speculate that the main prosodic unit in Ethiopic languages might be larger than the word.) Hetzron (1997:539) makes similar observations about the irrelevance of stress in Outer South Ethiopic, but does note that stress in Peripheral West Gurage regularly falls on what I take to be the rightmost moraic trochee. In Modern South Arabian languages (Lonnet and Simeone-Senelle 1997), stress patterns vary from language to language. Stress is Mehri resembles the unbounded Classical Arabic pattern. A word final consonant is extrametical. Stress falls on the rightmost heavy syllable, and if there is no heavy syllable, it is initial. Jibbali and Soqotri have lost the quantity contrast for vowels. In Jibbali, stress appears to fall on the syllable cognate with that bearing main stress in Mehri (though the conditions have become opaque); otherwise, there is no particularly prominent word stress. In Soqotri, stress is initial. 3.3.4 Modern Aramaic Hoberman (1997) gives a reasonably detailed account of stress in the modern Eastern Aramaic languages with which he is familiar. He reports penultimate stress, but with some morphological exceptions. Imperatives, for examples, have initial stress. Vocalised semiconsonants are extrametrical, as in: g-málu:s&-a 'she dresses' (< lwas&a 'to dress') The enclitic copula is extrametrical, and a number of verbal suffixes appear to eschew 13
This directionality is necessary to prevent stress a l w a y s falling on the leftmost syllable in the rightmost sequence of light syllables, if there is one. 14/2/02
Draft Only: Do Not Quote Or Circulate Without Permission
LIN325 Introduction to Semitic Languages Chapter 3: Phonology 16 stress. Stress in Modern Mandaic is reported as penultimate, when the word has at least two unreduced syllables. For Western Aramaic, Jastrow (1997:336) reports what I interpret as a rightmost moraic trochee main stress foot, with an extrametrical word-final consonant 3.3.5 Biblical Hebrew Biblical Hebrew in the Tiberian tradition is the only older Semitic language for which stress is consistently marked. Fully pointed Tiberian texts have a system of points called ta'amim, appearing on the stressed syllable of ever word. The choice of ta'am is an indication of relative phrasal stress. In some longer words, secondary stresses are also marked. The description of main stress in Biblical Hebrew has been the subject of a large body of literature over the last few decades, in both generative and non-generative frameworks. I might note here the historically-oriented papers by Blau (1978, 1979, 1981), dissertations by Prince (1975), Rappaport (1984), and Churchyard (1999), and the monograph by Malone (1993). This level of interest is a consequence of the particular problems posed by Tiberian Hebrew stress. Minimal pairs like TH äÞî÷Þ /qa#m a#/ 'she arose' /qa#ma#/ 'rising (fs)' demonstrate that stress is superficially contrastive. Nonetheless, stress is largely predictable and probably fully predictable, given a sufficiently abstract analysis. In general, TH main stress is final if the final syllable is closed, and penultimate otherwise: ka#tab
'he wrote'
ka#tabti”
'I wrote'
k«tabtem
'you (pm) wrote'
Forms like: ka#t«bu# keleb
'they wrote' 'a dog'
appear exceptional1 4. The latter is assumed to reflect an underlying (and historical) *kalb, as evidenced by forms like: kalbi” 'my dog', and a process of epenthesis inserting /e/ in syllable-final cluster after main stress has been assigned. Examples like the former, with final stress on an open syllable, are assumed to reflect a process shifting stress rightward off a light penult. That process would be opaque but for the behaviour of such forms in pause, and for the process of pretonic lengthening that is True exceptions, like the 1s independent pronoun éÏë–ðÞ/?a# à no#ki”/ 'I', are less amenable to resolution.
14
14/2/02
Draft Only: Do Not Quote Or Circulate Without Permission
LIN325 Introduction to Semitic Languages Chapter 3: Phonology 17 characteristic of Tiberian Hebrew. Under pause (sentence- or verse-finally, and before the major verse-internal, atnax, disjunctive accent), forms with final stressed open syllables in context have their stress 'retracted' to the penult: ka#t«bu# 'they wrote' in context is ka#ta#bu# (with penultimate stress) in pause. Pretonic lengthening is a process by which a light (short open) pretonic syllable becomes heavy by lengthening its vowel. The forms ka#tab 'he wrote' and ka#tabti” 'I wrote' show pretonic lengthening, but k«tabtem 'you (pm) wrote' does not. The distribution of pretonic lengthening is accounted for if we assume that stress is penultimate in all words with a final open syllable: katab
katabti”
katabtem
katabu#
katab
katabti”
katabtem
katabu#
Main Stress
ka#tab
ka#tabti”
-------------
ka#tabu#
Pretonic Lengthening
-------
---------
-------------
ka#tabu#
Stress Shift
-------
---------
k«tabtem
ka#t«bu#
Vowel Reduction
Vowel reduction applies to light syllables preceding the main stress, reducing the vowels of such syllables either to shwa or to a hatef (ultrashort) vowel, a member of the set {a&, e(, o(}. The hatef vowels appear if the onset is a pharyngeal or laryngeal consonant. In a sequence of pretonic light syllables, reduction applies alternately, right-to-left, as in: wa.y«.dab.b«.ru#•
'and they spoke'
Blau (1978) observes that at least from a diachronic perspective, Tiberian Hebrew stress can be viewed as exclusively penultimate, if one takes into consideration the fact that (most) consonant-final forms with final stress at some point had a short final vowel that was later lost.1 5 History also provides an account of stress differentiated minimal pairs like TH äÞî÷Þ /qa#m a#/ 'she arose' /qa#ma#/ 'rising (fs)', noted above. The former is a verbal form in which we Blau observes (ibid.:113) that, from a diachronic perspective the main stress in forms like øÍaAÆ /dabbe#r/ 'speak! (ms imp)' is exceptional, since there is no evidence that such forms ever had a final vowel. He postulates a stress shift from a closed penult to a closed penult to account for such cases. But given that related imperfectives may have had a final vowel, marking mood, (a fact he does not note), there may be other accounts for these imperatives. 15
Blau has also suggested (1979) that in pre-Hebrew, the stress pattern must have been like that of Classical Arabic, but I am not certain that I find his reasoning compelling. 14/2/02
Draft Only: Do Not Quote Or Circulate Without Permission
LIN325 Introduction to Semitic Languages Chapter 3: Phonology 18 would expect stress shift to the ultima, except that the penult is heavy. The latter is a nominal form reflecting an earlier *qa#mátV, in which the stress was penultimate before loss of the short final vowel case suffix. Both forms, then, reflect historical penultimate stress. 3.3.5.1 Metrical Accounts of TH Stress and Vowel Reduction To my knowledge, two distinct prosodic accounts of Tiberian Hebrew main stress have been proposed. One (from Prince 1975, to McCarthy 1979, and Prince 1985) involves an iambic main stress foot at the right edge of the word, with word-final vowels extrametrical. That rule stresses a final closed syllable, and a penultimate open syllable otherwise. The other account (in Rappaport 1984 and Churchyard 1999) employs trochaic main stress. In Churchyard's analysis, the main stress foot is a non-iterative syllabic trochee, in which a word final consonant "can fill the weak position of a trochee" (Churchyard 1999:32). For Churchyard, then, main stress is effectively penultimate, with a final consonant serving as a weak syllable. What is at issue in these accounts is not so much the treatment of main stress assignment per se as the interaction between main stress and vowel reduction. Ideally, these two processes should be given a unitary account, in terms of a single prosodic level1 6. But in Tiberian Hebrew, there are conflicts between stress feet and reduction feet, exemplified in forms like ka#t«bu#• 'they wrote' (in pause, ka#ta•b # u#). In order to account for the pretonic lengthening of the vowel in the first syllable, at some point in the derivation main stress has to be penultimate, falling on a syllable that is on the surface reduced. That is, at some point in the derivation, the strong syllable of a stress foot coincides with the weak syllable of a reduction foot. All metrical accounts of stress and vowel reduction in TH distinguish the two processes in some way. The problem is that reduction foot assignment must follow main stress assignment, yet the word-internal process of vowel reduction creates units (reduction feet) that are lower in the prosodic hierarchy than the main stress foot created by the wordlayer end-rule of main stress assignment. Most of the recent literature on TH main stress and vowel reduction is concerned with resolving this paradox. Most accounts, like Prince (1985) and Rappaport (1984) use right-to-left iambic reduction feet. Prince assigns those feet in what is effectively a 'second pass' can overwrite (and shift) a previously assigned main stress. Rappaport has recourse to separate but interacting metrical tiers. Churchyard (1999) uses moraic trochees in constructing reduction feet, and solves the ordering paradox by creating the relevant metrical structures top-down, rather than bottom-up. 3.3.6 Middle and Late Aramaic 16
p a c e Churchyard's apparent argument for an in principle distinction between the two, since "while a syllable which is not dominated by the head of a reduction structure contains a reduced vowel …, a syllable which is not dominated by the head of a Hebrew stress foot is merely unstressed" (1999:40). 14/2/02
Draft Only: Do Not Quote Or Circulate Without Permission
LIN325 Introduction to Semitic Languages Chapter 3: Phonology 19 Main stress in Biblical and Jewish Literary Aramaic is very similar to that in TH, except that with no process of stress shift off a light penult, stress is more consistently penultimate in forms with final open syllables. There are, however, numerous, largely morphologically conditioned, exceptions: JLA àÞîÇé /yo#ma#/ '(the) day', úÏá’úÙ[k k «tabit] 'I wrote'. There are distinct Western (Maronite) and Eastern (Nestorian) traditions for stress assignment in reading Syriac. The Maronite tradition is similar to Tiberian Hebrew and earlier Jewish dialects of Aramaic: stress a final closed syllable, otherwise stress the penultimate. The Nestorian tradition favours final stress, as does Classical Mandaic. The only formal metrical account of stress and vowel reduction in any Middle or Late Aramaic dialect is my own account (Harrison to appear) of Jewish Literary Aramaic. That account is simpler than those outlined above for TH, since it need not postulate any stress shift. 3.4 Vowel Systems of Semitic Languages 3.4.1 Proto-Semitic Vowels The phonemic vowel system of Classical Arabic and many modern Arabic dialects is that reconstructed for Proto-Semitic, three vowel qualities plus length: i
u
i#
u#
a a# Figure 3: Classical Arabic Vowel Phonemes Some evidence for vowels is provided in the cuneiform syllabaries of Akkadian and Eblaite, and the cuneiform alphabet of Ugaritic (see section 2.4). In the latter, there are three glyphs for the consonant alef, which comparative evidence suggests represent vowel contrasts. Ugaritic is assumed to have had four vowel contrasts, on the evidence of forms like riV riv /re#S/ 'head' (PSEM *ra?S). PSEM *a?C. > UGR e:C. by a process perhaps analogous to the raising of [a] to [e] in Arabic, termed imaale. Akkadian and Eblaite are assumed to have had identical four vowel systems, long and short: i u i# u# e
e# a a# Figure 4: East Semitic Vowels
though the status of /e/ is perhaps questionable. It is contrastive only in Akkadian, and then only in Sumerian borrowings. Otherwise in both Eblaite and Akkadian, [e] is a conditioned variant of [a]. 14/2/02
Draft Only: Do Not Quote Or Circulate Without Permission
LIN325 Introduction to Semitic Languages Chapter 3: Phonology 20 The length contrast of Proto-Semitic and Classical Arabic disappears in some modern Arabic dialects; for example, Moroccan Arabic as reported in Heath (1997). In Moroccan Arabic, older long vowels become short, and older short vowels merge on shwa1 7:
*i
*i:
*u i
*a
u
*u: *a:
« a Figure 5: Development of Moroccan Arabic Vowels Ge'ez, Amharic and Tigrinya show similar reduction of the Proto-Semitic length contrast, though complicated further by the monothongisation of the PSEM diphthongs *ay and *aw. The changes can be represented as follows:
*i
*i:
*u *a
i
ö
e *ay *aw
u
*u: *a:
o Î
a Figure 6: Development of Ethiopic Vowels Whether the vowel system of Tigre differs from that outlined above appears to be a question of analysis. Raz (1983, 1997) prefers to consider the Tigre analogs of /æ/ and /a/ a length contrast. The same analysis has been proposed to Harari. The latter may represent an intermediate stage between Proto-Ethiopic and languages like Amharic, since short /i/ and /u/ are preserved, though they alternate with /«/. 3.4.2 Representing Vowels 3.4.2.1 Mater Lectionis A consonant letter serving as a vowel marker is termed a mater lectionis (Latin 'mother of reading', pl. matres lectionis ). In Arabic, the position and, in most cases, the quality of a long vowel is indicated by one of the consonants alif Ç, ya ê, or waw è, corresponding to a#, i#, and u#, respectively. In Hebrew and Aramaic, consonants are also used to represent vowels, though with rather less consistency than in Arabic. With some exceptions, final vowels are marked in Hebrew and Aramaic, but internal vowels may or may not be 17
Heath (1997:209) also reports a short rounded [u(] (< earlier short /u/), but only in the environment of velars or uvulars.
14/2/02
Draft Only: Do Not Quote Or Circulate Without Permission
LIN325 Introduction to Semitic Languages Chapter 3: Phonology 21 marked by a mater lectionis. Those word-internal vowels marked by matres lectionis in Hebrew and Aramaic were historically long, though not all historically long vowels are necessarily so marked. In Hebrew, he ä, yod é, waw å, representing low, non-low front, and non-low back vowels, respectively. The consonant alef à is also used as a mater lectionis in Hebrew, though with less frequency. The same consonants are used in Aramaic, but alef is more frequent than he in marking low vowels, in most Aramaic orthographic systems1 8. 3.4.3 Tiberian Hebrew Vowels If there is a change characteristic of the vowel systems of Semitic languages, it is the expansion of the set of quality contrasts from the three reconstructed for Proto-Semitic into sets with as many as seven quality contrasts (in Modern South Arabian and some interpretations of Tiberian Hebrew). These expansions are typically accompanied by a reduction or elimination of the role of quantity in the vowel system. Tiberian Hebrew provides a good case study of these issues; indeed, almost an embarrassment of riches, since many of the changes to its vowel system are unusual and/or complex, and some not easily interpreted. 3.4.3.1 Tiberian Vowel Pointing The Tiberian Hebrew vowel pointing system was developed toward the end of the first millenium CE, roughly simultaneously with similar developments for Syriac. I illustrate the Tiberian vowel points here with the consonant bet, with and without matres lectionis, and followed by a transliteration:1 9 éÏá Ïá éÍá àÍá áÍ éÎá äÎá Îá Ìá
bi, bî be, be', bê bè, bèh, bèy ba
Èá áÜ Çá àÝá áÝ
bu, bû bo, bo', bô
éÞá äÞá Þá bå, bâ, båy
Table 7: Tiberian Hebrew Vowel Points, with Transliteration The Tiberian vowels might be viewed in two ways; as a 7-position system in which only quality is distinguished, or as a 5-position system distinguishing both quantity and quality:
18
See also section 2.9.2 Vowels in West Semitic.
19
In this transliteration, the circumflex represents the unmarked mater lectionis for a non-low vowel; other matres lectionis are represented by a following consonant. è represents segol, and å qames2.
14/2/02
Draft Only: Do Not Quote Or Circulate Without Permission
LIN325 Introduction to Semitic Languages Chapter 3: Phonology 22 i
u
i
o
e
u o
e
i#
u# o#
e#
•
E a
a
Quality-Sensitive
a#
Quantity-Sensitive
Figure 7: Two Interpretation of Tiberian Hebrew Vowels The seven distinct Tiberian vowel points correspond to the seven vowels of the qualityonly system2 0. In the quantity-sensitive system, the high-vowel points shureq and h2ireq with mater lectionis are always long. Without mater lectionis, h2ireq and qibbus2 are usually short, but may be long. The mid-vowel points s2e re and holem are long. Segol is the short counterpart of s2e re. In the quantity-sensitive interpretation, qames2 is ambiguous, marking both /a#/ and /o/. These correspondences are summarised in Table 8: Quality
Transcription Quantity
i
î
i”
i
i
i” ~ i
e
e e' ê
e#
E
è èh èy
e
a
a
a
•
å
a# ~ o
•
â, åy
a#
o
ô o' o
o
u
û
u#
u
u
u
Table 8 Interpreting Tiberian Hebrew Vowel Transcriptions The short /e/ and /o/ of the five-vowel system correspond to the /E/ and /•/ of the seven20
In the quality-only interpretation, matres lectionis are regarded as redundant, and shureq and qibbus2 as allographs. 14/2/02
Draft Only: Do Not Quote Or Circulate Without Permission
LIN325 Introduction to Semitic Languages Chapter 3: Phonology 23 vowel system; the /a/ of the seven-vowel system corresponds to both short and long /a/ of the five-vowel. Joüon (1991:37-38) attributes the quantity-sensitive analysis to the 12th century grammarian David Qimh2i, and suggests that Qimh2i may have been influenced by Arabic and/or Latin. Joüon's grammar (ibid. 34ff) favours the qualitative view. 3.4.3.2 Other Hebrew Vowel Pointing Systems Four different systems (ignoring intra-system variation) were developed for representing vowels in Biblical Hebrew texts: Tiberian, Babylonian, Palestinian, and Samaritan. The Tiberian and the Palestinian, distinguish seven vowels (ignoring both mater lectionis and the representation of reduced and ultrashort vowels). The Babylonian system distinguishes six2 1, while the Samaritan distinguishes only five. i
u o
e
•
E a
Figure 8: Hebrew Vowel Pointing Systems In Babylonian pointing, Tiberian /E/ and /a/ are conflated; in Samaritan, the pairs {/i/, /e/} and {/u/, /o/} are conflated. In some Samaritan manuscripts, /a/ and /å/ are confused. The same is true for some Palestinian texts, as is confusion of {/e/, /E/} and {/u/, /o/}. This suggests that the actual Palestinian and Samaritan pronunciation may have been similar. 3.4.3.3 Proto-Semitic Long Vowels and Diphthongs in Tiberian Hebrew The Proto-Semitic long vowels *i#, *u#, and *a# become TH /i#/, /u#/, and /o#/, respectively. The change PSEM *a# > o# is characteristic of Canaanite (as opposed to Aramaic) NW Semitic; for example: PSEM *Sala#m- 'peace, well-being', CA
sala#m -
TH
Salo#m
Arm
S«la#m
21
Babylonian pointing is discussed briefly in section 2.9.2.
14/2/02
Draft Only: Do Not Quote Or Circulate Without Permission
LIN325 Introduction to Semitic Languages Chapter 3: Phonology 24 Apparent exceptions (where *a# remains a# in Hebrew) may reflect Aramaic influence. The PSEM diphthongs *ay and *aw are regularly reflected as (typically plene; that is, with mater lectionis) /e/ and /o/2 2, respectively. The diphthong *ay becomes short /e/ (sometimes a#) when stressed in synchronic word-final position, alternating with /i/ non-finally, as in: äÎàY›é /yir?e/ 'he sees', äÞàøÞ /ra#?a#/'he saw', äÞàøÞ /ra#?i#ti#/ 'I saw' (PSEM *r?y 'to see'). Figure 8 summarises the major Tiberian Hebrew reflexes of Proto-Semitic long vowels and diphthongs. *i:
*ay
i#
u#
e^
o#
e
a#
*u: *a: *aw
Figure 8 PSEM Long Vowels and Diphthongs in Tiberian Hebrew In Figure 8, the Tiberian Hebrew reflexes are represented in the quantity-sensitive analysis. With one exception, the corresponding vowels in the quality-sensitive analysis can be obtained by removing the macron. The exception is /a#/, which is /•/ in the qualitysensitive analysis. 3.4.3.4 Proto-Semitic Short Vowels in Tiberian Hebrew The history of Proto-Semitic short vowels in Tiberian Hebrew is more complex, being sensitive to main stress and to syllable type. Major developments are shown in Figure 9 (in which the Tiberian Hebrew reflexes are show in the quality-sensitive analysis). In some forms, the original diphthongs remain; for example, *bayt 'house, Hebrew ú›éÌbá á y i t 'house', úéÍá b ê t 'house of'; *mawt 'death', Hebrew úœåÌmáwet î 'death', úÇî mot 'house of'. All such cases of which I am aware are reflexes of PSEM *CaGC- forms, which take an epenthetic vowel in TH to break up the consonant cluster. They may not be true exceptions.
22
14/2/02
Draft Only: Do Not Quote Or Circulate Without Permission
LIN325 Introduction to Semitic Languages Chapter 3: Phonology 25 Figure 9 Tiberian Hebrew Reflexes of Proto-Semitic Short Vowels *iá
*u
*a( C. *a
before geminates
i C.
u C.
e C.
o C.
*iá
*u
*i' C. *a
i
u
e
o
E
E C. a C. • C.
a
•
Philippi's Law Law of Attenuation Segolate Nouns
in Short Unstressed Closed Syllables
in Stressed and Open Pretonic Syllables
The Tiberian Hebrew reflexes of the Proto-Semitic short vowels are largely a function of stress and syllable structure. We begin by noting a number of phenomena already considered in section 3.3.5. First, immediate pre-Hebrew word stress appears to have been penultimate. Second, final short vowels were lost in Hebrew, with two immediate consequences: i. The only remaining (synchronic) final vowels2 3 were (phonologically) long. ii. Since the position of main stress apparently did not change in previously *CVCV# forms, most subsequently consonant-final words had synchronic final stress. Third, the constraint barring stress on a light penult (discussed in 3.3.5) was resolved through rightward shift of the stress, created another class of forms with synchronic final stress. 3.4.3.4.1 Short Vowels in Open Tonic and Pretonic Syllables In section 3.3.5, it was observed that, with some exceptions, shorts vowel in (pre-Hebrew) immediate pretonic open syllables were lengthened. PSEM *a becomes /a#/, written qamef, but PSEM *i and *u become mid e# and o#, respectively. For example: á’úÞ/ka# k tab/ 'he wrote' øÞò‘× /sÛe#Àa#r/ 'hair' äÞá]O /k«ro#b â/ 'near (fs)' where the initial syllables are pretonic. What was not observed in 3.3.5 is that, again with some exceptions, vowels in pre-Hebrew 23
ignoring cases like e# < *ay#
14/2/02
Draft Only: Do Not Quote Or Circulate Without Permission
LIN325 Introduction to Semitic Languages Chapter 3: Phonology 26 open stressed syllables also lengthened. Since the pre-Hebrew forms in question typically had a short final vowel, the resulting forms now appear to end in a CV:C superheavy syllable: á‘úÇk/ko#te#b / 'writing, one who writes' (< *ka#tib-) áÞáÍì /le#b a#b / 'heart' ( < *libab-) ïÝè÷Þ /qa#tªo #n/ 'he is small' ( < *qatuna) Tonic lengthening did not apply in 3sm perfective verb forms like á’úÞ/ka# k tab/ 'he wrote', and must have preceded (or been otherwise blocked in forms undergoing) stress shift from a light penult, as in: äÞáú" kÞ /ka#t«ba#/ 'she wrote'. 3.4.3.4.2 Philippi's Law and Short Vowels in Closed Tonic Syllables In closed stress syllables PSEM *a remains /a/ (as in á’úÞ/ka# k tabta#/ 'you (sm) wrote'; cf. CA îX9ò î:î2 /katabta/). PSEM *u seems to become /o#/, as in imperatives like ãÝîÚò /Àa&mo#d/ 'stand' (< *Àmud). If indeed such examples of *u > o# are not just analogy with related, and originally vowel final, members of the verbal paradigm, then one wonders whether what is involved is lengthening, and not just lowering of the stressed *u in a closed syllable. PSEM *i in closed stressed syllables does regularly lower, to /a/, a process termed Philippi's Law. Examples are: *kabid 'be heavy' ãÍáÞk /ka*be@d/ 'it (m) is heavy'
é!zEÌáÞk /ka#ba@dti/ 'I am heavy' í•zEÌá/k«badtE@ Ùk m/ 'you (pm) are heavy'
The form é!zEÌáÞk /ka#ba@dti/ 'I am heavy' is an instance of Philippi's Law, but í•zEÌáÙk /k«badtE@m/ 'you (pm) are heavy' should not be. It is not clear whether this last example is a matter of paradigm levelling or of lowering in a pretonic (and secondarily stressed, when followed by tonic CVC#) syllable. Forms like ãÍákÞ /ka#be@d/ 'it (m) is heavy' above, and ïÍäÝk /kohén/ 'priest' ( < PSEM *ka#hinArabic ìs6 ð Uî2 /ka#hinun/ 'a priest') are not exceptions to Philippi's Law. Since they originally had a final short vowel, their penultimate syllable was subject to tonic lengthening, obviating any later application of Philippi's Law. /gat/ 'wine press' originally ended in By contrast, forms like TH úÌa /bat/ 'daughter' and úžb a consonant cluster, so Philippi's Law applied to them even in the presence of a historical 14/2/02
Draft Only: Do Not Quote Or Circulate Without Permission
LIN325 Introduction to Semitic Languages Chapter 3: Phonology 27 final vowel: *bint 'daughter' úÌa /bat/ 'daughter'
é!zÏa
/bittiÛ/ 'my daughter'
(cf. *bin 'son') (cf. ïÍ a /ben/ 'son')
(cf. é›ðÙ/b«ni/ a 'my son',
ÐïÎa /bEn/ ~ ÐïÏa /bin/
'son of')
*gint 'wine press' úžb/gat/ 'wine press'
íé!z›b/gittiÛm/ 'wine presses' 3.4.3.4.3 Segolate Nouns A large class of PSEM noun stems had the shape *C1 VC2 C 3 -, where V is any PSEM short vowel. In Northwest Semitic, these nouns are termed segolate, because the final consonant cluster in their unsuffixed (absolute and construct singular) forms is broken by an epenthetic vowel that is usually Tiberian segol ; that is, /e/. After PSEM *y, the epenthetic vowel is /i/; before a gutteral, it is /a/. Some examples are: *kalb- 'dog' (cf. Arabic ìVòK2 î /kalbun/ 'a dog')
áÎìÎk /keleb/ 'a dog' éÏaÙìÌk /kalbi/ 'my dog'
*baÀl- 'master' (cf. Arabic ìqòFî! /baÀlun/ 'husband')
ìÌòÌa /baÀal/ 'a master' éÏìòÚ aÌ /baÀa(li/ 'my master'
*sipr '?calculation' (cf. Arabic ìÑòW' ð /sifrun/ 'numbers')
øÎôÍñ /se#p#er/ 'a book' éXÙôÏñ /sip#ri/ 'my book'
*qudS 'holy, holiness' (cf. Arabic ìÓòbï1 /qudsun/ 'holiness')
ÖCS /qo#deS/ 'holiness' ÇÖEÞ÷ /q•dSo#/ 'his holiness'
Not all *C1 V1 C 2 C 3 - nouns are reflected as segolates in Hebrew. Some appear as TH C 1 «C 2 V1 C 3 nouns, in which the original V1 C 2 has become C2 V1 , as in:
14/2/02
Draft Only: Do Not Quote Or Circulate Without Permission
LIN325 Introduction to Semitic Languages Chapter 3: Phonology 28 *bi?r 'well' (cf. Arabic ìd8ò ð! /bi?run/ 'well')
øÍàÙa /b«?e@r/ 'well' éXÍàÙa /b«?eriÛ/ 'my well'
Others have assimilated to a non-segolate morphological class: øÞä•ð /na#ha#r/ 'stream' (CA ìdòN5î /nahrun/ 'river'). In unsuffixed segolate noun forms the stress is penult in Tiberian Hebrew. In a derivational synchronic analysis, penultimate stress follows if segolate epenthesis is a relatively late process, following main stress assignment. The same analysis can be interpreted diachronically, with segolate epenthesis following stress shift. On the basis of Greek transcriptions of Hebrew segolates, Blau (1978:97, 99-103) argues that epenthesis must have preceded stress shift. He appears to adopt the synchronic analysis, claiming that TH segolates are morphophonemically monosyllabic. What TH vowel is expected in the unsuffixed forms of segolate nouns is a function of whether or not we expect tonic lengthening (of light stressed syllables) to have applied. TONIC PSEM
Open
Closed
Actual Segolate
*a
a#
a
e~a
*i
e#
a
e~e#
*u
o#
o#
o#
Table 9 TH Unsuffixed Segolate Noun Forms Table 3 gives us some inconclusive information regarding the relative chronology of segolate epenthesis and tonic lengthening. Since *u seems to default to holem /o#/ in all tonic syllables, open or closed (see above), we are not surprised to find /o#/ in reflexes of *CuCC- nouns. Tonic lengthening does not apply in *CaCC- segolates (except in pausal /na#p#eS/ 'soul', cf. contextual ÖÎôœð /nep#eS/), but we find segol /e/ (as in óÎñÎk forms like ÖÎô•ð 'silver', éÏtÙñÌk 'my silver') rather than patah2 (short /a/) except before pharyngeals/laryngeals and glides: ú›éÌ/baÛ a yit/ 'house' < *bayt, øÌò’Ö /SáÀar/ 'gate'. For *CiCC- nouns we would expect Philippi's law to apply, yielding /a/ in a closed syllable. Again, we find mostly segol /e/, as in çÌáœæ'sacrifice' (éÏçáÙ ›æ'my sacrifice'). I can only assume that the Tiberian segol for both *CaCC- and *CiCC- nouns is an assimilation to the epenthetic segol. The few *CiCC24
Segol before a gutteral is rare, but possible. An example is íÎçÌì /leh2em/ 'bread'.
24
14/2/02
Draft Only: Do Not Quote Or Circulate Without Permission
LIN325 Introduction to Semitic Languages Chapter 3: Phonology 29 nouns for which *i becomes /e#/ remain unaccounted for. Some, like /se#p#er/ 'book' (cf. Akk. sipru) may be borrowings. On balance, then, I conclude that segolate epenthesis is a relatively late historical process, following tonic lengthening. 3.4.3.4.4 Short Unstressed Vowels In the appropriate metrical environment (see section 3.3.5.1), PSEM short unstressed vowels reduce (to shwa or a hatef vowel) in synchronic open syllables. In a sequence of short open syllables, reduction applies in an alternating fashion. In the examples of such sequences of which I am aware, the unreduced member receives secondary stress, and the vowel reflected is that expected in a closed unstressed syllable. In unstressed syllables closed by a geminate consonant, the PSEM short high vowels *i and *u remain unchanged. In other closed unstressed syllables, they lower to /E/ and /•/. In Tiberian pointing the low back vowel /•/ < PSEM *u is represented by qamef, and read [o] rather than [a] in Sephardi pronunciation. (In Ashkenazi pronunciation, all qamef are pronouned [o].) This qamef is termed qamef xatuf ('snatched' or 'shortened' qamef). The qamef read as [a] in Sephardi pronunciation, and termed qamef gadol or qamef raxav ('large' or 'broad' qamef), is the result of (pre)tonic lengthening of PSEM *a. PSEM short *a in a closed unstressed syllable has two outcomes. Typically it remains /a/, but in many non-final closed unstressed syllables, *a raises to /i/ by a process sometimes termed the Law of Attenuation. Its operation can be seen in the form of the imperfective verb prefixes; for example, á&zÙë›é /yiktob/ 'he writes' (Arb. ïV:ï òJî7 /yaktubu/). The change must have been relatively late, since it is not reflected in Greek transcriptions of Hebrew names in the Septuagint. Consider proper names like ïÇÖÙî!Ö /SimSon/ 'Samson', ìÆAšâÏî /mig#dal/ 'tower, Migdal (place name) > Magdalene'. Attenuation is it not regular; in many closed unstressed syllables, *a remains /a/. Compare íÍl!Ö /Sille@m/ 'he finished/paid' and íÍl’Ö /Salle@m/ 'finish, pay (ms imp)', both of which should reflect an earlier *Sallim. (Cf. Arabic rñK' /sallama/ /he greeeted/ and rñK' /sallim/ 'greet (ms imp)')2 5. The variations changes affecting Proto-Semitic short vowels in Tiberian Hebrew give rise to vowel alternations in the Hebrew noun paradigm. The following are some examples with PSEM high vowels:
25
The Hebrew and Arabic forms are cognate, though their meanings are quite different. Note that the etymologically non-suffix (i.e. imperfective) stem *CaCCiC has replaced the etymological *CaCCaC in Northwest Semitic.
14/2/02
Draft Only: Do Not Quote Or Circulate Without Permission
LIN325 Introduction to Semitic Languages Chapter 3: Phonology 30 *libb 'heart'
*ðuqq 'statute'
áÍì /Èleb/ 'a heart' ÐáÎì /lEb/ 'heart of' éÏaÏì /libÈbi/ 'my heart'
÷Ýç /Èðoq/ 'a statute' •êOÞç/ð•q«Èka*/ 'your (sm) statute' éÆNÜç /ðuqÈqiÛ/ 'my statute'
*baqqiS2 6 'seek'
*kull 'all' (Arabic ñqï2 /kull/)
ÖÆLÌášé /y«baqÈqeS/ 'he sought' ÐìÞk /k•l/ 'all' (cf. ìÞk /Èkal/ 'he measured') •ê"ÖÆLÌá/y«baqqES«Èka* šé / 'he sought you' ìÝk /Èkol/ 'all' íÞlkÜ /kulÈla*m/ 'all of them' Here, qamef gadol and qamef xatuf are distinguished as /•/ and /a*/, respectively. In the following examples, note the effect of (pre-)tonic lengthening on PSEM *a, giving rise to TH qamef gado l: *√ktb 'write' á’úÞ2 7k/ka*Ètab/ 'he wrote'
*dabar 'word, thing' øÞáÞc /da*Èba*r/ 'word'
*yad 'hand'
*√Swb 'return'
é!záÙ ’úÞ/ka* k Ètabti/ 'I wrote' øÌáÆE /d«bar/ 'word of' í•zÙá’úÙ /k«tabÈtEm/ k 'you (2pm) wrote' éXÞáÆE /d«ba*Èri/ 'my word' ã•é/Èya*@d/ 'a hand' ãžé/yad/ 'hand of' éD•é/ya*ÈdiÛ/ 'my hand' íÎëEœé/yEd«ÈkEm/ 'your (mp) hand' úÇã•é/ya*Èdot/ 'hands' úÇãšé/y«Èdot/ 'hands of'
á'Ö /Sa*b/ 'he returned' é!zÙá’Ö/ÈSabti/ 'I returned' Èá'Ö /Sa*Èbu/ 'they returned' í•zÙá’Ö /SabÈtEm/ 'you (mp) returned'
*yamm 'sea' í•é/Èya*m/ 'sea'
í•é ¬ížé /yam, ya*m/ 'sea of' íéÏnžé/yamÈim/ 'seas' 26
This form should be Proto-Semitic *baqqaS, but the vocalism of the pi'el prefix conjugation has been generalised to both prefix and suffix conjugations in West Semitic.
Qal verb forms like á’úÞ/ka# k t ab/ 'he wrote' are a patterned exception to tonic lengthening; compare øÞáÞc /da#ba#r/ 'word'. 27
14/2/02
Draft Only: Do Not Quote Or Circulate Without Permission
LIN325 Introduction to Semitic Languages Chapter 3: Phonology 31 Note the absence of (pre)tonic lengthening in (non-final) closed syllables, and in construct noun forms. 3.4.3.5 Quality or Quantity in Tiberian Hebrew? The Tiberian masoretes inherited a consonantal text many hundreds of (and perhaps a thousand or more) years old, in which the position of original diphthongs, most final vowels, and (most) reflexes of PSEM long vowels was indicated by a mater lectionis. It does not, of course, follow that all such long vowels or diphthongs remained long in TH, or that new long vowels had not arisen subsequent to the consonantal text being fixed. The seven vowels of the Tiberian pointing system (ignoring shwa and the hatef vowels) are interpreted as a 7-position quality system in the Ashkenazi reading tradition. In the Sephardi reading tradition, s2ere [e] and segol [E] are collapsed as [e], and patah2 and qames2 as [a] (except for some instances of qamef xatuf , which are read as [o]). i
u
e
o
E
• a
i
u
e
o a
Figure 10 Tiberian Vowels in Ashkenazi and Sephardi Reading Traditions Joüon (1991) and Khan (1997 and elsewhere) argue that the Ashkenazi reading tradition roughly reflects the actual pronunciation of the Tiberian masoretes; that is, that their vowel system was purely qualitative. This interpretation is not the "traditional" one, as already observed in 3.4.3.1 above. The "traditional" interpretation of Tiberian pronunciation assumes the five vowel qualities of the Sephardi system, with contrastive vowel quantity; that is, a 10-vowel system. In this quantitative interpretation, all instances of s2e re, h2olem, and s&u req are long, while all instances of patah2 and segol are short. The vowel points hireq, qibbus2, and qames2 are ambiguous. With mater lectionis yod, hireq is always long [iÜ]. Otherwise it may be a long or a short high front vowel. qibbus2 may be a short or a long high back vowel, and qames2 may be [a#] or [o]. The purely qualitative interpretation of Tiberian vowel pointing has superficial appeal, but the traditional quantitative interpretation has one major advantage over it. The quantitative interpretation permits an account of vowel reduction and stress shift (see section 3.3.5.1) in terms of a contrast between light (short open) and heavy syllables. Khan (1997:93-99) gives a non-quantitative account of reduction, but I do not find it particularly convincing. 14/2/02
Draft Only: Do Not Quote Or Circulate Without Permission
LIN325 Introduction to Semitic Languages Chapter 3: Phonology 32 In an abstract synchronic analysis of Tiberian Hebrew phonology, it need not follow that the contrast between phonologically long and short vowels was directly reflected in a phonetic quantity contrast. Quantity may have been neutralised on the surface; There are to my knowledge no theoretical reasons to rule out an analysis in which an underlying quantity contrast, to which vowel reduction and stress shift are sensitive, is mapped onto a purely qualitiative vowel system on the surface. References Angoujard, J-P. 1990. Metrical structure of Arabic. Dordrecht : Foris Publications. Blau, Joseph. 1978. Hebrew Stress Shifts, Pretonic Lengthening, and Segolization: Possible Cases of Aramaic Interference in Hebrew Syllable Structure. Israel Oriental Studies 8:91106. Tel Aviv: Faculty of Humanities, Tel Aviv University. (reprinted in Blau 1998:104119). 1979. Some remarks on the prehistory of stress in Biblical Hebrew. Israel Oriental Studies 9:49-54. Tel Aviv: Faculty of Humanities, Tel Aviv University. (reprinted in Blau 1998:120-125). 1998. Topics in Hebrew and Semitic Linguistics. Jerusalem: Magnes. HSS 492.4 1998 TOP Churchyard, H. 1999. Topics in Tiberian Biblical Hebrew metrical phonology and prosodics. unpublished doctoral dissertation, Department of Linguistics, University of Texas at Austin. HC599DIS.PDF Goldsmith, John A. ed. 1995. The Handbook of Phonological Theory. Cambridge, Massachusetts: Blackwell. HSS 414 1995 HAN Hammond, Michael. 1985, Main Stress and Parallel Metrical Planes. in Niepokuj, Mary et al., eds. Proceedings of the Eleventh Annual Meeting of the Berkeley Linguistics Society. pp., 417-428. Berkeley: Berkeley Linguistics Society. P 410.5 P11 Harrison, S.P. to appear. Vowel quantity and stress in Jewish Literary Aramaic. to appear in Journal of Semitic Studies. Hayes, Bruce. 1995 Metrical Stress Theory: Principles and Case Studies. Chicago: University of Chicago Press. Hetzron, Robert. ed. 1997. The Semitic Languages. London: Routledge. HSS 492 1998 SEM Hoberman, Robert D. 1995. Current Issues in Semitic Phonology. in Goldsmith, John A. ed. 1995:839-847. 1997. Modern Aramaic phonology. in Kaye and Daniels, eds. (1997:313-335) Hudson, Grover. 1995. Phonology of Ethiopian Languages. in Goldsmith, John A. ed. 1995:782-797. Jastrow, Otto. 1997. The Neo-Aramaic Languages. in Hetzron, Robert, ed. 1997:334-377. Joüon, Paul. 1991. A Grammar of Biblical Hebrew. Part One: Orthography and Phonetics. (tr. and rev. by Takamitsu Muraoka). Rome: Pontifio Instituto Biblico. hssmn 492.482421 1991 GRA v1-2 Kaye, Alan and Peter Daniels, eds. 1997. Phonologies of Asia and Africa, vol. 1. Winona Lake, Ind. : Eisenbrauns. hssmn 414 1997 PHO v1 Khan, Geoffrey. 1997. Tiberian Hebrew Phonology. in Kaye and Daniels, eds. pp. 85-102. Lonnet, Antoine and Marie-Claude Simeone-Senelle. 1997. La phonologie des langues subarabiques modernes. in Kaye and Daniels, eds. 1997:337-372. Malone, Joseph. 1993. Tiberian Hebrew Phonology. Winona Lake, Indiana: Eisenbrauns. HSS Q 492.47 1993 TIB McCarthy. 1991. Semitic gutterals and distinctive feature theory. in Eid, M. and B. Comrie, 14/2/02
Draft Only: Do Not Quote Or Circulate Without Permission
LIN325 Introduction to Semitic Languages Chapter 3: Phonology 33 eds. Perspectives on Arabic Linguistics, vol. 3, pp. 63-91. Philadelphia: John Benjamins. Morag, Shelomo. 1962. The vocalisation systems of Arabic, Hebrew, and Aramaic. The Hague: Mouton. HSS 492 1962 VOC Qimhi, David. 1952. David Kimhi's Hebrew Grammar (Mikhlol) Systematically Presented and Translated by William Chomsky. New York: Bloch. Prince, Alan S. 1975). The Phonology and Morphology of Tiberian Hebrew. Massachusetts Institute of Technology dissertation. 1985. Improving Tree Theory. , in Niepokuj, Mary et al., eds. BLS 11. pp. 471-490. Berkeley: Berkeley Linguistics Society. Raz, Shlomo. 1997. Tigré. in Hetzron, R. (ed) pp. 446-456. Wright, W. 1874 (1975). A grammar of the Arabic language. Cambridge: CUP. 492.7 CAS
14/2/02
Draft Only: Do Not Quote Or Circulate Without Permission
