Diakonoff - Problems of Root Structure in Proto-Semitic

ARCHIV ORIENTALNI 38, 1970
403
.:;.il< l!
PROBLEMS OF ROOT STRUCTURE IN PROTO-SEMITIC.
I. M. Diakonofl, Leningrad
I
The interpretation of the root structure in Semitic (and in Semito-Hamitic)
is one of the most perplexing problems of Semito-Hamitic linguistics. It is also one
of its most essential problems, because only its solution can make possible a suf-
ficiently rigorous comparative study of Semitic and Semito-Hamitic languages; in
particular a reconstruction of "proto-forms" is virtually impossible without an under-
standing of the root structure at the Proto-Semitic level, and without such recon-
struction we have no adequate method for building up a scientific comparative and
historical grammar of the Semito-Hamitic languages.
An investigation into the possibilities of a reconstruction of a Common Semito-
Hamitic linguistic prototype should naturally start from the Semitic material, since
it is only in the Semitic branch of the Semi to- Hamitic (or Afro-Asiatic) linguistic
family that we find sufficient data on the situation which existed at its Ancient stage:
though there exists a considerable amount of very ancient evidence from the Egyptian
branch, this evidence is incomplete, because only the consonantal skeleton of the
words is apparent in writing; as to the other dialects of Common Semito-Hamitic,
only their very late and historically much modified descendants are known. Therefore.
although in the following we will keep in view the possibilities of an approach to the
reconstruction of Common Semito-Hamitic phenomena, our immediate task will non&
the less be a study of Common S emit i c material only. This task will, naturally
enough, include not so much the discovery of new facts, as a systematic interpretation
of facts already ascertained, mostly by other scholars.
It is well known that the Semitic roots can be neatly subdivided into verbal and'
primary nomtnal roots; in the verbal roots the vocalism has a functional role anal
changes according to certain apophonic laws which have lately been studied by
J. Kurylowicz;l while the vocalism of the primary nominal roots is in the Proto-Semitic
prototype a stable part 'of the root morpheme itself. This stability of the vocalism of
the root morpheme in non-derivative nouns is completely preserved in Akkadian and
can be reliably reconstructed in the Northern Central (North-Western) Semitic languages,
where the variations in depending upon status and number of the noun are
secondary.2 This stability is less obvious in the Southern Semitic languages, in the
first place, because of the profuse growth of the system of Pluralis fractus, and then
* The paper read at the session of the Semitic Cercle in January 1967 (see ArOr
36, 1968, 471).
1 J. Kurylowicz, Esquisse d'une theorie de l'apophonie en semitique, "Bulletin de Ia
societe linguistique de Paris", LIII, 1, 1958; id., L'apophonie en semitique, Wroclaw-
Warszawa-Krak6w, 1961.
2 The variations in the vocalism of Hebrew and Aramaic nouns depend mostly
upon the historical stress conditions at a comparatively late stage in the development
of the languages. One should note separately the infixation of -a- in the plural of
a certain set of patterns of Northern Central Semitic nouns, both primary and deverbal.
I. M.' DIAKONOFF
also because of the nearly unlimited freedom of forming new derivative verbs from
nouns, and of secondary nouns derived. from verbal roots. Besides, the stabiltty of the
primary nominal vocalism Is somewhat disturbed by the widely spread usage of
forming ;yarious word-base patterns for dtmtnuttve nouns by changing the vocalism
of the root morpheme not only tn derivative, but in primary nouns as well; this device
being observed also tn Northam Semitic languages.s
Nevertheless, the Southam Semitic languages preserve at least one feature typical
of the Proto-Semitic non-derivative nominal root morpheme, and perhaps its most
important feature, namely, the non-functional character of the vocalism (in the Sin-
gular). In other words, while the difference between the vocalism in the noun patterns
taz-, fl<l-, ftitz-, faiz-, faaz- derived from verbal roots corresponds to a difference
In their semantic and/or grammatical function, the difference tn the vocalism of various
primary nouns has no connection with their semantic.4
A system of six vowel phonemes (a, t, u, li., i, U) can, as is well known, rellably
be reconstructed for Proto-Semitic. In traditional Semitology it was thought that these
vowels are never a part of the root in Semitic, the root being reconstructed as con-
sisttng of consonants only. In the last decades, this point of view is being
more and more abandoned. At present it seems hardly possible to insist upon the
vowels being no part of the root in pronouns, numerals, and in fact, in primary nouns
generally; moreover, many Semitologists are now inclined to consider it possible for
a vowel be part\ of a verbal root.S But also at present there exist adherents
of the theory asserting that vowels are not and never have been part of a n y Semitic
1:oot morpheme.6
s What is meant here, is mainly internal inflection after the patterns C
1
i/uCl1C
3
-,
C
1
uC
2
ajC-, as e. g. Akkad. $Ul)ar- 'boy ( = servant)', l}ul)ir- ( < *l}ul)ajr-) 'youngster', cf.
l}e(J.r- ( < *l}al)r-) 'small'; Arab. kulatb- 'doggie', umatmc 'mummy'; the pattern C
1
i/uC
2
aC
3
-
is here no longer productive.
, 4 It is therefore completely wrong to classify together (as is still very usually
being done) e. g. Akkad. nasr- 'vulture', kal-b- 'dog', kasp- 'silver', 'akl- 'bread, food',
ward- 'slave', malk- 'prince', the first tw'O words being primary nouns with the vowel -a-
making part of the root, while the others are nouns derived from verbal roots, and the
vowel is functional in character (here it denotes a product or result of the action
of the verbs *ksp 'to break into crumbles', ,*'kl 'to eat', *wrd 'to descend', *mlk 'to
advise'; cf. also ri'm- 'wild ox', kibr- 'river bank' and contrast kisp- 'magic', milk-
'oounsel', sipr- 'work, charge' belonging to the verbal roots *ksp 'to make magiC',
*mlk 'to advise', *spr 'to send'; gammli.l- 'camel' and contrast kasSli.p-t- 'witch',
nappli.l.J- 'smith', dai{i}li.n- 'judge' belonging to *ksp 'to make magic', *npl) 'to plow
'bellows', *din 'to judge; to decide righteously'; or e. g. tn Hebrew: niisiir 'vulture',
kiiliiQ 'dog', kiisiip 'silver', agiig 'slave', and miiliils: 'king' belong to different nominal
types: the first two words contain *a as root vowel, while the last three are formed
from verbal roots: *ksp (not preserved in Hebrew; there exists a homonym), *"bd
'to work', and *mlk (not preserved in Hebrew except in Nipal; the Hebrew verb mlk
to reign' is a secondary denominative one). Even in setting up a simple catalogue of
word patterns one should not lose touch with the history of the Semitic languages.
s See, e. g., W. von Soden, GAG, 50 b, 73 c, 87 a-d, 103 b, 107 a etc.
6 Thus, A. M. Gazov-Ginzberg in one of his recent works (A. M. fa30B-fHH36epr,
C11eilbt .MOHOBOIWAU3.Ma B ce.MuTcKux BHeepa.M.MaruttecKux eAaCHbLX, KpaTl<He coo6llleHHli
HHCTHTyTa Hapo.D.OB A3HH, 86, 1965, pp. 90 sqq.) asserts that both in the Semitic verbal
and the Semitic nom'inal roots tt is only the consonants which determine the meaning,
while the vowel, if only it does not belong to the grammatical inflexion, is variable
and unstable, reflecting a primary condition of "monovocalism" when any vowel sound
was no more than a "Hilfsvokal" without a phonemic significance of its own, being
part of a "proto-phoneme" consisting of a consonantal element and ()f a neutral
vocalic 'by"sound. "Before the emergence of vocalic grammatical inflection, - writes
Problems of Root Structure tn Proto-Semitic
Obviously, the answer to the problem, what is the root morpheme to be separated
out of a word-base in Semitic languages, depends upon the analysis of the mor-
phonological structure of the word, in which alone the root morpheme can exist.
The notion "root" is certainly no more than an abstraction with no real content
unless under the conditiOns existing in real words.
Passing now to the study of the Semitic root, we will try to investigate both
groups of roots, the primary nominal and the verbal ones, separately.
II
Let us begin by considering the Proto-Semitic nominal root morphemes as a closed
system of facts, and by attempting to deduce the structurai principles of the formation
of nominal root morphemes exclusively from phenomena existing inside this system.
Of course, we shall not set up a complete list of Proto-Semitic nominal roots.
For our present purpose we can limit ourselves to the lists of the more important
nominal roots (mostly attested in all four groups of Semitic languages) which have
been published by G. Bergstrlisser, A. M. Gazov-Ginzberg, and myself.7 These lists are,
in Chapter III of this paper, classified in accordance with the phonological structure
pattems which will be established below.
Investigating the Semitic primary nominal roots as a selfcontained system, the
first thing which we discover is, that here there exists a structure of vocalism totally
different from the one reconstructed for Proto-Semitic as a whole - or, to be more
exact, for the diachronic level immediately prt:teding the division of Proto-Semitic
into four separate dialects.8 As is well known, and as has already been mentioned
here, six vowel phonemes can be reconstructed for Proto-Semitic: a, t, u, ii., rt, fl.
Gazov-Ginzberg, op. cit., p. 95 - monovocalism dominated the language as a whole".
It must be conceded that no fully satisfactory minimal pair contrasting only in vocalism
can apparently be found among the primary Semitic roots, either verbal or nominal.
However, postulating a "monovocalic" stage in the development of Semitic means raising
a number of questions that at present have no adequate answers. Thus, it remains
obscure how vocalic grammatical inflection could emerge (in the verb and in derivative
nouns) if the language as a whole was dominated by monovocalism, since it is pre-
cisely the different quality of the vowels that determines the semantic of the vocalic
inflection; or how it happened that the one proto-vowel resulted in different
vowels at the later stages of development of the language. This could happen either
because of functional differentiation - but then the way leading from monov,ocalism
to such differentiation remains to be shown - or by phonotactic reasons, in other
words, depending upon th'til position of the vowel; but the author asserts that the
proto-vowel can develop into different vowels in i d en t i c a 1 position. It seems
to me that the solution of obscure points in the vocalism of identical roots when
differently reflected in the individual Semitic languages should be sought, first, in
a correct reconstruction of the Proto-Semitic and Proto-Semito-Hamitic phonological
system, secondly, in a more_ exact reconstruction of the Proto-Semitic root patterns,
and also by taking the history of the individual language groups more rigorously into
account. It Is only after this that generalizations on the subject of the language
at its primaeval stage shol,Jld be attempted.
7 G. Bergstrasser, E-infiihrung in die semitischen Sprachen, Munchen, 1928, p. 181
sqq.; A. M. raaoB-fHH36epr, C/leabt MOHOB01Ca/IU3Ma , pp. 92-94; I. M. Diakonoff,
Semtto-Hamitlc Languages, Moscow, 1965 ( = SHL), pp. 32, 42-52, 56.
8 Here, as in my other publications, I classify the Semitic languages by dividing
this branch of the Semi to-Hamitic linguistic family into f our groups of languages:
the Northern Peripheral ( = North-Eastern], the Central ( = North-Western,
with at least two sub-groups, the Amorite-Canaanite and the Aramaic), the Southern
Central ( = Arabic), and the Southern Peripheral (South Arabian and Ethiopic].
456
. I.. M. DIAKONOFF
The situation is different in the sub-system of primary nominal root morphemes: for
this sub-system certain specific phonological rilles can be formulated, namely the
following (argumentation and illustrations are to be found in Chapter III):
1) With the exception of a few cases where vowel length appears to be secondary
(e. g. *tum- 'garlic', *ma'/i- 'water', *biib- 'door' J. long vowels are absent in the Proto-
Semitic primary nominal root morphemes.
2) While in the vowel scheme as reconstructed for Proto-Semitic generally, the
vowel u appears as a separate phoneme, in the sub-system of Proto-Semitic primary
nominal root morphemes u is an allophone of the i-phoneme appearing in contact
with the labial consonants b, p, m, sometimes also with g, q, k (when < *gw, *qw,
*kw? ],9 and in a few cases - in contact with the glottal stop '.10 Thus, in the sub-
system under consideration there seems to be only two vowel phonemes, a and i/u.
We may add that, as will be shown below, in unstressed (?] position the vowel a
lost its definite quality, and conld be reflected, in the different language groups,
either as a or as i. It is not impossible that the difference between the prototypes
of the a and i phonemes when under stress, was one of pitch, while their articulation
characteristics might have been identical. In our reconstruction of the proto-vocalism
in the sub-system in question we will use only the *a-symbol for the prototype of i/u
(without insisting on any definite phonetic significance of this notation), while re-
taining the a-symbol for the prototype of the historically attested a-vowel.ll
3] Lastly, a third phonetic rule obtains for the sub-system of Proto-Semitic
nominal root morphemes. It refers to the phonemes ', J, y, l, r, m, n. A primary
nominal root morpheme may contain either two different consonants or three, and
either one vowel or two; but if it contains three consonants and one vowel, then:
a) the morpheme invariably has the structure ClVCzC3,
b] either Cz or C3 is a phoneme of the group ', t, y, l, r, m, n;
c) if it is Cz that is a phoneme of this group, then V = a.
This leads to the supposition that here sonants of the Proto-Indoeuropean type
are to be reconstructed, i. e. phonemes which at a certatn stage in the development
of the language could function both as syllabic (vocalic] and as non-syllabic (con-
9 The Proto-Semito-Hamitic consonant series *gw, *qw, *kw seems to be established
reliably enough, mainly on the basis of Cushitic and Tchad data; but also in Semitic
there is evidence pointing to the original existence of these phonemes; apart from
their existence in Ethiopic where they might perhaps be due to the influence of
a Cushitic substratum, the fact itself that i > u not only in contact with b, p, m but
(in certain cases only!) also with g, q, k is significant.
10 For examples, see below in Chapter III.
11 One could see in this a complete analogy to the Proto-Indoeuropean "bivocalism";
it is well !mown that two vowels (in tl!.e narrower sense of the word) are usually
reconstructed for Proto-Indoeuropean: *o and *e; more convincing, however, seems
the reconstruction of *a and *a suggested by Pulleyblank who refers to the situation
in Sino-Tibetan languages (E. G. Pulleyblank, An Interpretation of the Vowel System
in Old Chinese and Written Burmese, "Asia Major", NS X 2, 1963, p. 221; cited after
B. B. HaaHOB, 06w,emtaoeaponeCictW!l, npac/taBHHCtW!l u aHaToAuCicKaH Jl3btKOBble cucTeMbl,
MocKaa, 1965, p. 21 sq.). V. V. Ivanov accepts Pulley blank's reconstruction, apparently
with good reason, and refers also to the similar situation in the very archaic North-
Western Caucasian (and, originally, also Southern Caucasian) languages; pointing out
that this tallies completely with the law of contrast between a primary compact and
a primary diffuse (an open and a closed] vowel phoneme as established by R. Jakobson
(Selected Writings, I, Phonological Studies, s'Gravenhage, 1962, pp. 358, 493, 531 ].
Problems of Root Structure in -Proto-Semitic 457
;
s::::nantal).
12
The following rule _may in this cas(! bf:l formulated (symbolizing a sonant
as S, a "pure" vowel as V, and a non-sonorant consonant as C):
in a *SV sec;uence the sonant is non-syllabic > ', 1, IJ, l, m, n, r;
in a *SC sequence the sonant is syllabic >,a', a!, a!J, al, am, an, ar
[ i. e. it resolves into a diphthong).
The sequences **VS and **SC cannot occur. This results from the general Se!JlitO-
Hamitic rule that a syllable cannot have a vccalic Anlaut. Neither can the sequences
*CSV and *SCC occur inside one syllable, which results from another general Semito-
Hamitic rule of syllable-formation, according to which nc syllable can begin with two
consonants or be clcsed by twe consonants [the sequence *CSV and *SCC being
equivalent to *CCV and *VCC ). If, however, the sec;uence in question is divided
between two syllables, then the rules stated above obtain: the syllable sequences
*CVC ISV, *CVC rev are possible, the sequences **VC I sv and **SC: cv are impossible.
As to the sequences CS* and CSC*, they constitute a case apart, which will be
discussed sepSrately below. For the present, it is to state that a *-SCC-
is ruled out even if the divisicn c-f the syllables is *CSC 1 CV.
The system of vocalism just described being obviously not in accc;-dance with
the situation recomtructed for Proto-Semitic at the diachronic level immediately
preceding its divis!on into dialects, we must assume that our inner reconstruction
[back from Proto-Semitic] reflects a certain earlier stage cf linguistic deve:opment.
However, as the six-vcwel system can reliably be reconstructed not only for Prate-
Semitic but also for Proto-Berbero-Libyan and is, therefc;-e, probably Proto-Semito-
Hamitic,B we have still tc go a step farther and to admit that our reconstructed
scheme cf vocalism as suggested for the sub-system cf primary nominal roct
phemes, i. e. seven sonants and two vowels proper, refers to a still earlier diachronic
level which we may conventionally call "Pre-Proto-Semito-Hamitic". It is in any
case net later than Proto-Semito-Hamitic, because there is evidence that a number
cf primary nouns belonging to the Common Semito-Hamitic vocabulary show a vo-
calism identical to that in Common Semitic primary nouns sub-system.14 But as the
Common Semito-Hamitic roots are not sufficiently investigated, we shall also in the
12 It is true that taking these phcnemes tc be scnants, we create a difficulty as
regards the phonetic interpretation of the prototype of the Sem. '; however, it seems
possible that the Indo-European "laryngals" may serve as an analcgy, showing as they
do a certain typological proximity to sonants. In the following, when reconstructing
the "Pre-Proto-Semito-Hamitic" forms, we will symbolize the seven phonemes of the
"sonant" group as *H, i, !J, l, m, n, r when non-syllabic, and as *If, J 11 {, l]l, 1), r
when syllabic.
13 Or, at least, common to the northern group of Semito-Hamitic languages, see
SHL, pp. 31 and 102 sq.
14 Thus, e. g. 'father': Sem. *Hab{b}- [?], Berb. abba [also ibba), both probably
< *abb- <' *'abb- [?), Cush. *abb- (according to the evidence of Agau and Somaii);
'mother': Sem. *Hamm-, Berb. imma < *'imm- (?), Tchad 'umma [if the last named
form is not a borrowing); 'man, husband': Sem. *mat-, Tchad mut-um; 'tongue': Sem.
*las- (for this reconstruction see below], Egypt. *las (thus should the Egypt. written
form ns be interpreted, where, on the evidence of Coptic, n = Ill]; 'tooth': Sem. *sann-,
Berb. Tamashek a-sin (probably < *-sin- < *-sinn-, cf. Sill) a-sannan 'thorn' < *hti-
sinn-tin-); analogous forms are attested in some Tchad languages; 'heart': Sem. *labb-,
Egypt. ib ( < *labb-), in Cushitic: Kaffa nibb-15 (but Galla labbe!]; in several Tchad
languages we find such forms as naffu, nafu etc.; 'water': Sem. *ma'l!-, *ma'/i-,
Berb. *-mti-, Cush. *-mti-; 'dog': Sem. *kal-b-, in Cushitic: Saho kare, Wiilamo kana;
Hausa kare; 'name' Sem. *sam-, Berb. i-sam < *-sim-, in Cushitic: Bec,tauye sim;
in Tchad languages are attested the forms sim, summ-o, sana, etc.
458 I. M. DIAKONOFF
following limit ourselves to illustrations taken exclusively from Semitic; it must
howevilr be kept In mind that the picture we are reconstructing is considerably older
than Proto-Semitic.
To make the following more comprehensible, one should start from the above
stated premise that a root morpheme never did and never could exist as such, outside
of a word as a fact of reality. Therefore, in order to study root morphemes, one
should consider (1) the laws of syllable formation, (2) the laws of syllable contacts.
Judging by the unanimous evidence of Semitic languages (which, on the whole,
is supported by the evidence of other Semito-Hamitic languages, in so far these New
Stage languages allow of making a reconstruction of the situation which must have
existed here at the Ancient and Middle Stages), the only possible types of syllable
were either -CV-, or ~ e v e ; we have all reasons to postulate this situation already
for Proto-Semito-Hamitic. Therefore, we may assume a similar situation to have
existed also at the "Pre-Proto-Semito-Hamitic" stage we are concerned with, although
i!n this case such assumption is to be modified to suit the specific system of vocalism
peculiar to this stage. In other words, we will assume that at the diachronic level
in question not only were syllables of the -CV- and -CVC- types possible, but also
equivalent syllable types, where either the consonant or the vowel was replaced by
a sonant. It will be shown that this assumption allows of reconstructing a consistent
theory of nominal root formation.
According to the rule formulated above (a sonant in contact with a vowel is
a consonant, a sonant in contact with a consonant is a vowel), only such combina-
tions of C, S and V are possible which exclude the clustering of either two con-
sonants or two vowels in one syllable; a syllable always beginning with no more
and no less than one consonant and never ending in two consonants. This means
that only the following sequence are admitted in one syllable:
CV and the equivalent sequences SV, CS, SS,
CVC and the equivalent sequences CVS, esc, CSS, SVS, SSG, SSS; other sequences
being impossible inside one syllable; thus a syllable of the *SC type is impossible
because here S = V and cannot begin a syllable'; a syllable of the *CSV type is
impossible because here S = C (or = V], and a cluster of either two consonants or
two vowels in one syllable is forbidden, etc.
This, however, is not enough, because not only the rules of syllable formation
but also the rules of syllable contacts must be taken into consideration. Starting
from the premise that two consonants cannot cluster inside one syllable and thus
can neither begin nor close a syllable we may formulate the following rule of syllable
contacts: not more than two consonants may cluster at a syllable boundary. Another
rule, the reasons for which will appear more clearly below, is that a sonant cannot
precede or follow a cluster of .two non-sonorant consonants; that is, sequences of the
types *CSC+CV, *CSC+CVC, CSC+ CSC, *CVC+CS, *CVC+CSC are forbidden.
As I have attempted to show elsewhere,15 a Proto-Semitic noun could exist either
in a zero case (this zero, or absolute case was used when the noun had no syntactical
connections, or when it was a predicate,16 and, at some very early stage, apparently
15 SHL, p. 57 sqq.; 11. M . .llbHKOHOB, f/31>tKU apeBHefi llepeaHefi A3UU, MocKBa, 1967
f = YDPA], p. 213.
16 Still thus in Akkadian; a similar form is assumed by the noun when governing
an attribute (status constructus); it is ,at present not quite clear whether the absence
of declension In the status constructus is a secondary or a primary feature.
Problems of Root Structure in Proto-Semitic
459
also when the noun was the subject of a state) ;17 or it could exist With a vocalic
suffixal inflection. Thus, Old Akkadian had, among others, an -U"inflection which
expressed, on the one hand, the locative, and on the other, the nominative (probably
originally the ergative j17a case, and an +inflection expressing the genitive (which,
probably, originally was a more general subjoined relative casej.18 Having postulated
that at the diachronic level which is being investigated in the present paper the
vovels t and u coincided in a proto-phoneme *a, we must, assume that at this level
there was no differentiation between the locative-ergative morph and the relative
morph; this, possibly, is corroborated by the fact that in Egyptian we encounter
a possessive construction of the transitive sentence (which is a variant of the ergative
construction with the sole difference in that the ergative case ~ the subj.ect of action
coincides formally with the genitive ).19 There are also other indirect indications of
the absence of differentiation between i and u tn Egyptian at an early stage.
20
But
17 This may be deduced from the observation that the personal affixes of the verbal
form expressing state (the stative, as in Akkadian, of which the "new perfective"
of the other Semitic languages is a later development; the qualitative-stative, also
called "pseudo-participle", in Egyptian, etc.) have developed from the direct case
of the personal pronouns, while the actor affixes of the verbal forms expressing
action cannot be related to such pronouns; from this follows that the subject of state
was originally expressed by the direct case, while the subject of action - by a quasi-
indirect, namely the ergative case (in ergative languages the case of the subject
of the state is usually expressed by zero inflection of the noun). The most ancient
Semito-Hamitic languages have the following typical features in common with lan-
guages characterized by the ergative construction of the sentence: (1) the existence
of formal contrast between the categories "action : state" and the absence of formal
contrast between the categories "active: passive" (the so-called SemitD>-Hamitic passive
is always of secondary origin and usually expresses the impersonal character of the
action or the state, not necessarily being in contrast to the active voice; it is not
an expression of the point of view of the logical object of action as opposed to the active
voice, expressing the point of view of its logical subject); (2) formal morphological
coincidence between the subject case (probably this was originally the actor case
only) ,and the locative (in Akkadian) or the genitive (thus probably in Egyptian);
( 3 J the existence in transitive verbal forms not only of a personal affix expressing
the actor, but also of personal affixes expressing the object (i.e., the subject of the
resulting state); moreover ( 4) there exist some indications that the object case
(Accusative) originally coincided with the zero case [in the noun, SHL, p. 58), or
with the direct case (in the pronouns; this direct case is termed nominative; however,
the pronouns of this form are not used directly to express the subject of an action,
but only for intonational emphasis; SHL, pp. 70-73, 86). In languages with an er-
gative construction of the sentence the object of action (or, better, the subject of the
state resulting from the action) is expressed in the same way as the subject of a state
in general. While in Semitic the connection between the direct case of the pronoun
and the subject of action is rather loose - cf. a construction like u sa imqut-su oattu
'and he, a fright fell upon him', - in Egyptian the corresponding form of pronoun
is not used for the subject of action at all.
17a It is true that the Akkadian nominative and locative differ considerably on the
syntactical level; however, this difference is probably of a later origin.
18 Thus in some Cushitic languages (Bec;lauye), in the Plural of Semitic nouns, etc.;
cf. also the Common Semito-Hamitic morpheme + used as denoting relation or pos-
session in adjectives (nisbah).
19 See on this SHL, p. 1814; 85; YDPA, pp. 246 sq., 248130.
2
0 Thus, the Semitic and Berber sequences -ku-, -gu- and -ki-, -gi- (and apparently
also *-kr-l equally correspond to Egyptian forms with palatalization of the consonant:
the pronouns Sem. -ki, -ku(y}ti(ti} - Egypt. -t, twt; Sem. *kr- 'kid' ~ Egypt. t,
where Egypt, t is (Jl:) or {/::); Berb. a-grur ( < *hti-gurar < *gurgflr-?) 'circle of
stones' ~ Egypt. g
1
g
1
-t 'council', where Egypt. g. is (g) or (g).
460
I. M. DIAKONOFF
be it as it <may, wee must certainly assume the existence 'both of zero inflection and
of vocalic in Pooto-Semito-Hamit1c.21
This means that, since the root morpheme could exist only inside a r e a I noun;
its structure must have been such as to make both a zero inflection and a vocalic
inflection of the noun possible. From the first condition it follows that any root
morpheme must either consist of one full syllable, or end in a full syllable; from
the second condition it follows that the structure of a monosyllabic root morpheme
(or of the second syllable of a bisyllabic root morpheme) must make a vocalic in-
flection possible without violating the rules of syllable contacts, i. e., the adding
of a vowel must result in two permitted syllables, viz. CV 1 C + V (or its equivalents;
CSjC+V, SSjC+V, SVjS+V, SVjC+V, SVjS+V, CSIS+V, SSjS+VJ. Thus, a root
morpheme of the type *C1VCzC3 is not allowed because, although it does not contra-
dict the rules of syllable contacts in the case of a vocalic inflection (*CIVCz!CJ+V),
lt does contradict the rules of syllable formation in the case of zero inflection.22
Neither is a root morpheme of the type *CV allowed, because according to the rules
of syllable formation and syllable contacts a sequence *CV + V is not permitted inside
one syllable, since a cluster of two vowels is forbidden; nor is such a sequence per-
mitted astride of a syllable boundary, because no syllable can begin with a vowel.
However, an analysis of the existant root morphemes shows, that a root morpheme
of the pattern CS was possible, apparently ebcause in this case the syllable boundary
would pass through the sonant, owing to its double (vocalic and consonantaO
nature: { CS I
jS+V, e. g. d7[!-
The sequence CS9F was
'blood' > dam-*
dajm-u.
thus not completely identical with the sequence CV9F;
the sonant had in this case a double function, both as a vowel and as a consonant.
This helps to make it clearer why, although a syllable of the pattem CSC (or its
equivalent) is permitted (also as root morpheme), a bisyllabic root morpheme of the
pattern CSC+CVC, CVC+CSC, CVC+CS and its equivalents are impossible, being in-
consistent with the rule forbidding the clustering of more than two consonants at
syllable boundary.
To sum up, a monosyllabic primary root morpheme can include: (1) two pho-
the first being either a non-sonorant consonant or a sonant, the second neces-
sarily a sonant; or (2) three phonemes, the first and third of them being either
a non-sonorant consonant or a sonant, and the second - either a vowel or a sonant:
(1) CS (or with S for C),
(2) C1VCz (or C1SCz or, in both cases, with S for C1 and/or for Cz).
A root morpheme may also be bisyllabic; in this case both its syllables must
conform to the same rules and also to the rules of syllable contacts, i.e. a sonant
21 Remnants of a system of case-inflection similar to Semitic are still extant e. g.
in Cushitic languages.
22 This means that all Semitic nouns of the pattern ClVCzCJ, if Cz or CJ is not
a sonant. must be classified as secondary formations [i.e. as derived from verbal
roots), and originating at a diachronic level later than the one at which the rules
stated above obtained. The same is valid in reference to nouns of the pattern C1iluSCz,
because in this combination S would, according to the same rules, be equivalent
either to a vowel, being in contact with a consonant, or to a consonant, being in
contact with a vowel; but this means that the sequence C1i/uSCz cannot represent
a primary nominal root morpheme, not constituting either one full syllable or two
full syllables of a permitted type.
461
must not precede or follow two consonants contacting over a syllable boundary; or,
in other words, if the first syllable ends in a vowel or in a syllabic sonant, then
the next syllable may be either of the type es or of the type eve; but If the first
syllable ends in a consonant or in a non-syllabic sonant, then in neither of the
syllables can the vowel be replaced by a sonant; thus, the only possible patterns are:
(3) eves (and cscs),
because ev ICS and eve liS+V, es ICS and CS 1 CS I }
I s + v are not forbidden sequences;
(4) evcvs (and CSCVC],
because CV 1 eve and CV 1 CV I' C + V, CS 1 eve and es 1 ev 1 C+ V are not forbidden se-
quences;
(5) cvccvc,
because eve I eve and eve I cv I C+ v are not forbidden sequences.
However, the pattem *CVCV is impossible, because the sequence *ev 1 CV + V is
forbidden; also *CVCCS is impossible, because even if we write *CVC 1 CS 1 }
I S+V , the se-
quence is forbidden, because the sonant S in its role of consonant follows the cluster
-CC-; by the same reason, forbidden are also the sequences *CVC 1 esc, *esc 1 eve,
*eve 1 css, *sse 1 esc, *css 1 esc, *css 1 css etc.
These five permitted types of nominal root morphemes have their equivalent
patterns: in the types (2), (3) and (4) V1 can be replaced by S, and in all five types
any C may be replaced by S.
There is one more type of primary nominal root morpheme, which can be con-
sidered to be an allomorph of the type (2). Namely, with the addition of a vocal
inflection the last consonant in the eve pattern is doubled: *C1VCz+V = C1VCzl CzV.
Of course, this phenomenon cannot be observed if the vowel of the root morpheme
is a syllable sonant, In accordance with the rule forbidding contact between a sonant
and a cluster of consonants at syllable boundary (the sequence *CSC 1 CV is Impos-
sible). For the sake of simplicity, we will denote the type ClVCz[Cz) as type (2a) CVC.
The word-base of the primary nouns in the historically attested Semitic languages
can also include word-formative suffix morphs:
-lim-, -an-, the so-called 'individualizing' suffix;
-{a)b-,23 suffix of nouns denoting harmful animals;
-[a)r-/-[a)l-,23 var. -llr-1-al-, suffix of nouns denoting animals used in production;
23 In postulating these suffixes we follow the observations of the late N. V. Jus-
manov (unpublished; cf. YDPA, pp. 4, 210). The relation between the forms -ab- and
-b-. -ar- and -r-, -al- and -1- lies on the same plane as the relation between the forms
-at- and -t- in the suffix of the feminine gender, i. e. at a certain period they must
have been in free variation. Gf. Arab. san-at- 'year' and Old Akkad. san-t- > satt-
'year'. The variant -iil-/-ar- also has its analogies: cf. kall-at- II kall-iit- 'bride,
daughter-in-law', >al)-at- 'sister'. It would probably be possible to establish the origin
of these variants, but the space does not allow us to divulge into this special problem
in the present paper. We will limit ourselves to bringing some examples of the suffixes
-b-, -r-, -l-: the names of the animals in question constitute the following series:
(a) animals of prey and harmful animals: >arn-ab- 'hare', dub-b- 'bear', gi'-b- 'wolf',
*gub-b- 'fly', kal-b- 'dog' (originally a wild one!), <aqr-ab- 'scorpion', tacl-ab- 'fox,
jackal'; (b) useful animals: >zmm-ar- 'ram', baq-ar- 'cow', *l}.am-ar- 'ass', kar-r- 'Iamb',
lal)-r- (also ral)-l-; cf. the Sumerian borrowing lal)ar) 'sheep', <aj-r- 'ass-foal', tay-r-
'ox'; (c) wild useful animals: >att-al- 'deer', *gam-{a)l-, *gamm-al- 'camel', naU-al-
'a sort of antelope', <tg-l- 'calf' (but originally 'a young antelope', cf. in Syriacl);
cf. also Hebr. ral}.el, Aram. ral}.l-0., Arab. ral)il- 'sheep'; but both the vocalism of the
462
I. M. DIAKONOFF
also the possessive S!Jffix + (ntsbah), the feminine gender suffix +, and, possibly,
a few more. In analysing the structure of the primary nominal root morpheme these
morphs must pot be regarded, as parts of the root.
III
We will now proceed to stating the arguments in favour of the rules indicated
above, and at the same time attempt to find out the reasons for certain seeming
anomalies in the formation of Semitic primary nominal roots.
24
word and the I suffix which does not fit with the wild animals, show that the
Akkadian form la?;lr Is the original. Names of animals without these suffixes should
probably be considered In Semitic either as originally their epithets, or as early
borrowings. Otherwise in the other Semito-Hamitic languages: cf. Cush. kana, kare,
Tchad kele, kare 'dog' II Sem. kal-b 'dog'; Egypt. t' 'fledgeling, cub' II Sem. kar-r-
'lamb'; Egypt. " 'ass' II Sem. 'ai-r 'ass-foal'; Cush. Bec;lauye na'i 'goat' II Akkad.
naUal 'a sort of antelope'; perhaps also Be9auye kam 'camel' II Sem. *gam-al
(and other variants 1 'camel'.
24 When reconstructing the phonemes of the sub-system of the primary Semito
Hamitic nominal root morphemes, we will use the following symbols [without for
mulating their .exact phonetic significance, which at the present level of scientific
development in Semitology would hardly be possible]:
Consonants:
!.L
m b p"l
pbJ
n d
t
t z s s
t;J.
ic)
l
$d) seJ
r
s

g"', ggJ kW, k
q: q y?hJ l].
l].
Hil h
Vowels: a a
tzz:'
u
!1 r
'
1j
.
Notes: (a) This phoneme can reliably be reconstructed for the Cushitic and Tchad
branches of the Semito-Hamitic linguistic family [see B. M. HJJJJH'I-CBHTbl'l, H3 ucropuu
tar)cKoco IWHCOHaHTU3Ma, 513blKH ApHKH>>, MocKBa, 1966, p. 9 sqq.; A. 5. )loJJronOJJb
CKHll, Marepua,1bl flO CpUBHUTeAbHOUCTOpUteCK011 1/JoHeTUKe KYCUUTCKUX R3blKOB, op. Cit.,
p. 65 sqq.]. but not at all reliably fur Semitic; cf., ho-wever, Akkad. pursil', Hebr.
par' lis, Arab. burYill 'flea'. (b) Southern Semitic, Egyptian, Berber etc. f, Northern
Semitic p. ( c] Arab. ;?: ( l; ]. Aram. t. Hebr., Akkad. ( d] Arab. rj. ( .fl Aram. g > ,
Hebr., Akkad. (l. ( e] Arab. s, Hebr. s, Akkad. (in the traditional transliteration] s,
probably (s). [f) Cf. note 43. Apart from the pronominal morphs
> Akkad. -su, -sa-, -s-ta- > -1-ta, Hebr. hil, ha, Arab. -hii, 'a-, -s-ta- J. some
nominal and verbal stems (although at present this would be hard to prove] may
appear to contain this phoneme, as e. g. Akkad. *ksd 'to overcome, to catch', Arab.
*kad 'to be near to something' etc. [*k'lf.Ld < Some peculiarities of Old
Akkadian orthography allow us to suppose that the phoneme was also heard in the
noun nis ( < 'people' and in the verb bs! 'to be'; the first root corresponds.
to Hebr. 'enos, Aram. 'iinds-d; Arab. 'ins etc., where the sIs probably was preserved
because of its final position in the syllable, cf. Arab. -st- for The verbal root
[?] may also witb time find its parallels in other Semitic or Semi to- Hamitic
languages. (g) Arab. g ( [: ). (h) The Proto-Semitic (still more the Proto-Semito-
Hamitic) origin of the Y phoneme is subject to considerable doubt; both in the Semitic
languages themselves and in the correlation between the different branches of the
family there reigns much irregularity (root variants?), e. g. Hebr. z'r 'to be small',
463
First of all, we w111 dwell upon the three phonetic rules stated above as
cally characteristic of the sub-system of primary rOQt morphemes.
(1) The absence of uowel length; Among our examples (which, as already men-
tioned, are non-derivative Semitic nouns, mainly selected from nouns attested in all
four groups of the Sem1t1c branch of the Afro-Asiatic family of languages, with the
aim of securing the representation of at least the Proto-Semitic level of the vocabulary)
only the following samples with long vowels were observed:
bli.b- 'door'. This word is absent in the other Semito-Hamitic languages; it should
be considered as a contraction of *ba'-ba'-,25 analogous to *kab-kab- 'star' > Akkad.
kakkab-, Amor. kabkab-, Hebr. Mehrl kebkib, Eth. kDkt!b, Arab. kay.kab-; cf.
*laj-lal- 'night, evening' > Akkad. lll- ( < *Zajl- ), Hebr. lajl-d (the old locative form
in -ah < Aram. Mlj-d ( <' *lajlaj-{h)ii?), Eth. Arab. lall-at-; *daj-dal- 'teat,
woman's breast' > Akkad. ( < *dald- ), Hebr. dad ( < *dadd-; <' *dald-?); Arab.
dajd- (also did-). In all the quoted instances a tendency towards contraction of the
form is observed; this is also generally typical of reduplications in Semitic.26 It should
be noted, that in some instances the nouns of this group can be connected with very
ancient verbal roots, cf., e. g., *bii' I *bil' 'to come, to enter', attested, besides Semitic,
also In Egypt. bw 'place', in Cushitic bli.{lJ 'to come, to enter', in Tchad ba{l) 'to go,
to walk'; cf. perhaps also Sem. *Ui'll 'to be weak, tired' (?).
diim- 'blood' is attested only as a dialectal (?) form in Akkadian, alongside of
dam-,27 all other Semitic languages have dam-, or at least reflect this form. There
are also other instances of an alternation a II a in the same root in different Semitic
languages: 'water' is in Akkad. mil, mamtl, Hebr. maJim, Aram. mattd, Eth. mill,
mil'-; despite of the long li also in other Semito-Hamitic languages, the vowel length
is here hardly original; the same applies also to the word 'sheep or goat': Akkad. tl,
u'u [ < *a'-?1, Hebr. ii, ( <' *aj-), Arab. M'- [cf. in 'cushitic: Somali sa 'a cow
or ox', in Tchad: Hausa sii 'bull'). Such vacillations vocalism may help to explain
also the anomaly of vowel length in the words kis- 'purse' and tum- 'garlic': they
should be regarded as the result of rhythmical Analogiebildung after the more usual
CaSC- and CVCS-patterns. Perhaps also Akkad. klis-, Hebr. kos 'bowl' belongs here;
cf., however, Arab. ka's- which may be the original form.28
Similar rhythmical reasons may be responsible for the variant 'iliih- 'god', parallel
Aram. zr 'to be few', Arab. zr 'to be scarce [of hair)'; Hebr. zr 'to be small',
Sab. iJYr, Arab. !JYr id.; Akkad. iJl:J.r 'to be small' (cf. Egypt. z 'son', where ' < *-r-
or even < *Yr-??). (i) Arab. hamzah, Hebr. 'dliip (').
25 It can hardly be a rhythmic lengthening of *bab- (cf. below on dam- II diim-,
ma'll- II mli'/t, etc.), because there are no other instances of nominal roots of the
pattern C
1
aC
1
while a prototype of the pattern *bljb would result in *ba'b-, a form
of which no traces can be found in Semitic.
26 Cf. the predominance of incomplete reduplications in the Semitic stirpes ("stem-
modifications") of the D-type (Akkad. D, R, Arab. II, IX, XI etc.) as compared to the
abundance of completely or almost completely reduplicated forms in the corresponding
stirpes in the other Afro-Asiatic languages (in Berber, in Egyptian, in Tchad and
partly in Cushitic, see SHL, pp. 100-101).
27 See AHw and CAD s. v.
28 D. 0. Edzard has in his review of SHL (RA 61, 1967, p. 149) suggested that
ka's- is a secondary formation by analogy, under the influence of the pattern ras-
'head'; there is no semantic connection, but a development of this kind may have
taken place as part of a general tendency towards rhythmic assimilation of the
pattern eve to the more frequent pattern cvsc or esc > case.
, ; L,l'-f..
to. tz-, but cf .. also 'man', Hebr. Aram. dnt:IJ-d, .. unlls-, ins-an-,
Sab. n.29
.,
At all events, primary nouns with a long vowel In the root are extremely rare,
and their vowel length Is apparently in alf cases of secondary origin.
(2) Secondary differentiation of i: u in non-derivative nominal roots. In the
material under review there are nearly no cases where the u-vowel can be attested
in the noun in question in all of the Semitic language groups, and does not alternate
with the t-vowel at least in some of them. It is common knowledge, that even In
verbal nouns, belonging to a much later diachronic level, noun patterns with i and
with u, or with i and with il, frequently alternate without essential change In the
semantic of the pattem.
In the sphere of primary nominal roots, in all examples which we have collected,
the vowel u (nearly always In alternation with i) is at1ested only in the following
sequences: (a) after b, p, m, (b] after g, q, k (apparently < Proto-SemitcrHamitic
*gw, "'qw, "'kW), (c] before b, p, m, (d] in a few cases after , only in roots of the
pattems HVCS (and possibly HSC].30 This last group (d) will be treated below in
another connection. As to the groups (a], (b] and (c), at first it seems that the
distribution of labialized and non-labialized forms among the different Semitic lan-
guages Is haphazard; it can, however, be surmised that this impression is due to the
scarcity of examples which can be used for its control,31 and also to our inadequate
knowledge of the real history of the Semitic languages and cultures at the pre-
historic level. Still, the labialized and non-labialized forms can be predicted, not
with a hundred per cent certainty, but at any rate with a high degree of probability:
1] sequence *btJ-, *gwa-, *qua-:
Akkad. bukr- 'first-born son' ( < *bak(W )r-?] ;32 bin-, also bun- 'son' (isolated ex-
amples, mostly from lexicological texts, probably a borrowing] ;33 bilr- 'well' ( < *bur-),
less frequently Mr ( < bir-; borrowing?); gulgul-t 'skull'; qutr- 'smoke'; Ju(b )bul-t-
'ear of com'.
29 The pattern *CaC- shows a tendency to further contraction in Allegro-Aussprache:
bin- 'son' > (i)bn-, 'people, > from whence also tns- with
ins-lin- as nomen unttatis and as a diminutive. Cf. the irregularities in the
word *sat- 'behind': Hebr. set ( < *Jit- ], but Pl. Mt-(}t; Arab. (i)st-, but the forms sat-
and sut- are also attested in certain combinations. (These forms are, of course, like
the form (t)st-, just so many cases of emergence of a "Hilfsvokal", and they are all
to be explained by the extreme contraction of the word-base which is so typical
of this root pattern); In Aramaic a prothetic aliip has developed, with a vacillation
in the vocalism, just as in the nouns with an original initial sequence *Ha- in the
root, on which see below.
30 Nouns with an u-vocalism but not belonging to the above categories are either
derived from verbs or of non-Semitic origin; the of kulb-ab- 'ant', attested only
in Akkadian, Is doubtful, in spite of the ancient Semitic suffix -ab- ( < *kwlb-ab-, or
borrowed?).
3
1 Which makes it impossible to take into account the influence of the positional
surroundings of the phoneme with sufficient precision and with adequate differentiation
of possible influences. '
32 That the second consonant might have been *kw is made probable by Eth.
bakwar [although this form is a result of a late Southern Semitic change in vocalism,
and does not directly continue the ProtcrSemitic *bak(W)r-; but even if we deduce
the Ethiopic form from *bakur-, the existence of the sequence -ku- may in this par-
ticular case point to a *kw in a still earlier prototype.
33 CAD s. v. Cf. bun- in Amorite and probably in Ugaritic (as in buniis- 'man' <
'son of man'?], alongside of bin-.
465
Hebr. ba!siJr 'firstling, first-born' ( < *buk[u]r-) ;34 ben 'son' ( < *bin-); brJr 'well,
pit' ( < *bur < *bu'r-) but also ba'er 'well' ( < *bt' [l)r-); gulgOl-lit 'skull' ( < *gulgul-t-);
qator-lit 'incense' ( < *qutur-t-, fern. of *qutr- or *qut[u]r-); 'ear of corn,
twig' ( < The discord of forms is obviously to be explatned by the fact
that the Hebrew language was formed by the superposition of an Amorite-Sutian
nomadic dialect (or dialects) over a strong affined substratum of the local language
of the settled Canaanites.35 The characteristic but no doubt comparatively late feature
of the development of CVzCS-patterns into CVzCVzS-pattems should be noted; it can
be compared with the Akkadian phenomenon of retaining the CVCVr- pattern while
the CVCVC/S pattern is as a rule contracted to CVCC/S;36
Aram. bu!sr-d 'first-born'; ber-d 'well', golgill{a J+d 'skull'; qitr-d, qetr-d37 'smoke';
sebbel-t-d 'ear of corn'. The non-labialized forms predominate but labialized forms
also occur (in contact with *kw, *gw? More probably, this phenomenon reflects the
historically attested mixed ethnic origin of the Aramaeans; many words may appear
to be borrowings from Akkadian; the number of proved Akkadisms in Aramaic is very
considerable);
Eth.: both i and u are reflected in Ethiopic as a, hence the Ethiopic examples
are mostly not informative;
Arab.: bikr- 'firstborn', {i}bn-,38 bin- 'son', 'well'; galag-at- 'skull',39 qitr- 'in-
cense', sunbul-at- 'ear of corn' (a borrowing from some dialect of the ancient settled
population?).
2) Sequence *kwa-:
Akkad. kull-at-; Hebr. kOl (< *kull-}; Aram. kull-; Eth. kWal; Arab. kull- 'totality, all'.
3) Seq,uence *ma-:
'Akkad. mut- 'man, husband'; Hebr. Plur, ma(im [which could be < *mut-zma or
even < *mat-zma, but cf. Amor., Can. mut- and Hebr. metlm Job 24: 12 [ < *mit-zma?].4
34 On the formation of Hebrew patterns of the type balsiJr < *bukur < *bukr-,
ba'.er < *bi'ir < *bi'r-, ra'em < *ri'im < *ri;m- alongside of patterns like 'ozlin <
*'wj.n-, 'eglil < *'igl- etc. formed according to rule, see also p. 22.
35 Similar discordant forms are found also in other Hebrew word-patterns, cf. e. g.
ganndg 'thief' < *ganniib- but gibbor 'strong man' < *gabbiir-. At a somewhat earlier
stage both sets of phenomena are characteristic, on the one hand, of Amorite dialects,
and on the other, of Canaanite (Amor. bun- 'son', Dagan 'the god Dagon', sim'O.l- II
sam'al- 'left side; north'; Gan. bin- 'son', ros- - or even rils-? - 'head' ( < *ras- <
*ra's- J.
36 W. v. Soden, GAG, 12b, 55d 6a etc.
37 The Aramaic examples are taken from different dialects, hence a certain in-
consistency in the forms quoted. The forms with e < i are mostly from Syriac.
38 On this and similar contractions see also pp. 1229, 15 and 1552, 1974, 23.
3
9 In connection with this form it might be asked - what is the prototype, *gwl-
gwl- or *gwaz-gwal-? It is to be supposed that the ,prototype of the Northern Semitic
forms is *gwaz-gwaz-, because the similar root *gwr- 'throat' (where *gw is postulated
in the supposition of the affinity of this root with the verbal root *gwr' 'to swallow',
cf. Eth. gwar'-e 'throat') results in Hebr. gar-gar-lit, Aram. gar-gar-t-d, gaggar-t-d.
As to the Arabic form [with a partial contraction of the reduplication], it may pos-
sibly continue a prototype *gwl-- It is not impossible that iii Akkadian the sequence
*gws-, *kws- resulted not in *gaS-, *kaS- as could be expected, but in *guS-, *kuS-;
besides the already cited example of kulb-ab- 'ant' we may perhaps also refer to
gangur+t- (an adjectival formation?) 'throat or stomach of animal (?]', where *-gwr-
in the second syllable > *-gwar- > -gur-, but in the first syllable this process was
prevented by the dissimilation -rg- > -ng-?
4
0 However, the form metlm might be regarded as a masoretic hypercorrection
( < *matzm?].
30 - Archiv Orientilni
466
I. M. DIAKONOFF
Note however, that in verbal nouns mi-G I C ... is always < *ma-C 1 C ... ; the prefix
mu- never alternates with *mi-; it seems that the change *ma- > mu- was Gommon
Semitic.
4) Sequence *-ab-:
Akkad. kibr- 'bank of river', libb- 'heart'; dub-b- 'bear' (borrowed?) ;41
Hebr. leb 'heart' ( < *Zibb-; legag- < *li/ublib- is probably a deminutive); dog 'bear'
( < *dub-b-); cf. the commentary to the Hebrew examples for the sequences *ba-,
*qwa-. Aram. libb-a, lebb-d ( < *Zibab-(hja?); debb-d 'bear'. Arab. lubb- 'heart',
dubb- 'bear'.
5) Sequence *-ap-:
Akkad. $Upr- 'finger-nail, claw'; Aram. tifir-d, tepr-d; Arab. f-ufr- [also ?ifr-).
6) Sequence *-am-:
Akkad. 'umm- 'mother', 'im-t- 'saliva; poison';42 sum- 'name'; cf. sum-ma 'if'
[ <
Hebr. 'em 'mother' ( < *'imm-), l!em-d 'poison' [ < *l!im-at-), sem 'name' ( < *sim-);
cf. 'im 'if' ( * ;43
Aram. 'imm-d 'mother', l!em-t-d 'poison', 'name' [ < *si/um-(h}ii], of. hen, 'in
'if' ( < ;43
Arab. 'umm- 'mother', l!um-at- 'poison', {ijsm- 'name';44 cf. 'in 'if'.44
Thus [apart from some inconsistencies, probably to be explained by inter-dialectal
borrowings or by some special conditions of phonetic surroundings which cannot be
identified because of the scarcity of examples), there are certain rules 'or distribution
of the reflexes of *a in contact with labials and labialized consonants:
*ba-, *qwa-, *q"'a- > Akkad. bu-, gu-, qu-; Hebr. bu-, gu-, qu- (but with a con-
siderable influence of an i-dialect); Aram. inconsistent[?);
* kWa-, * ma-
*-ab-
*-ap-
Arab. certainly bi-.
> probably ku-, mu- in all languages [Eth. kwa- ).
> Akkad. -ib-, Hebr. -ib-, Aram. -ib-, Arab. -ub-.
> Akkad. -up-, Hebr. -ip- (?],45 Aram. -ip-, Arab. -uf- [and -if-) ..
*-am- > Akkad. -um-, Hebr. -lm-, Aram. -im-, Arab. -um-.
The number of examples is not sufficient, but even those available prove clearly
the allophonic character of u in the sub-system of primary nominal rnot morphemes.
Let us now turn to the postulated specific character of the a-vowel in bisyllabic
roots and word-bases. Here we shall analyse the ancient bases with the word-formative
suffixes -til-, -tim-, -lin-, -ar-, as well as the primary roots of the patterns CV !CVC-
and CVICS- [CVICVIC+V and CVCIS+V). Here we must note that the specific
41 The bear is nnt an inhabitant of Mesnpotamia.
42 An inter-dialectal borrowing, or an example of an early case nf free variation
i II u? Eth. l!amot 'gall' is apparently a later derivative formation.
43 A particle of pronnminal origin. including the same phoneme which appears
in the pronoun 'him' and in the causative morphs (Akkad. -su, -sa-,
-s-ta-; Hebr. -hil, ha-; Aram. -h, ha-, 'a-; Arab. -hil, 'a-, -s-ta- ). The Auslaut of the
particle ( -m or -n) depends on the existence of mimation or of nunation in the lan-
guage in question; so, naturally, there is no labialization of *a in this word in lan-
guages with nunation [Aramaic, Arabic).
44 On the contraction (sandhi) see also below, p. 15 and 1552, 23.
45 I have not found any satisfactory example from Hebrew; but cf. 'nail, claw'
sip' parlin, a secnndary formation not yielding itself to easy interpretation ( < *?ippur-
-n-?); however, it includes the sequence -ip-, although we do not know its relative age
h this particular word. Cf. also qip_.Jor:]. 'hedgehog' (?).
467
phonotactic conditions reigning in Akkadian and Aramaic do not always allow of
distinguishing the patterns CVCVC and CVCC, and sometimes it is also difficult to
decide, whether the original semantic of the root is verbal or nominal; naturally,
if in a CJVCzCJ pattern neither Cz or GJ is a son{lnt, or if the sequence -VCz- is= a s ~
then there should be no doubt as to the verbal origin of the noun in question.
First we must separate the group of root morphemes in which either C1 or Cz
is one of the phonemes ' or ?z; these, as is well known, tend to give an a-colouring
to the neighbouring vowe].46 This group includes: ?za/abl- 'rope' (Akkad. 'ebl- < *?zabl-,
Hebr. ?zQ"Qiil, Aram. ?za/;11-a, Eth. ?zabl, Arab. ?zabl- J; *?za/aql- 'field' (Akkad. 'eql- < *?zaql-,
Aram. ?zaql-a, Eth. ?zaql, Arab. ?zaql-]; *ta'm- (Akkad. tem- 'understanding'; 'decree',
Hebr. ta'am 'taste', Aram. ta'iim-a 'decree', Eth. ta'm,47 Arab. ta'm- ],48 *la?z!/IJ.- 'chin,
beard' (Akkad. la?;!tl,49 Hebr. laM, Arab. la?zj- 'jaw', all apparently < *laM- but cf. also
Aram. Targ. ZO?zd < *laiJ.?z- < *la?ziJ.- (?],50 and also Arab. li?zj-at- 'beard' < *liM- which
may be the original variant (?]; *'a/aqr-b- 'scorpion' (Akkad. 'aqrab-, Hebr. 'aqra/;1,
Eth. 'aqrab, Arab. 'aqrab-, all apparently < *'aqr-{a)b-, but cf. also Aram. Syriac
'eqar[a)"Q-0.;51 ta/a'/-b- 'fox' (Akkad. seleb-, Arab. ta'lab-, but cf. Aram. ta'al-d < *ta'l-;
Hebr. stl'til < *tu'til- is a diminutive]. Vacillation is attested in the root of the word
'handful': Akkad. 'upn- ( < *?zupn-; the form ?;!apn- is a bormwing], Hebr. ?zopiin
( < *!lupn-], Aram. ?zupn-a, Eth. tzafn, ?zafn, but Arab. ?zafn-at-. Cf. also * sa/a'r- 'hair'
(Akkad. slir-t- - not *ser-t- as one would expect, but this is not without parallels -,
Hebr. se'dr < si' { a)r-, Aram. sa'r-a, Eth. sa'r- ( < *sa'r-] 'grass', Arab. sa'r{-at)-,
sa'ar{-at)- 'hair', si'r-at- 'do., of the pudenda'.
Then, apparently, a group of roots beginning with the sequence H ( > '] + unstable
vowel ('a-, 'i-, 'u-] should. be separated. It has so far not been possible to explain
the origin of this group; it may be connected with a weakening of the initial hamzah,
and with the resulting reduction of the following vowel under the conditions of
an emerging phrase-contraction of the sandhi-type?52
46
Therefore it is difficult to establish the original quality of the vowel, and so
in the reconstructed form the notation a/a is used; only in 'handful' the form to be
reconstructed is certainly *?zapn-, and in 'chin', possibly *Za?zi/IJ.-; and, still more
hypothetically, *'aqr- (??] in 'sco-rpion'.
47 Probably < *ta'm-.
4
8 This root (which is Common Semito-Hamitic: Sahc ta'am 'to swalluw' might be
a borrowing from Semitic but cf. also Be<;lauye tam 'to eat' and Hausa oan-oana 'to
taste'] is also attested as a verbal rnot, but possibly the nominal meaning is the
original one.
49 The irregularity of the form (*?z should result in ', not in ?;!) may mean that
this is a borrowing.
50 A variant root?
51
The sequence -CS- at morpheme boundary (the suffix beginning with -C] seems
to result either in -CaS I + C-, or (if there are favourable conditions as to the rules
of syllable contact], also in -C I Sa-l C-; perhaps stress conditions could play a role.
52
Another explanation might lie in the assumption of a labialised laryngal,
a phenomenon encountered in Proto-Indo-European. The root *HIJ-t 'woman' does
strictly speaking not belong here (pattern CSC]: Akkad. 'ass-at- 'wife' f < *'ant-at-],
Hebr. 'is sa (probably < *'ass-at- < *'ant-at-], Aram .. 'attat-a, 'atta ( < *'ant-at- J but
also 'itt at-a ( < *'int-at-?], Eth. 'anas-t, Arab. 'unta. The Arabic form is indubitably
a recent one, cf. Arab. 'unf- for 'nose' (alongside of 'anf-] while all the other Semitic
languages have forms continuing the prototype *'anp- < * HIJ-p-. The Aramaic 'ittat-a,
intat-a may, although with some difficulty, be traced back to a form with initial 'a-,
as well as the Hebrew form, but this 'a- also may be secondary, cf. the Hebrew
Plural na!Hm and cf. notes 83-84 in connection with the root *Hnp-. On the other
hand, such forms as 'iSsa might have developed under the influence of 'ls 'man' <
30*
468
I. M. DIAKONOFF
*Hahl-: Akkad. 'liZ- 'town, Hebr. 'ohiil ( < *'uhl-] 'tent'; Aram. Syriac
tahl-a 'nomadic tribe, band'; Arab. 'ahl- 'clan, people' (a > a under the
double influence of H- and -h].
*Hagn-: Akkad. 'uzn-; Hebr. 'ozan ( < *'ugn- J; Aram. 'udn-a and 'idn-a, 'edn-a;
Eth. 'azn; Arab. 'ugn-.
*Han[-: Akkad. 'int-, 'unt- 'utensil'; Hebr. 'onl 'vessel' ( < 'unJ-]; Arab. '!nii'-
'vessel' (a diminutive!].53
After the exclusion of these groups of nominal root morphemes, all the other
:primary nominal root morphemes of the patterns CVjCVC- and CVCjS-, as well as
the roots with the ancient word-formational morphs -liZ-, -lim-, -lin- and -lir- may be
subc;livided into two more groups: (a] with a stable vowel in the first syllable, and
(b] with an unstable vowel in the first syllable.
Cases of vacillation between u and i are classified with the first group, because
u is regarded as an allophone. To this group belong:
*baHr- 'well': Akkad. bur, ber-; Hebr. b6r, ba'er; Aram. ber-a; Eth. ba'r; Arab. bi'r-;
.*bak{w Jr- 'firstborn': Akkad. bukr-; Hebr. balfor; Aram. bu!fr-a, Arab. bikr-;
*dabas- 'treacle, honey': Akkad. [ < *dips- < *dib(ajs-J ;54 Hebr. dagas
(< *dibas-); Aram. dubS-a[< *dibas-(h}iiJ, Arab. dibs-;54
*anab- 'fruit': Akkad. 'inb..54 'fruit'; Hebr. enag ( < tnab-); Aram. tjenb{a J-t-a;
Arab. tnab- 'grapes'.
*qwatr- 'smoke': Akkad. qutr..55 'smoke'; Hebr. qatar-at 'incense'; Aram. q/tr-a56
'smoke'; Arab. qitr- 'incense.s6
*ra.Hm- 'wild ox': Akkad. rlm-, Hebr. ra'em { < *ri' (ijm-);
also ram-a) ;57 Ar_ab. ri'm- 'the oryx antelope'.
Aram. rem-a (but
:!
*(apr- 'finger-nail, claw': Akkad. $Upr-; Aram. tUepr-a; Eth. $afr; Arab . . (.and

*tah-(lim- J 's.ea, sea-side'; Akkad. ti'lim-t-, tam+; Hebr. tah6m ( < *tihlim-]; Arab.
tihiim-at-.
*l).am-{lir-} 'ass': Akkad. 'itner-; Hebr. l).Qm6r; Aram. l).amar-a; Arab. l).imlir; all <
*l).imiil".57a
It will be noted that the stable vowel of the first syllable is, in the examples
cited, *a = i/u.58 However, in contact with *H = ' also *a may appear as a stable
from an entirely different root, and the Aramaic variant 'ittata alongside of
'attata may have been influenced by Canaanite dialects.
53
Eth. naylij, probably < *yanlij < *'unlii is, if this reconstruction is correct,
an ancient diminutive, like the Arabic form.
54
A case of contamination of the patterns CVCVC- and CVCC- mentioned above.
55 Dissimilation t > t under the influence of q according to a general rule in
Akkadian.
5
6
It is not impossible that the word belongs to a verbal root. In Ethiopic and
Arabic we find ancient diminutives (with dissimilation]: Eth. qatlir-e, cf. Arab. qutlir-
'aloe-wood' (for incense], qutar- 'mist, cloud'.
57 An anomalous form, but in any case connected with the influence of ' on the
vocalism.
57a However, Akkad. Ass. dial. 'emar- might be < *l).am-ar-!
58 A stable vowel is also attested in the word *agw-[- (the suffix brings it also
into this group as regards its syllabic structure]. Cf. Hebr. egtil ( < *tgl-] 'calf',
Aram. egl-a 'calf; antelope'; Arab. igl-; in Ethiopic agwu, a secondary formation.
The Akkadian 'agtil- 'an equtne animal' has nothing to do with *agw-z-.
vowel: *Hat-iln 'she-ass': Akkad. atlin-, Hebr. 'dtiin ( <' atlin- ), Aram. lltdn-d, Arab.
'at lin-; *Hatar- 'place': Akkad. ar- ( <' *atar- ), Aram. atr-d, Eth. asar, Arab. atar-.S&
In the following examples the vowel is unstable:
*baraq- (?) 'Ugthntng': Akkad. berq- ( < *btraq-?); Hebr. Mrdq ( < *baraq-), Aram.
barq-d, Arab. barq-.59
*gwar-{lin- ]60 'troat': Akkad. girlin-, Hebr. glir6n ( <' gar-lin-), cf. the reduplicated
form garglir-lit ( < *gargar-t- ], Aram. Targ. (Hebraism} gdrlJn-d, Arab.
girii.n-.
*(/.llkar- 'man, male': Akkad. zikar-, Hebr. ( < *gakar- ), Aram. dtkr-d, Arab.
gakar-.
*ganab- 'tail': Akkad. ztbb-at- ( < *ztnb-at- <' *r;Jtnab-at- J; Hebr. zdndQ; Aram. dunb-d
[ < "'dubb-d < *dibb-d < r;Jinb-d < ginab-{h)li]; Eth. zanab; Arab. r;Janab-.
*gaqan- 'beard': Akkad. ziqn- ( < giqan-}; Hebr. zdqdn ( < *gaqan-); Aram. diqn-d,
l:laqn-d; Arab. r;Jaq{a)n-, giqan- 'chtn.61
*kanap- 'wing': Akkad. kapp- ( < *kanp- < *kanap-); Hebr. kdndp ( < kanap- J;
Aram. kin{a)p-d (< *kinap-); Eth. kanf; Arab. kanaf-.
*las-{iin-) 'tongue': Akkad. ltSlin-; Hebr. ld6n (< *la-lin-]; Aram. le!-dn-d;62
Eth. laslin; Arab. ltslln-.63 \
*batn- 'womb': Hebr. blitlin ( < *batn- J; Aram. bi/etn-d; Arab. batn-.
*nasr- 'vulture': Akkad. nar-; Hebr. nlislir ( <nar- J; Aram. nesr-d ( < *niSr-);
Eth. nasr; Arab. nasr-.
"'ragl- 'foot': Hebr. rliglil ( < *ragl- J; Aram. regl-d; Arab. rigl-.
'intoxicating drink': Akkad. stkar-; Mktir ( < *tkar-); Aram.
Eth. sakllr;64 Arab. sakar-.
saa As we shall see below, an original a in the first syllable should be unstable; this
probably means that here the prototypes should be reconstructed as *Hat-lin-, *Hatar-,
with *a > a under the influence of *H.
59 The example Is not altogether reliable. The Akkad. berq- may be regarded
as the result of a contamination of the types CVCVC and CVCC; in Aramaic we would
expect *berq-d, but Irregular alternations of the patterns CaGe- and CiCC- in Aramaic
are nothing out of the common owing to reasons stated below, p. 22. However, also
1n Phoenician we encounter *barqii. (apparently feminine} instead of an expected
*baraqli. It is possible that all the forms in question except perhaps the Hebrew
sh.1uld be traced to the derivative verbal root *brq, which could be an explanation
why tht .word does not fit into the of correlations for the pattern in question.
Cf. also the Lldvative nomen instr. in Ethiopic: mabraq.
60 The root is ht>re in the G
1
aG
2
[G
2
= S] degree, not *G
1
S, because in connection
With the suffix a forbidden !:Jiconsonantal Anlaut would arise; at least, this must be
the explanation of the form unless we decide that the suffixation is later than the
process *gwr- > gar-. Cf. the root variant gwar- (or a secondary verbal root *grr]
In the Hebr. ger-d 'cud', and the root variant *gwr In Hebr. garglir-lit, Aram. gargar-t-d,
gaggar-t -d 'neck, throat'.
61 In Arabic the phenomenon, already mentioned in connection with Hebrew
examples, of expansion of the pattern CVCC- (including< *CSC-, *CVCS- J > CVCVC-,and,
conversely, a contraction of the pattern cvcvc- > cvcc-, is common; however, the
secondary form usually exists as a variant of the main form; hence the variants:
gaqan- II r;Jaqn-, 'ur;Jn- II 'ugun-, II etc.
62 The reduplication is here probably the result of a rhythmical assimilation of
the pattem *GaG to the pattern *CaCC, -lin- still being felt as a suffix not belonging
to the root.
'63 See note 14.
64 An irregular [verbal?) form.
470
I. M. DIAKONOFF
The distribution of the vowels between the individual languages can be illustrqted
by a table:
Akkad. Hebr. Aram. Eth. Arab.
*gwar-{an-)
a i
*las-{ an-} a
*r;J_akar-
a a
*r;J_anab- a a a
*r;J_aqan- a i, a a, i
*kanap- a a a
i [i) a
* { baraq-} (i? l a [a?] (a?)
"'batn- a a
*nasr- a a a
*rag(
a
This table shows clearly that Akkadian, Aramaic and Ethiopic are typical
"i-dialects". and Hebrew a typical "a-dialect"; sporadic forms: Akkad. kapp-, nasr-,
Hebr. selsiir, Eth. zanab can be explained as results of interdialectal contacts, and
Aram. daqn-6. (and barq-6.] also as a result of contamination of forms, since the
form daqan, baraq may belong as well to *diqn-6, *birq-6. as to
*daqn-6, *barq-6.. The picture in Arabic is man complicated; however, the rule seems
to be that the first vowel is i if the second vowel is long but a if the second vowel
is short. Forms like r;J_iqan-, rig/- may be due to borrowing, perhaps from some South
Arabian dialect.
The stated situation is easiest explained in the supposition that the most ancient
stress lay on the second syllable (including the syllable formed by the final scnant].
In these conditions *a was preserved as i (or, in contact with a iabial or labialized
consonant, u ), while *a had a tendency towards reduction, which resulted, in the
different dialects, either in i or in a.
r 31 We can now give illustrations to the five main patterns of primary nominal
rc:Jt morphemes as formulated in Chapter II; at the same time they will illustrate
the third cf the main phonetic rules obtaining fer the sub-system of such morphemes:
the double {syllabic and non-syllabic} quality of the sonants ', L lJ, I, m, n, r.
For considerations of space economy we will limit ourselves tc reconstructed
forms, citing the historically attested forms only in cases when they cannot be deduced
from the prototypes with the help of such phonetic rules as are commonly known
to Semitolcgists, or those formulated in Chapter II above as specially referring to the
sub-system of primary root morphemes, and when additional factors have to J'Je taken
into account.
(1) Pattern CS: *m!j-65 'water'; *s!j-65 'sheep or goat'; 'a(foa1';66 *ty-r-
'bull, ox; *di-t- 'door-fold; *k!-b- 'dog';
6
6 *dTJl- 'blood'; *l:z'!l- 'father-in-law;66a S!l-t-
'year'; *krcr-67 'lamb'; with reduplication; *blj-b/j- [ > *btib-J 'door'; *d!-d!- 'teat,
woman's breast'; *g''r-g"'r-t- 'neck, throat' ;
68
*lj-lj [ > *la!l-J 'night, evening'.
65 See p. 11 and note 14. A Common Semito-Hamitic root.
66 Cf. note 23. Cf. Egypt." 'ass' [where
1
< *-r or *-' 11-iJ.
66a Dubious; cf. Arab. J:wm{1- and see below on *Hab{bj- and *Hal]{l]j-.
67 Akkad. kerr- ( < *karr- < *kr-r-), Hebr. kar ( < *karr- < *kr-r-). Cf. Egypt. t
'fledgeling, cub' (where t = /k'/ or /1';/, and
1
< *-r).
68 Cf. note 39.
471
(2) Pattern eve: [a) *Hat-(an-} [ < *Hat-(iin-J] 'she-ass',
*:;am- {a j[-70 'camel'; *g''ar-{ an- J 'throat' ;BB * jad-71 'arm, hand';
'tnngue'; *qas-t- 'bow;12 *sap-t_7; 'lip'; with reduplication:
*kakkab- I 'star'.
*buq-(ajr- cow, cattle';69
*lul]-r- 'sheep'; *Zas-(an-}
*kub-kab- ( > *kalj.kab-,
(b)
74
*Hnl- 111 'ilah-) 'god'; *ban- 'son'; *rjal/-b-75 'wolf'; *l;lam-t- 'poison'; mat-
man, [?) 'people'; *sam- 'name'; *sat- 'behind'; *tah-(am-} 'sea, sea-
s'de'; */:lam-{t.r-j 'ass' [cf. also the numeral *tan- 'two', and also 'iam!n 'if'); with
reduplication: *g:al-g'-"al-t- 'skull'; compound root: *san-ba/-t- 'ear of corn';75a cf. also
the pattern eveeve.
(c) *kl;ls- (?) 'bowl';76 *rl;ls- [?) 'head';77 *$/;ln- (?JIB 'goat and sheep'.
(d) 'house'; *IJt-. 'lion';79 *'Jn- 'eye, source'; 'sjb- 'gray hairs, old age;
old man, elder'; 'urine'.
(e) *!ifm- 'day'; qys- 'bow.BO The *myt- 'death', and 'voice' may
be derived from verbal roots (?); tlfr- 'bull, ox' should be analysed as 1\f-r-, see above
tm Only in the Palestinian Aramaic dialect is a form buqr-Q attested, probably
under the influence cf the contamination of the eveve and evee and of eaee and
Ci/uec patterns usual for Aramaic.
;o This word is attested in the variants *gam-1- and *yam-a/-: Hebr. gdm6.1 <
*gam-a!-, A ram. gam/-6. ( < *gam-a/- or < *gam-1-), Arab. gam/- and gamal-. The
Akkadian form does not fit intc the pattern (gamma!-, cf. the borrowing into Sumerian
ANSE.GAM.MAL); in any case, the pattern eac- [gamm-) is an allomorph of the pat-
tern eae-.
7
1 Akkad. 'id- [ < *!ad-), Hebr. jad ( < *!ad-), Aram. tag-6., 'id-6., Eth. 'ad, Arab.
ia.'J-. This word exhibits a number of irregularities in several Semitic languages, and '
c:1e might surmise that it is an expansion of an original one-consonant root *d-,
analogous to *p- 'mouth', which roct is also frequently expanded, cf. Eth. 'af, ::;ushitic
cf, Hausa 'afi etc.
7
2 Cf. also Arab. qalj.s- ( < *qalj.s-) - a root variant?
73
The isolated Arab. sifat-, alongside of the regular safat-, might be an ancient
diminutiv3. Does the Aram. sept-a. also belong here? The Hebrew st. constr. dual.
sipte is, cf course, < *saptaj.
7
4 The nouns cf this pattern tend to contraction in rapid speech ( Allegro-Aus-
sprache), hence the Arabic forms {i}bn-, (i}sm-, {i}st-, (iJin-; cf. also note 29 abuve
en the fate of the root [?). It is but natural that the word 'god' should be less
s-ubject to such contraction, cf. however Arab. allah 'Gcd' < *al-'ilO.h-.
Is Heb1. [ < *rji' (i Jb-], cf. ba'er, ra'em. See pp. 22 below.
;sa Cf. Eth. qwznfaz, Arab. qunfug- [Hebr. possibly qippog?) but Aram. qupd-d
'hedgehog' ( < *qupparjd?).
- See ncte <8.
77
Akkad. res- ( < *ra's-), Ugar. ris /ra's-/, Hebr. r6s ( < *ra's- ), Aram. res-a,
riS-ii, Eth. ra's, Arab. ra's-. The Aramaic form might be traced to *ra's- or *ri's-,
the Etniopic probably only to *ri's-, although forms resulting from *ra's- also exist
in this group nf Semitic languages. In Hebrew are attested alsc the derivatives r1s6n
I< *ri's-tln-) and res1t ( < *ri's-it- ). Such forms are probably late and belong to some
"i-dialect", and perhaps it is to the influence of such patterns as *ri's-an- formed
by analcgy to girt.n-, lisiin-, etc. that the variant *ri's- is due, which we encounter
alongside of *ra's-?
iB Akkad. $l!n- ( < [?]; possibly < under the influence of s?); Heb_r.
$6n. Aram. 'an-d [ < *ga'n-(hjii), both regular; but Arab. c;ta'in- secondary?
79 *Zajt- 'lion' is perhaps affined to *la'-b- 'lion' (which may be supposed to be
the original form from whence the historically attested lab'- has ist origin, cf. note 23
and 96; this is an "areal" word, cf. not only the Egypt. rw - read /lalj./? - but also
Greek leon, Latin l2o); cf. the numerous alternations ' II /, 1 !I IJ. both in nominal and
verbal roots. However, Prof. D. 0. Edzard has suggested that the original form is *najt-
{attested in Akkad. nes-) while lajt- is formed by analogy to lab'-.
80 Only in Arabic [qalj.s-); a borrowing, or a root variant to qas-t-?
472 I. M. DIAKONOFF
(f I I have found no reliable examples fo.r the root pattern esc with l for s.at
[g) *qT[I!z- 'flour'; *sT[I- (?]82 'sun'.
(h) *H1)p- (?]83 'nose'; *H1)t- (?] 'woman, wlfe'.84 Both examples are subject to
doubt.
[i] *Hr$- 'earth'; *rq-tl5 'green, yellowness'; *mrH-,86 'rs- 'couch'; *qrb-87 'in-
testines'; *qrn- 'horn'; *zr'- 'seed'.
[2a) Pattern evC: [a] *Hab{b}- [?]88 'father'; *Hal){l.J/-88 'brother'; baqq-
kall-at- 'bride. daughter-in-law'; *kapp- 'palm of hand', 'amm- 'uncle'; 'forefather';
'tribe'; *sarr-89 'chief'; *$arr- 'rival, enemy'; *amm-89 'grass'. Here probably belong
also * HaU-al- 'deer', *naU-al- 'a sort o.f antelope';9 cf. ncte 70 en *gam-al-.
(b) *Hamm- 'mother'; *Zabb- 'heart'; 'tree', 'Iog-wood'; *ann-92 'tooth';
*tall- 'shadow, shade' (perhaps derived from a verbal rcct).
(3] Pattern eves: (a) *HahZ-;93 *ba'l- 'lord (of the house]', 'husband'; !zabl- 'rope';
*l)aql- 'field'; *ragl- 'foot; *ta'l-ab- 'fox';"'ta'm-;95lfr'atm- 'bone'; *batn- 'womb'; *nasr-
'vulture'. The vocalism depends in several cases on the contact of the vowel with
81
The root quoted in SHL, p. 32 does not exist; it should be read $il'-;
the for.rn is derivative [from a verbal root *?I').
82 The phonemic structure of the root is not clear: Akkad. ams-, Hebr. iimiis
[ < *sams-], Aram. sims-a, samas (see p. 22 below], Sab. sms, Arab. sams-; the proto-
type is probably *51[1 [Bergstrasser] or [with metathesis in Arabic] *sT[Is.
8
3 The phonemic structure of the rcot is subject to doubts; possibly, the crigin
of the word 'anp- [in Arabic also 'inf-, 'unf-, alongside of 'anf-] is similar to that
of 'ins- [ < *nis- < *nis] through the zero degree of vocalism, because the root
is prcbablv in affinity with a whole cluster c.f verbal roots with the primary cell *np
connected with the semantic of 'blowing', like npl) [the latter, however, possibly
< *pill)). Also, it is not clear whether 'anp- is connected with p- 'mouth', with Egypt.
fncj 'nose' and
84 In regard to this root doubts similar tc these expressed in the preceding ncte
may be 8xpressed; cf. Hebr. Sg. 'iss-a ( < 'ant-at-}, Pl. nas-lm ( < *na(-ima).
85 This root is attested also as a verbal one [*(lrq]; hence it is difficult to decide
whether such variants as Arab. (laraq- 'left', Hebr. jaraq 'green', 'vegetables' alongside
of Akkad. (larq-, Hebr. !iiriiq- ( < *!arq-], Aram. tarq-a '[yellow-)green', or conversely,
or both, are derived from the verb. The root is Common Semito-Hamitic [Egypt. i'q-t
'vegetable', Berb. i-uray 'became yellcw'].
83 Meanings: Akkad. mt:r- < *mar'- 'son', Aram. mar'-a [the most ancient form],
mari-a etc. 'lord, master', Arab. mar'- 'man'.
Bl Arab. qalb- 'heart'; cf. Egypt. q'b- 'intestines'.
83 The phonemic structure of these roots is not clear: Akkad. 'ab- but with zero
inflectir.n 'abu, 'abi, Pl. 'abb-ii (and similarly with 'al)-; fern. 'al)-M- ); Hebr. 'til!
[ < *'ab- J but in bound form 'abii-; Aram. 'aQ-, 'ii!2-a, 'ab{b }-a, Pl. 'aQah-at-a; Arab.
'ab-, 'abil; 'al)-, Pl. 'il)(liln-, Eth. 'al.Jw- and others; in the languages of the other
branches of the Semito-Hamitic family, forms with reduplicated -b- ar8 attested in Berber
and Cushitic; Hausa 'uba. Several solutions are possible: ( 1) root * Hab- with rhythmical
assimilation to the pattern eaee-/eaee- ( cf. *'amm- 'mother'); ( 2) root *Hob- with
assimilation -by- > -b{b}-; (3] suppletive root *Hab- II *Habb-. Besides 'ab{bj- and
'al){l.JJ- another noun of the same type is Akkad . .,i${$}- 'tree', see below note 91.
Cf. Arab. (lam{yj- 'father-in"iaw', (lam-tit- 'mother-in-law'.
89 Common Semito-Hamitic roots: Egypt. sr, Hausa sar-k-; Egypt. sm etc.
90 In Akkadian only ( alsc- in the Cushitic language Bec;lauye: na'i 'goat').
9! Suppletive only in Akkadian and Ethiopic: Akkad. II 'i$$; Eth. falj, Pl. 'aljay;
in the other Semitic languages formed according to the pattern eae-: Hebr. ( < *'i$ ],
Arab. 'ilj-at; but Aram. 'a'-a < *'a'-a < *'ig-{h}ti.
92, Common Semito-Hamitic.
93 See p. 17 above. Probably < *Hahl-.
94 Gazov-Ginzberg (CJie!l.bi .\IOHOBOK8JII13\Ia .. , p. 9315] refers to a Hebr. *bitn-
which is unknown to me; the form with a suffix, e. g. bitn-1, is, of course, < *batn-i.
9
5 A Common Semito-Hamitic roct, possibly verbal, cf. note 48.
Problems of Root Structure in Proto-Semitic 473
the consonants ', h, ?z,
0
More obscure are: *labH-96? 'lion'; *laM/IJ, or ,perhaps,
better *laMIIJ 'chin, beard' .97
(b) *Hani- 'vessel'; *batm-, var. *batn- 'pistachio nut'; *raHm- 'wild ox'; *Hal}.n-
'ear'; *ba'r- 'well'; qwatr- 'smoke'; *tapr- 'finger-nail, claw;98 sa
0
r- (?) 'hatr.99
(c) Not clear is the vocalism of the following roots belonging to the same
pattern: *kVli- 'vessel, ship'; *pVri/'- 'shoot, scion, offspring';100 *cvqr-ab-101 'scorpion'.
(4) Pattern eveve: (a) *Hatar- 'place' ( < *Hatar- ?); *baraq- (?) 'lightning';
*l}.akar- 'man, male'; *l}.aqan- 'beard'; *kanap- 'wing'; *lahab- 'flame' (derivative from
a verbal root?); *!iakar- 'intoxicating drink';102 *!iamai/'- 'heaven, sky.103
(b) *daba- 'date treacle, honey'; *canab- 'fruit, grapes'.
c) The existence of primary nominal roots of the pattern eaeae must be regarded
as doubtful.104
(5) Pattern eveeve: see under eve.
The patte11n eveeve seems always to be the result of a compounding of roots;
it is very rare, the roots belonging to it tend to be adjusted to other more frequent
patterns, and the original type is often preserved in one or two languages only;
the pattern is otherwise reflected irregularly.
IV
It goes without saying that the list of primary nominal roots in Semitic is not
limited to the cited above, especially if we add roots which are preserved only in one
or two of the historically attested languages, and therefore cannot so far be proved
to date from Proto-Semitic or even from Proto-Semito-Hamitic. However, the examples
in Chapter III will, probably, give a sufficiently clear picture of the validity of the
rules of phonological structure in the sub-system of primary nominal root morphemes,
as formulated in the form of theses in Chapter II. The number of roots with other
patterns of structure that might claim to be primary nominal roots is very limited.
But it may be expedient to dwell briefly on some anomalous and doubtful forms
historically attested in the individual Semitic languages.
We have already noted some processes which were called forth by rhythmic
analogy (*dlJl > dam- > dam-; *mlf.- > *ma'- > *ma'-; *slj- > *sa'- > *sa'-, etc.).
An important circumstance which could bring about changes in the primary vocalism
9
6 The root is reflected not quite regularly in the various languages; Akkad. *lab'-
(and later, according to rule, > lab-, labb- ), Aram. lab'-, but Hebr. lab'i', Arab. lab-llt-,
labil'-at- 'lioness'. A metathesis < *la'-b-?
97 See note 46.
9
8 On these roots see above, p. 1334, 15 and 1761. The Akkad. :jupparu is an in-
frequent in Akkadian but quite regular occurrence of Pluralis fractus.
99
The Akkadian form ar-t- reveals no traces of c (we should expect *er-t- or
*e'er-t-); however, similar phenomena are also observed with some other Akkadian
roots containing c in the Proto-Semitic prototype, e. g. 'aqrab- instead of *'eqrab-;
it should be also taken into account that the primary nominal character of the root
~
by no means certain. Cf. p. 15.
1
00 The attested forms are ambiguous from the ety.mological point of view, and
the roots are possibly verbal; of these at least the second root is Common Semito-
Hamitic.
101 See notes 46 and 51.
102 On this group of roots see p. 17 above.
1
0
3
The vowel length is here apparently due to analogy with *mlj II *mJ 'water':
both words are used as pl. (or dual.) tantum and even rhyme. Akkad. !iama'-il II
!iamam-il, !iamii; Hebr. !idmaiim (dual); Aram. a maUd; Eth. samai, Arab. sama'-;
cf. sama 'above', 'rain' etc. in the Tchad languages.
104 See p. 22-23, 28 below.
474
I. M. DIAKONOFF
seems to have been cases of zero inflection. While the CVC pattern (with a "pure"
vowel) would result in CVCo!F, and the same happened to the eve pattern, the patterns
with sonants (CS, CSC, CVCS) could be subjected to certain changes. Theoretically,
the patterns CS and eves should not cause any difficulty, because S is here syllabic
and, consequently, these patterns should yield such forms as CS9t > CaS9t, eves* >
CVCaS9t)05 This is, in fact, what we observe at least in two Semitic dialects: in the
Assyrian dialect of Akkadian, where the eves- pattern with zero inflection results
in 'uzan, $Upar etc.,106 and apparently also in Aramaic, at a very early stage, where
the corresponding forms were, at it seems, *'irj.an, *tipar, whence in the historically
attested period 'edan, 'idn-il, tapar, tipr-il, and then also, by analogy, also such forms
as *samas > *samas were accompanied by such forms as sims-a instead of the ex-
pected *sams-il; by the same reason *batn- 'womb' > betn-il, butn-il etc. The result is,
that in Aramaic the original patterns C
1
iC
2
C
3
and C
1
aC
2
C
3
are often reflected ir-
regularly.
In most of the Semitic dialects, however, the patterns eves- with zero inflection
began to acquire a vocalism analogous to the CSC-pattern. In this case, a devalopment
*CSC9t > *CaSC* was impossible: although so long as S was a vowel, the sequence
*CSC was a normal syllable and therefore could combine with both zero and vocalic
inflection, as soon as it became a diphthong aS, the S became non-syllabic = con-
sonantal, and therefore the rule forbidding two consonants at the end of a syl-
lable came into action; this meant that CSC# had to develop into CaSaC9t (also
CaCaS9t and by analogy even CaCaS* when the pattern was CVCS]. In several in-
stances this vocalism became normal for the word in question, even with vocalic
inflections, hence the Hebrew forms *buk[u]r-, *zi'[i]b-, *ri'[i]m-, or the Arabic
sa'[a]r- etc.
Then, phonotactic conditions existing in Akkadian and Aramaic brought about
a virtual contamination of the patterns CVCVS and eves, as well as CaSaC and CSC;
with a zero inflexion one would, theoretically, expect to be able to distinguish between
patterns of the types CaCiS9t and CaSiC9t, on the one hand (probably these were
usually derivative nouns), and patterns of the types CaCaS* and CaSaC9t, on the
other: the last named types corresponding also to the original CaCS- and esc-
patterns. But in practice this led to a merging of patterns ( cf., e. g., Akkad. kars-,
karas- 'stomach', Aram. kirs-li, kars-li, kares9t, Eth. kars but Hebr. kilre, Arab. k;'Vis-
105 However, it seems that S9t could be reflected not only as -aS9t but also as -Sa.
Hence, through Systemzwang, may have emerged, as a more general rule, the free
variation: zero inflection II -a inflection, from which may be deduced the ending -a
(parallel with zero) of the predicative form of the noun in the most ancient Akkadian
(as attested by the earliest borrowings into Sumerian, and partly in Akkadian PN) and
in the most ancient Amorite; and hence, ending -a of the stative - "new perfective"
( = *predicative, or zero case use of the *verbal noun) in Arabic, Ugarit!c and Amorite
as contrasted to zero in the corresponding forms in Akkadian (and Canaanite?);
also the -a-inflection of the Accusative if, as we have suggested, the Accusative was
a development of one of the variants of the "absolute" case (zero case), expressing in
languages with an ergative construction of the sentence both the subject of a state in
general. and also the subject of a state resulting from an action, i. e. the direct object
of a transitive verb; such an "ergative construction" seems to have preceded the his-
torically attested "nominative construction" of the sentence in Semitic ( cf. SHL,
p. 58; YDPA, p. 213). Possibly the suffix variants R-b- II R-ab-, R+ II R-al-, R-r- II R-ar-,
R-t II R-at- might also be traced to this phenomenon (R being the root morpheme -
originally with final S, and then, by analogy, with any final consonant. However,
according to I. J. Gelb, -a- is the original feminine gender morpheme; then -b-, +,
-r-, -t- must belong to another class of morphemes).
106 Thus also, by analogy, in derivative nouns of the patterns CwCzC3, C1i/uCzC3.
Problems of Root Structure .in Proto-Semitic
475
and even kirs-; or Akkad. malk-, malik*, Hebr. miiliils < *malk-, Aram. malk-ii,
malels*, Arab. malik- 'king'; Akkad. napis-t-, Hebr. niipiis < *naps-, Aram.' naps-a,
Eth. nats, Arab. nats- ]. The two last words are pretty certainly derivative; generally
speaking, any nominal base pattern which may be written as C
1
VC
2
C
3
, where :vc
2
s
and C:: S, is suspicious from the point of view of the probability of its primary
origin, even if the verbal root in question has not been preserved in the historically
attested languages; still more so if there does exist such a verbal root, even although
traditionally the verb has been thought to be secondary, and the noun to be primary;
cf. Akkad. 'isd-, Hebr. 'iisiid ( < *'asd- J 'sole, foundation'; also Akkad. 'iSk-, Ugar.
"usk-, Hebr. 'iisiils (<*'ask-], Aram. 'esk(a}-t-ii, Eth. 'askit.'testicle' (however, 'V-
may here be originally prothetic, cf. note 29]; kasp- 'silver' etc.; the same is true
of the pattern *CaSC; for instance, Common Semitic *bark- 'knee'; Akkad. 'ikk-
r < *]J.ink-], Hebr. ]J.els ( < *]J.ikk- < *]J.ink-], Aram. ]J.ikk-ii, ]J.enk-ii, but Arab. ]J.anak-
'palate', etc.
At a very early period "re-vocalisations" of primary nominal roots to form diminu-
tives must have appeared, formed after the patterns CaCaC, CaCiC; the first pattern
lost later its productivity, and the forms in question lost their diminutive connotation,
as e. g. *'intis-, *'untis- 'man' (in contrast to animal]; and by contrast and analogy,
also (?] 'iltih- 'god'; 'inti' 'vessel', rjubtib- 'fly', *tu'til- 'fox,l06a *Zibab- 'heart' etc.
Especially numerous are the anomalies in word-base formation from primary
nominal roots in Arabic. This is natural, because the system of internal vocalic
inflection (also as a method of word-formation] has in Arabic developed much wider
than in other Semito-Hamitic languages; beginning with the formation of diminutives
and of forms of Pluralis tractus [which originated in the sphere of verbal nouns,
where it was a device of contrasting abstract and concrete nouns, as well as nouns
in the collective sense and in the sense of individual objects formed from the same
.root and with an identical basic semantic content; later it overflowed into the sphere
of primary nouns), the internal inflection was then made use of for forming nouns
from verbal roots, a process which is much easier in the Southern than in the Northern
Semitic languages, where the main stirps ["stem") of a verb is seldom formed from
a nominal root. Then from new verbal roots numerous secondary nouns were formed
in their turn [ e_ g. Arab. 'arjin- 'ear'), sometimes surviving the verb itself from which they
originated. Alongside of all this, also the other processes mentioned above are observed
in Arabic, as well as in the other Semitic languages: thus, already in Classic Arabic
a form 'urjun- is attested besides the regular form 'urjn-, cf. in Hebrew the forms
*bukur-, *zi'ib-, *ri'im- instead of the expected forms *bukr-, *zi'b-, *ri'm. Typical of
Arabic is also the already mentioned phenomenon of contraction of often used words of
the type CaC in rapid coherent speach [sandhi), leading to the emergence of such forms
as *bin- > bn- > (i}bn- etc.;l07 in the following such forms may receive, first in the init-
ial, later in all positions in the sentence, a prothetic, inorganic 'i-, 'u-; hence the var-
iants Arab. 'ant-, 'inf-, 'unf- 'nose'; cf. also 'ins- 'mankind' and Akkad. nis- 'people'
[with 'ins-tin- 'human being, man' as a nomen unitatisj,lOB etc., etc.
106a Hebr. sil'iil (not *sa'6l!] reflects *tu'al- in the same way as ganniil]- beside
gibb6r- < * gabbiir-.
10
7 Cf. the traditional pronunciation Hebr. snajim, Aram. tri'!n for sanajim, taren
'two' and the subsequent total loss of sayii mobile both in New Hebrew and in Neo-
Aramaic.
108 In this connection can such forms as Arab. 'ibham-, bihtim-, Akkad. 'ubtin
( < *'ubh-iin-) alongside of Hebr. bohiin < *buhn- etc. be explained. It seems that
the origin of this word [meaning 'thumb') must be reconstructed approximately as
476 I. M. DIAKONOFF
Lastly, we must mention that in Proto-Semitic and Proto-Semlto-Hamitic as in all
"proto-languages", there must have also existed root variants.109
Of course, there remains a certain though small number of roots which so far
defy explanation and classification, as e. g. the one-consonant root *p- 'mouth': Akkad.
pi-, Hebr. pii-, Aram. piim-d, pumm-d, Eth. 'a( Arab. fii, fumm-, famm- etc. This root,
by the way, is Common Semito-Hamitic.
v
The sub-system of Semitic verbal roots and of word-bases derived from these
differs from the sub-system of primary nominal roots not only typologically but also
chronologically. The whole verbal sub-system certainly could not have existed as we
know it before the creation of the six-vowel phonological structure, because It is just
the vocalism, and, moreover, this particular system of vocalism !and not the one
which may be reconstructed for the sub-system of nominal roots] which is the means
allowing to differentiate the various grammatical and forms inside the
verbal system, .including the verbal nouns. It it to this diachronic level that the
mechanism of the Semitic apophony interpreted by J. Kurylowicz belongs. Therefore
(in spite of some points of our disagreement with Kurylowicz's reconstruction, because
some of his relative dates for certain phenomena in Semitic seem to me too late cr
too early in the light of the data of other Semito-Hamitic languages) in this paper
we hardlv need touch upon apophonic laws; instead, we will attempt, by inner re-
construction, to reach an earlier, conventionally speaking, "Pre-Proto-Semito-Hamitic"
diachronic level. The task is here considerably more difficult than in the sphere of
primary nouns, and we will limit ourselves to a preliminary sketchy sug-
gesion of a solution. It should be stressed that any solution has at present to be,
and perhaps will always remain, highly hypothetic.
What is the reason for the fact that the type of the historically attested Semitic
verbal root is, in principle, quite dissimilar tc the primary nominal root? In a verbal
root there may exist three consonants, none of which need tc be a scnant even if they
cluster, and it is usually impossible to establish any particular vowel as belonging
to the root. [Sometimes the vowel of one particular verbal fcrm, usually that cf the
"Old perfective" - the Western Semitic "Imperfect", is taken to be part cf the rcct,
although logically one verbal form would seem as good as another].l1
follows: root *bah- plus the already mentioned "individualizing" suffix -am- II -&n- >
*bih-am/n- or, with the typical contraction of the roots of the GaG-pattern, > bh-ti.m/n-
with a prothetic 'i/u-, cf. the numerous analogies mentioned above; at the same time
a word like *biham- could easily be taken for a diminutive of the pattern GaGaG,
hence a "non-diminutive" form *bahm/n- would be "reconstructed"; other forms cf
the word (which exist e. g. in Arabic] are obviously secondary. More complicated are
other cases, e. g. the case of the word 'finger' which probably may be explained in
a similar way: Arab. 'i?ba'-, Hebr. 'G.$ba', Aram. 'G.$bt1'-d, Eth. 'a$btl'-t but Aram.
Syriac $eb'-d. Cf. also the enigmatic words Akkad. 'bird', Hebr. $ippor, Aram.
$ippar- 'bird', where a series of phonetic changes must be supposed tc have taken place.
109 F.. g. a number of root variants with alternation *' II *j; *lit- II *nit- 'lion';
g''r- I' g''ar- il gwar- 'throat' [a case reminding of Indo-European ap<fphonic
0
correla-
tions); $1]r ii zYr ii :;;'r; Akk. lit- < *Zi'-t-, Hebr. /e'd '[wild] cow', but Arab. Za'q, cf.
Cushitic Za'/y, Tchad Za, Zo 'cattle'.
110 As I have attempted tc show elsewhere [SHL, pp. 79 sqq.; YDPA, pp. 229 sqq.],
the conjugated forms of the Semito-Hamitic verb based on the semantic contrast
"imperfective : perfective" [or "durative : punctual"], and on the formal contrast "full
vocalism : contracted vocalism" appear to be as old as the system of Semito-Hamitic
prefixal conjugation itsP.lf. But if this proves tc be true, then there is no reascn
477
To answer the question, we should, as in the case of the primary nominal root,
start from the premise that a root can exist only as part of such lexica-grammatical
forms that really exist in the language.
In the first approximation we may exclude from our analysis the forms of the
Participle of Action and of the ma:jdar's, because these forms, as attested in the
individual Semito-Hamitic languages, cannot be traced back to Common Semito-Hamitic
prototypes, and therefore do not belong to the diachronic level of Proto-Semito-Hamitic
[or, in most cases, even of Proto-Semitic J .111 The same is true of the so-called pas-
sive,112 which also cannot be traced to the Proto-Semitic and Proto-Semito-Hamitic
level.
Thus, we will have to analyse only: [1) the form of the Participle of State [CaCaC-,
CaCiC-, CaCuC-, with the variants CaCaC-, CaCzC-, cacac- j, which in its predicative use
is identical with the stative [the form which preceded the New Perfect of the later
Semitic languages); and [2) forms connected with prefixes of the pattern CV-1 SV-
(e. g., ma- which serves to form certain nouns, or ta- which serves to forrn
the 3rd person of the, finite verb of action J.
Let us now consider the second case; it will appear directly, that the peculiarities
of the verbal structure are connected with the peculiarities of the
Se!Ilfto-Hamitic verbal inflection [namely, inflection by prefixation). We will also
take it for granted that the suffix of a \,'erbal form may be either zero' or vocalic.
Then, !iCCording to the rules of syllable formation as formulated above, the following
patterns are not forbidden for a verbal root:
. (1)
[2J
[3]
[4)
Sa+CS{+V /
S,a+c;VC(+VJ
Sa+CVCVC( + V j
SatCCVC(+V}
(Sa I CS, SaC I SV) '
rsa eve, sa 1 cv 1 CVJ
r s,a 1 cv 1 pvc, sa 1 cv 1 cv 1 qv J
(SaCICV<;;', SaCICVICVJ. .
Equivalent variants with sonants are also possible: in pattern [2] a sonant can
stand both for C and for V, in pattern [ 3 J for C and/or for one of the vowels, and
patterns [lJ and [4) for. C only.
We can thus make a very important statement, namely, that the conditions of
existence of the verbal root inside a word do not contradict the possibility of its
including three consonants, none of which is a sonant, even if two of them are
in contact.
to consider the vocalism of the "old perfective" [or the Akkadian Preterite J as
belonging to the root, more than the vocalism of the "old imperfective" [or the
Akkadian, Present). At the diachronic level when this verbal system originated, it
will evidently be correct to consider that there did not exist any stable vocalism
which could be regarded as belonging to the root, but that the vocalism had an
apophonic character. However, turning to earlier levels of the language, it will be
obvious that the personal affixes must have originally existed as separate lexemes,
and in this case there is no reason to believe that the root morpheme with verbal
semantic would formally differ in any way from that with nominal semantic; this
brings us to the necessity of reconstructing the primary verbal root morpheme as, in
principle, identical in pattern with the primary nominal root morpheme; which means
that it must have included a vowel. In the so-called verba mediae infirmae we can
apparently even at the historical level find certain remnants of the primary root
vocalism which, in the different contrasting conjugated forms, did not wholly depend
on the laws of apophony.
111 Even as stable a pattern as that of the Participle of Action (CaCiCJ is not
attested in the Southern Peripheral Semitic languages, to say nothing of the other
branches of Semito-Hamitic.
112 Cf. YDPA, p. 249 sqq.
478
I. M. DIAKONOFF
For Proto-SemitoHamitic these identical patterns of finite verbal forms can really
be reconstructed on the basis of historically attested facts, namely:
*ta-C1aCz-. - perfective, verb of motion or of transitive action;113
*ta-CtaCz-, *!a-G1CzaC3- - verb of intransitive action;114
*ta-CaCaC- - imperfective, verb of transit{ve action.115
The bisyllabic root pattern *Sa+ I CS 1 eve, *Sa+ I CS 1 CV i C+V is also possible,
and is actually observed in the case of the so-called "quadriliteral" verbs.
Apparently, the mechanism of pattern correlation was brought to uniformity; the
leading factor must have been the transition from external to internal inflection. The
situation in Hausa may serve as a sort of typological model [this does not mean
that Hausa r e t a ins the original situation in this respect; it is much more probable
that a secoodary development has here led to results similar to the situation in the
proto-language). In Hausa, at the end of a finite verb [which here consists of a form
of the conjugated auxiliary verb which actor prefix, and of a ma$dar), there is added
a vocalic inflection, variously characterizing the verb as stative,
116
transitive etc.
Let usassume that in ,Pre-Proto-Semito-Hamitic" the vocalic suffix -V could similarly
be the exponent of intransitive action [-a) or of motion and transitive action (-a).
Then, as a result of a very usual process of absorbing the external inflection into
the word-base and tui'ning it into an internal inflection, we would, proceeding from
patterns known to us from the nominal root sphere [CS, CVC, CSC, CVCS, CVCVCJ,
11
6a
encounter such forms as: *!a-CaS, *!a-cas; *ta-CaC, *ta-CaC; *ta-CaSaC, *ta-Casac;
*ta-CaCaS, *ta-CaCaS, with the connotations of intransitive or transitive action, re-
spectively. The necessity to distinguish, for verbs of action, forms expressing mo-
mentary action [punctualis, perfectivus) from those expressing duration of the action
[durativus, imperfectlvus), very naturally would bring about a system contrasting
contracted [momentary) forms to forms with full vocalism [ durative] :111
113 Sem. *ia-kun- *ia-qtul- la-d rib- ia-sqi- Berb. 1-ssan 1-fras i-gmi etc. Cushitic:
Bec,iauye e-dlr; 1-fdlg; tl:qin; Hausa ta-n auxiliary verb; 'the other
verbs are composed of a ma$dar and of the prefix conjugation of the auxiliary verb].
Note that while the Semitic languages differentiate the pattern *ta-C1CzuC3 [the normal
type of the transitive verb] and the pattern *ta-CICziCJ [verbs of motion and of tran-
sitive surface action), the Cushitic languages have only the latter type which there
is the normal one for the transitive verb - probably another evidence of the secondary
character of the vowel u.
114 Sem. *ta-?zzan-, *!a-lbas- etc., medial if a not in contact with ', h, ', ?z. In
Berber this type is usually indistinguishable from the preceding, because *a, *i, *u >
Berber a or zero.
115 Semitic Akkad. I' parr as (apparently with a secondary stress which is respon-
sible for the reduplication of the second radical, since the expected form *'iparas
would, according to rules obtaining in Akkadian, result in *'ipras, a form indistin-
guishable from the preceding pattern]. In spite of a wide-spread contrary opinion,
I assume that the similar forms of imperfective with full vocalism in Southern Peri-
pheral Semitic (Mehri, Ethiopic, etc., see SHL, 90 sqq.] are identical with this Akkadian
form in their origin; cf. also the Berber imperfective (ad} ifras, (ad} igmat ( < *ta-faras,
*ta-gamai) etc.; Cushitic: Bec,iauye e-ndir, (a-jfandig etc., Somali ta-qan; Tchad: Hausa
ja-na; all these are forms of the imperfective of the transitive [in Hausa: auxiliary]
verb.
116 Thus, according to an oral communication, are the Hausa verbal forms in -u
very convincingly interpreted by B. G. Lemesko.
These patterns should also be regarded as being the origin of the Participle
of State.
117 Cf. A. M. rasos-fHHs6epr, CuMBoAuKa KparKocru - iJoAcOTbt B ocHoaax ceMur
cKux cnpHcaeMblX opM, IV CecCHH no llpesHeMy BocroKy, TesHCbi noKJianos>>, Moscow,
1968, p. 18 sq.
*ja-CwCz:
*ta-CaSwSl, *ta-CwCzaCz: *ja-CaS-, *ja-CaCz- (or *ja-CS1aS1, *ja-C1CzaCz),
*ta-cacas: *ta-ccas, *ta-ccas,
*ja-CaSaC: *ja-CSaC,
479
which practically is the verbal system which has been preserved in the historically
attested Semito-Hamitic languages.
This situation made it possible to create new verbal roots with forms like
*ja-CwCzataCi: *ja-C1Cza/C!C3. In the already mentioned Hausa language we may
observe the phenomenon of adding various consonantal morphs to the verb; these
morphs are gradually becoming part of the lexeme, and of the verbal root in ques-
tion_l18 We may assume that the mechanism of inclusion of the third non-sonorant
radical into the Proto Semitic (or Proto-Semito-Hamitlc) verbal root119 was somewhat
similar, for instance:
*ja-CwCz+CJ = *ja-C1CzaCJ, *ja-CwCzaCJ,
*ja-C1aCz+ C3 = *ja-C1CzaC3, *ja-C1aCzaC3, etc.
In cases where there were no third radical added, if the original root was bicon-
sonantal, the primary pattern *ja-CVC might have been preserved (sometimes with
extension to *ja-CVC undel' the influence of the rhythmical pattern of triconsonantal
stems], or there might be created, by analogy, forms of the pattern *ja-C1CzV:
'*ja-CwCzV. Seeing that such forms were impossible with a vocalic suffix, because
of the forbidden sequence *-CVV, a non-syllabic sonant, articulationally affined to the
vowel in question (as -' to -a-, -i/IJ to -a-) would be added (of course, if the last
phoneme of the root was not already a sonant from t'he beginning); or, again, the last
consonant of the root might be repeated: *ja-C1CzVCz (or *ja-ClVCz]: *ja-CwCzVCz.
It is well known how often the verba mediae infirmae, tertiae infirmae, and secundae
geminatae alternate in Semitic languages.
The differentiation between quasi-normal triconsonantal verbal roots with the
"weak" consonants ', j, IJ as one of their real radicals,120 and "weak roots" with
an imaginary "weak" radica!121 (a fact attested in Semitic, in Egyptian etc.], although
never quite consistently carried out, because of the loss of its cause, corresponds in
principle to the original differentiation between the roots of the patterns *C1ljCz,
C11)Cz, C1VCzEJ etc., and the roots of the ClVCz pattern.
The emergence of forms like *kis-, *ma'-, *ja-kun- which have their origin in
rhythmical conditions was, probably, the impetus which in the long run caused the
creation of long vowels as separate phonemes. This, along with the differentiation
also of u as a special phoneme, greatly extended the possibilities of vocalic inflection,
both external and internal. Such extension was most necessary because of the scarcity
of other means of word-formation in Semito-Hamitic languages. But the creation
of a complicated, ramified and balanced system of verbal stems and nouns of verbal
origin (mainly distinguished only by their vocalism from words of affined but different
semantic, derived from the same rnots) brnught about the functional polarity between
118 Another side of the same process seems to have been the elimination of such
patterns as *ja-1 cs, *ja-C I s-v, esc, *ja-CS I C-V, because they do not allow of
contrasting momentary and durative forms; once *ja-CaS is contrasted to *ja-CaS (and
*ja-CaS r a JC to *jaC I SaC) there is no room for a form like *ja-G I s-v (or, respectively,
*ja- I CS I C-V) in the same paradigm.
119 Cf. YDPA, p. 240.
120 Like Akkad. i-!i'al, i-!ia'al, u-parri', or Hebr. ji$1Jal;l.
121 Like Akkad. i-kun, i-ku:an (or i-kiin; the symbol : denotes a glide), i-bn!, i-banni,
or Hebr. ja-qilm, ji-bnii.
480 I. M. DIAKONOFF: Problems of Root Structure in Proto-Semitic
consonants and vowels so typical of the Ancient and Middle Stages of Semito-Hammc.
In order not to break up the associative connections between the manifold nominal
and verbal patterns derived from one and the same root, a rigorous conservation of
the consonantal skeleton of the root was indispensable; by this reason positional
(phonotactical) changes of consonants were at that stage exceedingly rare, the pho-
nological system of Semito-Hamitic consonants remaining astonishingly stable in. the
course of millennia.
At the same time, the secondary origin of the patterns of vocalism with long
vowels [e .. g. the patterns *facaz-, *fa<il-, *facaz- as compared with *fa<ul-, *fa<il-,
'*fa<al-]122 can still be felt clearly enough, and this has repeatedly been noted in
Semitological works; the .:,arne can be said of the secondary character of u as com-
pared to i and of the existence of an ' original "bivocalic" stage in Pvoto-Semtto-
Hamitic_l23 It can also be noted, that the spectrum of word-formational patterns in
such an archaic Semitic language as Akkadian reveals very few examples of patterns
with an i- or u-vowel in the first syllable, while the few such patterns there are
seem to be of later origin than the others (always excepting the very ancient pattern
CaCS- and its imitation, C1aC2C3). It seems, tn general, that those patterns of deriva-
tive word formation in Semitic should be considered the most ancient, which cor-
respond in structure to the primary nomi!Jfll root morphemes, viz. the pattern *facz-,
'*fich *fa<al-; already a like it may be necessary to consider as the
resuH of the infiltration of an external inflection into the. word-base.124
The hypothetic character" of our results should once more be ,stressed here; even
.if they are. accepted, they will need corrections and amendments; wha\
see:{IIS to me really important, that the structure of the root should be explained
proceeding ,from the actual conditions that had existed in .wo:rtof the root
was a part. I should also like to remind the reader, that the suggested reconstructions
refer to a diachronic level different from that to .which the l.aws of apophony studied
by Kurylowicz relate. And it seems to me important to state once again, that the
prospects of further research in reconstruction of the history and pre-history of the
Semitic branch of the (or semito-Hamitic) linguistic f,amily seem to me
not very promising, unless the data of the other branches are considered, and the
reconstruction of the phenomena in question is attempted at the Proto-Semito-Hamitic
level - and, as far as possible, .. achieved. The. s.ttuation is identical to that in Indo-
European linguistics, where the study of one particular branch of the family can
lead to only limited results, unless the facts of the affined branches are drawn into
the discussion.
122 See J. Aro, Parallels to the Akkadian Statiue in the West Semitic Languages,
"Studies in Honor cf B. Landsberger" (Assyriological Studies N2 16), Chi cage, 1965,
p. 407 sqq. In such cases, too, rhythmic and intonational factors could have played
a certain role in the formation of vnwel length.
123 The bivocalism in Proto-Semitic was a firmly established fact already fer
G. Bergstrasser, Einfilhrung .. . , p. 5. Note that there are traces of free variation
in Semito-Hamitic not only between i and u but alsc between ! and (and, corre-
spondingly, j and yJ; see the examples cited in SHL, pp. 40, 86, 8670 (the relative
suffix +I I -lj-, the morph of the masculine gender -I I i-).
m Viz., the relative morph *-a?

Diakonoff - Problems of Root Structure in Proto-Semitic

Uploaded by

Document Information

Original Description:

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Diakonoff - Problems of Root Structure in Proto-Semitic

Uploaded by

Copyright:

Available Formats

ARCHIV ORIENTALNI 38, 1970

You might also like