On Verbal Negation in Semitic*

By Na’ama Pat-El, Austin

Summary: While there have been many studies of negative expressions and negation in a
number of Semitic languages, negation is not normally used in comparative Semitic lin­
guistic and is avoided in discussions about subgrouping. This study reviews the nega­
tion of verbal predicates in an attempt to reconstruct the original Semitic set of negation
particles. It concludes that verbal negation in Semitic was originally a set containing two
particles with distinct functions: *ʾ l was used to negate indicative verbal forms, while *lā
was used to negate non-indicative ones. Many languages generalized the indicative nega­
tion particle to negate all verbal forms, a move which subsequently led to the loss of the
original syntactic distinction. The reconstruction of the innovations in Semitic corrobo­
rates the subgrouping of Semitic suggested by Hetzron (1976) and modified by Porkho­
movsky (1997) and Huehnergard (1991; 2005), and in fact validates the modification of
Hetzron’s proposal.

1. Introduction

Every Semitic language has its own set of negation particles and there is
no one set that is shared by all of them, sometimes even within the same
sub-branch. For the negation of nouns a number of prepositions as well as
negation particles are used. In order to limit the discussion to a manageable
amount of data, I will deal here mainly with verbal negation and will try to
avoid negation particles that are unique to only one language.1
While there have been many studies of negative expressions and negation
in a number of Semitic languages, negation is not normally used in compara­
tive Semitic Linguistics and is generally avoided in discussions about sub­
grouping. The reason for this is not hard to see: the number of particles and
their distribution make it a very hard task to discuss even basic reconstruc­
tion. Take for example Central Semitic: Arabic uses mā and lā and OSA uses
ʾal. Or even a more tightly related branch, Canaanite: Phoenician uses bal
and ʾal and Hebrew lō and ʾal. In addition, in Akkadian, the set of negation
particles seems different than that of WS: ula or ul and lā. Many scholars

­* I wish to whole-heartedly thank John Huehnergard, Rebecca Hasselbach,

Lutz Eberhard Edzard and Daniel Birnstiel for their very helpful comments on
earlier versions of this paper. All remaining errors are mine alone.
18 Na’ama Pat-El

(e.g., Testen 1998) have tried to argue for a relationship between ʾal and ul.
Similarly, some scholars (e.g., Lipiński 2001, Pardee 2003–2004) suggested
that lā and ʾal are variants, i.e. the latter is identical to the former with an
ʾa- prefix.
However, the main problem is not only the etymology of the original
Semitic negation particle, but primarily the usefulness or even feasibility of
using negation for subgrouping.1 This is indeed a puzzle. On the one hand,
we have an essential feature which should be used for subgrouping, on the
other hand, the category shows such lexical and syntactic diversity, that it is
hard to see the connection between forms in the different languages.
There have been two main attempts to reconstruct negation in Semitic:
Walker (1896) and Faber (1991). The former is more comparative in na­
ture while the latter is also historical and reaches all the way to Proto-Afro-­
Asiatic (PAA). Walker collected an impressive number of negation particles
from a number of Semitic languages, however his definition of negation is a
very broad semantic one.2 He divides negation into four categories: negation
of pure dissent; negation by association, negation by transference of force;
negation by suggestion. Walker’s study shows a rather antiquated form of
historical linguistics, where, for example, a feature such as vowel length was
claimed to represent a meaning or an emotional attitude.
Faber made an attempt to connect the negation particles in Semitic to
supposedly related forms in other Afro-Asiatic branches and to explain
the apparent morphological similarity between some of the Semitic nega­
tions and interrogative particles, mostly mā and ʾay-. She concluded that
Semitic had three original negation particles: *lā, *ʾayn and *ʾay. Faber
further claimed that the particle ʾal is a WS innovation and developed to
a general negation only in South Semitic, but its original function, which
was preserved only in some NWS, is a prohibitive deverbal particle “don’t!”
(Faber 1991, p. 422). Central Semitic innovated bal as a negation particle,
which Faber takes as a compound negation particle of as yet unknown ori­
gin (Faber 1991, pp. 415–416).
Faber’s reconstruction is problematic on a number of levels. First, her
Afro-Asiatic reconstruction of Arabic mā as *mba is not strictly needed or
meaningful. Even if PAA had a negation *mba, Semitic most probably did

1 Indo-Europeanists have no qualms about using negation in subgrouping. Note for

example Forston (2004, pp. 133, 149) and Mallroy/Adams (2006, p. 62) for the recon­
struction of a negation particle and its syntax in Indo-European.
­2 A comparative table of the Semitic forms may be found in Walker (1896, p. 235).
The table reflects what Walker perceived as etymological relations rather than functional
ones; for example, Hebrew mā and Arabic mā, or Hebrew lôʾ and Syriac law are presented
as one and the same negation with very little discussion or analysis.
 On Verbal Negation in Semitic 19

not. In Semitic mā is an interrogative. Al-Jallad (2008) has noted that the

use of negative mā is restricted to Central Semitic and is productively used
only in Arabic, while it appears only in relics in other Central Semitic lan­
guages. Al-Jallad notes that it is unlikely that a Proto-AA particle is only
preserved in Arabic but is otherwise completely missing from any other lan­
guage. He further suggests an intra-Semitic syntactic development, which
accounts for the movement of mā from a WH-element to a negation p ­ article.3
As for the Central Semitic bal-, it seems to be closely connected to Akka­
dian balu(m) and Ethiopic ʾənbala, both of which have the same basic syntax
and semantics (negative preposition) as Central Semitic bal-; in light of this,
bal- should be considered common Semitic.4
Faber’s suggestion to consider ʾal a prohibition (‘don’t’) which then
spread to all positions in South Semitic is also shaky. It is not clear why a
highly restricted negation particle such as ʾal would spread to all positions.5
Therefore, the reconstruction and distribution suggested by Faber (1991,
p. 421) should be revised in favor of a reconstruction that will account for
the morphosyntax of the Semitic negation, before attempting to reconstruct
a common Afro-­Asiatic system.
In this paper, I would like to suggest a reconstruction, which I believe
explains the situation in all the different languages. At a minimum, all the
branches of Semitic exhibit as their verbal negation particle lā, or ʾal or both;
therefore, these particles will be the focus of this study.6 The following dis­
cussion will concentrate mostly on distribution and function.

2. Review of the data

The negation particle lā is found in all branches but Ethiopic, as the negation
of nominal forms, besides its function in the verbal system. Many languages
developed special prepositions for nominal non-predicative substantives, yet
very often lā is also found: OAkk. lā pātiḥtum ‘not opened’ (Hassel­bach
2005, p. 175); OB lā watar ‘no more’ (von Soden 1995, pp. 220, § 122 a); He­
brew lōʾ ʾĕlōah ‘a non-God’ (Deut. 32:17); Ugaritic l ib ‘no enemy’ (1.3:III:37);
­3 Al-Jallad objects to the derivation interrogative > negation, which was suggested
by Wehr (1953) and later Rubin (2005). He suggests that mā moved from being a sub­
ordinating particle (indefinite relative) to being a negation particle.
­4 For the development of bal- in NWS, see Pat-El (forthcoming).
­5 In fact, as I shall attempt to show below, restricted negations in the Semitic lan­
guages tend to be abandoned in favor of more general negations.
­6 I will not deal with *ʾayn here as it is not strictly a negation particle but a negative
predicate, and therefore a different element than other negation particles. Cf. Dahl (1979)
for examples of similar structures in other languages.
20 Na’ama Pat-El

Arabic min lā šayʾ in ‘for nothing’; Aramaic l-hn ‘not so’ (Sefire I B 36). This
function is missing from Ethiopic, OSA and probably Phoenician.7 Gener­
ally, ʾ l is not used for negation of nominals;8 the only language which uses ʾ l
to negate non-predicative nominal forms is Biblical Hebrew, where it is very
rare9: hārê ba-gilbōaʿ ʾal ṭal wə-ʾal māṭār ʿălêkem ‘mountains of the Gilboa,
let there be no-dew and no-rain on you’ (2Sam 1:21). Even languages, which
use only ʾ l in the verbal system, like OSA, prefer to use other prepositions
to negate their nominal forms, rather than using ʾ l.10
In this section I will briefly review the distribution of lā and ʾal (and/or
their reflexes) in the Semitic languages. This section is an attempt to concen­
trate on the shared morphosyntax of the languages, thus by definition, not
all negation forms and patterns are included. Furthermore, the semantics of
the negated verbs is for the most part not relevant for the arguments posited
here, as the emphasis is placed on the combination of a negation particle with
a certain verbal form. Thus, for example, the fact that in Arabic lam + Jussive
is a negation of the past is not as important as the fact that lam negates the
morphologically apocopate form, rather than the suffix conjugation (kāna).

A) Akkadian
In East Semitic, negation particles distinguish between matrix verbs (i.e.,
verbs in main clauses), negated by ula (later ul),11 vs. dependent verbs (i.e.,
verbs in subordinated clauses), negated by lā. The particle lā has several addi­
tional functions, most importantly it negates the durative to create a negative
imperative and cohortative (modal negation) and is the common negation in
sentences with interrogative particles (von Soden 1995, p. 220, § 122 a).
ZÍD.BA-su ù-la e-bı -́ ı ś ‘I will not provide for his flour rations’ (Hassel­bach
2005, p. 175, § 4.3.6)
lā tanaddin ‘do not give!’ (von Soden 1995, p. 133, § 81h)
aššat awīlim ša zikaram lā īdû ‘someone’s wife, who did not sleep with any
man’ (von Soden 1995, p. 267, § 165e)

 
­7 However, see Gibson (1982, III, p. 109) and Pat-El (forthcoming) for a possible
remnant of lā in the nominal system of Phoenician.
 
­8 This is also true of Akkadian ul(a).
 
­9 Qimron (1983, p. 473) suggests that examples of ʾal without a following verb are
variations of ʾal + PC and are, therefore, not negation of non-verbal forms.
­10 One exception is some modern Ethio-Semitic languages (Amharic, Gurage and Ga­
fat) which use alä ‘without’ to negate nominal forms. In Tigre and Tigrinya, the same
particle negates nominal predicate: ʾalä-bu ‘there is not’.
­11 The earliest texts show only ula. See data in Hasselbach (2005, p. 175, § 4.3.6).
A rare form uli is also attested.
 On Verbal Negation in Semitic 21

However, the negation particles in Akkadian do not neatly correspond to

the distinction matrix-dependent forms and both ul(a) and lā may be found
in non-matrix sentences, depending on the type of subordination. The par­
ticle ul(a) may occur in some types of adverbial subordinated clauses, mainly
those introduced by kīma and ištū (von Soden 1995, p. 276, § 172 d).12 This
may indicate that the difference between lā and ul(a) is not necessarily re­
lated to the syntactic independence of the predication. Therefore, we may
more accurately say that lā negates indicative verbs only when they are sub­
ordinated, but otherwise it mostly negates non-indicative verbs.
The particle ula is not found in WS as far as we know. Several scholars
have attempted to connect the etymology of this form to other known par­
ticles. Some scholars (e.g. Kienast 2001, p. 398) have assumed that ula is a
derivation from lā, with a conjunctive waw: *u-lā.13 This is problematic in
terms of syntax, as ula occurs in syntactic contexts where the conjunction u-
does not typically occur.14 Frequently there have been attempts to connect
Akkadian ula and WS ʾal. Testen (1998, p. 128) suggests that Akkadian ul
­12 Note, however, that conditional sentences introduced with šumma are negated with
lā. The particle ul(a) is the regular negation of the verb in the apodosis of conditional
clauses, which is a main clause.
­13 Similarly, there are some scholars who argue that ʾal and lā are one and the same
form (Kienast 2001, p. 398; Lipiński 2001, p. 464, § 47.8; Pardee 2003–2004, p. 380).
This assumption seems to stem from Blake (1911), who posits a vocalic liquid /l/ in Se­
mitic, which is only attested in the negation particles. Blake suggested that an unaccented
lā lost its final vowel to become a syllabic consonant /ļ/. Then, this consonant developed
a prosthetic vowel, which was pronounced with a glottal stop. This suggestion seems
rather ad-hoc phonologically (although ad-hoc phonetic changes are not uncommon in
negation particles). In addition, since the independent function of ʾal and lā seem to be
reconstructed at least to Proto-WS, there seem to be no reason and no evidence to assume
a derivation of one from the other. Note, that if the claims proposed in this study are true,
lā and ʾal have very different functions.
­14 Note, however, Middle Arabic w-lā as a negation particle of the perfect. For exam­
ple: w-lā zāla ‘he didn’t stop’. See Blau (1995, p. 140 for more examples and discussion).
For examples in Classical Arabic, see Bergsträsser (1914, p. 40). This use is a generaliza­
tion of the common Classical use of wa-lā as a continuation of any previous negation: raʾā
ʾanna ḏālika l-badana lam yuḫlaq lahū ʿabaṯan wa-lā qurina bihī li-ʾamrin bāṭilin ‘he saw
that this body was not created for him jokingly, nor was it attached to him for no reason
at all’ (Wright 1967, p. 303, where many other examples are provided). We do not know
if this was the case in Akkadian, but Middle Arabic provides an example where w-lā ap­
pears in syntactic contexts where wa- does not appear. Similarly, in modern Arabic one
finds w-lā iši ‘nothing’ and w-lā marrā ‘never’ either independently or in contexts where
wa- does not normally occur (I thank Dr. D. Birnstiel for these examples): (mā) šuft-iš
er-raǧul ‘I did not see the man’ vs. (mā) šuft-iš w-lā waʾad ‘I did not see anyone’. A similar
pattern is attested in Classical Ethiopic: wa-ʾ i-tarfa wa-ʾ i-ʾaḥadu (Judg. 4:16) ‘not even
one was left’ (see Dillmann 1907 [2003], pp. 508–509, § 197 a). While these cases prove
that it is possible for w-lā to develop into an independent negation marker, there is no
evidence that such a development took place in Akkadian.
22 Na’ama Pat-El

corresponds to WS ʾal, where the vowel difference is, according to him, also
attested in other particles (Akk. lu ~ G. la-). This is a very elegant solution;
however, since ula is the older form in Akkadian, not ul, the correspondence
should be Akkadian ula – WS ʾal. Akkadian ul is probably an apocopated
form of ula and thus cannot be the basis for the etymology suggested by
Testen. Lipiński (2001, p. 464) connects both ul and ʾal to Berber war. The
relationship between ula and ʾal seems logically possible, but the etymologi­
cal explanations given so far are not convincing. Hence, the fact remains that,
Akkadian has a set of two particles, each with a distinct function.

B) Ethio-Semitic
In Classical Ethiopic, one particle, ʾ i, is used as a common negation particle
for all verbal forms, indicative as well as modal. This particle is also the
regular negation of nominal forms, like ʾ i-yāʾməro ‘ignorance’.15
ʾ i-faqada yəkšətā ‘[Joseph] did not want to expose her’ (Matt. 1:19)
ʾ i-təfrāh našiʾotā la-Māryām fəḫərtəka ‘don’t be afraid to take Mary to be
your wife’ (Matt. 1:20)
ʾ i-taḥaṣṣəṣi ʾəmənna masāfenta za-Yəhudā ‘you are not lesser than the rulers
of Judea’ (Matt. 2:6)
Other Ethio-Semitic languages use this prefix invariably as a nominal nega­
tion: Gurage e-š ‘no’, Amh. yällämm ‘there is no …’16.
The particle ʾal is not used in Classical Ethiopic as a verbal negation; how­
ever, it is used in some non-verbal predicative forms (ʾal-bo ‘there is not’) and
in relics, like ʾakko ‘(emphatic) no’< *ʾal-kona ‘no, on the contrary’. This relic
clearly shows that ʾal was a verbal negation in Proto-Gəʿəz. Indeed, in Amharic
and other modern south Ethio-Semitic languages, al …-əmm is the negation
of the indicative perfect and imperfect (Leslau 1995, pp. 292, 302)17: pf.: al-
säbbärn-əmm ‘we didn’t break’, an-nəsäbr-əmm ‘we are not breaking’ etc. An
­15 This particle is probably etymologically related to similar negation particles in other
branches: Hebrew ʾ î-nāqî ‘not clean’ (Job 22:30); Phoenician ʾ bl /ʾi-bal/ ‘no’, ʾy šm ‘noth­
ing’ (KAI 14:5); Old Aramaic ʾ-šm ‘no-name’ (Sefire I C 24). The connection of ʾ i to ay
(attested in Akkadian, MSA and Phoenician) is unclear.
­16 This particle is inflected like allä ‘[be]’, i.e., like a verb, and like a verb it can take
suffixes. The negation suffix (ə)mm is positioned at the end of the form: yällä-čč-əňň-
əmm (NEG-3fs-1s-NEG) ‘I don’t have it (fs)’. See Leslau (1995, p. 528) for the full verbal
paradigm of yällämm.
­17 This is the case for the 1st person singular of the impf.; other persons have the form
a-(…əmm). The final -əmm is elided under certain conditions, for example when the clitic
-əmm is attached to any other member of the sentence, or when the imperfect is subordi­
nated. The negation of the Jussive is also a-, but this prefix causes gemination when at­
 On Verbal Negation in Semitic 23

important distinction in Amharic is between modal and subordinated on the

one hand, and indicative and non-subordinated on the other: the former are
negated with al without final -əmm, while the latter carry both.18 The əmm-
less negation is common in speech and in many of the environments where the
modal (non-subordinated) imperfect would occur (Goldenberg 1966, p. 86).

C) Modern South Arabian

Most MSA dialects use a variant of ʾal to negate all verbal forms. Only Soqoṭri
distinguishes the negation of the indicative (āl) and the prohibitive (ʾa/ʿa/
ḥa).19 It seems that a generalization similar to the one in Ethio-Semitic took
place here, too when one particle took over the entire system and pushed the
other out. Unlike Ethio-­Semitic, MSA retained lā for some restricted func­
tions, mostly as a negative answer, negating entire sentences (when placed
after the sentence) and as an emphatic element (Simeone-Senelle 1994,
p. 189). Lā is also the negation of nominal predicates:
Meḫri:
hābū bə-Kāśən lawnhəm lēybən lāʾ ‘The people in Qishn, their color is not
white’ (Simeone-Senelle 1994, p. 190)
Ḥarsūsi:
anīytəh seh gedt əlāʾ ‘This behavior is not good’ (Simeone-Senelle 1994, p. 191)
In the dialects of Meḫri, Jibbāli and Hobyot both particles may appear in
a double negation construction: Meḫri ʾəl… lā, Jibbāli el… lo (Johnstone
1975, p. 28; Simeone-Senelle 1994, p. 201; Rubin 2010, p. 264 ff.);20 how­
ever, lā always appears after the predicate (mostly at the end of the sentence),
and only ʾal may precede it21:
tached to t- (2nd person) or n- (1st pl.), so we may well assume that it comes from *ʾal. See
also Cohen (1970, p. 170).
­18 This distinction is missing from the typological study of Aikhenvald/Dixon (1998,
p. 64), who note that Amharic marks tense distinctions between past and perfect only in
the positive, not in the negative. In fact, the ‘perfect’ is a complex verb with an auxiliary,
and negation can only appear on the bare form of the verb for historical reasons.
­19 One local dialect (ʿAbd al-Kūri) uses āl (with the subjunctive) as the negation of the
imperative (Simeone-Senelle 1994, p. 190; 1997, p. 141).
­20 The double negation construction was also an option in Hobyot, but has simplified
following the well-known Jespersen Cycle (see below). It is still permissible with older
speakers (Simeone-Senelle 1994, pp. 201–202) so the simplification must have been rela­
tively recent.
­21 Rubin (2010, p. 265) notes that in nominal sentences in Meḫri, nominal and pro­
nominal subject are often fronted: hēt əl hēt ḥōkəm lā (91:29) ‘You are not the ruler’. In
verbal sentences, əl is positioned directly before the verb with some regular exceptions.
24 Na’ama Pat-El

Jibbāli:
el hemm yedre lo ‘he could not get up’ (B II 5, Wagner 1953)
Meḫri:
ʾəl səbēb lā ‘it’s not my fault’ (Simeone-Senelle 1997, p. 413)
In Ḥarsūsi and the Yemeni dialect of Meḫri only, lā is the regular nega­
tion and ʾal seems to be missing; however, even there, lā is always in final
­position:
Ḥarsūsi:
əkhōl ʾəġəter lā ‘I cannot speak’ (Simeone-Senelle 1997, p. 414)
Meḫri:
tijira ḫāmer lā ‘don’t drink wine!’ (MI 31:30 apud Wagner, p. 13)
šay ḳawt lā ‘I have no food’ (73:4 apud Rubin 2010, p. 266)
In all the Semitic languages the verbal negation, regardless of its form, is pre­
posed, and even in MSA one of the negation particles, ʾal, is preposed. Thus,
we may safely conclude that the post-positioning of lā in MSA is an innovation
of this branch; since all MSA dialects attest to it, the change must be dated to
Proto-MSA. There are two options to reconstruct to process: (1) Proto-MSA
moved lā to final sentence position and then some dialects used a preposed
ʾal in addition; (2) all dialects generalized a preposed ʾal and a post-posed lā,
and some dialects dropped ʾal at some later point. I think the latter possibility
is more reasonable, given the data. First, post-positing lā is a clear innova­
tion, while pre-positing ʾal is an inherited feature, shared by all the Semitic
languages (see below). Since, unlike ʾal, lā has other functions beyond verbal
negation, there are scenarios that may explain the post-positioning of lā, but
it would be almost impossible to explain abandoning and re-introducing ʾal
back into the system of some dialects. Finally, Ethio-Semitic, the closest rela­
tive of MSA, shows regular generalization of pre-positive ʾal.
Therefore, I assume a common MSA double negation *ʾal… lā. The later
developments in MSA are according to Jespersen’s cycle, similar to the pro­
cess attested in French, which produced the verbal negation ne… pas, where
pas is slowly taking over as the main negation particle, while ne may be
omitted > … pas (je veux pas vous aider ‘I don’t want to help you’, y va
pas ‘don’t go there’).22 Similarly, we can reconstruct for Ḥarsūsi and Yemeni
Meḫri a basic structure *ʾal V (lā) where ʾal is the main negation and lā is a

­22 See van Gelderen 2008 for a broader discussion of the phenomenon.
 On Verbal Negation in Semitic 25

special negation, used in some specific circumstances.23 While the force of

ʾal slowly reduced, lā became obligatory and finally, in some dialects, ʾal was
dropped completely, leaving post-positive lā as the sole verbal negation: V lā
(Stage III of Jespersen’s Cycle).24
But how did lā get to its post-predicate position in Proto-MSA? One pos­
sibility is that lā was re-introduced into the verbal system from its regular
function as a negative response. The position of lā at the very end of the
sentence, and not necessarily after the element it negates, may support such
a process.25 Regardless of its origin, it should be considered a Proto-MSA
feature. It is important to note that attempting to argue for the originality
of the Ḥarsūsi pattern (…lā) is untenable. In short, we may assume that
MSA generalized ʾal, and went through some subsequent developments in
its daughter dialects.

D) Old South Arabian

OSA uses only ʾ l for verbal negation and ġyr (cf. Arabic ġayr) and rarely dʾ
for the negation of the noun; lā is not attested at all.26
hn ʾ l tʾḫd fḥlt nfshw ‘if he does not give himself up…’ (R 4088/4)
f-ʾ l ymnʿw ‘let them not prevent’ (R 4815/6)

­23 Such a phenomenon is also attested in Ṣanʿānī Yemeni Arabic. Watson (1993, p. 261)
notes that when the full form -ši is used (rather than -š) “it often emphasizes the negative
element”. Although Meḫri is spoken in Yemen, the development in Meḫri is probably not
a result of language contact since the South Arabian dialects in Yemen are spoken around
the border with Oman in West Yemen.
­24 Many languages allow or require co-occurrence of verbal negation consisting of
two negations: verbal and nominal, e.g., Italian: non ho visto nessuno ‘I have seen no-one’
(Weiss 1999). The North Germanic languages are an interesting case in point. Old Norse
used a pre-verbal negation né and a post verbal negation -a(t) from a nominal negation
pattern: Gunnarr né kemr-at ‘Gunnar doesn’t return’ (van Gelderen 2008, ex. 16b).
The post-verbal negation -a(t) may not have been an original negation in Old Norse
(some suggest that it is derived fron nt. *ainata ‘one’), but by the time it was used as a
verbal negation, it was no longer recognized as anything but a negation. Later -a(t) can
appear as the sole negation of the verb: kemr-a nu Gunnarr ‘Gunnarr will not come now’
(van Gelderen 2008, ex. 18 a).
­25 Similarly, several French linguists noted that from its earliest attestations, French pas,
as well as other particles used in similar structures (aucun, point, jamais etc.), were com­
monly used as a negative answer: avez-vous rencontré des amis? aucun ‘have you met any
friends? none’ (Grevisse 1975, p. 937, § 877 a ; Winters 1987, p. 43), though this is not the
common explanation for the development of pas, aucun etc. as verbal negations in French.
­26 There are four examples of lm in the Haramic dialect and some examples of lhm/lm
in Minean before an imperfect, which seem to correspond to Arabic lam with the jussive,
used as the negation of the perfect (kāna – lam yakun): w-lm yġtsl ‘he didn’t wash himself’
(C 523/7). Koogan and Korotayev (1997, p. 239) suggest that this is a borrowing from
26 Na’ama Pat-El

E) Arabic27
In Classical Arabic, the regular negation of the perfect is mā,28 while non-in­
dicative verbs as well as the imperfective29 are negated by lā and variations
thereof (lam, lan, lammā). Mā may also negate the imperfect (Reckendorf
1921, p. 45); however, in this position, the verb is co-temporal with another
verb.30 This is mostly the case when the negated verb is in a subordinated clause.31

Arabic (such a suggestion is not mentioned in Beeston 1984). A borrowing seems reason­
able in light of the small number of examples and the general tendency of this dialect to
borrow North Arabian forms.
­27 Note that the discussion here is regarding morphosyntax, not semantics; hence, it is
inconsequential that a combination such as lam + Jussive is semantically the negation of
the past or that lā + Jussive is the negation of the imperative, but rather that the negation
of these verbal forms is lā or variations thereof and not mā.
­28 Most scholars agree that the interrogative particle mā was grammaticalized to a ne­
gation particle and is not an original Semitic negation particle; scholarly disagreement lies
in which process specifically led to the change interrogative > negation particle. Rubin
(2005, p. 50) suggests a process of grammaticalization: what evil is in my hand > there is
no evil in my hand (see also Wehr 1953). Recently, Al-Jallad (2008) has argued that mā
entered the negation system originally in subordinated clauses, through its function as an
indefinite relative particle. He further showed that there is some evidence from NWS to
support this reconstruction for the Central Semitic branch, not just for Arabic, though
this function of mā in NWS has largely been abandoned. This is a very attractive sugges­
tion which nicely connects Arabic to other Central Semitic languages; however, the data
from the Qūrʾan does not support it since in the Qūrʾan, the negation particle mā, with
very few exceptions, is excluded from subordinated clauses. See also Reckendorf (1921,
p. 45): “Im Nebensatz ist mā verhältnismäßig selten”. Lucas (2007) suggests that mā was
reanalyzed through its function as a grammatical subject in sentences with quasi-verbal
predicates (like fī- etc.).
­29 The indicative status of yaqtulu is not exactly equivalent to qatala. Its semantics is
mostly dependent on other verbs in the clause (simultaneity) or on particles (sawfa), while
qatala has regular independent indicative functions such as past and performative, and, of
course, it has no paradigmatic or historic connection to non-indicative forms, as yaqtulu does.
­30 This is obvious from Reckendorf’s discussion (1921, p. 45), which describes a verb
negated by mā as carrying mostly a present meaning, while past and future are also pos­
sible. It is likely that such a co-temporal usage was the impetus for the spread of mā to
negate all verbs.
­31 There are a number of cases where mā could conceivably be understood as the
interrogative mā, rather than the negation particle. For example: [quli nẓurū māḏā fī
s-samāwāti wa-l-ʾarḍi] wa-mā tuġnī l-ʾāyātu wa-n-nuḏuru ʿan qawmin lā yuʾminūna
(Q 10:101) [Say: “look what is in the heavens and in the earth!”] “But no signs or warn­
ings help people who do not believe”; or: “what help will signs and warnings be to people
who do not believe?” The negative reading is adopted by Arberry (1955), the interroga­
tive reading – by U. Rubin (2005). Such ambiguity is of course the basis of the path of
change suggested by A. D. Rubin (2005). There are, however, two main reasons to reject
it: first, mā functions as a negation particle in other Central Semitic languages (Canaanite,
Aramaic), not just in Arabic, so Arabic inherited the negation particle mā, rather than
developed it. Furthermore, as Al-Jallad (2008) already noted, Rubin offers no syntactic
 On Verbal Negation in Semitic 27

Lā may also negate the perfect, when the perfect express a wish, i.e. when
it expresses modality (Reckendorf 1921, p. 43; Fischer 2001, p. 103, § 182 b,
318). In addition, lā is used to negate nominal forms, alongside the preposi­
tions ġayr, bi-lā and dūna, as well as nominal predicates (lā šakka ‘there is
no doubt’). Arabic is the only West Semitic language where ʾal is not attested
at all.
–– pf. (past tense): mā qāla hāḏā ‘he did not say that’
–– pf. (modal): wa-llāhi lā fataḥtu hāḏā l-bāb ‘by God, I will not open this
door!’; lā ʾaʿǧaltu-ka wa-lā qātaltu-ka wa-lā fataktu bi-ka ‘I will not
force, fight or fatally deceive you’
–– impf.: lā yakrimu s-saḫiyyu l-baḫīla ‘a generous man shall not respect a
stingy one’
–– subj.: li-kay-lā taʾsaw ʿalā mā fātakum ‘that you may not grieve over
what escapes you’; yanbaġī ʾan lā yuḍʿ ifa n-nafsa ‘he must not weaken
the spirit’.
–– Juss.: lā yuḥzin-kumu llāhu ‘may God not cause you grief’
–– energ.: lā taqūlanna ‘don’t say a thing!’
Modal verbs are also negated by lam or lan, two negation particles which are
probably derived from lā. The particle lan negates the subjunctive to express
the negative future, or an ‘emphatic’ negation (lan taf ʿalū ‘you will never do
it’), while lam negates the jussive to express the negative perfect (lam taʿ lam
‘you did not learn’).32 Lan is assumed to be built on a combination of lā
and ʾan, a particle which conditions the subjunctive (Wright 1967, I, p. 287,
§ 362hh; Fleisch 1990, II, p. 476, § 149 t). Both lam and lan are never used
to negate the predicate of a nominal sentence. It seems that the only mor­
phologically marked indicative form negated by lā in Arabic is the imperfect
(yaqtul-u), while other forms negated by lā and its reflexes are clearly modal.
The functions of Arabic lā correlate with some of the functions of Ak­
kadian lā: in both languages, lā is used to express negative imperative or
cohortative (durative in Akkadian; jussive in Arabic), and in both lā is used
in oaths33:
Arab.:
lā raʾayta šarran ‘may you not see evil!’; wa-llāhi lā ʿaṣaytu rabbī ‘by God, I
will not disobey my lord!’

motivation for the change. After all, ambiguous readings are common, but they do not
normally result in syntactic change.
­32 Lam has a poetic variant lam-mā with the same function.
­33 Note that in Akkadian, the verb in oaths of this type is marked as subordinated
(with final -u).
28 Na’ama Pat-El

Akk.:
TUKUL da-me al-su la a-sa-ga-nu ‘I shall not raise the weapon of blood
giš

against him!’ (apud Hasselbach 2005, p. 175).

F) NWS languages
In most NWS languages, ʾal has a very restricted function. It mostly negates
the modal imperfect (“Jussive”) as the negative counterpart of the impera­
tive (when the subject is the 2nd person) or the cohortative and jussive (when
the subject is 1st or 3rd persons). In Ugaritic, Old Aramaic and Hebrew, lā
negates all other verbal forms and is a common nominal negation particle
as well34:
Hebrew35:
u-lə-ʾādām lōʾ māṣāʾ ʿēzer kə-negdô ‘He did not find a companion to Adam.’
(Gen. 2:2)
lōʾ ʾōsīp ləqallēl ʿ ôd ʾet hā-ʾădāmā ba-ʿăbûr hā-ʾādām ‘I will not curse the
earth any more on account of man’ (Gen. 8:21)
­ʾal tabbīṭ ʾaḥărêkā ‘don’t look back!’ (Gen. 19:17)36

­34 Deir ʿAllā has a number of occurrences of l (/lā/) but only two of ʾ l (I:6, 7). Hackett
(1984, p. 104) suggested that one of these occurrences is with a jussive: ʾ l thgy (I:7). The
other instance of ʾ l (šm ḥšk w-ʾ l ngh ʿ lm, I:6–7) is more problematic, since ʾ l occurs before
a noun (ngh ʿ lm ‘eternal light’). Hackett (1984, pp. 43–44, 105) suggests an ellipsis of a
volitive verbal form, since ʾal normally does not negate nouns in NWS. Thus, despite scant
evidence, Deir ʾAllā seems to show the expected NWS situation, similar to Old Aramaic
and Classical Hebrew. Due to the minimal attestation of negation in this language, it will
not be used in this discussion.
­35 The function of ʾ l as a modal negation particle in Hebrew is also emphasized by its
use with the particle nāʾ. ʾal + nāʾ with a verb occurs 19 times in the Bible, where it usually
occurs in proximity to another nāʾ either on the verb or in another syntagm, like hinne-
nāʾ (Qaddari 1985, p. 199; Fassberg 1994, pp. 41–42). Note that lōʾ does not occur before
nāʾ, nor does it negate the cohortative or the jussive in Biblical Hebrew. The particle lôʾ
may function as the negation of the cohortative, but than the verb negated is an indica­
tive verb, rather than subjunctive: lôʾ taʿǎśe (Ex. 20:4) vs. ʾal taʿaś (Gen. 22:12). Qaddari
(1985, pp. 200–201) also mentions some very rare nominal functions of ʾal: wə-yāśēm lə-
ʾal millātî (Job 24:25) My word was set to naught.
­36 The form following ʾal is not necessarily the paradigmatic form of the jussive; how­
ever, the function of ʾal with the imperfect is a vetitive. There are occasional examples of
the jussive preceded by lōʾ rather than the expected ʾal (e.g., 1Kg. 2:6), which Waltke/
O’Connor (1990, p. 567, § 34.2.1 d) suggest is indicative of conflation of the imperfect and
the jussive in the Masoretic tradition. See Hoftijzer (1985) and Huehnergard (1988,
p. 21) regarding the weakening and eventual marginalization of the indicative -n plurals
(< *yaqtulūna).
 On Verbal Negation in Semitic 29

wə-ʾal yədabbēr ʿ immānû ʾălōhîm pen nāmût ‘let God not speak with us di­
rectly, lest we die’ (Ex. 20:19)
ʾal ʾerʾe bə-môt hay-yāled ‘let me not see the child die’ (Gen. 21:16)
Old Aramaic:37
w-l-yšmʾ Mtʿʾ l ‘M. will not obey.’ (S 1 B 21)
zy l-ydʿ ‘one who does not knows (act. participle)’ (S II C8)
w-ʾ l tštq ḥdh mn mly sprʾ zn[h] ‘Let not one word in this inscription be silent.’
(S 1 B 8–9)
Ugaritic:
lm l likt ‘why did you not send?’ (2.63:7)
w mnkm l yqḥ spr mlk hnd ‘no one shall take this royal document’ (2.19:12–13)
ʾal tqrb l bn ʾ ilm mt ‘do not come close to the son of Il, Mot!’ (1.4:VIII:15–16)
Hebrew exhibits a conundrum, where two syntagms, lō yiqtôl and ʾal yiqtôl
appear concurrently. Gesenius (1910, p. 317, § 107 o), and others following
him, perceive the difference as a matter of emphatic strength: lōʾ with the PC
is understood as more emphatic than ʾal with the PC.38 It is possible that the
difference between indicative and non-indicative forms in Biblical Hebrew
evolved to be mostly syntactic, due to the degradation of the morphological
distinctions between the jussive forms and imperfect forms.39 A similar de­
velopment in Aramaic reached its peak in Middle Aramaic, where lā spread
to negate all verbal forms (see below). In Biblical Hebrew, the process did not
reach a natural end and a full spread, à-la Aramaic, because Biblical ­Hebrew

­37 The particle lā is written as a prefix l- in Old Aramaic, Ugaritic and Deir ʾAllā. The
final -ʾ found in other Central Semitic languages is probably an indication of a long vowel,
rather than an etymological consonant (Garr 1985, p. 199).
­38 This is apparently not unique to Hebrew. Huehnergard (1988, pp. 21–22) noted a
similar phenomenon in Old Babylonian: ul tanaddin ‘you will not give’ vs. lā tanaddin
‘do not give’. Huehnergard suggests that the Hebrew and Akkadian evidence suggests
that the difference between yaqtul and yaqtulu/yaqattal is not mood, but rather aspect:
yaqtul is punctual (perfective) and yaqtulu/yaqattal durative (imperfective). Thus, the dif­
ference between indicative and injunctive “had to be marked in some way other than by
the morphology of the verb and probably occurred at the level of the phrase” (ibid. p. 21).
­39 Qimron (1986, p. 80, § 400.13 b) notes that the pentateuchal prohibitions with lōʾ are
normally formulated with ʾal in Qumran Hebrew. See for example: lōʾ tassîg gəbûl rēʿăkā
(Deut. 19:14) ‘do not move your neighbor’s boundary’ vs. ʾ l tsg gbwl ʿwlm (Pr 22:28) ‘do
not move the eternal boundary’. Qimron (1983, p. 479) suggests that the reason for the
elevated use of ʾal in some Qumran documents in what seems to contradict the Biblical
usage may be explained by the loss of mood in the Hebrew of that period. He further sug­
gests that the lack of ʾal in Aramaic at that time rendered this particle a more literary air.
30 Na’ama Pat-El

ceased to exist as a spoken variant before the conclusion of the process. The
existence of two types of negation with the PC synchronically must have
resulted in functional differentiation, which Gesenius formulates in terms
of emphasis. The question is, of course, which one of them is original and
which developed later.
The only consistent exception to the NWS set is Phoenician, which seems
to lack lā,40 and instead uses bal for all indicative verbal negations. Bal is a
known nominal negation particle in Semitic (cf. Akkadian balu- ‘without’)
and a rare verbal negation in Ugaritic and Hebrew. The particle ʾal is used
with the modal imperfect to express a negative imperative:
Phoenician:
bl pʿ l ‘he did nothing’ (KAI 24:2)
bl tdrkn ‘you do not tread’ (KAI 27:8)
ʾ l ʾ l tptḥ ʿ lty w-ʾ l trgzn ‘do not open my sarcophagus and do not upset me’
(KAI 13:3–4)
Therefore, except for OSA, the distinction in the Central Semitic set of nega­
tion particles is indicative / non-indicative. Arabic has a rich modal system,
where this is evident; while the modal system reduced substantially and so
did the function of ʾal, the difference is at least minimally kept: the jussive is
negated by ʾal to express the negative imperative and all indicative verbs are
negated by a separate particle (lā or bal).

3. The problem

Akkadian and Central Semitic have a set of two particles corresponding to

two verbal functions, while the other languages use for the most part only
ʾal and variations thereof; however, in MSA both particles are used for verbal
negation, which means that two particles, ʾal and lā, are attested in every
branch of Semitic. In addition, if we reconstruct only lā, we are faced with
the problem of explaining why ʾal is the most common particle in verbal
negation in WS, missing only in Arabic, while lā is missing from three lan­
guages independently.41 Since all branches attest to a dual set of verbal nega­

­40 This is the common view. For a possible example of lā in Phoenician, see Gibson
(1982, III, p. 109). In Pat-El (forthcoming) lā is also argued to have been retained in
Phoenician for nominal negation.
­41 Contra Garr (1985, p. 175) who claims that most Semitic languages prefer to use lā
to negate their finite verbs. As the data above shows, in fact most Semitic languages prefer
ʾal for this function.
 On Verbal Negation in Semitic 31

tion, even if only as a relic, it is more reasonable to reconstruct a dual set for
Semitic: lā and ʾal.
This then raises two questions: which verbal distinction do these particles
denote and what is the function of each particle within the set. In addition,
it is important to substantiate why ʾal is preferred as a generalized negation
marker across WS branches. These questions are further complicated as Ar­
abic and NWS show two contradictory sets, where in Arabic, lā is the nega­
tion of the non-indicative, while in most NWS languages it is the negation of
the indicative. The following discussion will present a solution, accounting
for the issues raised above and explaining subsequent developments in the
daughter languages.

4. Discussion

The Akkadian distinction matrix/non-matrix is unknown in WS.42 There­

fore, let us assume that the functional set indicative/non-indicative is at least
Proto-WS, and potentially Proto-Semitic. This distinction is solidly attested
in CS and is in line with the fact that most Semitic languages maintain a mor­
phological or syntactic distinction between indicative and non-­indicative
forms. Such a set is very often reduced to a minimal distinction: indicative-
imperative, where the only unambiguous modal form is the imperative and
other modal distinctions are marked syntactically.43
As was shown above, most CS languages have two particles, but MSA,
OSA and Ethiopic have only one, which is almost exclusively ʾal. It may be
beneficial, therefore, to look at languages, which generalized one of their
negation particles during their attested history, that is, Arabic and Aramaic.
Both nodes show changes, loss of features and innovations, which can by and

­42 The only exception is found in the modern Ethio-Semitic languages. Some of these
languages have a different negation for subordinated verbs vs. independent verbs; however,
subordinated and negated verbs do share some similarities: in which show two types of
imperfects, simple and compound with an auxiliary; see for example Amharic: yämmə-
tsäbr set ‘a woman who breaks’, a-tsäbr-əmm ‘she does not break’ (Leslau 1967, p. 121).
The compound imperfect in Amharic is təsäbr-alläčč ‘she breaks’. These distinctions seem
to reflect a secondary distribution and function assignment, which developed in the Ethi­
opic branch and does not go back to a previous node. Another possible exception is Clas­
sical Arabic, where mā is generally excluded from subordinated clauses, while lā can occur
in both main and subordinated clauses. Since mā is a secondary development, this too may
be a local innovation.
­43 Such is the case in many Neo-Arabic dialects, where various moods are marked
with prefixes of various origins. These prefixes very often originate from auxiliary verbs,
originally conveying modality. For an overview and discussion of modal prefixes in Neo-
Arabic see Stewart (1998).
32 Na’ama Pat-El

large be dated. Moreover, Arabic dialects show a variety of changes, which

may hold clues to preferences in early Semitic.44 In the following pages I will
attempt to describe and discuss the changes in each of these dialect groups.
These changes may be indicative of the changes which preceded the earliest
attestations of MSA, OSA and Ethiopic.

A) Aramaic45
Old and Official Aramaic dialects exhibit the expected WS set: indicative lā /
non-indicative ʾal.
Official Aramaic:
EgA: hn nḥt l-Mnpy ʾ l tšbq l-ʾšn ‘If you come down to Memphis, do not leave
[anything] to ʾNŠ.’ (C 42:11)
w-hn lʾ yhb hmw lk šlḥ ʿ ly ‘If he did not give them to you, inform me.’ (C 42:10)
There are some indications that ʾal was still used in Middle Aramaic:
Qum: w-kʿn ʾ l tdḥl ʾnʾ ʿmk ‘therefore, don’t be afraid. I’m with you’ (1Q20, 22:30)
However, in some Middle Aramaic dialects and later, ʾal disappears and lā
is used for all functions, including to form the negative imperative. Some
Aramaic dialects innovated new negation particles, which are based on lā
(Pat-El 2006), but their distribution is not conditioned by verbal modality:
Middle Aramaic:
TO: lā tidḥal ʾAbrām ‘fear not, Abraham’ (Gen. 15:1 Cf. Hebrew: ʾal tîrāʾ ʾAbrām)
Late Aramaic:
Syriac: magīstros dēn ʾemar da-šrabhōn d-hāllen lā ta[d]kar lī ‘the magistrate
said: don’t remind me of the affairs of these people’ (Joshua the Stylite, 76:3–4)
The particle ʾal is not attested in any Aramaic dialect later than the first cen­
tury.46 Additionally, Neo-Aramaic dialects use lā for all verbal and nominal

­44 The similarity in the structure of the Semitic languages makes this type of “internal
typology” more helpful, and by far more reliable than typology. Repeated developments
have been pointed to in many Semitic languages, the most well known example being the
development of the “perfect” in Late Aramaic and Neo-Aramaic dialects from a deverbal
adjective, similar to the original WS perfect, also from a deverbal adjective (Akkadian sta­
tive). I will discuss more examples in a monograph to be published by Brill soon.
­45 See also Folmer (1995, p. 520) for a brief review of the distribution of verbal nega­
tion particles until and including Official Aramaic.
­46 Beyer (1984, p. 152) notes that the last attestation of ʾal in Aramaic is in 37 BCE, in
a text from Qumran; therefore, he suggests that ʾal was replaced by lā by the first century.
 On Verbal Negation in Semitic 33

functions, like Late Aramaic dialects. Note the following from the Jewish
dialect of Arbel:
ʿUmar ʾ ixālan la k-xilwāle ‘Umar didn’t eat our food’ (Khan 1999 B:102)
la qatlinnox ‘I won’t kill you’ (Khan 1999 B:149)
la zdimun ‘don’t be afraid’ (Khan 1999 Y:182)
Of the original Aramaic set ʾal/lā, the particle which was used for indicative
negation, lā, spread to all positions, despite the fact that the verbal system
in Middle Aramaic and later dialects carried the same tense-mood distinc­
tions: perfect/imperfect-imperative. The particle which was used to negate
the non-indicative verbal form, ʾal, was completely abandoned. Hence we
may formulate the change as: indicative lā > indicative/non-indicative lā.47

B) Arabic48
In Classical Arabic two particles are used: mā, primarily for the perfect, and
lā for modal verbs (jussive, subjunctive and energic) and the imperfect; how­
ever, already Middle Arabic shows signs that lā is dropping out of natural
use, while the use of mā spreads. In Judeo-Arabic, lā is hardly ever used for
the negation of verbal forms and the particle mā negates the imperfect and
the imperative, which are negated by lā in Classical Arabic. Other particles,
like lam, are rare and were probably not used in the vernacular (Blau 1995,
p. 142).
Some Neo-Arabic dialects generalized mā for all indicative verbal forms,
and lā was restricted to the negation of the imperative. For example, in Le­
vantine Arabic, mā negates the perfect and imperfect: ma ʾakalna šī ‘we
didn’t eat a thing’, ma raḥ-qūl šī ‘I won’t say a thing’. The particle lā is still
used, but it is restricted to the imperative; however, mā can also be used in
the negation of the imperative: ma tinsa hayda ‘don’t forget it!’, lā tqūl hayk
‘don’t say it!’.49 That is, lā is slowly pushed away from the verbal system.
In some Neo-Arabic dialects, the process described above for Levantine
Arabic is already complete. In Cairene Arabic mā …š is used as a general ne­
gation particle, while lā is not normally used for verbal negation (Woidich

­47 Some Neo-Aramaic dialects also use na for negation, though rather rarely. This par­
ticle is probably of Persian origin.
­48 There have been several pragmatic studies on Arabic Negation. Dahlgren (2006)
compared the use and distribution of mā and lam in the Qurʾān and concluded that lam “is
used in more varied contexts than mā” (p. 74) and that mā “is mainly used in main clauses
of realis declarative and conditional sentences” (pp. 74–75). See also Larcher 1994.
­49 I am informed by my colleague, Dr. Mahmoud Al-Batal, that the use of lā rather
than mā, for the negation of the imperative in Lebanon depends on dialectal preferences.
34 Na’ama Pat-El

1968; 2006). The particle lā/laʾ is used as a negative response, or as an opening

to a contradiction, as opposed to ʾaywa ‘yes’ (Woidich 2006, pp. 346–347)50:
–– pf.: mā rafaʿt-əš ʾīdi ‘I didn’t raise my hand’ (Brustad 2000, E10)
–– impf.: mā bi-yiʿgibū-š il-ʿagab ‘nothing pleases him’ (Brustad 2000, E3)
–– imp.: mā tidfaʿī-š ʾaktar min mitēn ‘don’t pay more than two hundred’
(Brustad 2000, E1)
A similar situation is attested in Moroccan Arabic, where only mā …š is
used to negate the verb: mā-ža-š ‘he didn’t come’, ma-nemšiw-š ‘we will not
go’, ma-temši-š ‘don’t go!’. The imperative, however, may be negated by lā
as well as mā. Harrell (1962, p. 153) suggests that the use of lā is for moral
admonishment: la-temši-š ‘you shouldn’t come’.51 In Moroccan, lā is found
otherwise only in nominal contexts: ma-klit la xubz w-la lḥem ‘I ate neither
bread nor meat’. Note that the verb is negated by mā, but the nominal ob­
jects are negated by lā.
In most Arabic dialects lā was either restricted to negate only the im­
perative, or was pushed out of the verbal system altogether. Therefore we
may say that lā, the particle originally used for non-indicative negation, was
either severely restricted or has been abandoned. We may also note that the
dialects, where the indicative negation particle, mā, spread to all positions,
are not in contact with each other and the process happened in each one of
them independently. Thus, here too we observe the generalization of the
indicative negation particle mā for all positions52: indicative mā > indicative/
non-indicative mā.

­50 Note, however, that lā has not disappeared from Cairene Arabic; the negation par­
ticle is still used in restricted and sometimes marginal constructions, like wala (< wa-lā),
which negates the element following it (Woidich 2006, pp. 342–344), and it is still at­
tested in the Classical pattern “no … and no …” or “neither … nor …” (Woidich 2006,
pp. 345–346; Brustad 2000, pp. 310–311). Brustad (2000, p. 295) notes that in some dia­
lects in Egypt lā is still used for negated imperative. She also notes that in dialects that
use both lā and mā to negate the imperative, the difference between them is pragmatic.
Brustad suggests that when lā …š is used to negate an imperative, it “carries less impera­
tive force”. This could be a similar distinction as the one found in Hebrew and Akkadian;
however, this is most probably not an original distinction in Egyptian Arabic, but rather
one that developed there independently. Furthermore, it seems that at least in those Leba­
nese dialects where both mā and lā are used to negate the imperative, there is no such
pragmatic distinction.
­51 Cf. Gesenius’ comments (1910, p. 317, § 107 o) about the difference between ʾal and
lôʾ negating the imperative in Biblical Hebrew, where according to him ʾal with the SC
denotes a simple imperative (don’t steal), while lôʾ with the SC is a more emphatic negation,
suitable for divine commands (you shall not steal).
­52 The negation of the non-verbal predicate is identical to the negation of the indicative
verbal predicate in all the dialects, e.g. Moroccan: huwa ma-ši hna ‘he’s not here’.
 On Verbal Negation in Semitic 35

C) Provisional conclusions
The same type of change is observed in both Aramaic and Arabic: the old in­
dicative negation particle (Aramaic lā, Arabic mā) was generalized at the ex­
pense of the non-indicative particle, which was abandoned in Aramaic and
in several Arabic dialects. Even in Arabic dialects, which did not lose their
non-indicative negation, its function has reduced significantly. If Arabic and
Aramaic are representative of the reduction in the negation system, we may
conclude that the indicative negation tends to be generalized.53
Following this conclusion, the generalization of ʾ l, found in Ethio-­Semitic,
MSA and OSA, suggests that ʾ l was originally the indicative negation parti­
cle. We may note that lā was normally not preferred for generalization un­
less special conditions occurred.54 A system where lā is the non-indicative
negation particle fits well with the set of negation particles found in Arabic,
where lā is the negation of the non-indicative verb,55 and in Akkadian, where
lā functions as the negation of the prohibition, when negating the durative
(e.g., lā terrubī ‘do not enter’).
Therefore, we may reconstruct a set: ʾ l (indicative) / lā (non-indicative).
This system is attested almost intact in Arabic and if we agree that ʾal and
ula are somehow related, than this set is also attested in Akkadian. Many of
the remaining languages (Ethio-Semitic, MSA, OSA, Neo-Arabic and Late
and Neo-Aramaic) have generalized their indicative negation particle.56 Such
­53 Note that the process of generalization is not necessarily related to loss of morpho­
logical distinctions in the verbal modal system; Gəʿəz and Amharic have morphologically
distinct forms of the jussive and Neo-Arabic and Neo-Aramaic distinguish between in­
dicative and subjunctive, though these languages do not use distinct forms of negation. A
contrary example is Biblical Hebrew, where the jussive and imperfect are indistinguish­
able morphologically in the sound verb, but the syntactic difference between the jussive
and imperfect is retained for all forms in the negation. Similarly Akkadian, see footnote 34.
­54 For Aramaic see B) above; for Ḥarsūsi it was suggested that lā was preferred as it
was the second element of a double negation pattern (“Jespersen’s Cycle”). See above for
details.
­55 As was mentioned above, the Arabic system has exceptions, which suggest that it
is breaking down. In section 2.E it was shown that lā may also negate a morphologically
indicative verb, whereby a modal meaning is adduced, while mā is attested negating an
imperfect, when the verb is co-temporal with another verb. Since mā became the general
negation particle it was suggested above that mā + impf. was the motivation for the spread
of mā to negate all verbs (while, of course, the modal system was eroding). Therefore, the
Arabic system should be considered an intermediate phase between the inherited WS sys­
tem and the new modern Arabic one.
­56 The conclusion suggested here that the choice of negation particle is dependent on
mood seems to be rare typologically. Aikhenvald and Dixon (1998), for example, claim
that polarity, the ability to negate a clause, is positioned at the beginning of the depend­
ency hierarchy (ibid., p. 68) and is not dependent on any other category (e.g., number, case
etc.). Both polarity and mood are associated with the clause (ibid., p. 73) and both are at
36 Na’ama Pat-El

a set, therefore, accounts for the situation in the daughter languages, except
the NWS ones, which will be dealt with in the next section.57

D) Negation in NWS
All NWS languages use ʾal as the negation of the imperative and most use
lā as the negation of the indicative, which is the exact opposite of the situ­
ation in Arabic and Akkadian and contradicts the reconstruction for WS
suggested above. Obviously, the NWS languages did not simply invert the
function of the particles, without any motivation. Since all NWS languages
show the same functionally inverted set, the change there must have oc­
curred in Proto-NWS.
CS inherited the WS system; however, while Arabic retained lā for the ne­
gation of non-indicative verbs, OSA generalized the indicative ʾal to all posi­
tions. It is possible that this generalization was shared by its sister language,
Proto-NWS. Nevertheless, like in other languages, lā remained functional
in the nominal system and probably also as a negative response.58 This func­
tion of lā is not evident in the OSA inscriptions and we must, therefore, as­
sume that lā was lost completely in this branch. OSA uses only prepositions
(bly/blt) for nominal negation.59
NWS then “borrowed” lā back from the nominal system to function as
its indicative negation. The incentive for this innovation was probably the
negation of the nominal predicates with lā.60

the upper level of the dependency hierarchy Aikhenvald and Dixon have devised. The
positive is always the unmarked term, while the negative is always marked. This seems to
be correct in Semitic; however, the choice which negation particle is used is dependent on
mood. Note also that in a previous study (Pat-El 2006), I have shown that in Syriac nega­
tion marks more distinctions than positive statements.
­57 As a point of comparison, we should mention Greek and Latin, both of which have
a dual negation set similar at least in part to what is proposed here. In Greek, mē is called
prohibitive while ou(k) is the negation of facts (indicative). Moreover, like in Akkadian (ul)
and Amharic (al-…əmm), the indicative negation, ou(k), is the negation in the protasis of
conditional sentences (Seiler 1971, p. 81). Similarly, in Latin, nē is used with imperatives
and optatives while nōn is used with indicative and potential subjunctive (Gildersleeve
1895, p. 287). Note that nōn is derived from nē. As was argued above, this is probably not
the case in Semitic.
­58 Such a situation is attested in MSA and many Neo-Arabic dialects.
­59 Akkadian uses balu(m) and lā; Arabic uses ġayr and lā; Ethiopic uses ʾ i and ʾal; MSA
uses be, biśi and other particles; NWS uses ʾ i, lā and bal-. Kurdish Neo-Aramaic dialects
use both lā and bē ‘without’.
­60 In other Semitc languages, verbal and nominal negation often overlap. Some use a
preposition additionally: This function of lā as the negation of nominal predicates is sup­
ported by evidence from Akkadian and Arabic, where the predicate of a nominal sentence
 On Verbal Negation in Semitic 37

Biblical Hebrew61:
lōʾ šālēm ʿăwōn hā-ʾĕmōrî ‘The iniquity of the Emorites is not complete’
(Gen. 15:16)
lōʾ ṭôb had-dābār ʾăšer ʾattā ʿōśe ‘What you are doing is not good’ (Ex. 18:17)
Ugaritic:
yn d l ṭb ‘bad wine (lit. wine which is not good)’ (4.213:1)
d l nʿm ‘one who is not pleasant’ (2.50:19)
Aramaic:
l-ṭb hʾ mk ‘he is not better than you’ (S III 22)
zy l-ydʿ ‘one who does not know (ptcl.)’ (S II C 8)
A similar process of introducing features from nominal system into the ver­
bal system probably accounts for the sporadic use of bal as a verbal negation
in both Ugaritic and Hebrew (for Phoenician, see below), and for the process
of nominal negation > verbal negation accounts also for Gəʿəz ʾ i-.62 Both
particles are nominal negations in most languages; bal is a negative prepo­
sition in Ethiopic and Akkadian and ʾ i- is used as a nominal negation in a
number of languages.63
The introduction of a new indicative negation particle pushed ʾal to an in­
creasingly restricted function, subsequently to be lost in Aramaic, but main­
tain a specialized function in Old Aramaic, Ugaritic and Biblical Hebrew.
is negated with lā: la KUR Ḫamata šunu ‘they are not from Ḫama (ABL 331 r. 5); lā
d-dāru ǧāmiʿatun ‘the house is not united’ (Reckendorf 1921, p. 44).
­61 It is quite possible that the allomorphy found in Hebrew with the PC (ʾal yqtl and
the “emphatic” lā yqtl) may be an indication that Hebrew never quite lost lā, but rather
restricted it to prohibitions. The re-introduction of the nominal lā and the subsequent
restriction of ʾal caused a functional differentiation and individuation (lā being a “more
emphatic” prohibition), which is generally lost in later phases of Hebrew, as the system
stabilized.
­62 Note that introduction of elements from the nominal system into the verbal system
is found even in the morphology of the Semitic languages. E.g., in Samaritan Aramaic,
the 2nd person perfect suffix on III-weak is -k, whereas it is normally -t in Aramaic (and
generally in CS). This new suffix is probably borrowed from the 2nd person possessive
suffix, which is -k everywhere in CS (Goldenberg 1977, p. 478). A similar movement of
-k from the nominal system to the verbal is attested in some Ethio-Semitic languages as
well (Apple­yard 1996, p. 214) and in Neo-Assyrian: parsāka < parsāta (von Soden 1995,
p. 122, § 75 c). Another type of movement from one system to another is attested in Old
Assyrian, where sometimes the pronominal possessive suffix for 1cs -ī is used instead of
the pronominal objective suffix -nī: lá-mı -̀ dı ́ ‘teach me’ (Hecker 1968, p. 75, § 49).
­63 Hebrew ʾ î-nāqî ‘not clean’ (Job 22:30); Phoenician ʾ bl /ʾi-bal/ ‘no’; Old Aramaic
ʾ-šm ‘no-name’ (S I C 24). Similarly Soqotry ʾē.
38 Na’ama Pat-El

Syntactically, the re-introduction of lā as an indicative negation particle is

a rather significant innovation compared to Arabic and OSA, which either
retained the Proto-WS system or generalized one of its members. Note the
following table, which lists the changes stage-by-stage:

Table 1: reconstruction of the negation set CS > NWS

CS P-NWS NWS
nominal *lā *lā
indicative *ʾal lā
non-indicative *lā *ʾal ʾal

Some further changes are observable later in NWS. Phoenician, the only
NWS language which does not use lā as a verbal negation, “borrowed” from
its nominal system a different nominal negation particle: bal. This particle
is found as a rare verbal negation particle in Biblical Hebrew and Ugaritic,64
but it never gained status as a regular verbal negation in these two l­ anguages.65
Despite the differences between Phoenician and the other languages, the
NWS system remained otherwise intact, with the basic distinctions between
indicative and non-indicative negation still functional. The Phoenician situ­
ation is thus reminiscent of the Classical Arabic system, where the indicative
negation particle ʾal was replaced by mā, but otherwise the system marks the
same distinctions inherited from an earlier phase.
The innovation of ʾal as negation of the imperative sets NWS apart from
other CS languages. This type of innovation is another indication that Ara­
maic is closer to Canaanite than to Arabic, which does not share this inno­
vation.66 The innovation in NWS, however, does not affect the position of
OSA, as loss of a feature cannot be used as a diagnostic tool.

­64 For example, Biblical Hebrew: yuḥan rāšāʿ bal lāmad ṣedeq ‘grace is given to the evil
person, [though] he does not learn justice’ (Is. 26:10); Ugaritic: bl išlḥ… ‘I will not send’
(1.14:V:21).
­65 Since Aramaic did not participate in this innovation, the process of introducing bal
into the verbal system must have happened later than the re-introduction of lā. It, there­
fore, is unlikely that there was a competition in these languages between lā and bal which
ended with Phoenician generalizing bal, while the other languages generalized lā.
­66 Other innovations of NWS are (1) double plural of segholates nouns (a-insertion and
sound plural); (2) verbal templates C qattila and D haqtila; (3) 1pl suffix -nū̌ on the perfect,
but -nā ̌ as object and possessive suffix. This features were proposed by Huehner­gard
(1991, 2005) on the basis of Hetzron (1976), with some adjustments. Note that the inno­
vation in Phoenician (bal) could be a lexical replacement, which is a rather weak evidence
for subgrouping.
 On Verbal Negation in Semitic 39

Note the following genetic tree, which shows the developments from
Proto-Semitic to the daughter Semitic languages:

Chart 1: Negation of the finite verb in Semitic

PS
? indep. *ʾal/ul
? dep. *lā

WS ES
ind. *ʾal Akk.
non-ind. *lā main ula/ul
subord. lā

Ethiopic MSA
G. ʾ i Meh./Soq. al
Amh. ʾal Ḥar. lā CS

CA OSA NWS
ind. mā ʾl ind *lā
non-ind. lā non-ind. *ʾal

Can. Ug. Aram.

ind. l- ind. l-
non-ind. ʾ l non-ind. ʾ l
BH Ph.
ind. lōʾ ind. bl
non-ind. ʾal non-ind. ʾ l

What immediately strikes one is the similarity of this genetic tree to the
one suggested by Hetzron (1976) and modified by Porkhomovsky (1997)
and Huehnergard (1991; 2005), and in fact it validates the modification
of Hetzron’s proposal.67 Hetzron suggested that Arabic and Canaanite

­67 Note that MSA and Ethiopic do not share the same innovations (Eth. ʾ i- for ver­
bal negation; MSA double negation and post-positive lā). OSA generalized the negation
40 Na’ama Pat-El

(Arabo-Canaanite) branched off from CS, separating from Aramaic. The

innovation proposed here of indicative lā, non-indicative ʾal is shared by Ca­
naanite, Aramaic and Ugaritic, but not by Arabic. This is consistent with
Huehnergard (1991), who posited NWS (Aramaic, Canaanite, Ugaritic)
which branched off from CS, thus separating from Arabic and OSA.68
A few additional notes are in order: first, this paper argues that nega­
tion should be used as a diagnostic feature for subgrouping; however, not all
changes can be used for this aim. The loss of lā in both OSA and Ethiopic
is not a proof of genetic affinity between them; while Gəʿəz innovated by
using ʾ i to negate all verbs, in OSA and modern Ethio-Semitic ʾal is a reten­
tion.69 Indeed, both Ethiopic and OSA lost lā as a result of generalization,
but loss and retention cannot be used for subgrouping. In any case, OSA did
not share the innovation attested in Gəʿəz and therefore is not part of its im­
mediate node as far as this feature is concerned.
Second, we tend to think of lā as the quintessential Semitic negation, but
in fact most MSA languages, OSA, Phoenician and all the Ethio-Semitic lan­
guages do not use it at all. On the other hand, the particle ʾal is missing only in
Arabic and Akkadian. So, if anything, ʾal is the most common negation parti­
cle in Semitic. Is it related to lā, as Blake, Pardee, Lipiński and others sug­
gested? It seems unlikely, since they have distinctive functions and different
distribution in early Semitic, hence, a development lā > ʾal, if it happened at all,
happened prior to Proto-Semitic, as all Semitic languages attest to a dual set.
Third, the connection between Akkadian ul(a) and WS ʾ l seems to be
only a conjecture, albeit rather tempting, at this point. On the one hand, the
functions of Akkadian ul(a) at least partially correspond to those in our re­
constructed WS ʾ l. On the other hand, from the data in most WS languages
(minus OSA) we can safely reconstruct the vowel in WS to be /a/, while the
Akkadian one is /u/.70 Consequently, and despite the compelling consonan­
tal and syntactic evidence, I remain agnostic.

­particles ʾal as well as most Modern Ethio-Semitic languages, but it is problematic to rely
on generalizations in a set of two elements. Huehnergard (2005, pp. 168–169) illustrates,
using the suffix-conjugation ending k/t, why such two-way leveling is an invalid argument
for subgrouping.
­68 OSA did not share the innovation in Arabic (indicative mā) nor the one in NWS
(indicative lā, non-indicative ʾal).
­69 Moreover, note that the innovation of ʾ i to negate verbs in Ethiopic is restricted to
North Ethiopic (Gəʿəz, Tigre and Tigrinya), while in South Ethiopic ʾal is used as the ver­
bal negation. It is, therefore, possible that the generalizations and innovations took place
in this branch internally, perhaps long after it was split from other WS languages.
­70 Even with a possible OSA *ʾul, we would have to assume, following Blake (1911), an
ad hoc vocalic PS *l, which appears only in this form and which is realized randomly as a
different vowel in the daughter languages.
 On Verbal Negation in Semitic 41

Finally, even if we only reconstruct a functional set, and not necessarily a

formal set, the possibility that PS had a dependent vs. independent distinc­
tion needs further support, mostly by understanding the function of Ak­
kadian -u/na and the function and distribution of the verbal modal forms. I
leave it for now as a hypothetical possibility.

5. Conclusions

I have suggested that verbal negation in Semitic was a originally set contain­
ing two particles with distinct functions: *ʾ l was used to negate indicative
verbal forms, while *lā was used to negate non-indicative ones. Since the
relationship between Akkadian ul(a) and WS ʾ l is still not resolved and since
Akkadian negation particles mark distinctions that WS does not (namely,
main vs. non-main predication), I cautiously suggest that this reconstruction
is Proto-WS, rather than Proto-Semitic.
Many languages generalized the indicative negation particle to negate all
verbal forms, which subsequently led to the loss of the original syntactic
distinction. The NWS languages show a further innovation, where after
an initial generalization in P-NWS (apparent in OSA), the nominal nega­
tion lā was re-introduced from the nominal system into the verbal system
for negation of the indicative. I have suggested that the incentive for the re-
borrowing was the function lā has across Semitic as the negation particle of
nominal predicates.
Finally, and perhaps most importantly, the reconstruction suggested
above confirms the validity of Hetzron’s tree, and its later modifications by
Porkhomovsky and Huehnergard. It was further argued that the NWS
set (non-indicative ʾal, and lā/bal for indicative forms) should be considered
a diagnostic feature, which connects Phoenician, Aramaic, Ugaritic, He­
brew and Deir ʿAlla (NWS), and separates them from OSA and Arabic. An
innovation on a smaller scale separates North Ethiopic (< innovation of *ʾ i-
for verbal negation) from South Ethiopic (< retention of *ʾal).
42 Na’ama Pat-El

Abbreviations

AA Afro-Asiatic OAkk Old Akkadian

Aram. Aramaic OSA Old South Arabian
BH Biblical Hebrew PC prefix conjugation
CA Classical Arabic Ph. Phoenician
Can. Canaanite PS Proto-Semitic
CS Central Semitic SC suffix conjugation
ES East Semitic TO Targum Onkelos
MSA Modern South Arabian Ug. Ugaritic
NWS North West-Semitic WH interrogative
OB Old Babylonian WS West Semitic.

Bibliography

Aikhenvald, A. Y./R. M. W. Dixon 1998: “Dependencies Between Grammatical

Systems.” In: Language 74, pp. 56–80.
Al-Jallad, A. 2008: “A Syntactic Account of the Development of the Negative
Adverb ma(h) in Central Semitic.” Paper read at thirty-sixth annual meeting
of the North American Conference on Afroasiatic Linguistics in Chicago, IL.
Appleyard, D. L. 1996: “Ethiopian Semitic and South Arabian: Towards a Re-ex­
amination of a Relationship.” In: Israel Oriental Studies 16, pp. 203–228.
Arberry, A. J. (1955). The Koran Interpreted. New York.
Beeston, A. F. L. 1984: Sabaic Grammar. Manchester.
Bergsträsser, G. 1914: Verneinungs- und Fragepartikeln und Verwandtes im Kūrʾān.
Leipzig.
Beyer, K. 1984: Die aramäischen Texte vom Toten Meer. Göttingen.
Blake, F. R. 1911: “Vocalic r, l, m, n in Semitic.” In: JAOS 31, pp. 217–222.
Blau, J. 1995: A Grammar of Medieval Judaeo-Arabic [Hebrew]. Jerusalem.
Bond, O. 2007: “Towards a Canon for Negation.” In: P. K. Austin/O. Bond/​
D. Nathan (eds.): Proceedings of Conference on Language Documentation
and Linguistic Theory. London, pp. 39–49.
Bron, F. 1994: “Note sur la négation en sudarabique épigraphique.” In: Matériaux
arabes et sudarabiques 6, pp. 183–185.
Brustad, K. E. 2000: The Syntax of Spoken Arabic: A Comparative Study of Moroc-
can, Egyptian, Syrian, and Kuwaiti Dialects. Washington.
Cohen, M. 1970: Traité de langue amharique (Abyssinie). Paris.
Croft, W. 1991: “The Evolution of Negation.” In: Journal of Linguistics 27, pp. 1–27.
Dahl, Ö. 1979: “Typology of Sentence Negation.” In: Linguistics 17, pp. 79–106.
Dahlgren, S. O. 2006: “Sentential Negation In Arabic”. In: L. Edzard/J. Retsö
(eds.): Current Issues in the Analysis of Semitic Grammar and Lexicon II.
Wiesbaden, pp. 64–78.
Dillmann, A./C. Bezold 2003 [reprint of 1907]: Ethiopic Grammar. London.
 On Verbal Negation in Semitic 43

Faber, A. 1991: “The Diachronic Relationship between Negative and Interrogative

Markers in Semitic”. In: A. S. Kaye. (ed.): Semitic Studies in Honor of Wolf
Leslau on the Occasion of his Eighty-Fifth Birthday. Wiesbaden, pp. 411–429.
Fassberg, S. E. 1990: “Negative Purpose Clauses in Biblical Hebrew: wə-lōʾ yiqtôl,
pen yiqtôl” [Hebrew]. In: M. Goshen-Gottstein/S. Morag/S. Kogut.
(eds.): Studies on Hebrew and Other Semitic Languages Presented to Profes-
sor Chaim Rabin on the Occasion of his Seventy-Fifth Birthday. Jerusalem,
pp. 273–294.
­—  1994: “NĀ and Its Cognates” [Hebrew]. In: S. E. Fassberg (ed.): Studies in Bibli-
cal Syntax. Jerusalem, pp. 36–73.
Fischer, W. 2001: A Grammar of Classical Arabic. Tr. J. Rodgers, New Haven.
Fitzmyer, J. A. 1995: The Aramaic Inscriptions of Sefire. Rome.
Fleisch, H. 1990 [reprint of 1961]: Traité de philologie arabe. II vol. Beyrouth.
Folmer, M. L. 1995: The Aramaic Language in the Achaemenid Period: A Study in
Linguistic Variation. Leuven.
Forston, B. W. 2004: Indo-European Language and Culture. Oxford.
Garr, W. R. 2004 [reprint of 1985]: Dialect Geography of Syria-Palestine, 1000–586
B.C.E. Winona Lake, IN.
van Gelderen, E. 2008: “Negative Cycles.” In: Linguistic Typology 12, pp. 195–243.
Gesenius, W./E. Kautzsch et al. 1910: Gesenius’ Hebrew Grammar. Oxford.
Gibson, J. C. L. 1971: Textbook of Syrian Semitic inscriptions. Oxford.
Gildersleeve, B. L. 1895: Latin Grammar. London.
Goldenberg, G. 1966: The Amharic Tense System. PhD diss. Jerusalem.
­—  1977: “The Semitic Languages of Ethiopia and their Classification.” In: BSOAS 40,
pp. 461–507.
Grevisse, M. 101975: Le bon usage. Grammaire française avec des remarques sur la
langue française d’aujourd’hui. Paris.
Hackett, J. 1984: The Balaam Text from Deir ʿAllā. Chico, California.
Harrell, R. S. 1962: A Short Reference Grammar of Moroccan Arabic. Washington.
Hasselbach, R. 2005: Sargonic Akkadian. A Historical and Comparative Study of
the Syllabic Texts. Wiesbaden.
Hecker, K. 1968: Grammatik der Kültepe-Texte. Rome.
Hetzron, R. 1974: “La division des langues sémitiques.” In: A. Caquot/D. ­Cohen
(eds.): Actes du premier congrès international de linguistique sémitique et
­chamito-sémitique. Paris 16–19 juillet 1969. Paris, pp. 181–194.
­— 1976: “Two principles of Genetic Reconstruction.” In: Lingua 38, pp. 89–198.
Hoftijzer, J. 1985: The Function and Use of the Imperfect Forms with Nun Para-
gogicum in Classical Hebrew. Assen/Maastricht.
Hoftijzer, J./K. Jongeling 1995: Dictionary of the North-Semitic Inscriptions.
2 Vol. Leiden.
Huehnergard, J. 1988: “The Early Hebrew prefix-Conjugations.” In: Hebrew
Studies 29, pp. 19–23.
­— 1991: “Remarks on the Classificatioon of the Northwest Semitic Languages.” In:
J. Hoftijzer/G. van der Kooij (eds.): The Balaam Text from Deir ‘Alla Re-
Evaluated. Leiden, pp. 282–293.
44 Na’ama Pat-El

­— 2005: “Features of Central Semitic.” In: A. Gianto (ed.): Biblical and Oriental
Essays in Memory of William L. Moran. Rome, pp. 155–203.
­— 2009: “Trees and Waves: On the Classification of the Semitic Languages.” Paper
read at thirty-seventh annual meeting of the North American Conference on
Afroasiatic Linguistics in Albuquerque, NM.
Johnstone, T. M. 1975: “The Modern South Arabian languages.” In: Afroasiatic
Linguistics 1(5), pp. 1–29.
­—  1977: Ḥarsūsi Lexicon and English-Ḥarsūsi Word-List. London.
­—  1981: Jibbāli Lexicon. London.
­—  1987: Mehri Lexicon and English-Mehri Word-List. London.
Khan, G. 1999: A Grammar of Neo-Aramaic: The Dialect of the Jews of Arbel. Leiden.
Kienast, B. 2001: Historische semitische Sprachwissenschaft. Wiesbaden.
Kogan, L. E./A. V. Korotayev 1997: “Sayhadic (Epigraphic South Arabian).” In:
R. Hetzron (ed.): The Semitic Languages. London, pp. 220–241.
Larcher, P. 1994: “Mā faʿala vs. lam yaf ʿal: une hypothèse pragmatique.” In: Ara-
bica 51, pp. 388–415.
Leslau, W. 1967: “Hypothesis on a Proto-Semitic Marker of the Imperfect in Gur­
age.” In: JNES 26, pp. 121–125.
­—  1995: Reference Grammar of Amharic. Wiesbaden.
Lipiński, E. 2001: Semitic Languages: Outline of a Comparative Grammar. Leuven.
Lucas, C. 2007: “Jespersen’s Cycle in Arabic and Berber.” In: Transactions of the
Philological Society 105(3), pp. 398–431.
Mallory, J. P./D. Q. Adams 2006: The Oxford Introduction to Proto-Indo-­
European and the Proto-Indo-European World. Oxford.
Palmer, F. R. 2001: Mood and Modality. Cambridge.
Pardee, D. 2003–2004: Review on J. Tropper: Ugaritische Grammatik. In: AfO 50,
pp. 1–404.
Pat-El, N. 2006: “Syntactical Aspects of Negation in Syriac.” In: JSS 51(2),
pp. 329–348.
­— forthcoming: “On Negation in Phoenician.” In: R. Holmstedt/A. Schade
(eds.): The Phoenician Language: Studies in Writing and Linguistics. Winona
Lake, IN.
Porkhomovsky, V. 1997: “Modern South Arabian languages from a Semitic and
Hamito-Semitic perspective.” In: Proceedings of the Seminar for Arabian
Studies 27, pp. 219–224
Qaddari, M. Z. 1985: “Studies in Diachronic Syntax: the Negation Particle ʾal”
[Hebrew]. In: Studies in Language I. Jerusalem, pp. 197–210.
Qimron, E. 1983: “The Negation Particle ʾal in Ancient Hebrew Sources” [He­
brew]. In: M. Bar-Asher et al (eds.): Studies in Language presented to Zeev
Ben Ḥayim. Jerusalem, pp. 473–482.
­—  1986: The Hebrew of the Dead Sea Scrolls. Atlanta.
Reckendorf, H. 1921: Arabische Syntax. Heidelberg.
Rosenbaum, G. M. 2002: “The Particles ma and lam and Emphatic Negation in
Egyptian Arabic.” In: W. Arnold/H. Bobzin (eds.): “Sprich doch mit deinen
Knechten aramäisch, wir verstehen es!” Wiesbaden, pp. 583–598.
 On Verbal Negation in Semitic 45

Rubin, A. D. 2005: Studies in Semitic Grammaticalization. Winona Lake, IN.

­—  2010: The Mehri Languages of Oman. Leiden.
Rubin, U. 2005. The Qurʾān [Hebrew]. Tel Aviv.
Seiler, H. 1971: “Abstract Structures for Moods in Greek.” In: Language 47(1),
pp. 79–89.
Simeone-Senelle, M.-C. 1994: “La négation dans les langues sudarabiques mo­
dernes.” In: Matériaux arabes et sudarabiques 6, pp. 187–211.
­— 1997: “The Modern South Arabian Languages.” In: R. Hetzron (ed.): The Se-
mitic Languages. London, pp. 378–445.
von Soden, W. 1995: Grundriss der akkadischen Grammatik. Rome.
Stein, P. 2003: Untersuchungen zur Phonologie und Morphologie des Sabäischen.
Rahden.
Stewart, D. J. 1998: “Clitic Reduction in the Formation of Modal Prefixes in the
Post-Classical Arabic Dialects and Classical Arabic sa-/sawfa.” In: Arabica
45/1, pp. 104–128.
Stroomer, H. 1999: Mehri Texts from Oman: based on the field materials of
T. M. Johnstone. Wiesbaden.
Testen, D. 1998: Parallels in Semitic linguistics: The Development of Arabic la- and
Related Semitic Particles. Leiden.
Tropper, J. 2000: Ugaritische Grammatik. Münster.
Voigt, R. M. 1987: “The Classification of Central Semitic.” In: JSS 32(1), pp. 1–21.
Wagner, E. 1953: Syntax der Mehri-Sprache unter Berücksichtigung auch der ande-
ren neusüdarabischen Sprachen. Berlin.
Walker, D. A. 1896: The Semitic Negative, with Special Reference to the Negative
in Hebrew. Chicago.
Waltke, B. K./M. P. O’Connor 1990: An Introduction to Biblical Hebrew Syntax.
Winona Lake, IN.
Watson, J. C. E. 1993: A Syntax of Ṣanʿānī Arabic. Wiesbaden.
Wehr, H. 1953: “Zur Funktion arabischer Negationen.” In: ZDMG 103, pp. 27–39.
Weiss, H. 1999: “Duplex Negatio non semper Affirmat: a theory of double negation
in Bavarian.” In: Linguistics 37(5), pp. 819–846.
Winters, M. E. 1987: “Innovations in French Negation.” In: Diachronica 4(1),
pp. 27–52.
Woidich, M. A. 1968: Negation und negative Sätze im Ägyptisch-Arabischen.
München [Inaug.-Diss.].
­— 2006: Das Kairenisch-Arabische. Eine Grammatik. Wiesbaden.
Wright, W. 31967: A Grammar of the Arabic Language. Cambridge.