An Amazigh substrate in Moroccan Arabic:
A sociolinguistic reconstruction of agentives1
Ahmed Ech-Charfi
Mohammed V University in Rabat
ﻣ ﺧص
ﺣﯾث،ﺗﺗطرق ھذه ا دراﺳﺔ ﺗﺻﺎل ﺑﯾن ا ﺗﯾن ا ﻣﺎزﯾ ﯾﺔ وا رﺑﯾﺔ ا ﻣ رﺑﯾﺗﯾن وﺗرﻛز ﻰ أﺳﻣﺎء ا ﺎل ﻰ ا ﺻوص
ﯾﺗ ق ا وﺟﮫ ا ول ﺑﺻﯾ ﺔ " ﺎل" ا ﺗﻲ ﺗم ﺗوﺳﯾ ﮭﺎ.ﺗ ص إ ﻰ أن ھﻧﺎك ﺗﺄﺛﯾر ﻣﺎزﯾ ﯾﺔ ﻰ ا رﺑﯾﺔ ﯾﺗﺟ ﻰ ﻲ أرﺑ ﺔ أوﺟﮫ
و ﻛن ﻛذ ك ﻛﺛﯾرا ﻣن ا ﺻ ﺎت، ﻛﻣﺎ ھو ا ﻣر ﻲ ا ﺻﺣﻰ وﺑﺎ ﻲ ا ﮭﺟﺎت ا رﺑﯾﺔ،ﺗﺷﻣل ﯾس ط أﺳﻣﺎء ﻣ ﺗ ف ا ﺣرف
أﻣﺎ ا وﺟﮫ ا ﺛﺎﻧﻲ ﯾرﺗﺑط ﺑﺎﺳﺗ ﻣﺎل ﻣﯾم "ﻣ ﺎل" ﺷق أﺳﻣﺎء ا ﺎل ﻣن ا ﺎل. ﻛﻣﺎ ھو ا ﺣﺎل ﻲ ا ﻣﺎزﯾ ﯾﺔ،ا ﺛﺎﺑﺗﺔ وا ﺎدات
ﯾﺗﺟ ﻰ ﻲ وﺟود ﺻﯾ ﻣ ﺗ ﺔ ﺗ ﺗﻣد ا وﺟوه ا ﺛ ﺛﺔ ا ﺳﺎﺑ ﺔ، أﻣﺎ ا وﺟﮫ ا راﺑ. وھو ا وﺟﮫ ا ﺛﺎ ث، ﺑل ﺣﺗﻰ ﻣن ا ﺳﻣﺎء،ا ﺛ ﺛﯾﺔ
ﺣﺎو ت ا دراﺳﺔ إ ﺎدة ﺑﻧﺎء ﺗﺎرﯾ ا ﺗﺻﺎل ﺑﯾن ا ﺗﯾن ﻰ أﺳﺎس أن ا ﻣﺎزﯾ، وﺑﻧﺎء ﻰ ھذه ا ﻣ طﯾﺎت.ﻛن ﺑﻣ ﻧﻰ واﺣد
ﻣﻣﺎ ﺷﺟ ﮭم ﻰ ﺗﺣوﯾل ﺳﻣﺎت ﻛﺛﯾرة ﻣن ﺗﮭم،ا واﺋل م ﯾﻛن دﯾﮭم ﻣﺎ ﯾﻛ ﻲ ﻣن ا ﺗﺻﺎل ﺑﻣﺗﻛ ﻣﯾن رب ذ ا ﺔ ﻧﮭم
ﻣﻣﺎ ﺷﺟ ا ﻣﺎزﯾ، ﯾر أن ﯾﺎم ا دو ﺔ ا درﯾﺳﯾﺔ وﺗﺄﺳﯾس ﺎﺻﻣﺗﮭم ﺎس ر ا رﺑﯾﺔ إ ﻰ ﻣﻧز ﺔ أ ﻰ.ا م إ ﻰ ا ﺔ ا ﮭدف
" و ﮭذا أو وا اﺷﺗ ﺎق أﺳﻣﺎء ا ﺎل ﺑﺎﺳﺗ ﻣﺎل ﺻﯾ ﺔ "ﻣ ﺎل" واﺳﺗﺑد وھﺎ ب " ﺎل.ﻰ ﺗ ﺿﯾل ا ﺻﯾ ا ﺻﺣﯾﺣﺔ ﻰ ا د ﯾ ﺔ
و د ازداد ھذا ا ﺗوﺟﮫ.ا ﺗﻲ ا ﺗ دوا أﻧﮭﺎ ا ﺻﯾ ﺔ ا ﺻﺣﯾﺣﺔ ر م أﻧﮭﺎ ﺗﺣﻣل ﻲ اﺳﺗ ﻣﺎ ﺗﮭﺎ ﻛﺛﯾرا ﻣن ﺻ ﺎت ﻧظﯾرﺗﮭﺎ ا ﻣﺎزﯾ ﯾﺔ
.ﺑ د ا ﮭﺟرات ا ﮭ ﯾﺔ
Abstract
This paper proposes to approach the history of Arabic-Amazigh contact on the basis of the (ir)regularities
of Moroccan Arabic participles in comparison with Amazigh agentive nouns. It is argued that MA has
developed a category of agentives copied from Amazigh. Four pieces of evidence are advanced in support
of an Amazigh substrate in this category: the extension of the “fə al” pattern, originally associated with
occupations, the use of the prefix “m” with active participles derived from triliteral verbs, the use of the
same prefix with participles derived from nouns, and the survival of variations with a mixture of these
processes. On the basis of these facts, a sociolinguistic reconstruction of the Arabic-Amazigh contact is
attempted. The “m” prefix is argued to be an early morphological transfer from Amazigh, when not
enough Arabic input was available to Amazigh bilinguals. After the foundation of Fez as the capital of the
Idrisids, correct Arabic became more prestigious, a fact which probably encouraged the erasure of salient
Amazigh loans, among which was the agentive “m” prefix. The development of the “fə al” class was
probably an attempt by the bilinguals to use what they thought was the correct Arabic participial form,
though in fact it was a copy of the Amazigh agentive class. This tendency was further invigorated by later
Hilali migrations.
1
I would like to thank Karim Bensoukas and Ángeles Vicinte for having read and commented on a draft of this
paper. Discussion with Bensoukas, in particular, in connection with earlier versions contributed to introducing many
changes that resulted in the present version. It goes without saying that all shortcomings remain mine.
©
The International Journal of Arabic Linguistics (IJAL) Vol. 4 (2018), Issue 1 (special) (pp. 103-137)
0.
Introduction
Although Arabic and Amazigh have been in contact in North Africa for more than a
millennium, the linguistic effects of this contact are not obvious and not much research has been
done so far to unearth them. This is more true in the case of Arabic than in that of Amazigh:
whereas lexical and grammatical borrowings from Arabic into Amazigh are easily noticeable
even for ordinary speakers, the effect of Amazigh on Arabic tends to be less visible. So far, only
short lists of Amazigh loans have been identified in North African Arabic by different researchers
(e.g. Tilmatine 1999), and many of the grammatical aspects suspected to be of Amazigh origins
are contested. Perhaps the best example of such aspects is the reduced vowel system of Moroccan
Arabic (MA), first advanced to be a clear result of the Amazigh influence only to be doubted later
(cf. Aguade 2010). What complicates the situation further is that the early contact between the
two languages is historically obscure: there are very few documents from that period, and later
sources are hardly reliable. Therefore, the reconstruction of the prehistory of North African
Arabic promises to be very challenging.
This paper will attempt to reconstruct the formation of MA and trace the Arabic-Amazigh
contact through the morphology of participles. Participles in MA and agentive nouns in
Moroccan Amazigh (MAm) are derived by more than one morphological process, some of which
will be argued to result from contact between the two languages. Particular attention will be
given to the prefix “m-” found both in MAm and in a class of MA agentive nouns; this formal
similarity will be claimed to be the source of some irregularities in MA participles. On the basis
of such linguistic clues, sociolinguistic aspects of the early Arabic – Amazigh contact will be
analyzed, relying on insights from contact linguistics and historical sociolinguistics. The basic
claim is that the Amazigh agentive prefix “m-” was borrowed by early Arabic – Amazigh
bilinguals because of lack of authentic input, but this morphological process was later on
terminated and its effects gradually undone, probably because of the rise of the political influence
of Arabs, despite their small number in comparison with Amazigh natives. The termination of the
“m-” prefixation, however, did not lead to the adoption of Arabic participial forms but, rather, to
the development of a new agentive category on the model of Amazigh agentive nouns. The new
category used the “fə al” pattern but copied the semantic functions of the corresponding
Amazigh class.
104
The paper will be articulated as follows. Section 1 provides a theoretical overview of the
interaction between linguistic, historical and sociolinguistic knowledge, with a focus on North
Africa, and expounds some principles and methods of sociolinguistic reconstruction. Section 2
presents briefly the agentive formation in MAm. Section 3 advances arguments for an Amazigh
substrate in the formation of MA participles. Section 4 is an attempt at sociolinguistic
reconstruction. Section 5 recapitulates and concludes.
1.
Linguistics, history and sociolinguistic reconstruction
In contact areas like North Africa, where different peoples speaking different languages
have come into contact since antiquity, it is not possible to trace the origin of a particular
structure or the etymology of a lexical item without testing the plausibility of the hypothesis
against facts of history. When facts of history are well-studied and safely supported by reliable
sources or compelling archeological evidence, contact linguistics becomes dependent on history;
but when a relevant period is still obscure or when the events reported by later sources are
unreliable, history can be dependent on linguistic and dialectological research as much as these
are dependent on history. Such is the current state of our knowledge about the Arabization of
North Africa since the 7th century CE.
Research on the Arabic-Amazigh contact is still full of controversies despite its relatively
long history, which started more than a century ago and was developed by colonial
dialectologists. Very often, lack of historical knowledge about the migration and settlement of
different groups opens the gate for speculation, thus weakening all possibilities for consensus.
Given this situation, it is not surprising that some dialectologists turned into historians. Caubet
(1998) raises the issue, reviewing a number of works by early French dialectologists of North
Africa, and stressing the important contributions they have made to historical knowledge. In a
subsequent paper, Caubet (2004) makes the same point again with a focus on the light that
sociolinguistics can shed on dialect contact and koinézation that may have happened in the past,
for there is no reason to believe that the Arabic spoken by early or later settlers was uniform. The
Amazigh languages themselves were diverse and some of their speakers at least were as nomadic
and mobile as the conquerors (e.g. the case of “Botr” discussed in Camps 1983). But although
controversy still reigns supreme, there seems to be at least one point researchers agree on,
namely, that Arabic-Amazigh contact forms layers corresponding to contact between different
105
groups in different periods of history. This consensus is summarized by Lévy (1995:59), who,
after discussing some aspects of the Jewish Arabic dialect of Tafilalt, concludes:
On ne peut faire de dialectologie sans histoire de peuplement, sans histoire
politique […], sans histoire économique […], culturelle (Zawiyas), soit sans chercher les
facteurs du changement linguistique (arabisation, re-berbérisation), sans établir les
problèmes du bilinguisme, du contact des parlers, des niveaux de langues …2
In the case of Morocco, he asserts firmly that “dialectologie arabe et berbère sont intimement
liées”3 and that we cannot do one without the other.
Amazigh speakers (i.e. Berbers) are the original inhabitants of North Africa, but the
region has known successive waves of colonization. Greeks and Phoenicians founded colonies
along the Mediterranean and the Atlantic coasts and, later on, the Romans managed to lay their
control over part of inlands. When the might of Rome weakened, the Visigoths, a Germanic
people, also invaded the parts that had been annexed by the Romans. The last power to control
these parts before the Arab conquest was the Byzantine Empire. The cultural and the linguistic
influence of these peoples on the local inhabitants is uncertain and, although some scholars have
identified a number of Latin loanwords in Amazigh (cf. the references in Bougchiche 1997), we
will probably never manage to reconstruct any good number of grammatical borrowings or
semantic calques from the languages of those ancient colonizers. But that does not mean that the
contact had no impact. Heath (2015), for example, who cites some dialectologists who suspected
a Late Latin substratum in Amazigh and North African Arabic, claims that the analytic genitive
marker in MA is a borrowing from this variety. This demonstrates the difficulty to trace substrata,
superstrata or adstrata with any degree of certainty4.
The Arab colonization itself did not happen in a single period but extended over many
centuries; nor were the Arab settlers of one and the same stock. Generally, two waves of Arab
migrants are recognized, following a classification originally made by Ibn Khaldoun. The first
wave is associated with the early conquest which started in the mid-7th century CE but was not
complete in the case of Morocco till the first decades of the 8 th century. The second wave
2
“We cannot do dialectological research without referring to the history of settlement, without political […],
economic […] or cultural (Zawiyas) history; that is, without investigating the factors of linguistic change
(Arabization, re-Berberization), or a good understanding of the problems of bilingualism, dialect contact, stylistic
levels, etc.”
3
“Arabic and Berber dialectology are intimately connected.”
4
Tilimatine (1999:101) notes in the same vein that “Derrière un mot d’apparence arabe, peut en effet se cacher un
mot berbère, qui lui-même peut provenir du grec par l’intermédiaire du latin.” [Behind an apparently Arabic word
may lie a Berber word, which may itself be of Greek origin but entered the language through Latin]
106
corresponds to Bedouin migrations of the period between the 11 th and the 13th centuries. In terms
of dialects, the first conquerors are assumed to have spoken sedentary varieties of Arabic while
the later migrants are known to speak Bedouin varieties. One of the major differences between
the two groups concerns the pronunciation of the phoneme corresponding to the grapheme ق,
which is realized as [q] or [ ] in the first but as [g] in the second groups of dialects. A finergrained classification, however, could identify differences in each of the two groups. Marçais
(1938), for instance, notes a number of isoglosses separating urban from rural sedentary dialects,
arguing that the rural varieties exhibit many Amazigh substratal features indicating, thus, that the
speakers of this group of dialects were probably Amazigh speakers who shifted toward Arabic.
This can be a good example of what dialectology can contribute to social history. A similar
classification of the Bedouin group could also identify further categories within the group (cf.
Boukous 1995; see also the papers in Aguadé, Cressier and Vicente 1998). Whether these minor
categories correspond to specific groups of migrants or to stratal residues will remain open to
further research. Lévy (1998:194) warns that “c’est une tautologie que de rappeler que les parlers
d’une même couche historique ne présentent pas les mêmes traits, ce qui en ferait un seul et
même dialecte”5.
Faced with this very complex situation and a drastic shortage of historical sources about
the early contact of Arabic and Amazigh, an attempt will be made to use what we know about
sociolinguistic variation and contact-induced change to reconstruct some aspects of the formation
of early MA and the role that Amazigh might have played in the process. The comparative
method has long been recognized to have limitations not only in an adequate reconstruction of
proto-languages, which is its primary goal, but also in tracing social domains in which
innovations happened and the limits of their propagation (cf. the papers in Durie and Ross 1996,
Romaine 1982, among others). Criticism leveled against this method targeted some of its
fundamental concepts among which are the very notions of “language” and “dialect” and the
family tree model they gave birth to. It has been noted, for instance, that innovations are often
assumed to happen in different languages and that, when enough innovations occur in a dialect,
that dialect may diverge into a distinct language. Linguists, however, have recorded many cases
in which linguistic change extends over many languages (i.e. language areas). Similarly,
5
“It is a tautology to state again that varieties from the same historical layer do not exhibit the same features which
would make of these varieties part of a single dialect.”
107
sociolinguists tend to consider language varieties as mere abstractions over patterns of human
behavior among groups of individuals at different levels of social organization. Therefore, the
domain of linguistic change should not be a language variety but rather a speech community
socially defined (e.g. in terms of social networks). From this perspective, tracing the propagation
of linguistic innovations in history amounts to tracing the formation and/or the disintegration of
social networks, irrespective of whether they were speaking the same or different
dialects/languages (sometimes referred to simply as “lects”).
The fundamental principles of sociolinguistic reconstruction have been, and are still
being, developed within what is known as “socio-historical linguistics” or “historical
sociolinguistics”. Since the seminal work by Romaine (1982), a lot of researchers have tried to
develop new methodologies, building on previous ones, in order to reconstruct past
sociolinguistic situations, something that techniques of the comparative method alone cannot
achieve. In this respect, work by Ross (1996, 1997, 2005), Croft (2000), Toulmin (2009, 2012),
Trudgill (2010), among others, is considered to provide the main tenets and the method for
sociolinguistic reconstruction (for application of this approach to Arabic, see Magidow 2013). It
is beyond the scope of this paper to discuss these works; instead, I will use the “Social Network
Model”, as developed by Ross (1997). Ross focuses on the propagation of particular innovations,
which he refers to as “PEvents”. These are instantiated by a set of utterances, referred to as
“PNetworks”, which need not be structurally homogeneous. A change in the structure of a speech
community constitutes a “speech community event” (SCEvent) which may affect the domain of
future innovations. The primary mission of the historical sociolinguist is to reconstruct the
prehistory of a language through a sequence of SCEvents. Toulmin (2012:507-8), one of the
adepts of this model, summarizes its method in the following steps:
IReconstruct linguistic innovations;
II- Scrutinise in as much detail as possible the dialectological range of the
innovations.
III- Evaluate whether the innovations are diagnostic of Propagation Events. Look
at: linguistic complexity of the innovation, ecological distinctiveness of the novel
variant, and the historical sociolinguistic plausibility of propagation within the range.
IV- Investigate whether there are linguistic grounds […] or textual grounds for a
particular sequencing of changes.
V- Investigate socio-historical grounds for sequencing the PEvents in a certain
way. [….]
VI- Bring together and harmonise the sequencing of PEvents based on linguistic,
textual and socio-historical criteria, within a unified account of language history.
108
It is obvious that linguistic reconstruction in this model goes hand in hand with the reconstruction
of the speech community. In other words, instances of linguistic change and their range are used
as clues to uncover the borders and the dynamism of the community.
In this paper, although various aspects of the Arabic-Amazigh contact will be invoked, the
focus will be particularly on the morphology of active participles corresponding to Amazigh
agentive nouns. The (ir)regularities of this class will be used as clues to reconstruct some stages
in the formation of MA.
2.
Amazigh agentive nouns
The agentive morpheme is no more productive in varieties of MAm. Relying on the few
agentive nouns that have survived in the language, some scholars (cf. Aspinion 1953, Chafik
1991, Bensoukas 1994, 2014; Krim 2013) have tried to reconstruct agentive noun formation in
Amazigh. According to them, agentive nouns are derived from verbs or other nouns by the
prefixation of the agentive morpheme “m-”. In cases where the root contains a labial sound, the
prefix usually changes into the corresponding dental nasal, most likely as an instance of
dissimilation. Here are some examples:
1- amksa “shepherd”
ks
“to shepherd”
amnaj “rider”
ni
“to ride”
anzdam “wood collector”
zdm “to collect wood”
amsaγ
sγ
“buyer”
“to buy”
In these and similar cases, the derived nouns express the meaning of agentivity, irrespective of
whether the verb is transitive or intransitive. The derivation, however, is not limited to verbs
taking subjects with agent roles, but can occur also with other verb types. This fact is exemplified
by the following:
2- amzdaγ “inhabitant”
zdγ
“live”
amaraj “lover”
iri
“love/want”
amzwaru “first”
zwr
“outrun”
amggaru “last”
gur
“come last”
In these examples, the agentive nouns do not denote agents in the strict sense of the term. It
seems that “m-” was subject to semantic extensions that included secondary uses and, thus,
separated it gradually from its primary sense. This kind of extension is attested in many other
109
languages, including English and Arabic (compare, e.g. English “lover”, “villager”, etc. and
Arabic “muħibb” (lover), “ ārif” (knower)).
What is of more relevance to the present study is that the Amazigh agentive affix was also
used to ascribe properties to entities, a function that is not usually associated with the equivalent
Arabic construction. These are some examples:
3- amẓluḍ “poor”
ẓlḍ
“to be poor”
amʒʒuḍ “bald”
ʒʒḍ
“to be bald”
amddallu “debased”
ddull
“to be humiliated”
anfraraγ “askew”
frγ
“to be askew”
Apparently, this type of extension opened the door to further innovations that included even
patients. A few of such cases, such as those in (4), did survive in the language, suggesting that the
semantic load of “m-”, though being prototypically agentive, included also non-agentive
meanings:
4- amazan “messenger”
azn
“send”
amaẓẓal “circumcised”
ẓẓl
“circumcise”
amuḍin
“patient”
aḍn
“be ill”
imnsi
“dinner”
ns
“spend the night”
ls
“wear”
timlsit
“clothes”
When we take into consideration the fact that Amazigh does not have a passive equivalent of
agentive nouns, we may get an explanation why various meanings gradually became associated
with one and the same morphological construction. In other words, in order to refer to a
messenger, for example, using periphrasis, one has to use a clause like “wən i-tt-wazanən” (the
one who is sent) or some other complex and awkward structure. In cases like those under 4,
semantic extension of an extant morphological form was preferred to periphrasis.
Unsurprisingly, this system of derivation must have lasted for some time after contact
with Arabic. Evidence supporting this conclusion comes from Amazigh agentive nouns based on
Arabic roots, as in the examples under 5:
5- amḥsad
“envious”
amsjjḥ
“wanderer”
anʒdam
“leprous”
amʃḥaḥ
“stingy”
anḍ uf
“weak”
amnnakṛ
“denier”
amʃγal
“worker”
am awn
“helper”
110
All these items, and many others more, are derived from Arabic roots. The speakers who derived
these and similar words must have been bilingual in the two languages to a considerable degree.
This is so not because some Arabic items were borrowed in Amazigh, but because the derivation
requires a mastery of the morphological process of the receiver language as well as an ability to
extract roots from lexical material in the donor language. This remark will be returned to later.
Apart from this obsolete morphological system, MAm varieties today use the template
CCCaC borrowed from Arabic to form agentive nouns (but see Krim 2013 for a slightly different
opinion). The template was probably abstracted from the large number of Arabic borrowings
referring mainly, but not exclusively, to various occupations, like those under 6 below:
6- agzzar
“butcher”
anʒʒaṛ
“carpenter”
axyyaṭ
“tailor”
afllaḥ
“farmer”
abnnaj
“mason”
asbbaγ
“painter”
aγḍḍaṛ
“traitor”
aqmmaṛ
“gambler”
Some similar agentive nouns derived from Amazigh roots are given under 7:
7- axwwan
“thief”
anbbaḍ
“ruler”
akrraz
“plower”
agnnaj
“tailor”
azddam
“woodcutter”
aʃwwal
“reaper”
It should be noted, however, that this alternative process itself is not very productive in the sense
that it is not often applied to Amazigh roots, especially that speakers are dependent on Arabic for
the expression of novel activities and occupations. But for our purposes, suffice it to note that
contact between Amazigh and Arabic has resulted in grammatical borrowing of a morphological
pattern from the second language into the first as well as the disposal of the equivalent Amazigh
pattern.
With this brief diachronic and synchronic sketch of the morphology of agentive nouns in
Amazigh, we can now tackle the corresponding classes in MA to consider the extent to which
contact has resulted in mutual influence.
3.
Moroccan Arabic participles
3.1-
Form and meaning
Participles in MA, as is the case in most varieties of Arabic including the standard, can be
either active or passive (cf. Harrell 1962; Heath 1987, 2002). Both of them are derived from
111
verbs, and their forms vary, depending on whether the verb root is triliteral or quadriliteral. When
the verb is triliteral, the active participle takes the form CaCC and the passive mCCuC; but when
it is quadriliteral, the active and the passive participles have the same form, namely mCCCC. For
lack of space, weak verbs and irregular cases (e.g. kla “eat”, qṛa “read”, ba “sell”) will be
overlooked since the processes involved in their derivation are not very relevant to the points
being made here. The following examples illustrate the two classes of forms:
8- Verb
Active participle
Passive participle
ḍrb
“hit”
ḍarəb
məḍrub6
γlb
“win”
γaləb
məγlub
krkb
“roll”
mkərkəb
mkərkəb
mṭərṭəq
mṭərṭəq
ṭərṭəq “explode”
As can be noticed, the active and the passive participle forms for quadriliteral verbs have
merged. In Modern Standard Arabic (MSA), as was the case in Classical Arabic (CA), the
difference between the two is expressed by different vowels in the last syllable, namely ‘i’ for the
active participle (e.g. mustahlik “consumer”) and ‘a’ for the passive (e.g. mustahlak
“consumed”). The reduction (or deletion) of vowels in closed syllables in MA (irrespective of
whether this process was inherited from an old Arabic variety or developed after contact with
Amazigh) resulted in a semantic ambiguity of the outcome forms such as the last two examples
under (8).
Apart from the complexities of their formal derivation, active and passive participles have
a number of functions. One of the most important of these functions is to express aspect, when
combined with verbal auxiliaries, particularly “kan” (to be). The kind of verbs that can occur in
such constructions, as well as the aspect that is expressed with each type, is too complicated to
deal with here at any significant detail. Again, suffice it to report that MA is not different from
CA or other Arabic varieties in this respect (cf. Brustad 2000; Fassi Fehri 2004; Mughazy 2005;
Cowell 1964; Boneh 2010; Prochazka and Batan 2016; Hallman 2017; for MA, see Caubet
1991). The following are examples illustrating this function:
6
The schwas in the MA examples throughout the paper are epenthetic and are usually inserted to break illicit
consonant clusters. The mechanisms of schwa insertion are discussed by Benhallam (1989); some dialectal variations
in the phenomenon are discussed in Ech-Charfi (2008).
112
9- kan ṛaʒə
10- kan qaṛi
l-ktab
was returning
was reading Def-book
(He was on his way back)
(He had already read the book)
As can be noticed, the participles in these examples have verbal functions in the sense that they
denote activities rather than entities. But while the participle in (9) denotes an on-going activity
(i.e. continuous aspect), that in (10) expresses the completion of the activity in relation to the past
reference point (i.e. past perfect). It seems that the appropriate interpretation of such uses of
participles depends on verb semantics, linguistic context as well as pragmatic information (cf. the
references cited above).
Where MA seems to diverge from CA and other Arabic dialects concerns mainly the
nominal functions of participles, particularly that of agentive nouns. It is this aspect of the
language that probably exhibits an Amazigh substrate, as will be argued immediately. This
hypothesis will be backed by four arguments expounded in the following sub-sections: the
emergence of a distinct pattern for the derivation of agentive nouns (3.2); the use of the prefix
“m” with triliteral verbs (3.3); the derivation of agentive nouns from nominal roots (3.4); and the
survival of variation in the derivation of some agentive nouns (3.5).
3.2- The agentive pattern “fəʕʕal” in MA
In CA, most agentive nouns have the same pattern as active participles. From the triliteral
root “ lm”, for example, we get “ ālim” (scientist), and from the augmented form “ llm”, we get
“mu allim” (teacher) (cf. Ryding 2005; Wright 1895). A small class of nouns, however, have the
pattern CaCCāC (i.e. fa āl), namely, those denoting some occupations. Examples of this class
include “fallāḥ” (farmer), “naʒʒār” (carpenter), “ṭabbāx” (cook), etc. According to Arab
grammarians, this pattern is not productive (cf. Hassan, n.d.). In modern Arabic varieties,
however, the class of CaCCāC nouns has been extended to various degrees. Cowell (1962:30),
for example, cites some cases from Syrian Arabic which normally correspond to active participle
equivalents in CA. These include “ṛa āṣ” (dancer), “bajjā ” (seller), “sawwā ” (driver)7, which
are meant to be illustrative rather than exhaustive. Similar examples are also found in other
Middle Eastern varieties.
7
Viz. “rāqiṣ”, “bā i ” and “sā iq”, respectively.
113
In comparison, MA has not only extended the CCCaC class, but also made of “fə al”
one of the most productive agentive patterns, if not the most productive one. As a result of this
productivity, the pattern now is no longer limited to occupations only, but also expresses all sorts
of meanings that are not strictly speaking agentive. The following are some examples illustrating
some of these meanings (see Appendix A):
11- xəwwaf
“fearful”
xaf
“to be afraid”
nə as
“sleeping”
n əs
“to sleep”
wəkkal
“gluttonous”
kla
“to eat”
What distinguishes these examples from their corresponding active participles is that they do not
denote activities in a particular point in time, but rather habits associated with the referent. In this
respect, such cases are similar to occupation nouns with which they share the same pattern. It is
probable that the pattern was originally extracted from occupation nouns illustrated above and
extended its meaning in ways exemplified by the items in (11).
While the pattern CCCaC is basically associated with triliteral roots, in which case the
second consonant is geminated, it has served also for the derivation of agentive nouns from
quadriliteral roots, as in the following examples:
12- bərgag
“spy”
bəznas
“drug dealer”
qəfqaf
“fearful”
ʃəmkar
“street boy”
sərbaj
“waiter”
ʃəfnaʒ
“doughnut maker”
təmtam
“stutterer”
təftaf
“sb. trembling”
Most of the items that undergo this pattern of derivation apparently have non-Arabic origins; they
are either borrowed (e.g. ‘sərbaj’ from French ‘service’) or coined, mainly by a process of
reduplication that is still productive in MA; hence, their informal character. The only difference
between the cases in (11) and those in (12) is that the second consonant is doubled in the first,
since the base is triliteral, but not in the second, where the base has four consonants. In both
cases, however, one and the same template is used as a basis for derivation.
In what respect does this development bear witness to the influence of the Amazigh
language? The morphology and the lexical material of agentive nouns discussed and illustrated so
far are all of Arabic origin. Even the extension of the functions and semantics of the occupation
noun pattern can be argued to result from internal factors leading to reanalysis. The recording of a
similar phenomenon in other Arabic varieties, though to a comparatively lower degree, can serve
114
as evidence in favor of such an argument. The problem, however, is that the similarity between
MA and MAm agentive nouns is too striking to be explained by unidirectional influence only. As
was explained in Section 2 above, MAm has completely lost its original pattern of agentive nouns
and replaced it by the very CCCaC pattern that has become prevalent in MA. In order to defend
the hypothesis that grammatical borrowing happened unilaterally from MA into MAm, a
sociolinguistic situation must be postulated in which MA was spoken by a dominant majority of
monolinguals while MAm was spoken by a minority who had to be bilingual in both languages in
order to communicate with the dominant group. But historical records do not support such a
hypothesis; MAm was a majority language up until the first half of the 20th century. For this
reason, the Amazigh substrate hypothesis will be preferred.
If this hypothesis is correct, the development of a class of agentive nouns in MA would be
the consequence of the mapping of a morphological class that was distinct in Amazigh but not in
Arabic. Unlike Arabic participles that had a number of verbal and nominal functions, Amazigh
agentive nouns were less opaque in that they denoted typically the agent of an activity and, less
typically, experiencer, patient and similar roles. The Arabic nominal category that seemed more
similar to Amazigh agentives was that of occupation nouns8. The typical function of this category
is to denote actions habitually performed by their agents, just like Amazigh agentives.
Unsurprisingly, this category was reanalyzed to serve as an equivalent of a first language
category, namely that of agentive nouns. The fact that MA agentives, like those under (11) and
(12), can denote permanent qualities, habits or roles other than that of an agent, exactly like their
Amazigh equivalents, as the examples under (2) and (3) testify, is a good indication that the two
classes of derived nouns probably did not develop independently of each other.
3.3- The derivation of triliteral passives.
As was shown in Section 2, participles in MA take a CaCC pattern when derived from
triliteral roots (e.g. gləs “to sit”, galəs “sitting”). In comparison, those derived from quadriliteral
roots take a prefixal “m-”, instead. There are, however, a number of exceptional cases that take
8
The “fə al” participles can sometimes have a verbal function and be used, together with “kan” (to be), to express
the past continuous, as in “kan xəddam” (He was working). But this is limited to a very small number of verbs (e.g.
“lə ab” (playing), “xəddam” (working), “wəllaj” (returning)) and it may be a characteristic of some dialects but not
others. Its restriction may even be an indication that it is a recent innovation.
115
this prefix even though they are derived from triliteral roots. These are some examples (cf.
Appendix B below):
13- ḥsəd “envy”
məḥsad
“envious”
gəz “be lazy”
mə gaz
“indolent”
dbəl “wither”
mədbal
“withered”
zin
məzyan
“good”
“good”
One can only wonder why these and similar cases take the prefix “m-” associated with
quadriliteral participles when the template CaCC of triliteral participles is still productive in the
language. But when we recall from Section 2 that MAm also used to have a similar prefix by
means of which agentive nouns were derived, we can easily suspect its role in the exceptional
character of these cases. The formal and the functional similarity between the Arabic and the
Amazigh prefixes could have created some confusion to learners of Arabic as a second language
and encouraged negative transfer that was fossilized in a later stage. As a matter of fact, some of
the exceptional cases are actually derived from Amazigh roots and are still used in the language,
as are the following examples:
14- zləṭ
“to be poor”
məzluṭ
“poor”
ḥḍər “learn”
məḥḍər
“student in a Quranic school”
kʃəf
məkʃuf
“ill-fated”
mhawʃ
“tranced”
“fade out”
hawʃ “play”
These and similar cases testify that interference from Amazigh did happen. Since the Amazigh
agentive prefix occurred with all sorts of roots, irrespective of their length, unlike the
corresponding Arabic morpheme, which occurred only with quadriliteral roots, it is very likely
that it tended to be overgeneralized during the first stages of learning. The items in (13) are clear
cases of overgeneralization, for they cannot have been heard from native speakers if the variety
of Old Arabic from which MA descended is similar to CA in this respect. The exceptional items
in which it has survived were probably too frequent to be regularized later when enough native
input was available to the learners. (On the effect of frequency on irregularity, see Bybee, 1985).
Another reason why the Amazigh agentive prefix could have been borrowed into MA is
of a semantic nature. Semantically, Arabic participles are distinguished by their reference to
temporary states, as was explained above. The items in Appendix B, however, denote permanent
qualities or habits and, therefore, they are not compatible with active participles. In comparison,
116
the Amazigh agentive construction is more suitable in that it can also denote permanent qualities
and habits, though not typically. Accordingly, Amazigh-Arabic bilinguals must have exploited
the formal similarity of the prefix “m-” in the two languages in order to express a grammatical
meaning that is made available only by the Amazigh counterpart. It should be born in mind that
even in CA, Arab grammarians were explicit that participles denoted temporary states (ḥudūth),
as opposed to adjectives (ṣifa muʃabbaha), which are characterized by permanence (thubūt) (cf.
Fassi-Fehri 1993; Hassan, n.d.). Arabic adjectives, however, do not have a single derivational
pattern and today, even advanced students of CA experience a lot of difficulty in learning them 9.
Therefore, there is good reason to suppose that early Amazigh-Arabic bilinguals experienced
similar difficulty and that the Amazigh agentive prefix provided them with an easy way out10.
Of course, there was another option which could have served the purpose, namely the
CCCaC pattern discussed in the previous section. As was pointed out, this pattern served
basically for the derivation of occupation nouns, but other types of agentives are also mapped on
it (cf. Appendix A). Therefore, all the items in Appendix B could in principle take a CCCaC
form, but they do not. The question is why. A possible thesis, which is defended in this paper, is
that these were formed on the basis of the Amazigh agentive pattern and, for one reason or
another, they were never regularized. As a consequence, there are cases in which a word takes the
“m-” prefix while its synonym, antonym or other related words do not (e.g. məʃḥaḥ/zəqṛam
“stingy”). In a coming section, this type of variation will be discussed in more detail and argued
to result from imperfect language learning, which was probably caused by lack of adequate input
and/or weak social control on the learning process.
Another piece of compelling evidence in support of an Amazigh substrate in MA
participles comes from cases derived from nouns, and these are discussed in the following
section.
3.4- Agentives derived from nouns.
In Arabic, participles are derived from verbs. Agentives or other nominals that are derived
from nouns have different morphological patterns, most of which have survived in MA but will
9
This category of adjectives in CA is expressed by a number of patterns most of which are unproductive. Examples
are “ḥasan” (good), “ʒajjid” (excellent), “sa īd” (happy), “qaliq” (worried), “ a raʒ” (crippled), “ aṭʃān” (thirsty),
etc. Each of these examples instantiates a distinct pattern.
10
The point, however, would hold only if the Variety from which MA descends is similar to CA in distinguishing
between adjectives and participles.
117
not concern us here (e.g. nisba). What is of more interest is that the patterns we have discussed so
far are based essentially on verbal roots. But as the examples in Appendix C testify, MA
agentives with the prefix “m-” can also be derived from nominal roots.
As can be noticed in Appendix C, some cases have an ambiguous derivation but others
can only be based on nouns. Examples of the ambiguous cases are:
15- məbrad “sensitive to cold”
məṛzaq “lucky”
mnaṣṣ
“half-full”
məqnaṭ
“boring”
The first of these can be derived either from “brəd” (to become cold) or from “bərd” (cold); the
second from “tnaṣṣ” (to be half-full), in which case it would be a regular quadriliteral passive
participle, or from the noun “nəṣṣ” (half); the third from “ṛzəq” (to give) or the equivalent noun
meaning (livelihood); the fourth from “qnəṭ” (to be bored) or from the noun “qənṭ”, meaning
boredom. In both cases, however, they do not conform to the Arabic regular derivation: if they
are based on verbs, these verbs are triliteral (apart from the dubious case of “mnaṣṣ”) and,
therefore, their participles are expected to have a CaCC pattern; and if they are based on nouns,
they are even more aberrant for the reason already mentioned. In some cases, such as “m əkkəs
/mə kas” (stubborn), two forms are available, each of which gives away its base: the first is
clearly based on the augmented form “t əkkəs” (to become stubborn), which suggests that the
second (and similar cases like those in (15)) is based on the corresponding noun.
In comparison, other cases, examples of which are cited below, are undoubtedly derived
from nouns:
16- məʒdam
məʒwa
“leprous”
məbraṣ
“affected with vitiligo”
“ravenous”
məṣnan
“exuding a foul underarm odor”
The bases of these participles do not have a corresponding verb, and, therefore, there is no way in
which a verbal derivation can be defended. To express a proposition that someone suffers from
the above defects, the phrase “fih/a + N” (s/he has + N) is usually used (e.g. fih l-ʒdam “he is
leprous”). This fact indicates clearly that the pattern mCCaC can be based on both verbal and
nominal roots. In fact, many of the cases in Appendix B can equally be argued to have a nominal
base, and the pattern can be considered simply as agentive, irrespective of the nature of its base.
What is of more interest is the origin of this extension to nominal roots when a similar derivation
in Arabic is restricted to verbal roots.
118
If we assume an Amazigh influence, the extension will come as no surprise. In addition to
the formal similarity between the Arabic participial prefix and its Amazigh agentive counterpart,
which was already pointed out, Amazigh agentives can also have a nominal base. These are some
examples, all of which are based on Arabic loans:
17- amssuq
“market visitor”
suq
“market”
anʒdam
“leprous”
ʒdam
“leprosy”
amʒʒu
“ravenous”
ʒu
“hunger”
anbraṣ
“sb. with vitiligo”
bərṣ
“vitiligo”
Of course there are many examples derived from native Amazigh nouns; the choice of Arabic
loans in (17) has been made on purpose to draw attention to the striking similarity between the
two languages as far as agentive nouns are concerned. Since this derivational process was not
allowed in CA and is not widely attested in Arabic dialects spoken outside North Africa, there is
a good ground to postulate an Amazigh substrate.
We will now turn to another aspect of MA agentive nouns that possibly points to
imperfect learning, which is variation in agentive formation.
3.5- Variation in MA Agentives.
All languages contain some degree of internal variation, and completely regular
paradigms in which each form corresponds to a single meaning can only be found in simplified
languages like pidgins and creoles. But in monolingual speech communities, linguistic
irregularity tends to be minimal, and very often that irregularity can be explained by reference to
diachronic facts. When neither internal nor diachronic explanations can be plausibly advanced,
external factors such as language contact should be returned to for plausible explanations of
irregular paradigms. This seems to be the case with MA agentives.
We have already encountered some aspects of this variation. The coexistence of different
derivational patterns for the same class is one example. We have seen, for instance, that in
addition to the traditional patterns corresponding to triliteral and quadriliteral roots, MA has
extended the use of CCCaC to derive occupation and other types of nouns. The same pattern
serves also to derive similar nouns from quadriliteral roots. It is true that, as was mentioned
earlier, this pattern is semantically distinct in that it denotes habits and permanent qualities while
the traditional participial patterns denote temporary states. But this is not always the case, and
119
active participles that behave like agentives are not hard to find, as the following examples
testify:
18- ḥakəm “governor”
ḍaləm
“oppressor”
ḍamən “guarantor”
qabla
“midwife”
ʒari
najb
“substitute”
“messenger”
These are ambiguous at least in two ways: on one hand, they can be interpreted as perfective or
progressive and, thus, they behave like active participles in general; but on the other hand, they
can also denote occupations or habits, as the English translations indicate. In this latter sense,
they are more like agentives than participles. Why these should have a participial rather than a
regular agentive pattern remains unexplained.
Besides the active participle form, agentives can also take the prefix “m-”. As was
illustrated above, even synonyms or antonyms can have different patterns. For example, while
“xajb” (ugly, bad) is a regular active participle, “məzjan” (beautiful, good) takes the prefix “m”.
Considering that the two correspond to the verbs “xjab” (become ugly) and “zjan” (become
beautiful), it is not clear why they should have different forms. One can even wonder why such
adjectives as “xəwwaf” (fearful) or “zəḥḥaf” (crippled), for example, should not take the agentive
prefix and have the forms “məxwaf” and “məzḥaf”, respectively, since they are likewise based on
triliteral roots.
Actually, there are cases in which two alternative forms are attested. Examples of such
cases are listed in Appendix D. As can be noticed from the Appendix, the agentive forms with
“m-” correspond to a variety of patterns. In what follows, the discussion will be limited to these
examples for the sake of brevity:
19-
kaməl / məkmul “perfect”
zamət / məzmut
“damp, muggy”
ʃəkkak/məʃkak “suspicious”
ləḥḥas / məlḥas
“bootlicker”
ḥrəʃ / məḥraʃ
hbil / məhbul
“crazy, mad”
“rough/ kind of bread”
The first two pairs alternate between an active participle form and an agentive form. The two
forms are usually used alternatively without any significant semantic distinction. In fact, both
forms can sometimes be used idiomatically for exaggeration or some other pragmatic effect, as in
“kaməl u məkmul” (very perfect). In the second pair, the agentives alternate with nouns based on
the CCCaC pattern originally associated with occupation nouns. Again, the two patterns express
the same grammatical meaning in this case, which is that of permanent qualities. As to the last
120
pairs, alternation is between agentives and adjectives that do not have any systematic derivational
pattern. Since these adjectives form an irregular class, it may be argued that the corresponding
agentive variants are innovations that resulted from attempts to regularize the class. Although
regularization is a tendency that often motivates language change, what we notice in MA
agentive nouns seems to go beyond internal factors in that the change does not take a clear
direction. On the one hand, many of the agentives with the “m-” prefix are innovations since they
diverge from the norms of Arabic participles, as explained in the previous two sections. On the
other hand, this class of agentives is itself no more productive in MA and survives in the
language only as an exceptional class. If this class resulted from an attempt at regularization,
what halted the process? Besides, the extension of the pattern of occupation nouns was also
apparently motivated by regularization, if only partly, which means that the same motivation was
drawing in different directions. How did the occupation pattern win over the “m-” prefix?
Since both MA and MAm are not written languages and, therefore, lack written records
from different periods of history, we can only speculate on the emergence of variation in
agentives which has survived in the two languages. The sociolinguistic situation in which the
Amazigh prefix “m-” was borrowed and replaced later by the “fə al” pattern will be discussed in
the following section.
4- Sociolinguistic reconstruction.
It should be recalled from Section 1 that the Speech Community Model, as summarized
by Toulmin (2012), begins by reconstructing linguistic innovations and their dialectological
range, and then evaluates whether these innovations are diagnostic of particular propagation
events (PEvents) and whether there is any linguistic or other evidence that favors a particular
sequencing of these PEvents, including socio-historical records. The innovations that we are
dealing with are the borrowing of the Amazigh agentive prefix “m-” in the derivation of some
MA active participles and the extension of the pattern CCCaC (i.e. fə al) for the derivation of
agentive nouns. In the first case, the propagation range is rather limited since the agentive prefix
“m-” suspected of Amazigh origin is attested only in a limited class of participles while the others
are regular. Besides, most of the irregular participles that take this prefix are found mainly in the
121
sedentary group of dialects11; conservative Bedouin varieties, in comparison, do not exhibit a
similar number of instances12. So, it seems that this is a case of termination, rather than
propagation, of an innovation. The second case, however, has become one of the most productive
processes for the derivation of agentive nouns in MA, as has already been noted. Its range of
propagation seems to include not only traditional sedentary dialects, but extends also to a large
number of Bedouin dialects spoken in the Atlantic plains, Angad plains and the eastern sides of
the Atlas Mountains. It is most likely to be found further east in other parts of the Maghreb. As
was mentioned earlier, the pattern has become productive even in MAm and probably other
Amazigh varieties. Are these two innovations diagnostic of PEvents corresponding to particular
SCEvents? And how can they be sequenced in a plausible socio-historical scenario?
If the prefix “m-” found in triliteral active participles is of Amazigh origin, its borrowing
into MA would be a contact-induced innovation. Contact linguistics, however, has yielded very
few generalizations so far. Thomason (2001:61), for example, asserts that the only prediction that
can be made with absolute confidence is that “contact-induced language change cannot occur
unless there is language contact”, which is trivial. In the domain of derivational morphology, not
much can be deduced from the borrowing of agentive morphemes. Such borrowings have been
attested in cases of long term contact as well as in situations where there was very little contact.
Matras (2009: Chap.8) cites examples in which the agentive affix is borrowed, including Kedang
(borrowing from Indonesian), Tetun Deli (from Portugese), Quechua and Tagalog (from Spanish)
and Hebrew (from various European languages). Matras also cites the case of the Turkish
agentive suffix “–ci/çi” borrowed into Iraqi Arabic; it is surprising that the same suffix is found
also in Algerian Arabic though contact with Turkish in North Africa was not intensive (cf.
Marçais 1977). In the case of MA, what is surprising is not borrowing the Amazigh agentive
suffix so much as the ending of the process. Therefore, some major social event must have stood
behind this termination.
Historians seem to agree that Berbers converted quickly to Islam, but their Arabization
happened much later. According to Camps (1983), while North Africans converted massively to
11
Most of the examples in the appendices are taken from Harrell, Fox and Abu-Talib’s (1962) bilingual dictionary of
Moroccan Arabic and English. In the introduction, the compilers assert that “the entries are based on the speech of
educated Moroccans from the cities of Fez, Rabat and Casablanca. Except for minor variations, their usage may be
taken as typical of urban speech in general.”
12
“məzjan” (good), for example, for which MA is known among speakers of other Arabic dialects, is not found in
Hassaniya, a Bedouin variety spoken in southern Morocco and Mauritania. Instead, “zin” is used.
122
Islam between the 7th and the 9th centuries CE, Amazigh remained the dominant language in the
region until the 11th century when the influx of the Hilali tribes began. Before these waves of
Bedouin migrations, Arabic was confined to urban centers, which were not numerous. The Arab
conquerors settled first in Tunisia, where they built the city of Kairouan. From there, they raided
areas to the west and managed to submit Tangier, Volubilis and a few other cities where they
held garrisons. But they were not able to lay complete control over these western parts, which,
even during Roman times, were more of buffer zones than integral parts of the empire. The
conquest of Morocco was carried out most probably by Berber soldiers recruited in Tunisia by
Arab chiefs and, therefore, must have been bilingual in Amazigh and some form of Arabic. Heath
(2015:8) quotes the historian Sánchez asserting that:
… the first Arabic speakers of the zone must have been the very Berbers who accompanied the Arab chiefs, most of whom subsequently left the zone. The first Arabization
was carried out by these presumably Arabized and therefore bilingual Berbers.
On the basis of this analysis, the local population who wanted to learn Arabic must have had
limited access to native input, which encouraged simplification of the target language as well as
interference from the mother tongue. It was probably under these conditions that the Amazigh
agentive morpheme was introduced, together with other grammatical and phonological traits that
are often argued to be Amazigh substrates in MA.
There were, however, two important events in the history of Morocco that not only
boosted the status of Arabic, but also encouraged Arab migration. The first relates to the rise of
the Idrisid state (788 CE) along with the foundation and the promotion of Fez to the status of a
capital city. Historians report that Idris II, who was the real founder of the city, decided to rely
more on Arabs in matters of administration, probably with the intention to minimize the role of
Berbers (Le Tourneau 1973). The capital became slowly an Arab city, and the arrival there of a
group of migrants from Cordova (817-8 CE) and another group from Kairouan (825 CE)
contributed to this major shift. When we take into consideration that the Idrisids controlled much
of present-day northern Morocco and western Algeria, we can conclude that the Arabization of
the capital city must have constituted a major event in the social and cultural life of Moroccans in
that period13. It was probably for the first time after the Arab conquest, or even long before that,
that they had a cultural center which they could take for a model. Another major event was the
13
A similar event happened in Muslim Spain when the caliph Abd ar-Rahman III adopted a policy that encouraged
the homogenization of Andalusi Arabic (cf. Vicente 2011).
123
arrival of the Hilali tribes, who boosted the process of Arabization. If we had to look for a PEvent
that could explain the termination of the productivity of the Amazigh agentive prefix, and other
substratal features, these two would be appropriate ones; for prestige must have been associated
not only with Arabs, but also with native Arabic speech.
The sociolinguistic interpretation of grammatical borrowing, however, shows some
disagreement. Thomason and Kauffman (1988) distinguish between cases in which imperfect
learning plays a role in contact-induced language change and those in which it does not. In cases
where imperfect learning plays no role, contact usually results in borrowing of non-basic
vocabulary but may affect the grammatical structure of the recipient language when bilingualism
is intensive and attitudes are favorable towards the dominant group. By contrast, when a
subordinate group learns a target language (TL) imperfectly, presumably because of lack of
sufficient contact with native speakers, change may result in the stabilization and
conventionalization of an imperfect version of TL. In language shift, for example, there is usually
very little or no lexical borrowing from the group’s L1 into TL, but many structural aspects of L1
are transferred into TL, including phonology and morphology. In some cases, native speakers of
TL, for one reason or another, also adopt the new version of TL (cf. Thomason 2001 for more
illustration). Instead of lexical and grammatical borrowing, Ross (1997) distinguishes between
phonological restructuring and what he calls “metatypy”. Phonological restructuring occurs when
speakers use a language that is not their mother tongue and so transfer their accent to TL. The
elements transferred are often pronunciation routines that are often beyond conscious control, as
is noticed in second language acquisition in general. As to metatypy, Ross (1997:241) maintains
that it includes “(a) the reorganization of a language’s semantic patterns and ‘ways of saying
things’, and (b) the restructuring of its syntax (i.e. the patterns in which morphemes are
concatenated to form words, phrases, clauses and sentences)”. Metatypy is characteristic of
contact situations in which speakers have to use more than one language continuously so much so
that “there is a strong tendency for them to reduce the cognitive and linguistic processing burden
by bringing their […] languages’ construals of reality into line with each other” (ibid.). Ross
claims also that metatypy is evidence of change in a minority language usually toward the
dominant or the inter-community language in contrast with phonological restructuring, which
happens during language shift as a consequence of interference from L1. He argues further that
“[s]ince metatypy and language shift are alternative outcomes of bilingualism, they cannot affect
124
a single language simultaneously” (Ross 1997:248). In other words, when speakers drop their L1
and become monolingual in TL, their version of TL will exhibit some substratal phonological,
but not grammatical, features. Before we compare the two opposing views in relation with the
Amazigh agentive prefix attested in some MA participles, it is worthwhile to put this item in the
context of the Arabic-Amazigh contact in general.
Although Arabic and Amazigh have been in contact for more than a millennium, they do
not exhibit equal traces of contact. Amazigh, in particular, has been influenced by Arabic at all
levels of its structure, sometimes affecting even basic vocabulary (cf. Kossmann 2013). In
comparison, although researchers seem to assume that Amazigh must have been behind the very
uniqueness of North African Arabic, very few significant substratal features have been identified,
and in many cases, there is controversy as to whether those features are Amazigh substrates or
mere linguistic innovations (cf. Aguadé 2010; Tilmatine 1999, 2011; Bensoukas 2016, this issue).
With regard to vocabulary, research has come up so far only with a derisory list of putative
Amazigh loanwords in North African varieties of Arabic. Such a list, like the one cited in
Tilmatine’s (1999) overview, is likely to include items that may not be familiar to MA speakers
(see also Souag, this issue). The area that apparently shows the heaviest Amazigh influence, at
least on MA, is phonology (cf. Heath 1987, 2002; Durand 1998, Elmedlaoui 1998, Bensoukas
and Boudlal 2012a, b, Dell and Elmedlaoui 2002, Boudlal, this issue, among others). One aspect
of this phonological convergence includes consonants, neutralization of vowel length,
phonotactic constraints, banning of schwas in closed syllables and, indeed, the whole syllable
structure.
If we follow Ross’s (1997) analysis summarized in the preceding paragraph, this would be
a clear case of language shift. The problem, however, is that some cases that would be classified
by Ross as instances of metatypy are also attested in MA. Examples are word order, the feminine
circumfix “t….t” (used mainly with nouns of occupations), the genitive marker “n” (of) (though
limited to some kinship terms like “bb°ay n X” (father of X)), etc. in addition to so many calques
(i.e. instances in which the lexical matter is Arabic but the semantic or the phrasal organization is
Amazigh) (cf. Colin 1963-6; Lafkioui, this issue; Zellou 2011). Many of these features, however,
are controversial; Tilmatine (1999), who cites many of them, is often forced to resort to hedges to
make claims sound more of probabilities than assertions. So, how can we interpret the borrowing
125
of Amazigh morphology, including the agentive prefix, in order to reconstruct the sociolinguistic
contact-induced change within a plausible socio-historical scenario?
It should be noted that, although Arabo-Islamic culture was prestigious, Arabs themselves
were never demographically superior to native Berbers in North Africa. As was noted earlier, the
first conquerors were essentially warriors and there is no evidence that they took their families
with them (cf. Heath 2015 and references cited therein). Even the Bedouin migrants who are
often held responsible for the Arabization of Berbers are estimated by Camps (1983) to have been
only around 80 000 in number at the outset, and their migration extended over two centuries.
Besides, not all of them settled in Morocco. The only plausible conclusion is that, although
Berbers outnumbered Arabs, they shifted gradually toward Arabic. The process started
presumably with Arabs’ neighbors and, when these have been Arabized, they serve as
intermediary to Arabize their next neighbors, etc. Ross (1997), who excludes the possibility to
have metatypy and language shift simultaneously, seems to have in mind the more frequent
situation where speakers of a minority language are forced to shift to the dominant language and
culture. In such situations, it is indeed difficult to imagine why shifting speakers would transfer
grammatical structure or material from their L1 into the inter-community language “as this would
run counter to its use as an inter-community language” (Ross 1997:247).
In the case of North Africa, however, the circumstances in which Arabic became an intercommunity language are a little peculiar. We must suppose that, for a considerable period of
time, Amazigh was used as a means of communication between different tribes since Amazigh
varieties spoken in a given region were to a large extent mutually intelligible. When Arabic
became an inter-community language (and this had never been a general case except in present
times), it was presumably already shaped by early shifters who did not have enough contact with
native speakers to undo the effect of imperfect learning. The emergence of early MA would
probably correspond to what Thomason and Kaufman (1988) call “abrupt creolization”, a
situation where a language develops from “extreme unsuccessful acquisition of TL”. In such a
situation, TL vocabulary is usually inserted in L1 phonological and grammatical structure. If
metatypy and language shift cannot occur simultaneously, the only possibility left to account for
their co-occurrence in MA is to hypothesize, along with Versteegh (1984), the development of
MA from an early Arabic-Amazigh Creole which, later on, was gradually de-creolized. No
uncontentious evidence, however, has survived from this early and obscure stage.
126
This scenario, indeed, seems to be supported by the morphology of MA agentives.
Remember that the major PEvent we have to account for is the termination of the Amazigh prefix
“m-” and its substitution with the “fə al” pattern. There seems to be no reason why the use of
the prefix “m-” should have been limited only to the items in Appendices B (e.g. məḥsad
“envious”) and C (e.g. məʒdam “leprous”), which must have been innovations since they did not
accord with Arabic rules. Therefore, it is plausible to advance that this prefix was used with many
more cases which were subsequently replaced by other forms believed to be more correct. As was
argued earlier in this section, the establishment of Fez as the capital of the Idrisid dynasty, with a
significant population of Arab stock, was presumably the major event that triggered the reversal
of the effects of prior Creolization, or at least the intensity of the process. The variation
exemplified by Appendix D is probably an indication that reversal was still in progress long after
that period and that the irregular items in Appendices B and C have escaped regularization
possibly because of their high frequency.
Of course, once the agentive formation with “m”-prefixation has been cancelled, it must
be replaced by an alternative morphological process. In a situation where there are enough native
speakers of Arabic, we would expect a return to the structure of Arabic participles as found in CA
and other Middle Eastern dialects. Instead, what we notice in MA is the development of a special
class of agentives with the pattern “fə al” (viz. Appendix A), exactly like its equivalent in all
MAm dialects, including those spoken in remote and mountainous areas. What is specific to this
class is that both the pattern and its lexical matter seem Arabic, unlike “m-”, which must have
been salient as a non-native feature. Apparently, the borrowing of the pattern “fə al” into MAm
follows the general path of contact-induced change which, after long-term contact between a
dominated and a dominant language, results in grammatical borrowing (cf. Thomason 2001:
Chap.4). But a deeper analysis of MA “fə al” agentives reveals that the morpho-syntax and the
semantics of this class are “copied” from Amazigh. As was pointed out in the first section,
Amazigh agentives are essentially active and nominal in character, unlike Arabic participles,
which can be active or passive and can express aspect. Therefore, MA “fə al” agentives and
their MAm equivalents, be they in a native or in a borrowed form, are parallel in almost every
respect. If MA were a minority language, this result would be expected since its speakers would
have to be bilingual in MAm too. But it was MAm that was the less prestigious language and its
127
speakers who had to learn the inter-community language. So, how did a morpho-syntactic class in
the dominated language end up in the dominant language?
It seems that the termination of the Amazigh agentive prefix and the extension of the
“fə al” class were successive PEvents. As was explained earlier, the rise of an Arab community
and their growing political influence must have been accompanied by an acute awareness of what
is or is not pure Arabic. This positive attitude toward Arabic led to the purification of extant
varieties from all sorts of salient non-native features, including the agentive prefix “m-”. In
counterpart, the extension of the “fə al” class from occupation nouns to agentives in general was
perhaps an attempt from the Amazigh bilinguals to speak what they thought was correct Arabic
without being conscious that they were calquing their native agentive class onto their version of
Arabic. Since the number of bilingual Berbers most probably far exceeded the number of the
more powerful monolingual Arabs, accommodation to the native speakers was not always
required. Actually, because of their demographic disadvantage, monolingual Arabs may have
frequently been forced to accommodate to non-fluent speakers of Arabic. Therefore, it seems that
both language shift and metatypy can occur simultaneously in complex contact situations.
Ross (1997) does make reference to complex contact histories in which shift and
metatypy are sometimes argued to co-occur, but he doubts that the two processes occurred
simultaneously. Megleno-Rumanian (north-west of Salonica, Greece) is a case in point. Like
other Rumanian varieties of the Balkan, Megleno-Rumanian acquired an essentially Slav
phonological system, thus indicating that it was originally developed by Slav shifting speakers.
But the language has borrowed also a number of morphemes from Bulgarian. Ross claims that
phonological restructuring and grammatical borrowing happened at different periods of contact:
the first during language shift and the second long after that under intensive bilingualism in
which “speakers treated Megleno-Rumanian and Bulgarian as related lects and used the affixes in
the ‘wrong’ lect” (Ross 1997:249). Irrespective of the cogency of this analysis, it is not clear how
it could apply to the MA-MAm contact. According to the analysis presented here, the Amazigh
prefix “m-” had been borrowed into MA during language shift but it was later on eliminated and
only vestiges of it have survived in the language. Even the morphological pattern that replaced it
was apparently developed by speakers of Arabic as a second language. This situation is certainly
distinct from those Ross had in mind, and Thomason and Kaufman’s (1988) analysis of language
shift seems more suitable to MA.
128
In any case, the Arabic-Amazigh contact seems to present a far more complex situation
than one that could fit easily in traditional models. What is most peculiar about this situation is
that it apparently involves the adoption of Arabic by a majority of Amazigh speakers in complex
social conditions over long periods of history. A better understanding of similar situations will
certainly contribute to more accurate models of language contact.
5- Conclusion
Agentive nouns in MA are found to exhibit a variety of morphological patterns that
probably reflect the history of contact between Arabic and Amazigh in Morocco. In both
languages, this class of derived nouns has lost regularity and productivity. This paper has argued
that, although there are very few Amazigh agentive loans in MA, the structure and organization
of agentive patterns bear witness to an Amazigh substrate testified particularly by the use of the
prefix “m-” with triliteral roots and the derivation of active participles from nominal roots. On the
basis of these vestiges, an attempt was made to reconstruct the sociolinguistic situation in which
MA developed, a situation which was claimed to have been characterized by widespread
bilingualism and language shift. More specifically, the agentive “m-” prefix was argued to have
been borrowed from Amazigh at an early stage of contact, when speakers of Arabic were not
numerous enough. During this period, the Amazigh – Arabic bilinguals apparently did not have
access to authentic Arabic input and/or there was no social control on the use of incorrect Arabic.
The m-prefixation process was probably productive since the items in which the prefix is attested
are irregular (i.e. incorrect) and cannot have been heard from native speakers. It was only later,
when an Arab state was formed under the Idrisid dynasty and Fez was founded and made the
capital of the state, that more Arabs migrated to Morocco and gained more prestige, which must
have encouraged Amazigh bilinguals to speak what they believed to be correct Arabic. Later
Hilali migrations reinforced this tendency. The termination of m-prefixation as a process to
derive active participles probably happened in these circumstances, and many of its effects must
have also been undone, saving only very frequent cases.
The similarities between MA and Amazigh agentive nouns, however, raise many issues
regarding the factors that caused them. The lack of detailed historical records as well as written
records of the two languages does not help resolve these issues, and only future dialectological
investigations can hopefully provide some clarification.
129
References
Aguadé, J. (2010). “On vocalism in Moroccan Arabic dialects”, in Juan Pedro Monferrer-Sala &
Nader Al Jallad (eds.). The Arabic Language across the Ages. Wiesbaden: Dr; Ludwig
Reichert Verlag.
Aguadé, J., P. Cressier and Á. Vicente (eds.) (1998). Peuplement et Arabisation au Maghreb
Occidental (Dialectologie et Histoire). Casa de Velazquez : Universidad de Zaragoza
Aspinion, R. (1953). Apprenons le berbère. Rabat: F. Moncho
Benhallam, A. (1989). “Moroccan Arabic Syllable Structure”. Langues et Littératures VIII, pp.
177-91,
Bensoukas, K. (1994). Tashlhit Agentive Nouns: An Optimality-Theoretic Approach. D.E.S
Thesis: Faculty of Letters – Rabat
______(2012). “The Morphology and syntax of negation in Amazigh: Synchronic variation and
diachronic change.” in Boumalk, A. & Laabdellaoui, R. (eds.), Faits de Syntax Amazighe,
publications of the Royal Institute of Amazigh Culture, pp. 37-53.
______(2014). Featural dissimilation in Tashlhit : Avoiding the repetition of labial and round.
Rabat: Mohamed V University.
______(2016). “bu-nouns in Moroccan Amazigh and Moroccan Arabic: A Case of Borrowing” In
Boudlal, A. et al. (eds.), Proceedings of the 3 rd International Conference on Cultures and
Languages in Contact, 481-507. El Jadida: Publications of FLSH
______(this issue). “Concurrent cognate and contact-induced plural traits in Afro-Asiatic
Amazigh id- and Arabic -at plurals”. International Journal of Arabic Linguistics 4/1, pp.
59-102
Bensoukas, K. & Boudlal, A. (2012a). “The prosody of Moroccan Amazigh and Moroccan
Arabic: Similarities in the phonology of schwa”. In T. Borowsky, S. Kawahara, T. Shinya
and M. Sugahara (eds.), Prosody Matters: Essays in Honor of Lisa Selkirk. London:
Equinox Publishing, pp. 3-42.
______(2012b). “An Amazigh substratum in Moroccan Arabic: the phonology of schwa”. In K.
Bensoukas (ed.), Issues in Amazigh Linguistics, Langues et Littératures 22 : pp. 179221. Rabat: Publications de la Faculté des Lettres
130
Boneh, N. (2010). “Perfect constructions in Syrian Arabic”. In P. Cabredo- Hofherr and B. Laca
(Eds.), Layers of Aspect (pp. 23-42). Standord: CSLI Publications.
Boudlal, A. (this issue). “Labial dissimilation as an Amazigh substratum in Moroccan Arabic”.
International Journal if Arabic Linguistics 4/1; pp. 10-38.
Bougchiche, L. (1997) Langues et littératures berbères des origines à nos jours : Bibliographie
internationale et systématique. Paris : Ibiss-Press.
Boukous, A. (1995). Société, Langues et Cultures au Maroc: Enjeux Symboliques. Rabat :
publications de la faculté des lettres.
Brustad, K. (2000). The Syntax of Spoken Arabic: A Comparative Study of Moroccan, Egyptian,
Syrian and Kuwaiti Dialects. Washington DC: Georgetown University Press.
Bybee, J. (1985). Morphology: A Study of the Relation between Meaning and Form.
Amsterdam/Philadelphia: John Benjamins Publishing Company
Camps, G. (1983). “Comment la Berbèrie est devenue le Maghreb arabe, in Revue de l’Occident
musulman et la Méditerranée, n° 35, pp. 7-24
Caubet, D. (1991). “The active participle as a means to renew the aspectual system: a
comparative study in several dialects of Arabic”, A.S. Kaye (ed.), Semitic studies in honor
of Wolf Leslau on the occasion of his eighty-fifth birthday November 14th , 1991.
Wiesbaden: Harrassowitz. 209-224
______(1998). “Histoire et dialectologie”, in Langues et Littératures (Contact et évolution
historique des langues au Maroc) XVI ; pp.61-70.
______(2004). “Dialectologie et Histoire au Maghreb: pour une sociolinguistique historique”, in
Trames de langues, Usages et métissages linguistiques dans l’histoire du Maghreb, ss la
dir. de J. Dakhlia, Paris, Maisonneuve et Larose, pp. 59-70.
Chafik, M. (1991). arba wa arba ouna darsan fī al-lugha al- amāzīhgiyya. Manshourāt
Akādīmiyyat al-mamlaka al-maghribiyya.
Colin, G. S. (1963-6). “Quelques calques syntaxiques et sémantiques sur le berbère dans le parler
arabe du Maghreb” in : GLECS 10 ; pp. 173-6
Cowell, M. (1964). A Reference Grammar of Syrian Arabic. Washington D.C.: Georgetown
University Press.
Croft, W. (2000). Explaining Language Change: An Evolutionary Approach. London: Longman
131
Dell, F. and M. Elmedlaoui (2002). Syllables in Tashlhiyt Berber and in Moroccan Arabic.
Dordrecht: Kluwer
Durand, O. (1998). “Le vocalisme bref et la question de l’accent tonique en arabe marocain et
berbère”, in Langues et Littératures (Contact et évolution historique des langues au
Maroc) XVI ; pp.117-36.
Durie, M. and M. Ross (eds.) (1996). The comparative method reviewed: Regularity and
irregularity in language change. Oxford: Oxford University Press
Ech-Charfi, A. (2008). “The full vowel/schwa alternation: An Optimality-Theoretic approach to
variation in Moroccan Arabic”. Languages and Linguistics 22, pp.93-112.
Elmedlaoui, M. (1998). “Le substrat berbère en arabe marocain: un système de contraintes”, in
Langues et Littératures (Contact et évolution historique des langues au Maroc) XVI ;
pp.137-66.
Fassi Fehri, A. (1993). Issues in the Structure of Arabic Words and Clauses. Dordrecht and
Boston: Kluwer Academic
______(2004). “Tense/Aspect Interaction and Variation”. In J. Guéron & J. Lecarme (Eds.), The
Syntax of Time (pp. 235-257). Cambridge, Mass.: MIT Press.
Hallman, P. (2017). “Participles in Syrian Arabic”, in Ouali, H. (ed.), Perspectives on Arabic
Linguistics XXIX (pp. 153-179). Amsterdam: John Benjamins Publishing Company.
Harrell, R.S. (1962). A Short Reference Grammar of Moroccan Arabic. Washington, D.C.:
Georgetown University Press
Harrell, R.S., T. Fox and M. Abu-Talib (1962). A Dictionary of Moroccan Arabic: Moroccan –
English. Washington: Georgetown University Press.
Hassan, A. (n.d). Al-Nahw Al-Wāfī. Cairo: Dār al-Ma ārif bi Misr
Heath, J. (1987). Ablaut and Ambiguity: Phonology of a Moroccan Arabic Dialect. New York:
State University of New York Press
______(2002). Jewish and Muslim Dialects of Moroccan Arabic. London and New York:
Routledge Curzon.
______(2015). “D-possessives and the origin of Moroccan Arabic”, in Diachronica 32:1, pp.1–
33.
Kossmann, M. (2013). The Arabic Influence on Northern Berber. Leiden: Brill
132
Krim, A. (2013). “Formation du nom d’agent dans le parler berbère de Ksar tit n’Ali : Etude
morpho-phonologique”. Revue des Etudes Berbères 009, pp. 453-64.
Lafkioui, M. (this issue). “Preposed topic specifcation in Berber : An innovation induced by
contact with Arabic”. International Journal of Arabic Linguistics 4/1, pp.
Le Tourneau, R. (1973). Fès avant le protectorat : Etude économique et sociale d’une ville de
l’occident musulman. Rabat : Publications de l'Institut des Hautes Etudes marocaines,
tome XLV
Lévy, S. (1995). “Problèmes de géographie dialectale : Strates st Buttes témoins”, in
Dialectologie et Sciences Humaines. Publications de la Faculté des Sciences Humaines de
Rabat; pp.51-60
______(1998). “Le substrat berbère en arabe marocain : traits et tendances”, in Langues et
Littératures (Contact et évolution historique des langues au Maroc) XVI ; pp.185-98.
Magidow, A. (2013). Towards a Socio-Historical reconstruction of Pre-Islamic Arabic Dialect
Diversity. Unpublished PhD dissertation: University of Texas at Austin.
Marçais, W. (1938). “Comment l'Afrique du Nord a été arabisée", in Articles et Conférences,
Paris, A. Maisonneuve, 1961, pp.171-192.
Marçais, Ph. (1977). Esquisse grammaticale de l’arabe maghrébin. Paris : A. Maisonneuve.
Matras, Y. (2009). Language Contact. Cambridge : Cambridge University Press
Mughazy, M. (2005). “Rethingking Lexical Aspect in Egyptian Arabic”. In M. Alhawary & E.
Benmamoun (Eds.), Perspectives on Arabic Lingusitics XVII-XVIII (pp. 133-172).
Amsterdam: John Benjamins Publishing Company.
Prochazka, S. and I. Batan (2016). “The functions of active participles in Šawi Bedouin dialects”,
in Grigore, G. and G. Bituna (eds.), Arabic Varieties: Far and Wide (Proceedings of the
11 th International Conference of AIDA). Universitati din Bucuresti. pp. 457-466
Romaine, S. (1982). Socio-historical linguistics: Its status and methodology. Cambridge:
Cambridge University Press
Ross, M. (1996). “Contact-induced change and the comparative method: cases from Papua New
Guinea”. In: M. Durie and M.D. Ross (eds.), The Comparative Method Reviewed:
irregularity and regularity in language change. New York: Oxford University Press; pp.
180–217
133
______(1997). “Social Networks and Kinds of Speech-Community Event”. In: Archaeology and
Language I: Theoretical and Methodological Orientations, pp. 209-61. London:
Routledge.
______(2005). “Diagnosing prehistoric language contact”. In: Raymond Hickey (ed.), Motives
for language change, pp. 174-98. Cambridge: Cambridge University Press.
Ryding, K. (2005). A Reference Grammar of Modern Standard Arabic. Cambridge: Cambridge
University Press
Souag, L. (this issue). “Berber etymologies in Maltese”. International Journal of Arabic
Linguistics 4/1, pp. 189-223.
Thomason, S. (2001). Language Contact: An Introduction. Edinburgh: Edinburgh University
Press.
Thomason, S. and T. Kaufman (1988). Language Contact, Creolization and Genetic Linguistics.
Berkley: University of California Press
Tilmatine, M. (1999). “Substrat et convergences: le Berbère et l’arabe Nord-Africain”, in estedios
de dialectologia nortafricana y andalusi 4, pp. 99-119
______(2011). “Berber and Arabic Language Contact”. In Weninger, S. et al. (ed.), The Semitic
Languages- An International Handbook. Boston: Walter de Gruyter, pp. 1001-1014
Toulmin, M. (2009). From linguistic to sociolinguistic reconstruction: the Kamta historical
subgroup of Indo-Aryan. Canberra: Pacific Linguistics, Research School of Pacific and
Asian Studies, the Australian National University’s Center for Research on Language
Change.
______(2012). “Historical Sociolinguistic Reconstruction Beyond Europe: Case Studies from
South Asia and Fiji”. In: Juan Manuel Hernández-Campoy and Juan Camilo CondeSilvestre (eds.), The Handbook of Historical Sociolinguistics. Wiley-Blackwell.
Trudgill, P. (2010). Investigation in Sociohistorical Linguistics: Stories of Colonization and
Contact. Cambridge: Cambridge University Press
Wright, W. (1895). A Grammar of the Arabic Language. (2 Vols.) Cambridge: University Press
Versteegh, K. (1984). Pidginization and Creolization: The case of Arabic. Amsterdam and
Philadelphia: John Benjamins.
134
Vicente, Á. (2011). “Formation and Evolution of Andalusi Arabic and its Imprint on Modern
Northern Morocco”, in Mohamed Embarki and Moha Ennaji (eds.), Modern Trends in
Arabic Dialectology. Trenton: The Red Sea Press; pp. 185-209.
_____ (2012). “Sur la piste de l’arabe marocain dans quelques sources écrites anciennes (du XIIe
au XVIe siècle), in M. Meouak, P. Sánschez and Á. Vicente (eds.), De los manuscitos
midievales a internet: la presencia del árabe vernáculo en las fuentes esctitas. Zaragoza:
University of Zaragoza ; pp.103-120.
Zellou, G. (2011). “Moroccan Arabic borrowed circumfix from Berber: Inverstigating
morphological categories in a language contact situation”. In Gregor Perko (ed.), Les
frontières internes et externes de la morphologie. Ljubljana: Znanstvena založba
Filozofske fakultete, pp. 231-244.
135
Appendix A
bəkkaj
“crybaby”
həḍḍar
‘talkative”
bərgag
“spy”
nə as
“sleeper”
ḥəggar
“humiliating”
wəkkal
“gluttonous”
hərrab
“fugitive”
məkkar
“malicious”
məjjaz
“observant”
qərraj
“studious”
qəmmar
“gambler”
zəḥḥaf
“crippled”
ʃəmʃam
“smelling”
ʃəffar
“thief”
təftaf
“trembling”
xərbaq
“foolish”
xəwwaf
“fearful”
qəfqaf
“timorous”
γəddar
“treacherous”
təmtam
“stutterer”
Appendix B
mədbal (dbl)
“withered”
məhjab (hib)
“scary”
məḥsad (ḥsd)
“envious”
məbrad (brd)
“sensitive to cold”
məlḥas (lḥs)
“bootlicker”
mər ud (r d)
“fearful”
məqnaṭ (qnṭ)
“boring”
mərzaq (rzq)
“lucky”
məʃḥaḥ (ʃḥḥ)
“stingy”
məzjan (zin)
“good”
məs ur (s r)
“rabid”
mət us (t s)
“unlucky”
məzrub (zrb)
“hasty”
məẓluṭ (ẓlṭ)
“poor”
məʒlaq (ʒlq)
“greedy”
məlhuṭ (lhṭ)
“greedy”
mə gaz ( gz)
“indolent)
məγjar (γir)
“jealous”
mə jaf ( if)
“easily disgusted” məʃkak (ʃkk)
136
“suspicious”
Appendix C
məhrar (hərr)
“sensitive to touch”
məṣnan (ṣnan)
“exuding
məʒdam
məzhar (zhər)
underarm məṣwab
“lucky”
“polite”
odor”
(ṣwab)
“leprous”
məʒhəd (ʒəhd)
“strong”
“mangy”
məbraṣ (bərṣ)
“affected with
(ʒdam)
məʒrab (ʒərb)
vitiligo”
məʒwa (ʒu )
“ravenous”
məsrar (sərr)
“charming”
mə kas ( ks)
“stubborn”
mə nad ( nad)
“stubborn”
mə lal ( əlla)
“sickly”
məkraʃ (kərʃ)
“ravenous”
Appendix D
hbil / məhbul
“crazy”
zəγbi / məzγub
“ill-fated”
ḥrəʃ/ məḥraʃ
“rough”
zin / məzjan
“good; beautiful”
kaml / məkmul
“perfect”
ʒi an / məʒwa
“ravenous”
ləḥḥas / məlḥas
“bootlicker”
γəṭṭaja / məγṭaja
“lid”
wasə / mtisə
“wide”; “large”
ḍijjəq / mḍijjəq
“narrow”
zamt / məzmut
“muggy”
zərban / məzrub
“hasty”
zwiwn / məziwn
“cute”; “good”
qənṭan / məqnaṭ
“(tends to be) bored”
ḍ if / məḍ uf
“weak”; “slim”
ḥajl / mḥajl
“aged (of food)
137