The Arabic Language

13.2 Code-mixing and the middle variety
Various efforts have been made to subdivide the continuum between the two 
extremes of Standard Arabic and dialect into intermediary varieties. Ferguson 
already referred to the emergence of a mixed middle form in the diglossia situa
-
tions he described. This middle level is often mentioned in publications about the 
sociolinguistics of Arabic. Arab linguists refer to it as ‘the middle language’ (
al-luġa 
al-wusṭā
), or as ‘the colloquial of the intellectuals’ (
ʿāmmiyyat al-muṯaqqafīn
); other 
terms that refer to this phenomenon are ‘Educated Standard Arabic’, ‘arabe 
médian’, etc. This middle language is supposed to be a form of Standard Arabic 

---

244
The Arabic Language
that does not use case endings, follows the colloquial pronunciation and freely 
introduces colloquial words, while retaining the general structure of the standard 
language.
In his study of the sociolinguistic situation in Egypt, Badawī (1973) rejects the 
strict dichotomy of Ferguson’s model with its two varieties, H and L, which in his 
view does not correspond to the Egyptian situation (and presumably not to other 
regions of the Arabic world, either). Instead, Badawī sets up five different levels, 
each with its own characteristic set of features setting it apart from the other 
levels of the continuum (see Table 13.1). 
I
fuṣḥā at-turā
t
‘Classical Arabic’
only used in Qur
ʾ
ā
nic recitation
II 
fu
ṣḥā al-ʿ
a
ṣ
r
‘Modern Standard Arabic’
the standard form of the language used in writing and sometimes on formal 
occasions in speaking
III 
ʿāmmiyyat al-mu
t
aqqafī
n
‘colloquial of the intellectuals’
the formal spoken language of educated people
IV 
ʿāmmiyyat al-mutanawwirī
n
‘colloquial of the literate’
the informal spoken language of educated people
V 
ʿāmmiyyat al-
ʾ
ummiyy
ī
n
‘colloquial of the illiterate’
the language in which the illiterate talk
Table 13.1 The five levels of Egyptian Arabic according to Badawī (1973) 
It is obvious that Ferguson’s model cannot be applied without modification to 
the actual linguistic situation in the Arab world. But the association with socio
-
economic groups that Badawī proposes is doubtful. There is not much empirical 
research on the social distribution of speech levels in Egypt, or for that matter in 
any Arab country. Elgibali (1985) shows that in accordance with Badawī’s predic-
tions there is a continuous flow in the distribution of sociolinguistic markers, 
such as the realisation of /q/ and /ṯ/, the use of aspectual markers in the verb, 
and word order and declensional endings. Yet only the upper and lower level 
(Ferguson’s H and L, Badawī’s level V and I) could be called discrete levels with 
a characteristic set of features. The middle part of the continuum cannot be 
divided into separate levels. In Elgibali’s test results, the informal register of each 
level has the same distribution as the formal level of the second lowest level. For 
instance, the distribution of the /q/~/ʾ/ variable among the five levels of spoken 
Egyptian turned out to be as in Table 13.2. The same distribution applied to the 
other sociolinguistic markers in his study.

---

Diglossia 
245
oral level 
source material 
/q/
percentage
informal 
msa
e.g., newscasting, university
45
lectures (arts) 
formal educated colloquial 
e.g., university lectures (science) 
44
informal educated colloquial 
e.g., intra-group conversations 
35
among professors
formal literate colloquial 
e.g., popular television programmes 
34
informal literate colloquial 
e.g., intra-group conversations
22
among shopkeepers 
formal illiterate colloquial 
e.g., conversations between 
23
employers and employees 
informal illiterate colloquial 
e.g., intra-group conversations
0
among workers 
Table 13.2 Distribution of /q/ in spoken Egyptian (after Elgibali 1985) 
This means that Arabic speech production is always characterised by a high 
degree of code-mixing. Various efforts have been made to discover constraints 
governing this speech behaviour. Two approaches may be distinguished: a global 
and a local analysis of discourse (Eid 2007: 433–4). Global analysis of hybrid texts, 
that is, texts combining elements from different levels of speech, focuses on the 
correlation between code-mixing and such factors as identity and group member-
ship (see below, section 13.3); this allows us to draw conclusions about the manip
-
ulation of language for various purposes (see section 13.4).
Local analysis of hybrid texts identifies specific reasons for switches within 
a stretch of discourse. In her account of the search for constraints in diglossic 
speech, Mejdell (2012b) points at the basic asymmetry between a dominant 
language, usually the first language of the speakers, and a standard language. In 
this connection, she formulates the Dominant Language Hypothesis, which states 
that lexical items of the dominant standard variety may be combined freely with 
grammatical markers from the non-dominant dialect variety, but not the other 
way round: standard markers are almost never combined with dialect items. The 
hypothesis was tested by Mejdell (2006) in an investigation of the mixed style in 
an academic context. The variation patterns in her data confirm the hypothesis. 
Forms like 
bi-yuktab 
‘it is written’, in which the colloquial aspectual particle 
bi-
is 
used with a standard passive, frequently occur, while a hypothetical *
yuʾāl 
‘it is 
said’ does not.
According to Mejdell, this principle is also behind the directionality constraints 
found by Eid (1988) in her study of code-mixing. A standard relative pronoun, for 
instance, may not be followed by a dialect word, while a dialect relative pronoun 

---

246
The Arabic Language
may be followed unproblematically by a standard form. In other words, combina
-
tions like *
allaḏī ṛāḥ 
‘the one who went away’, are disallowed, whereas combina
-
tions like 
illi ḏahaba 
are perfectly acceptable. Likewise, a standard negation may 
not be combined with a dialect verb (*
lan ḥa-tuʾaf 
‘she will not stop’), whereas a 
dialect negative marker may be combined with a standard verb, as in the form 
ma-sa-taqif-š
‘she will not stop’ with dialect negative circumfix 
ma- … -š
and 
standard future marker 
ḥa- 
(Boussofara-Omar 2003: 40). 
The Dominant Language Hypothesis also seems to operate in the Matrix 
Language Frame Model proposed by Myers-Scotton (1993). This model distin-
guishes between content (lexical) morphemes and system (grammatical) 
morphemes and introduces a strict rule that only one language involved in the 
code-switching provides the system morphemes; this is the matrix language. The 
other language involved, the embedded language, provides (some of) the content 
morphemes. The examples quoted above contain system morphemes from both 
languages, as in
 ma-sa-taqif-š
, which appears to invalidate the rule. Subsequent 
versions of the Matrix Language Model have tried to explain these problematic 
examples (Boussofara-Omar 2003). According to Myers-Scotton (2010), these 
newer versions still provide the right predictions about the way code-mixing 
takes place in Arabic diglossia, but she seems to acknowledge that in a situa
-
tion of diglossia, code-mixing has special properties. According to some scholars, 
diglossic mixing or switching should be differentiated therefore from ‘normal’ 
code-mixing (Boussofara-Omar 2006b).