PhiN 7/1999: 1

Cemi Belkacemi (Manchester)

A View on Relativisation in Arabic

It has been suggested by Abubakr (1970) and Haddad/Kenstowicz (1980) that Arabic relativisation is purely a matter of definiteness. Most works adopting this approach do not exceed the traditional stage of analysis which limits itself to description. The purpose of this article is to provide a contrastive study of the determiner versus the relative marker, and an account of the movements involved in the relativisation process. It will be argued that a phonological treatment based on assimilation inevitably encounters serious difficulties at the syntactical level. The discussion in this paper draws on work by Haddad/Kenstowicz (1980), Killean (1972), Wise (1975), Abubakr (1970) and on Majdi (1990). We will look firstly at some background information on the question of relativisation in Arabic. Secondly, we will turn to the properties of the definite marker and examine them in contrast to those of the relative marker. Finally, we will provide an analysis based on Chomsky (1986) and conclude by rejecting the claim made about relativisation being a matter of definiteness in Arabic.

1 Background

In some studies of Arabic dialects, there is a tendency to treat the relative marker (Rm) simply as another form of determiner (Abubakr 1970: 162). This treatment derives from the phonological phenomena of assimilation and can be illustrated by studies carried out by Haddad and Kenstowicz (1980) on Lebanese Arabic, examples of which are reproduced below:

In the above examples, it can be seen that what is produced as a Rm in the first instance , reduces to the phoneme /l/ (1.a). If followed by a "solar" letter (1.b), it assimilates the phonetic qualities of that consonant.

PhiN 7/1999: 2

Abubakr’s Sudanese study categorically refutes the existence of an element called Rm and states that it is the determiner which fulfils the task of linking HNs to their respective embedded sentence. He sees that relativizing is essentially a matter of agreement in definition (1970:163) as his examples demonstrate:

Example (2b) exhibits a relationship whereby we supposedly have a determiner that attaches to a verb. However, because is in the active form, this in our view breaches the syntactical rules of Arabic. This is to be differentiated from , which is a participle and allows such a combination. Killean (1972:146) holds similar views to ours by specifying that instances of relative marker reduction do occur in Arabic but only when a noun or noun-like form immediately follows the relative pronoun. This is explained in the following statement: "whenever a noun or noun-like form immediately follows the relative pronoun, it [the relative pronoun] will become an enclitic which is prefixed and pronounced just as the definite article is" (p.146). A similar stance is adopted by Wise who sees for instance that Egyptian Arabic reduces the Rm only when followed by an adjective or verbal adjective [i.e. a participle] (Wise 1975:89-90).

To give our analysis a broader scope, we can further examine the respective characteristics of determiner and relative marker in yet another Arabic-language representative of North Africa (Algerian Arabic), and deal with the issue of similarities and/or differences exhibited by both Dm and Rm. First, we shall look at the determiner.

2 Determiner vs Relative marker

2.1 The determiner

Determiners form "a set of closed-system items that are mutually exclusive with each other". They occur before the noun head and in "choice relation" only, i.e. they occur one instead of the other (Quirk, 1972: 137). In the light of the above definition, we shall begin by studying the Arabic determiner in its simplest form and then, for purposes of comparison, in an environment similar or close to that of the Rm. Pre-determiners aside, there are two main determiners in Arabic: 0 and

PhiN 7/1999: 3

occurs in noun-phrases (NP) of the type:

whereby it introduces specific entities. Because only definite entities are accompanied by a determiner in Arabic, it is more appropriate to call it a definite marker (Dm). Indefinite entities are accompanied by zero (0) marking.

As the Arabic indefinite marker is morphologically speaking built within the word, our schematic representation needs to take this feature into consideration:

The above rule suggests that the determiner is inherent to and part of the indefinite noun in Arabic as proposed in (4.a-b), hence the formulation of DP as a "base-constituent". This "base-constituent" associates with for definite entities as demonstrated earlier by (3.a,b) and also with adjectives to form an even larger constituent (the adjective being the simplest form of noun modification). The examples below are a sample of this type of modification.

Since the constituents of any one string share a common determiner, our rule will be rewritten as follows:

PhiN 7/1999: 4

The abbreviatory device of parentheses allows us to underline the optional status of the adjective. This rule conflates those before and links the marker Dm () or (0) to the first constituent (or head) which governs in "a chain" the second constituent This form of government is what obtains a DP instead of a sentence.

Using strings (4a) and (4b) and restricting the determiner's government to the first constituent will lead to a change from an adjectival DP to a sentence, as illustrated by (8.a) and (8.b) below:

The absence of a verb, or in this case the covert marking of the verb "to be", shows that:

  1. The application of D to both constituents yields a DP (6).
  2. Its restriction to the first constituent results in a sentence (8).
  3. Its restriction to the second constituent results in a genitive case, as can be observed in (9.a) and (9.b).

However, in structures of the pattern:

where we are dealing with an overtly expressed verb 'to be' in the past tense, any attempt to attach Dm "" to it would result in ungrammaticality.

PhiN 7/1999: 5

This suggests that within the subcategorization framework of the Arabic language, Dm does not attach to the verb. Its structural representation below projects its lexical properties which exclude verbal forms:

Dm [ DP/Adj ]

We shall come back to this point at a later stage when we deal with the relative clause. For now, we shall turn our attention to the relative marker.

2.2 The Relative Marker

Traditional grammar regards the relative pronouns as entities which "introduce relative clauses postmodifying nominal heads" (R. Quirk 1972: 214). In English, these include two series: wh-pronouns and that or zero, and neither series has number or person contrast. Wh-series, however, has animateness (personal/non-personal) contrast and case (nominative/accusative and genitive) contrast.

In Arabic (dialects included), the relative marker derives from what in traditional literature, "qualifies to be the relative adjective "who, what" which [in Classical Arabic] is variable in gender and in number and, in the dual only, also in case" (Cantarino 1975:162). Notice the inflections the Rm undergoes to observe the above rule:

Now, for comparative purposes with the determiner, let us consider the following example:

PhiN 7/1999: 6

The sentence contains a complex NP with a RC. The head noun (HN) is modified by a clause (CP) and its internal structure consists of an IP preceded by what we consider to be the Rm . The above structure can be illustrated by (16) below:

One can observe by comparing (4.a) and (11) that in the former the DP is followed by an adjective preceded by , whereas in the latter the HN is modified by an adjectival clause introduced by . It is this difference (or closeness) in structure that shall constitute a point of focus for us. We say closeness because if compared to the French structure

17. La maison qui est grande...

the latter undergoes, in its transition from RC to a simple DP, a sequence of transformations, the first of which deletes the auxiliary "be" thus yielding:

18. *La maison qui grande ..............

The second shifts the adjective from a postnominal to a prenominal position thus resulting in:

19. La grande maison [qui.............]

which represents our target DP: la grande maison as suggested in Wardhaugh (1977: 120-121, 141-142). This is obviously an old treatment of the RC but its purpose here is merely to point out the fact that because Arabic tolerates variation in its word order, the RC reduction requires fewer transformations than for instance its French counterpart above. This, it can be argued, is motivated further by the fact that:

  1. the aux. 'be' is covertly expressed in the present tense
  2. the adjective is of post-nominal character.

And so, to arrive to the widely accepted assumption that a construction such as:

is a reduced form of (15) which is a relative construction, it suffices to posit a reduction transformation which simply reduces to . Although this symmetry holds true in the case of adjectives (cf. Culicover,1976: 179; Benvéniste, 1966: 213), if we extend it to embrace verbs as suggested by Abubakr (1970:163) in his example (2.b) it would somewhat be questionable at the syntactical level. We shall deal with this point in the analysis below where further evidence is presented.

PhiN 7/1999: 7

3 Analysis

We shall start our analysis by considering the following examples:

The HNs in (21), (22), and (23) share one common feature: they are all modified by an adjectival RC. It is common knowledge that adjectives have a dual class membership (class-cleavage in Bloomfield’s terms), hence their co-occurrence with determiners (below) is a consequence of that fundamental property.

On those grounds, the adjectival RC reduction is a process that does not violate any structural or morphological rule. From the perspective of Arabic structure, the data presented by Abubakr (1973) and Haddad-Kenstowicz (1980) raises questions regarding discrepancies between phonological and syntactical functions. While we accept RC reduction in certain contexts, we dispute it at verb level and maintain the existence of RC in Arabic. Further evidence for refuting the claim that relativisation is primarily a matter of definiteness may be drawn from other Rms which, unlike cannot be subjected to the reduction transformation. Relative markers like , (Cantarino 1975: 148-192) carry specific syntactic properties related to wh-words. One of these properties is the wh-question as illustrated below:

PhiN 7/1999: 8

While it can be noted that pronouns have been shown to share one property with wh-words, the same cannot be said of the Dm preceding verbs. Notice:

It is generally acknowledged in the literature (Radford 1988: 499-508) that wh-words can fulfil the function of specifiers at CP level. This can be seen in:

Both examples show a PP at a lower level. Given that wh-movement can apply to both wh-NP and wh-PP constituents (cf. Radford,1988), the application of one or the other may be used without constraints in English as illustrated below:

However, similar movement would encounter difficulties in Arabic. For example, applying wh-movement to 31.a (below) which contains a PP:

shows that the fronting of the NP yields 31b (previously 28)

PhiN 7/1999: 9

The corresponding derivation (31b) now contains a wh-pronoun as the leftmost constituent of CP, while at IP level a trace in the form of a resumptive pronoun referring to has been postposed to the preposition saving it from being stranded or orphaned. The lack of a trace in the extraction site results in ungrammaticality as shown in 31.c below:

It can therefore be concluded that the stranding of prepositions is not permitted in Arabic, thus highlighting the fact that the latter is a subcategorization principle preserver. Equally unacceptable would be any attempt to shift the preposition to the left of at CP level.

With , on the other hand, no trace can be left in the extraction site.

This consequently leads to the preposition being pied-piped:

thus ending up as the leftmost constituent of CP to the left of C which now contains instead of .

Although (31.b) and (31.f) are synonymous, the fact that one exhibits the element and the other the element leads to the adoption of the view that has properties which distinguish it from the rest of the wh-phrase. One of the consequences of this difference is that this element cannot be said to originate in the object position of PP and subsequently to move to Spec position. Rather, it originates directly under Spec (Ouhalla, 1994 chp.3).

It can therefore be suggested that the two derivations (31.b) and (31.f) vary along the following parameters:

1. The first is accounted for by positing that a rule of wh-movement applies in the course of the derivation of such a sentence to generate wh-NP at CP level and the trace at IP-level in order not to orphan the preposition whose combination with is not accepted at CP level.

PhiN 7/1999: 10

2. The second derivation with is a structure whereby the wh-NP is used as the object of the preposition and both are moved along to the front of the clause at CP level.

The above analysis has shown that is not a determiner but a relative marker which must derive from what Cantarino defines as the relative adjective. The loss of (gender, number and case) inflection in dialects gives it a "readiness" for reduction, but its generation as already noted above clearly illustrates that and are in complementary distribution, hence equal in their function as modifiers.


Abubakr, E.R. (1970): The noun phrase in the spoken Arabic of Sudan, unpublished doctoral thesis, University of London.

Bahloul, M. (1993): "The Copula in Modern Standard Arabic (SA)", in: Eid, Mushira (ed.), Perspective on Arabic Linguistics V. Papers from the First Annual Symposium on Arabic Linguistics, Salt Lake City, Utah 1987. John Benjamins Publishing Company, Amsterdam/Philadelphia (Current Issues in Linguistic Theory, 101).

Benveniste, E. (1966): Problèmes de linguistique générale. Volume 1, Paris, Gallimard.

Borsley, R. D. (1991): Syntactic Theory: A Unified Approach. Edward Arnold.

Cantarino, V. (1975): Syntax of Modern Arabic Prose II. Bloomington, Indiana U.P.

Chomsky, N. (1986): Barriers. Cambridge, Mass: MIT Press.

Comrie, B. (1981): Language Universals and Linguistic Typology. Oxford, Basil Blackwell.

Haddad, G. / Kenstowicz, M. (1980): "A note on the parallels between the definite article and the relative clause marker in Arabic". Studies in the linguistic sciences 10:2.

Haegeman, L. (1994): Introduction to Government and Binding. Second Edition, Blackwell.

Killean, C. (1972): "Arabic Relative Clauses", in: Peranteau, P. M.; Pevi, J. N. and Phares, G. C. (eds.), The Chicago which hunt. Chicago, C.L.S.

Majdi, B. (1990): "Word Order and Proper Government in Classical Arabic" , in: Eid, Mushira (ed.), Perspective on Arabic Linguistics I. Papers from the First Annual Symposium on Arabic Linguistics, Salt Lake City, Utah 1987. John Benjamins Publishing Company, Amsterdam/ Philadelphia (Current Issues in Linguistic Theory, 63).

Maxwell, D.N. (1979): "Strategies on relativization and NP accessibilities". Language 55:2.

Ouhalla, J. (1994): Transformational Grammar. Edward Arnold.

Radford, A. (1988): Transformational Grammar. C.U.P.

Radford, A. (1997): Syntactic theory and the structure of English. C.U.P.

Wardhaugh, R. (1977): Introduction to Linguistics. New York: McGraw-Hill.

Wise, H. (1975): A transformational grammar of spoken Egyptian Arabic. Oxford, Basil Blackwell.