41

Chapter 2 Wh-movement in Modern Standard Arabic

2.0. Introduction The aim of this chapter is to investigate the general syntactic properties of whmovement in Modem Standard Arabic within the framework of the minimalist theory. In sections 2.1 and 2.2 of this chapter, we will look at wh-movement in simple and multiple wh-questions and also wh-movement in free relatives. Section 2.3 will be concerned with word order constraints in wh-questions. Subsequently, section 2.4 will analyze wh-movement within the copy theory of movement taking into consideration data not only from MSA but from other varieties of Arabic also; and it will attempt to account for variations in these languages. The last section of this chapter will summarize and conclude what has been done throughout the chapter. In MSA, as in most of the languages of Indo-European origin, wh-questions are formed by fronting a wh-pronoun or a wh-phrase headed by an interrogative determiner. Following common usage, we will refer to these as "wh-pronouns" and "wh-phrases", and to the transformational process of constituent question formation as "whmovement". In terms of more recent syntactic theories, a wh-element is assumed to land in a designated position which is provided for the functional vocabulary of the language. The structure of a clause can at least be analysed in terms of three layers: the lexical or thematic layer of the V-system, represented by VP; on top of this comes the

42

functional layer of the I(nflectional) system, represented in IPffP; and finally, above the inflectional

phr~

comes the C(omplementizer) system or CPo CP specifies the

force of the clause; in other words it specifies what type of clause it is. The MP takes a derivational approach to the generation of sentences. And it moves from bottom-up. Wh-phrases are moved from the domain of the V-system, in which they originate and function as arguments or adjuncts of any sort, to the specifier of CP [Spec, CP]. This movement is shown in the following example:

1. [cp Wh-phrasei

t

k c IIp [vp

ti ... ]]]]

I

The Wh-phrase leaves a trace t. The chain which is formed by wh-movement has roughly the following characteristics: if the wh-phrase corresponds to an argument, the trace is in an A(rgument) position, while the wh-phrase itself is in a non-A, or A'position. The wh-phrase is semantically an operator which binds the trace, which in turn is interpreted like a logical variable. In natural language, the operator is usually a phrase which consists of the operator proper, the pure wh-part, and a restrictive part. Thus, who is composed of the features [wh] and [person], what is composed of [wh] and [thing], which student is composed of[wh] and [student], etc. Once the operator is in Spec, CP,

it has scope over the proposition expressed by IP. The proposition contains a trace, and is thus an "open" proposition. As such it does not express a truth value but rather a property. According to standard assumptions, the wh-phrase also involves an existential operator. This operator is actually responsible for variable binding. By involving

43

existential quantification, the question makes a commitment to the existence of an entity which corresponds to the phrase which is missing from the clause. I As Chomsky (1977) has shown, wh-movement is only one manifestation of a larger class of movements which comprises of topicalization, relativization, clefting, comparative constructions, etc. In this study, we will confine ourselves to movements

related to question formation by wh-movement. It is important to bear in mind that not all wh-phrases undergo movement in a single sentence, while at the same time their interpretation seems to remain the same as if the phrase had been moved. In the whmoving languages this is shown in multiple wh-questions. Only one wh-phrase moves to Spec, CP, while all the others remain "in-situ." With reference to wh-questions, world languages can be broadly grouped into two types: wh-fronting languages and wh-in-situ languages. That is, in some languages of the world a wh-phrase moves overtly before PF to Spec, CP, while in some other languages it does not; but it does eventually at LF. Greenberg (\966) relates whmovement to word order. According to him, VSO languages, such as Arabic, always have wh-movement while SOY languages never do; many SVO languages, such as English, have wh-movement too but some languages, such as Mandarin, do not. Languages without overt wh-movement are referred to as wh-in-situ languages. In general, Arabic interrogative words are placed at the beginning of the sentences. Questions are straightforward in Arabic. There is no inversion of word order as in the case of English auxiliaries. There are no auxiliaries in MSA but verbs appear immediately after the wh-words in main clauses; while in the embedded clauses it is not 1 See

. 'iJ.

Bayer (2004).

44

so. The most common interrogative question words in Modem Standard Arabic are listed below as foIlows.:2

oAyy (which) e.g.: ayya kitaab-in

tuhibb-u

which-book-gen. like- you

ann taqra'a-hu? to read-it

'Which book do you like to read?'

oKam (how many, how much) e.g.: i) kam

tuffahat-an akalt-a?

how-many apples-gen. ate-you-msg 'How many apples did you eat?'

turid-u

ii) kam

maal-an?

how-much want-you money-acc. 'How much money do you want?'

oKayfa (how) e.g.: kayfa tadhab-u ila al-madrasat-i? how

go-you

to the-school-gen.

'How do you go to school?' 2

For more details see Ryding (2005), p. 401.

45

oMadha (what) e.g.: madha akaIt-a bil-amsi? what

ate-you yesterday

'What did you eat yesterday?,

oLi-madha (why, what for) e.g.: li-madha tuhibb-u al-sibahat-a? why

like-you the-swimming-aee.

'Why do you like swimming?'

oMa (what) e.g.: rna

ismuk-a?

what name-you-gen. 'What is your name?'

oMan (who, whom) e.g.: i) man fa'a1-a hadha? who did

this-ace.

'Who did this?' ii) man tuhibb-u akthar? who like-you more

'.

'.

46

'Whom do you like more?'

oAyna (where) e.g.: ayna taqa'a-u baladuk-a? where is

country-you-gen.

'Where is your country?'

o Mataa (when) e.g.: mata ja'a-a abuuk-a? when came

father-you-gen.

'When did your father come?'

oHal (introduces yes/no questions) e.g.: hal anta mariidh-un? are you sick-nom.

'Are you sick?'

oA- (introduces yes/no questions) e.g.: a-hadha aliyy-un? is-this

Ali-nom.

'Is this AliT

47

2.1. Wh-movement in Different Types ofWh-questions As a wh-movill.g language and as we have already seen, MSA has wh-movement which involves the movement of wh-elements from their original position into Spec, CP. For example, a declarative sentence in Arabic featuring normal surface word order would be VSO as shown in (2). 2. ishtara aliyy-un sayyarat-an. bought Ali-nom a-ear-ace •Ali bought a car. ' The direct object, sayyarat-an "a car" , of the verb, ishtara "bought", normally follows the subject; however, when the direct object is a wh-phrase, the wh-phrase generally appears at the beginning of the sentence as in (3): 3. madha ishtara aliyy-un?

what bought

Ali-nom

•What did Ali buy?' There are hardly any auxiliaries in Modem Standard Arabic and hence there is no aux-inversion. In Italian also there are no auxiliary verbs. But in certain dialects of Arabic there is inversion. In MSA declarative non-interrogative clauses, unlike in English, an auxiliary

need not follow a wh-word. However, in English, when an

auxiliary verb does not occur as it happens in simple tenses, a form of "do" must be used in questions and negatives. The auxiliary verb (or a form of "do") occurs immediately after the subject and the wh-phrase in declaratives and wh-questions respectively as shown in the following example:

48

4. a. He bought bread. b. What did he buy?· In Modem Standard Arabic, the occurrence of the auxiliary verb yakunu, 'be' is optional and it occurs in clauses with nominal predicates. These clauses generally have present tense reading even without the linking verb yakunu. Observe the following examples: 5. man qabala aliyy-un? who met

Ali-nom

'Whom did Ali meet?' 6. man (yakunu) ash-akhs-u who (is)

the-person

alladhi qabala-hu aliyy-un? who

met-him

Ali-nom

'Who is the person whom Ali met?' The above examples (5-6) have the same meaning but different structures represented in (7) and (8) respectively:

7. [cpmanj [qabala aliyy-un tj]] 8. [cp manj [yakunu [[ash-akhs-u [cp alladhi qabala-hu aliyy-un]]] [ tj ]]]] Example (8) has a verb yakunu, 'be' which functions like an auxiliary. This linking verb occurs, as we notice, only optionally in cleft sentences and wh-questions with nominal predicates. There are no other auxiliary verbs in Arabic corresponding to the English

have, is or do.

49

2.1.1. Wh-movement in Simple Wh-questions In simple wh-questions, only one wh-phrase is base-generated, whereas in multiple wh-questions, there are multiple wh-phrases, although only one of them moves overtly to Spec, CP position. In this section, we will consider simple wh-questions in which only a single wh-word or wh-phrase is base-generated. Wh-questions, as we have seen so far, use a strategy involving a wh-movement rule which moves the wh-element from its original position into the clause initial position, i.e. into Spec, Cpo Examine the following examples: 9.a. man; madaha t; a1iyy-an?

t

I

who praised

Ali -acc

'Who praised Ali?' b.

mr;

madaha aliyy-un

r

who praised Ali-nom 'Who did Ali praise?'

In the above example (9a), we find that there wh-extraction from the subject position and in (9b) there is wh-extraction from the object position. If we look at the following examples (10-11) below, we see that the wh-element in MSA can be an adjunct wh-word or -human wh-word which moves from the original position into Spec, Cp to form a simple wh-question in Arabic. 10. mfa; dhahabat fatimat-u

when went

ila a1-qariyat-i

r

Fatima-nom to the village-gen

'When did Fatima go to the village?'

50

II. madfa j akala al-walad-u what

ate

r

the boy-nom

'What did the boy eat?' Although madha is a single word it does stand for a maximal projection since it has the status of a pronoun. In the case of the simple wh-question formation, the wh-element can be not only a single word but it can also be a whole phrasal category (NP or PP). In such cases, the wh-movement has to move the whole major category from the original position into Spec, CP in order to avoid ungrammatical outputs. This means that it is not just the whword (containing wh-feature) that moves, the entire phrase has to pied-pipe along with the wh-word. 3 Consider the following examples (l2a-c): 12. a. ayya kfab-inj

qara'a ahmad-un

which-book-gen. read

F

Ahmed-nom.

'Which book did Ahmed read?' b. kitab-a manj

t

qara'a ahmad-un tj?

book-acc.-who read

I

Ahmed-nom.

'Whose book did Ahmed read?' c. i1a aynaj dhahaba ahmad-un tj? + I to-where went Ahmed-nom. 'Where did Ahmed go?' Thus, as is clear from the above examples (12a-b), an entire NP moves into Spec, CP, whereas in the example (12c), it is an entire PP which moves to Spec, CP. It 3

The notion of pied-piping and prepositional stranding will be discussed in details in chapter 4.

51

is not just the wh-word that moves, the entire phrase has to pied-pipe along with the whword. However, all the examples we have discussed in this section are examples of simple wh-questions since each of them contains a single wh-element. In the following subsection 2.1.2, we will consider the wh-movement in MSA multiple wh-questions. These questions include the direct or indirect wh-questions. We will prove that only one wh-phrase has to move to Spec, Cp in the case of the multiple wh-questions as in the case of simple wh-questions.

2.1.2. Wh-movement in Multiple Wh-questions In MSA, direct or indirect questions may have more than one wh-element in the same clause. This means that there are more than one questioned constituent in a clause. Examine the following examples: 13. man ishtara madha? who bought what 'Who bought what?' 14. man akhbara man bi-madha? who informed who with what 'Who informed whom of what?' 15. la

'arif-u

man fa'a1a madha /i-man.

Neg. know-I who did

what to-who

'I don't know who did what to whom.' 16. sa'a1a a1iyy-un ahmed-an

man daraba man.

asked Ali-nom Ahmed-acc who beat who

52

, Ali asked Ahmed who beat who.' In the above examples. (13-16), there are more than one wh-element in each question; (13-14) are direct questions, whereas (15-16) contain indirect questions. As for the whmovement in multiple wh-questions in Arabic, only one wh-element moves into Spec, CPo If both subject and object NPs are wh-phrases, the wh-element which should move to Spec, CP is the subject NP and that the wh-object cannot undergo wh-movement in Arabic multiple wh-questions. Observe the following structures of the examples of direct multiple wh-questions in (13-14) which are repeated here as (17) and (18): 17. [mani [ti ishtara madha]) who

bought what

'Who boUght whatT 18. [manj [akhbara tj man bi-madha]] who informed who with what 'Who informed whom ofwhatT In the above examples (17) and (18), we notice that only one wh-element has been moved overtly into Spec, CP whereas the other(s) remain in situ. In MSA, wh-in-situ is also used in echo questions, in simple wh-questions, with a very limited distribution, and in mUltiple wh-questions. We still focus here on the third construction (i.e. multiple wh-questions) and we will attempt to identify which wh-element will move to spec, CP and which will remain in situ. Consider the following example (19): 19. man ishtara madha? [manj [ti

ishtara madha]}

who-nom bought what-acc

53

'Who bought what?' Thus, in Arabic ,structures such as (19) repeated below as (20),

'c' has a strong

wh-feature that attracts only one wh-element into Spec, CP (questions regarding strong wh-feature, etc. will be discussed in Chap 4). Once this requirement is met, the strong wh-feature gets deleted, which means there is no further wh-movement in this clause. 20. [c

lip man

ishtara madha]]

who-nom bought what-acc 'Who bought what?' From the above example (20), we notice that in Arabic wh-questions C has a strong whfeature that attracts only one wh-element into Spec, CPo One of them has to move to Spec, CP and the other has to stay in situ. But which one? Either man 'who' or madha 'what' can satisfy this requirement. But, in the minimalist account, according to Chomsky's Attract Closest Principle (1995), which states that "a head attracts a given kind of constituent attracts the closest constituent of the relevant kind.,,4 Here, the movement of man 'who' is preferred since it is closer to

'c' than madha 'what' because

the distance it must travel is shorter than the distance madha would need to travel in order to check the same feature. This captures the contrast found in the following pair of sentences: 2 La. man; t; ishtara madha? b. *madha; man ishtara t;? In this respect, MSA does not differ from English or Gennan, which fonn

4

.z!- ,

See Chomsky (1995).

54

multiple wh-questions the same way, as examples (22) and (23) show: 22. Who bought what?

English

23. Wer hat was gekauft?

German

If we look at mUltiple wh-questions with a wh-object in situ and a DP-subject (regardless of the fronted wh-word), it is clear that the Arabic pattern in (24a-b) corresponds to English (25) and German (26) respectively: 24. a. mata ishtara al-walad-u when

madha?

bought the boy-nom what-ace

'When did the boy buy what?' b.*mata ishtara

madha al-waladu?

when bought what-ace the boy-nom 'When did the boy buy what?' 25. When did the boy buy what? 26. Wenn hat der Peter was gekauft? Furthermore, a closer look at the ungrammatical (27) below reveals that Superiority violation (in the GB framework) - or the recent Attract Closest Principle (ACT) - in Arabic would also lead to a violation of the Shortest Move Economy Principle (SMC): 27. *[cp madhai bp akhbart-a what

informed-you

[cpman

who

li-yaktub-a td]]? to-write

'What did you inform whom to write what?' The ungrammaticality of (27) above can be accounted for in terms of the Shortest Move Condition (SMC) by which Chomsky (1995) points out that the wh-movement in

55

sentences such as (27) can be excluded in terms of a single condition which he calls the

Shortest Move Condition, which states that a moved category cannot cross over another c-comrnanding category of the same type. (27) is ungrammatical because it involves an overt movement of the lower wh-question word madha 'what' from the direct object position of the embedded clause across a filled A-bar specifier position, which is occupied here by man 'whom'. Thus, the movement of madha 'what' in (27) is longer and also crosses over a c-comrnanding wh-element, namely man 'whom'. According to ACP, (27) can be rendered grammatical if man 'whom', which is the first wh-question word on the right, moves overtly to the root Spec, CPo The reason why the movement of

man is preferable is that such movement is shorter in the sense that it does not cross over another c-comrnanding question word. This is illustrated again in the grammatical sentence in (28) below:

28. [cpmani who

bp akhbart-a

ti

informed-you

li-yaktub-a madha]]? to-write

what

'Whom did you inform to write what?' Thus, we have seen in the above example (28) that the movement of the higher whelement man 'who' involves a shorter movement to make its movement less costly than the movement of the lower wh-question word madha 'what'. This obeys both Attract

Closest Condition and Shortest Move Condition, the economy conditions on whmovement. We will examine these and other constraints on wh-movement in MSA in depth in chapter 3.

56

Now, in the following section, we are going to discuss the MSA wh-movement in free relatives to show whether they are similar to the wh-questions or not. We will also give some information about relative clauses in Arabic. 5

2.2. Wh-movement in Free Relatives Arabic has two different types of relative clauses which include the definite and indefinite relative clauses. The definite relative clauses employ relative pronouns such as aUadhi, aUati, etc. to introduce the relative clauses (29); but the indefinite relative clauses do not employ relative pronouns (30). Both these types have a head NP or an antecedent which precedes the relative clause. Consider the following examples: 29. a'arif-u ar-rajul-a

aUadhi ishtara al-jubnat-a.

know-I the-man-acc who

bought the-cheese-acc

'I know the man who bought the cheese.' 30. a'arif-u rajul-an

ishtara al-jubnat-a.

know-I a-man-acc bought the-cheese-acc 'I know the man who bought the cheese.' From the above examples, we notice that the relative pronoun in MSA follows a definite antecedent only as illustrated in (29). This means that the presence of the relative pronoun in MSA is obligatory when the head of the relative clause is present as a definite noun. There is also another type of relative clauses which can be introduced either by a relative pronoun or by a wh-phrase like man, rna, etc. In other words, Arabic has what 5

See Ryding (2005} for more details on Arabic relative clauses.

57

are called free or headless relative clauses which are similar to wh-questions since they include wh-elements such as man and rna and involve wh-movement. These relative pronouns lack a head NP or an antecedent. For instance, the wh-word man 'who' is used in relative clauses to refer to human entities only as in (31), whereas the wh-word

rna 'what' is used in the same context to refer to non-human entities (32). 31. a' arif-u man; saraqa t; al-masrif-a. know-I who robbed the-bank-acc 'I know who robbed the bank.' 32. a'arif-u rna; tufakkir-u fi-hi;. know-I what think- you of-it 'I know what you think of.' In examples (31) and (32) we notice that the relative clauses are different from the relative clauses discussed in the previous examples (29) and (30) since they are headless and thus the relative pronouns man and rna do not refer back to any nouns and so lack antecedents. With regard to wh-movement, we find that the wh-relative pronouns man and

rna behave like the wh-phrases. They move from their base-generated positions to Spec, CP. In (31), man has moved from the subject position to Spec, CP leaving a trace, whereas rna in (32) has moved to Spec, CP leaving a resurnptive pronoun behind. Furthermore, the relative pronoun man can be used to refer to both masculine and feminine nouns. This can be identified by the verb following man. Consider the following examples:

58

33. sa-usa'id-u man ya-fa'alu al-khayr-a will-help-I who he-do

lil-aakharina.

the-charity-acc for-others

'I will help who does charity for others.' 34. sa-usa'id-u man ta-fa'alu al-khayr-a will-help-I who

she-do

lil-aakhariina.

the-charity-acc for-others

'I will help who does charity for others.' From the above examples, we find that thus the relative pronoun man can be used to refer to both masculine and feminine nouns. This can be identified by the following verb prefix. This is the case in Arabic where the prefix ya- is attached to the beginning of the infinitive verb to refer to the masculine nouns, whereas the prefix ta- is attached to the beginning of the verb to refer to the feminine nouns. But the entities that ma refers to carmot be identified by the following verb because ma is marked as a third person masculine singular. Consider the following example: 35. a'arif-u ma hadatha know-I what happened 'I know what happened.' Thus, the wh-word man can be used in MSA relative clauses just like the relative pronoun alladhi and can have the same grammatical function as the subject of the relative clause. Consider again the following examples: 36. a. a' arif-u alladhi saraqa know-I who

robbed

al-masrif-a the-bank-acc

'I know who robbed the bank.'

59

b. a'arif-u man saraqa al-masrif-a know-I who robbed the-bank-acc 'I know who robbed the bank. ' In the above examples, we notice that the Arabic wh-phrase man can have the same meaning as the relative pronoun alladhi and, hence, can be used in the same context and have the same grammatical function too. But, still there are some differences between the relative pronoun alladhi and while the wh-phrase man as clear from the above examples (36a-b). First, the relative pronoun alladhi refers back to a pronoun or noun phrase as its antecedent and agrees with it in number and gender, whereas the wh-phrase man is headless and so lacks antecedents. Second, the relative pronoun alladhi can refer back to both human and non-human entities, while the whphrase man refers back to +human NPs only. Finally, there is another important difference between the relative clauses introduced by the relative pronouns such as alladhi. allati. etc. and those introduced by the wh-phrases such as man and mao In the former, the relative pronouns alladhi, allati. etc. function like complementizers and hence they are generated in the head C position and there is no movement of alladhi in this. Rather there is movement of a null operator from the position of the NP that is relativized. In the latter case, the wh-phrase moves from its base-generated position to the clause-initial position, that is Spec, CPo This way it is very similar to the wh-question. And this clause also acts as an island for whextraction in MSA.

60

2.3. Word Order in Wh-questions In this section; I will discuss the different types of word order in MSA with relation to wh-movement strategy. I will assume SVO to be the underlying structure from which the surface word orders are derived in MSA. I will also show that Word order in Modem Standard Arabic declarative sentences may have more than one form. Then, I will show that a wh-phrase is followed by a verb and that is the main surface word order in the formation of wh-questions in MSA. In other words, we will deal with these different word orders to show the main word order in the formation of MSA whquestions, which is the main point of the following subsection. We will also examine word order constraints in Arabic wh-questions.

2.3.1. Word Order in Simple Wh-questions In this subsection, we will begin with the word order in MSA simple whquestions and then move on to the word order in multiple wh-questions. Word order in Arabic declarative sentences may have more than one form, i.e.

vsa, sva, avs,

and

vas. This means that Arabic enjoys a great deal of freedom in word order because of its rich inflections. We argue here that the wh-element is immediately followed by the verb and this is the main word order in the formation of Modem Standard Arabic wh-questions, which is the main point of this subsection. Let's, first of all, look at different varieties of the word order in Modem Standard Arabic declarative sentences. Consider the following example:

61

v

o

s

37. katab-a aliyy-un 'al-qissat-a. wrote

Ali-nom the-story-acc

'Ali wrote the story.' Sentence (37) above has VSO order. However, the order of the constituents of the same sentence can be changed to SVO order as in (38) below:

s

v

o

38. aliyy-un katab-a al-qissat-a. Ali-nom wrotethe-story-acc. 'Ali wrote the story.' The above sentence (38) indicates the SVO order. The difference between VSO and SVO orders does not lie only in the different structure of the words, but also in the matter of subject-verb agreement, which we will demonstrate shortly. A number of languages display a phenomenon known as "agreement asymmetry" in which the agreement on the verb differs depending on whether the subject occurs pre- or post-verbally. The asymmetry is generally between partial and full agreement. A well-known asymmetry of this type is found in Arabic in which a post-verbal subject triggers partial agreement of person and gender as in (39) whereas a pre-verbal subject triggers full agreement of person, gender, and number as in (40). The following examples are taken from Miyagawa (2010). 39. qadim-a

(I*qadim-u) al-awaladu.

came-3msg came-3mpl the-boys-3mpl 'The boys came. '

62

40. al-awaladu

qadim-u

the-boys-3mpl came-3mpl

(/*qadim-a) [I]. came-3msg

'The boys came. ' Here are some more examples. 41. al-awlad-u

akal-u

the-boys- 3 mpl

al-mawz-a

ate-3mpl the-banana-acc.

'The boys ate the banana. ' 42. akal-a ate-3msg

al-awlad-u

al-mawz-a

the-boys- 3mpl the-banana-acc.

'The boys ate the banana.' In these, we notice that difference in agreement results in different word orders. This means that person and gender agreement occurs in MSA - that is, partial agreement without number features - when the subject is in the post-verbal position. But number agreement, along with the person and gender agreement, is present only when the subject occurs in the preverbal position. So, in MSA full agreement occurs on the verb when the subject moves to preverbal position, presumably to Spec, TP. 6 However, other orders are also possible in Arabic clauses. Consider the following example:

v

o 43. al-qiassat-a the-story-acc

kataba-ha aliyy-un. wrote-it

'Ali wrote the story.'

6

s

See Miyagawa (2010).

Ali-nom

63

Sentence (43) above has the OVS order. In this order the c1itic -ha has been added to the main verb.

Consid~r

v

o

the following sentence also:

S

44. katab-a a1-qiassat-a wrote

the-story-acc

aliyy-un Ali-nom

'Ali wrote the story.' Thus, the above sentence (44) has the VOS word order. In spite of the different word orders of the above sentences, they have the same meaning. This indicates the richness of Arabic inflectional Morphology. What is noticeable in these examples is there is a kind of permutation in the constituents only. But in (43), in addition to the permutation of the constituents, a c1itic -ha should be attached to the main verb. The clitic itself refers to the object of the declarative sentence. 7 Some scholars have suggested that VOS is the unmarked word order in Arabic. If we take this as an accepted word order in Arabic, we may have confusion in the case of those nouns which have no case markers at the end. Take the following declarative sentence: 45. dharab musa

eisa.

hit-past musa

eisa

'Musa -hit Eisa.' The sentence (45) is grammatical as a declarative and is accepted by the speakers of Arabic. But there is a problem with such sentence where a kind of confusion will arise in the mind of the listener in which he or she cannot identifY the subject of the sentence

7

See Homeidi (2004).

64

from the object. Since there are no case endings on the subject or the object NPs, it leads to ambiguity abput who hit whom. Sentence (46a) below gives the message that the perpetrator of the beating is Musa, while Eisa is the victim of beating; whereas (46b) gives the message that the perpetrator of the beating is Eisa, while Musa is the victim of beating. Thus, sentence (45) can have two meanings repeated as (46a-b) below: 46. a. dharab

musa

hit-past musa-nom.

elsa. eisa-acc.

'Musa hit Eisa.' b. dharab

musa

hit-past musa-acc.

eisa. elsa-nom.

'Eisa hit Musa.' This kind of ambiguity arises because of the lack of case markers on the subject and the object NPs. However, this ambiguity can be resolved pragmatically at LF and clauses with ambiguous reading will receive multiple interpretations and so multiple representations. This cannot be resolved at the PF level. There have been several studies on the word order in Arabic. And Mohammad (2000) summarizes the suggestions on the Arabic word order in literature as follows: "Four word orders have been suggested as the basic word orders for Arabic: Arabic is a flat VSO language (cf. Fassi Fehri 1982; Ayoub 1981-1982; Saad 1975 and Bakir 1980); Arabic is a VOS language (cf. Anshen & Schreiber 1968 and Majdi 1990). Arabic is V-initial with subject and object being unordered with respect to each other (cf. EI-Yassin 1985); and finally Arabic is

65

an SVO language (cf. Emonds 1980 and Borer& Tuller 1985). Thus the four positions can be repuced to two: SVO and VOS." 8 Besides, Shlonsky (1997) suggests that VSO is the basic word order in Arabic. It would be difficult to account for Arabic word order within the minimalist framework if we analyse Arabic as a flat language with VSO as the main word order. Although several word orders are possible in Arabic, in all of these studies, SVO and VSO feature as two prominent word orders in Arabic, as is suggested by Mohammad (2000). Although SVO and VSO are two competing word orders in Arabic, one of them should be assumed as an underlying order from which both the surface SVO and VSO orders can be derived. If we assume VSO as the underlying order, we will face problems in using the binary analysis. In the binary approach, the head and the complement should stay together no matter which follows the other. Hence SVO would be a better choice for the candidacy of an underlying order. Accordingly, we will assume SVO as the underlying order for Arabic. This has several advantages. This will fall in line with Chomsky's (1995) proposal that SVO is the universal underlying word order from which the surface orders of different languages are derived. Secondly, it will comply with Arabic data, since SVO is a frequently occurring word order in Arabic. With these assumptions, we can derive VSO and SVO in the following way. In the SVO order, the inflectional head T has both case and full agreement features matching with those of the subject NP; this triggers operation Agree by which the subject NP raises to Spec, TP to check its case and agreement features. And in the VSO clause, the head T has only partial agreement features, in which case it fails to

• See Mohammad (2000).

66

attract the Subject NP and hence S stays in situ. Thus, we arrive at VSO order. Since T has strong tense features in Arabic, V raises to T in both SVO and VSO orders. This means that the V features ofT are strong T in both SVO and VSO orders. Accordingly, MSA verbs always raise to T since the verbal features in VSO and SVO word orders are strong. However, the subjects remain in situ in the case of the VSO order but raise to Spec, AgrP in the case of the SVO order. This is because the N features of Agr are weak in the VSO order and strong in the SVO order.9 But now the question is whether one has to assume an AgrsP for Arabic. We have seen that agreement does not occur consistently in all clauses. Secondly, in Chomsky's minimalist framework (1995), Agr is eliminated from UG for various reasons. So, on the light of the MP, Chomsky (2000) argues that the Agr node has no role in feature checking, since it has neither case nor theta features. As such, Agr and Tense are both located in T. Thus, we can account for both VSO and SVO orders without assuming an Agr phrase. Consider the following examples: 47. katab-a ahmad-un wrote

al-qissat-a

Ahmad-nom

VSO

the-story-acc

'Ahmad wrote the story. ' 48. ahmad-un

katab-a

Ahmad-nom wrote

al-qissat-a

SVO

the-story-acc

'Ahmad wrote the story.'

9

In SVO order, the verbs raise to Agr to get their strong verbal features checked ofT.

67

Thus, when the head T has full agreement features matching with those of the subject NP, it attracts

S, resulting in SVO order (48) and when the Phi-features on the

head T fails to match with those of the subject NP, it fails attract the subject which results in VSO order (47). While the verb raises to T in both VSO and SVO orders (Benmamoun 2000; Mohammad 2000; Fakih 2006, among others), the subject remains in situ in VSO order and raises to Spec, TP in SVO order (Chomsky 1995). This is clear in the structures of sentences (47) and (48) which are represented in the following tree diagrams (49) and (50) respectively: 49. TP

~T'

Spec

~VP

T

I~ Spec V'

katab-ai

~

V

\ Ahmad-un

V-to-T Movement (VSO)

Obj

I

ai-qissat-a

68

50. TP

~T I~ Ahmad-un T VP Spec

I~ Spec V' ~Obj V \ I al-qissat-a

katab-a

t;

NP'UbJ Raising (SVO)

Thus, the above structures (49-50) show that there is a V-to-T movement in both VSO and SVO word orders since T has strong tense feature. We can, therefore, assume SVO to be the underlying word order from which both SVO and VSO are derived. IO With these assumptions, we will now examine the word order in MSA whquestions. Word order in wh-questions is also derived from the underlying word order, Spec-head-complement. As we have seen in the earlier discussions, a wh-phrase, whether it is an argument or an adjunct, has to move to the Spec of CPo In the discussion that follows I will show that there is one general constraint on word order in wh-questions: "wh-expressions should immediately be followed by the main verb." Consider the following examples:

10

The conclusions about the word order in Arabic wh-questions are due to my supervisor.

;'.-.

69

51. a. madha; katab-a aliyy-un t;? what

wrote. Ali-nom.

'What did Ali write?' b. *madha; aliyy-un katab-a t;? what

Ali-nom wrote

'What did Ali write?' c. *madha; t; kataba-ha aliyy-un? what

wrote-it Ali-nom

'What did Ali write?' d. madha; katab-a t; aliyy-un? what wrote

Ali-nom

'What did Ali write?' In the above examples, we notice that (Sib) and (Sic) are ungrammatical, while (Sla) and (Sid) are grammatical and are accepted by the speakers of Arabic. In all these examples only those in which the verb immediately follows the wh-phrase are grammatical. We also notice that in these examples, it is the object wh-phrase that has moved to the clause-initial position. Now, we will see whether the verb occurs immediately after the subject wh-phrase or an adjunct wh-phrase. Consider the examples below in which a subject wh-phrase has moved to Spec, CPo 52. Awa walad-in qad a'amila

which-boy

has done-3msg

al-wajib-a? the-homework-acc

'Which boy has done the homework?'

70

53. AY.JIa alwaladi qad which boys

a'amil-u

al-wajib-a?

have done-3mpl the-homework-acc

'Which boys have done the homework?' In the above sentences, the verb occurs immediately after the subject wh-phrase. And rightly it shows full agreement with the subject, since the subject is in the preverbal position. Now consider the examples below (54-55) in which an adjunct wh-phrase has moved to Spec, CP:

54. mata dahaba-t when went-3fsg

al-bayt-a?

al-banat-u

the-girls-nom. 3fpl the-home

'Wben did the girls go home?'

55. mata dhahaba-t

al-bint-u

i1a

when did go-3fsg the-girl-nom.3fsg to

al-bayt-i? the-home-gen

'When did the girl go homeT In obedience to the verb-second constraint, the verb here also occurs immediately after

the adjunct wh-phrase mata. The subject, on the other hand, stays in situ within vP, because it cannot raise to Spec, T to occupy a position between the wh-phrase and the verb. Hence, accordingly, the verb also shows partial agreement since it precedes the subject NP. These two sentences (54-55) with an adjunct wh-phrase at Spec, CP give strong evidence for the verb second order in wh-questions. This is true of indirect whquestions also as the following examples show. 56. a. sa'ala-ha asked-her

aliyy-un madha kana ismu-ha. Ali-nom what

was

name-her

71

'Ali asked her what her name was. ' b. *sa'ala-ha aliyy-un

madha ismu-ha kana

57. a. atasa'al-u man tazawaja aliyy-un. wonder-I who married Ali-nom 'I wonder who Ali married.' b. *atasa' al-u man aliyy-un tazawaja. In the above examples (56-57) of the indirect questions, we see that a clause in which the verb immediately follows the wh-phrase is alone acceptable. This is the way Arabic distinguishes an interrogative from non-interrogative clauses. We will now turn our attention to the relative order of other constituents in whquestions. In Modern standard Arabic declaratives, both VSO and VOS orders are equally acceptable. This is so even in the simple wh-questions as can be seen in the examples below. 58. a. mata istara

al-walad-u al-kitab-a?

VSO

when bought the boy-nom the-book-acc 'When did the boy buy the book?' b. mata ishtara al-kitab-a al-walad-u ?

VOS

when bought the book-ace the-boy-nom 'When did the boy buy the book?' Sentences (58a) and (58b) show that in Arabic simple wh-questions two word orders are equally grammatical, i.e. wh-word VSO and wh-word VOS respectively. The subject and object NPs can be freely shuffled around, as shown in the sentences above.

72

However, this is not possible in multiple wh-questions as the following ungrammatical sentence (59b) shows.. 59. a. mata ishtara al-walad-u

madha?

vsa

when bought the boy-nom what-ace 'When did the boy buy what?' b. *mata ishtara madha

al-walad-u?

vas

when bought what-ace the boy-nom 'When did the boy buy what?' From a minimalist perspective, only one order that is, (V)Sa, is possible in multiple wh-questions because, according to Chomsky's Superiority Condition, recently known as Attract Closest Principle (ACP), mata 'when' in (59a) is an adjunct whphrase which has moved to Spec, CP, since it is closer to Spec, CP and is higher than the argument wh-phrase madha 'what'. By contrast, the vas order is not possible in the multiple wh-question as in (59b) because, according to Chomsky's Impenetrability Condition (IC), the wh-phrase mata 'when' originates in the domain of another

complementizer which c-commands it II And in this case the wh-phrase mata is not closer to the CP than the wh-phrase madha. That is why the above example (55b) is ruled out. Thus, the foregoing discussion shows that the verb has to follow the wh-phrase in Modem Standard Arabic wh-questions. But does it seem to be true of yes/no questions also? This is our point of discussion in the following subsection. In other words, in the following subsection, I will try to give further evidence for our assumption 11

More details on Phase Impenetrability Condition will be given in chapter 3.

73

that the verb has to follow the interrogative words in all types of questions in MSA, including yes/no questions also.

2.3.2. Word Order in YES/NO Questions In this subsection, I will give further evidence for our assumption that the verb has to follow the interrogative words in all types of questions in MSA. In Arabic, both declaratives and yes/no questions can be introduced by an overt complementizer, inna 'that' and hal 'if respectively as the following examples illustrate: 60. a. inna al-walad-a taraka al-bayt-a. that the-hoy-ace. left

Declarative

the-house-acc.

'The hoy left the house.' b. hal taraka al-waJad-u if

left

the-boy-nom

al-bayt-a?

YES/NO Question

the-house-acc.

'Did the hoy leave the house?' In the declarative sentence (60a), we find that it is introduced by the Comp inna which corresponds to the English "that". In Arabic, this complementizer inna is always followed by an NP. A verb cannot occur after this Compo This is because inna is a case assigning Comp and the NP that receives case from it has to meet the adjacency requirement. The Comp inna, as we notice assigns an accusative case to the NP that follows it, although the NP is the subject of the clause (60a). But in the example (60b), a yes/no direct question, the verb occurs after the Comp or particle hal, 'if' which denotes that it is a yes/no question. What we have to notice here, as is clear from the previous

74

examples, is that in many languages of the world there is no syntactic difference between a declarative sentence and a yes/no question except the existence of question particle which is phonetically realized. 12 Let's take another example: 61. sa'alt-u-hu hal taraka al-walad-u asked-I-him if left

al-bayt-a.

the-boy-nom. the-house-acc.

'I asked him if the boy left the house.' Sentence (61) contains an indirect question corresponding to the direct question in (60b). In both examples we find that the verb taraka "left" occurs immediately after the comp hal "if' and this is another piece of evidence that the word order in Arabic yes/no questions is that V follows the question particle. This means that verbs occur clause initially in MSA yes/no questions. We can use another yes/no question tool to make sure if the verb occurs clause initialIy or not. 62. a-taraka al-walad-u if-left

aI-bayt-a?

the-boy-nom. the-house-acc.

'Did the boy left the house?' If we have a look at the above sentence (62), we find that the verb occurs clause initially, i.e. the Comp a- "if' is introducing the sentence and is followed immediately by the verb.

However, an interesting complication arises in relation to the Arabic data in (60a) above. The sentence is introduced by the transitive finite complementizer inna 'that'; and the subject al-walad-a 'the boy' is assigned accusative case in accordance

12

See Farghal (1986).

75

with Radford's suggestion (in his Minimalist Syntax (2004», "A transitive head assigns accusative case to a uoun or pronoun which it c-commands." By contrast, sentence (60b) is introduced by the intransitive finite complementizer hal 'if; this is not a transitive comp and here assigns nominative case to the subject al-walad-u (which therefore carries the nominative suffix -u rather than the accusative ending -a). This leads again to Radford's observation that, "An intransitive finite complementizer assigns nominative case to a noun or pronoun expression which it c-commands."

13

The

accusative NP the boy is still the subject of the clause irrespective of the case it gets. In MSA, different types of Comps are used to introduce matrix as well as subordinate clauses irrespective of whether they are declaratives or interrogatives (direct or indirect questions). Consider the following examples:

63. inna al-walad-a

dhaldy-un.

that the-boy-acc. clever-nom. 'The boy is really clever.' 64. hal taraka al-walad-u

if

left

the-hoy-nom.

al-bayt-a?

VSO

the-house-acc.

'Did the boy leave the house?' 65. a'alimt-u anna alwalad-a knew-I

13

that the-boy-ace.

taraka al-bayt-a. left

SVO

the-house-ace.

a. In the case of the example (600), the verb has to have full agreement with the subject the boy. b. the word order in this case is SVO since the comps innalanna are always followed by noun (the

subject)

76

'I knew that the boy left the house.' 66. a-ya'arifu ahadu-hum inn taraka al-walad-u if-know

anybody

if left

al-bayt-a?

VSO

the-boy-nom. the-house-acc

'Does anybody know if the boy left the house?' In the example (63) above, the main clause starts with the overt Comp inna 'that', whereas example (64) is introduced by the overt interrogative Comp hal 'if. 14 In sentence (65), the subordinate clause is introduced by the overt Comp anna 'that'; 15 whereas in sentence (66) we notice that both the main and the subordinate clauses are introduced by two overt Comps a- and inn respectively. In his paper "On Wh-Movement", Chomsky (1977) assumes that clauses begin with Comps which in turn fall into two categories, i) Phonetically filled Comps and ii) phonetically empty comps. Chomsky further specifies Comps as either [+wh] or [-wh]. This assumption holds good for Arabic also. This means that Arabic Comps of CPs have the [+wh] feature which attracts the wh-elements to move into Spec, CPo But Comps which are signaled out as having [-wh] feature can either be phonetically occupied as in (67) or phonetically empty as in (68).

67. inna al-walad-a

fi

aI-bayt-i.

that the-boy-acc. inside the-house-gen. 'The boy is really inside the house.'

14

The Arabic Comp inna is used clause initially to make the sentence emphatic.

" The Arabic Comp anna introduces only the subordinate clauses.

77

68. al-walad-u

fi

al-bayt-i.

the-boy-nom. inside the-house-gen. 'The boy is inside the house. ' If we look at the previous examples (64) and (66) above, we find that they are introduced by yes/no question particles hal and a- respectively. These Comps cause no movement at all. This means that they have no [+wh] feature but are still interrogatives. In other words, yes/no questions are made by placing a particle at the beginning of a sentence without any kind of inversion. Furthermore, yes/no questions show no movement at all. That is to say the difference between the yes/no question and the statement is attributed to the presence of the particles hal and a-. This leads us to consider yes/no questions in Modern Standard Arabic as having [-wh] feature but [+interrog] feature. To conclude, the word order in Modern Standard Arabic declarative sentences may have more than one form, i.e. VSO, SVO, OVS, and

vas. lbis means that Arabic

enjoys a great deal of freedom in word order because of its rich inflections.

In

interrogatives in general and wh-questions in particular the verb has to immediately follow the wh-phrase irrespective of whether the wh-phrase is the subject, object or an adjunct. Thus, in this section we have discussed and examined the different types of word order in MSA and some of their constraints, etc. In the following section we will examine the wh-movement within the copy theory of movement.

78

2.4. Wh-movementwithin the Copy Theory of Movement This section ailnS to examine and explain the properties of wh-movement in MSA within the framework of Chomsky' more recent Copy Theory of Movement. Chomsky (1995) points out that the copy theory of movement simplifies syntax in two ways. First, the theory eliminates the need to postulate new objects (i.e. traces) beyond usual lexical items. In this sense, it brings us closer to a view of syntax as a recursive procedure that does not access anything but items from the lexicon. Second, the copy theory turns movement into a simpler operation; it is practically identical to the elementary structure-building operation, i.e. merge, differing only in one aspect; that is, it takes as input an object that has served as input for an earlier merge. So this way move is nothing but re-merge.

As is known, to Chomsky (1995), the copy theory of movement involves movement of an element that leaves behind a full copy of itself, rather than a trace. Besides, the copy theory of movement supports Chomsky's Inclusiveness Condition (IC) (1995) which limits the power of syntax to rearrangements of lexical items, thus banning syntax from creating new objects. This means that traces are primitive examples of creationism in syntax which violate the IC. In addition, Chomsky (1995) indicates that copy theory of movement noticeably turns movement into a simpler operation by reducing the number of theoretical primitives in the theoretical inventory. Copies are preferred to traces because copies are not new primitives; thus replacement of traces by copies leads to an overall simplification of the grammar and this by itself explains why the copy of theory became one of the pillars of the MP.

79

More importantly, the copy theory can account for both wh-in-situ languages and wh-moving languages by showing that there is movement even in wh-in-situ languages and what is deleted at Spell Out is the highest instance of wh-phrase (Reintges, 2007). This is the way the copy theory unifies wh-moving languages and whin-situ languages. What is more is that it can also account for unacceptable structures occurring in child language. Above all, the copy theory eliminates the need for reconstruction at LF since all the copies of the moved element are available at LF either as null copies or as overt copies. This way, Chomsky demonstrates that it is possible to treat reconstruction as an LF phenomenon and that there is no need for the non-interface levels of representation.

2.4.1. From Trace Theory to Copy Theory We will now take a closer look at the copy theory of movement and see how wh-movement in Arabic can be handled within this framework. In this subsection, we will show how the copy theory can handle ambiguous readings when Binding conditions apply at LF, and show how the copy theory can possibly account for the parametric differences between languages with overt wh-movement and languages with wh-in-situ. In the Government and Binding (GB) model (Chomsky 1981), the displacement property of constituents was considered as having an operation which moved an element from one structural position to another, leaving behind a co-indexed trace which was conceived of as a phonetically unrealized category that inherited the relevant interpretation properties of the moved element. However, the trace was like a pro-form, a new object created during the course of derivation. In this subsection, I am

80

going to discuss wh-traces, i.e. traces resulting from A-bar movement. 16 We will be looking at data from Arabic and English. Consider the following examples: 69. I wondered who(m); John kissed tj.

English

70. tasa'a1t-u

Arabic

manj qabbala aliyy-un tj.

wondered-I who kissed

ali-nom.

'I wondered who(m) Ali kissed.' In the aforementioned examples (69) and (70), we find that the wh-elements (who in English and man in Arabic) have been moved from the object positions of the embedded clauses into Spec, CP of the same clauses in both languages. As I mentioned before, in the Minimalist Program (1995), Chomsky assumes two interface levels of derivation: PF and LF. He incorporates the copy theory of movement into the MP; therefore, within the copy theory of movement, the notion about the nature of movement and the properties of "traces" left behind are significantly different from those found in the earlier trace theory. According to the copy theory, a trace is a full copy of the moved element that is deleted or that receives null spell-out in the phonological component (in the case of overt movement), but is still available for interpretation at LF. Under the copy version of trace theory, patterns in (69) and (70) appear as in (71) and (72) at LF respectively: 71. I wondered [who] John kissed [who]. 72. tasa'a1t-u

16

[man] qabbala a1iyy-un [man].

There are some other types of traces resulting from A-movement (NP-traces) and traces resulting from

head movement operations such as V -to-T movement in French.

81

Regarding the structures of the above sentences at PF, we find that only the highest copy is spelt out phonologically, whereas the lower copy receives null-spell out and consequently is not realized at PF, but is present syntactically and hence available as a null copy of the move element - for LF interpretation. So the null copy of the moved element is like an empty category that is present in syntax. The conceptual underpinning for the revival of the copy theory of movement is provided by the Inclusiveness Condition (lC) which indicates that no new objects are created during the course of derivation. 17 Replacement of traces by copies thus leads to an overall simplification of the grammar. Besides being compatible with the IC, the copy theory has the advantage of allowing binding theory to be stated exclusively in LF terms and dispensing with the operation of reconstruction, i.e. the LF -operation that "puts back" the moved constituent into the position of its trace. The latter is exemplified in (73).18 73. a. Which picture of himself did John destroy? b. [Which picture ofhimseJtli did John destroy t;? c. [Which picture of himself] did John destroy [which picture of himself]? Under trace theory, the moved wh-phrase which picture of himself must be reconstructed in its trace position t; as in (73b) in order for the anaphor himself to be bound by John. Under the copy theory, a full copy of the displaced wh-phrase is available at LF (73c). Consequently, the anaphor himself, being part of the copy in the

17

See Chomsky (1995).

" See Crover & Nunes (2007).

82

"trace" position, can be locally bound by John. This means that under the copy version of trace theory there is no need for an operation like reconstruction in the grammar. The introduction of the copy theory, however, raises some new questions. For instance, one must determine which copies reach the interfaces, that is which copies of the moved element is deleted and which receives overt spell out and why this is so. To put this in the words ofCrover and Nunes (2007): "Elements undergoing movement are generally pronounced and interpreted only in one position and the pronunciation and interpretation positions do not have to coincide. To ensure this under the copy theory, it is standardly assumed that all but one copy of an element X undergoing overt movement is deleted in PF and LF, so that only one copy of X remains at the interface levels. The question is then which copy should survive deletion." According to

Bo~kovic

(2001, 2002, 2004 a, b) and Nunes (1999, 2004), traces

(copies structurally lower in the syntactic representation) may be phonetically realized. The idea of Nunes (2001, 2004) is that the phonetic realization of multiple links of a chain is permitted as far as linearization - understood as the application of Kayne's

(1994) Linear Correspondence Axiom (LCA) - can operate. In particular, multiple copies may be allowed when morphological reanalysis makes some copies invisible to the LCA. 19 In his analysis of data from Coptic Egyptian (CE), Reintges (2007) argues for a new type of wh-in-situ, in which the copy privileged for phonological realization is the

'9

See Crover & Nunes (2001).

83

lowest member of the wh-chain, while the head of the chain as well as the intermediate copies are left unprono}.lllced. Coptic can be described as a wh-in-situ language in which the wh-clefting and wh-fronting are available as marked wh-interrogative strategies. The wh-in-situ pattern is marked morphologically by "relative tenses", so called because a relative marker appears in front of the tense-aspect-mood inflection. Reintges argues that the wh-in-situ and wh-fronting structures in Coptic are both derived by applications of wh-movement before Spell-Out. It is proposed that the simultaneous pronunciation of the topmost wh-copy and the relative marker are prohibited by an economy filter on the morpho-syntactic encoding of wh-dependencies, which is similar to the "Doubly-filled Comp" Filter in English. The apparent distinction between whmovement and wh-in-situ constructions lies in the deletion of the wh-element or the relative marker. What is interesting about Reintges' account is that it unifies both wh-in-situ languages and wh-fronting languages by showing that there is overt wh-movement in both types of languages and that it is only at PF, i.e. at the Spell Out stage, that these languages differ depending upon which instance of the moved wh-element is deleted. We will now take a closer look at the copy theory in order to consider which copy is deleted in languages like Arabic and English and why. In the copy theory, it is standardly assumed that all but one copy of an element X undergoing movement is deleted in PF and LF, so that only one copy of X remains at the interface levels. The question is then which copy should survive deletion. It is generally assumed that at LF we have at least some choice in deciding where deletion should take place in nontrivial

84

chains, with a preference for deletion in the head of operator-variable chains. Consider, for instance, the ambiguity of the anaphor himselfin the following examples: 74. JOhni wondered which picture of himselfilj Jimj bought. 75. a. John wondered [cp[which picture of himself] bp Jim bought [vAlish )lister!! sf himself]]] b. John wondered [cp [which )lieffire sf himself]

[IP

Jim bought [whieh picture of

himself1]] Chomsky (1993) argues that the upstairs reading of himself is obtained after the tail of the wh-chain is deleted, as shown in (75a), where himself is locally bound by

John.2o On the other hand, under the downstairs reading, himself is deleted in the head of the wh -chain and remains in its tail, as shown in (75b), where himself is locally bound by Jim. This is the case in MSA, also. Let's consider the following example from Arabic: 76. tasa'ala

aliyy-un [cpray'ya siwari-hi]

wondered Ali-nom.

['ajaba-t aIunad-un [8}:;'B siw8I'i hi)]]

which-picture-gen.-he liked-it Ahmed-nom.

'Alii wondered which picture of himself; Ahmed liked. ' In the above example (76), we find that in Modern Standard Arabic the lower copy a,v.ya siwari-hi is also deleted as is in the case of English; that is to say that lower copies cannot be phonetically realized in the overt wh-movement both in English and

20

Chomsky (1993) actually assumes that LF deletion in cases such as (75a) needs to leave a variable

(without internal sbUcture) behind, at least for O-theoretic reasons. This however might become unnecessary under

Bo~kovic

and Takahashi's (1998) feature approach to O-roles (see also Lasnik 1995,

Hornstein 1999,2001, Manzini and Roussou 2000, etc., among others.).

85

Modem Standard Arabic. As has been shown in the foregoing sections above, on the LF side we have the choice in deciding where deletion should take place. But it is assumed that no such choice is available in PF; the head of a chain is always the sole survivor as is clear in the following examples (77a-e). 77. a. [[the student] was arrested [tAli stuallllt]] b. * [[the Sftidellt] was arrested [the student]] c. * [[the stl!at!Bt] was arrested [the student]] d. *[[the stuaem] was arrested (tfle student]] e. * [[tAt! student] was arrested [the stl!dem]] Thus, there is a preference for deleting traces rather than the heads of chains. This preference is considered from the point of view of economy considerations. Nunes (1995, 1999, and 2004) provides an approach involving a combination of a convergence requirement with economy considerations. His proposal is that a syntactic structure containing a chain cannot be linearized if we assume Kayne's (1994) Linear Correspondence Axiom (LCA), according to which the linear order of terminals is

dependent on asymmetric c-command. Take the structure in the following example (78): 78. [[the book] [was [found [the book]]]] Given that the higher occurrence of the book in the above structure (78) asymmetrically c-commands was, the LCA dictates that both the and book should precede was. Likewise, given that was asymmetrically c-commands the lower

86

occurrence of the book, it should precede the and book in compliance with the LCA. Similarly, given that tbe higher occurrence of the book asymmetrically c-commands the lower one, we get the undesirable result that the book should be required to precede itself. In the Minimalist Framework (Chomsky 2000, 20001), movement is an operation that does not introduce a trace but rather leaves behind a complete copy of the moved element, with the result that structures formed by movement exhibit multiple copies of the moved element. In the phonology, it is determined which of these copies is phonetically realized, the usual assumption being that only the highest copy is pronounced. However, as demonstrated by Reintges (2007), the copy theory of movement allows for an option which is not available in the trace theory; namely, languageparticular conventions might dictate that the copy privileged for pronunciation is the lowest, as opposed to the highest. This possibility has been productively explored by Groat and O'Neil (1996), Pesetsky (1997), Boskovic (2001), Bobalijk (2002), and others. On this view, certain types of wh-in-situ instantiate "hidden" movement configurations, created by the wh-movement in narrow syntax followed by the pronunciation of only the lowest copy. As for as languages like Modem standard Arabic and English are concerned, only the highest copy of the moved element is pronounced, and all other copies are deleted or receive null spell out. Since both Arabic and English are wh-movement languages.

87

2.5. Conclusions We conclude that Arabic is similar to English in that wh-questions in Arabic share many characteristics with those of English. That is to say, they both have strong edge feature which require the wh-element to move from its base-generated position to Spec, CP, i.e. to the clause-initial position. Sections 2.1-2.1.2 in this chapter examined different types of wh-questions, such as simple wh-questions, multiple wh-questions, etc. Our study shows that when there are multiple wh-elements in a single clause only one moves to the clause initial position whereas the rest stay in-situ just like in English. And the one closest to the CP moves in obedience to the

Attract Closest Condition. In section 2.2., we dealt with the wh-questions in MSA free relatives. We found those differences between RCs introduced by relative pronouns such as

alladhi, allali, etc. and those introduced by wh-phrases such as man, mata, etc. In the fonner, the relative pronouns alladhi, allati. etc. function like complementizers and hence they are generated at the head C position and there is no movement of alladhi in this. Rather there is a movement of a null operator from the position of the NP that is relativized. In the latter case, the wh-phrase moves from its base-generated position to the clause-initial position, that is spec, CPo This way it is very similar to the whquestion. And this clause also acts as an island for wh-extraction.

In sections 2.3-2.3.2 we also looked at the varieties of word order in MSA whquestions - simple and multiple wh-questions - and declarative sentences. An important characteristic of the wh-questions in Arabic is that the verb immediately follows the moved whphrase. A discussion of the main features of the Arabic yes/no interrogative system has also shown that the verb in yes/no questions also occurs immediately after theCOMP

hal which optionally occurs in the clause-initial position of the interrogative sentence.

88

The points made here apply not only to Modern Standard Arabic but to the other varieties of Arabic as .well. 21 Thus, we have seen that Arabic yes/no question formation has a structure similar to that of a corresponding declarative sentence, since a declarative sentence can also frequently have VSO order. There is no movement or verb inversion; instead, a complementizer is inserted at the beginning of the declarative sentence to form Arabic yes/no question. 22 That is to say that inversion in Arabic does not turn a declarative into a question, as it does in English. It is the rising intonation with or without

hal and the morpheme

Q-

that signals an utterance as a yes/no question, whereas in English

inversion turns a sentence into a question with or without rising intonation. All said and done, the declaratives have other orders too, where as in yes/no questions the verb follows the yes/no Comp which distinguishes the interrogative from the non-interrogative. The important point to be noted here is that both wh-questions and yes/no questions share an important characteristic; that is, the verb occurs immediately after the wh-phrase in whquestions or a COMP in yes/no questions, which need not be the case in noninterrogatives.

21

Modern Standard Arabic is generally a written fonn, whereas Colloquial Arabic is generally a spoken

fonn. llIn some cases, Arabic declarative sentences can be considered as VesINo questions without the use of question particles such as hal 'if. This is by rising the intonation. Consider the following example: i. katabt-a

al-wajib-a?

wrote-you the-homework-acc 'Did you write the homework?' Thus, rising intonation is the main detenninant of whether or not a sentence is a yeslno question. In this respect, Arabic yeslno questions bear close resemblance to English declarative questions.

89

In section 2.4, we discussed wh-movement in MSA within the framework of Chomsky' more recent Copy Theory of Movement. A general discussion of the development of the Copy Theory comparing it with the Trace Theory was given. We have stated that in the Government and Binding (GB) model (Chomsky 1981), the displacement property of constituents was considered as resulting from an operation which involved moving an element from one structural position to another, leaving behind a co-indexed trace which was conceived of as a phonetically unrealized category that inherited the relevant interpretation properties of the moved element. But in the Minimalist Program (1995), Chomsky assumes only two interface levels of derivation: PF and LF. He incorporates the Copy Theory of Movement into the MP. According to the copy theory, a moved wh-element leaves behind a full copy of itself, rather than a trace. The conceptual underpinning for the revival of the copy theory of movement is provided by the Inclusiveness Condition (IC) by which he indicates that no new objects are created during the course of derivation. Traces are prime examples of creationism in syntax and, as such, violate the Inclusiveness Condition. Replacement of traces by copies leads to an overall simplification of the grammar. Under the copy theory, it is standardly assumed that all but one copy of an element X undergoing movement is deleted/receives null spell out in PF and LF, so that only one copy of the moved wh-phrase remains overt at the interface levels and the other copies are available in a null form. This means that all copies of the moved element are available for LF interpretation.