Remarks on the Arabic complementizer ‘inna Observações sobre o complementizador ‘inna em árabe

: A Standard Arabic (SA) complementizer known as ‘inna poses a restriction on word order in the clause it introduces and induces accusative Case-marking on the otherwise nominative preverbal NPs. 1 Following Chomsky’s (2001) account of the morphosyntax of Case, this paper argues that ‘inna is a Case assigner and thus it carries an uninterpretable Case feature that determines the value which it assigns to an unvalued Case feature concerning accessible goal within A-bar projection. The paper shows that this argument captures the restriction imposed on ‘inna- clauses.


Introduction
Arabic allows both subject-initial and verb-initial clauses. As sentences (1a) and (1b) demonstrate, respectively, the verb can either precedes or follows the subject: 1. a. l-'awlaad-u qara'-uu l-kitaab-a the-boys-NOM read.3PM the-book-ACC 'The boys read the book.' b. qara'a l-'awlaad-u l-kitaab-a read.3SM the-boys-NOM the-book-ACC 'The boys read the book.' The verb also shows full agreement in subject-initial-clauses (1a), but partial agreement in verb-initial-clauses (1b) in person and gender only, and not in number as the former does.
In addition, Arabic is considered as a subject pro-drop language. The verb shows full agreement when its subject is not overt: 2. qara'-uu l-kitaab-Acc read.3PM the-book-Acc 'The boys read the book.' However, subordinate clauses introduced by 'inna are restricted to subject-initial clauses. 2 A typical example is given here: 3. qultu 'inna l-'awlaad-a qara'uu l-kitaab-a said.1S that the-boys-ACC read.3PM the-book-ACC 'I said that the boys read the book.' As can be seen, 'inna is followed by an accusative NP which can be interpreted as a subject of the following verb. However, this accusative NP is not always interpreted as a subject. The following shows that the accusative NP is interpreted as an object: 4. qultu 'inna l-kitaab-a qara'a=hu l-'awlaad-u said.1S that the-book-ACC read.3S-it the-boys-NOM 'I said that (as far as) the book, the boys read it.' Notice that there is a pronominal clitic attached to the verb, a similar construction with no pronominal clitic is ungrammatical: 5. *qultu 'inna l-kitaab-a qara'a l-'awlaad-u said.1S that the-book-ACC read.3S the-boys-NOM 'I said that (as far) the book, the boys read it.' Notice also that you cannot have a gap in the preverbal position. As stated above, preverbal subjects are optional in null CPs, but it is not possible to have a gap in the embedded preverbal position (here and subsequently the paper marks gaps by '__'): 6. *qultu 'inna ____ qara'-uu l-kitaab-a said.1S that read.3PM the-book-ACC 'I said that the boys read the book.' Thus, the example above, where the subject has been omitted, is ungrammatical. However, a similar example with a pronominal clitic attached to 'inna is grammatical: 7. qultu 'inna=hum qara'-uu l-kitaab-a said.1S that-they read.3PM the-book-ACC 'I said that the boys read the book.' Note also that verb-initial clauses cannot occur in the domain of 'inna: 8. *qultu 'inna qara'a l-'awlaad-a l-kitaab-a said.1S that read.3SM the-boys-ACC the-book-ACC 'I said that the boys read the book.' This paper attempts to provide an account of accusative Case checking in this construction, in which the accusative preverbal NPs are embedded under a Case-assigning complementizer. In addition, it will account for the restriction that is imposed by 'inna on its clauses. It is important to mention that 'inna can introduce complement clauses; 'inna in this type of clauses signals that the clause under its domain is subordinated to the matrix clause. In addition, 'inna can introduce independent clauses (or sentence-like) in that the complement clauses are similar in form to what it would have as an independent clause without 'inna. However, in both cases, clauses under the domain of 'inna have an identical form.

Theoretical background
This section introduces the major concepts of the Minimalist program (MP) which has played a crucial role in the analysis introduced in the paper. Chomsky (1995) argues that the human language faculty consists of lexicons and derivational systems. There are two major operations: Select and Merge which operate over a group of lexical items named Numeration to form syntactic structures. The language faculty allows these syntactic structures to appear only in a binary set. Of the two, Merge is that operation which acts free in the syntactic component of Language (CHOMSKY, 2004, p. 108). It is a combinational operation which forms a syntactic object by merging two linguistics expressions (α and β) and form a new unified linguistic expression, resulting in the structure:

Merge and Move
In this sense, Merge is a recursive structure-building process operating on linguistic expressions based on their selectional features. For example, an X is a head and carries an uninterpretable feature which requires it to merge with a ZP to form an XP, resulting in deleting X's selectional feature. This is the first instance of Merge, called External Merge. The other instance is Internal Merge which is understood as Move (CHOMSKY, 2001). Move deals with linguistic expressions and phrases. It applies to the merged linguistic expressions; it places a copy of the object in another position. Move is triggered by the requirement to satisfy the Edge feature (EF) of a specific functional head. Move is required to take a place early in the syntax before the operation Spell Out which transfers the structure to the phonological component (PF) and the semantic component (LF). 3

Interpretable vs. uninterpretable features
Features are divided into two kinds of features: interpretable and uninterpretable features. Some of them are legible by semantic component, whereas others are not. Those with semantic component would get a semantic interpretation and thus would be interpretable, whereas the others that would not get a semantic interpretation, would be uninterpretable dues to the absence of the semantic component.
Likewise, functional and lexical categories too have a set of features. Functional heads carry 'formal features' such as person, number and gender (CHOMSKY, 2001). They are uninterpretable, and thus enter the derivation unvalued as they have no effect on semantic interpretation of heads such as C, T, and v at LF. By contrast, features on nominal expressions are important for their semantic interpretation, and thus would enter the derivation valued. However, the Case feature on nominal expressions has no semantic role, and thus it is uninterpretable at LF.

Agree
Unlike Merge, Agree is concerned with features rather that with lexical items. Its crucial function is to value these features which enter the derivation unvalued and to delete uninterpretable features that have no semantic content. Agree establishes a relation between Probe and Goal both of which have to be active with uninterpretable features. 4 In order to value its unvalued uninterpretable features, Probe searches for an active Goal in its c-commanding domain. A Probe is an uninterpretable feature carried by a minimal projection, while its Goal is an interpretable feature of the same type carried by a maximal projection.

Case Assignment
In the Minimalist Program, Case assignment continues to play a major role in the derivation of syntax. Abstract Case, in the original presentation of Case theory in Chomsky (1980), is related to the morphological property Case. The formal features that regulate the distribution of NPs are the same features that are overtly considered as Case morphology in some languages. Within GB framework, Chomsky (1981) proposed the Case Filter as a solution to the ambiguity of the distribution of lexical NP subjects in infinitive clauses in English as illustrated in (10) The subject cannot be overt in (10a,e), but this restriction is relaxed when the infinitival clause functions as a complement of a specific class of matrix verbs like the verb believe (10b), or when the infinitival clause includes the prepositional complementizer for (10c,d). Where the overt lexical NP subject is not permitted, the subject of the infinitive is considered as a silent pronominal element PRO. This assumption is the key of the Case Theory which proposes that all lexical NPs require Case (CHOMSKY, 1981, p. 49). A rather basic proposal of Case assignment for English is in (11): 11. a. subject of tensed clause: nominative b. object of verb: accusative c. object of preposition: accusative (or oblique) This is to say that verbs and prepositions have the distinctive properties of being Case assigners and this property accounts for why only verb and preposition in English take NP complements. Nouns and adjectives are not Case assigners and therefore are restricted to CP and PP complements (BOBALJIK;WURMBRAND, 2012, p. 46).
For Minimalism, the central study of Case Theory is to investigate the differences between nominative and accusative Case assignments and to develop a uniform theory with them. To achieve that, Chomsky (1991) proposal was to assimilate accusative Case assignment to the similar type of structural configuration as nominative, namely Spec(ifier)-head relation (for more information, see KOOPMAN, 2006). 5 The proposal assumes that all Case assignments are subject to c-command and locality, 6 the relation is later termed Agree (CHOMSKY, 2000). 7 This suggests that all subjects in Spec-IP (Spec-TP, in the most recent Minimalist works) are moved there from a lower position (this proposal is originally proposed in KOOPMAN; SPORTICHE, 1991).
The mechanism of the uniform Case assignment considers the functional versus lexical differences in the Case assigners. Chomsky (1991) and Johnson (1991) assume that VP-external functional projection is responsible for Case on objects, and this assumption in turn leads to unify a proposal that Case is assigned by functional heads (see WURMBRAND, 2001, for empirical evidence).
Under the most recent Minimalist conceptions, Case is generalized as part of a system of uninterpretable features that takes a place at the core of the linguistic coding of what Chomsky (2004, p. 110) called 'duality of semantics': one part is the argument structure and the second is the information structure. So, Case features permit the proper working of the Probe-Goal system, a feature-checking mechanism that was not understood in GB framework (see also PESETSKY; TORREGO, 2001, for another speculation).
However, the literature (e.g. HARLEY, 1995;SCHUTZE, 1997;OUHALLA, 1994;AOUN et al., 2010) shows that Universal grammar contains a notion of 'Default Case' which has a mechanism different from the one discussed above. The notion of 'default Case' is advanced in Marantz (1992) as the Case that does not interact with the Case Filter or 'feature-Checking'. So, what is mechanism of the Default Case' It is the mechanism that is used to spell out NPs that are not in association with the mechanism of the feature-checking. Thus, I assume that the model of grammar schematically follows three nominals through the syntactic derivation: two with an uninterpretable (ACC or GEN) feature to be checked and one with no Case (NOM) feature. The NPs with NOM Case feature survives at Spell-out level, given that it never had any uninterpretable features to be checked.

Discussions
In section (1), it has been shown that the adjacent of the complementizer must be in accusative Case: 12. qultu 'inna t-tabiib-a waSala said.1SM that the-doctor-ACC arrived.3SM 'I said that the doctor arrived.' 'inna heads finite clauses and the embedded preverbal 'subject' NP bears the accusative Case as it is obvious from the accusative Case on T-tabiib-a (-a is an accusative marker).
The fact that the embedded 'subject' is assigned an accusative Case raises an important question about its status with respect to the Case-assigment. Following the assumption that heads are endowed with Case features which must be checked, Mohammed (2000) assumes that both T and C assign their Case feature on the embedded preverbal NP (T-tabiib-a, the doctor, in (12)). He adds that the Case feature overtly shown on the NP is the one assigned by the highest projectional head which is C in this sense. However, it is not clear how it is possible for the head of CP to assign the accusative Case to an element that is located in the specifier position of TP (Spec-TP). The idea that an NP can get more than one Case is required to be constrained by some locality conditions, otherwise it is difficult to prevent C (the highst head) from assigning its Case to some element in a lower position. Consider the following example where the predicate precedes the subject, the subject is located in a lower position: 8 13. qultu 'inna=hu waSala t-tabiib-u said.1SM that-he arrived.3SM the-doctor-NOM 'I said that the doctor arrived.' Here, the expletive -hu is attached to 'inna but the subject (t-tabiib-u, the doctor) is in a position following the verb and it is in nominative Case that is formally assigned by T. The Case here is structural and not inherent since inherent Cases are limited to lexical elements that get a thematic role from the Case assigner which is not the case in (13). In the following, I will argue that the embedded preverbal NP is not located in Spec-TP, but rather in a position located between CP and TP. It is more likely in Spec-TopP.

Valued Case features
Given that Case is an unterpretable feature which needs to be checked and deleted (CHOMSKY, 1995), and that NPs in Spec-TP must be assigned Case being in an argument position (CHOMSKY, 1981(CHOMSKY, , 1986 and that the head of Spec-TP (or agreement) assigns the nominative Case under some versions of Case Theory, the status of the accusative NP in Spec-TP would be difficult to explain. Consider the ungrammaticality of the following example in which a gap occurs in the position of direct object of the verb: 14. *'inna l-kurat-a rakala l-walad-u __ that the-ball-ACC kicked.3SM the-boy-NOM 'That, the ball, the boy kicked.' Accoerding to Pesetsky and Torrego (2001), the ungrammaticlaity of (14) can be explained as follows: complementizers have some uniterpratable features that must be licensed by an element with interpretable features, presumply the complementizer 'inna has an uninterpretable Case feature which must be discharged. As simplified in (15), the accusative NP in (14) is a focus-fronting that raises from its original position to the specifier of a functional projection, named FocP, in order to receive a contrastive reading.

[ ForceP [ Force 'inna][ FocP l-kurat-a [ Foc Ø [ FinP rakala l-walad-u l-kurat-a ]]]]]
Here, the fronted NP, l-kurat-a, is Case-marked in its canonical position; the accusative Case is a reflection of the sharing properties between the fronted NP and the associated gap. So, if the accusative NP l-kurat-a, is assigned its accusative Case by virtue of being a nominal goal to the lexical verb, rakala, the Case feature of the complementizer 'inna would remain unchecked. The resulting derivation crashes, as we see from the ungrammaticality of (14).
However, the sentence in (14) can be repaired by inserting a pronominal clitic that is cliticized onto the verb as an accusative direct object of the verb. A typical example of this construction is given below (the considered clitic is in boldface): 16. 'inna l-kurat-a rakala=ha l-walad-u that the-ball-ACC kicked.3SM-it the-boy-NOM 'The ball, the boy kicked it.' Derivationally, the Case feature of the lexical verb is checked against the Case feature on the pronominal clitic. The NP, l-kurat-a, therefore, is not a fronted focused object, but rather it is a left-dislocated topic that occurs in A'-position. Aoun et al. (2010, p. 191) state that clitic-left-dislocations are realized by the appearance of a NP in the left peripheral position of the clause and it is associated with a pronominal clitic inside the clause. Assuming this to be so, the left-dislocated-topic, l-kurat-u, is active because its Case feature has not yet been valued. Assuming the complementizer 'inna is a Case-assigner, it carries an uninterpretable Case feature which determines the value that it assigns to an unvalued Case feature on the accessible goal (RADFORD, 2009, for more discussions). Consequently, the complementizer 'inna will enter the derivation carrying a feature which enables it to assign accusative Case to the left-dislocated-topic, l-kurat-u, the goal, which has an unvalued Case feature.

Left-Dislocation like Property
First, embedded preverbal accusative NPs obligatory occur in the kind of peripheral position that left-dislocated phrases do. This can be captured by the observation that the embedded preverbal accusative NPs can occur to the left of the copula kaan ("be"), but not on the right. Consider the contrast: 17. a. 'inna t-taalib-a kaana fii l-jaami'at-i that the-student-ACC was in the-university-GEN 'The student was at the university.' b. *'inna kaana t-taalib-a fii l-jaami'at-i that was the-student-ACC in the-university-GEN 'The student was at the university.' The contrast between (17a) and (17b) suggests that the accusative NP should be in a position higher than T.
Second, Soltan (2007) and Alotaibi (2015) argue that preverbal subjects are taken to be genuine topics that are associated with a null resumptive pronoun, pro, in the clause (see also MOHAMMAD, 2000;FASSI FEHRI, 1993;AOUN et al., 2010). This approach also assumes that NPs that appear in structural positions where there is no Case assigner they are assigned nominative Case by a default mechanism. Consider the contrast between the following examples: 18. a. jaa'a l-'walaad-u came.3SM the-boys-NOM 'The boys came' b. *jaa'uu l-'walaad-u came.3PM the-boys-NOM 'The boys came' SA has both preverbal and postverbal subjects and that they differ with respect to the agreement fact. The former triggers number, person and gender agreement, while the latter triggers only person and gender agreement. (18a) is grammatical because the verb, jaa'a 'came' agrees with its NP subject in person and gender, but not in number. The verb, however, in (18b) agrees in number as well and hence the sentence in ungrammatical. Now consider the following contrast with subject-initial clauses: 19. a. l-'walaad-u jaa'uu the-boys-NOM came.3PM 'The boys came.' b. *l-'walaad-u jaa'a the-boys-NOM came.3SM 'The boys came' (19a) is grammatical because of that the subject triggers full agreement, and (19b) is ungrammatical as the subject triggers partial agreement. This suggests that subject-initial clauses (19a) have a pro subject in a post-verbal position. Thus, full agreement is expected with a clause that is includes a pro subject.

The Non-Identity Effects
Following Miller and Sag (1997), I assume that clitics in Arabic are affixes realizing an otherwise unexpressed argument, and it is not a result of some superficial cliticization (see McCLOSKEY, 2006, for more details). I shall call this kind of arguments pro. This pro is in fact the resumptive pronoun. Under the copy theory of movement, it should be clear that the relation between the accusative NP following 'inna and pro is not generated via movement. The theoretical assumptions of Minimalism assume that movement leaves a copy with identical syntactic features. Adger and Ramchand (2005) argue that movement can be involved in cases where the apparently displaced constituent shows the same copy in the base position. More precisely, if the element in the higher position shares its corresponding in the lower position in respect to agreement, selection and Case-marking, then it can be said that the derivation involves movement, otherwise it should involve basegeneration account. In 'inna-clauses, the distribution of Case-marking between the accusative NP and the pro at the foot of the dependency is not the same. This is supported by the following examples for both independent and dependent 'inna-clauses, respectively: 20. a. 'inna r-rajul-a hajama 'alai=hi l-'sad-u that the-man-ACC attacked.3SM on-it the-lion-NOM 'The man, the lion attacked him.' b. qultu 'inna r-rajul-a hajama 'alai=hi l-'sad-u said.1S that the-man-ACC attacked.3SM on-it the-lion-NOM 'I said that the boys read the book.' Here, the topicalized prepositional object bears an accusative Case which is distinct from the one that is associated with in its base position. The accusative Case on the embedded NP would be surprising under the movement account. This would argue that the accusative NP does not originate in an argument position of the lower predicate, but rather it originates in A'-position, namely the specifier position of Topic projection. (20) is diagrammed in (21) According to our analysis the accusative Case on the preverbal NP r-raji-a 'the man' is assigned under c-command by an appropriate kind of head. So, since the complementizer 'inna c-commands the subject, r-raji-a, and since 'inna is a transitive complementizer, it follows that the NP r-raji-a, the man, will be assigned accusative Case at the stage of derivation shown in (21). However, 'inna can also be followed immediately by a PP which can intervene between 'inna and the accusative NP. The following example expresses this fact:

qultu
'inna fii l-bait-i rajul-an said.1SM that in the-house-GEN man-ACC 'I said that there is a man in the house.' Mohammad (2000, p. 22) observes that in 'inna-clauses no thing can intervene 'inna and the preverbal accusative NP except Prepositional Phrases (PPs). A plausible question arises is: why PPs and not others' A similar case is found in Italian sentence structures. For instance, Belletti (2004, p. 26)  As mentioned in Belletti (2004), the sentences in (23) must be pronounced with continuous intonation without a special break between S and its complement. 9 As for Belletti (2004), he suggests that XPs can be PPs because the PPs do not absorb Case as they need no Case, while NPs need Case. Therefore, Belletti (2004) assumes that the intervening of the subject with its already checked Case between XP and the responsible of the Case assignment would cause a Defective Intervention Effect (DIE) (CHOMSKY, 2000, p. 123)  DIE 9 However, one reviewer suggests that sentences in (23) wouldn't label as ungrammatical. The reviewer adds that they are perfectly acceptable with focus on MARIA; they would be possible in a focus configuration.
If this analysis is on the right track, it is possible to extend the same analysis to account for the contrast between (22) and (5, repeated in 25).
25. *qultu 'inna l-kitaab-a qara'a l-'awlaad-u said.1S that the-book-ACC read.3S the-boys-NOM 'I said that (as far) the book, the boys read it.' If PPs in Arabic do not have a Case feature, then their intervention between the Case assigner 'inna and the accusative NP would not cause any problem as the uninterpretable Case feature on NPs would be able to undergo feature checking.

Conclusion
This paper has investigated the behaviour of 'inna-clauses in Standard Arabic. The findings of this study reveal that 'inna assigns the accusative Case to the closest NP via Agree. It has also been argued that the accusative NP following 'inna occupies Spec-TopP and not Spec-TP as proposed in the literature. In addition, this study has accounted for the fact that PPs and not NPs can intervene between the complementizer 'inna and the accusative NP. It has explained this contrast in term of the Defective Intervention Effect in which a probe-goal relation holding between the Probe ('inna) and the Goal (the accusative NP) is blocked by an intervening active goad such as NPs but not PPs. NPs triggers a Defective Intervention Effect that bars the complementizer from entering into Agree relation with its goal,. On the other hand, PPs do not absorb a Case and therefore they are not problematic for the derivation.