Marking Mirativity in Syntax: Minimalsit Mechanisms

This paper offers a generative minimalist investigation to the derivation and interpretation of mirativity information in Central Najdi Arabic (CNA), arguing that grammar of CNA morphologically marks mirativity in syntax by means of the discourse particle wara. Implementing minimalist mechanisms (Chomsky 2001), it is shown that wara instantiates a functional, discoursal projection MrvP in the left periphery of the sentence, articulating the feature [Mrv] at the PF-interface. LF-interface analyses demonstrate that [Mrv] on wara is interpretable/ valued, while the counterpart on the subject DP that wara marks is uninterpretable/unvalued. Agree between wara and the subject DP creates a PF-chain wara>SubjectClitic>SubjectDP that results in the subject DP being marked with and interpreted mirativity at LF-interface. Further explorations show that movement of the subject DP across wara is only legitimised if the subject DP has a discourse, information structural feature beyond [Mrv]. Evidence for this claim comes from the fact that when wara marks the subject DP with mirativity, the subject DP remains in situ. Thus, on minimalist empirical groundings, movement is argued to be motivated by interpretive reasons beyond mirativity. Further analyses show that Agree between wara and subject is of mutual manner; wara u-[φ]-probes the subject goal, while the goal seeks valuation of u-[Mrv] on it (Alshamari 2017).


Cross-linguistic, Minimalist Background on Discourse Particles
Characteristics of the pragmatic distribution and the semantic scope of discourse particles have motivated significant minimalist research with respect to their syntactic position (Zimmermann 2011;Biberauer et al. 2014;Bayer & Trotzke 2015;Bayer 1996;Bayer & Struckmeier 2017). This line of research has since focused on the interaction of discourse particles with the clause internal syntactic items, giving rise to the widely held assumption that discourse particles carry information structural values that affect the interpretation of the sentence as well as its internal constituents. This can be wide scope, in which case a particle scopes over the whole proposition expressed by the clause, or narrow scope, in which case a particle scopes over certain parts of the proposition expressed by the clause. Hence, discourse particles contribute to the interpretation of the propositional content of the utterance (Biberauer & Sheehan 2011;Biberauer et al. 2014;Bayer & Struckmeier 2017).
Within minimalist practice, this view has given insights to mapping syntax to semantic/pragmatic interface system, Published  initiating a new view on the nature of discourse particles. Discourse particles start life in the CP domain, the locus of discourse information. They are articulated in the spine of CP domain, scoping and carrying discourse values which trigger syntactic operations like agreement and movement that affect, for instance, word order (Cruschina 2009;Miyagawa 2010;Alshamari 2017a,b;Alshamari & Holmberg 2019a,b;Jarrah 2017Jarrah , 2019. Having recently been developed within the minimalist practice, research adopting cartographic approach (Rizzi 1997) has been fruitful in detecting the position of discourse particles and mapping the left periphery of a huge range of language cross linguistically, investigating information structural values like topic, focus, modality which are encoded in syntax as overt functional heads in the C domain (Thumair 1989;Haegeman 1993Haegeman 1994Ouhalla 1997;Aikhenvald 2004;Zimmermann 2009;Paul 2009;Cruschina 2009;Coniglio & Zegrean 2010;Bayer & Obenauer 2011;Zimmermann 2011;Biberauer & Sheehan 2011;Biberauer et al. 2014;Struckmeier 2014;Hack 2014;Bayer & Trotzke 2015;Bayer & Struckmeier 2017;Alshamari 2017;Jarrah & Alshamari 2017;Alshamari 2021).
Research in minimalist practice has abstracted away from descriptive, semantic-pragmatic accounts to explaining the syntax of discourse particles. Hence, a discourse particles instantiates a syntactic category on its own spelling out a discourse feature of a certain discourse value it carries at the interface system (Bayer & Trotzke 2015) and heading a functional, discoursal projection of various articulated types of C-features that express information structural values like Focus and Topic (Thumair 1989;Ouhalla 1992Ouhalla , 1994Ouhalla , 1997Ouhalla , 1999Bayer 1996;Miyagawa 2010;Zimmermann 2009;Paul 2009;Coniglio & Zegrean 2010;Bayer & Obenauer 2011;Zimmermann 2011;Biberauer & Sheehan 2011;Biberauer et al. 2014;Bayer & Struckmeier 2017). 1 Having provided fairly reasonable minimalist-based background on discourse particles in a cross-linguistic view, let us move to some theoretical assumptions and practice on the syntax of discourse particles in recent minimalist studies. Consider the following Italian clause containing a discourse particle, from (Coniglio & Zegrean 2010: 15) 2,3 : (1) Tanto il libro non lo leggo PRT the book Neg it I.read 'I won't read the book in any case.' In (1), the discourse particle tanto takes wide scope over the whole clause expressing the proposition, coloring the clause with a value of structural information-expressing the speaker's' attitude towards the content of the proposition with kind of confirmation. This can be represented in (2) below (the notational arrow indicates wide semantic scope of tanto): (2) The discourse particle tanto is argued to be merged in the C head of CP, which in (2) is the syntactic position that hosts (encodes/expresses) extra-sentential information value and discourse packaging confirmation (Biberauer et al. 2014).
With the background alluded to above, in the next sub-section, we highlight some key facts about the discourse particle wara and show, while discussing the relevant data on descriptive basis, how wara interacts with the associate clause and the clause internal syntactic item, the subject DP. 4

Descriptive Facts about the Discourse Particle Wara: Pragmatic Distribution and Syntactic and Morphosyntactic Properties
Consider the sentences in (3). 5 (3) a. wara ʔal-katib j-anʃir qisˤusˤ PRT DEF-writer.M 3SG.M-publish.PRS story.PL 'The writer publishes stories (I'm wondering why!).' b. wara-h ʔal-katib j-anʃir qisˤusˤ PRT-3SG.M DEF-writer.M 3SG.M-publish.PRS story.PL 'The writer, he publishes stories (I'm wondering why he does so!).' c. ʔal-katib wara-h j-anʃir qisˤusˤ DEF-writer.M PRT-3SG.M 3SG.M-publish.PRS story.PL 'The writer, it is him who publishes stories (I'm wondering why he does so!).' In (3a), wara surfaces clause-initially, scoping over the clause and functioning as a syntactic-pragmatic device expressing speaker's mirativity, following Aikenvald (2004), expressing the state of a affairs that the speaker is surprised by the proposition that 'the writer be expected to publish some other sort of work, possibly, novels or books, but not stories'. In (3b), wara also surfaces clause-initially but displays some morphosyntactic properties; wara is suffixed with a clitic agreeing in φ-features with the subject DP that appears to the right of wara. In (3c), wara maintains this morphosyntactic behavior but is preceded by the subject DP. The cases in (3b,c) involve what seems to be φ-agreement (Alshamari 2017) with different positions of the subject DP that wara seems to in an agreement relation with.
We follow the consensus that discourse particles are first merged in CP domain, the topmost syntactic position projected when discourse is activated in the course of the derivation of the relevant sentence. we will therefore obey the assumption that wara is merged at C, hosts the mirativity information and is spelled out as wara at PF interface while at LF interface it signals the presence of a constituent with mirativity value. Following this line of logic, given (3a), we propose that wara is first merged in C and takes wide scope over the clause, marking the whole proposition expressed by the clause with mirativity (we assume that the lexical verb incorporates in T due to the fact that Arabic maintains rich morphology (Ouhalla, 1988(Ouhalla, , 1994(Ouhalla, , 1997). 6 The structure in (3a) can be schematized in (4). We propose that the functional, discoursal projection that wara heads is MrvP 7,8 .
Information of mirativity is expressed on the whole clause due to a feature on wara we term [Mrv], being involved in numeration and processed in the computation of the sentence. This feature is spelled out at the PF-interface as wara while at the LF-interface it conveys the speaker's being mirative towards the proposition (we will turn to the characteristics of the [Mrv] feature and the feature valuation when discussion reaches data involving more articulated structure).
The scenario, though, becomes more challenging when observing that wara has an expressive property associated to it; wara seems to interact with clause internal syntactic items that have nominal categorical nature, rather than propositional, sentential nature, as in (3b), in which wara is attached with an agreement marker h agreeing in φ-features with the subject DP, in what seems to be φ-content spell out operation (Alshamari 2017). Furthermore, we observe that wara, while still hosting the subject-agreeing clitic, is preceded by the subject DP, as in (3c), in which the subject DP appears to the left of wara, indicating that it does not show up in Spec TP since it c-commands wara, which in syntax is explained by means of movement of the subject DP across wara, to the Spec position of a functional, discoursal projection.
As we have seen in (3b), the subject is marked by wara without undergoing any movement. Given this logic, at the interface system, in theory, there is no reason to license movement of the subject DP in (3c), unless there is motivation associated to the discourse interpretive properties of the subject DP itself (Ouhalla 1997;Chomsky 1995Chomsky , 2001Boskovic 2007;Holmberg et al. 2017). Within minimalist considerations, agreement is a strategy that normally overtly exists. In the spine of minimalism, movement is treated as an operation that occurs in syntax for reasons related to expressiveness of language, too, like agreement. 9 However, movement is activated in case overt agreement is not available (Alshamari 2017). That is, movement is a last resort strategy used to satisfy an interface requirement in case no overt agreement is possible (Chomsky 1995(Chomsky , 2001. Using the logic of Miyagawa (2010), movement is activated in syntax in order to leave a record for the LF-interface system that there has been an agreement relation in the derivation of the sentence that could not be overt. With this background, this research will tackle the inquiry stated in the next sub-section.

Research Inquiry
Within formal minimalist assumptions, including economy conditions and interface requirements (Chomsky 1995(Chomsky , 2001Ouhalla 1997;Rizzi 1997;Bošković 2007;Holmberg et al. 2017) and given the syntactic and morphosyntactic facts that wara displays, the emerging inquiry is: (i) what is the LF-interface account for the spell out of the clitic on wara in (3b) and (3c) and, (ii) what motivates and authorizes movement of the subject DP across wara in (3c), given that wara marks the subject DP while in situ in (3b), where no movement takes place?
The rest of the paper will be structured as follows. Section 2 touches on the logic of determining the featural grid of wara, providing evidence from syntax and interface requirements that [Mrv] on wara is valued. Additionally, it highlights on basic principles of the minimalist approach, including mechanisms of Agree and Move (Chomsky 1995(Chomsky , 2001. Section 3 provides a minimalist investigation to the syntax and morphosyntax of wara, reflecting on its interpretive properties as mirativity marker with respect to the clause internal syntactic items marked as mirative. Section 4 shows some key impli-cations the findings of the analyses impose to the minimalist approach, including the assumption that wara marks the subject DP with mirativity while the subject is in situ, while movement of the subject DP is only triggered by the existence of other discourse features, including [CF], [A-Top] and [S-Top] on the subject DP. Section 5 concludes the paper.

VALUE OF [Mrv] ON WARA AND ITS INTERFACE LEGITIMACY
Features play a crucial role in standard minimalist assumptions. A feature is the information carried by a syntactic item, be it lexical, functional or discoursal. Interpretive properties of syntactic items, lexical or functional, overt or null, are characterized by features, like functional, expressiveness or discourse, information structural nature, for instance. Within the computational system and interface system, features such as Case-feature, Q-feature, Wh-feature and φ-features or δ-feature drive syntactic operations, including agreement and movement, which are motivated for interpretive reasons. We dedicate the following subsection to elaborating on features as a central notion in the minimalist practice.

Feature Distinction: Interpretability and Uninterpretability
Features in the minimalist theory come into types: interpretable and uninterpretable.
An interpretable feature enters the derivation with a specified, interface-interpreted value that is determined in Lexicon and whose value contributes to the interpretation of the expression with which it is associated. Thus, if a feature F is interpretable/valued, it is valued in lexicon and enters the derivation with a determined value. An example can be tense feature [T] on T°. This means that the feature [T] is on the right syntactic position where it is fully interpreted and licensed at the interface system (Chomsky 1995(Chomsky , 2001. Another instance of interpretable/valued feature is the occurrence of φ-features on a DP argument, where this is interpreted as the agreement relation third singular person. On the other hand, an uninterpretable/unvalued feature F doesn't have a value set in Lexicon. It doesn't contribute to the interpretation of the expression it is associated with and enters the derivation unvalued. This can be the instance of φ-feature on T°, where it is not interpreted at the interface system. Put differently, the interface system doesn't read uninterpretable/unvalued features. Interpretability/uninterpretability distinction is motivated by requirements of the interface system, PF and LF, imposed on the computational system, syntax. If a feature F is interpretable at a given interface, it contains content that this interface can interpret. Thus, LF-interface interprets features that have semantic import like φ-features and δ-features like Focus. The PF-interface, on the other hand, interprets articulatory related features like stress and contour. 10 We will follow the convention used in Chomsky (2001)  Let us now see how this can explain the nature of the featural grid of wara and their legitimacy at the interfaces. 12

Make up of [Mrv] on Wara
Let us assume that wara has a v- [Mrv] feature, which suffices to license wara at the LF interface as mirativity marker. On the grounds that wara hosts a clitic whose grammatical information expresses φ-content of the subject DP, it is not a huge leap to assume that wara has a set of φ-features, which we maintain are unvalued. On the other hand, the subject DP ʔalkati, by theory, has a set of v- [φ] features and an occurrence of u- [Mrv] feature. This can be summarized in the table below: This account of the characterization of the features on wara and the subject DP is based on conceptual and empirical reasoning. It is held in the cartographic approach (Rizzi 1997(Rizzi , 2006 that an occurrence of a δ-feature like [Foc] on a C-head is a discourse criterion, which triggers movement of a constituent that re-merges at the Spec position of the discourse criterial phrase, like Foc°. This is taken in the minimalist approach to be an instance of unvalued feature, a property which motivates probing and, sometimes movement (Chomsky 2001). We though abstract away from this assumption and argue that [Mrv] on wara is an occurrence of a valued feature, basing our assumption on conceptual (semantic) and empirical (syntactic) evidence as follows. Firstly, wara is merged on C°, which has discoursal properties and is the habitat of δ-feature, which directly suggests that [Mrv] on wara is valued. Secondly, this is directly captured and supported by the nature of [φ] on wara¸ being assumed unvalued because [φ] are not interpretable on functional, discoursal heads (in analogy with the standard assumption that [φ] are unvalued on T° (Chomsky 1995).
Further support of this contention comes from theoretical consequences. That is, if we obey the assumption that [Mrv] on wara is otherwise unvalued, we are at a risk of violating a crucial condition in the minimalist theory. That is, wara would end up with only intrinsically uninterpretable, unvalued features, hence, C° which has wara on it, has no instance of interpretable valued feature, a situation that Chomsky (1995: 349-355), discussing Agr, concluded was unwanted, for theoretical and empirical reasons. In this respect, he argues that a head with only uninterpretable, unvalued features would crash at the interface system. Thus, the epiphenomenal generalization the theory provides now us with is that [Mrv] on wara is valued (we will thus see that wara probes, in the sense of Chomsky (2001), by virtue of the u-[φ] features it has).
Having formalized a reasonable generalization about the features on wara, and the associate arguments, it is now relevant to highlight on the minimalist mechanisms and strategies used in the investigation of the data at hand. This is discussed in the next sub-section.

Theoretical, minimalist mechanisms and strategies
Current standard version of minimalism takes the syntactic operations Agree (agreement) and Move (movement), to be triggered by features (Chomsky 2001). In the computational system, i.e. syntax, syntactic items enter the derivation of a linguistic object (phrase, clause or a sentence) with features associated to them for expressiveness properties, being lexical, functional or discoursal which are, in a formal way, bundles of values to be interpreted at the LF-interface system. As discussed in sub-suction 2.1, there are two types of features, and it is the u-feature that triggers Agree and Move. In the derivation of a sentence, a u-feature needs to delete before the Transfer point of the derivation. To accomplish this, interface system imposes the condition that the u-feature receive a value during the derivation. Consequently, the u-feature becomes an active probe, which searches in the c-command domain for a syntactic item that has a matching but valued instance of the same feature. We list in (5) the conditions regulating Agree Operation, as developed in Chomsky (2001: 122).
(5) A probe α agrees with a goal ß provided that: a. α has an unvalued feature. b. ß has a matching valued feature. c. ß is active by virtue of having an unvalued feature. d. α c-commands ß. e. There is no potential goal γ intervening between α and ß. f. There is no phase-boundary between α and ß. In case the matching goal is not visible, meaning that it is in a lower distant phase with respect to the probe, as (f) claims, or there is an intervening syntactic item that can serve as a potential goal, as (e) claims, the goal moves up in structure high enough to become visible. 13 The goal also moves when it has an occurrence of u-feature on it. It moves to escape its phase transfer and to find a potential probe to value its u-feature.
Having characterized an account of the featural grid on wara and highlighted the minimalist techniques activated in computation, let us now examine the derivation of the clauses involving wara, including Agree and Move operations held therein.

ANALYSING WARA: MINIMALIST THEORY OF AGREEMENT
We arrive at this point of the argument to the inquiry as to how to legitimize the sentences in (3) above.

Wara and the Clitic
With the grip on the theoretical assumptions at hand, let us start with (3b), repeated below as (6), where wara surfaces clause-initially and is suffixed with a clitic agreeing in φ-features with the right-adjacent subject DP.
(6) wara-h ʔal-katib j-anʃir qisˤusˤ PRT-3SG.M DEF-writer.M 3SG.M-publish.PRS story.PL 'The writer, he publishes stories (I'm wondering why he does so!).' Activating Agree in (5), (6) shows the point of the derivation at which wara, having a set of u-[φ] features, oper-ates as a probe and searches in its c-command domain for a syntactic item that has matching instance of v- [φ] features to agree with. The subject DP in this case operates as a goal that, additionally, seeks valuation for its u- [Mrv]. This is schematized in (7): In standard formalisation of Agree (Chomsky 2001

), wara u-[φ]-probes the subject DP. v-[φ] on the subject DP values the u-[φ] on wara and, simultaneously, v-[Mrv]
on wara values u-[Mrv] on the subject DP. As a consequence of this Agree, a PF-chain is created, composed of wara cliticised by the φ-content of the subject DP and c-commanding it and the subject DP. This PF-chain then functions as a detecting device to determine that mirativity is marked on the subject DP, by the use of the clitic. This conclusion suggests that the clitic is a PF-product derived in computation and PF-interface to allow LF-interface to read what wara has marked in computation as mirative.
At the end of the derivation, all the instances of u-features (u-[Mrv] on the subject DP and u-[φ] on wara) have now been valued and deleted at the Transfer point of the derivation and only the instances of v-features (v-[Mrv] on wara and v-[φ] on the subject DP) survive and are read, interpreted and legitimised at the interface system. It follows from this that deleting u-[Mrv] on the subject DP doesn't mean the subject DP is not involved in the interpretation of mirativity. Rather, mirativity interpretation is delivered by the chain which accommodates all the instances of features. We elaborate on this logic in the next sub-sub-section.

u-Feature-cooperation: Expressiveness Characteristic of Wara
The process we introduce as u-feature-maintenance revolves around the assumption that an instance of u-feature on a syntactic item, which gets deleted in syntax, is maintained at the LF-interface system by means of a PF-interface system operation. Witness the schemata in (8) below. (

8) [wara v-[Mrv] u-[φ] [Subject DP u-[Mrv] v-[φ]]].
We see in (8) that the remaining, surviving features are those with valued nature-v- [Mrv] on wara and v-[φ] on the subject DP. What we are concerned with here is u-[Mrv] on the subject DP, which is deleted while the subject DP needs to link up with an instance of a syntactic item carrying mirativity interpretation. What we propose takes place within the strategy of u-feature-cooperation is that a deleted feature can be maintained for interpretive, expressiveness purposes needed at the LF-interface system. This can be done by activating PF operations, creating PF-items like clitics or stress, for instance. In our case, the scenario is as follows. v- [Mrv] on wara values u- [Mrv] on the subject DP and the latter deletes, but identity of the subject DP as mirative at the LF-interface system is maintained. This occurs because, in turn, v-[φ] on the subject DP values u-[φ] on wara. This, in particular, makes a desired consequence. As we have seen, though u- [φ] on wara is deleted, it has PF content, the clitic. This clitic serves as a means of cooperation between wara and the subject DP, or, put differently, between the features on them. This clitic, being a by-product of a u-future on wara, turns the subject DP detected and recorded as mirative at LF-interface system. In other words, the instances of u-features, u-[Mrv] on the subject DP and u-[φ] on wara, cooperate at the LF-interface and maintain mirativity interpretation on the subject DP.
We now move to the last issue this paper tackles, movement of the subject DP across wara. But before we do, let us elaborate more on the logic of marking the subject DP in situ. In this regard, witness what happens when the subject DP marked as mirative expresses third plural feminine, as in (9).
(9) wara-hin ʔal-katib-at j-anʃir-i qisˤusˤ PRT-3PL.F DEF-writer-3PL.F 3PL.F-publish. PRS-3P.F story.PL 'The writers, they publish stories (I'm wondering why they do so!).' Recapitulating the argument in section 2, LF-interface doesn't read φ-content on a functional head, since it makes no sense, following Miyagawa (2010). The sentence in (9) shows that the PF operation of spelling out of the value of φ-content of a DP on a functional head, like wara, depends on the value of the content of [φ] of the syntactic item wara marks. The consequence is that the relevant syntactic item is read in situ but need not move to hold a Spec Head configuration to agreement with wara but rather is marked in situ and is identified on wara by the clitic. This assures that linguistic operations in the derivation are subject to economy conditions (Chomsky 2001(Chomsky , 2008(Chomsky , 2013, involving a minimal number of syntactic operations (Collins 2001). Under this view, the PF-interface interprets (spells out) this φ-content (the clitic) as a record for LF-interface to read the subject DP as the mirativity without activating movement, which is only motivated for additional interpretive reasons, as will be investigated and discussed in the following sub-section.

Movement of the Subject DP Across Wara to the Left Periphery
Consider (3c) above, repeated below as (10): (10) ʔal-katib wara-h j-anʃir qisˤusˤ DEF-writer.M PRT-3SG.M 3SG.M-publish.PRS story.PL 'The writer, it is him who publishes stories (I'm wondering why he does so!).' Recall that we follow the theory that in Arabic the lexical verb re-merged at T° (Ouhalla 1988(Ouhalla , 2014(Ouhalla , 1997. Holding that wara is located to the left of the subject DP in (3a) and (3b), it follows that the subject DP moves as high as Spec TP but never to the left periphery across wara. In the spirit of the minimalist practice, movement is triggered by discourse interpretive reasons (Chomsky 2001). This makes the scenario in (10) quite challenging because it involves apparent movement of the subject DP, which needs a good theory to explain, since the subject DP doesn't need to move there given that CAN grammar allows marking the subject DP by wara while it is in situ.
Having concluded that the subject DP in marked with mirativity without movement across wara to Spec MrvP, we expect that movement of the subject in (10) is not triggered by mirativity interpretation. We therefore follow the logic of Miyagawa (2010) who argues that movement is conceptually conceived of as a functional relation established between two syntactic items in the structure. For Miyagawa (2010), in absence of overt agreement movement (i.e Spec-Head relation of agreement) is required by the computational system to be held in syntax as a record for the interfaces (for semantic and information-structure interpretation) that there has been a functional relation taking place in computation.
Following this line of logic, in (10), and all the cases at hand in fact, the subject was merged at Spec TP, whence it values u-[φ] on wara, in case wara u-[φ] the subject DP. The subject DP is then expected to have moved further because it has a u-feature on it. We therefore propose that the subject DP has a focus feature [Foc], interpreting the subject DP as an entity expressing new, non-presupposed information in the sense of Holmberg & Nikanne (2002). This [Foc] on the subject DP is the power that triggers movement of the subject DP and this movement is to Spec FocP in the left periphery of the clause, motivated because the head Foc° of FocP is null at the PF-interface, in the manner of Ouhalla (1997). We represent this in (11) below. 14 (11) Consider now (12) below, which has more moved structure.
(12) ʔal-katib qisˤusˤ wara-h janʃir DEF-writer.M story.PL PRT-3SG.M 3SG.M-publish.PRS 'The writer, it is stories that he publishes (I'm wondering why he does so!).' Notice that the object DP qisˤusˤ intervenes between the subject DP and the cliticised wara. Merger of the object DP qisˤusˤ, which functions as focus here, provides syntactic empirical evidence that the subject DP is distant from wara, at which it functions as a topic in the sense of Reinhart (1981), hence, the conclusion that the subject DP is not re-merged at the Spec MrvP but further above in the structure at the Spec of TopP. On derivational groundings, again, the syntax of (12) evidences the fact that the subject DP has already entered Agree with wara at an earlier step of the derivation, as (13) illustrates.
Following Ouhalla (1997), in his analysis of focus particles in Arabic, the generalization we can formulate for (12) is that the object DP qisˤusˤ moves to the Spec of the Foc°, whose head CF° is not morphologically realized, i.e. null. This behaviour is supported by principles of phase theory (Chomsky 2001), the assumption that a syntactic item moves out of its phase only if it has an unvalued (Bošković 2007;Holmberg et al. 2017). We therefore propose that the subject DP has also moved out of its phase, to Spec TopP to value the u-[Top] on it.
The same conclusion holds true in cases involving more articulated structure in the left periphery, providing evidence that the subject showing up across wara is re-merged at the Spec position of a discourse, information structural projection higher than MrvP. Consider (14).
(14) ʔal-katib zad wara-h j-anʃir qisˤusˤ DEF-writer S-Top PRT-3SG.M 3SG.M-publish. PRS story.PL 'As for the writer, he publishes stories (I'm wondering why he does so?).' Zad is a Shifting topic marker in the sense of Frascarelli & Hinterhölzl (2007), marking a topic constituent that is interpreted as being shifted to in the ongoing discourse. In (13), the subject DP maintains shifting topic interpretation via movement in syntax to the Spec position of S-TopP, headed by the marker zad, holding a Spec Head relation of agreement with zad. With the clitic on wara, and within derivational considerations, we propose that the subject DP had been into Agree relation with wara prior its movement, the point at which hit had the interpretation mirativity, and then to re-merged at Spec S-TopP for shifting topic interpretation. (14) is represented in (15) below.
The theory we have developed here, then, is that the subject DP is marked by wara at Spec TP, where it is visible to wara in syntax. In cases where the subject DP shows up across wara, the subject DP is at the Spec position of a functional projection, assigned a discourse value beyond mirativity, depending on the value the head has, for instance, [Foc], [Top] or [S-Top].
Having formulated the generalization that when wara is cliticised and marks the subject DP, it follows that in cases of bare, non-cliticised wara, in (3a) above repeated blow as (16), it is the whole proposition expressed by the TP that is marked with mirativity.
(17) ʔal-katib zad wara j-anʃir qisˤusˤ DEF-writer S-Top PRT 3SG.M-publish.PRS story. PL 'As for the writer, he publishes stories (I'm wondering why?).' Following our theory so far, derivationally, we expect the subject DP to value u-[φ] on wara. We though see that wara is not cliticised, indicating that it was not into an Agree relation with the subject DP. What is more, the subject DP has moved to Spec S-TopP, being now interpreted as S-Topic. If wara had u-[φ]-probed the subject DP, the theory expects wara to be cliticised. Being bare, then, it follows that what wara probes the TP in syntax, being the whole proposition at the pragmatic interface, which is explained because TP does not have φ-content inherently.

THEORETICAL IMPLICATIONS OF THE RESEARCH
The generalizations formulated and the analyses made above lay the groundwork for showing that CNA marks mirativity in a minimalist, economical way, in which the discourse particle wara and marks the subject DP in situ, without recourse to movement. The output of this research also adds that natural language can achieve kinds of discourse marking via the morphologically realizing discourse features, like [Mrv], spelling it out as the discourse particle, like wara, which functions as a discourse detector signaling the presence of discourse marked syntactic item. In this regard, one insight the syntax of wara provides us with is the fact that wara is an expressiveness-device, which has φ-content that can be spelled out to detect the mirative syntactic item in narrow syntax (in case this syntactic item has φ-content), a process which results in banning movement of the relevant mirative syntactic item to the left periphery. This characteristic of wara indicates that discourse particles seem to have developed the property of agreement rather than being functional heads that trigger movement of some kind syntactic item that they agree with/mark. This directly leads us to the idea that discourse particles can be good evidence to the more intricate view of the minimalist practice, in particular, the economy property of human language, being optimal, in that movement is only triggered if there is no way to make agreement overt in syntax. What this research significantly raises is that a DP, once valued an instance of u-feature on a head, it must still move as long as it has a u-feature on it, suggesting that movement is restricted to and triggered by the existence of unvalued features.

CONCLUSION
This paper sets an investigation to the syntactic operation of morphologically marking the information structural notion mirativity in CNA by means of the discourse particle wara, using minimalist mechanism and strategies (Chomsky 2001). Analyses advanced show that mirativity marking is carried out in a minimalist, economical manner, where a constituent is marked with mirativity without moving to the Spec position of MrvP headed by the mirativity marker wara, the locus of mirativity information. Discourse properties of wara have led to the assumption that the discourse feature [Mrv] on wara is valued. The expressiveness property of wara, i.e. cliticisation, adds up a set of [φ], which by theory are unvalued, via which wara operates a probe, φ-probing the subject DP. Agree between wara and the subject DP results in deriving a PF-chain, in which the subject DP is detected at the LF interface as the constituent marked with and expressing mirativity. With examinations to structures with more articulated discourse heads and involving movement of the subject DP across wara to the left periphery, it is shown that such movement is not triggered by [Mrv] on wara but by other discourse features, including [CF] and [S-Top]. Hence, wara serves as economical discourse device that mars and detects mirative constituents in syntax.  (2015) for the argument that a discourse particle instantiates the projection PrtP, standing for Particle Phrase. 2. Interlinear glossing for all data in this paper is in accordance with Leipzig Glossing Rules available at https:// www.eva.mpg.de/lingua/pdf/Glossing-Rules. 3. Data from other sources are stated as they are in the original work, with respect to glossing. 4. Agreement content like third person singular and pragmatic interpretive values like confirmation and mirativity are notated in italicised font throughout the paper. 5. All occurrences of wara, and other instances of discourse particles in cross-linguistic contexts, are in bold font. 6. We use the convention of italicising the term mirative when it refers to the information structural value. 7. We assume that wara instantiates a functional projection with the information mirativity in the left periphery, labelled as MrvP, where wara spells out the feature [Mrv] on the head Mrv°. 8. The notation '°' refers to the head category. Thus, T head of TP projection is notated as T° and C head of CP projection is notated as C°. 9. There is indeed LF movement but we are dealing with overt, syntax movement here. 10. We follow Chomsky (2001) view on the characterisation of features. Thus, interpretable features are lexically valued while uninterpretable features are lexically unvalued and therefore need to receive a value in syntax. 11. Throughout the paper, interpretable/valued feature is notated as u-F while uninterpretable/unvalued feature is notated as v-F. 12. Q = question feature, Wh = Wh-feature, φ = agreement features, δ = discourse feature. 13. We will not explicate phase theory here, for reasons related to space. We therefore refer the reader to Citko (2014). 14. The convention < x > indicates the extraction position of moved x.