The syntax of negative imperatives in Yemeni Arabic: A phase approach

Abstract

Negative imperatives in general have received less research compared to positive imperatives cross-linguistically, which is perhaps due to their complex nature. This article studies Yemeni Arabic (YA) negative imperatives, probing their syntactic and discourse properties, and proposing a syntax–discourse interface amalgamated with phase approach to the analysis of these structures. Syntactically, negative imperative structures in YA are bipartite in nature, consisting of the preverbal negative particle lā and the postverbal negative suffix – š, thus paralleling negative imperatives in French, for instance. We propose that the head Neg is split into NegP and NegClP as a result of Agree and interface properties between syntax and discourse. NegP is represented by lā and –š represents NegClP (=Clitic Negative Phrase). NegClP ensues from the clitic nature of –š. We also argue that the thematic subject of imperatives is a 2 (person) pro. Discoursally, we argue that the overt (pro)nominal constituent showing up preverbally is not a subject, but rather a C-domain element, precisely a topic. The latter merges in Spec,TopP and pro in Spec,vP, coreferentially correlated at the syntax–discourse interface, where coreferentiality takes the form of Agree as Match, which results in full interpretation of pro. The proposal accounts straightforwardly and elegantly for the bipartite negative imperative facts in YA, and other modern Arabic varieties. It has also adequately been applied to English and French, and hopefully, it can be applied cross-linguistically.

Keywords:

• negative imperatives
• bipartite negation
• Yemeni Arabic
• French
• Neg splitting
• syntax-discourse interface
• coreferentiality
• cartography
• Agree as Match
• phase theory
• feature inheritance

1. Introduction

Negation in natural languages is a universal phenomenon, and all scholars seem unanimous about this (see e.g., Aikhenvald, 2010; Isac, 2015; Rupp, 2003; Wiltschko, 2021; Zeijlstra, 2007). However, every language expresses negation by using a particle, word, or even a particular structure consisting of more than one specific word (see e.g., Isac, 2015; Zeijlstra, 2007). In this article, we will study negative imperatives in Yemeni Arabic (henceforth, YA), and see how negation in this structure is formed. In fact, imperatives in general have been extensively studied across languages, and various approaches have emerged (see e.g., Alcázar & Saltarelli, 2014; Bennis, 2006; Beukema & Coopmans, 1989; Downing, 1969; Han, 1998; Isac, 2015; Jensen, 2003; Kaufmann, 2012; Rupp, 2003; Shormani, 2017a, 2021; Thorne, 1966; van der Wurff, 2007; Zanuttini, 1991). However, negative imperatives are less investigated cross-linguistically, compared to positive imperatives, which is perhaps due to their complex nature (see e.g., Alcázar & Saltarelli, 2014; Isac, 2015; Kaufmann, 2012; Shormani, 2017a, 2021).

Thus, this article investigates negative imperatives in YA, and attempts to provide a novel approach to the analysis of these structures, based on their syntactic and discourse properties and functions. It involves data from YA. It will be limited to negative imperative structures of the type in (1).¹

The examples in (1) are ordered from simple, as in (1a) to the most complex, as in (1f). The structural simplicity of (1a) lies in involving simply the negative particle lā “not”, the verb ti-txabar- š “talk”. The most complex structure in (1) is (1f). In this structure, for instance, there is an NP occurring preverbally, the negative particle lā, the verb ti-ktub-ayn, the object clitic infix -uh “it” and the postverbal negative suffix – š. As the examples in (1) show, negation in YA is a bipartite phenomenon (we return to this issue in section 2).

The example in (1a) is said to be the “normal” structure of negative imperatives virtually cross-linguistically, where the subject is null/covert. In other words, imperative constructions, be they negative or positive, are deemed to be “subjectless” structures, invariably across languages (see e.g., Alcázar & Saltarelli, 2014; Bennis, 2006; Beukema & Coopmans, 1989; Downing, 1969; Han, 1998; Jensen, 2003; Kaufmann, 2012; Radford, 2009; Rupp, 2003; Shormani, 2021; van der Wurff, 2007). Thus, if imperatives are subjectless invariably across languages, then the subject of imperative could be taken as a null/covert pronoun. We take this null/covert pronoun to be pro. Our proposal is supported theoretically and empirically. Theoretically, since T enters the derivation with an unvalued EPP feature, the subject here pro is expected to value the unvalued EPP feature of T, or otherwise the derivation will crash. The idea that T in negative imperatives has an EPP feature is minimalist in nature, simply because it makes dealing with imperatives like dealing with other clause-types such as declarative, interrogative, etc., which gives rise to non-construction-specific postulations (cf. Jensen, 2003; Shormani, 2017a, 2021). Empirically, strong evidence ensues from the syntactic and semantic “activeness” of pro, which is, in fact, a property not any empty category has it (see e.g., Beukema & Coopmans, 1989) (we return to this issue in section 3).

If the subject of negative imperatives is pro the question, then, is: what are the syntactic categories and functions of the NPs: ʔantah “you”, ʕali “Ali”, ʕyāl “boys” and banāt “girls” in (1a-f), respectively? We propose that the NP occurring preverbally is a topic, a left periphery constituent and functions as the reference/antecedent of pro. The former is a C-domain element, while the latter is a T-domain element. They are linked at the syntax–discourse interface, via coreferentiality as a mechanism of linking both constituents. Coreferentiality takes place as Agree as Match (cf. Frascarelli, 2007; Pesetsky & Esther, 2007; Shormani, 2017b). (we return to this point in section 5).

It turns out that discourse has a role to play in imperative constructions. From a discourse perspective, and given the conversational nature of imperatives (see e.g., Wiltschko, 2021), it is expected that imperative structures have interpretive import, which is manifested via performative functions (say, performing speech acts) imperatives perform (see e.g., Aikhenvald, 2010; Aloni, 2007; Bianchi & Frascarelli, 2010; Han, 1998, 2001; Krifka, 2001, 2008; Lambrecht, 1996; Portner, 2007; Shormani, 2017a, 2021; Zhang, 1990) (we return to this point in section 4).

The rest of this article goes as follows: section 2 discusses the nature of negation in YA, including negative imperatives. Section 3 lays out our proposal from a syntactic perspective. Section 4 presents the proposal from a discourse point of view. Section 5 discusses the syntax–discourse interface in negative imperatives. Section 6 concludes the paper and provides further implications.

2. The nature of negation in YA

Central to negation in YA is the fact that it is a bipartite phenomenon, having the structure “lā/mā … -š” consisting of the negative particle lā/mā which occurs preverbally and -š, which occurs postverbally. This is exemplified in (2) below:

In (2a), the preverbal negative particle used in a declarative construction is mā along with the postverbal negative suffix -š. In imperative constructions, however, the preverbal negative particle is lā used also along with the postverbal suffix -š.² Thus, since lā is a word, in the sense that it is neither prefixed, nor suffixed to the verb, and since -š is suffix, we will gloss the former as “not”, and the latter as “NEG” throughout.

A word on the morphology of the verb in imperatives is in order here. Apart from the negative particle -š, in YA the verb in imperative has an imperative morphology. However, this verb morphology differs in both positive and negative imperatives. In positive imperative, the prefix ʔi- is used, while in negative imperatives the suffix ti-. The two verb morphology affixes are exemplified in (3a) and (3b), respectively.

In these examples the ʔi- can be taken as a 2 person positive imperative prefix, prefixed to, say, the root ʕ-m-l “do”. However, in negative imperatives, ti- could be said to substitute ʔi- in negative imperatives. In this way, Arabic seems to resemble languages like Italian, in which the verb has imperative morphology (see e.g., Isac, 2015, p. 76).³

2.2. Grammaticalization of the negative particle -š

Considering again the examples discussed in the section above, there arise several questions: i) what is -š?, ii) what is -š’s syntactic category/status? and iii) where does it come from, or otherwise what is its origin? In what follows, we will try to answer these questions.

Some Arabic scholars maintain that -š originates from the word “šay” “thing”, which is found in Classical Arabic (see Lucas, 2010, 2015). Lucas (2010), for instance, observes that “-š alone can be said to have completed its journey from independent lexical item to fully-fledged, unmarked negator” (p.166). What Lucas means by this is that -š has undergone a grammaticalization process. The idea that -š is a grammaticalized morpheme from the word “šay” “thing” is evidenced form Classical Arabic, namely, the Holy Qur’an. Consider the following examples (slightly modified from Lucas, 2010, p. 180):

It is a fact that the word šay “thing” is used in (4) and (5) as an adverb, but with a negative scope, i.e. as “a negative polarity adverb” (Lucas, 2010, p. 181). Note also that both occurrences of šay accompany the negative particle lā. Thus, it is clear that the ancestor of the enclitic -š is šay. That is to say šay has been grammaticalized into -š, having a negative sense or scope. This negative scope of šay has retained its status in -š even though it has undergone a grammaticalization process.

The negative scope the noun šay has is also evident in YA, consider the following examples:

Although šay in (6a) and ʔay šay “anything” in (6b) have the same negative meaning/scope, there is still some sort of difference, which is that ʔay šay expresses more emphasis than šay. The discourse of both (6a) and (6b) could be that of an exam situation. The two interlocutors could be a student and his classmate, where the first student asks the second about the difficulty of the exam, or whether the second student has written well in the exam. The second student wants to say that the exam was so difficult that he/she didn’t write anything.⁴ The meaning of ʔay šay in (6b), and šay in (4) and (5) in the Classical Arabic examples could be “at all” (cf. Lucas, 2010, p. 184), which has a negative scope, as we have just pointed out.

The negative structure with -š equivalent to those in (6, a & b) above is given in (6c). In this very example, -š, as a negative particle, is not used alone, but rather in combination with other negative particles such as mā, lā etc., hence forming a bipartite negative structure in the form of “mā/lā … . -š”, which is similar to the negative structure “ne … pas” in French. Though pas in French is a word in the sense it is not an affix, there is also some sort of similarity between -š and pas. It is also held that the French pas has undergone a grammaticalization process from the noun pas “step” (see e.g., Hansen, 2014, p. 188; Lucas, 2015, p. 79). This grammaticalization or loss and renewal has been referred to as “Jespersen’s cycle” (Hansen, 2014; Lucas, 2015).

Thus, there is a parallelism between negation in YA and French that both languages have a bipartite negative structure (see also Lucas, 2010, 2015). However, the difference between them is that while in French the preverbal negative particle ne is optional, lā in YA is not. That is, lā must be present as will be clear in subsequent sections.

2.3. Negative particles in standard and colloquial Arabic varieties

This section briefly spells out the difference between standard Arabic (SA) and colloquial varieties such as Egyptian Arabic (EA), Jordanian Arabic (JA), Platonian Arabic (PA), Moroccan Arabic (MA), examining the types of negative particles used in each variety. Consider the SA examples in (7).

As is clear from the above examples, one can postulate that negative imperative in SA has the structure “lā + verb” with nothing else.⁵ If lā is deleted, the resultant structure is ungrammatical as in (7d).

In modern Arabic varieties, on the other hand, the dominant negative structure is “lā/mā … -š”, as we have pointed out earlier regarding YA. Thus, - š, as a negative particle (and in all clause types), is used in almost all modern varieties of Arabic.⁶

2.4. Deletion of lā/mā

Now, we are in a position to examine a very important aspect in modern Arabic varieties including those discussed in this article. This aspect is the deletion of the first negative particle, viz. lā/mā. Consider the examples in (9).⁷

The examples (9a) through (9d) show that the negative particle lā/mā can be deleted and negation can be expressed by – š in imperatives in these four dialects.⁸ However, in YA the deletion of the lā is not possible as shown by the ungrammaticality of (9e).⁹

Note that while the first negative particle lā/mā can be deleted in some Arabic varieties, as we have just seen, negation can be expressed by the negative particle/clitic – š per se. Note also that – š cannot be deleted at all, be it in YA, or any other Arabic dialect. All the dialects mentioned above, except YA, are similar to French in the deletion of the first negative particle ne. Likewise, these Arabic dialects are also similar to French in the fact that –š/pas cannot be deleted.

3. Syntactic properties of negative imperatives

In this section, we will discuss the syntactic properties of negative imperatives. In section 3.1, we will focus on the subject of negative imperatives. Section 3.2 discusses tense and T’s features in imperatives, and Section 3.3. spells out the position of NegP in imperatives.

3.1. The subject of negative imperatives

Apart from negative or positive imperatives, there are several proposals in the literature that imperatives are “subjectless” constructions (see e.g., Alcázar & Saltarelli, 2014; Bennis, 2006; Beukema & Coopmans, 1989; Downing, 1969; Han, 1998; Jensen, 2003; Kaufmann, 2012; Radford, 2009; Rupp, 2003; van der Wurff, 2007). However, the fact that imperatives are a close-type like declarative, interrogative, etc. and from a minimalist perspective, imperative should be treated in the same way as other clause types, imperative should have a subject of some sort to satisfy the EPP feature of T. The idea that T has an EPP feature is minimalist in nature, simply because it makes dealing with imperatives like dealing with other clause-type structures, which gives rise to nonconstruction-specific postulations (cf. Jensen, 2003). Thus, if imperatives are subjectless invariably across languages, as we have just pointed out, then the subject of imperatives could be taken as pro.¹⁰

A recent study that has advocated the idea that pro is the subject of imperatives is conducted by Shormani (2021). Shormani bases his proposal on several proposals in the literature that the subject of imperatives is pro (see e.g., Bennis, 2006; Beukema & Coopmans, 1989; Jensen, 2003; Rowlett, 2014; Rupp, 2003).

As far as this study is concerned, there is good evidence that pro is the subject of negative imperatives. The evidence consists in the and semantic syntactic activeness pro exhibits in negative imperatives. Consider the following examples (cf. Shormani, 2021, p. 23).

Thus, the syntactic and semantic activeness of pro is clear in (10), where it binds different syntactic categories. As the examples in (10) show, pro controls the anaphora in (10a), the reciprocal in (10b), the internal anaphora in (10c), and the secondary predicate in (10d). we also have evidence from English which supports the Arabic examples in (11) that pro is syntactically and semantically active. Though English is not a pro-drop language, pro’s semantic and syntactic activeness is reflected in examples like (11) (from Bennis, 2006, p. 108).

In (11) examples, pro controls the reciprocal in (11a) and the anaphora in (11b).

Along the above discussion, another support of the fact that the subject of imperative is pro comes from structures such as:

(12a) and (12b) are grammatical. In (12a), for instance, pro is the subject of imperative and no preverbal NP occurs. In (12b), the subject is the pronoun ʔantah which bans pro to occur. The fact that (12a) and (12b) are grammatical means that there is nothing violating any principle of grammar, or of forming a well-formed negative imperative. However, the grammaticality of (13a) indicates that the preverbal pronoun ʔantah is a topic and not a subject, which is not the case in (13b). In fact, the ungrammaticality of (13b) has several things to say: i) the preverbal NP, here the pronoun ʔantah is not a subject but rather a topic and ii) there is a strong adjacency between lā and the verb.

As for the licensing of pro in imperatives, we propose that the functional head, licensing and determining the interpretation of pro, is Top^o in the C-domain¹¹ (we return to this issue in section 5).

3.2. Negative imperatives, tense and T’s features

In the literature, there are several proposals on tense availability in imperative structures (see e.g., Isac, 2015; Jensen, 2003; Shormani, 2021; Zanuttini, 1997).¹² Even those who claim that imperatives have no tense (see e.g., Platzack & Rosengren, 1998) find it difficult to account for the occurrence of temporal adverbs such as now and tomorrow. That imperatives lack tense was also questioned by some linguists who argue that lacking T in imperatives posed a problem of the compatibility of temporal adverbs (see also Isac, 2015, p. 51, for a discussion supporting this view).¹³ As the Minimalist theoretical assumptions adopted here, i.e. phase theory, etc., it seems difficult to postulate that only imperative clauses do not have tense, which certainly leads to construction-specific violation.

We propose that negative imperative has tense, hence TP projection. Our proposal that negative imperative has tense ensues from the use of temporal adverbs such as “now”, “latter” and “tomorrow”. Consider the examples in (14).

As the examples in (14) show, the tense is either present, as in (14a) or future, as in (14b), but not past as the ungrammaticality of (14c) shows. There are scholars (see e.g., Zanuttini, 1996) who advocate that there is strong relation between negation and Tense, and this relation is manifested in languages like Italian.¹⁴

If negative imperatives have tense, hence TP projection, the question, then, is: what are the features that T has? To understand what features T is endowed with in negative imperatives, let’s consider the following examples:

What the facts in (15a-d) show gives us enough room to postulate that T has number, person and gender, hence φ-features, and thus T in negative imperatives seems to be φ-complete (cf. Chomsky, 2000, 2001; Shormani, 2021). However, it should be emphasized that the person feature of T in imperatives is always 2 person, unlike other clause-types which can be 1, 2 or 3. In addition, since T is φ-complete, it should also have Case, a nominative Case in particular to value pro’s unvalued Case feature. Since negative imperatives is a clause-type, T is expected to have an EPP feature, which is valued via V-to-T movement (cf. Alexiadou & Anagnostopoulou, 1998; Shormani, 2021).

3.3. The position of NegP

Given our postulation so far that negative imperatives in YA have the negative structure “lā …-š”, there should be two negative projections: one hosting the preverbal lā, the other hosting the postverbal -š. So, the question is: where are these two NegPs positioned in the hierarchical clausal structure?

In fact, the position of the positioning of NegP in nonnegative constructions has been investigated, and it raises a hot debate and controversy since Pollock (1989) and Belletti (1990). For example, Pollock (1989, p. 397) proposes that NegP is positioned between TP and (AgrP) VP (given the generative assumptions at that time). In addition, Zanuttini (1991) proposes that there is a functional projection, namely NegP which encodes negation, whose head, i.e. Neg^o, takes TP as a complement. This means that NegP is above TP. As far as Arabic is concerned, Benmamoun (2000, p. 112) proposes that NegP is generated between TP and VP, which means that NegP is below TP, and takes VP as a complement.¹⁵ In addition, Shormani (2015) holds that NegP is above TP. Though these proposals may be adopted, the fact that they concern SA, in which negation is not bipartite, makes adopting them problematic.

As far as negative imperatives are concerned, there are several proposals. We will only discuss two of these proposals, namely Zanuttini (1991) and Rowlett (2014). The former proposes that NegP has two positions. She proposes two NegPs: NegP1 is situated above TP and NegP2 below TP. Rowlett (2014), furthermore, proposes that in French negative imperatives, which are to some extent similar to those in YA, NegP is projected below IrrP. However, Rowlett claims that ne is a head merging in Neg^o and pas merges in a projection lower within VP and then raises to Spec,NegP, which cannot be applied to negative imperatives in YA.¹⁶ Thus, it seems that Zanuttini’s (1991) proposal could be adopted in our analysis, as we will see in the section to follow.

3.4. The proposal

Thus, given our conclusions in section 2 that negation in YA is bipartite, having the structure “lā … . -š”, and given examples of negative imperatives like (16) below:

we will adopt Zanuttini’s (1991) proposal in our analysis. Zanuttini’s (1991) proposal is given below in (17) (see also Isac, 2015, p. 19):

Zanuttini (1991) proposes the negation projection in (17) to account for negative imperatives in Romance languages. As is clear in (17), Zanuttini proposes that the structure of negation in imperatives has two NegPs, which are:

realized in two distinct structural positions: above and below the functional projection TP, the syntactic realization of tense. Italian non, for example, is the head of a functional projection labeled NegP-1, which selects TP as its complement, while French pas is the specifier of a functional projection labeled NegP-2 which occurs lower than TP in the structure (p. viii).

Zanuttini (1991) holds that NegP1 hosts preverbal negative particles, whereas postverbal negative particles are hosted by NegP2, adding that the former are heads and the latter are phrases.^17,18

Thus, we propose two NegPs, the first of which is above TP, and the second below TP. However, given the fact that -š is a negative clitic, we propose that the second NegP takes the form NegClP. Our proposal is schematized in (18) as the negative imperative structure in YA.

We consider the negative lā a head in negative imperatives, merging in Neg^o. NegP selects a TP as a complement. NegClP, however, is projected as a complement of T, as a consequence of the fact that -š is both a negative particle and a negative clitic. Finally, NegClP selects a vP/VP as its complement due to the nature of the verb in YA negative imperatives. As for the derivation process, it will be clear when we apply our proposal to deriving some examples presented so far.

3.5. Why should there be two NegP projections?

The question now is: why should there be two NegP projections? It is true that Zanuttini’s (1991) analysis was pre-minimalist, i.e. basically a P&P analysis, but since then it has been very influential in accounting for bipartite negation. However, under minimalist assumptions, it is possible to account for the two NegP projections in terms of Neg splitting proposed by Poletto’s (2008) and advocated in de Clercq’s (2018) and Tilleson’s (2019) works. A very important remark to start with is that in a negative structure, be it imperative or else, the negation is expressed in a projection NegP, given that negation is a feature, and this feature has to be projected (cf. Chomsky, 2005, 2008; Shormani, 2017b). However, with bipartite negation, there are two negative particles, viz. lā and -š in YA, for instance. Thus, building on Poletto’s (2008), de Clercq’s (2018) and Tilleson’s (2019) works, we argue that the two NegP projections are a result of Agree, which splits NegP into two NegPs, namely NegP1 and NegP2.

Furthermore, in YA negative imperatives, there seems to be a role played by interface between syntax and discourse. This role is manifested by the person feature, i.e. 2 person of the addressee who is supposed to perform the action expressed by the negative verb. Bearing this in mind, we argue that the role of the 2 person feature, along with Agree (as Match), results in splitting of Neg into two NegP projections (cf. also Zeijlstra, 2007). The Agree resulting in the splitting of Neg can be thought of as Match (manifested as coreferentiality) between the null head Top, which has a scope over Neg (cf. 18) and -š, but not between Top and lā, given the fact that while lā can be deleted in some Arabic verities like EA, JA, PA, etc., -š cannot.¹⁹

Given that TopP is the topmost projection, TopP has a scope over NegP (and the entire projection), which in turn suggests that the interface between syntax and discourse comes to play only when Agree is established between Top, Neg, T, and pro. We also assume with Shormani (2021) that Top’s features as well as those of Neg are semantically uninterpretable, but they are syntactically valued (see also Pesetsky & Esther, 2007). The overall interpretation of Top, Neg and pro comes to play only when TopP is merged, that is, when the whole imperative structure is completely derived, and sent to Spell out operation (as will be clear in section 4). The Agree can take the form in (19).

The result of this Agree operation is the splitting of the head Neg into NegP and NegClP, as far as YA and other modern Arabic varieties are concerned. Thus, the interface manifested in the splitting of Neg results in two negative projections, though semantically there is only one sense of negation, perhaps along the lines argued for in Zeijlstra (2007).²⁰

Given also the nature of -š in YA that it is a clitic, Neg will be split into NegP and NegClP (cf. 18), but not NegP1 and NegP2 as is found in works on Romance languages such as Italian, French, etc.²¹

Now, let’s take the example in (20a) and derive it applying our proposal in (18). Thus, the structural representation of (20a) will be (20b).

As far as syntax and syntactic properties of negative imperatives are concerned, lā merges in Neg^o. pro merges in Spec,vP and remains there throughout the derivation simply because its features are valued via Agree relation with T^o. The verb ti-ktub “write” merges in V (not shown in (20b) for ease of exposition), then moves to v, and finally lands in T^o, where it checks its tense feature with T^o. The negative clitic particle -š merges and remains in NegCl^o. The -š may be said to affix onto the verb tiktub in the morphology module, applying M-merger (see e.g., Matushansky, 2006; Shormani, 2014).

From a minimalist perspective, our proposal seems to be minimalist in nature, employing the very needed machinery apparatus of the grammar, in accordance with the Economy Principle, which states that structural representations should have as few (syntactic) operations as possible (Chomsky, 1995, et seq).

However, it should be pointed out here that the structure in (20b) represents only the derivation of the negative imperative syntactic properties. It does not, however, represent structures such as (21) and (22), which are complex. In (21), for instance, the object pronominal clitic -hā is suffixed to the verb ti-qtul “kill”, and the negative clitic particle -š is suffixed to it. The negative imperative structure in (22) is a little bit more complex than (21) due to having the topic banāt “girls”, the plural feminine suffix -ayn and the object NP d-dars “the lesson”.

However, we will postpone deriving such structures to a later stage after we discuss the discourse properties of negative imperatives. This is the focus of the next sections and subsections.

4. Discourse properties of negative imperatives

One very important feature of imperative, be it negative or positive, is that there is an addressee/hearer and speaker- the two interlocuters of a conversation, and given the conversation nature of imperatives (cf. Waltereit, 2002), one could assume that discourse is an essential component of imperatives. Several scholars argue that discourse plays a crucial role in the interpretation of an imperative structure (see e.g., Frascarelli & Jimènez-Fernàndez, 2021; Neeleman & Szendrői, 2007; Shormani, 2021; Zanuttini et al., 2012). Consider the following examples:

In the examples in (23), the interpretation/meaning of negative imperatives is not clear in terms of the reference/antecedents of pro. In other words, it is also difficult to know or identify the addressee. Though the inflections attached to the verb ti-ktub “write” may help us understand this addressee to some extent, but still we do not know who is the addressee which pro refers to (cf. Koeneman & Zeijlstra, 2014). In other words, from a syntactic point of view we can identify the addressee that he is a 2 person, masculine singular entity, but still “who” is this person? (cf. Shormani, 2017a, 2021).

However, with the (pro)nominal occurrence in (24a) and (24b), it is easy to identify the reference of pro. For example, in (24a), pro refers to the pronoun ʔantah “you” and in (24b) the reference of pro is banāt “girls”. Following Portner’s (2007) argument, Shormani (2021, p. 21) argues that there are performative/interpretive requirements enforcing the (pro)nominals to occur in (24a) and (42b). These performative/interpretive requirements can be: i) “the authority of the speaker over the addressee” Portner’s (2007, p. 361), ii) the identification of the addressee, as in (24b) (cf. van der Wurff, 2007). This suggests that ʔantah and banāt in (24a) and (24b) “are required by a performative function or interpretive import, which makes explicit a discourse property (Shormani, 2021, p. 21).” According to Isac (2015, p. 39), “[i]t has been noted repeatedly in the literature … that imperatives have a modal dimension, given that they present a proposition as a possible and desirable state of affairs.”

Furthermore, from a discourse perspective, imperatives can express a wide range of speech acts such as command, order, advice, request, etc. (cf. Shormani, 2019). This is, in fact, advocated by a number of linguists (cf. also Aikhenvald, 2010; Aloni, 2007; Han, 1998, 2001; Shormani, 2017a, 2021; Zhang, 1990). Zhang (1990, p. 11), for instance, identifies several ways in which imperatives are used. For example, from a syntactic viewpoint, an imperative can mean “a class that is parallel to declarative and interrogative”, and from a discourse perspective, an imperative structure can mean “the pragmatic notion of directive, including commanding, ordering, advising, requesting, suggesting, that is parallel to notions such as assertives, expressives and so on” (cf. also Bianchi & Frascarelli, 2010; Shormani, 2017a, 2021).

However, the question is: are negative imperatives similar to positive imperatives in expressing these or some of these speech acts? Consider the following examples:

The speech acts performed by the negative imperatives in (25a-e) are command, order, advice, request, respectively. For example (25a) can be said in a military situation where the two interlocuters, viz. the speaker and addressee are a commander and a soldier, respectively. The speech act in (25c) is an advice, say, a father or a mother advises his or her son not to “sleep late” to wake up early. The speech act performed by (25e) can be that of a teacher and a student, a manger and an employee, etc. And, of course, the interpretation of the speech act(s) performed by the negative imperative(s) can vary depending on who are the two interlocutors, and the situation(s) in which the negative imperative is said (cf. Shormani, 2019).

Thus, we assume that the preverbal (pro)nominals in negative imperatives are topics, left dislocated elements, but not the (thematic) subjects of imperatives. Furthermore, if we try to replace pro with the NP ʕali in (26b), the result is an ungrammatical structure, as in (26c). Consider the examples in (26):

These examples provide us with strong evidence that the preverbal NP is a topic, rather than the (thematic) subject, which is pro. However, in imperative structures, this topic could be thought of as the logical subject which is considered the (discourse/world) reference of pro.²²

Thus, based on these facts and evidence, we can conclude that topics are part of the underlying structure of negative imperatives. Therefore, we propose (27) as the structural representation of negative imperatives in YA.

Our proposal in (27) is based on the cartography-based approaches to projection, which assume that features and information coding factors have structural representations and project their own projections. This is in line with Rizzi’s (1997, p. 283) postulations that the C-domain is “the interface between a propositional content (expressed by IP) and the superordinate structure (a higher clause or, possibly, the articulation of discourse, if we consider a root clause).”

Thus, the TopP in (27) represents the information structure/coda of negative imperatives (cf. Benincà, 2001; Erteschik-Shir, 2007; Shormani, 2017b). The topic will merge in Spec,TopP. Other information on the (re)merging of other elements has been discussed in section (3.4) in relation to the syntactic structure (20).

4.2. Top’s syntactic features

In this section, we will briefly discuss Top’s syntactic features, and see what features the head Top^o is endowed with in our proposal. Given the fact that the difference between positive and negative imperatives has to do mainly with syntax, discourse in both structures seems to be the same. In other words, the only difference concerns the NegP and NegClP which are mainly syntactic properties, which have nothing to do with discourse. Based on this, we will adopt Shormani’s (2021) postulations regarding Top’s features. Shormani (2021, pp. 32–34) proposes that the head Top^o has the following set of features:

As for [Adrs] feature, Shormani (2021, p. 33) proposes (29) (cf. Frascarelli, 2007; Rizzi, 2006; Shormani, 2017b).

The [Adrs] feature ensues from the fact that since the topic is the addressee, and since it merges/is based-generated in Spec,TopP as a C-domain element, Top^o has an [Adrs] feature, which is checked by the preverbal NP. Given this, Top^o will yield a discourse property; it also links the topic in C-domain with pro, in T-domain, the information structure, with the propositional structure, respectively. Given that C is represented by Top in our analysis, and since CP is a phase, it follows that TopP is a phase, and so, the [Adrs] feature could be thought as a cartographically constituent in the left periphery, having an “‘information structure primitive’ solely needed as an information/discourse requirement”. If this is on the right track, it follows that [Adrs] is an Edge Feature in Top^o by analogy with C, “which is valued via (re)merging a topic in Spec,TopP” (Shormani, 2021, p. 33, cf. also; Chomsky, 2005, 2008, p. 139).²³

As for [Spcty] feature, we assume that it arises/is acquired via two factors: i) the topic NPs bind a 2 pro, hence they acquire [Spcty] feature from being coreferentially linked with a 2 pro, as in (30a), but not 1 or 3, as in (30b& c):

And ii) these NPs are specific in imperatives due to being addressed entities, which is again a property, solely derived from/by discourse (cf. Erteschik-Shir, 2007).

As for [φ-features; Case; Tense] features that the head Top^o is endowed with, we have strong empirical evidence that C, here Top, is endowed with these features, theoretically and empirically. Theoretically, given Chomsky’s (2005, et seq) minimalist theoretic notions that T inherits these features from C in the narrow syntax, we have some sort of evidence that our Top has these features. Empirically, Shormani (2017b, pp. 152–153) provides us with strong empirical evidence that C, has φ-features. Consider the examples in (31)

In (31a-d), C (llaði, llað-iina, llati and llāti, respectively) agrees with the constituent it introduces, namely r-rajul, r-rijālu, al-bintu and al-banātu, respectively, in all φ-features.

As for Case, though Case marking in YA has been lost specifically in nominal NPs, Standard Arabic provides us with strong evidence that C in Arabic has Case as the examples in (32) show:

In (32a), the topic ʔallah occurs with a default nominative Case; however, in (32b) it occurs with an accusative Case assigned by the C ʔinna.

We also assume that C has Tense feature, though Arabic, be it SA or YA, does not show this. However, there are some languages in which C is endowed with a (T)ense feature.

Adger (2007, p. 34) argues that C in Irish exhibits a past and non-past tense contrast, as illustrated in (33).

As can be observed, C shows tense contrast; it is go in (33a), but gur in (33b). The former is present and the latter is past. Given this, it is possible to assume that T (even in imperatives) inherits C’s tense feature.

Thus, given our postulation that the head Top^o is endowed with an Edge Feature, which is valued via re-merging an NP in Spec,TopP, and given our discussion above that C, Top in our story, is φ-complete, we can postulate that TopP is a phase, whose head is Top^o. Along these lines, Chomsky (2005, 2008) proposes that based on LF interpretation purposes Lexical Items (LIs) enter the derivation endowed with an Edge Feature. This feature enables them to enter the computation. In Chomsky’s (2008, p. 139), “[a] property of an LI is called a feature, so an LI has a feature that permits it to be merged. Call this the edge-feature (EF) of the LI.” In our system, the fact that [Adrs] feature counts as an EF is motivated by LF interpretation purposes, and stems from the behavior of the topics. The topic is assumed to be the addressee, and the logical subject of imperatives, which greatly contributes to the interpretation of imperative constructions. And based on antecedent reasons, “T, φ-features and Tense appear to be derivative, not inherent: basic tense and also tenselike properties (e.g., irrealis) are determined by C (in which they are inherent)” (Chomsky, 2008, p. 143). This antecedent factor is at the heart of the proposal pursued here. In other words, assuming that the notion “antecedent” implied in Chomsky (2008) refers to a constituent in the A’-dependency domain, it is possible to argue that this antecedent is the topic as intended in our story. Strong evidence of this antecedent can be observed in the examples presented in (31), where the topic, the relative pronoun and the verb agree in all φ-features (see also Shormani, 2017b; Shormani et al., 2018).

Thus, if the head Top^o is a phase head, having an [Adrs] feature, then it will have two probes, namely Agree Feature and Edge Feature (cf. Chomsky, 2008, p. 148). The former concerns φ-features, in that Top^o probes for valuing its unvalued φ-features via Agree with pro in Spec,vP. The latter, however, is satisfied by (re)merging a topic NP in Spec,TopP.

In addition, though the head Top^o has several features, not only [Adrs], we have taken [Adrs] as an EF. This is simply because it is discoursally more prominent than the [2Pers] feature. In the literature (see e.g., Bennis, 2006; Zanuttini, 2008; Zanuttini et al., 2012), it is postulated that pro is licensed by a higher head taking the [2Pers] feature as more prominent than any other feature. As Shormani (2021, p. 34), points out, “it is possible to assume the [2Pers] feature to be an EF of Top^o, the [Adrs] feature seems to be more logical than [2Pers]”. He adds that from a theoretical perspective, “the [2Pers] feature cannot simultaneously be assumed as a φ-feature and an EF one … since the [2Pers] is part of the φ-composition of the head Top^o, it is theoretically untenable to take it as an edge feature as well”. Thus, if Top^o is the head of a phase, it “may be the locus of agreement, selecting T and assigning it (unvalued) φ-features” (Chomsky, 2005, p. 18).

5. Information coda and propositional coda: where they meet in imperatives?

As we have argued so far, there are two structures in our analysis: information coda and propositional coda. The former represents the C-domain, and the latter the T-domain. We have also argued that the logical subject, i.e. the topic is “housed” by the C-domain, particularly in Spec,ToP, while pro is housed in T-domain, precisely in Spec,vP. Given our postulation that TopP is a phase and since vP is also a phase, the Agree relation established between the two constituents is some sort of a phase Agree.

Since the topic NP merges in Spec,TopP, and since it is linked with pro in Spec,vP, there should be a mechanism under which both constituents are linked for full/ultimate interpretation of pro (cf. Moon, 2001). As we have alluded so far, this mechanism is coreferentiality, and following Shormani (2021), this coreferentiality between the topic in the information coda, i.e. TopP, and pro in the propositional coda, i.e. TP, takes the form of Agree as Match, which results in a local A’-chain.

Thus, given all this, let’s now see how our proposal accounts for negative imperatives like (34):

We select the example in (34) for its complexity to examine to what extent our proposal can account for it. It is, in fact, a complex negative imperative structure in the sense of having a topic banāt, the verb which has several suffixes: -ayn indicating a feminine 3 person and a plural number, -uh which functions as an object of the verb, and -š as the negative postverbal suffix, and pro the thematic subject of the structure. The structural representation of (34) is, thus, (35) below:

The derivation of (35) proceeds as follows. Given the minimalist bottom-up derivation, the derivation of (35) starts by merging the object NP with V^o which projects into V¹, (this is not shown in (35) for ease of exposition), which merges with v^o and projects into vP. The structural derivation proceeds until the topmost element, i.e. TopP enters the derivation, where the topic NP banāt merges in Spec,TopP. The object-clitic -uh originates in NP as a complement of V^o, then moves to Spec, NegClP, its landing site. The reason for -uh to move to Spec,NegClP is necessitated by word order, and suffixation, since it is suffix suffixed to the verb ti-ktub-ayn; this is shown by the red arrow. The latter merges in V^o, moves first to v^o, and then raises to T^o, where it acquires Tense. pro merges in Spec,NegClP, and remains there throughout the derivation process. The postverbal negative suffix -š merges in the head NegCl and “stays” there. lā merges in Neg^o, where it checks its Neg feature. Since it is a head, it has to merge in Neg^o. The topic banāt “girls” merges in Spec,TopP.

There are two types of Agree relation to be established. The first Agree relation is established between T and pro; in this Agree relation, T functions as a probe and pro as a goal. As a result of this Agree relation all unvalued feature get valued and deleted. in this Agree, pro’s unvalued Case feature is also valued by the valued Case feature of T. The second Agree as Match relation is established between the topic banāt and pro where all unvalued features get valued and interpreted. As a result of this Match Agee, pro is fully and ultimately interpreted as banāt.

It follows that when the topic is merged, it matches and values (and interprets) T’s and pro’s features. And given the antecedent nature of [Adrs] feature, it is likely that pro obtains the feature specifications of the topic before/during Transfer to the interfaces (cf. Chomsky, 2004, 2008). This is likely to result in a local A’-chain. In other words, given that the topic is hosted in the C-domain, and pro in the T-domain, coreferentiality due to Agree as Match between the topic, Topº, Tº and pro results in a local A’-chain.

6. Conclusions and further implications

This article has investigated a topic in the syntax of negative imperatives taking YA as the data of investigation, a topic that has not been tackled before. The article proposes that the thematic subject of these constructions is pro while the logical subject is a topic occurring preverbally (for more on the logic of imperatives, see e.g., Hansen, 2013). The negation in negative imperative in YA takes the form of a bipartite structure “lā … – š”, where the negative particle lā is preverbal and -š a postverbal negative suffix, thus paralleling negative imperatives in French, for example. While lā is projected as NegP,– š is projected as NegClP given the clitic nature of –š. We have also accounted for Neg splitting into these two projections arguing that this splitting is a by-product of Agree and discourse coreferentiality (cf. 19). The operation Agree is taken as Match correlating the topic in the informational coda, i.e. C-domain with pro in the proposal coda viz. the T-domain, thus, coreferentially linking the topic and pro. This coreferentiality results in the ultimate interpretation of pro. Coreferentiality takes place at the syntax discourse interface via Agree as Match. In the course of the article, the origin of the postverbal negative particle –š was discussed, concluding that –š results from the grammaticalized process šay “thing” has undergone.

The controversial aspect in negative imperatives, and bipartite negation in general, viz., the assumption that there are two Neg projections has been accounted for in our analysis. Thus, answering the question: why should there be two NegP projections?, and following Tilleson (2019), we have proposed that Neg in bipartite negative imperatives is split into two projections, one representing lā and one -š. Differently from Tilleson (2019), we have proposed that the former is NegP and the latter NegClP, given the nature of -š as a clitic in YA.²⁴ We have also argued that the splitting of Neg is a by-product of syntax–discourse interface. Syntax is represented by Agree as Match and discourse by coreferentiality between the 2 person of the addressee, namely the topic, the (null) head Top, which has a scope over Neg (cf. 18 & 19) and -š, but not between Top and lā, given the possibility of deleting lā in some Arabic verities like EA, JA, etc.²⁵

As for the feature-composition of T and Top, we have argued that both are φ-complete (cf. Chomsky, 2000, 2001). For example, T in negative imperatives has all the features possessed by T in any other clause-type, be it declarative, interrogative, etc., namely, φ-features, Case and tense, which gives rise to non-construction specific postulation, a recent minimalist notion (cf. Chomsky, 2005; Shormani, 2021). Very recently, Shormani (2021) proposes that imperatives have tense, but only present or future, but not past. Shormani bases his proposal on some temporal adverbs like “now” and “tomorrow”. He also cites some Arabic traditional grammarians such as (Sibawayhi, 1938; Wright, 1898) who argue that “the actual ‘happening’ of the speech act performed by an imperative verb is being carried out after or at the time of speaking. They actually base their argument on some temporal adverbial modification as in do it now/tomorrow, but not *do it yesterday” (p. 25).

Thus, T in our system inherits Agree Feature from C in the narrow syntax, because as a phase head, Top^o “may be the locus of agreement, selecting T and assigning it (unvalued) φ-features” (Chomsky, 2005, p. 18). And based on antecedent reasons, “T, φ-features and Tense appear to be derivative, not inherent: basic tense and also tense like properties (e.g., irrealis) are determined by C (in which they are inherent)” (Chomsky, 2008, p. 143).

As for Top; furthermore, and given that antecedent factor is at the heart of the proposal pursued here, assuming that the notion “antecedent” implied in Chomsky (2008) refers to a constituent in the A’-dependency domain, it is possible to argue that this antecedent is the topic as intended in our story. Strong evidence of this antecedent can be observed in the examples presented in (32), where the topic, the relative pronoun and the verb agrees in all φ-features (see also Shormani, 2017a, 2021). Given the coreferentiality proposed here, and given also Feature Inheritance postulation, C’s features, Top’s here, are argued to transmit from Top to T, and finally to pro. This inheritance of features results in valuing/interpreting the unvalued/uninterpretable features of these elements, hence the ultimate interpretation of pro in negative imperatives.

The question now is: is it possible to apply our proposal to data from other languages? We have stated so far that sentential negation, and negative imperatives in YA is somehow akin to that of French. As for the subject of negative imperatives in French, several scholars (see e.g., Isac, 2015; Rowlett, 2014; Rupp, 2003) propose that the subject of imperatives in French is pro. French has several types of negative imperatives like those in (36) (cf. Rowlett, 2014):

We notice that the French negative imperatives in (36) are simple like (36a), somehow complex, as in (36b), and complex, as in (36c). The complexity of structures like (36b) and (36c) arises from having enclisis and proclisis, respectively. The former is represented by le “it/him” occurring before the verb fais in (36b), and the latter by le suffixed to fais. (36c)’s complexity can also be ascribed to having to two verbs, viz. fais and tomber and a topic Jean. Let’s apply our proposal to (36c), which is the most complex in the set of the examples in (36). Thus, according to our proposal, (36c) will have the structural representation in (37) below²⁶:

The example in (36c) is considered a complex negative imperative in French, due to the fact that it has a proclisis -le. This proclisis functions as the object of the verb. There are also two verbs in this structure, viz. fais “do” and tomber “fall”. The former is finite, while the latter is nonfinite (say, infinitive). Therefore, the tomber merges in V^o and remains there. However, fais merges in v^o and then moves to T^o, where it acquires/checks tense. The proclisis -le first merges in NP, as the object of the verb tomber and then raises to Spec, NegClP much the same way as -uh in Arabic in (34) above. Thus, if our proposal can account for such structures as (36c), it follows that it can account for simple structures like those in (36a & b), simply because (36c) is more complex than (36a & b) even when a topic is present. The rest of the derivation is identical to the Arabic example in (35). The only difference is the absence of ne which cannot occur in such structures as we have pointed out so far.

Given the English gloss of Arabic examples throughout the paper, it seems that our proposal can be applied to English negative imperatives as well. According to our proposal, (38) will have the structural representation in (39):

Thus, it seems that our proposal straightforwardly accounts for negative imperatives in English as well. In (39), the derivation starts by merging the DP a ward as a complement of the verb say. The verb say merges in V^o and remains there for nothing compels it to move up as it is nonfinite, it has nothing to check. pro merges in Spec,vP, as the thematic subject of the imperative structure. The Aux. verb do merges in T^o where it checks its tense feature. For word order purposes, it then moves to Neg^o to adjoin to the negative particle n’t. As the topic of the structure, the NP Ali merges in Spec,TopP. An Agree as Match is established between Ali and pro. Ali functions as the reference of pro by means of coreferentiality relation at the interface, where syntax and discourse are linked. This coreferentiality at the syntax–discourse interface results in the full interpretation of pro, where pro acquires the feature-specifications of the topic, here Ali.

Thus, it seems that our proposal accounts for YA negative imperative data, and other modern Arabic varieties, given their similar properties. Furthermore, elegantly accounting for French and English negative imperative data gives further strength to our analysis. Bearing this in mind, our proposal could also be applied cross-linguistically, given the similar properties of negative imperatives invariably across languages, more specifically the subjectless property, i.e. pro as the thematic subject, performative import, etc.

This article, however, does not include some cases of negative imperatives in YA. One of these cases is conjoined negative structures, as in (40):

In this example, two negative imperatives have been coordinated, viz., lā ti-ftahh fam-ak and lā ti-tkalamkalima with the conjunction word wa. wa is the conjunction word whose function is to conjunct two words, phrases or sentences in Arabic, in general.

Another case we have not tackled in this article is negative imperatives in embedded clauses in “say-clauses”, as in (41):

In (41), the negative imperative lā ti-tkalam occurs as an embedded constituent in the “say-clause” qulk lak “I told you”. These cases are interesting and challenging. However, we leave them for future studies.

Acknowledgments

Earlier versions of this article were presented in the International Conference of Linguistics, held during 21–22 of October, 2019, Goethe University Frankfurt, Germany, and in Cyprus Acquisition Team (CAT), on the 17th of February, 2020. We thank the audience there, specifically Phoevos Panagiotidis and Kleanthes Grohmann, for their insightful comments. We thank Phoevos Panagiotidis for his discussion and feedback. We also thank three reviewers of Cogent Arts & Humanities for their valuable comments. The usual disclaimers apply.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

This research was supported by Deputyship for Research and Innovation, Ministry of Education in Saudi Arabia (IFKSUOR3-211).

Notes

1. The following abbreviations are used throughout this article: 1, 2, 3 = first, second and third person, respectively, Acc = Accusative, Adrs = Addressee, Agr = agreement, ASP = aspect, C = complementizer, EF= edge feature, EPP = extended projection principle, F = feminine, FUT= future, Gen = Genitive, Gend = gender, Neg = negative, Nom = Nominative, Num = number, P&P = Principles and Parameters, PL = plural, POS= positive, PRS= present, PST= past, S = singular, Spec = specifier, Spcty = specificity, SVO = subject verb object, T = tense, TOP = topic, TopP = topic phrase, u = unvalued, V = verb, v = v in vP, v = valued, VSO = verb subject object. Other abbreviations and/or acronyms used in the text are introduced in the first use.

2. It should be noted here that in YA, the preverbal negative particle mā can be used in negative imperatives, but it is not very common. The very common negative particle used in imperative is lā. However, the postverbal negative particle -š is used in all clause-types, viz. declarative, imperative, interrogative, etc.

3. In the literature, bipartite negation is referred to as a discontinuous negation, that is mainly due to the fact that there are two negative particles (see e.g. Zanuttini, 1991; van der Wurff, 2007; Poletto, 2008; Hagemeijer, 2009; Isac, 2015, among many others).

4. In fact, there are a lot of Arabic dialects that use -š as a negative particle including Egyptian Arabic, Moroccan Arabic, Palestinian Arabic, Syrian Arabic, Jordanian Arabic.

5. It is worth stating here that SA is a very rich inflected language. The gloss shows that (7a) is an imperative structure and the subject is interpreted as 2 masculine, singular pro, while it is 2 feminine, singular pro, and 2 masculine, plural pro in (7b) and (7c), respectively.

6. This example is taken from Lafkioui (2013, p. 67)

7. cf. Aoun et al. (2010). We are supposing here that in MA – š in negative imperatives can be deleted. The evidence can be inferred by analogy from declaratives, where mā/lā in MA can be deleted. Chatar-Moumni (2013, p. 222) provides examples of deleting – š as in (ia &b):The original gloss is as follows:(i) a. bi-t-ħib-ššiγlil-baytasp.elle.imp.aime-négtravail la-maison“Elle n’aime pas le ménage.”b. beddi-šveux-moi-nég“Je ne veux pas.”In (ia & b), mā is deleted and the negative function is performed by – š per se. The fact that negation is expressed is clear from the French gloss, namely n’aime pas “doesn’t like” and ne veux pas “don’t want”.

8. The negative structure in (9) is similar to the modern French use of pas, in that pas is used alone, i.e. without the preverbal negator ne, as in (i):(i) écris pas!Write NEG“Don’t write!”The fact that it is possible to drop lā in Jordanian and Palestinian negative constructions may be due to language contact between French and these varieties of Arabic, as Jordan and Palestine were French colonies for a long time.

9. There are also some peripheral dialects of Arabic such as Cypriot Arabic, spoken in Cyprus, where lā/mā is deleted in negative constructions, and -š is used per se (see e.g., Shormani, 2023).

10. Given that imperative structures are finite, any other empty categories cannot serve as the subject of imperatives. For example, Beukema and Coopmans (1989, pp. 420–422) points out that there are four categories of null constituents, viz. trace of NP-movement (NP-t), trace of Wh-movement, (Wh-t), PRO and pro. However, they provide empirical evidence that the only possible null constituent to function as the subject of imperatives is pro. The reader is advised to read their convincing arguments based on which they reject the other three alternatives. Kratzer (2009, p. 189, fn.2) also distinguishes between pro and PRO, stating that “pro can be born with all its features in place, in which case it is referential”, which is exactly what we are suggesting in our story (see also Chomsky, 1995, pp. 106–109, for arguments distinguishing pro from PRO (cf. Also Shormani, 2021).

11. It has been held in the literature that there is a functional head in the C-domain which licenses and interprets pro. For instance, Beukema and Coopmans (1989) holds that the functional head which is involved in the licensing and interpretation of pro is a tenseless INFL. Jensen (2003) proposes an imperative-flavoured-T. Bennis (2006) advocates an imperative C, and Zanuttini (2008) prefers a Jussive T. In fact, Zanuttini’s analysis was also assumed by Zanuttini et al. (2012).

12. Jensen (2003) proposes that imperatives have a tense feature, but different from that of declaratives, ascribing this tense feature “to the presence of an imperative-flavoured-T° that competes with prototypical-declarative-T° for this functional position” (p. 158).

13. Isac (2015, p. 51) argues that there is a problem if we assume that imperatives lack tense, stating that though imperatives can be interpreted in the present or future tense, they “cannot be interpreted in the past” providing the following examples:(i) a. Do try this!b. *Did try this!Isac comments on this fact as “[t]he incompatibility between Past Tense and imperatives could be explained by the idea that putting an obligation on an addressee to bring about a past state of affairs is pointless.”

14. Zanuttini (1996, p. 189) provides examples from the northern Italian dialect of Piedmontese in which negative imperatives show tense, as in (i):(i) Parlanen!talk-2SG NEG“Don’t talk!”

15. Note that though Benmamoun (2000) discusses what we call bipartite negations (including those of negative imperative) in Moroccan and Egyptian Arabic, he left it here without specifying bipartite negative projections.

16. Rowlett (2014) proposes a clitic projection to account for structures like Fais-le!, which is a positive imperative. However, cliticization like this does not exist in French negative imperatives for such structures as in *Ne fais-le pas! are ungrammatical.

17. However, our proposal differs from Zanuttini’s (1991) proposal in several aspects: first we consider both negative particles heads, and occupy heads of NegPs; second, her proposal adopts mainly P&P while ours adopts Minimalism, specifically Phase Theory. In the latter approach, we have also accounted for the assumption that Neg … NegCl are a result of Agree, whereby Neg is split into Neg and NegCl (cf. Hagemeijer, 2009).

18. A very important observation to be made here is that in YA, and all Arabic varieties, there is also some sort of tense in the negative particle lā. In Arabic dialects, specifically Standard Arabic, lā indicates and negates present, lan future, lam past, and mā is used to negate past and present constructions (see Shormani, 2015).

19. This line of thought can also be supported by the assumption that lā is a polarity item (cf. Benmamoun, 1997; Pollock, 1989).

20. Zeijlstra (2007) has also attributed the double negation to interface properties, but he assumes that this interface is between syntax and semantics. However, from a syntactic perspective, one could not ignore the syntactic representation of two Neg projections, and consider only one, which is in line with the current minimalist assumptions (cf. Hagemeijer, 2009; Tilleson, 2019).

21. In the literature on bipartite negation, double negative projection analysis has been advocated and applied cross-linguistically, regardless of the nature of both projections. Most scholars apply double NegPs, some others apply PolP along with NegP, yet some others apply AspP, along with NegP, etc., Thus, it seems that double negative analysis is dominating bipartite negation investigation (see e.g. Bell, 2004, for Afrikaans; Aboh, 2004, for Gbe; Pollock, 1989; Rowlett, 2014; Tilleson, 2019, for French; Hagemeijer, 2009, for Santome; Tilleson, 2019, for Sgaw Karen and Ojibwe (and French), among many other scholars and languages).

22. Note that if we replace “Ali” in (26b) by the pronoun ʔantah, the structure is grammatical as (i) shows:(i) lāti-qraʔ- šʔantahbi-sautʕālinot2-read.MS- NEG youwith-voice loud“You, don’t read loudly!”In fact, at the moment we do not have an answer to the question: why is a pronominal NP, but not a nominal one, is allowed to be used in examples like (26b)? We will leave it for future studies.

23. In case of the unavailability of the preverbal NP, i.e. the negative imperative occurs without a topic, we assume Shormani’s (2021) postulations that in this case a null topic, i.e. pro, can be assumed. In (i), for instance, the interpretation of the thematic subject pro and the topic pro will be determined by the discourse in which the imperative is said.(i) pro lā ti-tkalam-šproNEG 2-talk.MS-NEG“Don’t talk!”Suppose (i) is said by an invigilator in the exam, and suppose it is preceded by a sentence like It is an exam, so nobody is allowed to speak. We will expect that the interpretation of the null topic pro in (i) is students.

24. Note, however, that in Tilleson’s (2019) work, for instance, it is argued that Agree takes place between a null head and an overt focus negator (p. 68). However, since in YA the lower negative particle is not null, we assume that Tilleson’s analysis may not be applied to YA. Rather we adopt Hagemeijer’s (2009) postulations that “Neg1…Neg2 is an instance of a non‐local relation of agreement between functional projections” (p. 158) (cf. also Wiltschko, 2021).

25. Given that a feature in minimalism projects its own projection (cf. Chomsky, 2005, 2008; Shormani, 2017b), and since Neg is a feature, it is possible to posit that in bipartite negation when Agree takes place, it results in splitting Neg into NegP and NegClP (see also de Clercq, 2018; Poletto, 2008; Tilleson, 2019).

26. Note that that the preverbal negative particle ne is and must be absent in such French constructions, as in (36c) (Rowlett, 2014). Rowlett argues that ne cannot co-occur with proclisis.