On the Syntax of Sentential Negation in Yemeni Arabic

In this paper we explore the system of negation in modern Arabic dialects with a particular focus on Yemeni Arabic (Raymi dialect). The data observed in this dialect incorporate important and novel facts related to the syntax of sentential negation in Arabic. This includes the distribution of negation patterns and the interaction between negation and negative polarity items, which challenges the two widely adopted analyses for sentential negation in Arabic: The Spec-NegP analysis and the discontinuous Neg analysis. In this paper we argue that neither analysis can provide an adequate account of Raymi Arabic facts. Instead, a more recent analysis, the Spilt-Neg analysis, can accommodate them. In addition, in the study we provide empirical evidence in support of the Higher-Neg analysis, wherein Neg is projected higher than T in the derivation.


Introduction
The syntax of negation in Arabic is as extremely diverse as the varieties of the language themselves. Negation can be expressed in various ways that use different patterns across the varieties of Arabic (note 1). Negative constructions in these varieties range from those containing a single negative marker, such as Modern Standard Arabic (henceforth, MSA) as in (1), Gulf Arabic, Hijazi Arabic and Syrian Arabic, to those containing two negative markers (bipartite negation), such as Moroccan Arabic as in (2) (note 2), Egyptian Arabic, Palestinian Arabic, Yemeni Arabic (henceforth, YA) and so forth.
(YA/Raymi dialect) construction maa…ši in YA, focusing on Raymi dialect, before considering the equivalent constructions in other varieties of Arabic. In section 3, we look at the previous analyses of negation in Arabic to determine whether or not they can accommodate the facts related to negation in YA (Raymi dialect). We then, in section 4, discuss the categorial and functional status of -ši to determine its position in the clausal structure. In section 5, we provide a preliminary analysis along the lines of Soltan's (2011Soltan's ( , 2014 Spilt-Neg analysis. Finally, we conclude the paper in section 6.

Negation in Raymi Dialect
Although few studies have been conducted on negation in YA (see, e.g., Mansoor, 2012;Simeone-Senelle, 1996;Vanhove, 1996), to the best of our knowledge, none have been conducted on the syntax of negation in Raymi dialect (note 6). Negation in this dialect is expressed by either the discontinuous negative form maa… ši or the non-discontinuous negative form maa-ši, which can both be used to negate sentences containing verbal, nominal, adjectival and prepositional predicates as examples (7-10) illustrate, respectively.
(YA-Raymi dialect) NEG-NEG she in the market 'She is not in the market.' As the above examples demonstrate, both the discontinuous negative elements maa-x-ši and the non-discontinuous negative elements maa-ši are used to negate all types of predicates. This is not the case in other Yemeni dialects and in most Arabic varieties, as will be discussed shortly. In addition, there is no semantic or pragmatic difference between the two configurations maa-ši and maa…ši in (7-9) above (note 7). However, the second negative marker -ši can sometimes appear at the end of the clause, but this seems to be restricted to the context of oath only, as illustrated by the following examples:  Watson (1993, pp. 121, 226) reported some examples from YA (Sanʕani dialect) where the non-discontinuous negative elements maa-ši are used in two cases: first, to provide a negative answer to yes/no questions as in (16), and second, in elliptical contexts as in (17) there people take off.3.PL DEF-skin and-people NEG-NEG 'There are people who take off the skin and some people who don't. (Watson, 1993, pp. 121, 226) This is very much the situation in a southern dialect in Saudi Arabia (henceforth, SA) known as Zahran dialect. The non-discontinuous negative elements maa-ši appear in negative answers to yes/no questions. Interestingly, ši can appear in positive answers to yes/no questions as well. Consider the following examples:  (Vanhove, 1996, pp. 4, 7) Vanhove (1996, p. 4) noted that ši is used in association with maa in the Yaafiʕ dialect to serve specific purposes, among which is denying existence. She termed maaši 'the negative marker of existence'. Furthermore, she noted that the negative elements maa-ši in all the examples she recorded are placed either before an indeterminate noun (23) or after a determinate noun (24) or a demonstrative pronoun (25).  (Vanhove, 1996, p. 4) Note that in Sanʕani, Abyani and Yaafiʕi dialects, the negative particles maa and ši are realised only as non-discontinuous elements, in contrast to the data from Raymi dialect shown earlier. However, Vanhove (1996, p. 2) observed that maa-ši can occur discontinuously in Yaafiʕi dialects when the non-clitic ši means 'nothing', as illustrated by the following example.

Maa…ši in Other Dialects and Varieties of Arabic
(26) maa ʔasuuk ši NEG found-1.SG nothing 'I did not find anything.' (Vanhove, 1996, p. 2) The negative elements maa-ši are also attested in other Arabic varieties such as Moroccan Arabic, in which sentential negation is marked with both the non-discontinuous form ma-ši (with short vowels) in the context of non-verbal predicates and the discontinuous form ma-v-ši in the context of verbal predicates, as noted in (2). This is different from the case in YA (Raymi dialect), as discussed earlier. However, Ouhalla (2002, p. 304) reported some examples of negative clefts in Moroccan Arabic, in which the non-discontinuous form ma-ši is used to negate sentences containing verbal predicates such as the following: (27) ma-ši qrat Nadia l-ktab.
NEG-VAR read Nadia the-book 'It is not the case that Nadia read the book.' * 'Nadia did not read the book.' Note that the interpretation here is semantically different. The example in (27) does not negate a statement but corrects it by letting the listener suppose the unsaid, that Nadia bought, borrowed, threw or wrote the book. However, similar examples of such readings are not found in the Yemeni dialect of Riamah.
As for MSA, the equivalent construction would be the one introduced by a single negative particle maa, which can be used to negate in a wide range of contexts. Thus, it can negate sentences with verbal predicates in the past and present (habitual only) tenses, as shown in (28a) and (28b), respectively. It can also negate sentences with non-verbal predicates, namely nominal as in (27c), prepositional as in (28d) and adjectival phrases as in (28e).
The preceding discussion is summarised in Table 1, which focuses only on the use of the negative marker maa and its variants maaši, ma-ši, ma-š, mi-š, muš, muu and so forth in the varieties of Arabic.
To sum up, in YA (Raymi dialect) the negative elements maa-ši are used continuously and discontinuously to negate all sorts of sentences. Furthermore, they are realised as two negative elements and not as a single complex form consisting of two parts: maa + -ši. Moreover, the second negative marker -ši can appear in pre-predicate position and in post-predicate position. The question that arises here is how these facts related to negation in YA can fit within previous analyses of negation in Arabic. Let us now consider these analyses to determine whether they can accommodate these facts. ijel.ccsenet.

Previou
One of the Chomsky' Aoun et a located bet that in Raymi dialect the first particle maa always precedes the predicate, whereas the second particle ši appears in different positions. This suggests that the negatives maa and ši are two independent markers occupying different positions and not a single complex form consisting of two parts (the prefix maa-and the suffix -ši) generated in Neg°. Finally, Benmamoun's analysis faces problems accounting for negation in future tense sentences in some Arabic dialects. In fact, this issue was first observed by Soltan (2007, p. 185) in Egyptian Arabic, where the negative marker miš precedes the future tense marker as in (34) below. The situation in Raymi dialect is slightly different, as the examples in (35) illustrate. We will return to this later. It can be inferred from the preceding discussion that Benmamoun's (2000) analysis, referred to in the literature as Low-Neg Analysis, cannot account for all the facts related to sentential negation in Modern Arabic varieties. There is, however, an alternative analysis to Low-Neg Analysis proposed by Soltan (2007), where NegP is located in a position higher than a TP, along the lines suggested by Fassi Fehri (1993) and Shlonsky (1997). This analysis is referred to as High-Neg Analysis (Soltan, 2011) and is sketched in (39) below. Ample empirical evidence from different Arabic varieties supports High-Neg Analysis over Low-Neg Analysis (see Benmamoun et al., 2013; Soltan, 2011 for more information).

The Analysis
The negation patterns attested in YA (Raymi dialect) suggest that the projection of Neg must be in a position higher than T in the course of the derivation. Furthermore, the interaction between NPIs and the negative element -ši in this dialect is quite similar to that observed in Egyptian Arabic, as (58-60) show. As noted above, the distribution of the negation patterns within this dialect does not follow from the contrast observed in many other Arabic dialects between verbal and non verbal predicates or between past and non --past tense sentences. It seems that there is no restriction on the contexts in which the discontinuous maa…ši and the non-discontinuous maaši patterns occur (note 14). Thus, we argue that a modified version of the Spilt-Neg analysis can account for the distribution of the negation patterns in this dialect. Soltan (2011Soltan ( , 2014 assumed that the negative marker maa is semantically negative, whereas -ši is formally negative because it developed diachronically from the adverbial usage of the noun šayʔ 'thing'. Thus, under this analysis maa is treated as a polarity head that originates in Pol and carries the interpretable negative feature [iNeg], whereas -ši is treated as a negative head that originates in Neg and carries an uninterpretable negative feature [uNeg] (note 15). Soltan (2014) pointed out that the uninterpretable negative feature on Neg is valued via a modified version of Agree (Chomsky, 2000(Chomsky, , 2001 between Pol and Neg. Furthermore, he argued that negation patterns are better dealt with as the result of morphological head movement and that -š can be deleted under certain conditions. He proposed the following head movement algorithm, which applies in the mapping from syntax to morphology (i.e. a post-syntactic rule): (61) a. In contexts where Neg is adjacent to a hosting head H, H moves to Neg and then to Pol, and the circumfixal maa-H-š pattern arises.
b. Otherwise, Neg incorporates into Pol, giving rise to the miš pattern.
- (Soltan, 2014, p. 104) A modified version of this algorithm can be adopted to account for the negation system in Raymi dialect. We assume that both steps in (61) are available for the negation patterns in this dialect. Thus, the discontinuous maa…ši pattern (cf. the example in 5a) is derived under step (a) as illustrated in (62), whereas the non-discontinuous maa-ši pattern (cf. the example in 5b) is derived under step (b) as illustrated in (63) As expected, only the NPI ʕumr appears in non-negative contexts such as questions and conditionals (64) and as a fragment answer (66b), whereas the NPI ʕaad does not. This suggests that the NPI ʕumr is non-negative and that the NPI ʕaad is lexically negative.
Based on Soltan's (2012Soltan's ( , 2014 analysis, the overt realisation of -ši depends on the availability of formal negativity. Thus, -ši disappears only in the presence of a non-negative NPI like ʕumr but not in the presence of a negative NPI like ʕaad. As for the contrast between (59a) and (59b), it can be explained in terms of 'locality'. This means that -ši disappears only if the NPI ʕumr originates within the local domain (i.e. 'close by' as in 59a) but not when it originates outside the local domain (i.e., in a distant position as in 59b) (cf. Soltan, 2014).

Conclusion
In this paper we discussed negation in Raymi dialect (a variety of YA), which has not been explored prominently before. The aim was to broaden the discussion about the syntax of sentential negation in Arabic. The distribution of the negation patterns observed in this dialect is somehow different from those attested in other Arabic varieties. Both the discontinuous negative pattern maa-x-ši and the non-discontinuous negative pattern maa-ši are used to negate sentences containing verbal predicates and non-verbal predicates. Unlike the situation in many Arabic varieties, there is no contrast between verbal and non verbal predicates or between past and non --past tense sentences with respect to the distribution of the negation patterns in Raymi dialect. In addition, NPIs are not always in complementary distribution with the negative enclitic -ši; it is not always omitted when an NPI occurs. These facts, among others, have posed challenges to the Spec-NegP analysis and the discontinuous Neg analysis, which have been widely adopted for negation in Arabic. We provided some empirical evidence to strengthen the argument in favour of the Higher-Neg analysis, whereby the Neg projects in a position higher than T. Finally, we showed that a morpho-syntactic analysis such as the Spilt-Neg analysis is the best candidate to account for most of the facts related to negation in this dialect.