Analyzing word order variation and agreement asymmetry in SVO and VSO structures of Standard Arabic: Towards a unified account

Abstract Standard Arabic (SA) exhibits two common word orders: subject-verb-object (SVO) and verb-subject-object (VSO). In SA, agreement is contingent on word order. In SVO, for instance, the verb agrees in all phi-features (phi features = person, number, and gender) with the subject preceding it. In contrast, in VSO order, the verb only agrees with the post-verbal subject in person and gender but not in number, resulting in partial agreement. The purpose of this article is twofold. First, it attempts to provide a unified account for the subject-verb alternation based on a distinction between ʔal-ʤumllah ʔal-iʔsmiyah “the nominal sentence” and ʔal-ʤumlah ʔal-fiʕliyah “the verbal sentence”. I assume that there is a Functional Phrase (FP) in the CP area, which is responsible for licensing the sentence as nominal or verbal. In light of this assumption, it will be shown that the feature on the Functional head (F) is the locus of word order variation in SA. Second, the current paper aims to account for the agreement asymmetry that is dependent on word order. In a probe-goal configuration, I propose that the subject in (SVO) order values the verb’s unvalued phi (ɸ) features, thereby obtaining full agreement. On the other hand, in a verb-initial clause (VSO), I suggest that because the verb moves to the Functional head (F), which is endowed with an valued singular number feature, the verb always has a singular feature, regardless of the subject (singular, dual, or plural) that follows it. As a result, partial agreement takes place between the verb and the post-verbal subject.

(1) a. ʔal-muʕlim-uuna HaDar-u ʔal-iʔʤtimaaʕ-a The In SA, SVO as shown in (1a) is the derived (marked) word order, whereas VSO (1b) is the unmarked (canonical/basic) word order. 1 Data from SA reveal that agreement is sensitive to word order.For example, in SVO ordering as in (1a), the verb triggers full agreement with the subject preceding it.
In contrast, in a verb initial clause (i.e.VSO), the verb partially agrees with the subject following it as (1b) above shows.
It is interesting to note here that when the agreement, which is associated with word order, is reversed, this results in ungrammatical sentences.In other words, if partial agreement is forced in SVO pattern, and full agreement in VSO, the resulting sentences are ruled out.The following examples are illustrative.Standard Arabic, as mentioned by Al-Horais (2009), is not unique in displaying agreement asymmetry and word order alternation.This phenomenon (i.e.agreement mismatch) has been observed in a number of unrelated languages (see Corbett, 2006 for French;Franck et al., 2006 for French and Italian).
In summary, the following research questions are the focus of this paper: (i) How does the proposed analysis account for the subject-verb alternation in SA?
(ii) How does the proposed analysis account for agreement asymmetry, which is sensitive to word order in SA?
The present paper is divided into four sections.Section 1 outlines the data and demonstrates the difference between SVO and VSO configuration with respect to the agreement between the subject and the verb.Section 2 reviews a number of previous generative treatments of SV/VS alternations and the agreement asymmetry associated with word order.In section two, we show a common problem that these analyses share.

Previous analyses
Since the subject appears in a pre/post-verbal position (see section one), much of the discussion in the generative linguistic tradition revolves around the nature and position of the subject.There are two research strands with respect to the SV/VS alternation in the literature.According to the first viewpoint, the genuine subject occupies two positions.In VSO, the subject remains in situ, in the specifier position of the VP, whereas it raises to Spec, TP in SVO pattern (Aoun et al., 1994;Benmamoun, 1992Benmamoun, , 2000;;Bolotin, 1995;Fassi Fehri 1993;Mohammad, 1990Mohammad, , 2000)).The opposing viewpoint, on the other hand, assumes that a genuine subject appears only postverbally, occupying Spec, VP.This analysis indicates that the preverbal subject in SVO configuration is not a genuine subject, but rather an A bar element, a topic phrase (Akkal 1996; Alexiadou and Anagnostopoulou  1998, Al-Balushi 2011, Al-Balushi 2012; Al-Horais 2009; Fassi Fehri 1993 2 ; Ouhalla 1991, 1994,  1997 3 Plunkett 1993; Soltan 2006, 2007) among many others).Both analyses assume the verb moves from V to T. With respect to agreement asymmetry, there are a number of studies that account for this phenomenon.To name a few, Mohammad (1990), for example, attributes agreement mismatches to elements occupying Spec, TP.He believes that in VSO, Spec, TP is always filled with a null pro which is masculine singular.Because of this, and under Spec-head configuration, the verb only partially agrees with the empty expletive pro which always has masculine singular features.In contrast, in SVO, because a full lexical DP occupies the Spec, TP, the verb triggers full agreement with the subject preceding it.Mohammad's study (often referred to as a Null Expletive Analysis in the literature) was actually refuted by a number of researchers.For instance, it has been argued that the notion that partial agreement results from a Spec-head relation between the verb in T and the pro in its Spec position (i.e.Spec, T) is inaccurate.According to Soltan (2007) Ouhalla (1994) assumes the nominative Case on the pro is transmitted to the postverbal subject.It is, however, implausible to make this assumption, as noted by Coopmans (1994).A null pro may have accusative case marking if preceded by ʔinna (an accusative case assigner), yet the postverbal subject has nominative case marking.If the case transmission assumption were true, the postverbal subject would have an accusative case.In his attempt to account for the full vs. partial agreement, Benmamoun (2000)a argues that the full agreement in SV order derives from a Spec-head relation that occurs between the verb in T and the subject in Spec, TP.On the other hand, partial agreement in VS order, as proposed by Benmamoun, is a result of a merger between the verb and the subject following it.According to Benmamoun (2000), in VS order, the number feature is spelled out periphrastically by the merged lexical subject, hence the verb does not inflect for number.However, Soltan (2007) provides a counterexample against a merger that requires adjacency between the verb and the subject. 4He observes that, despite the fact that the object and adverbial phrases may intervene between the verb and the subject, partial agreement occurs.
Utilizing an operation Agree, a recent work of Chomsky (2000Chomsky ( , 2001)), Soltan (2006Soltan ( , 2007) ) proposes an explanation for agreement asymmetry in Arabic.Soltan argues that agreement mismatch is determined by the features on T.He assumes that T has three features: (i) an obligatory gender feature (or CLASS feature), (ii) optional phi-features (typically person and number), and (iii) an optional EPP (Extended Projection Principle) feature.In VSO, T lacks the optional EPP and phifeatures (person and number); hence, partial agreement is maintained and the subject stays in situ (i.e. in Spec, VP).On the other hand, in SVO, T is endowed with three types of features, namely gender (or CLASS) feature, EPP feature, and phi-features, resulting in full agreement between the subject and verb.Soltan's analysis, however, faces challenges.First, the notion that full vs. partial agreement is dependent on the presence/absence of the number feature on T violates the constraint of "no optionality"; a requirement that eliminates optionality in the derivation according to the economic principle.Another flaw in Soltan's approach is the assumption that a pro subject can be identified by entering into an Agree relation with T that has unvalued phi-features.This premise, however, has been called into question.According to Holmberg (2005), as cited by Al-Horais ( 2009) "it is mistaken under the minimalist feature-valuing approach, which is adopted by Soltan's analysis, to assume that pro is in need of identification simply because within this theory of agreement, it is obviously not possible for an inherently unspecified pronoun to be specified by the φ-features of I [i.e.The common problem that the previous treatments of SV/VS alternation share is that they tie the SV/VS alternation to the presence/absence of an EPP feature on T. In their views, in SVO, T has an EPP feature that triggers movement of the subject to Spec, TP.In VSO, they assume that T lacks the EPP feature so the subject remains in Spec, VP.The problem with this analysis is that it is against the main merit of MP, which assumes "no optionality".To elaborate, in MP, movement takes place for reasons.For example, the subject moves from Spec, VP to Spec, TP to satisfy the EPP feature on T. In SA, as shown above, the derivation of SVO seems to fit with the idea that movement takes place for a reason, which is an EPP requirement of T. However, in VSO, where T lacks an EPP feature, there appears to be no need for movement to take place, so the subject remains in Spec, VP.Based on this, movement in Standard Arabic seems to be optional; it is determined by the presence/absence of EPP.In MP, nonetheless, optionality is excluded within the numeration due to the economy principle that bans superfluous steps in derivation.Under the principle of economy, the numeration which requires movement of the subject (SVO) cannot be used to generate another different linear word order (i.e.VSO).To avoid "no optionality" constraint, I propose, based on AlShammiry's (2016) analysis, that there are two distinct features on the Functional head (F) that are responsible for SV/VS alternation.When (F) is endowed with a nominal feature, the numeration generates an SVO construction.On the other hand, when (F) bears a verbal feature, the numeration produces a VSO construction.
Having explained the problems raised by previous treatments of Arabic word order alternation and agreement asymmetry, I will offer in the next section a fresh analysis based on the nominalverbal distinction to address the issue at hand.

Nominal-verbal-distinction-based analysis
During the eighth and ninth centuries, there were two prominent schools of Arabic grammar: 1. the Basri School and 2. the Kufi School.According to these schools, Arabic sentences are categorized into two types: i. ʤumal iʔsmiyah "nominal sentences" and ii.ʤumal fiʕliyah "verbal sentences".In Basris' view, 6 nominal sentences are the ones initiated by a noun.On the other hand, verbal sentences are the ones that start with a verb.Following Al-Basrah categorization of sentences in SA, we will show in the following two sub-sections that the word order variation arises in Arabic can be ascribed to the distinction between nominal and verbal sentences

AlShammiry's (2016) analysis
Using Chomsky's (1993Chomsky's ( , 1995Chomsky's ( , 2000) ) Minimalist Program (MP), AlShammiry (2016) provides a fresh account of the controversial topic regarding word order variations in Modern Standard Arabic (MSA) and the Saudi dialect.Based on a distinction between verbal and nominal sentences in Arabic, AlShammiry (2016, p. 3) assumes that ''there is a Head Phrase (HP) at the left periphery of the clause where the subject surfaces to check the nominal head feature of the HP and to mark the clause nominal''.Under AlShammiry's assumption, (3) would have the derivation given in (4) both sentence (3) and the tree diagram 7 are quoted from AlShammiry (2016, p. 3).
(3) ʔar-riʤaal-u HaDar-u The-men-NOM attend.PAST-PLM 'The men attended.' AlShammiry assumes that the DP ʔar-riʤaal-u "the men" is base generated in Spec, HP to license the sentence nominal.Furthermore, he argues that TP has two specifiers (specs), one filled with an expletive that checks the gender feature and the other filled by a full lexical DP subject or a subject clitic -u that checks the number feature.He argues that the verb in MSA comes from the numeration uninflected.He also adds that the number feature in the verb is weak, thus partial agreement is obtained in VS construction.Despite its novelty in identifying SV/VS alternation in Arabic through nominal-verbal distinction, AlShammiry's assumption has several flaws.First, AlShammiry wrongly analyses the morpheme suffixed to the noun ʔarriʤaal-u "the men", which is a nominative case marker, as a subject clitic that checks the number feature on the verb.Second, he does not clarify why the verb in the VS pattern comes from the lexicon uninflected, whereas it is fully inflected in the case of SV order.Furthermore, the idea that the null expletive pro checks the gender feature on the verb is questionable.According to Chomsky (1995), only full lexical DPs are able to value the unvalued phi-features on T. Finally, AlShammiry does not account for the verb appearing in a clause initial position preceding the subject, deriving, thus, a VS word order.He only discusses subjects that appear in initial clauses (i.e.SV constructions) where he assumes that they surface in the left periphery, specifically in Spec, HP.
In the next sub-section, I provide an alternative minimalist analysis in accordance with Chomsky's (1995) MP and building on AlShammiry's ( 2016) assumption.

An alternative analysis
In the current proposal and for the sake of simplicity (in line with MP's simplicity principles), I eliminate the two specs of TP assumed by AlShammiry (2016).Second, I assume, following AlShammiry, that there is a Functional Phrase (FP) projection (a Head Phrase in the terminology of AlShammiry, 2016) located in the CP area above TP, whose functional head (F°) is responsible for licensing the sentence as verbal if it is initiated by a verb or nominal if it is initiated by a noun.Based on this assumption, it follows that SV/VS pattern is dependent on the feature on (F), the head of the Functional Phrase (FP).When the functional head bears an unvalued nominal (u-nom) feature, the resulting word order is SVO.On the other hand, VSO is obtained if (F) contains an unvalued verbal (u-vrb) feature.Furthermore, I propose that the verbal feature on (F) is strong, thus attracting the verb to it, in a similar fashion of V to T movement.The modified version of AlShammiry's (2016) model is depicted in (5). (5) Having discussed the new version of AlShammiry's (2016) model, now let us see if this model accounts for the facts in SA.Recall from section one that SA allows the subject to precede/ follow the verb, thus obtaining SVO/VSO order.Consider (1a, b), repeated here as (6a, b) for ease of exposition.Under nominal-verbal distinction-based analysis, ( 6a) is nominal because it begins with the noun ʔal-muʕlim-uuna "teachers".Accordingly, (6a) will have the derivation along the lines in ( 7). (7) According to the tree diagram shown in ( 7), the derivation of (6a) proceeds as follows.The head (F°) is a probe by virtue of having an unvalued nominal (u-nom) feature and the DP subject ʔal-muʕlimuuna "the teachers" is a goal. 8In a probe-goal configuration, the Functional head establishes an Agree relation with the subject that values F's unvlaued nominal feature.
On the other hand, (6b), which is a verbal sentence because it starts with a verb, will have the derivation demonstrated in (8). 9 (8) The schematic representation in (8) reveals that the VSO structure is obtained because the Functional head (F°) contains a strong unvalued verbal (u-vrb) feature.The verb HaDara "attended" undergoes two successive movements as the dotted show.First, it head-moves from V to T to satisfy the tense feature on (T°), the head of TP.Second, and in a similar fashion of V to T movement, the verb HaDara "attended" moves from T to F to satisfy the verbal feature on the head F.
The last issue to be addressed here is the full vs. partial agreement displayed by SA as (6a) and (6b) show, respectively.In SV construction (6a), where full agreement obtains between the subject and the verb, I assume that the subject values the unvalued phi (u-ɸ) features on the verb.The phi features are morphologically realized as a suffix (appear in bold, see (6a)).In VS structure, as shown in (6b), the number feature on the verb is invariant; it is always singular irrespective of the subject following the verb.Here are more examples that illustrate this point.Post-verbal subjects are singular in (9a), dual in (9b), and plural in (9c).Despite this, the verb does not inflect for number; it has a singular default number feature.
On the other hand, if the order in (9a-c) is reversed, meaning the subject precedes the verb, the verb will inflect for the number.Based on these observations, the partial agreement in VS construction can be attributed to the fact that the F, the head of FP, bears, in addition to the unvalued verbal feature (u-vrb), a default singular number feature that overrides the number feature already assigned to the verb.In (6b), there appears to be an overriding agreement; the verb agrees with the plural subject first, resulting in the suffix plural -u attached to the verb, but then, as a result of movement, the verb ends up in a new position (F), where it receives another number feature, the singular number.Therefore, (6b), which is a typical example of partial agreement, will have the derivation along the lines in ( 11). (11) It is worth noting that Arabic is a null subject language whereby the subject, under certain pragmatic situations, can be omitted.Consider (6a) repeated below as (12a) and its null subject counter example in (12b).At first glance, (12b) seems to cast doubt on our analysis, which assumes that in VS pattern, the verb always carries a singular number feature regardless of the subject following it.In (12b), however, the data clearly show that the verb inflects for plural number, which is morphologically spelled out as -u.In fact, the main reason for why the verb in (12b) inflects for number is the fact that null subject clauses lack a Functional Phrase projection.To illustrate this point, null subject clauses are always verbal because they lack a subject, so the necessity for a Functional Phrase, which is responsible for SV/VS alternation, is eliminated.Therefore, with the absence of a Functional Phrase whose head always has a singular number feature, the verb in (12b) is able to inflect for plural number.Now, I shall discuss how the verb values its unvalued phi-features.I assume that in null subject clauses there must be a covert (i.e.invisible = Ø) subject in Spec TP corresponding to an already established overt subject.Further, I suggest that the covert subject transmits its phi-features to a null pro in Spec, VP.In support of my hypothesis, I note that if the phi-features transmitted by the covert subject to the null pro do not match the phi-features already established in the discourse by the overt subject, the resulting sentence is ruled out.Consider (12a), repeated below as (13a), and its null subject counterpart scenario (13b).Based on this, the derivation of (12b) proceeds as follows.T, being a probe, establishes an Agree relation with a null pro, which gets its phi-features from a covert subject in Spec T (shown by the dotted lines in ( 14)).A null pro values T's unvalued phi-features, and T, in return, values an unvalued case of the null pro.This being the case, I claim, following incorporation analysis (cf. Fassi Al-Horais, 2009;Fass Fehri 1993;Holmberg 2005Holmberg , 2010;;Platzack 2004;Roberts, 2007) that the null pro, which is in Spec, VP, incorporates with the verb in T, forming one chain.The phi-features of the null pro surface on the verb as a suffix -u.Based on this premise, (12b) will have the derivation in ( 14). ( To sum up, the current proposal seems to be not only brings SV/VS alternation and agreement asymmetry associated with it into one unified analysis but also accounts for situations whereby the subject is null.Unlike AlShammiry's (2016) model, the new version is shown to be more parsimonious (in line with Chomsky's simplicity and economy principles).I do not discuss word order variation in dialectal Arabic due to space and time constraints.In the future, it is suggested that this analysis be extended to data in spoken varieties of Arabic.

Conclusion
In this article, I have provided a new analysis, based on AlShammiry's ( 2016) assumption, to one of the most controversial topics in Arabic generative grammar literature, namely word order alternation and agreement mismatch in Standard Arabic.There are two common word order permutations in SA: SVO and VSO.It is shown that, in SVO configuration, the verb agrees in all phi-features (person, number, and gender) with the pre-verbal subject.On the other hand, in VSO order, the verb only agrees with the subject in gender and person but not in number, thereby obtaining defective (i.e.partial) agreement.It is shown that the widely held assumption (cf.among others, Alahdal, 2018;Al-Humari, 2023;Fakih 2016, Mohammad 1990, 2000;Soltan 2006Soltan , 2007) ) that ties word order variation to the presence or absence of EPP feature on T is unworkable because it violates the main merit of "no optionality" inspired by Chomsky (MP;1993, 1995 andsubsequent work).As such, we offered a fresh account that is influenced by a nominal-verbal distinction put forward by Basri's scholars and recently adopted by AlShammiry (2016).Under the nominal-verbal distinction-based assumption, SV/VS alternation is determined by the type of the feature on the Functional head (F).When the head (F) has an unvalued nominal feature (u-nom), the obtained order is SV.On the other hand, if (F) bears an unvalued verbal feature (u-vrb), the resultant sentence would be VS.With respect to agreement mismatch, in particular, partial agreement in VS, we proposed that the Functional head (F) is endowed with a default number feature, which is singular.Given this, the verb on F does not inflect for number.This research also accounted for cases where the subject is null.It is assumed that the verb agrees with a pro, which inherits its phifeatures from a covert subject in Spec, TP.The current proposal shown not only brings SV/VS alternation and agreement mismatch into one environment (i.e.unified analysis) but also turns out to be more parsimonious (in line with economy and simplicity principles).

Notes
1. Arabic varieties, however, use SVO as an unmarked word order.See among others, Shlonsky (1997) for Palestinian Arabic;Fassi Fehri (1993) and Aoun et al. (2010) for Moroccan Arabic; Aoun et al. (1994) for Lebanese Arabic; Al-Shawashreh (2016) for Jordanian Arabic. 2. Fassi Fehri (1993) assumes that both analyses are possible.The preverbal subject can be a true subject occupying Spec, TP when it is indefinite, non-specific.On the other hand, the preverbal subject is interpreted as a topic if it is definite or specific.3. Ouhalla (1997) believes that the preverbal subject could be a topic or a focus depending on its interpretation.4. Consider the following examples, where an object and an adverbial phrase can intervene between the verb and the subject (intervening phrases are underlined).i. HaDara bilʔams Zayd-un come.3SGM yesterday Zayd-NOM "Zayd came yesterday."ii.Qaraʔa ʔal-kitaab-a Zayd-un read.3SGMthe-book-ACC Zayd-NOM "Zayd read the book" (Soltan 2007, p. 48). 5. Due to space and time constraints, I will not discuss conceptual and theoretical problems that recent studies face.However, the common problem of recent and earlier studies is that they tie the SV/VS alternation to the presence/absence of the EPP on T, something that goes against the main merit of MP, which assumes "no optionality".6. Basris' classification of verbal and nominal sentences has been widely adopted in modern Arabic linguistics literature (see among others, Fassi Fehri 1993, Plunkett 1993, Hassan 1961, Al-Ghalayyini, 1974).7. AlShammiry uses IP, an old convention that corresponds to TP, a recent convention used in the literature.Thus, I replace IP with TP in (4).
8. Whether the preverbal subject is a true subject occupying Spec, TP, or a topic phrase located in Spec, TopP, this does not invalidate our assumption.However, in this paper I adhere to the view that locates the pre-verbal subject in Spec, TP. 9.The dotted lines in (8) show the path of verb movement (V-T-F).10.The number features suffixed to the verb are bolded in (10b-c).However, in (10a), the masculine singular number feature is infused with the verb, rendering it morphologically null.11.To distinguish NOM, which stands for nominative case from nom, which stands for nominal, I gloss the former in small caps.

'
Teachers attended the meeting.'
Section 3 offers a new account based on a distinction between nominal and verbal sentences in SA, thus I call it a nominal-verbal distinction-based analysis.This section is split into two parts.The first part discusses the first version of a nominal-verbal distinction-based analysis proposed by AlShammiry (2016).The second part offers an alternative analysis, a modified version of AlShammiry's (2016) model.Section 4 concludes.