On low topics in Najdi Arabic: a rejoinder to Alshamari and Jarrah (2022)

1 Introduction

Since at least Rizzi’s (1997) seminal work, an enormous body of literature has accumulated to maintain the claim that what was formerly known as (C)omplementizer (P)hrase (CP) in the left periphery of the clause should be split into a strict cascade of functional projections, each with a distinct discursive function (i.e., topicality, focushood); see Rizzi and Cinque (2016) for a semi-exhaustive, cross-linguistic list of works capitalizing on this cartographic approach to information structure. According to this theoretical development, the fixed cascade of discourse features looks like the representation in (1).

According to this representation, CP is decomposed into a number of projections: while FocP denoting focused elements is sandwiched between two topic projections, ForceP (i.e., elements marking the illocutionary force of the sentence) and FinP (i.e., elements encoding the [non]-finiteness of the sentence) are projected at the two ends of the continuum. While the representation in (1) has undergone several changes and modifications since 1997, it remains the epitome of the cartographic approach to information structure (see Rizzi and Bocci [2017] for an updated version of left-peripheral cascades of projections). In reaction to Rizzi (1997), many authors have voiced concerns about the restrictiveness of the Rizzian model because there are languages that make use of discourse features within the area of IP (Belletti 2001, 2004; Cecchetto 1999; Villalba 2000). The original observation is that the rigid template argued to apply to the left periphery can be extented to the low IP, or the “centre periphery” in terms of Camacho (2003). As per this template, the following representation holds.

Against the backdrop of this historical context, Alshamari and Jarrah (2022), echoing Jarrah and Abusalim (2021) for Jordanian Arabic (JA), argue in favor of discourse-oriented projections between IP and vP in Najdi Arabic (NA), in accordance with the low IP desiderata. Although illuminating, we argue in this article that there are somber prospects for the low IP analysis when applied to right dislocation in NA and JA, and that an IP-external analysis is warranted here.

2 Right dislocation in Najdi Arabic and Jordanian Arabic

Najdi Arabic and Jordanian Arabic exhibit what is termed a “Clitic-Object Construction” as illustrated in (3). JA, unlike NA, also has a construction dubbed “Object-Clitic Doubling” as shown in (4) (Alshamari and Jarrah 2022).^[1]

As correctly pointed out by Alshamari and Jarrah (2022: 17), sentences of the kind illustrated in (3) are similar to Clitic Right Dislocation (CLRD). However, Alshamari and Jarrah (2022) do not provide an explicit account along these lines, lumping together the constructions in (3) and (4). Jarrah and Abusalim (2021) remain silent on this matter altogether. Yet a thread running through the literature on clitic-resumed constructions is that there is a dichotomy between (3) and (4) (Samek-Lodovici 2015). The former is often regarded as a case of CLRD, while the latter is taken to be an example of Clitic Doubling (CD), a construction which is widely attested in Levantive Arabic (including JA).

An anonymous reviewer doubts the validity of  this conclusion, pointing out that the sentence in (4) exemplifies a case of Differential Object Marking (DOM). This is indeed in accordance with much recent works, where it is argued that CD and CLRD in Levantive Arabic do overlap with DOM; see for instance  Zarka (2023). In this article we propose that DOM and CD should be treated as distinct articulations, given the fact that DOM is a multifactorial phenomenon which may involve (in)animacy, definiteness, specificity, agentivity, telicity and affectedness; see especially contributions to Kabatek et al. (2021). What is more, Driemel (2023) shows that DOM is ipso facto a pseudo-noun incorporation in disguise, given the assumption that the two configurations display similar properties in some languages, e.g., insofar as they target the same class of arguments, and can be sensitive to the strong-weak definiteness distinction. As it is beyond the scope of this article to discuss these developments in their entirety, they are left to future research.

Note that CLRD is easily recognizable thanks to a terminological definition of dislocation suggested by Lambrecht (2001: 1050). As per this definition, dislocation is characterized by there being a constituent which typically behaves as an argument within a predicate-argument structure, and can instead occur outside the boundaries of the core IP, either to its left (i.e., Left Dislocation), or to its right (i.e., Right Dislocation). Lambrecht further points out that this definition concomitantly involves four criteria: (i) an extra-clausal position of a constituent, (ii) a possible alternative intra-clausal position, (iii) a pronominal coindexation, and (iv) a special prosody. Although these criteria do not constitute necessary conditions for marking a given configuration as a dislocated one (except for the criterion depicted in (i), which is taken by Lambrecht as constitutive of dislocation), they squarely apply to NA and JA barring the prosody of dislocation. More specifically, a cataphoric argumental clitic is obligatory (as shown by the starred parentheses in [3]), which is simultaneously related to a constituent occurring outside the boundaries of the phrase containing the predicate (i.e., IP), in accordance with the criteria (i), (ii) and (iii).^[2]

This is not the case, however, when we consider CD, which exhibits properties setting it apart from related structures such as CLRD (Anagnostopoulou 2006). In particular, a prime feature of CD is that the clitic can cooccur with the DP argument in an IP-internal position without inducing a violation of the θ-Criterion, as evidenced by (4). What is more, CD is known to comply with the so-called ‘Kayne’s Generalization’ according to which CD should co-occur with a preposition/case marker, which precedes the doubled element. The example in (4) illustrates this point with an example from JA, where the preposition/case marker la immediately precedes the doubled element. A further illustrative feature of CD is that it only instantiates DPs in its derivation, in contrast to CLRD, which exhibits flexibility with respect to employing different kinds of phrases (Samek-Lodovici 2015). Overall, the emerging picture is that CD and CLRD should be seen as distinct phenomena with well-defined differences. For an exhaustive discussion of these differences, see, among others: Alzayid (2022), Anagnostopoulou (1994), Cinque (1990), Fernández-Sánchez (2020), Iatridou (1995), Kechagias (2011), and Rizzi (1986).

3 Right dislocation in Najdi Arabic and Jordanian Arabic is IP-external

Having established the existence of CLRD in NA and JA, we are now in a position to tackle the question of the derivation. One possible analysis is to argue that CLRD (i.e., Clitic Object Construction in Alshamari and Jarrah’s parlance) in NA and JA is IP-internal, an argument which is implicitly advocated by Alshamari and Jarrah (2022: 17), citing Cecchetto (1999). This claim, however, does not stand to scrutiny. For the sake of brevity, we present three pieces of evidence originally due to Samek-Lodovici (2006, 2015), which are taken here as precursors drawing on data from Italian. A first piece of evidence comes from the behavior of (N)egative (P)olarity (I)tems. NPIs are a class of lexical categories which must be in the scope domain of neg-words for the sake of licensing (see Giannakidou [2006], Moscati [2006], Zanuttini [1991, 1997], and Alqassas [2021] for a recent overview of the phenomenon of polarity sensitivity in Arabic). As predicted, NPIs in JA (5a) and NA (5b) are licensed under a structural condition defined in a c-command relation between a neg-word and an NPI (i.e., the NPI must be in the c-command domain of the neg-word).

Yet NPIs cannot be right dislocated in JA and NA, as exemplified in (6).

Barring ad hoc assumptions, this can be accounted for by locating right dislocated elements IP-externally. More specifically, the NPI in (6) does not stand in the c-command domain of the neg-word, giving rise to a failure in establishing the required c-command relation between the neg-word and the NPI. Under clause-internal analyses of RD assuming RD-ed elements to be invariably below I⁰, the failure of NPI licensing is not predicted (Samek-Lodovici 2015).

An anonymous reviewer claims that it is well-established in the literature that NPIs cannot be right-dislocated because such elements cannot be topicalized. This claim does not hold though, given that there are languages where NPIs can be right dislocated. Catalan and Italian are cases in point, as illustrated in (7) and (8).

This points to the fact that the licensing condition of NPIs is syntax-driven, defined in a c-command relation between a licensor and a licensee, which is quite orthogonal to information-structural considerations; see Samek-Lodovici (2010) for the original observation. Note, incidentally, that what is well-established in the literature is that recalcitrant topics, or ‘atypical topics’ in terms of Alzayid (2022), such as quantifiers, indefinites and by extension NPIs, can be dislocated under specific discourse conditions pertinent to specificity. See among others, Cinque (1990), De Cat (2007), Dimitrova-Vulchanova and Hellan (1999), Dobrovie-Sorin (1990), Gutiérrez-Rexach (2000, 2001, 2002), Kallulli (2000), and Suñer (1988).

In contradistinction to MSA, in many Arabic varieties, JA and NA included, full agreement obtains regardless of any word order alternations (Aoun et al. 2010).

What is relevant to the ongoing discussion is that both NA and JA display rich agreement in VOS order, insofar as the verb exhibits full agreement when the subject undergoes right dislocation, militating against the claim that Arabic VOS order displays only partial agreement between the subject and the verb, as advocated by Benmamoun and Lorimor (2006). Note that the subject in (11a) and (12a) is right-dislocated because it follows a VP with a thematic bound form cliticized to the verb which satisfies the valency of the predicate.

Under the assumption that movement to specIP is agreement motivated (Chomsky 2001; Miyagawa 2010), contrary to proposals assuming that movement to specIP is case-triggered (Boeckx 2000; Epstein and Seely 1999), this contrast is explained by assuming that the right-dislocated subject in (11a) and (12a) is not below I⁰ as per the clause-internal analysis, but that it goes through specIP to check agreement features. Another cogent analysis, which is independently advocated by Cardinaletti (2002), is to assume that rich agreement in (11a) and (12a) follows from the presence of a preverbal pronominal pro doubling the dislocated subject, and at the same time establishing an agreement relation with I⁰ (see also Samek-Lodovici [2015] for an analysis along these lines on data from Italian). An alternative straightforward analysis is to simply assume an adjunct status for the subject (and hence an IP-external analysis), implying that there is no room for this element in the skeleton of the core IP, unless one is ready to make stipulations to salvage the θ-Criterion and Chomsky’s (1965) sub-categorization principle. The best path to follow, therefore, is to assume that the dislocated subject occurs in a position external to IP.

An anonymous reviewer points out that the Arabic agreement facts in VOS order are a non-starter with respect to postulating an external-IP analysis. Our claim is guided by the argument that A-movement is motivated by the satisfaction of the EPP (i.e., the idea that specIP must be filled), in contradistinction to case-theoretic accounts; see Levin (2015) and references therein for extensive discussion. If we apply this analytic choice to the present context, the derivation goes as follows: the subject undergoes A-movement to specIP so as to satisfy the EPP condition; this would amount to deriving SVO order, which is known to be the archetypal word order in Arabic triggering subject-verb full agreement (Soltan 2007). For discourse structural reasons, the subject is then right dislocated to an IP-external position. Taken together, the conclusion is that the overt subject is not below I⁰, as predicted by the low-IP analysis. Note, incidentally, that Alshamari and Jarrah (2022) gloss over cases where subjects are right-dislocated and simultaneously co-occur with pro subjects in VOS word order, as illustrated in (11a) and (12a). Still, the lingering question persists: can the low IP analysis be upheld under such a scenario? (see Rizzi [2018] for a recent, interesting take on this issue).

A last piece of evidence speaking in favor of an external-IP analysis comes from the interaction of wh-phrases and foci in JA and NA. By way of illustration, consider the examples in (13).

This behavior is at odds with the predictions of the cartography program, which presupposes that wh-phrases and foci cannot cooccur (Rizzi 1997; Shlonsky 2000). Under an IP-external analysis of RD, this behavior is expected (Samek-Lodovici 2006). In particular, given our assumption that RD-ed elements, flagged here by wh-phrases, are generated IP externally, the wh-phrase is not embedded under another focus (see Eilam [2011] for the claim that, contrary to the widely held assumption in the literature, wh-phrases can be topics under certain conditions). Proponents of the internal-IP analysis of RD thus would have  to explain how the interaction between wh-phrases and foci is possible without inducing ill-formed strings.

Two anonymous reviewers raise a question about the dislocability of the wh-phrase. To approach this question, three comments are in order. First, as argued extensively and convincingly by Samek-Lodovici (2015), clitic doubling is not a condition on a well-formed right dislocation even if there is a suitable one. This essentially precludes the postulation of null clitics; see Cardinaletti (2002) for a strong case against null clitics. Second, since at least Rizzi (1997), the incompatibility of focus and wh-operators is a well-established generalization in the literature. One way to salvage this generalization, pace Samek-Lodovici (2006, 2015), who regards examples of the kind illustrated in (13) as a reason to debunk this generalization in toto, is to reanalyze confounding data in a way which retains the explanatory power of this generalization by assigning a different analysis to the co-occurrence of focus and wh-operators, locating wh-operators IP-externally. On this analysis, the impression of the incompatibility vanishes altogether; see Borise (2023) for a recent treatment where it is proposed that although wh-phrases and focus appear preverbally in Georgian, they should be derived in a different fashion. Third, the implicit assumption of a right-dislocation analysis for wh-operators amounts to furnishing an argument for the in-situ position of sentence-medial focus in Arabic, which happens to be rightmost. The logic goes as follows: sentence-medial focus in Arabic occurs rightmost, with apparent medial focalization on the surface being a consequence of a right-dislocation process. This claim (i.e., topical, right-dislocated wh-phrases as evidence for the in-situ focus) has originally and explicitly been advocated by Samek-Lodovici (2006). On the rightmost foci in general, see Büring (2001), Dehé (2005), and Zubizarreta (1998).

4 Conclusions

Completing our discussion of the plausibility of the low IP area in JA and NA, we conclude by pointing out that the low IP analysis à la Alshamari and Jarrah (2022) and Jarrah and Abusalim (2021) is indeed attractive since it proposes a completely novel analysis of RD in Arabic, where RD-elements are located in an intermediate position in the area between IP and vP. Given the serious drawbacks pointed out in this article, however, their proposal is rejected pending an analysis of how the empirical glitches can be tackled.