Clause-initial negators that reject aspects of a prior utterance rather than simply reversing its truth conditions are widely attested (Horn 1985; Larrivée 2018; Martins 2021; Tsiakmakis & Espinal 2022). In Portuguese, for example, the fixed expression uma ova ‘like hell’ introduces an objection at the clausal left edge without contributing truth-functional negation (1).

Following Horn (1985), such uses instantiate metalinguistic negation (MN): the speaker uses negation to object to a prior utterance or utterance type, including its presuppositions, scalar implicatures, discourse appropriateness, or choice of expression. I do not take this contrast to mean that ordinary/internal negation is merely a simple truth-functional inversion in Horn’s sense. Rather, the relevant distinction is between ordinary/internal negation, which contributes negative content inside the clause, and marked/external uses in which negation targets the assertability or appropriateness of a prior move. A central question is whether MN is reducible to a special pragmatic deployment of ordinary negation (Iwata 1998; Yoshimura 2013), whether it reflects a distinct clause-edge operator in the left periphery (Larrivée 2018; Martins 2021; D’Antuono 2024), or whether it belongs to a broader semantic-distributional distinction between types of negation (Bar-Asher Siegal 2026). The approach adopted here is structural, but it is compatible with the latter view: MN/objection is encoded by a high operator, with its characteristic distributional and polarity profile following from its height in the clausal spine.

In what follows, I use objection as a structural cover term for clause-edge rejection of utterance-level commitments hosted in ObjectionP (ObjP). This is the single structural category assumed in the analysis. Within that structural domain, I distinguish MN proper from denial as interpretive subtypes rather than as separate syntactic heads. MN proper refers to the corrective/scalar subtype that favors, and in many contexts effectively requires, an overt rectifier (“not X, Y”) (Horn 1985; Repp 2009; Pitts 2011; Martins 2020). Denial is used here for utterance-level rejection that need not impose an overt rectifying alternative, including presupposition-cancelling and proposition-rejecting cases. Thus, the analysis is partly unifying and partly distinguishing: it unifies MN proper and denial structurally under Obj, while preserving the descriptive contrast between repair-enforcing corrections and denial cases in which repair may remain implicit.

Three empirical diagnostics anchor the discussion of Jordanian Arabic (JA). First, clause-initial miš/mū supports corrective rejection, naturally pairing with a rectifying continuation (2). Second, miš/mū appears in scalar environments, rejecting a weaker predicate and forcing a stronger alternative (3).

Third, negative-sensitive items such as wala ‘not even one’ and ʕumr- ‘ever’ provide a baseline for the descriptive system: they occur under ordinary negation and track negative concord/polarity licensing, but they do not behave as independent high objection operators. Together, these diagnostics indicate that clause-edge miš/mū under the objection/corrective construal is not simply truth-functional negation: it is restricted in embedding and, crucially, it fails to license negative concord and NPIs.

Empirically, I show that JA exhibits a structural split between (i) descriptive negation in the lower clausal/predicational domain (verbal ma…-š; miš/mū in nonverbal predication), with wala-phrases and ʕumr- behaving as NC/polarity items, and (ii) objection/metalinguistic negation realized clause-initially (miš/mū), with a stable cluster of polarity and embedding diagnostics.1 Theoretically, I argue that MN in JA is encoded in the left periphery rather than arising as a purely pragmatic overlay. I adopt a cartographic architecture (Laka 1990; Rizzi 1997) and draw on recent work on high/illocutionary negation (Martins 2021; Tsiakmakis & Espinal 2022; D’Antuono 2024), extending that map with Semitic evidence (cf. Bar-Asher Siegal 2015; Bar-Asher Siegal & De Clercq 2019).

The rest of the paper is organized as follows. Section 2 reviews MN and denial and develops the diagnostics used here. Section 3 applies these diagnostics to JA, establishing the descriptive baseline and the MN profile. Section 4 proposes the analysis, locating descriptive NegP inside TP and ObjP above ForceP, and derives the observed polarity, ellipsis, and embedding patterns. Section 5 concludes.

Horn (1985) distinguishes ordinary/internal negation from metalinguistic (“marked/external”) negation. The former contributes negative content within the clause, whereas the latter targets the assertability or appropriateness of a prior utterance or utterance type, including its presuppositions, implicatures, discourse commitments, register, or lexical choice, and canonically invites repair (“Not A, (but) B”). Subsequent work situates MN within a broader family of denial and objection phenomena, but the term denial has not been used uniformly in the literature. In some approaches, denial refers broadly to rejection of a prior discourse move; in others, it is restricted to the rejection of a proposition or presupposition. To avoid this ambiguity, I use objection for the structural cover category and reserve denial for utterance-level rejection that need not impose an overt rectifying alternative. MN proper, by contrast, is the corrective/scalar subtype characterized by correction or rectification. Thus, throughout this paper, denial and MN proper are not separate syntactic heads; they are interpretive subtypes within the same high Obj domain. Crucially, both MN proper and presupposition-cancelling denial can be realized by clause-initial miš/mū in JA, and both motivate a CP/speech-act-edge analysis via the shared height diagnostics developed below. To make the typology concrete, it helps to juxtapose denial, MN proper, and plain descriptive negation in a minimal discourse set. Larrivée’s (2018) contrasts are representative (4).

A central line of work has refined Horn’s dichotomy by identifying diagnostics that distinguish MN/high negation from ordinary propositional negation (Larrivée 2012). Horn explicitly ties metalinguistic negation to objection to a previous utterance, not a proposition, and highlights its role in cancelling non-at-issue content (including presuppositions) and in removing scalar upper-bounding implicatures (e.g., It isn’t warm, it’s hot). Larrivée (2018) develops this into an analysis of MN as a corrective speech-act operator tightly linked to information structure (in particular, contrast and the availability of salient alternatives). Moeschler (2018) complements this perspective by proposing semantic/pragmatic criteria that separate descriptive vs. metalinguistic construals (including how MN interacts with entailments, conversational effects, and contrastive continuation patterns), supporting the idea that MN is not identifiable merely by “a corrective feel,” but by a cluster of interpretive and distributional properties that repeatedly track clause-edge behavior across languages. These diagnostics are naturally captured if the grammar distinguishes a low, TP-internal negation head from a higher operator in the left periphery. Here I treat clause-edge objection/denial (hosted in ObjP) as distinct from low truth-functional negation (Neg in TP), in the sense that the high operator targets utterance-level commitments and non-at-issue content rather than simply reversing truth conditions (Horn 1985; Larrivée 2018).

The literature divides broadly into (i) semantic/pragmatic accounts that treat MN as a special use of ordinary negation (e.g., echoic/interpretive approaches) and (ii) syntactic/cartographic accounts that derive MN’s distributional profile from structural height in the clausal spine. Pragmatic approaches capture the speaker-oriented character of MN, but they often leave underexplained why MN so robustly correlates with clause-edge placement, root restrictions, concord failure, and tight coupling to focus-driven correction. Cartographic approaches treat these correlations as structural: a high operator merges in the CP-field (or a speech-act–related layer of the left periphery), outside the TP domain in which propositional negation and concord are computed. I adopt this structural-height approach as the backbone below, while using the denial literature to keep the semantic/pragmatic space explicit. Section 3 applies the diagnostics to JA, and Section 4 implements the height split in the language-specific analysis.

A third position is developed by Bar-Asher Siegal (2026), who argues that the relevant contrast is not exhausted by a choice between pragmatic use and syntactic position. On his account, natural language distinguishes two types of negation, NEG1 and NEG2. NEG1 is internal/predicate-oriented: it takes a predicate as input and contributes negative information about the topic of the sentence. NEG2, which includes external negation, is proposition-oriented: it takes the root proposition as input and comments on its truth value. Crucially, this distinction is both semantic and distributional. It is supported by environments in which PPIs can scope under negation, by German light negation, and by languages such as Jewish Babylonian Aramaic and Sicilian in which historically external negators develop distinct forms and high positions. The present analysis of JA is compatible with this third approach: the proposed Obj head is not simply an extra syntactic label, but a structural implementation of a high, utterance-level negative operator whose distribution differs from low descriptive Neg.

Independent cross-linguistic work supports the existence of negation-related operators above TP. Martins (2021) argues that MN markers are a main-clause phenomenon and gives clear clause-edge configurations in which the MN marker precedes a complementizer in Romance, consistent with a left-peripheral locus for the relevant operator (and, importantly, incompatible with treating it as a TP-internal negator). For example, French mon œil ‘like hell’ occurs at the left edge, immediately before que (5).

High negation-related operators are independently attested beyond MN narrowly construed, including in analyses of external, emphatic, taboo, and demonic negation. I do not treat these as instances of MN proper. They are relevant here only as evidence that negation-related material may occupy positions above low TP-internal NegP. A fuller comparison with expletive, taboo, and demonic negation is left to future work.

De Clercq’s (2020) nanosyntactic account is relevant here because it argues that negative markers are not structurally uniform: they can be decomposed into smaller features and associated with different scope positions. In particular, De Clercq proposes a fine-grained classification of negative markers and argues that negative markers of different scopal types can be treated syntactically, with multiple negative projections available in the clausal spine. I do not adopt the full nanosyntactic machinery here, but the central point supports the present analysis: positing a distinct high position for an objectional negative marker is not ad hoc if negative markers independently vary in structural size, scope, and position.

The negative concord literature supplies a useful diagnostic partition. In strict NC systems, n-words typically require co-occurrence with an overt clausal negator and are licensed in the low clausal domain. I assume concord is implemented via Agree with a [neg] feature on low Neg⁰; high Obj (the head of ObjP) lacks that feature and therefore cannot license concord. If a negation-related operator merges outside the TP/NegP concord domain, whether an MN/objection head or an independently attested taboo/demonic negator, it is expected to fail the licensing profile characteristic of strict concord. This is the structural core of the “height diagnosis”: concord failure is not an accident of lexical idiosyncrasy, but a consequence of merging outside the concord domain.

A cartographic implementation makes these dependencies explicit. I assume a split CP in which the left periphery contains Force and an articulated topic/focus field (cf. Laka 1990; Rizzi 1997), and I adopt a speech-act–compatible extension in which stance/commitment operators occupy a high region of the C-domain (Speas & Tenny 2003; Hill 2007; Haegeman & Hill 2013; Tsiakmakis & Espinal 2022). The cartography assumed here places the objection operator above ForceP, as a dedicated head that selects a ForceP complement (6).

On this view, the core diagnostics follow from height. Root restriction follows because the high operator is anchored in the left periphery associated with clause typing and speech-act structure: embedding under complementizers and attitudes is either unavailable or constrained under the intended objection/denial construal. Concord and NPI licensing failures follow because the high operator does not enter into the local configuration in which concord is computed (and thus cannot serve as the licensor for n-words or other polarity items in strict NC systems). Moreover, the frequently reported apparent “double negation” sequences are analyzed as structural layering, while their final interpretation may also yield truth-functional cancellation: a high objection operator can target a prior negative utterance type, while a low negator contributes ordinary ¬p inside TP. This layering can be represented schematically as in (7). Finally, focus association is not incidental: the corrective profile of MN proper typically requires a salient alternative set and therefore interacts closely with the focus field.

Despite advances, three gaps remain. First, detailed empirical and analytical progress is still skewed toward Indo-European, although relevant work now exists for several non-Indo-European languages, including Japanese (Yoshimura 2013), Korean (Yoon 2026), Malay/Indonesian (Kroeger 2014), Egyptian Arabic (Mughazy 2003; 2008), and Jewish Babylonian Aramaic (Bar-Asher Siegal 2015; Bar-Asher Siegal & De Clercq 2019). Semitic data nevertheless remain comparatively sparse, especially with respect to the full diagnostic cluster used here. Second, standard diagnostics for MN (negative concord behavior, clause-type restrictions, embedding profile, and focus association) have not been systematically applied across Arabic varieties, where multiple negative strategies and negative-sensitive items interact. Third, there is no consensus on how to locate objectional operators within the Arabic left periphery, specifically, whether they merge above Force or lower in the C-field. The analysis pursued here adopts a structural stratification view in which the key contrast is height (Obj vs. Neg). Work on Semitic argues independently for external/high negation that targets discourse commitments rather than truth conditions (Bar-Asher Siegal 2015; Bar-Asher Siegal & De Clercq 2019), providing a semantic and cartographic precedent for the split argued for here (see also Lucas 2007 on the diachrony and morphosyntactic variation of Arabic negation; Mughazy 2008 on denial and information structure in Egyptian Arabic).

Negation in Arabic is often described in terms of a descriptive system, in which verbal clauses are negated with ma…-š and nonverbal predicates with miš/mū. Arabic also has a rich inventory of negative-sensitive items (e.g., wala-phrases, ʕumr- ‘ever’) that interact with negative concord (e.g., Alqassas 2015; Al Khalaf 2017; Al Khalaf & Yasin 2026). By contrast, MN, understood since Horn (1985) as an objection to a prior utterance rather than to its propositional content, has received much less explicit attention in the Arabic negation literature. Cross-linguistically, MN shows a stable cluster of diagnostics that distinguish it from truth-functional negation (e.g., Japanese: Yoshimura 2013; French: Larrivée 2018; Romance: Martins 2021), and related work on external/high negation and denial converges on the same space of phenomena (van der Sandt 1991; Geurts 1998; Repp 2009; Kroeger 2014; Lee 2015; 2016).

Within Arabic, one of the few sustained treatments is Mughazy (2003), who argues that Egyptian Arabic (EA) overtly encodes the descriptive vs. metalinguistic split: discontinuous ma…-š expresses ordinary propositional negation, whereas continuous miš supports MN readings. Mughazy highlights three diagnostics that are particularly clear in EA: (i) MN fails to combine with emphatic oath particles (w-alla:hi), (ii) MN supports “double negation” layering, and (iii) MN is systematically tied to rectification and contrastive prominence. An illustrative case comes from scalar/cardinal environments: EA allows continuous negation with a rectifier (… meš …; …) but resists adding the oath particle in the same MN configuration. This EA baseline matters here because Jordanian Arabic (JA) reproduces the same diagnostic profile under clause-edge miš/mū.

The present paper does not claim that JA and EA differ fundamentally in the core diagnostics discussed here. Rather, JA patterns with EA in allowing clause-edge miš/mū-type negation in corrective and presupposition-cancelling environments, in resisting negative-concord/NPI licensing under the objectional construal, and in showing sensitivity to oath/emphatic contexts. The contribution of the present paper is therefore not that JA diverges radically from EA, but that JA independently exhibits the same high-negation profile and allows the contrast to be formalized as a height split between low Neg and high Obj. Where no difference between JA and EA has been identified, I treat the two varieties as converging in the relevant respect.

JA patterns with EA most transparently in correction. In (8a), the negation does not deny the truth of the proposition ‘I am the actor’; instead, it rejects the choice of ʔanā as the relevant subject expression and replaces it with huwwe. Utterances like (8a) can be coerced into a truth-functional denial (‘it’s false that it’s me’), but the intended point is narrower: in MN proper the negation targets an utterance-level commitment (reference/choice of expression) and comes with a privileged alternative that repairs that commitment (Horn 1985; Repp 2009). What triggers the metalinguistic construal is not the presence of a salient prejacent (which can also hold under descriptive negation), but the requirement or strong preference for rectification. As Horn (1985) emphasizes for English Not John, Bill, corrective/scalar MN typically requires a repair continuation; in JA, dropping the rectifier in fragment corrections like (8a) yields a pragmatically incomplete result under the MN-proper parse. Fully clausal paraphrases are available and behave the same way (8b); the fragment shape is expected because correction naturally licenses ellipsis.

MN also appears in scalar environments, as in (9). Here the speaker does not deny the truth of sakhin ‘warm’; rather, they object to an inappropriately weak predicate. This is the kind of strengthening Horn (1985: 127) and Yoshimura (2013) treat as diagnostic of MN: the rectifier asserts a stronger alternative, cancelling a scalar implicature rather than negating a proposition. This fits Moeschler’s (2018) discussion of MN as operating over entailments and scale structure. The familiar garden-path effect (not pretty … she’s beautiful) is compatible with this: a low descriptive parse is initially available, but is overridden once the rectifier forces a metalinguistic construal (Horn 1985; Nedwick 2014; Larrivée 2018).

JA miš can also target presupposed content under an appropriate context, as in (10). The objection here targets the presupposition that the person who arrived was her husband. This parallels Horn’s (1985) classic pattern (e.g., The king of France is not bald) and EA cases where miš cancels presupposed content (Mughazy 2003). The convergence across correction, scalar strengthening, and presupposition cancellation supports treating clause-initial miš in JA as a genuine MN/objection operator. At the same time, presupposition-cancelling denial differs from MN proper in one respect that becomes important below: it does not necessarily require an overt rectifier (“not X, Y”). Rectification is robustly obligatory in corrective/scalar MN proper, but denial can be satisfied by an implicit repair or by a follow-up that need not focus a contrastive alternative (van der Sandt 1991; Geurts 1998; Repp 2009; Larrivée 2018).

Context: The addressee has just said/assumed that the person who arrived was her husband.

Another diagnostic comes from negative concord / polarity licensing. In JA, argumental wala ḥada and ʕumr belong to the negative-sensitive inventory of the descriptive negation system. With TP-internal ma…-š, both are acceptable under the intended NPI/NCI construal, as in (11). Under clause-edge miš with an MN-proper corrective construal, the same items are degraded, as in (12). Because the negative-sensitive expression is held constant across the two environments, the contrast tracks negation height (Neg vs. Obj), not properties of the item itself. This sets up the strict-NC expectation: in strict concord systems, n-words are licensed in the low clausal domain by a clausemate TP/NegP-internal negator (Zeijlstra 2004; Giannakidou 2006; Larrivée 2021). I take concord and NPI licensing to involve Agree with Neg⁰ in NegP (see also Al Khalaf & Yasin 2026); Obj lacks the relevant features and therefore cannot license wala ḥada or ʕumr even under c-command. Mughazy (2003) reports the same split for EA: MN-miš fails to license negative-sensitive items, unlike ma…-š. JA shows the same pattern, and (11) illustrates the expected availability of NC/NPI licensing under ma…-š.

Speakers judge (12) unacceptable under the corrective reading. This mirrors EA, where MN blocks ʔabadan, and is consistent with the broader claim that MN lies outside the licensing domain of polarity items (Larrivée 2018; Martins 2021). The contrast also helps separate MN from superficially similar strategies in Arabic, since wala-phrases and ʕumr- are better analyzed as NPIs/NCIs (Benmamoun 1997; Alsarayreh 2012; Alqassas 2015; Al Khalaf 2017), not as independent negators.

To provide an even tighter minimal pair, clause-edge miš still fails to license the same argumental wala ḥada in a simple clause (13).

(Wala-phrases also have minimizer uses in some varieties, but those are independent of the concord-licensing diagnostic exploited here; the test cases above keep wala ḥada in its argumental ‘(even) one person’ use.)

Apparent “double negation” provides further evidence for structural layering, but the point must be formulated carefully. In EA, MN allows sequences like miš ma…-š because the higher operator can target a prior negative utterance while the lower negator contributes descriptive negation (Mughazy 2003). JA behaves similarly, as in (14). Truth-conditionally, the final interpretation may yield the familiar cancellation inference: rejecting “I did not drink” is compatible with “I drank.” The diagnostic point is therefore not that cancellation is absent, but that the surface sequence shows two structurally distinct negative operators: a high objection operator and a low descriptive negator (Chung 2007; Collins 2018; De Clercq & Vanden Wyngaerd 2019; De Clercq 2020).

Embedding restrictions reinforce this view. Moeschler (2018) and Martins (2021) note that MN is speech-act-oriented and therefore typically unavailable under embedding. MN in EA is degraded under attitude predicates, and JA shows the same pattern. Speakers judge (15) unacceptable under the intended corrective reading, consistent with the claim that MN in JA is root-only, as expected if the relevant operator merges in the CP/speech-act layer (Larrivée 2018; Martins 2021). By contrast, apparent embedding improves under direct quotation/echo report, which is expected and does not undermine root restriction. The contrast between (15) and (16) indicates that what is root-restricted is the MN-proper repair configuration (ObjP), not the availability of lower miš as descriptive negation in embedded clauses.

Another diagnostic emphasized by Mughazy (2003) is MN’s incompatibility with emphatic oath environments. In EA, adding an oath particle is acceptable with truth-functional negation (discontinuous ma…-š), but it becomes infelicitous in the relevant MN configuration with continuous miš. JA shows the same pattern in (17)–(18): speakers strongly prefer wallāhi ma…-š for emphatic descriptive denial, as in (18), while wallāhi miš is degraded under the MN-proper corrective parse. The question mark in (17) marks degradation under the intended corrective construal, not an absolute ban on the string in all possible contexts. This contrast supports the view that MN is not “emphatic negation” (contra pragmatic-strengthening approaches) but a distinct corrective operator. The incompatibility does not plausibly reduce to linear adjacency; one natural hypothesis is that wallāhi itself anchors (or activates) a high commitment operator in the speech-act layer, leaving no available slot for Obj under the MN parse.

Finally, information structure further supports the analysis. As Larrivée (2018) shows for French, MN is systematically tied to focus: the rejected constituent is discourse-old and is rectified by a focused alternative. JA shows the same focus dependency in fragment and clausal corrections. I do not build an argument from connective choice here, since JA bass can appear in both fragmentary and fuller corrective continuations. The relevant diagnostic is therefore not the connective itself, but the presence of a focused rejectee and an overt rectifying alternative.

Thus, the cumulative evidence shows that JA, like EA, encodes clause-edge objection/denial, including the corrective/scalar subtype (MN proper). Clause-initial miš/mū behaves as predicted by Horn’s (1985) diagnostics and subsequent refinements (Iwata 1998; Larrivée 2018; Moeschler 2018; Martins 2021): it targets utterance-level commitments rather than simply contributing low propositional negation, cancels implicatures and presuppositions, resists NPI licensing, permits structurally layered “double negation” configurations, is root-only under the MN-proper parse, is degraded with emphatic oath contexts under the MN parse, and requires focus and overt rectification in MN-proper cases.

Crucially, MN must be kept apart from negative-sensitive expressions such as wala-phrases and ʕumr-, which belong to the descriptive negation system and show the concord/polarity behavior familiar from the Arabic literature. Establishing this split for JA adds independent JA evidence to the Arabic empirical record, complements the EA facts, and adds a Semitic case to the broader typology of clause-edge negation-related operators. The diagnostics line up with CP-edge analyses of MN/objection (Larrivée 2018; Martins 2021) and with speech-act cartography that accommodates high operators in the left periphery (Tsiakmakis & Espinal 2022). The claim defended here, under the objection/denial construal, is that JA miš/mū is restricted to the clausal left edge, and the ObjP analysis is designed to derive precisely that rigidity.

Section 3 motivates a diagnostic cluster for clause-initial miš/mū in Jordanian Arabic (JA): it licenses rectificatory continuations (“not X, Y”), supports scalar strengthening and presupposition cancellation, is root-only under the objectional construal, fails to license wala-phrases and other polarity items, permits structurally layered “double negation” with lower ma…-š, and systematically associates with contrastive focus and repair sequencing. At the same time, wala-phrases and ʕumr- ‘ever’ pattern as NPI/NC-sensitive elements in the descriptive system, not as independent high negators. These facts argue for a height-based split between (i) a low negation head that participates in ordinary propositional negation and polarity licensing and (ii) a CP-edge objection/denial operator (Obj) that hosts MN-proper as a subtype. (I use Obj as a mnemonic label for the higher operator; nothing hinges on the label itself.)

This split does not require collapsing all metalinguistic uses into a single interpretive subtype. Within the Obj domain, JA distinguishes at least two interpretive subtypes: MN proper (corrective/scalar) and denial (including presupposition-cancelling and proposition-rejecting uses). They share the same structural height but differ in whether an overt rectifier is required. Preserving this subtype contrast (cf. Larrivée 2018) allows the analysis to keep the core structural claim (Neg vs. Obj) while capturing the asymmetry between “repair-enforcing” corrections and denial cases where repair can remain implicit.

I adopt a cartographic analysis that combines CP-edge accounts of MN/objection (Larrivée 2018; Martins 2021) with speech-act cartography that permits speaker-oriented operators in the left periphery (Speas & Tenny 2003; Hill 2007; Haegeman & Hill 2013; Tsiakmakis & Espinal 2022). The goal is to derive the Section 3 diagnostic cluster from structural height (Obj vs. low Neg) and its interaction with the focus field, without reclassifying independent clause-edge taboo/emphatic negators in other languages as MN.

This paper has argued that Jordanian Arabic distinguishes truth-functional negation from a clause-edge objection operator realized by clause-initial miš/mū. The evidence comes from a convergent diagnostic profile: in corrective and scalar configurations, miš/mū requires (or strongly favors) an overt rectifier (miš X, Y), yields presupposition-cancelling readings, and is restricted to root environments. In the same objectional construal, miš/mū fails to license negative concord and NPIs, including argumental wala-phrases and ʕumr-, despite c-command. Finally, miš/mū can occur outside a TP containing low ma…-š, producing the familiar “double negation” surface pattern. Such examples may yield truth-conditional cancellation in the final interpretation, but their syntactic significance lies in structural layering between high Obj and low Neg. The analysis captures these properties by locating objectional miš/mū in Obj⁰ above ForceP, outside the TP/NegP domain where concord and ordinary polarity licensing are computed. Descriptive negation remains low (Neg⁰ inside TP), realized by ma…-š in verbal clauses and by miš/mū in nonverbal predication. The apparent syncretism of miš/mū is therefore analyzed as height-sensitive spell-out at the structural and distributional levels. This formulation does not by itself explain the lexical-semantic relation between the low and high uses of miš/mū, which remains an open question, as acknowledged in Section 4.1. Within the high domain, the paper maintains a distinction between MN-proper and denial: MN-proper introduces a repair requirement and contrastive prominence (implemented here via an active [uFoc] dependency plus paratactic ForceP sequencing), whereas denial may lack an overt rectifier even though it shares the same structural height and the same polarity/embedding diagnostics.

The analysis also clarifies the status of ObjP. Obj is not intended as a head whose only possible function is JA metalinguistic negation narrowly construed. It is the structural position in which JA realizes clause-edge objection to prior commitments. The present paper has focused on MN proper and presupposition-cancelling/proposition-rejecting denial because these are the contexts in which the diagnostics are clearest. Future work should examine whether the same high region is implicated in other polarity-reversal or response phenomena, including biased interrogatives and particles comparable to French si. This is a natural extension of the analysis rather than a premise on which the present argument depends.

Beyond JA, the results support two broader claims. First, the Arabic left periphery is articulated enough to host a commitment/objection head above Force, in line with cartographic approaches that locate certain discourse-related operators in the CP/speech-act region. Second, the JA pattern strengthens the view that negative concord is local to low clausal structure: a clause-edge negation-related operator can contribute denial/objection while remaining invisible to concord and NPI licensing. By adding independent JA evidence to the Arabic empirical record, complementing the Egyptian Arabic facts, the paper provides a Semitic case in which the full cluster—root restriction, concord failure, repair sequencing, and stacking with low negation—tracks a single height contrast between Obj and Neg. Several questions remain open. Comparative work across Levantine and Egyptian varieties can test how stable the high/low split is when the exponent is formally identical (miš/mū), and whether the repair requirement correlates with prosodic or connective strategies (e.g., bass, bal) across dialects. More generally, the interaction of Obj with clause type (questions, biased wh, and other Force-related environments) should help refine its placement relative to other CP-edge heads and clarify the extent to which objection in Arabic is packaged as a dedicated syntactic operator rather than as a pragmatic construal of low negation.