A great deal of work over the last two decades has explored strategies of agreement with coordinate structures whose DP-conjuncts have different φ-feature specifications. However, most of these studies have focused on agreement with conjoined subjects. We investigated verbal agreement with conjoined DPs in the Bantu language Zulu, in which object-DPs optionally agree with the verb; this allowed us to compare agreement with conjoined subjects and conjoined objects. Zulu also has a number of interesting properties that constrain the space of possible models of conjunct agreement, including the fact that all agreeing DPs have moved out of their base position. In a four-part experimental syntax study conducted with a large base of speakers, we demonstrate that Zulu patterns with languages such as Slovenian and Bosnian/Croatian/Serbian, in which preverbal agreement allows a greater range of options than postverbal agreement. We show that our results provide evidence in favor of the Distributed Agreement model proposed in earlier work on conjunct agreement, which allows for agreement to be partially established based on the linear relation between the agreement controller and the target at PF.
We present a formal analysis of an understudied postverbal completive marker -le in Mandarin, distinct from the perfective -le. We propose that completive -le forces strict telicity via maximal affectedness of the theme. Building on degree-based approaches, we show that -le interacts with a covert partitive operator to link event progress to both the extent of the theme and an abstract path scale. This analysis explains the marker’s unique referentiality constraints and semantic effects. The findings demonstrate that Mandarin completive markers possess distinct lexical semantics and reinforce the crosslinguistic link between object referentiality and the completion of accomplishments.
The placeholder verb construction, which has emerged relatively recently in colloquial Georgian, offers a unique case study of grammaticalization in progress: the emergence of a structurally innovative type of verb within a language whose verbal morphology is already quite complex. Derived from reanalysis of the event-anaphoric verb phrase ‘do that,’ placeholder verbs exhibit morphological variability which is quite uncharacteristic for the language. We advance a few structural analyses to account for this variation, differing primarily in how the demonstrative pronoun ‘that’ has been reanalyzed: as an incorporated theme; as a functional head of an extant category; as a novel functional head; or a sort of anaphor for a subword morphological constituent. Each analysis accounts for a different subset of the attested patterns, so we hypothesize that multiple grammars of the construction coexist across speakers. Theoretically, this is reasonable state of affairs, since the reanalyzed demonstrative has no unambiguous structural analogue in the standard Georgian verb. Empirically, evidence for the multiple-grammars hypothesis comes from a morphological acceptability study. Participants cluster into two major groups, whose judgement patterns correspond well to predictions of the extant-functional-head and morphological-anaphor analyses. We speculate on why our participants seem not to instantiate the other proposed placeholder-verb grammars, positing certain language-specific and language-general analytical biases relevant to the acquisition of this emergent construction.
While the basic/underlying word order in Chinese languages is SVO, SOV patterns are also attested, and in Mandarin frequently arise for reasons of focus or (inner) topicalization (Ernst and Wang 1995; Paul 2005). In contrast to such optional OV, in many varieties of Wu Chinese an obligatory OV sequencing often occurs, which cannot be attributed to information structure. Instead, the positioning of the object relative to the verb regularly varies according to the interaction of aspect, definiteness, and affectedness. In Jiaxing Wu Chinese, with predicates that are telic, objects that are both definite and affected must occur preverbally (OV), whereas indefinite objects and objects that are definite but unaffected naturally follow the verb (VO). We develop an analysis of the licensing of objects in OV/VO alternations, in which V-Asp and V-V resultative verb constructions are suggested to pattern like impersonal passives in French, with the elimination of structural Accusative case but retention of inherent Partitive case (Belletti 1988). The paper expands the typological description of the crosslinguistic phenomenon of object shift, provides support for the role of differentiated forms of abstract case in a morphologically caseless language, and highlights the role that affectedness may play in morphosyntactic phenomena.
Argument alternations are an empirical mainstay in studying the building blocks of verb meaning and how such meanings link to a verb’s grammatical properties. Event structures are widely adopted for exploring these questions, where a verb’s meaning is decomposed into an event template (defining grammatically coherent verb classes) and a root (filling in idiosyncratic verb-by-verb details). An alternation is analyzed as one root occurring in two templates, the latter supposedly predicting most of the properties of the alternation. Yet, idiosyncratic roots also have a significant effect in determining which alternations they occur in and the two variants’ interpretations. Beavers and Koontz-Garboden (2020) proposed that roots can also introduce the same templatic information as templates, in ways that explain such root-specific effects through compositional interactions of root and template meanings. We explore the explanatory power and limits of this ‘root semantic’ approach by expanding its empirical scope to an alternation similar to a dative alternation in Kinyarwanda (Bantu JD61; Rwanda), which are derived via the applicative morpheme -ir. The broad productivity of -ir in forming ditransitives, coupled with the significantly wider range of root-specific interpretative and grammatical effects, provides an ideal case study for the role of idiosyncratic verbal meaning in governing argument alternations. We show that a root semantic approach naturally captures the diversity of root-specific effects in Kinyarwanda, justifying it as part of a general theory of alternations. However, ingredients from work that has syntactified constraints on alternations or posited allosemic variation in templatic meaning are also needed, resulting in a multifactorate understanding of alternations with root semantics at its core.
Georgieva et al. (2021) argue that negation in Meadow Mari and Udmurt requires the Distributed Morphology conception of grammar with a postsyntactic level of grammar where extrasyntactic operations take place. I show that, by modifying one of their assumptions, a purely syntactic head movement account becomes available that uses no operations other than those of the phrasal syntax. I further show that suspended affixation in Meadow Mari does not require postsyntactic operations, either, contra Guseva and Weisser (2018). I argue that the default view of grammar should be one where there are no operations other than those of the phrasal syntax, and there is only one level of grammar, the morphosyntax. Cases where extrasyntactic levels and mechanisms have been proposed, like Mari and Udmurt, can and should be analyzed in purely syntactic terms.
In Ji’an Yunlou Gan Chinese, the fused motion verb combination luai- k^h ə ‘come-go’ presents a paradox. It is syntactically low, at the vP or VP edge, but interpretatively high, affecting speech-act structure. It activates both speaker and hearer roles, shifting second-person subjects to inclusive interpretations in directives and acting as an attention-getter or ‘negotiator’ in commissives. Pursuing analogies to Basque, Japanese, and Jingpo, we propose a movement-free analysis, following Cheng and Sybesma (2003, 2004) and Cheng (2019) for a similar paradox in Chinese: luai- k^h ə carries a bundled feature complex anchored to the Speech-act Phrase (SAP).
On the view that ellipsis sites contain full – just unpronounced – syntactic structure, it follows that two ellipsis sites should be able to mutually license one another for identity. This paper supports the structural view of ellipsis by exploring this prediction in relation to a series of otherwise recalcitrant counterexamples. With ‘double ellipsis’, two ellipsis sites can satisfy identity for one another without formally relating to overt antecedent material. This analysis allows structural identity to be maintained in the face of apparent mismatches, including argument structure in sluicing, Dahl’s many clauses puzzle and elliptical answers to certain questions. It means that ellipsis sites must contain syntactic structure to evaluate for identity, and can do so even where no overt antecedent material is available.
In this article, I investigate the case marking pattern of durative and multiplicative adverbials of Korean, which show a complex case pattern. Providing evidence from two phenomena (intervention effects of negative concord items, and the relative scope of the subject and the event), I show that case marking on the adverbial correlates with the position of the grammatical subject. I then derive the correlation from a successive-cyclic Dependent Case model (Levin 2017), which is a strictly structural model of case. While some previous researchers have taken case-marked adverbials as evidence that a purely syntactic model of case is insufficient, this article shows that a syntactic approach with the right configurations can achieve much more than what might initially seem possible.
This paper resolves an apparent paradox arising in an autosegmental account of Irish initial consonant mutation: the unexpected distribution of preverbal tense particle d’. Although d’ is usually taken to be the “trigger” of mutation, the conditions for its insertion seem to depend on the post-mutation form of the following word. It thus appears that the mutation trigger enters the derivation only after mutation has taken place. I consider two modularity-compatible solutions that differ only in where they place the burden of explanation for this pattern (i.e., on the morphosyntax or the phonology). Using evidence from both modules, I conclude in favor of the “phonological solution”: preverbal d’ is an underlyingly floating segment that is inserted in all phonological environments; and its distribution follows from the regular phonology. I thus demonstrate that (i) the Irish preverbal d’ pattern is compatible with strict modularity; and (ii) a methodology that commits to a priori modular assumptions can lead to important insights concerning the division of labor at the morphosyntax-phonology interface.
The partition of clause structure into two locality domains (CP and vP) has been standard in syntactic theory since the conception of Phase Theory (Chomsky 2000, 2001). A common type of evidence for the domains comes from footprints of successive cyclic movement, observed at the edges of both CP and vP. However, the two domains famously differ in their opacity for syntactic operations such as A-movement and ɸ-agreement, with vP being significantly more transparent than CP. This asymmetry has led some authors to question the phasehood of vP, especially in the context of alternative explanations for intermediate movement through vP (i.a. Keine and Zeijlstra 2025). This paper argues that we need not rely on intermediate movement as evidence for a clause-internal phase – operational opacity can likewise be observed in the clausal middle field. Evidence presented here comes from Zimbabwean Ndebele, in which VoiceP is argued to be opaque for A-movement and ɸ-agreement. The paper considers two alternative explanations for the absence of A-relations across VoiceP: one in terms of the failure of probing and the other as the absence of a matching goal. Both are ruled out on empirical grounds.
This paper develops a novel analysis of clausal complementation based on the distribution of Greek declarative clauses introduced by oti and pu. These clauses present three puzzles: (i) they are in complementary distribution after most verbs; (ii) they are licensed as internal arguments and derived subjects but are banned from the external argument position; and (iii) complement pu-clauses impose a previously unnoticed stativity restriction on the matrix verb, a restriction absent in their adjunct uses. I argue that (ii) and (iii) receive a unified explanation if oti and pu of complement clauses select a light noun in their specifier (see also Arsenijević 2009; Moltmann 2019, 2024). This noun must be licensed via incorporation into the matrix verb (Hale and Keyser 1993), an operation that succeeds from the complement position but fails from the external argument position, thus explaining the internal–external argument asymmetry. The semantic distinction between content (oti) and situation (pu) explains their distribution after verbs (see also Bondarenko 2022; Elliott 2020; Kratzer 2006; Moltmann 1989, 2013a,b, 2021; Moulton 2009, 2015). Finally, the stativity restriction reveals a distinction in the way arguments and adjuncts are licensed (Bruening 2013; Hunter 2015; Zeijlstra 2020; Hewett 2023; Neeleman et al. 2023, i.a.). Beyond resolving the puzzles in Greek, this analysis settles the open question of whether embedded clauses occupy the same argument position as DPs. Drawing on multiple diagnostics, I show that they can, thereby directly challenging approaches (e.g., Bondarenko 2022) that deny this possibility. The analysis carries broader implications for the role of selection in clausal complementation, the syntax of nominalized clauses, and the Syntax/Semantics interface.
This paper motivates a hierarchical, size-based view of finiteness in languages without inflectional morphology, such as Chinese. Drawing on an array of crossclausal dependencies of ‘again’-type and aspectual elements in Mandarin and Cantonese, we argue that finiteness distinction is best understood as properties arising from clause size differences rather than a simple featural distinction. Empirically, Mandarin preverbal you ‘again’ and Cantonese postverbal -faan ‘again’ exhibit exceptional scopal behavior, which is not found with other Chinese ‘again’-type elements that are lower than the outer aspect. This behavior is possible only across nonfinite clauses with a size smaller than TP (i.e., vP), resembling restructuring crosslinguistically. Adopting the split aspect approach and a hierarchy of complement clause sizes, we propose that you and -faan move to and agree with an outer AspP above vP, respectively. The locality constraint stems from intervention by the outer AspP in the complement clause, which is present in larger nonfinite clauses such as TPs in addition to full-fledged finite CPs. Furthermore, we argue that the minimal size of nonfinite clauses is vP, which contains an inner AspP and blocks restructuring-like patterns with inner aspectual elements. Our findings uncover a fine-grained cartography of ‘again’-type elements and aspectual elements in Chinese languages on the one hand and support a gradient distinction in finiteness marked by complement size over an all-or-nothing dichotomy on the other.
Auxiliary Switch (AS) is an under-studied phenomenon found in Romance restructuring clauses. It is contingent on the climbing of reflexive clitics, but sometimes remains insensitive to non-reflexive clitics. Drawing on the diachrony of French and comparative synchronic data from Italian and Sardinian, I argue that AS reflects a robust syntactic dependency between embedded and matrix domains, instantiated through Agree. The analysis adopts the hypothesis that auxiliary selection is governed not by argument structure but by person value identity. The main conclusions are that have is the default auxiliary in these languages, restructuring is not a phenomenon per se, and the locus of variation resides in the featural make-up of functional heads (cf. Borer-Chomsky Conjecture).
Crosslinguistically, sentences with wh-dependencies across the boundary of a definite depiction nominal object are degraded. We consider two accounts – DP phasehood and the Specificity Condition – which make different predictions about definite islands in wh-movement and wh-in-situ languages, especially around whether the choice of main verb can neutralize the definite island effect. We tested these predictions with parallel experiments for English (wh-movement) and Mandarin Chinese (wh-in-situ). Confirming existing reports, we observed that both languages show a definite island effect. Additionally, in English, but not Chinese, there is a neutralization effect related to the choice of main verb. However, the effect is only partial, contrary to previous claims. A combination of the PIC and the Specificity Condition accounts for both languages, even though it would be more parsimonious to have only one constraint. To support this conclusion, we ran a third experiment on Chinese wh-indefinites to provide independent evidence for the Specificity Condition.
All tonally derived verbs in Mwaghavul have ideophonic bases and are derived through one of two lexically determined alternations. In the first alternation, every tone-bearing unit (TBU) of the ideophone surfaces with a mid tone. In the second alternation, every nonfinal TBU of ideophones surfaces with a mid tone, while the final TBU surfaces with a high tone. To express pluractionality, both types of ideophones are reduplicated and follow the second pattern: every TBU of the reduplicant bears a mid tone, and every TBU of the base bears a high tone. While verbs derived through mid-tone overwrite share a tonal melody with underived verbs, only derived verbs have a mid–high melody. Beyond verb formation, the tonal alternations introduce various meanings, conveying the distinguishability of identical elements. We propose that the alternations are triggered by lexically specific allomorphs of a verbaliser, whose exponents are mid and mid–high tonal melodies, respectively. The restriction of the mid–high melody to derived verbs reflects systematicity. For the derived pluractional verbs, the tonal dissimilarity between a reduplicant and its base depicts the distinguishability of otherwise identical elements. We consider the implications of the patterns for linguistic theory, particularly the relationship between grammatical integration and expressiveness.
Elbourne has proposed that future research should investigate the possibility of eliminating syntactic categories from linguistic theory and letting their functions be taken over by independently motivated semantic categories such as semantic types (‘A program for eliminating syntactic categories’, 2024, Linguistic Inquiry). The current article applies this research programme to the case of adjectives. It argues that adjectives should be conceived of as functions from noun denotations to noun denotations (a unique type). Adjectives in predicative position are dealt with by giving a suitable denotation to the copula. Various arguments against this analysis of adjectives are reviewed and dismissed.
This paper provides a novel analysis of asymmetries in gender agreement with conjoined subjects, accounting for a range phenomena in the Xhosa and Shona languages. For all Xhosa conjoined gender-mismatched singular DPs, the gender of plural agreement with P depends on whether the conjuncts share the feature [human], [inanimate], or [animal]. I argue that this reflects a stacked nP structure within Bantu nominals: whatever a DP’s visible gender, it includes an nP core with one of these three semantic genders. Thus the internal structure is [nP1 n1 [nP2 n2+root]]; n2 is always the bearer of a semantic gender. n2 values agreement if: (i) the conjoined DPs have mismatching visible n1 gender features but the same n2 cores, or (ii) the visible gender of the two conjuncts matches but is formally uninterpretable, assuming with Adamson and Anagnostopoulou (2025) that resolved agreement is based on the intersection of conjuncts’ interpretable features. A diagnostic for (un)interpretability of gender emerges: uniform gender singular conjuncts pair with semantic agreement iff their visible genders are uninterpretable, so excluded from the intersection. For others there is systematic matching agreement because they are uniformly interpretable, even in combination with arbitrary members since in such cases, n1 is i[entity]. I show that in Bantu languages with a large number of genders, those exhibiting matching plural agreement with uniform conjoined singulars are the exception, and not the rule.
Quotative inversion (QI) clauses in Setswana and English—two SVO languages—show a marked OVS word order in which a quote appears preverbally and the agentive external argument appears postverbally. In this paper, I present empirical evidence from both languages showing that the preverbal quotative subject moves to its landing site via A-movement, while the agentive subject remains in-situ. While moving an internal argument to an A-position above an external argument may seem to pose a minimality problem, I show that QI constructions are best conceived of as smuggling derivations in which the internal argument is moved over the external argument via the movement of a larger constituent containing the internal argument (Collins 2005, 2024; Storment 2025). I support this proposal empirically and theoretically, drawing on novel data from both English and Setswana analyzed under the Strong Minimalist Thesis (Chomsky et al. 2023). I additionally draw parallels to other inversion phenomena in both languages, showing that these constructions are best conceived of as inverse voice phenomena.
