This paper provides empirical support for the tripartite VP structure (Voice-v-V), drawing on evidence from Japanese verbal phenomena. We examine two structural distinctions: (i) the demarcation between VRes and v, and (ii) the separation of v and Voice. Japanese VP anaphora permits transitivity mismatches in both causative-inchoative and inchoative-causative orders, suggesting anaphoric expressions target VPRes encoding a result state, independently of v. This contrasts with English do so anaphora, which only allows mismatches in the causative-inchoative order. The light verb su 'do' alternates with nar 'become' in some contexts (stative predicates) but not others (Sino-Japanese verbs, deverbal compounds), a pattern accounted for if they realize distinct heads: vCAUS/INCH and Voice. VP ellipsis further corroborates this analysis, consistently employing su regardless of antecedent transitivity. Our analysis establishes four verbal domain patterns in Japanese: (a) complete head movement in Native verbs, (b) v-Voice complexes without verbal roots, (c) partial movement with non-Native roots, and (d) partial movement with Native roots in specific contexts. These findings support the tripartite VP structure and illuminate the relationship between structure, head movement, and morphological realization in Japanese verbal domains.
This paper argues for a novel view of the Norwegian noun categorization system, in which a noun can belong to one of four classes. This goes against the traditional generalization that Norwegian is a three-gender language (divided into masculine, feminine and neuter). The empirical focus point of the paper is on procedures in which count nouns are transformed into mass nouns (grinding), or vice versa (portioning). I present data that shows that, while nouns with masculine and feminine gender can undergo these procedures, neuter nouns are locked to one of two countability values (count or mass). Based on this data, I argue that what is generally called "neuter gender" is actually a form that appears when a noun does not belong to a gender class. What we end up with are four noun classes in Norwegian: masculine, feminine, count and mass.
It is argued that the passive construction, common in the SOV Ob-Ugric sister languages of Hungarian displaying fused grammatical and discourse functions, was a frequent construction in Proto-Hungarian, as well, but its use gradually declined in the Old and Middle Hungarian periods, and it completely disappeared in the Modern Hungarian era. The attrition of passive voice is attributed to the growing opacity of the passive construction, and to the existence of more transparent alternative patterns. It is shown that both components of passive inflection, the -(t)At causative morpheme and the unaccusative -ik subject agreement paradigm, were ambiguous in a wide range of cases, imposing additional cognitive load on the language user. However, theme promotion to topic role and agent focusing, performed via passivization in Ob-Ugric, could also be realized by active sentences displaying object topicalization and subject focusing in Old Hungarian. Agent demotion could be attained by three means, representing different degrees of agent demotion: (i) the impersonal construction, involving a 3PL plural verb and a pro subject; (ii) the use of a detransitivizing verbal suffix; (iii) a -vA participle phrase + be-verb construction expressing result state. It is claimed that these patterns could be learned more easily than the passive, which led to their gradual increase in the primary linguistic data determining the grammars of subsequent generations.
This study examines whether animacy affects the acceptability and processing of island-violating wh-questions in Saudi Arabic, and whether this effect interacts with tail type-gap or resumptive pronoun (RP). The question is motivated by competing accounts of island effects: syntactic accounts argue that Arabic what-questions should be uniformly unacceptable due to violations of movement constraints, while who-questions should be acceptable only when paired with RPs, as the dependency is established through binding rather than illicit movement (Aoun & Benmamoun & Choueiri 2009). Processing accounts, in contrast, predict that more accessible fillers, such as animate expressions, should ameliorate, though not fully repair, island effects by facilitating the integration of fillers at dependency tail (Kluender 1991). To address this, we carried out a self-paced reading experiment combined with an acceptability judgment task. Our design manipulated two factors: the animacy of the filler (who vs. what) and the type of tail (gap vs. resumptive pronoun) in adjunct island-violating dependencies. Results revealed that animacy led to a modest but significant boost in acceptability, though only when dependencies terminated in an RP. At the same time, animate fillers consistently resulted in slower reading times, regardless of tail type. These findings are difficult to reconcile with both purely syntactic accounts and memory-based processing accounts: they challenge syntactic accounts, which predict a clear acceptability contrast between structurally licit and illicit dependencies, and memory-based processing accounts, which predict faster reading times at integration site when the filler is animate (i.e. highly accessible). Instead, we argue that the data are better explained by a multivariate account: animacy enhances referential salience, allowing for an interpretable anaphoric dependency with RPs when syntactic resolution is blocked. The resulting anaphoric dependency improves acceptability but imposes a processing difficulty. Taken together, the results are consistent with multivariate accounts of island effects (Chaves & Putnam 2020; Huang & Almeida & Sprouse 2025), in which syntax, processing and discourse jointly shape acceptability and comprehension.
Clause-embedding predicates alternating between a complement expressing a belief and a complement expressing an intention seem to be quite widespread both inside and across languages. The two different meanings are expressed by different morphosyntactic types of complement clauses, in line with the more general observation that the semantic distinction between representational and preferential attitudes tends to be reflected by certain morphosyntactic differences between the complement clauses used. While this association between form and meaning generally holds in Basque, too, it is not completely strict: both indicative clauses and absolutive verbal nouns can be found expressing not only one, but marginally also the other meaning, raising the question whether in these cases other disambiguating factors play a role, possibly resulting from the different semantic properties of beliefs and intentions. The results of an online questionnaire show that, with very few exceptions, the type of complement clause is decisive in determining the meaning of a complement, corroborating the morphosyntactic split between representational and preferential attitudes. While indicative clauses in intent contexts seem in fact to express beliefs rather than intentions, verbal noun complements in belief contexts are in general either interpreted as intentions or judged ungrammatical. However, it seems that the use of a verbal noun complement in a belief context can become more acceptable when it is past tense, discourse-given or does not have an overt subject. Thus, while there are canonical associations between the type of attitude and the type of complement clause, under certain conditions they may be neutralized.
Adpositions may govern case, and there is dramatic variation in the case values found. Different adpositions often govern different cases, which means that we need to partition (or split) the adpositions according to the case(s) they take, as in German. But this is just one possibility. The typological point of interest here is that the items governed (the governees) may also be a determining factor, as when nouns and pronouns require different cases, in Mari for instance. Furthermore, we may need to specify both the adposition and the governee (Russian). And then, rather than one fixed outcome, there may be alternating outcomes, for each of the three types of partition (adposition-based, governee-based, and splits based on both). At the extreme, even when we have specified a particular adposition, and an individual noun, we may still need further conditions. Here corpus work reveals the interesting interaction of determining factors, notably in Finnish. Our exploratory canonical typology lays out the possibilities, providing clear differentiation of the attested systems and subsystems. The specific example of adpositions can then be used as an entry point to approach unresolved issues in the more general area of differential marking.
