This paper aims to develop the empirical and theoretical basis for the necessity of motivating a cartographic approach (Rizzi 1997; Cinque 1999) to the clausal structure of nonsententials (NSs) in Mandarin. Especially noteworthy about NSs is that they are able to encode clause type information, illocutionary force and the discourse roles speaker and hearer/addressee, though their structure is considerably reduced. Following the line of reasoning in Sigurðsson & Maling (2012) and Tsai (2015a, 2015b), I show that NSs have a fully-fledged CP structure, according to the effects exerted upon their interpretation. Adopting Haegeman’s (2014) sa*P analysis of the discourse particle in West Flemish, I argue that NSs contain a suprasentential structure, a Speech Act layer, dominating ForceP and responsible for the encoding of the speaker- hearer/addressee relation sensitive to the immediate context. Crucially, it is argued that the discourse properties surrounding NSs pertain to the articulated CP structure of NSs. The major consequence of the proposed analysis is to show that the theory of discourse is closely tied to the architecture of grammar in general, adding weight to the view that the transparent syntax-discourse mapping results from a set of functional projections layered in the CP periphery.

This work investigates the division of labor between mood and illocutionary force in syntax by examining three modal construals encoded by the speaker-oriented adverb zuìhǎo ‘best’ (deontic, epistemic and evidential) in Mandarin, and accounts for a cluster of syntactic and pragmatic properties it is associated with. Very much in line with Tsai’s (2015a, 2015b and 2015d) modal system in Mandarin, it is observed that each type of zuìhǎo can co-occur with its matching modal auxiliary in the fashion of Cinque’s (1999) ‘location-in-Spec’ hypothesis and encodes a certain type of illocutionary force. One persistent question is how zuìhǎo substantiates illocutionary force in syntax, while its designated position is not situated in the licensing domain of ForceP. As far as the left periphery is concerned, this work argues for a conspiracy between syntax, semantics and pragmatics to ensure the success in activating the Bidirectional Agree relation between speech act, force and mood. We argue for a speech act layer (Sa*P) externally merging to CP (Speas and Tenny, 2003), whose head values the uninterpretable speech act feature [uSa] on Force0 via the Bidirectional Agree to trigger its interface with the utterance content (CP). Meanwhile, following Kempchinsky’s (2009) idea, it is further argued in this work that Force0 hosts the uninterpretable feature [uW] which has to be checked and valued by the modal construals of zuìhǎo to determine the irrealis-realis mood. An immediate implication is that ForceP serves as a gateway to not only mood but also speech act at the interface. Several issues involved in dealing with zuìhǎo are discussed.

This paper develops a syntax-pragmatics interface analysis of imperative clauses overtly marked by two grammatical categories of qing ‘please’ in Mandarin and refines the division of labor among directive force, clause typing and deontic modality jointly computing the interpretative properties of qing imperatives. We present a cluster of properties to differentiate between the two categories of qing and observe that qing1 denotes obligation imposed on the addressee by the speaker, while qing2 denotes permission with which the addressee is allowed to perform an action or make true a state of affairs according to a set of norms. It is argued that qing1 is an imperative mood head, while qing2 is an imperative adverb, but both are endowed with a similar internal composition and extent of the phrasal hierarchies of the CP periphery, and their disparate imperative properties can be ascribed to the addressee-oriented and subject-oriented deontic modality (Tsai & Portner 2008). Following Haegeman & Hill's (2013) version of the Speech Act Phrase, we claim that a speech act layer externally merges to the topmost position of ForceP to drive the syntax-pragmatics interface computation of the speaker-addressee relation and to mediate the imperative mood and clause typing represented in the CP layer.

By drawing on data of two types of VP dislocation in Mandarin, this paper argues that these two types represent a syntactic incarnation of XP-split constructions providing an interesting argument for the conception of information structure starting in the numeration and multiple Spell-Out of copies regulated by the checking of strong features over the course of a syntactic derivation. Building on Fanselow and Cavar's ( 2002) distributed deletion analysis, I show that Type 1 involves two autonomous nominals that form a nominal predicate structure in a VP, and the VP is copied and internally merged to a topic position, followed by multiple Spell-Out of the copies whose information structural features are checked. In contrast, Type 2 is derived via base-generation. I argue that the two types are derived employing the same set of operations in the computational system without extra postulations. Issues involved in dealing with Mandarin topic constructions are addressed.*This article is a revised version of Chapter 2 and 3 of my PhD dissertation. I wish to express my gratefulness to Hsiao-hung Iris Wu, Wei-tien Dylan Tsai, Wei-wen Roger Liao and Chyan-an Arthur Wang for their constructive comments on the earlier versions of the article and two anonymous reviewers of Studia Linguistica for their valuable suggestions which make this article close to being in the vicinity of satisfaction and adequacy. I owe deeply to Chris Hsu, Mindy Kuo, Larry Liu and Grace Liu for discussing the data with me. The remaining inadequacies, conceptual or empirical, are solely my own responsibility.