《2018年度博士論文佳作獎》

姓名：鄭偉成（國立清華大學語言學研究所）

指導教授：蔡維天教授、吳曉虹教授

論文題目：漢語句法-言談介面研究

中文摘要

本論文研究言談訊息是否以形式特徵存在於句法運算系統裡。我探討漢語的關聯主題和非語句結構等兩種現象，指出其語意的詮釋性必須依賴言談語境以及透過補詞層和詞彙層的左緣結構來體現其句法特性。我進而主張漢語有兩種手段觸發句法－言談介面銜接。

        漢語的關聯主題在前人文獻已經探討甚多，其句法位置和訊息結構特徵有著不同的分析。我提出新的觀察，指出漢語關聯主題應該分成三大類。我根據句法測試及其訊息結構主張漢語的關聯主題句應該視為XP-分裂結句。我採用Fanselow and Cavár’s (2002)分布刪除分析，提出關聯主題句是透過一連串特徵檢驗過程而生成的。XP由兩個詞組所組成，此兩詞組在詞列階段已經和主題特徵與焦點特徵分別合併。在句法階段，則是分別在補詞層和詞彙層左緣功能投射組進行特徵檢驗。關聯主題句的生成顯示主題和焦點在詞列階段進入句法操作，透過特徵檢驗手段現行於句法結構樹上。

非語句的結構儘管缺乏完整的句法骨架，但是可以蘊含子句標示、施為念力以及說話者/聽話者等言談訊息。承接Sigurðsson & Maling (2009)和Tsai (2016)的精神，我指出漢語非語句有完整補詞層結構，以句法形式實體化言談訊息。然而，不同於施用於關聯主題的特徵查核機制，我主張言談訊息並非在詞列階段就以形式特徵存在，取而代之的是透過言談行為投影層，分別由兩個言語行為功能投射組組成，分別體現了說話者與聽話者言談中的攸關性。

本文所探討的兩種現象指出漢語擁有兩種手段使句法與言談銜接。第一種手段為在詞列階段，將訊息結構相關的特徵與詞彙項目合併。接著，在句法階段，透過在補詞層/詞彙層的左緣的相關功能投射組完成特徵檢驗。第二種手段則是透過言談行為投影層內的兩個言語行為功能投射組，將說話者與聽話者等相關言談概念，帶入句法層面。換言之，說話者與聽話者這兩種言談概念並非以形式特徵出現於詞列階段，而是由言語行為功能投射組來體現該概念。本文所提的兩種介面手段和漢語句法高分析性息息相關。除此之外，本文更進一步指出言談可以內化為語法的一部分。

英文摘要

The goal of this dissertation is to investigate whether discourse notions are active in the syntactic computation or outside the domain of it. I document two phenomena in Mandarin, Aboutness Topic (AT) and Nonsententials (NS), whose interpretative import is acutely sensitive to the discourse context and needs to be syntactically substantiated in the articulated peripheral structure of CP/vP. In particular, I propose that there are two means to activate the syntax-discourse interface permitted by the computation system.

Mandarin AT has been discussed at great length in the previous scholarship, and its designated position and information structural properties receive a variety of analyses. Nonetheless, I offer novel observations showing that AT in Mandarin can be analyzed on a par with XP-split constructions, according to a battery of diagnostics and its information structural makeup. Following Fanselow and Cavár’s (2002) analysis, I propose that AT results from successive feature-checking processes in which a XP, whose sub-parts are merged with two disparate features (topic and focus) in the numeration, permits its subparts to undergo feature checking in corresponding functional projections and to be spelt out differently in the left periphery of CP/vP. The case of AT concludes that topic and focus are accessible in the numeration and corresponding functional projections are merged for feature-checking purposes in order for the derivation to converge.

NSs, whose syntactic structure is considerably reduced, are able to encode clause typing information, illocutionary force and the involvement of speaker and hearer/addressee. Following the line of reasoning in Sigurðsson & Maling (2009) and Tsai (2016), I propose that NSs have a fully-fledged peripheral structure of CP, according to the effects exerted upon their interpretation. Different from the feature-checking mechanism activated for AT, I argue that no discourse properties are accessible in the numeration and drive the derivation. Rather, a speech act layer, a supra-sentential layer, merges to and dominates ForceP, and is responsible for the encoding of the relevant discourse properties.

The two phenomena of inquiry suggest two means to activate the syntax-discourse interface. On the one hand, the syntax-discourse interface can be activated by merging lexical items with strong informational structural features that have to be checked by corresponding function projections in the periphery of CP/vP, along the lines of Aboh’s (2010) view that the interface starts with the numeration. On the other hand, a cluster of discourse properties, such as speaker and hearer/addressee, cannot be treated as formal features driving the derivation and, however, can be concretized by another supra-sentential layer, the speech act layer. The two means are made available due to the analyticity of Mandarin syntax. The major consequence of this work is to show that the theory of discourse is closely tied to the architecture of grammar in general.

論文貢獻

Five contributions are summarized as follows.

Theoretically, this dissertation offers solid evidence in support of the view that discourse notions (topic, focus and speaker-hearer relations) are visible to syntax, in contrast with Kučerová & Neeleman’s (2012) Rule Mapping Approach that discourse notions are the reflex of information structure derived via a set of mapping rules from syntax structure and Horvath’s (2010) Strong Modularity Hypothesis that discourse is outside the domain of grammar. Different from previous works that also endorse the same view, this work further points out that there are two means of activating the syntax-discourse interface in Mandarin. First, it is argued in this work that discourse notions (topic and focus) can be recast as formal features and enter into syntactic derivations in the numeration by merging lexical items with discourse-encoded features, and these features have to be checked off for interpretability at the interface by a layer of functional projections along the spine of clausal structure. This line of thinking amounts to showing that the syntax-discourse interface starts with the numeration, where an array of necessary discourse features is numerated in accordance with the speaker’s intension, and corresponding functional projections are merged in the course of derivation to check off these features for them to become interpretative at the interface. Second, certain discourse notions (the discourse roles speaker and hearer) cannot be treated as formal features, because they do not induce any interpretative effect and are acutely sensitive to the discourse. This is taken to show that these discourse notions are externalized to the syntactic computation. Nonetheless, this does not mean that they have to be dealt with on independent grounds; instead, they are substantiated by a supra-sentential layer (sa*P and SAP) serving the gateway between discourse and syntax. Following the sa*P analysis (Speas 2003; Haegemen and Hill 2013; Haegemen 2014), I argue that the syntax-discourse interface can be also activated by the merge of a speech act layer which is responsible of integrating the current proposition (CP) into the discourse (Speech Act Layer), where speaker and hearer/addressee are concerned. From a typological perspective, the availability of two means of activating the syntax-discourse interface is tied to the highly analytic syntax of Mandarin (Huang 2015).

This dissertation serves as the first attempt at convincing non-theoreticians and WYSIWYG (what you see is what you get) linguists that the theory of discourse is closely tied to the architecture of grammar in general. Specifically, discourse notions need to be substantiated by syntactic structures. Mandarin, defined as a highly analytic language (Huang 2015), has a fully-fledged structure of CP, consisting of an array of functional projections and serving the gateway between syntax and pragmatics/discourse (Tsai 2015a, 2015b). Following the cartographic approach (Rizzi 1997; Cinque 1999), this work proves that topic and focus project TopP and FocP respectively and are merged along the clausal spine ranging from vP to CP. In addition to consolidating the two well-established left peripheries (CP and vP), this work further argues for a supra-sentential layer above the sentential structure of CP, responsible for encoding the speaker-hearer relation sensitive to the immediate context. This natural affinity between syntactic structure and discourse cannot be readily captured by linguists that resort to discourse as a domain outside of grammar while marginalizing the component of syntax.

The first half of this dissertation is dedicated to addressing the unsolved issue of Aboutness Topic (AT) in Mandarin, whose precise information structural makeup, distribution and derivation remains a piece of the jigsaw in previous studies. This dissertation provides a thorough and critical review of competing analyses of AT. Besides, I make novel observations about AT, including the recursive nature of AT in two peripheries (CP and vP), lexical identity effects, island effects, its interpretative dependency with respect to a c-commanding focused element and three types of AT, which are not previously identified. I motivate Fanselow and Cavár’s (2002) distributed deletion analysis and suggest that AT results from successive feature-checking processes in which a XP, whose sub-parts are merged with two disparate features (topic and focus) in the numeration, permits its subparts to undergo feature checking in corresponding functional projections and to be spelt out differently in the left periphery of CP/vP. This analysis is advantageous to the existing analysis of AT in Mandarin and offers a more explanatory account of the derivation of AT and its connectedness effect with the main clause. The first documented case of AT suggests that the numeration determines information structure along the line of Aboh(2010), and functional projections in two peripheries (CP and vP) determine the distribution of lexical items that bear information-structural features. The proposed analysis also points out that TopP and FocP are recursive along the clausal structure, but their distributions are asymmetrically constrained. TopP and FocP are freely allowed to occur in the CP periphery, whilst their occurrence is more restricted in the left periphery of vP.

The second half of this dissertation investigates nonsententials (NSs) in Mandarin, which have not been noticed in previous studies. 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. Precisely, NSs can be regarded as a syntax-discourse construal whose interpretative import has to be substantiated by syntactic structure. The sa*P analysis of NSs advanced in this work has three implications. (i.) A cluster of properties about NSs in Mandarin sheds light on a long-standing debate about whether NSs are derived via base generation (a direct interpretation approach) or movement (a movement account). In addition to evaluating the competing analyses of NSs in English, I show that the presence of sentence-final particles, discourse particles and the vocative embodied in the syntactic structure of NSs lends support to the movement account. (ii.) It is moved in this dissertation that sa*P and SAP constitute a supra-sentential layer, which merges to the sentential layer (CP-TP-vP) and serves a gateway of integrating the propositional content into the immediate context. The sa*P analysis provides strong evidence for the syntactic representation of discourse particles and the vocative above ForceP. (iii.) The proposed analysis of NSs supports Sigurðsson & Maling’s (2009) claim that grammatical categories (heads and/or features) are commonly present and syntactically active even though they remain silent. Precisely, the missing parts of NSs are syntactically unmarked/silent. This line of thinking pursued here also finds an instant rapport with Cinque’s (1999:127) generalization that ‘the entire array of functional projections [is] present in every sentence.’ Typologically, Mandarin fits the two views advanced above.

Overall, the discussion in this dissertation provides significant implications for syntactic theory and the modulation of grammar. First, with the nature of analytic syntax, Mandarin provides a good testing ground for articulating the fine structure of CP/vP peripheries of Mandarin in the spirit of Cinque’s cartographic enterprise. Different from synthetic and agglutinating languages, overt spelt-outs of discourse particles and the discourse offers a window on how discourse notions are syntactically substantiated in a speech act layer and how the layer interacts with the proposition (CP) to active the syntax-discourse interface in a more transparent way. Besides, the distribution of topic and focus shows that discourse is syntactically represented by overt movement in Mandarin, which in turn suggests TopP and FocP. This view lends weight to a generalized transparent mapping between form and interpretation in Mandarin, which is to syntacticize as much as possible the interpretative domains. Second, the view that discourse can be syntactically represented is more compatible with the T/Y model of the Minimalist Program (Chomsky 1995) in the sense that specific operations (external and internal merge) and phrase structures are active at narrow syntax in order to concretize discourse notions at the interface.