A Cognitive-Functional Approach to Topic Constructions

in Beijing Mandarin

by

Linjun LIU

Dissertation

Submitted to the Faculty of Graduate Studies of

The University of Manitoba

in partial fulfilment of the requirements of the degree of

DOCTOR OF PHILOSOPHY

in LINGUISTICS

Department of Linguistics

University of Manitoba

Winnipeg, Manitoba

Copyright © January 2009 Library and Archives Bibliotheque et 1*1 Canada Archives Canada Published Heritage Direction du Branch Patrimoine de I'edition

395 Wellington Street 395, rue Wellington OttawaONK1A0N4 Ottawa ON K1A 0N4 Canada Canada

Your file Votre reference ISBN: 978-0-494-53003-0 Our file Notre reference ISBN: 978-0-494-53003-0

NOTICE: AVIS:

The author has granted a non­ L'auteur a accorde une licence non exclusive exclusive license allowing Library and permettant a la Bibliotheque et Archives Archives Canada to reproduce, Canada de reproduire, publier, archiver, publish, archive, preserve, conserve, sauvegarder, conserver, transmettre au public communicate to the public by par telecommunication ou par I'lnternet, preter, telecommunication or on the Internet, distribuer et vendre des theses partout dans le loan, distribute and sell theses monde, a des fins commerciales ou autres, sur worldwide, for commercial or non­ support microforme, papier, electronique et/ou commercial purposes, in microform, autres formats. paper, electronic and/or any other formats.

The author retains copyright L'auteur conserve la propriete du droit d'auteur ownership and moral rights in this et des droits moraux qui protege cette these. Ni thesis. Neither the thesis nor la these ni des extraits substantias de celle-ci substantial extracts from it may be ne doivent etre imprimes ou autrement printed or otherwise reproduced reproduits sans son autorisation. without the author's permission.

In compliance with the Canadian Conformement a la loi canadienne sur la Privacy Act some supporting forms protection de la vie privee, quelques may have been removed from this formulaires secondaires ont ete enleves de thesis. cette these.

While these forms may be included Bien que ces formulaires aient inclus dans in the document page count, their la pagination, il n'y aura aucun contenu removal does not represent any loss manquant. of content from the thesis.

1*1 Canada THE UNIVERSITY OF MANITOBA

FACULTY OF GRADUATE STUDIES

COPYRIGHT PERMISSION

A Cognitive-Functional Approach to Topic Constructions in Beijing Mandarin

BY

Linjun LIU

A Thesis/Practicum submitted to the Faculty of Graduate Studies of The University of

Manitoba in partial fulfillment of the requirement of the degree

Of

Doctor of Philosophy

Linjun LIU © 2009

Permission has been granted to the University of Manitoba Libraries to lend a copy of this thesis/practicum, to Library and Archives Canada (LAC) to lend a copy of this thesis/practicum, and to LAC's agent (UMI/ProQuest) to microfilm, sell copies and to publish an abstract of this thesis/practicum.

This reproduction or copy of this thesis has been made available by authority of the copyright owner solely for the purpose of private study and research, and may only be reproduced and copied as permitted by copyright laws or with express written authorization from the copyright owner. ACKNOWLEDGEMENTS

Throughout writing this dissertation I was fortunate to have the assistance and support of a number of people who influenced the direction this study has taken. Terry Janzen, my supervisor, has given me countless insightful comments and suggestions, and has been the source of much encouragement in my PhD programme. The other members of my committee, Hongyin Tao, Jila Ghomeshi, David Pentland, and Hua Li, have made many excellent suggestions.

Many thanks go to the Department of Linguistics, University of Manitoba, for fostering the cognitive and functional linguistic study. Any errors are, of course, my own.

My thanks also go to Beijing Language and Culture University for granting me a four-and-a-half-year leave, without which my pursuit of a degree in Canada would have been impossible.

The writing of this dissertation was supported by a University of Manitoba Graduate Fellowship and a J. G. Fletcher Award from the Faculty of Arts, University of Manitoba.

Finally, nothing could have been accomplished without my family's support, encouragement, patience and perseverance. Thank you.

Linjun Liu

The University of Manitoba January 2009 A Cognitive-Functional Approach to Topic Constructions

in Beijing Mandarin

by

Linjun LIU

Abstract

Mandarin Chinese has long been accepted as a topic-prominent language. However, a consensus has never been reached on how to define or characterize the notion of topic.

Given the pragmatic orientation of the notion and the heterogeneity of topic-comment constructions, a cognitive-functional approach has been adopted in this research to analyze a spoken corpus of Beijing Mandarin. On the basis of what has been found about structurally identified topics in the corpus, I have argued that topic constructions are best taken as specific instantiations of the schema of Conceptual Reference Point for the ensuing comments.

Spoken data are more indicative of online processing than are written texts. Taking advantage of this, I have examined how pauses and pause particles are interrelated with topic marking and topic verbalization. A comparison between marked grammatical subjects and unmarked ones has revealed that the former are more heavily coded.

Following the iconicity principle or the accessibility theory as proposed by Givon (1983),

Ariel (1988, 1990), etc., topics convey more information than subjects.

iii The investigation into the discourse use of several high frequency pause particles, namely ne, ba, ma, a (and its phonetic variants), and two lexical topic markers, i.e., dehua and laijiang, shows that they, like Japanese wa or eyebrow raise in ASL, are all polysemous: they fulfill different discourse functions by marking topics of different semantic roles and information statuses. These topic markers are also observed to occur successively and co-occur with each other in topic marking. The richness in topic markers and topic marking manners further evidences topic as a complex category that demands a schematic characterization.

Following detailed analysis of topic instantiations in the spoken corpus, a schematization of topic as a complex linguistic category is expounded first horizontally then vertically. Topic as a schema is then characterized as follows: topics of all semantic roles may fulfill the function of conceptual reference point in spontaneous speech, and the reference point may sometimes be objective, especially when the relationship between topic and its target referent(s) resembles some objective ones, such as the possessive or the part-whole relationship.

IV Table of Contents

List of Abbreviation x

List of Tables xi

List of Figures xii

Chapter 1 Introduction 1 1.1 The Issue 1 1.2 Cognitive Linguistics as an Alternative 2 1.2.1 Outlining cognitive linguistics and its relevance to this research 2 1.2.2 Topic as conceptual reference point 3 1.3 A Corpus Approach 4 1.4 Thesis Chapters 8 1.5 Summary 11 Chapter 2 Review of Previous Studies on Mandarin Topics and Proposal of the Cognitive Approach as an Alternative 12 2.1 Findings on and Controversies over Topic 13 2.1.1 Topic versus subject 13 2.1.2 Syntactic topic versus pragmatic topic 15 2.1.3 Central versus marginal properties of topic 18 2.1.4 Topic identification 24 2.1.5 Summary 34 2.2 A Cognitive Approach to Topic 35 2.2.1 An overview of cognitive linguistics 35 2.2.2 Cognitive Grammar on reference point 38 2.2.3 Cognitive Grammar on topic 40 2.2.4 Insights of the reference point model for topic study in Mandarin 43 2.3 Findings in Other Topic-Prominent Languages 47 2.3.1 Tagalog 47 2.3.2 Japanese 49 2.3.3 American Sign Language 51 2.3.4 Section summary 52 2.4 Summary 52 Chapter 3 Pause and Pause Particles as Topic Markers and a Coding Property Based Approach to Topic Identification 53 3.1 Pauses as Topic Markers 53 3.1.1 Some general observations 53 3.1.2 Shi's (1995) detailed discussion on pausing and topic marking 55

v 3.1.2.1 Obligatory subtypes 56 3.1.2.2 Optional subtypes 60 3.1.2.3 A brief discussion 62 3.2 Pause Particles as Topic Markers 63 3.2.1 Some general observations 63 3.2.2 Some specific findings 64 3.2.3 Summary 68 3.3 Discourse Use of Pause Particles in Beijing Mandarin 68 3.3.1 Zhang and Fang (1996) on pause particles 68 3.3.2 Yuan (2002) on pause particles 70 3.3.3 Summary 71 3.4 A Coding Property Based Approach to Topic Identification 72 3.4.1 Coding properties as the primary concern 72 3.4.2 Topic identification 74 3.4.3 Summary 80 3.5 Defining Some Grammatical Terms 80 3.5.1 Clause 80 3.5.2 Topic-comment constructions (TCs) 82 3.6 Summary 83 Chapter 4 Methodology of a Corpus Study 85 4.1 Data: The Beijing Corpus 85 4.2 Corpus Analysis Procedure 87 4.3 Research Goals 90 4.4 Summary 91 Chapter 5 Pausing and Topic Marking 92 5.1 Pausing in Nominal Constructions 92 5.2 Pausing in Double-Subject Constructions 96 5.2.1 Occurrence of pauses 96 5.2.2 Semantic relationship between topic and subject 98 5.2.3 Word order flexibility and its impact on construction semantics 101 5.2.4 Possession as the prototype relationship between topic and subject 102 5.2.5 Distribution of double-subject constructions 106 5.2.6 Summary 107 5.3 Pausing in TOP-subject Constructions 108 5.3.1 Frequency of occurrence 108 5.3.2 Distribution of TOP-subject vis-a-vis SP 109 5.3.2.1 Within TCs 110 5.3.2.2 Within subordinate clauses of CCs 113 5.3.2.3 Within LDs 114 5.3.2.4 Within TOP-subject constructions , 116 5.3.2.5 Summary 117

vi 5.3.3 Coding forms of topic in TOP-subject vis-a-vis subject in SP 117 5.3.4 Discourse functions of TOP-subject constructions 126 5.3.4.1 Referencing to some findings in English 126 5.3.4.2 Context of use of TOP-subject in Beijing Mandarin 128 5.3.4.3 Information status of topic in TOP-subject 131 5.3.4.4 Summary 136 5.4 Summary 137 Chapter 6 Topic Markers in Beijing Mandarin 139 6. 1 Pause Particle ne 139 6.1.1 Frequency and context of occurrence 139 6.1.2 Discourse functions of He-marked topics 143 6.1.2.1 INTRODUCING 144 6.1.2.2 CONTINUING 145 6.1.2.3 REPEATING 146 6.1.2.4 PROMPTING.... 147 6.1.2.5 CONTRASTING 148 6.1.2.6 PICKING UP 149 6.1.2.7 LISTING 150 6.1.2.8 SUMMARIZING 151 6.1.2.9 Summary 151 6.1.3 Semantic roles of «e-marked topics 152 6.1.3.1 Summary 155 6.1.4 Topic ordering in multiple topic constructions 156 6.1.5 Summary 159 6.2 Pause Particle ba 160 6.2.1 Frequency and context of occurrence 160 6.2.2 Discourse functions and semantic meanings of Aa-marked topics 161 6.2.2.1 CONTRASTING ba 161 6.2.2.2 ILLUSTRATING ba 166 6.2.2.3 SUPPOSING ba .....168 6.2.2.4 Summary 170 6.2.3 Semantic roles of Z>a-marked topics 170 6.2.4 Summary 171 6.3 Pause particle a and its phonetic variants 172 6.3.1 Frequency counts and contexts of occurrence 172 6.3.2 Discourse functions of topics marked by a and its phonetic variants 175 6.3.2.1 Summary 179 6.3.3 Semantic roles of topics marked by a and its phonetic variants 179 6.3.4 Summary , 186 6. 4 Pause particle ma 186 6.5 Lexical topic marker dehua 'if and laijiang 'speaking of 190

Vll 6.5.1 Frequency counts and contexts of occurrence 191 6.5.2 Discourse functions and semantic roles of topics marked by dehua and laijiang/laishuo 193 6.5.3 Summary 199 6.6 Discussion 200 6.6.1 Diversity in topic marking 200 6.6.2 Information status of marked topics and function of topic marking 204 6.6.3 A cognitive interpretation of topic marking: first attempt 208 6.7 Summary 209 Chapter 7 A Unified Interpretation of Topic_as Conceptual Reference Point 211 7.1 Reference Points are more than entities 212 7.1.1 ENTITY topics 213 7.1.2 Setting indicating topics 217 7.1.2.1 TIME topics 217 7.1.2.2 PLACE topics 220 7.1.2.3 MANNER topics 221 7.1.2.4 Setting indicating topics as Reference Points 222 7.1.3 CONJUNCTIVE topics 222 7.1.4 RESPECT topics 224 7.1.5 INTERPERSONAL topics 229 7.1.6 Summary 232 7.2 Reference Points can be multiple 233 7.2.1 Multiple topics as Reference Point Chains 233 7.2.2 Multiple topics as Reference Point Networks 236 7.2.3 Other multiple topics 238 7.2.4 On the limit to the number of marked topics in one clause 240 7.2.5 Summary 244 7.3 Schematizing Topic-Comment Constructions 245 7.3.1 Horizontal extension 245 7.3.1.1 Subject as prototype Reference Point 245 7.3.1.2 First extension from prototype Reference Point: topics in double-subject constructions 248 7.3.1.3 Second extension from prototype Reference Point: circumstantial topics 250 7.3.1.4 Further extension from prototype Reference Point: INTERPERSONAL and TEXTUAL topics 253 7.3.1.5 RESPECT topics 254 7.3.2 Vertical schematization 255 7.4 Summary 259 Chapter 8 Conclusion . 260 8.1 A Brief Review 260

Vlll 8.2 Revisiting the Controversies over Mandarin Topics 261 8.3 Conceptual Reference Point as a Conceptual Archetype 265 8.4 Areas of Further Study 266 8.5 Summary 270

Appendix: Characteristics of Monologs Comprising the Corpus 271

References 273

IX List of Abbreviations

Abbreviations Term AD Adverb AP Adjectival phrase ASL American Sign Language ASP Aspect marker AUX Auxiliary cc Complex/compound clause CL Classifier COMP Comparative CSC Complex stative construction {de) DE Nominalization construction (DE) Nominalization construction with de omitted DEMON Demonstrative pronoun DSC Double-subject construction FS False start GEN Genitive {-de) Interj. Interjection IU Intonation unit LD Left dislocation LD2 A subset of LD LBM Left boundary marker NOM Nominalizer {de) NP Noun phrase PART Pause particle for topic marking PF Pause filler PP Prepositional phrase PREP Preposition RC Relative clause REDUP Reduplication REP Repetition RBM Right boundary marker SFP Sentence-final particle SP Subject-predicate construction TAG Tag question TC Topic-comment TOP Topicalization VP Verb phrase List of Tables

Table 1.1 Properties of oral vs. written discourse 6 Table 5.1 Distribution and frequency counts of TOP-subject vis-a-vis SP in the Beij ing Corpus 110 Table 5.2 Coding forms and frequency counts of subject in SP vs. topic in TOP-subject in the Beijing Corpus 119 Table 6.1 Linguistic Context of Use of Particle ne within the Scope of Clause 141 Table 6.2 Discourse Functions of ne in Different Contexts .144 Table 6.3 Semantic Roles of «e-Marked Topics in Different Contexts 153 Table 6.4 Linguistic Contexts of Use of Particle ba within the Scope of the Clause 161 Table 6.5 Discourse Functions of ba in Different Contexts 161 Table 6.6 Semantic Roles of 6a-Marked Topics in Different Contexts 171 Table 6.7 Linguistic Contexts of Use of Particle a and its Phonetic Variants within the Scope of the Clause 174 Table 6.8 Discourse Functions of a and its Phonetic Variants in Different Contexts 176 Table 6.9 Semantic Roles of o-Marked Topics in Different Contexts 180 Table 6.10 Semantic Roles of fa-Marked Topics in Different Contexts 180 Table 6.11 Semantic Roles of wa-Marked Topics in Different Contexts 180 Table 6.12 Semantic Roles of wa-Marked Topics in Different Contexts 181 Table 6.13 Semantic Roles of >-a-Marked Topics in Different Contexts 181 Table 6.14 Linguistic Contexts of Use of Particle ma within the Scope of the Clause 187 Table 6.15 Discourse Functions of ma in Different Contexts 188 Table 6.16 Semantic Roles of mo-Marked Topics in Different Contexts 188 Table 6.17 Linguistic Contexts of Use of Particle dehua 192 Table 6.18 Linguistic Contexts of Use of Particle laijiang within the Scope of the Clause 192 Table 6.19 Discourse Functions of dehua in Different Contexts 193 Table 6.20 Discourse Functions of laijiang in Different Contexts 194 Table 6.21 Semantic Roles of dehua-Marked Topics in Different Contexts 196 Table 6.22 Semantic Roles of laijiang-Marked Topics in Different Contexts 196 Table 6.23 Frequency counts of marked ENTITY topics 204 Table 7.1 Frequency Counts of Top 10 Semantic Roles of Marked Topics 214

xi List of Figures

Figure 2.1 Topic as Specifier of TP 16 Figure 7.1 Schema Network of Topic as Reference Point 258

XI1 Chapter 1: Introduction

In this dissertation, I present a corpus study of topic constructions in Beijing

Mandarin, i.e., the variety of Mandarin as spoken by Beijing locals, and propose that

topic as a complex category is best interpreted in terms of Langacker's conceptual

reference point model (1991, 1993, 1998).

1.1 The Issue

Topic has been a major concern within Chinese grammar for more than four decades.

Ever since Li and Thompson (1976) laid the foundation for a typology on the basis of the

grammatical relations of subject-predicate and topic-comment, Mandarin Chinese has

been predominantly taken as a topic-prominent language, that is, in Mandarin Chinese,

"the basic constructions manifest a topic-prominent relation" (Li and Thompson

1976:459). However, a consensus has never been reached on how to define topic as a

notion, and different accounts have been made on its properties and its opposition with

subject. Also at issue is its syntactic and pragmatic status and how it should be identified.

In short, topic in Mandarin remains a notion in need of better characterization.

This reality can be attributed to the interface status of topic, in that it embodies syntactic, semantic as well as pragmatic properties. Therefore, it is not unexpected for the same construction to be interpreted very differently, especially when approached from different perspectives. The diversity in topic-comment (TC)1 construction encoding

1 For abbreviations used in this study, refer to List of Abbreviations, p.ix.

1 further complicates the issue. It seems that a rigorous definition in terms of necessary and

sufficient conditions does not fit for such a loaded notion. A new approach is thus in order.

In what follows I discuss what insights cognitive linguistics in general can give us in addressing the issue of topic in Mandarin Chinese.

1.2 Cognitive Linguistics as an Alternative

1.2.1 Outlining cognitive linguistics and its relevance to this research

Cognitive linguistics is a cover term for a variety of approaches, methodologies and emphases which are unified by a set of assumptions. The foundational point of cognitive linguistics is simply that "language is all about meaning" (Geeraerts 2008:3); that is, language is viewed as an instrument for organizing, processing and conveying information.

Cognitive linguistics makes several basic assumptions, and two of them are highly relevant to this research. One is that Cognitive Grammar views language as a usage-based system (e.g. Langacker 1987, 1990, 1999; Bybee and Hopper 2001; Taylor 2002).

Langacker (2001:143) unambiguously articulates that linguistic units are "abstracted from usage events, retaining as part of their value any recurring facet of the interactive and discourse context", and thus linguistic structures "incorporate discourse expectations and are interpretable as instructions to modify the current discourse state". The second assumption is that Cognitive Grammar "blurs many of the distinctions traditionally made

2 in linguistic theory" (Taylor 2002:30), and one of the blurred distinctions is that between semantics and pragmatics. In Cognitive Grammar, pragmatic aspects are "incorporated into the conventualized meaning of an expression. Because conventualization is a matter of degree, the distinction is a graded one, with no clear cut-off point between the entrenched meaning of an expression and its context-dependent interpretation" (Taylor

2002:30).

This first assumption is relevant here because this research takes a bottom-up approach; that is, by analyzing the actual use of topic constructions in a corpus of Beijing

Mandarin to abstract the linguistic conventions for topic encoding and decoding. This is exactly the approach that cognitive linguistics adopts for explorations of linguistic structures and ultimately theories and grammar. The relevance of the second assumption comes from the interface feature of topic. Understanding a notion that crosscuts syntax, semantics and pragmatics calls for an approach that takes into account all three, even though traditionally, they have been seen as necessary distinctions.

1.2.2 Topic as conceptual reference point

The reference-point phenomenon, according to Langacker (1993:5), is "so fundamental and ubiquitous in our moment-to-moment experience that we are largely oblivious to it", and he technically defines topic as a kind of "subjective reference point": the topic is used "to establish mental contact with another entity" (1991:314). From the cognitive perspective, topic is a notion beyond the syntactic level. Langacker (1998)

3 addresses the organization of reference point at various levels and makes a distinction between a clause-internal discourse topic and a subject, by holding that "although they may be conflated in a single nominal expression, these represent distinct functions which have to be kept separate for analytical purposes" (1998:18). He argues that a discourse topic can serve as reference point for any number of comment clauses, and the nominal establishing it is usually external to each of them. By contrast, a subject specifically functions as such only for a single clause, as an inherent part of its internal structure.

However, the two functions converge at the level of the clause, which, according to

Langacker, is "the smallest unit expressing a full proposition, and the largest one with a single overall trajectory2" (1998:20).

While Langacker's reference-point model offers a coherent, integrated view of topic and clarifies the relationship between topic and subject as affinitive but non-equivalent,

Langacker himself also acknowledges that detailed analyses are required of individual languages of diverse types in order for the model to be widely accepted as a viable and well-supported characterization of topic. Therefore, this research examines how topic is used in a spoken corpus and how different types of topic constructions can be understood under a single conceptual reference point model.

1.3 A Corpus Approach

In order to address the issue of topic and its use in discourse, I have decided to use a

2 Actions done by the trajector (Langacker 1987:268), where trajector is roughly the focal or most important participant in a given sentence (Evans 2007).

4 spoken corpus of around 200,000 Chinese characters,3 which is randomly sampled from a 1.8 million character corpus, A Corpus of Modern Spoken Beijing Mandarin.4

There are at least two advantages to using a spoken corpus. For one thing, to abstract the linguistic conventions for topic use in discourse is an approach favoured by cognitive linguistics for exploring linguistic structures. Given the inherently dynamic nature of conceptualization (Langacker 1998, 1999), conceptual structure emerges and develops through processing time. In this manner, naturally occurring data best show processing activities. A focus on spoken language data is thus one important consideration for the present research. Speech units, according to Chafe (1979, 1987, 1994, etc.), cut across several dimensions of language production, including prosody, syntactic structure, cognition, and interaction, and have a cognitive basis, whose formation reflects speakers' focus of consciousness.

The second advantage is that spoken data can better show how topics function in discourse. Givon (1995:29-30) outlines the cross-linguistic features of oral and written discourse as in Table 1.1. It should be noted, however, that the features are in terms of

"distributional tendencies (relative frequency)", rather than "in absolute presence or absence".

3 Due to the Chinese writing system, the size of the corpus is measured by character, which can be either a word or part of a word. 4 See Chapter 4 (p.80) for details.

5 Table 1.1 Properties of oral vs. written discourse parameter Oral-informal Written-formal Syntactic complexity conjoined embedded Grammatical morphology sparse abundant Word order flexible, pragmatic rigid, grammatical Processing speed slower, halting faster, fluent Context dependent higher lower

The properties of oral-informal discourse will provide more hints on how the speaker conceptualizes the situation and packages the information in discourse. For instance, a flexible and pragmatic word order may result in more TC constructions in oral data than in written texts. Similarly, halts in processing can be topic markers in many cases.

One feature that Brown and Yule extract for English can be intriguing to this study as well (1983:16-17): Whereas written language sentences are generally structured in subject-predicate form, in spoken language it is quite common to find what Givon (1979b5) calls topic-comment structure, as in the cats + did you let them out.

If English, a subject-prominent language in Li and Thompson's (1976) sense, tends to use

TC constructions in the spoken variety, it might be interesting to see what spoken data will reveal about TC constructions in topic-prominent languages, such as in Mandarin

Chinese.

In addition to these theoretical considerations, there is also a very practical one for using spoken data, namely, spoken data abound in the use of topic markers. Many studies, such as Chao (1968), Li and Thompson (1981), Tsao (1979, 1990), Chu (1998), etc.,

5 Brown and Yule's referencing.

6 maintain that a topic may be followed by a pause or a pause particle. In written texts, such topic markers may not be explicitly coded; in spoken discourse, however, a whole range of topic markers is available. On the basis of a conversational analysis, Tao

(1996:52) finds that "particles correlate to an overwhelmingly large degree with the end of intonation units", with NP, VE (i.e., verb expression) and XV (i.e., argument verb combinations) being three major types of intonation units in Mandarin Chinese. In addressing the issue of the syntax of speech units, Tao makes a proposal of a "dynamic process" by which cross-speech unit predications are made (1996:181-182): In this process, multiple speech units are combined to accomplish the task of referring and predicating, constituting a two-stage predicating process. This process is commonly found in our conversation data. ... many arguments are separated intonationally from the verb predicate, in line with the 'one new idea at a time' constraint. For these NP IUs,6 there is no syntactic rule that specifies the separation of the argument NP from the predicate, yet this is a common phenomenon in Mandarin discourse. To capture this fact in Mandarin discourse, only the proposal of the dynamic process makes it possible.

According to his frequency count, such nominal IUs make up 28.7% of all IUs , and these nominal IUs are evenly distributed across two types: attachable NPs (47.2%) and detached NPs (52.8%). By attachable NPs, Tao refers to "argument NPs that are intonationally separated from their verbal predicate", and detached NPs are those

"non-argument NPs with no syntactic relationship with any verbal predicate" (Tao

1996:75). Though Tao does not discuss his IU NPs in terms of topic constructions, it is known that Mandarin topics can be both "attachable" to and "detached" from the comment predicate. Moreover, referentiality, which is thought to be a fundamental feature

6 Intonation Units.

7 of topic, is one of the three general functions that Tao identifies for NP IUs. Thus Tao not only highlights the features of spoken Mandarin at the prosodic phrase level, but offers a good starting point for further study on spoken corpora as well.

1.4 Thesis Chapters

Following this introduction, Chapter 2 first reviews the major findings of previous studies on topic, with the controversies over its definition, characterization and identification being highlighted. Then a cognitive approach is proposed as an alternative.

It is argued that Cognitive Grammar's assumption that language is usage-based suits this bottom-up corpus study. The reference-point proposal made by Langacker opens up the possibility for a uniformed interpretation of topic-comment constructions in Mandarin. A brief mention is also made of topic studies in some other topic-prominent languages, with a view to understanding topic from a cross-linguistic perspective.

Chapter 3 begins with a detailed discussion on the functions of pause and pause particles in topic marking in Mandarin Chinese. Some principles for topic identification are then proposed for this corpus study. The chapter closes with the characterization of some grammatical terms which are useful for this study.

Chapter 4 is the methodology chapter, where the corpus data used for the research is described and the data analyzing procedure introduced. This chapter also further outlines the research goals of this dissertation.

Chapter 5 focuses on the use of pauses as topic markers in three construction types,

8 namely the nominal construction, the double-subject construction and the TOP-subject construction,7 with a view to revealing when pauses occur in language use and what impact a pause has on the message encoded in the pause-marked expression. Although

Shi (1995) classifies nominal constructions as a type where the occurrence of a pause is obligatory, the present corpus shows that the great majority are produced without pauses.

Likewise, the double-subject construction also tends to be produced in one intonation unit; that is, more often than not there is no pause within the construction. A close examination of pause-marked grammatical subjects vis-a-vis unmarked subjects reveals that the occurrences of pauses and pause particles correlate with the information status of the pause-marked element. The clause-initial element in TOP-subject constructions tends to be more complicated in coding form and accordingly more information-loaded than its counterpart in subject-predicate constructions. In this manner, the occurrence of pause turns out to be the requirement of the cognitive constraints on information flow. It is perhaps for the same reason that pauses are less frequently observed in nominal and double-subject constructions, given the fact that most of their instantiations in the present corpus are fairly short, thus not very heavily loaded with information. The affinity of prototypical double-subject constructions to possessives further reduces the rates of occurrence of topic markers.

Chapter 6 presents a comprehensive study of high frequency topic markers used in

Beijing Mandarin. Pause particles ne, ba, a (and its phonetic variants) and ma, and two

7 See Section 3.5.2 for a characterization of these construction types.

9 lexical topic markers dehua 'if and laijiang 'speaking of are analyzed in terms of frequency of use, distribution patterns and discourse functions. Following this, the diversity in topic marking is highlighted. It is also argued that definiteness may not be a fundamental property of marked topics; but rather, to ensure the identifiability of what is encoded in the topic expression seems more basic a function of topic marking. Finally, topic marking is cognitively interpreted as a conjoined requirement of the speaker's information packaging and the hearer's message interpreting. Thus it is an

"intersubjective" activity.

In Chapter 7, I propose a unified interpretation of all types of topics as conceptual reference points in discourse. The chapter begins with a detailed description of the diverse semantic meanings that topic expressions convey in the spoken corpus and how they function as reference points. Then multiple topic marking is discussed, with the finding that topic marking can occur either linearly or in a crosscutting manner. Finally the schematization of topic as a complex linguistic category is expounded first horizontally then vertically. Topic as a schema is then characterized as follows: topics of all semantic roles may fulfill the function of conceptual reference point in online speech, and the reference point may sometimes be objective, as opposed to Langacker's (1991) subjective interpretation, especially when the relationship between topic and its target referent(s) resembles some objective ones, such as the possessor-possessed relationship.

Chapter 8 concludes the study and suggests possibilities for future study concerning topic in Mandarin.

10 1.5 Summary

In this chapter, I first raise issues concerning the study of topic in Mandarin and then propose a cognitive approach as an alternative. Following this, an introduction is made to the corpus used for this research, with the advantages being highlighted. A chapter-by-chapter description constitutes the last major section of this introductory chapter.

11 Chapter 2: Review of Previous Studies on Mandarin Topics and

Proposal of the Cognitive Approach as an Alternative

Ever since Li and Thompson (1976) laid the foundation for a typology on the basis of the grammatical relations of subject-predicate and topic-comment, Mandarin Chinese has been predominantly taken as a topic-prominent language; that is, in Mandarin Chinese,

"the basic constructions manifest a topic-prominent relation" (Li and Thompson

1976:459). However, a consensus has never been reached on how to define topic as a notion within the language. Different accounts have been made on its properties and its opposition with subject. Also at issue is its syntactic and pragmatic status and how it should be identified. This reality can be attributed on the one hand to the fact that the construction is approached from different theoretical angles, and on the other to the heterogeneity of topic-comment constructions in the language. Below is a summary of the findings in previous studies, with the important controversies highlighted, which is then followed by a proposal that the category of topic is best understood as a conceptual reference point schema for ensuing comments (Langacker 1991, 1993, 1998) and that different types of topic constructions are instantiations of the same schema. Studies in some other topic-prominent languages are also briefly discussed towards the end of this chapter with a view to understanding Mandarin topics in a cross-linguistic context. 2.1 Findings on and Controversies over Topic

2.1.1 Topic versus subject

Topic is perhaps the most frequently used term in discourse analysis and many

definitions have been proposed. The most generic definition of sentence topic is perhaps

Hockett's "what the sentence is about" (1958:21). This definition, however, has been

criticized as being too vague to provide rigorous criteria for topic identification (e.g.

Chafe 1994). This partially explains why many other definitions under various

terminological guises have been proposed. When talking about Mandarin grammar, Li

and Thompson (1981:15) start from Hockett's definition, stating that "basically, the topic

of a sentence is what the sentence is about", and by taking Chafe's (1976) viewpoint they elaborate "aboutness" as to say that "a topic sets a spatial, temporal, or individual

framework within which the main predication holds" (1981:85). On the basis of this characterization, they identify the following four properties for Mandarin topic

(1981:87): a. A topic is typically a noun phrase (or a verb phrase) that names what the sentence is about. b. A topic is definite or generic. c. A topic occurs in sentence-initial position. d. A topic may be followed by a pause or a pause particle.

In contrast, the subject of a sentence in Chinese is defined as "the noun phrase that has a 'doing' or 'being' relationship with the verb in that sentence" and the precise nature of this relationship "depends on the semantic makeup of the verb" (Li and Thompson

13 1981:87). As for the properties of Chinese subjects, a summary is found in Tsao

(1990:170): a. Subject is always unmarked by preposition. b. By position, subject can be identified as the animate NP to the left of the verb; otherwise, the NP immediately before the verb. c. Subject always bears some selectional relation to the main verb of the sentence. d. Subject tends to have a specific reference. e. Subject plays an important role in the following co-referential nominalization or deletion processes: reflexivization, imperativization, and Equi-NP deletion.

If the above characterization is followed, topic at first look seems to differ from subject in a systematic manner. For instance, topic, unlike subject, is not constrained by animacy or selectional relations. When the coding form is concerned, topic can take marking (e.g. a pause or pause particle), whereas subject can never be marked by preposition. Also, topic is not as sensitive as subject to such grammatical processes as reflexivization, imperativization, etc. Upon closer examination, however, these differences are no more than tendencies. When faced with spontaneous speech, the definitions and properties often fall short of expectation. As a result, scholars have very divided opinions on the opposition between topic and subject in Mandarin Chinese.

According to Shi (2001), there are three different views. The first view is represented by

Chao (1968), Li (1990) and Xu (1997), arguing that there is only topic and that subjects in Mandarin are in fact topics. Take Chao (1968) for instance: he equates topic-comment with subject-predicate. The second, represented by Li and Thompson (1976), Tsao (1977) and Shen (1999), perceives both topic and subject, with each showing different properties. The third view acknowledges only subject and is most advocated by Chinese

14 grammarians such as Zhu (1985) and Lii (1984). The third view is somewhat unusual,

but since both Zhu and Lii are grammarians, they have chosen to take some typical

topic-comment constructions, such as topicalization, as evidence of syntactic flexibility.8

The reason for the co-existence of these conflicting views is multifold. First, topic is a cover term for varying types of constructions. More importantly, the discrepancies can be attributed to the status of the notion of topic—it lies at the juncture of syntax, semantics and pragmatics. The complex status of topic as a notion can also explain why so many linguists are addressing the same issue, and it is not infrequent for different approaches to come to very similar conclusions or for one approach to result in different findings (Xu and Liu 1998:3). Perhaps Sanders (1984:222) is right in stating: "It's simply true in general that empirically significant concepts are inherently incompatible with rigorous definition, i.e., in terms of necessary and sufficient conditions, except within the specific context of a particular scientific theory".

2.1.2 Syntactic topic versus pragmatic topic

Just as researchers have conflicting views on how to define topic vis-a-vis subject, the syntactic or pragmatic status of the notion also remains an unsettled issue. Some researchers take topic as a syntactic unit. To them, a sentence or clause is usually made up of a topic followed by a comment (Li and Thompson, 1976, 1981; Chafe, 1976; Givon,

1983; among others). Xu and Liu (1998:275) propose that topic has the same syntactic

8 All quotations from literature published in Chinese are translated and transliterated (where applicable) by me.

15 status as subject and object and that the prominent sentence pattern of Mandarin is TSVO

(topic-subject-verb-object) in contrast with TC or SVO (subject-verb-object). Their

definition is purely syntactic: topic is what is inside the TP (topic phrase) but outside the

IP (inflectional phrase). Diagrammatically, this is shown in Figure 2.1 (Xu and Liu

1998:34):

Figure 2.1 Topic as Specifier of TP

TP

Spec .. T'

NP T IP I I A Zhege ren a wo taoyan. This man PART I hate. 'This man I hate.'

Lambrecht (1994), though not focused on Mandarin, explicitly states that information

structure is "a component of grammar, more specifically of sentence grammar", thus

identifying topic as sentential or clausal.

In contrast, Tsao's PhD dissertation A Functional Study of Topic in Chinese: The First

Step towards Discourse Analysis (1979) marks the first explicit proposal that topic is

discoursal. Accordingly, he modifies Li and Thompson's (1976) characterization of topic

as taking sentence initial position by arguing that topic takes the initial position of a topic chain which can extend over a series of clauses or sentences (1979:38). Chu (1998), on the other hand, argues for a more restricted view. He contends that, because topic is mainly a discourse notion, it can be identified only on the level of discourse when it

16 serves as an interclausal link. To him, there is no point in talking about a topic within a single sentence.

Between the two extremes is Gundel's (1985:86) definition of sentence topic from both the syntactic and pragmatic perspective: A construction, C, is the syntactic topic of some sentence S, iff C is immediately dominated by S and C is adjoined to the left or right of some sentence S' which is also immediately dominated by S. An entity, E, is the pragmatic topic of a sentence, S, iff S is intended to increase the addressee's knowledge about, request information about or otherwise get the addressee to act with respect to E.

It seems that topic to Gundel is both a syntactic and pragmatic unit. In a similar manner,

Shi, a formalist, views the TC construction in Chinese as "a grammatical device used to fulfill certain discourse functions and it is derived9 from basic sentence structure via syntactic operation" (Shi 2000:386). She thus defines topic and comment as below (ibid): A TOPIC is an unmarked NP (or its equivalent) that precedes a clause and is related to a position inside the clause; a topic represents an entity that has been mentioned in the previous discourse and is being discussed again in the current sentence, namely, topic is what the current sentence is set up to add new information to. The clause related to the topic in such a way is the comment.

Janzen's (1998) empirical study of marked topics in American Sign Language (ASL), also a topic-prominent language, seems to support a dialectic approach. While acknowledging topic as falling within the clause domain and thus having a bearing on syntax, Janzen maintains that "marked topic constructions in the present study are best described as 'pivots' in the organization of discourse" (1998:vii). This finding is very interesting, given the observation that the clause is a basic unit of discourse (e.g.

9 In the formal approach to Mandarin grammar, there are two distinct views. One takes topic as base-generated (e.g. Xu and Langendoen 1985) and the other interprets it as derived (e.g. Huang 1982). Shi (2000) echoes the second view.

17 Langacker 1999). Janzen's study is also interesting in that it is based on naturally occurring ASL discourse and that his marked topics are examined in specific contexts rather than as isolated units. This research also studies Mandarin topic constructions in actual use, and it is expected that such an approach can to some extent resolve this controversy over the syntactic or pragmatic status of topic.

2.1.3 Central versus marginal properties of topic

This section presents a review of several lists of topic properties as proposed by different researchers and raises the question whether some properties are central while others are marginal.

It has been mentioned in Section 2.1.1 that Li and Thompson (1981) recognize four properties for Mandarin topics at the clause level. When in the context of discourse, Tsao

(1979:38-39, 1990:48) proposes that the following six properties differentiate topics from subjects: a. Topic always takes the clause-initial position of a topic chain; b. Topic may be separated from the rest of the sentence with pause particles a (yd), tie, me and ba; c. Topic is always definite; d. Topic is a discourse notion and it may and often does extend its semantic domain over a number of clauses; e. Topic controls pronominalization or deletion of co-referential nominal phrases within the topic chain; f. Topic, unless being the functional subject simultaneously, is syntactically independent of such processes as reflexivization, passivization, Equi-NP deletion,

10 It should be noted that Janzen's and Langacker's "clause" is more or less the same in reference with "sentence" as used in other studies discussed in this subsection. For instance, Xu and Liu's example sentence cited here can be referred to as a clause as well in Janzen's and Langacker's term. In the present study, "clause" is used to refer to the verb together with its core argument(s). See Chapter 3 for details.

18 verb serialization and imperativization.

Shi (2001) examines Mandarin topics against the background of world languages and emphasizes the syntactic behaviour of topic versus subject. He draws the following conclusion regarding topic properties vis-a-vis subject properties (2001:65):.

subject topic 1. cannot be marked by focus mark 1. can be marked by focus marker shi shi 2. cannot be prompted by questi< 2. can be prompted by question words words 3. can be used in matrix clauses only 3. can be used in matrix and subordinate clauses 4. referring to a definite entity 4. agent of an action or the entity of an attribute

Shi illustrates his point with the following two sets of examples (2001:62):

(l). a. itej^M^iMiW^.

Ta-baba congzaodaowan dou bu zhanjia.

His-father from-morning-to-evening all not at home

'His father barely stays at home.' b. M&ggL^mmn^m

Shi ta-baba congzaodaowan dou bu zhanjia.

SHI his-father from-morning-to-evening all not at home

'It is his father who barely stays at home.' c mhkw-^mu^f^^

Shui congzaodaowan dou bu zhanjia?

Who from-morning-to-evening all not at home

'Who barely stays at home?'

19 Ta-baba congzaodaowan dou bu zhanjia de shi

his-father from-morning-to-evening all not at home NOM thing

ren-ren dou zhidao.

person-REDUP all know

'Everyone knows that his father barely stays at home.' (2). a. jbg&m, /A«M»«o

Ta-baba ma, congzaodaowan dou bu zhanjia.

*Shi ta-baba ma, congzaodaowan dou bu zhanjia.

*Shui ma, congzaodaowan dou bu zhanjia? d. *&£&&, JA^-nmuT^m^mAAmmo

* Ta-baba ma, congzaodaowan dou bu zhanjia de shi ren-ren dou zhidao.

(1) shows the properties of subject. In (1), ta-baba 'his father' is the subject of (l)a, and because of its subject status, it can be preceded by the focus marker shi in (l)b, prompted by question word shut 'who' in (l)c, and used in the subordinate clause headed by deshi

'thing'. In contrast, (2) illustrates the syntactic constraints on topic use. The occurrence of pause particle ma grants the subject NP ta-baba in (1) topic status, and the topic status of the NP renders (2)b, c and d ungrammatical.

A more comprehensive summary of topic properties in Mandarin is found in Xu and

20 Liu (1998:28-32):

Semantic features: 1. Topic is what the ensuing part of the sentence is about; 2. Topic is the agent, theme or another argument of the main verb or the verb in a subordinate clause; it can also be circumstantial elements of time or place. Syntactic features are: 1. Topic takes sentence initial position; 2. Topic is pre-positioned before comment; 3. Topic can be subsequently omitted; 4. Topic may be followed by a pause; 5. Topic can be marked; 6. Promoted topic may have its co-referring element in the original position; 7. Topic cannot take natural sentence stress; 8. Topic can be shared by more than one sentence. Pragmatic features are: 1. Topic is definite; 2. Topic is given information; 3. Topic is shared information of the addresser and the addressee; 4. Topic is activated information; 5. Topic is what the addresser wants the addressee to take as center of attention; 6. Topic is opposed to focus and thus cannot be focus.

It should be noted that the semantic, syntactic and pragmatic features are Xu and Liu's categorization. Other scholars may have different classifications. As far as the present study is concerned, these terms can be roughly defined as follows: syntax is the study of how words combine to form sentences and the rules which govern the formation of sentences; semantics is the study of how meaning in language is created by the use and interrelationships of words, phrases, and sentences, without reference to the users and communicative functions of sentences; and pragmatics refers to the study of the use of language in communication, particularly the relationships between sentences and the contexts and situations in which they are used.

21 On basis of what they have found in standard Mandarin and the Shanghai dialect, Xu and Liu propose the following characterization of topic (1998:280-281):

Coding features: 1. Topic is pre-positioned in terms of word order; 2. Topic is or can be followed by a pause; 3. The pause following topic can be materialized with a pause particle; 4. Topic does not carry focus stress. Semantic-functional features: 1. Topic fulfills three discourse functions, namely framework function (i.e., setting the temporal, spatial or individual framework1' for the forthcoming discourse), relevance function (indicating how the forthcoming discourse is relevant to the encoded topic) and start-point function (offering a starting point for the comment); 2. In order to fulfill its discourse function, topic is required to be definite, generic, universally quantificational, or definite existentially quantificational in terms of referential ity; indefinite referential expressions are thus incompatible with topic; 3. Topic usually conveys known, shared or situational ly identifiable information, especially information that has just been activated; non-shared new information does not occur in this position; 4. Topic is not restricted by semantic roles, so long as the expression meets the functional-informational requirements as stated above.

Perhaps Chu (1993) is the first and only one to organize different properties into a hierarchy of relative significance in topichood perception, or in Chu's terminology, primary, secondary and non-essential attributes. For the prototype of Mandarin topics, he presents the following list of hierarchical attributes (1993:37): A. Primary Attributes of a Topic a. Being nominal b. Serving as an interclausal link B. Secondary Attributes of a Topic a. Being specific/referential

" This seems to resonate with Chafe's (1976) Chinese style topics.

22 b. Occupying the sentence-initial/preverbal position C. Non-Essential Quality: Bearing no selectional relations to the predicative verb

One obvious advantage of this attribute hierarchy is that the topic status of an expression can be calculated in terms of the number of primary and secondary attributes. Also, the hierarchy allows for competition of topichood within the domain of clause or sentence, given that the sentence-initial position is an attribute secondary to being nominal.

Nonetheless, this hierarchical list might need some substantiation, if only on account of the fact that the topic expression in Mandarin may very well be coded as a verb phrase

(e.g. Li and Thompson 1981:87). If being nominal is a primary attribute of Mandarin topics, how can a VP win the competition for topichood over an NP that occurs in the same clause or sentence?

It is obvious from the above discussion that the findings are as diverse as the researchers. The proposed properties cover a range of aspects which includes the coding form, the syntax, semantics as well as pragmatics of the topic expression. It seems that the only consensus that has been reached concerns the initial position of the element and its function which is very vaguely put as "aboutness". As a result of the diversity in topic property characterization, topic identification is more or less the researcher's personal choice. The following section introduces the various topic constructions as identified by different researchers. 2.1.4 Topic identification

Li and Thompson (1981:88-93), on basis of their definition and characterization of topic, identify the following types of TC constructions. First, topic may co-occur with a different subject.

(3) j8PKffl%B£#a7.

nei - zhi gou wo yijing kan - guo le.12

that - CL dog I already see-EXPCRS

'That dog I have already seen.'

In (3), nei-zhi gou 'that dog' is the topic and wo 'I' the subject. Topic and subject can also be conflated. In (4), wo T is both the topic and subject.

(4) H^MlIo

wo xihuan chi pingguo.

I like eat apple

'I like to eat apple.'

Meanwhile, they interpret the so-called double-subject constructions as consisting of both topic and subject. In the following example, xiang 'elephant' is the topic and bizi 'nose' the subject:

(5) £#iH£.

xiang bizi chang.

12 Examples used in this section are taken from their respective authors, unless otherwise noted. In this particular example, EXP and CRS are Li and Thompson's glosses, used for experiential aspect and currently relevant state respectively.

24 elephant nose long

'Elephants' noses are long / Elephants have long noses.'

Chafe is among the first to discuss Chinese style topics, which, according to his

observation, are to "limit the applicability of the main predication to a certain restricted

domain...Typically, it would seem, the topic sets a spatial, temporal, or individual

framework within which the main predication holds" (Chafe 1976).

A more detailed and language-specific discussion is found in Chen (1996). In addition to such TC constructions as topicalization (TOP) and left-dislocation (LD), which are

also observed in languages such as English, Chen (1996) makes a finer differentiation and identifies three subtypes of TCs in Mandarin Chinese, namely instance topics, frame topics and range topics. The characterization for each type follows (1996:399): An instance topic represents an instance of the object about which a predication is made and assessed. It is typically a definite entity in the cognitive inventory of referential entities in the context. A frame topic is one that provides the spatial, temporal and individual frame within which the proposition expressed by the remaining part of the TC, typically a predication made of another expression in the sentence, normally that of the subject, holds true. A range topic is one that delimits the range of a variable of which the predication is made.

For illustration, he gives the following examples, the underlined indicating the topic:

Instance topic: (6) ^atf]e^it*7o

Lao Li women yijing qing lai le.

Lao Li we already invite come CRS

25 'Lao Li, we have already invited here.'

Frame topic: (7) ±frnmm?immmTo

shang ci jiaoyou haizimen dou lei ji le.

last CL outing children all tired extremely CRS

'On the last outing, the children were all exhausted.'

Range topic:

(8) MlM$3%%°

wujia Niuyue zui gui.

price New York most expensive

'Speaking of prices, New York is the most expensive.'

Chu (1998) focuses more on the formation of topics, holding that topics can be formed in Chinese either in an unmarked or a marked way. According to Chu (1998:261),

"When a topic is introduced and goes on without any special morphological or syntactic marking, it is called an unmarked topic. They are the ones that cannot be recognized as topics when they are first introduced. They become topics only when they are picked up by a co-referring nominal, pronominal, or zero in a subsequent clause". Chu

(1998:266-273) identifies six types of marked topics, for which he argues, "As the topics are clearly marked morphologically or syntactically, they can be easily recognized within the boundaries of the clause or sentence in which they occur without any larger context".

The six marked topic types are illustrated below (Chu, 1998:266-271):

26 Type 1: Double-nominal construction; Type 2: Nominal followed by a pause particle; Type 3: Introduced by prepositions like zhiyu ('as far as ... is concerned'), duiyu ('as for'), etc.; Type 4: Comparison; Type 5: The ilian...dou/ye'' construction;13 Type 6: The ba- and 6e/-sentences.14

It should be noted, however, that other scholars may have different interpretations for these constructions. For instance, Tsao (1990) takes the double-nominal construction as a primary topic followed by a secondary topic that has a bearing within the clause boundary and the ^-construction as a secondary topic holding sway beyond the clause boundary.

Shi (2000) identifies six types of so-called dangling topics which are not subcategorized by the verb in the comment and are therefore not related to any position inside the comment. The first type is an idiomatic expression where the interpretation is a non-literal phrasal one. The underlined is the topic in all six examples.

(9) Ml^#^#3&»

Tamen wo kan ni ni kan wo.

they I look you you look me

'They look at each other.'

In (9), the topic expression taman 'they', according to Shi, is not related to any position inside the two-clause comment, where the subject and object in the first clause wo kan ni

'I look at you' is wo T and ni 'you' and the subject and object in the second clause ni kan

13 This is a construction to give emphasis to the element occurring in the slot represented by the dots. 14 Ba is a means to pre-pose the patient of an action while bei introduces the agent in the passive.

27 wo 'you look at me' is ni 'you' and wo T. Therefore, the topic is a "dangling" one and the sentence as a whole has an idiomatic reading.

The second type is characterized with an initial NP followed by a WH-word licensed by the emphatic morpheme dou 'all'. These constructions are usually considered as a subtype of double-nominative constructions (Teng, 1974).

(10) Mli£»*o

Tamen shui dou bu lai.

they who all not come

'They (topic), none of them are coming.'

The third type is overtly marked by adverbs such as xingkui 'fortunately', kexi

'pitifully', andyao-bu-shi 'if not for', etc. (ii) mmm^m^i^&Tm*

na - suo fangzi xingkui qu-nian mei xia-xue.

that-CL house fortunately last-year not snow

'That house, fortunately it didn't snow last year.'

The underlined NP in this example is the topic, setting an individual frame for the ensuing part of the sentence. The adverb xingkui is used to highlight the implied meaning, namely, that the house is in very poor condition and a snow may make it collapse.

The fourth type is best represented by the following example where the argument for the dangling status of the topic is that it is not in a subject-predicate relation with the comment.

28 (12) mmmifc^mytmrn-^Ao

Zhe-jian shiqing ni bu neng guang mafan yi - ge ren.

this-CL matter you not can only bother one-CL person

'This matter (topic), you can't just bother one person.'

The subject of the comment clause is ni 'you', which immediately follows the underlined topic expression, and the predicate is the remaining part of the clause. The topic expression specifies the "frame" of the comment proposition as 'this matter', in contrast with any other matter.

The fifth type consists of two sets of NPs but has no verb to head the predicate.

(13) iS#5?-/rH+jjfet)S.

Na - zhong douzi yi-jin sanshi-kuai qian.

that-CL beans one-CL thirty-CL money

'That kind of beans (topic), one catty15 is thirty dollars.'

The two sets of NPs in (13) are the underlined topic NP and what remains of the clause, with no verbs occurring in between.

The last type is what Chen (1996) refers to as the range topic. The same sentence is reproduced below as (14), where the topic wujia 'price' defines the range of the comment clause, i.e., the comment is true only with respect to prices.

(i4) Mjja^SMo

Wujia Niuyue zui gui.

15 A traditional Chinese measurement equal to half a kilogramme.

29 price New York most expensive

'Speaking of prices, New York is the most expensive.'

Tsao (1979:46) identifies 15 types of topic-comment constructions, the most

inclusive in the available literature. Under closer examination, I think that the 15 types

fall under three subgroups depending on how the topic is related to the comment, namely

core-argument topics, non-core-argument topics and those related to a constituent within

the comment expression. Below are Tsao's examples, where a is a pause particle and the

bracketing indicates optional occurrence of the pause particle.

Core-argument topics

(15) 3£H (PR), Rl^*#&.

Zhangsan(a), zuotian lai kan wo.

Zhangsan, yesterday come see me

'Zhangsan came to see me yesterday.'

(16) &&# (PR), Xxto

Zhe - ben shu (a), zhen nan.

this-CL book, really difficult

'This book is really difficult.'

The topic expressions in these two examples are the functional subjects of their respective comment expressions.

(17) &*# (PR), ftB^THl. j£»ffio

Zhe- ben shu (a), wo vijing nian le san - bian, haishi bu dong.

30 this-CL book, I already read ASP three-CL, still not understand

'This book, I've read three times, but still can't understand.'

(18) ^M ("R). S£egj£7-ffr*L.

Lisi_(a), wo yijing song le yi-fen li.

Lisi, I already give ASP one-CL gift

'Lisi, I've already given a gift.'

(19) &;£4$g|jtt4J (PR), 3£i«^o

Zhe-ben shu gen na-ben shufa). wo dou yao mai.

this-CL book and that-CLbook, I both want buy

'This book and that book, I want to buy both.'

The topic expression in these three examples fulfills the role of functional object.

Non-core-argument topics

(20) B£^ (OH), &H*#f£o

Zuotian (a), Zhangsan lai kan wo.

yesterday, Zhangsan come see me

'Yesterday, Zhangsan came and see me.'

The topic specifies the time frame of the comment. (2i) mikm (DH) mT&nu±, ra

Na-kuai tian (a) daozi zhang-de hen da, hen zhiqian.

that-CL field rice grow-AUX very big, very valuable

'That field, rice grows very big; it is very valuable.'

31 (22) itM. (W. W'Nfeg.

Beijing (a), you ge Gugong.

Beijing, exist CL Palace Museum.

'Beijing has a Palace Museum.'

(23) ikPM£ (BR), W'Nfeg.

Beijing cheng li (a), you ge Gugong.

Beijing city inside, exist CL Palace Museum

'Inside the Beijing city, there is a Palace Museum.'

The topics in (21) to (23) specify the spacial frame of the comments.

(24) gam («), *T-^AO

Zheli zheshi (a), lai le yi-ge ren.

here now, come LE one-CL person.

'Here and now comes a person.'

In (24), the topic specifies both the place and time of the comment.

Topics related to some constituent in the comment

(25) JIH-A («), MiW, EISLR&O

Zhe-ge ren (a), tounao jiandan, sizhi fada.

this-CL person, head simple, limbs strong

'This person, his head is simple; his limbs are strong.'

In (25), the topic is in a whole-to-part relationship with the two subject expressions, tounao 'head' and sizhi 'four limbs' respectively, in the comment.

32 (26) ±S (PR), ife*, tlW, API.

Zhongguo(a), di da, wu bo, renkou duo.

China, field broad, produce rich, population many

'China is broad in territory, rich in produce and big in population.'

Here, the topic is also the whole with respect to the three aspects of it described in the

comment. (27) mm^m? («), -^mwm, -^mp±, asw-^gm.

Ta de san-ge haizi (a), yi-ge dang liishi, yi-ge dang hushi,

he POSS three-CL children, one-CL be lawyer, one-CL be nurse,

haiyou yi-ge xue jianzhu.

still one-CL study architecture.

'His three children, one is a lawyer, another is a nurse, and still another studies

architecture.'

The three entities introduced by the topic are elaborated one by one in the comment.

(28) Z^+TNit («), 7£3j±ito

San-shi-liu ii (a), zou wei shang-ji.

three-ten-six strategem, leave be good-strategem

'Among the thirty-six stratagems, leaving is a good one.'

(29) M, CR). ^t&HSft^SH.

Yu_(a),jinqiangyu xianzai zui-gui.

fish, tuna now most-expensive

33 'Speak offish, tuna is now the most expensive.'

The topics in (28) and (29) specify the class, with the subject in the comment being a

member of the class.

From the above discussion, it can be seen that topic identification varies with the

researcher. Li and Thompson (1981) acknowledge conflated topic and subject; by contrast, Shi (2000) focuses on dangling topics only. While Chu's (1998) identification is more structure-based, Tsao (1979) and Chen's (1996) are more semantically oriented. The subjectivity in topic identification calls for some criteria that can be followed in a more rigid manner.

2.1.5 Summary

To summarize, topic in Mandarin remains a notion in need of better characterization.

Given its interface status, it embodies syntactic, semantic as well as pragmatic properties.

Therefore, it is not unreasonable for the same construction to be interpreted very differently, especially when approached from different perspectives. The diversity in TC construction encoding further complicates the issue. It seems that a rigorous definition in terms of necessary and sufficient conditions does not fit for such a loaded notion. A new approach is thus in order. In what follows I discuss what insights cognitive linguistics can give us in addressing the issue of topic in Mandarin Chinese. 2.2 A Cognitive Approach to Topic

2.2.1 An overview of cognitive linguistics

Cognitive linguistics is a cover term for a variety of approaches, methodologies and emphases which are unified by a number of assumptions. Foremost among these assumptions is the belief that language forms an integral part of human cognition, and that any insightful analysis of linguistic phenomena will need to be embedded in what is known about human cognitive abilities. Accordingly, cognitive linguistics aims for a cognitively plausible account of what it means to know a language, how languages are acquired and how they are used. The foundational point of cognitive linguistics is simply that "language is all about meaning"; that is, language is viewed as an instrument for organizing, processing and conveying information (Geeraerts 2008:3). To cognitive linguists, a formal semantics based on truth conditions is inadequate for describing the meaning of linguistic expressions. Langacker (2008:29) suggests two reasons: "One reason is that semantic structures are characterized relative to knowledge systems whose scope is essentially open-ended. A second is that their value reflects not only the content of a conceived situation, but also how this content is structured and construed."

Of the basic assumptions that cognitive linguistics makes, two are highly relevant to this research. One is that Cognitive Grammar views language as a usage-based system

(e.g. Langacker 1987, 1990, 1999; Bybee and Hopper 2001; Taylor 2002). In a usage-based model, "substantial importance is given to the actual use of the linguistic

35 system and a speaker's knowledge of this use; the grammar is held responsible for a speaker's knowledge of the full range of linguistic conventions, regardless of whether these conventions can be subsumed under more general statements" (Langacker

1987:494). This assumption is relevant because this research takes as its focus the actual use of topic constructions in a corpus of spoken Beijing Mandarin in order to abstract the linguistic conventions for topic encoding and decoding. This is exactly the approach that cognitive linguistics adopts for explorations of linguistic structures and ultimately theories of grammar.

The second assumption is that Cognitive Grammar "blurs many of the distinctions traditionally made in linguistic theory" (Taylor 2002:30), and one of the blurred distinctions is that between semantics and pragmatics. As is evident from the discussion above, topic is a notion that lies at the juncture of syntax, semantics and pragmatics, thus understanding its function and use calls for an approach that takes into account all three, even though traditionally they have been seen as necessary distinctions. In Cognitive

Grammar, pragmatic aspects are "incorporated into the conventualized meaning of an expression. Because conventualization is a matter of degree, the distinction is a graded one, with no clear cut-off point between the entrenched meaning of an expression and its context-dependent interpretation" (Taylor 2002:30). Langacker (2001) also makes it explicit that linguistic units are abstracted from language use and retain as part of their value any recurring facet of the interactive and discourse context. This is also a mainstay of functional approaches to linguistics which claim that linguistic elements and patterns

36 that are frequently used in discourse become conventionalized as grammar (e.g. Hopper

1998, Bybee and Hopper 2001).

Under the first assumption, the notion of schema is defined as "the commonality that emerges from distinct structures when one abstracts away from their points of difference by portraying them with lesser precision and specificity" (Langacker 1999:93). A schema is thus acquired through exposure to actually occurring expressions which instantiate it, and the schematic pattern becomes entrenched through repeated activation. While this abstraction or schematization is a vertical process, category formation (i.e., making extensions from prototype) is something horizontal. According to Langacker (1993:2), both processes "can be carried to various lengths: we thus encounter chains of extensions, as well as successive schematizations that reach progressively higher levels of abstraction". Langacker also holds that there is a close relationship between the two processes: "extension from a prototype is commonly based on some perceived similarity between two structures; the apprehension of their similarity amounts to the extraction of a schema that both structures instantiate" (ibid). Consequently, in the formation of a complex category, horizontal extension from the prototype tends to accompany vertical schematization.

Traditionally, various forms to topics have been labeled with the pragmatic-functional tag of "aboutness". Inclusive as "aboutness" is, the term is not free from being vague. In the following section, therefore, I introduce the notion of "conceptual reference point" from Cognitive Grammar and explore what insights it may bring to topic study.

37 2.2.2 Cognitive Grammar on reference point

Langacker (1993) proposes that a cognitive reference point is a construct whose

general psychological significance is as apparent as the notions of force dynamics, image

schemas, subjective vs. objective construal,16 and correspondences across cognitive

domains17 or mental spaces.18 "It appears, in fact, that reference points are fundamental

to both linguistic and cognitive organization" (1993:2), and "the reference point

phenomenon is so fundamental and ubiquitous in our moment-to-moment experience that

we are largely oblivious to it" (1993:5).

Conceptualization as viewed by Langacker (1997, 1999) is "inherently dynamic", in

that "it develops and manifests itself through processing time, the specific nature of its

temporal development being essential to its values. ... One aspect of dynamic

conceptualization is our very general cognitive ability to direct attention to some entity19

for purposes of then establishing 'mental contact' with another" (1998:7). In this context,

Langacker speaks of a conceptual reference point (R) as "providing mental access to a target (T) found in its dominion (D) ".20 When talking about its operation, Langacker

In talking about construal, Langacker takes an entity as 'subjective' to the extent that "its role as observer is maximized, and its role as object or observation is minimized" (1987:493); in contrast, an entity is "objective to the extent that its role as observer is minimized and its role as object of observation is maximized" (1987:491). 17 Any coherent body of conceptual content which serves as an essential background for, or is presupposed by, some individual concept or conceptual process (Cruse 2006:52). 18 Originally introduced by Fauconnier to explain a range of semantic phenomena mostly connected with reference. They are temporary, limited packages of conceptual content set up to assist understanding and guide action on particular occasions (Cruse 2006:104). 19 Entity as used by Langacker is "a convenient cover term for anything we might conceive of or refer to for analytical purposes: things, relations, locations, points on a scale, sensations, interconnections, values, etc. It is especially important to note that an entity is not necessarily discrete, separately recognized, or cognitively salient" (Langacker 1987:198). 20 In the reference point model, the target is "the object that the viewer seeks to locate", and the dominion is a region anchored by the reference point (Langacker 1991:170).

38 continues, "By serving as an initial focus of attention, the reference point tends to activate an array of associated entities which collectively constitute its dominion. This enables attention to shift to a secondary focus within it, the target (which activates its own dominion and may in turn function as a reference point)" (1998:7). In this manner, the reference point has both a certain cognitive salience and a dynamic aspect to it.

Van Hoek (1997:58-59) identifies three general considerations that underlie the reference point patterns, namely prominence, linear word order and conceptual connectivity:21 Prominence. X is likely to be taken as a reference point relative to Y if X is prominent in the context which includes Y. This reflects the basic nature of a reference point as something which is selected on the basis of salience and used as a starting point from which to make mental contact with other, less salient entities.

Linear Word Order. Other things being equal, X is more likely to be taken as a reference point in relation to Y if X is introduced into the conceptualizer's awareness before Y is. As with prominence, the role of linear order is motivated by the very conception of a reference point as something which is identifiable and available to the conceptualizer before the conceptualizer makes mental contact with the things in the reference point's dominion.

Conceptual Connectivity. The extent of a reference point's dominion is determined by the interaction of conceptual (semantic) connectivity and linear word order. Connectivity between two nominals is determined by the relations in which they participate. Nominals are strongly interconnected when they participate in (i.e., elaborate) an explicit interconnecting relation, as in the complement chain. Nominals are more weakly interconnected when they merely co-occur within a single linguistic unit of some kind, such as a sentence or a conceptual paragraph.

21 Though van Hoek proposes these three considerations as constraining anaphora, I see no reason why they cannot be generalized to other nominal phenomena like topic control, especially given the fact that they are depicted in very general terms. As a matter of fact, van Hoek also holds that "the basic components of the reference point model—prominence, semantic connectivity, and linear order—are posited as cognitive and linguistic universals" (1997:228).

39 Moreover, the reference point-target relationship can be manifested at different levels of organization: "Roughly speaking, the higher the level of organization, the looser the connection may be between a nominal reference point and a process in its dominion"

(Langacker 1998:16). Of particular interest here is the level of clause. As the expression of a single proposition, a clause is "the smallest unit of discourse over which a topic functions as the active reference point; at the same time, a clause is the largest unit for which we can posit a single overall trajector" (Langacker 1998:19). Therefore, clause constitutes a unit ideal for grammatical as well as discourse analysis.

In this particular research, spoken data are used to demonstrate the dynamicity of conceptualization, and clause taken as the primary unit of analysis in order to see how grammar and language use interact in discourse organization.

2.2.3 Cognitive Grammar on topic

Langacker acknowledges that the very definition of reference point recalls the notion of topic: "a salient entity evoked for purposes of mentally accessing another can also be thought of as providing a mental 'address' to which some notion is 'delivered' (i.e., as being what an expression 'is about'). Or to put it another way, a reference point's dominion—the range of conceptions that it itself tends to evoke—can equally well be thought of as providing a context with respect to which an expression is interpreted (or into which its content is integrated)" (1993:24).

Langacker technically defines topic as a kind of "subjective reference point". By

40 "reference point", he means that the topic is used "to establish mental contact with another entity". He then explains (1991:314): This reference point is subjective in two respects. For one thing, it remains offstage and often unmentioned. Rather than being explicitly discussed, an established topic is presupposed as part of the common background that the speaker and hearer rely on for making sense of the material presented overtly. A topic is also subjective in that the basis for its reference point function resides in speaker/hearer knowledge per se. In contrast to possessive expressions, where the possessor serves as reference point by virtue of some objective relationship it bears to the possessed (ownership, part/whole, kinship, etc.), it may be the organization of knowledge itself—even a relationship of class membership—that allows a topic to serve this function.

Langacker (1993, 1999:194-196) elaborates the relationship between reference point and target. While noting that reference point and target are both things in a possessive relationship, he holds that "the clearest cases of topics are those in which the target is not a thing, but is, rather, clausal or processual in nature, and where neither the reference point nor the relation it bears to the target is part of the objective scene being portrayed"

(Langacker 1993:25). He thus makes clear the "subjective function" of topic, that is,

"specifying which domain of speaker/hearer knowledge should be accessed in order to properly interpret the processual target" (Langacker 1993:25).

Langacker also identifies four topicality factors, and among the four, Langacker ranks the entity's semantic role as the most objective "in the sense of being intrinsic to the event described (not just a matter of how it is construed)" (1991:306), followed in turn by the empathy hierarchy, definiteness and figure/ground organization in order of declining objectivity or ascending subjectivity (Langacker 1991:306-307).

From the cognitive perspective, topic is a notion beyond the syntactic level.

41 According to Langacker, "Whereas subject is a clause-level construct, topics are clearly a

discourse phenomenon. We can distinguish two phases in a topic's career.22 At some point in the flow of discourse, an entity is first established as a topic, often by a special marking or grammatical construction... Once established, a topic holds sway over a certain stretch of discourse (perhaps until another one replaces it). There may be little need to mention the topic explicitly during this phase of career" (1991:313). The career of a topic actually reflects the dynamic aspect of the reference point model.

In addressing the levels of organization for reference point, Langacker (1998) illustrates topics at various levels, from Brown and Yule's (1983) discourse topics to clause-internal topics. He makes a distinction between a clause-internal discourse topic and a subject, by holding that "although they may be conflated in a single nominal expression, these represent distinct functions which have to be kept separate for analytical purposes" (1998:18). He argues that a discourse topic can serve as reference point for any number of comment clauses, and the nominal establishing it is usually external to each of them. By contrast, a subject specifically functions as such only for a single clause, as an inherent part of its internal structure (Langacker 1998:19): It does serve as a reference point with respect to the clausal process, but the reference point role of a subject or object per st—qua subject or object—is independent of any particular discourse status. Instead, a subject (or object) specifies the trajector (or landmark) of a profiled relationship, which functions as reference point with respect to the very act of conceptualizing that relationship. This is the closest connection possible between a nominal reference point and a processual target. Access to a profiled relation via its focal participants is intrinsic rather than contingent.

22 Chu (1998:261) proposes a three-phase hypothesis.

42 He then concludes that "starting from the extremes, these two functions converge at the level of the clause—the smallest unit expressing a full proposition, and the largest one with a single overall trajectory" (1998:20).

While Langacker's reference point model offers a coherent, integrated view of topic-like elements at the lexical, grammatical and discourse levels, and explains and clarifies the special relationship among topic, subject and possessor as affinitive but non-equivalent, Langacker himself also acknowledges that detailed analyses are required of individual languages of diverse types in order for the model to be widely accepted as a viable and well-supported characterization of topic. Therefore, this research examines how topic is used at different levels in spoken Mandarin and how different types of topic constructions can be understood as a conceptual reference point model.

2.2.4 Insights of the reference point model for topic study in Mandarin

Though Langacker's cognitive account of topic as conceptual reference point is not specifically oriented for topic-prominent languages like Mandarin, it does offer some guiding principles for analyzing naturally occurring data from topic-prominent languages.

First, the proposal that the reference point-target relationship can be manifested at different levels of organizations can help resolve the controversy over topic as a sentential or discourse unit: topic can be both sentence internal and external, and when sentence internal, it is a focal participant and inherent to the profiled relationship.

Second, the affinities and non-equivalence between topic and subject as revealed

43 under the reference point model may settle another controversy over the subject/topic categorization of the sentence-initial NP in Mandarin. Although topic is in general external to the comment clause, it can also be conflated with the subject, in which case the topic holds sway for just one clause and has a central grammatical role to play.

Third, it offers some guiding principles for topic identification, especially when the data are "dynamic" in the terminology of cognitive linguistics. Langacker distinguishes two distinct phases in a topic's "career", the establishing phase and the holding phase respectively. At the establishing phase, he mentions two often observed means, special marking and grammatical construction; once established, a topic can take a zero form in an ensuing stretch of discourse. Though topic marking is not obligatory in Mandarin, when topic markers do appear, they should be taken as valid means for topic identification. When spontaneous speech is concerned, topic markers abound, and so do zero-form topics. In addition to these two coding features of topic construction,

Langacker also hints at the possible semantic relationships between topic and the clause-internal entity (e.g. class membership). This also helps with topic identification, though how the reference point model operates in TC constructions in spoken Beijing

Mandarin has to date not been clear.

Fourth, by talking about the subjectivity of topic as reference point, Langacker takes into consideration the two parties of communication as well as people's world knowledge, which can in turn help resolve the dispute over the given/new information status of topic, in that the topic expression offers cognitve grounding for the comment. As advocated by

44 Chafe (1976, 1987, 1992, 1994) and Du Bois (1980, 1987) among others, properties of

information flow have to do with the way in which speakers package the flow of thought

into the flow of speech, whereby ideas are activated in speakers' consciousness and produced in accordance with the speaker's assumptions about what the hearer knows, as well as with other factors in the interactional environment. When it comes to topic-comment constructions, in general, the topic appears sentence-initially and the comment follows it, and this is cognitively grounded, because the comment is related to the topic either by talking about the topic or being inferentially linked to it. That a topic is assumed to be definite or generic (e.g. Li and Thompson 1976, 1981, Tsao 1979, etc.) can be explained in terms of the condition that the topic must be identifiable to the participants. So long as it is identifiable, it is not necessary for the topic to be syntactically related to a position inside the comment clause; instead, it can be only pragmatically related to the comment as a whole, where the comment itself can usually occur independently and consist of multiple predications.

Finally, Langacker gives primacy to the semantic role of topic over other topicality factors. This stance may help explain why topic may conflate with subject (e.g. Li and

Thompson 19981) as well as how other elements with a non-subject role can also become topical—the joint work of the other three topicality factors, namely, empathy hierarchy, definiteness and figure/ground organization, may motivate a non-subject element to the topic position.

In fact Langacker's interpretation of topic as conceptual reference point can be more

45 enlightening than it appears. First, the reference point-target relationship can be both subjective and objective. In human cognition, time and space are two basic domains and probably among the most frequently used reference points, which, I think, explains why temporal-spatial expressions tend to occur early in discourse. In some cases, their discourse-initial position can be a matter of conventionalization or repeated entrenchment.

For instance, time adverbials such as once upon a time, long long ago, etc. are observed without fail to occur at the very beginning of a narrative, indicating that time is ready reference point for narratives. Double-subject constructions (i.e., where topic and subject co-occur and both expressions have claim for subjecthood) such as xiang bizi chang

'elephants have long noses' (literally 'elephants noses long') stand in the middle of the continuum of subjective and objective construal, because the two nominal elements usually co-occur and occur in one order instead of the reverse (e.g. Chen 2004). These constructions are therefore conventionalized expressions, and most probably not entirely the outcome of the speaker's subjective construal. Also, Langacker (1998) has argued that topics are affinitive to possessors in possessive constructions. Since the possessor is the objective reference point for the possessed in Langacker's terms, it is possible for some topics to be objective reference points for their comments as well.

Moreover, this reference point view on topic resonates with "a major property" of topic expressions; that is, "they are referential" (Nikolaeva 2001:5), or to activate the referent. As Maslova and Bernini (2006) put it, "(T)he only palpable common property of all topic constructions is that they allow the speakers to activate a referent without

46 including it into the scope of assertion (or another illocutionary operation), i.e., to separate reference to an entity from the act of assertion."

While the cognitive interpretation of topic provides both practical guiding principles and a schema for more language-specific studies, how the schema operates in a particular language is yet to be learned, especially given the fact that topic constructions are both complicated and prevalent in Mandarin. To Langacker, reference point organization represents "so basic a cognitive ability that there may indeed be no linguistic phenomenon that does not involve it in some way" (1993:35). This research, therefore, represents some practical work aimed at revealing how topic functions as conceptual reference point in real discourse.

2.3 Findings in Other Topic-Prominent Languages

In this section, a brief introduction is made to topic studies in three other topic-prominent languages, with a view to understanding Mandarin topics in a cross-linguistic context.

2.3.1 Tagalog

In contrast with Shi's (2000:386) structural definition of topic in Mandarin, Hirano

(2001) uses the term topic in his paper for the phrase that is separated from the state of affairs to be known or judged (Shibatani 1991:97), and accordingly defines topic in terms of the following properties (2001:4):

47 a. A topic is followed by a topic marker or by a pause. b. Atopic is definite or generic. (Kuno 1973:39, Li and Thompson 1981:85) c. A topic usually occupies the sentence-initial position. d. A topic is not only an NP but also an Adverbial. e. A topic and its comment can be related in terms of either subcategorization or the pragmatic notion relevance. It is not always the case that a topic is syntactically related to a position inside the comment clause. Based on this, Hirano comes to the following findings regarding topic in Tagalog

(2001:22): a. Tagalog makes a clear distinction between topic and subject. The topic stands sentence-initially, and is followed by the topic marker ay or by a pause, while the subject is preceded by the subject marker ang in the IC224 position of situational sentences.25 The topic can be an NP or and ADV while the subject must be an NP. b. Two types of topic exist, i.e., internal topics (IT) and external topics (ET). The IT is a subject marked by the subject marker ang and in the Id position. The ET is a genuine topic that occurs sentence-initially. f. Tagalog has two types of topic construction, i.e., syntax-based (SB) and pragmatics-based (PB) topic constructions. This distinction shows that Tagalog has a genuine topic that cannot be moved sentence-initially from a position inside the comment clause.26

This study is interesting not only because it concerns Tagalog, also a topic-prominent language like Chinese, but also in its topic property characterization and the dual type classification of topic constructions. Of Hirano's five defining properties, (a), (c) and (d) are coding oriented, and to my knowledge Hirano is the first to state explicitly that topic can be adverbial although many others have hinted at it by saying that topic may set the temporal-spatial framework for comment (e.g. Chafe 1976, Li and Thompson 1981,

a, b and fare Hirano's ordering. 24 IC2 is the short form for the second immediate constituent. 25 Constantino's (1970, 1971 and 2000, as seen in Hirano) terminology. The example given is: Tumakbo ang bata'. (': glottal stop) ran the child 'The child ran.' The other two types are indefinite sentence and definite sentence, respectively illustrated as Bata 'ang tumakbo and Ang bata 'ang tumakbo. 26 This is in contrast with Shi's (2000) proposal that dangling topic constructions are all syntax-based.

48 Janzen 1998, among others). The dual type classification seems to echo the dispute over the relationship between subject and topic in Mandarin Chinese as discussed in Section

2.1. There is, however, a fundamental difference between the two languages: whereas

Tagalog marks its subject explicitly with the marker ang, subject in Mandarin Chinese is virtually lacking in any formal marking, especially given the argument that pause functions as the demarcation line between subject and topic (e.g. Shi 1995, Xu and Liu

1998:283-285).27 This stance of Hirano's and his double way of topic marking, namely ay or a pause, may help bring together the different views on the relationship between subject and topic in Chinese. (30) is one of Hirano's examples illustrating topic marking in Tagalog (p. 17).

(30) Si Lito ay nag-ba-basa na ng komiks. (IRT:3)

TOP/(SUB) TM AV-UF-read already ACC comic book28

'As for Lito, he is already reading a comic book.'

Hirano explains that Si Lito is followed by the topic marker ay, thus it is a topic. However,

Hirano suggest that the expression also "correlates" with the verb nag-ba-basa, which makes it a subject as well.

2.3.2 Japanese

Japanese is known as a language with a specialized topic marker wa. In present-day

27 Details to be seen in Chapter 3. 28 In this gloss line, TOP stands for topic, SUB for subject, TM for topic marker, AV for agent voice, UF for unfinished, and ACC for accusative.

49 Japanese, three functions have been identified for wa, namely contrastiveness, topic

creation and topic maintainence (Hinds, Maynard and Iwasaki 1987:x). In a study of particle choices in Japanese, Hinds (1983) finds that NPs marked by the so-called topic-marker wa form an intermediate category between noun phrases marked by ga and ellipted noun phrases in exhibiting topic continuity. In other words, ellipsis is the unmarked form of topic continuation and the NP marked by the particle ga exhibits the least amount of topic continuity, which is consistent with its role as the indicator of new information in discourse. The inspiration of Hinds' study for the present research is that pause particles in Mandarin may also be put onto a topicality scale composed of different forms of verbalization (including zero verbalization), especially when taking into consideration the fact that wa is a specialized topic marker while pause particles in

Mandarin are not so specialized.

Iwasaki (1987) chooses oral Japanese discourse as the object of study, which makes his study particularly relevant to the present one. Iwasaki focuses on two related issues, the pragmatic status of wa-marked noun phrases and the functions of these noun phrases in discourse, and finds that "identifiability" characterizes war-marked noun phrases more accurately than the often cited anaphoric or generic category. He also proposes that the primary discourse function of wa is "scope-setting", complemented by three derived functions, namely providing "multiple predications", indicating "negative scope" and

"contrastive element" in the discourse context. Iwasaki's "scope-setting" is very similar to "frame-setting" as identified for Mandarin (e.g. Chafe 1976), and "multiple

50 predications" echoes what Tsao (1979, 1990) and Chu (1998) among others have discussed regarding topic chaining in Chinese. As for wa as a "contrastive element", the contrastive function is among the most recognized semantic-pragmatic functions of topic constructions cross-linguistically. Given the spoken variety of the data, this study on

Japanese provides a good reference for a pause particle analysis in spoken Mandarin.

2.3.3 American Sign Language

As a signed language, ASL is also taken as topic prominent. Janzen (1998, 1999) analyzes marked topics in ASL as grammaticalized from yes-no questions. This analysis is very close to Chao's (1968) observation of topic-comment constructions as questions and answers in Mandarin. Yuan (2002) echoes a similar viewpoint when he argues that the TC construction in Mandarin grammaticalizes from two clauses, one reduced in form to function as the topic and the other the comment. By examining spontaneous spoken discourse, it is possible for this research to illuminate more instances of question-answer functioning as topic-comment, which will in turn show how a question is reduced in form to become a topic.

Janzen's categorization of the information status as conveyed by marked topics in

ASL is enlightening as well. According to Janzen's analysis, topic may contain information from three different "domains", namely the "pragmatic domain" (which is assumed to be "accessible to discourse participants, but is newly introduced into the discourse), the "syntactic-anaphoric domain" ("occurring subsequent to a previous

51 mention in the text"), and the "text-organizational domain" (where highly grammaticalized topics function to "organize parts of the text rather than to organize information reflecting the experiences of the participants") (1998:vi). If this categorization turns out to be applicable to spoken Mandarin as well, the information status of topic will be less controversial: the topic referent needs to be anchored in discourse, but the anchoring can be done either anaphorically or pragmatically.

2.3.4 Section summary

To summarize, studies of topic constructions in other languages have shown some commonalities, irrespective of the basic word order or manner of articulation of the language. Therefore, it might not be groundless to anticipate a more schematic account of topic, first within an individual language and then cross-linguistically.

2.4 Summary

This chapter first reviews previous studies on topic, with the controversies over its definition, property, status and identification highlighted. Then a cognitive approach is proposed as an alternative. It has been argued that the reference point model as proposed by Langacker opens up the possibility for a unified interpretation of TC constructions in

Mandarin. A brief mention is also made of topic studies in some other topic-prominent languages, with a view to understanding topic from a cross-linguistic perspective. Chapter 3: Pause and Pause Particles as Topic Markers

and a Coding Property Based Approach to Topic Identification

It can be seen from the discussions in Chapter 2 that pause and pause particles are topic markers in Mandarin (e.g. Chao 1968, Li and Thompson 1981, Tsao 1979, etc.), but how they function in discourse remains unaddressed. Therefore, this chapter first presents a review of previous studies on these topic markers. Following this, a coding property based approach to topic identification is proposed for this particular research. Some terms useful for this study are introduced towards the end of the chapter.

3.1 Pauses as Topic Markers

3.1.1 Some general observations

According to Chao (1968:67), "(T)he full sentence has two ICs:29 a subject and a predicate, separated from each other by a pause, a potential pause, or one of the four pause particles: a (~ia), ne, me, and bd\ Given the fact that Chao equates subject-predicate with topic-comment, the above statement means that topic can be followed by a pause or one of the four pause particles. Likewise, Li and Thompson

(1981:86) also take pause as a formal property of topic by holding that "a topic can be separated from the rest of the sentence (called the comment) by a pause or by one of the pause particles~a (or its variant ya), me, ne or 6o~although the use of the pause or the pause particle is optional".

29 Immediate constituents.

53 Xu and Liu (1998:283) also maintain that pause marking is a feature of topic. They argue that expressions typical of subject in semantics and functions are not to be marked by pauses. To them, there are at least two contexts under which pauses are not allowed: one is when the interrogative pronoun functions as the agentive subject, as in (l)d, and the other is when the subject of the relative clause is the agent of the predicate verb, as in

(2)b.

(1) a. MH$T^°

ta dasui le chabei.

He break ASP teacup

'He broke the tea cup.'

b. ilk Off), fl"#7£*F.

ta (ya). dasui le chabei. c. Jtfr#7»?

shui dasui le chabie?

Who break ASP teacup

'Who broke the tea cup?'

d.*it (PJ?), trwr^?

*shui (ya), dasui le chabei?

(2) a. Mife^&frllirWJWo

zhe shi ta dang-guo zhiqing de difang.

this is he be-ASP educated youth NOM place

54 'This is the place where he used to work as an educated youth.' b.*i&jjfe (iff), ^mnmmijo

*zhe shi ta (va). dang-guo zhiqing de difang. (Xu and Liu (1998:283-285))

In these two examples, the pauses are annotated as commas, and the optional uses of pause particles are indicated by the parentheses. The ungrammaticality of (l)d is due to the fact that the agentive subject is the interrogative pronoun shui 'who' (cf. the agentive pronoun subject in (l)b). In (2), the predicative is coded as a relative clause, where ta 'he' is the agent of the predicate verb dang 'be' within the relative clause. The unacceptability of (2)b is because the linguistic context does not allow topic marking. On basis of this observation, Xu and Liu conclude: when an element that can function as a subject is marked by a pause and/or a pause particle, it can be taken as a topic. They also note that the statement is not to be reversed, given the fact that topics are only optionally marked with pauses and/or pause particles.

3.1.2 Shi's (1995) detailed discussion on pausing and topic marking

A more comprehensive discussion of the relationship between topichood/subjecthood and pausing is seen in Shi (1995). Shi (1995:102) follows a structural approach and takes the sentence-initial position as the main criterion for subject identification; when a grammatical subject is followed by a pause, it turns into a topic. Within his framework, post-subject pauses fall under two categories, the obligatory and the optional.

55 3.1.2.1 Obligatory subtypes

Of the obligatory category, there are six subtypes. (3)-(5) are what Shi uses to

illustrate the first three subtypes.

(3)-ft I*I§o

wo, Wang Dayong.

I, Wang Dayong.

'I am Wang Dayong.'

(4) JIM. i&m.

daovan. Xie Jin.

Director, Xie Jin

'The director is Xie Jin.'

(5) &&, ^Hift-trftfo

Beijing. Zhongguo de shoudu.

Beijing, China ASSOC capital

'Beijing is China's capital.'

It can be seen that all three sentences consist of two nominal expressions. Given that there is a comma in each sentence to separate the two nominals, the first nominal is a topic and the second a comment. There are, however, some differences between them: the first nominal expression in (3) is deictic; the relationship between the two nominals in (4) is in an equative relationship, with the copula word missing; (5) is also an elliptical

56 copular sentence, with the second nominal naming a feature of the first one.

Since the fourth subtype needs to be addressed in more detail, the fifth and sixth

subtypes precede. The fifth subtype involves Shi's subjects, or topics when pause-marked,

that are in themselves subject-predicate constructions. (6) &%ikm$tmmi, ^¥&^ifcitt.

Zhang laotou song wo zhexie li. bushi pingbaiwugude.

Zhang old man send me these gifts, not for no reason

'It is not for no reason that old man Zhang has sent me these gifts.'

The underlined part, or the subject in Shi's terminology, is in itself a ditransitive clause in

itself, where Zhang laotou 'old man Zhang' is the subject, song 'give' the verb, wo 'me' the indirect object, and zhexie li 'these gifts' is the direct object.

Unlike the fifth subtype, the sixth subtype is characterized by some complication in the comment.

(7) j&EM* ftx^an*

ta xiongdi, zuowen bu-ken xie.

His brother, composition not-willing write

'His brother is unwilling to write his composition.'

In (7), the comment, i.e., the part after the comma, takes the form of a topic-comment construction, a topicalization construction to be specific. The grammatical object zuowen

'composition' occurs before the verb phrase buken xie 'not willing to write'. The comment expression in (7) is thus more complex in structure than the VP comment

57 expression in (6).

Now we can address Shi's fourth subtype. Shi refers to this subtype sweepingly as double-subject clauses. Under closer examination, however, I think that they belong to different construction types. (8) ^gjfrg, xm&n*]w<>

Lao Li ta-ba, gongzi di-de kelian.

Lao Li his father salary low-CSC poor

'Lao Li's father, his salary is so small.' (9) ±&gfi^r, mim^tWRmro

Shanghai Zixingche Chang, women dou guan ta jiao Fenghuangchang.

Shanghai Bicycle Factory, we all PREP it call Phoenix Factory

'Shanghai Bicycle Factory, we all call it the Phoenix Factory.' HO) ((»-£)), MM*1&ft%%o

Shuangchengji, na shi ben shijie mingzhu.

A Tale of Two Cities, that is CL world famous writing

'A Tale of Two Cities, that is a world classic'

xiongdi lia, yige xihuan wenxue, vige xihuan shuxue.

Brother two, one like literature, one like maths

'The two brothers, one likes literature and the other maths.' naci dizhen. zhengfu yixiazi pai-qu-le hao-jibai ren.

That time earthquake, government at a time send-go-ASP quite a few hundred people

'That earthquake, the government sent quite a few hundred people at one single

time.'

(8) is typical of the double-subject construction30 where Lao Li taba 'Lao Li's father' is the topic and gongzi 'salary' the subject, with the salary being the father's instead of Lao

Li's. (9) and (10) are two instances of left dislocation (LD). The difference between them is that the co-referential pronoun in (9) is the object while the demonstrative na 'that' in

(10) is the subject. (11) is a TC construction. Within the example, the two occurrences of yige 'one' are the set members of the sentence-initial NP xiongdi lia 'the two brothers'.

(12) is typical of Chafe's Chinese style topics: the sentence-initial NP names an event and the ensuing clause states something "about" it, with no selectional restrictions required of the sentence-initial NP and the predicate verb.

When it comes to the contexts in which pauses are not allowed, Shi (1995) lists three.

The first one is the same as Xu and Liu's interrogative pronoun functioning as the agentive subject; the second is the so-called subjects of universal reference, as in (13); and the third is when the subject is indefinite in reference, as in (14).

(l3)i|tP^iAWa>hAc

Shui dou bu-renshi zhege-ren.

Who all not-know this-person

30 Irrespective of the name of the construction, following Li and Thompson (1981), the first element in this construction type is referred to as topic and the second as subject.

59 'No one knows this person.'

yi-zhi wuva zai muse li fei-guo zheli.

One-CL crow in twilight inside fly-past here

'A crow flew over here in the twilight.'

3.1.2.2 Optional subtypes

Optional pauses, according to Shi (1995), occur in sentences of the following types. asm (,) 5+#7o

wo (.) wushi-sui le.

I fifty-years of age ASP

'I'm already fifty years of age.'

Zhe-jiwei (,) doushi wo-de laoxiang.

This-a few all are I-POSS townsman

'These people are all my townsmen.' (l7)fegA (,) JTM^o

Ta-airen (,) bing-le hao-jige-yue.

His wife sick-ASP quite-a few-months

'His wife was sick for quite a few months.'

(18)% (,) Jtt?&.

60 Wo C) duzi teng.

I stomach ache

'I've got a stomachache.'

(19)a. ^ (,) jftjg>fc&.

Ni (.) shuohua tai-kuai.

You speak too-fast

'You speak too fast.'

b. Ni shuohua (.) taikuai.

(20)a. ^ (,) i&Mfficfeo

Ni (.) shuohua shuo-de tai-kuai.

You speak speak-CSC too-fast

"You speak too fast."

b. Ni shuohua (,) shuo-de tai-kuai.

(2l)#£fP/Mr (,) (^&iffeM&Tio

Chunsheng he Xiaoqing C) yiwang shui ye mei jian-guo shui.

Chunsheng and Xiaoqing before who too not see-ASP whom

'Chunsheng and Xiaoqing, neither has seen each other before.'

Sentence (15) is a nominal clause, given that both the subject and the predicate take the form of a nominal phrase. (16) is a copular sentence with an optional pause after the subject. (17) differs from (16) in that the predicate consists of the verb bing 'be sick',

61 rather than the copula 'shi'. (18) is another instance of the so-called double-subject

construction, with wo T being the topic and duzi 'stomach' the subject. (19)a is both

similar to and different from (18). The two sentences are similar, in that the part after the

optional pause in each case states one aspect of the initial NP; the difference is also

conspicuous: while the element immediately following the optional pause is nominal in

(18), it is verbal in (19)a. (19)b has a different pausing pattern from (19)a, and the

resulting expression becomes a statement on a habitual behavior of the addressee's

manner of speaking, i.e., being too fast in speed. (20) resembles (19) in pausing pattern as

well as in semantic meaning. The only difference lies in the repetition of the verb shuo

'speak', which is probably the reason why some scholars refer to this type of construction

as a copying TC (e.g. Xu and Liu 1998). (21), like (10), is also an instance of LD, with an

adverbial of time yiwang 'before' inserted between the initial NP and the agentive

interrogative pronoun shut 'who'.

3.1.2.3 A brief discussion

From the above description, it can be seen that there is some overlapping between

Shi's obligatory and optional pausing. First, there seems no fundamental difference between (3)-(5) and (15) in coding form. All four sentences fall under the same category of nominal constructions, though the semantic relationship between the two NPs varies from one example to another. Second, (8) and (18) are both double-subject constructions, differing only in the alienability of the subject from the topic. Third, (21), like (10), is

62 also an instance of LD, i.e., there is a pro-form within the clause which is co-referential

with the sentence-initial NP. Then arises the question: why does the same construction

allow for different pausing patterns? Although Shi has made it explicit that pause is a

psychological thing whose existence is not necessarily in opposition to the absence of one

(1995:103), we may wonder when pauses actually occur in a particular construction type

and what impact the pause has on the message in language use. Such issues make up the

theme of Chapter 5. In what immediately follows, I present the findings of previous

studies on the use of pause particles as topic markers.

3.2 Pause Particles as Topic Markers

3.2.1 Some general observations

According to Chao (1968:795), particles in Mandarin are like suffixes and

interjections in taking the neutral tone. They are always unstressed and always bound,

being enclitic to the preceding syllable and in construction with the whole preceding phrase. Chao divides particles into phrase particles and sentence particles, with a few occurring in either function, i.e., either to mark a phrase or a sentence.

Chao also holds that there can be a potential pause between subject and predicate, and four pause particles, namely a (~ia), ne, me and ba, may occur to fill the pause. In talking about subject and predicate as question and answer, Chao observes that the four pause particles "all occur as interrogative particles at the end of questions" (1968:81). It is also important to note that Chao equates subject-predicate with topic-comment. As a corollary,

63 the four interrogative particles are topic markers.

The Chinese grammarian Zhu Dexi (1982:95) accepts Chao's interpretation of the

grammatical meaning between subject and predicate in a Chinese sentence as topic and

comment. He points out that there can be an optional pause between subject and predicate or topic and comment, and that the pause can be verbally coded as a, ne, ba, me, etc.

Likewise, Li and Thompson (1981), Tsao (1979, 1990), among others, all take these four particles as potential topic markers.

The above studies, however, present no elaboration on the use of these particles as topic markers. The exact number of pause particles also remains an issue, especially given the fact that the orthographic conventions for some homophonic particles are often confusing, and that scholars have different interpretations for the phonetic variants of particle a in different contexts (e.g. Zhang and Fang 1996, Yuan 2002, etc.). Below are some studies focusing on pause particles occurring in the middle of the sentence, with

Fang (1994), Zhang and Fang (1996) and Yuan (2002) primarily on the Beijing dialect of

Mandarin, and Xu and Liu (1998) mainly on the Shanghai dialect.

3.2.2 Some specific findings

According to Fang (1994), pause particles in Beijing Mandarin originate from sentence-final particles (henceforth SFPs). This observation echoes to some extent Chao's

(1968) proposal that the four pause particles can all be used at the end of questions as interrogative particles. The difference is that Fang does not limit the origin of pause

64 particles to interrogative particles only. Xu and Liu (1998:103) even suggest that the pause particle shi in the Shanghai dialect has evolved from the homophonic copula verb.

Irrespectively of the origin, an essential difference between pause particles and SFPs is that the former mark phrases and the marked phrases must be followed by some other element(s) (Xu and Liu 1998:105). That is to say, a marked topic must be followed by a comment. Given the enclitic nature of pause particles, it is also self-evident that pause particles do not occur clause- or utterance-initially.

When it comes to the distribution of pause particles in discourse, Zhang and Fang

(1996:37-38) find that pause particles can follow either constituents or non-constituents of a sentence. Among sentence constituents, the preceding elements can be functional subjects (those having a 'doing' or 'being' relationship with the predicate or comment), topics, conjunctives, adverbs, prepositional phrases and some textual elements. When it comes to non-constituents, they use the following sentence for illustration (1996:38): (22) mnn&'X&kBM, &^m.n°

zanmen dei gei zaiqu renmin na, chu-ge-hao-zhuyi.

We must for disaster area people PART put forth-CL-good-idea

'We must put forth a good idea for the people in the disaster area.'

In this sentence, the part preceding the pause particle na consists of the functional subject zanmen 'we', the modal verb dei 'must', and the prepositional phrase gei zaiqu renmin

'for the people in the disaster area'. This sequence as a whole is by no means just a single constituent of the sentence. It is out of the observation that the occurrence of the particle

65 may not correspond precisely to any syntactic or semantic configuration that Zhang and

Fang propose that some pause particles, such as a and ba, are indicators of the speaker's division of primary information (or focus, which follows the particle) and secondary information (or theme, which precedes the particle) (1996:39).

When the Shanghai dialect is concerned, Xu and Liu (1998:114) conclude that the pause particle may occur (1) after any elements preceding the predicate verb of a clause

(main clauses and sub-clauses included, but not relative clauses), but not after adverb or adjective adverbials, and (2) between two elements which ensue the predicate verb but are in a predicating relationship. We have seen many instances of the first type in

Mandarin. (23) is an example from the Shanghai dialect illustrating the second type (Xu and Liu 1998:111). (23)aif«i, ftM^Ift.

Wo qing Xiao Zhang mo, fuze yewu-gongzuo.

I ask Xiao Zhang PART be in charge of professional-work

'I ask Xiao Zhang to be in charge of professional work.'

In this sentence, the predicate verb is qing 'ask', and the pause particle mo marks the proper name Xiao Zhang, which is then predicated by the ensuing VP. In Mandarin, a pause particle is also acceptable in this slot, though most probably it is not mo.

Moreover, Zhang and Fang (1996), Xu and Liu (1998), and Yuan (2002) all acknowledge that pause particles can be used more than once to mark different elements within the scope of a single clause or sentence. This phenomenon calls into question

66 Zhang and Fang's proposal that pause particles are speakers' division of primary and secondary information, because the primary information does not occur until all pause particle-marked elements have been produced. A better explanation is therefore yet to be found.

Xu and Liu (1998) also present a good summary of the functions of topic marking across different languages. In languages like Japanese and Korean, topic markers tend to occur on two occasions: one is when the discourse shifts to a new topic, and the other when a contrastive topic is introduced (Gundel 1988, as seen in Xu and Liu 1998:224). A study on Bunun, a native language in Taiwan, shows that topics in the language serve two functions, namely to set the conceptual background and to contrast two conceptual frameworks (Zheng 1991, as seen in Xu and Liu 1998:224-5). On the basis of such cross-linguistic studies and their own observation of the Shanghai dialect, Xu and Liu propose that pause particles have a double function in discourse. On the one hand, the occurrence of a pause particle helps arouse the listener's attention to the ensuing stretch of utterance; on the other, pause particles facilitate the introduction of topics, especially those conveying semi-active information. They then reduce this double function to one: pause particles are used to enhance the topicality of the marked element (1998:220). The use of topic markers can promote indefinite information to topichood by means of marking. In line with Shi's (1995:97) claim that the pause after a functional subject can highlight the topic status of the subject, Xu and Liu maintain that pause particles are more powerful and more specialized means of topic marking than pauses.

67 3.2.3 Summary

This section has first acknowledged the topic marking status of pause and a set of pause particles. Then some specific findings on the origin, context of occurrence and discourse functions of these topic markers are reviewed. The following section focuses on the discourse use of each individual pause particle in Beijing Mandarin, i.e., the variety of

Mandarin as produced by Beijing locals.

3.3 Discourse Use of Pause Particles in Beijing Mandarin

Pause particles are not only more specialized means of topic marking; they are also assumed to have labor divisions between themselves. In what follows, I introduce some specific findings on the use of different pause particles in Beijing Mandarin, with a view to better orienting pause particle analysis in this research.

3.3.1 Zhang and Fang (1996) on pause particles

Zhang and Fang (1996:ch.4) put particles which occur in the middle of a sentence under three categories, topic markers, quasi topic markers, and non-topic markers. In the first category, there are two particles, a and ba, which have lost their mood-indicating functions when occurring in the middle of the sentence. They acknowledge that ba is one of the six phonetic variants of a,31 and functions in a similar manner to a. The reason for a and ba being taken as two separate topic markers is that ba has become a social variant

31 Other variants are na, nga, ra, ya, and wa.

68 of a, given the fact that Beijing locals aged forty and younger prefer to use ba to a, though a is the universally accepted form. To support this argument, Zhang and Fang illustrate that a and ba may occur after numerous types of thematic32 or topic constituents. When marked topics of different metafunctions co-occur, the following order is observed: Textual topic > interpersonal topic > experiential topic.33

The second category includes me, ma and ne. In some contexts, these behave very much like a and ba, conveying no mood-indicating information; in other contexts, however, they still maintain some of their functions as SFPs. Zhang and Fang also note that me is rarely used in spoken Beijing Mandarin. Instead, ma seems to have taken its place. Moreover, ma and ne seem to be in complementary distribution, in that ma tends to occur early in a stretch of utterance while ne usually appears towards the end of the stretch, especially when a topic shift is in order or a new perspective is to be taken.

There are two members in the third category, namely la and ya, which Zhang and

Fang take as pure mood particles. These two particles, according to Zhang and Fang, are very restricted in distribution, i.e., they usually occur to mark coordinate items.

In short, not all pause particles are topic markers according to Zhang and Fang: a and ba are full topic markers in Beijing Mandarin, me, ma and ne are quasi topic markers, and la and ya non-topic markers.

32 In Halliday's (1985, 1994) terminology. 33 Systemic grammar takes the clause as a "composite affair" or a combination of three different structures, namely the transitivity structure, the mood structure and the theme structure, which are derived from three distinct functional components, i.e., metafunctions (Halliday 1994:179). Also see p.71.

69 3.3.2 Yuan (2002) on pause particles

Yuan (2002) is another recent study focusing on the spoken variety of Beijing

Mandarin. According to his study, the frequently used topic markers in the language include ne, ba, a and dehua 'if. Of the four, ne is the most frequently used topic marker and may be paired with lexical topic markers such as shuodao 'speaking of, etc. The main discourse function of ne is to shift the topic and establish a new starting point for the ensuing discourse. Accordingly, it is more frequently observed late than early in a discourse stretch. On account of its topic-introducing function, ne is inclined to be used in utterances with more than one topic, or utterances of a listing or contrasting nature. For the same reason, ne is optional in the first clause of a coordinate sentence of contrast, but prone to occur in the second clause. The reverse situation, according to Yuan, is rarely seen.

The particle ba for Yuan is a frequently used topic marker in Beijing Mandarin, and its basic discourse function is to introduce topics of illustration, comparison and supposition, ita-introduced topics can be framed with lexical means such as na...laishuo

'take ... for example', biru 'for instance', jiaru 'if, etc. The discourse function of ba is to introduce a new topic to serve as the starting point for the ensuing discourse. Therefore, ba can be used either in the opening or the closing clause of a stretch of utterances. On account of this discourse function, ba and ne seem to be in complementary distribution; that is, ba tends to be used in the early occurring clause(s) while ne in the late occurring

70 one(s).

Another frequently used pause particle in Beijing Mandarin is a (wa and na as its phonetic variants), which can be used in opening as well as closing clauses of a discourse stretch to introduce various types of topic.

According to Yuan (2002), the topic marker dehua evolved from its suppositive homophone, and expresses a similar meaning to laijiang 'speaking of. He maintains that dehua is not as well grammaticalized as ne, a or ba, in that it is often coupled with lexical expressions such as shuo 'speak' to the front and ne to the rear in use. In the meantime,

Yuan also acknowledges that in modern spoken Beijing Mandarin, independent use of dehua has become a common phenomenon and its overall frequency is on the increase.

For Yuan, dehua as a topic marker is similar to ne in discourse function and distribution pattern, which to some extent explains the co-occurring tendency of the two.

Yuan also briefly comments on the prepositional topic markers guanyu 'as for' and duiyu 'to/for', asserting that such expressions are seldom used in speech, unless in pseudo-intellectual expressions. He also observes that the question tag bushima may also be used to introduce a topic in Beijing Mandarin. All in all, Yuan presents a fairly wide range of topic markers in the language of concern.

3.3.3 Summary

From the above discussion, it can be seen that the two studies have both similar and divergent findings on the number and use of pause particles as topic markers in the same

71 language, i.e., Beijing Mandarin. It is true that they have different focuses of attention, namely, Zhang and Fang focus on the specialization of different pause particles while

Yuan is concerned more with the discourse use and distribution of high frequency topic markers. Nonetheless, the discrepancies are conspicuous enough to catch our attention.

Also, their studies are more qualitative than quantitative; that is, no specific frequency counts are given to support their arguments. If semantic change in grammaticalization is a function of language use and thus a usage-based phenomenon, as cognitive linguists argue (Evans 2007:97), it would be best for these pragmatically motivated linguistic elements to be studied in naturally occurring discourse in a bottom-up manner. Chapter 6 of this dissertation represents such an endeavour.

3.4 A Coding Property Based Approach to Topic Identification

3.4.1 Coding properties as the primary concern

According to what has been discussed in Section 2.1, two consensuses seem to have been reached on topic in Mandarin, namely, the topic precedes the comment and the topic can be separated from the comment with a pause or pause particle. These two consensuses, under closer scrutiny, collapse into one—the coding property of the topic expression. When studying subject cross-linguistically, Keenan (1976:324) maintains that subject can be defined in terms of coding properties, behavior/control properties, and semantic properties. Of the three types of properties, the coding properties play a central role on account of the fact that the behavior properties may differ from language to

72 language and semantic properties from construction to construction. Hence the Promotion to Subject Hierarchy (Keenan 1976:324):

Coding Properties > Behavior and Control Properties > Semantic Properties

Of the coding properties, Keenan lists position, case marking and verb agreement; of the behavior properties are deletion, movement, case changing properties, control of cross-reference properties, etc; and the sematic properties include agency, autonomous existence, selectional restrictions, etc. Since topic also embodies features in these three aspects, I assume that the coding properties of topic may take primacy over its behavior properties and semantic properties as well.

Of the coding properties (CPs) listed by Keenan, position is of relevance to Mandarin, in that Mandarin topics precede comments. In this study, another coding property stands out, that is, the frequent occurrence of pause and pause particles in the spoken corpus.

Given what has been discussed in Sections 3.2 and 3.3, pauses and pause particles can be used as reliable clues for topic identification. Another item that can be added to the CP list for Mandarin is that some grammatical constructions have established topic-comment status irrespective of the use of pause and/or pause particles. These construction types include the double-subject construction, topicalization, left dislocation, etc. In fact, to take typical TC constructions as means of topic identification is in agreement with

Langacker's (1991:313) observation that "grammatical construction" is an often used means of topic establishment. In total, this study identifies three CPs for topic identification, namely initial position, pause and/or pause particles, and grammatical

73 constructions.

There are at least two advantages in granting CPs the primary status in identifying topics in Mandarin. First, the CPs are more tangible, which renders them more reliable than other less obvious properties. Perhaps this is why the two consensuses reached by various studies on topic characterization are both regarding coding properties. Second, these CPs, pause and pause particles in particular, may reveal how Mandarin speakers organize information. Tao (1996:52) has shown that "particles correlate to an overwhelmingly large degree with the end of intonation units", and according to Chafe

(1979, 1987, 1994), intonation units or speech units cut across several dimensions of language production, including prosody, syntactic structure, cognition, and interaction, and have a cognitive basis, whose formation reflects speakers' focus of consciousness.

Therefore, through observing pause and pause particle use, we can see how cognition is at work in discourse organization.

3.4.2 Topic identification

By granting primacy to the coding properties, topic in this study is first identified by its early occurrence in the clause. As for where to draw the demarcation between topic and comment, different researchers seem to have different decisions. I find that

Halliday's (1985, 1994, 2004) proposal can be taken as a good starting point. Halliday uses the dichotomy of theme34 and rheme, roughly equivalent to the opposition between

34 Halliday defines Theme as "the element which serves as the point of departure of the message; it is that with which the clause is concerned" (Halliday, 1985:38). Here his "concerned" may very well be synonymous with "about". When

74 topic and comment, to account for the information structure of the clause. According to the configuration, a theme can be simple or multiple. A simple theme is one forming a

single constituent in the structure of the clause, and a multiple theme consists of two or more constituents fulfilling different metafunctions, namely the experiential (constructing a model of experience), the interpersonal (enacting social relationships) and the textual

(creating relevance to context) (Halliday 1994:36). Halliday specifies the following principle for theme identification, "If the initial element in the clause does not function as

Subject or Complement or CIRCUMSTANTIAL Adjunct, then the Subject, Complement or Adjunct next following is still part of the Theme" (1985:53). As a corollary, all his themes, both simple and multiple, end with subject, complement or circumstantial adjunct, or in his terminology, the experiential element, which he refers to as "the topical Theme".

To avoid confusion and align with the three metafunctions that Halliday proposes for all languages, I use experiential topic in place of Halliday's topical Theme. Below is a clause

Halliday gives for illustration (1985:55, my highlight): on the other hand maybe on a weekday it would be less crowded conjunctive modal topical textual interpersonal experiential Theme Rheme

In this clause, there are three Themes, each fulfilling a different metafunction, as indicated in the third line.

it comes to its identification, he says that "As a general guide, the Theme can be identified as that element which comes in first position in the clause" (Halliday, 1985:39), again resonating with the usual position of topic occurrence.

75 Xu (1996) makes two modifications to Halliday's principle for theme-rheme segmentation, and the one of concern here is that Xu extends Halliday's topical Theme to process participants. The resulting analysis of the same clause above would be as follows

(again my highlight).

on the other maybe on a weekday it would be less crowded hand conjunctive modal circumstantial topical textual interpersonal experiential Theme Rheme

Thus Xu has extended topic to the argument NP by juxtaposing it with the earlier occurring circumstantial elements.

In this research, I take the middle course by counting as topic only the marked process participants while leaving out those unmarked. This is in line with the primary status of CPs in topic identification as argued for in Section 3.4.1. Furthermore, Xu and

Liu (1998:111) also maintain that all content expressions preceding the main verb can be topics, on the basis of the fact that these expressions can all be marked by pause particles.

The dividing line then is drawn in accordance with the availability of a pause and/or a pause particle.

To extend the topic expression to include marked process participants is an interesting stance. Topic, as discussed in Chapter 2, is quite a heterogeneous notion, covering argument entities as well as temporal-spatial frames. Although argument entities

76 behave very differently from temporal-spatial adverbials, they do share at least one thing in common; that is, they can all take marking in Mandarin. Given the fact that scholars have the most divided opinion on whether or not to count adverbial temporal-spatial expressions as topics, I, following Xu (1996) and Xu and Liu (1998), choose to be more inclusive in identification criterion formulation and leave this issue open until after examination of the corpus data.

It is evident by now that topic in this research is identified by position to include all elements performing the textual and interpersonal function, and by marking to include marked circumstantial elements and process participants. A special mention needs to be made of preverbal object roles. Since they are marked by the grammatical construction, they are taken as constant topics disregarding the availability or not of a pause and/or pause particle after them.

This delineation of topic has at least two implications. The first one is that participant topics thus identified are all marked, either with a pause/pause particle or with a particular grammatical construction. Section 3.1 has reviewed Xu and Liu's (1998) observation of the conflicts between pausing and subjecthood. Detailed analysis of the corpus (see Chapter 5 and 6) also shows that this delineation of topic is meaningful in revealing the information status of the expression defined as topic.

The second implication is that there can be more than one topic in one clause or sentence. Actually this is exactly what is frequently observed in the corpus: the topic marking particles can be used several times in a row in one single clause. Below are two

77 examples from the corpus. The context of use of each example is given within the parentheses, where the part preceding the colon indicates the code number of the speaker and the figure following the colon the number of the paragraph in which the example utterance occurs. This notation applies to all examples from the Beijing corpus. (24)X%mM%±n%, mmiim, m$t^&)im&o (057-H20:5)

Youqi wo na laoda ya, zhe-xifur ne.

Especially I that first child PART, this-wife PART,

ta jiu youdiar xiangfa.

she then a little dissidence

'Especially my first child, his wife has some dissidence.'

(25)£^M±M> aiitefeiMg, A§B?£±3gil, ^fe£MmWL. (043-N50:2)

Zai xuexi-shang a. muqian laishuo ne,

in study-on PART, presently speaking PART,

cong ziji zinii shang kan ne,

from self children on see PART,

you henduo de xiangfar.

have many NOM comments.

'On study, speaking of the present time, when my own children are concerned, I have

lots of comments to make.'

In (24), there are two pause particles, ya and ne respectively, bringing one entity after another to the foreground. First, the eldest child of the speaker is introduced into

78 discourse by means of the possessive construction, and his wife through the proximal demonstrative zhe 'this'. If the topic entities in (24) are introduced in a chained model, the marked topics in (25) approach the comment proposition from different perspectives.

The first intonation unit specifies the aspect of concern, the second the time dimension, and the third the people involved. Altogether, three pause particles are used in a row, marking three consecutive topics.

It might be of interest to examine the two examples against the so-called "aboutness" property of topic. In (24), the comment is more 'about' the wife rather than the first child of the speaker, given the fact that the experiencer of the negative feeling of dissidence is the wife's and that the coding form ta 'she' is also in agreement in gender with the wife.

But this leads to several questions: is the first NP a topic in the sense of 'aboutness'? If yes, how? If not, why does it occur in sentence-initial position, a position typical of topic occurrence? As the linguistic coding stands, how does it contribute to the information flow in the context? And in what way is it effective? Impressionistically, there is a chain of referents in this example, from the speaker to her first child, then from the first child of the speaker to the child's wife, following the order of occurrence. To put it in Langacker's terminology, the first entity is a reference point for the second, and the second for the third. A reversed order may not work. (25) is somehow different form (24), in that the order of occurrence is not a prerequisite. The order of the three particle-marked elements is variable, without hindering the grammaticality or acceptability of the utterance. The three topics as identified here show how the speaker narrows down his focus of attention,

79 while keeping all three dimensions (namely the activities of study, the present time, and

the persons concerned) of the subject matter on stage simultaneously. To put it differently,

the comment is equally related to all three topics. Cognitively speaking, it seems that a

multiple reference point system is at work in utterances such as this.

Since these two examples have shown very different manners of topic reification and

reference system establishment, a multiple-topic analysis for this research is needed to

reveal more of the nature of topic in spoken discourse.

3.4.3 Summary

This section first establishes the primacy of coding properties in topic identification

in spoken Mandarin. Then a CP-based approach to topic identification is proposed for

this corpus study, which is in turn followed by a brief discussion of the implications of

such an approach.

3.5 Defining Some Grammatical Terms

Once the criteria for topic identification have been established, some structural terms

that are essential to this research need to be clarified.

3.5.1 Clause

According to Langacker (1998:19), a clause is the smallest unit of discourse over which a topic holds dominion as the active reference point, and at the same time the

80 largest unit for which a single overall trajectory is posited. Therefore, it is essential to

have a clear definition of clause. Unfortunately, clause is perhaps never a well defined

term in Chinese grammar (maybe in other languages as well), and it can be even trickier

to define a clause in spoken discourse. For a useful heuristic method to deal with the

corpus, I define clause as a verb plus its core argument(s), with modifiers optionally

present (Dixon 1979, Du Bois 1987, among others). As a matter of fact, core arguments

can also be optionally present in the clause. For instance, subjectless clauses are both

grammatical and frequent in Mandarin Chinese. In this manner, clauses in all elliptical

forms are also put under the same umbrella category, the most typical coding forms

including verb and adjective phrases. With some simplification, clauses with subjects are referred to as SP (i.e., subject-predicate) and clauses without subjects as elliptical SP in this dissertation.

There are two points to be borne in mind with regard to the definition of clause as verb plus its core argument(s). One is that relative clauses are not taken into account, given the fact that they are embedded in complex NPs and function at a lower level. The other is that some NPs, when used obviously in a predicating manner, are also counted as clauses. Very often, they are the outcome of ellipsis, as represented by (25).

(25)ft\h£U)l-M, -Tj&tfeo (295-L12:3)

Zai chuqu wanr yitang, viwan-kuai-qian.

More go outside play once ten thousand-RMB yuan-money

'If a trip is also planned, you'll need ten thousand RMB yuan.'

81 The underlined expression is an NP, predicating on the expenditure of a wedding where a

(honeymoon) trip is a potential component.

Closely related to simple clauses are utterances consisting of more than one clause.

In this dissertation, clause combinations explicitly marked with conjunctives are called either compound or complex sentences, depending on the conjunctive(s) used in the sentence.

3.5.2 Topic-comment constructions (TCs)

TC is used in this dissertation to refer to a topic-comment construction. Following the CP-based approach to topic identification as formulated in Section 3.4, a TC may begin with quite varied linguistic forms, such as adverbials (of time, place, manner, etc.), conjunctives of various kinds (e.g. transition, reason, result, concession, condition, and the like), modal components (or Halliday's interpersonal elements), etc.

Below the category of TC, there are several sub-sets deserving some special mention.

First, topicalization (TOP) and left dislocation (LD) are two structurally defined construction types. According to Prince (1984), both TOP and LD contain a preclausal

NP and a clause following, but they differ formally: TOP contains a gap in the clause which corresponds to an argument position that the preclausal NP can be construed as filling, whereas LD contains an argument-position pronoun which is co-referential with the preclausal NP. Prosodically, however, the two constructions have analogous patterns and each has two prosodic peaks: one occurs within the preclausal NP and the other

82 within the predicate expression. On the basis of the grammatical role of the preclausal NP, both TOP and LD can be sub-divided into TOP-subject vs. TOP-object and LD-subject vs. LD-object. In addition, there is another subset of LD, namely Prince's LD2, which

"triggers an inference on the part of the hearer that the entity represented by the initial NP stands in a salient partially-ordered set (poset) relation to some entity or entities already evoked in the discourse model"(Prince 1997:126).

The double-subject construction is identified by the co-occurrence of a topic and a subject irrespective of the name to the construction. According to Li and Thompson

(1981:93), there is a 'part-whole' relationship between the topic and the subject in the double-subject construction. Otherwise, this type of construction is "just like all the other topic-comment sentences being considered: the topic is the definite noun phrase that is what the sentence is about, and the subject is the noun phrase in a 'doing' or 'being' relationship with the verb". Chapter 5, however, shows some complications of the construction in real-time spoken discourse.

Existential clauses constitute another subset, given the fact that they usually begin with an adverbial of place.

Some other minor subsets are characterized as individual cases when they are under discussion.

3.6 Summary

In this chapter, I first present a review of previous studies on pause and pause

83 particles with regard to topic marking in Mandarin. After arguing for the primary status of

CPs with respect to SPs and BPs, three coding properties are proposed for topic identification in the spoken corpus. Some grammatical terms are clarified towards the end of the chapter as a preliminary for corpus analysis.

84 Chapter 4: Methodology of a Corpus Study

In this chapter, I first introduce the corpus used for this study, which is then followed

by a brief discussion of the corpus analysis procedure. The research goals are further

outlined towards the end of the chapter.

4.1 Data: The Beijing Corpus

The data used for this research are part of a 1.8 million character corpus, the Corpus

of Modern Spoken Beijing Mandarin (henceforth the Beijing Corpus), which is sponsored

by the Language Research Institute at the Beijing Language and Culture University,

Beijing. All language informants are Beijing locals, sampled across all urban and near

suburban areas of the city. They were asked to talk on six different topics, and their

monologues were recorded without intervention from the language investigator. The

corpus takes the form of 374 monologues recorded in the 1980s and transcribed in

Chinese characters, maintaining slips of the tongue, repetitions, and redundant and

unfinished stretches of utterances without any editing. Given the size of the corpus and

the broad sampling of informants, the findings of this research are expected to apply to

speakers of the modern Beijing dialect of Mandarin in general.

Altogether 45 monologues, with a total of slightly over 200,000 characters, were randomly sampled from the Beijing Corpus for the purpose of this research. Following the corpus's designation, these monologues are on six different topics, three on topic A

(living conditions, health, sightseeing, love and marriage, and social welfare), twelve on

85 topic B (family, education, extraordinary experiences, procreation and social security),

nine on topic C (education, employment, fighting poverty, funeral and travel), nine on

topic D (work, income, climate, currency and information), three on topic E (daily

schedules, market, recreation and transportation), and nine on topic F (personal

experiences, price, physical activities and visiting of people). The age of the 45

informants ranges from 25 to 84, and there are 24 males and 21 females. The informants are of three different nationalities, namely Han, Hui and Man, and live in different parts of Beijing city. They have different education background, 6 with primary education, 31 with secondary education, and 8 with college or university education. They work in different professions, such as teacher, official, engineer, worker, driver, etc. For further details, see Appendix 1.

The monologues from the corpus are available as raw data, that is, as transcripts from recordings without tagging or annotation. The transcripts are punctuated to indicate pauses and other prosodic features, such as the different intonations, sentence boundaries, etc. In this research, I segment each and every monologue into clauses by referring to the transcribers' punctuation pattern but without changing the original narrative order. For examples appearing throughout this dissertation, the transliteration and translation are done by myself.

However, handling transcribed texts is not unproblematic. Brown and Yule (1983) argue that "however objective the notion of 'text' may appear as we have defined ('the verbal record of a communicative act'), the perception and interpretation of each text is

86 essentially subjective" (1983:11). They continue to suggest that "once the analyst has

'created' a written transcription from a recorded spoken version, the written text is

available to him in just the way a literary text is available to the literary critic" (1983:12).

In this sense, it is probably not a disadvantage for me, the analyst, to have access to the

transcripts only, because the subjectivity is at least somehow undermined by the transcribers' neutral manner; that is, they did not have any particular research objective when doing the transcribing work. This point might be especially important for this study because pauses have been taken as topic markers by many studies (e.g. Li and Thompson

1976, 1981; Tsao 1979, 1990; etc.), and this study relies heavily on pauses and pause particles for topic identification.

4.2 Corpus Analysis Procedure

The corpus secured for this research is a raw corpus, which means that tagging or labeling must be done manually. First, the monologues were divided up into clauses following the characterization of clause as proposed in Section 3.5. Then the clauses were analyzed into various construction types and tagged accordingly. Since the principles formulated for topic identification in this research allows for more than one topic in one clause, a clause is sometimes segmented more than once for all topics to be tagged. When the construction resulting from initial segmentation remains more complex than an SP or elliptical SP (i.e., a VP, AP or occasionally an NP), the segmentation procedure is repeated until the comment consists of an SP in its full or elliptical form only. (1) below

87 is an utterance that needs segmentation only once, whereas (2) shows how segmentation

must be repeated for all topics to be tagged.

(1) ^Bjg #^7M€7Vo (274-T30:2)

Yi-yi-tian ne zheng-bu-liao liang-mao-ba.

One-REP-day PART earn-not-AUX two-CL-eight

'For one day, we could earn no more than twenty-eight cents.' —^m #^7W^A Yi-vitian ne zheng-bu-liao liang-ma-ba Topic: time Comment: VP As the underlining in (1) shows, the clause is divided into two parts: the first part is the particle ne marked topic, indicating the time frame, and the second a VP fulfilling the grammatical role of predicate or the pragmatic function of comment. The clause as a whole is thus tagged as an instance of TC.

(2) ffiVM, MU1EE, &MM%-&)1. (057-H20:5)

Suovi ne, tamen-lia ne, wo jiu pian-zhe yidianr,

so PART they-two PART I AUX be partial to-ASP a little

'So, the two of them, I'm a little partial to." mmm suovi ne tamen-lia ne wo jiu pian-zhe yidianr Step 1 Topic: conjunctive Comment: TOP-object Step 2 Topic: object Comment: SP This clause occurs in a context talking about familiy members, where the first child and his wife are contrasted with other children of the speaker. Although this example, like (1), is also a TC, segmentation must be conducted twice for the final result. First, the

88 conjunctive word suoyi 'so' is segmented and tagged as a textual topic following

Halliday's terminology, leaving the remaining part a TOP-object construction. When the

grammatical object tamen lia 'they two' is set off, the resulting construction is an SP. The

practice as illustrated in (1) and (2) not only can show the frequency of occurrence of

each construction type; more importantly, it also reveals how extensively a particular

construction type distributes across different linguistic contexts, especially given the

arguments already put forward by Tsao (1979, 1990), Li and Thompson (1981), Shi

(2001), etc. on how topic differs from subject in this respect.

In order to address the issue of whether pause is a reliable demarcation between topic and subject or not, a grammatical subject that is separated from the predicate by a pause and/or pause particle is tagged as topic following the CP-based approach to topic identification and one immediately followed by the predicate without a pause or pause particle as subject. The frequency of use, context of occurrence, together with the coding form for both cases are then compared to verify or reject some previous claims on the demarcation (Shi 1995, Xu and Liu 1998, among others). Since the corpus takes the form of transcripts, the actual pauses cannot be measured. When a pause particle is articulated by the language informant, it can be taken as a topic marker without any question; when just a comma is observed in the transcript, the transcriber's punctuation is accepted as denoting where pauses exist even though native speakers of Mandarin sometimes differ in their punctuation pattern.

When it comes to topic-marking pause particles, tagging is done by the particle.

89 Since pause particles and SFPs may take the same coding form, the tagging needs to be

limited to only those that mark the topics by leaving out the clause-final ones. When

tagging is done, the coding form of particle-marked topics is examined against the

context of use. A frequency count is necessary to show the distribution tendency of the

particle in discourse use and the degree to which the particle has specialized as a topic

marker. The same procedure is repeated for each pause particle.

In this research, the following pause particles are tagged and analyzed: ne, ba, a (and

its phonetic variants la, na, wa and yd), and ma. Also analyzed are two lexical topic

markers dehua 'if and laijiang 'speaking of.

In addition to these pause-marked and pause particle-marked topic constructions, the

well-established topic constructions (such as double-subject constructions, LDs and

TOPs, etc.) are tagged individually and analyzed in their specific contexts, so as to give

a panoramic view of topic constructions in Mandarin and highlight how they schematize

into a complex linguistic category.

4.3 Research Goals

I have several research goals for this corpus study. First, I hope that this corpus study can help reveal the nature of topic in naturally occurring discourse and that this method of research can tell how topic compares to subject. For instance, definiteness, while often stated as a criterial feature of Mandarin topics (e.g. Tsao 1979, Xu and Liu 1998, Shi

2001, etc.), cannot be taken as a given in this study. It is highly probable for online

90 processing to come up with topics that are indefinite in reference. Second, I expect this

study to be able to demonstrate how pause particles are used for topic marking in modern

spoken Beijing Mandarin and resolve some of the controversies over the number of topic marking particles and how labor is divided among them. Subsequently, more can be

learned about the essence of topic marking, especially when taking into consideration the fact that topic markers are more often than not polysemous. It is well-known that

Japanese wa and eyebrow raising in ASL both occur in contexts other than to mark topics and have other semantic roles in these cases. Third, a large variety of topic constructions is expected. Xu and Liu (1998) argue that the Shanghai dialect is more topic-prominent than standard Mandarin, given that more topic constructions are found in that dialect.

Yuan (2002), however, does not agree with Xu and Liu, because he finds that standard

Mandarin is also very rich in topic construction. This corpus then is expected to reveal some new forms of topic coding. Finally, with the availability of a full range of topic constructions and the varying frequencies of occurrence of each type, a roadmap of topic evolution can be hopefully drawn up, thus highlighting topic as an abstract schema and different forms of coding as specific instantiations.

4.4 Summary

This chapter first describes the corpus data used for the present research. Then a brief introduction is made to the data analyzing procedure. The chapter ends with a short outline of the research goals of this corpus study.

91 Chapter 5: Pausing and Topic Marking

In this chapter, the use of pauses as topic markers is examined in three construction types as identified in the Beijing Corpus, namely the nominal construction, the double-subject construction and the TOP-subject construction, with a view to revealing when pauses occur in language use and what impact the pause has on the message that is encoded in the pause-marked expression.

5.1 Pausing in Nominal Constructions

A nominal construction is a clause that consists of two NPs, with the second NP predicating on the first NP. Following Shi's (1995) categorization, nominal constructions fall under the category of obligatory pausing after the first NP; that is, nominal constructions are TC constructions, where the first NP is the topic and the second NP the comment. This corpus analysis, however, shows differently. At the clause level, there are altogether 41 instances of nominal constructions that are identified in the Beijing Corpus, with 40 having no pause between the two (occasionally three) NPs. The only exception can be attributable to the fact that the speaker stumbled in his speech, as illustrated in (1).

(1) StfS —^WStfH&'h A'Nh&t, (001-N07:3)

Zhege, yitian de zhege-zhege, ba-ge xiaoshi,

PF one day NOM PF-PF eight-CL hour

'Each day, (they worked) eight hours.'

Here the speaker is talking about the number of hours that the rickshaw pullers had to

92 work per day during the old times, but he stumbles, as seen in the clause-initial zhege and two more occurrences of the same pause filler preceding the second comma. Without the stumble, the clause would read yitian ba-ge xiaoshi, literally 'one day, eight hours', which is both grammatical and appropriate in the context. The use of de in the original clause is somewhat perplexing. Primarily, de functions to link the modifier and the modified, which makes this utterance sound very much like an NPyitian de bage xiaoshi

'eight hours of a day'. A closer perusal reveals that there is no comment or predication on this NP in the ensuing discourse, given that the next clause has shifted its topic to the rickshaw. Therefore, the nominal construction reading seems more appropriate. The conclusion is then that in Beijing Mandarin, a physical pause is far from necessary in nominal constructions; rather, nominal constructions are seen predominantly without a pause in between.

Then arises the question: what is the relationship between the two constituent NPs, subject-predicate or topic-comment? Given the fairly rigid definition of subject in this research, i.e., an element that has a 'doing' or 'being' relationship with the VP, not all nominal constructions are SPs. In terms of semantic relatedness, the two NPs can be subject-predicate, possessor-possessed, or topic-comment.

(2) %-X I IW° (120-H15:1)

laoda jiaoshi.

The first child teacher

'The first child is a teacher.'

93 (3) £—&&} I H^g|. (246-L23:4)

cha-vidianr-de san-mao-qian.

Bad-a bit-NOM three-CL-money

'Not so good ones are thirty cents (per half kilogram).'

(4) a I 5£ffic (001-N07:3)

zhe jiaotong.

This transportation

'This is transportation.'

(2)-(4) are instances of nominal constructions in the subject-predicate relationship. In each case, the clause can also be said with a copula in between, thus making explicit the

'being' relationship. It should be noted that the nominal constructions in spoken Beijing

Mandarin show some variation in the coding form. In (2), the subject is in the form of an

NP; the first constituent in (3) is a case of nominalization marked by the nominalizer de; in (4), the demonstrative zhe 'this' sums up the preceding discourse and also serves as the subject in the clause where it occurs. In (5) below, the two NPs are in a possessive relationship.

(5) Mil I TO, (246-L23:4)

tamen si-iian,

They four-CL

'They have four rooms.'

In (5), tamen 'they' is the possessor of the four rooms. In the following example, the first

94 NP is best taken as the topic of the second NP.

(6) nWCT. | £+A^^c (204-L53:4)

siji-gong wu-shi-jiu-kuai-qian.

Worker at Level 4 fifty-nine-CL-money

'The salary of workers at Level 4 is fifty-nine RMB yuan.'

Strictly speaking, the second NP in (6) is not a predicate of the first one. The first NP is metonymically used for the worker's salary, instead of the worker himself.

WmXmMl I ff A II 5?K M)L II ->N (141-X03:17)

wo-fuqin tamen ger wu-ge. jier yi-ge.

My-father they brother five-CL sister one-CL

'My father has four brothers and one sister.'

In this example, the first NP can only be understood as a topic, with several comments following, and each comment is a nominal construction in itself. Thus, in (7), we see nominal constructions occur not only consecutively, but recursively as well.

If the above finding is compared with Shi's (1995) discussion, it seems that his obligatory pausing (see Section 3.1 for details) does not apply to actual language use.

This fact, however, does not necessarily suggest that Shi is wrong in making his argument.

If pausing can be both physical and potential, it is highly probably for the speaker to produce a nominal construction without a prosodically realized pause, though many native speakers would acknowledge the existence of a potential pause between the two

NPs in sentences like (2)-(4) as realistic.

95 The reason for pauses not to occur in the present corpus is at least twofold. On the

one hand, the two nominals in this construction type may exhibit the subject-predicate

relationship, as in (2) to (4), which in general is incompatible with pause marking in

Mandarin (e.g. Xu and Liu (1998:283) among others). On the other, the non-occurrence

of pause could be attributable to the length of the identified nominal constructions—they

are relatively short in linguistic coding, with no subject or topic expressions being heavily

modified.

In any case, it seems that native speakers of modern Beijing Mandarin tend to produce nominal constructions without pausing in between.

5.2 Pausing in Double-Subject Constructions

5.2.1 Occurrence of pauses

Irrespective of the contradiction in terms, the double-subject construction is characterized by the co-occurrence of a topic and a subject, and considered a typical topic-comment construction in Mandarin Chinese (e.g. Li and Thompson 1981, Chu 1998,

Tsao 1979, among others). A comprehensive cross-linguistic characterization of this type of construction is found in Langacker (2007:243): (a)NPl andNP235 both have some claim to subjecthood. (b)NP2 and the predicate can themselves constitute a clause. (c)The entire structure also has clausal status. (d)NPl is a topic with respect to the inner clause. (e)NPl is typically a possessor with respect to NP2. (f) Body-part relationships between NP1 and NP2 are prototypical.

35 In Mandarin, the coding forms for the topic and subject are not limited to NPs only.

96 (g)NPl is commonly an experiencer with respect to the inner clause; (h) These sentences are always stative. ' (i) They often translate naturally with have, despite the absence of a possessive verb.

In the corpus secured for this study, there are altogether 99 clauses identified as double-subject constructions. Though not very large in number, they show some complications of the construction in real-time spoken discourse.

According to Shi's (1995) categorization, double-subject constructions, like nominal constructions, also belong to the obligatory pausing type. The present corpus, however, shows a very different picture. Of the 99 instances, 62 have no pause within the clause.

Of the 37 instances where pauses occur, nine are marked by a comma. If the possibility for the comma to be the transcriber's choice cannot be ruled out, the instances with an actual pause would be even fewer. Chappell (1996) compares the intonation contour of double-subject constructions beginning with pronouns and those beginning with NPs. On the basis of an analysis of 21 colloquial narratives elicited by Mary Erbauge at National

Taiwan University, Taibei, in 1976, Chappell proposes the following refinement on the syntactic configuration of double-subject constructions in Mandarin Chinese: when the topic takes the form of a pronoun, it consistently falls within the same intonation unit with the subject NP; when the topic is coded as an NP, it typically falls outside the intonation unit of the subject NP (1996:490-491). In this study, 22 double-subject constructions are found to begin with a pronoun, and of them, 18 are conjoined with the comment clause with the remaining four marked by a pause particle. Therefore, this corpus shows some difference from the data Chappell uses. In any case, the general

97 tendency in Beijing Mandarin is for the topic and subject in double-subject constructions to occur in one intonation unit.

5.2.2 Semantic relationship between topic and subject

The present corpus shows that the semantic relationship between the topic and subject constituents in the construction in question is more varied than has been acknowledged. For instance, Li and Thompson (1981:93) maintain that there is a

"part-whole" relationship between the topic and the subject, and Langacker (2007:243) takes body-part as the prototype. Upon closer examination, the semantic relationship is much richer than part-whole or body-part: the subject in many clauses specifies a particular aspect of the topic; the two constituents are also seen in the predication relationship, where the first one can be either the agent or the patient of the action conveyed by the second. The part-whole relationship can be even reversed; that is, the second constituent can be the whole while the first part of the whole. (8) to (12) illustrate various semantic relationships as observed in the present corpus.

(8) M MW^fXn-m, (057-H20:3)

ta shenti bu tai-hao ba,

She body not too-good SFP

'She is not very healthy.'

This is a typical instance of double-subject construction where the second NP is part of the first one in denotation.

98 (9) j& Bi^fWifg, (120-H15:5)

ni shuijiao dou bu tashi,

You sleep even not free from anxiety

'You can't even sleep free from anxiety.'

The semantic relationship between ni 'you' and shuijiao 'sleep' is subject-predicate, i.e.,

the former undergoes the process of the latter.

(10)iEM&iM±> ^±i^^Wlo (113-D21:2)

tebie zhe da-haizi. kaoshang daxue bu rongyi.

Especially this first-child enroll university not easy

'Especially, it's not easy for the first child to get enrolled in a university.'

The predicate bu rongyi 'not easy' is both on the NP zhe da-haizi 'the first child' and the

activity indicated by the clausal subject, i.e., kaoshang daxue 'getting enrolled in a

university'. The relationship between the topic and subject is thus agent-action.

(ll)MMtt, M$M> &£3J^S«tfnJ, (309-X12:7)

tamen-de sangshi, jiejue ne, jiushi dao jiali weiwen,

They-GEN funeral arrangements solve PART namely go household condole

'As for their funeral arrangements, the solution is for us to go to their households and

express our condolence.'

The logical relationship between the topic and subject in this example is one of patient-action; that is, the topic is the grammatical object of the subject VP.

(12)'^M MP?ffi^¥A, (057-H20:5)

99 laosi liangkouzi dou shi junren,

The fourth child the couple both are army men

'As for the fourth child, the couple (the child and his wife) are both in the army.'

In reverse to the presumed relationship between the two NPs, i.e., the topic represents the whole while the subject part of the whole, the topic in (12) laosi 'the fourth child' can

only be half of the couple as indicated by the subject liangkouzi 'the couple'. Cognitively, the topic is used to introduce the couple of which the topic referent is a half. There is also a good reason to orient the couple by means of laosi 'the fourth child', rather than the couple, because as an offspring of the speaker, the fourth child is more accessible than the couple of which he makes a half. In this manner, a chain of reference points is observed in this example, first from the situationally anchored speaker to her fourth child, then from the child to the couple, i.e., the child and his wife. This chained reference relationship is more explicitly coded in (23) of Chapter 3, reproduced below as (13).

(l3)±%&.m%±W, &£!», j&Wtfl&JLM&o (057-H20:5)

Youqi wo na laoda ya, zhe-xifur ne,

Especially I that first child PART, this-wife PART,

ta jiu youdianr xiangfa.

she then a little dissidence

'Especially my first child, his wife has some dissidence.'

Here, the speaker is overtly coded through the use of wo in the possessive NP wo na laoda 'my first child', and the chain starts from the speaker to her first child, and ends at

100 the wife of her first child.

From the above illustration, it can be seen that the semantic relationship between the topic and subject of double-subject constructions in Beijing Mandarin is indeed more varied than the prototypical part-whole relationship.

5.2.3 Word order flexibility and its impact on construction semantics

Given the chained reference relationship as shown in (12) and (13), it seems that the topic and subject need to occur in a particular order in double-subject constructions. This supposition is indeed true with some, but not with others.

(8')shenti ta bu tai-hao ba,

(9') shuijiao m dou bu-tashi,

(10') ?kaoshang daxue, tebie zhe da-haizi bu rongyi.

(11') *jiejue ne. tamen-de sangshi. jiushi dao jiali weiwen,

(12') *liangkouzi laosi dou shi junren,

It is very interesting to see that the primed examples show a uni-directional scale from (8') to (12') in acceptability judgment. The more typical the construction is, the more acceptable the swapped double-subject construction. A tentative explanation is that in clauses such as (11) and (12), the time factor plays a more important role than in other utterances; that is, the solution of a problem cannot precede the problem, or only after one is married can there be a couple. In contrast, a person must be co-existent with his or her body, as in (8'). Likewise, sleep is among the most conventionalized activities of

101 mankind, i.e., human beings inevitably perform this activity. That (10') is of marginal acceptability, I suppose, is because the VP, though indicating a conventionalized activity, explicitly specifies the result of the action, thus also involving the time scale significantly.

In addition to this limited freedom of swapping between the topic and subject, there are also some changes in the semantics of the construction after swapping. Take (8) and

(8') for example. In (8), the 'body' must be the topic referent's, whereas in (8'), the body is used for general reference and the referent of the original topic is just one instance of all that have a body. The same is true with shuijiao 'sleep' in (9), where it indicates the activity undergone by the preceding pronoun; in (9'), by contrast, it can refer to anyone's activity so long as the activity can be categorized as 'sleep'. In this sense, the topic and subject exhibit a quasi-possessive relationship in instances such as (9), where the subject is a VP rather than an NP.

5.2.4 Possession as the prototype relationship between topic and subject

Of the 99 instances of double-subject constructions in the Beijing corpus, 76 show the possessive relationship. Therefore, this study supports Langacker (2007:243) and Li and Thompson's (1981:93) claim that body-part or part-whole is the typical relationship between the first two constituents in the double-subject construction. Meanwhile, utterances such as (9) have extended the connotation of possession to include the agent-action or experiencer-sensation relationship. Perhaps it is exactly the analogy

102 between the double-subject construction and the possessive construction that has

prevented an actual pause from occurring in the speeches as produced by the language

informants of the present corpus.

The observation that double-subject constructions are analogous to possessive

constructions has at least two consequences. In the first place, the construction resulting

from the swapping procedure is no longer a double-subject construction; it would be best

taken as a more general TC construction, because the first constituent after swapping just

specifies the 'frame' within which the ensuing proposition is to be interpreted (e.g. (9'));

it is no longer an instance that can be predicated upon, or in Langacker's terminology, it

has given up its "claim to subjecthood". In the second place, the so-called double-subject

constructions may constitute a challenge to Langacker's 'subjective' interpretation of

topic in contrast with possessives where he argues that the possessor serves as reference

point by virtue of some "objective relationship it bears to the possessed" (Langacker,

1991:314). Since double-subject constructions are a subset of topic-comment

constructions, if they align with the possessive construction and show some objectivity in

referencing, perhaps topics are not entirely subjective in their reference-point function.

Nonetheless, it should be borne in mind that the topic and subject in the

double-subject construction are not to be equated with possessives, not only because the possessive relationship cannot cover instantiations like (10), (11) and (12), but also

because double-subject constructions exhibit the following features.

First, pause is seen to occur after the topic in 37 clauses out of a total of 99 instances,

103 and 28 of them are filled with pause particles or lexical topic markers. Moreover, ten of the 28 clauses also have a pause and/or a pause particle after the second constituent or the subject. (10) is actually an instance of the first case and (11) the second. (14) can be compared with (8).

(UMm, m±MMM2M> #®J3S. (288-N40:3)

ta ne, zhege jiaoxue nengli ya, tebie qiang.

She PART PF teaching ability PART especially strong.

'Her teaching ability is especially strong.' Like (8), the topic ta 'she' is in a possessive relationship with the subject jiaoxuenengli

'teaching ability'; nonetheless, a pause occurs in between, which contrasts with (8).

Moreover, a second pause is seen to follow the subject. Consequently, the two pauses used in this clause have turned this construction into a 'shelled' TOP-subject construction, following how topic is identified in this research. Within the inner shell, jiaoxue nengli

'teaching ability' is the topicalized subject, with tebie qiang 'especially strong' as the comment. When it comes to the outer shell, ta 'she' serves as the topicalized subject, with the ensuing TOP-subject construction functioning as its predicate. That the clause is put under the category of double-subject construction is on account of the prototypical possessive relationship between the two NPs which precede the predicate adjective phrase.

Second, double-subject constructions also allow for insertion between the topic and subject, a phenomenon which is not observed in possessives. The underlined parts in the following examples show how elements of different metafunctions can be inserted

104 between the topic and subject. (i5)8&&, &#H*m, &ji*m, &m&&bm%, (253-154:12)

Zhe gonggong, gonggongqiche ne, zhe iinian na,

This FS bus PART this several years PART

fazhan ye shi bijiao duo,

development also is comparatively much

'The buses have also been fairly developed during the past several years.'

(16)j5Sjftji&Slf£A£, ffiXfcfeift, t£0!l^>fi?J. (232-T17:2)

houqin-de zhexie gongzuorenyuan,

Logistics-NOM these staff

xiangduilaishuo, bili shi shao-de.

relatively speaking proportion is small-CSC

'Relatively speaking, the proportion of the logistic staff is small.'

(\7)mnWB=±&i&P!'i&, (211-N52:4)

women na-fang vige shi dishi di,

Our that-house for one thing is terrain low

'That house of ours, for one thing, is located in a low terrain.'

If Halliday's terminology is used, the insertion in (15) is an experiential element, specifying the time dimension, that in (16) an interpersonal insertion, and in (17) a textual or discourse organizer. Therefore, the relationship between the topic and subject in double-subject constructions is looser than that in possessives.

105 5.2.5 Distribution of double-subject constructions

In addition to functioning independently in discourse, double-subject constructions

are also observed to occur within SP, LD and other TC constructions.

(mAMm&$ti$HWM, (204-L53:2)

renjia dou shuo wo piqi hao a,

Others all say I temper good SFP

'Others all say that I'm good-tempered.'

(I9)%m, ^TTtt^ggo (302-T49.-4)

huo ya, ta iiiiexing biiiao qiang.

Goods PART it seasonality fairly strong

'As for the goods, their seasonality is fairly strong.' vom&^jimumz, $tMw^mm\ (i83-x46:3>

woziger qingkuang dehua, wo shenti shi bucuo a!

Myself situation if I body is not bad SFP

'As for myself, my health is not bad.'

In (18), the underlined double-subject construction occurs within an SP clause, functioning as the object of the verb shuo 'say'. (19), as a whole, is an LD-subject, where the pronoun ta 'it' refers back to the clause-initial NP huo 'goods'.36 (20) is an explicitly marked TC construction, where the construction of concern here serves as the comment of the topic NP which is marked by the lexical topic marker dehua. Thus, this typical TC

36 Generally speaking, Mandarin Chinese does not mark number explicitly.

106 construction is seen to be more than a main-clause phenomenon; it may occur in other

linguistic contexts as well. In Section 5.3, a more varied distribution pattern of another

TC construction, i.e., TOP-subject, is presented and discussed in detail.

5.2.6 Summary

To summarize, the double-subject construction constitutes a subset of topic-comment

constructions in Beijing Mandarin. Contrary to some previous claims that pause is

required in such constructions, this corpus study shows that the great majority of

double-subject clauses as produced by native speakers of Beijing Mandarin are without pauses. This phenomenon may be attributable to the analogy between the construction in

question and possessives. In prototypical cases, the topic and subject in double-subject

constructions are in the possessor-possessed relationship, thus instantiating one kind of

Langacker's objective relationship, which in turn constitutes as a challenge to

Langacker's interpretation of topic as 'subjective' reference point. In the meantime, double-subject constructions differ from possessives, in that the topic and subject are related in a more varied manner, and pauses and insertions are allowed in between.

Double-subject constructions are also observed to occur within other construction types in the present spoken corpus, which calls into question the claim that TC constructions are main-clause phenomena. 5.3 Pausing in TOP-subject Constructions

The construction that best shows the function of pausing is most probably

TOP-subject, because without the pause, a TOP-subject would be an SP. Therefore, this

section focuses on a discussion of TOP-subject in contrast with SP, with a view to

showing how pause functions in naturally occurring discourse and whether it is a reliable demarcation between subject and topic in Mandarin Chinese. Occasional reference is made to English, a subject-prominent language, in order to highlight the pragmatics of

TOP-subject in a topic-prominent language.

5.3.1 Frequency of occurrence

Among a total of 13,000 or so clauses in the Beijing corpus, there are 708 instances

(or 5.4%) of TOP-subject constructions and 2198 instances of SPs. When segmentation is further conducted, another 711 TOP-subject constructions and 2686 SPs are observed to occur within other construction types. The numbers themselves may not tell very much about the corpus or about Mandarin; but when contrasted with English, the difference is evident.

Gregory and Michaelis (2001:1678-1679) use English data from the Switchboard

Telephone Speech Corpus and find 44 instances of TOP out of a total of 32,805 clausal utterances. As a percentage, TOP is no more than 0.134% of all clausal utterances in spoken English in Gregory and Michaelis's study. Further, their use of "TOP" is as a cover term for both TOP-subject and TOP-object, which means that the proportion of

108 TOP-subject is actually even smaller than 0.134%. In this corpus, 284 TOP-object clauses are observed to occur in addition to the 708 TOP-subject constructions. If added up, the proportion of TOP in the Beijing corpus is 7.63% of all clauses, or 57 times that in

English.

The frequency count of TOP-subject in Mandarin also varies with the genre. A case study of a novelette written in the spoken style of Beijing Mandarin (Liu 2007) shows that TOP constructions count for 1.5% of all clauses (cf. 5.4% in this spoken corpus), and of the TOP instantiations, 12.5% are TOP-subject clauses, in contrast with 70% found for the spoken corpus. The most likely reason is a stylistic difference between planned and spontaneous speech. Even though the novel is written so as to imitate a spoken variety of the language, its written form might suggest that it has in fact been carefully thought out by the author. However, further study along this line would be necessary to present a clearer picture of TOP usage differences in various genres of Mandarin discourse.

Gregory and Michaelis (2001:1678) also maintain that TOP and LD are "generally main-clause phenomena". They find only two exceptions to this characterization, and the example they give is an instance of LD which is used in a subordinate clause. In the following section, I try to show how widely TOP-subject distributes across construction types, subordinate clauses included.

5.3.2 Distribution of TOP-subject vis-a-vis SP

According to many linguists' observations, topic-comment constructions have very

109 different behavioral properties from SPs. Just as Gregory and Michaelis (2001:1678) argue for "the main-clause phenomena" for TOP and LD in English, Shi (2001) among others also maintains that while subject can be used in both matrix and subordinate clauses, topic can be used in matrix clauses only. Nonetheless, Table 5.1 below shows how widely TOP-subject, a subset of topic-comment constructions, is distributed across different construction types in the Beijing corpus.37

Table 5.1 Distribution and frequency counts of TOP-subject vis-a-vis SP

in the Beijing Corpus Context of use SP TOP-subject total TC 1644 555 2199 Subordinate clause in CC 341 9 350 LD-subject 74 4 78 LD-object 7 1 8 TOP-subject 12 4 16 total 2078 573 2651

From the table, it can be seen that TOP-subject, like SP, can occur in various linguistic contexts, from TCs and subordinate clauses in CCs to such typical topic-comment constructions as LDs and TOPs. Different cases are illustrated below under separate subheadings.

5.3.2.1 Within TCs

TOP-subject and SP are both used in various TC constructions. (21) and (22) show the opposition of the two constructions in TCs beginning with a conjunction.

37 Note that (18)-(20) above show how double-subject constructions can occur within other constructions as a constituent.

110 (21)Mi| jfmM^Mo (071-D57:6)

danshi iiaqian xiangdang gui.

But price fairly expensive

'But the price is fairly high.'

(22) MM. Mmm.-Wm, j&ifcb&ffftl. (176-N38:1)

suoyi jiating iiaovu ne. haishi biiiao hao-de.

So family education PART passably fairly good-NOM

'So the family education is fairly good.'

The conjunction in (21) is danshi 'but' and that in (22) suoyi 'so'. Following the conjunctive word, an SP is used in (21), where jiaqian 'price' is the subject. In (22), the topic jiatingjiaoyu 'family education' is separated from the comment by the particle ne.

Likewise, the first element of the TC can also be a time adverbial, with SP or

TOP-subject ensuing, as illustrated in (23) and (24).

(23)S£ fe&j^qig, (057-H20:3)

xianzai ta zai zixue ne.

Now he is self-study SFP

'Now he's self-studying.'

(24) BM. 3frSE%QB&&g&. (176-N38:1)

houlai wo-fuqin ne zai tielu.

Later on my-father PART be railway

'Later on, my father worked on the railway lines.' in The topic in TC may also be an explicitly stated domain for the following SP or

TOP-subject comment expression.

(25)WlMniPMmMi#, ?Eftl3iniRi£&£. (260-X39:6)

iiu cong zhishi fanchou laiiiang. si-de zhishi bijiao duo.

Just from knowledge category speaking dead-NOM knowledge fairly much

'When knowledge categories are concerned, mechanical knowledge is fairly much.'

(26)&^lkM^I%, ngd-JMPl, WJg^fg^JCjj, (043-N50:3)

zai jiuye litou ne, zanmen danwei ne.

In employment inside PART our unit PART

you zhaokao zhege jiashu.

have enroll PF dependents

'When it comes to employment, our employer enrolls our employees' dependants.'

The topic in (25) is lexically marked with cong...laijiang 'when X is concerned', and that in (26) is framed by the prepositional phrase zai...litou 'inside'. In both examples, the comment statement is to be interpreted with reference to the domain delineated by the preceding topic phrase, regardless of the coding form of the comment (SP in (25) and

TOP-subject in (26)).

Within TCs, there are also instantiations of shelled TOP-subject, as shown in (27). (27)TO m^xmmm-^mi', &&m&\k%mffi®L-R&)io (225-D47:2)

buguo liang-ge-ren ne fuvang yi-ge haizi.

But two-CL-people PART raise one-CL chilid

112 zenmezhao ye bi bie-de shaowei haodianr.

in any case too COMP other-NOM slightly better

'However, for a couple to raise one child is in any case a bit better than raising more

than one.'

Within this TC example, there are two levels of TOP-subject construction. On the lower level, liang-ge-ren, which is separated from the ensuing VP with the particle ne, is the topic; and on the higher level, the whole clause liang-ge-ren ne fuyang yige haizi 'one couple raise one child' is the topic of the ensuing comparative construction headed by bi.

Also, at this higher level, TOP-subject is seen to occur as a subordinate clause.

5.3.2.2 Within subordinate clauses of CCs

(28) and (29) show how TOP-subject is used within subordinate clauses.

(28)E&^2ttfi«, ftfegWTS^. (169-T58:5)

youyu sudu kuai a, ta zhangwo-bu-liao zhege che.

Because speed fast SFP he control-not-AUX this motorcycle

'Because the speed was very fast, he wasn't able to control the motorcycle.' a9)&mm&±m&, mrn^^-m^, #a##ift, RTJI&$&^*&. (274-T30:2)

suiran xianzai dajia dai, daidai38 de shoubiao ha. geshigeyangde.

Although now people wear NOM watch PART of various kinds

keshi wo shizhong shebude.

38 Stumbling of speech.

113 but I all along hate to part with

'Although the watches people are now wearing are of various kinds, I hate to part

with (my old watch) all the time.'

(28) and. (29) are two instances of complex sentences from the corpus. In (28), the underlined SP construction occurs in the subordinate clause headed by the conjunction youyu 'because', where sudu 'speed' is the subject and kuai 'fast' the predicate. (29) is a complex sentence of concession linked by the paired conjunctive suiran...danshi

'although...but'. In the subordinate clause, a TOP-subject is seen to follow the clause initial conjunctive word suiran, where the topic is the relative clause together with its head noun, separated from the comment AP geshigeyangde 'of various kinds' with the particle ha and an ensuing pause. The conclusion is that TC constructions are more than main-clause phenomena in Beijing Mandarin.

5.3.2.3 Within LDs

Since LD is a structurally defined construction type, it seems self-evident that SP will occur below LD. In this corpus, however, some complications are observed; that is,

TOP-subject constructions are also observed to occur within LD. (30) and (31) illustrate the opposition of SP and TOP-subject in LD-subject.

(30)WjSAqig, ifcWStfS PISJE, (141-X03:12)

vouxie ren ne, ta you zhege, ng yuanjian,

some people PART he have this PF foresight

114 'Some people, they have the foresight.'

(3n&fiUL, %Wfa, MM±> ffl^&#ff^o (001-N07:2)

Chaguanr. Longxugou,

Teahouse Longxu Slum

jiu zhege. dou shi Laoshe zuo-de.

precisely this all is Laoshe wrote-NOM

'Teahouse, Longxu Slum, these precisely are written by Laoshe.'

In (30), the underlined pronoun ta 'he' is co-referential with the clause-initial NP

which is marked by the pause particle ne. By contrast, they/w-emphasized demonstrative

zhege 'this' in (31) is separated from the ensuing VP with a pause, as indicated by the

comma, thus turning an SP into a TOP-subject.

The following two examples illustrate how SP and TOP-subject are used in

LD-object.

(32)Jfe%/MgJLPE, 3IMS W^m^Aimmffiit, (267-H24:3)

wo zhe xiaohair ba,

I this child PART

nage, nage jialiren bijiao jiaoguan ta,

PF PF family members rather pamper him

'This child of mine, my family members rather pamper him.'

(33)MlS±Zi£> ft£, ^mW, MMIB. (253-T54:4)

women zhege gongzuo, jiushi,

115 we this work PF

you-lingdao, ye zhua zhege dongxi.

have-leader too stress this thing

'This work of ours, some leaders stress it.'

The underlined parts in both examples are co-referential, thus making the example utterances LDs. The difference between the two utterances is that in (32) the second line

is an SP while in (33) it is a TOP-subject, given the fact that there is a comma separating the agentyoulingdao 'some leaders' from the verb.

5.3.2.4 Within TOP-subject constructions

The TOP-subject construction is also seen to occur within TOP-subjects.

(34)fe^]jg#±>M> mE3\TimmfiL, WL>&mmm\m%Lm.o (232-T17:6)

ta-de houbei liliang. xianzai dao-le zhezhong chengdu.

It-POSS reserve power now reach-ASP this degree

iiu bixu vinqi zanmen-de zhuvi.

then must catch we-POSS attention

'That its reserve power has now reached this level must catch our attention.'

In this example, the double-underlined segment in the first line is a TOP-subject within which the part preceding the comma is the topic. When the utterance is taken as a whole, the TOP construction in the first line functions as the topic of the comment in the second line, again separated with a pause. This constitutes another instance of

116 subordinate-clause-phenomenon TC construction.

5.3.2.5 Summary

To summarize, TOP-subject constructions are not only relatively high in frequency of use in spoken Beijing Mandarin, they have also shown very similar, if not parallel, distribution patterns to SPs, i.e., they can occur in various construction types. Moreover,

TOP-subject is also seen to occur in subordinate clauses, a property claimed exclusively as belonging to SP by many linguists. This finding, however, may not be as surprising as it appears, because the two constructions are differentiated only by the existence or non-existence of a pause and/or a pause particle. Otherwise, they would be identical in configuration. Given the situation, we need to find out if it is necessary to make the current differentiation. If the differentiation is more than something artificial, what does the pause tell us about information flow in naturally occurring discourse? Such questions are addressed in the following section.

5.3.3 Coding forms of topic in TOP-subject vis-a-vis subject in SP

There are several scales that relate the coding form with the information status of the starting point of a clause. One is Givon's scale in the coding of topic accessibility

(1983:17): Most continuous/accessible topic Zero anaphora Unstressed/bound pronouns or grammatical agreement Stressed/independent pronouns

117 R-dislocated DEF-NP's Neutral-ordered DEF-NP's L-dislocated DEF-NP's Y-moved NP's ('constrastive topicalization') Cleft/focus constructions Referential indefinite NP's Most discontinuous/inaccessible topic

If Givon's discussion is "formulated in terms of the speech receiver's...perspective"

(Givon 1990:895), Ariel's accessibility hierarchy is more from the perspective of the speech initiator (1991:449): Low accessibility Full name + modifier Full name Long definite description Short definite description Last name First name Distal demonstrative (+ modifier) Proximal demonstrative (+ modifier) Stressed pronouns + gesture Stressed pronouns Unstressed pronouns Zeros High accessibility

According to Chafe (1994:179), what is involved in the hierarchy is "the nature of the language necessary to make a shared referent identifiable in a given context". Chafe also takes Ariel's observation as "a valuable extension" of his discussions on identifiability that "could profit from an application to conversational examples". Given these guidelines, this section focuses on the coding forms of topic in TOP-subject vis-a-vis subject in SP and how differences in coding form impact information packaging in naturally occurring discourse.

118 Table 5.2 Coding forms and frequency counts

of subject in SP vs. topic in TOP-subject in the Beijing Corpus Context of use Below TC Below CC total Coding form SP TOP-s Sub-clause main clause SP TOP-s SP TOP-s SP TOP-s AD 16 6 4 5 0".;.W 0 0 0 31 AP I 0 0 0 Ots. 0 0 0 1 CC I 3 0 0 0 0 0 0 4 clause t 9 0 1 0 •:•*.. 0 0 1 12 DE 72 • 39 18 6%- 0 1 1 172 (DE) 1 1 6 0 0 -.?=; 0 0 0 8 DEMON 164 5 42 7 o ;•••:- 0 7 0 225 double-subject 0 . 2 0 0 0 0 0 2 construction V'i existentials 7 - 5 (> 3 W ' :1 0 0 22 LD 0 1 f) 1 •o^v;; 0 0 0 2 noun 185 47 106 44 ^57 ' 0 38 3 480 NP 644 411 51)3 316 m 10 61 20 2027 PP 26"-. 44 12 34 % 0 0 2 121 pronoun 880 56 "T26 50 248 4 219 2 2185 RC 20' 11 13 12 vf; 1 2 2 62 SP 16 17 15 14 0 1 0 3 66 TC 0 0 0 1 ® 0 0 0 1 textual i ; 0 1 0 # 0 0 0 2 expressions TOP-object -> :©• 2 2 • 0 0 0 0 6 TOP-subject 0 . 2 () 1 ,o: 0 0 0 3 VP 4ft. 47 40 40 •:2J 0 5 4 187 total 2086 708 1511 547 379 17 333 38 5619

It can be seen that the coding forms in Table 5.2 are not entirely based on the categories listed in Givon's scale or Ariel's hierarchy; instead, they are tailored for their actual occurrence in the spoken corpus and listed in the table in alphabetical order.

Altogether, 4309 SPs and 1310 TOP-subjects are counted. Proportionately, SPs are about

3.3 times more frequent than TOP-subject constructions. Given this proportion, many coding forms listed in the table become interesting. The first category is the clause, where

11 out of a total of 12 instances are found in TOP-subject constructions, with only one

119 occurring in SP. NP is the next major category that occurs disproportionately to the frequency counts of the two types of constructions. The total number of NPs occurring in

SP is 1090, with that in TOP-subject being 757, i.e., the proportion has dropped to 1.44:1.

When it comes to the coding form of prepositional phrases (PPs), the proportion becomes

1:2. Further down the list, the proportion for the instances of relative clauses (RC) in SP and TOP-subject is 1.38:1, that of SP 0.89:1, and that of VP 1.05:1. However, demonstrative pronouns, nouns and pronouns are more frequently observed in SP constructions than in TOP-subject constructions. 213 instances of demonstrative pronouns are observed in SPs vs. 12 in TOP-subjects (or 17.8:1 proportionately); for the coding form of noun, the proportion is 386 to 94 (or 4.1:1); and the proportion for pronouns is 2073 to 112 (i.e., 18.5:1).

If the coding forms are divided into two groups on the basis of their frequencies of use in the two construction types, demonstrative pronouns, nouns and pronouns are in the proportionate group, whereas clauses, NPs, PPs, relative clauses, SPs and VPs are in the disproportionate group. Then emerges the picture: subject in SP tends to be coded in simpler forms than topic in TOP-subject in Beijing Mandarin. In terms of Givon's topicality scale, the subjects that have a 'doing' or 'being' relationship with the predicate in the corpus are more topical than the topics in TOP-subject constructions; or in Ariel's terminology, the subjects are more accessible than the topics.

This finding may sound paradoxical at first, but upon closer scrutiny, it reflects the requirement of discourse organization. According to Langacker, two "phases" can be distinguished in a topic's "career", namely the establishing phase and the holding phase.

120 He argues that "at some point in the flow of discourse, an entity is first established as a topic, often by a special marking or grammatical construction...Once established, a topic holds sway over a certain stretch of discourse (perhaps until another one replaces it).

There may be little need to mention the topic explicitly during this phase of career"

(1991:313). For the Chinese language, Chu holds that the topic formation process consists of three stages, introduction, pick-up and continuation (1998:261). Though different in the number of the phases of a topic's "career", both Langacker and Chu maintain that a topic needs to be established in the first place. Givon (1983) has put forth an iconicity principle to explain the selection of heavy forms (e.g., full NPs) vs. light forms (e.g., unstressed pronouns and zero forms): the harder the referent is to process, the more grammatical material must be assigned to it. In this manner, pauses and pause particles constitute convenient means of topic establishment. Cross-linguistically, Janzen

(1998:vii) comments that topic marking in American Sign Language does not signal topic maintenance, and that topics tend to be more fully specified than other more topical elements in the discourse.

At a stage when the topic is not yet established, the referent of the topic expression tends to be less accessible, thus closer to the lower end of Givon's scale or the higher end of Ariel's hierarchy. Actually, it has been repeatedly seen that an established topic will be picked up and continued in the form of a subject or a zero, which evidences Langacker's and Chu's observation. An excerpt from a monologue illustrates the point very clearly.

(35) W, EtfS IfcfeflUmig, £, WtfLEiU. >l>JMo (071-D57:2)

a, zhege, wo laobar ne, shi, you xuyao-gao, xinzangbing.

PF PF my spouse PART FS have blood pressure-high, heart disease

121 Ta ne. shi basan-nian, tiqian jiu lixiu-le.

He PART is 83-year in advance already retire-ASP HI, SIS&£JkB£ft, ^#TM¥^7c

Ng, dao xianzai ta yijing li, lixiu-le liang-nian duo le.

PF till now he already FS retire-ASP two-years more ASP

Fan'er cong lixiu yihou ne,

On the contrary from retire later PART

youyu zhege meiyou gongzuo de yali le,

because PF without work NOM tension ASP

suoyi zhege, shenti ne, zhege, huifu-de hai bijiao hao.

So PF body PART PF recover-CSC still relatively good

Ta tian-tian zaochen chuqu qu liuliuwanr qu.

He day-day morning go outside go stroll go

'My husband has high blood pressure and heart disease. He retired as early as in

1983. Till now, he has been retired for more than two years. After retirement, he no

longer has the tension from work, so his health has recovered fairly well, which is

not quite expected. Every morning, he'll go and have a walk.'

From this example, it can be seen how topic is established and continued in naturally

122 occurring discourse. First, the topic, as underlined in line 1, is introduced with reference to the speaker in the form of a possessive construction. In the second line, the same referent is coded in the form of a pronoun, a more accessible form than the NP in line 1, but still separated the grammatical predicate with the pause particle ne. When it comes to

line 3, ta 'he' is no longer marked, i.e., it has turned from topic to subject. With the discourse developing further, the referent becomes off-stage, or subjectively construed in

Langacker's terminology. In line 4, the grammatical subject of lixiu 'retire' is made implicit, and the same is true with the verb meiyou 'not have' in line 5. The possessor of shenti 'body' in line 6 is also made covert. In the ensuing sentence, i.e., after the referent remains covert for three consecutive clauses, it is overtly coded again in the form of the pronoun ta 'he', functioning as the subject. This example as a whole shows how the coding form is related to topic/subject status. The overt coding of the subject in the last line deserves some attention. Since the referent is highly topical in the context, it seems unnecessary to code it overtly. However, the three-clause distance from its previous mention may have reduced its topic availability. The interference from another topic in the preceding clause (i.e., shenti 'body') further affects its topic status (see Givon 1983).

The overt coding under such circumstances then seems a better choice than leaving it to contextual inference.

Given the above discussion, it seems that topics in TOP-subject constructions are more information-loaded than subjects in SPs in the spoken variety of Beijing Mandarin.

Since the two constructions differ only in the availability or not of a pause after the clause-initial element, pauses turn out indeed to play a role in information packaging; that is, they tend to mark information-rich elements. This corollary is not only evidenced in

123 online speech; it also has a cognitive basis. When online speech is concerned, the intonation contour and the pause break are reflexive and indicative of the operation of the human mind, because, as a result of the way the human mind handles information flow, speakers typically deliver no more than one new idea in each intonation unit (Chafe 1994, ch. 9). In observing natural spoken language, Halliday states that "the tone group...is not only a phonological constituent; it also functions as the realization of something else, namely a quantum or unit of information in the discourse" (Halliday 1985:274-275). Ho

(1993) uses data consisting of spontaneous but primarily monologic speech collected from interviews where native speaker informants were asked to engage in narrating, explaining, and describing activities,39 and observes that disfluencies typically concentrate at the theme/rheme (i.e., topic/comment) juncture in Chinese thematic and information structure. Tao (1998) also argues that the cognitive effort required in activating and verbalizing new information is suggested as the major factor that causes the delay of the production of speech and the resulting split of a syntactic unit (e.g. the agent/subject (A/S) arguments from VP) across IUs. When the A/S arguments are separated from the VP, TOP-subject constructions emerge. In addition to the cognitive constraints on information flow, Tan and Tao (1998) suggest the degree of structural integration between syntactic constituents as another factor. For example, the prevalence of elliptical clausal IUs in speech seems connected to the less integrated constituent relation in the subject-predicate structure (as compared to that in the verb-object

39 Very much like the corpus secured for this research in genre.

124 structure). The phenomenon observed here is also in agreement with several principles or constraints on subject. One is Chafe's Light Subject Constraint (1987, 1994). After stating that many information units employ a subject to express the starting point, Chafe

(1987:37) proposes the constraint: (1) A starting point is usually a given referent. (2) Occasionally a starting point is an accessible referent. (3) A starting point is rarely a new referent, and then only at the beginning of a major section of a discourse.

Very similar to Chafe's Light Subject Constraint is Du Bois's Given A Constraint, i.e.,

"avoid introducing a new referent in the A-role argument position", where A is the transitive subject (Du Bois, 1987). Lambrecht's Principle of Separation of Reference and

Role is also a good account of this phenomenon: "the grammatical principle whereby the lexical representation of a topic referent takes place separately from the designation of the referent's role as an argument in a proposition" (1994:185). Lambrecht explains that the communicative motivation of this principle can be captured in the form of a simple pragmatic maxim: "Do not introduce a referent and talk about it in the same clause".

Therefore, it is expected for the grammatical subject to be separated from its predicate by means of a pause and/or a pause particle when it is complicated in coding form or rich in information load, so that it occurs in a different intonation unit from its predicate. In a similar vein, it is not meaningless or just something artificial to make the distinction between SP and TOP-subject.

125 5.3.4 Discourse functions of TOP-subject constructions

5.3.4.1 Referencing to some findings in English

Not much research has been conducted on the discourse function of TOP in

Mandarin Chinese. The TOP construction has been used either to show the flexibility of syntax in the language (as in Lii 1986), or taken as non-canonical (especially in comparison with the SVO sentence pattern) and grammatically constrained in use (e.g.

Chen 2004). However, there are two studies on the functional opposition of TOP vs. LD in English that are applicable to this discussion. Since TOP is a structurally defined construction and observed in numerous languages all over the world, the findings in

English may be used here for reference, and the contrast between Mandarin and English in the use of the same construction may highlight the differences between topic-prominent languages and subject-prominent languages.

Prince (1997:128) proposes a 'double discourse function of topicalization': 1. Topicalization triggers an inference on the part of the hearer that the entity represented by the initial NP stands in a salient partially-ordered set relation to some entity or entities already evoked in the discourse-model. 2. First, if the entity evoked by the leftmost NP represents an element of some salient set, make the set-membership explicit. Then in all cases, the open proposition resulting from the replacement of the tonically stressed constituent (in the clause) with a variable is taken to represent information saliently and appropriately on the hearer's mind at that point in the discourse, the tonically stressed constituent representing the instantiation of the variable and the new information.

To put it simply, Prince believes that the second part of the double function "attributes a

'focus/presupposition' or 'focus/focus frame' (Partee 1991) information structure to the

126 proposition conveyed by a Topicalization...". Her example is reproduced below as (36)

for convenience of reference:

(36)'She had an idea for a project. She's going to use three groups of mice. One, she'll

feed them mouse chow, just the regular stuff they make for mice. Another, she'll feed

them veggies. And the third she'll feed [e] junk food.'(Prince, 1997:129, Example

12)

In this example, the open proposition is she '11 feed the third group X, where the variable

X\s junk food.

When it comes to the preclausal NP denotata, Gregory and Michaelis find that the

majority in TOP are "(a) previously mentioned and (b) fail to persist as topics". Therefore,

they conclude that the TOP pattern is not used to establish new topics (2001:1969). They

have also examined the use of pronominal preclausal NPs in TOP and find that none of them in TOP sentences is a deictic pronoun. The example that they give is reproduced as

(37) for illustration (2001:1698).

(37) B: — and they do a lot of fishing there. Of course what I used to like to do though is

go snorkeling for scallops -

A: Oh, yeah.

B: -in some of the bays, or uh, or oy~, for oysters.

A: That would be fun.

B: And, uh, that I was able to do,40 when I could go in and go down and actually

40 Unlike Prince, Gregory and Michaelis have bolded the whole clause, but in this example that is the topic element.

127 pick them off the bottom. (Gregory and Michaelis, 2001:1698, Example 35)

They further conclude that TOP is most welcome in anaphoric contexts. Their example

39 is reproduced below as (38) (2001:1702).

(38) Context: A has just outlined some possible policies for local school board.

B: Uh huh. That's some pretty good ideas. Why don't you do something with those?

You should run for a local school board position.

A: That I'm not sure about. I've got a lot of things to keep me busy.

From the above discussion, it can .be seen that TOP is more backward-looking than forward-looking in English. What it is like in Mandarin deserves some attention.

5.3.4.2 Context of use of TOP-subject in Beijing Mandarin

In this section, I examine the context in which TOP-subject utterances occur.

Altogether, 464 instances in the corpus follow a sentence boundary represented by a period in the transcription. In addition, 25 more occur at the beginning of a paragraph and another eight at the very opening of a monologue. Conversely, those following a comma

(i.e., a pause, but not a sentence boundary) are 197 cases. The remaining 15 cases are preceded by a question or an exclamatory expression. When the ensuing punctuation is eximined, 300 TOP-subject constructions end in a comma, with 127 of them persisting on for a scope of two to seven clauses. (39) illustrates how the scope extends before the demonstrative pronoun appears in the last line to refer back to the topical referent.

(39)i®, m^B&^w, II^OT^-'J^MO (OSO-NI4:I)

128 ng. nage nanhaizi ne. ng shi Shifuma Yixiao biyede.

PFthat boy PARTPF is Shifuma No. 1 Primary School graduate

Yixiao biye, kao-de shi Xuanwu Shisizhong.

No.l Primary School graduate sit for an exam-NOM is Xuanwu 14 High School

Shisizhong chuzhong biye,

No. 14 High School junior high school graduate

Wanle, gaozhong41 shi kao-de nage Nansanzhong.

Then senior high school is sit for an exam-NOM that No. 3 Boys'High School

Nansanzhong biye ya, kao Tianjin Daxue, JixieXi.

No. 3 Boys' High School graduate PART exam Tianjin University, Mechanics Dept.

Xianzai zai Beijing Sanlingsansuo gongzuo.

Now at Beijing No. 303 Research Institute work

You liangge xiaohair, yige nanhair, yige nuhair.

Have two children one boy one girl

41 Though this NP occurs in the subject position, it is by no means the agent of the action indicated by following verb. It is best taken as a topic instead of a subject.

129 15 TE^ei—-o

Zhe shi Laoer.

This is Laoer

'As for the boy, he gradated from No. 1 Shifuma Primary School. Upon graduation,

he got enrolled by Xuanwu No. 14 High School. He finished his junior high school

there, and for senior high school, he got enrolled by No. 3 Boys' High School. After

he graduated from No. 3 Boys' High School, he went to Tianjin University, the

Mechanics Department. He now works at Beijing No. 303 Research Institute. He has

two children, one boy and one girl. This is the second child.'

What is observed here obviously goes against Gregory and Michaelis' finding for English that the topic in TOP constructions does not persist. Clearly, the 300 TOP-subject clauses ending in commas persist. As a matter of fact, the topics in those ending in periods may also maintain their topical status in the ensuing discourse. (39) is a very good illustration of this point. Of the seven clauses that share the same topic, only one ends with a comma while six close with a period.

(35) is an alternative form of persistent topic to (39). In (35), the topic established in the first line by means of an NP is picked up in a simpler form (i.e., a pronoun) in the second line and persists as the topic in the zero form till the second last line. The topic persists in (35) through step-by-step simplification in the coding form or gradual increase in accessibility. Like TOP-subject constructions, SPs also end in periods although the sentence boundary itself does not hinder the subject from persisting in ensuing clauses.

130 Therefore, in the spoken variety of Beijing Mandarin, not only can TOP-subject be used to establish topic, but established topics may very well persist over the scope of several sentences.

5.3.4.3 Information status of topic in TOP-subject

When it comes to the information status of the clause-initial constituent in

TOP-subject, a whole range is observed. Of those following a comma, 108 cases (54.8%) convey discourse-old information, with the remaining 89 (45.2%) expressing accessible, generic or indefinite information. Of the three subtypes, those expressing accessible information constitute the majority, totaling into 79 cases or 88.8%. Below are some examples.

(40)$J5Blg. ^Wt^-^)IT, (288-N40:5)

Ranhou ne, ta jiu you yige erzi,

Then PART he at once have one son

Zhege erzi ne, ye bushi ta-ziji qinsheng-de,

This son PART too not himself of one's own

'Then he had a son, but this son was not his own.'

In this example, the second clause begins with zhege erzi 'this son', which repeats the discourse entity introduced in the preceding clause. In this way, it conveys discourse-old information.

131 (41)M_gJMnA£XWPI. (204-L53:l)

Erqie women renyuan you youxian,

Also our staff too limited

Yi-ge-ban ba. yejiu sanshi-duo-ren,

One-CL-shift PART only thirty-over-people

'Also, we are very tight in manpower. There are only thirty odd people in one shift.'

In (41), the staff as a whole is mentioned in the first line. The second line then begins with part of the staff, namely a shift. Given the fact that the NP yigeban 'one shift' is never mentioned in previous discourse but somehow anchored by the 'staff in the immediate context, it is best interpreted as accessible information.

(42)$Lm»k, 1> %Mm&iE, WftJ^i^fcttgtfHiBftJ, (288-N40:5)

Jiating ha, ng, jiating ba fanzheng, youdejiating shi bijiao hexiede,

Family PART PF family PART nonetheless some families is fairly harmoneous

youde jiating ne, jiu bushi tebie hexie.

Some family PART then not extremely harmonious

'When it comes to families, some are fairly harmonious, whereas others are not

quite.'

In (42), the clause-initial NP in the TOP-subject construction in the second line is modified by youde 'some', which makes the referents of the NP indefinite. The

132 expression just denotes that there are such families, without specifying which ones fall under this category. It should be noted, however, that the topic expression, though indefinite in reference, is in contrastive use with the indefinite subject in the preceding clause.

The situation with the TOP-subject constructions following a period mark is very similar to those following a comma in the information status of the clause-initial constituent. 266 instances (57.3%) repeat a referent that has occurred in earlier discourse;

198 (42.7%) clauses involve discourse-new (i.e., not explicitly coded in previous discourse) information at the starting point, and of the 198 clauses, 181 (91.4%) are accessible in information status, with the remaining being either generic or indefinite, seen in (43) to (46) below.

(43)-^7NPJLAO (127-L27:1)

Yigong shi liu-kour-ren.

Altogether is six-CL-people

TNPAA^, m^m, WSA, WH^®^,

Liu-kour-ren ne, you fuqin, you airen, you san-ge-haizi,

Six-CL-people PART have father have spouse have three-CL-children

'Altogether there are six people. The six people include father, spouse and three

children.'

It is quite obvious that the first NP in the second clause in (43) repeats what is mentioned in the second half of the preceding utterance, thus conveys discourse-old information.

133 (44)&ftj*i#. ik-MA%i$&&tkmmo (197-X22:3)

Zongdelaijiang, Beijing qihou haishi bijiao hao de.

Generally speaking Beijing climate rather fairly goodNOM

Dongtian ne. ye bushi tebie hanleng,

Winter PART too not extremely cold

xiatian ne. suiran re yixiege,

Summer PART though hot somewhat

ye bushi shuo tebie yanre, hai bijiao, bijiao heshi de.

too not speak very hot still fairly REP comfortable NOM

'Generally speaking, Beijing's climate is fairly good. It's not very cold in winter.

Though it's hot in summer, it's not very bad. It's still comfortable.'

In this example, the two underlined NPs are accessible from the expression qihou

'climate' in the first line—the four seasons of the year are among the most frequently accessed referents in talks on climate.

Yiban de haizi. xiang women tongxue shenme-de ha,

Average NOM children like our classmates the like-NOM PART

fB^ftW—*P£A*JEffS]. (309-X12:3)

134 dou deng deiyou yi-nian-duo de shijian.

all wait have one-year-over NOM time

'Average children, like my classmates, all waited more than a year's time.'

In this clause, the sentence-initial NP is generic in reference. The information status of the topic expression in (46), however, is intriguing.

(46)ni, jEW^^gADE,

ng, hai youxie nianqingren ba,

PF still some young people PART

^S.a^3itl^"Mfltl^iK, (239-N06:6)

quefa zhege daode fangmian de , zhishi,

lack PF moral respect NOM knowledge

'Well, there're still some young people who lack moral knowledge.'

The clause-initial constituent has a double interpretation. It can be either an NP or an existential expression. In either case, it denotes the existence of a subgroup of young people, but without specifying the referents. In other words, the referents of the topic expression are not definite. However, the speaker and the listener do not have much difficulty in identifying the subgroup under discussion. Utterances like this, therefore, seem to suggest that identifiability is a more primary property of topic than definiteness.42

Of the 25 paragraph-initial TOP-subject constructions, the information conveyed by the clause-initial constituent is either discourse-old (eight instantiations in total) or

42 More discussion on the property of identifiability is found in Chapter 7.

135 pragmatically/semantically accessible (17 instantiations). When the eight TOP-subject constructions taking the monologue-initial position are concerned, six are situationally oriented, with the other two related to the given interview topics. (47) and (48) are examples.

(47)Pjt, %QgJ6, Pjt, ®.G&%fttt&MmWL'W, (015-D25:1)

Ai, wone. shi, ai, meiyou shi shou-guo shenme gaodengjiaoyu,

Oh I PART is PF not is receive-ASP what advanced education

'As for me, I haven't received any advanced education.'

This clause occurs at the very beginning of a monologue, and the particle marked topic wo T is an entity in the communication situation.

(mm^mnmm, -^WAPAA. (267-H24:i)

Nage women-j ia ba. yigong you liu-kour-ren,

PF my family PART altogether have six-CL-people,

'My family has six people altogether.'

This TOP-subject construction also occurs monologue-initially, and the clause-initial NP is anchored through the given topic for the interview, namely Topic B: family, education, extraordinary experiences, procreation and social security.

5.3.4.4 Summary

The discussions and illustrations above show that what holds true for English does not apply to Mandarin Chinese. The denotata of the clause-initial NPs need not be

136 previously mentioned, because they can be deictic pronouns (e.g. when the speaker uses

T to refers to himself or herself) or generic referents. Though TOP-subject constructions

are "welcome" in anaphoric contexts (given that more than half of them involve second

mention of some referent(s)), they also occur in non-anaphoric contexts, for instance, at

the very beginning of a monologue. TOP-subject constructions are also observed to

persist as topics, and the scope can be fairly broad. Given these contrasts, it can be

concluded that the TOP-subject construction behaves very differently in Mandarin, a

topic-prominent language, from its counterpart in English, a subject-prominent language.

It seems to have a more important role to play in languages where topic-comment

constructions have gone beyond sheer pragmatic motivations with occasional use and

become a frequent phenomenon or simply the norm.

5.4 Summary

This chapter examines the use of pauses as topic markers in three construction types

in the Beijing corpus. Although Shi (1995) classifies nominal constructions as a type

where the occurrence of a pause is obligatory, the present corpus shows that the great

majority are produced without pauses. Another construction type, namely the

double-subject construction, also tends to be produced in one intonation unit; that is,

more often than not there is no pause within the construction. Moreover, the distribution of TOP-subject, which is identical in configuration with SP except for the occurrence of a pause after the grammatical subject role, almost parallels that of SP.

137 These linguistic phenomena put together seem to call into question the discourse function of pausing as the demarcation between topic and subject, but a closer examination reveals that the demarcation is far from artificial. The occurrences of pauses and pause particles turn out to correlate with the information status of the pause-marked element. The clause-initial element in TOP-subject tends to be more complicated in coding form and accordingly more information-loaded than its counterpart in SP. In this manner, the occurrence of pauses turns out to be a requirement of the cognitive constraints on information flow. It is perhaps for the same reason that pauses are less frequently observed in nominal and double-subject constructions, given the fact that most of their instantiations in the present corpus are fairly short, thus not very heavily loaded with information. The affinity of prototypical double-subject constructions to possessives further reduces the rates of occurrence of materialized pauses.

138 Chapter 6: Topic Markers in Beijing Mandarin

This chapter presents a comprehensive study of high frequency topic markers used in the Beijing corpus. Pause particles ne, ba, a (and its phonetic variants) and ma and two lexical topic markers dehua 'if and laijiang 'speaking of are analyzed in terms of frequency of use, distribution pattern and discourse function. Following this is a discussion on the diversity of topic marking in the language, the information status of the marked expression, and the cognitive significance of marking in naturally occurring discourse.

6.1 Pause Particle ne

6.1.1 Frequency and context of occurrence

According to the clause-by-clause tagging completed on the sampled corpus of

Beijing Mandarin, ne is found to be the most frequently used pause particle. In all TC constructions, ne appears 1422 times, among which 1279 instances are followed by a comma and 129 without. Moreover, ne is also found to co-occur with other topic marking expressions such as cong...laijiang 'speaking of, cong...laishuo 'speaking of, dehua

'if, laikan 'looking from...', laijiang 'speaking of, laijiang dehua 'if speaking of, laishuo 'speaking of, zaijiushi 'then is', etc.

When it comes to the linguistic context of use for particle ne, Table 6.1 presents an overview. With some simplification, the linguistic context of the clause within which ne is used as a topic marker is identified by how the clause is bounded by punctuation marks

139 in the transcribed corpus, with "left boundary marker" (or LBM) designating the

punctuation mark that precedes the clause and "right boundary marker" (or RBM) the

punctuation mark that closes the hosting clause (i.e., where the particle in question

occurs). When the LBM of a «e-hosting clause is a comma, the we-marked topic occurs in

the continuing part of a discourse stretch. By contrast, when the LBM is a period, the

we-marked topic occurs in the opening part of the immediately following discourse

stretch. When it comes to the RBM, the comma means the we-marked topic occurs early

or in the middle of a discourse stretch and the period means the we-hosting TC clause

closes a discourse stretch. Also, clauses that contain ne are found to occur at paragraph

boundaries and occasionally at monologue or discourse boundaries. Other boundary markers, such as question marks and exclamation marks, are also observed to bound

He-hosting TC constructions, but in much lower frequency than commas and periods.

In Table 6.1, nel means the particle ne is used to mark the first topic element in the clause. Likewise, nel marks the second topic, ne3 the third, and we-FiNAL the last one. It should be noted, however, that «e-FlNAL may conflate with nel or ne3; that is, ne-FiNAL may be the second pause marker in some cases and the third in others. For example, if one instance of ne is used to mark the second element of a clause and this element happens to be the grammatical subject or object of the predicate, this ne is counted as we-FlNAL, rather than ne-2. If the marked element is not the grammatical subject or object, it is tagged as nel, not «e-FiNAL. To put it differently, we-FiNAL is confined to key argument roles such as subject and object. To group topic-marked key argument roles

140 under a separate category is to highlight their particular status as "focal participants" of the clausal process. According to Langacker, "Access to a profiled relation via its focal participants is intrinsic rather than contingent" (1998:19). Occasionally, four or even

more marked topics may co-occur, in which case each marking is counted separately.

Table 6.1 Linguistic Context of Use of Particle ne within the Scope of Clause

Boundary nel nel nei We-FINAL marker LBM RBM LBM RBM LBM RBM LBM RBM comma 229 439 26 56 6 54 98 period 724 544 120 99 19 15 166 128 question 14 10 2 1 6 3 mark43 exclamation 1 2 1 mark paragraph 30 7 9 1 1 9 7 boundary monologue 5 2 1 1 1 boundary not clear44 1 total 1003 1003 157 157 21 21 238 238

From Table 6.1, it can be seen that «e-marked TC clauses are mainly bounded by commas and periods. However, the frequency counts show that there are fewer commas to mark the left boundary of we-marked TC constructions than there are to mark the right boundary. For nel, the figure is 229 to 439; for nel, 26 to 56; for ne3, zero to six; and for we-FiNAL, 54 to 98. Since commas are used to mark unfinished utterances, the figures show that ne is more frequently used in the opening or middle part of a discourse stretch.

Questions are also observed in this spoken corpus. However, they are not targeted for the listener. They are mainly used as rhetorical questions which expect either no answer or the speaker's own answer. Therefore, the occurrence of questions does not necessarily conflict with the nature of monologues which make up the present corpus. 44 This is a carryover from the transcript where the transcriber used the ellipsis mark to indicate a stretch of intelligible utterance from the language informant.

141 The frequency counts of periods show a similar tendency; that is, there are more periods to function as LBMs than as RBMs for He-marked TC constructions. The figures are 724 vs. 544 for ne\, 120 vs. 99 for nel, 19 vs. 15 for ne3 and 166 vs. 128 for rce-FlNAL.

Although the punctuation pattern in Mandarin Chinese is fairly flexible and the topic chain may go beyond the scope of a sentence as bounded by the period, the consistency in the frequency counts of comma and period as LBM or RBM for «e-marked TC clauses may suggest that ne is used mostly to open a topic chain. This finding turns out to be in conflict with what Yuan (2002) and Zhang and Fang (1996) have claimed for the particle ne, i.e., that it tends to occur in the closing instead of the opening clause of a discourse stretch or a clause serial.

This conflict is unambiguously evidenced by two «e-hosting TC clauses that are right bounded by a question mark. In (1) and (2), ne is used to mark topics that commence, rather than close, a discourse stretch.

(i) x&mnmh, rt^-kmm&jim, thM/s? (O43-NSO:2)

Youqi zuo fumu de, duiyu zinii de kanfar ne, shenme ne?45

Especially do parent NOM as for children NOM comments PART what SFP

'Especially as a parent, my comments on my children, what are they?'

(2) StfS ffi&tofomm, WLM, ag^Ai^liJfhMJS? (141-X03:10)

Zhege, iiuve de wenti ne, jiushi, woziger ganjuedao shenme ne?

PF employment NOM issue PART PF myself feel what SFP

45 Particle ne may also occur at the end of a clause or sentence as a sentence final particle (SFP) (e.g. Li and Thompson 1981:300-307, etc.), but with a different function.

142 'As for the issue of employment, well, how I feel about it?'

Both of these two examples introduce a new topic into discourse, namely the speaker's opinion on children in (1) and the speaker's feeling on employment in (2). If a comparison is made between the two clauses, (2) is more global than (1), in that it occurs at the beginning of a paragraph where the discourse shifts from one interview question

(i.e., education) to the present one (i.e., employment). In well-planned discourse, these two clauses would most probably combine with the ensuing clause(s) to form a statement, which is reminiscent of Chao's interpretation of subject and predicate in Mandarin as question and answer (1968:81-2).

The function of ne as a marker on discourse-new topics is also evidenced by the use of we at paragraph as well as monologue boundaries: not only can ne appear in paragraph- as well as monologue-opening clauses, but also its frequency of use in such contexts outnumbers that in paragraph- or monologue-closing clauses. The total number for paragraph-opening ne is 49, with that for paragraph-closing purposes being only 15. The numbers for monologues are 6 and 3 respectively. Here is the question: what motivates the occurrences of ne in these contexts? The corpus data show that ne, like Japanese wa, is polysemous, and the different pragmatic functions of ne predetermine its use in different contexts.

6.1.2 Discourse functions of ne-marked topics

According to Yuan (2002), the basic function of ne as a topic marker is to shift the

143 topic by establishing a new starting point of utterance. Upon closer examination of the spoken data, ne can be used to mark topics of varying discourse functions. In Table 6.2, eight different discourse functions are identified for «e-marked topics in the Beijing

Corpus, though some of them may also be roughly understood as topic "shifting", given the fact that Yuan does not elaborate the term in his paper. In what follows, the eight categories are briefly characterized and illustrated in order of frequency of occurrence in the Beijing corpus.

Table 6.2 Discourse Functions of we in Different Contexts

Function nel nel nei #f

6.1.2.1 INTRODUCING

The He-marked topic expression introduces a new dimension to the discourse, with discourse-new entities being the prototypical instantiations. An expression indicating an action, activity or event may also be introduced by ne to topichood. Adverbial components like time and place are also frequently observed. Of all categories, topics performing the INTRODUCING function constitute the most frequently occurring one. This category may also be taken as topic shifting, in that the topic is shifted to a newly

144 introduced element. The underlined part in (3) is an example of this category of topic.

(3) &^-^HiPA^£, AU^bUfefeft. (183-X46:10)

Zheme shi yige chadui xuesheng, cong dongbei huilai-de.

This is one work in the country student from northeast return-NOM

O, shengao ne. zhe-ger, gen wo yibanr-gao,

oh, height PART this-height with I same-tall

'This one was a student who once worked in the countryside and came back from the

northeast. In height, she was about the same as me.'

A female student is under discussion in (3). In the second line, one dimension of the

student, namely her height, is introduced by means of a we-marked topic expression.

6.1.2.2 CONTINUING

The we-occurring TC clause can also be a continuation of previous discourse, which

is characteristic of narrative discourse. This category may also be taken as shifting the topic to a later stage of discourse development, as shown in (4).

(4) m^^m, ffiS'N&S'K ®JftIf£l!IBX^gg>NL*flft, &m, (162-H37:1)

O houlai ne, shuo zhege shuo zhege,46

Oh later PART speak PF speak PF

bie-de gongzuo ne you xuyao ziger qu-zuo, shiba,

46 Stumbling of speech.

145 other-NOM work PART again need self go-do TAG

'Later on, I was needed for some other work, you know.'

In (4), particle ne is used to mark the sentence-initial time adverbial, which moves the narrative onto a new stage of development. Other marked CONTINUING topics may also be adverbials of manner as well as textual elements such as conjunctions.

6.1.2.3 REPEATING

REPEATING topics constitute a subset of CONTINUING topics, with the marked topic being active or readily accessible in the context. It should be noted, however, that this functional category does not "shift" the topic to a different one, as illustrated in (5).

(5) mm, m^MZ^UTW, (267-H24:l)

Wone. zai nage caijingxueyuan gongzuo,

I PART in PF Institute of Finance and Economics work

ng gao nage biyi, bitour fanyi.

PF do PF translation text translation

Wo ne shi waiyuxueyan nage yingyuxi biye-de.

I PART is Foreign Languages Institute PF English Department graduate-CSC

'As for me, I work at the Institute of Finance and Economics as a translator,

translating written texts. I graduated from the English Department of the Foreign

146 Languages Institute.'

There are two underlined instances of wo T in this example, each topic-marked by particle ne. The second one simply repeats the first one, without changing or shifting the topic.

6.1.2.4 PROMPTING

PROMPTING topics are another subset of CONTINUING topics which contrast with

REPEATING instances in that the referent of the marked topic has remained dormant for some time in discourse.

(6) 'bfo%&=Pm>, ft#51-®. (113-021:2)

Xiao-de zhe haizi ne, renwu geng zhong yixie.

Little-NOM this child PART task more heavy some

'As for the younger child, his study tasks are even heavier.'

When the discourse context is referred to, the previous mention of the child occurs 14 clauses earlier when the speaker talks about his two children. The intervening clauses are mainly on the older child, without explicit referring to the younger one. Prompted by the particle ne, the child comes to the foreground and remains topical for four consecutive clauses. In this sense, PROMPTING also serves the function of shifting the topic.

In addition to PROMPTING a discourse-old topic, ne can also be used to prompt a discourse-new topic on condition that the topic is somehow anchored in a broader context.

For instance, (7) occurs at the very beginning of a monologue.

147 (7) mto^At&mm, wtmfe&m, m, tmn^, (274-T30:D

Wo-de geren jingli ne, jiushi zenmeshou ne, ng, bijiao jiandan,

I-GEN personal experiences PART PF how to say PART PF fairly simple

'My personal experiences, how to say, are fairly simple.'

The topic prompted by ne, though occurring discourse initially, is anchored in a double manner. Linguistically, it is prompted by means of the possessive construction, where the possessor is the situationally available speaker wo T; contextually, the prompted topic is on the list of subject matters for the interview.

6.1.2.5 CONTRASTING

By CONTRASTING, 1 mean that the He-marked TC clause is to contrast with an element in previous discourse, as in (8)

(8) m%tt%tfimmWm#Ltfy&ft&&M, ^i> (169-T58:2)

Yinwei sizhe-de jiashu a yaoqiu-de tiaojian bijiao gao, shiwa,

Because the dead-GEN dependant PART require-NOM condition fairly high TAG

duifang ne you wuli-di fudan zheme zhong-de tiaojian,

the other party PART too unable afford so heavy-NOM condition

'The bereaved usually demand high compensation (for traffic accidents), but the other

party (the trouble maker) is unable to afford it.'

Here, the oe-marked trouble maker is to contrast with the a-marked the bereaved, in that

148 they are the two opposing parties of the traffic accident. In addition to entity NPs, transitional conjunctive expressions constitute another subtype of CONTRASTING topics.

This category may also be taken as a subtype of shifting, where shifting is accomplished through contrasting.

6.1.2.6 PICKING UP

PICKING UP is also a subset of the CONTINUING or shifting function, in that the

«e-marked topic "picks up" what is newly introduced in the immediate context, prototypically as the object of the predicate verb of the preceding clause (Du Bois 1987,

Chafe 1994:ch.9, among others).

(9) M^iPPr a BMmmxr, (309-X12:4)

Minzhengke ta dou you ziji-de fuligongchang.

Civil Affairs Office it all have self-GEN welfare factory

zhe fulizongchang ne, jiushi jiejue zhexie canjirenyuan de jiuyewenti,

this welfare factory PART precisely solve these disable GEN employment issue

'The Civil Affairs Offices all have their own welfare factories, and these welfare

factories aim precisely at solving the employment issue of the disabled personnel.'

The He-marked element in the second line picks up the entity newly introduced into the discourse as the object of verb you 'have' in the preceding clause (as underlined). This is in line with Langacker's 'life career' claim about topic (1991:313) or Chu's three-stage

149 process of topic formation (1998:261), in that the entity (welfare factories in this case) is first introduced in a comment, then immediately picked up as the new topic for what ensues.

6.1.2.7 LISTING

Under the functional category of LISTING, the we-marked topic expression mainly functions as a text organizer, the typical examples including shouxian 'first of all', yifangmian 'on the one hand', zaiyige 'for another', etc.

(10)£&^-ffiAB/g, fc^fiS^f^o (309-X12:l)

Zai zhe-fangmianr ne, jiushi xuyao xuexi.

In this-respect PART precisely need learn

Yige ne. jiushi wenhua-de xuexi,

one PART is culture-NOM learning

zaiyige ne, jiushi xuyao xiang lao-tongzhi xuexi

another PART is need from old-comrades learn

yixie jiushi gongzuo-shang-de hao-de jingyan, hao-de fangfa.

some PF work-on-NOM good-NOM experience good-NOM method

'In this respect, what I need is precisely learning. One is cultural learning, and the

other is to learn good experience and good working methods from my seniors.'

150 The underlined parts in this example list the two aspects of learning. Of course, a list may very well go beyond two items. The listed items may also be human entities, for instance, when talking about the family members in a one-by-one manner. When a new item is listed via topic marking, topic shifting is realized at the same time.

6.1.2.8 SUMMARIZING

By SUMMARIZING, I mean that the we-marked topic sums up previous discourse, thus a special case of the CONTINUING function, but not that of shifting.

(ii)ffiji&iS±MMM' KcM&^ffiA. tt£ft%m, bkvXmif* (127-L27:5)

Tongguo zhexie-ge wenti ne. jiu shuoming zhe fangmianr,

Through these-CL issue PART AUX show this respect

shehuizhi'an ne, bi yiqian hao.

public security PART COMP before good

'All these show that the public security now is better than before.'

The we-marked topic zhexie wenti 'these issues' sums up what has been covered in a stretch of 47 clauses.

6.1.2.9 Summary

From the above explanation and illustrations, it can be seen that the pause particle ne is not only high in frequency of use in spoken Beijing Mandarin, but also compatible with topics of varying discourse functions. A closer look at Table 6.2 reveals that the

151 frequency counts for different function types are not evenly distributed across the four topic marking positions, namely ne\, ne2, nei and «e-FiNAL. For instance, the category of

CONTINUING ranks the most frequent for ne 1 but next to the least frequent for we-FiNAL;

conversely, the category of INTRODUCING is fairly high in all four contexts. Also, PICKING

UP and SUMMARIZING are seen only in ne\ and HC-FINAL. In order to address this unevenness of the functional categories across varying contexts of use, an analysis of the semantic roles of the marked topics is in order.

6.1.3 Semantic roles of ae-marked topics

In this section, the semantic roles of He-marked topics are addressed. Table 6.3 lists the frequency counts of the semantic categories of «e-marked topics in different linguistic contexts.

152 Table 6.3 Semantic Roles of we-Marked Topics in Different Contexts

Meaning ne\ nel neh «e-FINAL total ENTITY 304 23 1 193 527 TIME 287 81 6 3 377 CONJUNCTIVE 157 1 158 TEXTUAL 60 3 2 65 RESPECT 45 9 8 62 MANNER 34 16 1 51 EVENT 39 2 2 4 47 PLACE 28 9 2 4 43 ACTION/ACTIVITY 8 2 2 22 34 INTERPERSONAL 15 7 22 QUESTION 14 14 MEANS 3 1 2 6 PERSPECTIVE 3 2 5 CIRCUMSTANCE 3 3 BASE 2 2 CONDITION 1 1 2 PHENOMENON 1 1 total 1003 157 21 238 1419

From Table 6.3, it can be seen that the semantic role of ENTITY with a total of 527 instances is by far the most frequent among all meaning categories. This is not unexpected though, especially given the fact that an entity is more central to a designated predication, be it an agent, a patient or a theme. The predominant use of ENTITY topics in the column of we-FiNAL (193 out of a total of 238, or 81.1%) can be accounted for in a similar vein. Since this position immediately precedes the predicate, it is the default position for an argument role. It should be noted that ENTITY is used here to denote concrete objects that are discrete and prominent for conceptualization.

47 Entity in Cognitive Grammar refers to "anything that may be conceived of or referred to for analytical purposes; e.g. a thing, relation, sensation, or point on a scale. An entity need not be discrete, separately recognized, or cognitively salient" (Langacker 1987:489). In this study, however, entity is mainly used to refer to concrete objects, or "discrete" things in Langacker's terminology.

153 The second most frequently occurring category is TIME, totaled into 377 instances. Its distribution, however, is different from that of ENTITY, in that marked time expressions are mostly observed in ne\ and ne2, 287 and 81 instances respectively. The proportion of

TIME topics in ne3 is also relatively high (6 over 21). The only exception is WC-FINAL.

This phenomenon is actually attributable to the semantic role of TIME being circumstantial: if a circumstantial role is prototypically optional, it may be omitted or occur in a peripheral position, without giving rise to ungrammaticality, latency or a change in the meaning of the verb. How topic is identified for this particular research (i.e., any pause-marked element(s) preceding the predicate) also predetermines this distribution pattern. Like TIME, the categories of PLACE, MANNER and MEANS also play a circumstantial role and tend to occur early in a topic serial, as shown in Table 6.3.

However, these elements, unlike CONJUNCTIVES to be discussed below, do not rule out the possibility of being predicated upon. Therefore, they are also seen to occur in the FINAL topic position, though very infrequently (once for MANNER, four times for PLACE, none for MEANS or CIRCUMSTANCE).

The next semantic category in order of frequency count is CONJUNCTIVE. Except for one case,48 CONJUNCTIVE topics occur exclusively as the first topic of a TC construction.

This is predictable from the nature of conjunctives: conjunctives are words and

48 This case turns out to be preceded by the question weishenme ne? 'why?'. This example together with the 14 instances under the category of QUESTION is typical of the speech style of a small number of the language informants. Though a question in form, such utterances function only as prompters for ensuing discourse, which explains why these expressions are taken in this dissertation as question-form topics, rather than independent clauses. This phenomenon is reminiscent of Chao's (1968) interpretation of subject and predicate as question and answer. Except for this special case, all conjunctive expressions occur clause initially.

154 expressions which join words, phrases, or clauses together (Richards, Piatt and Piatt

2000). Because the analysis in this research is at the clause level, the conjunctives are to link clauses together. According to Li and Thompson, "adverbial backward-linking elements must occur at the beginning of their clause" (1981:653). Since most of the conjunctives are adverbial in nature, the clause-initial position seems to be their default slot. Cognitively, it would be more realistic to orient the listener on how the ensuing proposition is to be related to the preceding one through the early use of a conjunctive expression than to postpone it until after the statement is presented.

The next category, TEXTUAL topic, is external to the clausal proposition by conveying discourse organizing information. Also somewhat external to the clausal message per se is the INTERPERSONAL category, which expresses the speaker's judgment regarding the relevance of the message to the discourse context. The RESPECT category, as the term indicates, specifies the respect of an object or states of affairs that is to be commented on.

Likewise, other categories, such as EVENT, PERSPECTIVE, CONDITION, etc. all give prominence to a certain element of the clausal process.

6.1.3.1 Summary

This section has presented the semantic roles of we-marked topics and the distribution tendency of different semantic roles in certain topic positions. Since some topics seem to have default positions of occurrence, it would be interesting to address the issue of topic ordering in multiple topic constructions.

155 6.1.4 Topic ordering in multiple topic constructions

According to Zhang and Fang, the order of multiple topics is in general from textual topic to interpersonal topic and to experiential topic in spoken narratives (1996:26), which agrees with Halliday's proposal (1985, 1994). When the present spoken corpus is concerned, the general trend seems to be verified, though Table 6.3 shows three textual elements marked by ne2 and two marked by we-FlNAL. A closer look reveals that the marked topics preceding these textual elements are also textual in metafunction; that is, in these clauses, the textual element has been instantiated more than once.

(12) RTJiBIg, %U±M, &^iJ, ^Pi? (015-025:1)

Keshi ne. shiiishang ne, da-bu-dao, shiwa?

But PART in fact PART reach-not-reach TAG

'But in reality, (the standard) can't be reached.'

This example begins with the CONJUNCTIVE expression keshi 'but', an instance of

Halliday's or Zhang and Fang's textual theme/topic, which is then followed by the interpersonal element shijishang 'in fact', with the experiential element (a certain standard) being retrievable from previous discourse context. Sometimes, the first textual element may take the form of a time adverbial, which, upon closer examination, functions as a text organizer, rather than an indicator of the chronological ordering of the event, as in (13).

156 JSJ&ff'Cs JiPio (127-L27:6)

Ranhou ne. you, iushuo ne. Lugouqiao zheng,

Then PART FS it's said PART Lugouqiao Bridge FS

zhunbei shishuo zuowei yihou shishuo Kangrizhanzheng de

prepare so to speak serve as later so to speak Anti-Jananese War NOM

zhege quanguo zhanlan zhongxin, shiwa.

PF national exhibition center TAG

'Then it's said that the Lugaoqiao Bridge is designed to become a national exhibition

center for the commemoration of the Anti-Japanese War.'

The first underlined element is often used to show the chronological order of an event. In this particular context, however, it is a near synonym to 'in addition', functioning to relate the ensuing discourse to what precedes.

Nevertheless, the present corpus does show some complications to the ordering of the topics fulfilling the three metafunctions, as (14) to (16) show.

(Umitm, %*r—^RE, MMT&tti%M%, n^M, n^&M$k, (274-T30:3)

Wo juezhe. lingwai yige ne. jiushi shuo de zhe shucai,

I feel other one PART is speak NOM this vegetable

meinian de, meinian zhe shucai,

every year NOM every year this vegetable

'I feel that the other is the vegetables, the vegetables of each year.'

(15)MMM> &%m, fiz±E, m&%M:?tfimkmMt/A-gJlo (057-H20:5)

157 Jiehun wan de. ranhou ne. zai yige ne.

Marry late NOM then PART still one-CLPART

jiushi shenghaizi ye dou bijiao na-shenme yidianr.

that is give birth to a child too all comparatively that-what a little

'The late married, then, for another thing, are all comparatively late in giving birth to

a child.'

j@[. (141-X03:14)

Yudao-le,49 guanyu duidai kunnan de wenti ne.

Meet ASP as for treat difficulty NOM issue PART

wo iuede ne, yeshi you liangzhong, zhege genben de taidu wenti.

I think PART too have two-CL PF prime NOM attitude issue

'As for how to treat difficulties, I think, there are also two prime attitudes.'

In (14), the first element wojuezhe 'I feel' constitutes an interpersonal topic. The second one, however, is more textual than experiential, because the speaker is listing the things whose prices are on the increase, with the previous discourse about fish and meat.

Therefore, in this example, the interpersonal element precedes the textual. (15) puts the experiential element jiehun wande 'the late married' before the textual ranhou 'then'50 and zaiyige 'still another'. Again we see double use of the textual element in one single clause. (16) resembles (15) in that the experiential topic (i.e., guanyu duidai kunnan de

49 This is actually a false start of speech. 50 This expression is used in the same manner as that in (13).

158 wenti 'how to treat difficulties') precedes the interpersonal (yvojuede 'I think'). However, it also differs from (15), given that the clause-initial element is marked by the preposition guanyu 'as for', instead of an agentive entity. In any case, these three examples have shown that the ordering of multiple topics is not as fixed in spontaneous speech as in the written variety. Perhaps the double use of textual elements is a feature particular to spontaneous speech.

6.1.5 Summary

To sum up, ne is the most frequently used topic-marking pause particle in Beijing

Mandarin, and distributes extensively across all topic-marking positions, namely from the very first topic element to the element immediately preceding the predicate of the clause.

Also, it can mark topics of many different functional types, and the marked element can be very diverse in semantic role. It seems that ne is a versatile topic marker in Beijing

Mandarin, and it may occur either at the beginning (e.g. when introducing or shifting the topic) or towards the end (e.g. when in contrastive use) of a discourse stretch as produced by native speakers of Beijing Mandarin. This finding, however, is somewhat different from what some previous studies, such as Zhang and Fang (1996) and Yuan (2002), have concluded on the use of we, i.e., that it is a phenomenon in the continuing rather than the opening clause of a discourse stretch.

The following section presents what has been found about particle ba in the present corpus.

159 6.2 Pause Particle ba

6.2.1 Frequency and context of occurrence

On the basis of the corpus analysis, ba is the second most frequently used pause particle to mark a topic expression in Beijing Mandarin. As shown in Table 6.4, there are

196 instances of ba marking the first topic of a clause, 54 the second, 13 the third and 82 the FINAL topic or the topic immediately preceding the predicate. In total, 345 6a-marked topics are observed. In distribution, ba, like ne, also marks topics in various positions.

Given the fact that more 6a-marked TC clauses are bounded by periods to the left, the analysis results seem to confirm Zhang and Fang (1996) and Yuan's (2002) statement that ba is typically a phenomenon at the opening of a stretch of utterance. The same trend is further evidenced by fewer occurrences of periods at the right boundary of the clause in which the particle ba occurs. The predominant use of 6a-marked TC clauses as paragraph openers (26 instances vs. nine instances occurring at the end of a paragraph) and the exclusive use of the construction under scrutiny as monologue openers (5 instances) offer more support for the same conclusion that 6a-marked topics are prototypically observable at the early stage of information conveyance. This finding is also in agreement with the fundamental function of topic, i.e., to serve as the starting point for ensuing discourse.

160 Table 6.4 Linguistic Contexts of Use of Particle ba within the Scope of the Clause

Boundary bal bal bai 6<*-FINAL marker LBM RBM LBM RBM LBM RBM LBM RBM comma 39 78 5 18 1 3 11 36 period 131 113 47 34 10 8 64 42 question 5 3 1 1 mark paragraph 18 2 1 2 2 2 5 3 boundary monologue 3 1 1 boundary total 196 196 54 54 13 13 82 82

6.2.2 Discourse functions and semantic meanings of £a-marked topics

Functionally, ba has its own features. What distinguishes ba from ne is the rare occurrences of the CONTRASTING category and the relatively high frequency count for the

ILLUSTRATING category (see Table 6.5).

Table 6.5 Discourse Functions of ba in Different Contexts

Function bal bal bai 6a-FINAL total INTRODUCING 85 13 3 36 137 REPEATING 38 21 7 31 97 ILLUSTRATING 37 4 9 50 CONTINUING 23 14 1 38 PICKING UP 9 1 2 12 PROMPTING 1 1 1 2 5 CONTRASTING 3 3 SUPPOSING 2 2 SUMMARIZING 1 1 total 196 54 13 82 345

6.2.2.1 CONTRASTING ba

As discussed in Section 6.1.2, CONTRASTING is a basic function of we-marked topics,

161 and a total of 150 instances cross-cutting three different positions fall under this category.

When it comes to ba, only three 6a-marked topics are found to contrast with preceding discourse, and they uniformly occur in the first topic position, as illustrated in (17).

(17)$P4^AHE, Wftj83fffiA&4>7&IEJl, '(190-D16:4)

Rujin-de ren ba. guan de zhe-fangmian ren

Today-NOM people PART take care of NOM this-respect people

jiu shao-le fanzhengshi,

thus few-ASP nonetheless

'Nonetheless, there're fewer people nowadays to care for things in this respect.'

This clause occurs in a context where the speaker is contrasting people of 'nowadays' with those about ten years earlier. Before this clause, there are six clauses referring to people in old times. The 6a-marked topic expression takes the clause-initial position, and no anomaly would result if this ba were replaced by ne, as in (17'), given the fact that ne is more typical of marking CONTRASTING topics than ba.

(17') ruiin de ren ne, guan de zhe-fangmian ren jiu shao le fanzhengshi,

The shuffling use of ba and ne can be illustrated with the following excerpt from a monologue.

(wains jguEPE, &JE, mmnxftMHwim, &^m&bmm, Q9O-DI6:5)

Nage, vuanlai ba, fanzheng, wo canjia gongzuo de shihour ba.

PF previously PART in any case I join work NOM time PART

zhege huanjing ye bijiao youyue,

162 PF environment too fairly good

'Previously when I took up my first job, the environment was fairly good.'

Zhege nage zai jiguan libianr ba. fanzheng tiaojian ye bijiao hao.

PF PF in office inside PART anyway condition too fairly good

'In the office, the working conditions were fairly good.'

Ng, houlai__ne, jiu women jiu dao nage Wuqi Ganxiao Hubei,

PF then PART then we then to that May 7 Cadre School Hubei

Dao nar qu nage dao ganxiao qu-le.

To there go PF to cadre school go-ASP

'Later on, we went to the May 7 Cadre School in Hubei Province.'

Dao Ganxiao qu de shihour ne, nahuir dangshi ba, fanzheng ziji ba.

To Cadre School go NOM time PART that time then PART anyhow self PART

ng, fanzheng nahuir jiu juede haoxiang nianqing ma,

PF anyhow that time just feel as if young SFP

'At the time when we were going to the Cadre School, in any case, I just felt that I was young,'

163 yinggai duanlian, yinggai dedao zhege sixiang-shang he shenti-shang de duanlian.

should temper should get PF mind-on and body-on NOM temper

'and that 1 should be tempered mentally and physically.'

suoyi nahuir ba, hen jiji-di yaoqiu qu le.

so that time PART very actively-AUX ask go ASP

'So then I actively asked to go.'

Erqie ne. dangshi, dangshi nahuir ba.

Also PART then then that time PART

women nage yewubanzi ba, lingdaojiu mei-shuo rang wo qu,

our that leading group PART leader AUX not-speak let me go

'What's more, at that time, the leading group didn't ask me to go.'

jiushishuo ba, zhunbei gei wo liuzai, liuzai jiguan litou jixu gongzuo.

that is PART prepare GEI51 me keep REP office inside continue work

'That is to say, they prepared to keep me in the office and let me go on with my

work.'

Danshi ne, nahuir jiu jiang, hai, fanzheng nianqing-ren,

51 Gei is an object-preposing preposition. Then PART that time just say interj. anyhow young-people

yinggai duanlian-duanlian qu,

should temper-REDUP go

'But it was preached at that time that young people should go and get tempered.'

suoyi ba, jiushi yaoqiu qu le.

so PART just ask go ASP

'Therefore, I asked to go.'

There are altogether 10 clauses in this example, where ten ba and four ne particles are

used to mark different topics. In the first clause, ba is used twice to introduce the time of

an event, and the two 6a-marked topics are repetitive, a typical feature of spontaneous

speech. The third ba occurs in the second clause, marking the expression which

elaborates the working environment as first mentioned in the previous clause. Ne first appears in the third clause, marking the time adverbial houlai 'then' and moving the discourse on to a new stage of development. In the following clause, ne repeats its function by marking a more specific time than 'then', with the ensuing £>a-marked element referring back to the time. The second ba in the same line marks a situationally and textually available referent—the speaker herself. In the sixth line, ba is used again to repeat the time. In the seventh line, the textual element erqie 'also' is marked by ne, again moving the discourse forward. Later on in this clause, the same time adverbial is marked by ba. The second ba in the same clause, in contrast, reintroduces an agentive entity, i.e.,

165 'the leading group'. The ensuing clause reiterates the proposition of the seventh line, where ba is used again to mark a textual topic. The second last line begins with the contrastive conjunctive expression danshi 'but', expectedly marked by ne. The last line sums up the previous discourse by marking the resultative conjunction suoyi 'so' with ba.

This particle-by-particle analysis highlights the functional opposition between the two particles: while ne more often than not collocates with contrasting elements, ba tends to occur in topic repeating or reestablishing contexts. In this sense, these two particles do complement each other in discourse use.

6.2.2.2 ILLUSTRATING ba

In performing the ILLUSTRATING function, unlike ne, which is more often than not used as an independent topic marker, ba is prototypically coupled with lexical means such as bifangshuo 'for instance', (ni)xiang 'for example', nikan 'such as/like', biru

'such as/like', haobi 'such as/like', na...laishuo 'take...for example', etc. Tao

(1996:166-8) calls these attention-getting expressions, which, I think, is in agreement with the prime function of topic marking, namely to emphasize the topic status of the marked expression. When ba is used for the purpose of illustration, it is observed to follow topics in different positions, as illustrated in (19) and (20).

(19){f!i^N%WPE, ^&M;g», m, IlWtfi^t-. (288-N40:2)

Xiang zhege jiaoyu ba, xuexiao de laoshi he, ng,

Like PF education PART school GEN teacher and PF

166 jiating de jiaoyu ne bu-tongyi. (db3059)

home NOM education PART not-integrated

'Take education for example, school education and home education are not

integrated.'

The clause-initial element specifies the topic of the ensuing comment, and the topic expression is framed by preposition xiang 'like' and particle ba.

-fc+, -fc-h -fc+£#l£e£;&. (197-X22:2)

Dangshj.....de...gpngzi....ne, nage, xiang lao-jiaoshi ba.

Then NOM salary PART PF like old-teacher PART

you si-wushi sui de lao-jiaoshi ba,

have four-fifty age NOM old-teacher PART

gongzi yiban zai qi qishi, qishi, qishi duo kuai qian zuoyou.

salary generally at FS seventy REP REP more CL money or so

'At that time, when salary is concerned, take experienced teachers of forty to fifty

years of age for example, their salary was about seventy yuan.'

This clause as a whole is similar to the LD-subject construction, in that the sentence-initial element (underlined with the dotted line) is repeated later on as the functional subject of the clause, but not in the form of a pronoun. In between are two ba-marked topic expressions repeating and reinforcing each other. The first one is complemented with xiang 'like', highlighting the ILLUSTRATING function of the marked

167 topic. This example also shows that ba can be used to mark a clause-middle topic in a consecutive manner.

6.2.2.3 SUPPOSING ba

If Table 6.5 is compared with Table 6.2, it can be seen that a new category, that of

SUPPOSING, appears in Table 6.5. This discourse function, I think, can be taken as a carryover from the "suppositions as alternatives" function of the sentence final particle ba as discussed by Chao (1968:807), who uses the following sentence for illustration:

Bu-gei qian ba, buhaoyisi bana; gei qian ba. you gei-bu-qi.

Not-give money PART bashful take for nothing give money PART but can't afford

'Suppose I don't pay for it, 1 am ashamed to take something for nothing; and if I am

to pay for it, I can't afford it.'

When ba is used in the middle of a clause, the Z>a-marked element tends to indicate some kind of action or activity, also a carryover feature of ba as an SFP. What distinguishes the topic marker ba from the SFP ba, as shown by the following two instantiations in the data, is the marked element constituting a clausal constituent, rather than standing alone as a clause. mm&^n^mm&Mm, &fg.v&7WLm, (302-149:6)

Xianzai dengvu ni yao zai zuzhi jinlai ba, yijing wan le jiushi,

Now equal you will again organize purchase PART already late ASP namely

168 'Making more purchases now is already too late.'

mmitm^, m^mm±££xm®&U7fm« (267-H24:5)

Ranhou women, ng, erqie nage iiushi bu, bu-kongzhi renkou ba.

Then we PF also PF namely FS not-control population PART

jiushi dui guojia nage jiushi, ng, gao sihua nage,

PF to country PF PF PF build four modernizations PF

zhegejinxing shehuizhuyi jianshe ye hen buli.

PF undergo socialism construction too very harmful

'Then not controling the population will be detrimental to the four modernizations in

our country, and to socialism construction as well.'

In both examples, the ba-marked phrases denote some action or activity, but unlike conditional clauses, they both serve as the grammatical subject of the remaining part of the sentence. That is, what is late in (21) is the supposed action of making new purchase, and the detrimental thing in (22) is not to control population. Chao's example sentence, in comparison, is very different. The grammatical subject of the two clauses following the two 6a-closed conditionals in Chao's example can by no means be what is denoted by the

^/-closed conditional clause; it has to be a situationally available or contextually retrievable referent, most probably the speaker himself. Therefore, constructions with clausal topics, such as (21) and (22), are not to be equated with complex sentences with conditional clauses.

169 6.2.2.4 Summary

From the discussions above, it can be seen that ba functions differently from ne in discourse in Beijing Mandarin. While ne is versatile in topic marking, ba is rarely used for CONTRASTING purposes. Also, ba tends to co-occur with some lexical expressions in fulfilling the ILLUSTRATING discourse function.

6.2.3 Semantic roles of Zw-marked topics

When it comes to the semantic meaning of Z>a-marked topics, many different categories are identified (see Table 6.6). Both functional and circumstantial roles are found to be marked by ba. Like ne, ba is also found to mark ENTITY topics (165 cases) and TIME adverbials (89 cases) most frequently in Beijing Mandarin. Likewise, ba can also mark interpersonal (as in (23)) and textual elements (as in (24)). nimmmmxmmM, mmji&ikTfeji&m, a83-x46:4)

Ju ta liiyou de ren huilai shuo ba.

According to he travel NOM people come back say PART

dao nar shi bi ganhuor hai-lei.

get there is COMP work still-tiring

'According to those coming back from a traveling trip, it's even more tiring traveling

than working.'

(24)MJJB>N &K)3fcTftHE, ^SiP>h#;frffi1ft&aE£*fgfi

Erqie nage, zongdelaishuo ba.

170 Moreover PF generally speaking PART

jiali nage ge-fangmian qingkuang haishi bucuo de.

family PF each-respect situation still fairly good CSC

'Moreover, generally speaking, my family is fairly good in every respect.'

In (23), the INTERPERSONAL topic is marked by ba. In (24), the fo-marked element is

textual in discourse function, in that it sums up the previous discourse and closes the

paragraph in which it occurs.

Table 6.6 Semantic Roles of Aa-Marked Topics in Different Contexts

Meaning ba 1 bal ba3 6fl-FINAL total ENTITY 84 11 2 68 165 TIME 57 25 4 3 89 ACTION/ACTIVITY 10 2 6 18 PLACE 12 3 3 18 RESPECT 11 2 1 14 EVENT 8 2 1 1 12 INTERPERSONAL 4 3 7 MANNER 1 5 1 7 TARGET 3 1 4 MEANS 1 2 3 CONJUNCTIVE 2 2 BASE 1 1 2 CIRCUMSTANCE 1 1 2 TEXTUAL 1 1 2 total 196 54 13 82 345

6.2.4 Summary

To recapitulate, ba is the second most frequently occurring pause particle in Beijing

Mandarin. Although it can be used to mark topic expressions of varying discourse functions, it shows a tendency to complement particle ne in distribution, in that ba is

171 typically an early discourse phenomenon while ne tends to mark a late occurring topic.

On account of this distribution, ba is rarely used to mark a contrastive topic; instead, it is mainly used for illustration and topic introducing or reintroducing purposes. It has also carried over its "suppositions as alternatives" function as an SFP, with thus marked topics to be predicated by the ensuing comments. Finally, the semantic roles of 6o-marked topics are also very varied in naturally occurring discourse.

6.3 Pause particle a and its phonetic variants

According to Zhang and Fang (1996), particle a has five phonetic variants in addition to ba,52 namely la, na, nga, wa, and ya. Though nga is a phonetic possibility, its corresponding written form does not exist. Therefore, this section examines the discourse use of a, la, na, wa and ya.

6.3.1 Frequency counts and contexts of occurrence

Table 6.7 below shows the frequency counts of each particle in the four different contexts as identified for ne and ba. From the table, it can be seen that a and ya are relatively high in frequency of use, 177 and 151 instances respectively, with a being more extensively distributed across contexts. Of a total of 56 possible contexts where a particle may appear, as indicated by the shaded fields in the table, a appears in 30 such contexts,

52 This sampled corpus does not show any consistent difference in the use of a and ba among language informants of different age groups. Rather, the preference for a or ba is more an idiosyncratic feature than a social variance as claimed by Zhang and Fang (1996:ch.4). Since ba occurs much more frequently than others in this study, it has been presented separately in Section 6.2.

172 while ya appears in 22. In contrast, la and wa are not so frequently observed in the corpus, with the former appearing 12 times and the latter 21 times. In any case, these two particles show a common feature of distribution, namely, they are rarely used to mark the

FINAL topic (once only for each). Since the FINAL topic position is typical of core argument roles, the inference seems to be that la and wa, unlike other particles, are probably not good candidates to mark topics fulfilling the grammatical subject or object role. The other particle, na, stands in between the two pairs.

173 Table 6.7 Linguistic Contexts of Use of Particle a and its Phonetic Variants

within the Scope of the Clause Boundary Particle 1 Particle 2 Particle 3 Partic e-FINAL particle marker LBM RBM LBM RBM LBM RBM LBM RBM a 22 32 8 13 1 3 12 14 la 3 2 1 lf-- 1 comma na 6 7 1 1 2 ,4 wa 5 5 \ • ya 27 26 5 to 3 4 . 13 a 44 51 16 , 13 7 4 37 41 la 5 6 2 2 1 period na 19 20 9 7 wa 10 12 1 1 1 1 1 1 ya 47 59 14 9 4 -2 35 24 a 2 1 3 1 la question na - mark 1 wa 1 ya 2 1 " a .1 1 ~ la exclamation mark na wa ya a 11 X 1 1 4 • \ la paragraph na 2 boundary 1 1 1 wa 3 ya 9 1 1 a a 4 1 1 la monologue na boundary wa ya 2 I a 1 la not clear53 na wa ya a 85 85 27 27 8 8 57 57 la 8 8 3 3 1 1 total na 28 28 0 0 1 1 12 12 wa 18 18 1 1 1 1 1 1 ya 87 87 19 19 5 5 40 40

174 6.3.2 Discourse functions of topics marked by a and its phonetic variants

The discourse functions of a and its phonetic variants are presented in Table 6.8.

There are at least two features characterizing this table. First, the CONTRASTING discourse function is very infrequently performed by this set of pause particles. Of the five particles, only one instance of a and one instance of ya is used to mark a contrastive topic. Instead, this set of particles functions predominantly to mark topics that are illustrative. As shown in the table, the instantiations of ILLUSTRATING topics outnumber discourse CONTINUING topics, ranking below the INTRODUCING and REPEATING categories only. If Table 6.8 is compared to Table 6.5, it is not difficult to see that particle ba as a topic marker also exhibits these two features. This finding seems to support the claim that ba is also a phonetic variant of a. The functional opposition between this set of pause particles and ne offers further support for taking ba, a, la, na, wa and ya as one single group. It is just on account of a much higher frequency of occurrence that ba is presented separately in this dissertation.

175 Table 6.8 Discourse Functions of a and its Phonetic Variants in Different Contexts Function particle Particle 1 Particle 2 Particle 3 Particle-FINAL total a 7 4 1 12 la 1 1 CONTINUING 54 na 10 10 (36) wa 4 1 5 ya 6 2 8 a 1 1 la 0 CONTRASTING na 0 (2) wa 0 ya 1 1 a 8 2 10 la 6 3 9 ILLUSTRATING (48) na 1 1 wa 3 3 ya 18 5 2 25 a 49 15 5 29 98 la 1 1 2 INTRODUCING (190) na 10 1 6 17 wa 8 1 9 ya 39 5 1 19 64 a 4 2 6 la PICKING UP (13) na 2 2 wa 1 1 ya 4 4 a 1 la PROMPTING na (1) wa ya 1 a 16 6 2 26 50 na 5 6 11 REPEATING (112) wa 2 1 3 ya 18 7 2 21 48 a 85 27 8 57 177 la 8 3 1 12 Total na 28 0 1 12 41 (402) wa 18 1 1 1 21 ya 87 19 5 40 151

Of the particles listed in Table 6.8, la and ya seem to be the best candidates for marking ILLUSTRATING topics. Of altogether 12 uses of la, nine have the ILLUSTRATING function. Though the frequency count of ILLUSTRATING ya is not as high in percentage as la, its absolute number of occurrences is larger (25 times altogether). Meanwhile, the

other three are also seen to perform this function. In this manner, the ILLUSTRATING

function can be taken as a cross-cutting function for particle a and its phonetic variants,

ba included.

Somewhat incongruent with the general picture is the prevalent use of na for the

CONTINUING discourse function. As shown in Table 6.8, na is not a very frequently used topic marker, but ten instantiations are used to mark CONTINUING topics, outnumbering all others, including the more frequently used a and ya overall. Below are three instances of na marking CONTINUING topics of varying types.

(25)Jtfc^fpj+Ao (204-L53:4)

Houlai yi-nian na dao shiba.

Later one year PART reach eighteen

'One year later, my salary reached eighteen yuan.'

(26)MMS> Ml, ^TOIfggene, &IEo (057-H20:2)

Suoyi na, wojiu ng, jinliang-di jiancha ziji ba,55 fanzheng.

So PART I just PF try one's best-AUX criticise self SFP nonetheless

'So I just tried to do self criticism, so to speak.'

(27)%3fHBt.gJiWftJ*tk;fr;L, &%&, Wmm* (015-D25:7)

Wo kan na. haishi youde difangr, haishi zuo, zuo-de-dao de.

1 see PART still some place still FS do-AUX-reach CSC

'In my opinion, some places were able to manage it.'

177 The wo-marked element in (25) is a time adverbial, that in (26) a conjunction (i.e., a textual element), and that in (27) an interpersonal expression, all falling under the category of CONTINUING topics. The difference between CONTINUING na and ne is perhaps better seen when a substitution is done, as in (25'), (26') and (27').

(25') houlai yinian ne dao shiba.

(26') suoyi ne wo jiu ng, jinliangdi jiancha ziji ba, fanzheng.

(27') wo kan ne, haishi youde difangr, haishi zuo, zuo-de-dao de.

Intuitively, these substitutions are all acceptable; however, there would be a slight change in the intonation contour: Me-marked topic expressions tend to be produced in a higher contour than no-marked ones (Shi 1995:116). In fact, I doubt whether (25') and (26') can take a falling contour, though the falling contour is compatible with na. This observation seems to verify Chao's observation of a as "pause for hearer", namely, "(T)he deliberate pause is to give the hearer time to let what is said sink in, as distinguished from ... ne to bring up specific points for both speaker and hearer to consider" (Chao 1968:806). A rising contour is in general more inviting than a falling one. On the other hand, a falling contour may indicate that the speaker is more interested in performing the narrating task without much regard to the hearer, though he or she also attends to the hearer's needs for time to process what has been uttered.

The use of na as speaker's self-containment is also reflected in the relative high frequency count of na to mark interpersonal topics. Of the 28 instances of no occurring as the first topic marker, three, or more than one-tenth, mark interpersonal topics. Since the

178 interpersonal element is basically the speaker's viewpoint on the statement, most probably no participation from the hearer is anticipated. In addition, a and ya are also found to mark interpersonal topics. As for the fact that no interpersonal topics are followed by la or na, the low frequency of the overall use of these two particles may pose as a ready explanation.

6.3.2.1 Summary

The discourse functions of the topics marked by a and its phonetic variants are quite similar to those of 60-marked topics, which seems to support the claim that ba is also a phonetic variant of a.

6.3.3 Semantic roles of topics marked by a and its phonetic variants

The semantic role analysis of the topics marked by this set of pause particles also reveals that ENTITY is by far the most frequently occurring category (see Tables 6.9 to

6.13). Altogether, 93 ENTITY topics are marked by a, five by la, 24 by na, eight by wa, and 85 by ya. ENTITY is also most frequent for ne- and io-marked topics. Similarly, the circumstantial role categories, such as TIME, PLACE, etc. are also found to be instantiated by almost all pause particles. This is not difficult to explain though. Given the narrative nature of the monologues, participants and settings are indispensable components. For reasons of discourse organizing and processing necessities, they may take marking in linguistic coding. Since the entities are prototype figures, that is, the most prominent

179 components of a construed situation, they may typically be seen by speakers as deserving of more attention, and thus iconically occur in the marked form.

Table 6.9 Semantic Roles of a-Marked Topics in Different Contexts

Meaning a\ al a3 a-FINAL total ENTITY 46 4 43 93 TIME 12 16 1 1 30 ACTION/ACTIVITY 10 1 8 19 RESPECT 8 1 3 2 14 PLACE 3 3 2 1 9 INTERPERSONAL 2 1 1 4 ISSUE 3 3 PHENOMENON 2 2 PERSPECTIVE 1 1 SUPPOSITION 1 1 TEXTUAL 1 1 total 85 27 8 57 177

Table 6.10 Semantic Roles of /a-Marked Topics in Different Contexts

Meaning la\ lal lab /fl-FINAL total ENTITY 3 1 1 5 ACTION/ACTIVITY 2 1 3 PLACE 1 1 2 INTERPERSONAL 1 1 PROPOSITION 1 1 total 8 3 0 1 12

Table 6.11 Semantic Roles of «a-Marked Topics in Different Contexts

Meaning na\ ndl nab WO-FINAL total ENTITY 12 1 11 24 TIME 9 9 INTERPERSONAL 3 3 ACTION/ACTIVITY 1 1 2 RESPECT 2 2 CONJUNCTIVE 1 1 total 28 0 1 12 41

180 Table 6.12 Semantic Roles of wo-Marked Topics in Different Contexts

Meaning wa\ wdl wdi Wfl-FINAL total ENTITY 1 1 8 TIME 7 1 8 PLACE 1 2 ACTION/ACTIVITY 1 RESPECT 1 MEANS 1 total 18 1 1 1 21

Table 6.13 Semantic Roles ofyo-Marked Topics in Different Contexts

Meaning ya\ yal yah Jfl-FINAL total ENTITY 49 4 1 31 85 TIME 17 6 1 24 ACTION/ACTIVITY 8 1 1 7 17 ISSUE 3 2 2 7 PLACE 2 3 2 7 RESPECT 5 1 6 MEANS 1 1 2 INTERPERSONAL 1 1 PERSPECTIVE 1 1 TARGET 1 1 total 87 19 5 40 151

The frequent use and extensive distribution of this set of particles may pose a challenge to Zhang and Fang's categorization of ya and la as mood-indicating particles, but not as topic markers (1996:49). Their major argument is that these two particles are very restricted in distribution, namely that they are more often than not used to follow juxtaposed items. The above analysis has shown otherwise. The two particles in question, ya in particular, have been observed to fulfill various discourse functions by marking topics of diverse semantic roles. Although la is predominantly used for illustrating purposes (nine out of 12 instances), it is also seen to introduce subject or object entities without evoking juxtaposition. (28) is an example of la marking the first topic, and (29) the FINAL topic.

(28)£»%'«PS, ^M&mmAm, (029-X36:4)

Hui xiyan-de la, you yan wo jiu zhaodai renjia,

Able smoke-NOM PART have tobacco I AUX treat them

'If the visitor smokes, I'll treat him to a cigarette if I have.' (29)@^ft-^, mrxm^m, wz^Mmm, Mra*-^ (253-154:4)

Yinwei shenme,56 zhe guangda chengke la.

Because what this numerous passenger PART

dui zhe shoupiaoyuan ya, yijian bijiao da yixie.

to this bus conductor PART complaint rather big some

'The reason is that a great number of passengers have a lot of complaints about the

bus conductor.'

(28) is coded as an LD-object construction, where the clause-final pronoun renjia 'them' refers back to the la-marked referring expression. (29) is a multiple topic construction: the conjunctive expression yinwei 'because' is marked by a pause (as indicated by the comma) at the clause-initial position and the topic marked by la takes the second position.

Although the following element is also marked (by ya), it is not taken as a topic on account of the fact that it occurs after the grammatical subject of the predicate, i.e., the fa-marked element.57

182 Zhang and Fang (1996:49) also argue that the ILLUSTRATING function is not performed by the pause particle la or ya; instead, the illustrating reading is the outcome of the sentence pattern. Therefore, the existence of the pause particle is just to express the mood of the speaker. This stance of Zhang and Fang is reminiscent of Li and Thompson's discussion of Chao's (1968) ten different meanings of a/ya. According to Li and

Thompson, only some of the ten meanings concern a/ya as sentence-final particle, while

"most of those meanings can be understood on the basis of the meaning of the sentence to which a/ya is attached. In other words, most of those various meanings should not be attributed to a/ya" (1981:314). The question then is: do particles have a meaning independent of the expression to which they are attached? It appears that a great number of, if not all, particles, ne included, could be omitted without affecting the grammaticality or acceptability of the clause.58 Conversely, the pause particles examined so far are all polysemous, at least in that they can co-occur with topics of varying discourse functions.

In this spirit, I think that Xu and Liu's interpretation of the function of topic marking is more realistic. According to Xu and Liu (1998:219-220), pause particles have a double role to play in topic-prominent languages like Mandarin Chinese. First, they are used to prelude more important information coming up in the comment; second, they function as topic introducers, though the introduced topic may not be brand new information. Xu and

Liu then summarize the double function in one: pause particles are to enhance the topichood of an expression, and different particles tend to go with topics of different information status. This is in fact the reason why this dissertation has been avoiding

183 equating the semantic role of the topic expression with that of the topic-marking particle, despite the fact that the particle forms an indispensable whole together with what it marks.

The decision to treat the semantic role of the topic expression and the topic marker separately has its theoretical advantages. First, this research identifies topic mainly on the basis of its coding form (marking in particular) and its relative position to the predicate expression within the scope of the clause. Therefore, it is a theoretical requirement to take all qualifying pause particles as topic markers. This approach to topic identification turns out verifiable with naturally occurring language data. An example with particle la illustrates this point.

(30)BR, ffift-^Pft, $fm, &M«^o (050-N14:2)

A, xiang shenme yu la. xia la, zhe wo genben jiu bu-mai.

PFlike what fish PART shrimp PART this I at all AUX not-buy

'Things like fish and shrimps, I don't buy them at all.'

In (30), it is quite obvious that la is used for illustrating purposes. The use of xiang 'like' makes the purpose explicit, while shenme 'what', an indicator of an incomplete list, also occurs before the two la-marked phrases. Meanwhile, the entire utterance constitutes a typical TC construction, i.e., an LD-object construction, where zhe 'this' which follows the pause refers back to the two coordinate /o-marked expressions. The point is that the illustrating function of la does not necessarily go against the topic-marking function.

Rather, the two may be integrated in appropriate linguistic contexts (before the predicate

184 in this example). Thus, to interpret la as a topic marker helps maintain theoretical unity

for this research.

More importantly, however, is the fact that many pause particles are used both clause

finally and in the middle of the clause, and that the same particle when used in the middle

of a clause may fulfill the same function as when used clause finally. For instance, the

CONTRASTING function is fulfilled by the sentence-middle ne as well as the clause-final ne.

Chao's (1968:802) "question with a specific point" is contrastive in nature, as seen in the

following example he gives for illustration:

^#feJI#; «?59

Ta hui la tiqin; ni ne?

He can play violin; you SFP

'He can play the violin, how about you?'

In this sentence, ni 'you' is contrasted with ta 'he'. Likewise, Li and Thompson's

(1981:300-7) interpretation of ne as "response to expectation" also implies contrast between we-closed utterances with the hearer's claim, expectation, or belief. In contextualizing the sentence tamen you san-tiao niu ne_ 'They have three cattle', Li and

Thompson say that the expression "would be a perfect response" to the hearer's statement that " 'they' don't have any money and are very poor" (1981:301). Shei studies the discourse functions of ne as an SFP, and concludes that "the SFP ne is primarily used in discourse to manage the Idea Structure, by contrasting ideas or by asking other participants to complete a proposition, ..." (2005:5). Therefore, CONTRASTING is a

185 function fulfilled by particle ne irrespective of its position of occurrence.

Both points taken into consideration, it should be more realistic to study the pause particle as it is used in context than simplistically attributing some functions to the clause-final position and others to the clause-middle position and in turn judging some particles as topic markers and others as not.

6.3.4 Summary

To summarize, particle a and its phonetic variant la, na, wa and ya can all occur to mark topic expressions. Of the five, a and ya are more frequently occurring than the other three. This set of pause particles also shows some functional opposition. The two high frequency ones (a and ya) are found to be more extensive in distribution as well.

Although all can be used to introduce or enhance an ENTITY topic, la is used more for

LISTING and ILLUSTRATING purposes, while na indicates self-containment on the part of the speaker. The discourse use of these particles undermines the priority of a certain discourse function (e.g. the LISTING function of la and ya) in judging the topic marker status of a certain particle. It would be more meaningful to examine how the marking contributes to the topichood of the marked expression and subsequently to discourse development.

6. 4 Pause particle ma

According to Zhang and Fang (1996), ma seems to have taken the place of me in

186 Beijing Mandarin, on basis of the fact that me is rarely used by the Beijing locals. They also note that ma and ne seem to be in complementary distribution, in that ma tends to occur early in a stretch of discourse while ne usually appears toward the end of the stretch, especially when a topic shift is in order or a new perspective is to be introduced. This statement, however, is only partially evidenced by the present corpus.

Table 6.14 Linguistic Contexts of Use of Particle ma within the Scope of the Clause

Boundary ma\ mal mdi WO-FINAL marker LBM RBM LBM RBM LBM RBM LBM RBM comma 9 24 1 1 period 38 24 6 5 1 1 2 1 question 1 1 1 mark paragraph 2 boundary total 49 49 7 7 1 1 2 2

First, ma is not a high frequency particle in this sampled corpus. In total, 59 instances are used in the middle of a clause, with the hosting clauses occurring at the beginning of a discourse stretch.

Second, the distribution of wo (see Table 6.14) differs to some extent from all other pause particles discussed so far. Of the 59 instances of topic marker ma, 49 follow the first topic, seven the second, one the third, and two the FINAL. In contrast, the frequency counts of other particles following the FINAL topic expressions are all fairly large, second only to those marking the first topic. For instance, there are 238 instances of «e-marked

FINAL topics compared to 157 for the second topic and 21 for the third; the respective

187 figures for ba, likewise, are 82, 54 and 13. This imbalance, I think, can be attributed to the different discourse functions and semantic roles of the marked topics, as shown in

Table 6.15 and Table 6.16.

Table 6.15 Discourse Functions of ma in Different Contexts

Function mai mat mai /MO-FINAL total REPEATING 22 2 1 25 INTRODUCING 13 1 1 1 16 PICKING UP 8 8 CONTINUING 5 3 8 LISTING 1 1 PROMPTING 1 1 total 49 7 1 2 59

Table 6.16 Semantic Roles of ma-Marked Topics in Different Contexts

Meaning mai mal mai OTA-FINAL total ENTITY 30 2 32 TIME 11 3 1 15 ACTION/ACTIVITY 2 1 3 RESPECT 3 3 REASON 1 1 2 EVENT 1 1 INTERPERSONAL 1 1 MANNER 1 1 PLACE 1 1 total 49 7 1 2 59

First, the INTRODUCING function (recorded 16 times altogether) is not the most frequently observed for ma; the REPEATING category has taken its place with a total of 25 instances. The PICKING UP category records eight first topics that are marked by ma.

When the semantic roles of the marked topics in these three categories of discourse functions are examined, the results are as follow: ten ENTITY roles are under the

188 INTRODUCING category, six under the PICKING UP category, and 13 under the REPEATING category. With another instance falling under the LISTING category, altogether 30 ma-marked first topics are ENTITY roles. Besides, there are two more ENTITY roles marked by ma in the second topic position: one follows a conjunction and the other is more a revision of the first /no-marked ENTITY topic. In contrast, neither of the two ma-marked

FINAL topics are ENTITY roles: one indicates a sort of action and the other an event. The general impression then is that ma-marked ENTITY topics tend to occur early, most favorably in the first topic position. This tendency is probably attributable to the given information status of the marked topics. It is self-evident that ENTITY topics under the

PICKING UP and REPEATING functional categories convey discourse-old information. When the ten ENTITY topics in the INTRODUCING category are concerned, all are anchored in discourse already, as illustrated in (31) and (32).

(3\)iMmmMMTm, m&aq, m, TH&IUL. (oi5-D25:5)

Tamen fa de na piaozi ma, jiushi jiao, ng, Zhunbei Piaor.

They issue NOM that bill PART just call PF Preparatory Bill

'The bills they issued were just called Preparatory Bills.'

(32)MJBm^m, &M;bm^%i&)o (197-X22:6)

Shangdian shichang ma, haishi bijiao fanrong de.

Shop market PART still fairly prosperous CSC

'As for the shops and markets, they are fairly prosperous.'

The ma-marked entity in (31) is introduced into discourse by means of a relative clause

189 where the subject tamen 'they' is discourse-old information. Also, currency is the subject

matter of the paragraph in which this clause occurs. (32) occurs at the very beginning of a

paragraph, where the topic is shifted from transport to business, both among the interview topics. Therefore, the conclusion seems to be that ma is usually used to mark early

occurring topics. When the marked topic is an entity, it needs to be either known or

sufficiently anchored in discourse. This feature of clause-middle ma is very much like what Shei (2005:9-10) has observed for sentence-final ma: ma56i is concerned primarily with the information state and mutual knowledge.... The fundamental function performed by ma5 seems to be to mark some information as particularly obvious and relevant to the current situation. Sometimes it also points to some background knowledge retrievable from the information provided, thereby bringing this knowledge to the foreground as shared knowledge in the discourse domain.

In this light, it may not be appropriate to separate clause-middle and clause-final particles completely or to reject the topic marker status of some particle just because it still conveys some mood information. When the perspective of cognitive linguistics is taken, the resemblance here might be accounted for by the frequency effect: on the one hand, the fairly low frequency of use is inadequate to bleach the mood sense of the particle ma; on the other, the most grammaticalized topic marker among all particles should be ne, not a or ba,62 given the fact that ne is by far the most frequently used pause particle and found to perform the most diverse discourse function.

6.5 Lexical topic marker dehua 'if and laijiang 'speaking of

This section focuses on the discourse use of two lexical topic markers, dehua 'if;

190 supposing' and laijiang 'speaking of (and its variant laishud). According to Yuan (2002), topic marker dehua evolves from its suppositive homophone and expresses a similar meaning to laijiang 'speaking of. This corpus analysis reveals that these two topic markers not only express a similar meaning, they also co-occur with each other and with other pause particles. For instance, dehua and laijiang are observed to be followed by a, ba, ne and ya. Of the four particles, ne is the most frequent one, with 20 instances following dehua, 30 following laijiang, and 14 following laishuo. Furthermore, there are

12 instances of laijiang dehua, three of laijiang dehua ne and one of laijiang dehua ya, but none of laishuo dehua. When distribution is concerned, the two lexical expressions and their co-occurring pause particles do share something in common. This finding aligns with Yuan's (2002) observation that dehua and ne may co-occur in Beijing Mandarin and that dehua as a topic marker is similar to ne in discourse function and distribution pattern.

However, the present corpus presents a fuller picture than what Yuan has observed. In what follows, I discuss in detail how similar or different these two lexical topic markers are and how they are related to their co-occurring pause particles.

6.5.1 Frequency counts and contexts of occurrence

The frequency counts and contexts of occurrence of dehua and laijiang/laishuo are presented in Table 6.17 and Table 6.18 respectively. Table 6.17 Linguistic Contexts of Use of Particle dehua

within the Scope of the Clause Boundary dehual dehual dehual dehua-FiNAL marker LBM RBM LBM RBM LBM RBM LBM RBM comma 14 22 2 6 3 5 period 49 43 15 11 1 5 4 question 2 4 1 1 mark paragraph 4 2 2 1 1 boundary total 69 69 20 20 1 1 9 9

Table 6.18 Linguistic Contexts of Use of Particle laijiang

within the Scope of the Clause Boundary laijiang] laijiang! laijiangi laijiang-FWAL marker LBM RBM LBM RBM LBM RBM LBM RBM comma 14 37 8 12 1 2 3 5 period 56 35 37 32 8 6 10 7 question 1 1 1 mark paragraph 5 4 2 1 1 1 boundary monologue 1 1 boundary total 77 77 46 46 9 9 14 14

From the tables, it can be seen that dehua shows up 99 times in the corpus, 69 marking the first topic, 20 the second, one the third, and nine the FINAL (Table 6.17).

Laijiang and its variant laishuo are used 146 times altogether, 77 times following the first topic, 46 the second, nine the third, and 14 the FINAL (Table 6.18). It seems that laijiang/laishuo is somewhat more frequently used than dehua in Beijing Mandarin.

192 6.5.2 Discourse functions and semantic roles of topics marked by dehua and laijiang/laishuo

It has been mentioned in Section 6.5.1 that laijiang/laishuo has a higher frequency of use than dehua. Proportionate to the frequency of use, laijiang/laishuo also fulfills more discourse functions than dehua. The categories identified for dehua, as shown in Table

6.19, include CONTINUING (41 instances), REPEATING (25), INTRODUCING (20),

CONTRASTING (7), ILLUSTRATING (5), and PICKING UP (1) in order of frequency count.

Unlike dehua, the REPEATING category (61 instances) ranks the first in frequency of use for laijiang/laishuo, followed by the category of INTRODUCING (57 cases), CONTINUING

(15), PICKING UP (3), PROMPTING (2), and one instance for each of the CONTRASTING,

LISTING and SUMMARIZING categories (as in Table 6.20).

Table 6.19 Discourse Functions of dehua in Different Contexts

Function dehua\ dehual dehuab dehua-FiNAL total CONTINUING 33 8 41 REPEATING 14 7 4 25 INTRODUCING 11 3 1 5 20 CONTRASTING 6 1 7 ILLUSTRATING 4 1 5 PICKING UP 1 1 total 69 20 1 9 99

193 Table 6.20 Discourse Functions of laijiang in Different Contexts Function laijiangl laijiangl laijiangl laijiang-FmAL total REPEATING 26 26 3 6 61 INTRODUCING 33 12 5 1 57 CONTINUING 9 5 1 15 ILLUSTRATING 4 1 5 PICKING UP 1 1 1 3 PROMPTING 1 1 2 CONTRASTING 1 1 LISTING 1 1 SUMMARIZING 1 1 total 77 46 9 14 146

When Table 6.19 and Table 6.20 are compared, three categories are found to be particular to laijiang/laishuo, namely the LISTING, PROMPTING and SUMMARIZING. Since none of these three categories is a major one, the occurrence of these three categories may be attributable to the more frequent use of the lexical topic marker. Of the shared categories, the CONTINUING category stands prominent for dehua. A closer look reveals that it is mainly due to the frequent occurrence of a manner adverbial zheyang(r) 'thus; in this way'. Of the 33 dehua-marked CONTINUING first topics, 26 are realized by this expression; and of the eight CONTINUING second topics, six are this manner adverbial.

This observation also explains why MANNER is the highest frequency category among all semantic roles for dehua (34 in total, Table 6.21). Moreover, the CONTINUING discourse function of dehua is also enhanced by its collocation with a typical continuing conjunction suoyi 'so'. Of the six dehua-marked zheyang(r) topics that occur in the second topic position, five are in collocation with suoyi. Therefore, dehua is a typical

194 marker of CONTINUING topics, and the expression of zheyang(r) dehua has been

somewhat fossilized in fulfilling this function.

Dehua is also observed to mark CONTRASTING topics. A frequency count shows that six de/zwa-marked CONTRASTING topics occur in the first-topic position and one more in the second-topic position. The contrasted topic can be either experiential (as in (33)) or textual (as in (34)) in metafunction. (33)p£, &a£iaa, MaMi^, -±, mtim, &m, ^m^mum,

^Pio (092-L44.-5)

Ai. zai-guoqu dehua. jianzhi yaoshi shuo shi, yishang,

PF in the past if simply if speak is FS

ben bei qu ba, shiba, zhege jiu bu-hao-zou shide, shiwa.

head for north go PART TAG PF just not-good-walk as if TAG

'In the past, whenever heading for the north is mentioned, the poor road condition

would jump into people's mind.'

(34)£MIM> »TM£, ^Pi? (141-X03:7)

Buran dehua, ta naozi dou hen-kong, shiwa?

otherwise if she brain even very-empty TAG

'Otherwise, she would even be empty-headed, wouldn't she?'

In (33), the past time is contrasted with the present time, and the marked topic is an experiential element. The marked element in (34) is conjunctive in nature, serving as a textual link to previous discourse. In fact, expressions like buran 'otherwise' in (34) are

195 antonymous to expressions like zheyang 'thus'. In this way, dehua can be used to move

the discourse forward in two opposing directions.

Table 6.21 Semantic Roles of dehua-Marked Topics in Different Contexts

Meaning dehuaX dehual dehual dehua-FiNAL total MANNER 27 1 34 ENTITY 15 2 9 26 TIME 7 3 1 11 ACTION/ACTIVITY 7 7 CONJUNCTIVE 5 5 PLACE 2 3 5 RESPECT 1 2 3 INTERPERSONAL 2 1 3 EVENT 1 1 2 BASE 1 1 ISSUE 1 1 TEXTUAL 1 1 total 69 20 1 9 99

Table 6.22 Semantic Roles of/a///7mg-Marked Topics in Different Contexts

Meaning laijiangl laijiangl laijiangl laijian -FINAL total ENTITY 24 12 3 11 50 RESPECT 15 11 3 29 TIME 8 8 1 17 BASE 8 3 1 12 PERSPECTIVE 8 1 9 PLACE 1 6 1 1 9 TEXTUAL 7 1 8 INTERPERSONAL 2 4 6 ACTION/ACTIVITY 2 1 3 CONJUNCTIVE 1 1 EVENT 1 1 MANNER 1 1 total 77 46 9 14 146

What distinguishes laijiang/laishuo from other topic markers, dehua included, is its

196 frequent use to mark topics expressing the semantic role of RESPECT (see Table 6.21 and

Table 6.22). There are altogether 29 instances, second only to the category of ENTITY (50 in all). In contrast, the total use of dehua after this type of topic is three instances only. In fact, the category of RESPECT is not so frequently observed among the topics marked by all other topic markers as by laijiang. Besides, the BASE category and PERSPECTIVE category are also proportionately large for laijiang in frequency count, which are 12 and nine instances respectively. As mentioned above, laijiang/laishuo is a lexical topic marker with a consistent lexical meaning 'speaking of. This is a feature very different from particle topic markers, in that the meaning of the latter is more context dependent.

Given its fairly definite lexical meaning, laijiang is semantically compatible with topics of the RESPECT, BASE and PERSPECTIVE categories, as illustrated respectively in (35), (36) and (37).

(35)^&^Btfi]^iIE^-S^i#, WJ&£$L (253-T54:13)

Ta zhege shiiian baozheng fangmian laijiang , jiu da-bu-dao.

It PF time guarantee respect speaking of AUX reach-not-reach

'Speaking of time guarantee, it can't be obtained.'

(36)M±££&A^ffiBfi. £-#0*. (141-X03:11)

Duivu daduoshur laijiang ne, shi yi-jian haoshi.

For majority speaking of PART is one-CL good thing

'For the great majority, it's a good thing.'

(37)hk^A1%mm#m, m, M»Jfr£*#^lftlULo (260-X39:l)

197 Cong geren jiaodu laijiang ne. ng,

From personal perspective speaking of PART PF

ye mei yudao shenme da kanke de shir.

too not meet what big frustration NOM thing

'Speaking of myself, I haven't met with serious frustrations (in my life).'

(35) is an instance of RESPECT topic, where laijiang is used independently to mark the topic. The marked topic in (36) gives the base for the comment; that is, the impact of the thing is based on whether the great majority of people will benefit from it or not. (37) shows a PERSPECTIVE topic, namely the speaker takes herself as the perspective of commenting. It should be noted that laijiang is complemented by pause particle ne in (36) and (37). Also complementing the topic marker under discussion are two prepositions, duiyu 'for' in (36) and cong 'from' in (37). In fact, cong...laijiang is a frequent collocation in Mandarin Chinese. Given the fact that cong indicates the starting point, no anomaly would be incurred if it is used in (35) to frame the topic expression together with laijiang.

Because dehua and laijiang/laishuo are lexical topic markers, they are not as well grammaticalized as pause particles. As Yuan (2002) has argued, dehua is often coupled with lexical expressions like shuo 'speak' to the left and particle ne to the right in use.

Yuan also acknowledges that in modern spoken Beijing Mandarin, independent use of dehua has become a common phenomenon and its overall frequency of use is on the increase. This corpus study aligns with Yuan's comments. For example, there are 69

198 instances marked by dehua; and of these, 48 are marked by dehua alone. The corpus also shows a more grammaticalized use of dehua in topic marking; that is, de- in dehua may take zero articulation as (38) shows.

(38)&|£iU£«fPfo7$5i0ft> (120-H15:5)

Zheyangr hua ni jiu yingxiang le shixian sihua,

Thus HUA you AUX affect ASP realize four modernizations

'In this way, the realization of the four modernizations will be affected.'

There are two instances of (de)hua in the corpus. Though both of them are produced by one single speaker, they are far from unacceptable to native speakers of Beijing Mandarin.

Actually, this is a phenomenon repeatedly observed in languages, i.e., high frequency items are more reduced in form (e.g. Bybee and Hopper 2002, Bybee 2001), especially given the fact that de is never a stressed syllable in dehua. When it comes to laijiang/laishuo, however, no such reduction is observed.

6.5.3 Summary

In summary, dehua and laijiang/laishuo, as lexical topic markers, are both similar to and different from each other. They are similar in that they both have a definite lexical meaning, and that they can co-occur with pause particles like a, ba, ne and ya besides the possibility of occurring together, with each making its own contribution to the discourse meaning. The co-occurrence phenomenon reveals that the two lexical topic markers and their co-occurring topic-marking particles can be used in the same context. The fact that

199 the discourse functions identified for dehua and laijiang/laishuo overlap significantly

with those of the pause particles a, ba, ne and ya offers additional support for this

supposition. However, dehua and laijiang/laishuo also show a functional opposition in discourse use. While dehua tends to mark topics that move the discourse forward progressively or adversatively, laijiang/laishuo most often marks topics introducing

RESPECT, BASE and PERSPECTIVE.

6.6 Discussion

In this chapter, a wide range of topic markers have been addressed in terms of frequency of occurrence, distribution and discourse function. While each topic marker has its own feature in discourse use, they can also complement each other in topic marking or co-occur to mark one single topic expression. As a result, topic marking has become a complicated phenomenon in Beijing Mandarin. Likewise, the marked topics are also heterogeneous in both semantic meaning and discourse function. This section is therefore devoted to a recapitulative discussion of the diverse manners of topic marking and the information status of marked topics as observed in this corpus.

6.6.1 Diversity in topic marking

By diversity in topic marking, I mean two things: the diversity in topic markers and the diversity in the manner of topic marking. The former has been fully illustrated in previous sections in this chapter, but even that is far from the entire picture. Broadly

200 speaking, "topic marker" is a cover term for pauses, pause particles, initial word order, and some lexical expressions (Zhang and Chen 2006). Chapter 5 mainly addressed pauses as topic markers, and in Section 6.3, some higher frequency pause particles and two lexical expressions are discussed in detail. A number of additional items, however, also function as topic markers, which are not analyzed in this study. For instance, Tao's attention-getting expressions, such as ni kan 'You see', ni xiang 'for example', ni zhidao ba 'you know', ni shuo (literally 'you say', mainly to introduce an entity into discourse), etc. (1996:166-168) are also observed in the clause-initial position, functioning in a similar manner to topic markers. Here, I would prefer to interpret them as topic markers as well, if only because they can fulfill the topic marking function. Their uncertain status is just because they are not as grammatical ized as others. The four attention-getters mentioned above are respectively instantiated 21 times, 54 times, once64 and 44 times.

Of even lower frequency of use are some "typical" topic markers, like zhiyu 'as for', guanyu 'about', and duiyu 'for'. The frequency counts for them are two, eight and 11 respectively. They may be "typical" in some genres of language use, but not in the spoken variety of Beijing Mandarin. In any case, "topic marker" is a cover term for a wide range of expressions, from semantically empty pause particles to fossilized expressions that convey definite meanings.

When it comes to the manner of topic marking, a range has been observed as well.

First, a clause may have several topics in a row. This has been evidenced by the use of almost all topic markers discussed in this chapter, pause particles and lexical topic

201 markers alike. As each table has shown, topic marking may occur in different positions within a clause, from after the first element to the element immediately preceding the predicate. Second, all kinds of topics can be marked. In terms of metafunctions, marked topics can be experiential, textual or interpersonal. Although in general the three types of topics occur in a certain order, i.e., from textual to interpersonal and then to experiential, there are also variations to this general order. Also, topics of the same metafunction may occur more than once in one single clause. Example 13 in Chapter 5 illustrates a chained reference relationship between two marked experiential topics, and the example is reproduced below as (39) for convenience of reference.

(39)X%mm%±W, &mW\M, «W£JL*I&o (057-H20:5)

Youqi wo-na-laoda ya, zhe-xifur ne,

Especially my first child PART, this-wife PART,

ta jiu youdianr xiangfa.

she then a little dissidence

'Especially my first child, his wife has some dissidence.'

In this example, a chain of referents is observed, which starts from the speaker to her first child, and ends with the wife of the first child. It should be noted that the chaining is unidirectional, that is, from the most accessible to the least accessible.

In addition to the chained reference relationship, a topic may be approached from different angles, as illustrated by (40). m))k&mm9Lm&&T&M%, mm&m&i%. ii<»«o (m-x46-.2)

202 Cong iiankang qingkuang xianzai laiiiang dehua.

From health condition now speaking of if

na shi pubian dehua, qingkuang dou hen-hao.

that is general if condition all very-good

'Speaking of the health condition, nowadays, in general, it's very good.'

The topic of this clause is approached from three different angles. First, it is singled out

through the nominal modifier jiankang 'health', which occurs very early in the clause.

Then the time dimension is delimited to the time of speaking. Finally, the speaker

constrains the scope within which the comment is to be interpreted by means of the

interpersonal element nashi pubian dehua 'if in general'. This is a phenomenon also noticed by Tao when he talks about referent anchoring NP IUs: "Another prominent type of NP IU includes those which are used by the speaker to describe to the hearer the

intended referent from different angles, usually in a step by step manner" (1996:91).

Although his main concern is the flexibility available between the production of NP referents and that of the predicate, utterances such as (40) reveal that topics used in this manner need not be NPs only; PPs and APs may also be employed for such purposes.

In terms of the semantic roles fulfilled by the marked topics, participant roles and circumstantial roles alike are observed. Of all role types, ENTITY is by far the most frequently occurring type, with 1015 instances marked by different topic markers (see

Table 6.23).

203 Table 6.23 Frequency counts of marked ENTITY topics Topic ne ba a ya laijiang ma dehua na wa la total marker Frequency 527 165 93 85 50 32 26 24 8 5 1015 count

This is not unpredictable though, given that entities are the most prominent elements in a construed situation. Also prevalent are marked TIME and PLACE topics—entities and events exist or occur in a certain place at a certain time. From time to time, other circumstantial dimensions (like MANNER) may also become prominent enough to take marking. In short, topic marking is frequent enough in Beijing Mandarin to make it a topic-prominent language. The diversity in topic markers and topic marking manners substantiates the status of the language as topic-prominent.

6.6.2 Information status of marked topics and function of topic marking

So far, the information status of marked topics has been addressed only indirectly when talking about the discourse functions of marked topics. As a corollary of categorization, topics under the REPEATING, PROMPTING and PICKING UP categories convey discourse-old information. Somewhat predictable are those in the CONTINUING,

CONTRASTING and ILLUSTRATING categories. As indicated by the category names, a marked topic performing the CONTINUING function is a continuation of a particular element in the preceding discourse. Likewise, a CONTRASTING topic is anchored through contrast, and an ILLUSTRATING one through the illustration of a claim in a prior clause.

204 The most evasive is the category of topics fulfilling the INTRODUCING discourse function.

Therefore, in what follows, I focus on this category and illustrate the information status of the marked topics with examples where particle ne is the marker of the first topic (see

Tables 6.2 and 6.3).

First, I address ENTITY topics. There are altogether 72 instances of first topics marked by ne, and 56 of these can be taken as accessible from the discourse or situational context.

More specifically, two topics make reference to the speaker self, which, of course, is always situationally available though it may be a first mention introduced into the discourse. In this sense, these may be taken as situationally available topics when marked.65 A topic may also be accessed by means of the possessive construction or the part-whole relationship between the introduced entity and a discourse-old one. World knowledge constitutes another means of accessing a topic. For instance, when a language informant talks about her two daughters, she uses the NP lia-guye 'two sons-in-law' as the topic for a clause (085-H14:2). Though the reference has never been made in previous discourse, world knowledge enables the hearer to interpret this expression as pragmatically anchored—it is normal for people with daughters to have sons-in-law as well. Modification is also an effective means to access an entity, because it can single out some entity or entities with a certain attribute. For example, laonian-ren 'old-age people'

(043-N50:4) stands in contrast with young or middle-aged people. It should be noted here that the referents of the expression are more generic than definite, which may stand as a challenge to the widely held assumption that topics should be definite in information

205 status (e.g. Tsao 1979, Xu and Liu 1998, Shi 2001, etc. as detailed in Chapter 2).

A second challenge to the assumption comes from topic NPs where the number or amount, rather than the referent, is the major concern. There are two instances: one is shikuaiqian 'ten yuan' (141-X03:19) and the other yikuaiqian 'one yuan' (141-X03:19).

In both cases, the focus is more on the amount of the money indicated by the two numerals (shi 'ten' and yi 'one' respectively) than on the money itself.

The most serious challenge, I think, is from NPs that are indefinite in reference.

There are 14 such examples (about 20%) among this set of topics, typically coded with youde/youxie 'some', xuduo 'many', or the existential you.

(4nWjSAq>S, f&W&'h 1!SJAL, (141-X03:12)

Youxie ren ne, ta you zhege, ng yuanjian,

Some people PART he have PF PF foresight

'Some people, they have foresight.'

(42)JLM±^S2LE!. &*' ?l£BBrtgB%^A;ra#, (134-D09:3)

You liang-ge xiaohair ne. jiushi. henduo-bu neibu dianyingr bu-gai kan,

Have two-CL child PART PF very many-CL internal movie not-should watch

'There are two young children who should not have watched many rated movies.'

In (41), the underlined topic is indefinite in reference, but interestingly, the pronoun ta

'he' is used to refer back to the NP as if it were definite. The ne-marked topic in (42) is similar in reference to (41), emphasizing the existence of two such children, instead of specifying which two. Thus it seems that topics need not be definite. Sometimes,

206 definiteness is simply irrelevant, as in those emphasizing the number/amount, and under circumstances such as (41) and (42), etc. So long as the information can be activated

(Chafe 1994:ch.6), an expression can enjoy the topic status. It should be noted, however, that the marking in such utterances tends to be obligatory, though it may be reduced to a pause only. In this manner, topic markers play the role of enhancing the topic status of the marked expression.

Also irrelevant to the notion of definiteness are the semantic roles of

ACTION/ACTIVITY and EVENT. They can by no means be measured in terms of definiteness, but they are pervasive as marked topics in the corpus data. Given these phenomena, perhaps Shi (1995) is right in interpreting topic as early-known information, or Haiman's

(1978) observation that once said, the information is known, which is a more unified interpretation for topic as a cover term for a range of topic expressions and functions.

Complementary to early-known information is permanently-known information, or in Erteschic-Shir's terminology "permanently available topics", that is, the temporal and spatial parameters (2007:17). Of the 121 we-marked first topics of TIME, three subtypes can be identified. One is represented by zai zhege guocheng zhong 'in the course of

(141-X03:3), which orients the time with an ongoing event narrated in previous discourse.

The second type indicates an indefinite time, typically coded as youshihou 'sometimes'

(211-N52:4). The third type is the most numerous (51 instances), 33 of which appear as xianzai 'now', with several more expressed in the temporal correspondents to proper names (e.g. qijiunian 'the year of 1979' (204-L53:4)). The PLACE category is moderate in

207 number (14 instances), and more than half are anchored with reference to the 'here' of the speech event or simply to the speaker (e.g. women fujin 'near our place' (197-X22:6)).

There are also generic references like gongyuanr-li 'in a park' (162-H37:5) where no actual park is conceptualized. Proper names like Beijing are also observed to occur clause initially (e.g. 015-D25:7).

To summarize, discourse-old information is not the only type of information that can be topicalized in Beijing Mandarin. Neither is definiteness an obligatory feature of marked topics in the language. Generic and indefinite elements, once marked, may also function as the starting point for a proposition.

6.6.3 A cognitive interpretation of topic marking: first attempt

Speech production can be viewed as starting from the generation of a dynamic series of communicative intentions, each of which is instantly decomposable into a set of sub-goals (Levelt 1989). In achieving the sub-goals, there is 'the one new idea at a time' constraint as proposed by Chafe (1987, 1994). Tao (1996:ch.5) uses this constraint to account for the phenomenon that the introduction of the referent and the predication about the referent are accomplished in two separate intonation units. This, I think, can be extended to topic-comment constructions. It has been argued in Chapter 5 that pause marked topics tend to be more complicated in coding form than the unmarked, which is in agreement with Givon's iconicity principle on the selection of heavy versus light forms; that is, the harder the reference is to process, the more grammatical material must be

208 assigned to it. Although not all marked topics are heavily coded, to grant an expression topichood may need more material, such as topic marking. Therefore, it is not unreasonable for a topic marker to occur, and more often than not be accompanied by a pause.

Pause-complemented topic markers may have another function to fulfill. According to Huang (1993), the more (filled or not filled) pauses there are in a speaker's speech, the more likely it is that the speaker is trying hard to come up with accurate or appropriate information at that particular point of speaking. If the comment is indeed the nucleus of a topic-comment construction, it would be more likely for there to be a pause between the topic and comment. Though the length of pausing in the corpus data, which takes the form of transcripts, cannot be measured, it is intuitively true that many pause particles can take a prolonged articulation.

In short, the use of topic markers has a cognitive basis. It is not only necessary for the speaker to package his or her information; it also functions to guide the hearer in how to interpret the packaged information.

6.7 Summary

In this chapter, some frequently used pause particles together with two lexical topic markers are studied one by one at the clause level, with necessary reference made to earlier or later occurring discourse. In the discussion section, the diversity in topic marking is highlighted. It is also argued that definiteness may not be a fundamental

209 property of marked topics; to enhance the topic status of the marked expression seems

more basic a function of topic marking. Finally, topic marking is cognitively interpreted as a conjoined requirement of speaker's information packaging and hearer's message

interpreting. In next chapter, a unified account of various topic constructions as identified

in the Beijing Corpus is made in terms of the schema of conceptual reference point for the ensuing comments.

210 Chapter 7: A Unified Interpretation of Topic

as Conceptual Reference Point

This chapter addresses the possibility of whether different types of topics can be

covered under one single schema, that of Conceptual Reference Point or Reference Point

for short. If each type of topic can be taken as an instance of the same schema, a unified

interpretation of different types of topics can be achieved. From another perspective, different types of topic-comment constructions may also enrich the notion of Reference

Point. Langacker (1991:Chapter 7) makes a binary distinction between "subjective" and

"objective" Reference Point, and takes topic as a subjective Reference Point. Given the whole range of topic expressions as marked by pauses and pause particles, it is possible to observe some as objective Reference Points as well, especially given the fact that the topic expression and an entity within the clause may be in such objective relationships as possessor-possessed, part-whole, etc. In this manner, this chapter not only echoes Chapter

3 by justifying the heavy reliance on topic markers in topic identification, but also develops Langacker's Reference Point interpretation of topic with respect to Beijing

Mandarin.

First, it is highlighted that the Reference Point function can be fulfilled by semantic roles other than entities. Then I point out that Reference Point can be multiple in one single clause. Finally, the schema of Conceptual Reference Point is delineated both horizontally and vertically.

211 7.1 Reference Points are more than entities.

Langacker defines topic as a kind of "subjective reference point" (1991:314). In

Cognitive Grammar, Reference Point is a notion related to the ability to utilize the idea of one entity in order to invoke another closely related one. In more technical terms, a

Conceptual Reference Point provides "mental access to a target (T) found in its dominion (D)" (Langacker 1998:7), where the target is "the entity which is identified in an utterance by virtue of invoking a particular reference point" (Evans 2007:209) and dominion "the possible set of targets that a given reference point can invoke" (Evans

2007:63). When talking about the operation of Conceptual Reference Point, Langacker

(1998:7) maintains: By serving as an initial focus of attention, the reference point tends to activate an array of associated entities which collectively constitute its dominion. This enables attention to shift to a secondary focus within it, the target (which activates its own dominion and may in turn function as a reference point).

In this definition, Reference Point has both a certain cognitive salience and a dynamic aspect to it. There is, however, a subset of topics which operate somewhat differently from the 'initial focus' type, namely those conflated with the subject. Such topics serve as

Reference Point with respect to the clausal process, which is "the closest connection possible between a nominal reference point and a processual target". Langacker takes this process as "intrinsic rather than contingent" (Langacker 1998:19), in that the subject functions as "starting point for the very purpose of conceptualizaing the profiled clausal process", or "the target clause does not exist independently from the reference point"

212 fulfilled by the subject (Langacker 1999:371).

From the above discussion, it can be seen that Cognitive Grammar has limited the

semantic role of Reference Point to entities only, though Langacker (1987:198) maintains that "the term entity must be understood in a maximally general sense". He uses entity as

"a convenient cover term for anything we might conceive of or refer to for analytical purposes: things, relations, locations, points on a scale, sensations, interconnections, values, etc." (1987:198). General as the term is, if the Reference Point phenomenon is indeed as fundamental and ubiquitous as claimed by Langacker, most probably it may be fulfilled by items with other semantic roles as well. It has been illustrated repeatedly in

Chapter 6 that the semantic role of marked topics can be diverse enough to include textual and interpersonal elements, which also contribute to discourse semantics but are not covered by Langacker's "entity". Given the significance of semantic meaning in cognitive linguistics in general and the semantic role as the most important topicality factor in Cognitive Grammar (see Chapter 2), I discuss in what follows the diverse semantic roles fulfilled by topic expressions, and explore how they serve as Conceptual

Reference Points for another entity in the clause in which they occur or for the processual target (i.e., the clausal process) per se.

7.1.1 ENTITY topics

As far as the present corpus is concerned, the clause-initial elements, i.e., the topics, are far richer in semantic role than discrete entities. Besides ENTITY topics, there are

213 ACTION/ACTIVITY, RESPECT, TIME, PLACE, MEANS, and MANNER topics as well as topics fulfilling textual and interpersonal discourse functions. ENTITY topics,66 due to their high frequency of occurrence, are best taken as the prototype of all kinds of topic expressions in the corpus. As introduced in Chapter 2, Cognitive Grammar views language as usage-based. In a usage-based model of language, the relative frequency with which a particular word or construction is encountered by the speaker is held to affect the nature of the language system, and a linguistic unit that is more frequently encountered becomes more entrenched in the language system.

Table 7.1 shows the frequency counts of the first ten most frequently occurring semantic roles fulfilled by topics marked by the pause particles discussed in Chapter 6

(i.e., ne, ba, a, la, na, wa, ya and ma) and the two lexical topic markers dehua and laijiang/laishuo.61

Table 7.1 Frequency Counts of Top 10 Semantic Roles of Marked Topics Semantic role ne ba a la na wa ya ma dehua laijiang total ENTITY 527 165 93 5 24 8 85 32 26 50 1015 TIME 377 89 30 9 8 24 15 11 17 580 CONJUNCTIVE 158 2 1 5 1 167 RESPECT 62 14 14 2 1 6 3 3 29 134 ACTION/ACTIVITY 34 18 19 3 2 1 17 3 7 3 107 PLACE 43 18 9 2 2 7 1 5 9 96 MANNER 51 7 1 34 1 94 TEXTUAL 65 2 1 1 8 77 EVENT 47 12 1 2 1 63 INTERPERSONAL 22 7 4 1 3 1 1 3 6 48 total 1386 334 170 11 41 20 140 57 97 125 2381

From the table, it can be seen that marked ENTITY topics rank by far the highest in

214 frequency (1015 instances, or over 40 percent of all marked topics), and that ENTITY

topics can be marked by all high frequency topic markers. Below are two ENTITY

examples marked by the most used pause particle ne, where the Reference Point function

is fulfilled somewhat differently.

(1) r.mm,

Er-de ne, xiaohair cai san-sui,

Second-NOM PART child only three-years of age

'As for my second child, his kid is only three years old.'

(2) If!, 3»Wlttire^7rW^7o (281-H40:5)

Gongchang ne, xianzai youde gongchang yijing kaishi fudong le.

Factory PART now some factory already begin float ASP

'As for factories, some have begun to raise the salaries already.'

In (1), the underlined topic, i.e., the second child of the speaker, is used as the Reference

Point for his own child, which functions in turn as the grammatical subject of the clause.

(2) is somewhat different, in that the topic expression is generic, rather than specific, in reference. It specifies the category of which the target is a subset. There are two points worthy of noting in (2). The first is that a time adverbial occurs between the Reference

Point and the target, and the second is that the target is not a definite expression. But since the category or domain has been delineated by the topic expression, the target expression may very well be taken as anchored or accessible information in discourse.

This corpus analysis shows that two other categories can also be taken as ENTITY

215 topics, namely the ACTION/ACTIVITY category and the EVENT category, in that the two categories specify abstract (vs. concrete) entities. This is not unexpected though.

According to Tomasello (2003:196), "(N)ominals are used by people to make reference to

'things'. In many theories, the prototype is concrete objects (people, places, and things).

But it is well known that nominals may be used to refer to basically any kind of entity at all, real or imagined. Thus, when the needs arise, there are ways of construing actions, properties, and relationships as if they were things, on analogy with concrete objects." It seems that Tomasello's "thing" is synonymous with Langacker's entity. In the following two examples, (3) is an instance of the ACTION/ACTIVITY category and (4) the EVENT category. Since both topics are also grammatical subjects, they serve as Reference Points for the processual target.

(3) %, %M*m, tttfcffljf, (309-X12:3)

Ng, kao gaozhong ne, bijiao kunnan,

PF take entrance exams senior high school PART fairly difficult

'It's fairly difficult to pass the entrance exams for senior high school.' (4) mMT&#m&&M&&£Am, MM, ^mmmi^^m, (302-149:3)

Jiushishuo zhan guitai iiechu zheme-duo ren ne.

That is to say stand counter encounter this-many people PART

jiushi, keyi shuo kaikuo-le yanjie a,

PF may say broaden-ASP horizon SFP

'That is to say, as shop assistants, we can get to know people of various kinds, which

216 can be said to have broadened our horizons.'

In (3), the marked topic as underlined indicates an action or activity, i.e., taking the exams, but since it is common for people to say that this book or that book is difficult, an analogy can be made between the action/activity of taking exams with the entity of book.

The underlined topic expression in (4) expresses an event, namely getting to know people through serving as a shop assistant. Also, this EVENT topic can be analogized to any other

ENTITY topics that can broaden people's horizons. Topics in these two semantic categories have one thing in common; that is, the speaker wants to comment on, or predicate, the action/activity or the event. Langacker (1987b) notes that the discourse function of identifying the participants in events and states of affairs requires language users to construe whatever they wish to talk about as a "thing", so it can be referred to, irrespective of its "true" ontological status. Therefore, these topics may very well be interpreted as entities. When this interpretation is taken, the total occurrence of ENTITY topics is 1185 instances, half of the total given in Table 7.1 or 48% of all marked topics.

Thus, ENTITY topics are indeed prototypical of marked topics in Beijing Mandarin.

7.1.2 Setting indicating topics

7.1.2.1 TIME topics

The category following ENTITY in terms of frequency count is TIME, a category depicting the setting of events, which instantiates 580 times in total in the present corpus.

The marked topics expressing the semantics of time are found to take different forms,

217 ranging from adverbials, as in (5), (6) and (7), to nominals, as in (8) and (9). (5) m^m, nW^m%, (288-N40:3) Ranhou ne. xie nage biaoyu,

Then PART write PF slogan

'Then we wrote the slogans.'

(6) jfertaugqig, mmm^Mmmxft, MM*, (I34-DO9:D

Lai-le yihou ne, wojuedeyingdang shi genghao-di gongzuo, shi'a.

Come-ASP later PART I think should is better-AUX work TAG

'After coming here, I think that I should work better, shouldn't I.'

In (5), the TIME topic occurs in the middle of a clause serial and is coded in the form of an adverb, indicating the chronological order of the narrated event. In (6), the topic consists of the verb lai 'come', the perfect aspect marker le, and the adverbial expression yihou

'later'. A common feature shared by these two examples is that both are explicitly related to the preceding discourse by the time order.

(7) Mfefci&P/g, VC%mt$-m%, &Mm—&)L&3a (162-H37:3)

Jiu-xianzai-laishuo ne. bi bie-de danwei ne, haishi cha-yidianr de.

on-now-speaking of PART COMP other-NOM unit PART still bad-a little CSC

'Speaking of the present time, our unit is still a bit worse than others.'

The feature of interest in this example is that the expression xianzai 'now' indicates the time of speaking, an indispensable element of the speech event, or in Erteschik-Shir's

(2007:ch. 2) classification, a permanently available topic. Also note that the time topic is

218 framed with jiu...laisshuo 'speaking of, which is a lexical topic marker accompanied by the pause particle ne. The double marking emphasizes the topic-marked time xianzai

'now' with respect to other points on the time scale.

(8) 4-^-m, %Xm, ^-f^iko (141-X03:2)

Jinnian ne. laoda ne, shi chuzhong biye.

This year PART first child PART is junior high school graduate

'This year, my first child will graduate from junior high school.'

In (8), the «e-marked time topic takes the form of a nominal phrase, though it is adverbial in function.

(9) ^m^m, iYMMo (281-H40:3)

Jinnian de dongtian ne. bijiao hao.

This year GEN winter PART fairly good

'It's fairly good this winter.'

The topic expression in this example is somewhat different from (5) to (8), in that the utterance can be interpreted as a TOP-subject construction where the «e-marked topic resembles an entity and functions as the grammatical subject. However, the topic expression can still be taken as a time adverbial, because the speaker's concern here is the weather in the topicalized season. In other words, the marked topic expression is used metonymically for one aspect of it, i.e., the weather of the season. With cases like (9),

TIME constitutes a fuzzy category whose boundaries cannot be clearly defined.

219 7.1.2.2 PLACE topics

Also part of the setting is the category of PLACE topics. There are 96 instances of

PLACE topics, ranging from prototypical adverbials of place to entity-like place-denoting nominals.

(10)PJt, £MMMM&$k° (274-T30:l)

Ai, zai nongcun ne cha-guo-dui.

PF in countryside PARTjoin-ASP-team

'Well, I was once sent down to the countryside (to live and work with the production

teams).'

This is a prototypical example of PLACE topic, given that the preposition zai 'in' is used to precede the place-indicating NP.

(UMtPM, SfcfRftlltJt#?c. (015-D25:7)

Beijing ne, zui taoyan-de jiushi chuntian.

Beijing PART most annoying-NOM is spring

'In Beijing, the most annoying season is spring.'

This example is less typical, in that the place adverbial is coded in the form of a proper noun, rather than the canonical prepositional phrase. Moreover, this clause can be classified as either a TC or double-subject construction. As a TC, the underlined topic is adverbial in function; as a double-subject construction, the underlined is the topic and the ensuing nominalization the subject. In the latter case, the two nominal expressions are in

220 the part-whole relationship, with Beijing being the whole and the nominalization a part or

an aspect of the city.

(l2)%^mWtm&MAL7^n)m%til$lo (099~T27:2)

Gedi ne jiushiye chengli-le zhunmenr yanjiujigou.

Various places PART PF also establish-ASP specialized research institute

'Various places have established specialized research institutes.'

This example is more marginal than (11), because the place-indicating expression can be

understood as the grammatical subject under a metonymic reading, i.e., the place is used

for the people living there. As evidenced by the naturalness of the English translation, this phenomenon is commonly observed in different languages. Clauses like (12) show that

PLACE as a topic category is also fuzzy at the boundaries.

7.1.2.3 MANNER topics

Another category also adverbial in nature is the MANNER category, with 94

instantiations in the present corpus. The most frequently marked expression is zheyang

'in this way', as in (13). Other adverbial expressions are also possible, as in (14).

(UY&ftJlM, &$#&#£AftJi^itMJg. (309-X12:3)

Zheyangr ne, jiu nengjiejue haoduo-ren de zhege jiuye wenti.

Thus PART AUX can solve many-people GEN PF employment issue

'In this way, many people's employment issue can be solved.'

(14)JUtAiM' WLM%^, &^ftfttt&&^mg&ti}MMJ9k, (309-X12:5)

221 Manmanr de ne, jiushi ta ne,

Gradually AUX PART PF he PART

ye buzai zhao shenme zhege juweihui de mafan le ha,

too no longer look for what PF community committee GEN trouble ASP SFP

'Gradually, he no longer found fault with the community committee.'

7.1.2.4 Setting indicating topics as Reference Points

The three categories of adverbial topics mentioned so far, namely adverbials of time, place and manner, operate as Conceptual Reference Points in a different way from ENTITY topics. As discussed above, an ENTITY topic is used to invoke either another entity within its domain or the processual target per se when the entity is a core argument (often the subject) of the host clause. In contrast, adverbial topics constitute the targets of the domains they themselves invoke, given the fact that time and place are two basic domains of human conceptualization and manner is also inherent in events and states of affairs.

They fulfill the function of Reference Point by relating the clausal process to these peripheral aspects. These topics are subject to the construal of the speaker, and their occurrences, unlike core-argument topics, are contingent rather than intrinsic.

7.1.3 CONJUNCTIVE topics

The third most frequently occurring category is conjunctives, with a total of 167 instances. The CONJUNCTIVE topics can develop the discourse either progressively or

222 adversatively, and include such expressions as yinci/suoyi 'therefore', erqie 'also',

danshi/keshi 'but', yaoburan 'otherwise', etc. These conjunctives can be combined with

the textual expressions to form a unified category, given that they all serve as connectives

between utterances and reflect aspects of the speaker's epistemic stance towards the

sequencing of discourse. If a distinction is made, the conjunctives are more local

discourse markers whereas the TEXTUAL topics are more global discourse organizers. In terms of Reference Point, topics in this category refer their target clause to previous discourse.

(15)g[»^, $OMJiI#MttS*, (197~X22:2)

Jiaoshi doushijiaoshi doushi gongzuo-liang bijiao da,

Teacher all REP work-load fairly big m\>jM, i&inmxmwtM-m, (ddi366)

suovi ne, tamen de gongzi jiu gao yixie,

so PART they GEN salary AUX high some

'All the teachers there have a comparatively heavier work load, so their salaries are

somewhat bigger.'

(141-X03:2)

Zaiyige ne. jiushi cong shehui zhegejiaodu-shangkan ne,

Another PART PF from society this angle-on look PART

zhe xuexi ne, queshi shi weiranchengfeng de.

223 this study PART indeed is become the order of the day CSC

'For another, when society is taken as a whole, to study has indeed become the order

of the day.'

In (15), the two clauses connected by the underlined conjunctive suoyi 'so' are both given, and the result of a bigger salary in the second clause is on account of a heavier work load mentioned in the first clause. In contrast, (16) occurs at the beginning of a paragraph which elaborates the proposition coded in this multiple-topic clause, with the preceding paragraph focusing on the importance of study. That is, it takes two paragraphs to fully develop the given subject matter of everyday study. The comparison between (15) and

(16) shows that both conjunctives and textual elements may be used to connect utterances, but the latter seem to operate at a higher level (i.e., the paragraph level) of discourse organization than the former (i.e., at the clause level).

7.1.4 RESPECT topics

The next category in terms of frequency of use is RESPECT topics which, as the category name indicates, specify the respect in which the comment is to be interpreted.

Though not very high in frequency count (62 instances), this category of topics perhaps best fits the characterization of 'aboutness'. Like other semantic categories, RESPECT topics also have central as well as marginal instantiations. e { (17)£&±MLM±M> &Rm%r %i& gm, U34-D09:4) Zai haizi jiaoyu shang ne. wo zhineng shi yange di guanli,

224 In child education on PART I only can is strictly AUX manage

'As for children's education, I can only manage them strictly.'

(1K)&WW, &&&}¥%&&*¥&. (302-T49:l)

Jiaovu ne. jiu wo de xueli shi gaozhong biye.

Education PART only I GEN education is senior high school graduate

'When education is concerned, only my education background is a senior high school

graduate.'

Of (17) and (18), the former is more standard in coding form than the latter, in that the respect is explicitly framed by the prepositional expressionzai...shang 'as for; when ... is concerned' in (17). In contrast, the RESPECT topic in (18) is coded in the form of an NP, a topic form commonly observed in Mandarin TC constructions. The comment in both cases is 'about' the RESPECT topic, but the relationship is fairly loose; that is, the topic is by no means a core argument of the predicate verb. In contrast, the RESPECT topics in the following two examples have a closer relationship with their respective comment.

(I9)&, &4>ffl%>Hfeg;frSJm)S, a^fcmtt&^fa, (309-X12:10)

Zai, zai Zhongguo zhege dili fangmianr ne. zhege zhishi bijiao pinfa de,

FS in China PF geography respect PART PF knowledge rather poor CSC

'When geography of China is concerned, my knowledge is rather poor.'

(20)&%>KE£;frMJL#2rMJL'%, Umm* (036-N44:8)

Zai zhege weisheng fangmianr ge-fangmianr ne, dou ting-hao de.

In PF hygiene respect each-respect PART all very-good CSC

225 'When hygiene is mentioned, every respect is pretty good.'

In (19), though the topic is still framed by the PP zai...fangmianr 'in ... respect', the subject in the ensuing clause (zhishi 'knowledge') takes modification from the RESPECT topic; that is, the knowledge has to be geography-related. But for the prepositional expression, the entire utterance might be classified as a case of double-subject construction where a part-whole relationship would be observed between the two NPs.

(20) perhaps shows an even closer relationship between the RESPECT topic and the comment. Although the topic expression still takes the form of a PP, it functions as the grammatical subject of the ensuing adjective phrase. In other words, this clause is both a

TC and a TOP-subject construction.

This interpretation appears contradictory to the claim that subjects are not marked by prepositions (e.g. Tsao 1990:170). However, (20) is a naturally occurring utterance, and it is by no means a unique instance in the present corpus. There are a number of other preposition-marked expressions serving as the grammatical subject of the predicate.

(2l)WLM, MiH^U§/«, XlMftW, &Jtl£8t£&$WiS&to. (2ll-N52:6)

Jiushi. cong wo geren jiaodu ne. dui liixing ne,

PF from I self angle PART to travel PART

yeshi bijiao bijiao ganxingqu de.

too fairly REP interested CSC

'That is, from my own perspective, I'm fairly interested in travelling.'

In this example, the underlined topic specifies the speaker's perspective from which the

226 comment is to be interpreted. Meanwhile, it is also the grammatical subject or

experiencer of the feeling of 'interested', despite the fact that it is marked by the

preposition cong 'from'. (22))k$t&Am&mw,m%iMm, mm^m° (2ii-N52:4)

Cong wo-airen he wo-mama de guanxi ne. haishi bucuo de.

From my spouse and my mother NOM relation PART is not bad CSC

'The relationship between my wife and my mother is not bad.'

(22') wo-airen he wo-mama de guanxi ne. haishi bucuo de.

The preposition cong 'from' in (21) is a component of the frame expression cong...jiaodu

'from ...angle', but in (22) the preposition is used independently. Moreover, it is dubious whether it still performs the function of a preposition; it seems more like a pre-positioned topic marker. If cong is omitted (as in (22')), the semantic or pragmatic meaning of the utterance is not much affected.

(23)kk^*7kmMifc1j)L, 7K«o (183-X46:5)

Cong Miyun Shuiku de difangr, shuiyuan ma.

From Miyun reservoir NOM place water source SFP

'The place of Miyun Reservoir is a water source.'

(23) resembles (22) in almost every respect except that the copula verb is missing.

Preposition-marked grammatical subjects are also observed to be predicated by content verbs, as in the case of (24), where the verb isguo 'live a life'. avmn&mm, kk&mE&mmimmmm, fu?i&#fM£*iftj. (085-HI4:2)

227 Women zhe jiating. cone wo zhe sa haizi he women-de Hating a.

We this family from my this three child and our-GEN family PART

dou hen guo-de hen xingfu de.

all FS live-CSCvery happy CSC

'This family of ours, my three children's families and my own all live a happy life.'

Given these examples, it seems that the functional role of subject is not entirely

incompatible with preposition marking, especially when the preposition-marked subject

is separated from the predicate with a pause or pause particle. To put it differently, if a

grammatical subject gains the status of topic by means of topic marking, it can then take

a preposition. Perhaps this is exactly how topic differs from subject, namely topic accepts

such marking while subject usually does not. However, the boundaries are by no means clear-cut, especially when the semantic role is taken as the most objective criterion of categorization.

Also, cong is not the only preposition that can be used in this manner. For instance, speaker 260-X39 tends to use both cong and zuowei 'as' before nominal expressions that fulfill the grammatical subject role; speaker 071-D57 favors the use of laijiang (dehua)

'(if) speaking of...' after NPs; and speaker 106-X14 uses nishuo 'you say' to precede NP expressions as an attention getter. The illustrative preposition xiang 'like' is found across almost all monologues. Although such expressions tend to be separated from the predicate with a pause, it should be noted that a pause may not be obligatory, as in the case of (25).

228 aswmmmm&T&m.Gmmo (26o-x39:3)

Cong women jiaodu xianzai hai meiyou jiedao.

From our perspective now yet not receive

'We haven't received (the file) yet.'

In this example, the underlined element is not separated from the rest of the clause with a

pause or a pause particle. According to how topic is identified in this research, the

element does not qualify as a topic. Irrespective of the wording, the element is

semantically the goal of the verb jiedao 'receive', thus the grammatical subject of the

clause. As a result, we have a preposition-preceded subject. Chao is perhaps very unusual

to propose that actor subjects can be introduced by prepositions, on the basis of the

observation that "the subject is but a minor sentence" in the sense that the subject in a full

sentence can be interpreted "as the topic and as the question" and the predicate "as the comment on the topic and as the answer to the question (Chao 1968:82). As such, he argues that subject "can take, among other forms, the form of a preposition and an object"

(1968:85). Expressions such as (25) seem to show that subjects are not completely

incompatible with prepositions.

7.1.5 INTERPERSONAL topics

The only major category unaddressed by now is INTERPERSONAL topics which occur

48 times in this spoken corpus. INTERPERSONAL topic as used here is in line with

Halliday's definition, that is, "any combination of (i) vocative, (ii) modal, (iii)

229 mood-marking" (1994:53-4). This category is, therefore, more inclusive than Fraser's

(1988) "discourse markers" or Blakemore's (1987) "discourse connectives".

(26)£l5±t|)ggjg£ffiiPJi—#£it, (H3-D21:9)

Shiiishang ne zhexie dongxi dou shi yi-zhong anwei,

In fact PART these things all are one-CL comfort

'In fact, these things are all kinds of comfort.'

The underlined topic expression falls under the category of discourse markers, and refers the utterance in which it occurs to the preceding discourse by means of verification.

Pio (092-L44:l)

Dangran ne. shi da gongxiaoshe ne,

Of course PART is big supply and marketing co-operative PART

shuoshi cong zhege fangmianr ma, zong-de-kanqilai,

it's said from this respect PART generally speaking

haishi bu-cuo de, shiwa.

still not-bad CSC TAG

'Of course, generally speaking, bigger supply and marketing co-operatives are not

bad in every respect.'

The clause-initial element dangran 'of course' is a modal adjunct, highlighting the obviousness of the ensuing statement. The grammatical subject in this example is the second «e-marked expression where shi is best taken as an intensifier or attention getter

230 rather than a copula (despite the gloss). The marking here effects contrast between bigger

and smaller supply and marketing co-operatives. Thus, the INTERPERSONAL topic is used to refer the hearer to the stance or attitude of the speaker on the present utterance, without

linking back directly to a prior utterance. a&m'vtnm, mmw, &xm, $t£&wfM+£-Htftj. (295-LI2:2)

Wo iuede ne, jiushi-shuo, zhe ren ba,

I think PART that is to say this person PART

jiushi meiyou shenme shiquanshimei de.

simply without what flawless and perfect CSC

'1 think, that is, when people are concerned, no one is flawless and perfect.'

In this research, expressions like the first underlined topic in (28) are interpreted as

INTERPERSONAL topics when occurring clause-initially. There are at least two reasons for so doing. One is that such expressions can occur either at the beginning or in the middle of an utterance, which helps justify a non-matrix clause reading of such expressions.

More importantly, it indicates that the ensuing predication is the personal opinion of the speaker, and can be reworded as 'in my opinion' without any anomaly being incurred in semantic meaning. The second underlined element also fulfills the interpersonal function, referring the hearer to a reading of the clause as rewording. Thus, in this example, we see consecutive occurrence of INTERPERSONAL topics in one single clause.

It can be seen from illustrations above that INTERPERSONAL topics, though not great in number, are also coded in varying forms, with some more clause-bounded (such as (27)

231 and (28)) and others more discourse-oriented (as in (26)). In any case, the hearer is

'referred' to the ensuing proposition by this type of topic expression.

7.1.6 Summary

To recapitulate, ENTITY topics are the most frequently occurring topic type, which is in agreement with Cognitive Grammar's characterization of topic as an entity that invokes another entity within its domain or the very act of conceptualizing the profiled relationship when the topic is conflated with the subject. However, other semantic categories are also observed to function as topics and refer their host clauses in different ways to the discourse context. Action/activity and event expressions are also found to serve as Reference Points for the processual target, very much in the same manner as

ENTITY topics. Adverbial topics such as TIME, PLACE, MANNER, etc. are used to refer the clausal process to the adverbial or circumstantial aspects. TEXTUAL topics, conjunctions included, anchor the present utterance to the prior one(s). Also of this anchoring function are some INTERPERSONAL topics (i.e., those discourse-marker interpersonal expressions), with other INTERPERSONAL topics directing more clause-internally. RESPECT topics, crosscutting adverbial and entity topics, are perhaps the exemplar category that specifies explicitly the domain within which the target is to be identified. All in all, Conceptual

Reference Point appears to be successful in characterizing the array of topics that have been identified in this corpus study: topics of varying semantic meanings all refer their target clause to the discourse context, with each functioning in its own manner.

232 7.2 Reference Points can be multiple.

To this point, we have seen numerous examples of single Conceptual Reference

Point, one of which is (1) in Section 7.1, repeated here for convenience of reference.

(1) ~mm, 'hWl*^.$, (057-H20:5)

Er-de ne, xiaohair cai san-sui,

Second-NOM PART child only three-years of age

'As for my second child, his kid is only three years old.'

As mentioned in Section 7.1, a prototypical Reference Point operates by activating an array of associated entities and shifting attention to the target. In (1), the second child of the speaker is a Reference Point for the target, his three-year-old child. In this section, however, the focus of attention is on multiple topic marking and how they function as

Reference Point Chains and Networks.

7.2.1 Multiple topics as Reference Point Chains

In addition to the one-step mode of activation as in (1), the Reference Point may also operate in a chained manner; that is, one Reference Point may shift the focus of attention to its target, and the target may in turn serve as the Reference Point for a new target. This phenomenon has been illustrated by (13) in Chapter 5, which is reproduced below as

(29). {29)im$iM%Xm, &m&)im, »W£JL*Jl&o (057-H20:5) Youqi wo-na-laoda ya, zhe xifur ne,

233 Especially my first child PART, this wife PART,

ta jiu youdianr xiangfa. (057-H20)

she then a little dissidence

'Especially my first child, his wife has some dissidence.'

In this example, a Reference Point Chain is observed, which starts from the speaker to her first child, and proceeds to the wife of the first child. In (30), the entities in the

Reference Point Chain are made up of both human and non-human entities. eons:, immm, gftjt&gAi, $, m^^mmnmM, a£#*, m±MR

MB)Lim&~m, (113-D21:2)

Ai, ta-muqin ne. iiushi wo-airen ne,

PF his mother PART that is my spouse PART

ai, ta suo congshi de gongzuo a, xianzai kanlai,

PF she AUX undertake NOM work PART now appear

wanshang xuexi de shir bijiao qing yixie.

evening study NOM thing comparatively light some

'As for his mother, that is, my wife, what she's doing now doesn't need her to do

heavy study work in the evenings for the moment.'

In this example, the Reference Point Chain consists of four links, i.e., the four underlined elements. Since the previous discourse has had the wife's child as the topic, the Reference

Point Chain starts from the child (coded in the pronominal form ta 'he') in the first topic expression ta-muqin 'his mother'. The second topic expression, like appositives generally,

234 has a clarification purpose. Then attention is extended from the human entity to her work, and finally to a section of her work.

A closer examination of (29) and (30) shows that the chaining is unidirectional in both examples and that the reverse order is most probably unacceptable. In (29), the chain starts from the situationally available speaker T and through two steps ends with the wife of the first child of the speaker. The chain in (30) begins with the already activated entity, i.e., the speaker's child, and closes at a section of the child's mother's work. Thus, the general rule is for the chain to proceed from the more accessible to the less accessible.

Nonetheless, this unidirectional chaining rule does not necessarily imply that the initial focus must be discourse-old or situationally available to the speech event as is the case in (29) and (30). In (31) we see an example of a more generic type of topic introduction:

OnW-iEfe, MM, Ma4-££«BE, MUmTM, (288-N40:5)

You vi-iia ha, jiushi, ta zhege gonggongpopo ba.

Have one-family PART PF her this parents-in-law PART

dou ting nenggan de,

both very capable CSC

'There was a family, and the parents-in-law of the woman were both very capable.'

This utterance occurs at the beginning of a story told by the speaker for illustration purposes. The existential you 'have' in the utterance-initial position indicates that the referent is neither definite nor specific. But the existential construction itself is an

235 efficient means for introducing new entities into discourse (e.g. Du Bois 1987). When it is marked by the ensuing pause particle ha, the newly introduced entity is promoted to topichood. Once the entity has been established in the mental space of the hearer, it can function as the Reference Point for the ensuing target entities. Of course, the successful establishment of the association between the two underlined expressions presupposes such world or encyclopaedic knowledge that parents-in-law live together with the young couple in some households.

Given the unidirectional feature, Reference Point Chains are in themselves reflections of dynamic conceptualization, just as Langacker has noted (1999:363-364): Invoking a reference point relation is thus an inherently dynamic process involving a shift in focus from the more readily accessible reference point to a target accessed through it. Since the target is then in focus, it has the potential to function in turn as reference point in its own right, giving access to another target. We thus encounter chains of reference point relations, ...

In short, multiple topics may form Reference Point Chains in establishing mental contact with the target. A Reference Point Chain usually moves from more accessible to less accessible entities. Occasionally, a newly introduced entity may also function as the starting point of a Reference Point Chain on condition that the entity can be established in the communicator's mental space.

7.2.2 Multiple topics as Reference Point Networks

In addition to Reference Point Chains, a target may also be approached from different angles, as illustrated in (32).

236 aim&mmmM&xw&ift, mM&mmft,- mumm. d83-x46:2)

Cong jiankang qingkuang xianzai laijiang dehua.

From health condition now speaking of if

na shi pubian dehua. qingkuang dou hen-hao.

that is general if condition all very-good

'Speaking of the health condition, nowadays, in general, it's very good.'

The topic of this clause is approached from three different angles. First, the nominal phrase jiankang qingkuang 'health condition' is singled out from the domain of all given topics of the interview. Then the time dimension is delimited to the time of speaking.

Finally, the speaker constrains the scope with respect to which the comment is to be interpreted by means of the interpersonal element nashi pubian dehua 'if in general'.

Multiple topic marking as illustrated in (32) is far from unique in the corpus; rather, it is common occurrence in spontaneous speech. Of a total of around 13,000 clauses in the present corpus, 1568 are found to have a second topic, and 235 have a third topic. In the subset of TC constructions, there are 1237 clauses beginning with a conjunctive expression, 405 of which are followed by another marked topic fulfilling as diverse roles as grammatical subject or object, adverbial of time, place, manner, etc., as well as a second textual element or an interpersonal element. Because of this wide variety of elements available as topics, the target, be it an entity or a clausal process, can be approached from different angles. What differentiates this mode of referencing from the chained mode is that there is one target only, and the only target stays where all activated

237 domains intersect. In other words, the target seems to lie at the center of a network formed by different topics.

It has been proposed in Section 7.1 that some types of topics are more discoursal than clausal. For instance, the CONJUNCTIVE functions primarily to link the utterance in which it occurs to the previous clause(s). TIME topics, in contrast, may operate in a double manner: if it is with respect to a time or event introduced in previous discourse, it is both discoursal and clausal; if it is just a point or a span on the time axis, it is usually more clausal. In this manner, the Reference Point Network can anchor its host clause at both local and global levels.

7.2.3 Other multiple topics

It should be noted that not all multiple topics fall neatly under the Chain or Network categories. From time to time, the speaker may struggle a while in establishing the topic or the Reference Point.

flfoo (260-X39.-2)

Dangran ni yao shuo, cong geren iiaodu. cong geren shenghuo a.

Certainly you AUX say from personal angle from personal life PART

jiushi cong jiating iiaodu. shibi yao shou yixie yingxiang.

that is from family angle surely AUX suffer some influence

'Of course, if you consider from the perspective of individuals or personal life, to be

238 exact, the family, it's bound to be affected.'

This utterance is produced when talking about inflation, and it takes the speaker some effort to pinpoint the best Reference Point for the target process. First, the speaker produced the interpersonal element dangran 'certainly; of course' to show his attitude towards the target statement he is to make. After producing the attention getter niyaoshuo

'if you say', he makes two trials before he finds the right "entry" with which to approach the target. Perhaps this explains why he has chosen to mark it with the emphatic jiushi.

Roughly, the multiple topic in this example seems to approach the target from two angles, thus forming a network.

This struggling-for-topic phenomenon is found more frequently when the speech shifts to a new subject matter, as is the case with the following example.

5£#&i#PE, &, AJtJfeifl, &1E, fttt¥ttMnn&M'%. (190-D16:2)

Fanzheng. si, women-jia haizai ba, jiushi shuo de zhege.

In any case PF our family child PART that is speak NOM this

qishi women-jia libianr ba, zhege xuexi,

in fact our family inside PART PF study

fumu laijiang ba, hai, cong wo laijiang, fanzheng,

parents speaking of PART FS from me speaking of in any case

wo dui xuexi bushi zhua de tebie jin.

I to study not seize AUX especially tight

239 'In any case, my child, well, in fact in my family, in study, as a parent, speaking of

me, anyhow, I don't study very hard.'

This sentence occurs at the beginning of a paragraph where the speaker shifts her speech

to study, a different subject matter for the same interview. As the underlining shows, there

are as many as eight topics preceding the propositional statement, when all repetitions

and shuffles are counted. It might be interesting as well to note that the pause particles

used in this clause are unanimously ba; none is ne. As mentioned in Chapter 6, Chao

(1968:806) takes ne as pause for both speaker and hearer to consider some specific point

brought up by the we-marked expression. Then it seems that the speaker here is struggling

for the right "entry" himself, without inviting the listener to collaborate.

In terms of Reference Point mode, I think that (34) combines the chain and the

network mode. The chain consists of women-jia 'our family', fumu 'parents' as

indispensable components of the family, and wo T as a parent, leaving out women-jia

haizi 'our child' as an unsuccessful trial. The network is formed by the interpersonal

elementfanzheng 'in any case', the participant role in the chain and xuexi 'study', a given

subject matter of the interview. In short, the two modes of Reference Point organization

may co-work in one single TC construction.

7.2.4 On the limit to the number of marked topics in one clause

It can be seen from the above discussion that topic can occur a number of times in one single clause. In (34) for example, there are as many as eight topics functioning as

240 Reference Points for the target clause, although some of them may be understood as false starts, slips of the tongue, or changes of ideas, features which are typical of spontaneous speech. The question then is how many topics there can be in a single clause. According to Yuan (1996), five different semantic elements can occur in a sequence of topic positions in Mandarin Chinese, namely the agent, the adverbial of time, place or manner, the patient, the instrument, and the patient of an embedded clause. If Halliday's (1985,

1994) three metafunctions are referred to, these five types all fall under the experiential metafunction, which leaves open the possibility that there can also be topics fulfilling the other two metafunctions, namely the textual and the interpersonal metafunction. Since examples above have shown that both TEXTUAL and INTERPERSONAL topics can occur more than once in a single clause, the number of potential topics may very well exceed five. Furthermore, as adjuncts, adverbial expressions of time, place, manner, etc. may also occur more than once or co-occur with one another. Therefore, the number of potential topics in a clause is perhaps in theory unlimited, especially when taking into consideration the features of spontaneous speech.

In practice, however, the number of consecutive topics is subject to the limited cognitive capacity of the speaker and the hearer. Topics, marked topics in particular, are perhaps not the least energy-demanding type of information. It has been shown in

Chapter 5 that marked grammatical subjects (i.e., topics in TOP-subject constructions) are more complicated in coding form than subjects in SP constructions. If Givon's (1983) iconicity principle is followed, the referent of the topic in TOP-subject should be harder

241 to process than the subject in SP. Langacker also refers to topic as the "initial focus of attention" (1998:7); that is, the speaker and hearer's attention must be focused on the topic expression at least for a while. Xu and Liu (1998:116) go one step further by proposing a functional category, i.e., topic focus, implying that topic marking functions to promote the information value of the marked topic. This corpus analysis also shows that topic markers, diverse as they are, fulfill one common function, that is, enhancing the topic status of the marked expression. Everything considered, topics most probably demand fairly large processing capacity, and therefore cannot occur in an unlimited number despite their recursive potential. There are altogether 4711 clauses in the present corpus that have marked topic(s), excluding such special constructions like TOP, LD, etc.

1568 of these clauses also have a second topic, but only 235 clauses are observed to take a third topic. Cases like (34) are indeed rare.

Speakers appear to consciously avoid using too many topics in one single clause from time to time. As a compromise, the speaker may resort to rhetorical questions, as is shown in (35).

(35) g BMtmWT;££8#JLI15J. (043-N50:2)

Ziji jiali de zinti shi zheyangr a.

Self family NOM children is like this SFP

'My own children are like this.'

zai xuexi-shang a. muqian laishuo ne,

242 in study-on PART at present speaking of PART

cong ziji zinu shang kan ne, you hen-duo de xiangfar.

from self children on look PART have very many NOM comments

'When study is concerned, for the moment, as for my own children, I have lots of

comments.'

Youqi zuo fumu de, duiyu zinu de kanfar ne, shenme ne?

Especially do parent NOM as for children NOM opinion PART what SFP

'Especially as a parent, my comments on my children, well, what are they?'

This excerpt occurs at the beginning of a paragraph where the subject matter is shifted to the speaker's opinion on children's education. A close examination of the three clauses in

(35) reveals that nothing specific or substantial about the subject matter has yet been said by the speaker to the hearer. It is more like the speaker is talking to himself. He first produces an SP where the predicate shi zheyangr a 'is like this' is cataphoric in reference.

In the second line, he produces three topics in a row, but after finding that the comment remains hard to come by, he closes this utterance with a very general expression you henduo de xiangfar 'have lots of comments'. The speaker's struggle for an appropriate comment continues in the third clause. After producing two more topics to no avail, he utters a question directed at himself. It is not until after these three clauses that the speaker succeeds in coding his thought in words. This example shows that the speaker may use various techniques to limit the number of topics in one single clause. A good

243 speaker will not overstretch himself/herself as well as the hearer(s) cognitively, and he/she will observe the convention of closing an utterance with a comment (sometimes one in a fashion), without leaving the topic or topics stranded. In other words, Reference

Point-Target organization is a process rooted in human cognition.

7.2.5 Summary

To recapitulate, topic as Conceptual Reference Point has multitudinal manifestations in naturally occurring language data. Marked topics in the spoken variety of Beijing

Mandarin have maintained the distinction between core-argument Reference Point and non-core-argument Reference Point. When the marked topic is the grammatical subject or object of the clause, it serves as an intrinsic Conceptual Reference Point for the clausal process; otherwise, it refers the target clause to a contingent aspect, such as a circumstantial, interpersonal or textual element. In the latter case, this corpus study reveals that the Reference Point function can be realized in two complementary modes, the chain mode and the network mode. In the chain mode, the first topic is used as a

Reference Point for the second one, and the second for the third, and so on and so forth, until the clausal target is accessed. A Reference Point Chain will be thus formed, and in general the chain follows a certain direction, in most cases from more accessible referent(s) to less accessible one(s). In the network mode, topics of different

(meta)functions are used in a consecutive manner to delineate the interpretation of the clausal target. Some topics used in this mode are to link the present utterance to previous

244 discourse (like conjunctives) while others are more local by referring the clausal target to the time axis for example. In both modes, topics may occur recursively, but the limited cognitive capacity on the part of the speaker and hearer will put an upper limit to the total number of actually occurred topics.

7.3 Schematizing Topic-Comment Constructions

It can be seen from the above discussion that topic as identified in this research is broader than topic as defined by Langacker and many others. But given the fact that

Conceptual Reference Points are more than entities, as illustrated in Section 7.1, it remains a possibility to interpret all topic constructions identified in the spoken corpus as instantiations of one single schema, the schema of Conceptual Reference Point, with some as prototype categories and others as extensions.

Given the network model proposed by Cognitive Grammar, a complex grammatical category is formed through horizontal extension and vertical schematization (Langacker

1999, Evans 2007, among others). In this section, I first address how topic as a category extends horizontally, and then explore how it schematizes vertically.

7.3.1 Horizontal extension

7.3.1.1 Subject as prototype Reference Point

Langacker (1999) agrees with Givon (1984:138) that the subject is the primary clausal topic and the object the secondary clausal topic, given the following reasoning:

245 "For one thing, the elements unproblematically identified as subjects and objects are full,

grounded69 nominals. Whereas simple nouns merely furnish type specifications,

nominals designate particular instances of those types and function in discourse to

coordinate mental reference to these instances. Furthermore, the subject and object relations hold at the clausal level of organization, and the clause is a basic unit of discourse" (1999:28-9). Among the major construction types identified for this spoken corpus, SP makes up the second largest subset (2196 instances), following only TC (4711

instances). This outcome can be attributed to at least two reasons. For one thing, topic is identified very broadly in this research; that is, all elements preceding the predicate, marked subjects included, are taken as topics. As a natural result, TC as identified here is very much a heterogeneous subset of clauses. All circumstantial elements as well as interpersonal and textual expressions are included. When all these peripheral elements are removed, 1623 of the 4711 TCs are found to be SP constructions, with an additional 1685 being VPs. For another, Mandarin is a language where the subject is an optional element of the clause. Elliptical SPs (i.e., VPs) are found to occur extensively in the corpus. They constitute the third largest group of all construction types, with 2053 instantiations, only marginally less frequent than SPs. Moreover, the 708 instances of TOP-subject would also fall within the subset of SP without the marking. Therefore, SP is indeed a construction type that abounds in actual language use in Beijing Mandarin, and the subject can be said to function in discourse to coordinate mental reference to these clausal processes.

246 The situation with object topics is somewhat different. At the clause level, that is, as the result of initial segmentation (see Chapter 4), only 322 objects are observed to occur

clause-initially. Among these 322 instances, 284 are TOP-object constructions and 38 are

LD-object constructions. The low frequency count of object topics is not without good reason. Du Bois (1987) has demonstrated that adult speakers of a number of languages

show a very consistent pattern, using what he calls "preferred argument structure". The basic idea is that new information, expressed by a lexical noun phrase, is typically introduced either in an intransitive utterance (often in the specialized presentational or existential construction) or else as the object within a transitive utterance. He continues that it is very rare for the subject of a transitive utterance to carry new information expressed in a lexical noun phrase, which means that typically there is only one lexical noun phrase per clause. Tao's (1996) study on conversational Mandarin supports Du

Bois's findings. In investigating the preferred clause type in conversational Mandarin discourse, Tao finds that "most of the verbs are low on the transitivity scale, and clauses rarely have two overt arguments. In the one overt argument, full nouns are more likely to be found in the O role in high transitivity clauses, as they typically convey new information and/or inanimate referents" (1996:126). Here "the transitivity scale" represents a "simplified version" of Hopper and Thompson's (1980) Transitivity

Hypothesis and "the O role" means the object role.

There are two points, I think, immediately responsible for the low frequency count of objects as topics. First, if most verbs are low on the transitivity scale, the total number of

247 object uses will not be very large. Second, the limited number of objects will more often

than not convey new information, which makes the object role incompatible with the

topic position, a position typically reserved for given or accessible information. Although

this spoken corpus is composed of monologues rather than conversations, the transcripts

still reveal a bipartite feature, namely, the speaker also takes the interviewer as his or her

addressee irrespective of the fact that the latter makes few contributions to the speech

event. Given the nature of this corpus and the fact that Du Bois's findings are more cross-linguistic than language-specific, the comparatively infrequent use of objects as topics can be taken as a natural outcome of the preferred clause type and the preferred

information status of topic in Mandarin. When put in terms of Conceptual Reference

Point, objects, given their new information status, are likely targets, i.e., the places that the speaker is trying to get to, instead of the Reference Points. That objects are also seen to function as topics can be explained by the fact that the target clause may sometimes not exist independently from the Reference Point introduced by the object, such as in

TOP-object constructions.

7.3.1.2 First extension from prototype Reference Point: topics in double-subject constructions

The first extension from the prototype Reference Point are topics in double-subject constructions. According to Langacker's (2007:243) characterization, both NPs70 in this construction type have some claim to subjecthood, and although NP2 and the predicate

248 can themselves constitute a clause, the entire structure also has clausal status. Therefore, the first NP or the topic is both intrinsic and contingent to the clausal process, which in turn justifies this construction type as the first extension to prototype topics, i.e., subjects and objects.

This type of construction, however, is not very frequently observed in the present corpus (103 instances in total), and many of the instantiations that do appear are to some extent fossilized expressions. It has been illustrated in Chapter 5 that the prototypical relationship between the topic and subject in this construction type is possessor-possessed, and that the two elements tend to occur in a particular order. There are two immediate consequences of these findings on clausal process construal. First, if the two elements are in the possessive relationship, a kind of objective relationship in Langacker's view, the first element (i.e., the topic) will serve as an objective Reference Point for the target or the possessed. This corollary is then in conflict with the "subjective reference point" interpretation of topic as proposed by Langacker. Furthermore, in some cases, the topic and subject in double-subject constructions occur in a fixed order, which implies that construal is not always the conceptualizer's choice. To put it differently, it is not always

"the organization of knowledge itself (Langacker, 1991:314) that allows a topic to serve its function. These two points may be combined into one, namely, topics are not always

"subjective" Reference Points if possessives are taken to show a kind of objective relationship.

Observations on marginal instantiations of a category may constitute further evidence

249 for rejecting some double-subject constructions as the outcome of subjective construal.

According to Taylor (1998:198), "more marginal instantiations tend to be subject to various idiomatic constraints that are absent with more central members. Constructions, in other words, tend to be less productive at their periphery than at their center".

Double-subject constructions, as noted above, are at the margin of SP and TC constructions. Therefore, it is not something unexpected that they are constrained by word order along with some other factors. They are not very productive either, with a total of 103 instantiations, which is less than one percent of all clauses in the sampled corpus.

7.3.1.3 Second extension from prototype Reference Point: circumstantial topics

Further extension results in topic-comment constructions beginning with all kinds of circumstantial elements, such as time and place of the predicated event. It has been noted above that adverbial expressions as adjuncts can occur and recur at the clause-initial position. Though Cognitive Grammar has not made it explicit that adverbials may serve as Reference Points, it has been acknowledged that "status as a participant, setting, or location is not, however, intrinsic to an entity, but depends on how it is construed in the context of the overall situation" (Langacker 1999:68). If it is just a matter of construal, the setting or location may come onstage and serve as a subjective Reference Point for the clausal process. A study by Ford (1993, as seen in Ford, Fox and Thompson 1998:125) examines the particular work of adverbial clauses in interaction, "(S)ignificantly,

250 adverbial clauses may be placed either before or after the material they modify. In naturally occurring interaction, the use of such clauses in initial position can be a resource for constructing longer turns". This function performed by clause-initial adverbial clauses is compatible with that of topic, which, once established, may "hold sway through a substantial stretch of discourse, until some other topic is established to supplant it" (Langacker 1993:24). Also, fronting is one of the three general considerations identified by van Hoek that underlie Reference Point patterns, "(O)ther things being equal, X is more likely to be taken as a Reference Point in relation to Y if X is introduced into the conceptualizer's awareness before Y is" (van Hoek 1997:58-9). Within a sentence, the main verb, combined with its subject and object(s), forms the prominent core of the sentence, or its profile. Meanwhile, there are adverbials as well. It is true that not all clause-initial adverbials are clauses; there are also phrases. However, the difference in coding form may not have much impact on the discourse function of adverbials. All things considered, adverbials of time, place, manner, etc. may very well enjoy the status of Reference Point for the clausal event.

Nonetheless, it should be noted that adverbial topics, unlike ENTITY topics, are used not to invoke another entity, but profile a particular aspect of the predicated event and access the clausal target via the adverbial topic. In the case of a time adverbial, the topic refers the event to the time scale or the chronological order of the narrative; when it is an adverbial of place, it specifies the location of the clausal process; an adverbial of manner, when occurring in the topic position, gives prominence to the named manner among a set

251 of possible alternatives which most probably constitute the world knowledge of the two

parties in communication. In these cases, the topic expression takes initial focus of

attention by standing out as the target of the domain (i.e., all possible notions) that itself

can establish mental contact with. The reason why these elements do not need to be

invoked by another element, I suspect, is that they are either available in the communication situation or in the speaker/hearer's encyclopedic knowledge. According to Langacker, time and space are two basic domains, meaning that they cannot be fully reduced to another, and of the two, time is "in some sense more fundamental than space"

(1987:148-9).72 Although not all researchers agree with Langacker in this regard, it is most probably true that time is more relevant than space to narrative discourse. This is actually reflected in the present corpus, in that TIME topics are more frequently seen than

PLACE topics. As shown in Table 7.1, TIME topics instantiate 580 times, whereas PLACE topics do 96 times only.

When it comes to MANNER topics, they function in the same way as TIME and PLACE topics, in that they also stand out as the target that they themselves invoke. Though the exact size of the inventory of manner is hard to tell, both speaker and hearer know that things always happen in a particular manner. Experience will tell them which manners are more likely and which are not. Other adverbial types, say adverbials of means, I think, will operate more or less the same as manner adverbials.

252 7.3.1.4 Further extension from prototype Reference Point: INTERPERSONAL and

TEXTUAL topics

Still farther away from prototype topics than the adverbials are clause-initial interpersonal elements. As introduced in section 7.1, this subset of topic is heterogeneous.

Some have an impact on the interpretation of the clause in which they occur, while others serve as discourse markers and link their respective host clause to previous discourse.

Those impacting the interpretation of their host clauses seem more closely related to the clause itself, thus closer to the prototype categories than those serving as discourse markers. At about the same distance from the prototype categories with the second group of INTERPERSONAL topics are TEXTUAL topics, whose main function is to link the host clause to previous clause(s).

INTERPERSONAL and TEXTUAL topics, I suppose, fulfill their Reference Point function in a similar way to adverbials of manner; that is, by choosing the interpersonal or textual expression as the starting point of an utterance, the speaker has picked the expression out from an inventory of possible alternatives and granted the expression the status of an initial focus of attention. Also like adverbials of manner, the inventory may not be a closed class, but the commonly occurring instances have already been conventionalized and internalized as part of the communicator's world knowledge. For instance, a clause is most probably related to the discourse context in terms of condition, cause, justification, elaboration (e.g. additivity), contrast (adversativity), or topic shift.

253 7.3.1.5 RESPECT topics

The remaining subset is RESPECT topics. This subset differentiates from other topics

in that it specifies the domain explicitly, and more often than not, the domain is abstract

rather than concrete. For instance, the invoked domain in (17) in Section 7.1 is children's

education, and that in (18) is education in general, and thus more abstract than in (17).

Also, the domains in (19) and (20), geography of China and hygiene respectively, are

conventionalized entities, not something in the physical world. But there is indeed some

difference between (17)-(18) and (19)-(20): the relationship between the RESPECT topic

and the grammatical subject is fairly loose in (17) and (18), but much tighter in (19) and

(20): in contrast with (17) and (18), where the topic specifies no more than the respect,

the topic in (19), i.e., zai Zhongguo zhege dili fangmianr 'when geography of China is

concerned', can be taken as a modifier of the subject zhishi 'knowledge', and that in (20), zai zhege weisheng mangmianr ge fangmianr 'when hygience is mentioned', can be

understood as a grammatical subject. Except for the prepositions that are used in these two topic expressions, they might be re-classified as double-subject or TOP-subject

constructions. Therefore, RESPECT topics should be farther away from the prototype categories than double-subject constructions but not as far as INTERPERSONAL or TEXTUAL topics. It should be noted, however, that the boundaries are by no means clear-cut and that there are always boundary instantiations. 7.3.2 Vertical schematization

I have described above how topic as a complex category extends horizontally. In what follows, my focus is on how the category schematizes vertically.

At the bottom tier of the network for topic, there are the different kinds of topics as seen in TOP, LD and various TC constructions. At the second tier, I identify three sub-categories, respectively referred to as Argument Topics, Circumstantial Topics and

External Topics. Argument Topics mainly consist of topics fulfilling argument roles, such as those in TOPs and LDs, where the topics serve as the Reference Point for an entity in the comment clause or the clausal process.

Circumstantial Topics cover topics functioning to orient the clausal process to some situational factors, such as time, place, manner, etc. Once occurring clause-initially, they gain prominence and serve as an initial focus of attention.

At the boundary of Argument Topics and Circumstantial Topics are the topics in double-subject constructions and RESPECT topics. When the topic in a double-subject construction has a stronger claim for subjecthood, it may be taken as an instance of

Argument Topics; otherwise, it falls under the category of Circumstantial Topics. Given the fact that the great majority of double-subject constructions are conventionalized expressions, i.e., most of them resemble the prototype topics, more instantiations would fall under the category of Argument Topic rather than Circumstantial Topics. RESPECT topics operate in a similar manner: when the respect indicated by the topic is closely

255 related to the subject, say in a modifier-modified relationship, the topic may be classified as an argument topic; when the relationship is a loose one, it might be better to be taken as a circumstantial topic. As far as this corpus is concerned, the majority of RESPECT topics are only loosely related to the subjects they invoke, which implies that the category as a whole is best taken as falling under the subcategory of Circumstantial Topics.

External Topics include INTERPERSONAL and TEXTUAL topics. Like Circumstantial

Topics, the topics in this category highlight a particular facet with respect to which the clausal process is to be interpreted, but they are external to the clausal proposition per se.

Hence is the name of the sub-category. INTERPERSONAL topics are concerned either with the speaker's stance on the clausal process or the relatedness of the present clause to previous discourse. In the former case, the INTERPERSONAL topic shares something in common with Circumstantial Topics. In the latter case, the meaning conveyed by the topic expression is completely external to the proposition of the host clause; the topic serves just as a super-clausal discourse linker. TEXTUAL topics are homogeneous in discourse function, unanimously indicating how the host clause is related to the preceding clause(s). In this light, the defining feature of External Topics is that the topic does not contribute to the propositional meaning of the host clause. Topics in this subcategory prototypically refer the hearer to discourse organization at a more or less global level.

Above these three subcategories is the abstract schema Topic. On the basis of what has been discussed so far, the schema has the following features. As an overarching one, a topic expression is a Reference Point for its host clause, with the type of Reference

256 Point being determined by its semantic meaning. Then the topic should occur early in the

clause: if not in the initial position, it should precede the comment or the information

nucleus, serving as a Reference Point either for a core argument of the clause or the

clausal process as a whole. As a derived feature, it may take marking, either materialized

or not, which functions to enhance the topic status of the marked expression. In extreme

cases, indefinite referents may be used as topics, so long as the speaker can conceptualize

them as plausible Reference Points and mark them as identifiable. Also, topichood may

be realized more than once in one single clause. There are at least two modes of multiple

topic organization, the chain mode and the network mode.

The schema of topic as delineated above is diagrammatically illustrated in Figure 7.1.

At Tier 1 are topics that are represented by the various semantic-functional roles that they

fulfill in discourse; Tier 2 consists of the three subcategories as elaborated above; and

Tier 3 is the schematic interpretation of Mandarin topics as Reference Points. It should be

noted that the possessor/modifier and respect categories are two boundary cases.

Although topics functioning as possessors and modifiers are not arguments per se, they

are entities that are closely related to arguments. Therefore, they fall more under the

subcategory of Argument Topics than under the other two subcategories. RESPECT topics, as mentioned above, may in some non-prototypical cases be interpreted as possessors or modifiers; that is, they are also related to Argument Topics, but not in a major way. A dotted line is therefore used to show this secondary relatedness.

257 Figure 7.1 Schema Network of Topic as Conceptual Reference Point

Tier 3 Topics

Tier 2

Her 1 possessor subject respect time place manner inter­ textual object /modifier personal

bO The schema as proposed here extends Langacker's characterization in two dimensions. First, the semantic meaning of the topic expression has been extended to

include more than entities; interpersonal and textual elements are also taken as instantiations of the schema. Second, the construal of some topic-comment constructions is not entirely subjective. At least, many double-subject constructions and utterances with multiple topics that are organized in the chain mode are unlikely to be construed in a different way, in that the occurrence of one topic expression presupposes early occurrence of another.

7.4 Summary

In this chapter, I have proposed a unified interpretation of all types of topics as

Conceptual Reference Points in discourse. The chapter begins with a detailed description of the diverse semantic meanings that topic expressions convey in the spoken corpus and how they function as Reference Points. Then multiple topic marking is discussed, with the finding that topic marking can be done either in the chain or the network mode.

Finally the schematization of topic as a complex linguistic category is expounded first horizontally then vertically. Topic as a schema is then characterized as follows: topics of all semantic roles may fulfill the function of Conceptual Reference Point in online speech, and the Reference Point may sometimes be objective, especially when the relationship between a topic and its target resembles some objective ones, such as the possessor-possessed relationship.

259 Chapter 8: Conclusion

This chapter concludes the corpus study of topic in Beijing Mandarin by

summarizing the findings in the preceding chapters and suggesting a number of areas for

further study.

8.1 A Brief Review

Mandarin Chinese has long been accepted as a topic-prominent language. However,

a consensus has never been reached on how to define or characterize the notion of topic.

Given the pragmatic orientation of the notion and the heterogeneity of topic-comment

constructions, a cognitive-functional approach has been adopted by this research to

analyze a spoken corpus of Beijing Mandarin. On the basis of what has been found about

structurally identified topics in the corpus, I have argued that topic constructions are best

taken as specific instantiations of the schema of Conceptual Reference Point for the

ensuing comments.

Spoken data are more indicative of online processing than are written texts. Taking

advantage of this, I have examined how pauses and pause particles are interrelated with topic marking and topic verbalization. A comparison between marked grammatical

subjects and unmarked ones has revealed that the former group is more heavily coded.

Following the iconicity principle or the accessibility theory as proposed by Givon (1983),

Ariel (1988, 1990), etc., topics convey more information than subjects. This finding also supports the argument that pause is a reliable demarcation between topic and subject (e.g.

260 Shi 1995, Xu and Liu 1998:283).

The investigation into the discourse use of several high frequency pause particles,

namely ne, ba, ma, a and its phonetic variants, and two lexical topic markers, i.e., dehua

and laijiang, shows that they, like Japanese wa or eyebrow raise in ASL, are all

polysemous: they fulfill different discourse functions by marking topics of different

semantic roles and information statuses. They are also observed to occur successively and

co-occur with each other in topic marking. The richness in topic markers and topic marking manners further evidences topic as a complex category that demands a schematic characterization.

Following detailed analysis of topic instantiations in the spoken corpus, a schematization of topic as a complex linguistic category is expounded first horizontally then vertically in Chapter 7. Topic as a schema is then characterized as follows: topics of all semantic roles may fulfill the function of Conceptual Reference Point in spontaneous speech, and the Reference Point may sometimes be objective, especially when the relationship between topic and its target referent(s) resembles some objective ones, such as the possessive or the part-whole relationship.

8.2 Revisiting the Controversies over Mandarin Topics

In Chapter 2, four controversies over Mandarin topics have been delineated, namely how topic is related to subject, topic as a syntactic or pragmatic notion, what the central and marginal properties of topic are, and how it is identified. Given the cognitive

261 approach in general and Langacker's interpretation of topic in particular, some of the controversies have been resolved. First, topic is technically defined by Langacker

(1991:314) as a kind of "subjective reference point" for the ensuing comment, where

"subjective" has a double meaning (see Chapter 2). It should be noted, however, that this research has shown that the reference point may also be "objective" in some TC constructions, such as double-subject constructions, because the topic and subject in many double-subject construction are in a possessive or part-whole relationship, which

Langacker himself classifies as "objective" (1991:313).

Another controversy resolved by the cognitive approach itself is over the syntactic or pragmatic status of the notion. Langacker unambiguously claims that "whereas subject is a clause-level construct, topics are clearly a discourse phenomenon" (1991:313). In his

1998 paper, he expounds that topics can be found at various levels of organization, from discourse topics in Brown and Yule's (1983) terminology to clause-internal ones. He also points out that topics and subjects may converge at the clause level, but still represent distinct functions: the Reference Point role of the subject, unlike that of the topic, is

"independent of any particular discourse status". Therefore, topic is maintained exclusively as a discourse or pragmatic notion.

When it comes to the controversy over topic identification, this research, I think, has made a meaningful attempt. It is obvious that my identification of topic is structure-based and tailored for spoken data where topic markers abound. In the meantime, topic identification as practiced in this research aligns with the guiding principles of Cognitive

262 Grammar for topic identification as well. Langacker maintains that topic can be

established in.two often observed means, namely special marking and grammatical

construction. In addition to these two coding features of topics, Langacker also grants

semantics the status as the most important topicality factor, and hints at the possible

semantic relationships between topic and the clause-internal entity (e.g. class membership)

(see Chapter 2 for details). Topic marking is the theme of the whole study, Chapter 6 in particular, and Chapter 5 has focused on three grammatical constructions, with the

semantic role of the marked topic being a constant concern throughout the corpus

analysis and results interpretation. Although modifications may be necessary for this coding property based method to be applied to written texts, topics as identified in this corpus study constitute a category which, broad as it is, can be uniformly accounted for with the schema of Conceptual Reference Point for the ensuing comment.

As for the controversy over the central and marginal properties of Mandarin topics, the linguistic data seem to point to identifiability as the central property of all topic expressions. It has been shown in previous chapters that elements conveying discourse-new or indefinite information may also occur clause-initially and function as topics. Section 6.2.2 has argued with examples from the corpus that discourse-old information is not the only type of information that can be topicalized, and that definiteness is by no means an obligatory feature of marked topics, especially given the fact that the ACTION/ACTIVITY and EVENT semantic categories are a crosscutting phenomenon of topics marked by all high-frequency topic markers in the corpus.

263 Likewise, the elements fulfilling the textual and interpersonal metafunctions may also

resist characterization in terms of "definiteness"; they are just members of more or less

closed sets which constitute part of the speaker and listener's world knowledge. The same

is true with proper names and the chronological organization of narratives as well. For

instance, when John is produced as topic by the speaker, the listener needs to know who

the referent is. Otherwise, he or she needs to negotiate with the speaker until the referent

becomes "identifiable". Therefore, the central property of topic should be identifiability.

However, by establishing identifiability as the central property of topic is not to

exclude other properties altogether; instead, properties such as definiteness, being old in

information status, etc., are also relevant in characterizing topics. A prototypical topic

should be definite in reference and given in information status. This has been evidenced

repeatedly in previous chapters, especially in Chapter 6. For example, ENTITY topics far

outnumber ACTION/ACTIVITY or EVENT topics, which implies that the great majority of

topics are definable in terms of definiteness. Also, topics conveying given information are

also much more frequently observed than those conveying new or indefinite information.

These findings not only explain why definiteness, given information, etc. are taken by many researchers to be the defining properties of topics, but also provide support for

Cognitive Grammar's distinction between the function fulfilled by topic and subject respectively: given the intrinsic nature of subject and object roles to a clausal process, they are more often than not the Reference Point for accessing the relationship profiled by the clause, which explains why the number of marked ENTITY topics is the largest

264 among the frequency counts of all topic categories. By contrast, when the semantic role

of a topic becomes more contingent, i.e., more contextually determined, to the clausal

process, its frequency of occurrence tends to decrease.

In short, a cognitive approach to naturally occurring language data has resolved many

of the controversies over Mandarin topics.

8.3 Conceptual Reference Point as a Conceptual Archetype

From the previous section, it can be seen that the Reference Point model has provided

a favorable account for Mandarin topics, and the model seems very powerful in

explaining complex linguistic phenomena such as topic and subject. In fact, a reference

point, to Langacker, is a construct which enjoys similar psychological significance to

notions such as force dynamics, image schemas, subjective vs. objective construal, and

correspondences across cognitive domains and mental spaces, etc., and it is fundamental

to both linguistic and cognitive organization (1993:1-2). In the same paper, Langacker

also argues that possessive constructions, quasi-possessive constructions (e.g. the mere juxtaposition of the possessor and the possessed as in Papago, a native American

language, and constructions where two noun phrases appear to have equal claim to

subject status,73 etc.), and non-possessive constructions, such as topic, can all be

understood as Reference Point phenomena. He also presents a Reference Point

interpretation of metonymy (1993:30):

265 ...metonymy is basically a reference-point phenomenon. More precisely, the entity that is normally designated by a metonymic expression serves as a reference point affording mental access to the desired target (i.e., the entity actually being referred to).

Moreover, van Hoek (1997) represents a comprehensive study of anaphora in terms of

Conceptual Reference Point, and this research consists of some practical work on topic in one particular language, i.e., Beijing Mandarin.

If the above mentioned linguistic phenomena are classified, there can be roughly two categories, those of syntagmatic manifestations and those of paradigmatic manifestations.

Under the former category are the possessives, quasi-possessives and topics, and under the latter category is metonymy, with anaphora being the former when within the sentence boundary and the latter when across the sentence boundary. Since so diverse a range of linguistic phenomena can all find satisfactory explanation following the

Reference Point model, the model is most probably significant enough to be a conceptual archetype, especially given the fact that Langacker has noted towards the end of his 1993 paper that most broadly, novel experience is interpreted with reference to previous experience.

8.4 Areas of Further Study

This study has brought into light a unified cognitive account of different types of topics in Beijing Mandarin. However, many areas have been identified for further study.

1. The present study has examined a range of high-frequency topic markers in the spoken corpus. However, there are still many others that remain unaddressed, such as the

266 particle ha, the frame topic marker jiu...(lai)shuo 'as far as .... is/are concerned', some

topic introducing verbs, etc. They occur less frequently; nonetheless, they function to

mark topics in a more or less similar manner to those discussed in Chapter 5 and 6. For

instance, the verb wen 'ask' in (1) functions to introduce the entity wofuqin 'my father'

into discourse.

(1) M^*, ffe^HMS mmWLWt&^ltio (064-L26:2)

Wen wo fuqin, ta shi nage, na-shihour jiu zai chezhan.

Ask my father he is PF that-time just at station

'If you ask about my father, he was working at the station at that time.'

This sentence occurs at the beginning of a paragraph, where the speaker shifts the

discourse topic from her own family to her father. Keenan and Schieffelin (1976) have used the term "directives" to refer to this kind of construction and "locating verbs" for the verbs involved. They note that directives are commonly used by speakers to bring a referent into discourse (1976:249). Therefore, the topic marker status of such expressions is beyond doubt. The issue is how many such expressions are used in Beijing Mandarin and how grammaticalized they are through discourse use.

2. The second area for further study is the relationship between topic marking particles and sentence final particles. It is not difficult to see from the examples used throughout the dissertation that most, if not all, of the topic-marking pause particles are also used sentence finally as mood-indicating particles. It has been noted in Chapter 3 that some researchers, such as Fang (1994), maintain that pause particles in Beijing

267 Mandarin originate from sentence-final particles, but how the topic-marking function of a particle is related to its mood-indicating function remains to be uncovered. Chao

(1968:795-814) discusses the use of a number of particles, but does not elaborate how sentence-middle particles are related to sentence-final ones. Li and Thompson (1981) have devoted a whole chapter (i.e., Chapter 7) to sentence-final particles and conclude that the function of such particles is "to relate to the conversational context in various ways the utterance to which they are attached and to indicate how this utterance is to be taken by the hearer" (1981:317). A comparative study of sentence-middle and sentence-final particles, especially those that can occur in both contexts, may throw some light on the relationship between the topic-marking and the mood-indicating functions of particles.

Chapter 6 has suggested that ba should be best taken as a phonetic variant of a, because, except for a higher frequency of occurrence, ba does not differ systematically from other particles in the a-set. While the phonological context may have some bearing on the distribution pattern of this set of particles, a deeper understanding of their interrelatedness may lie with a comparative study as sketched above.

3. Another area deserving of further study is the information status of subjects.

Chapter 5 has shown that topic-marked grammatical subjects are more complex in coding form than unmarked subjects. Following the iconicity principle, the former should be heavier in information load than the latter. This corollary is then in conflict with a widely held belief that topics tend to convey given information whereas subjects can convey

268 both given and new information. In fact, many observations, such as Chafe's (1987, 1994)

Light Subject Constraint, Du Bois' (1987) Given A Constraint, Lambretch's (1994)

Principle of Separation of Reference and Role, etc., all point to a "light subject". What is lacking then seems to be an empirical study on the information status of subjects. The fact that topics also convey new information may facilitate a better understanding of the information status of subject.

4. It has been made clear in the methodology chapter that the present corpus is available in the form of transcripts; that is, no audio-visual clues can be relied upon for analytical purposes. If such clues were available, many more observations could have been made. For instance, the length of a pause would be very useful in revealing how much cognitive effort is involved in coding an expression as a topic. Also, eyebrow raise, a facial expression, has been shown to be a topic marker in ASL (Janzen 1998, 1999).

Although ASL is a signed language, verbal communication by means of spoken languages is never exempted from facial expressions or body gestures. Therefore, to incorporate audio and visual information into speech analysis would be very enlightening.

5. Last but not least, the present study is limited to narrative monologues only. Even though the findings are informative about topic expressions in Beijing Mandarin, investigation into a wider range of genres, interactive genres included, would permit a closer look at the linguistic construction under discussion. At least, topic negotiation, which is missing in monologues, would be observable in conversations. Therefore, a

269 more varied corpus would bring us closer to the essence of the phenomenon that we are interested in.

8.5 Summary

In this chapter, I first highlight the major findings of the present research. Following this, the four controversies over Mandarin topics are revisited. After reviewing a range of linguistic phenomena that can be successfully explained in terms of Conceptual

Reference Point, I suggest that this research on a particular construction in a particular language provides support for Langacker's proposal that Reference Point is a conceptual archetype. Finally, several areas are suggested for further study on Mandarin topics.

270 Appendix: Characteristics of Monologs Comprising the Corpus nation­ education code age gender district/area profession topic ality level 001-N07 84 m Hui Niujie primary community cadre E 008-L50 71 m Han Lugouqiao primary driver D 015-D25 68 m Man Dongcheng primary technician D 022-H44 65 m Han Haidian primary shop assistant D technical primary school 029-X36 64 f Man Xicheng secondary F teacher school 036-N44 62 m Hui Niujie primary bus conductor C 043-N50 60 m Han Niujie senior high bus conductor C 050-N14 60 f Han Niujie senior high official F 057-H20 58 f Han Haidian senior high teacher B 064-L26 57 f Han Lugouqiao senior high doctor B 071-D57 56 f Man Dongcheng senior high policeman A 078-T50 55 m Hui Tianqiao junior high driver B 085-H14 55 f Han Haidian junior high official B 092-L44 54 m Han Lugouqiao primary shop assistant D 099-T27 53 m Han Tianqiao college doctor F technical 106-X14 53 f Han Xicheng secondary official D school 113-D21 52 m Man Dongcheng university engineer C 120-H15 50 m Han Haidian junior high official B 127-L27 48 m Han Lugouqiao junior high doctor B 134-D09 47 m Han Dongcheng senior high official B 141-X03 45 m Han Xicheng junior high worker C 148-N21 44 m Han Niujie university teacher F 155-L03 43 m Han Lugouqiao junior high worker F primary school 162-H37 42 m Han Haidian junior high D teacher 169-T58 41 m Han Tianqiao senior high policeman C primary school 176-N38 40 f Han Niujie senior high B teacher 183-X46 38 m Han Xicheng junior high shop assistant A 190-D16 38 f Han Dongcheng junior high official C 197-X22 37 f Han Xicheng university teacher D

271 nation­ education code age gender district/area profession topic ality level 204-L53 36 f Han Lugouqiao senior high bus conductor D 211-N52 35 m Hui Niujie junior high driver C 218-T59 35 f Han Tianqiao junior high policeman C 225-D47 35 f Han Dongcheng junior high shop assistant F 232-T17 34 m Han Tianqiao college official E 239-N06 34 f Hui Niujie junior high worker E 246-L23 31 m Han Lugouqiao university high school teacher A 253-T54 30 m Han Tianqiao junior high bus conductor B primary school 260-X39 30 m Han Xicheng senior high F teacher 267-H24 30 f Han Haidian university teacher B 274-T30 30 f Han Tianqiao senior high doctor F primary school 281-H40 29 f Han Haidian junior high D teacher technical primary school 288-N40 28 f Han Niujie secondary B teacher school 295-L12 27 f Han Lugouqiao junior high community cadre F 302-T49 26 f Han Tianqiao senior high shop assistant B 309-X12 25 f Han Xicheng college community cadre C (adapted from List of Language Informants of A Corpus of Modern Spoken Beijing

Mandarin)

272 References

Ariel, Mira. 1988. Referring and accessibility. Journal of Linguistics 24: 65-87.

Ariel, Mira. 1990. Accessing noun-phrase antecedents. London: Routledge.

Blakemore, Diane. 1987. Semantic Constraints on Relevance. Oxford: Blackwell.

Brown, Gillian and George Yule. 1983. Discourse Analysis. Cambridge: Cambridge

Universigy Press.

Bybee, Joan, and Paul Hopper. 2001. Introduction. In Joan Bybee and Paul Hopper (eds.)

Frequency and the Emergence of Linguistic Structure, 1-24. Amsterdam/

Philadelphia: John Benjamins Publishing Company.

Bybee, Joan. 2002. Word frequency and context of use in the lexical diffusion of

phonetically conditioned sound change. Language Variation and Change 14.

261-290.

Bybee, Joan. 2003. Cognitive processes in grammaticalization. In Michael Tomasello (ed.)

The New Psychology of Language Vol. 2,145-168. Mahwah/London: Lawrence

Erlbaum Associates.

Chafe, Wallace L. 1976. Givenness, contrastiveness, definiteness, subjects, topics, and

point of view. In Charles N. Li (ed.) Subject and Topic, 25-55. New York: Academic

Press.

Chafe, Wallace L. 1979. The flow of thought and the flow of language. In Talmy Givon

(ed.) Discourse and Syntax. Syntax and Semantics. Vol. 12, 159-181. New York:

Academic Press.

273 Chafe, Wallace L. 1987. Cognitive constraints on information flow. In Russell Tomlin

(ed.) Coherence and Grounding in Discourse, 21-51. Amsterdam/Philadelphia: John

Benjamins.

Chafe, Wallace L. 1992. Information flow. In William Bright (ed.) Oxford International

Encyclopaedia of Linguistics, Vol. 2, 2215-218. Oxford: Oxford University Press.

Chafe, Wallace L. 1994. Discourse, Consciousness, and Time: The Flow and

Displacement of Conscious Experience in Speaking and Writing. Chicago:

University of Chicago Press.

Chao, Yuen-ren. 1968. .4 Grammar of Spoken Chinese. Berkeley/Los Angeles: University

of California Press.

Chappell, Hilary. 1996. Inalienability and the personal domain in Mandarin Chinese

discourse. In Hilary Chappell and William McGregory (eds.) The Grammar of

Inalienability: A Typological Perspective on Body Part Terms and the Part-Whole

Relation, 465-527. Berlin/New York: Mouton de Gruyter.

Chen, Ping. 1996. Pragmatic interpretations of structural topics and relativization in

Chinese. Journal of Pragmatics 26:389-406.

Chen, Ping. 2004. Hanyu shuangxiang mingci ju yu huati-chenshu jiegou (Double

nominal constructions and topic-comment constructions in Mandarin). Zhongguo

yuwen (Journal of Chinese Philology) Vol. 6:493-507.

Chu, Chauncey C. 1993. The prototypicality of topic in Mandarin Chinese. JCLTA

28.1:25-48.

274 Chu, Chauncey C. 1998. A Discourse Grammar of Mandarin Chinese. New

York/Washington: Peter Lang.

Cruse, Allen. 2006. A Glossary of Semantics and Pragmatics. Edinburge: Edinburgh

University Press.

Dixon, Robert M.W. 1979. Ergativity. Language 59:59-138.

Du Bois, John W. 1980. Beyond definiteness: The trace of identity in discourse. In

• Wallace L. Chafe (ed.) The Pear Stories: Cognitive, Cultural, and Linguistic

Aspects of Narrative Production, 203-274. Norwood: Ablex.

Du Bois, John W. 1987. The discourse basis of ergativity. Language 63.4:805-855.

Erteschik-Shir, Nomi. 2007. Information Structure. Oxford: Oxford University Press.

Evans, Vyvyan. 2007. A Glossary of Cognitive Linguistics. Edinburgh: Edinburgh

University Press.

Fang, Mei. 1994. Beijing kouyu zhong yuqici de gongneng yanjiu (A functional study of

pause particles in spoken Beijing Mandarin). Zhongguo yuwen (Journal of Chinese

Philology) 239.2:129-138.

Ford, Cecilia E., Barbara A. Fox, and Sandra A. Thompson. 1998. Social Interaction and

Grammar. In Michael Tomasello (ed.) The New Psychology of Language Vol. 2,

119-143. Mahwah/London: Lawrence Erlbaum Associates.

Fraser, Bruce. 1988. Types of English discourse markers. Acta Linguistica Hungarica

38:19-33.

Geeraerts, Dirk. 2008. A rough guide to cognitive linguistics. In Dirk Geeraerts (ed.)

275 Cognitive Linguistics: Basic Readings. Berlin/New York: Mouton de Gruyter.

Givon, Talmy (ed.). 1983. Topic Continuity in Discourse: A Quantitative Cross-Language

Study. Amsterdam/Philadelphia: John Benjamins Publishing Company.

Givon, Talmy. 1995. Functionalism and Grammar. Amsterdam/Philadelphia: John

Benjamins Publishing Company.

Gregory, Michelle, and Michaelis, Laura. 2001. Topicalization and left-dislocation: A

functional opposition revisited. Journal of Pragmatics 33:1665-1706.

Gundel, Jeanette. 1985. 'Shared knowledge' and topicality. Journal of Pragmatics

9:83-107.

Halliday, Michael A.K. 1985/1994. An Introduction to Functional Grammar. London:

Edward Arnold.

Hinds, John. 1983. Topic continuity in Japanese. In Talmy Givon (ed.) Topic Continuity

in Discourse: A Quantitative Cross-Language Study, 43-93. Amsterdam/

Philadelphia: John Benjamins Publishing Company.

Hinds, John, Senko K. Maynard, and Shoichi Iwasaki J. 1987. Perspectives on

Topicalization The Case of Japanese Wa. Amsterdam/Philadelphia: John Benjamins

Publishing Company.

Hirano, Takanori. Subject and topic in Tagalog. (http://homepage.ntu.edu.tw/

~gilntu/data/workshop%20on%20Austronesian/2%20hirano.pdf. retrieved on Nov.

18th, 2006)

Ho, Yong. 1993. Aspects of Discourse Structure in Mandarin Chinese. Lewiston: Mellen

276 University Press.

Hockett, Charles F. 1958. A Course in Modern Linguistics. New York: Macmillan.

Hopper, Paul. J. 1998. Emergent grammar. In Michael Tomasello (ed.) The New

Psychology of Language Vol. 1, 155-175. Mahwah: Lawrence Erlbaum Associates.

Hopper, Paul. J, and Sandra A. Thompson. 1980. Transitivity in grammar and discourse.

Language 56:251-299.

Huang, C.-T. James. 1982. Logical Relations in Chinese and the Theory of Grammar.

PhD. Dissertation. Cambridge Mass: MIT.

Janzen, Terry. 1998. Topicality in American Sign Language. PhD Dissertation, University

of New Mexico.

Janzen, Terry. 1999. The grammaticization of topics in American Sign Language. Studies

in Language 23.2:271-306.

Keenan, Edward L. 1976. Towards a universal definition of "subject". In Charles N. Li

(ed.) Subject and Topic, 303-333. New York: Academic Press.

Keenan, Edward L., and Bambi Schieffelin. 1976. Foregrounding referents: a

reconsideration of left dislocation in discourse. Berkeley Linguistic Society

2:240-257.

Lambrecht, Knud. 1994. Information Structure and Sentence Form. Cambridge:

Cambridge University Press.

Langacker, Ronald W. 1987. Foundations of Cognitive Grammar, Vol. 1. Stanford:

Stanford University Press.

277 Langacker, Ronald W. 1987b. Nouns and verbs. Language 63:53-94.

Langacker, Ronald W. 1990. Concept, Image, and Symbol. The Cognitive Basis of

Grammar. Berlin/New York: Mouton de Gruyter.

Langacker, Ronald W. 1991. Foundations of Cognitive Grammar, Vol. 2. Stanford:

Stanford University Press.

Langacker, Ronald W. 1993. Reference-point constructions. Cognitive Linguistics 4:1-38.

Langacker, Ronald W. 1998. Topic, subject, and possessor. Linguistic Notes from La Jolla

19:1-28.

Langacker, Ronald W. 1999. Grammar and Conceptualization. Berlin and New York:

Mouton de Gruyter.

Langacker, Ronald W. 2001. Discourse in Cognitive Grammar. Cognitive Linguistics

12-2:143-188.

Langacker, Ronald W. 2007. Ten Lectures on Cognitive Grammar. Beijing: Foreign

Language Teaching and Research Press.

Langacker, Ronald W. 2008. Cognitive Grammar. In Dirk Geeraerts (ed.) Cognitive

Linguistics: Basic Readings, 29-67. Berlin/New York: Mouton de Gruyter.

Levelt, Willem J.M. 1989. Speaking: From Intention to Articulation. Cambridge: The

MIT Press.

Li, Charles N., and Sandra A. Thompson. 1976. Subject and topic: A new typology of

language. In Charles N. Li (ed.) Subject and Topic, 457-489. New York: Academic

Press.

278 Li, Charles N., and Sandra A. Thompson. 1981. Mandarin Chinese: A Functional

Reference Grammar. Berkeley/Los Angeles: University of California Press.

Liu, Linjun. 2007. Topicalization and Left Dislocation in Beijing Mandarin: a case study.

Generals paper. Winnipeg, MB: University of Manitoba.

Lii, Shuxiang. 1986. Hanyu jufa de linghuoxing (On the flexibility of Mandarin syntax).

Zhongguo yuwen (Journal of Chinese Philology) Vol. 1:1-9.

Maslova, Elena, and Giuliano Bernini. 2006. Sentence topics in the languages in Europe

and beyond. In Giuliano Bernini and Marcia L. Schwarz (eds.) Pragmatic

Organization of Discourse in the Languages of Europe, 67-120. Berlin/New York:

Mouton de Gruyter.

Nikolaeva, Irina. 2001. Secondary topic as a relation in information structure. Linguistics

39-1:1-49.

Prince, Ellen. 1981. Toward a taxonomy of given-new information. In Peter Cole (ed.)

Radical Pragmatics, 223-255. New York: Academic Press.

Prince, Ellen. 1984. Topicalization and left-dislocation: A functional analysis. Annals of

the New York Academy of Sciences 433:213-225.

Prince, Ellen. 1997. On the function of left-dislocation in English discourse. In: Akio

Kamio, ed., Directions in Functional Linguistics, 117-143. Philadelphia: John

Benjamins.

Richards, Jack C, John Piatt, and Heidi Piatt. 2000. Longman Dictionary of Language

Teaching & Applied Linguistics. Beijing: Foreign Language Teaching and Research

279 Press.

Sanders, George. 1984. Adverbials and objects. In Frans Plank (ed.) Objects: towards a

theory of grammatical relations, 221 -241. London: Academic Press.

Shei, Chris. 2005. Analysing Chinese Sentence-Final Particles Using Academia Sinica

Balanced Corpus of Modern Chinese. Proceedings from the Corpus Linguistics

Conference Series Vol. 1, No. 1 Corpus Linguistics.

(www.corpus.bham.ac.uk/PCLC)

Shi, Dingxu. 2000. Topic and topic-comment constructions in Mandarin Chinese.

Language 76:383-408.

Shi, Youwei. 1995. Zhuyu hou tingdun yu huati (Post-subject pause and topic). Zhongguo

Yuyan Xuebao (Linguistic Transactions of China) 5:97-123. Beijing: The

Commercial Press.

Shi, Yuzhi. 2001. Yufa de Xingshi he Liju (Forms and Motivations of Chinese Grammar).

Nanchang: Jiangxi jiaoyu chubanshe (Jiangxi Education Press).

Shibatani, Masayoshi. 1991. Grammaticization of topic into subject. In Elizabeth C.

Traugott and Bernd Heine (eds.) Approaches to grammaticalization 2, 93-133.

Amsterdam/Philadelphia: John Benjamins Publishing Company.

Tan, Pack-Ling, and Hongyin Tao. 1998. Taiwan guanhua xushiti zhong yunlii danwei de

yufa goucheng qiji guilu chutan (Patterns of syntax-prosody correspondence in

Taiwan Mandarin narrative discourse). Yuyan Yanjiu (Language Studies) 34:1-22.

Tao, Hongyin. 1996. Units in Mandarin Conversation. Amsterdam/Philadelphia: John

280 Benjamins Publishing Company.

Tao, Hongyin. 1998. A preliminary study of patterns of syntax-prosody mismatch in

Mandarin narrative discourse. In S. F. Huang (ed.), Selected Papers from the

Second Conference on Languages in Taiwan, 183-205. Taipei: Crane.

Taylor, John. 1998. Syntactic constructions as prototype categories. In Michael Tomasello

(ed.) The New Psychology of Language Vo\. 1, 177-202.

Taylor, John. 2002. Cognitive Grammar. Oxford: Oxford University Press.

Teng, Shou-hsin. 1974. Double nominatives in Chinese. Language 50:455-473.

Tomasello, Michael. 2003. Constructing a Language. Cambridge: Harvard University

Press.

Tsao, Feng-fu. 1979. A Functional Study of Topic in Chinese: The First Step toward

Discourse Analysis. Taipei: Student Book Co.

Tsao, Feng-fu. 1990. Sentence and Clause Structure in Chinese: A Functional Perspective.

Taipei: Student Book Co.

van Hoek, Karen. 1997. Anaphora and Conceptual Structure. Chicago and London: The

University of Chicago Press, van Hoek, Karen. 2003. Pronouns and point of view: cognitive principles of coreference.

In Michael Tomasello (ed.) The New Psychology of Language Vol. 2, 169-194.

Xu, Liejiong, and D. Terence Langendoen. 1985. Topic structures in Chinese. Language

61:1-27.

Xu, Liejiong, and Liu Danqing. 1998. Huati de jiegou yu gongneng (Structure and

281 Function of Topic). Shanghai: Shanghai jiaoyu chubanshe (Shanghai Education

Press).

Xu, Yulong. 1996. Hanying pianzhang zhong juzi zhuti de shibie (Topic identification in

English and Chinese discourse). Waiguo Yu (Foreign Languages) 106:3-9.

Yuan, Yulin. 1996. Huatihua ji xiangguan de yufa guocheng. (Topicalization and related

grammatical process). Zhongguo Yuwen (Chinese Philology) 253:241-254.

Yuan, Yulin. 2002. Hanyu huati de yufa diwei he yufahua chengdu—jiyu zhenshi ziran

kouyu de gongshi he lishi kaoliang (Grammatical status and degree of

grammaticalization of topic in Chinese: Synchronic and diachronic considerations

based on real and natural spoken data). Yuyanxue Luncong (Linguistic Discussion)

25:82-115. Beijing: Shangwu yinshuguan (The Commercial Press).

Zhang Bojiang, and Fang Mei. 1994. Hanyu Gongneng Yufa Yanjiu (A Functional

Grammar of Chinese). Nanchang: Jiangxi jiaoyu chubanshe (Jiangxi Education

Press).

Zhang Meilan, and Chen Siyu. 2006. Qimo michu Beijing kouyu zhong de huati biaoji

(Topic markers in the Beijing dialect during the period between the end of Qing

Dynasty and the beginning of the Republic of China). Shijie Hanyu Jiaoxue

(Chinese Teaching in the World) 76:63-73.

Zhu Dexi. 1982. Yufa Jiangyi (Lectures on Grammar). Beijing: Shangwu yinshuguan

(The Commercial Press).

282