Abstract Entity Anaphora in Argumentative Texts: Pragmatic Features and Referent Interpretation [1st ed. 2023] 9819946298, 9789819946297

Citation preview

Donghong Liu

Abstract Entity Anaphora in Argumentative Texts Pragmatic Features and Referent Interpretation

Abstract Entity Anaphora in Argumentative Texts

Donghong Liu

Abstract Entity Anaphora in Argumentative Texts Pragmatic Features and Referent Interpretation

Donghong Liu Southeast University Nanjing, China

ISBN 978-981-99-4629-7 ISBN 978-981-99-4630-3 (eBook) https://doi.org/10.1007/978-981-99-4630-3 Jointly published with Central China Normal University Press The print edition is not for sale in China (Mainland). Customers from China (Mainland), please order the print book from: Central China Normal University Press. ISBN of the Co-Publisher’s edition: 978-7-5622-3812-6 © Central China Normal University Press 2023 This work is subject to copyright. All rights are solely and exclusively licensed by the Publisher, whether the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now known or hereafter developed. The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication does not imply, even in the absence of a specific statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use. The publishers, the authors, and the editors are safe to assume that the advice and information in this book are believed to be true and accurate at the date of publication. Neither the publishers nor the authors or the editors give a warranty, expressed or implied, with respect to the material contained herein or for any errors or omissions that may have been made. The publishers remain neutral with regard to jurisdictional claims in published maps and institutional affiliations. This Springer imprint is published by the registered company Springer Nature Singapore Pte Ltd. The registered company address is: 152 Beach Road, #21-01/04 Gateway East, Singapore 189721, Singapore

Contents

1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.1 The Definition of Abstract Entity Anaphora . . . . . . . . . . . . . . . . . . . . 1.2 Differences Between Abstract Entity Anaphora and Other Kinds of Anaphora . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.2.1 Abstract Entity Anaphora and Concrete Individual Anaphora . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.2.2 Abstract Entity Anaphora, Extended Reference and Text Reference . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.2.3 Abstract Entity Anaphora and Deep Anaphora . . . . . . . . . . . 1.2.4 Abstract Entity Anaphora and Associative Anaphora . . . . . . 1.3 The Necessity of the Study of Abstract Entity Anaphora in Argumentative Texts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.3.1 The Limitations of the Previous Study of Abstract Entity Anaphora . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.3.2 Characteristics of Argumentative Text . . . . . . . . . . . . . . . . . . . 1.3.3 Argumentative Text and Abstract Entity Anaphora . . . . . . . . 1.4 Research Purpose and Significance . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.5 Organization of the Book . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 Pragmatic Features of Abstract Entity Anaphora . . . . . . . . . . . . . . . . . 2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.2 The Written Argumentative Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.2.1 The Sources of the Argumentative Texts . . . . . . . . . . . . . . . . . 2.2.2 The Criteria of Data Collection and the Content of Data Analyses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.3 Pragmatic Functions of Abstract Entity Anaphors . . . . . . . . . . . . . . . 2.3.1 Recapitulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.3.2 Obscuration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.3.3 Demarcation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1 1 4 4 5 5 6 7 7 8 9 10 11 13 13 13 13 14 16 17 21 22

v

vi

Contents

2.4 Text Environment of Abstract Entity Anaphora . . . . . . . . . . . . . . . . . 2.4.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.4.2 Text Environment of English Abstract Entity Anaphora . . . . 2.4.3 Text Environment of Chinese Abstract Entity Anaphora . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.4.4 Comparisons Between English and Chinese Abstract Entity Anaphora . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.5 Pragmatic Tendency of Abstract Entity Anaphora . . . . . . . . . . . . . . . 2.6 Backwards Anaphora . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.7 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

24 24 25

3 Discourse Structure and Discourse Anaphora . . . . . . . . . . . . . . . . . . . . . 3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2 Rhetorical Structure Theory (RST) by Mann and Thompson . . . . . . 3.2.1 Main Idea . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2.2 Comment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.3 Rhetorical Structure by Fox . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.3.1 Main Idea . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.3.2 Comment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.4 Linguistic Discourse Model (LDM) by Polanyi . . . . . . . . . . . . . . . . . 3.4.1 Main Idea . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.4.2 Comment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.5 Other Related Studies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.6 Discourse Representation Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.6.1 Main Idea . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.6.2 Comment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.7 Taking Stock . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

39 39 40 40 42 43 43 46 47 47 49 51 53 53 56 57

4 SDRT and Abstract Entity Anaphora . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.2 Discourse Relations and Topic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.2.1 Discourse Relations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.2.2 Topic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.3 SDRS Construction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.3.1 Determining the Basic Constituents . . . . . . . . . . . . . . . . . . . . . 4.3.2 Constraints on Possible Sites for Constituent Attachment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.3.3 SDRS Updating . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.3.4 Constituent Revision . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.4 SDRT and Abstract Entity Anaphora . . . . . . . . . . . . . . . . . . . . . . . . . . 4.4.1 Constraints on Abstract Entity Anaphora . . . . . . . . . . . . . . . . 4.4.2 AVAILABILITY and Constituent Revision . . . . . . . . . . . . . . 4.4.3 More Application to Abstract Entity Anaphora . . . . . . . . . . . 4.5 Application of SDRT to Chinese Abstract Entity Anaphora . . . . . . . 4.6 SDRT and Backwards Anaphora . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

59 59 60 60 61 62 62

29 31 32 35 37

63 64 66 68 68 70 73 80 85

Contents

vii

4.7 Support from Cognitive Psychology . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.7.1 Situation Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.7.2 Resonance Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.8 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

87 88 89 90

5 Inadequacies of SDRT and Suggested Solutions . . . . . . . . . . . . . . . . . . . 5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.2 Inadequacies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.2.1 Discourse Relations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.2.2 Topic Construction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.2.3 The Role of the Topic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.2.4 Explicit Referent Identification . . . . . . . . . . . . . . . . . . . . . . . . . 5.2.5 Implicit Referent Identification . . . . . . . . . . . . . . . . . . . . . . . . . 5.2.6 Reference Ambiguity Resulting from Double Attachment Sites . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.2.7 Long-Distance Anaphora . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.3 Suggested Solutions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.3.1 Simplification of the Discourse Relations in SDRT . . . . . . . . 5.3.2 Topic Construction of Binary Structures . . . . . . . . . . . . . . . . . 5.3.3 Topic Dominating Role . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.3.4 Explicit Referent Interpretation . . . . . . . . . . . . . . . . . . . . . . . . 5.3.5 Proposition Abstraction for Implicit Referent . . . . . . . . . . . . 5.4 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

93 93 94 94 94 95 96 97 97 98 99 99 105 110 111 117 121

6 Processing of Reference Ambiguity and Long-Distance Anaphora . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.2 Solutions to Reference Ambiguity . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.2.1 Context and Intension . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.2.2 Tendency Investigation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.3 Long-Distance Anaphora . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.3.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.3.2 Theoretical Bases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.3.3 Hypothesis Formulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.3.4 Instances of Application . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.3.5 The Role of Semantic Chains in Referent Interpretation . . . . 6.4 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

123 123 123 123 125 135 135 136 139 143 153 154

7 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157 7.1 Achievements of This Study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157 7.2 Limitations and Suggestions for Future Research . . . . . . . . . . . . . . . . 159 Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161

Chapter 1

Introduction

1.1 The Definition of Abstract Entity Anaphora Abstract entities originally fall into a philosophical category, yet they have linguistic aspects and are discussed by such linguists as Lyons (1977), Asher (1993), Vendler (1967), Smith (2003) and Peterson (1997). But Lyons’ abstract entities correspond only in part to the other linguists’. Lyons classifies entities in the world into three kinds. First-order entities are physical objects which are publicly observable and relatively constant in their perceptual properties. Second-order entities mean events, processes, state-of-affairs etc. that occur or take place rather than exist. They can be observed and have a temporal duration. Third-order entities refer to such abstract entities as propositions that are outside space and time. The abstract entities proposed by Asher et al. cover not only third-order entities but also second-order entities. According to Asher (1993), abstract entities refer to propositions, concepts, facts and events, as can be seen in the following underlined parts. (1.1) John does not believe that [Mary is treating him fairly]. But Fred is certain of it. (proposition) (Asher, 1993: 226) (1.2) [Every Swiss farmer who owns a donkey beats it]i . But thati ’s not true of an Austrian farmer and any donkey he owns. (concept) (Asher, 1993: 249) (1.3) [No one heeded the government’s curfew orders]i . Thisi shows that the Junta can no longer control the people. (fact) (Asher, 1993: 245) (1.4) [The Ashers were predictably short of groceries the day of the party. Nicholas Asher went out to get some, got lost and arrived back only after the party had ended] i . Because of thisi , the committee made sure that the Ashers never gave party for the Society again. (event) (Asher, 1993: 234)

Those abstract entities in the above examples, except events, are pure abstract entities since they “have no spatio-temporal location, usually no causal efficacy and not perceived by the senses” (Asher, 1993: 1). Events, contrarily, have spatiotemporal location as well as causal efficacy. Despite this, events are closely related to abstract entities in that the sentences describing events may express propositions and facts. For example, the antecedent of (1.4) expresses the proposition “Nicholas failed © Central China Normal University Press 2023 D. Liu, Abstract Entity Anaphora in Argumentative Texts, https://doi.org/10.1007/978-981-99-4630-3_1

1

2

1 Introduction

to supply enough groceries for the party”. For that reason, Asher (1993) subsumes events into the category of abstract entities. In the continuum of abstract entities events have the lowest degree of abstractness while propositions the highest degree. The term “abstract entity anaphora” is initiated by Asher (1993) in his book Reference to Abstract Objects in Discourse. It refers to the anaphora whose antecedent is abstract entity. According to Chu (1998), the term “anaphora” is generally used to denote a device to refer back to what has been mentioned previously. Strictly speaking, the anaphors in most contexts do not actually refer back to the previous expressions, but rather refer to the same entities or concepts that the previous linguistic forms are used to refer to. So it is more accurate and reasonable to treat anaphors as coreferring with some previous expressions than as referring back to those expressions. As for abstract entity anaphora, the anaphors refer back to the propositions or concepts expressed by the antecedents instead of the linguistic forms of the antecedents (see the above examples). In this sense, abstract entity anaphora is counted as one kind of anaphora. Asher (1993) does not advocate making an accurate differentiation among the types of abstract entity anaphora. He holds the view that abstract entities are the products of the thought of human beings and have indefinite and vague limits between the different types. Apart from that, different types of abstract entities such as propositions and facts may be fused to constitute the antecedent, for example (1.5). Moreover, it is hard to differentiate the sentences directly expressing propositions from those conveying facts. For instance (1.1), if we remove the main clause containing the verb “believe” which typically expresses a proposition, the rest part “Mary is treating him fairly” may either express a proposition or a fact. For the above three reasons, Asher (1993) incorporates fact anaphora into proposition anaphora, which makes up the major type of abstract entity anaphora. (1.5) [The fact that Sam ran off with his secretary surprised no one. Mary’s expectation that he would be unfaithful is also a pretty grim commentary on his character]i . All of iti makes plain the typical, sad state of modern marriages. (Asher, 1993: 59)

Therefore, in this dissertation three kinds of abstract entity anaphora are discussed: proposition anaphora, event anaphora and concept anaphora. Generally speaking, verb phrases serve as the antecedents of concept anaphora while clauses, sentences and sentence clusters as the antecedents of proposition and event anaphora. According to the different linguistic forms and syntactic features of antecedents, Xu Yulong (许余龙) (2004) classifies anaphora into eight types:

1.1 The Definition of Abstract Entity Anaphora

3 noun anaphora

nominal anaphora noun phrase anaphora verb anaphora verb anaphora verb phrase anaphora clause anaphora sentential anaphora discourse segment anaphora adjective anaphora adverb anaphora

Abstract entity anaphora we are discussing here involves clause anaphora, discourse segment anaphora, verb phrase anaphora and some of noun phrase anaphora. Not all of the noun phrase anaphora is abstract entity anaphora. Only if the noun phrase contains a proposition, for example “the belief that Mary is treating Jim fairly”, such anaphora is treated as abstract entity anaphora. The simple noun phrases such as “the critical part of the American character” are excluded from abstract entity anaphora, no matter how many words they contain. Abstract entity anaphors are mainly demonstratives (this, that), demonstrative noun phrases (this NP, that NP), sentential pronoun it and full noun phrases (the NP). Apart from that, verb phrase ellipsis, gapping and sluicing may also be used as anaphors for verb phrase anaphora (Asher, 1993). Take for example (1.6), (1.7) and (1.8). (1.6) Fred hit a home run, and then Sally did too. (Asher, 1993: 227) (1.7) Mary beat Fred at chess on Thursday and Sam on Friday. (Asher, 1993: 227) (1.8) A: Fred’s playing the Mozart K. 333 sonata.

B: Yeah, but not very well. (Asher, 1993: 227). However, the colloquialism of such usage is in conflict with the object of this study—written English. And according to Asher (1993), they are not the major forms of abstract entity anaphors. Even if there happens to be any verb phrase ellipsis or gapping or sluicing in argumentative essays, the limited number may be too small to be considered. In view of that, the three kinds of anaphors are beyond our concern in this study. For the same reason, zero anaphor in Mandarin Chinese is also excluded although occasionally it can be used to refer to abstract entity, such as (1.9). (1.9) In addition, some wives nowadays can stand together with rich men to take pictures at a port or a conference, or go forward to break a wine bottle before a steam ship or a plane is launched (Maybe only a miss can do it. I’m not sure of the detailed requirement). Ø the advantage of females. —from Xun Lu’s On Women’s Liberation 还有, 现在有些太太们, 可以和阔男人并肩而立, 在码头或会场上照一个相; 或者当汽 船飞机开始行动之前, 到前面去敲碎一个酒瓶 (这或者非小姐不可也说不定, 我不知道 那详细) 了, ø也还是做女子的便宜的地方。 —鲁迅. 关于妇女解放.《鲁迅文集》

4

1 Introduction

1.2 Differences Between Abstract Entity Anaphora and Other Kinds of Anaphora Contrary to abstract entity anaphora is concrete individual anaphora, that is, mainly nominal anaphora which has been widely studied. Several linguists have mentioned abstract entity anaphora at different degree. But they have only touched upon some aspects of abstract entity anaphora and employed different terms to describe them, such as “extended reference”, “text reference”, “deep anaphora” and “associative anaphora”(Liu 刘东虹, 2006). In this section abstract entity anaphora will be differentiated from these terms respectively.

1.2.1 Abstract Entity Anaphora and Concrete Individual Anaphora As is mentioned above, both second-order entities and third-order entities (Lyons, 1977) are treated as abstract entities contrary to concrete individuals (first-order entities) by Asher (1993). The major difference between abstract entity anaphora and concrete individual anaphora lies in the fact that the antecedents of the former express abstract propositions and concepts which sometimes should be abstracted from surface linguistic forms, for example (1.4). Moreover, abstract entities “depend crucially upon the structure of the language that we use to talk about them” (Lyons, 1977: 445), the interpretation of abstract entity anaphora is determined by discourse relations (see Chap. 4). Take for example (1.10). Only when the discourse relation (cause-consequence) between the two sentences is determined the referent of “that reason” can be identified with the previous sentence. (1.10) [The vast majority of Americans still believe that honesty is an important part of the American character]i . For that reason i , there are numerous watch-dog committees at all levels of society.

On the contrary, in most cases concrete individuals corresponding to Lyons’ (1977) first-order entities can be perceived and their property is determined by the information outside the discourse. For instance, “cars”, “men” and “houses” can be seen or touched independently of the discourse and their images may immediately occur to the interpreter when they are mentioned. Although on some occasions concrete individuals can not be identified directly yet they can be perceived indirectly by some clues in the discourse, for example the invented characters in novels and fairy tales. Their properties and features are determined by the information contained in the discourse rather than by discourse relations. Therefore, they still belong to concrete individual anaphora. The abstract entity antecedents may be phrases, clauses, sentences or sentence clusters, as can be seen in the examples in Sect. 1.1. That is quite different from concrete individual anaphora whose antecedents are mainly noun phrases and they

1.2 Differences Between Abstract Entity Anaphora and Other Kinds …

5

seldom involve clauses, sentences or sentence clusters. Abstract entity anaphors are mainly this (NP), that (NP), it and the NP while for concrete individual anaphora, personal pronouns and zero anaphor can also refer anaphorically as well as those anaphors used for abstract entity anaphora.

1.2.2 Abstract Entity Anaphora, Extended Reference and Text Reference Extended reference and text reference are initiated by Halliday and Hasan (1976) who subsume them into a special kind of personal reference. The referring expressions are it, this and that. The referents of extended reference may not only be participants (a particular person or object) in a narrow sense, but also a process or a sequence of processes or repeated phenomena. The antecedents may be clauses or sentences as well as simple nouns or noun phrases. In this respect extended reference share some similarities with abstract entity anaphora. But the different point is that abstract entity anaphora excludes simple nouns and noun phrases. Only if the noun phrases express propositions can they be viewed as the antecedents of abstract entity anaphora, for example (1.11). In addition, extended reference may also refer to time and place such as (1.12) and (1.13). This property diverges from abstract entity anaphora. (1.11) [The claim that Susan got a C on the test]i was surprising. John didn’t believe it i . (1.12) Just at this moment Alice felt a very curious sensation. (Halliday & Hasan, 1976: 67) (1.13) I came from [Wolverhampton]i . —Thati ’s where I came from too. (Halliday & Hasan, 1976: 67)

Extended reference refers to surface phenomenon while the referent of text reference is “metaphenomenon”, i.e. the proposition or concept expressed by surface phenomenon. The “referent is not taken up at its face-value but is being transmuted into a fact or a report” (Halliday & Hasan, 1976: 52). For example (1.14), the anaphor “it” at the end of the passage refers to “the benefit of curtseying”. (1.4) in the above may also be regarded as text reference. Therefore, it can be seen that text reference is a kind of abstract entity anaphora whose antecedent is sentence clusters. (1.14) [The Queen said:] [“Curtsey while you’re thinking what to say. It saves time.”]i Alice wondered a little at this, but she was too much in awe of the Queen to disbelieve iti . (Halliday & Hasan, 1976: 52)

1.2.3 Abstract Entity Anaphora and Deep Anaphora Deep anaphora is put forward by Sag and Hankamer (Ariel, 1990; Kehler & Ward, 2004). Different from surface anaphora, deep anaphora is not controlled by syntactic factors but by pragmatic factors. It requires the semantic coherence of the recovered

6

1 Introduction

entity. The deep anaphors include personal pronouns, sentential it and zero anaphor (the following examples may serve as illustration), among which only sentential it bears some resemblance to it in abstract entity anaphora. Consider the example (1.16) which has something in common with the concept anaphora in (1.2). Ariel (1990) maintains that the antecedent of deep anaphora may not be encoded in linguistic forms, i.e. deep anaphora has no explicit antecedent. The anaphors are interpreted by reference to the interpreters’ discourse model that may contain entities evoked by both physical context and linguistic context. In other words, deep anaphora only requires a semantic referent of the appropriate type and the referent can be evoked situationally without any explicit linguistic introduction. In this sense deep anaphora bears some resemblance to Asher’s concept anaphora. To illustrate, according to Asher (1993), C-abstraction (see Chap. 4) is necessary for referent identification in (1.16), by which the referent “paint sb. all over with tincture of sth.” is abstracted and identified with “it”. (1.15) I had been waiting at the bus stop for over twenty minutes. At last I saw the bus coming, but as I moved out to get on, he just kept going even though he obviously saw me. I was so angry. (Qin 秦洪武, 2001: 60) (1.16) Paul painted Harry all over with tincture of iodine, and Mary did it to me with strawberry jam. (Ariel, 1990: 59) (1.17) Here is some syrup for you. Shake ø before using. (Ariel, 1990: 59)

In addition, abstract entity anaphora includes some of the surface anaphora such as verb phrase ellipsis. This indicates that abstract entity anaphora has wide difference from deep anaphora in some respects despite the similarities.

1.2.4 Abstract Entity Anaphora and Associative Anaphora Associative anaphora is also called indirect anaphora (Wang 王军, 2004) or bridging reference (Matsui, 1998; Wilson & Matsui, 1998). The referent is not expressed explicitly in previous discourse. As Charolls (1999: 312) put it, “associative anaphora is characterized by the presence of definite NPs for which the definite article can be justified in reference to a preceding NP with which it stands in a part-whole relation”. For example (1.18), the definite NPs in bold type are easily interpreted since the commonly shared representation of “letter” includes among its components: an envelope, one or more stamps, a postmark, etc. (1.18) A letter was awaiting Sherlock Holmes. The envelope was crumbled, the stamp was half off and the postmark indicated that it had been sent the day before. (Charolls, 1999: 312)

However, the relationship between the antecedent and the anaphor is not merely a part-whole one; there may be other kinds of relation such as action-result, type-token and metonymy. Generally speaking, associative strength between the antecedent and the anaphor plays a major role in the inference for the appropriate referent. To be

1.3 The Necessity of the Study of Abstract Entity Anaphora …

7

specific, the inference rests on the existence of a general and stereotypical relation between the entity referred to by the definite NP and the entity previously introduced in the mental model. What should be noted is that there is no coreferential relationship between the antecedent and the anaphor. Apotheloz et al. (1999) maintain that the anaphor reveals the state of discursive memory of the speaker, which is inconsistent with the state constructed by the previous discourse. Similarly, the referents of abstract entity anaphora are not explicitly encoded in linguistic forms, such as the concept to be abstracted by C-abstraction and the proposition to be yielded by constituent revision (see Chap. 4). Referent interpretation also relies on the information conveyed in the discourse. Moreover, abstract entity anaphora involves the state of discursive memory at times. However, the critical difference between the two kinds of anaphora lies in whether the anaphors co refer with the antecedent. Therefore, although abstract entity anaphora bears resemblance to associative anaphora in referent interpretation, they never overlap just like two parallel lines.

1.3 The Necessity of the Study of Abstract Entity Anaphora in Argumentative Texts 1.3.1 The Limitations of the Previous Study of Abstract Entity Anaphora To date there has been a large amount of research in anaphora in various approaches, such as syntactic approach (Huang, 1996; Reinhart, 1981; Chomsky, 1981), semantic approach (Reinhart, 1983; Kamp, 1981), pragmatic approach (Horn, 1984; Levinson, 1991; Huang, 1994), cognitive approach (van Hoek, 1997; Ariel, 1990; Givon, 1983; Xiong 熊学亮, 1999; Xu 许余龙, 2002), psychological approach (Wiley et al., 2001; Mckoon et al., 1993; McDonald et al., 1995) and even computational approach (Mitkov, 2002; Kibble, 2001). However, nearly all of those researches have focused on nominal anaphora i.e. concrete individual anaphora (since such nominal anaphora hardly involves propositions but mainly involves first-order entities), but little attention has been paid to abstract entity anaphora. Although some types of anaphora as mentioned in Sect. 1.2 cover one or two kinds of abstract entity anaphora, after all these types of anaphora display more difference from abstract entity anaphora. Of course, the anaphora with non-NP antecedents has been mentioned sporadically. Chu (1998) differentiates anaphora both in a broad sense and in a narrow sense, and categorizes verb anaphora, adverb anaphora and clause anaphora into the anaphora in a broad sense. Xu Yulong (许余 龙) (2002)mentions eight types of anaphora among which are included some forms of abstract entity anaphora. Cornish (1989), He Ziran (何自然) (2006), Hu Zhuanglin (胡壮麟) (1995) and Gao Yanmei (高彦梅) (2002) have mentioned sentential anaphora. He Zhaoxiong (何兆熊) (2002) subsumes such usage of this and that into

8

1 Introduction

“discourse deixis” to refer to a discourse entity either in the previous text or in the subsequent one (this corresponds to anaphora and cataphora that we are discussing). Webber (cited in Kehler and Ward, 2004) points out that the demonstrative that is often used to access a referent constructed from information conveyed in more than one clause. To describe it, Webber appeals to a process of Circumscriptive Reference that identifies the collection of discourse entities and creates a new entity representing them as a unit. And the newly-created entity is available for subsequent reference. Webber’s idea perhaps sheds some light on Asher’s SDRT since topic construction in SDRT resembles Circumscriptive Reference (see Chap. 4). In Chinese anaphora study Wang Canlong (王灿龙) (1999) points out that “这(this)”and “那(that)”may refer to events. Although these studies touch some parts of abstract entity anaphora to different degrees, none of them provides all that is needed for an in-depth exploration of what they have mentioned. Comparatively, Asher (1993) makes a systematic and further study of abstract entity anaphora. He labels it as a special type of anaphora different from concrete individual anaphora and formulates a theory—Segmented Discourse Representation Theory (SDRT) to deal with abstract entity anaphora. But his instances only involve English and most of them are invented for the purpose of illustrating his theory, so the instances seem too perfect to be plausible. Moreover, his theory has not been tested by on-line instances and by other languages. Notwithstanding the inadequate attention and study, abstract entity anaphora as a kind of anaphoric phenomenon, still flourishes, especially in argumentative texts.

1.3.2 Characteristics of Argumentative Text Argumentation is defined by Van Eemeren et al. (1997) as using language to justify or refute a standpoint, with the aim of securing agreement in views. Argumentation was first distinguished in the theory of the syllogism, later refined in the philosophical theory of argumentation and now reappears in various forms not only in everyday conversations but also in scholarly discourse, editorials, opinions and comments in newspapers. Usually, claim-plus-support arrangement is referred as an “argument”. Take for example (1.19). (1.19) A recent study found that women are more likely than men to be murdered at work. 40% of the women who died on the job in 1993 were murdered. 15% of the men who died on the job during the same period were murdered. (Van Eemeren et al., 1997: 208)

An argument usually means “the use of a statement in a logical process of argumentation to support or weaken another statement whose validity is questionable or contentious” (Kopperschmidt, 1985: 159). In the process of argumentation a claim may be supported in many ways at the same time, as can be seen in Toulmin’s six elements—claim, data, warrant, backing, modal qualifier, rebuttals (Van Dijk,

1.3 The Necessity of the Study of Abstract Entity Anaphora …

9

1997). Nowadays American rhetoricians have developed a set of simpler categorization, using the three basic terms instead—claim, support and warrant, with “support” covering a wider scope of meanings, not merely referring to data on which the claim rests but also the practice of providing evidence to the claim (Qu 曲卫 国, 2005). Such structural characteristics of argumentative texts entail various kinds of discourse relations to connect the sentences and highly logical structures so as to achieve persuasiveness. For example (1.19), the argument involves such kind of discourse relations as Elaboration or Instance.

1.3.3 Argumentative Text and Abstract Entity Anaphora One of the reasons why abstract entity anaphora has not received recognition probably lies in the discourse genres that the researchers have selected as research material. Most of the data of concrete individual anaphora come from narrative writing such as novels and folklores. However, abstract entity anaphora frequently occurs in commentary and exposition, according to Asher (1993). It therefore remains a “largely virgin territory” (Asher, 1993: 9) waiting to be explored. In order to prove the high frequency of abstract entity anaphora, we have selected about 10,000 words at random respectively from argumentative texts➀ , expository texts➁ and narrative texts➂ , and have made a comparison in terms of the frequency of abstract entity anaphora. The result is shown in Table 1.1. The frequency of abstract entity anaphora is the highest in argumentative texts (even higher than expository texts) and lowest in narrative texts. The reason maybe lies in the particular properties of argumentative writing. Argumentative writing requires highly logical relations. Argumentative texts rely more on discourse relations than narrative texts which rely heavily on topic chains that are usually the major characters or the events related to the major characters (Chu, 1998). Similar to expository essays, argumentative essays have major information as well as minor information and the information may be neutral or peripheral. That involves various discourse relations. Moreover, English argumentative writing requires critical thinking, i.e. thinking independently and objectively from multiple perspectives, supporting the arguments with evidence, statistics, examples and quotations (Liu 刘东虹, 2005; Liu, 2005). That entails Elaboration, Instance, Explanation, Result, Commentary, and other discourse relations. In addition, in order to reinforce the rhetorical effect Parallelism and Contrast will be employed. As is Table 1.1 Frequencies of abstract entity anaphora Argumentative texts

Expository texts

Narrative texts

Total words

10, 093

10, 152

10, 113

Frequency (per 10,000 words)

53

32

23

10

1 Introduction

mentioned above, abstract entity anaphora is closely related to discourse relations and the interpretation of abstract entity anaphors largely depends on discourse relations. Apart from that, argumentative texts have their own argumentative devices such as Toulmin’s six elements and contemporary American rhetorician’s three elements. By this means, the writer’s standpoint may be embodied throughout the whole discourse and some sub-standpoints may be presented in one or several paragraphs. Some of the functions may also be performed by abstract entity anaphors which can be used to restate the standpoints, to summarize arguments and to make conclusions. In addition, almost all the writing techniques in expository texts may be employed in argumentative texts. That is likely to render all kinds of abstract entity anaphora appear in argumentative texts and thus make the frequency of abstract entity anaphora higher than that in expository texts. Just because of the high frequency of abstract entity anaphora in argumentative texts, we decide to select argumentative texts as our data.

1.4 Research Purpose and Significance This study will explore the pragmatic functions, text environment and tendencies of abstract entity anaphora with a pragmatic view on the basis of the data analyses, and analyze the anaphora resolution with SDRT. Furthermore, in order to make up the inadequacies of SDRT, some proposals will be put forward. The significance of this study lies in five aspects. First of all, this study expands the scope of anaphora study. The previous studies of anaphora have only sporadically mentioned some parts of abstract entity anaphora but they have never labeled abstract entity anaphora as a separate type of anaphora until Asher (1993) does the work in his book Reference to Abstract Objects in Discourse which centers on English instances alone. Different from Asher, this study makes an investigation into both English and Chinese abstract entity anaphora in argumentative texts and makes comparisons of them. Second, by mean of both quantitative and qualitative studies the pragmatic functions and pragmatic tendencies of abstract entity anaphors are explored. This is an advancement based on the previous studies and a characteristic of this study as well. Third, the establishment of abstract entity anaphora may elicit other related studies. As we know, concrete individual anaphora has been extensively and intensively studied in syntactic, semantic, pragmatic, psychological and computational approaches. Abstract entity anaphora may also be further studied in these similar approaches in the future. Fourth, this study widens the genre type of data and the researchers of anaphora will pay more attention to argumentative texts once neglected in the past. Argumentative texts exist in various forms in our life, for example the columns of opinion and comment in newspapers and magazines, academic articles and literary comments. All of that may serve as the data in anaphora study.

1.5 Organization of the Book

11

Finally, the most important work is the challenge and improvement of SDRT. Our work is not limited to the application of SDRT to Mandarin Chinese and backwards anaphora. As a result of our improvement, SDRT may be used to explain more phenomena in abstract entity anaphora.

1.5 Organization of the Book Two aspects—pragmatic features and referent interpretation—are involved in this book which consists of seven chapters. The first chapter gives a brief introduction to abstract entity anaphora and argumentative text, distinguishes abstract entity anaphora from other kinds of anaphora, and justifies the necessity of the research into abstract entity anaphora. Chapter 2 discusses the pragmatic features of abstract entity anaphors both in English and in Chinese. More than 160,000-word English and Chinese argumentative texts have been selected and analyzed. The data analyses show that abstract entity anaphors are not utterly constrained by Ariel’s accessibility hierarchy. Moreover, abstract entity anaphora has different text environment from concrete individual anaphora. From the statistics of probability, pragmatic tendencies of abstract entity anaphors have been found. The next five chapters deal with referent interpretation. Chapter 3 is the review of literatures on discourse structures and discourse anaphora, including Polanyi (2001)’s Linguistic Discourse Model, Fox (1987)’s Rhetorical Structure and Mann and Thompson’s (1988) Rhetorical Structure Theory. Finally, Discourse Representation Theory is introduced and commented since it is the forefather of Segmented Discourse Representation Theory. This chapter is placed after Chap. 2 which deals with the pragmatic features just for these reasons: this dissertation mainly handles two problems—pragmatic features and referent interpretation, and Chap. 3 provides the literatures on discourse structures and discourse anaphora closely related to Chap. 4 but loosely related to Chap. 2. Chapter 4 introduces Segmented Discourse Representation Theory which is formulated by Asher (1993). To date it is the most effective theory to handle the full range of abstract entity anaphoric phenomena. Then we also apply this theory to Chinese abstract entity anaphora and backwards anaphora, both of which prove its effectiveness. Support from cognitive psychology indicates that SDRT is valid and reasonably designed. Chapter 5 points out that Segmented Discourse Representation Theory is not a perfect theory although it is effective to analyze most of abstract entity anaphora. Four problems in SDRT construction and three problems in anaphora resolution are analyzed, and solutions to five of the problems are proposed. Chapter 6 provides solutions to the rest two problems. An investigation is implemented with 250 freshmen as participants. As a result of the quantitative analyses, tendencies in referent interpretation are found. And long-distance anaphora particular to argumentative texts can be solved in this chapter.

12

1 Introduction

Chapter 7 summarizes the whole study, points out the limitations and makes suggestions for future studies.

Notes ➀ Washington Post (editorial). April 1–6, 2005. College English Intensive Reading (argumentative texts). ➁ Washington Post (news). April 1–6, 2005. China Daily (news). August 10, 2005; June 7, 2006. ➂ College English Intensive Reading (short stories).

Chapter 2

Pragmatic Features of Abstract Entity Anaphora

2.1 Introduction This chapter explores the pragmatic features of abstract entity anaphora, based on the written argumentative data from both English and Chinese argumentative texts. Section 2.2 describes the written argumentative data, including the sources of the texts, the criteria of data collection and the content of data analyses. Section 2.3 discusses the pragmatic functions of abstract entity anaphors, as a result of qualitative analysis of the data. Section 2.4 yields the text environment by quantitative analyses. In this section comparisons will be made between the findings of abstract entity anaphora in this study and the previous studies of concrete individual anaphora, i.e. between the findings from the English sources and the study of Ariel (1990), and between the findings from the Chinese sources and the study of Xu Yulong (许余龙) (2002). Section 2.5 explores the pragmatic tendencies by means of unified analyses as to antecedent types and text environment and more importantly by means of probability calculation. In Sect. 2.6 data of backwards anaphora is collected and analyzed. Section 7 makes a summary of this chapter.

2.2 The Written Argumentative Data 2.2.1 The Sources of the Argumentative Texts The written argumentative data are composed of 103 English articles (about 80,000 English words) and 70 Chinese articles (about 80,000 Chinese words). The argumentative texts have been selected from various sources in an attempt to have a broad range of argumentative texts represented. Most of the data are the editorials and the articles in the columns of opinion and comment from the newspapers since editorials and speeches are the typical texts in which a person makes an argument (van © Central China Normal University Press 2023 D. Liu, Abstract Entity Anaphora in Argumentative Texts, https://doi.org/10.1007/978-981-99-4630-3_2

13

14

2 Pragmatic Features of Abstract Entity Anaphora

Eemeren, 1997). The rest are argumentative essays from magazines and textbooks. The sources ultimately chosen are the following: English Argumentative Data. Daily Telegraph (www.telegraph.co.uk/opinion). The Guardian (www.guardian.co. uk/commentisfree). The New York Times (www.select.nytimes.com/2006…/opinion). China Daily (www.chinadaily.com.cn). The Times (www. timesonline.co.uk/comment). Washington Post. College English Intensive Reading. Chinese Argumentative Data. Nanfang Weekly 《南方周末》 ( 众议) (//www.nanfangdaily.com.cn). People’s Forum 《人民论坛》 ( ) (//paper.people.com.cn). People’s Opinion Weekly 《人民观点周刊》 ( ) (//opinion.people.com.cn/GB). Xun Lu’s Essays 《鲁迅文集》 ( ) (//www.tianyabook.com/luxun). Pictures and News Weekly 《周末画报》 ( (随笔栏目)). Special Attention 《特别关注》 ( ).

2.2.2 The Criteria of Data Collection and the Content of Data Analyses 2.2.2.1

The Criteria of the Data Collection

Anaphors As is mentioned above, all of the abstract entity anaphors may be used for concrete individual anaphora. However, first of all, in this study the anaphors should refer to abstract entities instead of concrete individuals. Such anaphora as (2.1) and (2.2) must be excluded. (2.1) CCTV recently reported [a poverty-stricken student named Mohan Liu] who was admitted to Beijing University though his hard-working and with the philanthropists’ help. She did part-time in the university and set up “Mohan Fund” … although I admire this country girli who has fought against her misfortune, never given in and in return helped others with love, I have a kind of disappointment: as an excellent fourth-year student in the History Department of Beijing University, she still concentrates all her attention on charity donation and moral exhortation, especially when she has well known the situation of education in China though her practice in person. — 71 > 41). The demonstratives predominate in the intermediate text position, with its frequency in Previous Sentence much higher than that of pronouns and definite descriptions (222 > 47 > 15). So the results of the two studies conflict in demonstratives. However, it is the disagreement that indicates the difference between concrete individual anaphora and abstract entity anaphora. The comparison makes clear the pragmatic features of abstract entity anaphors. Different from concrete individual anaphora where pronouns are most frequently used, abstract entity anaphora uses single demonstratives and demonstrative NPs as major anaphors, among which this (124) and this NP (178) predominate in frequency and are most frequently used in intermediate text position. In addition, based on the total of text position in . Tables 2.1 and 2.2, another comparison is made between concrete individual anaphora (CIA) and abstract entity anaphora (AEA) pertaining to the sequence of the text position where antecedents locate. CIA: Previous S (390/ 51.7%) > Same Para (157/ 20.8%) > Same S (118/ 15.6%) > Across Para (90/ 11.9%). AEA: Previous S (286 / 61.8%) > Same S (91/ 19.7%) > Across Para (73 / 15.8%) > Same Para (13 / 2.8%). The two sequences display the unique similarity that the predominant position is Previous Sentence. The greatest difference is the position of Same Para in which concrete individual anaphora is frequently seen but abstract entity anaphora is scarce. This wide divergence also supports the view that English abstract entity anaphora as a special type of anaphora should be treated separately.

2.4.3 Text Environment of Chinese Abstract Entity Anaphora 2.4.3.1

The Study of Chinese Concrete Individual Anaphora by 许余龙

Xu Yulong (许余龙) (2004) uses as his data source 18 Chinese folklores with altogether 11,000 words as well as 855,000-word newspaper articles. Xu Yulong (许 余龙) (2004) does not merely separate single demonstratives “这” and “那” from demonstrative NPs “这NP” and “那NP”, but also treats the demonstratives (NPs) in subject/topic position as High Accessibility markers and those in other positions as Intermediate Accessibility markers. In other words, he classifies demonstratives into eight kinds in his data analyses: “这 (subject/topic)”, “这(other)”, “那(subject/topic)”, “那(other)”, “这 NP(subject/topic)”, “这NP (other)”, “那NP (subject/topic)”, “那NP (other)”. Adopting Ariel(1990)’s parameter, Xu Yulong (许余龙) (2004) has found that all of the single demonstratives “这” and “那” in his data occur in the subject/topic position. The text environment of the single demonstratives is shown in Table 2.4.

30

2 Pragmatic Features of Abstract Entity Anaphora

Table 2.4 Text environment of Chinese Demonstratives (Xu许余龙, 2004: 231, revised) Single demonstratives

Text position

Total

Same S

Previous S

“这 (subject/topic)”

280 (53.7%)

241 (46.3%)

521 (100%)

“那 (subject/topic)”

49 (89.1%)

6 (10.9%)

55 (100%)

Total

329

247

576

Only two text positions are involved, which reveals the difference between Chinese and English single demonstratives. Table 2.4 also shows that the total frequency of “这(subject/topic)” is much higher than that of “那(subject/topic)”. Moreover, Xu Yulong (许余龙) (2004) has found that the use of demonstrative NPs is different from that of single demonstratives. In his folklore data all of the demonstrative NPs in subject/topic position or as modifiers of the subjects unanimously refer to the entities in the same sentence. The demonstrative NPs in other positions typically refer to the entities in the previous sentence or in the same paragraph. However, the data about the demonstrative NPs is not provided. And his study of demonstrative NPs does not involve the newspaper articles which are supposed to be abundant with abstract entity anaphora. Consequently, what Xu Yulong (许余龙) (2004) centers on is mainly concrete individual anaphora.

2.4.3.2

The Present Study of Chinese Abstract Entity Anaphora

Taking the classification of demonstratives by Xu Yulong (许余龙) (2004) and Ariel’s (1990) parameter about text environment, we have collected data from 80,000-word Chinese argumentative texts and have studied the text environment of both single demonstratives and demonstrative NPs in subject/topic position and in other positions. Different from Xu Yulong (许余龙) (2004), we have provided all of the related data of the statistics and marked the primary and the secondary positions for each kind of anaphor, as is shown in Table 2.5. Table 2.5 Text environment of Chinese demonstrative (NP) Demonstrative (NP)

Text position Same S

Previous S

Same para

Across para

Total

这 (subject/topic)

77 (45.8%)

80 (47.6%)

3(1.8%)

8 (4.8%)

168 (100%)

这 (other)

0

1 (100%)

0

0

1 (100%)

这 NP (subject/topic)

6 (13.3%)

32 (71.1%)

5 (11.1%)

2 (4.4%)

45 (99.9%)

这 NP (other)

17 (35.4%)

16 (33.3%)

8 (16.7%)

7 (14.6%)

48 (100%)

那 (subject/topic)

9 (37.5%)

15(62.5%)

0

0

24 (100%)

Total of text position

109 (38.1%)

144 (50.3%)

16 (5.6%)

17 (5.9%)

286 (99.9%)

2.4 Text Environment of Abstract Entity Anaphora

31

By comparing this study with that of Xu Yulong (许余龙) (2004), we have found the similarities. First, the total frequency of “这(subject/topic)” is much higher than that of “那(subject/topic)” (and even several times higher than that of the other anaphors). There is little difference between the frequency of “这(subject/topic)” in Same S and that in Previous S. Second, “那(other)” hasn’t been found in this study as well as in Xu Yulong (许余龙) (2004). Finally, the instance of “这(other)” is rather scarce. Xu Yulong (许余龙) (2004) has not found any in his study and there is only one instance in this study. This is the case. (2.31) [Instead of being imprisoned in a cage, the bird may appear to have a new position of power when it is placed on a perch. But in reality, it is still like a toy for others, taking orders from others and every bite and every sip is preordained.] i Thisi is accordance with the saying: “One meal received, one’s will surrendered.” —Lu Xun. On women’s liberation. Xun Lu’s Essays [拿一匹小鸟关在笼中, 或给站在竿子上, 地位好象改变了, 其实还只是一样的 在给别人做玩意, 一饮一啄, 都听命于别人。] i 俗语说: “受人一饭, 听人使唤”, 就是这 这i。 — 鲁迅. 关于妇女解放.《鲁迅文集》

There are more differences between the two studies. To begin with, “这NP(subject/ topic)” in this study chiefly refers to the entity in the previous sentence while in the study ofXu Yulong (许余龙) (2004) all of the instances of “这NP(subject/topic)” refer to the entities in the same sentence. Moreover, the frequency of “这NP(other)” in the position of Same Sentence (primary position) is approximate to that in Previous Sentence (secondary position). “这NP(other)” in this study may typically refer to the entity either in the same sentence or in the previous sentence rather than typically refer to the entity in the same paragraph as is described byXu Yulong (许余龙) (2004). Finally, no instance of “那NP (subject/topic)” or “那NP (other)” has been found. In addition, in this study the sequence of the text position of antecedents is displayed as follows: Previous S (144 / 50.3%) > Same S (109 / 38.1%) > Across Para (17 / 5.9%) > Same Para (16 /5.6%).

2.4.4 Comparisons Between English and Chinese Abstract Entity Anaphora By means of comparison, we attempt to find out the similarities between English and Chinese abstract entity anaphora. The first similarity is revealed in the frequency of demonstratives. The anaphors this (NP) and “这(NP) (subject/topic)” occur statistically far more frequently than the other anaphors. The second similarity consists in the sequence of text position. Both English and Chinese demonstratives display the same sequence: Previous S > Same S > Across Para > Same Para. And in both languages the frequency of Previous S is by far higher than that of any other position.

32

2 Pragmatic Features of Abstract Entity Anaphora

This similarity indicates that when used to refer to abstract entities demonstratives are not necessarily constrained by Accessibility. The reason may be attributed to the variety of antecedent — not only nominal phrases but also clauses, sentences and sentence clusters. These antecedents maybe influence the distance from the anaphors. Abstract entity anaphora whether in English or in Chinese, is different from concrete individual anaphora in terms of text environment and the frequency of demonstratives. It is necessary to treat it as a separate type of anaphora and to explore some pragmatic tendency in it.

2.5 Pragmatic Tendency of Abstract Entity Anaphora The frequency of the antecedent types for each kind of anaphor in both Chinese and English abstract entity anaphora have been collected and the percentage of each type of antecedent has been produced. In Tables 2.6 and 2.7 most of the anaphors have two typical antecedents marked in bold type. In Table 2.6 only one antecedent type of “ 这(other)” is marked because it is the unique instance; only Clause of “这NP(other)” is marked since its frequency is by far the highest leaving that of the others far behind (more than three times higher). In Table 2.7 three types of antecedent for this are marked because of their approximate frequency; only one type respectively for that and that NP is marked — Clause for that and Sentence for that NP since the frequency is several times higher than that of the others. Table 2.6 shows that clauses, sentences and chunks compose the major antecedent types of Chinese abstract entity anaphora, altogether making up 95.8%; contrarily, no instance of noun phrase as antecedent has been found. To be specific, the most typical antecedent of “这(subject/topic)” is Clause and the second most typical one is Chunk; the most typical antecedent of “这NP(subject/topic)” is Sentence and then Clause; “这NP(other)” mainly refers to clause; for “那(subject/topic)”, Table 2.6 Antecedent types of Chinese abstract entity anaphora Anaphors

Antecedent types Chunk

Sentence

Clause

NP VP

Total of anaphors

这 (subject/ topic)

45 (26.8%) 36 (21.4%)

85 (50.6%) 0

2 (1.2%)

这 (other)

0

1 (100%)

0

0

0

1 (100%)

这 NP(subject/ topic)

0

24 (53.3%) 21 (46.7%) 0

0

45 (100%)

这 NP(other)

2 (4.2%)

9 (18.8%)

8 (16.7%) 48 (100.1%)

那 (subject/ topic)

3 (12.5%)

11 (45.8%) 8 (33.3%)

0

2 (8.3%)

81 (28.3%)

0

12 (4.2%) 286 (100%)

Total of 50 (17.5%) antecedent types

29 (60.4%) 0

143 (50%)

168 (100%)

24 (99.9%)

2.5 Pragmatic Tendency of Abstract Entity Anaphora

33

Table 2.7 Antecedent types of English abstract entity anaphora Anaphors

Antecedent types Chunk

Sentence

Clause

NP

VP

Total of anaphors

This

33 (26.6%)

31 (25%)

40 (32.3%)

20 (16%)

0

124 (100%)

That

3 (12%)

5(20%)

17 (68%)

0

0

25 (100%)

This NP

53 (29.8%)

47 (26.4%)

16(9%)

35 (19.7%)

27(15.2%)

178 (100.1%)

That NP

5 (20.8%)

14 (58.3%)

3(12.5%)

2(8.3%)

0

24 (100%)

It

0

9 (12.7%)

2(2.8%)

39(54.9%)

21(29.6%)

71 (100%)

The NP

14 (34.1%)

13 (31.7%)

8 (19.5%)

4(9.8%)

2(4.9%)

41 (100%)

Total of antecedent types

108 (23.3%)

119 (25.7%)

86 (18.6%)

100 (21.6%)

50 (10.8%)

463 (100%)

the most typical antecedent is Sentence and then Clause. However, things are not utterly the same in English abstract entity anaphora (see Table 2.7) as in Chinese. Noun phrases often serve as antecedents, especially for the demonstrative (NP). But clauses, sentences and chunks are still the major antecedents (77.6%) in abstract entity anaphora. The most typical antecedent of this as well as that is Clause, with Chunk and Sentence serving as the second most typical ones for this; Chunk is most typically used for this NP and Sentence is the secondary one; Sentence serves as the most typical antecedent for that NP; as for it, NP is the most typical antecedent while VP the secondary one; for the NP anaphor, Chunk and Sentence are most frequently employed as antecedent. In order to explore the pragmatic tendency, unified analyses have been made as to antecedent types and text environment of anaphors. For the text environment of English abstract entity anaphora, only the figures in bold type in Tables 2.2 and 2.3 are considered as the typical text positions. And the figures in bold in Table 2.7 indicating typical antecedent types are taken into account. Apart from that, the frequency of each typical antecedent in each typical text positions has also been collected for each kind of anaphor. Based on that, probability has been calculated (see Table 2.8). Take this for example. One of the typical antecedents of this is Chunk occurring 33 times (see Table 2.7); and a typical text position of this is Previous Sentence as is shown in Table 2.3; then we find 29 Chunks in the position of Previous Sentence. Consequently, the probability is (29/33) 0.88. The figures in bold in Table 2.8 indicate high probability which manifests the pragmatic tendency of these anaphors. Both this and this NP are mainly used to refer to the sentence or the chunk freshly mentioned, and this may also refer to the clause

34

2 Pragmatic Features of Abstract Entity Anaphora

Table 2.8 Probability of typical antecedent(s) for each English anaphor Chunk

Sentence

Clause

This

Previous S Same S

(32/40) 0.8

That

Same S

(16/17) 0.94

This NP

Previous S

NP

VP

(29/33) 0.88 (30/31) 0.97

(43/53) 0.81 (31/47) 0.66

Across Para (11/53) 0.21 (16/47) 0.34 That NP Previous S It

Previous S

The NP

Across Para Previous S

(8/14) 0.57 (24/39) 0.62 (19/21) 0.9 (11/13) 0.86 (12/14) 0.86

in the same sentence. That tends to code the clause in the same sentence while that NP the previous sentence. It is primarily employed to refer to the verb phrase and also the noun phrase in the previous sentence. The NP favors the sentence across the paragraph or the chunk most recently mentioned. The same unified analyses have been made for Chinese abstract anaphora. In view of the limited number, NP and VP are counted out in the probability calculation. Table 2.9 displays the probability and pragmatic tendency. “这(subject/topic)” chiefly refers to the chunk most recently mentioned and the clause in the same sentence. “这NP(subject/topic)” favors the previous sentence and “这NP(other)” the clause in the same sentence. “那(subject/topic)” tends to be identified with the previous sentence and the clause in the same sentence. As “this (other) (这(other))” is rarely used, the occurrence in this study does not constitute a tendency despite the high probability. Table 2.9 Probability of typical antecedent(s) for each Chinese demonstrative (NP) Chunk

Sentence

Clause

这 (subject/topic)

Previous S

这 (other)

Previous S

(1/1) 1

这 NP (subject/topic)

Previous S

(20/24) 0.83

Same S

(4/21) 0.19

这 NP (other)

Previous S

(13/29) 0.45

Same S

(16/29) 0.55

(43/45)0.96

Same S

那 (subject/topic)

Previous S Same S

(77/85) 0.91

(9/11) 0.82 (7/8) 0.88

2.6 Backwards Anaphora Table 2.10 English backwards anaphora

35

Anaphor

Antecedent

This

4

Sentence

Chunk

This NP

1

The NP

2

2.6 Backwards Anaphora What we have discussed is forwards anaphora. Usually the antecedent locates before the anaphor so that the anaphor refers back to it. By backwards anaphora, the anaphor appears earlier than the antecedent and it refers forwards to the antecedent. Seven cases of backwards anaphora have been found in English argumentative texts. The anaphors involve proximal demonstrative (NP) and full NP, and only sentences and chunks serve as the antecedent. As for the selection of anaphors, Fillmore (1997) maintains that that requires identifiability by both the speaker and the addressee while this may refer to what is accessible only to the speaker. Therefore, this may be used in backwards anaphora while that may not. For illustration, consider (2.32) in which “this” is used as the anaphor and (2.33) whose anaphor is demonstrative NP “this advice”. Moreover, besides this and this NP, the NP is used to refer to the first mentioned entity for example (2.34), which is a rare occurrence in concrete individual anaphora. In all the three cases of this NP and the NP the anaphors locate in the post-verbal position which is closer to the antecedent. (2.32) I do not know if this i is true: [press reports suggested that Mr Banda, an illiterate 32year-old farmer who had already lost two children to malaria and a wife in childbirth, regularly cycled many miles to visit his son]i. — Jenny McCartney. What’s best for baby… is often forgotten. Telegraph 05/11/ 2006 (2.33) He offered this advice: [“Hold a picture of yourself... in your mind’s eye, and you will be drawn toward it. Picture yourself vividly as defeated, and that alone will make victory impossible. Picture yourself as winning, and that will contribute immeasurably to success. Do not picture yourself as anything, and you will drift...”] ---Eugene Raudsepp. Daydream a Little. College English (2.34) And in our own day we are used to hearing the traditional complaintsi : [“I can’t wait for my vacation,” “I wish I could stay home today,” “My boss treats me poorly,” “I’ve got too much work to do and not enough time to do it.”] i. — Leonard R. Sayles. Why People Work. College English

Six cases of backwards anaphora have been found in Chinese argumentative texts, three of them take clause as antecedent and the other three use sentence as antecedent. In five of the cases the anaphors are the derivation of “this NP(other) (这NP(other))” —“like this NP (other) (这样的NP (other))”, for example (2.35) and (2.36); and in one case (2.37) the anaphor is the derivation of “this (other) (这 (other))” — “like this (这样 (other))”. Of course, Fillmore’s explanation may also be applied to

36

2 Pragmatic Features of Abstract Entity Anaphora

Chinese anaphor selection. In addition, the reason why “like this NP(other) (这样 的NP(other))” is preferred in backwards anaphora may be that its position is closer to the subsequent antecedent. (2.35) Somebody has made this analogyi : [university graduates from the 1950s were at a level equivalent to today’s doctoral students, while those from the early 1980s were at a level equivalent to today’s master’s students.] i —Guo Lichang. The debate on blue collar and white collar is not important. People’s Forum, 09/16/2006 曾经有人打过这 这样的比方i , [说上世纪50年代的大学毕业生其水平相当于现在 的博士, 80年代初的毕业生相当于现在的硕士。] i — 郭立场. “蓝领”与“白领”之争并不重要.《人民论坛》 , 2006-09-16 (2.36) Reflecting on the history of China from the late Qing Dynasty to the 50th year of the Republic of China, Liang Qichao once made this sigh with emotioni : [“After fifty years of scientific input, the only person in our country who can achieve an international level as a scholar is Dr. Wu Xinglian (i.e. Wu Liande)!”] i —Lin Guanzhen. Don’t forget the first doctor od modern medicine in China. Nanfang Weekly, 03/08/2007 梁启超回顾晚清到民国50年历史, 发出这 这样的感慨i : [“科学输入垂五十年, 国 中能以学者资格与世界相见者, 伍星联(即伍连德)博士一人而已!”]i — 林冠珍. 莫忘中国现代医学第一人. 南方周末, 2007-03-08 (2.37) It can be said like thisi : [the acquisition of material goods and the utilization of wealth are, in essence, a process of continuous allocation and consumption of social resources.] i —Jiang Meng. The association of executives’ skyrocketing salaries in Lenovo. “People’s Forum,” 09/16/2006 可以这么说 说i , [物质获取与财富利用, 其实质就是社会资源不断分配与消耗的 过程]i。 — 蒋萌. 联想高管暴涨薪酬的联想.《人民论坛》 , 2006-09-16

As far as the typical constructions, Ariel (1990) suggests that there are two types of backwards anaphora, each allowed under different circumstances. Dependency is crucial in the first type where the antecedent is newly introduced into the discourse. As the antecedent is a new entity the interpreter has to rely on the antecedent to interpret the pronoun. Thus, the pronoun clause should be dependent on the antecedent clause so as to facilitate the interpretation of the pronoun. In this case the pronoun clause is usually subordinate clause. Take for example (2.38). The pronoun clause is an adverbial clause subordinate to the main clause containing the antecedent “a child of my acquaintance”. On the contrary, independency of the pronoun clause characterizes the second type since the entities form part of the discourse already. In this case the pronoun clause is usually a Dominant clause. To illustrate, in (2.39) the referent of “he” is previously mentioned in the discourse and its interpretation does not depend on the antecedent “John”. (2.38) When shei was five years old, [a child of my acquaintance]i announced a theory that she was inhabited by rabbits. (Ariel, 1990: 158) (2.39) Hei lied to me, and [John]i was my friend. (Ariel, 1990: 158)

2.7 Summary

37

Ariel’s view does not come out singly and it is also borne out in van Hoek (1997). From the backwards anaphora corpus, van Hoek (1997: 112) has found the following characteristics: the antecedent is the most prominent nominal conception in the sentence; the pronoun is semantically backgrounded relative to the antecedent; the larger phrase containing the pronoun is also subordinated to a more prominent linguistic unit. Nevertheless, in all of the cases in this study the antecedents are new entities without exception. According to Ariel and van Hoek, the anaphoric clause should be subordinate to or dependent on the antecedent clause. On the contrary, the anaphoric clause or sentence in this study is Dominant. This may be attributed to the use of demonstrative (NP) and full NP instead of pronoun. After all, it is another signal marking the difference between abstract entity backwards anaphora and concrete individual backwards anaphora.

2.7 Summary In this study more than 160,000-word English and Chinese argumentative texts are selected and analyzed. Both qualitative and quantitative analyses have been employed. As a result of qualitative analysis, pragmatic functions of abstract entity anaphora have been found: recapitulation, obscuration and demarcation. The quantitative analyses include the frequencies of all kinds of anaphor and antecedent, the text position, the probability that the typical antecedents of each kind of anaphor appear in the typical text position. Various comparisons have been made between the findings of abstract entity anaphora in this study and the previous studies of concrete individual anaphora. The data analyses show that abstract entity anaphors are not utterly constrained by Ariel’s Accessibility Hierarchy. Moreover, abstract entity anaphora exhibits some similarities in English and in Chinese. For example, the frequencies of this (NP) and “这(NP)(subject/topic)” predominate in the use of anaphors. Abstract entity anaphora whether in English or in Chinese, is different from concrete individual anaphora in terms of text environment and the frequency of demonstratives. From the statistics of probability, pragmatic tendencies of abstract entity anaphors have been found. Both this and this NP chiefly refer to the sentence or the chunk most recently mentioned, and this may also refer to the clause in the same sentence. That tends to be identified with the clause in the same sentence while that NP the previous sentence. It tends to code the verb phrase and also the noun phrase in the previous sentence. The NP favors the sentence across the paragraph or the chunk most recently mentioned. “这(subject/topic)” mainly refers to the chunk freshly mentioned and the clause in the same sentence. “这NP(subject/topic)” favors the previous sentence and “这NP(other)” the clause in the same sentence. “那(subject/topic)” is primarily

38

2 Pragmatic Features of Abstract Entity Anaphora

employed to code the previous sentence and the clause in the same sentence. However, the occurrence of “这(other)” in this study does not form a tendency. Apart from that, differences also exhibit between abstract entity backwards anaphora and concrete individual backwards anaphora. Therefore, backed up with firm empirical findings from above analyses, we may come to the conclusion that abstract entity anaphora should be treated as a special category in its own right.

Chapter 3

Discourse Structure and Discourse Anaphora

3.1 Introduction The last two decades have seen a tremendous upsurge in work on discourse production and comprehension. Central to the work has been the belief that there is a strong relationship between the structure of discourse and the treatment of anaphora. The literatures reviewed in this chapter are concerned with “discourse anaphora” i.e. the anaphora across sentences and beyond syntactical control. Moreover, most of these literatures emphasize that discourse structures consist of constituents glued by discourse relations, and in return discourse relations may segment a discourse. Included in this chapter are Rhetorical Structure Theory by Mann and Thompson (1987, 1988), Rhetorical Structure by Fox (1987), Linguistic Discourse Model by Polanyi (2001), Discourse Representation Theory by Kamp (1981), and other theories merely concerning discourse structure such as episode model, attention model and distance model. In addition, although Polanyi’s LDM is not applied to analyze discourse anaphora, it provides a detailed view of discourse structure and SDRT owes much to LDM in its SDRS construction. For this reason, LDM is introduced in this chapter. Apart from that, while DRT does not depend on discourse relations, it deals with discourse anaphora and lays a foundation for SDRT. Therefore, DRT is also introduced. By the way, since this chapter serves as the sinfonia of SDRT and has little relation to Chap. 2, the literature review of discourse structure and discourse anaphora is located just before the introduction of SDRT.

© Central China Normal University Press 2023 D. Liu, Abstract Entity Anaphora in Argumentative Texts, https://doi.org/10.1007/978-981-99-4630-3_3

39

40

3 Discourse Structure and Discourse Anaphora

3.2 Rhetorical Structure Theory (RST) by Mann and Thompson 3.2.1 Main Idea Rhetorical Structure Theory advanced by Mann and Thompson (1987, 1988) has become one of the most popular theories in the last decade. The key elements of RST are rhetorical relation and text spans. A text span is any portion of a text realized by a unit which is typically a clause. Usually, one of the text spans connected by a rhetorical relation is nucleus (N) and the other is satellite (S). There are a few exceptions: some relations are multinuclear. What the nucleus expresses is more essential to the writer’s purpose than the satellite. The satellite is incomprehensible if independent of the nucleus but not vice versa. Consequently, deleting the nuclei of the rhetorical relations holding among all text spans in a text may produce an incomprehensible text, while deleting all the satellites may yield a text still comprehensible. Apart from that, Mann, Marthiessen and Thompson (1992) make a distinction between hypotaxis and embedding. Embedding includes restrictive relative clauses, subject and object clauses, and clausal complements to verbs and adjectives. These clauses can not be treated as satellites; instead they are part of the nucleus. Hypotaxis including adverbial clauses can be treated as satellite to the nucleus—the main clause. Central to RST is rhetorical relation, which can hold between two non-overlapping text spans. Mann, Matthiessen and Thompson (1992) have found 24 relations, including 21 nucleus-satellite relations (e.g. Evidence, Elaboration and Restatement) and 3 multinuclear relations (i.e. Sequence, Contrast and Joint). Most of the relations are asymmetric nucleus-satellite relations. The definition of a relation consists of two fields: constraints and effect. The constraints lie in three aspects: on the nucleus, on the satellite and on combination of the two. Take Evidence for example (Mann et al., 1992:48): Constraints: a. Constraints on the Nucleus: The reader might not believe the nucleus to a degree satisfactory to the writer b. Constraints on the Satellite: The reader believes the satellite or will find it credible c. Constraints on the combination of Nucleus and Satellite: The reader’s comprehending the satellite increases his or her belief of the nucleus Effect: a. The reader’s belief of the nucleus is increased b. Locus of the Effect: Nucleus

The relation definition does not constrain the order of spans. For nearly all of the relations, both of the possible orders—nucleus-satellite and satellite-nucleus—are found. The Locus of Effect may be the nucleus or both the nucleus and the satellite. When it is the former, as in Evidence relation, the satellite supports the nucleus but does not contribute to it and nuclearity reflects the supporting role of the satellite.

3.2 Rhetorical Structure Theory (RST) by Mann and Thompson

41

When the locus of effect is both the nucleus and the satellite, as in Circumstance relation, nuclearity reflects the symbiotic role of both the nucleus and the satellite, and expresses particular characteristics of the subject matter. In this case, the satellite sets a subject matter framework for the nucleus to be interpreted. In addition, an RST structure is “rhetorical” in that the text structuring relations reflect the writer’s selection of organization and presentation of a text, which is determined by the writer’s purpose, assumptions about the reader and the subject matter of the text. According to Mann, Matthiessen and Thompson (1992), rhetorical relations can be assembled into rhetorical structure trees on the basis of five types of structural patterns, RST schemas. In the following schemas, the horizontal lines represent text spans while the vertical and diagonal lines represent identifications of the nuclear spans. The arrows link the satellite to the nucleus of a rhetorical relation and are labeled with the name of the rhetorical relation. Most of the rhetorical relations are assembled based on schema (a). Schema (d) covers the cases where a nucleus is linked to multiple satellites by different rhetorical relations. Schemas (b), (c) and (e) represent multinuclear relations or paratactic relations. In (c) and (e) the nuclei are identified by convention only since there is no satellite and hence no arrow. These structural patterns are available at every scale from the largest (the whole text) to the smallest scale (two-clause combination) (Fig. 3.1). Some scholars use RST to explain anaphora. For example, Chen Ping (陈平) (1987) analyzes the referential selection in Mandarin Chinese by means of RST hierarchical structures. In the example (3.1), the two clauses between the dashes are asides, however after which zero anaphor is used continually. And the whole discourse appears to keep intact and fluent without being interrupted by the asides. Chen Ping (陈平) attributes this phenomenon to the subordination of the asides (to the previous clause), and points out the constraint of this phenomenon—simple

Circumstance S

Contrast

N

N

(a)

N

N

N

(b)

Motivation S

Joint

Enablement N

S

(d) Fig. 3.1 RST schemas (Mann et al., 1992, revised)

(c)

Sequence N

Sequence N

(e)

N

42

3 Discourse Structure and Discourse Anaphora

Sequence

a

b

c

d

h

i

Explanation e Cause f

g

Fig. 3.2 Discourse structure graph of example (3.1)

discourse structure of the asides. Employing RST, Chen Ping (陈平) (1987: 374) draws the graph of the discourse structure of (3.1) (Fig. 3.2). (3.1) Sister-in-law Zhang was cookinga , # serving teab , #treating the guests c , #adding soupd , #changing chopstickse — Lao Li was so excited in eating f that his chopsticks were dropped to the ground twiceg — #selecting the fat to eat h and #praising her own cooking skillsi . —Lao She. Divorce 张大嫂做菜a , #端茶b , #让客人c , #添汤d , #换筷子e — 老李吃高兴了f , 把筷子掉 在地上两回g — #自己挑肥的吃h , #夸自己手艺i。 — 老舍.《离婚》

3.2.2 Comment RST provides a general way to describe the relations among the text spans of a text whether or not the relations are explicitly signaled grammatically or lexically. RST as an analytical tool has been used for wide range of text types such as narrative and expository texts. Fox’s (1987a) study is a development of RST. Chen (1993), Chu (1998) and Wu Qizhu (吴启主) (2001) have applied RST to Chinese discourse analysis. Despite its advantages in the study of text coherence, RST has some inadequacies. Marcu (2000) argues that RST lacks two things: a formal specification that would be used to distinguish between well-formed and ill-formed rhetorical structure trees; algorithms that would determine all the possible rhetorical analyses of a given discourse. Marcu attributes the problem to a lack of compositionality in RST, which would explain the relationship between rhetorical relations holding between large textual spans and that holding between elementary units. RST fails to prescribe how the rhetorical relations holding between minimal units relate to the relations

3.3 Rhetorical Structure by Fox

43

holding between larger text spans. Thus, it seems unreasonable to join two partial trees in order to create a bigger tree just because there is a rhetorical relation that holds between two minimal units belonging to two spans. Moreover, RST can never determine the precise boundaries of the spans over which the relation holds. As a result, readers may not produce unambiguous text spans or rhetorical relations. In addition, the classification of discourse relation is so fine-grained that readers have to choose among long lists of relations. And “the lists are often unmotivated and unconstrained” and “they may lead to analytic problems” (Sanders et al., 1993: 117). Last but not the least, RST can only be used to explain the selection of anaphoric expressions rather than anaphora resolution.

3.3 Rhetorical Structure by Fox 3.3.1 Main Idea 3.3.1.1

Basic Units

Fox (1987a) maintains that any treatment of anaphora must seek its understanding in the hierarchical structure of the text. Texts are regarded not merely as strings of clauses but as groups of hierarchically organized clauses with various informational and interactional relations to one another. The basic and smallest unit is proposition which is more abstract than clause or sentence. However, Fox’s proposition is different from Kintsch’s (Carroll, 2005; Gui 桂诗春, 2002). The latter consists of two or more concepts and some kind of relation between them. For example, the sentence “Harry was hit by George” conveys the proposition “Hit (George, Harry)”. More complex sentences convey more than one proposition. The sentence “George got his salary, paid for the bill and bought a toy for the boy” could be represented as three separate propositions: Got (George, salary) Paid (George, bill) Bought (George, toy, boy) But Fox regards the sentence as one proposition which provides some reference for SDRT. And propositions are not necessarily equivalent to clauses. Relative clauses are not treated as separate propositions but as belonging to their modified clauses while adverbial clauses are separate propositions. That is in line with RST.

44

3 Discourse Structure and Discourse Anaphora

3.3.1.2

Rhetorical Relations

In Fox’s R-structures, texts are treated as hierarchically organized groups of propositions. Most R-structures consist of more than two units—the nucleus and the adjunct, with the nucleus representing the main goals of the writer and the adjunct supplementing the nucleus. Fox describes eleven R-structures—Issue, Conditional, Circumstance, Reason, Concession, Opposition, Purpose, Response, List, Narrate, and Contrast. All R-structures have a nucleus but unnecessarily have adjuncts. Of all the structures, Issue structure is the most powerful organizing unit. Issue structure contains the relations such as elaboration, background and evidence which hold between the nucleus and the adjunct. R-structures may build upon one another. So each unit may be realized either by a proposition or by another R-structure. In drawing R-structures, Fox first labels the rhetorical relation at the top (such as issue in Fig. 3.3), and then draws several lines from the top. The straight line descending from the top represents the nucleus and the arcing lines coming out of the bottom of the nucleus line represents the adjuncts. The label at the arcing line displays the relation between the nucleus and the adjunct. Take (3.2) for example. (3.2) ➀ James S. Albertson has been appointed acting academic vice president by the Regents following President Saxon’s recommendation. ➁ The appointment is effective from March 1 until a permanent academic vice president is named. ➂ Academic Vice President Donald C. Swain earlier was named president of the University of Louisville.

In addition, floating relations such as Summary, Conclusion, Result and Assessment, “are not associated with any particular R-structure” and “can occur between a proposition of any R-structure and another proposition” (Fox, 1987a: 88). (There are some problems in Fox’s words and this will be commented in Sect. 3.3.2) (Fig. 3.4). Fig. 3.3 R-structure of example (3.2) (Fox, 1987a: 87)

Issue elaboration

background

Summary

Fig. 3.4 R-structure of floating relation (Fox, 1987a: 88)

Text

3.3 Rhetorical Structure by Fox

3.3.1.3

45

Anaphora

Fox distinguishes active and controlling propositions. The former refers to a proposition whose R-structure partner is being produced. When the nucleus is produced the adjunct is active; when the adjunct is being produced the nucleus is active. For instance, when a post-posed conditional clause is being produced the proposition of the previous main clause is active; otherwise, the pre-posed conditional clause is active when the main clause is being produced. When a proposition is active, its Rstructure partner which has previously been produced is controlling. In Fig. 3.5, when the nucleus of the Concession structure is being produced the nucleus of the Issue structure is active. Then it becomes controlling when the adjunct of the Concession structure is being produced. And at this time the nucleus of the Concession structure is active. The basic patterns for anaphora in expository texts are presented as follows: a pronoun is used to refer to a person if that person is previously mentioned in an active or controlling proposition; otherwise a full NP is used. In other words, by using a pronoun the writer manifests to the reader that the intended referent is in an active or controlling proposition, whereas by using a full NP the writer indicates to the reader that the intended referent is outside of these units. It is not merely the case that the context induces the writer to use an appropriate anaphoric form; the writer may actively create a context through the use of a certain anaphoric form. It is not simply distance or prominence that triggers the use of one anaphoric form over the other. But rather, rhetorical structure plays an enormous role in determining the distribution of full NP and pronouns. Another important phenomenon identified by Fox is return pop in which a proposition is tied back to a physically distant proposition rather than the immediately preceding one. Return pop shows that the writer is returning to an earlier, superordinate node. The pattern frequently seen is nucleus-adjunct-adjunct, but the second adjunct is for the nucleus instead of the preceding adjunct. As can be seen in Fig. 3.6, ➆ is tied back to ➀, the most distant proposition, rather than following the preceding one ➅. According to Fox, return pops are often done with NPs, thus pronouns are neither necessary nor sufficient for creating return pops. Although hierarchical structure relations play a significant role in the patterning of anaphora, they are not responsible for every use of all the anaphoric devices. Fox Fig. 3.5 Active and controlling propositions (Fox, 1987a: 96)

Issue evidence concession

46 Fig. 3.6 R-structure of return pop (Fox, 1987a: 103)

3 Discourse Structure and Discourse Anaphora

Issue elaboration

background background

Joint List

also acknowledges that the choice of anaphoric devices may be influenced by nonstructural factors, such as categorization of the referent, further information about the referent and comparison and contrast of the referent with other people.

3.3.2 Comment Fox’s theory provides a new way to look at discourse behavior. It makes an attempt to explore the relationship between language, cognition and society. Its smallest unit is larger than that in Polanyi’s (2001) LDM, and the diagrams of Fox are more palpable and intuitionistic than that of RST, and thus it is easier to process texts and construct rhetorical structures. However, it also has some weaknesses. To begin with, despite its palpability and plainness, the diagram lacks full hierarchical display, especially the diagram of a simple text. In example (3.2), although an adjunct ➁ is an elaboration of the nucleus ➀, Fox still puts it at the same level as ➀. Moreover, Fox holds that “floating relations can occur between a proposition of any R-structure and another proposition” (Fox, 1987: 88). But contradiction arises that in her examples of Summary, Conclusion and Assessment, these relations occur between the last proposition and the previous text. To be accurate, floating relations can occur not merely between a proposition and another proposition, but also between a proposition and an R-structure or even a text (several R-structures). However, Fox fails to subsume these R-structures composing the text into a superordinate node. As for anaphora, Fox’s theory is limited to third-person singular human reference and full NP to track a participant through a discourse. Thus, other forms of anaphor have not been studied and her theory does not entirely apply to other anaphoric forms. Furthermore, Fox only focuses on the principles which cause a full NP or a pronoun to be used. Consequently, referent identification or interpretation is not solved. It is not clear how to assign an appropriate referent to a pronoun. Despite the claim that the theory takes into consideration the relationship between language, cognition and society, it appears that Fox can hardly explore the relation from cognitive perspective. She renders great emphasis on static structures rather than reflect the dynamic aspect of discourse production and comprehension.

3.4 Linguistic Discourse Model (LDM) by Polanyi

47

3.4 Linguistic Discourse Model (LDM) by Polanyi 3.4.1 Main Idea LDM (Polanyi, 2001) distinguishes two types of basic discourse units: the elementary discourse constituent unit carried by the propositional content and the extrapropositional discourse operator such as but, because, on the other hand, by the way, and any proper names used as vocative. Discourse segmentation is based on the basic discourse units. For example, the sentence “I was reading but Mary was sleeping” can be analyzed under the LDM as a three-unit structure consisting of two elementary discourse constituent units (I was reading and Mary was sleeping) and a discourse operator but. As for complex discourse units, three types of higher structures are recognized: coordination, subordination and binary. Continuing activities such as lists, topic chains and narratives (including Written Yiddish Anecdote consisting of a title and Yiddish Anecdote which are made up of orientation, action, question and answer) are coordination structures. When a second unit is added to the first one, the information carried in the propositions will be compared and conjoined, and higher-level and more general information will be inferred at the coordination node. When more units are added, further comparisons will be made. Finally, all the units as the specific instances of the same general topic will be accommodated under the same higher-level mother node like example (3.3). (3.3)

C(oordination) John's bad habits

John addicts to alcohol

he likes smoking

he loves gambling

Subordination structures include elaborations on a point just made (reported speech or thought is another type of elaboration), digressions to discuss something else, asides, appositives, interruptions, parenthetical elements, and sentential subordination which consists of a matrix clause and its subordinated clause. The subordinated unit is usually regarded as the right daughter Y in an elementary tree like (3.4). But the superordinate unit does not dominate Y since the subordination relation is expressed by the label of the mother node. Different from the coordination case, the mother node inherits all the information of its left daughter X. S(ubordination) X

(3.4)

X

Y

48

3 Discourse Structure and Discourse Anaphora

S(ubordination)

Fig. 3.7 Binary structure

B(inary) a

c b

A binary structure consists of two elementary discourse constituent units “related logically (e.g. if/then, then/if, or, therefore), rhetorically (e.g. sum up), or interactionally (e.g. question/answer, warrant/response, error/repair)” (Polanyi, 2001: 270). A binary node dominates its two units. In example (3.5), the discourse parsing begins with building up a binary node, and then c is subordinated to the binary node as a digression (Fig. 3.7). (3.5) a. If you go to the library before half past four b. you may borrow the book; c. otherwise, you’ll wait until next week.

Under a coordination node there are usually two or more daughters, while under a subordination node or a binary node there are exactly two daughters. The discourse parse tree has propositional discourse constituent units at the leaves while operators are treated as clitics attached to the propositional hosts. One distinctive feature of the discourse parse tree is “right open”, that is, the attachment is allowed only at the right edge. Polanyi (2001: 271) maintains that “…discourse pops which resume an interrupted constituent will always close off the interrupting (elaborating or otherwise subordinated) constituents and make it impossible to attach (coordinate or subordinate) any subsequent DCUs to them.” Polanyi also describes the discourse parsing procedure. When a new discourse unit dk + 1 is provided, its relationship to the immediately preceding discourse unit dk should be considered. For instance, if elaboration relation holds between two discourse units dk and dk + 1, then dk + 1 is attached as the right sister of dk under a newly created S node. If there is no semantic or syntactic match, the dk + 1 is treated as an interruption of all ongoing discourse activities and is adjoined as the right daughter of a newly created S node. In this case, dk + 1 may probably introduce an entirely new discourse activity. So a higher-level mother node will be created to close off the old discourse activity and begin a new one.

3.4 Linguistic Discourse Model (LDM) by Polanyi

49

Written Yiddish Anecdote

Title

0

C-Yiddish Anecdote

C-Orientation/Narrative

a

b

c

C-Action/Narrative

B-mod

d

g

B(e or f)

e

S(h)

h

j

i

S(m)

m

f

p

S(n)

n

S(q)

q

s

t

u

r

o

Fig. 3.8 Discourse parse tree of example (3.6) (Polanyi, 2001: 272)

To illustrate the explanatory power of LDM, Polanyi reanalyzes Prince’s (1993) example and demonstrates its discourse parsing tree. First, Polanyi (2001: 272) segments the text according to LDM criteria: (3.6) What my father did. (a) A guy once went by an inn. (b) He left his horse and wagon outside (c) and went alone into the inn. (d) Inside the inn, (e) he ordered a couple of eggs (f) or some chicken (g) and ate it. (h) Then he got up (i) to travel further. (j) He goes outside the inn. (k) He looks around. (l) There’s no horse and no wagon. (m) He thought, (n) there was probably a thief among the people in the inn (o) that had stolen the horse and wagon. (p) He goes back into the inn (q) and shouts (r) “The horse and wagon should be returned.” (s) The thief got scared. (t) He quickly went out (u) and brought back the horse and wagon.

The discourse relations are of various types: List, Elaboration, Sum up, Subordinate Clause, Operator/Sentence, Interruption, Reported Thought, If/Then, Because, Antecedent/Consequence, Yiddish Anecdote, and Written Yiddish Anecdote. The discourse parse tree is shown in Fig. 3.8.

3.4.2 Comment LDM provides significant mechanisms for systematic investigation of discourse-level linguistic phenomena, which conduce to the interpretation of simple and complex units. Moreover, although no specific cognitive claim is made, within the LDM framework the relationship between discourse structure, sentence form and memory

50

3 Discourse Structure and Discourse Anaphora

limitations is investigated. Despite the obvious benefits, LDM can not avoid some weak points. First, the “right open” idea can not explain all the cases. In argumentative texts contrast and comparison are the techniques frequently used to support claims. Sometimes a series of contrast are employed with different agents of the sentences appearing by turns. In this case, the closed off unit may be resumed by the attachment of a new one. For example, whenever House B in (3.7) closes off House A, the closed off unit is reopened by a new unit describing House A. It is inappropriate to treat each pair of contrast as an entirely new discourse activity since the last two sentences requires considering House A and House B separately. (3.7) House A is in Palo Alto and House B in Portola Valley. House A has 2 bath, and House B, 4. House A is on a quarter acre, with a lovely garden, while House B is on 4 acres of steep wooded slope, with a view of the mountains. The owner of House A is asking $425K. The owner of House B is asking $600 K. That’s all I know about House A. And that’s all I know about House B. — Webber (1998: 115).

Another case may also explain the inadequacy of the “right open” idea. Again in argumentative texts, especially in Chinese, the writers tend to make concessions before putting forward stronger arguments. In the following example, the topic about women is introduced by Confucius’ words and emphasized that Chinese mothers were disparaged by the men other than their sons. And then Lu Xun(鲁迅) made a concession to list the progress in women’s liberation. According to Polanyi, the previous discourse activity in the first paragraph is closed off and the interrupting unit is no longer available. But the sentence “However, we can still hear it often that females at workplace moan in pain and the critics make a mockery towards modern women (不 不过我们还常常听到职业妇女的痛苦的呻吟, 评论家的对于新式女子 的讥笑)” reopens the closed off unit. Therefore, the attachment is not necessarily at the right edge and occasionally it can be back to the left. (3.8) … Confucius said “Only women and small-minded people are difficult to live in peace with. They will be conceited if one gets closely to them, or grumbling if one estranges them.” In his philosophy, women and small-minded people are grouped together, but it’s hard to say whether his own mother is also included. Later Taoist scholars were always respectful of their mothers at least superficially. However, the mothers in China are still treated in a contemptuous way by all men other than their sons. This is so called the achievements done after the May Fourth Movement greatly advocating the female revolution. However, we can still hear it often that females at workplace moan in pain and the critics make a mockery towards modern women. When they leave their boudoirs and enter society, they became new sources of jokes and discussion for the public. That’s because although they get exposure to the real world, they still have to rely on others to make a living. Therefore, it’s inevitable that they should bear the naggings and even insults from others. If we carefully observe the naggings of Confucius, we can find that his “difficulties” came from the nourishment. Therefore, neither “gets closely” nor “be estranged” is proper. —Lu Xun. On Women’s liberation. Xun Lu’s Essays

3.5 Other Related Studies

51

孔子曰: “唯女子与小人为难养也, 近之则不逊, 远之则怨。”女子与小人归在一 类里, 但不知道是否也包括了他的母亲。后来的道学先生们, 对于母亲, 表面上 总算是敬重的了, 然而虽然如此, 中国的为母的女性, 还受着自己儿子以外的一 切男性的轻蔑…… 这是五四运动后, 提倡了妇女解放以来的成绩。不过我们还常常听到职业妇女 的痛苦的呻吟, 评论家的对于新式女子的讥笑。她们从闺阁走出, 到了社会上, 其实是又成为给大家开玩笑, 发议论的新资料了。 这是因为她们虽然到了社会上, 还是靠着别人的“养”; 要别人“养”, 就得听人的 唠叨, 甚而至于侮辱。我们看看孔夫子的唠叨, 就知道他是为了要“养”而“难”, “ 近之”“远之”都不十分妥帖的缘故。 — 鲁迅. 关于妇女解放.《鲁迅文集》

Second, Polanyi fails to make it clear the information of which daughter the mother node of a binary structure may inherit. Third, discourse segmentation does not appeal to full stops. As a result, the basic constituent units may be very small and the segmentation work becomes trivial. Linguistic research in punctuation (Dale, 1991) has suggested that many uses of certain marks act as signals of discourse structure. However, LDM requires the actions of the narrative mainline to be represented as daughters of the same coordinate node, regardless of sentential boundary. That disarranges the discourse structures signaled by punctuations. Because of its trivialness, the segmentation fails to represent the writers’ cognitive state. In example (3.6), the third sentence is segmented into four units, with (g) as a sister of the mother node of the other three units. That sounds unreasonable to single out (g) because it continues the action of (e) and both of the actions take place inside the inn. Furthermore, the actions of (p) and (q) which belong to the same sentence are treated separately. It is the same case with (t) and (u).

3.5 Other Related Studies Other studies related to discourse anaphora touch upon discourse structure more or less although they never involve discourse relations. The discourse structures these studies deal with are episodes, paragraphs and discourse segments. In this section some representative ones are selected and introduced briefly. Givón (1983) represents the traditional view of discourse anaphora. Givón “makes explicit mention of something that sounds slightly structural and hierarchical” (Fox, 1987a: 158): human discourse is multi-propositional and chains of clauses are combined into thematic paragraphs within which continuity in discourse is discussed. In his distance model, referential distance, potential interference from other referents and topic persistence are the three measurements of topic continuity, among which referential distance is the major one. Referential distance refers to the number of clauses between a given anaphor and its antecedent. By means of quantitative measurement of referential distance (i.e. the number of clauses between subsequent mentions of a given referent), Givón displays an iconic relationship between referential distance and the amount of coding material used to maintain the continuity:

52

3 Discourse Structure and Discourse Anaphora

with the increase of referential distance, more coding material is required. However, in Givón’s view, all clauses are equal in the contribution to the measure of continuity whether a clause is an explanation of the previous one or an evidence of a claim. The discourse thus is composed of a string of clauses which follow one another but never form larger units to perform communicative functions. Since what really counts in Givón’s model is distance, attention is equally allocated to the old and new information. It is assumed that discourse structure is not hierarchical but flat and linear. Clancy (1980) and Marslen-Wilson et al. (1982) argue for episode model wherein the use of anaphora is considered to be a function of a particular discourse structure— episode. Although the notion of discourse structure is not fully embodied in his account, Clancy (1980) demonstrates the association between discourse units and full noun phrases by providing a detailed look at anaphora in short spoken English and Japanese narratives. Episode boundary is the main discourse structure which affects referential choice and indicates a new line of action starts. Marslen-Wilson et al. (1982) aver that referential choice is governed by discourse structure and the context of speaking. They have found the general pattern of anaphora that full noun phrases and proper names are employed for initial reference at the beginning of an episode while pronouns are used to maintain reference within an episode—an action sequence. This pattern is congruous with Fox (1987a) who resorts to discourse relations. The psychological evidence underlying this model is that at the beginning of an episode a particular referent is usually in a state of low focus whereas it is in a state of high focus within an episode. Although this model does not give a full account of the relationship between discourse structures and referential choice, it treats discourses as hierarchically organized rather than as strings of clauses and thus anaphora is no longer viewed as responsive to linear notion of distance as in Givón (1983). The weakness lies in the fact that the critical linguistic notion—episode—is not very well defined theoretically. Tomlin (1987) introduces attention model in which the use of anaphora is associated with the cognitive activities of attention and memory. An episode is defined ultimately “by the sustaining of attention on a particular paragraph level theme, a pragmatic instantiation of a rhetorical act” (Tomlin, 1987: 458). In his General Hypothesis Tomlin claims that referential choice may be determined by how focal attention is assigned in discourse production. When attention is sustained on a referent, a pronoun is used to maintain the reference within an episode. When attention is diverted, an episode boundary is reached and a full noun phrase is used to reinstate reference. The experiment of Tomlin and Pu (1991) corroborates the claim that full noun phrases are employed at the episode boundaries and pronouns are used within episodes. Chen (陈平, 1987) advances “micro continuity” and “macro continuity”. Gao (高原, 2005) makes a further distinction and calls paragraphs or episodes “macro discourse segments” while topic chains “micro discourse segments”. When a topic chain is interrupted, a more complicated referring form is used to refer to the subject in the subsequent clause. The break of the topic chain means the end of a micro discourse segment and that leads to the change of referring form. The referring form may be switched from zero anaphor to pronominals or to full noun phrases, or from

3.6 Discourse Representation Theory

53

pronominals to full noun phrases. Moreover, the weakened topicality of a referent under discussion (e.g. no longer a subject) will bring an end to the discourse segment and thus affect referential choice. In another case, insertion of some digression into a discourse segment may also cause a more complicated referring form to be selected.

3.6 Discourse Representation Theory Discourse Representation Theory (DRT) is first introduced in Kamp (1981). Diverging from traditional formal semantics, DRT emphasizes the dynamic feature of languages and the influence of context on the interpretation of sentences. DRT consists of three parts—syntactic tree, Discourse Representation Structure (DRS) and semantic interpretation. The syntactic rules essentially based on Generalized Phrase Structure Grammar serve to construct Discourse Representation Structures. To save space this part is left out. The information in a discourse is represented syntactically by a DRS which is graphically depicted as a box. And the major task of DRT is display the semantic representation of sentences with boxes. Therefore, what is the real concern in this paper is DRS, the core part.

3.6.1 Main Idea 3.6.1.1

DRS Construction

DRS construction procedure is also called DRS construction algorithm. When a discourse is being processed, sentences are entered one after another into the DRS (matrix DRS). Denied facts and temporary assumptions etc. are represented as separate DRSs embedded in the matrix DRS. The interpretation of each new sentence being currently processed relies on the previous information having already been processed. In return, the new sentence updates the old information and all of this becomes the prerequisite of the subsequent sentence processing. A DRS consists of two parts—universe and conditions. A set of variables in the top line of a DRS is called the “universe” of the DRS (UK) and the other items in the DRS are called “conditions” (ConK). DRS construction is restricted by a set of construction rules, for example the construction rule for negation and the construction rule for indefinite NPs. The former requires creating an embedded DRS prefixed by ¬ and to enter the negated sentence into the embedded DRS. The latter postulates the following rules: Introduce a “new” variable into the top line of the DRS. Predicate of this variable the descriptive content of NP. Predicate of this variable the rest of the sentence. Take for example (3.9). We begin with an empty box and enter the first sentence into it. According to the rule for indefinites, a new variable x is introduced into the top line of the DRS and is predicated of “man” and “came in”. In the same way we

54

3 Discourse Structure and Discourse Anaphora

enter the second sentence. According to the rule for negation, an embedded DRS prefixed ¬ is created and the negated sentence is entered into it. The DRS must be updated by taking the union of the universes and the condition sets. In the box “y = ?” is an incomplete condition and it means that the referent is not specified and needs resolving. The DRSs containing only complete conditions are called complete DRSs while those containing at least one incomplete condition are incomplete DRSs. (3.9) [A man]i came in. Hei didn’t dance with a woman.

As for quantification and conditional, the construction rules instruct us to create a complex condition containing a pair of DRSs combined by ⇒ which means that the right-hand DRS (consequent or nuclear scope) is asserted on the assumption of the left-hand DRS (antecedent or restriction). The donkey sentence in (3.10) and the sentence of (3.11) containing a universal quantifier are represented as follows. In the embedded DRS of (3.10), x and y are unspecified. So anaphora resolution is needed (Fig. 3.9). (3.10) If [a man]i owns a donkey, hei beats it. (3.11) Every student has a book.

Now semantic interpretation is considered. In DRT truth is not defined for sentences but for a DRS. The definition goes roughly as this: a DRS is true if and only if there is a value assignment which verifies all the conditions in it (Kadman, 2001:32). The definition induces existential quantification taking scope over the entire DRS. Kamp holds that a DRS is a partial picture of the world or the model and is embedded into the world/model. Matching a DRS with a part of the world or model is called “embedding functions”. In other words, a DRS is true if and only if there is an embedding function verifying all the conditions in it. Fig. 3.9 DRSs of examples (3.10) and (3.11)

u

v

x

z

man (u) donkey (v) u owns v

x=? y=? x beats y

x

y

students (x)

books (y) x has y

3.6 Discourse Representation Theory

3.6.1.2

55

DRT and Anaphora Resolution

Kamp’s ACCESSIBILITY constraint ensures that a pronoun will fall within the scope of the quantifier that binds its antecedent. We introduce ACCESSIBILITY as follows by summing up the definitions of Kamp (1981), Kadman (2001) and Asher (1993): Let DRS K = K 0 or K be subordinate to (graphically contained in) K 0 . i. if r and w are variables in the universe of DRS K , then r is accessible to w in K0; ii. if K graphically contains a DRS K  and w is in the universe of K  and r in K , then r is accessible to w in K 0 ; iii. if r is in the universe of K  and w is in K  and K  ⇒ K  is a complex condition of K , then r is accessible to w in K 0 . (i) is applied to simple DRS without embedded DRSs. In a simple DRS all the variables are in the same universe of K. (iii) may be illustrated by example (3.10) which contains a complex condition. So “man” may be accessible to “he” and “donkey” to “it”. (3.9) may instantiate (ii) in that the embedded DRS prefixed by ¬ is graphically contained by a larger DRS and the referent is in the larger DRS, thus it is accessible to “he”. But in (3.12) “man” is not accessible to “he” because it is the antecedent instead of the anaphoric pronoun that is in an embedded DRS. In addition, it is stipulated that ACCESSIBILITY depends on the scopes for quantifiers which determine the exact logical structure of the DRS. Every DRS has a quantifier attached to it and a quantifier can only bind variables that occur in the universe of the particular DRS to which it is attached. Accordingly, x is bound by the existential quantifier that only has scope over the DRS prefixed with ¬. The variables x and y are not bound by the same quantifier. (3.12) A man didn’t come in. He danced with a woman.

ACCESSIBILITY is the principal innovation in DRT analysis of anaphora, but it only provides some possibilities of identifying the appropriate referents. How can we know, for example (3.10), “he” refers to “man” and “it” to “donkey” rather than vise versa? Apart from scope which affects ACCESSIBILITY, anaphora resolution is also dependent on other syntactic factors. One of the syntactic factors is the agreement between the anaphoric pronoun and the antecedent noun phrase in gender and number and hence the anaphora problem in (3.10) is solved. Although Kamp (1981) does not state it explicitly, Kadman (2001: 39) makes a generalization about anaphora and scope: “Anaphora is possible only when the anaphoric element is under the scope of the quantifier that binds the antecedent variable.” That briefly summarizes Accessibility.

56

3 Discourse Structure and Discourse Anaphora

3.6.2 Comment What makes DRT distinguished from previous semantic theory is the context sensitive notion of meaning and the highly structured notion of context. In many cases, DRS is not the representation of a single sentence but rather of a series of sentences. That reflects the dynamic feature of information accumulation and transmission by making use of semantic cohesiveness between sentences. Moreover, DRT processes “donkey sentence” in a particular way, breaking away from the traditional semantics which treats indefinites as existential quantifiers. Instead, DRT postulates a complex DRS condition “K  ⇒ K  ”, describing the donkey sentence accurately and solve the problem of referent assignment. However, DRT has its own weak points. To begin with, DRT resorts to the agreement between the anaphoric pronoun and the antecedent noun phrase in gender and number. But DRT itself does not work when there are two antecedent noun phrases and the anaphoric pronoun agrees with both of them in gender and number as in (3.13). The theory provides no principle for the best or most likely choice. (3.13) A man met his father in the pub. He eschewed him.

Furthermore, the notion of semantic updating in DRT is rather simple. “To build a DRS for the discourse as a whole and thus to determine its truth conditions, one simply adds the DRS constructed for each constituent sentence to what one already had” (Asher, 1993: 256). Compared with a subsequent theory Segmented Discourse Representation Theory (see Chap. 4), the DRS construction procedure is inadequate to reflect the mental process of discourse reading comprehension, although DRT postulates a common form of mental representation for information. DRT is a relatively successful approach to abstract entity anaphora in a simple discourse. But it takes no account of discourse structure and its ACCESSIBILITY excludes the cases when a sentence and sentence clusters act as antecedents. So it seems ineffective to deal with the abstract entity anaphora in a complicated discourse, for example (3.14) and (3.7). (3.14) Human life expectancy gets longer and longer. At first glance this seems like good news. But hold it. Human life is not the only thing getting longer, so are television miniseries. Well you may say it only proves that Parkinson’s Law also fits human life: the entertainment expands to fit the time. (Asher, 1993: 340)

In (3.14) ACCESSIBILITY fails to assign the proposition in the first sentence to the anaphor “this”. Moreover, in the last sentence “it” refers to two aspects, that is, both human life expectancy and television miniseries are getting longer. The antecedent is not just the proposition expressed by a single sentence or even a sequence of adjacent sentences. But rather it may be the summation of the contents of non-adjacent sentences. DRT does not make that proposition accessible. Asher (1993) thinks that the critical reason is that the DRT researchers pay no attention to discourse structure and that leads to his leaping improvement of DRT.

3.7 Taking Stock

57

3.7 Taking Stock The theories discussed in this chapter relate to at least either discourse structure or discourse anaphora. Most of those concerning discourse structure consider a discourse as hierarchically organized (except Givón). And discourse relations may serve to glue the sentences together in a discourse and may also determine the hierarchy. Nevertheless, these theories lack an adequate and explicit hierarchical display of discourse. Of all the theories, only Polanyi’s LDM advances the processing of a high-level proposition—conjoining or inheriting. SDRT draws on this idea of proposition processing and makes a further development. Fox’s theory and Mann and Thompson’s theory embody plain and simple view of discourse hierarchy, without touching upon higher-level propositions. The other theories concerning discourse structure even fail to mention the hierarchy of proposition. Some scholars such as Fox, Polanyi, Mann and Thompson do not take a cognitive perspective and their theories never reflect mental representations of human beings. Others such as the advocates of episode model and attention model focus on cognitive and psychological factors, but they fail to manifest the dynamic process of discourse production and comprehension. The exception in this respect is DRT, which stresses the dynamic feature of language. That marks its difference from the other theories mentioned in this chapter, as well as the divergence from traditional semantics. Another aspect—anaphora resolution—also delimits the difference between DRT and the other theories. ACCESSIBILITY constraint in DRT makes appropriate referents accessible. However, the other theories show particular interest in the selection of anaphoric expressions. As for abstract entity anaphora which is under discussion in this dissertation, none of the theories can afford an effective resolution although DRT may be applied to account for only a few phenomena of the kind. In general, these theories either neglect discourse structure and discourse relations, or fail to provide resolution to anaphora. However, the antecedent of abstract entity anaphora may be several sentences clustering in a discourse and the anaphora resolution necessitates the consideration of discourse structure. Therefore, a theory taking into account of discourse structure and discourse relations is needed to deal with abstract entity anaphora.

Chapter 4

SDRT and Abstract Entity Anaphora

4.1 Introduction Basing his theory on DRT, Asher (1993) postulates Segmented Discourse Representation Theory (SDRT) to account for discourse structure and full range of abstract entity anaphora. Different from DRT, SDRT leaves out syntactic analysis, and thus consists of only two parts—discourse structure theory for abstract entity anaphora and model theory for abstract entities, with its focus on the former. Considering that model theory is the hotspot to which logicians usually show interest but linguists pay little attention and that this paper is confined to anaphora, model theory is accordingly excluded from this paper. Although SDRT is constructed in semantic terms and integrated with and interpreted by a semantic apparatus, this theory displays the cognitive structure of the interpreter. Moreover, Asher (1993) holds that the felicity of the anaphoric link depends on the interpreter’s ability to construct an appropriate scenario and to create an acceptable causal, temporal and other paths. Therefore, SDRT is at least in part a cognitive theory. Discourse segment, the central notion of SDRT, refers to a “chunk of text” which includes one or more sentences. Discourse segments are determined by a lot of factors—thematic content, logical and semantic structure, rhetorical structure of a discourse and the interpreter’s knowledge of the subject matter. Discourse segments may be used as anaphoric antecedents. A segmented DRS (SDRS) is made up of the simplest form of a group of DRSs, and a set of conditions on the DRSs which assert certain discourse relations holding of certain DRSs. These DRSs are constructed from segments of the text. To distinguish these DRSs from those in DRT, they are called constituents of SDRS or simply constituents. SDRSs are determined by a set of discourse relations. Section 4.2 distinguishes major notions about discourse relations and the notion of topic in SDRT. Section 4.3 explains the SDRS construction procedure while Sect. 4.4 introduces the constraints on abstract entity anaphora, the functions in referent resolution and the application to various kinds of abstract entity anaphora. Then, the application of SDRT to Chinese abstract entity anaphora is demonstrated © Central China Normal University Press 2023 D. Liu, Abstract Entity Anaphora in Argumentative Texts, https://doi.org/10.1007/978-981-99-4630-3_4

59

60

4 SDRT and Abstract Entity Anaphora

in Sect. 4.5. Section 4.6 displays another application of SDRT—application to backwards anaphora. Section 4.7 provides some support from cognitive psychology and Sect. 4.8 summarizes the whole chapter.

4.2 Discourse Relations and Topic 4.2.1 Discourse Relations To begin with, we must make clear the essential symbols particularly used in SDRT. Greek letters α, β, γ , δ etc. stand for constituents—DRSs or subDRSs. These letters are used in definitions, rules and axioms, but in boxes or constituent graphs k, k0 , k1 , k2 , etc. are employed. Shadow letters with primes or subscripts like K , K  , K 1 , K 2 , denote SDRSs. “ < ” represents the constituenthood relation while “