Variation in Metonymy: Cross-linguistic, Historical and Lectal Perspectives 9783110455830, 9783110453522

The monograph presents new findings and perspectives in the study of variation in metonymy, both theoretical and methodo

219 105 7MB

English Pages 380 [382] Year 2016

Report DMCA / Copyright

DOWNLOAD PDF FILE

Table of contents :
Contents
Introduction
Part One: THEORETICAL PRELIMINARY
1 Demarcation and variability of metonymy
1.1 Conceptual metonymy
1.1.1 Definition/demarcation of metonymy
1.1.2 Internal typology of metonymy
1.1.3 The conventionalization of metonymy
1.2 Metonymy: A variationist CL view
1.2.1 The cross-linguistic dimension
1.2.2 The historical dimension
1.2.3 The lectal dimension
1.3 Metonymy: Semasiology and onomasiology
1.3.1 The semasiological perspective
1.3.2 The onomasiological perspective
1.4 The methodological state of the art in metonymy research
1.4.1 Why CL research needs an empirical approach
1.4.2 Overview of methods in CL metonymy research
1.5 Summary: Research gaps
2 Metonymy in expressions
2.1 Models for analyzing figurative composite expressions
2.1.1 Compositionality and analyzability
2.1.2 Fauconnier and Turner’s blending model
2.1.3 Benczes’s analysis of noun-noun compounds
2.1.4 Geeraerts’s prismatic model
2.2 Metonymic mappings in Chinese expressions
2.2.1 Bringing the prismatic model to Chinese metonymic expressions
2.2.2 Various paths of metonymy in Chinese expressions
2.3 Some criteria for classification
2.3.1 Global metonymy versus parallel local metonymies
2.3.2 Internal metonymy versus external metonymy
2.3.3 Differentiation between alternative prismatic structures
2.4 Summary
Part Two: CASE STUDIES
3 The cross-linguistic perspective: Metonymies for PERSON
3.1 Methodology
3.1.1 The Chinese data resources
3.1.2 The English data resource
3.1.3 Summary of the methodology
3.2 Results
3.2.1 Variation in metonymic patterns for PERSON in general
3.2.2 Variation in metonymic patterns for different kinds of person
3.2.3 Variation in metonymic sources of a particular pattern
3.3 Discussion
3.3.1 Paragons in Chinese culture
3.3.2 The carriage-clothing system in Chinese culture
3.3.3 The residence system in Chinese culture
3.3.4 Culture-bound conceptualizations of body parts
3.3.5 Collectivism versus individualism
3.3.6 Section summary
3.4 Summary
4 The diachronic perspective: Metonymies for WOMAN
4.1 Methodology
4.1.1 The Corpus of Historical Chinese
4.1.2 Metonymy identification
4.1.3 Metonymy quantification and analysis
4.2 Diachronic variation in metonymic patterns
4.2.1 Exploring and visualizing the data
4.2.2 Targets with relatively stable diachronic variation
4.2.3 Targets with a dominant trend in diachronic variation
4.2.4 Targets with highly fluctuating diachronic variation
4.2.5 Section summary
4.3 Interactions with stylistic variation
4.3.1 Stylistic variation in metonymic patterns
4.3.2 Introduction of new metonymic items
4.3.3 Section summary
4.4 Summary
5 The lectal perspective: Metonymies for GOVERNMENT
5.1 Methodology
5.1.1 Data collection
5.1.2 The variables
5.1.3 The mixed-effects logistic regression model
5.2 The general regression model for GOVERNMENT
5.2.1 General impact of the predictors
5.2.2 Specific influence of fixed effects
5.2.3 The random-effect variable of verbs
5.3 The separate regression model for MAINLAND CHINESE GOVERNMENT
5.3.1 The separate mixed-effects model
5.3.2 The lectal variation between Mainland and Taiwan Chinese
5.4 Summary
Conclusion
Main findings and contributions
Limitations and further directions
References
Appendix
A. Chinese/English metonymic mapping datasets
B. Distributions of main patterns for specific targets
C. Distributions of sources under a particular pattern for specific targets
D. References for meaning identification in the corpus
E. Classification of source/target concepts in Case 2
F. Contributions of patterns for interpreting the MDS dimensions
Index
Recommend Papers

Variation in Metonymy: Cross-linguistic, Historical and Lectal Perspectives
 9783110455830, 9783110453522

  • 0 0 0
  • Like this paper and download? You can publish your own PDF file online for free in a few minutes! Sign Up
File loading please wait...
Citation preview

Weiwei Zhang Variation in Metonymy

Cognitive Linguistics Research

Editors Dirk Geeraerts Dagmar Divjak John R. Taylor Honorary editors René Dirven Ronald W. Langacker

Volume 59

Weiwei Zhang

Variation in Metonymy Cross-linguistic, historical and lectal perspectives

This book was partially sponsored by grants from MOE (Ministry of Education in China) Project of Humanities and Social Sciences (Project No. 15YJC740136).

ISBN 978-3-11-045352-2 e-ISBN (PDF) 978-3-11-045583-0 e-ISBN (EPUB) 978-3-11-045365-2 ISSN 1861-4132 Library of Congress Cataloging-in-Publication Data A CIP catalog record for this book has been applied for at the Library of Congress. Bibliographic information published by the Deutsche Nationalbibliothek The Deutsche Nationalbibliothek lists this publication in the Deutsche Nationalbibliografie; detailed bibliographic data are available on the Internet at http://dnb.dnb.de. © 2016 Walter de Gruyter GmbH, Berlin/Boston Printing and binding: CPI books GmbH, Leck ♾ Printed on acid-free paper Printed in Germany www.degruyter.com

| To my parents: ZHANG Zhiqing and WEI Guilan 

Acknowledgement This book builds on my PhD dissertation, “Variation in metonymy: A corpusbased Cognitive Linguistic approach”, which I defended at the Research Unit of Quantitative Lexicology and Variational Linguistics (QLVL) at the University of Leuven in 2013. Substantial revisions have then been made. I wish to thank the many colleagues, friends and family who helped me throughout these years. First and foremost, I would like to sincerely thank my supervisors, Dirk Geeraerts and Dirk Speelman, for giving me the opportunity to pursue a PhD in the QLVL research unit at KULeuven. Dirk Geeraerts has provided insightful supervision and feedback about my work and given me continuous inspiration and unfailing support. Similarly, my deep gratitude goes to Dirk Speelman for his critical and detailed comments at all stages of my research. He has kindly guided me through the field of quantitative methods for linguistics and offered me exceptional hand-on sessions on statistics. Without the help of these two generous scholars, I would not have been able to finish the book. I am also grateful to all my QLVL colleagues for their most valuable help and their interest in my work. Special thanks are due to, in alphabetic order, Alean Anishchanka, Andrea Pizarro Pedraza, Dirk De Hertog, Eline Zenner, Jocelyne Daems, Jack Grieve, Kris Heylen, Natalia Levshina, Tom Ruette and Yves Peirsman for their collaboration, encouragement and invaluable comments. I am deeply indebted to my PhD committee, Hubert Cuyckens, Kurt Feyaerts, Nicolas Standaert and Kathryn Allan, who helped me to improve the text with their very insightful comments. I also gratefully acknowledge Antonio Barcelona, Gitte Kristiansen, Stefan Evert and Josefien Sweep for their expert suggestions during various conferences and their interest in my project. My thanks also go to Dennis Tay for his encouragement and advice on the publishing of this book. I would also like to thank the anonymous reviewers who provided detailed feedback which shaped the final version of this book. I am grateful to Birgit Sievert, Julie Miess and Nancy Christ at De Gruyter for their help throughout this work. In addition, I would like to express my gratitude to the Institute of Linguistics at Shanghai International Studies University for providing a friendly and inspiring academic environment while I was revising the book. Finally, most importantly, my deepest gratitude goes to my parents and to my husband, Jia Wan. This book would not have been possible without their unconditional love and support.

List of Symbols Abbreviations



CL OED HTOED DCMS DCSW CHC MIP MC TC UDN PTT PD TYC MCGOV Cn Eng AIC



Cognitive Linguistics Oxford English Dictionary (online) Historical Thesaurus of the Oxford English Dictionary (online) Dictionary of Chinese Metonymic Senses Dictionary of Chinese Substitutive Words Corpus of Historical Chinese Metaphor Identification Procedure (by the Pragglejaz Group) Mainland Chinese Taiwan Chinese United Daily News Professional Technology Temple People’s Daily Tianya Club Mainland Chinese government Chinese metonymic mapping dataset English metonymic mapping dataset Akaike’s Information Criterion

Notation



“ ” ~ [ ] ( ) italics small caps % n

Meanings; English equivalents; quotations Headings in the taxonomic structures of the HTOED Glosses of Chinese expressions; source/literal readings of expressions in the context Target readings of metonymic expressions in the context Linguistic expressions; emphasis; romanized Chinese examples; terminology Metonymic/metaphorical sources, targets and patterns/mappings; concepts Relative frequency Absolute frequency

Contents Introduction| 1 Part One: THEORETICAL PRELIMINARY 1 1.1 1.1.1 1.1.2 1.1.3 1.2 1.2.1 1.2.2 1.2.3 1.3 1.3.1 1.3.2 1.4 1.4.1 1.4.2 1.5

Demarcation and variability of metonymy | 7 Conceptual metonymy | 7 Definition/demarcation of metonymy | 7 Internal typology of metonymy | 20 The conventionalization of metonymy | 26 Metonymy: A variationist CL view | 27 The cross-linguistic dimension | 27 The historical dimension|31 The lectal dimension|33 Metonymy: Semasiology and onomasiology | 35 The semasiological perspective | 36 The onomasiological perspective | 37 The methodological state of the art in metonymy research | 40 Why CL research needs an empirical approach | 41 Overview of methods in CL metonymy research | 42 Summary: Research gaps | 46

2 2.1 2.1.1 2.1.2 2.1.3 2.1.4 2.2 2.2.1

Metonymy in expressions | 47 Models for analyzing figurative composite expressions | 48 Compositionality and analyzability | 48 Fauconnier and Turner’s blending model | 50 Benczes’s analysis of noun-noun compounds | 53 Geeraerts’s prismatic model | 56 Metonymic mappings in Chinese expressions | 63 Bringing the prismatic model to Chinese metonymic expressions | 63 Various paths of metonymy in Chinese expressions | 65 Some criteria for classification | 79 Global metonymy versus parallel local metonymies | 79 Internal metonymy versus external metonymy | 85 Differentiation between alternative prismatic structures | 89 Summary | 92

2.2.2 2.3 2.3.1 2.3.2 2.3.3 2.4

xii | Contents

Part Two: CASE STUDIES 3 3.1 3.1.1 3.1.2 3.1.3 3.2 3.2.1 3.2.2 3.2.3 3.3 3.3.1 3.3.2 3.3.3 3.3.4 3.3.5 3.3.6 3.4

The cross-linguistic perspective: Metonymies for PERSON | 97 Methodology | 98 The Chinese data resources | 98 The English data resource | 110 Summary of the methodology | 119 Results | 121 Variation in metonymic patterns for PERSON in general | 121 Variation in metonymic patterns for different kinds of person | 127 Variation in metonymic sources of a particular pattern | 137 Discussion | 152 Paragons in Chinese culture | 153 The carriage-clothing system in Chinese culture | 155 The residence system in Chinese culture | 168 Culture-bound conceptualizations of body parts | 172 Collectivism versus individualism | 176 Section summary | 178 Summary | 179

4 4.1 4.1.1 4.1.2 4.1.3 4.2 4.2.1 4.2.2 4.2.3 4.2.4 4.2.5 4.3 4.3.1 4.3.2 4.3.3 4.4

The diachronic perspective: Metonymies for WOMAN | 181 Methodology | 182 The Corpus of Historical Chinese | 182 Metonymy identification | 185 Metonymy quantification and analysis | 188 Diachronic variation in metonymic patterns | 190 Exploring and visualizing the data | 191 Targets with relatively stable diachronic variation | 206 Targets with a dominant trend in diachronic variation | 230 Targets with highly fluctuating diachronic variation | 242 Section summary | 263 Interactions with stylistic variation | 265 Stylistic variation in metonymic patterns | 267 Introduction of new metonymic items | 276 Section summary | 281 Summary | 284

Contents | xiii

5 5.1 5.1.1 5.1.2 5.1.3 5.2 5.2.1 5.2.2 5.2.3 5.3 5.3.1 5.3.2 5.4

The lectal perspective: Metonymies for GOVERNMENT | 286 Methodology | 287 Data collection | 287 The variables | 298 The mixed-effects logistic regression model | 303 The general regression model for GOVERNMENT | 308 General impact of the predictors | 309 Specific influence of fixed effects | 310 The random-effect variable of verbs | 313 The separate regression model for MAINLAND CHINESE GOVERNMENT | 315 The separate mixed-effects model | 316 The lectal variation between Mainland and Taiwan Chinese | 317 Summary | 319

Conclusion | 322 Main findings and contributions | 322 Limitations and further directions | 326 References | 329 Appendix | 351 A. Chinese/English metonymic mapping datasets | 351 B. Distributions of main patterns for specific targets | 352 C. Distributions of sources under a particular pattern for specific targets | 357 D. References for meaning identification in the corpus | 361 E. Classification of source/target concepts in Case 2 | 361 F. Contributions of patterns for interpreting the MDS dimensions | 362 Index | 365

Introduction Conceptual Metonymy Theory, which was initiated by Lakoff and Johnson (1980), has been pervasive in the Cognitive Linguistics (henceforth CL) research of the last 35 years. Although there remain many important disagreements on a number of specific issues in Conceptual Metonymy Theory, cognitive linguists essentially agree on the conceptual nature and the fundamental importance of metonymy (Barcelona and Benczes 2011: 1). More specifically, CL scholars have reached a consensus on the notion that metonymy is a basic cognitive and conceptual mechanism. Another basic assumption in CL is that large areas of language and cognition are motivated by human embodiment, and therefore metonymy is experientially grounded (Lakoff and Johnson 1980: 39; Barcelona 2011: 8; Brdar-Szabó and Brdar 2012: 728). As Geeraerts and Kristiansen (2014: 205) point out, the experiential view of meaning in CL actually has at least two main aspects: first, “we are embodied beings” and “our organic nature influences our experience of the world”; and second, “we also have a cultural and social identity, and our language may reveal that identity”. In other words, the embodied basis of conceptual metonymy is bodily, cultural and social at the same time. Stressing the idea of “putting the body back into the mind” (Johnson 1987: xxxvi), cognitive linguists have shown much interest in demonstrating the importance of the body in concepts and conceptualizations. If we accept the idea that bodily experience is shared by humans, and in consequence universal to some extent, we should expect that human languages display some omnipresent similarities. In addition, the similar specifics of our cultural/social environments could trigger metonymies which are universal in different languages. Given this experiential view of metonymy, cognitive linguists naturally pay special attention to confirming the universality and omnipresence of metonymy in different languages and cultures. Much success has been achieved in uncovering the cross-linguistic availability of metonymy (e.g. Velasco 2000; Mol 2004; Radden 2004; Zhang 2008; Nissen 2011; Siahaan 2011). The notion that metonymic thought is embodied can explain the motivation for certain universally attested metonymies. The cross-linguistic attestation of certain metonymies, however, cannot conceal another important aspect of conceptual metonymy, which is its culture/social-sensitivity. As explained by Brdar-Szabó and Brdar (2012: 729), the acknowledgement of the standpoint that metonymies are universal in principle and present everywhere in language does not lead to the conclusion that metonymic processes are unconstrained and all human languages avail themselves of metonymic processes in exactly the same way. In

2 | Introduction

fact, metonymic processes could be constrained by a number of cultural and social factors, and the general principles of metonymy are not the same in all languages (Lakoff 1987: 78). At the same time, both the predominantly semantic perspective and the usage-based nature of CL foster a social conception of language (Geeraerts 2005; Geeraerts et al. 2010a), and hence motivate an examination of metonymy from a sociolinguistic point of view; that is, an exploration of metonymy variation in its cultural-social environment. In brief, the present book chiefly explores variation in metonymy from an onomasiological perspective, asking: given a concept, how do we name it? The theoretical objective of this book is to provide a systematic way of disclosing how certain cultural-social factors exert an influence on the onomasiological variation in metonymic conceptualization. As Geeraerts (2006c: 31) points out, the importance of socially and culturally determined variation for CL must inevitably lead to an empirical revolution, which will require the “adoption of the quantitative, empirical methodology that is dominant in sociolinguistic research at large”. Hence, the methodological objective of the present book is to demonstrate the importance of including corpus data in metonymy research as well as employing quantitative statistical methods to explore and visualize the variation. In particular, this book intends to broaden the perspective of CL research on metonymy by raising the following three overarching research questions: 1. To what extent does conceptual metonymy exhibit variation? 2. Where is metonymy variation most likely to occur? 3. What are possible driving forces of metonymy variation? The book consists of two main parts: Part One “Theoretical preliminary” and Part Two “Case studies”. Part One consists of Chapter 1 and Chapter 2, which detail the theoretical background of the present book. Chapter 1 mainly deals with the basic notions of conceptual metonymy in the framework of CL and gives an overview and evaluation of previous research on metonymy from different theoretical and methodological perspectives. Chapter 2 primarily reviews previous research regarding the demarcation of metonymy in linguistic expressions and provides the methodological guidelines used to identify metonymies in the case studies. Part Two constitutes the main body of the book and reports on three case studies in Chapters 3, 4 and 5. Chapter 3 examines the cross-linguistic variation in the metonymic conceptualization of PERSON (Case 1); Chapter 4 discusses the historical variation in the metonymic conceptualization of WOMAN (Case 2); Chapter 5 focuses on the lectal variation in naming options with respect to the

Introduction | 3

metonymy (Case 3). In all three chapters, I will introduce the methodology used in the given case study, describe statistically significant variation, and discuss the possible driving forces behind this variation. The three case studies are designed for the purpose of filling the research gaps that will be summarized in the last section of Chapter 1, and they are interlinked theoretically (see Figure 1). First, all of the case studies take an onomasiological perspective, focusing on variation in naming a concept (see Section 1.3.2 for a detailed summary of onomasiological research on metonymy). While both Case 1 and Case 2 address the choice of different metonymies for a particular target, Case 3 studies the alternation of metonymic versus literal designations for a given concept. The choice of the metonymic pattern/source is worth detailed investigation, as the perspective imposed by it constrains the way in which the target is viewed (Barcelona 2011: 13). At the same time, alternation in the choice of metonymic versus literal naming is also interesting, as it helps us understand the function of metonymy. Second, Case 1 attempts to detect important cultural effects at the level of cross-linguistic variation; it is reasonable to further explore the historical and lectal variation in metonymy, given the fact that culture by definition has historical and social dimensions (see Case 2 and Case 3). At the same time, Case 2 and Case 3 intend to show that metonymy variation is also due to another set of cultural influences, namely stylistic effects (see Feyaerts and Brône 2004). Third, the arrangement of the three case studies can be understood from a “time” perspective. Case 1 provides a contrastive study on metonymy in two typologically unrelated languages from a panchronic perspective; Case 2 zooms in on one specific language and investigates diachronic variation in metonymy, focusing on language uses in the history of Chinese language and culture; Case 3 extracts a particular time scale in the history of Chinese, providing a synchronic comparison of variation in two lectal varieties1 of contemporary Chinese.

PLACE NAME FOR GOVERNMENT

|| 1 “Lectal variety” refers to all types of language varieties or lects, such as regional dialects, sociolects, basilects, acrolects, idiolects, registers and styles (Geeraerts 2006c: 30; Kristiansen and Dirven 2008a: 47).

4 | Introduction





    

  



    





  

Fig. 1: Theoretical interlinking of the case studies

Finally, in the final chapter of the book, I will summarize its main findings and contributions, indicate its limitations and discuss directions for future research.

| Part One: THEORETICAL PRELIMINARY Overview of the theoretical framework

6 | Overview of the theoretical framework

Chapter 1 begins with a review of Conceptual Metonymy Theory, in which major issues in CL metonymy research will be introduced and the working definition and typology of metonymy in the present study will be given. Subsequently, three main perspectives in the previous research on variation in metonymy are reviewed. A summary of previous metonymy research from both semasiological and onomasiological perspectives is then given. Next, the methodological state of the art in contemporary metonymy studies and the advantages and drawbacks of various methods are discussed. Following on from the literature review, the research gaps that remain to be filled will be identified. Chapter 2 focuses on the demarcation and identification of metonymies in linguistic expressions, especially in Chinese composite expressions. This chapter first compares three different models for analyzing figurative composite expressions, because we need a reliable way to identify metonymic mappings in expressions in the case studies. The prismatic model is then chosen for the present case, and several criteria for identifying metonymic mappings in Chinese expressions will be discussed. In brief, Part One situates the present book in the broader context of conceptual metonymy research, and explains how the research questions formulated in the Introduction were inspired by previous metonymy studies.

1 Demarcation and variability of metonymy This chapter gives an overview of metonymy research in CL from both theoretical and methodological perspectives. The main aim of this book is to open up an important but underexplored territory in contemporary metonymy studies, namely, how metonymy differs in various dimensions. The chapter clarifies a number of issues in detecting metonymy variation, which are related to the range, typology and conventionalization of metonymy. Section 1.1 summarizes the core notions of Conceptual Metonymy Theory. Section 1.2 reviews the previous research on variation in metonymy. In Section 1.3, two perspectives in semantic studies as well as the relevant metonymy research are sketched. Section 1.4 briefly introduces the methodological state of the art in metonymy research. Finally, Section 1.5 outlines the research gaps.

1.1 Conceptual metonymy In the field of CL, metonymy is regarded as a cognitive phenomenon rather than figures of speech, as in traditional rhetoric (see Arata 2005). While the CL community essentially agrees on both the conceptual/cognitive nature and the fundamental importance of metonymy (Lakoff and Johnson 1980; Panther and Radden 1999; Barcelona 2003a; Benczes et al. 2011), a number of disagreements on specific issues remain (Barcelona and Benczes 2011: 1). Two main problems concern the external boundaries and the internal typology of metonymy. The former boils down to the definition of metonymy and how it is demarcated from other cognitive mechanisms like conceptual metaphor, the “active zone” and so forth (see Section 1.1.1); the latter pertains to the classification of metonymic patterns (Section 1.1.2). Another issue relating to conceptual metonymy rests in its conventionalization (Section 1.1.3).

1.1.1 Definition/demarcation of metonymy Although Lakoff and Johnson’s (1980) Metaphors We Live By was primarily devoted to metaphor, it contains one chapter on metonymy. This chapter inspired the growing interest in metonymy in CL and established Conceptual Metonymy Theory. So far, CL has largely been concerned with discovering the fundamental

8 | Demarcation and variability of metonymy

importance and the impressive range of metonymy (Panther and Radden 1999; Panther and Thornburg 2003, 2007; Peirsman and Geeraerts 2006b; Benczes et al. 2011; Koch 2011). Four uncontroversial properties have been proposed as the “core elements” of metonymy from the cognitive linguistic view: 1) it has a fundamentally conceptual nature; 2) it has an experiential basis; 3) it can be at the root of certain cognitive models; and 4) it involves experientially and conceptually contiguous elements (Barcelona 2011: 8). These properties are widely acknowledged by cognitive linguists and form the kernel of Conceptual Metonymy Theory. However, other properties of metonymy have not been universally accepted in CL, especially with regard to its definition/demarcation. Two mainstream views, domain-based and prototype-based, will be discussed below. These two views have different emphases with respect to the “core elements” of metonymy: the domain-based view focuses on the third property, while the prototype-based view shifts attention to the fourth property.

1.1.1.1 The domain-based view In an approach initiated by Lakoff and Johnson (1980: 36) and Lakoff and Turner (1989: 103), metonymy is often defined in terms of within-domain conceptual mapping. This definition depends to a large extent on the cognitive idea that concepts constitute more complex semantic-conceptual structures, that is, conceptual domains. In the CL literature, other terms such as domain matrix (Croft 2002 [1993]), frame/scenario (Blank 1999; Koch 1999a; Panther and Thornburg 1999), idealized cognitive model (ICM) (Radden and Kövecses 1999), and functional domain (Barcelona 2011) have been proposed to replace domain. Although using different terminologies, all these definitions try to define metonymy in contrast to metaphor by invoking the number of conceptual domains (or domain matrixes, etc.) involved in the conceptualization process: metaphor involves two domains, metonymy only one (Geeraerts 2010d: 215). For instance, if you call a crafty person a fox, you metaphorically map the properties of the animal domain onto the human domain. Conversely, if you call a stupid person a stupid head, you just use the body part of a person to refer to a person, and both the source BODY PART and the target PERSON belong to the same domain of human being. Croft (2002 [1993]) notes that metonymy sometimes seems to regularly map across domain boundaries. In Proust is tough to read, for instance, the source PROUST belongs to the domain of human being; the target PROUST’S WORK, to that of creative activity. Therefore, Croft (2002 [1993]: 177–179) proposes to replace domain with domain matrix and formulates a more dynamic view of metonymy:

Conceptual metonymy | 9

a metonymic mapping occurs within a single domain matrix, not across domains (or domain matrices), and metonymy involves “domain highlighting” by making primary a domain that is secondary in the literal meaning. Using the terms frame and scenario (see Fillmore 1975, 1977, 1985), Blank (1999), Koch (1999a) and Panther and Thornburg (1999) point out that in metonymy, target and source concepts are not isolated but parts of greater conceptual networks, that is, frames/scenarios. A more widely accepted definition of metonymy, inspired by Langacker’s (1993) active zone theory2 and Lakoff’s (1987) framework of ICMs, is proposed by Radden and Kövecses (1999: 21): “metonymy is a cognitive process in which one conceptual entity, the vehicle, provides mental access to another conceptual entity, the target, within the same idealized cognitive model”. In spite of its popularity (see Gibbs 1994; Barcelona 2002; Truszczyńska 2003; Benczes 2011), however, the single-domain approach3 has received much criticism (Feyaerts 1999; Riemer 2001; Haser 2005; Peirsman and Geeraerts 2006b; Panther and Thornburg 2007), mainly in regard to the efficacy of the notion of domain or domain matrix, which has not been clearly defined yet. As Panther and Thornburg (2007: 240) point out, “what constitutes one domain or domain matrix has to date not been satisfactorily elucidated in the literature”. Feyaerts’s (1999: 318) criticism emphasizes that semantic/conceptual structures as well as domains (matrices) are “experientially based and consequently to a large extent individually determined” and, further, that “drawing distinct boundaries around a domain (matrix) always reflects an arbitrary intervention by a linguist”. At the same time, it is easy to find counterexamples to the singledomain approach. Ruiz de Mendoza Ibáñez and Otal Campo (2002) suggest that the “domain highlighting” (Croft 2002 [1993]) phenomenon may also occur in metaphor. Feyaerts (1999: 319) points out that “one cannot exclude metaphorical mappings taking place within the boundaries of a domain matrix”. For example, the conceptual relationship between the concepts ANIMAL and HUMAN in the expression He is a fox is traditionally regarded as a conceptual metaphor because the mapping occurs across the two domains of animal and human. The question is, however, how it can be demonstrated that these two domains do not belong to the same domain matrix of mammal. Without certain theoretical

|| 2 The notion of active zones is introduced by Langacker (1984: 177) and defined as “those portions of a trajector or landmark that participate directly in a given relation”. 3 The “single-domain approach” refers to those definitions of metonymy that see it as taking place within a single cognitive domain (matrix), frame or ICM, in contrast to metaphor as crossdomain mapping.

10 | Demarcation and variability of metonymy

or practical criteria, it is hard to provide a clear description of a domain or domain matrix. Therefore, the demarcation of metaphor and metonymy should not be based only on consideration of the domains involved in the conceptual process (Riemer 2001: 383). Discerning the problem of the domain (matrix) hypothesis, Barcelona (2011: 52) refines the definition of metonymy as “an asymmetric mapping of a conceptual domain, the source, onto another domain, the target. Source and target are in the same functional domain and are linked by a pragmatic function, so that the target is mentally activated”. Then, Barcelona (2011: 53) also revises the definition of metaphor as having the properties of “a symmetric mapping” and “source and target…either in different taxonomic domains and not linked by a pragmatic function, or…in different functional domains”. Barcelona’s definitions modify two aspects of the previous definitions. First, he explicitly restricts the “domain” of metonymy to a “functional domain” (2011: 52). The example The piano is a real genius would be identified as a metonymy by this definition. Although PIANO and PIANO PLAYER are in two different taxonomic domains, they are in the same functional domain (i.e. a CONCERT or ORCHESTRA frame), and linked by a pragmatic function (i.e. INSTRUMENT–AGENT). The term functional domain is actually in accordance with the terms frame and scenario4. However, the notions of functional domain, frame/scenario and ICM, which are as ambiguous as domain, may not solve the problem of how to differentiate metonymy and metaphor either. Second, Barcelona also pays attention to the specific nature of conceptual mapping as a separate issue from that of the range of an expression’s extension vis-à-vis the boundaries of a functional domain; metonymy’s mapping is asymmetric and metaphor’s mapping is symmetric. Mapping implies a projection of structure from the source onto the target, so that the projected structure imposes (some of) its elements, traits and properties onto their counterparts in the target (Barcelona 2011: 12). In metaphor, both source and target share at least part of their abstract image-schematic structure, therefore, metaphorical mapping is symmetric. A metonymic source, however, does not project its conceptual structure onto that of the target by means of a systematic matching of counterparts, as in metaphor. The metonymic projection imposes a conceptual and linguistic perspective from which the target is activated, and this activation

|| 4 Barcelona (2011: 12) points out that the notion of “domain matrix” should be distinguished from that of ICM and “frame”, although it overlaps with them. He uses the term “functional domain” as equivalent to frames and ICMs, which are models of areas of experiential knowledge and “not exclusively attached to lexical or constructional meaning”.

Conceptual metonymy | 11

entails a shift of attention from source to target. In this sense, metonymy can be regarded as an asymmetric mapping. The symmetry of mappings in Barcelona’s terms should not be confused with two other notions: the directionality and reversibility of mappings. Symmetry of mappings is analogous to the number of mappings taking place in a metaphorical or metonymic process. Metaphors are characterized by whole sets of systematic conceptual correspondences between certain elements of the source domain, on one hand, and their counterparts that are elements of the target domain, on the other (Brdar-Szab and Brdar 2011: 221). Metonymy, however, normally works on the basis of a single correspondence instead of a set of correspondences as in metaphor (Ruiz de Mendoza Ibáñez and Peña Cervel 2002). This is why we regard metaphor as a symmetric mapping but metonymy as an asymmetric one. The notions of directionality and reversibility describe different properties of metaphor and metonymy. Metaphors typically employ a mapping from a more concrete concept as the source to a more abstract concept as the target. Therefore, metaphorical mappings are normally unidirectional, and the source and target are not reversible (Barcelona 2011: 15; Brdar-Szab and Brdar 2011: 220). Metonymic mappings are also unidirectional, which implies that the source and the target in metonymy do not simultaneously map onto each other (Barcelona 2011: 15). However, metonymy differs from metaphor in the reversibility of its mappings. Metonymy is a reversible process, as either of the two conceptual entities may stand for the other (Radden and Kövecses 1999: 22). Barcelona (2011: 15) uses two examples from Radden and Kövecses (1999: 31) to demonstrate that reversibility is not the same thing as bidirectionality5. Example (1) shows the reversibility of metonymy: metonymic mappings can proceed in either direction, from part to whole, or the other way round. However, in neither metonymy is the mapping bidirectional. For instance, in the sentence America will prevail, the source AMERICA activates the target UNITED STATES, but the target does not activate the source at the same time.

|| 5 Brdar and Brdar-Szabό (2011) seem to have a contrary view. They regard the directionality of mapping as a way to distinguish metaphor from metonymy. However, a closer look at their argument reveals that when they employ directionality as a differentiation criterion, they actually mean the reversibility of the mapping. They (2011: 220) argue that “metaphorical mappings are thus normally unidirectional, and the source and target are not reversible”. They (2011: 220–221) also notice that metonymy is a reversible process and that metonymic mappings can project in either direction, like CAUSE FOR EFFECT and EFFECT FOR CAUSE. But they never claim that the source and the target in a metonymy can map onto each other simultaneously, i.e. bidirectional mapping. To avoid confusion, I distinguish the notions of directionality and reversibility as suggested by Barcelona (2011).

12 | Demarcation and variability of metonymy



a. WHOLE THING FOR A PART OF THE THING America for “United States”

b. PART OF A THING FOR THE WHOLE THING: England for “Great Britain”

Although Barcelona’s (2011) modified definition of metonymy still hinges upon delineating the conceptual structure – functional domain – within which metonymy operates as the basis of the demarcation of metonymy from metaphor, it provides another possible way to define metonymy, which shifts the attention from the “single-domain” underlying the metonymic mapping to the specific nature of the mapping (Peirsman and Geeraerts 2006b: 273). This shift has been promoted to a large extent by Peirsman and Geeraerts (2006b) in their prototype-based model of metonymy.

1.1.1.2 The prototype-based view Attention to the nature of the conceptual relationship underlying metonymy can be traced back to structuralist semantics, which claims that metonymy is based on semantic relationships of co-occurrence and association, that is, contiguity or proximity, while metaphor is based on semantic relationships of similarity (Ullmann 1962: 212; Jakobson 2002 [1971]; Geeraerts 2010d: 63–64). The notion of contiguity in cognitive semantics goes beyond linguistic association and purely spatial proximity to include conceptual contiguity (Feyaerts 1999: 317; Dirven 2002: 91) as in CAUSE–EFFECT relations (Koch 2011: 266). Cognitive linguists have for some time resisted contiguity as a criterion to demarcate metonymy from similarity-based metaphor (Steen 2005: 4), but in recent years many of them have finally come round to the idea of metonymy as a cognitive relation of contiguity6 (see Koch 1999a, 2011; Peirsman and Geeraerts 2006b; Panther and Thornburg 2007).

|| 6 To constrain the scope of metonymy, Panther and Thornburg (2002: 282, 2007: 240–241) argue that the contiguous relationship between metonymic source and target displays the property of contingency/defeasibility, i.e. contiguity does not exist by conceptual necessity, and the source concept is still retrievable in the given context. For instance, if a waitress says, The ham sandwich has asked for the bill, the contiguous relation between CUSTOMER and HAM SANDWICH is a contingent link; it is not conceptually necessary that the ham sandwich belongs to that customer throughout or that he/she always orders ham sandwiches. The metonymic relation is based on contingent conceptual contiguity and thus in principle defeasible or cancelable.

Conceptual metonymy | 13

Noting that the notion of contiguity itself has prototypicality, Peirsman and Geeraerts (2006b) propose a prototype-based view of metonymy instead of a unitary definition. They describe metonymy in terms of contiguity and also look at contiguity from a prototypical point of view. Postulating spatial part-whole contiguity as the prototypical core, Peirsman and Geeraerts show how the set of traditionally recognized metonymic patterns derived mainly from the historicalphilological literature on semantic change can be connected to the prototypical core in terms of three underlying dimensions. First, along the dimension of strength of contact, contiguity decreases from the core of a spatial part-whole relation to containment, and then to contact and adjacency without contact as weaker forms of contiguity. The second dimension involves a shift of domain from the spatial and material domain to the spatio-temporal domain of actions, events and processes, and then to the domain of functional assemblies and collections. The third dimension, boundedness, extends the prototypical core, which is based on contiguity between concrete and bounded parts and wholes, in the direction of unbounded contiguous entities. Peirsman and Geeraerts’s (2006b) prototype-based approach to metonymy is “an interesting attempt to reconcile at the same time internal diversity and unity of metonymy” (Koch 2011: 270). Their solution makes two main contributions to the cognitive notion of metonymy. First, deviating from the “singledomain” approach, Peirsman and Geeraerts’s solution relies on the notion of conceptual contiguity. On one hand, their proposal avoids the difficulty of the vague notion of domain (matrix); on the other hand, by adopting a prototypical analysis of conceptual contiguity, they find a feasible strategy to delineate the notion of contiguity, which is in itself probably as vague as “domain”7. Second, they have suggested an efficacious classificatory scheme for a great number of metonymic patterns and investigated how these patterns may be related to each other (Peirsman and Geeraerts 2006a: 329). Their prototype-based model of metonymy thus provides a detailed and highly structured reference for the analysis of the internal typology of metonymy. Peirsman and Geeraerts’s (2006b) prototype view of metonymy has also received criticism. According to Croft (2006: 319), a weakness of Peirsman and Geeraerts’s account is that they fail to provide evidence that the cases they regard as prototypical metonymies are really more acceptable or frequent than others. Another critical comment is from Barcelona (2011: 29), who points out that Peirsman and Geeraerts do not discuss the related issue of the link between || 7 Proponents of the “single-domain” approach often mention the lack of clarity of the concept of contiguity as one of the main reasons against its use as a criterion (Geeraerts 2010d: 217).

14 | Demarcation and variability of metonymy

the source and the target, and thus do not clearly indicate whether or not they see metonymy as a mapping. The validity of these criticisms, however, is debatable. First, although prototypicality is normally linked to high usage frequency, they do not inevitably coincide (Handl 2011a: 91). For instance, Gilquin (2006: 175) found that the causation structures which are assumed to be prototypical “account for an astonishingly small proportion of the corpus data”. Second, Peirsman and Geeraerts (2006a: 332) point out that their prototype-based view of metonymy is not incompatible with but supplementary to a domain highlighting view (Croft 2002 [1993], 2006) in the sense that the prototypical model of contiguity primarily suggests a classification of metonymies and shows how the original source and the highlighted target can be related to one another. Koch (2011) combines the “single-domain” and the contiguity criteria in defining metonymy. He underlines the importance of contiguity as the basis of metonymy and at the same time describes metonymy in terms of figure-ground effects within conceptual frames, following Geeraerts’s (2006b: 16) restrictive interpretation of the notion of frame: “a semantic frame…is a coherent structure of related concepts where the relations have to do with the way the concepts cooccur in real world situations”. Koch’s definition highlights three important aspects of the notion of metonymy: 1) figure-ground effects (similar to the highlighting or perspectivization in the “single-domain” approach); 2) being within a conceptual frame; and 3) contiguity between two elements of a given frame or between the frame as a whole and one of its elements. If we incorporate Peirsman and Geeraerts’s prototype-based view into Koch’s interpretation, we could consider the third aspect as prototype-based contiguity. Koch’s definition, however, is not unassailable. The “figure-ground effect” property of metonymy would exclude the semantic phenomenon of facets from metonymy but include the active zone as a subtype of metonymy, as facets involve no figure-ground profile (Koch 2011: 273–274) but the active zone can be treated figure-ground effect (Truszczyńska 2003: 226). Several scholars in CL disagree with Koch on the relationship between metonymy and the phenomena of zone activation and facetization; see Ruiz de Mendoza Ibáñez (2011) and Geeraerts and Peirsman (2011). I will explore this issue in detail in the later discussion of the demarcation between metonymy and other related phenomena.

1.1.1.3 Working definition of metonymy So far, I have reviewed two basic views on metonymy in CL. Although these two views diverge from each other on particular matters, both recognize that meton-

Conceptual metonymy | 15

ymies are conceptual and not purely linguistic. The most important point for the purposes of this book is to establish a working definition of metonymy. What we need is a more precise, operational definition for linguistic metonymy, which is a metonymic expression that instantiates or realizes a conceptual metonymy. I adopt the definition proposed by Geeraerts (1997: 96) as my working definition: “Metonymy is a semantic link between two readings of a lexical item that is based on a relationship of contiguity between the referents of the expression in each of those readings”. This definition is different from those used to describe conceptual metonymies, which focus on the cognitive and conceptual nature of metonymy. In contrast, this working definition is mainly for identifying metonymies in the linguistic data collected in empirical studies. The reason to use this definition is twofold. First, it allows us to link metonymic concepts to reality by specifying that the contiguity is between the referents. Second, it is easier to manipulate in practice by examining the different readings of a lexical item in empirical linguistic research. This definition thus is useful for a bottom-up approach to identifying metonymic expressions and then formulating conceptual metonymies. Four points will contribute to the theoretical clarification of this working definition. First, the contiguous relations between the referents should be understood as a prototype-based category as suggested by Peirsman and Geeraerts (2006b). Second, metonymy allows us to think of one referent in terms of its relation to something else (Geeraerts 2010d: 214). Hence, a source and a target can be identified for a metonymy as they can for a conceptual metaphor. The referent the speaker intends to express in a certain context is a target, while the contiguous referent which triggers the target is a source8. Each source-target pair is labeled a metonymic pattern. Third, “metonymies are systematic in the sense that they form patterns that apply to more than just an individual lexical item” (Geeraerts 2010d: 214). Finally, this definition by no means denies the conceptual nature of metonymy or its experiential roots.

1.1.1.4 Demarcation between metonymy and other related phenomena After discussing the definitions of metonymy in CL, we already have a rough idea of where metonymy stops. Next, I would like to explicitly discuss the demarcation between metonymy and several other cognitive mechanisms, like metaphor, the active zone, facets and synecdoche. || 8 Different terms, like vehicle-target (Radden and Kövecses 1999) or donor-target (Goossens 1990), are also used as alternatives for source-target.

16 | Demarcation and variability of metonymy

(1) Metonymy versus metaphor The differentiation between metonymy and metaphor has been mentioned in the preceding discussion of the two views of metonymy. I have addressed the problems derived from the domain-based view for demarcating metonymy and metaphor. The vague notion of domain (matrix), inevitably causes the demarcation of metonymy from metaphor to run into difficulty. It is easy to find counterexamples of metaphors whose source domain is apparently included in the target domain9 (e.g. in the ANGER IS HEAT metaphor, BODY HEAT as a typical physiological effect of anger is part of the ANGER domain) or whose source and target domains both belong to the same superordinate domain (e.g. He is a fox). Within the domain approach, further differences between metonymy and metaphor have been proposed with respect to the symmetry of mappings as well as the reversibility of mappings (see “The domainbased view” above). The contiguity versus similarity criterion seems to be less problematic than the domain approach. With this criterion, CL reapproaches the traditional structuralist treatment of metonymy and metaphor. Metonymy is considered a contiguous relationship between the source and the target, unlike metaphor, which is based on similarity. We should of course bear in mind the prototypicality of contiguity, which provides the basis of metonymy. However, all the above-mentioned criteria have to be treated with caution in practice. No single criterion can definitely distinguish all metaphorical cases from metonymic ones. After all, even conceptual contiguity is not a necessary and sufficient definition of metonymy (Peirsman and Geeraerts 2006a: 329). At the same time, a linguistic expression may sometimes be identified based on context, background knowledge or the interpreter’s choice, as metaphorical, metonymic, or as a combination of both (Barcelona 2011: 43). Actually, many scholars have recognized that metaphor and metonymy may be located on a continuum with no clear-cut boundary between them (e.g. Goossens 1990; Geeraerts 2002; Deignan 2005a; Benczes 2006; Barnden 2010). Besides the demarcation of metonymy and metaphor, their interaction also attracts much interest in cognitive semantics. Goossens (1990) proposes the term “metaphtonymy” for the interaction of metaphor and metonymy in conventionalized expressions. He recognizes two types of interactions: metaphor from metonymy, and metonymy within metaphor/metaphor within metonymy. The former refers to the phenomenon of a || 9 The counterexamples of this type normally lie in EMOTION metaphors, which arise through the decontextualization or generalization of certain metonymies (Barcelona 2011: 37).

Conceptual metonymy | 17

metaphor with a metonymic origin, which describes a sequential operation of the two mechanisms. The latter involves a simultaneous, parallel type of interaction (Geeraerts 2010d: 220). The successive type of metaphtonymy is exemplified by the verb giggle. In “Oh dear,” she giggled, “I’d quite forgotten”, the verb giggle has a metaphorical interpretation of “to say as if giggling”, which is based on a metonymic process from the original meaning of giggle “to laugh continuously in a manner not uproarious” to the metonymic interpretation of “to say/express while giggling”. The simultaneous type of metaphtonymy can be understood from an example like catch someone’s ear. This idiomatic expression is metaphorical in the sense that the hunting scene evoked by catch has to be reinterpreted metaphorically; at the same time, within the metaphor, the object ear is used metonymically for the person’s attention. The successive type, which plays an important role in the semantic evolution of lexical items (Geeraerts 2010d: 220–221), may also be of the same kind of the chained metonymy or serial metonymy discussed in CL (see Nerlich and Clarke 2001; Hilpert 2007). Scholars have tried different approaches to uncover the successive and simultaneous steps of metaphor and metonymy in semantic extensions, such as Geeraerts’s (2002) prismatic model, which is chosen to identify metonymy in expressions in the present book and will be discussed at length in Chapter 2.

(2) Metonymy, active zone and facets Another widely discussed issue we need to clarify when identifying the scope of metonymy is the relationships between metonymy and its neighboring phenomena of zone activation and facetization. Langacker (2000b: 67) argues that “profile/active-zone discrepancy is a special case of metonymy”. This treatment of metonymy as a reference point and activation phenomenon, however, as has frequently been pointed out, makes the scope of metonymy too unconstrained for a more detailed semantic analysis (Paradis 2004: 246). Paradis (2004) holds that metonymy involves conceptual directionality of sense mappings; facetization is a matter of intraconceptual highlighting only, and resides within concepts or sense; zone activation involves a conventionalized pairing of a lexical item and a certain image schematic profiling of a sense or a facet of a sense. Therefore, “it [metonymy] stops at the level of senses. Facetization stops at the level of qualia structure and zone activation concerns all readings within qualia structure” (Paradis 2004: 262). Paradis’s three-way classification of metonymy, facetization and zone activation has been challenged by Ruiz de Mendoza Ibáñez (2011), who regards metonymy

18 | Demarcation and variability of metonymy

as a cover term for two basic cognitive operations, domain expansion and domain reduction, with domain reduction then based on two kinds of highlighting process, facetization and zone activation. Geeraerts and Peirsman (2011) investigate the relationship of these three mechanisms and point out that facetization is a subtype of metonymy, while zone activation appears to be a totally different phenomenon. They discuss three distinct phenomena, which are exemplified in (2). In the zone activation example (2a), there is no shift of reference in the noun phrases. Actually, the shift from the dog to its activated zone the dog’s teeth violates the selectional restrictions of bite, which requires a volitional subject. However, in example (2b), which is an uncontroversial metonymy, the kettle violates the selectional restrictions of boil, which requires a fluid as its subject. A shift of reference from the container (i.e. kettle) to its content (e.g. water) is necessary. According to Geeraerts and Peirsman, there is no reason to separate facetization, as in (2c), from metonymy. Facetization has the same pragmatic function of referential identification with metonymy; its behavior with respect to zeugma is the same as that of uncontroversial target-in-source metonymy (Ruiz de Mendoza Ibáñez, 2000); both metonymy and facetization involve similar semantic as well as pragmatic processes. (2)

a. Your dog bit my cat. (zone activation) b. The kettle is boiling. (uncontroversial metonymy) c. This book is boring. (facetization)

Geeraerts and Peirsman’s proposal contributes to a clearer delimitation of the notion of conceptual metonymy in CL and also resolves the terminological problems of metonymy and related phenomena. In the present book, I fully subscribe to their argument to treat facetization as a subtype of metonymy and zone activation as nonmetonymic.

(3) Metonymy versus synecdoche The scope of metonymy also involves a controversial point in the relation between metonymy and synecdoche. In historical-philological tradition, metonymy and synecdoche are sometimes kept apart (Hock 1991: 285; Campbell 2006: 257–260). In traditional rhetoric, synecdoche refers to part-whole relations and includes two subtypes: PART–WHOLE synecdoche (e.g. sail for ship) and SPECIES– GENUS synecdoche (e.g. bread for “food”) (Lausberg 1998: 572–573; Nerlich and Clarke 1999: 198–199). Many scholars, however, agree that PART–WHOLE synec-

Conceptual metonymy | 19

doche is metonymy (e.g. Lakoff and Johnson 1980; Koch 1999a; Nerlich and Clarke 1999; Radden and Kövecses 1999; Seto 1999; Ruiz de Mendoza Ibáñez 2000; Croft 2002 [1993]; Jakobson 2002 [1971]; Taylor 2003; Panther and Thornburg 2007). In Peirsman and Geeraerts’s (2006b) prototype-based model of metonymy, the spatial PART–WHOLE relation is even located at the core of the contiguity category. Once we accept PART–WHOLE synecdoche as a subtype of metonymy, which is less controversial, the question arises of where we should place SPECIES–GENUS synecdoche. The relation between SPECIFIC and GENERIC10 has been discussed widely in CL under different terminologies such as PARAGON/MEMBER–CATEGORY relation (Radden and Kövecses 1999; Barcelona 2004) and INDIVIDUAL–COLLECTION relation (Peirsman and Geeraerts 2006b), among others. One point of divergence in the various delimitations is whether or not they consider SPECIES–GENUS synecdoche as a subtype of metonymy. On one hand, scholars note the difference between a “part-of” partonomy relation and a “kind-of” taxonomy relation (Cruse 1991; Seto 1999: 93), therefore they tend to exclude the SPECIES–GENUS synecdoche from metonymy. Seto (1999) points out that unlike metonymy, which reflects contiguous relations between one entity and another in the world (i.e. entity-related transfer), synecdoche is a conceptual relation between a more comprehensive and a less comprehensive category (i.e. category-related transfer). In fact, although using the term “synecdoche”, Seto restricts his argument to SPECIES–GENUS synecdoche and leaves partonomy (i.e. PART–WHOLE synecdoche) to metonymy. On the range of metonymy, Koch (2011: 274–277) advocates the “split strategy” of separating the engynomic process of metonymy from the taxonomic processes of specialization and generalization (i.e. SPECIES–GENUS synecdoche). Koskela (2011) distinguishes between metonymy and vertical polysemy (i.e. a polysemous word with a broader and a narrower sense that occupy different levels in a taxonomic hierarchy). She argues that the GENERIC and SPECIFIC relation in vertical polysemy is a taxonomic vertical relationship in encyclopedic semantics and it presupposes a metaphorical conceptualization of categories (cf. Seto 1999): either as containers for the category members (cf. Lakoff 1987) or in terms of wholes, with the members constituting the parts that make up the category (cf. Radden and Kövecses 1999).

|| 10 This relationship also receives a metaphor treatment as GENERIC IS SPECIFIC; see Sullivan and Sweetser (2009).

20 | Demarcation and variability of metonymy

On the other hand, many scholars work with a vague borderline between SPECIES–GENUS synecdoche and metonymy 11 (Radden and Kövecses 1999; Barcelona 2003c, 2004; Peirsman and Geeraerts 2006b). With a metaphorical interpretation of taxonomic hierarchies as part-whole structures, Radden and Kövecses (1999) treat the SPECIES–GENUS (MEMBER–CATEGORY) relation as a type of metonymy operating in the Category-and-Member ICM. Barcelona (2003c, 2004) regards the phenomenon of using paragon names (well-known individual members of a category) for the category as motivated by a metonymic process. In the prototype-based model of metonymy, Peirsman and Geeraerts (2006b) conceive the INDIVIDUAL–COLLECTION relation as metonymy. The contiguity in this taxonomical relation, however, is assigned to the domain of functional assemblies and collections, which is a marginal or peripheral position along the dimension of domain within the prototype-based model. In the present book, my understanding of synecdoche stays very close to the solution proposed by Peirsman and Geeraerts (2006b). Briefly, synecdoche should be included under the umbrella of metonymy. PART-WHOLE synecdoche in the spatial domain presents the prototypical core of contiguities underlying metonymies, while SPECIES–GENUS synecdoche is a peripheral (untypical) case in the prototype-based model of metonymy.

1.1.2 Internal typology of metonymy In the previous section, I briefly discussed how metonymy is defined and where metonymy stops in the framework of CL. What is at issue in this section is the internal variety and typology of metonymy, which is also crucial to metonymy research in practice.

|| 11 Panther (2007: 241) discusses the demarcation problem of how meaning specialization and generalization relate to metonymy. Based on the contingency criterion (metonymic relation is contingent conceptual contiguity), he argues that it is problematic to regard generalization as a metonymic process (e.g. aspirin for “any pain-relieving tablet”), because aspirin as a hyponym of pain-relieving tablet and therefore aspirin entails the meaning of “a pain-relieving tablet”. Specialization, however, could qualify as metonymy (e.g. flower for “a rose”), because the relation between hyperonym (e.g. “flower”) and hyponym (e.g. “rose”) is contingent, and the superordinate concept does not semantically imply any of its hyponyms.

Conceptual metonymy | 21

Different typologies Inventories of metonymic patterns can be found already in traditional rhetoric and, further on, from the historical-philological tradition of the 19th century up until CL (Koch 2011: 268). In order to go beyond a mere listing of metonymic patterns, CL scholars have made efforts to categorize various metonymic patterns under a few more general headings according to particular criteria. One of the most popular taxonomies of metonymy in CL starts with a basic dichotomy proposed by Radden and Kövecses (1999): 1) a whole ICM and its part(s) and 2) parts of an ICM (Koch [2011: 267, 269] uses a somewhat different terminology for a similar taxonomy: FRAME–ELEMENT versus ELEMENT–ELEMENT). Within each general type, refined classifications are adopted according to the different ICMs metonymies operate in. Whole-part configurations are assumed to underlie the Thing-and-Part ICM, the Constitution ICM, the Event ICM, the Category-and-Member ICM, etc. Each ICM then may lead to different metonymic types. For instance, the Thing-and-Part ICM includes WHOLE FOR PART and PART FOR WHOLE as two metonymic variants. The Part-Part configuration includes the Action ICM, the Containment ICM, the Production ICM, the Control ICM, etc. Metonymic variants of CONTAINER FOR CONTENTS and CONTENTS FOR CONTAINER, for instance, operate in the Containment ICM. Because Radden and Kövecses’s classification relies heavily on the notion of ICM, again, practical constraints have been noted, as in the domain-based view of metonymy (see Section 1.1.1.1): the notion of ICM is not easy to delineate. Emphasizing the difference between synchrony and diachrony, Blank (1999) proposes a typology of metonymy based on co-presence and succession. He demonstrates that all types of metonymies can be subsumed under two general types of contiguity: contiguous relations between entities that are copresent within a frame (e.g. INVENTOR–PRODUCT, OBJECT–FUNCTION, PLACE–OBJECT) and those that are successive within one frame or two related frames (e.g. RESULT–CAUSE, ACTIVITY–AIM, INSTRUMENT–PRODUCT). Blank’s typology comprises three levels of abstraction: the two superordinate domains of co-present and successive contiguity at the highest level, an open list of “type of contiguity” or “contiguity schemas” (in other words, conceptual metonymies) at the intermediate level, and concrete linguistic metonymies at the lowest level. Briefly, Blank’s classification links the typology of metonymy with two very fundamental ways of conceptualizing real-life situations: as static frames or as dynamic scenarios. At the same time, the three levels of abstraction draw awareness to the schematicity of metonymy in its internal typology. However, one shortcoming in Blank’s approach should be mentioned: he offers little information on how he groups the “contiguity schemas” at the intermediate level.

22 | Demarcation and variability of metonymy

Seto (1999) distinguishes three general types of metonymy in terms of spatial versus temporal versus abstract domains. Then, in each domain, Seto further classifies different subtypes of metonymy. Although Seto’s classification excludes the category-relation (i.e. the SPECIES–GENUS synecdoche), his classification shows that the same relationship seems to appear in different domains of contiguity (Peirsman and Geeraerts 2006b: 274–275): the temporal domain of metonymy might mirror the spatial domain in the sense that there are WHOLE– PART, CONTAINER–CONTENTS and ADJACENCY subtypes of metonymy in the spatial domain on one hand, and subtypes of WHOLE EVENT–SUBEVENT and PRECEDING– ENSUING in the temporal domain on the other hand. The correspondence between the spatial and the temporal domains accords with the prototype-based view of contiguity as suggested by Peirsman and Geeraerts (2006b). The above-mentioned typologies of metonymy all follow a unitary classification criterion. Radden and Kövecses (1999) build their classification on the basis of the notion of ICM; Blank (1999) combines a dichotomous classification of types of contiguity with metonymic types; Seto (1999) groups different metonymies in terms of different domains along a continuum of spatial to abstract. In striving to go beyond unidimensional typologies, Peirsman and Geeraerts (2006b) put forward their prototype-based view of metonymy. They propose a prototypical definition of contiguity along three dimensions (i.e. strength of contact, domain and boundedness) and accordingly classify the various metonymic patterns they collected from an extensive survey of the literature on metonymy against a cross-classification of the dimensions. As an example, Figure 2 shows the classification of metonymic patterns against a cross-classification of the strength of contact and domain dimensions. In this way, a fairly welldifferentiated structure of various types of contiguities and, hence, of conceptual types of metonymy, is attained within the prototype model. Peirsman and Geeraerts’s approach also reveals the granularity/schematicity of metonymy. At the most general level, we could group various metonymies according to their positions along different dimensions. For instance, if we only take one dimension into account at a time, both a three-way classification of part-whole, containment and proximity metonymies and a four-way classification of space, time, event and assemblies/collections metonymies are possible. If we classify metonymies against two dimensions, we could have spatial part-whole metonymies, spatial containment metonymies, spatial proximity metonymies, etc. (see Figure 2). At a more detailed level, we can construct an inventory of metonymic patterns such as CONTAINER FOR CONTAINED, LOCATION FOR LOCATED, etc.

Conceptual metonymy | 23

  

   

    

      

   ­   ‚ ­    





­ €  

­  ­ ­  ƒ

­    

   ­     

­   ­  

­­   „„ 

­   „

­   „



­   

„­



 ƒ  „… ­   ƒ

 ­„   ­  † 

­   ‚‚­

­   

­  

Fig. 2: A prototype-based classification of metonymic patterns (adapted from Peirsman and Geeraerts 2006b; Geeraerts 2010d: 218)

In brief, Peirsman and Geeraerts’s typology makes two important contributions. First, based on the prototype account of the notion of contiguity, it manages to explain a wide variety of metonymic patterns and situate them in three different dimensions, which define their extensions from the prototypical center. Second, their prototype-based classificatory schema not only provides a typological architecture of metonymy category but also shows how different patterns connect to each other through the extensions starting from the prototypical core area. The metonymic pattern classification in the present book is mainly based on the metonymy inventory in Peirsman and Geeraerts’s prototype-based classification.

Schematicity of metonymy Concerning the typology of metonymy, one important issue needs to be mentioned, which is schematicity. The typologies of metonymy discussed here all involve different levels of granularity (in Blank [1999]’s term, “levels of abstraction”). We need to build a hierarchy of metonymic patterns that is virtually related to the hierarchical structure of categories (Allan 2006) in both sources and targets, and that in turn reflects the notion of schematicity in Cognitive Grammar. In Cognitive Grammar, the term schema refers to a superordinate concept/structure, and elaborations or instantiations or subcases refer to specific concepts/structures of the schema; see Figure 3. Schematicity pertains to level of specificity in the sense that “schematicity can be equated with the relation

24 | Demarcation and variability of metonymy

between a superordinate node and a subordinate node in a taxonomic hierarchy” (Langacker 1987: 68). For instance, the schema [FRUIT] and its instantiations [APPLE] and [BANANA] represent the same entity at contrasting levels of specificity. Each solid arrow in Figure 3 presents a relation of schematicity between a schema and its instantiations. ^ĐŚĞŵĂ ΀&Zh/d΁

/ŶƐƚĂŶƚŝĂƚŝŽŶϭ ΀WW>΁

/ŶƐƚĂŶƚŝĂƚŝŽŶϮ ΀EE΁

Fig. 3: Schema and instantiations (after Langacker 1987: 74)

The notion of schematicity has already been successfully applied in metaphor research. Clausner and Croft (1997: 278) demonstrate that metaphorical mappings are formulated at different levels of schematicity of both source and target domains. For instance, the conceptual metaphor ARGUMENT IS WAR, which involves a metaphorical mapping between the source domain WAR and the target domain ARGUMENT, may instantiate a large number of metaphorical expressions. In other words, the metaphor schema of ARGUMENT IS WAR generalizes a great variety of individual metaphorical instantiations such as attack the point, be right on target, wipe you out and many more (see Lakoff and Johnson 1980; Ritchie 2003). Given the idea that all concepts are schematic to some degree (Tuggy 2007: 84), when we discuss the source and target concepts of metonymy, we should take their schematicity into account. In other words, like hierarchical metaphorical mappings (Lakoff 1993: 222), metonymic mappings are also organized in “hierarchical structures” (Feyaerts 1999). For example, we know that LOCATION is a common metonymic source for PERSON (i.e. LOCATION FOR LOCATED). The category of LOCATION allows a schematic hierarchy (see Figure 4): LOCATION is the superordinate concept that is elaborated by a number of instantiations like PALACE or ROOM, which in turn function as intermediate schemata for more specific instantiations. At the bottom level of the hierarchy are the instantiations of specific linguistic expressions, which refer to the concept in the most precise, finegrained detail.

Conceptual metonymy | 25





 

   



   

       

  

Fig. 4: An example of the schematic hierarchy of the source concept LOCATION

The same goes for the target concept PERSON; see Figure 5. At the top, general level, the category PERSON is instantiated by some intermediate categories like WOMAN and OFFICIAL, which are in turn instantiated by more specific categories like BEAUTIFUL WOMAN, QUEEN and GIRL for the former, and COUNTY MAGISTRATE and OFFICIAL OF COURT for the latter. Each specific category can be further elaborated by a number of linguistic expressions.  



 

   





 

  

   

Fig. 5: An example of the schematic hierarchy of the target concept PERSON

Hence, at the very highest level of a metonymic pattern, we may have a LOCATION FOR PERSON metonymic pattern, based on the schematic hierarchies represented in Figure 4 and Figure 5. However, from a variational point of view, the highestlevel metonymy is too general to reveal something about the conceptualization of the target, let alone the variation in the conceptualization. LOCATION FOR LOCATED is a conventional metonymic pattern, and given PERSON as the located entity, it is quite natural to infer a metonymic mapping from LOCATION to PERSON. However, as we explore the schematic hierarchy, the lower-level metonymies may reflect more variation in the conceptualization of a target or a subcategory of the target. The hierarchical nature of source and target concepts makes the examination of variation in metonymy more interesting. It also raises questions. For instance, when a general source-target mapping (i.e. a metonymic pattern, like PIECE OF CLOTHING FOR PERSON) is identified in two languages, can this metonymic mapping function comparably for the more specific categories of PERSON, like WOMAN, OFFICIAL and SCHOLAR, in the two languages? Can the more specific

26 | Demarcation and variability of metonymy

subconcepts of PIECE OF CLOTHING, like ACCESSORY, HAIRSTYLE and CLOTHES, which can be metonymically mapped onto the target concept, share similar distributions in the two languages?

1.1.3 The conventionalization of metonymy Another frequently discussed notion in CL studies of metonymy is conventionalization (e.g. Radden and Kövecses 1999; Taylor 2003; Barcelona 2011; Handl 2011a). The conventionality of metonymy would certainly merit a study of its own. I will devote the remainder of this section to a brief discussion of factors that contribute to the conventionality of metonymies, beginning with Barcelona’s (2003b, 2011) summary of two of the main factors. First, metonymy conforms to one or more of the default high-level patterns, like PART FOR WHOLE, WHOLE FOR PART, PRODUCER FOR PRODUCT, etc.; see Section 1.1.2 for inventories of metonymic patterns (e.g. Radden and Kövecses 1999; Peirsman and Geeraerts 2006b). Second, a conventional metonymy must be socially sanctioned by the community of speakers (Ungerer and Schmid 2006: 130). The social sanctioning of metonymy depends, in turn, on two groups of principles: (i) The notion of salience12 is helpful for the description of the conventionality of metonymy. The “default-routes” of metonymic mappings are not arbitrary but constrained by laws of salience, which cover general human experience, perceptual selectivity and cultural preference. Referring to the laws of salience, Radden and Kövecses (1999: 44–52) propose a number of general cognitive and communicative principles affecting people’s preference for a certain metonymy in contrast to others. These principles help us understand why we select certain sources to access a target and why certain source-to-target routes have become conventionalized in a given language (Radden and Kövecses 1999: 52). For instance, in reality, humans take precedence over nonhumans and concrete objects are more salient than abstract entities; therefore, the principles of HUMAN OVER NONHUMAN and CONCRETE OVER ABSTRACT apply in this case.

|| 12 Salience has two distinctive usages: cognitive salience and ontological salience (Schmid 2007: 120). Cognitive salience concerns the current activation of concepts in actual speech events, while ontological salience describes the attention-attracting ability of entities in the real world.

Metonymy: A variationist CL view | 27

(ii) The presence or absence of specific principles such as rhetorical effects and social norms sometimes overrides the general cognitive and communicative principles favoring the conventionalization of metonymy. Importantly, the conventionality of metonymy can only be adequately described when both conceptual and linguistic levels as well as the schematicity of metonymy are taken into account. For instance, example (3a) manifests the low-level metonymy PICASSO FOR HIS WORK and the high-level metonymy PRODUCER FOR PRODUCT. This metonymy is motivated by the general cognitive principles HUMAN OVER NONHUMAN, CONCRETE OVER ABSTRACT and GOOD GESTALT OVER POOR GESTALT. It is highly conventionalized because it also follows a cultural principle in which works of art are regarded as unique products of well-known artists, Picasso in this case. The situation is different in example (3b): although it has the same high-level metonymy PRODUCER FOR PRODUCT motivated by the same cognitive principles as in (3a), the more specific level of JANE FOR HER WORK is not socially accepted or conventionalized, as the speech community does not agree that “a Jane” is an important and valuable work of art in art history (see also Taylor 2003). (3)

a. I have just bought a Picasso. b. I have just bought a Jane. (Barcelona 2011: 35)

1.2 Metonymy: A variationist CL view In this section, I will review the studies on variation in metonymy in terms of three dimensions recognized by cognitive linguists, although for the most part in metaphor research, and seldom studied systematically with respect to metonymy.

1.2.1 The cross-linguistic dimension Universal or regular cognitive patterns can be approached from a crosslinguistic perspective, by looking at the relationships between metonymic sources and targets in different languages, even typologically unrelated languages. The cultural nature of cognition and language has already been noticed by many scholars in cross-linguistic metaphor-oriented research (to name just a few, Dirven 1994; Deignan et al. 1997; Yu 1998, 2008a, 2008b, 2009; CharterisBlack and Ennis 2001; Boers 2003; Kövecses 2003, 2005; Maalej 2004;

28 | Demarcation and variability of metonymy

Dobrovol'skij and Piiraninen 2005; Aksan and Kantar 2007; Sharifian et al. 2008; Simó 2008, 2011; Ding et al. 2010). The great amount of metaphor research has provided a systematic way of disclosing how certain bodily and/or cultural factors exert an influence on metaphorical conceptualization. However, relatively fewer scholars have explored the cross-linguistic variation in conceptual metonymy in a systematic way. One of the reasons for the lack of investigation of cross-linguistic variation in metonymy in CL might be that the experiential grounding of conceptual metonymy is in general more obvious than it is in conceptual metaphor, as metonymy usually involves direct physical or causal associations (Lakoff and Johnson 1980: 39). Nevertheless, the role cultural models play in metonymy from both conceptual and linguistic perspectives is as important as it is in metaphor, in the sense that “a cultural model, as a representation of a community’s wisdom in a given domain, can be seen as both the synthesis and simultaneously as the source of many clusters of metaphors and metonymies and also of many individual metaphors and metonymies” (Niemeier 2008: 350). Recently, cognitive linguists have brought to light significant crosslinguistic differences in the availability of certain types of metonymies (e.g. Panther and Thornburg 1999; Brdar-Szabó and Brdar 2002; Barcelona 2003c; Brdar and Brdar-Szabó 2003; Radden and Seto 2003; Deignan and Potter 2004; Ibarretxe-Antuñano 2005; Sneesby 2009; Littlemore 2015: 161–190). Panther and Thornburg (1999) test whether and to what extent the POTENTIALITY FOR ACTUALITY metonymy is operative cross-linguistically by a comparative analysis of English and Hungarian. They check the linguistic relevance and distribution of this metonymy in seven conceptual and communicative domains (e.g. perceptual events, mental states and processes, indirect speech acts, acquired skills). The most important cross-linguistic difference exists in the conceptual domains of sense perceptions and mental states/processes. In these two domains, English exploits POTENTIALITY FOR ACTUALITY systematically, whereas in Hungarian it is either not exploitable or only weakly exploited; the link between POTENTIALITY and ACTUALITY is very strong in English, whereas in Hungarian, it is nonexistent or weak. Later, in Ibarretxe-Antuñano’s (2005) examination of the metonymic basis of conceptual metaphors using English, Spanish and Basque data, the metonymic basis POTENTIALITY FOR ACTUALITY for the metaphor AFFECTING IS TOUCHING can be found in the English but not in the Spanish and Basque data. Ibarretxe-Antuñano concludes that cross-linguistic conceptual metaphors may not always share the same metonymic basis in different languages. A series of cross-linguistic studies of media discourse by Brdar and BrdarSzabó ( Brdar-Szabó and Brdar 2003, 2011; Brdar 2006, 2007b; Brdar and Brdar-

Metonymy: A variationist CL view | 29

Szabó 2009) show that there are noticeable cross-linguistic differences in the use of capital name metonymy, which is a so-called “universal” conceptual metonymy. The corpus-based comparisons in English, German, Croatian and Hungarian reveal that English and German are liberal in the use of capital name metonymies, which is much less used in Croatian and Hungarian. In addition, Croatian and Hungarian discourses exhibit uneven distributions of the metonymy CAPITAL FOR GOVERNMENT at both the type and token levels. A close inspection of the data further shows that the marked differences observed are the result of constraints from conceptual, grammatical and discourse-pragmatic perspectives. The availability of capital name metonymy is dependent on some structural factors of the language systems. For example, Croatian, as a typical pro-drop language, has a rich agreement system and impersonal constructions, which could serve in a replacement pattern for capital name metonymy, and therefore might lead to the underuse of the metonymy type in question. Barcelona (2003c) explores the role of metonymy in the reclassification of proper names as common nouns in English, French, German, Italian and Spanish. For example, the proper locative name Washington may make reference to people somehow related to it; the person name Shakespeare may evoke the notion of “talented writers”. Barcelona discusses two general instances: names used as paragons and “partitive restrictive modification” of names (Quirk et al. 1985: 290). The use of names as paragons is motivated by a chain of two metonymies: CHARACTERISTIC PROPERTY FOR INDIVIDUAL and MEMBER/SUBCATEGORY FOR CATEGORY; the partitive restrictive modification on names is triggered by a WHOLE FOR PART metonymy. In both cases, the rules of grammar for names can be violated. Although these two metonymic phenomena seem to be shared by the five languages, cross-linguistic differences arise in their grammatical behaviors. The cross-linguistic differences appear to be determined by the peculiarities of their respective grammatical systems: English and German impose fewer constraints on the grammatical reclassification of paragon names as common names than French, Spanish and Italian (Barcelona 2003c: 26). Even cross-linguistic differences in body-related metonymies, which seem to clearly be heavily embodied, are attracting increasing attention. CharterisBlack (2001) explores conceptual metonymies of the body part hand (i.e. HAND FOR CONTROL, HAND FOR OUTCOME AND ACTION, HAND FOR TYPICAL BEHAVIOR and HAND FOR PERSON) in English and Malay. Although both languages show evidence for these metonymies, which provides some support for the view that the metonymic conceptualization is bodily grounded and there exists a prototypical universal conceptualization of the human hand, comparisons of their uses in English and Malay show a significantly different cultural resonance in each language. The

30 | Demarcation and variability of metonymy

“hand” may not necessarily convey the same connotation in the two languages due to their different cultural specifics. For instance, Malay speakers often place a negative evaluation on the HAND FOR CONTROL metonymy, whereas HAND FOR CONTROL is positively evaluated in English. In the same paradigm, CharterisBlack (2003) studies the metaphorical and metonymic uses of three oral body parts (MOUTH, LIP and TONGUE) in English and Malay phraseology. The crosslinguistic study shows that English has a tendency toward metonymy and hyperbole whereas Malay has a tendency toward metaphor and euphemism. These preferences might be a result of cultural differences in attitudes toward facial expressions and in stylistic preferences between English and Malay. Deignan and Potter (2004) compare metaphorical and metonymic expressions related to four body parts, nose, mouth, eyeheartin English and Italian using the Bank of Englishand two Italian corpora. The corpus data demonstrate that while universal bodily experience or similar cultural knowledge may motivate many metaphors and metonymies in both English and Italian, the process will not necessarily result in equivalent linguistic realizations in the two languages, for cultural and linguistic reasons. Sneesby (2009) studies metaphorical and metonymic expressions with hand and their uses in English, German and Spanish. The study provides a picture of how the human mind is metaphorically and metonymically organized in the three different linguistic communities. At the same time, it shows the different ways in which metaphors and metonymies reflect the folk or cultural models existing in the three languages. Littlemore (2015: 161–190) broadens the discussion on cross-linguistic and cross-cultural variation in metonymy and outlines how metonymy facilitates understanding and communication between people from different cultural/linguistic backgrounds. With ample examples, Littlemore further analyzes the problems that metonymy presents to language learners and to translators, given the heavily culture-bound nature of metonymy. Her work makes a crucial contribution to investigations of the comprehension and production of metonymy in cross-linguistic and cross-cultural communication. As discussed in this section, the experiential nature of the Cognitive Linguistic conception of metonymy involves both a physiological and a cultural kind of experience. The central role of culture, which has been discussed by many scholars (e.g. Holland and Quinn 1987), is essential and indispensable in people’s metonymic thinking and language, because it is impossible to fully explain our cognition and language without considering the cultural context in which they are situated. One might then wonder whether some concepts have always been metonymically construed in the same way at least with-

Metonymy: A variationist CL view | 31

in a single language or cultural sphere. This question brings us to the historical dimension of metonymy research.

1.2.2 The historical dimension Culture is, by definition, inherently historical and not stable across time. The experientialist nature of language and cognition entails both cultural and historical “situatedness” of human experience; that is to say, “both cultural and historical factors are likely to influence our cognitive patterns” (Geeraerts and Grondelaers 1995: 227). In consequence, if we want to examine metonymy from a cultural perspective, we should consider the diachronic dimension, i.e. exploring the real distribution of metonymies amid historically changing environments. Changes in culture/society through time could affect the use of metaphors and metonymies; hence, diachronic studies are useful to show cultural effects on people’s cognition and language. The important role of cultural-historical change in people’s conceptualizations has been attested by a number of metaphor studies (e.g. Sweetser 1991; Kay 2000; Tissari 2003, 2010; Musolff 2004; Kövecses 2005; Allan 2006; Koivisto-Alanko and Tissari 2006; Mischler 2008, 2009; Trim 2010, 2011; Benczes 2011; Cánovas 2015). Most of them have explored a conceptual metaphor or a subject with an expectedly high proportion of metaphors like emotions across time. The historical approach to metaphor has provided substantial evidence for cultural interpretation as a factor in observed metaphor variation. For instance, instead of explaining the uses of metaphors for emotions like ANGER and SADNESS on a purely physiological , scholars have drawn attention to the culturally/historically specific background of metaphorical expressions for the emotions, such as the theory of four humors, as in work by Gevaert (2002, 2005, 2007), Geeraerts and Grondelaers (1995), Geeraerts, Gevaert and Speelman (2011) and Ding (2011). Their findings question the universality of metaphorical conceptualizations across time. For instance, the metaphorical conceptualization of ANGER as HEAT is not a constant feature of the concept of ANGER in the history of English. A possible cultural change through time may account for the historical fluctuation of the use of the ANGER IS HEAT metaphor. Diachronic variation in metonymy has not been studied as much as diachronic variation in metaphor, but is drawing increasing recognition. In fact, many proponents of historical approaches in metaphor research have mentioned in passing the culturally/historically specific background of certain me-

32 | Demarcation and variability of metonymy

tonymies. For example, in studies of the semantic development of synonyms and near-synonyms for human categories, like WOMAN/FEMALE HUMAN BEING and MAN/MALE HUMAN BEING, Kleparski (1997) and Grygiel (2005) spare a few words for metonymy, although they emphasize metaphor and other causes like borrowing. Nevertheless, they notice the enormous role of metonymy in the rise of lexical items related to human categories at certain historical points, such as female/male proper names used in the sense of “woman” or “man”, especially in colloquial and vulgar contexts. Tissari (2008) studies the occurrences of the noun and verb respect in Early Modern English with respect to their conceptual metaphors and metonymies. The study examines the role of the concept VISION in the conceptualization of RESPECT and adds a historical perspective to the contemporary interpretation of RESPECT’s conceptualization, as in Kövecses (1990: 109–127). Focusing on the role of metonymy in triggering semantic change in historical synonyms of MAN and WOMAN, Kleparski (2000, 2004) and Grygiel (2007) provide a good number of historical metonymic transfers from the conceptual domain CLOTHES, BODYPARTS, PROPER NAMES, OCCUPATION/PROFESSION and related social roles like WARRIOR/SOLIDER to the conceptual domain MAN or WOMAN. Rusinek (2008a) examines lexical items that originally denote CLOTHES like skirt, pinafore, shawl or bloomer and then become historically synonymous with woman. Rusinek terms the historical rise of CLOTHING as WOMAN as a result of metaphorical processes. The semantic transfer, however, actually involves a conventionalized metonymy, PIECE OF CLOTHING FOR PERSON, rather than a similarity-based metaphor. Although these studies have to some extent considered the historical dimension of metonymy, that is, the role of metonymy in the diachronic semantic change of certain lexical items, they have not explicated whether a particular metonymic relation between a source and a target, such as CLOTHES FOR WOMAN, behaves in a similar way or not across time. The diachronic variation in metonymies of a particular type has been explored by Allan in a series of studies (2008, 2010). Allan (2008) shows that diachronic metonymy may play an equally important role as diachronic metaphor in accounting for contemporary synchronic polysemy. She presents a diachronic approach to uncover the way in which English lexical items for the concept groups of INTELLIGENCE, SENSES, DENSITY and ANIMALS have occurred over time via metaphorical and metonymic mappings. Based on data from the Historical Thesaurus of English, her study addresses the importance of including historical data (in this case, from Old English to Present Day English) in metaphor and metonymy research. Allan (2008: 185) demonstrates that historical data can offer insight into metaphoric/metonymic mappings that could not be disclosed

Metonymy: A variationist CL view | 33

by a purely synchronic approach. I have to mention, however, that Allan’s main interest is the motivation for mappings of different kinds of metaphor and metonymy, which she traces through etymological developments in the Historical Thesaurus of English database, rather than the real distributions of metonymic and metonymic expressions either diachronically or synchronically, which of course leaves an important research gap in quantitative metonymy research. Using data from the Oxford English Dictionary, Allan (2010) traces the historical sense development of groups of lexemes related to particular conventionalized metonymies, such as MATERIAL FOR OBJECT (e.g. glass for “vessel”), with a special focus on the results of metonymically motivated change and historical evidence for the way in which the change has arisen over time. The diachronic data show that the timescale on which MATERIAL lexemes become conventional and the life-spans of a particular metonymic OBJECT sense vary from one MATERIAL lexeme to another. She also points out that the historical development of metonymic polysemy13 might be specific to particular word histories and sensitive to both intralinguistic systemic factors and extralinguistic factors such as their cultural and historical contexts. Allan’s studies provide valuable insights into historical metonymy research, which is often neglected in the field of CL. She succeeds in showing that “a diachronic perspective can be helpful and valuable in formulating and testing theories of metonymy” (Allan 2010: 163). Little work investigates the diachronic development of metonymy, especially in contrast to the large amount of diachronic research on metaphor. There is a great need for research on the extent to which metonymies of a particular type follow a similar diachronic evolution pathway, as well as the extent to which the historical and cultural part of human experience might influence the historical development of metonymies.

1.2.3 The lectal dimension Variation in metaphor or metonymy has been studied in CL primarily from the cross-linguistic and historical perspectives. Another important but less investigated dimension of culture in metaphor/metonymy research is its social environment. Sociolinguistic diversity in metaphor/metonymy has not been well explored, but still, we may note an increasing interest in the role of lectal factors within CL research. There are two features of CL that inescapably lead towards || 13 Metonymic polysemy develops through the “lexicalization” of ad hoc metonymic usage in discourse (Koch 1999a: 140).

34 | Demarcation and variability of metonymy

the incorporation of sociolinguistic aspects in studies of language variation: the experiential nature of language and the usage-based nature of language (Geeraerts 2010c; Geeraerts et al. 2010b). In the first place, CL advocates that languages embody the historical and cultural experience of groups of speakers (and individuals) (Geeraerts 2006a: 5). Interaction in social groups and among individuals implies that there must be a social part of embodiment. Therefore, embodiment is bodily, historical, culturally and socially situated. In other words, meaning is not recognized by CL as involving isolated cognition, but as existing in social and cultural contexts. The “social” side of cognition and language has been noted by many cognitive linguists; see the edited collection of Frank et al. (2008). CL’s insistence on the social nature of meaning inevitably calls for a socially and variationally oriented form of research on language. In the second place, there is a growing consensus within CL that it is a usage-based approach to language (see Barlow and Kemmer 2000; Langacker 2000a; Tummers et al. 2005), which entails that it considers sociolinguistic variation. Empirical sources for a usage-based approach to linguistics primarily consist of corpus materials, which may be internally heterogeneous (Geeraerts 2005: 168); for example, texts collected from different dialects, national varieties, registers and so forth. One might want to know whether the variation displayed in such data is due to lectal factors or not. The social aspect of language variation has been generally acknowledged in CL and has resulted in the advent and development of a new realm of study known as Cognitive Sociolinguistics (e.g. Geeraerts 2005; Kristiansen and Dirven 2008a; Geeraerts et al. 2010a), but even here the relevance of lectal variation in metaphor/metonymy research is not yet well represented. Nevertheless, metaphor/metonymy, as socially-conditioned cognition, can reflect the lectal stratifications of people’s language usage, which suggests that investigations on metaphor/metonymy would be desirable in the field of Cognitive Sociolinguistics. Recently, a growing tendency in CL to stress the social nature of metaphor/metonymy can be observed, exemplified by studies by Takada et al. (2000), Berthele (2008, 2010), Ziegeler and Lee (2009), Shie (2011)andZhang, Speelman and Geeraerts (2011), as well as by the inception of the journal Metaphor and the Social World in 2011 (Cameron and Low 2011), which provides an important outlet for studies on metaphor/metonymy and their social interactions. Takada et al. (2000) examine the social and cultural background for Japanese WOMAN metaphors involving ANIMALS and PLANTS. They find that Japanese

Metonymy: Semasiology and onomasiology | 35

metaphors to describe women have undergone some changes in their interpretations and usages, in pace with changes in social norms. The empirical data – collected in a questionnaire survey of 112 native Japanese speakers of different ages, genders and professions – show that diversity in the interpretation and use of metaphors can be attributed to changes in Japanese society. For instance, older respondents were more likely to have a traditional or conservative interpretation of WOMAN metaphors. Focusing on the construction of complex folk cultural models underlying attitudinal metaphors, Berthele (2008, 2010) provides representative variationist investigations of the social-cognitive structure of metaphors in specific dialect and standard language environments. Ziegeler and Lee (2009) report a cross-dialectal study on a CAUSED ACTION FOR CAUSED ACTION+RESULT grammatical metonymy in causative constructions between Singaporean and Malaysian English. The authors investigate the relationship between the causativeresultative construction and a conventionalized scenario. The empirical data demonstrate that language contact features (e.g. the contact presence of a topicprominent discourse structure) in the dialects may influence the extent of conventionalized scenario usage. Shie (2011) studies the metaphors and metonymies in two editions of the same newspaper published for two different social communities, readers in the United States and learners of English as a foreign language in Taiwan. The textual survey reveals that the different readerships affect the use of metaphors and metonymies. The lectal productivity of a particular type of metonymy has been explored by Zhang, Speelman and Geeraerts (2011). Using data from a corpus built for the study, the authors explore the (non)metonymic usage of capital names in Mainland Chinese and Taiwan Chinese news articles. The quantitative analysis shows a notable difference between the two language varieties, especially in the usage of the CAPITAL FOR GOVERNMENT metonymy. There still remains much further work to be undertaken on the differences of metaphor/metonymy usage resulting from lectal factors. Given the socialcultural reorientation of embodiment and the recognition of the usage-based nature of meaning, CL must address the social dimension of metonymy.

1.3 Metonymy: Semasiology and onomasiology In the study of the relationship between words and their semantic values, we can distinguish two main perspectives, semasiological and onomasiological. Geeraerts et al. (1994: 5) explain the former thus: “The semasiological perspective takes its startingpoint in the word as a form, and describes what semantic

36 | Demarcation and variability of metonymy

values the word may receive”. It addresses the question of what the senses of words are and traditionally deals with semantic variability in the form of polysemy, vagueness, homonymy and so forth. Of the latter, the authors write: “The onomasiological perspective takes its startingpoint on the level of semantic values and describes how a particular semantic value may be variously expressed by means of different words” (Geeraerts et al. 1994: 5–6). It seeks to answer the question of what words are linked with a particular semantic value and traditionally focuses on the study of semantically related expressions such as near-synonyms and hyponyms. Put simply, semasiology moves from linguistic form to semantic variation, while onomasiology moves from meaning to word choice. The distinction between the semasiological and onomasiological perspectives can be extrapolated to metonymy research. The semasiological perspective on metonymy emphasizes the important role of metonymy in semantic change: words may have different meanings and the development of a certain meaning may be triggered by a metonymic process. The onomasiological perspective on metonymy involves the selection of the preferred metonymic source for a given meaning/concept. In this section, I will review research into metonymy from these two perspectives.

1.3.1 The semasiological perspective The semasiological perspective on metonymy can be traced back to an early stage in the history of lexical semantics, that is, the historical-philological semantics of the 19th century. Early linguistic research was mainly concerned with metonymy as a fundamental mechanism underlying semantic change and metonymic polysemy (Ullmann 1967: 80, 89, 231ff; Nunberg 1978: 144–145, 1995). Agreeing on the cognitive nature of metonymy, researchers in CL also highlight the crucial role of metonymy behind the semantic structure of language. In CL, metonymy has been widely recognized to be instrumental in sense development and meaning structuring both synchronically and diachronically. In general, metonymy research in the paradigm of CL is primarily being pursued from the semasiological perspective, involving topics such as the role of metonymy in meaning shifts from a theoretical point of view (e.g. Traugott and Dasher 2001; Taylor 2003: 122–130; Paradis 2011), serial metonymic extensions in lexicon and grammar (e.g. Nerlich and Clarke 2001; Hilpert 2007) as in belly (“a bodypart” → “pregnancy” → “offspring”), metonymic meaning construal in word formation (e.g. Panther and Thornburg 2002; Colman 2004;

Metonymy: Semasiology and onomasiology | 37

Benczes 2006; Kosecki 2007; Barcelona 2008; Basilio 2009; Janda 2011; Yoon 2011; Brdar and Brdar-Szabó 2014) and contextual meanings of metonymy (e.g. Koch 1999a; Milić and Vidaković 2007; Brdar and Brdar-Szabó 2009; Zhang et al. 2011), as well as the pragmatic meanings of metonymy (e.g. Panther and Thornburg 2005; Ruiz de Mendoza Ibáñez and Uson 2007) like implicature derivation or the illocutionary meaning of a particular metonymy. By a semasiological approach, we may learn how a specific word has acquired its lexical/grammatical metonymic sense and how a metonymy, as a mechanism of lexical innovation, triggers the compositional meaning of a compound or an idiom (Blank 2001: 7). When a semasiological perspective is brought to variation-oriented research on metonymy, the subjects investigated normally include the cross-linguistic availability or differences of a particular metonymy in motivating the distribution and function of its semantic or grammatical meanings (e.g. Panther and Thornburg 1999; Brdar and Brdar-Szabó 2009), the diachronic separation of stages where metonymic sense develops and changes (e.g. Allan 2008, 2010) and the lectal (constructional) semantic variation of metonymic patterns (e.g. Ziegeler and Lee 2009; Zhang et al. 2011).

1.3.2 The onomasiological perspective With regard to the onomasiological perspective on metonymy, we first must distinguish conceptual and formal interpretations. The former concerns the choice of a conceptual category for naming a referent; the latter, the choice of a lexical item designating a referent. For example, given a concept or referent of TROUSERS, speakers may make a conceptual choice for a specific name between trousers and jeans, or a formal choice between lexical items of trousers and pants to name the referent (Geeraerts 2010b: 823). Following Geeraerts’s (2010b) terminology, conceptual onomasiological variation is proposed for the case of trousers and jeans and formal onomasiological variation is used for the case of trousers and pants. Onomasiological studies help to discover the different conceptual or lexical “pathways” through which a concept or a group of concepts has developed by going back to the source concepts (Blank 2001: 7). Many previous studies on the cognitive/cultural model of emotions actually provide an onomasiological approach to lexical semantics, with an emphasis on how an emotion (e.g. ANGER, SADNESS, HAPPINESS) is expressed metaphorically in different cultural and historical contexts (e.g. Lakoff 1987; Geeraerts and Grondelaers 1995; Gevaert 2005, 2007; Mischler 2008; Ding 2011). Metaphor research from the onomasiological

38 | Demarcation and variability of metonymy

perspective, of course, goes beyond emotion concepts, as in Takada et al.’s (2000) study of how WOMAN is metaphorically conceptualized in Japanese in terms of ANIMALS and PLANTS. In a discussion including but not limited to metaphors, Kövecses (2006) considers concepts of WOMAN and MAN in American English slang to demonstrate how ideology connects with the socially and culturally important subject matter of WOMAN and MAN. He finds that the expressions used to designate the concepts frequently conceptualize them metaphorically as FOOD or ANIMALS or metonymically by BODY PARTS or PROPER NAMES. Intimately linked to the onomasiological perspective in the studies of metonymy is conceptual onomasiological variation; that is, given a referent, what conceptual categories might be chosen as metonymic sources for that referent by a metonymic process? For example, when we decide to designate a person metonymically, there is a categorical choice to be made for the metonymic source among his/her clothes, body parts, location and so on. This conceptual variation in naming a referent by metonymy is closely related to the selection of metonymic sources for a single concept or sets of related concepts. A series of studies by Kleparski (1996, 1997, 2000, 2004, 2005, 2007), Grygiel (2005, 2006, 2007), Rusinek (2008a, 2008b) and Kleparski and Borkowska (2007) have explored the disappearance or change of meaning of a number of historical synonyms for MAN/WOMAN categories from an onomasiological perspective. One noticeable feature of these studies is that they all focus on the impact of the mechanism of metaphor/metonymy on the development of senses related to the conceptual macro-category HUMAN BEING. Their studies also show that the onomasiological link between the macro-category MAN/WOMAN and the associated categories like PERSONAL NAMES, CLOTHES and BODY PARTS highlights different values specific to the attributes of the macro-category. Koch (1999b) carries out a cognitive-onomasiological investigation of the TREE-FRUIT frame from a cross-linguistic perspective. In many languages, the synchronic result of a change of meaning via contiguity is metonymic polysemy. For instance, in Russian, the word originally referring to “pear” could be used to designate “pear tree”. His study indicates that cognitive constants may affect cross-linguistic designational patterns and that designations of TREES and FRUITS are not created in a totally arbitrary way. Then, Koch (2008) conducts a diachronic cognitive onomasiological study on the possible lexical items (source concepts) designating the semantic domain EYE (EYELASH, EYEBROW, EYELID and EYEBALL). He discusses polygenetic semantic parallels in semantic change and provides a list of all the cognitive solutions to create lexical innovations in the language sample, such as taxonomic subordination (HAIR for EYELASH and EYEBROW) and metaphorical similarity (BALL for EYEBALL). Contiguity-based metony-

Metonymy: Semasiology and onomasiology | 39

my proves to be a typical relation holding between one of the target concepts and particular source concepts, such as EYELID for EYELASH, TO PALPITATE for EYELID, EYE for EYEBROW and PUPIL for EYEBALL. Koch’s onomasiological study contributes to “the investigation into typology and universals in the lexicon and to our understanding of cognitive constants in particular conceptual domains” (Koch 2008: 130). The importance of an onomasiological perspective for metonymy research has been addressed by Sweep (2012: 625). She argues that the onomasiological point of view has to be taken into account in order to reveal the reasons why a metonymic expression has been used. It is impossible for us to find out the pragmatic effects of metonymies, such as referring in the most economic and relevant way, causing euphemistic or humorous effects and so on, or to identify the reasons why speakers choose the metonymic expressions, unless we examine a metonymic expression in contrast with its nonmetonymic counterpart from an onomasiological perspective. ͞dŚĞůŽǁĞƌƉĂƌƚŽĨĂ ǁŽŵĂŶΖƐĚƌĞƐƐŽƌŐŽǁŶ͕ ĐŽǀĞƌŝŶŐƚŚĞƉĞƌƐŽŶĨƌŽŵ ƚŚĞǁĂŝƐƚĚŽǁŶǁĂƌĚƐ͟

 ŽŶŽŵĂƐŝŽůŽŐŝĐĂů ƐĞůĞĐƚŝŽŶŽĨŵĞƚŽŶLJŵŝĐ ƐŽƵƌĐĞƐǁŝƚŚƌĞƐƉĞĐƚƚŽ ͞ǁŽŵĂŶ͟ 

ƐĞŵĂƐŝŽůŽŐŝĐĂůŵĞĂŶŝŶŐĐŚĂŶŐĞ ǁŝƚŚƌĞƐƉĞĐƚƚŽ

͞ǁŽŵĂŶ͟

ƐĞŵĂƐŝŽůŽŐŝĐĂůŵĞĂŶŝŶŐĐŚĂŶŐĞ ǁŝƚŚƌĞƐƉĞĐƚƚŽ

͞ǁŽŵĂŶΖƐƵŶĚĞƌŐĂƌŵĞŶƚ͟

Fig. 6: Semasiological meaning change and onomasiological naming change

In a nutshell, it is very important to combine the semasiological perspective with an onomasiological one if we want to know whether a metonymy takes place or not in a certain context. The relation of the two perspectives is represented in Figure 6. Semasiologically, the lexical item skirt literally refers to a piece of clothing worn by women. Through the metonymic process of PIECE OF CLOTHING FOR PERSON, it may refer to the concept of WOMAN. Similarly, the lexical

40 | Demarcation and variability of metonymy

item smock literally expresses “a woman’s undergarment” and may metonymically designate WOMAN as well. The semasiological description of these two polysemous words focuses on meaning change with respect to each lexical item. Onomasiologically, the concept WOMAN can be expressed by skirt or smock, and this is of course a result of a PIECE OF CLOTHING FOR PERSON metonymic process. The onomasiological perspective can provide a description of the change of naming/designation with respect to the concept WOMAN. In this case, it deals with the selection of metonymic sources for the target WOMAN. If conceptual metonymy research is reoriented toward the onomasiological perspective, two types of questions naturally arise in such research: what is people’s lexical/categorical selection for a given target concept, and what factors govern people’s choice of a preferred (metonymic) source or the alternative? Answers to such questions are fundamentally linked to a contextualized, pragmatic interpretation of onomasiology (Grondelaers and Geeraerts 2003: 69– 70), which focuses on an investigation of use, that is, the actual choices made for a particular name as a designation of a particular concept. If scholars are interested in the choices of metonymies for a particular referent, they invariably, and unsurprisingly, need to situate the issue in multiple dimensions to gain a sense of possible cross-linguistic, historical and lectal variations.

1.4 The methodological state of the art in metonymy research The methodological situation in CL – including metonymy research, of course – is a mixed one, with a growing interest in empirical methods on the one hand, but a lingering dominance of traditionally analytic approaches on the other (Geeraerts 2006c: 44). The traditional introspection-based semantic methodology (e.g. Lakoff and Johnson 1980; Wierzbicka 1988; Talmy 2000a, 2000b), however, proves to be insufficient because it is based on two debatable assumptions: homogeneity in a linguistic community and direct access to meaning in the individual. It has been demonstrated that the worth of intersubjective comparison of interpretations based on our individual intuitions is doubtful, as is the possibility of direct access to semantic phenomena (Geeraerts 2010a). Gibbs (2007b: 3) also points out that although introspection can be valuable for constructing hypotheses, we have to be aware of the biased nature of any individual’s linguistic judgments. The insufficiency of introspective-based semantic methods has led to an empirical revolution in CL. As a matter of fact, a number of theoretical features of CL encourage an empirical approach, which has had a profound influence on contemporary metonymy research.

The methodological state of the art in metonymy research | 41

1.4.1 Why CL research needs an empirical approach In Geeraerts’s (2006c: 28–31) view, at least three features of CL naturally promote an empirical approach: its cognitive nature, its usage-based perspective and its contextualized conception of language. First, CL describes and defines itself as a cognitive science and regards language as a multifaceted phenomenon of cognitive, social and psychological dimensions. Hence, the methods adopted effectively in the cognitive sciences can be expected to prove valuable in CL. If we wish to gain a comprehensive understanding of possible language-mind-body interactions, we have to combine methods that address different aspects of cognitive/linguistic behavior and knowledge, such as the experimental techniques of psychology, computer modeling and behavioral and neurophysiologic research. A growing number of studies in CL have successfully combined interdisciplinary empirical methods, such as those of Gibbs (1994, 2006), Carlson and Hill (2007) and Coulson (2007). Second, CL stresses its essential nature as a usage-based linguistics; see Barlow and Kemmer (2000) for a very thorough discussion on the central notions of usage-based linguistics. Because it argues that knowledge of language and language structure arises out of language use (Croft and Cruse 2004: 1; Tomasello 2005: 5), CL needs empirical methods to study actual language usage. Authentic data of real usage can be collected by means of surveys, experiments and corpora (Tummers et al. 2005: 290). Their high level of objectivity makes corpus data especially useful as evidence of language production in Cognitive Linguistic research (e.g. Gries et al. 2005, 2010; see Glynn 2010: 23–24 for a brief summary of corpus-driven Cognitive Semantics studies). Quantification and statistical analysis is then needed for a systematic analysis of the data gathered (Tummers et al. 2005: 234). Third, the need for empirical methodology within CL is supported by its contextualized conception of language. The recontexualizing tendency in CL attaches importance to encyclopedic meaning, which is highly contextualized rather than an isolated phenomenon, as it is treated in formal semantics, for instance. The relevant context, however, does not just consist of other types of cognition, but also involves society and culture (Geeraerts 2006c: 29–30). In this sense, CL necessarily has to incorporate language variation research, from, for example, cross-linguistic, diachronic and lectal perspectives. Sociolinguistic diversity has been much less explored compared with the other two perspectives in CL; however, “a truly usage-based CL cannot ignore the qualitative and quantitative variation to be found within the standard and non-standard varieties of a language” (Kristiansen and Dirven 2008b: 3). The importance of social-

42 | Demarcation and variability of metonymy

ly and culturally determined variation for CL has a natural methodological consequence: the quantitative empirical methods that are dominant in sociolinguistic research at large should be adopted in CL. To summarize: the three core features of CL imply that it requires empirical underpinnings for its theories of language and thought.

1.4.2 Overview of methods in CL metonymy research If CL is indeed experiencing an empirical revolution, then metonymy research in CL should be practiced in an empirical way with actual usage data. A variety of methods have been employed in metonymy research in CL. In the following, I will introduce how the various approaches work in practice in conceptual metonymy research and point out the particular strengths of empirical methods.

1.4.2.1 Introspection Section 1.4.1 presented reasons for going beyond introspection in CL and discussed how several features of CL are leading to an empirical revolution. This, of course, does not imply a wholesale rejection of introspective methods. In fact, introspective and empirical research in CL could feed into each other (BrdarSzabó and Brdar 2012); the former dominates at the stage of theory formation while the latter has a decisive role in the stage of theory testing14. Since the first proposal of Conceptual Metonymy Theory, the introspective approach has retained a lingering dominance. Conscious introspection is often conducted on particular points of interest in metonymy such as its cognitive nature and definition (e.g. Lakoff and Johnson 1980; Radden and Kövecses 1999; Barcelona 2011), its demarcation from related phenomena (e.g. Seto 1999; Geeraerts and Peirsman 2011; Ruiz de Mendoza Ibáñez 2011) and its operation in grammar (e.g. Dirven 1999; Langacker 2009). Studies that practice introspection largely contribute to revealing the systematicity and universality of conceptual metonymies. Yet an introspective methodology has an obvious theoretical correlate: without empirical evidence from actual usage data, it is impossible to corroborate the diversity of metonymic usage along the various relevant dimensions. Another methodological consequence of introspection-driven research is that it often relies on qualitative analysis of scattered, decontextualized data for || 14 Geeraerts (2006c, 2010a) points out that empirical research involves a cycle in which the analysis of empirical data can also contribute to theory formulation or modification.

The methodological state of the art in metonymy research | 43

verification or falsification of a theory. In brief, being based on an individual’s intuition, introspection does not make it easy to provide an accurate picture of metonymy phenomena or explore the variation in metonymy uses from different perspectives; see Croft (1998) and Gibbs (2006) for further discussion on the limitations of introspection.

1.4.2.2 Empirical methods The introduction of empirical methods into CL has had a significant impact on metonymy research. The main types of empirical methods that are available for metonymy research include the dictionary-based approach, experiments and corpus methods.

(1) Dictionary-based approach Admittedly, dictionary-based research has several limitations (see Markert and Nissim [2006: 154] for a detailed discussion). For instance, dictionaries normally include only conventional metonymic senses; neither the inclusion of metonymic senses nor the supplementation of semantic contexts is systematic; dictionary material leaves less room for a multivariate quantitative investigation of metonymy in real language usage15. However, dictionary material, especially data from historical dictionaries, can provide valuable evidence for metonymy research with various research objectives. Several studies (e.g. Kleparski 1997; Kay 2000; Grygiel 2005; Allan 2006, 2008, 2010, 2011) have used the Oxford English Dictionary (henceforth OED) as evidence for research on metonymic semantic change. Providing dated attestations and semantic contexts (e.g. stylistic information such as dialectal or specialized usage) for most senses within the entry for each lexeme, the OED offers an opportunity to examine historical metonymic polysemy. At the same time, the advent of computers has made a major difference in dictionary-based research. Escaping the labor-intensive and time-consuming data collection needed for paper versions of dictionaries, scholars can easily find useful data with the help of the search engines provided by the online dictionaries. Another invaluable dictionary resource for work in metonymy on English is the Historical Thesaurus of the OED (henceforth HTOED), which reorganizes the OED’s

|| 15 Fischer (1997) uses the OED on CD-ROM as a corpus for historical semantic research. Hoffmann (2004) proposes using the OED quotations database as a corpus for linguistic research, especially for historical linguists.

44 | Demarcation and variability of metonymy

content into semantic fields. The HTOED provides new possibilities for the investigation of metonymy, because it offers a taxonomic classification of conceptual fields and, for each field, a list of synonyms with their dates of use under brief explanatory headings (Kay 2011). From a semasiological perspective, a search for source vocabulary can be based on checking variants under a HTOED headword (e.g. Allan 2008); from an onomasiological perspective, the taxonomic classification in the HTOED offers lists of lexemes (e.g. near-synonyms) denoting the same meaning, from which researchers can identify lexemes metonymically referring to the same target concept.

(2) Experiment Elicitation of online data from experiments is another way to collect data for metonymy research. A great advantage of conducting experiments to gather data is that it makes it possible to investigate not only the product but also the process of metonymy (e.g. Frisson and Pickering 1999; Gibbs 2007a; Zarcone et al. 2012). Experimental work (e.g. eye-tracking, priming experiments, comprehension and production tasks), while it is the dominant paradigm in psycholinguistic and neurolinguistic research, is less developed in metonymy research in CL compared with the other two types of empirical methods. One of the weaknesses of the experimental approach is that it cannot be used to investigate linguistic phenomena from a historical point of view, as we cannot find participants who are users of historical languages for data elicitation.

(3) Corpus methods Using corpora is the most promising empirical approach in metonymy research (e.g. Markert and Nissim 2003; Deignan and Potter 2004; Deignan 2005a; Stefanowitsch and Gries 2006). Corpus methods have been proposed as a reliable way to investigate both conceptual and linguistic metaphors (Steen 1999; Charteris-Black 2004; Deignan 2005b, 2008). Providing real data of natural language from a wide variety of language users, corpora offer an unprecedented opportunity to expand research on metonymy. More important, corpus techniques have led to new and more objective research methodologies that can specifically address the issue of variation in metonymy; for example, by investigating the effect of cultural models on metonymy over time or the importance of situational context in comprehending metonymy uses by social communities. Three main points concerning corpus methods for metonymy research in general will be discussed below.

The methodological state of the art in metonymy research | 45

First, the general goal of corpus compilation is to “represent as wide a range of variation as possible” (Biber et al. 1994: 4). Corpora should be designed to be representative of language use across various time periods, genres and social strata, for instance. A scientifically compiled corpus can provide reliable data for the investigation of metonymy as it was actually used in a situational context. Second, after selecting a representative corpus for a particular research objective, we need an objective way to extract and identify metonymies in the corpus data, because corpora normally do not annotate word senses as dictionaries do. Stefanowitsch (2006) summarizes several strategies for extracting metaphorical or metonymic expressions from corpora, including searching for source or target domain vocabulary (e.g. Hilpert 2006; Markert and Nissim 2006; Zhang et al. 2011), searching for sentences containing lexical items from both the source and target domains (e.g. Martin 2006) and searching from a corpus annotated for conceptual mappings (e.g. Steen et al. 2010). For the identification of metaphor, scholars have developed a scientific procedure known as the Metaphor Identification Procedure (MIP) (Pragglejaz Group 2007; Steen et al. 2010). In addition, Markert and Nissim (2002, 2006) have proposed practical procedures for metonymy identification and annotation. Third, because the empirical data offered by corpora are heterogeneous, quantitative methods are needed to investigate the variation in metonymy. In practice, a range of quantitative techniques (see Glynn and Fischer 2010) are available for exploring variation in language usage. The simplest way is to quantify the frequencies of metonymy by counting; see Hilpert (2006) and Janda (2011), for example. However, in order to find the factors that motivate the variation in language usage, we need more advanced quantitative techniques. Multivariate statistical methods can achieve this goal and have been employed successfully by many scholars. Arppe (2008) presents an overall methodological framework for studying lexical variation proceeding from univariate via bivariate to multivariate techniques; Hilpert (2011) proposes an innovative way of visualizing diachronic changes in semantics with the help of a multidimensional scaling technique. Multivariate techniques are also helpful in detecting lectal variation; see Levshina (2011), Ruette (2012), Zhang, Speelman and Geeraerts (2011) and Zenner (2012, 2013), among others. The variation in metonymy cannot feasibly be uncovered without authentic data from large corpora and multivariate statistical techniques. Current corpus methods provide both practices and techniques that can give a more reliable and objective picture of differences in metonymy uses. Going beyond the word or sentence level, corpus data enable us to investigate

46 | Demarcation and variability of metonymy

different dimensions of variation in metonymy in order to further reveal the discursive, ideological, social and cultural functions of metonymy.

1.5 Summary: Research gaps Based on this review of previous research on metonymy, several research gaps can be summarized as follows. First, generally, metonymy research has not gained as wide attention as metaphor research and is still a minor field in the context of CL. It is not uncommon for metonymy to be referred to only cursorily in a metaphor-oriented study, or simply ignored. Second, variation in metonymy is little investigated by CL scholars, who have for the most part been more interested in the embodied conceptual nature of metonymy and its universal availability. Third, while there has been much research on the significant role of metonymy in semantic extension from a semasiological point of view, relatively little has been written about the variation in metonymy from an onomasiological perspective. Fourth, it remains a matter of serious concern that there are few systematic and practical statistical tests and visualization techniques designed to investigate metonymy uses in authentic language. Accordingly, the present book reports on three case studies focusing on specific dimensions of variation in metonymy, which were designed to fill these research gaps. In Part Two, the three case studies will be discussed in turn in Chapters 3, 4 and 5.

2 Metonymy in expressions For single-word expressions, it is straightforward to identify the metonymic patterns that might have been at work behind their semantic changes and to recognize the metonymic processes by which new meanings are generated from existing words, if we check the semantic development of the word from a historical perspective. For example, in English, skirt originally refers to “the lower part of a woman's dress or gown, covering the person from the waist downwards”16; later, it develops a meaning of “a woman”. It is quite clear that this semantic change has proceeded based on a conventional metonymy, PIECE OF CLOTHING FOR PERSON. For metonymic composite expressions (such as compounds and idioms), the recognition of metonymy in the process of meaning construction is akin to the cases of single words, but in a more multilayered form in the sense that their meanings are both compositional and figurative. The semantic architecture of a figurative composite expression is compositional to the extent that it is composed of the constituent parts of the expression and is figurative to the extent that conceptual mechanisms like metaphor or metonymy act upon the meaning extension process. To identify the metonymic transfer in composite expressions, several models of meaning construction are proposed in CL. In Section 2.1, I will first discuss two basic notions related to the meaning structuring in composite expressions, compositionality and analyzability, with reference to Langacker’s (1987) Cognitive Grammar. I will then give a detailed account of three models of meaning construction and how they have been applied to the analysis of figurative composite expressions: the blending model of Fauconnier and Turner (1998, 2002), the model for English noun-noun creative compounds of Benczes (2006) and the prismatic model of Geeraerts (2002). Next, this chapter will discuss various paths of metonymy in Chinese expressions with numerous examples in Section 2.2. Finally, several differentiation criteria for type classification, which are crucial for the metonymy quantification in the case studies, will be proposed in Section 2.3.

|| 16 Definitions of English examples in this book are from the Oxford English Dictionary Online (http://www.oed.com/), unless otherwise indicated.

48 | Metonymy in expressions

2.1 Models for analyzing figurative composite expressions The metaphorical or metonymic motivation of the rich semantic meanings in composite expressions has been receiving increased attention in CL (e.g. Gibbs 1994; Geeraerts 2002; Benczes 2006; Langlotz 2006; Philip 2011). The previous research has yielded a number of models for interpreting the role of metaphor or metonymy in composite expressions. In what follows, I will first provide a cognitive-grammar account of compositionality and analyzability, and then review three models for figurative composite expressions in CL.

2.1.1 Compositionality and analyzability In the framework of Cognitive Grammar, Langacker (1987: 448) defines composition as “the relation between component structures and the composite structure that derives from them”. The notion is further clarified in Figure 7, in which he specifies three structures and three relationships: [A] and [B] stand for component structures, whereas [C] is the composite structure; the integration relationship between [A] and [B] is along the horizontal correspondence and the compositional relationships are established along the vertical correspondences of [A]–[C] and [B]–[C]. Langacker further claims a fourth structure [AB], which represents the expected compositional value of the integrative relationship [A]– [B].

 









Note: int, integration; com, composition Fig. 7 Composite construction (after Langacker 1987: 84, 450)

Langacker’s (1987: 452–453) notion of compositionality consists of two conceptual metaphors, the CONTAINER metaphor and the BUILDING BLOCK metaphor, but with broader applicability. With regard to the former, an expression is viewed as providing access to various knowledge systems. As to the latter, the component structures serve to motivate or highlight various facets of the meaning of the

Models for analyzing figurative composite expressions | 49

composite structure. Thus, the composite expression is thought to evoke a semantic network to which neither of the components provides direct access (Benczes 2006: 73). Additionally, Langacker (1987: 449) highlights the degree of compositionality and believes that “linguistic phenomena lend themselves more easily to a claim of partial rather than full compositionality”. Many other linguists (e.g. Cruse 1991; Fabb 1998) are also aware of degrees of compositionality and reject a dichotomous notion of it. Kavka (2009: 23) offers a cline of compositionality: fully compositional (e.g. shoot a bird, red ink), semi-compositional (e.g. shoot a film, red carpet), or non-compositional (e.g. shoot the breeze, blue blood). It is precisely this gradable nature of compositionality that makes it rather difficult to make a clear demarcation between compounds and phrases (Matthews 1974). Claiming that the notion of compositionality refers to the intrinsic complexity of a structure, Langacker (1987: 448, 457, 2000b: 13) distinguishes it from the notion of analyzability, which refers to a person’s ability to be cognizant of this complexity and apply some kind of analysis to a complex structure. In other words, compositionality deals with the degree to which the semantic value of a composite structure as a whole is added up from the values of its components, while analyzability concerns the extent to which components are discernible and isolated within a composite expression as well as a person’s awareness of the contribution that each component makes to the composite whole. In Philip’s (2011: 18) words, analyzability refers to “the ease with which [an idiom] can be taken apart and its meaning understood from the meanings and syntactic roles of its constituents”. Langacker (1987: 462, 2000b: 13) maintains that analyzability is also a matter of degree. For example, these expressions run the gamut from full analyzability to virtual opacity: diversifier > complainer > printer > computer > propeller > drawer. He indicates that printer may be invariably regarded as “something that prints”, computer tends to not be regarded as “something that computes”, and propeller may never be viewed as “something that propels”. In a word, compositionality and analyzability are different perspectives in Langacker’s schema of composite construction (Figure 7): “compositionality describes the bottom-up perspective on composition”, while “analyzability captures the composition process top-down” (Langlotz 2006: 89–90). The notions of compositionality and analyzability provide the fundamental theoretical basis of the three models for analyzing figurative composite expressions introduced in the following sections.

50 | Metonymy in expressions

2.1.2 Fauconnier and Turner’s blending model The core of blending theory (Fauconnier and Turner 1998, 2002) is the conceptual integration network. Instead of focusing on two-domain mapping as in Conceptual Metaphor Theory (Lakoff and Johnson 1980), the conceptual integration network involves a multiple-space integration process and prototypically is composed of four mental spaces17: two input spaces containing partial counterpart connections, a generic space containing what the inputs have in common and a blended space containing generic structure captured in the generic space and an emergent structure, whose elements are not found in the inputs. In blending, the projection of structures from inputs to the blend is partial and selective; that is, not all the elements from the inputs are projected (see Figure 8). To construct the emergent structure in the blend, three processes are involved. First, composition composes elements from the inputs and provides relations available in the blend that do not exist in the separate inputs. Second, completion brings additional structure to the blend by recruiting a great range of familiar and conventional conceptual structures and knowledge. Third, elaboration develops the blend through imaginative mental simulation according to principles and logic in the blend (Fauconnier and Turner 1998: 142–144, 2002: 40–44). Through these three operations, an emergent structure, which is nonexistent in the inputs, is built in the blend. To avoid improperly exaggerating the power of the blending theory (Coulson and Oakley 2001: 168), Fauconnier and Turner (1998: 162–177) suggest a number of “Optimality Principles” that constrain the process of blending. Space limitations prevent me from elaborating on these constraints in detail, but, in short, the optimality constraints make blending theory more effective and principled (Coulson and Oakley 2001: 186).

|| 17 Mental spaces are small conceptual packets constructed as we think and talk, for the purposes of local understanding and action. As partial assemblies, they contain elements and are structured by frames and cognitive models (Fauconnier and Turner 1998: 137; Fauconnier 2007: 351).

Models for analyzing figurative composite expressions | 51





 





Fig. 8: Conceptual blending (Fauconnier and Turner 1998: 143)

With its promise of accounting for various linguistic and cognitive phenomena, blending theory has been widely applied to analyzing composite expressions; see, among others, Sweetser (1999), Fauconnier and Turner (2002: 353–406) and Schmid (2011). Coulson’s (2001) study marks an important milestone as an insightful examination of conceptual blending in modified noun phrases. With detailed examples, she shows that the processes of meaning construction in nominal compounds (e.g. gun wound) and privative adjectives (e.g. fake guns) are well captured by the conceptual integration network. Coulson and Oakley (2003) then explore composite expressions that employ both metaphor and metonymy with the help of blending theory. They attempt to explicate the important role of metonymy in the process of conceptual blending by maintaining “connections between networks of mental spaces comprised of opportunistic juxtapositions of conceptual structure from distally related domains” (Coulson and Oakley 2003: 78). Metonymy thus optimizes conceptual integration as one of the optimality constraints, i.e. “metonymy tightening” (Coulson and Oakley 2001: 186) or the “metonymy projection constraint” (Fauconnier and Turner 1999: 85). A good example of the application of blending in composite expressions in which metaphor and metonymy interact is digging your own grave (Fauconnier and Turner 1998; Coulson 2001; Coulson and Oakley 2003). We know that the expression He is digging his own financial grave means “he is making lots of mistakes, which will lead him to failure”. The blend analysis is presented in Figure 9. One input space includes the elements of the digger, the action of digging, the grave, dying and so on. The other input space contains the ele-

52 | Metonymy in expressions

ments of the agent, the action of making mistakes, bankruptcy and so on, and is also the target space of the conceptual metaphor. Each input space further embodies some relations of the elements, and the relations in the two input spaces show a certain discrepancy: in the Dig Grave space, the action of digging is intentional and has the grave as its goal, but digging is not the cause of death and the depth of the grave is irrelevant; in the Mistake space, the action of making mistakes is unintentional (i.e. he is not aware of the mistakes and keeps making more mistakes), making mistakes is the cause of failing, and the amount of mistakes he makes is relevant. Elements in input spaces may be associated by conceptual metonymy or metaphor, such as grave’s metonymic link to death, which is metaphorically associated with failure. A CAUSE–EFFECT metonymy works in the Mistake space to the extent that mistakes lead to bankruptcy. We build a blend space by projecting elements and relations selectively from both inputs and get an idiomatic interpretation of “making some mistakes which will cause one to fail”. 

LJŝŶŐĂŶĚĚŝŐŐŝŶŐ ŐƌĂǀĞƐĨŽƌƚŚĞĚĞĂĚ

&ŝŶĂŶĐŝĂůŵŝƐƚĂŬĞƐ ůĞĂĚŝŶŐƚŽďĂŶŬƌƵƉƚĐLJ



 

ŝŐŐŝŶŐLJŽƵƌŽǁŶ ĨŝŶĂŶĐŝĂůŐƌĂǀĞ

 ĂƵƐĂůƐƚƌƵĐƚƵƌĞ /ŶƚĞŶƚŝŽŶĂůƐƚƌƵĐƚƵƌĞ &ƌĂŵĞƐƚƌƵĐƚƵƌĞ ǀĞŶƚ ƐƚƌƵĐƚƵƌĞ

Fig.9: The blend analysis of digging your own grave (after Fauconnier 2010: 58)

Looking back at Langacker’s (1987) distinction between compositionality and analyzability, Coulson (2001: 159–161) reminds us that conceptual blending is a

Models for analyzing figurative composite expressions | 53

noncompositional method of concept combination. However, it fulfills the requirement of analyzability, in the sense that “speakers need not deterministically derive the meanings of an expression from its component parts” and blended models can “be sensibly related to the input frames” (Coulson 2001: 160). Yet even though blending has been successfully applied to investigation of composite expressions, there remain several limitations on blending analyses. With regard to the approach taken by Coulson (2001), Benczes (2006: 62) points out two major limitations. First, the relation between input spaces and the composite components of the expression is not clear; second, Coulson did not explain how she decided what elements and relations to list in the input spaces. These two limitations are also found in other blending analyses. Another problem rests with the expressive and analytic power of blending theory in general. Benczes (2006: 59) argues that by relying too much on intuition in the process of analysis, blending tries to be too powerful and to explain everything. Gibbs (2001: 351–352) also draws attention to a number of questions for blending theory (e.g. how do specific blending operations, i.e. composition, completion and elaboration, function in the online creation of semantic meanings for expressions?) and suggests that psychological testing and theorizing should be incorporated into the application of blending theory. Additionally, Bundgaard et al. (2006) question the applicability of blending to compounding: blending analyses only consider meaning construal related to compounds from the exclusive “hearer’s point of view”. Also, Broz (2011) suggests that blending seems to be not an adequate descriptive or explanatory tool for the analysis of kennings18 on account of the difficulty of identifying the common characteristics of the two inputs, which determine the generic space and give rise to the blend.

2.1.3 Benczes’s analysis of noun-noun compounds With a heavy reliance on blending theory, Benczes (2006: 59) examines English creative noun-noun compounds and maintains that “blending theory is not enough in itself to explain the semantics of compounds”. Following Langacker’s (1987) account of composition, as well as Libben (2006) and Lamb’s (1998) con-

|| 18 A kenning, which is frequent in Old English poetry, may be loosely defined as “the poetic interpretation or description of a thing or thought by means of a condensed simile: and in Old English such a condensed simile normally takes the form of a compound” (Gardner 1969: 109). For instance, mere-hengest is a kenning that implies a simile comparing a ship to a horse (hengest), with the ship moving over the sea (mere) as a horse does over the land.

54 | Metonymy in expressions

nectionist approach19 to compounding, Benczes explores in a systematic way how metaphor and metonymy play a role in creative compounds. Diverging from Dirven and Verspoor’s (1998: 59–60) discussion of placing compounds on a cline of transparency20, Benczes (2006: 77) demonstrates that given the cognitive interpretation of metaphor and metonymy (Lakoff and Johnson 1980) as basic everyday processes of thought, “metaphorical or metonymical compounds are just as normal and everyday constructions as nonmetaphorical or nonmetonymical ones”. Benczes further points out that the difference between a nonmetaphorical compound such as apple tree and a metaphorical one such as red tape is not transparency but a matter of linguistic creativity: metaphorical or metonymic compounds involve a more creative process of word formation. In her analysis of creative English compounding, Benczes shows a systematic way of examining the meaning construction of metaphorical/metonymic noun-noun combinations with the help of blending theory, and uncovers the regularities behind the cognitive processes that yield creative compounds and the products of the processes (Benczes 2006: 3–4). One of Benczes’s examples, office-park dad (“a married, suburban father who works in a white-collar job”), will be illustrated next, to show how her analysis of creative compounds works. With the help of the blending model (see Figure 10), she finds that a conceptual metonymy operates on the first constituent, office-park: a defining property (office park, which denotes a large office complex located outside of residential areas) stands for a category (a certain kind of white-collar job, which is well-paid). This metonymy is rooted in the Office-Park ICM, which serves as one of the input spaces. The second constituent, dad, belongs to the DAD domain, which serves as the other input space. Each input space contains a number of elements from our everyday knowledge of the two, and the blend is a combination of the selective elements from both input spaces.

|| 19 The connectionist model of the mental architecture (Lamb 1998; Libben 2006) suggests that the mind relies on parallel processes of understanding a compound on the basis of the two individual constituents and understanding it as a whole unit (Benczes 2006: 85). 20 Dirven and Verspoor (1998: 59–60) classify compounds into fully transparent compounds like apple tree, partially transparent compounds like blackbird and nontransparent compounds or darkened compounds like red tape.

Models for analyzing figurative composite expressions | 55



 



   

          



 



   

  

  

       





   

Fig. 10 The blend analysis of office-park dad (Benczes 2006: 142)

„

 €ƒ

­ €‚ƒ

 €ƒ

Fig. 11: Schematic model of a creative compound with a metonymy-based modifying constituent (Benczes 2006: 141)

Compounds with metonymic modifiers like office-park dad can be schematically presented as in Figure 11. The compound is composed of two lexical items21 (i.e. symbolic units), each consisting of a semantic unit (X and Y), a phonological unit ([x] and [y]) and the established symbolic relationship between the two. The first constituent, the modifier, is an item whose semantic unit is embedded in the ICM. Given the single-domain definition of metonymy as a part of an ICM,

|| 21 Strictly speaking, office-park dad has three lexical items; however, we can simplify the combination of office-park as one symbolic unit.

56 | Metonymy in expressions

is related to the ICM by a conceptual metonymy. The composite meaning of the compound is a combination of the metonymic interpretation of the modifier and the meaning of the second constituent, the head. With a detailed analysis of numerous examples, Benczes looks at how metaphor and metonymy can affect the meaning of a compound’s constituents: the metaphoric/metonymic modifier (e.g. armchair, office-park dad), the metaphoric/metonymic head (e.g. jailbird, handwriting), the metaphoric/metonymic modifier and head (e.g. flame sandwich, phone neck), the metonymic compound as a whole (e.g. bearskin) and the metaphor/metonymy-based semantic relation between the modifier and the head (e.g. beanpole family, lamppost). She also investigates the interaction of metaphor and metonymy in compounds as in the expression picasso porn, which involves a PRODUCER FOR PRODUCT metonymy in the modifier and a metaphorical link between the modifier and the head. One of the contributions of her study is that Benczes (2006: 190) shows that the meanings of creative noun-noun compounds, however innovative and unusual, can be systematically analyzed in the framework of CL (i.e. Cognitive Grammar, blending theory) and that the systematicity of creative compounds can be based upon which part of the compound is activated by the conceptual metaphor or/and metonymy. In her study, Benczes explores not only the various ways in which metaphor and metonymy act upon the constituents of the expression, but also how metaphor and metonymy operate between the two constituents (i.e. the integration dimension in Langacker’s 1987 schema of composition; see Figure 7). X

2.1.4 Geeraerts’s prismatic model To account for the semantic complexity as well as the interaction of metaphor and metonymy in composite expressions, Geeraerts (2002) offers a general architecture with two distinctive dimensions in meaning structuring. The paradigmatic dimension involves the relation between the literal meaning of the composite expression as a whole and its figuratively derived meaning. It also involves the relation between the literal meaning of the constituent parts and the interpretation that those constituents receive within the derived meaning of the expression as a whole. The syntagmatic dimensions describe the relationship between the meaning of the constituent parts and the meaning of the composite expression as a whole. It can be envisaged both with regard to the literal meaning and with regard to the derived meaning (Geeraerts 2002: 199). The semantic

Models for analyzing figurative composite expressions | 57

relations in composite expressions are graphically represented by means of a prismatic structure as in Figure 12. 

 





  









1 Expression as a whole in its literal reading 2 First constituent item in its literal reading 3 Second constituent item in its literal reading 4 Expression as a whole in its derived, idiomatic reading 5 First constituent item in its derived reading 6 Second constituent item in its derived reading Fig. 12: The prismatic structure of a composite expression (after Geeraerts 2002: 200)

The syntagmatic dimensions deal with the semantic process of composition and integration in Langacker’s (1987) terms (see Figure 7). Geeraerts (2002: 200) suggests that the notion of compositionality may receive either a dynamic or a static interpretation on the syntagmatic dimension of both the literal panel and the idiomatic panel: Within the dynamic interpretation, compositionality is thought of as a syntagmatic derivational process in the course of which the meaning of a compound expression is computed on the basis of the meanings of the constituent parts of the expression. In contrast with this dynamic, bottom-up conception, a static interpretation can be envisaged that merely notes that a one-to-one correspondence between the parts of the semantic value of the expression as a whole and the meanings of the constituent parts of the expression can be detected, regardless of the question whether this correspondence has come about through a process of bottom-up derivation or through a top-down interpretative process. (Geeraerts 2002: 200)

58 | Metonymy in expressions

He then terms the dynamic bottom-up derivation compositionality and gives a new name, isomorphism22, to the static, neutral, nondirectional derivation. However, the paradigmatic dimension deals with the process of semantic extension, in which conceptual metaphor/metonymy or other interacting cognitive mechanisms like generalization may work on both levels of the constituents and the expression as a whole. The concept motivation applies to this derivability in the paradigmatic dimension from the original meaning to the transferred meaning, and it involves both constituental motivation and global motivation. In a nutshell, both isomorphism and motivation involve the transparency of some links (see Figure 12): isomorphism as syntagmatic transparency and motivation as paradigmatic transparency (Geeraerts 2002: 201). Table 1 summarizes the key notions in the prismatic model. Coinciding with transparency in different dimensions, both isomorphism and motivation are scalar notions. Thus, we may have partially or fully motivated and isomorphic expressions. Table 1: Summary of key notions in the prismatic model

Dimension

Semantic process

Transparency

Where in the prism the process works

syntagmatic

semantic composition

isomorphism (static, nondirectional compositionality)

literal/figurative planes

paradigmatic

semantic extension

motivation

global/constituental levels

The prismatic model simplifies the formal structure of composite expressions by including only two core lexical items. It works well for analyzing a compound, which normally contains two components. Take the Dutch compound schapenkop (literally “sheep’s head”, idiomatically “dumb person”) as an example (Geeraerts 2002: 214) to illustrate how the prismatic model works (see Figure 13). The first part, schaap, metaphorically triggers a “sheep-like” reading (senses 2 to 5 in Figure 13); the second part, kop, does not undergo a semantic change from senses 3 to 6, and the intermediate reading, sense 4, “(human) head like a sheep,” can be considered isomorphic from senses 5 and 6; the ultimate reading, sense 7, “stupid person,” maintains the isomorphism: “stupid” is related by

|| 22 Langlotz (2006: 109, 115) regards isomorphism as the top-down re-analysis of an idiom’s compositional path instead of a nondirectional one-to-one correspondence.

Models for analyzing figurative composite expressions | 59

metaphorical similarity to the reading of “sheep-like” (senses 5 to 8) and a PART FOR WHOLE metonymy links “head” to “person” (senses 6 to 9). 



 









 











1 sheep’s head 2 sheep 3 head 4 (human) head like a sheep 5 sheep-like 6 head 7 stupid person 8 stupid 9 person Fig. 13: The prismatic structure of schapenkop (sheep’s head) (after Geeraerts 2002: 214)

One of the contributions of the prismatic model is the proposal of the isomorphism notion. Describing a nondirectional syntagmatic process, isomorphism provides a way to deal with composite expressions from both a compositional path and an analyzability path. Hence, the notion of isomorphism integrates the two perspectives (both bottom-up compositionality and top-down analyzability) of Langacker’s composition schema in the process of meaning construction. Another contribution lies in the reinterpretation process advanced in the prismatic model. It may work on both syntagmatic and paradigmatic dimensions. Geeraerts (2002: 209–211) proposes that semantic interpretation processes are not always a question of bottom-up compositionality (i.e. the meaning of the component parts determines the meaning of the expression as a whole), but that they can also be top-down: the overall meaning of a composite expression determines the specific meaning of its component parts. At the same time, reinterpretation along the paradigmatic dimension (i.e. figurative-to-literal transfer) may be found, too: new meanings may come about through the search for motivation, so it is not only the case that literal meanings determine figurative ones; figurative meanings also determine literal ones. With the help of the notions of isomorphism and reinterpretation, the prismatic model offers a more fine-grained and integrated analysis of semantic relations in complex meaning structuring, in the sense of allowing the projecting of meaning along different pathways in the prism: bottom-up (traditional semantic compositionality) and front-to-back (traditional semantic shift/extension) or top-down and back-to-front pathways; see Figure 14.

60 | Metonymy in expressions

   







Up-arrow dashed line: compositionality Down-arrow dashed line: analyzability (reinterpretation process of compositionality) Up & down-arrow dashed line: isomorphism Double-headed solid line: motivation and its reinterpretation process Horizontal dashed line: integration (cf. Figure 7) Fig. 14: Pathways of meaning construction in a prism

Providing a clear picture of how various cognitive mechanisms like metaphor and metonymy interact with each other in the meaning construction of composite expressions (e.g. metaphor and metonymy in a continuum, with a parallel presence, or existing interchangeably), the prismatic model serves as a necessary addition to the blending model (Geeraerts 2009: 87). Compared with the blending model, the prismatic model has the following main advantages. First, blending scholars tend to highlight the mismatch between the input space and the target space. For example, in the blending analysis of the expression dig one’s own grave, we find disparity of meaning relations between the Dig Grave space and the Mistake space (see Section 2.1.2 for discussion). However, the blending theory fails to explain why the very discrepancies between the two spaces do not simply block the metaphorical conceptualization of the target concept through such an inappropriate source image (Geeraerts 2009: 103). The prismatic analysis of dig one’s own grave is presented in Figure 15. The semantic mechanisms at work in the transition of the prism are not difficult to identify. On the top line (global semantic shift): from 1 to 4, it involves a generalization: excavating a burial place is one of the things that one would generally do when preparing for a death or a burial. Then, from 4 to 7, it is a metonymy between time and causality: a temporal relation is replaced by a causal relationship, that is, doing the things that come chronologically before the death or burial stands for doing the things that causally come before the death or burial (Geeraerts 2009: 98). On the bottom lines (constituental semantic shifts): the semantic shift from 3 to 6 can be easily identified as a common metonymy of GRAVE for DEATH; and from 6 to 9 is a hyperbole of using DEATH as an image for

Models for analyzing figurative composite expressions | 61

any type of trouble. The global meaning that arises from the semantic transition from 1 to 4 can be easily distributed over the constituents, thus, to dig in 5 gains the meaning of “to prepare” by the process of reinterpretation. Then, the semantic switch from 5 to 8 is similar to the transition from 4 to 7, in that it is grounded on a time-causality metonymy. 















 











1 to dig one’s own grave to excavate the burial place where one’s own dead body will be laid to rest 2 to dig to excavate, to hollow out (something) 3 one’s own grave the burial place where one’s own dead body will be laid to rest 4 to dig one’s own grave to prepare for one’s own death or burial, to anticipate one’s own death, to do things that come before the death or burial of someone (viz. oneself) 5 to dig to make preparations, to prepare for (something) 6 one’s own grave one’s own death or burial 7 to dig one’s own grave to contribute to one’s own downfall or plight, to do things that causally come before the downfall or plight of someone (viz. oneself) 8 to dig to contribute to (something) in such a way that it becomes more likely 9 one’s own grave one’s own downfall or plight Fig. 15: The prismatic analysis of dig one's own grave (after Geeraerts 2009: 102)

Without the emphasis on the discrepancies between the source and the target, “the prismatic analysis can not only easily incorporate the emergent structure

62 | Metonymy in expressions

phenomena that are cited by the proponents of the blending analysis, but it can also show why these phenomena are readily accessible at all, given the target domain interpretation” (Geeraerts 2009: 101). Second, the blending approach neglects various stages of sense development as well as the distinction between sense development on a global level (i.e. within the expression as a whole) and sense development on a constituental level (i.e. within component parts) (Broz 2011: 184). However, the prismatic model provides a more detailed approach for composite expressions in the sense of allowing clearer identification of various semantic mechanisms (e.g. metaphor, metonymy, hyperbole, etc.) within different parts/levels of the expression and at different stages of sense development. As metonymies often come in succession, in many cases chained metonymies may operate in the semantic transition within an expression. The prismatic model seems to easily untangle the chained metonymies in the process of meaning construction of the expression. It must be added that in the prismatic model, successive metonymies along either of the three lines (the top line and the two bottom lines) can be identified as a chained metonymy, but it disallows identifying successive metonymies interweaving across different lines, such as the first metonymy on the bottom right line and the successive metonymy on the bottom left line, as a chained metonymy. Nevertheless, Benczes (2006: 180) points out that one drawback of the prismatic model is no other than the sequence by which metaphor and metonymy operate in the meaning of the composite expression. She argues that it is often rather difficult to identify which cognitive mechanism contributes the meaning of the expression first, metaphor or metonymy. Actually, as an explanatory or interpretative model, the prismatic approach does not insist on only one feasible prismatic structure of a fixed sequence of semantic projections for a composite expression; sometimes it does allow alternative prismatic structures with different sequences of various cognitive mechanisms, as long as it can explain the semantic shift in the expression properly. The main point here is not that the prismatic model is an indisputable model or a more powerful model than blending theory in general, but rather that these two models may be complementary approaches to meaning construction. Blending theory has strong explanatory power in wider applications like humor studies (Howell 2007) and discourse analysis (Oakley and Hougaard 2008), whereas the prismatic approach shows more promise of capturing details of meaning structuring and systematizing various semantic mechanisms in figurative composite expressions. Therefore, the methodology for tackling metonymy

Metonymic mappings in Chinese expressions | 63

in expressions that is applied in this book relies primarily on the prismatic model.

2.2 Metonymic mappings in Chinese expressions So far, I have discussed three models for analyzing composite expressions in CL. In this section, I will explain how to apply the prismatic model to analyze Chinese expressions behind which conceptual metonymy plays a role. First, the effectiveness of the prismatic model for analyzing metonymies in Chinese expressions will be briefly demonstrated in Section 2.2.1. Then, with a large stock of concrete examples, I will show that conceptual metonymy may act upon Chinese expressions in a remarkable variety of ways in Section 2.2.2.

2.2.1 Bringing the prismatic model to Chinese metonymic expressions As demonstrated in Section 2.1.4, the prismatic model gives a better view of the internal structure of a composite expression and presents clearly how the various components contribute to the overall meaning of the expression in sequence. The meaning of a complex word registered in dictionaries23 may be its overall literal meaning, and the meanings of the constituents are largely dormant when the word is in use. Hence, without an effective model to decompose the semantic structure or trace the semantic extension process in the expression, it is difficult to know on which part of the expression or at which stage of the semantic extension metonymies operate. What’s even worse, we may ignore some hidden metonymies in the meaning construal of the expression. For example, according to dictionaries, the literal sense of nian–gu (4) is “imperial carriage” and it may also metonymically refer to “emperor”; see example (4). Thus, the metonymy of “carriage” for “emperor” is identified here. But how do the two constituents reach the meaning of “imperial carriage” first? Is there any metonymic mapping involved during the process of building the meaning of “imperial carriage”? The prismatic model (see Figure 16) may help us decompose the composite expression and show the isomorphic processes of

|| 23 Chinese dictionaries used in this book are mainly the Great Chinese Dictionaries (Lou 1993), the Chinese Etymology Dictionary (Ci Yuan Editorial Board 1980) and the largest online Chinese dictionary,  Han Dian (Chinese Dictionary, http://www.zdic.net/).

64 | Metonymy in expressions

the compound AB getting the two readings based on A and B. Hence, for the compound nian–gu, we may find two metonymic mappings: “round log of a wheel” to “carriage” (PART FOR WHOLE) and “imperial carriage” to “emperor” (POSSESSED FOR POSSESSOR). In this book, the metonymy operating on the global level of the prism is termed global metonymy, while the metonymy operating on the constituental level is called local metonymy. AB  nian–gu24

(4)

Imperial carriages. Metonymically refers to the emperor. A  nian Man-driven carriages in ancient times. After the Qin and Han Dynasties, it refers to the carriage of an emperor or a queen specifically. B  gu The round log in the center of a wheel, which connects to the spoke and has a round orifice for the axle. Generally refers to a carriage.





   



 





1 imperial carriage–round log of a wheel (the round log of a wheel of an imperial carriage) 2 imperial carriage (as a modifier) 3 round log of a wheel 4 imperial carriage 5 imperial carriage (as a modifier) 6 carriage 7 emperor Fig. 16: The prismatic structure of expressions like  nian–gu

In addition, sometimes dictionaries only note the figurative meaning of a compound. With the help of the prismatic model, we may easily fill in the missing stages in meaning construction of the figurative meaning by checking the literal meanings of both constituents. For example, the compound jiao–shu has only || 24 All Chinese examples in this chapter are selected from the Dictionary of Chinese Metonymic Senses (Han 1995) and the Dictionary of Chinese Substitutive Words (Zhang 1993). Dictionary definitions of Chinese examples are mainly from the Great Chinese Dictionaries (Lou 1993).

Metonymic mappings in Chinese expressions | 65

one sense registered in dictionaries, which is “beautiful woman” (sense 4 in the prism, Figure 17). The prismatic model decomposes the meaning construal process: first, the two constituents compose an overall literal meaning expressing the salient attribute of a beautiful woman and then, the overall literal meaning undergoes an ATTRIBUTE FOR PERSON metonymy.  









1 sweet, charming, beautiful 2 sweet, charming 3 beautiful 4 beautiful woman Fig. 17: The prismatic structure of expressions like jiao–shu

2.2.2 Various paths of metonymy in Chinese expressions Based on Chinese examples, a number of paths in which metonymy operates in expressions will be discussed in this section. Generally, there are two main morphological groups of Chinese expressions under discussion: monomorphemic words and polymorphemic words (or complex words). For monomorphemic words, the acting metonymy is conspicuous and easy to identify. For a polymorphemic word, it is not enough to claim that metonymy exists in its semantic transition because that does not tell us much about how metonymy works in the meaning extension or in which part of the complex word metonymy plays a role. Thus, the prismatic model is employed to unravel the participation of metonymy in the meaning construction of polymorphemic words. The prismatic model allows us to tackle the problem of how the constituents construct the literal meaning of the composite expression as a whole, and at the same time it shows the correspondence between the constituents and their conceptual equivalents in the derived figurative meaning. Metonymies in Chinese expressions are classified into seven types according to the morphological structure of the expression and the locus of metonymy in the expression.

66 | Metonymy in expressions

Metonymy acts upon a monomorphemic word: Type 0 Most of the words in this type are monosyllabic words, but it also includes semireduplicates25, as in (5). Recognizing semi-reduplicates in contrast to normal coordinate compounds26 is crucial for the identification of metonymy. In practice, even with the help of phonological principles, it is hard to tell semireduplicates from coordinate compounds immediately. Thus, I resort to the lists of semi-reduplicates collected by Chinese grammarians, and the reduplication compound dictionaries (Gao 2001; Xie 2011), to tell semi-reduplicates from coordinate compounds. We also find chained metonymies in the semantic transfer of a monomorphemic word, as in (6). (5)

 shi [room] → wife (LOCATION FOR LOCATED)  yao–tiao [(of a woman) gentle and graceful] → beautiful woman (ATTRIBUTE FOR PERSON)

(6)

 xuan [tawny daylily] → mother’s residence → mother (LOCATED FOR LOCATION and then LOCATION FOR LOCATED)

Metonymy acts upon the first constituent of a complex word: Type 1 This type involves expressions whose first constituent undergoes a metonymic shift or successive metonymic shifts. For the compound rou–zhen (7) (Figure 18),

|| 25 Semi-reduplicates, i.e.  lian mian ci, according to Chinese grammarians (He and Jiang 2010: 26–29), are actually disyllabic monomorphemic words in Chinese. Normally, a Chinese syllable is a combination of a consonant plus a vowel. People build the semireduplicates mainly following certain phonological principles, e.g. repeating initial consonants (i.e. alliteration  cen–ci) or vowels (i.e. rhyming  yao–tiao) of the two syllables. One of the significant characteristics of semi-reduplicates is that we cannot divide the meanings of the two parts; in other words, they are semantically noncompositional. For instance, neither of the two constituents has its meaning independently (e.g.  xi–xu “to sign”) or only one of the parts has the meaning on its own and the other is just a dependent semantic unit (e.g.  hu– die “butterfly”;  die may refer to “butterfly” and  hu is meaningless independently). 26 Coordinate compounds in Chinese are those that show a logical coordination between the constituents (“and”), as in  shu–guo [vegetable–fruit] “vegetable and fruit”, or show a semantic relation of synonymy (e.g.  liang–li [pretty–beautiful] “pretty/beautiful”), antonymy (e.g.  hu–xi [exhale–inhale] “breathe”), redundancy (e.g. song–shu [pine–tree] “pine tree”) (Ceccagno and Basciano 2009: 481). Coordinate compounds may have both constituents as heads (endocentric coordinate compounds) or no head at all (exocentric coordinate compounds), and the two constituents have the same weight in indicating the meaning of the compound as a whole.

Metonymic mappings in Chinese expressions | 67

the metonymy PART FOR WHOLE acts on the first constituent, which is the modifier. The compound xuan–chun (8) shows a case in which successive metonymies and a metaphor interact: LOCATED FOR LOCATION and then LOCATION FOR LOCATED (2 to 5 and then 5 to 8 in Figure 19) work upon the first constituent of the compound, while the semantic extension of the second constituent involves a metaphorical mapping in the sense of FATHER resembling a long-lived Chinese tree in term of long living.  rou–zhen [flesh–array] → array of maid servants or concubines or

(7)

beautiful women xuan–chun [tawny daylily–long-lived Chinese tree] → mother and

(8)

father  

  











1 array of flesh 2 flesh 3 array 4 array of maidservants or concubines 5 maidservants or concubines 6 array Fig. 18: The prismatic structure of  rou–zhen [flesh–array]

 



 









 













1 tawny daylily–long-lived Chinese tree 2 tawny daylily 3 long-lived Chinese tree 4 mother’s residence and long-lived Chinese tree 5 mother’s residence 6 long-lived Chinese tree 7 mother and father 8 mother 9 father Fig. 19: The prismatic structure of  xuan–chun [tawny daylily–long-lived Chinese tree]

68 | Metonymy in expressions

Metonymy acts upon the second constituent of a complex word: Type 2 In this type, the occurrence of metonymy occurs in the second constituent of the expression in the form of a singular mapping or successive mappings. In the expression liang–shou (9) (Figure 20), the metonymy PART FOR WHOLE works on the second constituent, and the expression as a whole achieves the ultimate derived meaning of “two people”. In the expression lao–ge (10), two successive metonymies are identified in its second constituent (3 to 6 and 6 to 9 in Figure 21), first MATERIAL FOR OBJECT and then PIECE OF CLOTHING FOR PERSON. liang–shou [two–hand] → two people

(9)

 

  

 







1 two hands 2 two 3 hands 4 two people 5 two 6 person Fig. 20: The prismatic structure of  liang–shou [two–hand]

lao–ge [old–leather] → old soldier

(10)





 









 













1 old–leather 2 old 3 leather 4 old-armor made of leather 5 old 6 armor made of leather 7 old soldier 8 old 9 soldier Fig. 21: The prismatic structure of  lao–ge [old–leather]

Metonymic mappings in Chinese expressions | 69

Metonymy acts upon a complex word as a whole: Type 3 This type includes complex words where the word as a whole is metonymic. If we apply the prismatic model in this case, the metonymy only operates on the global level, i.e. the top line of the prism, and there are no semantic shifts on the constituental levels. The example dong–gong (11) is an attributive compound27 literally referring to the palace of the crown prince (i.e. the eastern palace) (Figure 22). The metonymy LOCATION FOR LOCATED acts upon the overall literal meaning and triggers the metonymic meaning of “crown prince”. Chang–xiu (12) is also an attributive compound with a literal meaning of “long sleeves” and it may metonymically denote the meaning of “dancing girl” through two successive metonymies (Figure 23): PART FOR WHOLE and then PIECE OF CLOTHING FOR PERSON.  dong–gong [eastern–palace (living quarters of the crown prince)] →

(11)

crown prince  







1 eastern palace 2 eastern 3 palace 4 crown price Fig. 22: The prismatic structure of  dong–gong [eastern palace]

(12)

 chang–xiu [long–sleeves] → dress with long sleeves → dancing girl

|| 27 Attributive compounds in Chinese are “those in which the constituents have a modifierhead relation” (Ceccagno and Basciano 2009: 481). The modifier could be an adjective/noun that expresses a property of the head (e.g.  zhu–ye [main–page] “homepage”,  di–jia [land–price] “land price”), an adjunct modifying the head (e.g.  kou–suan [mouth– calculate] “to calculate orally”), or a verbal modifier ( mai–dian [sell–point] “selling point”).

70 | Metonymy in expressions





 







1 long sleeves 2 long 3 sleeve 4 dress with long sleeves 5 dancing girl Fig. 23: The prismatic structure of  chang–xiu [long–sleeves]

This type also includes cases in which metaphor or other cognitive mechanisms play a role in the semantic extension process. For example, the compound tiao– geng (13) has a metaphorical meaning of “to govern national affairs” and then it may metonymically refer to “a prime minister in feudal China”. The global metonymy ACTION FOR AGENT acts on the compound as a whole (senses 4 to 7 in Figure 24). The local mappings of the example tiao–geng will be discussed later in this section.  tiao–geng [to season–soup] → prime minister in feudal China

(13)









  





 





1 to season–soup 2 to season 3 soup 4 to govern national affairs 5 to govern 6 national affairs 7 prime minister in feudal China Fig. 24: The prismatic structure of  tiao–geng [to season–soup]

Metonymic mappings in Chinese expressions | 71

Metonymy acts upon both constituents of a complex word on a constituental level: Type 4 This type deals with complex words where parallel metonymies operate on both the first and the second constituents. The cases in this category are coordinate compounds (see Footnote 26) mainly but not exclusively. For coordinate compounds, sometimes both constituents may share the same metonymic target through the same metonymic process or different metonymic processes. For instance, in the compound xu–mei (14), which literally refers to “beard and eyebrows”, both constituents go through a PART FOR WHOLE metonymic process and achieve a metonymic meaning of “a male” (Figure 25). The constituents of the compound guan–gai (15), however, undergo different metonymic mappings (PIECE OF CLOTHING FOR PERSON acts on guan and chained metonymies PART FOR WHOLE and then POSSESSED FOR POSSESSOR operate on gai), but both constituents reach the same target of “high official” (Figure 26). (14)

 xu–mei [beard–eyebrows] → a male

(15)

 guan–gai [official hat–circular covering on ancient vehicle] → high officials





  



 





1 beard–eyebrows 2 beard 3 eyebrows 4 a male 5 a male 6 a male Fig. 25: The prismatic structure of  xu–mei [beard–eyebrows]

72 | Metonymy in expressions





 









 











1 official hat–circular covering on ancient vehicle 2 official hat 3 circular covering on ancient vehicle 4 official hat and ancient vehicle 5 official hat 6 ancient vehicle 7 high official 8 high official 9 high official Fig. 26: The prismatic structure of  guan–gai [official hat–circular covering on ancient vehicle]

 bing–can [sick–handicapped] → sick person and handicapped per-

(16)

son  

  





 



1 sick–handicapped 2 sick 3 handicapped 4 sick person and handicapped person 5 sick person 6 handicapped person Fig. 27: The prismatic structure of bing–can [sick–handicapped]

The constituents of coordinate compounds may also refer to different metonymic targets. For example, both constituents in bing–can (16) undergo the ATTRIBUTE FOR PERSON metonymy (Figure 27), but with different targets: “sick person” and “handicapped person”, respectively28. Similarly, the compound bai–hei (17)

|| 28 Someone may argue that the two targets “sick person” and “handicapped person” do not constitute different meanings, but that there is only one schematic category “sick or handi-

Metonymic mappings in Chinese expressions | 73

(Figure 28) involves parallel metonymic chains (ATTRIBUTE FOR THING and then PIECE OF CLOTHING FOR PERSON) on the two constituents for different targets of “common people” and “monk”.  bai–hei [white–black] → common people and monk

(17)





 









 













1 white–black 2 white 3 black 4 white clothes and black clothes 5 white clothes 6 black clothes 7 common people and monk 8 common people 9 monk Fig. 28: The prismatic structure of  bai–hei [white–black]29

Attributive compounds that have metonymy-based modifiers and metonymybased heads, or subordinate compounds30 that have metonymic arguments and metonymic heads, may also be classified as this type. For instance, in ci–kun (18) (Figure 29), as an attributive compound, both the modifier and the head undergo metonymic processes and refer to the same target “queen mother”. The metonymy ATTRIBUTE FOR PERSON acts upon the modifier, while a metonymy chain works on the head: PART FOR WHOLE and then LOCATION FOR LOCATED. (18)

 ci–kun [kind–passageway inside the palace] → queen mother

|| capped person”. In this study, we followed the dictionaries to treat bing–can as a coordinative compound with two different targets on its constituents. 29 In Buddhism, white clothes are typical for common people and black clothes for monks. 30 Subordinate compounds in Chinese are “those in which constituents have an argumenthead (or head-argument) relations” (Ceccagno and Basciano 2009: 480). Two subtypes are further proposed: first, either a verbal or deverbal head projects an argument satisfied by the nonhead constituent, e.g.  xin–teng [heart–to grieve] “to be sorry/sad”,  tong–guan [clear out–customs] “clear the customs”,  zou–gao [climb–high] “rise”. Second, a relational noun is the head and the nonhead acts as a semantic argument saturating the noun head, e.g.  jing–sao [police–sister] “respectful term for a policeman’s wife”.

74 | Metonymy in expressions





 









 











1 kind–passageway inside the palace 2 kind 3 passageway inside the palace 4 kind inner court 5 kind 6 inner court 7 queen mother 8 queen mother 9 queen mother Fig. 29: The prismatic structure of  ci–kun [kind–passageway inside the palace]

Metonymy first acts upon the first constituent of a complex word and then acts upon the complex word as a whole: Type 5 Complex words of this type have a metonymy-based first constituent and a nonmetonymic second constituent at the first stage, and then the derived meaning of the complex word as a whole involves another metonymy process on the global level.  qing–gong [blue–palace] → crown prince

(19)

 

 

  





 



1 blue palace 2 blue 3 palace 4 eastern palace 5 eastern 6 palace 7 crown prince Fig. 30: The prismatic structure of  qing–gong [blue–palace]

For example, qing–gong (19) is an attributive compound, where the two constituents refer to “blue” and “palace” respectively. First, the metonymy ATTRIBUTE

Metonymic mappings in Chinese expressions | 75

FOR THING31 acts

upon the first constituent, which gets a derived overall meaning of “eastern palace” (sense 4 in Figure 30); then the metonymy LOCATION FOR LOCATED works on the complex word as a whole, triggering the final interpretation of “crown prince”. The compound e–mei (20) undergoes a similar but more complicated semantic extension process, in the sense of interweaving with metaphor. First, the modifier e [moth] involves a WHOLE FOR PART metonymic mapping; then the complex sense “antenna of moth–eyebrows” as a whole goes through a metaphorical mapping (4 to 7 in Figure 31) to get a derived meaning of “eyebrows like antennae of moths”; finally, this overall meaning further undergoes a PART FOR WHOLE metonymy for the meaning of “beautiful woman”.  e–mei [moth–eyebrows] → beautiful woman

(20)

 

 

 

  

 







1 moth–eyebrows 2 moth 3 eyebrows 4 antenna of moth–eyebrows 5 antenna of moth 6 eyebrows 7 eyebrows like antennae of moths 8 beautiful woman (person with eyebrows like antennae of moths) Fig. 31: The prismatic structure of  e–mei [moth–eyebrows]

Metonymy first acts upon the second constituent of a complex word and then acts upon the complex word as a whole: Type 6 There are also complex words where the second constituent first undergoes a metonymy and then another metonymy acts on the global level of the complex word. For example, in the compound fei–xi (21) (Figure 32), first, a MATERIAL FOR OBJECT metonymy works on the second constituent, and a metaphor operates from the sense “to fly” to “to roam” on the first constituent; then the derived

|| 31 According to the system of five-element categorization in Chinese philosophy (Yu 2009: 359), each of the four directions is associated with a color; the color for the east is blue.

76 | Metonymy in expressions

meaning of the complex word as a whole “roam with a monk’s staff made of tin” involves the metonymy ACTION FOR AGENT for the final meaning of “a monk”. In the compound Qi–e (22) (Figure 33), the metonymy ATTRIBUTE FOR PERSON first acts on the second constituent and then the whole compound goes through an INDIVIDUAL FOR COLLECTION metonymic process to its final metonymic reading of “singing girls”32. (21)

 fei–xi [to fly–tin] → monk

(22)

 Qi–e [Qi–beautiful] → singing girls   





  



 



1 to fly–tin 2 to fly 3 tin 4 to roam with a monk’s staff made of tin 5 to roam 6 monk’s staff made of tin 7 monk Fig. 32: The prismatic structure of  fei–xi [to fly–tin]





   



 



1 Qi State–beautiful 2 Qi State 3 beautiful 4 woman of Qi State 5 Qi State 6 woman 7 singing girls Fig. 33: The prismatic structure of  Qi–e [Qi–beautiful]

|| 32 The motivation for the INDIVIDUAL FOR COLLECTION metonymy in the example Qi–e is that in ancient China, women of the Qi State were considered good at singing; thus they are regarded as the paragon of singing girls.

Metonymic mappings in Chinese expressions | 77

 wu–qun–ge–shan [dancing–skirt–singing–fan] → traditional

(23)

female entertainers



 

  





 











1 dancing 2 skirt 3 singing 4 fan 5 dancing skirt 6 singing fan 7 dancing skirt and singing fan 8 traditional female entertainers 9 traditional female entertainers 10 traditional female entertainers Fig. 34: The prismatic structure of  wu–qun–ge–shan [dancing–skirt–singing–fan]

So far, all the examples for polymorphemic words are actually bimorphemic. Similarly, complex words with more than two morphemes can be simplified to include only two core constituents in the prism and then be classified into different types according to the locus of metonymy within the morphological structure of the expression. This shows another advantage of the prismatic model: it allows different layers of composition in polymorphemic words. For instance, the expression wu–qun–ge–shan (23) is composed of two cores (senses 5 and 6 in Figure 34), which are built from other component parts at the bottom level of the prism. In this sectin, I have classified metonymies in Chinese expressions based upon their various paths in meaning construction. I would like to stress the following points: First, there are many cases whose semantic readings are both isomorphic and motivated. In these cases, the global metonymy is directly inherited from the local metonymy, which has an extreme independence and whose metonymic motivation can be realized by the stand-alone constituent. In other words, we can identify a consistent one-to-one mapping between elements of the global meaning and the meanings of the constituents of the complex word (Geeraerts 2002: 202). Take rou–zhen, seen in example (7) and Figure 18, for instance; the global metonymic mapping from “array of flesh” to “array of maidservants or concubines or beautiful women” (senses 1 to 4) is directly inherited from the

78 | Metonymy in expressions

local metonymic mapping from “flesh” to “maidservants or concubines or beautiful women” (senses 2 to 5). As Geeraerts (2002: 203) states, the isomorphic relation identified in the prism should not be confused with the question of whether the isomorphically mapped meanings of the constituents of the complex word are themselves motivated or not. For instance, in the compound tiao–geng, seen in example (13) and Figure 24, while the second constituent “soup” maps onto the “national affairs” part of the global metaphorical reading “to govern national affairs”, there is no independent motivation for the semantic shift nor readily conceivable metaphors from “soup” to “national affairs”. In this example, there is no independent paradigmatic motivation at the bottom level of the prism for the second constituent. Second, it is common to find an interaction of metonymy and metaphor in the prism, like the example of e–mei; see example (20) and Figure 31. Besides metaphor, there are other semantic mechanisms that may interweave in the prism, such as generalization and specification. The classification here only takes the locus of metonymy into account and leaves out the various ways in which other conceptual mechanisms operate on the meaning construction of the complex words. Third, meaning construction in Types 0, 1, 2, 3 and 4 may yield two subtypes due to the immediate or serial application of metonymy. Types 5 and 6 only deal with cases in which at least two successive metonymies play a role. It is worth noting that immediate application of metonymy does not imply onestage semantic extension, that is, a single prism, because there might be a metaphor preceding or following the metonymy, which actually implies multiplestage semantic extension, that is, more than one prism; see example tiao–geng (13) in Figure 24. Fourth, Type 0 and Type 3 reveal the important role of metonymy in the semantic extension of a word, while Types 1, 2, 4, 5 and 6 demonstrate the function of metonymy not only in the sense development of a word but also in its inner semantic construal. More generally, the former group deals with metonymy on a purely lexical-semantic level, while the latter group deals with metonymy at an interface between semantics and morphology. In addition, Type 0 and Type 3 only involve global metonymies, while Types 1, 2, 4, 5 and 6 deal with both global and local metonymies at the same time. Finally, unlike Benczes in her 2006 study, I only look at the various paths of metonymy in meaning extension in a paradigmatic dimension in the prism. This is in accordance with Benczes’s classification in general, but excludes one of her classifications, the metonymy-based relation between the two constituents

Some criteria for classification | 79

of a complex word, which deals with the role of metonymy in the integration process of complex meaning (see Figure 14).

2.3 Some criteria for classification The classification of the various paths of metonymy in Chinese expressions in Section 2.2.2 is only programmatic. In practice, the classification is often less straightforward. At the same time, as explained in Section 2.1.4, the prismatic model may allow more than one prismatic structure, with flexible sequences of semantic projections for a composite expression. A number of complications arise in the process of identifying metonymy with the prismatic model. In the following sections, I would like to mention three complications in the differentiation of types.

2.3.1 Global metonymy versus parallel local metonymies The differentiation between Type 3 and Type 4 is tricky, but it is crucial for the quantification of metonymic mapping. For the metonymic mapping calculation, an expression treated as Type 3 has only one global metonymy or a global metonymic chain; however, an expression treated as Type 4 has two parallel local metonymies or metonymic chains on the constituental level. Hence, several criteria are needed to decide whether the meaning of the compound as a whole serves as a metonymic source (Type 3) or each constituent of the compound serves separately as a metonymic source (Type 4). Although, as I have argued above, Type 4 mainly includes coordinate compounds, not all the coordinate compounds belong to Type 4. For example, the coordinate compound  ren–liang [kind–good] should be classified as Type 3, because, with a literal meaning of “kind and good”, its metonymic interpretation is “a kind and good person”. We thus find that the ATTRIBUTE FOR PERSON metonymy can work on the literal meaning of a compound as a whole, namely, the overall literal meaning of the compound serves as the metonymic source. For a given coordinate construction AB, I formulate the following criteria for type classification. Criterion (1): If A and B each trigger a different target concept, the coordinate construction AB is classified as Type 4.

80 | Metonymy in expressions

In this case, two parallel local metonymies or metonymic chains are identified in the compound. Most coordinate compounds with an antonymous or opposite semantic relation between the constituents, such as the examples in (24), fit this case. Coordinate compounds with simple logical coordination between constituents, as in (25), may also trigger different target concepts. (24)

 liang–jian [good–low] → good people and low people  ren–tan [kind–greedy] → kind person and greedy person

(25)

 gu–ji [parentless–handicapped] → orphan and handicapped person  tang–shi [hall–room] → mother and wife

Criterion (2): If A and B trigger the same target concept by different metonymic patterns, the coordinate construction AB is classified as Type 4. For example, in the coordinate compound jin–juan (26), we can identify two local metonymic mappings from both jin and juan to the metonymic interpretation of “students” in the prismatic structure. The metonymic pattern PIECE OF CLOTHING FOR PERSON (or more specifically, ACCESSORY FOR PERSON) operates on jin and the metonymic pattern POSSESSED FOR POSSESSOR operates on juan. Thus, two different metonymies are identified in the semantic extension process of the compound jin–juan. (26)

AB  jin–juan [headscarf–book] 1. The headscarf and the book, used by students of the Imperial College. 2. Students. A:  jin [headscarf] → students B:  juan [book] → students

Criterion (3): If A and B trigger the same target concept by the same metonymic pattern but literally denote different concepts, the coordinate construction AB is classified as Type 4. For instance, both A and B in hong–fen (27) have the same metonymic reading of “beautiful woman” by the same metonymic pattern POSSESSED FOR POSSESSOR. Nonetheless, the two constituents literally denote different concepts: hong “blusher” and fen “powder”.

Some criteria for classification | 81

(27)

AB  hong–fen [blusher–powder] 1. The blusher and powder used by women for make-up. 2. Metonymically refers to a beautiful woman. A  hong [blusher] → a beautiful woman B  fen [powder] → a beautiful woman

However, given the hierarchical structure of categories (see Murphy 2002), the judgment of “different literal concepts” is too subjective to be sufficient for identifying type. To what extent are two constituents accepted as having different literal meanings? For example,  ying–hu [tassel of official hat–scepter]; on a specific level, A and B refer to two different concepts, TASSEL OF OFFICIAL HAT and SCEPTER. However, on a more general level, both A and B belong to the same concept category with a general sense of ESSENTIAL OBJECT POSSESSED BY AN OFFICIAL IN ANCIENT CHINA. Thus, to meet the third criterion, we need the following four supplementary criteria. Criterion (3)(i): If the coordinate construction has a generalized literal meaning as a whole registered in dictionaries, it is treated as Type 3. It should be pointed out that the generalized literal meaning does not equate to the overall literal meaning of a complex word. The former is a meaning derived by a process of generalization (sense 4 in Figure 35), which involves a semantic transfer on a paradigmatic dimension, while the latter is a meaning based on the process of integration (sense 1 in Figure 35). (28)

 gao–liang [fat meat–fine grain] Fat meat and fine grain. Generally refers to rich food. Metonymically refers to sons of rich families.  ge–mao [dagger axe–spear] A dagger axe and a spear. Generally refers to weapons. Soldiers, army.  chuan–yu [make a hole in a wall–climb a wall] To make a hole in a wall and climb the wall. Refers to stealing. Also refers to thief.

The logic of treating expressions like those in (28) as Type 3 is that the generalized literal meaning registered in dictionaries shows a certain degree of conventionality. People may conceptualize “sons of rich families” by the rich food they have for meals, “soldiers” by their weapons and a “thief” by his action of steal-

82 | Metonymy in expressions

ing without tracing their conceptualization back to what kind of food or weapon or stealing action it is. Another reason for not treating gao–liang, ge–mao and chuan–yu as Type 4 is that for each of these compounds, neither of the two constituents independently has a metonymic target category PERSON in dictionaries. The prismatic model of this case is represented as in Figure 35. 







  



 



1 fat meat–fine grain 2 fat meat 3 fine grain 4 rich food 5 rich food 6 rich food 7 sons of rich families Fig. 35: The prismatic structure of expressions like  gao–liang

In a word, a generalized sense of a compound’s two constituents registered in dictionaries indicates that the general concept evoked by the two constituents is accepted by the language users and entrenched in language use to a great extent so that it can be chosen as a metonymic source providing access to the target. In addition, this criterion is in accordance with the source identification for chained metonymies (for a detailed explanation, see Section 3.1.1). Criterion (3)(ii): If the metonymic interpretation of the coordinate construction derives from the integration of the semantic features (e.g. attributes, actions) of both A and B, it belongs to Type 3. For many coordinate compounds with a form of near-synonymous Adj + Adj, the metonymic target of the compound as a whole inherits the integrated attributes implied in both constituents. For instance, ren–sheng in (29) metonymically refers to “a benevolent and wise person” instead of “a benevolent person and a wise person”; you–ai metonymically refers to “a beautiful young woman” instead of “a young woman and a beautiful woman”, you–yuan metonymically refers to “a person in a secluded and remote place” instead of “a person in a secluded place and a person in a remote place”. In other words, the attributes of both A and B are integrated in the metonymic interpretation of the compound as

Some criteria for classification | 83

a whole. The metonymic target cannot be reached by decomposing the literal meanings of the constituents. (29)

 ren–sheng [benevolent–wise] Benevolent and wise. Refers to a benevolent and wise person.  you–ai [young–beautiful] Young and beautiful. Refers to a young and beautiful woman.  you–yuan [secluded–remote] Secluded and remote. Refers to a person who lives in a secluded and remote place.

This case also includes some examples with a form of V + V or N + N, as in (30). The metonymic target of gong–xiao is not “a person in charge of supply and a person in charge of distribution” but “a person who is in charge of both supply and distribution”, and the first metonymic target of zhu–cui is not “an ornament made with pearls and an ornament made of jade” but “an ornament made with both pearls and jade”, which then may metonymically refer to “a woman in a rich dress”. (30)

 gong–xiao [supply–distribution] → a person who is in charge of supply and distribution  zhu–cui [pearls–jade] → an ornament made with pearls and jade → woman in rich dress

Criterion (3)(iii): If both A and B, as monomorphemic words, independently have the same or a similar metonymic target as the metonymic target of the compound as a whole, it belongs to Type 4. This criterion overrides the previous two. For example, the compound qi–ai (31) contains two near-synonymous constituents, qi and ai, both of which literally refer to the attribute of “old, aged”, and metonymically denote “old people”. Also, as independent monomorphemic words, both qi and ai have a metonymic meaning of “old people” registered in dictionaries. Thus, for this case, two ATTRIBUTE FOR PERSON metonymies are identified. (31)

AB  qi–ai [old–elderly] An elder, a teacher. Generally refers to old people.

84 | Metonymy in expressions

A  qi old teacher, elder B  ai old age, a respectful form of address to old people

If only one of the constituents independently refers to the same or a similar metonymic target as the compound as a whole, criteria (3)(i) and (3)(ii) will not be overridden. For example, although the second constituent of the compound shou–wei (32) independently has a similar target as the compound, this compound is classified as Type 3 according to criterion (3)(ii), because the target has both actions expressed by A and B. (32)

AB  shou–wei [defend–protect] To defend and protect. Or a person who is in charge of defending and protecting. B wei [protect] To protect and guard. Or a person who is in charge of defending and protecting.

Criterion (3)(iv): If AB shares the same literal and metonymic meanings with its reverted-word-order counterpart BA, AB belongs to Type 4. Chinese abounds in cases in which AB is semantically fully compositional and has the same meaning as its reverted-word-order counterpart BA. Strictly speaking, the coordinate construction AB in this case is at best a peripheral category of compounding and tends to be phrasal. Normally, the word order of compounds is invariable; the sequence of constituents is fixed. Violation of the expected sequence of words would result in nonsense words or a different meaning, as is the case for English sunrise, common sense, lazybones and so on (Kavka 2009: 28–29). Likewise, according to Chao (1968: 372), apart from a few exceptions, Chinese coordinate compounds cannot be reversed. Thus, the coordinate construction AB and its reverted-word-order counterpart BA are better understood as coordinate phrases. Semantic equivalency between AB and BA proves to some extent that the sequence between A and B is freer than in common compounds, and such constructions share some characteristics of phrases. For example,  bing–jia [weapon–armor] literally means “weapon and armor” and has a metonymic interpretation of “soldier”; its reverted-word-order counterpart  jia–bing [armor–weapon] shares the same literal and meto-

Some criteria for classification | 85

nymic interpretations. For both bing–jia and jia–bing, the two constituents are coordinate in nature and functionally contribute in the same way to the overall meaning of the expression. Because the semantic cohesion between A and B in this case is quite loose, as in a free combination of words, it is reasonable to analyze the metonymic processes of the two constituents in the compound separately. Take the examples in (33), for instance; two ATTRIBUTE FOR PERSON metonymies are identified in the expression gu–jiu, and the same two metonymies are also identified in its reverted-word-order counterpart jiu–gu. Similarly, two LOCATION FOR LOCATED metonymies are found in jia–shi and also in its reverted-word-order counterpart shi–jia. (33)

 gu–jiu [former–old] → old friend  jiu–gu [old–former] → old friend  jia–shi [home–room] → wife  shi–jia [room–home] → wife

2.3.2 Internal metonymy versus external metonymy Within attributive compounds (i.e. modifier-head compounds), it is important to distinguish internal metonymy from external metonymy. The former belongs to Type 2 and the latter to Type 3. An example of Type 2 is the aforementioned expression lao–ge [old–leather] (10), and an example of Type 3 is chang–xiu [long–sleeves] (12). The similarity between [old–leather] and [long–sleeves] is that both compounds have a modifier-head structure, that is, they are both attributive compounds. However, they differ in the following way. For the former, the first constituent modifies the metonymic target of the second constituent. The schematic representation of a case like [old–leather] is modifier–head[SOURCE→TARGET] in the sense that the metonymic process works in its internal configuration and affects only one component of the whole bimorphemic combination. The metonymy in this case is an internal metonymy. For the latter, the first constituent modifies the literal meaning of the second constituent and then the meaning of the compound as a whole derives from a metonymic process. The schematic representation of a case like [long– sleeves] is [modifier–head]SOURCE→TARGET in the sense that the metonymic process acts on the external configuration and affects the whole compound rather than one component. The metonymy here is an external metonymy. In other words, metonymy happens in the meaning extension process of the constituent parts for

86 | Metonymy in expressions

the case of [old–leather] and of the composite expression as a whole for the case of [long–sleeves]. A set of examples of Type 2 is given in (34), for which Figure 36 presents the metonymic process.  da–dang [big–ornament on a eunuch’s hat] → eunuch in power

(34)

 ju–dang [great–ornament on a eunuch’s hat] → great eunuch, eunuch in power zhong–dang [intermediate–ornament on a eunuch’s hat] → intermediate eunuch  nei–dang [internal–ornament on a eunuch’s hat] → eunuch (in the palace)  gui–dang [noble–ornament on a eunuch’s hat] → noble eunuch    





 

1 big–ornament on a eunuch’s hat 2 big 3 ornament on a eunuch’s hat 4 a eunuch in power 5 big (metaphorically, powerful) 6 a eunuch Fig. 36: The modified prismatic structure of examples like  da–dang [big–ornament on a eunuch’s hat]

The expressions in example (34) all belong to Type 2, i.e. modifier– head[SOURCE→TARGET], with internal metonymy operating on the second constituents of the compounds. The reason for treating them as Type 2 is threefold. First, independently,dang [ornament on a eunuch’s hat] also has a metonymic interpretation of “eunuch” in dictionaries. Second, the interpretations of the compounds as a whole (sense 1 in Figure 36) build on the literal meanings of both A and B (i.e. “different kinds of ornaments”), and are not registered in dictionaries. Third, even though da–dang, ju–dang and zhong–dang may be interpreted in such a way that the first constituents modify the literal meaning of the second (i.e. “ornaments of three different sizes”), no evidence has been found to suggest an association between eunuchs of different ranks and the different sizes of

Some criteria for classification | 87

ornaments on their hats in ancient China. Thus, the reasonable interpretation is: the first constituent serves as a modifier of the metonymic target of the second constituent. All the expressions in (34) share the same metonymic mapping of “ornament on a eunuch’s hat” for “eunuch”, and then the first constituent modifies and differentiates the ranks of eunuchs. These examples also show that dang is a very productive constituent for forming compounds. Examples of Type 3 is given in (35), and their prismatic structure is illustrated in Figure 37.  zheng–fang [principal–room] → wife

(35)

 pian–fang [side–room] → concubine  hou–fang [back–room] → concubine  zheng–shi [principal–chamber] → wife  ce–shi [lateral–chamber] → concubine  bie–shi [other–chamber (rooms other than the principal room)] → concubine  wai–shi [outside–chamber (external room)] → concubine  







1 principal–room 2 principal 3 room 4 wife Fig. 37: The modified prismatic structure of examples like  zheng–fang [principal–room]

In the context of ancient China, fang refers to “room behind the hall but off to the side, for sleeping and rest” and shi refers to “room right behind the hall, for sleeping and rest” (Yang 1987: 185–186). Both words independently have a metonymic interpretation of “wife” registered in dictionaries. One may argue that, like the set of examples with dang (34), the first constituent modifies the metonymic target of the second constituent: as in pian–fang [side–room], the second constituent fang [room] metonymically refers to “wife”, and then the first constituent pian [side] modifies “wife”, hence “side wife”, which metaphorically reaches the meaning of “concubine”. However, this argument is problematic. In

88 | Metonymy in expressions

fact, all the expressions in example (35) should be classified as Type 3, i.e. [modifier–head]SOURCE→TARGET, with external metonymy working on the whole literal meanings of the compounds (see Figure 37). On the one hand, the literal meaning of the compound as a whole (sense 1 in the prism) is registered in dictionaries: zheng–fang literally means “principal rooms (in a courtyard, usually facing south)”, bie–shi literally means “chambers other than the principal room”, hou– fang literally refers to “back rooms in the building, usually for concubines”. On the other hand, historical records provide proof that the wife normally lived in the principal room and concubines lived in the side rooms in old China (Chen 1937). As mentioned earlier, both fang and shi may refer to “wife” independently. At the same time, they are productive constituents in compounds. Thus, there are also many cases involving fang and shi as Type 2 (i.e. modifier– head[SOURCE→TARGET]), in which an internal metonymy works on the second constituent of the compound and the first constituent modifies the metonymic target of the second; see another set of examples in (36). For these cases, no literal interpretation of the modifier-head combination (i.e. sense 1 in Figure 38) is registered in dictionaries.  bi–fang [my (modest)–room] → my wife (modest denotation)

(36)

 shao–fang [young–room] → young wife, i.e. concubine  xian–shi [deceased–chamber] → deceased wife  han–shi [nagging–chamber] → nagging wife    





 

1 nagging–chamber 2 nagging 3 chamber 4 nagging wife 5 nagging 6 wife Fig. 38: The modified prismatic structure of examples like  han–shi [nagging–chamber]

To be aware of the difference between these two cases is important, because it helps us to know exactly on which lexical item (i.e. one constituent or the composite expression as a whole) the metonymy works. In the quantification of metonymic mappings, all the expressions in example (34) are actually generat-

Some criteria for classification | 89

ed from one specific metonymic expression, dang for “eunuch”, and in example (36), they are generated from two specific metonymic mappings, fang for “wife” and shi for “wife”. However, in the seven examples in (35), seven different specific metonymic mappings with the same metonymic pattern of LOCATION FOR LOCATED can be identified.

2.3.3 Differentiation between alternative prismatic structures As briefly mentioned in Section 2.1.4, the prismatic model allows alternative prismatic structures with different semantic extension orders for a complex word. A few points should be noted regarding the identification procedure with this particular issue. Most importantly, the annotation/definition in dictionaries may help to a large extent to decide on the most reasonable prismatic structure of an expression. Two compounds, dai–e (37) and xuan–wei (38), provide an illustration. (37)

dai–e [dark pigment–moth] → beautiful woman

(38)

 xuan–wei [tawny daylily–women’s quarters] → mother  

 

 



 













  







1 dark pigment–moth 2 dark pigment 3 moth 4 eyebrows drawn by dark pigment and antenna of moth 5 eyebrows drawn by dark pigment 6 antenna of moth 7 eyebrows drawn by dark pigment and with a shape like antennae of moths 8 eyebrows drawn by dark pigment 9 eyebrows like antennae of moths 10 beautiful women 11 beautiful women with eyebrows drawn by dark pigment 12 beautiful women with eyebrows resembling the shape of antennae of moths Fig. 39: The plausible prismatic structure of  dai–e [dark pigment–moth]



90 | Metonymy in expressions

For the compound dai–e, both of the constituents dai and e independently have the meaning of “eyebrows” registered in dictionaries. Dai, with a literal meaning of “a dark pigment used by women in ancient times to paint their eyebrows (similar to kohl)”, metonymically refers to “women’s eyebrows” or “beautiful woman”; e, with a literal meaning of “moth”, metonymically refers to “beautiful eyebrows”. Thus, it seems to be treated as a coordinate compound of Type 4 with two parallel local chained metonymic mappings, see Figure 39. However, this interpretation is problematic if we check the definition of dai–e in dictionaries. It is untenable to treat dai–e as a coordinate compound with both parts as heads metonymically referring to a “beautiful woman”.  dai–e 1. See dai–mei [dark pigment–eyebrows]. 2. Refers to a beautiful woman.

The original sense of dai–e registered in dictionaries is  dai–mei, which implies that dai in dai–e and dai in dai–mei share the same morphological function. In addition, dai–mei literally refers to “eyebrows drawn by a dark pigment, especially women’s eyebrows”, which indicates that dai has a modification function. Thus, dai–e should be treated as an attributive compound with B as the head and A as the modifier instead of a coordinate compound. It belongs to Type 6, and its metonymic mapping process should be reinterpreted as in Figure 40.  

 

   



 



1 dark pigment–moth 2 dark pigment 3 moth 4 dark pigment–antenna of moth 5 dark pigment 6 antenna of moth 7 dark pigment–eyebrows, i.e. eyebrows drawn by dark pigment 8 beautiful woman Fig. 40: The modified prismatic structure of dai–e [dark pigment–moth]

Some criteria for classification | 91

  

 

















1 tawny daylily-women’s quarters 2 tawny daylily 3 women’s quarters 4 mother/woman’s room 5 mother’s room 6 women’s quarters 7 mother 8 woman 9 mother Fig. 41: The modified prismatic structure of xuan–wei [tawny daylily–women’s quarters]

For the compound xuan–wei, both constituents, xuan and wei, independently have the meaning of “woman’s residence” in dictionaries. Xuan, with a literal meaning of “tawny daylily”, metonymically refers to “mother’s room”, and wei literally refers to “women’s quarters”. One possible interpretation of the metonymic mappings in this expression is represented in Figure 41, in which it is treated as Type 4. However, the definition of the compound xuan–wei in dictionaries is:  xuan–wei See xuan–tang [tawny daylily–hall]; refers to mother.

The original sense of xuan–wei is  xuan–tang, which literally means “a hall planted with tawny daylily” and in which xuan serves a modification function. Thus, xuan–wei is also a modifier-head attributive compound with a literal meaning of “women’s quarters with tawny daylily” instead of a coordinate compound. The justified metonymic process within this expression is therefore as follows:  xuan–wei: the literal meaning “women’s quarters with tawny daylily” serves as a source for the metonymic target “mother” (Type 3).

If the annotations/definitions from dictionaries fail to help us choose among the different possible prismatic structures of an expression, the rule of thumb is to use the prism that is the simplest, in the sense of having the minimum number of stages in the semantic extension and the fewest interacting cognitive mechanisms.

92 | Metonymy in expressions

2.4 Summary This chapter focuses on the role of metonymy in the meaning construction of expressions. I integrate Langacker’s (1987) view of grammatical composition with the cognitive-linguistic theory of semantic extension and meaning construction. In connection with the specific issue of examining metonymy in expressions, I have reviewed the practicability of three models for analyzing the semantic structure of figurative composite expressions in CL: Fauconnier and Turner’s (1998, 2002) blending theory, Benczes’s (2006) framework of English noun-noun creative compounds and Geeraerts’s (2002) prismatic model. I gave the most attention to the prismatic model, which has the advantage of being able to deal with metonymic transfers in composite expressions in different dimensions and, at the same time, in a more systematic way than the other models. The prismatic model was successfully applied to explain the various paths of metonymy operating on meaning construction of Chinese expressions, which are equivocal from a blending point of view. The analyses have shown that metonymy can be employed in a systematic way to create complex semantics in expressions. Besides the metonymic semantic transfer found in Chinese monomorphemic words, various paths of metonymy have been identified in Chinese complex words: 1) metonymy acting upon the first constituent; 2) metonymy acting upon the second constituent; 3) metonymy acting upon the complex word as a whole; 4) metonymy acting upon both constituents; 5) metonymy acting upon the first constituent of the complex word and then acting upon the complex word as a whole; and 6) metonymy acting upon the second constituent of the complex word and then acting upon the complex word as a whole. Nevertheless, using this type classification requires much caution. In practice, the classification is quite intricate. I have proposed a number of differentiation criteria for systematically identifying metonymy in Chinese expressions in the prismatic model. The classification method discussed in this chapter will serve as a guideline for metonymy identification and quantification in the following chapters, especially for Case 1 and Case 2 (all metonymic expressions in Case 3 are proper names; thus they all belong to Type 3).

|

Part Two: CASE STUDIES

Overview of the case studies

94 | Overview of the case studies

Part Two constitutes the main body of the book and reports on three case studies in turn. In Part One, I showed that previous research on metonymy has not been entirely satisfactory, and I identified several research gaps (see Section 1.5). These three case studies address some of the research gaps. The first case study (Case 1) is centered on cross-linguistic variation in metonymy. It investigates variation in metonymies for the target PERSON between Chinese and English and intends to draw attention to the culture-specific diversities in metonymy between two languages with different cultural backgrounds. Metonymic expressions for PERSON were collected manually from two Chinese metonymy dictionaries and the Historical Thesaurus of the OED. The prismatic model proposed by Geeraerts (2002) for analyzing figurative composite expressions is employed to identify metonymic mappings in the expressions. Then, a statistical significance test (i.e. Fisher’s exact test) is adopted to do the quantification and comparison. Significant cross-linguistic variation found in the dataset is considered against the cultural-social background to illuminate the important role of culture in the metonymic conceptualization of PERSON. The second case study (Case 2) deals with diachronic variation in metonymies for WOMAN in the history of Chinese language and culture. Potential metonymic expressions for WOMAN selected from the first case study are compared against authentic data from the Corpus of Historical Chinese (developed by Peking University) for their real distributions in different historical periods and genres. A metonymy identification procedure adapted from the Metaphor Identification Procedure (henceforth, MIP) (Pragglejaz Group 2007; Steen et al. 2010) is used to identify metonymies in the corpus data, and the prismatic model is also used to further identify metonymic mappings in composite metonymic expressions. Multivariate statistical techniques (i.e. multidimensional scaling, Poisson regression) are employed to investigate the diachronic and stylistic variation in metonymies in the corpus data. Significant diachronic changes emerging in the historical corpus and the differences between literary and nonliterary discourse are then explored in light of historical changes in culture and society to detect potential factors behind the changes in metonymic conceptualizations of WOMAN. The third case study (Case 3) focuses on lectal variation in metonymy. It investigates the factors that influence the alternation of metonymic versus literal designations for the concept GOVERNMENT in the Mainland Chinese and Taiwan Chinese. This study is based on a corpus built for the purpose from newspaper articles and online forum postings in the two language varieties. The MIP-based metonymy identification procedure as well as Markert and Nissim’s (2003, 2006) metonymy annotation schema was employed to identify metonym-

Overview of the case studies | 95

ic readings in context (see Zhang et al. 2011). The valid observations were tagged for a set of factors, after which a confirmatory statistical technique (i.e. mixedeffects logistic regression) was applied. The analysis aims to show that the observed alternation is not a question of free variation, but the result of a complex interplay of a number of conceptual, grammatical, discursive and lectal factors. Table 2 presents an outline of the case studies. With respect to the subjects investigated, the three case studies concern different dimensions of variation in metonymy, and each focuses on a specific target concept. For data resources, Case 1 uses dictionary data, while both Case 2 and Case 3 investigate metonymies in context based on corpus data. With regard to data analysis, in all the case studies, metonymies are identified based on reliable procedures and different quantitative techniques are applied to achieve various objectives. By showing the multidimensional onomasiological variation in metonymy through exemplary case studies, this book calls for a more dynamic view of conceptual metonymy. Metonymy cannot simply be reduced to an effect of the universal and biological species-specific human experience. The usage of metonymy is highly contextualized and displays diversity in various degrees, and the variation in metonymy is codetermined by a number of cultural, historical and social factors. Table 2: Outline of the case studies

Case studies

Case 1

Case 2

Case 3

Dimension of variation

cross-linguistic

diachronic (stylistic)

lectal (stylistic)

Target concept

PERSON

WOMAN

GOVERNMENT

Data resources

Chinese metonymy dictionaries, the HTOED

Corpus of Historical Chinese

Corpus of Mainland and Taiwan Chinese

Metonymy identification

prismatic model

prismatic model, adapted MIP

adapted MIP, Markert & Nissim’s annotation schema

Quantitative techniques

statistical significance exploratory statistics test (Fisher’s exact (multidimensional test) scaling), confirmatory statistics (Poisson regression)

confirmatory statistics (mixed-effects logistic regression)

3 The cross-linguistic perspective: Metonymies for PERSON As discussed in Chapter 1, scholars in CL agree that metonymy, like metaphor, is rooted in embodiment. Given the two types of embodiment, physiological and cultural (see Maalej 2004, 2007), some scholars argue that metaphor tends to be more culturally embodied, while metonymy tends to exhibit a physiologically oriented type of embodiment (Maalej and Yu 2011: 14). Because all human beings share a basic body structure and have many common bodily experiences (Yu 2008b: 387), physiological embodiment suggests the existence of universal conceptual metonymies across different languages. To prove that metonymy is a fundamental cognitive process in everyday life and to show that metonymy is derived from human life and bodily experience, CL has often emphasized the ubiquity and universality of metonymy (Radden and Kövecses 1999; Radden 2005). In line with this paradigm, universally valid metonymic patterns or sources have been suggested for some target concepts, such as ATTRIBUTE/CLOTHING/BODYPART for PERSON (Kleparski 1997; Yu 2003a). Meanwhile, culture-specific diversity, which has been widely recognized as relevant in metaphor research, is still largely understudied in metonymy research. This chapter reports on a panchronic study of metonymies for PERSON between two languages of different cultures, Chinese and English, from an onomasiological perspective. The specific research questions this case study intends to answer include:  To what extent are the metonymic patterns for PERSON in Chinese different from those in English?  In terms of a particular kind of PERSON, do the metonymic patterns for it behave in a similar way between the two languages?  In a certain pattern, do metonymic sources share a similar distribution between the two languages?  If variation is found, what are the possible driving forces behind it? By answering these questions, this study aims to extend metonymy research in CL; as a fundamental cognitive process, metonymy deserves a closer scrutiny from a cross-linguistic perspective. This chapter is organized as follows. Section 3.1 will introduce the methodology adopted in this study. Section 3.2 will report the quantitative and qualitative results of the comparison. The potential cultur-

98 | The comparative perspective: Metonymies for PERSON

al factors behind the variation will be discussed in Section 3.3. Finally, Section 3.4 will offer a summary of the chapter.

3.1 Methodology 3.1.1 The Chinese data resources The starting point for this study was to build a list of expressions that potentially might metonymically signify the concept PERSON. Two dictionaries of Chinese metonymies were used as resources: the Dictionary of Chinese Metonymic Senses (DCMS) (Han 1995) and the Dictionary of Chinese Substitutive Words (DCSW) (Zhang 1993). The entries in the two dictionaries are metonymic expressions collected from regular Chinese dictionaries like The Great Chinese Dictionary (Lou 1993), classical literature, annotated works and so forth; they thus contain expressions coined in different historical periods33. The DCMS and the DCSW provide valuable resources and play a nontrivial role in metonymy research in Chinese; however, their use still raises a few problems. First, although they are called dictionaries of metonymy, not all of the items registered involve a real metonymic process in a CL sense. Many metaphorical expressions are included and labeled as fitting the pattern PROPERTY FOR THING, such as  xian–zao (literally “fairy waterweed”, figuratively “snowflake”),   yi–ye (literally “one leaf”, figuratively “one boat”) and  shuang–gou (literally “double hooks”, figuratively “woman’s feet”) (Han 1995). However, similarity of a property to a thing is precisely the basis of metaphor in a cognitive account: for these examples, the metaphors are SNOWFLAKE IS WATERWEED, BOAT IS LEAF and FOOT IS HOOK, respectively. Second, the pattern LATER WORD FOR EARLIER WORD in the two dictionaries actually links to the Chinese rhetorical device of separation ( ge–lie is the Chinese term). Pervasive in pre-modern Chinese, separation refers to the phenomenon in which people sever an idiom or a proverb from the Confucian classics into parts and use one part to express the meaning of the other (Ji 2006). LATER WORD FOR EARLIER WORD is a case of using the later part of a phrase to stand for the meaning denoted by an earlier part of the same phrase34. For example,   er–li (literally “be independent”), which is the last part of a famous proverb

|| 33 Personal communication with the editors (March 2009). 34 Chen (1997) regards this rhetorical device as  cang–tou [elliptical beginning].

Methodology | 99

from the Confucian Analects:  san–shi–er–li [thirty years old be independent], can denote the meaning of the first part, “thirty years old”. However, from a CL point of view, whether the rationale behind this phenomenon is based on a metonymic process or not is still open to discussion. One may argue that LATER WORD FOR EARLIER WORD follows the process FORMB FOR FORMACONCEPTA, which is analogous with Kövecses and Radden’s (1999: 23) sign, reference and concept metonymies schema. In this example, the association between the forms may be motivated by a PREDICATE-OBJECT grammatical relation, which is heavily based on the link of linguistic forms. For some other cases, there may be no conceptual link between the two forms at all. The rationale of using FORMB for FORMA is simply their coexistence in one utterance of FORMA– FORMB. Strictly speaking, it is not a typical metonymy from the CL perspective, as there is no conceptual contiguity between the two concepts but only a formal co-occurrence. It is better to treat this specific device as a type of ellipsis, which is motivated by the transfer of content based on the contiguity link of linguistic forms (Ullmann 1962; Geeraerts 2010d: 65). Thus, expressions labeled LATER WORD FOR EARLIER WORD in the two Chinese metonymy dictionaries are excluded in the current study. Third, the pattern FORM FOR CONNOTATION in the DCMS seems to be similar to Kövecses and Radden’s (1999: 24) sign metonymies (FORM FOR CONCEPT), but in fact they are quite different. The CONNOTATION here is not the concept evoked by the linguistic form, but an allusion to a whole story or the allegorical or literarily symbolic meaning of an expression gained from a certain context. For example, the expression  san–dao [three–knife] involves a story from The Book of the Jin Dynasty: Wang Jun: Wang Jun had a dream in which he first saw three knives hanging on the house beam, and then one more knife appeared. The official Li Yi heard of the dream and came to congratulate Wang Jun, saying, “san–dao orthographically means zhou [state], and one more knife semantically means  yi [increase]. Thus, the dream actually implies  yi–zhou [the State of Yi]”. (The word yi is a polysemy that here refers to both the proper name of the state and the verb “increase”.) Later, as expected, Wang Jun got a promotion as an official in the feudal province of the State of Yi. Based on this story, the expression san dao gains a number of symbolic meanings such as feudal province, promotion and the State of Yi, which are all included as metonymies in both metonymy dictionaries. In addition, the DCSW contains sections on “symbolic substitution” and “allegorical substitution”. Such expressions are not included in the present study. Next, I will explain how I selected the entries wherein a metonymy for PERSON intervenes.

100 | The comparative perspective: Metonymies for PERSON

Metonymy identification In Chapter 2, I explained several paths of metonymy in expressions based on the prismatic model. In this section, I will illustrate at length how I identified metonymies with the target PERSON in the prismatic model. First of all, for each expression collected from the DCMS and the DCSW, I checked its etymology in the Great Chinese Dictionaries (Lou 1993), the Chinese Etymology Dictionary (Ci Yuan Editorial Board 1980) or the largest online Chinese dictionary, Han Dian (Chinese Dictionary, http://www.zdic.net/). These dictionaries served as supplements to the two metonymy dictionaries for the identification of metonymies for PERSON. There are cases of an expression with more than one literal meaning that can serve as the metonymic source for PERSON and cases of one literal meaning triggering polysemous metonymic interpretations of PERSON. The dictionaries provide detailed etymological information that can facilitate deciding which literal sense motivates the metonymic interpretation of PERSON. For example, the expression  cang–tou has two possible literal interpretations, “a kind of black headscarf” and “gray hair”. The former has two metonymic senses registered in dictionaries: “soldier with black headscarf” and “servants, normally wearing black headscarf in ancient China”. The latter motivates the metonymic meaning of “old people”. The annotations in the dictionaries point to the right source-target correspondences: HEADSCARF FOR SOLDIER (PIECE OF CLOTHING FOR PERSON), HEADSCARF FOR SERVANT (PIECE OF CLOTHING FOR PERSON) and HAIR FOR ELDER (BODYPART FOR WHOLE). It should be pointed out that, in some cases, an expression may even have more than one feasible literal sense that can serve as the source for the same metonymic meaning. The simpler metonymic mapping, in the sense of having less successive metonymic mappings or a more conventional metonymic mapping, is chosen in this study. For example, huang–wu (39) has a metonymic meaning of “emperor”. It has two meanings that may be the metonymic source of “emperor”: “imperial carriage” and “imperial palace”. Although both IMPERIAL CARRIAGE and IMPERIAL PALACE for EMPEROR are conventional in Chinese35, for the expression huang–wu, IMPERIAL PALACE was identified as the source. As a single metonymy (LOCATION FOR LOCATED), IMPERIAL PALACE needs less cognitive effort, while as an intermediate source, IMPERIAL CARRIAGE involves a chained

|| 35 These two sources are realized by many other metonymic expressions for “emperor”, e.g.   long–yu [dragon–carriage],  luan–jia [imperial carriage],  ci–chen [Zi Chen Palace],  zhong–jin [imperial residence], etc.

Methodology | 101

metonymy: “silk canopy” → “imperial carriage” → “emperor” (PART FOR WHOLE and POSSESSED FOR POSSESSOR). (39)

 huang–wu 1. The yellow silk canopy of an imperial carriage in ancient times. 2. Imperial carriages. 3. Imperial palace. 4. Imperial power and position. 5. Metonymically refers to an emperor.

Second, for chained metonymies36, the immediate source in the chain instead of the initial source was identified as the source for PERSON. For example, the compound qi–wan (40) (Type 3) involves a metonymic chain of MATERIAL FOR OBJECT and then PIECE OF CLOTHING FOR PERSON. Here CLOTHES instead of FABRIC MATERIAL is identified as the source for PERSON. (40)

 qi–wan 1. Rich fabric of silk. Also refers to clothes made of rich fabric of silk. 2. Refers to rich families or sons of rich families, playboys (derogatory).

This consideration also justifies the criterion (3)(i) for cases of a coordinate construction AB when A and B imply the same target concept by the same metonymic mapping (see Section 2.3.1): take the expressions gao–liang [fat meat– fine grain] and ge–mao [dagger axe–spear] (28) for examples (see Figure 35). Even if we accepted the GENERALIZATION process as a special kind of metonymy, i.e. successive metonymic mappings of “fat meat” or “fine grain” → “rich food” (GENERALIZATION) → “sons of rich families” (PATIENT FOR AGENT) and “dagger axe” or “spear” → “weapon” (GENERALIZATION) → “soldier” (POSSESSED FOR POSSESSOR), the immediate sources or the generalized senses “rich food” and “weapon” instead of the initial senses “fat meat/fine grain” or “dagger axe/spear” would be the real sources for PERSON according to the source identification principle for chained metonymies. Third, for the case of an expression having more than one metonymic mapping for PERSON in its semantic change, all of them are included in the dataset of || 36 Two constraints have been proposed for chained metonymies (Hilpert 2006: 130): 1) all intermediate steps have to be productive, i.e. the intermediate stage of a chained metonymy has to be an isolated metonymic mapping that can be realized by other linguistic expressions or the intermediate target registered as an entry in dictionaries. This constraint prevents freely assuming extinct intermediate steps; 2) each metonymic link must be motivated by a strong experiential basis (Grady 1997).

102 | The comparative perspective: Metonymies for PERSON

metonymies for PERSON. It is quite common that in one expression there are multiple metonymic mappings for PERSON, even with different specific targets; see nei–chong (41) and Figure 42. AB  nei–chong [inner–favor]

(41)

1. Person who is favored by the emperor. Refers to imperial concubines. 2. Person who is favored by the emperor. Refers to influential inner court officials. 3. Person who is favored by the emperor. Refers to toy-boys. 4. Being favored by the emperor. A nei [inner] Inside, internal, interior. Imperial palace, residence of an emperor.





   





 



1 imperial palace–favor 2 imperial palace 3 favor 4 The emperor favors 5 emperor 6 favor 7 imperial concubines/court officials/toy-boys Fig. 42: The prismatic structure of  nei–chong [inner–favor]

In nei–chong, the first constituent nei originally means “inside, internal, interior” and has a meaning of “imperial palace”, which in turn serves as a metonymic source for the first target “emperor” (LOCATION FOR LOCATED, 2 to 5 in the prism). Then, this metonymy triggers the overall literal meaning of the compound (sense 4 in the prism), i.e. “the emperor favors (subject-predicate combination)”. Finally, the action “the emperor favors” activates the second metonymic target “imperial concubines/court officials/toy-boys” by an ACTION FOR PATIENT metonymic mapping (4 to 7 in the prism). Therefore, two metonymic mappings for PERSON can be identified here: 2 to 5 and 4 to 7 in the prism.

Methodology | 103

Metonymy quantification All metonymies for the target PERSON were quantified subsequent to their identification. The metonymy quantification in this section has four facets: 1) the number of metonymic expressions37 for PERSON collected from the dictionaries; 2) the number of metonymic items for PERSON identified after the prismatic analysis of the expressions; 3) the number of unique metonymic items that involve a metonymy for PERSON; 4) the number of metonymic mappings for PERSON identified in the unique items. This four-faceted quantification was achieved by the following four steps. Step 1: In total, 1794 metonymic expressions in which metonymies for PERSON play a role in the semantic construction were collected from the DCMS and the DCSW. The dataset at this stage is called the “dictionary-based dataset”. Step 2: By analyzing the prismatic structures of expressions in the dictionary-based dataset, I identified in which part of the expression a metonymy for PERSON acts. In Section 2.2, I discussed the various pathways of metonymy in an expression. For a Type 4 expression, I divided it into two constituents and treated the constituents separately as two single lexical items. For instance, I separated the four Type 4 coordinate compounds in example (42) into eight constituents, jia (n = 4), bing (n = 2), dun (n = 1) and rong (n = 1). All of them have the same metonymic interpretation of “soldier”. (42)

 jia–dun [armor–shield];  jia–bing [armor–weapon];  bing– jia [weapon–armor];  rong–jia [weapon–armor]

A polysemous expression is treated as different expressions and counted more than once. For instance, cang–tou has two metonymic readings, “soldier” and “servant”, with the same source of “black headscarf”. Subscript letters were used to label the different metonymic readings of the expression, for example, cang–toua and cang–toub. There are also cases with successive metonymic mappings for PERSON in one expression. For example, the compound chong– tong literally means “double pupils in each eye”, which the emperor Shun was believed to have. Thus, chong–tong metonymically refers to “the emperor Shun”, and then by an INDIVIDUAL FOR COLLECTION metonymic mapping it further refers to “an emperor”. In these cases, subscript numbers indicate the different

|| 37 In this book, expression mainly refers to single words, compounds or complex words collected from the dictionaries; item refers to a component part or constituent of a compound/complex word. When the expression is a single word or Type 3 compound, expression and item are interchangeable.

104 | The comparative perspective: Metonymies for PERSON

stages of semantic shift in the chained metonymies: chong–tong1 and chong– tong2. A more complicated example is nei–chong (41). It has three metonymic readings and for each reading two metonymies for PERSON can be identified in the prismatic structure: the local metonymy on the constituent nei for EMPEROR and the global metonymy on the compound as a whole for each specific metonymic reading. The dataset at this stage is called the “metonymic item dataset”. Step 3: The unique metonymic items for PERSON were quantified after excluding duplicate metonymic items (the same item with the same source and target readings). A closer glance at the data in the metonymic item dataset from Step 2 suggests that some expressions actually share the same underlying metonymic component. Such duplication is due to two facts. First, quite a few Chinese words are very productive as constituents of compounds. For instance, the set of expressions in example (34) in Section 2.3.2 are all Type 2 with the same underlying metonymic constituent dang. Such duplication may distort the data analysis for the purpose of a cross-linguistic comparison of metonymy for PERSON by boosting the numbers of certain metonymic patterns falsely. In other words, a frequent use of a metonymy found in Chinese may not imply that it is a preferred metonymic mapping but merely that it is a productive item that can serve as a constituent in many compounds. Second, duplication might be caused by the fact that in the dictionarybased dataset I also included monomorphemic words, which could serve as constituents of compounds, too. For instance, the two monomorphemic expressions in example (43), with the target of “soldier”, are found in the dictionarybased dataset: bing (n = 1) and jia (n = 1). Both of them also serve as components in the metonymic expressions in (42). After Step 2, in total, we seem to have ten metonymic lexical items in (42) and (43) with repetition: jia (n = 5), bing (n = 3), dun (n = 1) and rong (n = 1). Only four unique metonymic items, however, were actually identified in these two examples after Step 3, with no repetitions counted: jia (n = 1), bing (n = 1), dun (n = 1) and rong (n = 1). The dataset at this stage is called the “unique metonymic item dataset”. (43)

 bing [weapon];  jia [armor]

Step 4: Based on the unique metonymic item dataset, the quantification of metonymic mappings for PERSON was carried out. Strictly speaking, there is no oneto-one correspondence between a metonymic item and a metonymic mapping. As mentioned in Section 2.3.1, we may identify both global and local metonymies in one prism. However, to avoid double counting the ostensible metony-

Methodology | 105

mies in the prism, for one metonymic item only one metonymic mapping was counted38. Two points need to be clarified. First, if the global metonymy is directly inherited from the local metonymy – in other words, if the local metonymy itself has an independent paradigmatic motivation – only the local metonymy is quantified. For instance, for the examples rou–zhen (7) and fei–xi (21) (see Figure 18 and Figure 32), the global metonymies on the top lines of the prisms are ignored in the quantification procedure.  

  





 





1 knife–writing brush 2 knife 3 writing brush 4 officials who draw up appeals, indictments, etc. 5 officials who draw up appeals, indictments etc. 6 officials who draw up appeals, indictments, etc. Fig. 43 The prismatic structure of  dao–bi [knife–writing brush]

Second, if the local metonymy is gained through a reinterpretation process from the global metonymy – in other words, the local metonymic shift itself seems to be inapplicable and unmotivated – only the global metonymy is counted. This case is quite rare in the Chinese examples, however. For instance, dao–bi [knife–writing brush] metonymically refers to “officials who draw up appeals, indictments, etc.” through a conventional global metonymy INSTRUMENT FOR AGENT39 (Figure 43). Although the first constituent dao [knife] maps onto “officials” (senses 2 to 5) and the second constituent bi [writing brush] maps onto || 38 Polysemous items and items involving chained metonymies are treated as different items distinguished by subscript letters and subscript numbers, as mentioned in Step 2. 39 According to the third differentiation criterion discussed in Section 2.3.1, dao–bi is best treated as a Type 3 case. The reasons correspond to the third criterion’s first three supplementary criteria: first, it has a generalized literal meaning registered in the dictionary, i.e. “writing implement in ancient China”; second, the metonymic reading of “officials” inherits the semantic features of both constituents, i.e. the official uses both knives and writing brushes as writing implements; third, neither dao nor bi independently has a metonymic reading of PERSON registered in dictionaries. In this case, then, only one INSTRUMENT FOR AGENT is tallied, which is also in accordance with the results of the quantification procedure illustrated in this section.

106 | The comparative perspective: Metonymies for PERSON

“officials”, too (senses 3 to 6), neither of the semantic shifts, from “knife” to “officials” and from “writing brush” to “officials”, is conventional in the meaning of dao or bi. Only when the top-down isomorphic process (i.e. the reinterpretation process) is accomplished can the local metonymies be identified in the prism. Put simply, I avoided double counting of global and local metonymies from one prism in the quantification procedure. The dataset at this stage is called the “metonymic mapping dataset”. Table 3 summarizes the different Chinese datasets gained by the metonymy quantification procedure discussed here. For each dataset, a few examples and the real frequency in each dataset are provided. It is the metonymic mapping dataset on which the metonymy analysis is based. Table 3: Different Chinese datasets

Dataset

Dictionary-based dataset

Content

metonymic exmetonymic items pressions collect- identified after the ed from the DCMS prismatic analysis and the DCSW

Examples

Type 0 bing jia

Metonymic item dataset

bing jia

Unique metonymic item dataset

Metonymic mapping dataset

unique metonymic items after excluding duplicates

metonymic mappings identified in unique metonymic items

bing jia

bing for SOLDIER jia for SOLDIER

Methodology | 107

Dataset

Dictionary-based dataset

Metonymic item dataset

Content

metonymic exmetonymic items pressions collect- identified after the ed from the DCMS prismatic analysis and the DCSW

Unique metonymic item dataset

Metonymic mapping dataset

unique metonymic items after excluding duplicates

metonymic mappings identified in unique metonymic items

Type 1 rou–zhen

rou

rou

rou for WOMAN

Type 2 da–dang ju–dang gui–dang zhong–dang

dang dang dang dang

dang dang dang dang

dang for EUNUCH

Type 3 dong–gong cang–tou zhong–tong

dong–gong cang–tou cang–toua cang–toub chong–tong1 chong–tong2

dong–gong cang–tou cang–toua cang–toub chong–tong1 chong–tong2

dong–gong for CROWN PRINCE

cang–tou for OLD PEOPLE

cang–toua for SERVANT

cang–toub for SOLDIER

chong–tong1 for

EMPEROR SHUN

chong–tong2 for EMPEROR

Type 4 jia–dun jia–bing bing–jia rong–jia

jia dun jia bing bing jia rong jia

jia dun jia bing bing jia rong jia

dun for SOLDIER rong for SOLDIER

Type 5 nei–chong

nei in nei–chonga nei in nei–chongb nei in nei–chongc nei–chonga nei–chongb nei–chongc

nei in nei–chonga nei in nei–chongb nei in nei–chongc nei–chonga nei–chongb nei–chongc

nei for EMPEROR nei–chonga for IMPERIAL CONCUBINES

nei–chongb for

COURT OFFICIALS

nei–chongc for

TOY-BOYS

Type 6 Qi–e

e Qi–e

e Qi–e

e for WOMAN Qi–e for SINGING GIRL

108 | The comparative perspective: Metonymies for PERSON

Dataset

Dictionary-based dataset

Content

metonymic exmetonymic items pressions collect- identified after the ed from the DCMS prismatic analysis and the DCSW

Frequency based on examples in the table

16

Frequency in 1727 my datasets

Metonymic item dataset

Unique metonymic item dataset

Metonymic mapping dataset

unique metonymic items after excluding duplicates

metonymic mappings identified in unique metonymic items

29

18

18

2039

1812

1812

Metonymy analysis One of the most important issues for the analysis of metonymic mappings for PERSON is how to precisely label the source and target concepts. As mentioned in Section 1.1.2, metonymic mappings are organized in hierarchical structures (Feyaerts 1999). The classification of metonymic patterns at the highest level presented here is consciously modeled after the inventory of metonymic patterns in Peirsman and Geeraerts’s (2006b) work and the metonymic annotations in the DCMS and the DCSW. In one pattern, however, there may be much variation in the specific source and target concepts. For instance, in the pattern LOCATION FOR LOCATED, we may have different source concepts like PALACE or HOME and various target concepts like EMPEROR or WIFE. Establishing an exhaustive classification of source/target concepts with a view to their hierarchical structures is an issue of much interest, but it is not the major task of the present study. An alternative consistent classification can be achieved in the following ways. First, classification can for the most part be done based on the linguistic and encyclopedic experience of native speakers. I consulted several native Chinese speakers, all in the field of linguistics, for their opinions on some difficult concepts. Second I resorted to the hyponymy structures from HowNet and EHowNet40, which show not only the semantic similarities of two concepts but

|| 40 HowNet (http://www.keenage.com/) and E-HowNet (http://ckip.iis.sinica.edu.tw/taxonomy/) are Chinese frame-based entity-relation models like the WordNet.

Methodology | 109

also the semantic differences between them. However, I have to concede that the practical application of HowNet and E-HowNet in the classification of Chinese concepts is very restricted. As tools for developing semantic and categorical representation models for natural language processing, neither of the two uses a knowledge system that covers Chinese concepts in great quantity. Also, many concepts in the DCMS and the DCSW are archaic and obsolete, and thus are not included in either HowNet or E-HowNet, which both use contemporary knowledge systems. Third, definitions from ordinary Chinese dictionaries provide supplementary materials for classification. A dictionary often defines a sense in the form of “…a kind of a hyperonymy…” or “(a) modifier(s) MOD a hyperonymy”. For instance,  qing–yi is defined in the dictionary as “ hei–se–MOD–yi– fu (black MOD clothes)”; thus, the superordinate concept of qing–yi is CLOTHES. Last but not least, annotations of the DCMS and the DCSW are helpful for classification as well. The two metonymy dictionaries list the metonymic pattern of each entry. For example,  huang–guan [yellow–cap] is registered in the pattern ACCESSORY FOR PEOPLE; therefore, it is reasonable to classify the source of huang–guan into the group ACCESSORY. Both source and target concepts were labeled on different levels. The “higher” mappings are elaborated by “intermediate” mappings, which in turn have “lower” mappings as initiations in the hierarchic structure. The target concept on the highest level is always PERSON and the source concept on the lowest level is always the specific linguistic item. Table 4 shows two examples with different levels for the source and the target. In total, 63 different kinds of PERSON (Target) and 20 kinds of source categories (SourceCate) were classified in the Chinese metonymic mapping dataset (see Appendix A). In practice, if a more specific concept can be identified from its superordinate concept, the more specific concept is taken into account for labeling of the source/target, and it is excluded in the quantification of its superordinate concept. For example, PIECE OF CLOTHING, VEHICLE, WEAPON and INSTRUMENT are taxonomically subordinate concepts of ARTIFACT. On Level 1 of the source category, the frequency of ARTIFACT (n = 57), however, does not cover its subordinate concepts, PIECE OF CLOTHING (n = 318), VEHICLE (n = 50), WEAPON (n = 20) and INSTRUMENT (n = 14).

110 | The comparative perspective: Metonymies for PERSON

Table 4: Examples with different levels in the hierarchy for the source and the target in  Hierarchy

Terminology

Example

Terminology

Example

Level 1

SourceCate

PIECE OF CLOTHING

TargetCate

PERSON

Level 2

Source

ACCESSORY

Target

SOLDIER

Level 3

SubSource

HAIRPIN

SubTarget

GUARD

Level 4

Item

 jin–chai [gold–hairpin]





3.1.2 The English data resource Metonymic expressions for PERSON in English were collected from the Historical Thesaurus of the Oxford English Dictionary (HTOED) for a comparison with the Chinese data. The HTOED41 is a taxonomic classification and reorganization of the majority of senses and lemmas in the OED, which contains lexical items grouped by concept from Old English to present-day English. Because the DCMS and the DCSW contain expressions from different historical periods, it is reasonable to use the HTOED, which is also a historical dictionary, to build the English metonymy equivalents. The classification of entries in the HTOED has a hierarchical structure. For instance, at a relatively high level, we have a category of THE LIVING WORLD, which is a subcategory of a broader division, THE EXTERNAL WORLD. Scrolling through THE LIVING WORLD, we find categories at the next level, including ANIMALS, PLANTS, PEOPLE and so on. The category PEOPLE in turn contains categories at the next level for NATION OR NATIONS, PERSON and so on down the hierarchy. On a lower level, PERSON then contains categories like CHILD, OLD PERSON and so on. OLD PERSON in turn contains subcategories like OLD MAN, OLD WOMAN and so on. The taxonomic nature of the HTOED makes it a good resource for an onomasiological study in the sense of colleting linguistic expressions for a certain concept (e.g. Stenroos 2002; Allan 2006, 2008; Łodej 2012), as, in this case study, for the concept PERSON.

|| 41 Detailed information about the HTOED is available at http://www.oed.com/public/htoed. The online search interface is at http://www.oed.com/thesaurus.

Methodology | 111

Metonymy identification For the cross-linguistic study on metonymies for PERSON between Chinese and English, the HTOED was used to build the English equivalents by an onomasiological approach, namely, following a target-to-source direction. The data collection includes two main steps: searching for target concepts and searching for metonymic expressions for that target.

Step 1: Searching for PERSON (sub)categories in the taxonomic structure of the HTOED By searching for the term person in the headings of the taxonomic structure of the HTOED, I created a list of possible categories denoting the concept PERSON or a particular kind of PERSON; see example (44) for a partial list of the search results. (44)

~the external world > the living world > people > person ~the external world > the living world > cleanness > dirtiness > dirty person > [noun] ~the external world > abstract properties > action or operation > behaviour or conduct > good behaviour > kindness > [noun] > person

(45)

~society > the community > kinship or relationship > kinsman or relation > parent > [noun] ~society > education > learning > learner > [noun] > one who studies ~the mind > having or possession > taking > stealing or theft > thief > [noun]

Searching for the term person in headings, however, inevitably misses some categories whose headings do not contain the term person but actually express the concept PERSON or a certain kind of person; see example (45). Thus, a second search method, the translation method, was developed. The Chinese linguistic realizations of the targets collected in the Chinese dataset were translated into English. I searched for all the translated expressions in both words and headings of the HTOED. For instance, in the Chinese dataset, I

112 | The comparative perspective: Metonymies for PERSON

identified metonymic expressions with targets like mei–nv, shi–nv, zai–xiang, ge–nv, etc. (see Table 5), and then I translated them into English42. Table 5: Words searched in the HTOED based on the translation of Chinese target concepts

Chinese target concept

English translation

Actual words searched in the HTOED

 mei–nv

beautiful woman beautiful girl beautiful lady

woman girl lady

 shi–nv

maidservant

servant

 zai–xiang

prime minister in feudal China

prime minister

 ge–nv near-synonym:  ge–ji

female singer female performer

singer performer near-synonym: entertainer

One point that should be emphasized is that when I actually search in the HTOED, I do not restrict the search to only one English translation for each Chinese target concept. Three aspects of the translation method of searching extended its coverage: 1) near-synonyms of the Chinese target (e.g. ge–ji, as a near-synonym of ge–nv “female singer”) were also checked in the ChineseEnglish dictionaries or translation tools to find their English equivalents (e.g. “female performer” in this case); 2) different Chinese-English dictionaries or translation tools may have different English equivalents for one Chinese word (e.g. for the Chinese word mei–nv, we may find beautiful girl, beautiful woman, beautiful lady as the English equivalents); 3) because many target concepts from Chinese are very culturally pertinent or archaic to some extent, their English equivalents are too specific to find a match category in the HTOED. For instance, zai–xiang translates as “prime minister in feudal China”, a concept not found in the taxonomic structure of the HTOED. It is more effective to search for a more general expression such as “prime minister” in the HTOED. More precisely, when I searched for the translated words, I ignored the modifiers or parts that

|| 42 For the translations, I consulted the Chinese-English Dictionary (Wu 2010) and two online Chinese-English dictionaries (http://www.iciba.com/, http://humanum.arts.cuhk.edu.hk/Lexis/Lindict/). For English synonyms, I consulted the Roget’s Synonyms at http://www.roget. org/.

Methodology | 113

play a role of semantic restriction. For instance, instead of searching for “beautiful woman” and “female singer”, I searched for “woman” and “singer”, which are on a higher level in the taxonomic structure and embrace the more specific expressions.

Step 2: Searching for metonymic expressions in each PERSON (sub)category When PERSON (sub)categories in the HTOED were identified, all the entries of each PERSON (sub)category were checked manually to select the expressions with a metonymy playing a role in the semantic shift to the PERSON reading. Three points need to be highlighted concerning the selection procedure. First, one entry in the HTOED may have different subsenses of the same general sense. If the subsenses are classified into different categories in the taxonomic structure, I treat the entry as polysemous, and therefore as different expressions. However, if the subsenses from one general sense all belong to the same category in the taxonomic structure, I treat the entry as one expression in my dataset. For example, the entry help (46) has the sense of “action of helping” (sense 1), which is identified as the literal sense. The metonymic sense of help is the third sense in the HTOED, which in turn has four subsenses (senses 3a, b, c and d). (46)

help 1. The action of helping; the supplementing of action or resources by what makes them more efficient; aid, assistance, succour. ... 3. A person, or company of persons, whose office it is to render help. a. gen. Assistant; adjutant. b. An ally; pl. allies, auxiliary troops. Obs. c. A person employed to give assistance in household or other manual work; in U.S., a hired labourer or servant, esp. a domestic servant. d. The labour of hired persons; collect. the body of servants belonging to a farm or household. orig. U.S.

The HTOED offers the option of tracing the taxonomic structure of an entry for each sense. The taxonomic structures for senses 3a, b, c and d of help are shown in (47). In the end, three helps were included in the English metonymy dataset: the first help is sense 3a of the category of A SUBSIDIARY HELP > PERSON, the second help

114 | The comparative perspective: Metonymies for PERSON

has sense 3b of the category of MILITARY FORCES, and the third help includes senses 3c and 3d of the category of SERVANT. Despite this systematic approach, the hierarchical nature of the categories often makes the different-or-same decision difficult. For the purpose of building an English equivalent to the Chinese metonymy dataset, I relied on the classification of the target concepts in the Chinese dataset. The SERVANT category was identified as a target concept in the Chinese dataset, with no further subcategories. Therefore, although according to the taxonomic structures in the HTOED, senses 3c and 3d of help refer to two different kinds of SERVANT, these two senses are treated as one metonymic item with the same target concept of SERVANT. (47)

The taxonomic structures for the subsenses of sense 3 of help: a. ~the external world > abstract properties > action or operation > easiness > aid, help, or assistance > [noun] > subsidiary or contributory help > a subsidiary help > person b. ~society > armed hostility > military forces > the Army > part of army by position > [noun] > second line c. ~society > authority > fact of being subject to authority > service > servant > types of servant > [noun] > hireling d. ~society > authority > fact of being subject to authority > service > servant > [noun] > servants collectively > of a family or household

Second, there are cases in which the HTOED offers different taxonomic structures for the same sense of one entry. It is logical that the compilers of the HTOED sometimes labeled a single sense (i.e. one concept) with different taxonomic structures to reflect the concept’s different highlighted attributes. However, it complicates the metonymy analysis from an onomasiological perspective. I may reach the same sense of an entry several times from different PERSON subcategories. For example, I reached the entry drum with the sense 3a in (48) from both the category “... > soldier >...> signaler or musician” (SOLDIER) and the category “... > instrumentalist >...> drummer” (MUSICIAN). The rule of thumb in such cases is to keep just one category, which, for the purpose of the crosslinguistic comparison, is the one that is more frequent in the Chinese dataset. In this example, the definition itself actually suggests that the SOLDIER taxonomic structure is more proper, because the label “Mil.” (i.e. “military”) implies that this sense of drum is military-related.

Methodology | 115

(48)

drum 3a. Mil. One who plays the drum; a drummer ~society > armed hostility > fighting man or warrior > soldier > soldier with special duty > [noun] > signaller or musician ~society > leisure > the arts > music > musician > instrumentalist > percussion player > [noun] > drummer

The reason to keep only one taxonomic structure is threefold. First, the different taxonomic structures provided by the HTOED are not based on essentially different senses but on the different perspectives from which the compilers classified the concept into the taxonomic structure. Second, the various taxonomic structures usually differ in very subtle ways. The compilers of the HTOED seem to intend to cover as many taxonomic structures for one sense as possible. Sometimes, the division between various taxonomic structures of one sense is so vague that it is hard to distinguish them from each other. Third, in the Chinese dataset, I only provided one taxonomic structure for one target concept. In principle, I could also offer multiple taxonomic structures for one target concept, for instance, the target jin–wei “emperor’s guard” could have been given different taxonomic structures as in (49). However, in practice, this would result in unnecessary complications. Therefore, for example, for jin–wei, I only coded SOLDIER, which is more intuitive, as a more general target concept of “imperial guard” in the Chinese dataset. Following the same idea, for the English data, only one taxonomic structure for an entry with the same sense was selected. (49)

 jin–wei [emperor’s guard] ~person > person who works in the capital and the imperial palace > imperial guard ~person > strong and brave man > imperial guard ~person > soldier > soldier who guards something > imperial guard

Third, I have ignored the presence of metonymy in words formed by adding an affix, such as the examples in (50). Although Janda (2011) demonstrates that the semantic relationships between stems, affixes and the words they form can be analyzed in terms of metonymy, she also maintains that word-formational metonymy is located in a different place than lexical metonymy on a grammaticallexical continuum (see Langacker 1987: 18–19). Some other scholars also point out that very little metonymy takes place in word-formation per se (e.g. Brdar and Brdar-Szabó 2014). In English, there are relatively more suffixal word-

116 | The comparative perspective: Metonymies for PERSON

formation metonymies (e.g. with -er, -ling, -ing, -ee, -y, -ie, -ess, -or, -ist, -ier, -ian, etc.) than in Chinese43, where affixation can only apply to a very limited set of words. The Chinese metonymy dataset includes monomorphemic and polymorphemic Chinese words but does not cover the complex words that are built on affixation, i.e. derived words. Furthermore, suffixal word-formation metonymies fundamentally differ from lexical metonymies in the sense that the linguistic form of the former (e.g. sixteener) has only a metonymic reading, while that of the latter (e.g. skirt) has both literal and metonymic readings. To make the English dataset comparable to the Chinese, I excluded suffixal word-formation metonymies and only focused on metonymy in “semantic change with formal identity” (Koch 2011: 299). (50)

shorty, sixteener, shoveling, littling, sexagenarian, cheatee, ancientry, offlicensee, actor

In these two steps, metonymic expressions for PERSON were collected from the HTOED. A seeming drawback of this method is that it prevents us from capturing English metonymic expressions that cover all the target concepts found in the Chinese dataset, as the English data collection relied greatly on the taxonomic classification of the HTOED. However, an exhaustive search of the HTOED for all the expressions for PERSON or all the target concepts found in Chinese is not the purpose of the current study. The main purpose is a crosslinguistic comparison of metonymy for PERSON in Chinese and English. Therefore, we only need a comparable English dataset that covers the main target concepts found in the Chinese dataset and that keeps the sample sizes approximately balanced between the two languages.

Metonymy quantification The same procedure of metonymy quantification illustrated in Section 3.1.1 was adopted for the English data. The quantification here also has the four facets. Table 6 summarizes the four English datasets with some examples, as well as the frequency in each dataset.

|| 43 In Chinese, there are also some suffixes that denote the category of PERSON (Li 2003). For example, the suffixes  jiang,  shi,  yuan and  tu express a person with different occupations (craftsman, expert, member and trainee, respectively); and the suffixes  zi,  shi,  fu,  sheng and  han all express a male.

Methodology | 117

The classification of the different types of metonymic expressions follows the same logic as that explained in Section 2.2.2. English expressions, however, involve different features due to the morphological differences between the two languages. The morphological structures in the English expressions are much clearer and simpler than in the Chinese ones. In the English dataset, no expressions of Type 1, Type 4 or Type 544 are found, and for Type 6, the compound play-white is the only case. Type 0 mainly includes single words and derived words like comely and disabled, and Types 2 and 3 are mostly compound words. It should be noted that the boundary between compounds and single words in English is fuzzy. This is not a problem, however, for the present study, because for an expression of Type 0 or Type 3, the metonymy always works on the expression as a whole. If there is a metonymy acting on the constituental level, i.e. a local metonymy, the expression is classified as Type 2 or Type 6 according to which part of the expression undergoes metonymic mapping. Table 6: Different English datasets

Dataset

Dictionarybased dataset

Metonymic item dataset

Unique metonymic item dataset

Metonymic mapping dataset

Content

metonymic expressions collected from the HTOED

metonymic items identified after the prismatic analysis

unique metonymic items after excluding duplicates

metonymic mappings identified in unique metonymic items

Examples

Type 0 prison finger smock sneak millenary45

prison finger smock sneak millenary1 millenary2

prison finger smock sneak millenary1 millenary2

prison for PRISONER finger for THIEF (PICKPOCKET) smock for WOMAN sneak for THIEF (WHO WORKS BY STEALTH) millenary1 for PERSON millenary2 for OFFICIAL

|| 44 A plausible Type 4 expression (coordinate structure) in English is senior-junior. However, its metonymic reading is “a person old and young at the same time”. Therefore, according to the criterion (3)(ii) in Section 2.3.1, senior-junior is classified into Type 3. 45 According to the OED, millenary literally means “a total sum of one thousand” and metonymically refers to “an officer in command of a thousand men”. Two successive metonymies can be found here: “one thousand” for “a thousand men” (QUANTITY FOR PERSON) and “a thousand men” for “an officer” (CONTROLLED FOR CONTROLLER).

118 | The comparative perspective: Metonymies for PERSON

Dataset

Dictionarybased dataset

Metonymic item dataset

Unique metonymic item dataset

Metonymic mapping dataset

Content

metonymic expressions collected from the HTOED

metonymic items identified after the prismatic analysis

unique metonymic items after excluding duplicates

metonymic mappings identified in unique metonymic items

Type 2 new poor home help house help field hand farm-hand share-hand46 area-sneak

poor help help hand hand hand sneak

poor help help hand hand hand sneak

Type 3 iron-face leather-head log-head pickpurse red coat whitecoat

iron-face leather-head log-head pickpurse red coat whitecoata whitecoatb

iron-face leather-head log-head pickpurse red coat whitecoata whitecoatb

white play-white

white play-white

Type 6 play-white47

poor for POOR PERSON help for SERVANT hand for FARMER

iron-face for STUBBORN PERSON

leather-head for STUPID PERSON

log-head for STUPID PERSON

pickpurse for THIEF red coat for SERVANT whitecoata for PHYSICIAN whitecoatb for SOLDIER white for WHITE PERSON play white for BLACK PERSON

Frequency based on examples in the table

19

Frequency in 1267 my datasets

22

18

18

1342

1329

1329

|| 46 According to the OED, share-hand metonymically refers to a farm-worker or tenant who raises crops on shares. 47 According to the OED, play-white metonymically refers (in South Africa in the apartheid era) to a black or colored person who has succeeded in being accepted as a white person.

Methodology | 119

Metonymy analysis To label the source and target concept of each metonymic mapping in the English metonymic mapping dataset consistently, the taxonomic structures for both the source and target readings of an entry in the HTOED are employed as references. The hyponymy structure from WordNet 3.148 serves as a supplement for the labeling. The source and the target were classified on different levels (see Table 4), as with the Chinese data. The labeling in the Chinese dataset also provides a guideline for the English data. For instance, the literal sense of dumpy “adj. short and stout; deficient in length or stature” has the taxonomic structure of “… > body > bodily height > shortness” in the HTOED. I modified such structures to fit three levels of category (in this case, CHARACTERISTIC > PHYSICAL CHARACTERISTIC > SIZE), for the purpose of making the classifications between the two languages more uniform. One important comment on this procedure is that, although we can find alternative taxonomic structures for a given concept, the main requirement for this study is to keep the categories in essence the same on the same levels between the two languages. For instance, PHYSICAL CHARACTERISTIC should always be on Level 2 in both Chinese and English datasets. In all, 61 different kinds of PERSON and 19 kinds of source categories were classified in the English metonymic mapping dataset (see Appendix A).

3.1.3 Summary of the methodology So far, I have introduced the methodology of the first case study. Figure 44 presents a flow chart of the data collection procedures. It was neither practical nor the purpose of the current study to collect an exhaustive dataset of all metonymies for PERSON or to build completely corresponding datasets with regard to the target concepts in the two languages. The data collection methods adopted in this study, however, prove to be efficient for collecting possible metonymic mappings for a target concept quickly and in numbers large enough to do statistical analysis.

|| 48 WordNet 3.1 is available at http://wordnet.princeton.edu/.

120 | The comparative perspective: Metonymies for PERSON





    

 



      

  

    

 

    

      

 

  

  

             

  

   

    

   



Fig. 44: Summary of the data collection procedures

Table 7: Frequency information in Fisher’s exact test

Cn

Eng

row.total

X

a

b

a+b

non-X

c

d

c+d

column.total

a+c

b+d

a+b+c+d (n)

The statistical analysis presented in the next section is based on the two metonymic mappings datasets (see Table 3 and Table 6); the abbreviation  stands for the Chinese metonymic mappings dataset and the abbreviation for the English metonymic mappings dataset. In total,  contains 1812 metonymic mappings and  contains 1329 metonymic mappings. Fisher’s exact test is employed to examine whether the proportion of a particular metonymic pattern (or source) is significantly higher or lower than expected in a certain language. It works well for 2*2 contingency tables and can handle small sample sizes (i.e. less than 5 observations). Table 7 provides the frequency information that is needed for the calculation. I represent the cells by the letters a, b, c and d, standing for the absolute frequency of each classification and name the totals across rows and columns marginal totals; n stands for the grand total. Of all the frequencies listed in the table, a, b, c and d are the inputs for Fisher’s exact test.

Results | 121 Methodology

The null hypothesis is that the left and right columns ( and ) are drawn from the same sample, i.e. that X behaves similarly in both  and . The pvalue returned by the test indicates whether the null hypothesis should be accepted or rejected, which in turn reveals whether X exhibits significant crosslinguistic variation or not.

3.2 Results This section presents the results of the data analysis. Generally, three main kinds of variation are found in the datasets, which reflect the different levels of schematicity in a metonymic mapping, namely, variation in patterns for the general target PERSON, variation in patterns for specific kinds of PERSON and variation in sources of a pattern. Sections 3.2.1 to 3.2.3 will illustrate each kind of variation in turn.

3.2.1 Variation in metonymic patterns for PERSON in general Table 8 shows the frequencies of metonymic patterns found in  and  as well as the p-values calculated by Fisher’s exact test for each pattern. The significant level α was determined to be 0.05 in this study. In fact, a strict and conservative measure, using the Bonferroni correction, was taken for the significance level when multiple comparisons were conducted49. In total, 16 patterns are found in  and 17 in . The patterns found only in one language will be discussed first and then the patterns with significantly higher proportions than expected in one language will be explained. For each pattern, a few instances will be presented by way of illustration.

|| 49 The Bonferroni correction (see Abdi 2007) is used to counteract the problem of multiple comparisons in a set of data and to avoid Type I errors. For example, if an experimenter is testing n hypotheses on a dataset and chooses α = 0.05 as the significant level for a series of tests, then the Bonferroni correction would be to test each individual result at a significance level of α / n, namely 0.05 / n.

122 | The comparative perspective: Metonymies for PERSON

Table 8: Metonymic patterns for PERSON in Cn and Eng

Pattern

Cn n

%

Eng n

p-value

%

INDIVIDUAL FOR COLLECTION

376

20.8

68

5.1

0.000*

PIECE OF CLOTHING FOR PERSON

318

17.5

102

7.7

0.000*

LOCATION FOR LOCATED

289

15.9

36

2.7

0.000*

CHARACTERISTIC FOR PERSON

238

13.1

516

38.8

0.000*

ACTION FOR AGENT

213

11.8

158

11.9

0.911

POSSESSED FOR POSSESSOR

151

8.3

92

6.9

0.156

BODYPART FOR WHOLE

0.000*

136

7.5

223

16.8

ACTION FOR PATIENT

27

1.5

19

1.4

1

PATIENT FOR AGENT

22

1.2

16

1.2

1

INSTRUMENT FOR AGENT

15

0.8

44

3.3

0.000*

CONTROLLED FOR CONTROLLER

11

0.6

2

0.2

0.053

PRODUCT FOR PRODUCER

7

0.4

26

2.0

0.000*

QUANTITY FOR PERSON

6

0.3

5

0.4

1 0.249

EVENT FOR AGENT

2

0.1

4

0.3

INSTRUMENT FOR PATIENT

1

0.1

NA

NA

1

0.1

1

0.1

STATE FOR PERSON

NA

NA

16

1.2

MANNER FOR AGENT

NA

NA

1

0.1

1812

100

1329

100

TIME FOR AGENT

total

1

Note: ɑ = 0.05, n = 15, Bonferroni correction: 0.05 / 15 = 0.003; * indicates a significant difference.

Patterns only found in one language There is one pattern found only in , has only one instance (51). (51)

INSTRUMENT FOR PATIENT, which,

however,

 tiao–zi A brief note people used to get a prostitute in the old days. → A prostitute.

Two patterns, STATE FOR PERSON (52) and MANNER FOR AGENT (53), are found in  but not in . STATE FOR PERSON has 16 instances and takes up 1.2% of all patterns. Only one instance is found under the pattern of MANNER FOR AGENT.

Results | 123

(52)

wedlock The condition of being married; marriage as a state of life or as an institution. → A wife.

(53)

scribble-scrabble In a scribbling manner. → A scribbler.

Patterns with significantly higher proportions than expected in Chinese There are three patterns significantly associated with Chinese: INDIVIDUAL FOR COLLECTION, PIECE OF CLOTHING FOR PERSON and LOCATION FOR LOCATED. In other words, there is significant difference in the proportions of these patterns between the two languages and the proportion in  is higher than expected by chance. INDIVIDUAL FOR COLLECTION is the most productive pattern for PERSON in  (54). It takes up 20.8% of all patterns and includes 376 instances. The top five targets of this pattern in  are OFFICIAL (n = 44), PERSON WITH SPECIFIC PROFESSION (n = 42), BEAUTIFUL PERSON (n = 40), RECLUSE (n = 25) and SOLDIER (n = 19). In , only 68 instances are found under this pattern (55) and the percentage drastically falls to 5.1%, with the top five targets NEGATIVE PERSON (n = 9), A WOMAN50 (n = 8), SERVANT (n = 8), BEAUTIFUL PERSON (n = 6) and RICH PERSON (n = 4). (54)

 Hong Niang The name of a maidservant in The Romance of the Western Chamber. → A matchmaker.  Xi Shi Xi Shi, a beautiful woman during the Spring and Autumn period. → A beautiful woman.

(55)

Nestor The name of a Homeric hero famous for his age and wisdom. → An old man, esp. a wise old man. Jane Female Christian name. → A woman/girl/girlfriend.

|| 50 The target A WOMAN in Chapter 3 includes those cases with a general reference of a female and without a specific indication of being “a wife”, “a beautiful woman”, “a servant girl”, etc.

124 | The comparative perspective: Metonymies for PERSON

PIECE OF CLOTHING FOR PERSON has 318 instances in total, accounting for 17.5% of all the patterns in . Example (56) represents a polysemous metonymic expression with three different targets. The main targets of this pattern in  are OFFICIAL (n = 67), SOLDIER (n = 35), BEAUTIFUL PERSON (n = 31), SCHOLAR (n = 26), RELIGIOUS PERSON (n = 18), COMMON PERSON (n = 16) and SERVANT (n = 16). This pattern has 102 instances in  with a percentage of 7.7%. The example whitecoat in (57) is also a polysemous word with two metonymic readings. The main targets under this pattern in  are SOLDIER (n = 29), PERSON WITH SPECIFIC PROFESSION (n = 12), A WOMAN (n = 6), UNCHASTE PERSON (n = 6), SCHOLAR (n = 6) and LOW PERSON (n = 6). (56)

 qing–shan a. Black clothes worn by scholars in ancient China. → A scholar. b. Clothes of officials in Tang Dynasty. → An official. c. Clothes of the main female character in Chinese operas. → A female character in Chinese operas.

(57)

whitecoat a. A white coat. → A doctor or hospital attendant who wears a white coat. b. A white coat. → A soldier wearing a white or light-colored coat.

LOCATION FOR LOCATED ranks as the third most popular pattern for PERSON in  (58). It contains 289 instances with a percentage of 15.9%. Instances of this pattern are identified with the following main targets: SOVEREIGN RULER (n = 32), ROYAL PERSON (n = 31), WIFE (n = 23), OFFICIAL (n = 23) and KINSMAN (n = 23). In , 31 instances of this pattern are found (59) and take up only 2.3%. The main targets of this pattern in  are OFFICIAL (n = 5), SOLDIER (n = 3) and INHABITANTS (n = 3). (58)

 hou–gong The imperial harem/palace. → Imperial concubines.  yu–zhang The military tent in which the chief general lives. → A chief general.

(59)

pri A building or other facility to which people are legally committed as punishment for a crime or while awaiting trial. → A prisoner. woop woop A jocular name for a remote rural town or district. → An inhabitant of such a place.

Results | 125

Patterns with significantly higher proportions than expected in English There are four patterns whose proportions in  are significantly higher than expected: CHARACTERISTIC FOR PERSON, BODYPART FOR WHOLE, INSTRUMENT FOR AGENT and PRODUCT FOR PRODUCER. CHARACTERISTIC FOR PERSON is the most dominant pattern for PERSON in  (60). It has 516 instances and reaches the percentage of 38.9%. The top five targets of this pattern in  are NEGATIVE PERSON (n = 85), UNWISE PERSON (n = 25), CHILD (n = 25), INFLUENTIAL PERSON (n = 24) and ETHNIC GROUP (n = 24). In , CHARACTERISTIC FOR PERSON (61) is the fourth most frequent pattern and has 244 instances with a percentage of 13.5%. It mostly refers to the targets BEAUTIFUL PERSON (n = 54), NEGATIVE PERSON (n = 39), POSITIVE PERSON (n = 14), OLD PERSON (n = 12) and SOVEREIGN RULER (n = 11). (60)

common Free to be used by everyone/public. → A common woman/prostitute. dizzy Foolish/stupid. → A foolish man/a fool.

(61)

 xiong–xie Evil. → An evil person.  yan–ye (Type 3) Pretty and coquettish. → A beautiful woman.

BODYPART FOR WHOLE is the second most popular pattern for PERSON in  (62). In total, 223 instances are of this pattern, with a percentage of 16.6%. The main targets of this pattern in  are UNWISE PERSON (n = 99), NEGATIVE PERSON (n = 46), ETHNIC GROUP (n = 15), UNCHASTE PERSON (n = 8) and INTELLECTUAL PERSON (n = 7). In , this pattern has 135 instances (63) with a percentage of 7.5% and it mainly denotes the targets of BEAUTIFUL PERSON (n = 25), OLD PERSON (n = 15), PERSON WITH SPECIFIC PROFESSION (n = 11), A PERSON51 (n = 10) and KINSMAN (n = 8). (62)

pig-head → A stupidly obstinate head or mind; (now chiefly) a stupidly obstinate person. pumpkin-head → A person with a head compared to a pumpkin; a fool/a dolt.

|| 51 The target A PERSON includes those cases with a general reference of a human being.

126 | The comparative perspective: Metonymies for PERSON

(63)

 chang–mei Slim eyebrows. → A beautiful woman.  dan–chun Red lips. → A young man.

INSTRUMENT FOR AGENT is found to be significantly associated with . It has 44 instances and takes up 3.3% of all patterns (64). Among them, 34 instances refer to the target of MUSICIAN and four instances denote the target of PERSON WITH SPECIFIC PROFESSION. Only 15 instances of this pattern are found in  (65), accounting for a very low percentage of 0.8%. The main target of this pattern in  is PERSON WITH SPECIFIC PROFESSION (n = 7). (64)

pen A writing tool. → The person who uses a pen; a writer/an author. saxophone A brass wind-instrument with a clarinet mouthpiece. → One who plays the saxophone.

(65)

 lao–qiang An old opium pipe. → A person who is addicted to opium smoking.  zhu–bi A chief pen. → An editor in chief.

The proportion of PRODUCT FOR PRODUCER (66) in  is significantly higher than expected too. It has 26 instances overall and accounts for 2% of all patterns. The main target of this pattern in  is PERSON WITH SPECIFIC PROFESSION (n = 12). Most instances involve chained metonymies, in which the expression first undergoes a metonymic mapping of CHARACTERISTIC FOR THING for an intermediate target of LITERARY AND ARTISTIC CREATION, upon which PRODUCT FOR PRODUCER then acts for the final target of “a person who creates or produces it” (e.g. satiric, poetic, comic, tragic, lyric, epic, etc.). PRODUCT FOR PRODUCER is not a common pattern for PERSON in . Only 7 instances are found in , amounting to 0.4% of all patterns. (66)

tragic Of, pertaining, or proper to tragedy as a branch of the drama; of the nature of tragedy; composing, or acting in, tragedy. → a. A tragic actor. b. A tragic poet or author.

Results | 127

(67)

 liang–gong Good bows. → A person who is good at making good bows.  Zi Xu Short for Zi Xu Fu (The Poetic Essay of Zi Xu). → The author Shima Xiangru.

Table 9 summarizes the quantitative analysis on patterns for the target category PERSON. Although patterns of INSTRUMENT FOR AGENT and PRODUCT FOR PRODUCER display statistically significant differences between and , these two patterns have extremely low proportions in  (< 1%), so that the statistical results with regard to them may not be good estimates. Briefly, we have seen in this section that onomasiologically, for the general target PERSON,  and  do not share similar distributions of several patterns; semasiologically, patterns have different target distributions in the two languages. Table 9: Summary of quantitative analyses on patterns for the target category PERSON

Target Cn category Patterns only found in Cn (n > 5)

Eng Patterns with higher proportion than expected in Cn and > 1% in both languages

Patterns only Patterns with higher profound in Eng portion than expected in (n > 5) Eng and > 1% in both languages

PERSON in NA

INDIVIDUAL FOR COLLECTION,

STATE FOR

CHARACTERISTIC FOR PERSON,

general

LOCATION FOR LOCATED,

PERSON

BODYPART FOR WHOLE

PIECE OF CLOTHING FOR PERSON

3.2.2 Variation in metonymic patterns for different kinds of person If we look at a subordinate level of the target category, it is no surprise to find strong associations between patterns and specific targets (i.e. different kinds of PERSON) in each language. For instance, in both languages, the patterns for NEGATIVE PERSON are not shared by SOLDIER. Thus, patterns which are significantly associated with a language might be different across various specific targets. In this section, I will report on the patterns significantly associated with Chinese or English for a number of specific targets.

128 | The comparative perspective: Metonymies for PERSON

A target whose relative frequency amounts to more than 1% in both languages is considered for cross-linguistic comparison. The reason to set a cut-off point is to ensure enough samples in both languages for a statistical analysis. In all, 15 targets were selected for the analysis; see Table 10. For each target, a series of Fisher’s exact tests were further conducted to find out patterns significantly associated with one language for the target. One target, OFFICIAL, will be exemplified. Table 10: Targets with a percentage of more than 1% in both languages

Target

OFFICIAL

Cn n

%

217

12.0

Eng n 23

% 1.7

BEAUTIFUL PERSON

163

9.0

21

1.6

PERSON WITH SPECIFIC PROFESSION

145

8.0

69

5.2

SOLDIER

117

6.5

126

9.5

64

3.5

22

1.7

SCHOLAR SERVANT

56

3.1

44

3.3

NEGATIVE PERSON

55

3.0

187

14.1

A WOMAN

47

2.6

25

1.9

CHILD

36

2.0

44

3.3

INTIMATE PERSON

34

1.9

33

2.5

OLD PERSON

34

1.9

19

1.4

POSITIVE PERSON

32

1.8

18

1.4

YOUNG PERSON

20

1.1

22

1.7

UNCHASTE PERSON

19

1.0

38

2.9

ETHNIC GROUP

18

1.0

46

3.5

Results | 129

An example target: OFFICIAL        

                

               









  



Fig. 45: The pattern distribution for OFFICIAL in Cn vs. Eng Table 11: Main patterns for OFFICIAL and the p-value of Fisher’s exact test

Pattern

Cn

> 5% in PIECE OF CLOTHING FOR PERSON either Cn or INDIVIDUAL FOR COLLECTION Eng POSSESSED FOR POSSESSOR

n

%

n

67

30.9

2

44

20.3

NA

NA

35

16.1

1

4.3

p-value

% 8.7

0.028 0.216

ACTION FOR AGENT

32

14.7

NA

NA

LOCATION FOR LOCATED

23

10.6

10

43.5

0.000* 0.000*

CHARACTERISTIC FOR PERSON

total

Eng

1

0.5

8

34.8

217

100

23

100

Note: ɑ = 0.05, n = 4, Bonferroni correction: 0.05 / 4 = 0.013; * indicates a significant difference.

130 | The comparative perspective: Metonymies for PERSON

In total, for the target OFFICIAL, 217 instances of twelve patterns are found in  and 23 instances of six patterns in . Figure 45 presents the percentage of each pattern found in  and . A cut-off point of 5% was set to distinguish the more common patterns from the rarer ones in each language. Patterns that occur with low frequencies have less room to display significant quantitative or qualitative variation. Hence, only patterns with a frequency of more than 5% in either language are considered for comparison. Table 11 shows the distributions of the main patterns for OFFICIAL in the two languages as well as the p-values of Fisher’s exact tests. The p-values of Fisher’s exact tests indicate that CHARACTERISTIC FOR PERSON is significantly associated with  for the target OFFICIAL (68). Only one instance of this pattern is found in  for OFFICIAL (69). The proportion of LOCATION FOR LOCATED in  is also significantly higher than expected (70). This pattern is found in as well, but with a much lower proportion (71). (68)

supernumerary Applied to an official, officer or employee not formally belonging to the regular body or staff, but associated with it to assist in case of need or emergency. → A supernumerary official or employee.

(69)

 zhong–gui [inner court–powerful] → A high official in the imperial court.

(70)

privy seal The office or position of Keeper of the Privy Seal. → The Keeper of the Privy Seal.

(71)

 Xi–fu West Mansion → The head of state security.

The patterns INDIVIDUAL FOR COLLECTION (72) and ACTION FOR AGENT (73) for the target of OFFICIAL do not exist in , but have large shares in . (72)

 Yi–Zhou Refers to Yi Yin in the Shang Dynasty and Zhou Gongdan in the West Zhou Dynasty. → A chancellor of the state.

(73)

 fu–bi To assist a ruler in governing a country. → A prime minister in feudal China.

Results | 131

Table 12: Summary of quantitative analyses on patterns for the fifteen targets in Table 10

Target

OFFICIAL

Cn

Eng

Pattern only found in Cn (n > 5)

Pattern with higher proportion than expected in Cn

INDIVIDUAL FOR

NA

Pattern only found in Eng (n > 5) NA

COLLECTION,

CHARACTERISTIC FOR PERSON,

ACTION FOR AGENT BEAUTIFUL

Pattern with higher proportion than expected in Eng

LOCATION FOR LOCATED

NA

NA

NA

NA

NA

INDIVIDUAL FOR

NA

CHARACTERISTIC FOR

PERSON PERSON WITH SPECIFIC PRO-

PERSON, PRODUCT FOR

COLLECTION

FESSION SOLDIER

PRODUCER INDIVIDUAL FOR

NA

NA

PERSON

COLLECTION SCHOLAR

LOCATION FOR

CHARACTERISTIC FOR

NA

NA

LOCATED

CHARACTERISTIC FOR PERSON

SERVANT

NA

NA

NA

NA

NEGATIVE PER-

NA

PIECE OF CLOTHING

BODYPART FOR

ACTION FOR AGENT

FOR PERSON, CHARAC-

WHOLE

SON

TERISTIC FOR PERSON A WOMAN

LOCATION FOR

NA

NA

LOCATED,

INDIVIDUAL FOR COLLECTION

POSSESSED FOR POSSESSOR CHILD

NA

PIECE OF CLOTHING

NA

FOR PERSON

CHARACTERISTIC FOR PERSON

INTIMATE PERSON NA

NA

NA

NA

OLD PERSON

BODYPART FOR WHOLE

NA

CHARACTERISTIC FOR

NA

PERSON POSITIVE PERSON

INDIVIDUAL FOR

NA

NA

NA

COLLECTION YOUNG PERSON

NA

NA

NA

NA

UNCHASTE

INDIVIDUAL FOR

LOCATION FOR LOCAT-

PIECE OF CLOTH-

NA

PERSON

COLLECTION

ED

ING FOR PERSON

ETHNIC GROUP

NA

PIECE OF CLOTHING

NA

FOR PERSON

CHARACTERISTIC FOR PERSON

132 | The comparative perspective: Metonymies for PERSON

I followed the same logic to test the other fourteen targets. Appendix B lists the pattern distributions for each target successively and the Fisher’s exact tests for the main patterns of each target. Table 12 summarizes the results of the quantitative analysis of the main patterns for the targets. For each target, I present the patterns that are significantly associated with one language and the patterns that are common in one language (n > 5) but do not occur in the other. For the targets BEAUTIFUL PERSON, SERVANT, INTIMATE PERSON and YOUNG PERSON, no significant cross-linguistic variation in patterns was found between Chinese and English. I will give examples for those targets with significant variation in the patterns listed in Table 12 next.

PERSON WITH SPECIFIC PROFESSION

Fisher’s exact tests show that the proportion of INDIVIDUAL FOR COLLECTION in  is significantly higher than expected (74). The patterns CHARACTERISTIC FOR PERSON (75) and PRODUCT FOR PRODUCER (66) are significantly associated with English. (74)

 Chang Sang Chang Sang, a famous doctor in the Warring States Period. → A famous doctor.

(75)

mercantile Of or relating to merchants or traders/their trade. → A merchant; a person engaged in trade or commerce.

SOLDIER

The pattern INDIVIDUAL FOR COLLECTION is only found in  and it has 19 instances (16.2%) (76). The pattern significantly associated with  for SOLDIER is CHARACTERISTIC FOR PERSON (77). (76)

 Huo Piaoyao Huo Qubing, a great soldier in the West Han Dynasty. → A military officer.

(77)

effective Of personnel in the armed forces: fit for work or service. → An effective soldier.

Results | 133

SCHOLAR

In , there are eight instances (6.8%) of LOCATION FOR LOCATED for the target SCHOLAR (78); however, no such pattern is found in . The statistical analysis indicates that the pattern CHARACTERISTIC FOR PERSON (79) is significantly associated with English for SCHOLAR. (78)

 bei–men Northern gate. Refers to the Hanlin Academy. → Scholars in the Hanlin Academy.

(79)

worthy Distinguished by good qualities; entitled to honour or respect on this account; estimable. → A prominent scholar or theologian.

NEGATIVE PERSON

There are two patterns significantly associated with  for this target: PIECE OF (80) and CHARACTERISTIC FOR PERSON (81). The pattern ACTION FOR AGENT (82) is significantly associated with  for NEGATIVE PERSON. BODYPART FOR WHOLE (83) is a common pattern for this target (n = 46) (24.6%) in , while it has no instance in . CLOTHING FOR PERSON

(80)

 wan–ku Silk trousers. → (Derogatory) sons of the rich, playboys.

(81)

 jian–ning Crafty and fawning. → A crafty and fawning person.

(82)

do-nothing → One who does nothing; an idler.

(83)

dry-fist → A niggardly or stingy person.

A WOMAN

Two patterns, LOCATION FOR LOCATED (n = 16, 34%) (84) and POSSESSED FOR POSSESSOR (n = 7, 14.8%) (85) are only found in . The proportion of INDIVIDUAL FOR COLLECTION (86) is significantly higher than expected in  for this target. (84)

 gui–fang A woman’s bedroom. → A woman.

(85)

 zhi–fen (Type 3) Blusher and powder. Women’s cosmetics. → A woman.

134 | The comparative perspective: Metonymies for PERSON

(86)

lubra An Australian Aboriginal woman. → A woman.

CHILD

The pattern PIECE OF CLOTHING FOR PERSON (87) is significantly associated with  and the pattern CHARACTERISTIC FOR PERSON (88) with  for the target CHILD. (87)

dan–mao A kind of hairstyle for a child. → A child.52

(88)

girly-girly Girly, esp. in an exaggerated or affected manner. → A girl/a little girl.

OLD PERSON

For the target OLD PERSON, the pattern BODYPART FOR WHOLE (89) is significantly associated with  and the pattern CHARACTERISTIC FOR PERSON (90) with . (89)

 tai–bei [mackerel–back] Metaphorically refers to an old man’s back, which resembles a mackerel. → An old person.

(90)

old Having lived or existed a long time; not young or new. → An old person.

POSITIVE PERSON

For the target POSITIVE PERSON, the pattern INDIVIDUAL FOR COLLECTION has 16 instances (50%) in , while no instance of this pattern was found in .

|| 52 HAIRSTYLE FOR PERSON is treated as a subpattern under PIECE OF CLOTHING FOR PERSON. In Chinese traditional culture, hairstyle always goes with clothing customs. Expressions denoting hair of a certain color, however, are classified into a subcategory of BODYPART under the pattern of BODYPART FOR WHOLE. For instance, expressions like  bai–fa [white–hair], which do not denote an artificial and aesthetic hairstyle but a body part distinguished from other body parts like  dan–chun [red–lips], are classified under the pattern of BODYPART FOR WHOLE, as is the English example blond.

Results | 135

(91)

 Tian Shu Tian Shu, a virtuous man in the Jin Dynasty. → A virtuous person.

UNCHASTE PERSON

Seven instances (36.8%) with the pattern INDIVIDUAL FOR COLLECTION (92) are found in  but not in  for the target UNCHASTE PERSON, while six instances (15.8%) with PIECE OF CLOTHING FOR PERSON are found in  (93) but not in . In addition, LOCATION FOR LOCATED is significantly associated with  for this target (94). (92)

 Xiao Xiao Du Xiaoxiao, a famous prostitute in the Southern Qi Dynasty. → A prostitute.

(93)

vizard-mask53 A mask worn to conceal or disguise the face; a domino. → A prostitute.

(94)

qing–lou A brothel. → A prostitute.

ETHNIC GROUP

The pattern PIECE OF CLOTHING FOR PERSON (95) is significantly associated with Chinese and the pattern CHARACTERISTIC FOR PERSON (96) with  for the target ETHNIC GROUP. (95)

 zhan–qiu Fur coats. → A northern minority nationality in China.

(96)

red Red, the hue of a person’s skin. → A North American Indian.

To sum up, I have found that patterns significantly associated with one language are quite different from one target to another. The finding confirms a strong association between the pattern and the target. Table 13 summarizes the main findings in Table 9 and Table 12. Each target listed in Table 13 is there for one of two possible reasons: first, the proportion of the pattern in one language || 53 ACCESSORY FOR PERSON is treated as a subpattern under PIECE OF CLOTHING FOR PERSON in this study.

136 | The comparative perspective: Metonymies for PERSON

is significantly higher than expected; second, the pattern is only found in one language and with an absolute frequency of more than five. In this study, either of the two possibilities indicates a significant association with a language. In addition, a significant association also embodies two levels: for the target category PERSON or for specific kinds of person. In the discussion of the potential cultural factors behind the variation in Section 3.3, metonymic patterns that have proportions of more than 1% in both languages and display significant cross-linguistic variation for the general target category PERSON will be interpreted, and then I will zoom in on those patterns which also display significant cross-linguistic variation for specific targets. In other words, the metonymic patterns INDIVIDUAL FOR COLLECTION, LOCATION FOR LOCATED and PIECE OF CLOTHING FOR PERSON, which demonstrate significant associations with Chinese, and the metonymic patterns STATE FOR PERSON, CHARACTERISTIC FOR PERSON and BODYPART FOR WHOLE, which are significantly associated with English, will be compared in terms of the cultural background for the cross-linguistic interpretation. Table 13: Summary of the findings in Sections 3.2.1 and 3.2.2

Pattern

Target for which the pattern is only found in Cn (n > 5) or more frequent than expected by chance in Cn

Target for which the pattern is only found in Eng (n > 5) or more frequent than expected by chance in Eng

INDIVIDUAL FOR

PERSON in general, OFFICIAL, SOLDIER,

A WOMAN

COLLECTION

POSITIVE PERSON, UNCHASTE PERSON, PERSON WITH SPECIFIC PROFESSION

LOCATION FOR

PERSON in general, SCHOLAR, A WOMAN,

LOCATED

UNCHASTE PERSON

OFFICIAL

PIECE OF CLOTHING

PERSON in general, NEGATIVE PERSON,

FOR PERSON

CHILD, ETHNIC GROUP

PRODUCT FOR

NA

PERSON WITH SPECIFIC PROFESSION

ACTION FOR AGENT

OFFICIAL

NEGATIVE PERSON

POSSESSED FOR

A WOMAN

NA

STATE FOR PERSON

NA

PERSON in general

BODYPART FOR

OLD PERSON

PERSON in general, NEGATIVE PERSON

NEGATIVE PERSON

PERSON in general, OFFICIAL, PERSON WITH

UNCHASTE PERSON

PRODUCER

POSSESSOR

WHOLE CHARACTERISTIC FOR PERSON

SPECIFIC PROFESSION, SOLDIER, SCHOLAR, CHILD, OLD PERSON, ETHNIC GROUP

Results | 137

3.2.3 Variation in metonymic sources of a particular pattern So far I have found that for the general target category PERSON and a few specific targets, Chinese and English have different preferences on several patterns. The link between Section 3.2.1 and 3.2.2 involves an increase in the specificity of the target concept in a metonymic mapping; in this section, we look at an increase in the specificity of the source concept in a given pattern. I would like to answer the question: for a particular pattern, do Chinese and English have a similar distribution of metonymic sources? For each pattern, I first conducted a quantitative analysis to find the sources significantly associated with one language, and then I qualitatively analyzed several targets whose frequencies in one of the languages are low for a quantitative comparison. The distinction between the target category PERSON and special targets was maintained when the source distributions were explored. Five patterns which display significant crosslinguistic variation in sources will be discussed in turn.

PIECE OF CLOTHING FOR PERSON

Three sources were coded for the pattern PIECE OF CLOTHING FOR PERSON: CLOTHES (e.g. PANTS, COATS, ROBES, SKIRTS and GOWNS)54, ACCESSORY (e.g. HAIRPINS, BRACELETS, EARRINGS) and HAIRSTYLE (e.g. ROUND HAIRSTYLE, HAIRSTYLE WITH THE SHAPE OF BRANCHES). The general distributions of three sources for the target category PERSON are presented in Table 14. Two sources, CLOTHES and HAIRSTYLE, display significant cross-linguistic variation. In this pattern, CLOTHES (97) in  shows a significantly higher proportion than expected, as does HAIRSTYLE in  (98). (97)

brigander Body-armour for foot-soldiers. → A soldier wearing a brigander.

(98)

 niu–bi–zi The hairstyle of a Taoist priest, which looks like a cow nose. → A Taoist priest.

|| 54 The variation on a more specific level of the hierarchy of sources, i.e. subsource, would show more language-specific features. It is impractical, however, for a cross-language quantitative comparison, because many of the subsources are very culture-pertinent, i.e. the particular subcategory of CLOTHES is unique in one language or even in one historical time period of a culture. For example, in , there are many instances with APRON or CLOAK as the source, while in ,  ru [an open cross-collar shirt in ancient China] is often chosen as the source.

138 | The comparative perspective: Metonymies for PERSON

Table 14: Metonymic sources for PERSON of PIECE OF CLOTHING FOR PERSON

Source

> 5% in either Cn or Eng

Cn

Eng

n

%

ACCESSORY

120

37.7

28

27.5

CLOTHES

164

51.6

71

69.6

0.001*

34

10.7

3

2.9

0.015*

318

100

102

100

HAIRSTYLE

total

n

p-value

% 0.074

Note: ɑ = 0.05, n = 3, Bonferroni correction: 0.05 / 3 = 0.017; * indicates a significant difference.

We should bear in mind that the specific targets may influence the choice of sources. Therefore, I also analyzed the source distributions for several main targets of the pattern PIECE OF CLOTHING FOR PERSON. Targets with absolute frequencies in both languages of more than five were chosen for cross-linguistic quantitative comparisons. For this pattern, five targets, SOLDIER, SCHOLAR, SERVANT, A WOMAN and PERSON WITH SPECIFIC PROFESSION were selected. Also, a cut-off point of 5% was set for distinguishing common sources from rarer ones. The target SERVANT will be illustrated as an example. For the target SERVANT, no source has a proportion in one language significantly higher than expected (see Table 15). However, in , HAIRSTYLE takes the first rank among all sources. In total, nine instances (56.2%) with HAIRSTYLE as the source are found in  for SERVANT (99), but no instance with the same source is found in . Table 15: Metonymic sources for SERVANT of PIECE OF CLOTHING FOR PERSON

Source

Cn

> 5% in either ACCESSORY Cn or Eng CLOTHES HAIRSTYLE

total

Eng n

n

%

2

12.5

1

20.0

1

5

31.2

4

80.0

0.1194

9

56.2

NA

NA

16

100

5

100

%

Note: ɑ = 0.05, n = 2, Bonferroni correction: 0.05 / 2 = 0.025; * indicates a significant difference.

 

p-value

Results | 139

(99)

 ya–huan Hairstyle resembling a limb of a tree. → A servant girl.  ping–tou The hairstyle of a servant in ancient China, without a cap or headscarf. → A servant.

Table 16: Summary of quantitative analyses on sources of PIECE OF CLOTHING FOR PERSON for the target category PERSON as well as main targets

Pattern

Target Source only Source with high- Source only Source with high(n > 5) in both found in Cn er proportion than found in Eng er proportion than Cn and Eng (n > 5) expected in Cn (n > 5) expected in Eng

PIECE OF

PERSON in

CLOTHING

general

NA

HAIRSTYLE

NA

CLOTHES

SERVANT

HAIRSTYLE

NA

NA

NA

SCHOLAR

NA

NA

NA

NA

PERSON WITH

NA

NA

NA

NA

A WOMAN

NA

NA

NA

NA

SOLDIER

NA

NA

NA

NA

FOR PERSON

SPECIFIC PROFESSION

Appendix C provides similar analyses for the other four targets of the pattern PIECE OF CLOTHING FOR PERSON. Fisher’s exact tests show that no significant difference is found in sources for the other four targets between the two languages. Table 16 summarizes the results of the quantitative analysis. We can see that the source HAIRSTYLE shows a strong association with Chinese in this pattern for both the target category PERSON and the specific target SERVANT. Three other targets of this pattern are quite interesting for qualitative analyses: OFFICIAL, BEAUTIFUL PERSON and CHILD. In , there are 44 instances (65.7%) with the source ACCESSORY and 23 instances (34.3%) with the source CLOTHES for the target OFFICIAL in PIECE OF CLOTHING FOR PERSON. For the source ACCESSORY, HEADWEAR (n = 36) and WAISTBAND (n = 5) are the two main subsources. The subsource HEADWEAR then mainly contains four typical types of male headwear in ancient China (100): FORMAL HEADDRESS ( guan or  mian) (n = 27), STIFF HATS

140 | The comparative perspective: Metonymies for PERSON

( mao) (n = 2), HAIRPINS ( zan) (n = 2) and SOFT CAPS ( jin) (n = 1). In , however, the two instances of PIECE OF CLOTHING FOR PERSON for the target OFFICIAL both choose CLOTHES as the source (101). (100)

 guan–mian A formal headdress in ancient China. → An official.  jin–gou The soft hat and unlined garment worn by an official in ancient China. → An official.

(101)

grey clak → an alderman who has ‘passed the Chair’. blue coat → A person who wears a blue coat. (Taxonomy: ~holder of office > public officials)

For the target BEAUTIFUL PERSON, there are 31 instances of PIECE OF CLOTHING FOR PERSON in . Among them, twelve (38.7%) instances are with the source ACCESSORY, ten (32.3%) with CLOTHES and nine (29.0%) with HAIRSTYLE. For the source ACCESSORY, six instances are with the subsource of HAIRPINS (e.g. chai,  zan and  ji), three with FACIAL ADORNMENT, two with JEWELRY and one with SCARF (102). The source HAIRSTYLE mainly includes instances denoting a typical female hairstyle in ancient China such as compounds with the term  huan “bun of hair, a circular hairstyle” (103). In , only one instance of this pattern is found and it is with the source CLOTHES (104). In addition, for both Chinese and English instances, the subtarget of this pattern is BEAUTIFUL WOMAN. (102)

 yu–chai A hairpin made of jade. → A beautiful woman.  hong–jin A red scarf. → A beautiful woman.  yue–huang A yellow adornment on women’s foreheads in ancient China. → A beautiful woman.

(103)

 qing–huan A black circular hairstyle. → A beautiful woman.  cui–huan A black circular hairstyle. → A beautiful woman.  yun–huan A circular hairstyle resembling clouds. → A beautiful woman.

(104)

a bit of skirt → A woman; esp. an attractive one.

Results | 141

For the target CHILD, in total there are 15 instances of this pattern in . All of them are with HAIRSTYLE as the source (87). Only two instances of this pattern for CHILD are found in , and both of them are with CLOTHES as the source (105). (105)

brat A cloth used as an over-garment. → A child. pinafore An apron. → The wearer of a pinafore; spec. a child.

POSSESSED FOR POSSESSOR

Table 17 lists the main sources for PERSON of POSSESSED FOR POSSESSOR. Fisher’s exact tests show that ABSTRACT ENTITY (106) and WEAPON (107) are significantly associated with , while ARTIFACT (108) and VEHICLE (109) are significantly associated with . The source OFFICIAL TITLE (n = 22) is only found in , and its metonymic target is mainly a SPECIFIC PERSON who held this official title once (110). (106)

supermind A mind of exceptional capacity or ability. → A person possessing such a mind.

(107)

rifle A type of gun/usually fired from shoulder level. → A marksman (esp. a hunter) armed with a rifle.

(108)

 yu–bing A jade handle. → A horsetail whisk with a jade handle. → A man of letters.

(109)

tong–nian The carriage of the crown prince. → The crown prince.

(110)

 zhong–lei An official title in the Western Han Dynasty. → Liu Xiang, who held this title.  zhu–shi An official title in the Zhou and Qin Dynasties. → Lao Tzu.

142 | The comparative perspective: Metonymies for PERSON

Table 17: Main sources for PERSON of POSSESSED FOR POSSESSOR

Source

Cn

> 5% in either abstract entity Cn or Eng artifact official title

n

%

10

6.6

48 22

Eng n

p-value

%

42

43.3

0.000*

31.8

8

8.2

0.000*

14.6

NA

NA

vehicle

50

33.1

4

4.1

0.000*

weapon

19

12.6

36

37.1

0.000*

151

100

97

100

total

Note: ɑ = 0.05, n = 4, Bonferroni correction: 0.05 / 4 = 0.0125; * indicates a significant difference.

Next, the targets with absolute frequencies of more than five in both languages were selected for cross-linguistic comparisons: OFFICIAL and SOLDIER. Overall, there are 35 instances of POSSESSED FOR POSSESSOR for OFFICIAL in  and one instance in  (see Table 18). ARTIFACT (n = 17) is the most dominant source in . Three subsources are found in ARTIFACT in : SEAL (n = 11), FLAG (n = 4) and TABLET (n = 2) (111). VEHICLE (n = 13) is the second most popular source in . It has three subsources: CARRIAGE (n = 9), HORSE (n = 3) and BIRD (n = 1) (112). Neither ARTIFACT nor VEHICLE is found in  as the source for OFFICIAL. (111)

 zheng–yin A foursquare official seal. → A county magistrate. mian–hu (Type 4)55 The formal headdress and the tablet of an official. → An official.  sun–qi A flag with falcon pattern. → An official.

(112)

xiong–shi A crossbar in the shape of a crouching bear (in the front of a carriage). → A carriage. → A procurator.  cong–ma A piebald horse, normally ridden by a censor. → A censor.

|| 55 Two parallel metonymic mappings are identified in this expression, i.e. PIECE OF CLOTHING FOR PERSON and POSSESSED FOR POSSESSOR.

Results | 143

 shuang–fu Two mallards. → A magistrate.56 Table 18: Main sources for OFFICIAL of POSSESSED FOR POSSESSOR

Source

> 5% in either Cn or Eng

Cn

Eng

n

%

ARTIFACT

17

48.6

NA

VEHICLE

13

37.1

NA

NA

WEAPON

3

8.6

1

100

35

100

1

100

total

n

p-value

% NA 0.111

For the target SOLDIER, the statistical tests show no significant difference between the two languages on the choice of sources (see Table 19). Table 20 summarizes the quantitative analysis on sources of POSSESSED FOR POSSESSOR for the general target category PERSON and main targets in this category. The statistical results demonstrate that the sources ARTIFACT and VEHICLE are not only dominant for PERSON in general, but also for the target OFFICIAL in . Table 19: Main sources for SOLDIER of POSSESSED FOR POSSESSOR

Source

Cn n

> 5% in either Cn or Eng total

%

Eng n

p-value

%

ARTIFACT

7

24.1

6

13.3

0.348

VEHICLE

6

20.7

3

6.7

0.141

WEAPON

15

51.7

34

75.6

0.045

29

100

45

100

Note: ɑ = 0.05, n = 3, Bonferroni correction: 0.05 / 3 = 0.017; * indicates a significant difference.

|| 56 The motivation for this metonymic mapping is that, according to a legend in The History of the Later Han, a magistrate at that time usually rode two mallards instead of horses.

144 | The comparative perspective: Metonymies for PERSON

Table 20: Summary of quantitative analyses on sources of POSSESSED FOR POSSESSOR for the target category PERSON as well as main targets

Pattern

Target Source only (n > 5) in found in Cn both Cn and (n > 5) Eng

Source with high- Source only Source with higher er proportion than found in Eng proportion than expected in Cn (n > 5) expected in Eng

POSSESSED

PERSON in

ARTIFACT, VEHICLE

FOR

general

OFFICIAL TITLE

NA

ABSTRACT ENTITY, WEAPON

POSSESSOR OFFICIAL

NA

ARTIFACT, VEHICLE

NA

NA

SOLDIER

NA

NA

NA

NA

Another target, SOVEREIGN RULER, whose absolute frequency in  (n = 2) is too low for a quantitative comparison, indicates a great preference for the source VEHICLE (n = 27) out of all sources (n = 29) in  (113). (113)

 long–yu,  yu–jia,  huang–yu An imperial carriage. → The emperor.

LOCATION FOR LOCATED

The main sources for the target category PERSON of LOCATION FOR LOCATED are listed in Table 21. The source COURT (n = 17) refers to the general place where the officials had an audience with the emperor and where the emperor dealt with state affairs in a feudal society. It includes expressions denoting the meaning of “court as a whole” or “a specific part of the court” (114). This source is only found in . Another source only found in  is HOUSE (n = 16), which generally refers to a building for human habitation, typically one that is a residence of a family (115). A source only found in  is RANK (n = 5) (116), which can be metaphorically conceptualized as LOCATION. The proportions of the source ORIENTATION (117) between the two languages are significantly different, and its proportion in  is significantly higher than expected. The proportion of the source PALACE in  is significantly higher than the proportion in ; see (118). The example qing–gong [blue–palace] in (118) also reflects the fact that the “existence” of a contiguity relation between the metonymic source and target depends on anthropological, social, cultural and other parameters of relevance (Koch 2011: 264). We know that in ancient China, the residence of the crown prince was the eastern palace, which provides a metonymic motivation for re-

Results | 145

ferring to the crown prince by his palace. The contiguity between BLUE and EAST, however, is not necessarily universal but typical of ancient Chinese Wu Xing (Five Elements) thinking, in which the color BLUE corresponds to the cardinal direction EAST (see Footnote 31). Table 21: Main sources for PERSON of LOCATION FOR LOCATED57

Source

Cn

> 5% in either COURT Cn or Eng ORIENTATION

n

%

17

5.9

Eng n

p-value

%

NA

NA

13

4.5

8

22.2

HOUSE

16

5.5

NA

NA

NATURAL PLACE

20

6.9

1

2.8

0.489

INSTITUTION

13

4.5

6

16.1

0.011

PALACE

45

15.6

1

2.8

0.040

PLACE NAME

16

5.5

2

5.6

1.000

POSITION

18

6.2

1

2.8

0.707

RANK

NA

NA

5

13.9

ROOM

61

21.1

2

5.6

289

100

36

100

total

0.000*

0.025

Note: ɑ = 0.05, n = 7, Bonferroni correction: 0.05 / 7 = 0.007; * indicates a significant difference.

(114)

 zhong–chao The imperial court. → An official in the imperial court.

(115)

 bai–wu Houses with roofs made of cogon grass. → A common person.

(116)

tip-top The very top; the highest point or part; the extreme summit. → People of the highest quality or rank.

|| 57 NATURAL PLACE includes concepts like mountain, woods, cave, marshland, etc. POSITION refers to the way in which a person is arranged, and expressions with this source are literally prepositional, like  che–you [on the right of a carriage] (metonymically for “warrior on the right of a carriage”) and  zuo–you [left–right] “at one’s side” (metonymically for “courtier; attendant”). PLACE NAME includes locative expressions as proper nouns, e.g. Mao Lin (a county name).

146 | The comparative perspective: Metonymies for PERSON

(117)

western → A native or inhabitant of the west. southern → A native or inhabitant of the southern states of the United States.

(118)

 qing–gong The blue palace. → The eastern palace. → The crown prince.  chu–gong The palace of the crown prince. → The crown prince.

Table 22: Main sources for OFFICIAL of LOCATION FOR LOCATED

Source

Cn

Eng

n

%

6

26.1

NA

NA

2

8.7

2

20.0

0.567

INSTITUTION

8

34.8

6

60.0

0.257

POSITION

2

8.7

NA

NA

23

100

10

100

> 5% in either COURT Cn or Eng ORIENTATION

total

n

p-value

%

Note: ɑ = 0.05, n = 2, Bonferroni correction: 0.05 / 2 = 0.025; * indicates a significant difference.

Table 23: Summary of quantitative analyses on sources of LOCATION FOR LOCATED for the target category PERSON as well as main targets

Pattern

Target (n > 5) in both Cn and Eng

Source only found in Cn (n > 5)

Source with higher proportion than expected in Cn

Source only found in Eng (n > 5)

Source with higher proportion than expected in Eng

LOCATION

PERSON in

COURT, HOUSE

NA

RANK

ORIENTATION

FOR

general COURT

NA

NA

NA

LOCATED OFFICIAL

Next, I zoomed in on specific targets whose frequencies in both languages are more than five for cross-linguistic comparison. In this pattern, only one target, OFFICIAL, was selected for quantitative analysis. Fisher’s exact tests show no significant difference in the sources between the two languages (Table 22). However, the source COURT for OFFICIAL is only found in  (114). Table 23 sum-

Results | 147

marizes the results of the quantitative analysis on sources for PERSON and specific targets of LOCATION FOR LOCATED between the two languages. SOVEREIGN RULER, ROYAL PERSON and several feminine targets, which are frequent in  but rare in , were selected for qualitative analyses. For the target SOVEREIGN RULER (n = 32 in , n = 2 in ), the first two preferred sources in  are PALACE (n = 12) (119) and COURT (n = 8) (120). For the target ROYAL PERSON (n = 31 in ), the dominant source is also PALACE (n = 26) (118). For targets of feminine concepts like WIFE, CONCUBINE, BEAUTIFUL WOMAN, MOTHER and A WOMAN (121), 51 instances choose the source ROOM out of all sources (n = 67) of this pattern. (119)

 yu–dian The imperial palace. → An emperor.

(120)

 chao–ting The imperial court. → An emperor.

(121)

 nei–she An inner room. → A wife.  bei–tang A northern principal room. → A mother. gui–wei An inner room. → A woman.  pian–fang A side room. → A concubine.

BODYPART FOR WHOLE

In total, nine sources were coded for the pattern BODYPART FOR WHOLE58. Table 24 lists the distributions of sources with percentages of more than 5% in either language, as well as the p-values of Fisher’s exact tests. The sources FACIALPART (63) and HAIR59 (122) are significantly associated with Chinese and HEAD/BRAIN (62) with English for the target category PERSON. The source GENITALS (123) has 17 instances (7.7%) in  but no instance in .

|| 58 The sources INNER ORGAN (n = 3 in ) and SOUL (n = 1 in ) have low proportions, and are excluded from the quantitative comparison. 59 See Footnote 52. The three English instances with HAIR as the source are: blond, strawberry blond and ash-blond.

148 | The comparative perspective: Metonymies for PERSON

(122)

 er–mao Gray hair. → An elder.

(123)

pussy The female genitals. → A woman/women collectively, regarded as a source of sexual intercourse.

Table 24: Main sources for PERSON of BODYPART FOR WHOLE

Source

Cn n

> 5% in either Cn or Eng

n

p-value

%

FACIALPART

63

46.7

33

15.0

0.000*

LIMBS

24

17.8

21

9.5

0.031

BONE/BLOOD/FLESH/SKIN

11

8.1

13

5.9

0.393

GENITALS

NA

NA

17

7.7

HAIR

17

12.6

3

1.3

0.000*

HEAD/BRAIN

13

9.6

124

56.4

0.000* 0.588

TORSO

total

%

Eng

7

5.2

8

3.6

135

100

223

100

Note: ɑ = 0.05, n = 6, Bonferroni correction: 0.05 / 6 = 0.0083; * indicates a significant difference.

There is no target whose absolute frequencies are more than five in both languages. Hence, quantitative tests for specific targets were not conducted. Next, three targets of this pattern, BEAUTIFUL PERSON, NEGATIVE PERSON and UNWISE PERSON will be discussed qualitatively. The target BEAUTIFUL PERSON has 28 instances in , and the main source is FACIALPART (n = 24). Of the source FACIALPART, 16 instances are with the subsource EYEBROWS, four instances with FACE and two instances with EYES (124). Only one instance of BODYPART FOR WHOLE for this target is found in  (125), and its source is GENITALS. In Section 3.2.1, I mentioned that the targets UNWISE PERSON (n = 99) and NEGATIVE PERSON (n = 46) are the two most dominant targets of the pattern BODYPART FOR WHOLE in . In regard to the sources of these two targets, both of them have HEAD/BRAIN as the first choice, with 91 instances of HEAD/BRAIN for UNWISE PERSON (62) and 20 instances of HEAD/BRAIN for NEGATIVE PERSON (126). For NEGATIVE PERSON, the second dominant source is FACIALPART (n = 14) (127).

Results | 149

(124)

e–mei Women’s delicate eyebrows.→ A beautiful woman.  xiao–bai–lian A little white face → A beautiful male youngster.  ming–mou Bright eyes. → A beautiful woman.

(125)

beaver The female genitals or the pubic area in general. → A girl or woman/esp. one who is sexually attractive.

(126)

big-head → A conceited or arrogant person.

(127)

brass-face → An impudent person.

hot-brain → A impetuous/rash/reckless person. sharp-nose → A mean person.

CHARACTERISTIC FOR PERSON

Three sources were coded for the pattern CHARACTERISTIC FOR PERSON: AGE, NONand PHYSICAL CHARACTERISTIC. Table 25 presents the distributions of sources between the two languages. Fisher’s exact tests indicate that for the general target PERSON, NONPHYSICAL CHARACTERISTIC (e.g. MORAL TRAIT, PERSONALITY TRAIT) is significantly associated with  (60), while PHYSICAL CHARACTERISTIC (e.g. PHYSICAL AESTHETICS, SIZE) is significantly associated with  (128). PHYSICAL CHARACTERISTIC

Table 25: Main sources for PERSON of CHARACTERISTIC FOR PERSON

Source

Cn

> 5% in either Cn or Eng

CHARACTERISTIC (AGE) NONPHYSICAL CHARACTERISTIC PHYSICAL CHARACTERISTIC

total

Eng n

n

%

14

5.9

27

139

58.4

408

79.1

0.000*

85

35.7

81

15.7

0.000*

238

100

516

100

% 5.2

Note: ɑ = 0.05, n = 3, Bonferroni correction: 0.05 / 3 = 0.017; * indicates a significant difference.

(128)

p-value

 jiao–li (Type 3) Bright and beautiful. → A beautiful woman.

0.731

150 | The comparative perspective: Metonymies for PERSON

shan–chi Three Chinese feet. → A child.

Then, the source distributions for a few main targets (n > 5 in both languages) were compared between the two languages (see Appendix C). Take the target BEAUTIFUL PERSON as an example. Fisher’s exact tests (see Table 26) show that for this target the proportion of NONPHYSICAL CHARACTERISTIC (129) is significantly higher in  and the proportion of PHYSICAL CHARACTERISTIC (130) is significantly higher in .



(129)

lovesome Worthy of love; having qualities that inspire love; lovable. → A lovely or beautiful person.

(130)

 jiao–zi Beautiful looks and postures. → A beautiful woman.

Table 26: Main sources for BEAUTIFUL PERSON of CHARACTERISTIC FOR PERSON

Source

> 5% in either Cn or Eng total

Cn

NONPHYSICAL CHARACTERISTIC PHYSICAL CHARACTERISTIC

n

%

2

3.7

52 54

Eng n

p-value

%

5

38.5

0.002*

96.3

8

61.5

0.002*

100

13

100

Note: ɑ = 0.05, n = 2, Bonferroni correction: 0.05/2 = 0.025; * indicates a significant difference.

Table 27 presents the results of quantitative analysis on sources for PERSON and main targets of CHARACTERISTIC FOR PERSON between the two languages. It shows that the sources PHYSICAL CHARACTERISTIC and NONPHYSICAL CHARACTERISTIC have significantly different distributions between  and  for the target category PERSON as well as for the target BEAUTIFUL PERSON. The source PHYSICAL CHARACTERISTIC is preferred in , while NONPHYSICAL CHARACTERISTIC is preferred in . However, for other targets, no significant difference is found in the sources between the two languages.

Results | 151

Table 27: Summary of quantitative analyses on sources of CHARACTERISTIC FOR PERSON for the target category PERSON as well as main targets

Pattern

Target (n > 5) in both Cn and Eng

Source only found in Cn (n > 5)

Source with high- Source only Source with higher er proportion than found in Eng proportion than expected in Cn (n > 5) expected in Eng

CHARAC-

PERSON in

NA

PHYSICAL CHARACTER-

TERISTIC

general

NA

ISTIC

NONPHYSICAL CHARACTERISTIC

FOR PERSON

BEAUTIFUL

NA

PERSON

PHYSICAL CHARACTER-

NA

ISTIC

NONPHYSICAL CHARACTERISTIC

NA

NA

NA

NA

INTIMATE PERSON NA

NA

NA

NA

KINSMAN

NA

NA

NA

NA

LOW PERSON

NA

NA

NA

NA

NEGATIVE PER-

NA

NA

NA

NA

OLD PERSON

NA

NA

NA

NA

POOR PERSON

NA

NA

NA

NA

POSITIVE PERSON

NA

NA

NA

NA

SERVANT

NA

NA

NA

NA

YOUNG PERSON

NA

NA

NA

NA

A WOMAN

NA

NA

NA

NA

CHILD

SON

The main findings of the quantitative analyses in Section 3.2.3 on the variation in metonymic sources of five particular patterns are summarized in Table 28.

152 | The comparative perspective: Metonymies for PERSON

Table 28: Summary of quantitative analyses in Section 3.2.3

Target

Pattern

Source only found in Cn (n > 5)

Source with higher proportion than expected in Cn

Source only found in Eng (n > 5)

Source with higher proportion than expected in Eng

PERSON

PIECE OF CLOTHING

NA

HAIRSTYLE

NA

CLOTHES

in general

FOR PERSON OFFICIAL TITLE

VEHICLE, ARTIFACT

NA

POSSESSED FOR POSSESSOR LOCATION FOR

ABSTRACT ENTITY, WEAPON

COURT, HOUSE

RANK

ORIENTATION

LOCATED BODYPART FOR

NA

FACIALPART, HAIR

GENITALS

HEAD/BRAIN

NA

PHYSICAL CHARAC-

NA

NONPHYSICAL

WHOLE CHARACTERISTIC FOR PERSON SERVANT

PIECE OF CLOTHING

TERISTIC

CHARACTERISTIC

HAIRSTYLE

NA

NA

NA

NA

VEHICLE, ARTIFACT

NA

NA

COURT

NA

NA

NA

NA

PHYSICAL CHARAC-

NA

NONPHYSICAL

FOR PERSON OFFICIAL

POSSESSED FOR POSSESSOR LOCATION FOR LOCATED

BEAUTIFUL CHARACTERISTIC PERSON

FOR PERSON

TERISTIC

CHARACTERISTIC

3.3 Discussion So far, I have reported on the main cross-linguistic variation in metonymies for PERSON between the two language datasets. In this section, the possible underlying forces behind the variation will be discussed for the purpose of showing the influence of cultural elements on metonymies. Many scholars agree that the choice of metonymic pattern/source is not free but appears to be motivated or restrained by a number of principles, which in turn reflects the function of metonymy. For instance, Langacker writes (1993: 30), “Metonymy allows an efficient reconciliation of two conflicting factors: the need to be accurate, i.e., of being sure that the addressee’s attention is directed

Discussion | 153

to the indented target, and our natural inclination to think and talk explicitly about those entities that have the greatest cognitive salience for us”. Following Langacker’s claim, Radden and Kövecses (1999) have discussed in detail the cognitive and communicative principles that promote the most conventionalized or most “natural” metonymic patterns (in their term, vehicleto-target routes). The principles proposed by Radden and Kövecses (1999: 52) explain well the preferred pattern/source for a target and help us “understand why we select certain sources to access a target and why certain patterns have become conventionalized in the language”. For example, in both Chinese and English, we find that the pattern INDIVIDUAL FOR COLLECTION is frequent (> 5%). The CONCRETE OVER ABSTRACT principle proposed by Radden and Kövecses (1999: 45) could explain the reason for the preference of this pattern in both languages: the individual member (the paragon) is more concrete than the abstract category. In specific cases of this pattern, the IDEAL OVER NONIDEAL and TYPICAL OVER NONTYPICAL principles (Radden and Kövecses 1999: 48–49) further constrain the choice of source: we use an individual (an ideal/typical one or a paragon) in the culture to stand for the whole category. Generally, the principles provided by Radden and Kövecses could govern the default selection of the preferred metonymic source for the target, PERSON in this case; they serve as universal determining factors for the cross-cultural similarity of metonymic patterns/sources. However, as illustrated in Section 3.2, much cross-linguistic variation is found in metonymic patterns/sources for the same target. The variation found here is not baseless but influenced by a number of cultural elements. Section 3.3.1 will briefly discuss the possible reasons for the preference of INDIVIDUAL FOR COLLECTION in Chinese. Section 3.3.2 will introduce the carriageclothing system in Chinese culture and its influence on the variation in metonymy. Section 3.3.3 will focus on the residence system in Chinese culture and its role on the metonymic conceptualization of PERSON. Section 3.3.4 will inspect the variation relevant to body-part metonymies, followed by a discussion on cultural factors behind CHARACTERISTIC FOR PERSON/STATE FOR PERSON in Section 3.3.5. Finally, Section 3.3.6 will provide a summary of the discussion.

3.3.1 Paragons in Chinese culture It is in accordance with expectations that INDIVIDUAL FOR COLLECTION is significantly associated with Chinese, since Chinese culture has placed great emphasis on cultural icons or prototypes since ancient times. This cultural preference is rooted in the long tradition of Chinese biography and biographical literature

154 | The comparative perspective: Metonymies for PERSON

(Chen 1953). The portion of biography in Chinese historical writing is so large that, for example, in actual pages, 60% of the whole History of the Ming Dynasty is biographical in nature. Historical biographies mainly focus on the individual lives of officials, imperial concubines, writers, hermits, virtuous wives and filial sons (Nivison 1962: 457). The individual is typically described as a social stereotype in the biography and might then serve as a central point of reference representing the whole category. Table 13 shows that concepts which are more prone to be represented by a paragon in  are normally concepts with positive social values, like OFFICIAL, SOLDIER, POSITIVE PERSON and PERSON WITH SPECIFIC PROFESSION (esp. PHYSICIAN, WRITER, PAINTER, etc.). A cursory glance at the data suggests that for the target UNCHASTE PERSON, which seems to be a negative concept, the pattern INDIVIDUAL FOR COLLECTION is also significantly more frequent than expected in . However, a close scrutiny reveals that expressions for UNCHASTE PERSON of this pattern actually highlight the goodliness or the literary/artistic talent of the person and mask her unchastity; see (92) and (131). Semasiologically, the main targets of INDIVIDUAL FOR COLLECTION in  are mainly positive, while the most dominant target of this pattern in  is NEGATIVE PERSON (see Section 3.2.1). The findings here confirm a claim made by Zhang (1963: 125–126) in a discussion of the development of metonymy patterns in Chinese: “There are increasing numbers of cases using a positive persona or an advanced unit for a general term. …This rhetorical pattern has been used (in Chinese) since ancient times. There are also cases using a negative persona for a general term but these are relatively uncommon [my translation]”. (131)

 Du Qiuniang The beautiful prostitute in Du Qiuniang, written by Du Mu in the Tang Dynasty. → A prostitute.

It is worth noting that the results found in this study by no means suggest that INDIVIDUAL FOR COLLECTION in English for daily use is absolutely less popular than in Chinese. On the contrary, many studies show that INDIVIDUAL FOR COLLECTION is quite common in English language as well (e.g. Kleparski 1996, 1997; Barcelona 2004). The cross-linguistic difference found in this pattern might also be a side effect of the discrepancy of the compilation of the data resources from the DCSW/DCMS and the HTOED. In the guide to the HTOED, the editors explain explicitly that “proper names are not systematically covered by the dictionary,

Discussion | 155

though many are entered because the terms themselves are used in extended or allusive meanings, or because they are in some way culturally significant”60. However, as a pervasive rhetorical device discussed by many Chinese rhetoricians (see Zhang 1963; Yang 1980; Chen 1997), INDIVIDUAL FOR COLLECTION is still a hotspot in Chinese rhetoric research. It should be no surprise that when compiling the DCSW and the DCMS, the editors were prone to including instances of this pattern. Therefore, it is quite possible that the variation in respect to the pattern INDIVIDUAL FOR COLLECTION is also due to the unsystematic selection of proper names in the HTOED and the attention attached to “proper name for general term” in the DCSW and the DCMS.

3.3.2 The carriage-clothing system in Chinese culture Besides its practical and aesthetic functions, clothing has a crucial function in relation to social status in traditional Chinese culture. Although other cultures may also use clothing to indicate social rank or status (e.g. the Romans also introduced a variety of apparel bans), very few societies have developed a system with such strict social/ethical codes for clothing as in feudal China. I will first comment on the social function of clothing reflected in the Confucian classics, and then I will briefly introduce the carriage-clothing system in Chinese historical records. Finally, the influence of the carriage-clothing system on the preferred metonymies found in  will be discussed.

The clothing system and the Confucian li “rites” Confucianism, as an ethical and philosophical system, gained and retained great influence in all fields of Chinese society and culture. The dominance of Confucianism in Chinese history has shaped peoples’ minds and behaviors, and it also influenced the clothing system in the feudal society of China. Central to Confucian thought, the concept  li “rites” refers to ritual laws and moral standards established by governing classes to consolidate the social hierarchy and the patriarchal clan system. The cardinal political idea of Confucianism is  li–zhi “the rule of rites”, which advocates the operation of government through the rule of rites. In other words, all social ranks should obey the ritual principles and moral rules so that the country will endure. The purport of the

|| 60 http://www.oed.com/public/oed3guide

156 | The comparative perspective: Metonymies for PERSON

Confucian li is to define “what should distinguish people in different positions from each other” (Elvin 1985: 168), as in: [1]

“    ”

  

The Master said, “It is by the rules of ceremony [li] that what is doubtful is displayed, and what is minute is distinguished, that they may serve as dykes for the people. Thus it is that there are the grades of the noble and the mean, the distinctions of dress, the different places at court; and so the people (are taught to) give place to one another.” Book of Rites: Record of the Dykes61

The rituals and practices of li crystallized and embodied diverse topics of human life like learning, titles, mourning and governance. Among these topics, clothing is particularly essential. As far back as the Zhou Dynasty (1046–256 BC), official rules for clothing were set up in pace with the establishment of social hierarchies. At that time, clothing started to take on its essential role in the decrees and regulations of the Zhou Dynasty. The Rites of the Zhou, which describes rules of conduct mainly for the ruling class in the Zhou Dynasty, records a number of official titles and responsibilities that involve clothing, for example: [2]

  ­ The official for clothing: Taking charge of clothing for the emperor in sacrificial ceremonies; classifying them in accordance with the government decrees, and assigning them in line with different sacrificial rituals.  

Later, in the Book of Rites, which propagated the rule of rites in the Spring and Autumn Period and the Warring States Period (770–221 BC), the importance of clothing was further highlighted. Expanding to address all social activities and hierarchies, the Book of Rites systematically explained the rules of conduct in feudal China with much emphasis on clothing. The clothing system became

|| 61 The citations from the Book of Rites, the Analects, the Book of Poetry and Yi Jing (The Book of Changes) in this book are translated by James Legge, from the Chinese Text Project, http://ctext.org/liji/zhong-yong. All others are my translations.

Discussion | 157

very strict at this time, in the sense that violating the clothing regulations or wearing bizarre clothes was regarded as one of the “four misbehaviors”; see [3]. [3]





Using licentious music; strange garments; wonderful contrivances and extraordinary implements, thus raising doubts among the multitudes: all who used or formed such things were put to death. Book of Rites: Royal Regulations

The clothing system set the norms not only for CLOTHES but also for ACCESSORIES, especially HEADWEAR (e.g. CAPPING) and JADE (or PRECIOUS STONES), as explained in [4] and [5]: [4]

    Hence after the capping has taken place, provision is made for every other article of dress. With the complete provision of the dress, the bodily carriage becomes (fully) correct, the harmonious expression of the countenance is made perfect, and the speech is all conformed to its purposes. Book of Rites: The Meaning of the Ceremony of Capping

[5]

 ­



Anciently, men of rank did not fail to wear their girdle-pendants with their precious stones, those on the right giving the notes Zhi and Jiao, and those on the left Gong and Yu. Book of Rites: The Jade-bead Pendants of the Royal Cap

Chinese people have hence attached great importance to the social/ethical functions of clothing. The Book of Rites specifies that people should not only follow the clothing codes to choose the correct clothing, but also know the names and the social functions of the clothes; see [6]. [6]

€‚ƒ„…†‡ˆ‰ When a man had his robes on his person, and did not know their names (or the meaning of their names), he was ignorant indeed. Book of Rites: Smaller Rules of Demeanor

158 | The comparative perspective: Metonymies for PERSON

The carriage-clothing system in historical records Since the Zhou Dynasty, the concept of li has never completely disappeared from Chinese society. Every ruling class in each dynasty tried to unify people’s thought in accordance with the rule of rites by establishing and propagating rules of conduct including clothing regulations. The Twenty-Five Histories, which is a collection of Chinese historical books covering a period from 3000 BC to the Qing Dynasty in the early 20th century, is a considerable resource of Chinese history and culture. In the historical records of different dynasties,  yu–fu–zhi [carriage–clothing–records], starting from the Book of the Later Han, faithfully document every step clothing has taken towards its social and historical development (Hua 2001: 116). In Chinese historical records,  yu–fu is an umbrella term for clothes, carriages, flags, uniforms, headwear, footwear, accessories and so forth. In each dynasty, the clothing system was strict and divided people by rank. A man had to put on the dress of his own rank to obey the rule of rites: [7]

  Hence esteem is shown to the noble, honor is given to the honorable, and no one goes against this; only in this way can it be li. A man should not wear the clothes which are not of his rank; only in this way can he obey the rule of rites. Book of the Later Han: Carriage Clothing Records

A number of apparel bans for officials and common people in feudal China were recorded in the Twenty-Five Histories. These apparel bans cover various aspects of social life and involve the color, material and design of clothes and accessories; see [8] and [9]. [8]

    Common persons, merchants, craftsmen, unless appearing in performances on stage, are not permitted to dress in black or white clothes, or waistbands made of stiff fabrics; they are not allowed to wear purple. History of the Song: Carriage Clothing Records 

Discussion | 159

[9]

ㅅⅉಹಹ♗ⅳ䟆᧨ₜ㈦⍼䞷兲枵兊€兺処᧨㷱 幇共侯兀ಹಹᇵ㢝⚁ವᇶ On the clothing and caps for common people: …for men and women, they are not allowed to wear gold embroidery, brocade or figured woven silk material, fabrics made of ramie, and fine silk fabrics; but only ordinary silk, tough silk, white crepe silk…    

In sum, the carriageclothing system in feudal China reflects the strict social hierarchy of that time. Although as part of the culture, clothing could differ from one dynasty to another, as culture varies across time (Zhou and Gao 1987), in all dynasties of feudal China, there was little flexibility in clothing and great concern with dressing codes (Shen 2005). The society’s interpretation of the Confucian li largely determined an individual’s choice of clothing to match his/her social identity, which provided a strong motivation for metonymic links between a person and his clothing.

The carriage-clothing system and preferred sources for PERSON in Chinese The close connection between the carriage-clothing system and social status in Chinese culture has resulted in a number of metonymic expressions of the pattern PIECE OF CLOTHING FOR PERSON in Chinese, as introduced in Section 3.2.1. First, distinctions of social status in feudal China are reflected in the main targets of this pattern in , such as OFFICIAL, SOLDIER, SCHOLAR, RELIGIOUS PERSON, COMMON PERSON and SERVANT. Another main target of this pattern in  is BEAUTIFUL PERSON (all 31 cases are actually for BEAUTIFUL WOMAN), which shows that the distinction of male and female could also be achieved by judging clothing; see quotation [10] and example (132). In contrast, metonymies of PIECE OF CLOTHING FOR PERSON in  are mainly for targets like SOLDIER, PERSON WITH SPECIFIC PROFESSION and SCHOLAR, as in (133). Unlike in Chinese, with its great emphasis on clothing’s social functions, clothing metonymies in English tend to highlight the practical functions of occupational clothing. The metonymic relation between an article of dress and the person who wears it in English has been recognized by many scholars (Warren 1992; Kleparski 2000; Rusinek 2008a, 2008b). However, a person in an English-speaking society might change dress constantly and there might be no constant relation between the clothing and the wearer in the experiential world (Dirven 1985: 97). Hence, traditionally, English-speaking people have likely been less constrained by strict clothing

160 | The comparative perspective: Metonymies for PERSON

regulations than people in feudal China. It is possible that for this reason the metonymic pattern PIECE OF CLOTHING FOR PERSON in English is not as popular as in Chinese, but this interpretation can only be made very tentatively without evidence from a diachronic comparative study. (132)

 hong–qun /  hong–shang [red–skirt] → a beautiful woman  hong–xiu [red–sleeves] → a beautiful woman  hong–zhuang [red–clothes] → a beautiful woman

(133)

black gown → an academic/a Christian priest velvet-cap → a physician blue apron → a tradesman lob-coat → a farmer whitecoat → a physician/a soldier redcap → a member of the military police suit → a business executive

[10]

    Outside or inside, they [male and female] should not go to the same well, nor to the same bathing-house. They should not share the same mat in lying down; they should not ask or borrow anything from one another; they should not wear similar upper or lower garments. Book of Rites: The Pattern of the Family

 Second, as shown in Table 12, the proportions of PIECE OF CLOTHING FOR PERSON for the targets NEGATIVE PERSON, CHILD and ETHNIC GROUP in Chinese are significantly higher than expected, which is also a result of cultural influence. For the target NEGATIVE PERSON, all the six instances metonymically refer (derogatorily) to “sons of the rich, playboys” as in example (80), and all of them literally denote “fine silk material”, which was a unique clothing fabric for rich people according to the clothing system in feudal China (see [9]). All the instances (n = 15) for the target CHILD in  are with the source HAIRSTYLE, more precisely the specific hairstyles of early childhood; see examples (87) and (134). The ceremony of cap-

Discussion | 161

ping62 in the Chinese clothing system marks a great change concerning hairstyles between boys and adult men. In consequence, HAIRSTYLE is a salient feature distinguishing children from adults. The expressions of PIECE OF CLOTHING FOR PERSON for the target ETHNIC GROUP are all literally from the semantic fields of fur coats and woolen clothes, which were uncommon clothing for the ethnic majority according to the clothing regulations. The clothing records reveal that for the ethnic majority Han, the common clothes were normally made of fabrics such as silk, cotton and flax. Thus, metonymic links between uncommon clothes and ethnic minorities are motivated, as in (135). (134)

 jiao–guan,  liang–guan,  zong–guan

(135)

 zhan–cui

Hair twisted into the shape of two horns. → A child. Woolen clothes. → A northern minority nationality.  zhan–qiu Fur coats. → A northern minority nationality.

Third, the importance of headwear and jade (see quotations [4] and [5]) in the Chinese clothing system is supported by the finding that many expressions originally denoting HEADWEAR and JADE have metonymic interpretations of the wearers, as in examples (100) and (136). The expressions literally denoting the four main kinds of headwear (formal headdress, stiff hats, hairpins and soft caps) in ancient China show a close correspondence between the kind of headwear and the social status of the wearer. For example, the target OFFICIAL is mainly with the source FORMAL HEADDRESS (100), while the target SCHOLAR is mainly with the source SOFT CAPS (137). HEADWEAR is also a distinguishing characteristic between male and female (see quotation [11]), as in CAPS for males and HAIRPINS for females (138). (136)

 wu–yu The five kinds of jade used as authenticating objects of feudal lords. → Feudal lords.  ming–yu A jade accessory worn on one’s waist. → An official.

|| 62 According to the Book of Rites: Summary of the Rules of Propriety, “When he is twenty, we call him a youth; he is capped”.

162 | The comparative perspective: Metonymies for PERSON

(137)

⎡ ru–jin The soft caps for scholars. → A scholar.  jin–juan (Type 4) Soft caps and books. → A scholar.

(138)

 jin–chai A gold hairpin. → A woman. ⹡嗕 han–jing Hairpin made of bramble stems. → One’s own wife (self-deprecatory). 

[11]

㢞Ⲋ䶓᧨㓏ⅴ⒴䟆⃮ᇭᇵ䯋帿ವ⃟帿ᇶ By marriage, capping, and the assumption of the hair-pin, they maintained the separation that should exist between male and female. Book of Rites: Record of Music

Fourth, as illustrated in quotations [8] and [9], the apparel bans in feudal China deal with various aspects of clothing, like its color, fabrics and design (see also Zheng 2008: 6–14). Consequently, all these aspects could be used as metonymic sources to differentiate wearers of different social statuses. Table 29 briefly presents the metonymic correspondence between clothes in different colors and the wearers of various social classes. It shows that colors like RED and PURPLE were unique colors for people with high social status like emperors or high-level officials. These two colors were typical for formal dress, and people from other social ranks were not expected to wear these colors even for informal clothes; see [12]. [12]

⚪ₜⅴ兏偔毿᧨ₜⅴ⃉⅄ᇭᇵ幉幼ವ⃰⏩ᇶ The superior man [or man of honor] did not use a deep purple, or a puce color, in the ornaments of his dress. Even in his undress [informal clothes], he did not wear anything of a red or reddish color. The Analects: Among the Xiang and the Dang63

|| 63 A xiang was a group of 12,500 families; a dang a group of 500 families.

Discussion | 163

Table 29: Clothes of different colors as metonymic sources for people of various social classes

Colors of clothes

Expression example

Metonymic target found in 

Social stratification

SOVEREIGN RULER

highest social class

color color term category YELLOW

RED64

 huang

 huang–pao [yellow robe]

 zhe  chi  zhu

 zhe–huang–pao /  zhe– SOVEREIGN RULER pao [reddish brown robe]  zhe–huang–yi [reddish brown clothes]  chi–fei /  chi–fu [red long front cloth panel attached from the waist belt] chi–yi [red–clothes]

HIGH-LEVEL OFFICIAL,

zhu–fu [red long front cloth panel attached from the waist belt]  zhu–yi [red clothes]  zhu–zi [red–purple] “red clothes and purple ribbons”

HIGH-LEVEL OFFICIAL

HIGH-LEVEL OFFICIAL

PURPLE

 zi

 zhu–zi [red–purple] “red clothes and purple ribbons”  qing–zi [vivid purple] “vivid purple clothes and ribbons”

QING-

 qing

 qing–jin [qing-black–garment front],  qing–jin [qing-black– long gown]

BLACK65

upper-class

NOBILITY

SCHOLAR

middle-class

|| 64 The expressions in example (132) literally all refer to red clothes, but with the metonymic target of BEAUTIFUL WOMAN, which is not a high social rank in feudal China. In fact, the RED in example (132) and in the examples of Table 29 are actually different kinds of RED and have different linguistic realizations. For the former, the linguistic form of RED is  hong, while for the latter, the linguistic forms are  zhe,  chi or  zhu. 65  qing is polysemous color. It may refer to the basic color BLUE as in qing–gong (see Footnote 31), or BLACK as in example (56), or DARK GREEN as in  qing–shan [dark green– mountain]. When qing denotes the color of BLACK, it is classified into the color category of QINGBLACK to discriminate it from the color BLACK expressed by linguistic forms of zao and  hei.

164 | The comparative perspective: Metonymies for PERSON

Colors of clothes

Expression example

Metonymic target found in 

Social stratification

color color term category  qing–shan [qing-black– sleeveless jacket]

OFFICIAL

 qing–pao [qing-black–robe]

LOW-LEVEL OFFICIAL

 qing–yi [qing-black–clothes]

SERVANT, ENTERTAINER

 qing–pao [qing-black–robe]

LOW PERSON, POOR

SCHOLAR, LOW-LEVEL

low-class

SCHOLAR WHITE

 bai

 bai–yi [white–clothes]

BEADLE, COMMON PERSON, LAYMAN, LIQUOR SERVANT, SOLDIER IN INFORMAL CLOTHES

 bai–shan’er [white sleeveless jacket]

BLACK

 zao  hei

UNDERLINGS OF EUNUCH

 bai–pao [white–robe]

COMMON PERSON

 zao–yi [black–clothes]

JUNIOR OFFICIAL

middle-class

MONKS, GUARD OF

religious person

 hei–yi [black–clothes]

PALACE BROWN

 he

 cui–he [shaving off the hair– brown clothes] (Type 4)

MONKS

Table 30 and Table 31 further show metonymic correspondences between fabrics/designs of clothes and the wearers, which are in accordance with the clothing system in feudal China (see quotation [13]). Furthermore, the established differences in clothing with regard to color, fabric and design that were based on the social statuses of the wearers motivates coordinate constructions in which the two constituents are in an antonymous relation. These expressions metonymically indicate wearers from two opposite social stratifications, as in 0. [13]

 

The son of Heaven wore an upper robe with dragons figured on it; princes, a lower robe with axes embroidered on it; great officers, a lower robe with the symbol of distinction; and other officers [scholar-officials], a darkcolored upper robe, and the lower one red. Book of Rites: Rites in the Formation of Character

Discussion | 165

Table 30: Clothes of different fabrics as metonymic sources for people of various social classes

Fabrics of clothes

Expression example

brocade, damask

 jin–yi [clothes made of brocade]

leather, metal

 zu–jia [armor]  si–jia [armor]  mei–wei [leather-made armor]

wood

 mu–jia [wooden armor]

silk

 luo–yi /  luo–qi /  qi–luo [silk clothes]

cotton cloth

 wei–gao /  bu–wei [clothes made of cloth and girdle without precious stones]  bu–yi [clothes made of cloth]

gunny cloth

 ma–yi [gunny clothes]

coarse cloth

 bu–he /  he–yi /  ye–he [coarse clothes]

fur, wool

 zhan–qiu [fur coat]  zhan–cui [woolen coat]

ko-hemp cloth

 hui–fu [clothes made of ko-hemp cloth]

(139)

Metonymic target found in  NOBILITY SOLDIER

BEAUTIFUL WOMAN

POOR PERSON

COMMON PERSON

MINORITY NATIONALITY

 bai–hei White and black. → White clothes and black clothes. → Common people and monks.  hua–su Gorgeous and plain (in color, fabric or design of clothing). → Gorgeous clothes and plain clothes. → Noblemen and common people.  zi–huang Black and yellow. → Black clothes and yellow Taoist cap. → Monks and Taoist priests.

166 | The comparative perspective: Metonymies for PERSON

Table 31: Clothes of different design as metonymic sources for people of various social classes

Design of clothes

Expression example

court dress with dragon pattern

 gun–long–pao [court dress with dragon pattern]

clothes with colorful embroidery

 xiu–fu /  xiu–yi [clothes with colorful embroidery]

court dress

 chao–fu [formal court dress]

ceremonial court dress

 tai–gun /  shang–gun [ceremonial court dress]  tuan–shang [ceremonial court dress and headwear]

academics’ dress

 ru–fu /  ru–zhuang /  ru–yi /  shi–fu [clothes for scholars]

clothes with wide sleeves

 feng–yi /  feng–ye [clothes with wide sleeves]

square collars

 fang–ling [square collar]

cassock

 fang–pao /  jia–sha [cassock]

Taoist priest frock

 yu–yi,  guan–pei [Taoist priest frock]

short half-body garment  duan–yi [short coat]

Metonymic target found in  SOVEREIGN RULER

NOBILITY

COURT OFFICIAL OFFICIALS

SCHOLAR

MONK

TAOIST PRIEST COMMON PERSON

Fifth, the preference for the source HAIRSTYLE in  (see Table 14 and Table 16) is also culture-determined to some extent. As one aspect of clothing systems or fashion, HAIRSTYLE can be considered a signifier of social class, especially in feudal China (Wu 2008: 106). Generally, the source HAIRSTYLE indicates three kinds of PERSON: CHILD (n = 13), example (134); SERVANT (n = 9), example (99); and BEAUTIFUL PERSON (n = 9; all instances refer to BEAUTIFUL WOMAN), example (103). As explained above, the selection of HAIRSTYLE as the source for the target CHILD could be motivated by the importance of the ceremony of capping in feudal China. For the targets SERVANT and BEAUTIFUL WOMAN, the preferred source HAIRSTYLE is due to the fact that at that time, HAIRSTYLE was one of the distinguishing features that separated servants from other social classes or signaled the female identity. In , in total, only three instances are found with the source HAIRSTYLE for sporadic targets (140).

Discussion | 167

(140)

long-hair → A ‘brainy’ person/an æsthete/an intellectual. soap-lock A lock of hair made smooth by the application of soap. → One who wears such. kinky Having/full of kinks; closely curled or twisted. → Kinky hair. → A person with “kinky” hair.

The Chinese carriage-clothing system is also reflected in the source selection of the pattern POSSESSED FOR POSSESSOR, especially in the sources VEHICLE (CARRIAGE) and ARTIFACT (SEAL, FLAG); see examples (111), (112) and (113) and Table 18 and Table 20. In traditional ceremonies, carriages and flags were two important elements for emperors and high-level officials, and in the Book of Rites, we find a number of records on their regulation. The carriages and flags for emperors changed from season to season (see quotation [14]), and officials at different levels had their own carriages corresponding to their ranks. Quotations [15] and [16] indicate that in feudal China CARRIAGE and FLAG were also tokens of social status, so that expressions from these semantic fields could serve as metonymic sources for the possessors. For the source SEAL in , expressions are etymologically related to the official seal, which serves as a personal identification of the official by stating his name and official title. Likewise, the colors of the ribbons on the seals represent the different official ranks, as in (141). [14]

…… ……



The (period of) slighter heat arrives…The son of Heaven occupies the Ming Tang Grand Fang; rides in a vermilion carriage, drawn by red horses with black tails, and bearing a red flag. He is dressed in red robes, and wears carnation gems… Book of Rites: Proceedings of Government in the Different Months [15]

    The emperor rides in a carriage with colorful engraving; the highest-level officials ride in carriages with colorful painting; the great officials ride in black carriages without decoration; the scholar-officials ride in carriages made of bamboo; the ordinary people ride in carriages for service. Rites of the Zhou: Offices of Spring

168 | The comparative perspective: Metonymies for PERSON

[16]

 The emperor has a flag with patterns of the sun and the moon; the feudal lords with patterns of dragons and small bells; the highest-level officials with a falcon pattern; the great officials and the scholar-officials have motley flags… Rites of the Zhou: Offices of Spring

(141)

qing–shou Cyan ribbons. → An official seal with cyan ribbons. → A high-level official.  hei–shou Black ribbons. → An official seal with black ribbons. → A low-level official.

In sum, in a society with a strict social hierarchy, people could “judge your social status by your clothing ( )”66. The social function of the carriage-clothing system in Chinese culture has triggered many metonymic expressions of the patterns PIECE OF CLOTHING FOR PERSON and POSSESSED FOR POSSESSOR in Chinese. The cultural motivation behind these expressions echoes the statement of Jia Yi, a Confucian scholar (201–169 BC), in his representative work, New Book; see [17]. [17]

  ­€‚ƒ„…†  ‡ˆ‰Š‹„…ŒŠ People’s innate qualities/characters and appearances are all the same. The difference between noble and humble is not based on a person’s inborn personality or looks. Rather, it is rank, power, clothing and order that distinguish between noble and humble people, between the highclass and the low-class. New Book: Equality

3.3.3 The residence system in Chinese culture Beyond the clothing system, the Confucian li has had an effect on various other aspects of the society, one of which involves the important notion of LOCATION in || 66 From Jia Yi, New Book: Equality.

Discussion | 169

Chinese culture. The social hierarchy reflected in the residence system in ancient times also plays a role in the motivation of metonymies of LOCATION FOR LOCATED in Chinese. To maintain a feudal code of social hierarchy, the ruling class established regulations for residence, as illustrated in the quotation from Master Xun Zi in [18]. The expression  gong–shi, literally “the palace and the house”, embodies a broader sense in Chinese literature, “the general residence of people”. [18]

:…… The royal regulations: …clothing should be in accordance with regulations, and residence (gong–shi) should follow rules and measures… Xun Zi: Royal Regulations

Therefore, in a similar vein to the clothing-carriage system, residence regulations in feudal China evoked the coining of many expressions which literally express locative meanings for the metonymic meaning of “people located in the place”. Table 23 shows that the sources COURT and HOUSE exist in  but not in . A closer glance at their targets suggests that COURT has two possible metonymic readings: SOVEREIGN RULER and OFFICIAL (COURTIER); see examples (114), (119) and (142); HOUSE has sporadic targets like COMMON PERSON, POOR PERSON, NOBILITY, CONCUBINE, UNCHASTE WOMAN, etc.; see (115) and (143). From an onomasiological perspective, for the target OFFICIAL, COURT is a preferred source in  (see Table 22). In feudal China, the court was the place where a sovereign and his courtiers discussed state affairs. In other words, the court functions as the working place (i.e. office) for a sovereign and courtiers. The importance of working place in Chinese culture, as said in quotation [19], provides motivation for a metonymic reasoning of “working place” for “people working there”. In addition, “the different places at court” (see quotation [1]) also signal different grades of people at court (142). For the targets expressed by the source HOUSE, one can also safely assume that their metonymic links are motivated by one prominent feature, which is that people from different social ranks occupy different types of residences. [19]

  The Master said, “He who is not in any particular office [working place] has nothing to do with plans for the administration of its duties.” The

170 | The comparative perspective: Metonymies for PERSON

Master Zeng said, “The superior man, in his thoughts, does not go out of his place.” The Analects: Xian Asked (142)

 chao–you The right side of the court. → A courtier.  shang–lie The upper row at court. → A high-level courtier.

(143)

 zhu–men A red lacquered gate. → A wealthy house. → Powerful and wealthy families.  wai–zhai An external house. → A concubine.  sang–shu A mulberry-wood rotating shaft. → A door with this rotating shaft. → A poor house. → Poor people.

Leaving the general target category PERSON and the specific target OFFICIAL aside and looking at the targets SOVEREIGN RULER and ROYAL PERSON, PALACE and COURT are preferred sources for the former and PALACE is preferred for the latter. The court was the working place of an emperor, and the palace was his living place67. The palace was also the living place of the royal family (princes, queens, princesses, imperial concubines, etc.), as illustrated in [20]. In consequence, the specificity of the locative relation between COURT/PALACE and targets SOVEREIGN RULER/ROYAL PERSON triggered a number of metonymic expressions in Chinese, as in (144). [20]

 [The Imperial Palace has an inner part and an outer part.] The inner part [palace] has nine rooms for imperial concubines on nine different levels to live; the outer part [court] has nine rooms for courtiers at nine different levels to have an audience with the emperor. Rites of the Zhou: Office of Winter, Record of Trades

|| 67 Both COURT and PALACE refer to the main building in an ancient Chinese capital, i.e. the Imperial Palace: “the front for court and the back for residence ()” (Rites of the Zhou: Office of Winter, Record of Trades).

Discussion | 171

(144)

 zhong–gong The middle palace. → The queen.  dong–gong The eastern palace. → The crown prince.  xi–gong The western palace. → The imperial concubines.  jiao–fang Generally refers to the palace of imperial concubines. → The imperial concubines.

The rank distinction in relation to LOCATION goes far beyond the Imperial Palace in Chinese culture. Even in common people’s houses, both the social hierarchy and the patriarchal clan system influence people’s specific location in the house. First, the males (especially the husband) and the females (especially the wife) are located separately in the house according to the rule of rites; see [21] and (145). Second, a locative distinction between parents and children is also reflected in Chinese metonymies. The examples in (146) all literally refer to the parents’ room in a house:  tang “the principal room, the hall”. Because the principal room is particularly associated with parents, it can serve as a metonymic source for the target MOTHER/PARENTS. Third, examples found in  also reveal the differential status between wife and concubines in feudal China; see the examples in example (35). [21]

……  The observances of propriety commence with careful attention to the relations between husband and wife. They built the mansion and its apartments, distinguishing between the exterior and interior parts. The men occupied the exterior; the women the interior. …The men did not enter the interior; the women did not come out into the exterior. Book of Rites: The Pattern of the Family

(145)

 nei Inner. → A woman.  jian–nei [lowly–interior] → My wife (with a self-deprecatory tone).  wai

172 | The comparative perspective: Metonymies for PERSON

Outer. → A male, esp. the husband. (146)

 xuan–tang Mother’s room, i.e. the principal room. → The mother.  gao–tang A hall with a high ceiling. → Parents.

3.3.4 Culture-bound conceptualizations of body parts The results of this case study show that the proportion of BODYPART FOR WHOLE is significantly higher in  for PERSON in general and for NEGATIVE PERSON (Table 13). A closer look at the sources of this pattern reveals that the source HEAD/BRAIN is preferred in , while the source FACIALPART (especially EYEBROWS) is preferred in  for PERSON. Moreover, the source GENITALS is relatively frequent (7.7%) in , but no instance is found in . These findings will be discussed in detail below.

HEAD/BRAIN for NEGATIVE PERSON/UNWISE PERSON In , body parts are preferred sources for NEGATIVE PERSON. The brain or head is believed to be one of the seats of feeling/emotion in Western culture (LeDoux and Phelps 2008; Yu 2009: 365)68, and hence a container of negative emo-

|| 68 In Western culture, both the head/brain and the heart are seen as critical organs of the body, and both have been conceptualized in various metaphoric/metonymic terms. After a detailed examination of literature on the historical development of the concepts of HEAD/BRAIN and HEART in the West, Yu (2009: 365) concludes that present-day English regards the heart as the center of feelings and the brain as the center of thought, whereas in scientific terms the brain controls both thought and feelings. Although a separation between THOUGHT/MIND/ INTELLECT located in the head/brain and EMOTION/FEELING located in the heart still exists in Englishspeaking culture, the dichotomy tends to be blurred since developments in Western medicine have shown the role of HEAD/BRAIN in emotional experience. Some emotions/feelings are reenactments or simulations of sensory-motor experiences, which are located in the sensorymotor areas of the brain (cf. Barsalou et al. 2003). The scientific discovery of the role of the head/brain then triggers many metaphorical conceptualizations of HEAD/BRAIN as a container of emotions/feelings in English. In a series of studies on metaphors with Chinese body parts, Yu (2003b, 2007, 2009) finds that HEAD/BRAIN and HEART are also seats of mental experience (e.g. emotion, mind, thought) in Chinese culture. Although a large number of conventionalized expressions in Chinese provide evidence of the metaphorical conceptualization of MIND/EMOTION in relation to HEAD/BRAIN/HEART (see also Pritzker 2007), metonymies with HEAD/BRAIN and HEART for PERSON are not frequent in my datasets. One might speculate that

Discussion | 173

tions/temper. Consequently, expressions relating to HEAD/BRAIN may metonymically refer to people with persistently negative emotions like “an impetuous or quick-tempered person” or “a person of flighty and hasty character” (e.g. hothead, fling-brain). However, NEGATIVE PERSON in  is mainly metonymically conceptualized by clothing or personal characteristics; see examples (80) and (81). The noticeable cross-linguistic variation of this pattern also links to the popularization of INTELLIGENCE metaphors in Western culture (Feyaerts 1999; Allan 2006, 2008). The brain had already been widely acknowledged as the locus of intelligence by the Middle Ages (Allan 2008: 35). The reasoning behind the link is that “intelligence as a property of the mind is metaphorically reified as a concrete object and metonymically situated in the brain, which itself is situated in the head” (Dirven 2003: 84). Two poles of INTELLIGENCE are found in , CLEVERNESS and STUPIDITY, represented by the targets INTELLECTUAL PERSON and UNWISE PERSON, respectively. For the target INTELLECTUAL PERSON, only seven instances of BODYPART FOR WHOLE are found in  (e.g. egg-head, pointy-head, highbrow, master-brain, the brain(s)). As presented above (see Section 3.2.3), in total, 99 instances of BODYPART FOR WHOLE for the target UNWISE PERSON are found in , and 91 of them are with the source BRAIN/HEAD. The great quantity difference between the targets INTELLECTUAL PERSON and UNWISE PERSON in English accords with the generally more frequent development of derogatory vocabulary than ameliorative vocabulary (Ullmann 1962: 233; see also Allan 2008: 24). Morphologically, most instances (n = 87) with BRAIN/HEAD as the source for UNWISE PERSON are compounds with different linguistic forms of HEAD or BRAIN, and only four instances are single words (cabbage, clot, clod and noll). Among the compounds, 64 instances contain the word head, eight instances are realized with the word brain, eight instances with pate, three with skull, and sporadic instances with noll, poll, etc. As to the metaphorical source of STUPIDITY (i.e. THE MIND IS AN ENTITY), a high proportion of the instances are from the semantic fields of “pieces of wood” (e.g. blockhead, blockpate, log-head, noghead, woodenhead), of “food” (e.g. beefhead, cabbage-head, peabrain, puddinghead, pumpkinhead, chowderhead), of “animal” (e.g. oxhead, sheep’s head) and of “earth” (e.g. clod skull, clod-head, clod-pate, mudhead). This finding supports the research by Allan (2008) on how INTELLIGENCE is metaphorically conceptualized to a large extent. The semantic extension found here is the result of interaction

|| while the first step of metaphorical mappings from HEAD/BRAIN/HEART to EMOTION/MIND is common in Chinese, the further metonymic mapping between a person and the emotion/mind is underdeveloped to some extent.

174 | The comparative perspective: Metonymies for PERSON

between metaphor and metonymy: first, the BRAIN/HEAD is metaphorically conceptualized as the metaphorical source (WOOD, FOOD, etc.), and then it metonymically maps to the person with this brain/head, that is, a person with low intelligence. In addition, many expressions in this case are marked “derogatory” or “humorous” in the HTOED, which supports the notion that metonymy has pragmatic implications, producing extra meaning effects (see Herrero Ruiz 2011). However, in , the targets UNWISE PERSON and INTELLECTUAL PERSON are mainly conceptualized by paragons of the category instead of body parts.

FACIALPART

The high proportions of the source FACIALPART (Table 24) in both  (46.7%) and  (15.0%) are in accordance with the INTERACTIONAL OVER NONINTERACTIONAL principle (Radden and Kövecses 1999: 46). During interpersonal communication, facial parts, especially the face, are the most salient parts of our bodies when we interact with other people and thus serve as better reference points for a person than other body parts (e.g. LIMBS, TORSO) with which we do not interact directly; consider the English examples iron-face, bold-face, brazen-face and brass-face (which all denote the target NEGATIVE PERSON). Expressions relating to facial parts like MOUTH and NOSE are also found in , such as big mouth and sharp-nose for NEGATIVE PERSON. The source FACE is frequent in  as well; see (147). (147)

 wen–yan A kind face. → Mother.  ci–yan A kind face. → Mother.

The cross-linguistic variation in the source FACIALPART is largely determined by the preference for the source EYEBROWS (n = 18) in , while no instance with EYEBROWS is found in . The specific targets of EYEBROWS metonymies in Chinese are mainly subtargets of BEAUTIFUL WOMAN (n = 16). Chinese people believe that eyes and eyebrows can speak one’s emotions and deliver one’s feelings (consider the Chinese fixed expression  [eyes–eyebrows–deliver– feelings]) (see also Eberhard 1971). Furthermore, eyebrows can easily vary in shape, which leads to the eyebrows’ popularity as a metonymic source for the target BEAUTIFUL WOMAN (Zhao 2001: 86). The importance of eyebrows in Chinese aesthetic appreciation of women has been noticed since ancient times. Chinese writers are generous in praising the beauty of women’s eyebrows, and a number

Discussion | 175

of descriptions of eyebrows as a symbol of feminine beauty can be found in classical poetry; see [22]. [22]

…… …her forehead cicada-like; her eyebrows like (the antennae of) the silkworm moth; What dimples, as she artfully smiled! How lovely her eyes, with the black and white so well defined! Book of Poetry: Shuo Ren [the Western Zhou (1046 BC–771 BC)]

GENITALS

Another interesting finding related to BODYPART FOR PERSON rests in the crosslinguistic variation in the source of GENITALS. In total, 17 instances with this source are found in . Among them, eight instances with FEMALE GENITALS metonymically refer to targets of UNCHASTE WOMAN (n = 5, e.g. muff, tail), A WOMAN (pussy, quim) and BEAUTIFUL WOMAN (beaver). Nine instances with MALE GENITALS metonymically signify the targets UNWISE PERSON (n = 5, e.g. dork, bollock), INTIMATE PERSON (natural69), CHILD (pillicock) and A MAN70 (e.g. prick). Obviously, genitals are viewed as crucial body parts that are decisive in determining whether a person is a man or woman (Grygiel 2005: 252). Therefore, it is reasonable to use the female or male genitals to express a female or a male in the pattern BODYPART FOR WHOLE. Generally, the targets metonymically expressed by GENITALS have negative connotations (e.g. UNWISE PERSON) or are related to sexuality (e.g. UNCHASTE WOMAN). Needless to say, many instances of this case are taboo words in English culture and marked “derogatory” (Nesset 2010: 270). Expressions with genitals are also typically taboo in Chinese culture, and they are often used as swearwords or dysphemisms. Swearwords and taboo words, however, are not systematically compiled in ordinary Chinese dictionaries (Liu 1999), which provide the main sources of the DCSW and the DCMS. What has been overlooked in discussions of the cross-linguistic variation in GENITALS metonymies is that the registers of referring to a person by his/her || 69 natural: literally the genitals. Also plural, natural parts; metonymically refers to a mistress (OED). 70 The target A MAN includes those cases with a general referent of a male and without a specific indication.

176 | The comparative perspective: Metonymies for PERSON

genitals in English are mostly slang or colloquial words (marked “slang” or “colloq.” in the HTOED). As pointed out in Section 3.1.1, besides ordinary dictionaries, the original resources of the DCSW and the DCMS are mainly literary texts. The different registers from which the items were selected for the HTOED in contrast to the two Chinese metonymy dictionaries may also contribute to the cross-linguistic variation found here.

3.3.5 Collectivism versus individualism The division of cultures or societies as collectivist and individualist has been proposed as a way of interpreting cultural variability and cross-cultural psychology by many scholars (e.g. Hui and Triandis 1986; Triandis 1988, 1990; Schwartz 1990). According to Triandis (1995: 2), “collectivism is a social pattern consisting of closely linked individuals who see themselves as parts of one or more collectives (family, coworkers, tribe, nation)”, and in contrast, “individualism is a social pattern that consists of loosely linked individuals who view themselves as independent of collectives”. Briefly, individualism can be associated with the importance of “personal” issues, while collectivism can be regarded as inherently constituting one form of the “impersonality” concept (Yamamoto 2006). In Chinese culture, collectivism has been the chief leitmotiv71 for a long time, and it is often considered to be a product of the social and political ideology of Confucianism (Kim 1994, 1997). Confucian doctrines helped to shape the Chinese collective culture, which then built a strict hierarchical society based upon ties of kinship in the family and close personal relationships in the society. In Confucian morality, the individual is bound by his social role and the reciprocal obligations attached to it (Su et al. 1999: 28). Thus, social positions and ascribed status are important in the context of collectivistic societies in which individuals are in close relation to others and to the society (Kuwahara 2004: 24). Britain is generally considered an individualistic country. The religious philosophy of Protestantism that developed in the 16th century has been claimed to be responsible for the prevalence and prioritization of individualism over collectivism in Britain (Weber 1930 [1904]; Macfarlane 1978; Sampson 2000). Under the individualistic philosophy, “the self is defined independently of specific || 71 Based on cross-cultural studies, Hofstede (1984) concludes that Chinese societies are lower on individualism than most Western societies.

Discussion | 177

collectives in the sense that one is not expected to get involved ‘in the business’ of group members” (Triandis 1995: 10). In addition, individualists place an emphasis on high self-esteem and do not carry as many moral obligations as collectivists. We could speculate that the individualist culture in Britain plays an important role in the prevalence of CHARACTERISTIC FOR PERSON in  (see Table 13). For the target category PERSON in general and many specific targets like OFFICIAL, SOLDIER, SCHOLAR and so on, CHARACTERISTIC FOR PERSON is the preferred pattern in English. Table 25 further indicates that the cross-linguistic variation of this pattern mainly lies in the source NONPHYSICAL CHARACTERISTIC (e.g. irresistible, academic, unmarried), which is a typical personal attribute of an individual. At the same time, the fact that the pattern STATE FOR PERSON has attestations only in  (see Table 13) might also be attributed to the individualist culture in Englishspeaking society. For instance, the expressions overpower and nobleness, which originally refer to “a state or condition of being too powerful” and “the state or quality of being noble”, may metonymically designate a person in that state, i.e. “a superior” and “a noble person”. One could infer that in the English individualist culture, a person tends to be judged or identified by his or her personal attributes because he or she is seen as an autonomous individual. It was mentioned above that CLOTHING and LOCATION are two popular sources for PERSON in Chinese culture. Both may be related to how a collectivist tendency affects the salience of group members’ social status. The group-oriented ideology of the Chinese collectivist culture makes it natural to distinguish a person by his clothing or location rather than a personal attribute. Generally, the results found in this case study support the basic assumptions about these two different kinds of cultures as explained by Triandis (1995: 12): “Individualists thus construct their self-concept by focusing on personal attributes; collectivists focus on collective attributes, linking them to other people and to collectives”. And yet the likely correlation between the prevalence of collectivism/individualism and the metonymic source selections can be overridden by other factors. For example, this study’s results also show that even as a personal trait, the source PHYSICAL CHARACTERISTIC is preferred in . If specific targets are taken into account, the higher proportion of PHYSICAL CHARACTERISTIC (especially PHYSICAL AESTHETICS) in  is primarily due to metonymies for the target BEAUTIFUL PERSON (see Table 26). This variation can be accounted for by a variety of reasons. One of the most important reasons may lie in the rich Chinese aesthetic vocabulary referring to feminine beauty, which is an everlasting topic in Chinese literature. In the conventional CHARACTERISTIC FOR PERSON pattern, most of

178 | The comparative perspective: Metonymies for PERSON

these aesthetic near-synonyms can metonymically denote “a beautiful woman”; see (148). (148)

 jia–zhi,  jia–li,  jiao–mei,  jiao–li, jiao–ai,  miao–li,  shu–li,  yao–yan Beautiful, charming. → A beautiful woman.

A point to note is that the influence of individualism/collectivism on the distribution of metonymic patterns discussed here needs more scrupulous exploration. Individualism and collectivism are not notions with clean divisions, but are located on a continuum (Hofstede 1984). And, also, historically, the social patterns vary to some extent. Neither the Chinese nor the English data in this study are contemporary. Therefore, without a historical study on the social patterns in China and in Britain, the interpretation here remains tentative. Nonetheless, we do find that in the language of a codified society like feudal China, metonymies have more reference to the structured social world, which reflects a symbolic system (e.g. CLOTHING, LOCATION, VEHICLE, FLAG) shared as common knowledge by people in the society. Meanwhile, in English-speaking society, metonymies make more reference to individuals’ behavior or psychology (e.g. NONPHYSICAL CHARACTERISTIC, STATE). Understanding the underlying reasons behind these findings, however, will require research from multiple perspectives in the fields of sociology, anthropology and psychology, among others, and is beyond the scope of current study.

3.3.6 Section summary In summary, the main argument of Section 3.3 is that the cross-linguistic variation found in metonymies for PERSON between Chinese and English might be strongly influenced by many culture-pertinent factors. First, the preference for stereotypes and icons in Chinese culture might be responsible for the large number of metonymies of the pattern INDIVIDUAL FOR COLLECTION. Second, the Chinese carriage-clothing system, which determined a strict dress code in ancient China, seems to have led to a preference for the pattern PIECE OF CLOTHING FOR PERSON in general and several sources of POSSESSED FOR POSSESSOR (e.g. CARRIAGE, FLAG) in Chinese. Third, the popularization of Chinese metonymies of LOCATION FOR LOCATED may also have been indebted to the regulations on location/residence in feudal China. Fourth, culture-bound conceptualizations of body parts exist in each language. In English, HEAD and BRAIN are

Summary | 179 Discussion

common sources for targets related to INTELLIGENCE, as with the UNWISE PERSON, while the Chinese aesthetic appreciation of women has triggered an abundance of metonymies with the source of FACIALPARTS (especially EYEBROWS). Finally, I surmise that broad social patterns (specially, the contrast between individualism and collectivism) may have an influence on the salience of the sources NONPHYSICAL CHARACTERISTIC/STATE in English and the sources CLOTHING/LOCATION in Chinese. In addition, we should bear in mind that in this dictionary-based study, dictionary-related factors may also play a role in the metonymy distribution, especially for the pattern INDIVIDUAL FOR COLLECTION and the source GENITALS in BODYPART FOR WHOLE.

3.4 Summary This chapter investigated the cross-linguistic variation in metonymic patterns/sources for the target category PERSON between Chinese and English. Metonymic expressions were manually collected from the DCMS and the DCSW for Chinese and the HTOED for English. The prismatic model discussed in Chapter 2 was successfully employed in the metonymy identification in this case study. With the quantitative approach of Fisher’s exact tests followed by qualitative analyses, three main types of cross-linguistic variation, which occur at different granularities of metonymies, were found: variation in metonymic patterns for the general target category PERSON, variation in metonymic patterns for a specific kind of person, and variation in metonymic sources of a certain pattern. Possible driving forces behind the variation were discussed in terms of the following aspects. First, the cognitive and communicative principles for source selection discussed by Radden and Kövecses (1999) function as general cognitive rules for pattern/source choices: a salient attribute or salient aspect of a target tends to be chosen as the source. Cognitive salience varies among different kinds of people, thus strong associations exist between patterns/sources and specific targets of PERSON. Second, I considered the different types of variation against their cultural backgrounds. Although it is a fundamental cognitive mechanism, metonymy is not completely governed by the universal bodily experience. Cross-linguistic variation is to a large extent determined by cultural elements, as illustrated by the study’s findings of the marked influence of paragon, clothing and location notions in Chinese culture and the pervasive conceptualization of INTELLIGENCE

180 | The comparative perspective: Metonymies for PERSON

in terms of the HEAD/BRAIN and the emphasis on individuals’ behavior, psychology or state in English-speaking culture. This study by no means denies the important role of common bodily experience in the universality of metonymy, but it does intend to draw people’s attention to the culture-specific diversities in metonymy between two languages with different cultural backgrounds. Cognitive factors, communication principles and culture elements may jointly shape the metonymic conceptualization of PERSON. As Geeraerts and Grondelaerts (1995: 227) argue, the experiential basis of language and cognition actually includes both cultural and historical “situatedness” of human experience. And culture by definition is a historical notion; it is unstable and subject to diachronic and stylistic variations. Therefore, without a historical study, we cannot really uncover the cultural influence on people’s metonymic conceptualization. This fact is the motivation for the second case study, which explores metonymies in the history of Chinese language and culture.

4 The diachronic perspective: Metonymies for WOMAN The previous chapter has shown that metonymy, as a cognitive mechanism, is also a culturally relevant phenomenon. Geeraerts and Grondelaers (1995: 265) point out that the cultural-institutional nature of cognitive models implies their historical continuity, as cultural institutions carry their history along with them. Only when a diachronic examination has been conducted can we know the features of metonymies in a particular culture. In addition, purely dictionarybased research is methodologically flawed and inevitably fails to show the actual diachronic distribution of metonymies. Therefore, a detailed, bottom-up corpus-based approach is needed to examine the fundamental cognitive nature as well as many aspects of the linguistic nature of metonymy. The technological advances of corpus linguistics offer new possibilities for examining metonymy in real language usage of different historical periods. This chapter presents a corpus-based onomasiological study designed for the purpose of revealing the diachronic variation in the metonymic conceptualization of WOMAN in Chinese. The specific research questions in this case study are as follows:  Do the different subtargets of WOMAN display similar distributions of metonymic patterns across time and genre?  Are there any significant diachronic changes in metonymies for each subtarget of WOMAN in different time periods?  Are there any significant stylistic differences in metonymies for each subtarget of WOMAN between literary and nonliterary discourse?  If variation is found, what are the possible driving forces behind it? In answer to these questions, corpus data were used to reveal the variation in metonymic conceptualizations. This chapter will report on the diachronic case study in four sections. Section 4.1 will describe the methodology adopted to collect the data, quantify the variation and visualize the statistical results. Section 4.2 will focus on the differences in the conceptualization of different subtargets of WOMAN in the history of Chinese language and culture. Section 4.3 will inspect the stylistic variation in metonymy between literary and nonliterary discourse. Finally, Section 4.4 will provide a summary of the chapter, indicating its theoretical and methodological implications.

182 | The diachronic perspective: Metonymies for WOMAN

4.1 Methodology 4.1.1 The Corpus of Historical Chinese The data collection is based on a list of Chinese metonymic expressions for WOMAN72, collected from the previous case study, which contains 287 expressions with feminine targets. These “source expressions” (Stefanowitsch and Gries 2006: 2) were then  against a corpus to examine their distributions in real language usage. The corpus used in this study is the Corpus of Historical Chinese (CHC)73 developed by Peking University. The CHC comprises 17 million characters of classical texts in Chinese, covering the time period from the 11th century BC to the early 20th century and varying genres ranging from fiction, poems and academic prose to religious scriptures and historical works, among others. The corpus also provides the chronological period (i.e. dynasty) and genre information for most texts. For a part of the corpus, it does not give the chronological period74 of the original texts, which mainly fall into two main cases: works that were cowritten or revised by authors from different dynasties, such as You Xue Qiong Lin [the children's knowledge treasury], which was originally written by Chen in the Ming Dynasty and supplemented by Zhou in the Qing Dynasty, and compilations that span various chronological periods, such as the collection Shi San Jing Zhu Shu [notes and commentaries of the Thirteen Classics]. In practice, it is difficult to distinguish the languages from various dynasties in these cases. In this study, I simply coded the chronological time for these cases with the later dynasty. For example, in Yi Li Zhu Shu [notes and commentaries of the Book of Etiquette and Ceremonial], the notes are by Zhen in Han Dynasty and the commentaries are by Jia in the Tang Dynasty (zhu refers to notes on classic texts, and shu is the commentary on the notes). Hence, I coded this text as Tang Dynasty. The reason for this procedure is twofold: first, it is safe in the sense of being conservative in dating language usage, as some linguistic expressions coined in the later dynasty might not have existed in the earlier dynasty. Second, in some cases, linguistic expressions from the earlier dynasty seem to be archaic in

|| 72 The concept WOMAN here covers all feminine targets collected from the DCSW and the DCMS. Note that this is different from the concept A WOMAN in Chapter 3, see Footnote 50. 73 The corpus can be accessed at http://ccl.pku.edu.cn:8080/ccl_corpus/. 74 For example, one category labeled “[\txt] (Readings for primary school\Fireside Chats in the Evening.txt)” has no specific chronological information provided by the CHC.

Methodology | 183

terms of the language usage of the later dynasty. However, as they still appear in the work, they are comprehensible at least for a specific group of readers in the later dynasty. Table 32: Seven conflated time periods in the present study

Time

Dynasty

Chronological period

3U4+

Zhou Dynasty

1046 BC–256 BC

Spring and Autumn Period

770 BC–476 BC

Warring States Period

476 BC–221 BC

Western Han Dynasty

206 BC–9

67)

6RQJ

Eastern Han Dynasty

25–220

Six Dynasties Period

220–581

Sui Dynasty

581–618

Tang Dynasty

618–907

Five Dynasties Period

907–979

Northern Song Dynasty

960–1127

Southern Song Dynasty

1127–1279