182 33 26MB
English Pages 753 Year 2017
T h e Ox f o r d H a n d b o o k o f
P R AG M AT IC S
ii
OXFORD HANDBOOKS IN LINGUISTICS Recently published
THE OXFORD HANDBOOK OF CORPUS PHONOLOGY Edited by Jacques Durand, Ulrike Gut, and Gjert Kristoffersen
THE OXFORD HANDBOOK OF LINGUISTIC FIELDWORK Edited by Nicholas Thieberger
THE OXFORD HANDBOOK OF DERIVATIONAL MORPHOLOGY Edited by Rochelle Lieber and Pavol Štekauer
THE OXFORD HANDBOOK OF HISTORICAL PHONOLOGY Edited by Patrick Honeybone and Joseph Salmons
THE OXFORD HANDBOOK OF LINGUISTIC ANALYSIS Second Edition Edited by Bernd Heine and Heiko Narrog
THE OXFORD HANDBOOK OF THE WORD Edited by John R. Taylor
THE OXFORD HANDBOOK OF INFLECTION Edited by Matthew Baerman
THE OXFORD HANDBOOK OF DEVELOPMENTAL LINGUISTICS Edited by Jeffrey Lidz, William Snyder, and Joe Pater
THE OXFORD HANDBOOK OF LEXICOGRAPHY Edited by Philip Durkin
THE OXFORD HANDBOOK OF NAMES AND NAMING Edited by Carole Hough
THE OXFORD HANDBOOK OF INFORMATION STRUCTURE Edited by Caroline Féry and Shinichiro Ishihara
THE OXFORD HANDBOOK OF MODALITY AND MOOD Edited by Jan Nuyts and Johan van der Auwera
THE OXFORD HANDBOOK OF LANGUAGE AND LAW Edited by Peter M. Tiersma and Lawrence M. Solan
THE OXFORD HANDBOOK OF PRAGMATICS Edited by Yan Huang
For a complete list of Oxford Handbooks in Linguistics please see pp. 713–714.
The Oxford Handbook of
PRAGMATICS Edited by
YAN HUANG
1
iv
3 Great Clarendon Street, Oxford, ox2 6dp, United Kingdom Oxford University Press is a department of the University of Oxford. It furthers the University’s objective of excellence in research, scholarship, and education by publishing worldwide. Oxford is a registered trade mark of Oxford University Press in the UK and in certain other countries © editorial matter and organization Yan Huang 2017 © the chapters their several authors 2017 The moral rights of the authors have been asserted First Edition published in 2017 Impression: 1 All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, without the prior permission in writing of Oxford University Press, or as expressly permitted by law, by licence or under terms agreed with the appropriate reprographics rights organization. Enquiries concerning reproduction outside the scope of the above should be sent to the Rights Department, Oxford University Press, at the address above You must not circulate this work in any other form and you must impose this same condition on any acquirer Published in the United States of America by Oxford University Press 198 Madison Avenue, New York, NY 10016, United States of America British Library Cataloguing in Publication Data Data available Library of Congress Control Number: 2016943266 ISBN 978–0–19–969796–0 Printed in Great Britain by Clays Ltd, St Ives plc Links to third party websites are provided by Oxford in good faith and for information only. Oxford disclaims any responsibility for the materials contained in any third party website referenced in this work.
For my daughter Elizabeth, my wife Lihua, and in loving memory of my parents
vi
Contents
Preface and Acknowledgements List of Symbols and Abbreviations List of Contributors 1. Introduction: What Is Pragmatics? Yan Huang
xi xiii xvii 1
PA RT I S C HO OL S OF T HOU G H T, F O U N DAT ION S , A N D T H E OR I E S 2. Contextualism and Semantic Minimalism Anne Bezuidenhout
21
3. Neo-Gricean Pragmatics Yan Huang
47
4. Relevance Theory Deirdre Wilson
79
5. Formal Pragmatics Reinhard Blutner
101
6. Continental European Perspective View Jef Verschueren
120
7. The Sociological Foundations of Pragmatics Jacob L. Mey
132
PA RT I I C E N T R A L TOP IC S 8. Implicature Yan Huang
155
viii
viii Contents
9. Presupposition and Givenness Bart Geurts
180
10. Speech Acts Stephen C. Levinson
199
11. Deixis and the Interactional Foundations of Reference Jack Sidnell and N. J. Enfield
217
12. Reference Barbara Abbott
240
13. Context Anita Fetzer
259
PA RT I I I M AC RO -P R AG M AT IC S A N D C O G N I T ION 14. Cognitive Pragmatics Bruno G. Bara
279
15. Developmental Pragmatics Pamela R. Rollins
300
16. Experimental Pragmatics Raymond W. Gibbs, Jr
310
17. Computational Pragmatics Harry Bunt
326
18. Clinical Pragmatics Louise Cummings
346
19. Neuropragmatics Brigitte Stemmer
362
PA RT I V M AC RO -P R AG M AT IC S A N D S O C I E T Y /C U LT U R E 20. Politeness and Impoliteness Penelope Brown
383
Contents ix
21. Cross-Cultural and Intercultural Pragmatics Istvan Kecskes
400
22. Interlanguage Pragmatics J. César Félix-Brasdefer
416
23. Conversation Analysis Emanuel A. Schegloff
435
PA RT V I N T E R FAC E S 24. Pragmatics and Semantics Robyn Carston
453
25. Pragmatics and Grammar: More Pragmatics or More Grammar Mira Ariel
473
26. Pragmatics and Morphology: Morphopragmatics Wolfgang U. Dressler and Lavinia Merlini-Barbaresi
493
27. Pragmatics and the Lexicon Laurence R. Horn
511
28. Pragmatics and Prosody Julia Hirschberg
532
29. Pragmatics and Language Change: Historical Pragmatics Andreas H. Jucker
550
30. Pragmatics and Information Structure Gregory Ward, Betty J. Birner, and Elsi Kaiser
567
References Index
591 699
x
Preface and Acknowledgements
The aim of this handbook is to provide a collection of authoritative, comprehensive, thorough, insightful, and yet accessible, state-of-the-art critical surveys of current original research in pragmatics, by a group of the world’s most distinguished scholars working in the field. Unlike many other one-volume handbooks of pragmatics, this one has two distinctive characteristics. The first one is that the research overviews contained in the handbook are made from both the Anglo-American component and the European Continental perspective points of view. Secondly, the handbook surveys the most important sub- branches of (macro-)pragmatics in a systematic way, ranging from clinical through interlanguage to historical pragmatics. There are, however, a few topics that are not covered in the handbook. This is because either the contributors contracted were unable to meet the final submission date (I commissioned chapters on the philosophical foundations of pragmatics for Part I, pragmatics and philosophy for Part V, and a history of pragmatics as an Appendix, and had also to step in to write the chapter on neo-Gricean pragmatics at the last minute), or for various reasons, a suitable author could not be found for a particular subject. Rather than delaying publication further, I have decided to dispatch the handbook in its current form to Oxford University Press, in the hope that these chapters, together with newly commissioned ones will eventually be included in the second edition. The handbook is aimed primarily at scholars, researchers, and graduate and advanced undergraduate students specializing in linguistics, especially pragmatics and semantics. It will also be suitable for advanced undergraduate students taking a linguistics option as part of a language, humanities, or social science degree, and for scholars and researchers in the philosophy of language, cognitive science, artificial intelligence, computer science, informatics, neuroscience, language pathology, anthropology, and sociology. My greatest debt of gratitude is to my stellar set of contributors for their scholarship, patience, and above all, trust in me. I am particularly grateful to Deirdre Wilson for her constant encouragement and support. Thanks to Shuangshuang Chen and Danyang Zheng for their assistance with the references and index. I would also like to thank John Davey, Julia Steer, Vicki Sunter, and William Richards of Oxford University Press for their professionalism in preparing the work for publication. The preparation of the final version of the handbook was partially made while I spent part of my 2015 sabbatical/ research leave at the Faculty of Linguistics, Phonetics, and Philology at the University of Oxford. I am indebted to Martin Maiden for inviting me and to the faculty for providing me with an opportunity to do research in a stimulating and congenial environment.
xii
xii Preface and Acknowledgements Once again, I dedicate this book to my daughter and my wife: without their love, I would not have completed the handbook; and to the memory of my parents: during the dark days of Mao’s ‘Great Proletarian Cultural Revolution’ in China, it was they who, despite being persecuted, tried their best to bring me up, educate me, and teach me how to be a decent person. Yan Huang 黄衍 Auckland and Oxford December 2015
List of Symbols and Abbreviations
Symbols ~ negation +>
conversationally implicate
~ +>
do not conversationally implicate
+>>
conventionally implicate
~ +>>
do not conventionally implicate
⇝ presuppose ~ ⇝
do not presuppose
||-
entail
~ ||-
do not entail
< >
Q or Horn scale
[ ]
I scale
{ }
M scale
K
speaker knows that
* example sentence that follows is syntactically ill-formed or semantically anomalous ?
example utterance that follows is pragmatically anomalous
xiv
xiv List of Symbols and Abbreviations
Abbreviations 1 first person 2 second person 3 third person ACC accusative case AD
Alzheimer’s disease
ADHD
attention deficit hyperactivity disorder
AGR agreement ASD
autistic/autism spectrum disorder
AUX auxiliary CA
conversation analysis
CCSARP
Cross-Cultural Speech Act Realization Pattern
CP
cooperative principle
DA
discourse analysis
DEM demonstrative DET determiner D distance EBD
emotional and behavioural disorder
ERG ergative EXH exhaustivity fMRI
functional magnetic resonance imaging
FTA
face-threatening act
GCI
generalized conversational implicature
I imformativeness ILP
interlanguage pragmatics
L1
first language
L2
second language
LF Logical Form LHD
left-hemisphere damage
LOC locative M manner NEG negative
List of Symbols and Abbreviations xv NOM
nominative/nominalizer
NP noun phrase OT
optimality-theoretic
P power PART particle PCI
particularized conversational implicature
PF
Phonological Form, propositional form
PLI
pragmatic language impairment
Q quantity QUD
question-under-discussion
R
relation, absolute ranking
RHD
right-hemisphere damage
RT relevance theory SLA
second-language acquisition
SLI
specific language impairment
TBI
traumatic brain injury
TCU
turn-constructional unit
ToM
theory of mind
UC
unarticulated constituent
VP verb phrase The abbreviations used in the glosses of the original sources and texts are retained, except for those that have been altered for the sake of uniformity. For abbreviations that are non-conventional, language-specific, and/or idiosyncratic, consult the original examples and texts.
xvi
List of Contributors
Barbara Abbott received a PhD in linguistics from the University of California at Berkeley in 1976, and taught linguistics and philosophy at Michigan State University from 1976 to 2006, where she is Emeritus Professor of Linguistics and Philosophy. She is the author of Reference (Oxford University Press, 2010) as well as numerous articles on topics in semantics, pragmatics, and philosophy of language, including ‘Doing without a partitive constraint’, ‘Water = H2O’, ‘The formal approach to meaning’, ‘Support for a unique theory of definite descriptions’, ‘Presuppositions as nonassertions’, ‘Definiteness and indefiniteness’, and ‘Some remarks on indicative conditionals’. Mira Ariel is Professor of Linguistics at Tel Aviv University. Using a usage-based rather than a logic-based approach to language, her research focuses on the language– cognition interface (accessibility theory) and on the grammar–pragmatics interface (scalar quantifiers and so-called logical connectives). She has published numerous articles, as well as three books: Accessing Noun-Phrase Antecedents (Routledge, 1990), Pragmatics and Grammar (Cambridge University Press, 2008), and Defining Pragmatics (Cambridge University Press, 2010). Bruno G. Bara (MD, with a PhD in Medical Psychology) is Professor of Psychology at the University of Turin. He worked with Philip Johnson-Laird at Cambridge, studying syllogistic inference, and then with John Searle at Berkeley, studying communication acts. In 1993, he founded the Centre for Cognitive Science at the University and Polytechnic of Turin, which has since become an influential school in cognitive pragmatics. His research focuses on the mental processes underlying communication through different methods. In recent years, his main interest has been in social neuroscience. In particular, he has been working on the neural bases of social interaction in neurotypical and clinical populations. He is also a cognitive therapist in the constructivist vein; he founded two schools of cognitive psychotherapy in northern Italy, focusing on increasing the clients’ awareness about themselves through the client/therapist relation. He is the author of Cognitive Science (Psychology Press, 1995) and Cognitive Pragmatics (MIT Press, 2010) Anne Bezuidenhout is Professor of Philosophy and Core Member of Linguistics at the University of South Carolina, where she currently serves as Senior Associate Dean of Arts, Humanities, and Social Sciences. She has published on issues at the semantics– pragmatics interface, as well as in experimental pragmatics. Amongst her current projects are ones on the role of discourse coherence relations in pronoun resolution and
xviii
xviii List of Contributors on the nature of the process of conversational tailoring that interlocutors engage in the course of conversational exchanges. Betty J. Birner received her PhD in linguistics from Northwestern University in 1992. She held a two-year postdoctoral fellowship at University of Pennsylvania’s Institute for Research in Cognitive Science from 1993–1995, and is Professor in the Department of English at Northern Illinois University. Her books include The Discourse Function of Inversion in English (1996), Information Status and Noncanonical Word Order in English (1998, co-authored with Gregory Ward), and Drawing the Boundaries of Meaning: Neo- Gricean Studies in Pragmatics and Semantics in Honour of Laurence R. Horn (2006, co- edited with Gregory Ward). Reinhard Blutner is Emeritus Lecturer in artificial intelligence and cognitive philosophy at the University of Amsterdam. He began his scientific career in theoretical physics and shifted later to artificial intelligence, cognitive psychology, and linguistics. His main research interests lie in the semantics and pragmatics of natural language, but he has contributed to other fields as well, including the psychology of language, neural- symbolic integration, and quantum cognition. He has numerous publications, including an edited volume on optimality theory and pragmatics (Palgrave Macmillan, 2003) and a monograph on optimal communication (CSLI Publications, 2006). Penelope Brown is Senior Researcher Emeritus at the Max Planck Institute for Psycholinguistics in Nijmegen, the Netherlands. She has worked for many years in a Tzeltal Maya community in southern Mexico, on research that broadly addresses relationships between language, culture, and cognition and ranges across spatial language and cognition, conversational structure and inference, the systematics of social interaction, and child language socialization. She is (with Stephen Levinson) the author of Politeness: Some Universals in Language Usage, and editor (with Melissa Bowerman) of Crosslinguistic Perspectives on Argument Structure: Implications for Language Acquisition. She is currently writing two books based on her research in Mexico, one on Tzeltal conversation, the other on spatial language and cognition. Harry Bunt is Professor of Computational Linguistics at Tilburg University in the Netherlands. He started out as a theoretical physicist with a side interest in artificial intelligence, and worked for a number of years at Philips Research. He later moved into AI and language understanding. He published the monograph Mass Terms and Model-Theoretic Semantics (Cambridge University Press, 1985) and the edited volumes Computing Meaning 1–4 (Kluwer/Springer, 1999–2014). His research has been concerned with all aspects of language processing, with a focus on computational semantics and pragmatics, as well as with multimodal interaction, knowledge representation, context modelling and reasoning. With William Black he published the book Abduction, Belief, and Context in Dialogue (Benjamins, 2000). He developed the framework of Dynamic Interpretation Theory for dialogue semantics and pragmatics, including the DIT++ taxonomy of dialogue act types, and is the main author of the ISO 24617-2 standard for dialogue act annotation.
List of Contributors xix Robyn Carston is Professor of Linguistics at University College London and Research Coordinator at the Centre for the Study of Mind in Nature, Oslo. Her main research interests are in pragmatics, semantics, relevance theory, word meaning, and figurative language. She has published a monograph Thoughts and Utterances: The Pragmatics of Explicit Communication (Blackwell, 2002) and is preparing a collection of papers to be published under the title Pragmatics and Semantic Content (Oxford University Press). Louise Cummings is Professor of Linguistics at Nottingham Trent University in the UK. She conducts research in pragmatics and clinical linguistics. She is the author of Pragmatics: A Multidisciplinary Perspective (Edinburgh University Press, 2005), Clinical Linguistics (Edinburgh University Press, 2008), Clinical Pragmatics (Cambridge University Press, 2009), Communication Disorders (Palgrave Macmillan, 2014), Pragmatic Disorders (Springer, 2014), Communication Disorders Workbook (Cambridge University Press, 2014), and Pragmatic and Discourse Disorders: A Workbook (Cambridge University Press, 2015). She has edited The Routledge Pragmatics Encyclopedia (2010) and the Cambridge Handbook of Communication Disorders (2014). She has held Visiting Fellowships in the Department of Philosophy at Harvard University and in the Centre for Research in the Arts, Social Sciences, and Humanities (CRASSH) at Cambridge University. Wolfgang U. Dressler studied linguistics and classical philology in Vienna, Rome, and Paris. He was then briefly a teacher of Greek and Latin, and a Greek epigraphist, and after completing his habilitation taught at the University of Vienna (1968), at UCLA (1970), and at Ohio State University (1970–1971). He was afterwards Professor of Linguistics and Head of Department at the University of Vienna (1971–2008), and is now head of the working group ‘Comparative Psycholinguistics’ at the Department of Linguistics of the University of Vienna and of the Institute for Corpus Linguistics and Text Technology of the Austrian Academy of Sciences. A member of several academies and Dr. h.c. of Paris, Athens, and Poznan, he has worked with varying focus on diachrony, text linguistics, phonology, morphology, pragmatics, aphasia, and language acquisition. N. J. Enfield is Professor and Chair of Linguistics at the University of Sydney, Australia. He is also a research associate at the Max Planck Institute for Psycholinguistics in the Netherlands. He was leader of the European Research Council project ‘Human Sociality and Systems of Language Use’ from 2010 to 2014. His research on language, culture, and cognition, from both micro and macro perspectives, is based on extended fieldwork in mainland South-East Asia, especially Laos. His books include The Utility of Meaning (Oxford University Press, 2015), Natural Causes of Language (Language Science Press, 2014), The Cambridge Handbook of Linguistic Anthropology (Cambridge University Press, 2014, with Jack Sidnell and Paul Kockelman), Relationship Thinking (Oxford University Press, 2013), The Anatomy of Meaning (Cambridge University Press, 2009), A Grammar of Lao (Mouton, 2007), Roots of Human Sociality (Berg, 2006, with Stephen C. Levinson), Linguistic Epidemiology (Routledge, 2003), and Ethnosyntax: Explorations in Grammar and Culture (Oxford University Press, 2002).
xx
xx List of Contributors J. César Félix-Brasdefer is Associate Professor in the Department of Spanish and Portuguese, and also adjunct Associate Professor in the Department of Second Language Studies as well as the Department of Linguistics at Indiana University, Bloomington, USA. His research interests include pragmatics, discourse analysis, cross-cultural and interlanguage pragmatics, pragmatic variation, and (im)politeness theory. In addition to books and edited volumes, Professor Félix-Brasdefer has published numerous research articles and book reviews in a variety of scholarly journals. His most recent book is entitled The Language of Service Encounters: A Pragmatic-Discursive Approach (Cambridge University Press, 2015). Anita Fetzer is Professor of English Linguistics at the University of Augsburg, Germany. She received her PhD from Stuttgart University in 1993 and her habilitation in 2003, and is currently engaged in research projects on follow-ups in political discourse, and on the overt and non-overt realization of discourse relations. Her research interests focus on functional grammar, contrastive analysis, modality and evidentiality, and context. She has had a series of articles published on rejections, context, and political discourse. Her most recent publications are Contexts and Context: Parts Meets Whole (Benjamins, 2011, with Etsuko Oishi), Political Discourse in the Media (Benjamins, 2007, with Gerda Lauerbach), Context and Appropriateness (Benjamins, 2007), and Recontextualizing Context: Grammaticality Meets Appropriateness (Benjamins, 2004). She is the editor of the Pragmatics & Beyond New Series. Bart Geurts is Professor of the Philosophy of Language and Logic at the University of Nijmegen. He has authored and co-authored publications on a wide variety of topics in semantics and pragmatics, including a monograph on presupposition (Elsevier, 1999) and one on quantity implicatures (Cambridge University Press, 2010). Although his core business is semantic and pragmatic theory, he has been involved with experimental research as well, and has contributed to various other fields, including the psychology of language, the psychology of reasoning, and social cognition. Raymond W. Gibbs, Jr, is Distinguished Professor of Psychology at the University of California, Santa Cruz. His research interests focus on embodied cognition, pragmatics, and figurative language. He is the author of several books, including The Poetics of Mind: Figurative Thought, Language and Understanding (1994), Intentions in the Experience of Meaning (1999), Embodiment and Cognitive Science (2006), and (with Herb Colston) Interpreting Figurative Meaning (2012), all published by Cambridge University Press. He is also the editor of the Cambridge Handbook of Metaphor and Thought (2008), and editor of the interdisciplinary journal Metaphor and Symbol. Julia Hirschberg is Percy K. and Vida L. W. Hudson Professor of Computer Science and Chair of the Computer Science Department at Columbia University. She worked at Bell Laboratories and AT&T Laboratories—Research from 1985–2003 as a Member of Technical Staff and a Department Head, creating the Human–Computer Interface Research Department in 1994. She served as editor-in-chief of Computational Linguistics from 1993–2003 and co-editor-in-chief of Speech Communication from 2003–2006.
List of Contributors xxi She served on the Executive Board of the Association for Computational Linguistics (ACL) from 1993–2003, on the Permanent Council of International Conference on Spoken Language Processing (ICSLP) since 1996, and on the board of the International Speech Communication Association (ISCA) from 1999–2007 (as President 2005– 2007); she has also served on the CRA Executive Board (2013–2014). She now serves on the IEEE Speech and Language Processing Technical Committee, the Association for the Advancement of Artificial Intelligence (AAAI) Council, the Executive Board of the North American ACL, and the board of the CRA-W. She has been AAAI Fellow since 1994, ISCA Fellow since 2008, and (founding) ACL Fellow since 2011, and was elected to the American Philosophical Society in 2014. She is a winner of the IEEE James L. Flanagan Speech and Audio Processing Award (2011) and the ISCA Medal for Scientific Achievement (2011). Laurence R. Horn is Professor of Linguistics and Philosophy at Yale University, where he has taught since 1981. His linguistics PhD dissertation (UCLA, 1972, supervised by B. H. Partee) introduced the notion of scalar implicature. He has written over 100 papers and handbook entries on implicature, presupposition, negation, polarity, word meaning, grammatical variation, logic, and the semantics–pragmatics interface; his 1985 paper ‘Metalinguistic negation and pragmatic ambiguity’ was selected to appear in the LSA’s Best of Language compilation. He is the author of A Natural History of Negation (Chicago University Press, 1989; CSLI, 2001) and the editor or co-editor of Negation and Polarity (Oxford University Press, 2000), The Handbook of Pragmatics (Blackwell, 2004), Explorations in Pragmatics (de Gruyter, 2007), The Expression of Negation (de Gruyter, 2010), and Micro-Syntactic Variation in North American English (Oxford University Press, 2014). He is an elected fellow of the Linguistic Society of America and was past editor of the Outstanding Dissertations in Linguistics series for Garland and Routledge. Yan Huang received his PhD in Linguistics at the University of Cambridge. He also holds a DPhil from the University of Oxford. He is Professor of Linguistics at the University of Auckland, and Changjiang Scholar Chair Professor (appointed by the Ministry of Education, China) at Beijing Foreign Studies University. He has previously taught linguistics at the universities of Cambridge, Oxford, and Reading, where he was Professor of Theoretical Linguistics. He has also spent his sabbaticals/research leaves at Yale, Harvard, Cambridge, and Oxford universities, and at a number of top universities in Australia and China. His books include The Syntax and Pragmatics of Anaphora (Cambridge University Press, 1994, reissued in 2007), Anaphora: A Cross- Linguistic Study (Oxford University Press, 2000), Pragmatics (Oxford University Press, 2007), The Oxford Dictionary of Pragmatics (Oxford University Press, 2012), and Pragmatics, second edition (Oxford University Press, 2014). He has also published numerous articles and reviews in leading international journals of linguistics. He is on the editorial board of a number of international linguistics journals and research monograph series. He is the editor of this handbook, and will serve as the editor of a new international journal of pragmatics Brill Research Perspectives on Pragmatics in 2016.
xxii
xxii List of Contributors Andreas H. Jucker is Dean of the Faculty of Arts and Social Sciences and Professor of English Linguistics at the University of Zurich. Previously, he taught at the Justus Liebig University, Giessen. His current research interests include historical pragmatics, politeness theory, speech act theory, and the grammar and history of English. His recent publications include Handbook of Historical Pragmatics (Mouton, 2010, co-edited with Irma Taavitsainen), Communicating Early English Manuscripts (Cambridge University Press, 2011, co-edited with Päivi Pahta), English Historical Pragmatics (Edinburgh University Press, 2013, co-authored with Irma Taavitsainen), and Communities of Practice in the History of English (Benjamins, 2013, co-edited with Joanna Kopaczyk). Elsi Kaiser is Associate Professor in the Department of Linguistics at the University of Southern California. Her research focuses mostly on psycholinguistics, especially sentence processing and issues related to reference resolution, information structure, and the syntax–pragmatics–semantics interface(s). She is especially interested in how different kinds of information interact and are integrated during language processing and what this can tell us about the nature of the mental representations activated during processing. In her own research and in collaborative work, she has conducted work in a range of languages, including English, Finnish, Estonian, Dutch, German, Chinese, and Korean. Her research methodology is interdisciplinary in nature, and incorporates tools and insights from linguistic theory as well as behavioural experiments, including visual- world eye-tracking. Istvan Kecskes is Distinguished Professor of the State University of New York System. His research interest is in pragmatics, second-language acquisition, and bilingualism. He is the President of the American Pragmatics Association (AMPRA) and the Chinese as a Second Language Research Association (CASLAR). His book Foreign Language and Mother Tongue (Erlbaum, 2000), co-authored with Tunde Papp, was the first book that described the effect of the second language on the first language based on longitudinal research. His latest books are Intercultural Pragmatics (Oxford University Press, 2013), Research in Chinese as a Second Language (de Gruyter, 2013), and with Romero- Trillo, Research Trends in Intercultural Pragmatics (de Gruyter, 2013). Professor Kecskes is the founding editor of the journal Intercultural Pragmatics and the Mouton Series in Pragmatics as well as the Chinese–English journal Chinese as a Second Language Research published by Mouton. Stephen C. Levinson is Director of the Max Planck Institute for Psycholinguistics and Professor of Comparative Linguistics at Radboud University, Nijmegen. He is the author of over 270 publications on language and cognition, including the books Politeness (Cambridge University Press, 1987, with Penelope Brown), Pragmatics (Cambridge University Press, 1983), Presumptive Meanings (MIT Press, 2000), Space in Language and Cognition (Cambridge University Press, 2003), and has edited the collections Grammars of Space with D. Wilkins (Cambridge University Press), Language Acquisition and Conceptual Development with M. Bowerman (Cambridge University Press), Culture and Evolution with P. Jaisson (MIT Press), Roots of Sociality with N. Enfield (Berg), and a
List of Contributors xxiii new edition of Language, Thought, and Reality: Selected Writings of Benjamin Lee Whorf with P. Lee (MIT Press). His current research is focused on the cognitive foundations for communication and the relation of language to general cognition. He is Fellow of the British Academy and the Academia Europaea and has received a five–year ERC Advanced Grant in 2011. Lavinia Merlini-Barbaresi is Professor Emerita of English Linguistics and former Director of the PhD School in Linguistics at the University of Pisa. Her scientific interests are in text linguistics, language and discourse varieties, pragmatics, and semiotics, with a special focus on markedness and text complexity and on the pragmatic effects of morphology. She has published in national and international journals and collections, and has authored volumes, among which are Markedness in English Discourse and Morphopragmatics (co-authored with W. U. Dressler). She has also edited the volume Complexity in Language and Text. Jacob L. Mey studied medicine, philosophy, Dutch philology, and general and computational linguistics at the universities of Amsterdam, Nijmegen, Copenhagen, Helsinki, Oslo, and Prague. He acquired a licentiate in philosophy in 1951 and a PhD in linguistics in 1959; he was created Dr. Phil. H.c. in 1993 (Zaragoza) and 2006 (Bucharest). His main interests include the pragmatics of language, especially as it concerns oppressed groups. Among his recent works are a textbook Pragmatics, second edition (2001), and a study on literary pragmatics When Voices Clash (2000); his reflections on language and society are bundled in a Portuguese-language monograph As vozes da sociedade (2006). In 2012, the first part of his memoirs was incorporated in a festschrift bundle entitled Language in Life and a Life in Language. In 1977, he founded (with Hartmut Haberland) the Journal of Pragmatics of which he was editor-in-chief until 2010; in 2008, he edited the Concise Encyclopedia of Pragmatics, second edition (Elsevier). Jacob L. Mey is the author of numerous articles on pragmatics and other linguistic subjects. In 2010, he founded (with Hartmut Haberland and Kerstin Fischer) the journal Pragmatics and Society (Benjamins), of which he remains the chief editor Pamela R. Rollins is Associate Professor of Communication Disorders at the Callier Center for Communication Disorders, the University of Texas at Dallas. Her early research focused on the development of tools to understand pragmatic development in typical children. She used these tools to define learning and communication processes associated with acquisition of social communication and language in typically developing children and children with ASD. Currently her research focus is on developing and evaluating interventions for children at risk for and diagnosed with ASD, and defining learning and communication processes associated with ASD. Professor Rollins works with a robotics designer to create a Robots4Autism social skills curriculum. She is conducting research to understand the effects of robots on social interaction and communication in children with ASD. Emanuel A. Schegloff is Distinguished Professor Emeritus of Sociology and Applied Linguistics, UCLA. He was educated at Harvard College (BA, 1958 magna cum laude)
xxiv
xxiv List of Contributors and at the University of California, Berkeley (MA 1960, PhD 1967). Most of his teaching has been at Columbia University (1965–1972) and the University of California, Los Angeles (1972–2010), with research leaves at Rockefeller University (1972), the Netherlands Institute for Advanced Study in the Social Sciences and Humanities (1978–1979), and the Center for Advanced Study in the Behavioral Sciences, Palo Alto, CA (1998–1999), the last of these supported by a Fellowship from the John Simon Guggenheim Memorial Foundation. He is the author of over 100 publications, including the book Sequence Organization in Interaction: A Primer in Conversation Analysis (Cambridge University Press, 2007) and papers in many interdisciplinary volumes and diverse journals (e.g. American Anthropologist, American Journal of Sociology, Applied Linguistics, Discourse & Society, Discourse Processes, Discourse Studies, Gesture, Journal of Pragmatics, Language, Language and Speech, Language in Society, Linguistics, Research on Language and Social Interaction, Semiotica, Social Problems, Social Research, and Social Psychology Quarterly inter alia). Jack Sidnell is Professor of Anthropology at the University of Toronto. His research focuses on the structures of talk and interaction in a range of settings. In addition to research in the Caribbean and Vietnam, he has examined talk in court and among young children. He is the author of Conversation Analysis: An Introduction (Wiley/ Blackwell, 2010), the editor of Conversation Analysis: Comparative Perspectives (Cambridge University Press, 2009), and the co-editor of The Handbook of Conversation Analysis (Wiley/Blackwell, 2012, with Tanya Stivers), The Cambridge Handbook of Linguistic Anthropology (Cambridge University Press, 2014, with Nick Enfield and Paul Kockelman), and Conversational Repair and Human Understanding (Cambridge University Press, 2013, with Makoto Hayashi and Geoffrey Raymond). Brigitte Stemmer completed her studies in applied linguistics at the University of Bochum and obtained her medical degree at the University of Essen, Germany. After spending several years as a physician and clinical researcher in a neurological acute care and rehabilitation hospital, she was awarded Canada Research Chair in Neuroscience and Neuropragmatics at the Université de Montreal, Canada where she is a full professor and researcher at the Centre de Recherche, Institut universitaire de Geriatrie de Montreal (CRIUGM). She is an associate researcher at the Department of Psychology at Brock University, St. Catharines, Ontario. Jef Verschueren received a PhD in Linguistics from the University of California at Berkeley. After a long career as a researcher for the Flemish Fund for Scientific Research, he is now Professor of Linguistics at the University of Antwerp, Belgium. He is the founder and Secretary General of the International Pragmatics Association, and he directs the IPrA Research Center. His main interests are theory formation in linguistic pragmatics, intercultural and international communication, and language and ideology. In all these areas he has published extensively. Some recent publications include the annually updated Handbook of Pragmatics (John Benjamins, first published in 1995, now also available online), Debating Diversity: Analysing the Discourse
List of Contributors xxv of Tolerance (Routledge, 1998, co- authored with Jan Blommaert), Understanding Pragmatics (Edward Arnold, 1999), and Ideology in Language Use: Pragmatic Guidelines for Empirical Research (Cambridge University Press, 2012). Gregory Ward (1978, UC-Berkeley, BA in Comparative Literature and Linguistics (with honours); 1985, Penn, PhD) is currently Professor of Linguistics, Gender & Sexuality Studies, and Philosophy at Northwestern University, where he has taught since 1986 (and was Chair from 1999 to 2004). Professor Ward’s primary research area is discourse/pragmatics, with specific interests in pragmatic theory, information structure, and reference/anaphora. His scholarship includes over 175 talks, 75 papers, and four books: The Semantics and Pragmatics of Preposing (1996), Information Status and Noncanonical Word Order in English (1998, co-authored with Betty Birner), The Handbook of Pragmatics (2004, co-edited with Laurence R. Horn), and Drawing the Boundaries of Meaning: Neo-Gricean Studies in Pragmatics and Semantics in Honor of Laurence R. Horn (2006, co-edited with Betty J. Birner). Outside Northwestern, he has taught at eight Linguistic Society of America (LSA) Summer Linguistic Institutes. From 1986 to 1998, he was a consultant at AT&T Labs, working on intonational meaning. In 2004–2005, he was a Fellow at the Center for Advanced Study in the Behavioral Sciences, and from 2004 to 2007 he served as Secretary-Treasurer of the LSA. He was also elected Fellow of the LSA in 2009 and was the 2012 recipient of the E. LeRoy Hall Award for Excellence in Teaching in Northwestern’s Weinberg College of Arts and Sciences. Deirdre Wilson is Emeritus Professor of Linguistics at University College London and Research Professor of Philosophy and co-director (with Herman Cappelen) of the Linguistic Agency project at the Centre for the Study of Mind in Nature, University of Oslo. Her main research interests are in communication and theoretical pragmatics: her long-standing collaboration with Dan Sperber (Relevance: Communication and Cognition, Blackwell, 1986, second edition 1995; Meaning and Relevance, Cambridge University Press, 2012) has led to publications on a wide variety of pragmatic topics, from disambiguation and reference resolution to rhetoric, style, and the interpretation of literary works. Her novel Slave of the Passions was published by Picador in 1991 and she has just completed a second.
xxvi
1
Chapter 1
I n troduction What is Pragmatics? Yan Huang
1.1 What Is Pragmatics? Pragmatics is one of the most vibrant and rapidly growing fields in linguistics and the philosophy of language. In recent years, it has also become increasingly a central topic in cognitive science, artificial intelligence, informatics, neuroscience, language pathology, anthropology, and sociology. But what is pragmatics? Pragmatics may initially be broadly defined as in (1). (1) Pragmatics is the study of language use in context. However, though perhaps sufficient for the current purposes, such a definition may be too general and too vague to be of much use. This is because pragmatics is a particularly complex subject with all kinds of disciplinary influences, and few, if any, clear boundaries (see e.g. Levinson 1983: 5–35 and Ariel 2010 for attempts at defining pragmatics).1 In section 1.2, I shall provide two different definitions of pragmatics from two different theoretical points of view.
1
Ariel (2010) coined the metaphor ‘big-tent pragmatics’ to refer to the heterogeneous nature of pragmatics. Under big-tent pragmatics, there are two groups of pragmaticists: what she called ‘problem-solvers’ and ‘border-seekers’. Her own way of classifying pragmatics is to treat it as containing a set of inferences, as opposed to grammar as comprising a set of conventional codes. But such a classification is questionable, to say the least.
2
2 Yan Huang
1.2 Two Main Schools of Thought in Pragmatics Currently, two schools of thought in pragmatics can be identified: the Anglo-American and the European Continental traditions.
1.2.1 The Anglo-American component view Within the Anglo-American conception of linguistics and the philosophy of language, pragmatics may be defined as in (2). (2) Pragmatics is the systematic study of meaning by virtue of, or dependent on, the use of language. The central topics of inquiry include implicature, presupposition, speech acts, deixis, reference, and context, and the division of labour between, and the interaction of, pragmatics and semantics (see also Huang 2007, 2013c, 2014: 2, 2016). This is known as the component view of pragmatics. On this conception, a linguistic theory consists of a number of core components: phonetics, phonology, morphology, syntax, and semantics. Each of the core components has a relatively properly demarcated domain of inquiry. Pragmatics, then, is just another core component placed in the same contrast set within a linguistic theory. By contrast, other ‘hyphenated’ branches of linguistics, such as anthropological, educational, and sociolinguistics, lie outside this set of core components. The component view of pragmatics is to some extent a reflection of the modular conception of the human mind, namely, the claim that the mental architecture of homo sapiens is divided roughly into a central processor and a number of distinctive, specialized mental systems known as modules (e.g. Fodor 1983).2 Two prominent competing theories in the Anglo-American component 2 A particular version of Jerry Fodor’s modularity of mind thesis is the massive modularity of mind thesis. The term ‘massive modularity’ was introduced by Dan Sperber. According to this view, the human mind is largely, if not entirely, composed of modules. Two forms of the massive modularity of mind thesis can then be identified: strong and weak. On the authority of the strong massive modularity of mind thesis, the human mind does not contain any overarching general- purpose mechanism. In other words, every central process is modular. By contrast, the weak massive modularity of mind thesis maintains that while central processes are largely modular, there are also non-modular, general-purpose processes. The massive modularity of mind thesis is not, however, espoused by Fodor himself (e.g. Meini 2010). Furthermore, in the opinion of Gabriel Segal, modularity can be divided into diachronic and synchronic modularity. The former is a cognitive module that follows a genetically and developmentally determined pattern of growth. By comparison, synchronic modularity refers to a module that is static. Finally, there are competence and performance modules. A competence module, also referred to as a ‘Chomskyan’ or ‘information module’, is one that constitutes a system of mental representations. It is said to contain
3
Introduction: What Is Pragmatics? 3 camp are classical and neo-Gricean pragmatic theory (see e.g. Huang, this volume) and relevance theory (see e.g. Wilson, this volume).3
1.2.2 The European Continental perspective view Within the European Continental conception of linguistics, pragmatics is taken to present a functional perspective on all core components and ‘hyphenated’ areas of linguistics and beyond. (3) ‘Pragmatics is a general functional (i.e. cognitive, social, and cultural) perspective on linguistic phenomena in relation to their usage in forms of behaviour. [It] should be seen … as a specific perspective … on whatever phonologists, morphologists, syntacticians, semanticists, psycholinguists, sociolinguists, etc. deal with’ (Verschueren 1999: 7, 11, 1995: 12). This represents the perspective view of pragmatics, namely, the view that pragmatics should be taken as presenting a functional perspective on every aspect of linguistic behaviour. More or less the same can be said of the definition of pragmatics provided within the former Soviet and Eastern European tradition, under which pragmatics (called pragmalinguistics) is in general conceived of as a theory of linguistic communication, including how to influence people through verbal messages, i.e. political propaganda (Prucha 1983). Consequently, within the wider Continental tradition, the empirical orbit of pragmatics has been considerably widened, encompassing not only much that goes under the rubric of non-core branches of linguistics, such as sociolinguistics, psycholinguistics, and discourse analysis, but also some areas that fall in the province of neighbouring social sciences (see e.g. Huang: 2014: 5–7 for a critique of this school of thought, but see Verschueren, this volume, for a dissenting view). linguistic, biological, psychological, physical, and mathematical knowledge. It runs in contrast with a performance module, also termed a ‘computational module’. A performance module is one that functions as a computational mechanism. In other words, it is a device that processes mental representations (e.g. Carston 2010a). 3 In recent years, as a philosophical background to Anglo-A merican component pragmatics, there has also been an ongoing heated debate between contextualism and semantic minimalism in the Anglo-A merican tradition of philosophy of language and semantics and pragmatics. This debate can be traced back at least to the differences between philosophers in the tradition of ideal language philosophy, such as Gottlob Frege, Alfred Tarski, Bertrand Russell, the early Ludwig Wittgenstein, and Rudolf Carnap, and philosophers in the camp of ordinary or natural-language philosophy, like J. L. Austin, H. P. Grice, Peter Strawson, the later Ludwig Wittgenstein, and John Searle. For an overview of contextualism and semantic minimalism and various in-between positions such as (hidden) indexicalism, non-indexical contextualism or semantic relativism, situationism, and (strong) moderate relativism, see e.g. Huang (2013c, 2014: 307–11, 2016) and Bezuidenhout (this volume). See also e.g. Recanati (2005) and Preyer and Peter (2005, 2007).
4
4 Yan Huang However, there has recently been some convergence between the Anglo-American and Continental traditions. On the one hand, important work has been done on micro-pragmatic topics such as implicature, speech acts, and presupposition from a Continental perspective. On the other hand, research within the Anglo-American conception has been extended not only to some core topics in formal syntax such as anaphora and the lexicon in lexical pragmatics but also to certain ‘hyphenated’ domains of linguistics, such as computational, historical, and clinical linguistics, giving rise to computational, historical, and clinical pragmatics (see section 1.3). Furthermore, each side of the Anglo-American–Continental divide complements and has much to learn from the other. Whereas the strength of the Anglo-American school lies mainly in theory and philosophical, cognitive, and formal pragmatics, the Continental camp has much to offer in empirical work (empirical pragmatics) and socio-, cross-, and intercultural and part of interlanguage pragmatics.
1.3 Macro-Pragmatics I move next to what is called macro-pragmatics—the study of the use of language in all aspects. Current topics of inquiry in macro-pragmatics can roughly be divided into three groups: (i) cognitively oriented, (ii) socially and/or culturally oriented, and (iii) those that are not easily or neatly placed in the first two groups.
1.3.1 Group I: Cognitively oriented macro-pragmatics This category includes cognitive pragmatics, psycho-or psycholinguistic pragmatics (including both developmental and experimental pragmatics), computational pragmatics, clinical pragmatics, neuropragmatics, and part of interlanguage pragmatics. Cognitive pragmatics has its roots in the emergence of modern cognitive science— an interdisciplinary amalgam of philosophy, psychology, linguistics, anthropology, computer science, artificial intelligence, and neuroscience—in the 1970s. A typical example of cognitive pragmatics is relevance theory (e.g. Wilson, this volume). Another significant cognitive approach to pragmatics is cognitive pragmatics theory developed by Bruno Bara. Cognitive pragmatics theory offers an explanation of the cognitive processes that are involved in intentional verbal and non-verbal communication. The practitioners of the theory maintain that a ‘partner’ (addressee) in communication establishes the communicative intention of an ‘actor’ (speaker) by identifying the behaviour game that the actor intends him or her to play. Pragmatic phenomena are accounted for in terms of the complexity of the inferential steps that are needed to refer an utterance to a particular behaviour game and the complexity of the underlying mental representations. Cognitive pragmatics theory has been applied to studies of developmental pragmatics in children, the comprehension of pragmatic phenomena
5
Introduction: What Is Pragmatics? 5 in head-injured subjects, and pragmatic decay in subjects with Alzheimer’s disease (e.g. Bara 2010, this volume). In these cases, it overlaps with clinical, neuro-, and developmental pragmatics. Psychopragmatics or psycholinguistic pragmatics is the psycholinguistic study of aspects of language in use and mind. It is primarily concerned with the issue of how human beings acquire, store, produce, and understand the use of language from the vantage point of psychology. Within psychopragmatics, developmental or acquisitional pragmatics studies the empirical development of pragmatic competence in children, utilizing both observation and experiments (e.g. Rollins, this volume). Experimental pragmatics, another subfield of psychopragmatics, which deploys both psycho-and neurolinguistic methods, investigates, through carefully controlled experiments, such important pragmatic issues and theories as scalar implicature, felicity conditions on speech acts, reference, metaphor, neo-Gricean pragmatic theory, and relevance theory. The term ‘experimental pragmatics’ has two senses. In its broad sense, it refers to any investigation by experiment of any phenomenon or issue that is considered to be pragmatic. By contrast, in its narrow sense, the term makes reference to a recent (late 1990s and early 2000s) development in psycholinguistics, pragmatics, and the psychology of reasoning that experimentally investigates a particular set of issues at the interface between pragmatics and semantics. These issues, phenomena, and theories include scalar implicatures, default versus contextual versus structural inference theories, the felicity conditions on speech acts, reference, neo-Gricean pragmatic theory, relevance theory, and children’s pragmatic competence. Methodologies typically adopted in psychology and neuroscience, such as reaction times, eye movements, and event-related potentials, are used in experimental pragmatics (e.g. Noveck and Sperber 2004; Sauerland and Yatsushiro 2009; Meibauer and Steinbach 2011; see also Gibbs Jr, this volume). More recently, the scope of work in experimental pragmatics has been considerably widened. One case in point is concerned with the experimental testing of so-called ‘embedded implicature’ (e.g. Huang 2014: 68–73). The importance of psychopragmatics is that it has a crucial role to play not only in the formulation and development of pragmatic theories but also in the testing and revision of these theories.4 4
Notice the so-called ‘experimental paradox’—a well-k nown dilemma in experimental psycholinguistics including experimental pragmatics. The dilemma is that the more perfect an experiment, the less like the real speech situation it is, and the more likely that subjects of the experiment will produce unnatural responses. On the other hand, the more like the real speech situation the experiment is, the less easy it is for the experimenters to control the external factors that may interfere with the experiment. The consequence of this paradox is that it is almost impossible to design a perfect experiment (e.g. Aitchison 2003). Somewhat related is what a recent editorial in Nature (2015) calls the human ‘cognitive bias’, namely, ‘[t]he human brain’s habit of finding what it wants to find’, which ‘is a key problem for research’. As pointed out by the editorial, ‘One enemy of robust science is our humanity—our appetite for being right, and our tendency to find patterns in noise, to see supporting evidence for what we already believe is true, and to ignore the facts that do not fit.’ See also the other three relevant papers published in the same issue of Nature (vol. 526, no. 7572).
6
6 Yan Huang Computational pragmatics is the systematic study of the relation between utterances and context from an explicitly computational point of view, utilizing computational resources such as annotation standards, algorithms for language generation and interpretation, context models, inference tools, and machine learning methods. This includes the relation between utterances and action, between utterances and discourse, and between utterances and their uttering time, place, and environment. Two sides to the question of how to compute the relation between linguistic and contextual aspects can be identified. On the one hand, given a linguistic expression, one needs to work out how to compute the relevant properties of context. On the other hand, in the case of language generation, the task is to construct a linguistic expression that encodes the contextual information a speaker intends to convey. Given the relevant properties of the context, one needs to work out how to compute the relevant properties of the linguistic expression. This study of the relation between linguistic and contextual aspects requires the building-up of explicit computational representations at either side of the relation. A particularly important topic of inquiry in computational pragmatics is inference. Abduction, the resolution of reference, the generation and interpretation of speech acts, and the production and comprehension of discourse structure, and coherence relations have figured prominently in computational pragmatics (Bunt and Black 2000; Hobbs 2004; Jurafsky 2004; see also Bunt, this volume).5 Clinical pragmatics involves the application of pragmatic concepts, theories, and findings to the assessment, diagnosis, and treatment of pragmatic aspects of language disorders. It studies such pragmatic concepts and phenomena as Grice’s (1989a) cooperative principle and its attendant maxims, implicature, speech acts, inferences, context, non-literal meaning, deixis, and conversation and discourse from a clinical perspective. Pragmatic deficits have been examined in a variety of clinical groups, including children and/or adults with developmental language disorder, autism spectrum disorder, learning disability, left-or right-hemisphere damage of the brain, Alzheimer’s disease, and schizophrenia (Perkins 2007, Cummings 2009, this volume). In so far as most of these clinical groups are defined by an underlying neurological condition, and a large amount of research involves children, clinical pragmatics overlaps to some degree with developmental and neuropragmatics. Neuropragmatics is a recently developed branch of pragmatics that examines the neuro-anatomical basis of language in use. It is concerned with the relationship between the human brain/mind and pragmatics. It investigates how the human brain/ mind uses language, that is, how it produces and comprehends pragmatic phenomena in healthy as well as neurologically impaired language users. The majority of 5
Computational pragmatics is different from cyberpragmatics. Cyberpragmatics refers to a newly emerged research area in which Internet-mediated interactions are analysed mainly from a cognitive pragmatic point of view. A wide variety of interactions on the Internet are dealt with in cyberpragmatics. These include emails, web pages, chat rooms, social networking sites, blogs, 3D virtual worlds, instant messaging, and videoconferencing (e.g. Yus 2011).
7
Introduction: What Is Pragmatics? 7 neuropragmatic research has focused on aspects of pragmatics in adults with identifiable clinical disorders and brain pathology. The brain-damaged populations include patients with left-and right-hemisphere damage, traumatic brain injury, neurodegenerative disorders like Parkinson’s disease and dementia, and schizophrenia (e.g. Stemmer, this volume). This field of inquiry overlaps in particular with clinical and experimental pragmatics. Finally, I come to interlanguage pragmatics. What, then, is an interlanguage? Simply put, an interlanguage is a stage on a continuum within a rule-governed language system that is developed by second or foreign language learners on their path to acquiring the target language. This language system is intermediate between a learner’s native language and his or her target language. It gives rise to the phenomenon of what Slobin (1996: 89) called ‘first language thinking in second language speaking’. Interlanguage pragmatics lies at the interface between pragmatics and second language acquisition. It studies how non-native speakers of a language acquire and develop their ability to understand and produce pragmatic features in a second language, i.e. an interlanguage. Central research topics include pragmatic awareness, pragmatic transfer, the development of pragmatic competence, speech act production and comprehension, and the relationship between second language grammar and pragmatics (e.g. Kasper 2010). The sub-branch of interlangauge pragmatics that investigates the empirical acquisition and development of pragmatic competence in children is called developmental interlanguage pragmatics. The best-studied interlanguage is that developed by speakers of English as a second language. Other interlanguages that have been investigated include Chinese, German, Hebrew, Japanese, and Spanish (e.g. Huang 2013c, 2014).
1.3.2 Group II: Socially and/or culturally oriented macro-pragmatics In the preceding section, I surveyed a number of branches of cognitively oriented macro-pragmatics; in this section, I turn to the second group of branches of macro- pragmatics. This group includes mainly sociopragmatics, cultural, cross-, and intercultural pragmatics, and part of interlanguage pragmatics. Institutional, interpersonal, postcolonial, and variational pragmatics, and conversation analysis also belong to this category. Sitting at the interface between sociolinguistics and pragmatics, sociopragmatics studies the use of language in relation to society. One topic that has long been the focus of sociopragmatic research is politeness. Politeness, broadly defined so as to encompass both polite friendliness and polite formality, is concerned with any behaviour including verbal behaviour of an interlocutor to constitute and maintain his or her own face and that of the people he or she is interacting with. As pointed out by Brown (this volume), different aspects of this behaviour are captured by terms such as ‘manners’, ‘courtesy’, ‘tact’, ‘deference’, ‘sensibility’, ‘poise’, ‘rapport’, ‘urbanity’, ‘civility’,
8
8 Yan Huang and ‘graciousness’. More recently, a distinction has been introduced by Watts, Ide, and Ehlich (2005) between first-and second-order politeness. By first-order politeness or politeness 1 is meant the common-sense notion of normative politeness, that is, a judgement about whether a particular behaviour is polite or not in keeping with the norms of a sociocultural or speech community, made by lay members of that community. In contrast, second-order politeness or politeness 2 refers to a scientific concept of politeness, that is, an abstract, theoretical construct defined within a theory of politeness and impoliteness. Furthermore, it should be pointed out that there is a dynamic trade-off between the two notions of politeness. For example, second-order politeness, which is informed by first-order politeness, is a concept that is more inclusive than first-order politeness. Consequently, first-and second-order politeness should be studied hand in hand rather than in isolation. Finally, this dualistic distinction between an everyday and a technical sense of a notion can be applied to other concepts like face and impoliteness in the study of politeness and impoliteness (see also e.g. Terkourafi 2012; Huang 2013c, 2014). Currently, there are a variety of theoretical accounts of politeness. These include (i) the ‘social norm’ model, (ii) the ‘conversational maxim’ model, (iii) the ‘face-saving’ model, (iv) the ‘conversational contract’ model, (v) the ‘social practice’ model, and (vi) the ‘discursive’ or ‘postmodern’ model. Of these frameworks, Brown and Levinson’s (1987) classic ‘face-saving’ model—generally considered to have inaugurated the field of modern politeness and impoliteness study—remains the most influential one (see e.g. Huang 2014: 144–149 for a detailed overview of the model). On the other hand, interest in impoliteness has only surged recently, with 1998 being dubbed ‘the Year of Impoliteness’. By impoliteness is meant any face-aggravating behaviour relevant to a particular context. For some scholars, impoliteness has to be intentional (on the part of the speaker) and has to be perceived or constructed as intentional (on the part of the addressee). For others, intentions play no part in impoliteness. If intentions and recognition of intentions are involved, then rudeness rather than impoliteness occurs (e.g. Bousfield 2008; Terkourafi 2012). In Culpeper’s (2011) work, impoliteness has been classified into three types: (i) affective, (ii) coercive, and (iii) entertaining. Regarding the study of impoliteness, one of the major criticisms levelled against Brown and Levinson’s classic face-saving model of politeness is that it marginalizes impoliteness for the reason that impoliteness is seen as a phenomenon and/or concept that is related not to face itself but to the absence of it. Eelen (2001), for example, concluded that traditional theories of politeness have three main problems. First, they are conceptually biased towards the politeness end of the politeness–impoliteness distinction. Secondly, they conceptualize politeness and impoliteness as opposites. And thirdly, their conceptualizations of impoliteness are speaker-based, focusing largely on utterance production (see also Garcés-Conejos Blitvich 2010d). Consequently, in recent years, there has been an explosion of work in politeness and impoliteness research that seeks to theorize impoliteness in its own right rather than treating it merely as the reverse of politeness. On Garcés-Conejos Blitvich’s (2010d) view, current
9
Introduction: What Is Pragmatics? 9 theoretical approaches to impoliteness, which are broadly social constructionist in nature, can be divided into three categories: (i) the impoliteness 1 or the discursive/postmodern model, (ii) the impoliteness 2 or the universalist/context-sensitive model, and (iii) the blended genre model. Notice further that within the second category there are a number of analyses, such as (i) the frame-based view, (ii) the face-constituting theory, and (iii) work by Bousfield (2008) and Culpeper (2011). Whereas much progress has been made, research on impoliteness is still in its infancy. There are still many questions that need an answer. These include: (i) what is the best theoretical and methodological framework for the study of impoliteness, (ii) what is the relationship between impoliteness and politeness, and (iii) what is the role played by a speaker’s intention in impoliteness. Other topics that have attracted attention in sociopragmatics include social deixis, social conventions on the performance of speech acts, and social factors which constrain language in use, such as the overriding of conversational implicature by the Malagasy taboo on exact identification (e.g. Huang 2014: 42–43). From a macro point of view, the hand of societal pragmatics can be detected in any area that pertains in any way to society, dealing with topics as diverse as language in education, pragmatics and social struggle, and what is called critical pragmatics. Critical pragmatics refers to the work done in sociopragmatics that follows the tradition of critical linguistics, in particular critical discourse analysis. As in critical discourse analysis, in critical pragmatics, great emphasis is put on the relationship between language and social power and between language and ideology.6 Institutional pragmatics refers to an area of research in pragmatics which investigates the use of language in social institutions and institutionalized contexts, such as courtroom interaction, job interviews, and police interrogation. Cultural pragmatics, sometimes also known as anthropological or ethnographic pragmatics, is the systematic study of language use and its place in the functioning of human communities and institutions from a cultural or anthropological view, especially but not exclusively focusing on non-Western cultures. It overlaps with the ethnography of communication and ethnography of speaking. A particular variety of cultural pragmatics is ethnopragmatics. Ethnopragmatics is an approach to language in use that is semantically grounded in natural semantic metalanguage developed by Anna Wierzbicka and her associatess. Utilizing cultural scripts and semantic or reductive paraphrase explications as analytical tools, practioners of ethnopragmatics aim to find out more about speech practices and language use of particular, local cultures, contextualized and understood in terms of the beliefs, norms, and values of 6
Note that the term ‘critical pragmatics’ has a totally different sense in the philosophy of language and formal pragmatics. It is the term employed by John Perry and Kepa Korta to refer to the philosophical position that takes the contents of an utterance as central and critical to both pragmatics and semantics. According to critical pragmatics in its philosophical sense, language is a way of doing things with words. Meanings of linguistic expressions and contents of utterances derive ultimately from intentions. Language combines with other factors to allow human beings to achieve communicative goals (e.g. Korta and Perry 2011).
10
10 Yan Huang speakers themselves. In other words, the emphasis of ethnopragmatics is on culturally anchored analyses and explanations, thus rejecting what ethnopragmaticists have labelled ‘universal(ist) pragmatics’, namely, any pragmatic theory that views human communication as governed largely by a rich inventory of universal pragmatic principles with variations between cultures being accounted for in terms of local adjustments to and local construals of these universals (e.g. Goddard 2006). Somewhat similar to ethnopragmatics is ethnographic pragmatics defined in its narrow sense. It refers to the ethnographically oriented approach to context-sensitive language use associated particularly with the work of Michael Silverstein and his students. Research conducted in ethnographic pragmatics has concentrated on non- Western cultures, societies, and languages. A third variety of cultural pragmatics is emancipatory pragmatics. A recently emerged research framework, emancipatory pragmatics attempts to free the study of language in use from the confines of the theoretical and methodological orthodoxies grounded in the dominant thought and practice derived from Anglo-American and European languages and ways of speaking, with the attendant premises of individualism, rationality, and market economy, thus the term ‘emancipatory’. The focus of emancipatory pragmatics is also placed on non-Western languages and ways of speaking and on describing a language and/or culture strictly in its own terms (Hanks, Ide, and Katagiri 2009). Somewhat overlapping with socio-and cultural pragmatics is interpersonal pragmatics. Interpersonal pragmatics is a research arena that concentrates on the interpersonal and relational aspects of language in use, especially how interlocutors utilize language to establish and maintain social relations, and how interactions between interlocutors both affect and are affected by their own and others’ understanding of culture and society (e.g. Locher and Graham 2010). Whereas cross-cultural pragmatics presents a systematic comparison of different cultures on the basis of studying aspects of language use, intercultural pragmatics is concerned with the interaction between speakers from different cultures, speaking different languages (e.g. Kecskes 2013, this volume). Since the 1980s, a principal concern of cross-cultural pragmatics has been the issue of how particular kinds of speech acts, especially such face-threatening acts as requests, apologies, and complaints, are realized across different cultures and languages. One of the most influential investigations is the large-scale Cross-Cultural Speech Act Realization Patterns Project carried out in the 1980s. In this project, the realization patterns of requesting and apologizing in German, Hebrew, Danish, Canadian French, Argentinean Spanish, and British, American, and Australian English were compared and contrasted (e.g. Blum-Kulka et al. 1989). Since then, strategies for the performance of a variety of face-threatening acts in a much wider range of languages have been examined. These languages include Catalan, Chinese, Danish, Dutch, German, Greek, Hebrew, Japanese, Javanese, Polish, Russian, Thai, Turkish, four varieties of English (British, American, Australian, and New Zealand), two varieties of French (Canadian and French), and eight varieties of Spanish (Argentinean, Ecuadorian, Mexican, Peninsular, Peruvian, Puerto Rican,
11
Introduction: What Is Pragmatics? 11 Uruguayan, and Venezuelan). As a result of these studies, it has now been established that there is indeed extensive cross-cultural/linguistic variation in directness/indirectness in the expression of speech acts, especially in face-threatening acts, and that these differences are generally associated with the different means that different languages utilize to realize speech acts. These findings have undoubtedly contributed to our greater understanding of cross-cultural/linguistic similarities and differences in face-redressive strategies for face-threatening acts (e.g. Huang 2014). A sub-branch of cross-, intercultural, and/or interlanguage pragmatics is postcolonial pragmatics, which studies the use of the language of the colonizers in a postcolonial society or postcolonial societies. In a postcolonial society, a second (as opposed to a foreign) language is sometimes used in interaction, as in the use of English in contemporary India (e.g. Anchimbe and Janney 2011). Another recently emerged branch of pragmatics that has a close affinity with socio- and cross-and/or intercultural pragmatics is variational pragmatics. It endeavours to study and determine the influence or impact of macro-social factors such as region, social class, ethnicity, gender, and age, and the interplay of these factors on language use, especially pragmatic variation, in interaction. Construed thus, variational pragmatics also represents a research domain at the intersection of pragmatics and sociolinguistics, in particular dialectology (e.g. Barron and Schneider 2009). The exploration of speech acts has been extended to interlanguage pragmatics (e.g. Kasper and Blum-Kulke 1993a; Kasper 2010; Trosborg 2010). Of these studies, some have investigated how a particular type of speech act is performed by non-native speakers in a given interlanguage; others have compared and contrasted the similarities and differences in the realization patterns of given speech acts between native and non-native speakers in a particular (inter)language. Mention should be made of conversation(al) analysis (CA), sometimes also called ‘conversation(al) pragmatics’. Since Levinson (1983), CA has become a branch of macro-pragmatics. Grown out of a breakaway group of sociologists known as ethnomethodologists within micro-sociology, CA represents an empirical, procedural, and inductive approach to the analysis of (audio and/or video recordings of) naturally occurring, spontaneous conversations or ‘talks in (face-to-face) interaction’. It is concerned with the discovery and description of the methods and procedures that participants employ systematically to display their understanding of the structure of naturally occurring, spontaneous conversations in face-to-face interaction. In conversation, there are rules governing sequential organization, such as the turn-taking system, the formulation of adjacency pairs, and the mechanism for opening or closing a conversation. There are also norms regulating participation in a conversation, such as those for how to hold the ‘floor’, how to interrupt, and how to remain silent. Other interesting structural devices of conversation include the preference organization, the pre-sequence system, and the repair mechanism (e.g. Sacks 1992; Sidnell 2011; Scheloff, this volume). Given that conversation is the most important spoken manifestation of language, CA has to be closely linked to prosodic pragmatics—a study of how prosody-like intonation can affect the interpretation of a variety of linguistic
12
12 Yan Huang phenomena in relation to context (e.g. Hirschberg 2004, this volume). Furthermore, since rules, norms, and regulations for conversational interaction may vary from culture to culture, society to society, and language to language, CA may overlap with the ethnography of speaking and cross-cultural pragmatics. For some scholars, it is opposed to discourse analysis (DA). CA can further be divided into pure CA and applied CA. By pure CA is meant the type of CA that collects data from naturally occurring conversations. In other words, in pure CA the data is not arranged or provoked by the analyst, as in a psycholinguistic experiment or a sociolinguistic interview. Used in contrast to pure CA, ‘applied CA’ is a term that is employed with reference to the type of CA that studies specific types of conversational situation.
1.3.3 Group III I turn next to a group of branches and research areas of macro-pragmatics that are not easily and/or neatly placed in the first two categories. Historical pragmatics is a branch of macro-pragmatics that came to light in the 1990s. It is concerned with the investigation of language change between two given points in time in individual languages and in language generally from a pragmatic perspective. There are two main research trends that correspond roughly to the distinction between ‘external’ and ‘internal’ language change. The first, ‘external’ research strand is called ‘pragmaphilology’. Pragmaphilology represents primarily a ‘macro approach’ to the study of the pragmatics of historical texts at a particular point of time. The focus is on the wider changing social and cognitive contexts of the texts in which pragmatic change occurs. It is closely related to ‘historical discourse analysis proper’ in historical discourse analysis. The second, ‘internal’ research trend is diachronic pragmatics in its narrow sense. Diachronic pragmatics in this sense represents a ‘micro approach’ to change in pragmatic phenomena over time, concentrating on the interface between a linguistic structure and its communicative use across different historical stages of the same language. Furthermore, a methodological distinction is made between the ‘form-to-function’ and ‘function-to-form’ modes. The former, called ‘pragmalinguistic diachronic pragmatics’, is semasiological and the emphasis is on how a particular linguistic form has undergone functional changes; the latter, termed ‘sociopragmatic diachronic pragmatics’, is onomasiological and the focus is on how a particular pragmatic function has changed the form it uses. Diachronic pragmatics in the sense being described here is closely related to ‘diachronically oriented discourse analysis’ in historical discourse analysis. Since the boundary between pragmaphilology and diachronic pragmatics is sometimes not clear-cut, an intermediate category, dubbed ‘diachronic pragmaphilology’, has also been proposed. Furthermore, in addition to the two main approaches, there is a third research strand, which is labelled ‘pragma-historical linguistics’. Given that textual data is heavily used in both historical pragmatics and historical discourse analysis, there is a considerable overlap between the two fields. At the early stage of its development, historical pragmatics was
13
Introduction: What Is Pragmatics? 13 called ‘new philology’ or ‘diachronic textlinguistics’ (e.g. Traugott 2004a; Jucker, this volume; Jucker and Taavitsainen 2010). Next, historical sociopragmatics involves the interaction between historical pragmatics and sociopragmatics. According to some scholars, historical sociopragmatics is more closely related to the pragmaphilology research trend in historical pragmatics. It constitutes a systematic study of interaction between aspects of social context and particular historical language uses that give rise to pragmatic meaning. Historical sociopragmatics can be either synchronic or diachronic. Synchronic historical sociopragmatics studies how language use shapes and is shaped by social context at a certain moment of time in the past. By contrast, diachronic historical sociopragmatics traces how changes in language use shape social context, changes in social context shape language use, and/or changes take place in the relationship between language use and social context (e.g. Culpeper 2009/2011). Directly opposed to historical pragmatics is synchronic pragmatics. Synchronic pragmatics is a subfield of pragmatics that studies language use in general or in a particular language as it is, or was, at a particular point in time. In other words, synchronic pragmatics is concerned with the pragmatics of what Ferdinand de Saussure called an ‘état de langue’: that is, the pragmatics of the state of language at a particular point in time, regardless of its previous or subsequent history. The term ‘applied pragmatics’ has two senses. In its broad sense, applied pragmatics makes reference to any application of the concepts and findings of theoretical pragmatics to practical tasks such as the diagnosis, assessment, and treatment of pragmatic disorders, human–computer interaction, and the teaching and learning of a second and/or foreign language. In the last connection, the field is often called ‘second and foreign language (L2) pragmatics’. ‘Second and foreign language pragmatics’ is a term that is interchangeable with applied pragmatics in its narrow sense. It is part of instructional pragmatics, namely, pragmatics that is concerned with how to teach and learn pragmatics in language, especially second and/or foreign language, instruction (e.g. Ishihara and Cohen 2010). Applied pragmatics should not be confused with applying pragmatics. The latter is a term that is used within the Continental tradition of pragmatics for any dynamic, user-oriented, problem-solving activity employing pragmatic knowledge in a real-world context, especially in the arena of the social struggle. Described thus, applying pragmatics has an overlap with critical pragmatics in its sociological sense. A corpus is a systematic collection of naturally occurring spoken or written language or a variety of such a language, which can be searchable online. When it is accessible on a computer, it is called a computer corpus or corpora. By corpus pragmatics is meant the investigation of language use on the basis of the analysis of corpora. Corpus pragmatics forms part of empirical pragmatics. It can be divided into two types: (i) corpus- based and (ii) corpus-driven. In the former, researchers approach the corpora with a set of assumptions and expected findings. By comparison, the latter investigates linguistic forms and pragmatic functions that emerge from the corpora in order to discover things that have not been recognized. Much of the current research in corpus
14
14 Yan Huang pragmatics is corpus-based rather than corpus-driven. From a methodological point of view, corpus pragmatics can be either form-based (that is, it takes a linguistic structure as its starting point and examines the range of pragmatic functions the form serves in a corpus) or function-based (that is, it takes a particular pragmatic function as a point of departure and studies how such a function is actually realized linguistically). Finally, corpus-based or driven research in pragmatics can be either qualitative (treating corpora primarily as a source of natural data) or quantitative (studying patterns of frequency, distribution, and collocation using statistical techniques) (e.g. Andersen 2011; Rühlemann 2011). Literary pragmatics can be best described as covering an area of research rather than a well-defined unified theory. It represents a domain at the intersection of pragmatics, literary theory, and the philosophy of literature. It is the study of the use of linguistic forms in a literary text and the relationship between author, text, and reader in a sociocultural context from a pragmatic perspective, focusing on the question of what and how a literary text communicates. Two complementary aspects of literary pragmatics can be identified. On the one hand, how can the insights of pragmatic theories be employed for the study of literature, and on the other, how can the insights of literary pragmatics contribute to general pragmatic theories? Literary pragmatics can further be divided into two sub-branches: formalist and historical. Formalist literary pragmatics seeks to characterize literariness in terms of the pragmatic properties of literary texts, concentrating on formal analyses which are based on formal systems or pragmatic processes. Key research themes include speech acts in literary communication and free indirect discourse. In contrast with formalist literary pragmatics is historical literary pragmatics. Interdisciplinary in nature, historical literary pragmatics places an emphasis on the interconnections between literary studies, history studies, sociocultural studies, and pragmatic studies. For example, it uses the insights of Brown and Levinson’s (1987) face-saving model of politeness to characterize the relationship between language users, i.e. writers and readers, in a literary context (e.g. Pilkington 2010). Next, somewhat related to literary pragmatics is pragmatic stylistics or pragmastylistics. Pragmatic stylistics refers to the application of the findings and methodologies of theoretical pragmatics to the study of the concept of style in language, that is, systematic variations in usage in written or spoken language including those in literary texts among individual writers, genres, and periods (e.g. Black 2006). Originating in part from the work of J. L. Austin, legal pragmatics is concerned mainly with the study of legal documents and spoken legal discourse in the courtroom from a pragmatic point of view. Pragmatic features in written legal texts and spoken legal discourses that have been analysed include speech acts such as legal performatives, presuppositions, turn-taking, question–answer adjacency pairs, and silence. The sociopragmatic concepts of power and politeness and impoliteness have also been used in these studies (e.g. Kurzon 2010). Finally, feminist pragmatics represents an approach to the study of gender and language in use, incorporating insights from both feminism and pragmatics. Within this approach, it is assumed that on the one hand, if pragmatics is to provide a theoretical
15
Introduction: What Is Pragmatics? 15 framework for the investigation of gender and use of language, it has to be informed by the findings of feminist scholarship. On the other hand, pragmatics can inform feminist research on gender and language in a wide range of contexts (e.g. Christie 2000).
1.4 Organization and Content of the Handbook The remainder of this handbook is divided into five parts. Part I is concerned with schools of thought, foundations, and theories. In Chapter 2, Anne Bezuidenhout provides an overview of the recent progress in the debate between contextualism and semantic minimalism from three perspectives: (i) cognitive architecture, (ii) formal semantics, and (iii) conceptual analysis. Chapter 3 by Yan Huang gives a state-of- the-art survey of classical and especially neo-Gricean pragmatics, focusing on the bipartite model put forward by Laurence Horn and trinitarian model advanced by Stephen Levinson. The contribution assesses the role neo-Gricean pragmatics plays in effecting a radical simplification of the lexicon, semantics, and syntax in linguistic theory. Relevance theory is the topic of Chapter 4, in which Deirdre Wilson—one of its co-founders—provides an authoritative assessment of the current state of play of the theory and points out some new directions for research. Chapter 5 deals with formal pragmatics. In this chapter, Reinhard Blutner discusses three formal pragmatic frameworks: (i) optimality-theoretic, (ii) game-theoretic, and (iii) decision- theoretical pragmatics. Next comes Jef Verschueren’s contribution (Chapter 6) on the Continental European perspective view of pragmatics. While questioning the accuracy of the contrast between an Anglo-American component view and a Continental European perspective view, the author considers the contrast between Western-based conceptualizations of language use and views that are rooted in non-Western cultures and societies to be more important. Finally, Chapter 7 by Jacob Mey discusses the social foundation of pragmatics in a slightly personal way, outlining an emancipatory pragmatics. Part II deals with some central topics in pragmatics. Chapter 8 by Yan Huang examines implicature. The concept of implicature (both conversational and conventional) has its origin in the work of H. P. Grice (1989a). Since its inception, the notion of conversational implicature has become one of the single most important pragmatic ideas in linguistics and the philosophy of language. It has spurred numerous new concepts such as explicature (Sperber and Wilson 1986, 1995; Carston 2002), the ‘pragmatically enriched said’ (Recanati 2004a, 2010), and impliciture (Bach 2004, 2012) in various neo-and post-Gricean enterprises (but see Levinson 2000 and Huang 2007, 2014 for a dissenting view). Presupposition and givenness is the theme of Chapter 9. Presupposition is a proposition whose truth is taken for granted in the utterance of a sentence. The main function of presupposition is to act as a precondition of some
16
16 Yan Huang sort for the appropriate use of the sentence. This background assumption remains equally valid when the sentence that contains it is negated. In this chapter, Bart Geurts provides a survey of two theories of the givenness of presupposition. Chapter 10 by Stephen Levinson takes up speech acts. This notion, introduced by J. L. Austin (1962a), refers to the uttering of a linguistic expression, the function of which is not just to say things but actively to do things or to perform actions as well. Speech act theory was established by Austin and after his death, refined, systematized, and advanced by John Searle. It has since remained another foundation stone of pragmatics. Levinson’s contribution reviews some core issues in speech act theory, focusing on the role played by conversation structure, probabilistic linguistic cues, and plan or sequence inference in speech act recognition, and the centrality of deep recursive structures in speech act sequences in conversation. Chapter 11 by Jack Sidnell and N. J. Enfield is devoted to deixis. Deixis, or indexicality in the philosophy of language, is the phenomenon whereby features of context of utterance or speech event are encoded by lexical and/ or grammatical means in a language. Based on an analysis of deixis in both its most basic and elaborated forms, Sidnell and Enfield stress that the interactional foundation of all references including deixis involves directing the attention of others. Next, in Chapter 12, Barbara Abbott discusses reference. Reference is a three-place relation that involves speakers, linguistic expressions, and the entities the linguistic expressions stand for in the external world or in some mental representation. In other words, referring is an act of a speaker picking out a particular entity, denoted by the linguistic expression, in the outside world. It is performed through the speaker’s utterance of that linguistic expression on some particular occasion of use. Looked at this way, reference is largely a context-dependent aspect of utterance meaning and it therefore falls largely within the domain of pragmatics (e.g. Huang 2014). In this chapter, Abbott examines the way in which we use referring expressions to refer, what it is that a speaker is referring to, a speaker’s choice of referring expressions, and what are the factors that play a role in the addressee’s interpreting intended referents (see also Abbott 2010). Finally, Chapter 13 by Anita Fetzer on context ends this part. Context is one of the notions that is widely used in the pragmatics literature, but to which it is very difficult to give a precise definition. This chapter comments on some fundamental rethinking of this important notion in linguistics and its related subjects. In this handbook, I have divided macro-pragmatics into two groups: cognitively oriented and socially and/or culturally oriented. Cognitively oriented macro-pragmatics is the topic of Part III. In this part, six branches are covered: Chapter 14 on cognitive pragmatics by Bruno Bara, Chapter 15 on developmental pragmatics by Pamela Rollins, Chapter 16 on experimental pragmatics by Raymond Gibbs Jr, Chapter 17 on computational pragmatics by Harry Bunt, Chapter 18 on clinical pragmatics by Louise Cummings, and finally, Chapter 19 on neuropragmatics by Brigitte Stemmer. Part IV proceeds to look at socially and/or culturally oriented macro-pragmatics. Politeness and impoliteness is the topic of Chapter 20 by Penelope Brown—one of the co-founders of the face-saving model of politeness; cross-and intercultural pragmatics the topic of Chapter 21 by Istvan Kecskes; interlanguage pragmatics the topic
17
Introduction: What Is Pragmatics? 17 of Chapter 22 by César Félix-Brasdefer; and finally, conversation analysis (CA) the topic of Chapter 23 by Emanuel Schegloff—one of the pioneers of this branch of macro-pragmatics. The final part of this handbook, Part V, is concerned with pragmatics and its interfaces. The part begins with Robyn Carston’s discussion of the relationship between pragmatics and semantics in Chapter 24. Semantics and pragmatics are the two major branches of linguistics devoted to the study of meaning in language. This much is largely agreed upon. However, what constitutes the domain of semantics and that of pragmatics? Can semantics and pragmatics be distinguished in a principled way? Are they autonomous or do they overlap with each other? How and to what extent do they interact with each other? These are some of the questions that have puzzled, and are still puzzling, linguists and philosophers of language. In this chapter, the author offers a cognitive-scientific or relevance-theoretic approach to communicators’ pragmatic interpretative ability, and argues for the logical and temporal priority of pragmatics from three viewpoints: communicative, developmental, and evolutionary. The division of labour between pragmatics and grammar is the topic of Chapter 25 by Mira Ariel, who is of the view that the pragmatics–grammar divide should be drawn along a code versus inference distinction. In Chapter 26, Wolfgang Dressler and Lavinia Merlini-Barbaresi are concerned with morphopragmatics. They provide an account of the relationship between pragmatics and morphology on the basis of two major theoretical premises. First, pragmatics is not a secondary meaning derived from semantics. On the contrary, there is a priority of pragmatics over semantics. Second, morphology is capable of a direct interface with pragmatics, but not mediated through its semantics. Thus, certain morphological patterns may yield autonomous pragmatic meanings, independently of their denotative power. Next, Laurence Horn’s contribution in Chapter 27 outlines a neo-Gricean lexical pragmatic theory. It motivates a (Q- principle-based) constraint on lexicalization, reviews the role played by his R-principle (Horn 2004) in giving rise to the division of pragmatic labour, syntagmatic reduction, narrowing of meaning, euphemism, and negative strengthening, and presents pragmatic motivation for the lexical clone, un-noun, and un-verb constructions, and for the complementary avoid synonymy and avoid homonymy principles (see also Huang 2009, 2015a, b). In Chapter 28, Julia Hirschberg considers the relationship between pragmatics and prosody. She reviews a substantial amount of research on prosodic variation and pragmatic meaning in linguistics, computational linguistics, and psycholinguistics, focusing on the pragmatic influence of prosody on the interpretation of syntactic, semantic, and discourse phenomena. Next comes Chapter 29, in which Andreas Jucker takes a look at language change from a pragmatic point of view. He provides an analysis of the process of grammaticalization and pragmaticalization, and presents two case studies of the development of specific pragmatic entities. Finally, we come to Chapter 30 on pragmatics and information structure by Gregory Ward, Betty Birner, and Elsi Kaiser. In this chapter, utilizing a wide array of both corpus and experimental data, the contributors provide an insightful overview of the current research on information structure.
18
18 Yan Huang
Acknowledgements Part of the material contained in this chapter is drawn from Huang (2013c and 2016). I am grateful to Piotr Cap and Keith Allan for their comments on related material.
19
Pa rt I
S C HO OL S OF T HOUGH T, FOU N DAT IONS , A N D T H E OR I E S
20
21
Chapter 2
Con textua lism a n d Sem a n tic M i n i m a lism Anne Bezuidenhout
2.1 Introduction The recent debate between so-called contextualists and semantic minimalists about meaning/content is one that probably matters most to philosophers of language, even though the debate is not solely a philosophical one. The debate can be cast in several ways, and failure to appreciate these different ways of casting the debate may be one source of the apparent disagreements between the warring parties. There are at least three ways in which to approach the debate. Firstly, the debate can be thought of as one about the organization of the human language faculty and about how and when semantic and pragmatic mental resources are used in the course of ordinary conversational exchanges—or in verbal communication more broadly construed to include written as well as spoken communication. This is a debate about cognitive architecture and the processing it supports; and it relies on theories and methodologies from the psychology of language and psycholinguistics. These are largely empirical issues, although philosophers can have a valuable role to play in testing and challenging the conceptual foundations of the empirical frameworks and theories offered by psycholinguists and psychologists of language. A second way of casting the debate is as one about the logic of natural languages and about how to incorporate context sensitivity into a formal, compositional model of natural-language sentence-level meaning. That is, if we think that it is a valuable aim to produce a model-theoretic account for a natural language such as English that is capable of generating truth conditions for any arbitrary sentence of that language, then we will want to know what sorts of modifications must be made to accommodate expressions whose meanings depend in some way on the conversational context. Do the contextualists’ insights require us to give up this aim altogether? Or can we offer a truth-conditional pragmatics that accounts for context sensitivity but simultaneously
22
22 Anne Bezuidenhout preserves the aims of formal approaches to natural language? This second debate depends on theories and methods in formal semantics and pragmatics, which are also empirically informed disciplines, even if empirical evidence is brought to bear on formal models in a less direct way than it is brought to bear on cognitive theories of linguistic competence and performance. A third way of approaching these issues is from an analytic philosophy of language perspective. The aim here is to get clear about crucial concepts, such as the concepts of saying and implicating and how these are to be demarcated from each other. What is the mark of a conversational implicature and are there tests for telling implicatures apart from what is said on the one hand and from entailments and presuppositions on the other? What are we to make of Grice’s (1975: 56–57) distinction between generalized and particularized conversational implicatures? Is there a need to recognize a type of implicit content intermediate between explicitly expressed content and what is conversationally implicated, as Bach (1994) has argued by proposing his notion of impliciture? This third debate is a metaphysical debate that relies largely on the sorts of a priori methods that philosophers use in their conceptual investigations. It relies on the intuitions of the conceptual analysts themselves, and on their ability to devise possible scenarios (thought experiments) that can test these intuitions. These three debates should be kept apart, at least initially. Unfortunately, a separation between these various concerns has not always been observed, which has led to inappropriate criticisms. For example, some of the harshest critiques have been launched by minimalists from the conceptual analysis camp (e.g. Cappelen and Lepore 2005; Devitt 2013) against contextualists from the cognitive architecture camp (e.g. Sperber and Wilson 1995; Bezuidenhout 2002; Carston 2002). Prima facie this is inappropriate, as the aims and methods of practitioners from these various camps are so different. Ideally, these areas of research will produce outcomes that cohere with each other. However, it seems wise at the outset to get clear within each area what exactly the contextualist challenge is and how it is to be addressed, using the tools appropriate to those areas, before attempting a unified account across all these domains. There is a core set of intuitions (or ‘data’) that are common to practitioners in all three camps, concerned with classes of natural-language expressions that seem to exhibit ‘context sensitivity’—that is, expressions that show variation in ‘meaning’ (and what exactly it is that varies is itself one of the issues under dispute) depending on variation in contexts of use. However, what is done with this data is very different and conditioned by the divergent aims of these camps.
2.2 The Cognitive Architecture Debate The debate in this arena concerns the degree of autonomy of semantic processing from pragmatic processing. The debate as it concerns minimalism versus contextualism has
23
Contextualism and Semantic Minimalism 23 largely concentrated on the comprehension rather than the production end of language processing. Moreover, the debate is often cast in very Chomskyan terms, in the sense that it takes for granted a computational model of the language faculty and an internalist, mentalist conception of cognitive processing generally. Chomsky (1995a) claims that internalism is at odds with Lewis’s (1983) claim that ‘real’ semantics concerns word–world relations. This prima facie puts the cognitive architecture project at odds with the formal modelling project to be discussed below. (I say prima facie, because it is not clear that formal models are any more or less connected to the external world than cognitive models are or need be.) Against this background, there are two ways to be a minimalist. Some researchers are minimalists in both senses (e.g. Borg 2004), some only in one sense (Sperber and Wilson 1995), and some in neither sense (Bezuidenhout 2009, 2010). I will label these two versions of minimalism ‘Minimalism1’ and ‘Minimalism2’. Minimalism1 is the view that the language system is a module whose internal workings, which could of course be highly complex and involve submodules, are not affected by contextual information. The module has no access to information such as who is speaking to whom, when or where they are speaking, what is mutually manifest to the conversational participants, what they intend to accomplish by their speech acts, etc. Only the output of the language system at the so-called conceptual-intentional interface can be affected by contextual information. This output is a representation of the logical form (LF) of the linguistic string that was input to the syntactic parser. An LF representation is one that reveals the lexico-syntactic structure of the linguistic input string. This structural information is needed to assign what Sperber and Wilson (1995) call a propositional form (PF) to the input string. The PF of an uttered sentence classifies the utterance in terms of its truth-conditional content and is a ‘pragmatic development’ of the LF. Those who reject Minimalism1 argue that contextual information plays a role at the earliest moments of sentence comprehension and that information from all levels (phonological, syntactic, semantic, information-structural, and discourse- level information) interacts to produce a representation of the content of the uttered sentence. Minimalism2 concerns the LF representation itself and how close it is to a representation of the propositional form (PF) of the uttered sentence—that is, of the sentence that is being comprehended. Minimalism2 is the view that the LF representation just is a representation of the PF, lacking only the contextual assignment of values to any indexical elements. This presupposes that LF representations are already structurally and lexically disambiguated, so that there would be as many distinct LF representations as there are possible lexico-structural analyses of the input string. For example, suppose the uttered sentence is (1) below. One of its LF representations might be something like (2): (1) He has a bat. (2) [There is a y: Bat1(y)] Have(x(salient male),y)
24
24 Anne Bezuidenhout Here ‘Bat1’ should be understood as representing one of the possible meanings (e.g. the animal or the sporting equipment meaning) of the lexically ambiguous word ‘bat’. Context would, however, still be needed to assign a value to the free variable ‘x’. This reference assignment is semantically constrained, since the third-person singular masculine pronoun ‘he’ semantically encodes the information that the referent is a male and is salient in the conversational context. Arguably, however, this information is not a part of the propositional content of the uttered sentence but simply encodes a rule or procedure to help the comprehender determine the referent of the pronoun—and hence the propositional content of the sentence. Thus I have indicated this information in subscripted form in (2). It is to be understood as an instruction to the pragmatic system to fill out the LF representation in an appropriate way to yield a representation of the uttered sentence’s PF.1 Those who reject Minimalism2 argue that there is a much larger gap between the representation of the uttered sentence’s LF and the representation of its PF. Pragmatic processing is needed to go from the LF representation to the PF representation and this doesn’t involve simply assigning values to free variables or selecting amongst multiple possible lexico-structural analyses. It may also require various sorts of pragmatic modulation processes, such as enrichment or loosening (Carston 2002; Recanati 2010) or pragmatic transfer (Nunberg 1995). Such modulation processes are not mandatory and are not linguistically triggered by any element of LF. They are triggered by information available in the conversational context. For example, the word ‘have’ in (1) is not semantically ambiguous. Nevertheless, even if we fix on, say, the sporting equipment meaning of ‘bat’, (1) could be understood in several different ways, as suggested by the following possible conversational exchanges, in which the content enclosed by {} should be taken as implicitly communicated: (1’) Speaker A: Who has {possession of} a bat? Speaker B: He has a bat. Speaker A: Yes, but not on him. He left his at home. (1”) Speaker A: Who has {ownership of} a bat? Speaker B: He has a bat. Speaker A: That’s not his bat. He’s just holding it for someone. One has to understand something about the wider conversational purposes to understand whether what is at issue is bat ownership or physical bat possession. Think also of the subtly different understandings that are triggered depending on the object of 1 Korta and Perry (2011) defend a form of multi-propositionalism, according to which a sentence uttered in a context can express multiple propositions. They would argue that the semantic information encoded by ‘he’ is a part of the utterance-level truth conditions of an utterance of (1), although not of the utterance’s referential truth conditions. Multi-propositionalism and the procedural account of pronouns outlined in the paragraph above are likely to be empirically equivalent views, in the sense that they would not generate distinct processing predictions.
25
Contextualism and Semantic Minimalism 25 the verb ‘have’—such as ‘have a bat’, ‘have a haircut’, ‘have a meal’, ‘have a chat’, ‘have a showdown’. This is not a matter of semantic ambiguity but of polysemy. Hofstadter and Sander (2013) see the understanding of polysemy as a form of analogy-making, which they argue is fundamental to all categorization. Every time you categorize an object, you are extending the category in novel ways, by seeing the present situation as analogous to past ones in contextually relevant respects. Thus, polysemy is the norm rather than the exception. Carston (2013) argues for a similar position. She claims that all open-class words (e.g. nouns, verbs, and adjectives) are polysemous to some degree. She also argues that polysemy is best explained by assuming that the encoded (lexical) meanings of such open-class words are concept schemas rather than full-fledged concepts. Thus context will be needed to figure which concept the speaker intends to communicate by the use of a verb such as ‘have’ in a particular conversational context. The main contextualist point is that what is explicitly uttered by a speaker (or writer) may be incomplete in all sorts of ways, even once lexical and structural ambiguities have been resolved and contextual values have been assigned to all indexical elements. Speakers utter these incomplete forms with the expectation that hearers (or readers) will be able to access relevant contextual information in order to flesh out the explicitly encoded information and thereby arrive at full PF representations (or full enough representations—it may be that in certain cases an underdetermined form is ‘good enough’. On the idea of ‘good enough’ language comprehension, see Ferreira and Patson 2007; Frisson and Pickering 2007). PF representations are representations of the truth-conditional content or contents of the speaker’s utterance. (We already saw that Korta and Perry (2011) are advocates of multi-propositionalism. Sperber and Wilson (1995) too have long held that multiple PF representations will be recovered— not only the so-called ‘explicature’ of the speaker’s utterance, but also various higher- level explicatures.) It is possible also to reject both Minimalism1 and Minimalism2. Such a position argues that language processing is sensitive to contextual information from the very earliest moments of the comprehension process and that pragmatic processes may operate in a local way to pragmatically modulate word or phrasal meanings even before any sentence-level meaning can be generated. On this view it is not entirely clear what role an LF representation would play, since the notion of an LF representation belongs to the modularist conception of language processing. Such a representation is the output of the language system at the interface with the conceptual-intentional system and is by definition something unaffected by pragmatic factors. It is something arrived at by what Sperber and Wilson (1995) would call decoding processes, as opposed to inferential processes. Those who deny Minimalism1 should question the idea of a stage of processing at which such an LF representation is derived. Instead, they should argue that the only full representation derived is a propositional-level one. Moreover, there is even room for scepticism as to how frequently full propositional contents are derived in the course of ordinary conversations, since in many cases something much sketchier may suffice for the conversational purposes—see
26
26 Anne Bezuidenhout Gregoromichelaki et al. (2011) and Gregoromichelaki, Cann, and Kempson (2013) and the work on ‘good enough’ comprehension cited above. Sperber and Wilson have long challenged the idea that successful communication requires the transmission of precise, definite, and fully articulated propositional contents. See Sperber and Wilson (1995) and Wilson and Sperber (2012b) for discussions of the notion of weak communication and the idea that communication may on occasion involve simply making a set of assumptions manifest or increasing the salience of information that is manifest. Information that is manifest to person x at time t need not be explicitly represented by x at t but merely be able to be so represented at t. The question as to whether the language system is modular or not has been debated for over forty years by now, so summarizing that body of work would be a mammoth task and not particularly germane to the debate about contextualism. For the limited purposes of this Handbook essay, I will say just a few sketchy things about the sort of empirical evidence that is germane to the cognitive architecture debate between minimalists and contextualists. Firstly, this debate is perhaps better conceived as a debate between those who hold that language processing is serial, with information of various sorts coming online in stages (first semantico-syntactic information then pragmatic information), and those who hold that language processing is a process in which multiple constraints must be simultaneously considered, so that pragmatic considerations may very well have effects from the very earliest moments of comprehension. I will call these the staged- processing and the multiple constraint views respectively. Secondly, there is an extensive body of experimental work that is relevant to the contextualism–minimalism debate, although it wasn’t explicitly undertaken because of any awareness of this debate. There is relevant experimental work on pronoun resolution, discourse coherence, contextual frames and scripts, topic shifts in evolving conversations, focus constructions, etc. Some authors have used this existing experimental literature in defence of their contextualist views. See, e.g., Bezuidenhout (2009, 2010). No more will be said about this body of experimental work here. Thirdly, the sorts of linguistic constructions that have been experimentally investigated, where the impetus for the study does come from a concern to address the contextualist–m inimalist debate, are somewhat limited in scope. There is work on various sorts of non-l iteral language uses, such as the use of metaphor, irony, metonymy, and idioms. Another target of investigation has been scalar implicatures and other Gricean generalized conversational implicatures (GCIs). While this is not the place to thoroughly evaluate the current state of the empirical findings regarding the staged-processing versus the multiple-constraint view of pragmatic processing, I will point to some of the principal experimental findings regarding non-l iteral language use (section 2.2.1) and scalar implicatures (section 2.2.2).
27
Contextualism and Semantic Minimalism 27
2.2.1 Experiments on non-literal language use Raymond Gibbs and his colleagues undertook some of the earliest experimental work in this area, much of it summarized in Gibbs (1994). Gibbs was interested in the processing of idioms, ironical and metaphorical utterances, and indirect speech acts (e.g. indirect requests such as ‘Can you open the window?’). He reports data that seems to go against what Recanati (1995) calls the ‘literal-first serial’ model of processing and to support what we might call a ‘pragmatics-first’ model. On the basis of reading times for sentences presented in contexts biased to either a literal or a pragmatic meaning, Gibbs found that sentences are read fastest in contexts that support the pragmatic meaning. The staged-processing view claims that pragmatic content can be constructed only after the literal, semantically encoded meaning has been retrieved, which would predict longer reading times for sentences in pragmatic-biased contexts as opposed to ones in literal-biased contexts, which is just the opposite of what Gibbs found. Subsequent research has challenged Gibbs’ work in various ways. For one thing, he was not careful to control for novel as opposed to routinized uses of non-literal language. It may be that certain expressions have been used so frequently and successfully to convey metaphorical and ironical contents that they have become standardized forms for expressing these contents (e.g. ‘Lawyers are sharks’ or ‘He’s a fine friend’). In comprehending these forms of language, we would expect the pragmatic content to be more salient than the literal content. Rachel Giora and her colleagues have done extensive work on irony in particular, showing that Gibbs’ findings do not hold for novel ironies. See Giora (2003, 2004) and Peleg et al. (2004). Giora argues in favour of her Graded Salience Hypothesis, according to which what drives processing is the relative salience of information. In some cases, the most salient meaning may very well be the non-literal one, in which case it will be accessed first. In the case of novel ironies, on the contrary, the literal meaning may be more salient, leading to longer processing times in order to recover the ironical meaning, even in contexts supporting an ironical reading. There is a substantial literature on metaphor processing, much of it pre-dating the contextualist–minimalist debate. Nevertheless, some of this literature can be mined for insights that are relevant to this debate. The research of Glucksberg and his associates is most relevant. See Glucksberg (2001, 2003, 2004). The reason that Glucksberg’s work is particularly relevant is that his categorization account of metaphor, first developed by Glucksberg and Keysar (1990, 1993), is very similar to the relevance theory account of metaphor in terms of the creation of ad hoc concepts, as developed by Carston (2002), Wilson and Carston (2006), and Carston and Wearing (2011). Contextualists interested in defending the view that metaphorical meaning is directly expressed and not merely conversationally implicated (as Grice had earlier argued) have adopted the relevance theory account of metaphor in terms of ad hoc concepts. Thus, Glucksberg’s experimental work can be used to support the contextualist framework, as he has
28
28 Anne Bezuidenhout consistently found that metaphorical meanings are rapidly accessed and that they are more accessible than literal meanings under certain conditions. There has also been experimental work done on the processing of metonymy that has been explicitly undertaken in order to empirically investigate the contextualist notion of semantic underdetermination. Frisson and Pickering (2007) presented people with sentences containing familiar or novel metonymies and compared the reading times for these sentences with controls where a literal reading was called for. Frisson and Pickering found that familiar metonymies, such as ‘read Dickens’, were read just as fast as literal controls, such as ‘met Dickens’, although unfamiliar metonymies, such as ‘read Needham’, resulted in longer reading times. In addition, unlike what has been found for lexically ambiguous words, Frisson and Pickering found no word frequency effects for polysemous words. No extra processing effort was needed to process the subordinate (less frequent) meaning. They conclude that this is because initially only a single underspecified meaning is retrieved, which is the same in all cases. Only subsequently, during a ‘homing-in’ stage of processing, is the underspecified meaning fleshed out to retrieve the context-specific concept that was communicated. They also speculate that in some cases no fleshing out is needed because the underspecified meaning is ‘good enough’ for communicative purposes. Carston (2013) has construed this work as evidence for her contextualist position according to which all open-class words lexically encode concept schemas, rather than full-blown concepts.
2.2.2 Experiments on scalar implicatures Here again Raymond Gibbs’ work was pioneering. Gibbs and Moise (1997) showed that contents that have traditionally been classified as Gricean generalized conversational implicatures (GCIs) are in fact more accessible than literal or minimal meanings, whether sentences containing GCI triggers are presented in isolation or in contexts biased to a literal or enriched meaning. They used offline methods, such as forced- choice judgement tasks. Nicolle and Clark (1999) and Bezuidenhout and Cutting (2002) expanded on Gibbs’ work by controlling more carefully for various types of GCIs and using methods that could tap into online language processing, such as self- paced reading tasks. These studies found a processing advantage for contextually enriched meanings over literally encoded contents. More recent work in so-called experimental pragmatics has concerned itself with the comprehension of GCIs. This work (described below) has focused especially on one special subclass of GCIs, namely scalar implicatures (SIs). And even more narrowly, the focus has been on three types of SIs, namely SIs triggered by ‘some’, as in ‘Some of the apples are red’, implicating that not all are, SIs triggered by numerical adjective phrases, as in ‘Jane has three apples’, implicating that she has no more than three, and SIs triggered by ‘or’, as in ‘Jane has apples or oranges’, implicating that she doesn’t have both apples and oranges.
29
Contextualism and Semantic Minimalism 29 One exception to this focus on scalar inferences is the work by Ira Noveck and his colleagues. They have also investigated the sorts of generalized inferences that are derived by what Levinson (2000) calls the I-Principle, such as the implicature of temporal sequence generated by ‘Laurent broke his ankle and went to the hospital’ (see Noveck and Chevaux 2002). This work on SIs has been directed at testing three competing models of SI comprehension, namely (i) the Default Model according to which SIs are default inferences that are automatically derived by a listener/reader on encountering an SI trigger; (ii) the Contextual Model according to which SIs are only derived if the conversational context warrants it, or (iii) the Structural Model according to which SI derivation depends on structural or grammatical factors such as whether the SI trigger occurs in an upward-entailing (as opposed to downward-entailing) environment. The Default Model is associated with the work of neo-Griceans, and in particular with the views of Levinson (2000). The Contextual Model is associated with the work of relevance theorists, such as Sperber and Wilson (1995) and Carston (2002). The Structural Model is associated with the work of Crain and Pietroski (2002) and Chierchia (2004). Competing findings by experimentalists have often been cast as supporting one or another of these theoretical frameworks. For example, Bezuidenhout and Morris (2004), Katsos et al. (2005), Breheny et al. (2006, 2013), Bezuidenhout et al. (2009), and Degen and Breheny (2011) have argued that context affects whether and what sorts of inferences are drawn. On the other hand, Storto and Tanenhaus (2005) and Grodner et al. (2010) have argued in favour of the default inference view. Finally, Chierchia et al. (2006) and Panizza and Chierchia (2011) have argued in favour of the structural inference view. There is also work that has been said to support the contextualist view but that does not clearly do so. As Borg (2004) has pointed out, the work by Noveck and his colleagues seems rather to favour the minimalist viewpoint, Noveck’s own claims notwithstanding. For example, Noveck and Posada (2003) found evidence that suggests that pragmatic inferences are drawn only after retrieval of a literal meaning. And Bott and Noveck (2004) found that under time pressure adults tend not draw scalar inferences, suggesting that such pragmatic processing is effortful. They conclude that minimal meanings are retrieved first, with additional pragmatic inferences drawn only later and with effort. Although Noveck claims that these findings support relevance theory, they seem more in line with what Recanati (1995) calls the literal-first serial model of processing, which is friendlier to minimalism than to contextualism. Work in experimental pragmatics over the past five years has branched out considerably and is no longer focused just on SI triggers in simple clauses but has looked at such triggers in embedded clauses and has also compared them with presupposition triggers. See, for example, Geurts and Pouscoulous (2009), Chemla and Spector (2011), Cummins et al. (2012), Chemla and Bott (2013). There is also a thriving experimental pragmatics research community, the EURO-XPRAG network (see reference section for a link to this organization’s website), whose members are involved in regular
30
30 Anne Bezuidenhout research exchanges, conferences, and the sharing of work through online archives. This work covers a broad array of pragmatic phenomena, including lexical pragmatics, quantifier scope assignments, presupposition projection, referential processing, the processing of metaphor and irony in the context of child language acquisition, and so on. Overall, the most recent work in experimental pragmatics tends to show that a simplistic contrast between minimalism and contextualism is not amenable to direct experimental testing. There is unlikely to be any crucial experiment that will decisively show one or other of these broad stances to be correct. Instead, we can expect that a much more nuanced picture regarding the respective contributions of grammar and context will emerge from more refined experimental investigations.
2.3 The Formal Semantics Debate Formal semanticists, who are interested in sentence-level meaning and in producing compositional theories that generate truth conditions for natural-language sentences, take the contextualist challenge to be either to explain away the data, i.e. to cast doubt on the idea that there really are expressions that are context-sensitive in the relevant sense (section 2.3.1); to tame the data, i.e. to show that these expressions after all belong to well-understood categories of context-sensitive expressions, such as indexicals, that can be handled by traditional approaches (section 2.3.2); or to accommodate the data, i.e. to show that we can offer compositional theories that can handle context-sensitive expressions other than indexicals (section 2.3.3).
2.3.1 The ‘explaining away’ strategy Cappelen and Lepore (2005) adopt this first strategy. They argue that there is only a small set of context-sensitive expressions (e.g. ‘I’, ‘you’, ‘now’, ‘then’, ‘here’, and ‘there’). These all belong to the fairly well understood category of indexicals, which we know how to accommodate in a formal semantic model. Cappelen and Lepore (2005) present a set of tests for context sensitivity that they argue show that expressions that contextualists have made much of—expressions that are ‘semantically underdetermined’ and in need of pragmatic modulation—are not after all context-sensitive. Examples of expressions allegedly requiring pragmatic modulation include quantifier phrases in need of domain restriction, ‘contextuals’ such as ‘local’, ‘foreign’, ‘nearby’ that presuppose a point of view that must be figured out in context, and comparative adjectives such as ‘big’ that require a standard of comparison that is often not explicitly supplied. Cappelen and Lepore’s tests for context sensitivity are discussed in more detail in section 2.4.3 on the conceptual analysis debate.
31
Contextualism and Semantic Minimalism 31
2.3.2 The ‘taming’ strategy Stanley (2000, 2002) and Stanley and Szabo (2000) have adopted this strategy. This approach accepts contextualist intuitions as to the truth-conditional contents of utterances of sentences such as (3) to (10) below. Here we are to imagine appropriate conversational contexts that would yield the (unpronounced) pragmatic contributions to truth- conditional content that are represented by the words enclosed by ‘{}’ in the examples:
(3) June has finished {washing the dishes}.
(4) This table is strong enough {to support this pile of books}.
(5) All the bottles {in the room} are empty {of drinkable alcohol}.
(6) Every student {at Yale} runs.
(7) You’re not going to die {from this cut}.
(8) I haven’t eaten {lunch today}.
(9) France {the region} is {approximately} hexagonal {-shaped}.
(10) This meat is {almost} raw. However, Stanley and Szabo would deny that this variability is a matter of semantic underdetermination calling for contextual supplementation. Rather, such variability should be assimilated to ordinary cases of indexicality, except that the indexicality is hidden, in the sense that the indexicals belong to the underlying LF of the uttered sentence and remain unpronounced. This hidden indexical account has been worked out in most detail for cases of quantifier domain restriction, and so would handle the cases of context sensitivity illustrated in examples (5) and (6). Under Stanley and Szabo’s proposal, a sentence such as (6) would receive the syntactico-semantic analysis along the following lines: [S [NP [Det Every][N < student , f (i) >]][VP [V runs]]] The idea is that a common noun such as ‘student’ ‘co-habits a node with a contextual variable’ (Stanley and Szabo 2000: 251). The values of ‘i’ and ‘ f’ are provided by the context. In the case of (6), the net effect is that the set of those who are at Yale is intersected with the set of students to yield the restricted domain of the quantifier phrase ‘every student’ in (6). Thus the interpreted LF of (6) can be represented along the following lines: Everyx : student ( x ) & at ( x , Yale) runs ( x ) What evidence is there that common nouns cohabit nodes with contextual variables? Stanley and Szabo argue that the evidence for such hidden variable elements is that
32
32 Anne Bezuidenhout higher quantifiers can bind them. For example, take sentence (5) and embed it so that its quantifier is governed by a higher-level one, as in (11). Intuitively, (11) means something like (12), suggesting that there was a hidden variable f(i) cohabiting a node with the nominal ‘bottles’ in (5) and which was available for binding by the quantifier ‘every’ in the embedding phrase: (11) In every room in the house, all the bottles f(i) are empty. (12) [Every i: room(i) & in-house(i)]([Every x: bottle(x) & in(i,x)] empty(x)) This account of quantifier domain restriction in terms of the notion of nominal restriction is very elegant, but it is unclear how this method of dealing with contextual variability is going to generalize to all the different sorts of context dependence that contextualists have pointed to. For one thing, since the method depends on the idea of nominal restriction, it isn’t clear how it will be extended to sentences that do not contain common nouns and/or in which the contextual variability is due, for example, to the use of a verb or a prepositional phrase or a sentence connective, as in (13)–(15): (13) Sam is walking. (14) It is on top of that. (15) She took to the bottle and he left her. Sentence (13) could be used to say different things in different contexts, even keeping the referent of the proper name fixed. For instance, if Sam is recovering from a serious stroke, then (13) might be true, even if Sam is dragging himself along with great difficulty with the aid of a walker. On the other hand, if Sam is a hale and hearty person who is dragging himself along with the aid of a walker, perhaps as some sort of sick joke, then (13) would not be true. Similar remarks could be made regarding the context sensitivity of the prepositional phrase ‘on top of that’ in (14) and the sentential connective ‘and’ in (15). It surely is not plausible to claim that every part of speech (nouns, verbs, prepositions, sentential connectives, etc.) cohabits a node with a contextual variable; at the very least it is unclear that we are justified in positing so many unpronounced variable elements. A pragmatic account seems more parsimonious. Stanley (2000) makes it plain that he does not wish to explain all contextual variability by his method of nominal restriction. He aims merely to show that people have been too hasty in offering pragmatic explanations when a syntactico-semantic explanation of contextual variability is possible for some cases.
2.3.3 The ‘accommodation’ strategy Advocates of this strategy argue that contextual modulation is perfectly compatible with compositionality. Recanati (2010) has recently defended this view. He argues that
33
Contextualism and Semantic Minimalism 33 the meanings that undergo semantic composition in order to determine the (primary) truth-conditional content of an uttered sentence are the (optionally) pragmatically modulated meanings of its constituents. He introduces the idea of a modulation function that takes an expression and a context as an argument. When the modulation function that the context makes salient is applied to the character of the expression in that context, it yields the modulated meaning of that expression in that context. If the salient function is the identity function, then the expression receives its literal interpretation. That is, ‘literalness is treated as a special case of (zero) modulation’ (Recanati 2010: 45). For example, one way to interpret ‘The city sleeps’ is to understand ‘the city’ as being used metonymically to refer to the inhabitants of the city. We assume that in such a context c, a contextually salient metonymic function, call it g1, is applied to the character of the expression ‘the city’ in the context, and yields the modulated content The inhabitants of the city. It is this modulated meaning that is composed with the (zero) modulated meaning of ‘sleeps’ to yield the truth-conditional content of the whole. So the interpretation of the complex expression ‘The city sleeps’ in a context c can be represented as a function of the modulated meanings of its syntactic parts, as follows:
(
)
((
) (
I ’The city sleeps' c = f g 1 I (’the city’) , g 2 I (’sleeps’) c1 c2
))
Here c1 and c2 are intended to correspond to ‘sub-parts of the context c in which the complex expression … is used’ (Recanati 2010: 44). The functions g1 and g2 are the modulation functions made salient in those subcontexts. To get the intended interpretation, g2 must be the identity function, which yields the literal interpretation for ‘sleeps’ in subcontext c2. An alternative interpretation for ‘The city sleeps’ might apply the identity function to ‘the city’ and then be forced to apply a metaphorical modulation function to ‘sleeps’, yielding something like the interpretation that the city is free of traffic and noisy activities. The main issue regarding Recanati’s proposal is whether it does indeed capture the spirit of the principle of compositionality that formal semanticists place such value on. Recanati himself states that his proposal violates strong compositionality. It does not preserve the idea that the content of an expression is a function of the contents of its constituent expressions and the way they are put together syntactically, but merely that the content of an expression is a function of the modulated meanings of its constituent expressions and the way they are put together. Moreover, the content of the whole can itself be modulated in different ways depending on the context. However, Recanati argues that if we let the context be an extra argument of the composition function, we can say that the modulated meaning of a complex is a function of the modulated meanings of its constituents, together with the way they are put together syntactically and the context that provides for these modulated meanings. This yields a form of weak compositionality. Citing Pagin (2005: 313), Recanati argues that this
34
34 Anne Bezuidenhout weak form of compositionality is compatible with the aims of formal semantics, as the meaning function in most formal models is already assumed to take a context as one of its arguments. Whether this proposal does indeed preserve a robust enough notion of compositionality depends greatly on how the details are worked out, and especially on whether the notion of contexts and subcontexts can be rendered in a way that is formally tractable. This issue is too large to be examined within the confines of this Handbook essay. Those who are interested in pursuing the issue should consult the edited collection by Werning and Hinzen (2012), especially the essays in that collection by Pelletier, Westerståhl, and Pagin. Further discussions of Recanati’s views can be found in Frápolli (2007).
2.4 The Conceptual Analysis Debate The earliest salvos in this philosophical battle came from scholars sympathetic to certain Wittgensteinian themes about meaning as use. Travis (1985, 2000), for example, argues for the view that what is said (the truth-conditional content of a sentence uttered in a context) varies radically across contexts even keeping reference to particular individuals fixed. Consider, for example: (16) This kettle is black. Referring to a particular kettle, (16) might be true in one conversational context, but false in another, because in the first context the fact that the kettle is covered in thick soot is what is at issue, despite the fact that, were the soot to be removed, we would discover a red enamelled surface below, whereas in the second context, it might be the colour of the enamel covering that is at issue. Searle (1980, 1983) argues for a very similar view about the shifting of truth-conditional content depending on the changing background. Thus a sentence such as ‘The cat is on the mat’ will have different truth conditions depending on whether, for example, we are considering cats and mats in normal gravitational conditions on the surface of Earth or cats and mats in deep space. In the 1980s, the accepted view in philosophy of language distinguished cleanly between sentence meaning on the one hand, which was regarded as the province of semantics, and speaker meaning on the other, which was regarded as the province of pragmatics. Berg (2002: 351) calls this the Standard View of Semantics (SV): (SV) Every disambiguated sentence has a determinate semantic content, relative to an assignment of contents to its indexical expressions, and not necessarily identical to what may be conveyed (pragmatically) by its utterance.
That is, once structural and lexical ambiguities have been resolved and contextual values have been assigned to indexicals and demonstratives, the resulting content is
35
Contextualism and Semantic Minimalism 35 fully propositional and corresponds to the sentence’s truth-conditional content. Such sentence meaning also corresponds to what is strictly and literally said by an utterance of the sentence. Speaker meaning, on the other hand, corresponds to what is implicitly conveyed by saying something in some conversational context. Such implicitly conveyed meanings are Gricean conversational implicatures. Many competitors to SV have emerged over the past three decades. I am not going to survey all these alternatives here. For example, I won’t say more here about the views of neo-Griceans such as Levinson (2000) and Horn (1984, 1992a). These views have been especially influential in the realm of experimental pragmatics (discussed above). For detailed discussions of these neo-Gricean alternatives, see Huang (2007: ch. 7) or Huang (2014: ch. 8). Instead, I want to focus on the class of competitors to SV that have exercised philosophers of language in particular, namely those views that argue that pragmatic considerations intrude into the domain of what is said, and (apparently) into the domain of truth-conditional content. Travis (1985), Carston (2002), and Recanati (2004a, 2010) amongst others defend this view. In its extreme form, this pragmatic intrusion view claims that no sentence of a natural language such as English expresses a complete proposition independent of some context of use. There are no Quinean ‘eternal’ sentences. Weaker versions claim merely that many more sentences are semantically underdetermined than has traditionally been recognized. The phenomenon of semantic underdetermination is meant to be distinct from the phenomenon of indexicality. The claim is that even after all lexical and structural ambiguities have been resolved and after all indexical elements in a sentence have been assigned a reference, a sentence may still fail to express a complete proposition, and hence may fail to be truth-evaluable. Relevance theorists such as Sperber and Wilson and Carston, and Neo-Griceans such as Bach and Recanati, have drawn attention to a range of examples of semantic underdeterminaton or incompleteness, such as examples (3)–(15) discussed earlier, where the encoded meaning of a sentence falls short of a fully propositional meaning or at least of the directly communicated meaning. The communicated meaning involves pragmatic modulation of the encoded meaning. Bach (1994) distinguishes modulation processes of completion and expansion. Recanati (2004a) talks of processes of saturation, free enrichment, and transfer. Relevance theorists, such as Sperber and Wilson (1995) and Carston (2002), talk about processes of enrichment and loosening. Although at one level these people’s views are similar, they also disagree on some crucial issues. One major disagreement concerns cases that Bach would say call for expansion. In these cases, such as examples (7)–(10), Bach would say there is a complete proposition that is expressed. It is just that this (minimal) proposition is not the one directly communicated, and hence a process of expansion is necessary to arrive at the communicated proposition. Relevance theorists on the other hand would regard all of (3)–(10) as semantically incomplete. Enrichment processes are called for in all cases except (9) and (10), which call for loosening.
36
36 Anne Bezuidenhout There are also those who agree with the contextualists’ intuitions regarding such examples but who reject the idea of semantic underdetermination. Instead, they argue for a view according to which pragmatically supplied content is always linguistically controlled. If there is no explicit syntactic element that does the controlling, there must be a hidden element. This is the hidden indexical view that was discussed in section 2.3 and will not be discussed further here. Minimalists on the other hand want to argue that there is a complete, albeit minimal, proposition expressed by every well-formed natural language sentence. This minimal proposition is one that is expressed by that sentence in all contexts (although other contextually specific propositions may be expressed in addition to the minimal one). So, in example (16), ‘The kettle is black’, the same minimal proposition is expressed whether the context is focused on soot-covered or enamelled kettles. Carston (2013) takes the idea of semantic incompleteness to apply not only at the level of sentences but also of words. She argues that the encoded meaning of the word ‘black’ (an element of meaning that is common to all uses of ‘black’) is a concept schema rather than a full-blown concept. Supplementation with information from the conversational context is always required to yield the full ad hoc concept that the speaker has in mind and intends to convey in that context. This idea avoids commitment to a concept of being black, simpliciter. Moreover, since it does not commit itself to a full-blown concept of blackness that is expressed in every context of use of ‘black’, it allows us to see how ‘black’ could be used to cover a range of different cases, such as in our talk of black kettles, black tea, black hair, black skin, black smoke, black knights, black holes, black moods, and so on. Thirdly, it explains why, even if we confine ourselves to the category of kettles, there is no single way of being black (say having a black enamelled surface) that must be expressed in every context by the use of the phrase ‘black kettle’. (‘Kettle’, just as does ‘black’, expresses a concept schema on Carston’s view.) Returning to the minimalist claim about example (16), what is the minimal proposition that is allegedly common to the imagined contexts of use described above? One suggestion is that it is a proposition that is way-of-being-black neutral. On this view, the various ways things have of being black are not constituents of the truth-evaluable contents of (utterances of) (16), but rather enter the picture at the point at which such contents are evaluated for truth and falsity at various circumstances of evaluation. This view, known as relativism, is examined in more detail in section 2.4.4. However, inasmuch as it assumes that there is a fully propositional content that involves the idea of something’s being black, simplicter (as opposed to a propositional schema that is ‘gappy’ with respect to modes of being black), it faces problems. Another suggestion is that the common propositional content is a generalized one that can be made fully explicit, as in (17): (17) This kettle is black in some way. However, if (17) is supposed to be truth-conditionally equivalent to (16), it gets the truth conditions of (16) wrong. Suppose I am shopping for a kettle to go with my black
37
Contextualism and Semantic Minimalism 37 and white kitchen decor. My friend points to a red kettle with a black handle and utters (16). What she has said is false in the context of my shopping trip. However, if what she said corresponded to the content of (17), she would have said something true. Minimalists could respond that, in the imagined circumstances, the speaker did in fact say something strictly and literally true, but what she said was misleading, because it conversationally implied something false, namely that she had spotted something that would satisfy my desire for a kettle that continued the black-and-white theme of my new kitchen. Clearly we are not making much progress in this debate at this point, but are simply trading intuitions back and forth about what was said, about whether what was said is true or false or not truth-evaluable at all (because it is merely a propositional schema), and about whether there is a minimal (literal) proposition that captures a content that is shared across various contexts of use of a sentence such as (16). Cappelen and Lepore (2005) suggest that the question as to the nature of minimal propositions is a metaphysical one that can be set aside by those concerned with capturing the truth- conditional content of sentences such as (16). We can rest content with a disquotational treatment that gives truth conditions by means of biconditionals such as: ‘The kettle is black’ is true (in English) if and only if the kettle is black. However, inasmuch as the right-hand sides of such biconditionals use the contested sentences, it is not clear that we can put these questions aside, as the coherence of such biconditionals is at stake. Is there any less question-begging reason to accept the minimalist view? I will end this section by considering five ways that might be suggested for breaking this deadlock between minimalists and contextualists, which I will discuss in sections 2.4.1 to 2.4.5.
2.4.1 Grice’s tests Firstly, readers might have been thinking for a while now that there is an easy way to decide the issue between the contextualist and minimalist conceptions of saying. After all, Grice proposed various tests to determine whether some communicated content is a conversational implicature or not, namely, the tests of cancellability, non- detachability, and calculability. Why can’t these tests be applied to sentences such as (16) to see whether the contents that Travis say they express are implicatures or are part of what is said? However, as Travis (1985) shows, this cannot be done without begging the question against the contextualist. All Grice’s tests presuppose a conception of what is said, and hence can’t be applied until we’ve settled the debate between contextualists and minimalists. For example, cancellability is the test according to which we add the denial of the putative implicature to what is said and if the result is not semantically anomalous, the putative content must indeed be an implicature. Thus, if one can felicitously assert ‘Some of my friends are linguists; in fact, all of them are’, Grice would say this shows that ‘Not all of my friends are linguists’ is merely conversationally implied by ‘Some of my friends are linguists’. Clearly, in order to carry out this test, we already need to
38
38 Anne Bezuidenhout know what is said by the utterance of the sentence that is claimed to have the implicature in question. If one’s intuitions are that what is said by an utterance of ‘Some of my friends are linguists’ is ‘Only some of my friends are linguists’, then it will not be felicitous to utter ‘Some of my friends are linguists; in fact, all of them are’. Thus this cancellability test will not break the deadlock. Similar remarks apply to the other tests proposed by Grice.
2.4.2 Conceptual confusions A second strategy for breaking the deadlock is to argue that contextualists are conceptually confused about the notion of saying, and that in particular it is a mistake to assume, as (some) contextualists do, that what is said = proposition expressed = truth- conditional content. There are at least two directions in which to push this second line of attack: (i) argue for a thin conception of what is said that severs the connection between what is said and the proposition expressed; (ii) argue for a thick conception of what is said that severs the connection between what is said and truth-conditional content. Bach (1994) opts for the first of these substrategies. For Bach, what is said can be something incomplete, and hence something that is neither propositional nor truth- evaluable. This will be the case for all those sentences that he believes are semantically underdetermined, such as examples (3)–(5). There are also cases in which Bach would say that what is said is something propositional and truth-evaluable, but in these cases he is still at odds with contextualists. What contextualists identify as what is said, Bach prefers to call an impliciture (i.e. something that is implicit in what is said). This is directly communicated and is distinct from the minimal proposition, which is Bach’s candidate for what is said. For example, ‘I haven’t eaten’ expresses the minimal proposition that the speaker has never eaten, but it directly communicates the proposition that the speaker hasn’t eaten a meal recently. Bach argues vigorously for his views, marshalling many considerations in favour of his more minimalist conception of what is said, and for tying the conception more closely to sentence meaning than contextualists are willing to do. He argues that contextualists have conflated locutionary and illocutionary content, and that they are bringing extraneous psychological considerations into semantics. He also suggests that contextualists can’t adequately explain the following four-way division: it is possible to say something and mean something else instead, say something and mean something else in addition, say something and mean exactly that, and say something and mean nothing at all. Contextualists have argued against Bach’s minimalist conception of what is said on the grounds that it is psychologically implausible, and that it is unlikely to play a role in any cognitively realistic account of utterance interpretation. This objection is not decisive, because Bach (1995) has an explanation as to why minimal propositions might sometimes be psychologically inert. He claims that in cases such as ‘I haven’t eaten’ a
39
Contextualism and Semantic Minimalism 39 process of standardization takes the listener directly to the impliciture, which is what is directly communicated. The minimal proposition (namely, that the speaker has never eaten, which is what is strictly and literally said by ‘I haven’t eaten’) is bypassed. Here again it looks as though a priori considerations are not going to break the deadlock. Cappelen and Lepore (1997) adopt the second substrategy—the one that offers a thick conception of saying in order to break the link between what is said and truth- conditional content. They argue that our conception of what is said should line up with our practices of speech reporting. Indirect reports of speech that use the locution ‘X said that ___ ’ often go well beyond the truth-conditional content of the speech being reported. There may be only partial semantic overlap between the truth-conditional content of the reported utterance and the content of the complement clause used to report the content of the utterance. In particular, there are cases in which pragmatic features of a reported utterance can become semantically encoded in the complement clause of an indirect report of that utterance. For example, suppose that Bob is interested in dating his co-worker Jane. He asks his friend Stanley whether he should ask Jane out on a date. Stanley says: ‘Jane has three children’. Later, Suzie asks Bob why he hasn’t asked Jane out on a date yet. He replies: ‘Stanley said that Jane is married’. In this case, what Stanley conversationally implied to Bob has been incorporated into Bob’s speech report to Suzie. Cappelen and Lepore’s thick conception of saying cannot be used to break the deadlock between minimalists and contextualists. The notion of ‘saying’ that is integral to the common-sense practice of indirect speech reporting is indeed extremely elastic and ranges over contents that are directly expressed, contents that are entailed by what is directly expressed, and even content that is merely conversationally implicated by what is directly expressed. This just shows that ordinary people do not find it necessary to mark distinctions in the same way that theorists do. Appeal to this common- sense conception does not help settle the theoretical debate.
2.4.3 Cappelen and Lepore’s tests A third way to attempt to break the deadlock between minimalists and contextualists is to devise (allegedly non-question-begging) tests that can demarcate context- sensitive from context-insensitive expressions. The quest for such tests has become the focus of the philosophical debate about contextualism in recent years. For example, Cappelen and Lepore (2005) propose what they call the Inter- Contextual Disquotation (ICD) test. If an expression ‘e’ is context-sensitive, then any sentence S involving ‘e’ satisfies the following test: (ICD) There can be a false utterance of ‘S’ even though S.
In particular, the first-person pronoun ‘I’ appears to pass the test, because there can be a false utterance of ‘I am a woman’ even though I (the author of this Handbook entry)
40
40 Anne Bezuidenhout am a woman. This sentence would be false, for example, if uttered by the editor of this Handbook. Cappelen and Lepore argue that any expression that passes the ICD test should also be one for which we can produce a Real Context-Shifting Argument (RCSA). An RCSA for a sentence S is essentially a fleshed-out version of its ICD-instance that spells out the two contexts across which S varies in truth-conditional content. Cappelen and Lepore are playing off the fact that contextualists often defend their views by imagining varying contexts across which some sentence S shifts in truth conditions. Consider for example ‘Barack is ready’. In one context it could be that Barack is ready to play golf and in another that he is ready for a TV debate. Contextualists claim that there are different truth conditions across these two contexts, because ‘ready’ is context-sensitive and its content will be enriched in different ways in these different contexts. Cappelen and Lepore deny this, claiming that this sentence fails the ICD test and that there is no RCSA for it. Leslie (2007) argues, contra Cappelen and Lepore (2005), that expressions such as ‘enough’, ‘ready’, ‘every’, ‘tall’, and sentences such as ‘It’s raining’ easily pass the ICD and RCSA tests. She concedes to Cappelen and Lepore that other expressions, such as ‘knows’, ‘red’, ‘weighs 80 kg’ that contextualists have suggested are context-sensitive, fail these tests and so she agrees that these are context-insensitive. However, since she only wants to be a moderate, not a radical, contextualist, she is happy to make this concession. In other words, Leslie is willing to accept Cappelen and Lepore’s ICD and RCSA tests for context sensitivity but thinks they have mistakenly applied their own tests. On the other hand, others have argued that the ICD and RCSA tests are in fact question-begging. They trade on just the sorts of conflicting intuitions between contextualists and minimalists that led to the deadlock that we hoped the tests could break. The tests only seem convincing if one is already persuaded that minimalism is true and if one is antecedently persuaded that grammatically complete sentences such as ‘Barack is ready’ express complete propositions independently of context. See Bezuidenhout (2006). Cappelen and Lepore (2005) propose two further tests for context sensitivity, the Reporting and Collections tests, which I have labelled (RT) and (CT) respectively: (RT) Indirect speech reports are blocked for context-sensitive expressions. (CT) Collective uses are blocked for context-sensitive expressions.
They argue that expressions in their ‘basic set’ (viz. indexicals such as ‘I’, ‘now’, ‘today’) pass these tests and that all other expressions (such as ‘enough’, ‘ready’, ‘nearby’, or ‘tall’) fail. However, these tests can be challenged too. Cappelen and Hawthorne (2009), following Leslie (2007: 145–146), have argued that the Reporting and Collection tests are problematic for so-called contextuals, viz. expressions such as ‘nearby’, ‘local’, ‘foreign’, and so on. To determine the truth- conditional contents of utterances using such contextuals, one must first determine
41
Contextualism and Semantic Minimalism 41 the perspective or point of view from which the contextual is being used. Consider the following sentences containing the contextual ‘nearby’: (18) Barack is at a nearby café. (19) Michelle said that Barack is at a nearby café Note that in (18), the café could be nearby to Barack (the subject-based reading) or nearby to the speaker or nearby to the addressee or nearby to some third person or to some landmark (four possible context-based readings). And if we add a layer of indirect reporting, we add the possibility of the point of view being anchored to either the internal or the external subject—in example (19) this would be either Barack’s or Michelle’s point of view. Beginning first with the Collection test, consider the examples given in Table 2.1 for ‘nearby’. These examples show that ‘nearby’ patterns with indexicals such as ‘today’. Note that readers are to imagine that context changes from one use of the expression to the next. In the case of ‘today’, we are to imagine that it picks out two different days across its various uses (e.g. 8 June versus 9 June). Similarly, we are to imagine that the point of view for ‘nearby’ varies across uses (e.g. nearby the addressee versus nearby the Empire State Building). Obviously, if we imagine that the context is fixed across the different uses, then collective uses are possible, for indexicals as well as for contextuals. As far as the Reporting test is concerned, the conclusion is similar, given the assumption that the context changes for the original versus the reported uses of the expressions. Corazza and Dokic (2013) argue that there are other ways in which contextuals pattern with expressions that are acknowledged to be context-sensitive. In particular, contextuals exhibit behaviour that is similar to the behaviour of pronouns. Just as pronouns have deictic, bound, and anaphoric uses, so too do contextuals such as ‘nearby’. And just as pronouns in cases of ‘too’-ellipsis have both strict and sloppy readings, so do contextuals such as ‘nearby’. Consider the following examples, where the (a) uses involve the pronoun ‘he’ or ‘his’ and the (b) uses involve the contextual ‘nearby’: Deictic uses: a. He was driving a Mercedes. b. John was nearby. Anaphoric uses: a. A man pulled up at the curbside. He was driving a Mercedes. b. A car pulled up at the curbside. John was nearby. Bound uses: a. Everyone loves his Mercedes. b. Everyone picked out a nearby Mercedes. Ellipsis cases (both strict and sloppy readings possible): a. John loves his Mercedes and Bill does too. b. John went to a nearby café and Bill did too.
42
42 Anne Bezuidenhout Table 2.1 Report and Collection tests applied to context-sensitive expressions Collection test
Collection Reporting test blocked?
Report blocked?
Pure indexicals (e.g. ‘I’, ‘now’, ‘today’)
Barack is leaving today {June 8}. Michelle is leaving today {June 9}. *Barack and Michelle are both leaving today.
Yes
Barack: It needs to be done today {June 8}. *Michelle: Barack said that it needs to be done today {June 9}.
Yes
Contextuals (e.g. ‘nearby’, ‘enemy’, ‘ foreign’, ‘left’)
Barack is nearby {to the addressee}. Michelle is nearby {the Empire State Building}. *Barack and Michelle are both nearby.
Yes
Barack: Joe is nearby {to the addressee}. *Michelle: Barack said that Joe is nearby {the Empire State Building}.
Yes
Underspecified expressions (e.g. ‘enough’, ‘ready’, ‘cut’)
This {reinforced steel} table is strong enough {to bear the weight of a cannonball}. That {balsa-wood} table is strong enough {to bear the weight of a plastic toy}. *This table and that table are both strong enough.
Yes
Barack: This {balsa-wood} table is strong enough {to bear the weight of a plastic toy}. *Michelle: Barack said that this {balsawood} table is strong enough {to bear the weight of a cannonball}.
Yes
However, even though these tests do support the claim that contextuals are context- sensitive, on the assumption that pronouns are, we cannot make these tests be definitive of context sensitivity, as indexicals like ‘today’ would not pass them. Returning to Cappelen and Lepore’s Collection and Reporting tests, I would argue that underspecified expressions like ‘strong enough’ behave like contextuals and indexicals. The examples in Table 2.1 establish this, on the assumption that the condition for being ‘strong enough’ varies across the two uses; e.g. strong enough
43
Contextualism and Semantic Minimalism 43 to bear the weight of a cannonball versus strong enough to bear the weight of a plastic toy. For instance, suppose that Barack is talking to his children’s nanny. While gesturing towards a balsa-wood table, he utters the sentence ‘That table is strong enough’, intending to communicate that it is strong enough to bear the weight of a plastic toy. Michelle overhears the conversation and mistakenly thinks that Barack is talking about something that is strong enough to bear the weight of a Civil War cannonball that has just been donated to the White House, because earlier she and Barack had been discussing places to display the cannonball and that issue is still on her mind. If later Michelle is debating with a White House staffer whether that same balsa-wood table is suitable to act as a display location for the cannonball, she cannot felicitously support her position by saying: ‘Barack said this table is strong enough’. Similar scenarios would show that the collective use is infelicitous. One might try to carve out a special class of context-sensitive expressions by insisting that indexicals are expressions whose content is locked to the time, place, and world of the speaker’s utterance. Firstly, this move doesn’t help save the Collection and Reporting tests. These tests are not calibrated to this utterance-context-locking feature, since even expressions without this feature sometimes pass the tests (e.g. contextuals such as ‘nearby’ whose reference can be fixed from either the speaker’s or the addressee’s or an internal subject’s or a third person’s or some landmark’s point of view). Secondly, it would have the consequence that even indexicals are not context-sensitive expressions. Predelli (2005) calls the view that indexicals are always interpreted relative to the utterance context the Simple Minded view. However, indexicals can sometimes have ‘shifted’ interpretations, for example when used in recorded messages, in free indirect discourse, the historical present, and so on. In these uses their content appears to be fixed relative to a context other than the utterance context. For more on the debate on indexical shifting (which not everyone agrees involves context-shifting), see Predelli (1998, 2005), Corazza (2004), Schlenker (2004), and Recanati (2007). More recently, Cappelen and Hawthorne (2009) have proposed that a better test of context stability across contexts is provided by their Agreement-based diagnostics (2009: 54–67). Many others have recently entered the fray, with several arguing that even these argument-based diagnostics are question-begging and cannot be used to resolve the debate between minimalists and contextualists. See Barker (2011) and Caponigro and Cohen (2011) for further relevant commentary on these tests and their shortcomings.
2.4.4 Relativism A fourth popular way of breaking the deadlock between contextualists and minimalists is to defend what has become known as relativism. This is a way of dealing with
44
44 Anne Bezuidenhout Travis-t ype examples such as (16), ‘This kettle is black’, in a way that preserves intuitions about shifting truth values but keeps the propositional content (i.e. truth conditions) fixed across contexts. It also eschews hidden variables. Context sensitivity is not reduced to indexicality, as it is in Stanley’s approach. Relativism generalizes the Lewis–Kaplan idea that contents are evaluated as either true or false at circumstances of evaluation, which consist of World–Time pairs: . On this view, times are part of the circumstances of evaluation, not of the propositions that are evaluated at these circumstances. Tense markers are treated as temporal operators (PAST, PRES, FUT), which shift the time of evaluation of a proposition to a past, present, or future time as needed. Thus, ‘John will be late’ is represented as ‘FUT(Late(john))’, which is true at a world w iff there is a time t in the future in world w at which ‘Late(john)’ is true. This requires propositions expressed by sentences such as ‘John will be late’ to be temporally neutral. Similarly, take an utterance of (16) in context c 1 in which the amount of soot covering the surface of the kettle is what is at issue, versus a context c2 in which the colour of the kettle’s enamel finish is what is at issue. One might argue that there is a ‘manner of being F’ parameter that belongs to circumstances of evaluation, where F is a property such as being black. With that parameter set one way in a circumstance of evaluation, , the proposition expressed by (16) may come out true. With the parameter set another way, , (16) may come out false. The proposition expressed by (16) in both contexts is the same and does not contain any element corresponding to this mode of being black. It is mode- of-being-black neutral, just as it is temporally neutral (if we adopt an operator analysis of tenses). This view faces some of the same problems faced by minimalists who wish to argue that there is a complete proposition that is expressed across all uses of a sentence such as (16). It is unclear what the nature of such a proposition is. One might try to set this issue aside as a metaphysical question that is irrelevant to those studying the semantics and pragmatics of natural languages. I agree that metaphysics can be set aside by those engaged in psychological investigations of human language competence and performance. However, philosophers of language, who are in effect engaging in metaphysics, cannot set such questions aside. The attempt to offer a relativist theory, according to which context sensitivity is a matter of shifting circumstances of truth evaluation rather than of shifting truth- conditional content, is currently a growth area in the philosophy of language. For discussions of relativism, see Predelli (2005), Recanati (2007), the essays in Kölbel and Carpintero (2008), and MacFarlane (2009, 2012).
2.4.5 Situationalism Another approach to defusing the minimalist–contextualist debate is the situationalist view defended by Corazza (2007) and Corazza and Dokic (2012, 2013). They accept
45
Contextualism and Semantic Minimalism 45 the idea that the encoded contents of sentences fall short of the enriched contents that contextualists argue are directly expressed by the utterance of such sentences and that minimalists hold are merely ‘speaker-meant’ or indirectly communicated (e.g. conversationally implied) by such utterances. However, Corazza and Dokic deny the assumption shared by (some) contextualists and minimalists, namely that there is a fully enriched thought lying behind the incomplete or minimal content encoded in the sentence. Corazza and Dokic argue that the underlying thought is as minimal as the encoded content. This is because the speaker can exploit (features of) the conversational situation and offload the representational burden to the situation. This in turn means that the addressee does not need to build a representation of a fully enriched content in order to recover the speaker’s intended meaning. The addressee too can offload the representational burden to the situation. This is an externalist conception of communicated content, according to which utterance content is partly determined by features of the external environment in which the utterance occurs. Perry (1986) defends this view in a paper titled ‘Thought without representation’. He imagines a community of speakers, the Z-landers, who live in an isolated location and who know nothing about a world or other communities (even possible ones) beyond the bounds of their secluded home. (It seems they don’t tell stories about other imagined places and times; they have no creation myths or religion; they don’t believe in ancestral spirits or anything like that—in other words, they are not humans). When a Z-lander utters the sentence ‘It’s raining’, he does not express a thought with an unarticulated constituent corresponding to the location of the rain event. Hence his interlocutors need not enrich the encoded content of ‘It’s raining’ in order to understand the content the speaker intended to communicate. There is no need to articulate this location parameter, either in language or in thought, for there is no possible location other than Z-land for the rain event and the Z-landers are right there, in Z-land. The location of the rain event is determined by the situation in which the utterance occurs and the relation of the speaker’s and hearer’s thought tokens to that external location means that the location need not be internally represented by either of them. The principal problem facing this view is to explain, without an appeal to fine- grained mental representations, how the fine-grained contents that are communicated in real conversational contexts get communicated. Situations are of course also fine- grained, but which aspects of a fine-grained situation are the ones that are relevant in fixing the content of a thought that is tokened in that situation? Let us assume that Z- land has microclimates, just as places on Earth do. We all know that it can rain in one neighbourhood in your city and yet be dry a few blocks away. Thus even a Z-lander will on occasion want to communicate that it is raining in some specific subregion of Z-land rather than to communicate the more general claim that it is raining somewhere in Z-land. If the speaker can’t rely on the hearer to be able to mentally represent Z-land subregions, this more fine-grained content could not be communicated. The various philosophical stances taken in this conceptual analysis debate (minimalism, hidden indexicalism, relativism, situationalism, etc.) do indeed occupy
46
46 Anne Bezuidenhout interestingly different positions in the logical space of alternatives. However, what is less clear is that these distinctions make any psychological difference. Thus, if one is interested in an empirical account of natural-language competence and performance, it is hard to see whether, say, minimalism and relativism make testably different predictions about language processing. This is why I said at the outset that the debate between contextualists and minimalists (and their critics) is largely of interest to philosophers.
2.5 Conclusion In all three debates that I have identified above (the cognitive architecture debate, the formal semantics debate, and the conceptual analysis debate), the picture that has emerged is far more subtle and complex than some defenders of minimalism and contextualism might like to admit. It is increasingly clear that there will be no knockdown argument against contextualism and in favour of minimalism. While it is clear that the Standard View of semantics from the 1980s has to be given up, there are a range of positions intermediate between extreme minimalism on the one end and radical pragmatism on the other, and it is not clear exactly how far towards the contextualist end of the spectrum one is forced to go by the current evidence. For a characterization of this spectrum, see Recanati (2004a). Additionally, it is certainly the case that much of the early debate in this area was marred by the fact that the various parties to the debate were using terminology in inconsistent ways. Thus many disagreements were in fact merely terminological. It is to be hoped that as the debate moves forward, participants will attempt to keep track of evidence from multiple domains (both empirical and conceptual), in order to avoid setting up argumentative ‘strawmen’, which are all too easy to demolish. The truth is likely to be much more complicated, nuanced, and interesting.
47
Chapter 3
N eo - G r icea n Pr agm atics Yan Huang
In the William James Lectures delivered at Harvard University in 1967, H. P. Grice presented a panorama of his thinking on meaning and communication—what he called his ‘tottering steps’ (Grice 1989a: 4) towards a systematic, philosophically inspired pragmatic theory of language use, which has since come to be known as Gricean pragmatics (see Chapman 2005 on the life and work of Grice). Since its inception, the classical Gricean paradigm has encouraged numerous refinements, reinterpretations, and reconstructions, giving rise to various neo-Gricean enterprises. Consequently, the classical and neo-Gricean theory has revolutionized pragmatic theorizing and has to date remained one of the foundation stones of contemporary thinking in linguistic pragmatics and the philosophy of language. This chapter undertakes to present a state-of-the-art survey of neo-Gricean pragmatics. The organization of the essay is as follows. Section 3.1 discusses classical Gricean pragmatic theory. Next in section 3.2, I present neo-Gricean pragmatic theory, focusing on the bipartite model put forward by Horn and the trinitarian model posited by Levinson. Finally, sections 3.3, 3.4, and 3.5 examine the role played by neo-Gricean pragmatic theory in effecting a radical simplification of the lexicon, semantics, and syntax in linguistic theory, respectively.
3.1 Classical Gricean Pragmatics On a general Gricean account of meaning and communication (e.g. Grice 1989a), there are two theories: a theory of meaningn[on]n[atural] and a theory of conversational implicature. In his theory of meaningnn, Grice emphasized the conceptual relation between natural meaning in the external world and non-natural, linguistic meaning of utterances. He developed a reductive analysis of meaningnn in terms of the speaker’s reflexive intention, the essence of which is that meaningnn or speaker meaning is a matter of expressing and recognizing intention.
48
48 Yan Huang In his theory of conversational implicature, Grice suggested that there is an underlying principle that determines the way in which language is used maximally efficiently and effectively to achieve rational interaction in communication. He called this overarching dictum the cooperative principle and subdivided it into nine maxims of conversation classified into four categories: Quality, Quantity, Relation, and Manner. The names of the four categories are taken from Immanuel Kant (Grice 1989a: 26). The cooperative principle and its component maxims ensure that in an exchange of conversation, truthfulness, informativeness, relevance, and clarity are aimed at. (1) Grice’s cooperative principle and its constituent maxims of conversation a.
The cooperative principle Make your conversational contribution such as is required, at the stage at which it occurs, by the accepted purpose or direction of the talk exchange in which you are engaged.
b.
The maxims of conversation Quality: Try to make your contribution one that is true. (i) Do not say what you believe to be false. (ii) Do not say that for which you lack adequate evidence. Quantity: (i) Make your contribution as informative as is required (for the current purposes of the exchange). (ii) Do not make your contribution more informative than is required. Relation: Be relevant. Manner: Be perspicuous. (i) Avoid obscurity of expression. (ii) Avoid ambiguity. (iii) Be brief (avoid unnecessary prolixity). (iv) Be orderly.
Assuming that the cooperative principle and its associated maxims are normally adhered to by both the speaker and the addressee in a conversational interaction, Grice suggested that a conversational implicature—roughly, any meaning or proposition expressed implicitly by a speaker in his or her utterance of a sentence which is meant without being part of what is said in the strict sense1—can arise from either strictly observing or ostentatiously flouting the maxims. In Huang (2003, 2004b, 2007, 2014), I called conversational implicatures that are engendered by way of directly observing the maxims
1 Defined thus, conversational implicature is a component of speaker meaning rather than a pragmatic inference (Bach 2001, 2006a, 2012; Saul 2002; Horn 2004, 2012a, 2012b). By contrast, Sperber and Wilson (1995), Levinson (2000), Atlas (2005), Geurts (2010), and Chierchia (2013) are still treating conversational implicature as a pragmatic inference.
49
Neo-Gricean Pragmatics 49 conversational implicaturesO, and conversational implicatures that are generated by way of the speaker’s deliberately flouting the maxims conversational implicaturesF. A second Gricean dichotomy, independent of the conversational implicatureO/conversational implicatureF one, is between those conversational implicatures which arise without requiring any particular contextual conditions and those which do require such conditions. Grice (1989a: 31–38) called the first kind generalized conversational implicatures (GCIs), and the second kind particularized conversational implicatures (PCIs). Finally, Grice designed a battery of tests to facilitate the identification of conversational implicature. First, defeasibility or cancellability—conversational implicatures can disappear in certain linguistic or non-linguistic contexts. Second, non- detachability— a ny linguistic expression with the same semantic content tends to carry the same conversational implicature. (A principled exception is those conversational implicatures that arise via the maxim of Manne.) Third, calculability—conversational implicatures can transparently be derived via the cooperative principle and its attendant maxims. Fourth, non-conventionality— conversational implicatures, though dependent on the saying of what is coded, are non-coded in nature. Fifth, reinforceability—conversational implicatures can be made explicit without producing too much sense of redundancy. Sixth, some conversational implicatures may be indeterminate. They can be taken as conveying an open-ended range of implicitly expressed meanings relating to matters in hand. Finally, we have universality—conversational implicatures tend to be universal, being rationally motivated rather than arbitrary (e.g. Grice 1989a; Levinson 2000; Huang 2014: 39–43).2
3.2 Neo-Gricean Pragmatics 3.2.1 The rise of neo-Gricean pragmatics While revolutionary in nature, what Grice presented at the James Williams Lectures was no more than just a sketchy proposal, albeit an ambitious one. As pointed out by Lakoff (1995: 194) metaphorically: Grice himself provided an architect’s sketch, but the full-fledged habitable edifice is still under construction; the original blueprint must be continually extended and reinterpreted to meet the needs of those who will actually inhabit it.
2 Needless to say, the Gricean doctrine has a long and rich lineage. Some proto-Gricean notions can go back at least as far as the first-century bc rhetorician Dionysius and the fourth-century rhetoricians Servius and Donatus. These ideas were later reiterated by the nineteenth-century English philosophers John Stuart Mill and Augustus De Morgan. Much more recently, in the 1950s, Grice was also influenced by similar concepts put forward by his colleagues within the tradition of ordinary language philosophy at Oxford University (e.g. Horn 2006a, 2012a, 2012b).
50
50 Yan Huang Given Grice’s seminal but patchy proposal, it was no wonder that in the 1970s his ideas were considered by some scholars to be vague, superfluous, vacuous, unfounded, and even plain contradictory. Even Horn (1988a: 130), himself an eminent neo-Gricean, was of the following view: Grice’s original framework is clearly at best incomplete and at worst inadequate beyond repair to the task of predicting sets of nonlogical inferences … It is simultaneously too weak, in allowing the derivation of virtually anything by encompassing directly opposite maxims …, and too strong, in treating all calculable inferences monolithically.
Therefore, if the classical Gricean programme was to be taken seriously within linguistics and the philosophy of language, much work had to be done to systematize, rigidify, and develop the original concepts that had been adumbrated by Grice. It was partially to meet this challenge that various neo-Gricean pragmatic reformations were developed. What, then, have the neo-Griceans done to improve the classical Gricean pragmatic theory? A number of areas can be identified. In the first place, individual types of classical Gricean conversational implicature were systematized. Horn (1972) represented the first attempt to provide a systematic analysis of conversational implicature due to Grice’s first half of his submaxim of Quantity. He succeeded in providing a formalized account of what has since come to be known as Q-scalar implicatures. The next major breakthrough relating to the same Gricean submaxim came from Gazdar (1979). Inspired in part by Horn’s treatment of Q-scalar implicatures, Gazdar showed how Q-clausal implicature can be formalized in an equally elegant way. Later, Atlas and Levinson (1981) noted that if we apply the reasoning behind Q-scalar and Q-clausal implicatures to the conversational implicature arising from the second half of Grice’s submaxim of Quantity, we will get the wrong results. This led them to present the first formal analysis of this type of implicature by appeal to a novel principle of informativeness (‘Read as much into an utterance as is consistent with what you know about the world.’), hence the term I-implicatures. Second, formal mechanisms were devised to account for implicature projection and cancellation (e.g. Gazdar 1979; Horn 1984, 2009; Hirschberg 1991; Levinson 2000). For example, the constraints on Horn scales (see section 3.2.2), proposed by Levinson, successfully rules out *, *, and * as forming a genius Horn scale. In the same vein, the Levinsonian resolution schema (to be discussed in section 3.2.3) makes correct predictions for which type of conversational implicature overrides which type of conversational implicature under what circumstances. Third, the Gricean maxims of conversation were reinterpreted, and more recently, there have also been various attempts to integrate the classical and neo-Gricean pragmatic theories with other current linguistic theories such as decision theory (Merin 1999), game theory (Benz, Jäger, and van Rooy 2006), bidirectional Optimality theory (Blutner and Zeevat 2004), and Bayesian natural-language semantics (Zeevat and Schmitz 2015). Fourth, both a narrower (minimalist) and a broader (maximalist) characterization of what is said than the original Gricean concept have been put forward (for the former, see e.g. Bach 2012,
51
Neo-Gricean Pragmatics 51 Saul 2002, and Horn 2012a, and for the latter, see e.g. Levinson 2000 and Recanati 2010; see also Sperber and Wilson 1995). Fifth, contrary to the orthodox Gricean view, conversational implicature is allowed to intrude onto the semantic content of what is said (e.g. Levinson 2000; Recanati 2010; but see Bach 2012; Horn 2012a, 2012b; Geurts 2010 for a dissenting view). Finally, the whole Gricean mechanism of the cooperative principle and its constituent maxims has been subject to various attempts at reduction. Of all the reductionist models, the most influential are the bipartite Hornian and the tripartite Levinsonian neo-Gricean models of pragmatic principles.3 This reductionist approach4 is consistent with the spirit of a meta-theoretical desideratum known as ‘Occam’s razor’ which dictates that entities are not to be multiplied beyond necessity.
3.2.2 The Hornian model Horn (1984, 2012a, 2012b) put forward a bipartite model. On Horn’s view, all of Grice’s maxims (except the maxim of Quality) can be replaced with two fundamental and antithetical pragmatic principles: the Q[uantity]-and R[elation]-principles. (2) Horn’s Q-and R-principles
a. The Q-principle (addressee/hearer-based) Make your contribution sufficient. Say as much as you can (modulo the R-principle).
b. The R-principle (speaker-based) Make your contribution necessary. Say no more than you must (modulo the Q-principle). In terms of information structure, Horn’s Q-principle, which collects Grice’s first half of his submaxim of Quantity and his first two submaxims of Manner, is a lower- bounding pragmatic principle which may be (and characteristically is) exploited to engender upper-bounding conversational implicatures: a speaker, in saying ‘ … p … ’, conversationally Q-implicates that (for all he or she knows) ‘ … at most p … ’. In other words, as pointed by Horn (2012a), what is Q-implicated relies on what isn’t (but could have been) said. The locus classicus here is those conversational implicatures that arise from a semantic or lexical scale called a Q-or Horn scale.
3
Another influential reductionist model is, of course, Sperber and Wilson’s (1995) unitarian relevance theory. Given that relevance theory does not endorse at least some of the basic premises and goals of the Gricean programme, it is post-rather than neo-Gricean. See e.g. Huang (2012a, 2014) for a comparison between relevance theory and classical and neo-Gricean pragmatic theory. See also the debate between Laurence Horn and Robyn Carston carried out in Intercultural Pragmatics, vols 2 and 3, in 2005–6. 4 This version of neo-Gricean pragmatics was first outlined by Laurence Horn and Stephen Levinson in their joint seminar given at the 1987 Stanford LSA Linguistic Institute.
52
52 Yan Huang There are two types of Horn scale: positive and negative. A positive Horn scale is defined in (3) (Horn 1972; Atlas and Levinson 1981; Levinson 2000; Huang 2007: 38, 2014: 45). (3) Positive Horn scales A set of linguistic alternates such that S(x1) unilaterally entails S(x2), where S is an arbitrary simplex sentence frame, and x1 > x2, and where x1, x2, …, xn are a. equally lexicalized items, of the same word class, from the same register; and b. ‘about’ the same semantic relation or from the same semantic field. A few examples of Horn scales are given in (4). (I use ‘< >’ to represent a Horn scale.) (4)
a. Quantifiers: b. Connectives: c. Adjectives: d. Adverbs: e. Cardinal numbers: f. Articles: g. Modals: , h. Verbs:
We move next to a negative Horn scale, which is defined in (5) and exemplified in (6) (Levinson 2000: 82; Huang 2014: 46–47). (5) Negative Horn scales For each well-formed positive Horn scale of the form , there will be a corresponding negative Horn scale of the form , regardless of the relative lexicalization of the negation. (6) i.e.
Given a Horn scale, if a speaker asserts a lower-ranked or semantically weaker alternate (i.e. a rightwards expression in the ordered set), then he or she conversationally Q-implicates that he or she is not in a position to assert any of the higher-ranked or semantically stronger ones (i.e. leftwards in the ordered set) in the same set. Thus, the use of the positive (7a) gives rise to the Q-implicature in (7b), and the assertion of the negative (8a) generates the Q-implicature in (8b). (I use ‘+>’ to stand for ‘(ceteris paribus) conversationally implicate’.) (7) a. Some of the local residents considered the disaster a wake-up call. b. +> Not many/ most/ all of the local residents considered the disaster a wake-up call
53
Neo-Gricean Pragmatics 53 (8) a. Not many of the local residents considered the disaster a wake-up call. b. +> Not none of the local residents considered the disaster a wake-up call; some of them did An interesting point to note is that sometimes scalar expressions in the same semantic field form two Horn scales linked by a (sub)contrary relationship rather than one single, unified Horn scale, systematically engendering two opposite sorts of pragmatic enrichment (e.g. Horn 1989: 239–240; Levinson 2000: 86–87). This is illustrated by the paired Horn scales in (9). (9) a. Quantity: b. Frequency: c. Epistemic modality: d. Temperature: e. Preference: f. Evaluation: Finally, mention should also be made of the fact that under certain circumstances, a Horn scale can be inverted. For example, the order of the cardinal numbers in the Horn scale in (10a) is reversed. Consequently, (10b) gives rise to the Q-scalar implcature in (10c). (10) a. < … 6, 7, 8, 9, 10, … > b. John has cut down his smoking to ten cigarettes a day. c. +> John hasn’t cut down his smoking to less than ten cigarettes a day Now contrast (10) with (11). Notice that the latter constitutes a negative Horn scale. The assertion of (11b) then generates the Q-scalar implicature in (11c). (11) a. < … ~12, ~11, ~10, … > b. John hasn’t cut down his smoking to ten cigarettes a day. c. +> John has cut down his smoking to more than ten cigarettes a day Of particular relevance here is that the direction of Q-scalar implicatures engendered in an affirmative sentence like (10b) and in its negative counterpart like (11b) is exactly the opposite. The same is also true of the direction of scalar or pragmatic entailments. Having discussed Horn’s Q-principle, let me turn to his countervailing R-principle. The R-principle, which subsumes Grice’s second half of his submaxim of Quantity, his maxim of Relation, and his last two submaxims of Manner, and which is based on Atlas and Levinson’s (1981) principle of informativeness, is an upper-bounding pragmatic law which may be (and systematically is) exploited to invite lower-bounding conversational implicatures: a speaker, in saying ‘ … p … ’, conversationally R-implicates
54
54 Yan Huang that (for all he or she knows) ‘ … more than p … ’. An example is given in (12), adapted from Grice (1989a: 38). (12) John broke a finger yesterday. +> The finger was one of John’s own However, more recently, Horn (2004, 2006a, 2007a) has been of the view that the R-principle is not in itself subsumable under Grice’s cooperative principle, but under rationality. Viewing the Q-and R-principles as a mere instantiation of Zipfian economy (Zipf 1949), Horn (1984, 2007a) explicitly equated the Q-principle (‘a hearer-oriented economy for the maximization of informational content’) with Zipf’s Auditor’s Economy (the Force of Diversification, which tends towards a vocabulary of m different words with one distinct meaning for each word) and the R-principle (‘a speaker-oriented economy for the minimization of linguistic form’) with Zipf’s Speaker’s Economy (the Force of Unification, which tends towards a vocabulary of one word which will refer to all the m distinct meanings). The notion of Speaker’s Economy is further distinguishable between mental inertia or paradigmatic economy (économie mémorielle) and articulatory/physical inertia or syntagmatic economy (économie discursive), hence internally dialectic in its operation. The former is concerned with the reduction in the inventory of the mental lexicon; the latter, with the reduction in the number of linguistic units (Martinet 1964: 169; Horn 2007a: 173–174). While the Auditor’s Economy places a lower bound on the informational content of the message, the Speaker’s Economy places an upper bound on its form. Furthermore, Horn argued, quoting Paul (1899) and Martinet (1964) as support, that the whole Gricean mechanism for pragmatically contributed meaning can be derived from the dialectic interaction (in the classical Hegelian sense) between the two mutually constraining mirror-image forces in the following way. (13) Horn’s division of pragmatic labour The use of a marked (relatively complex and/or prolix) expression when a corresponding unmarked (simpler, less ‘effortful’) alternate expression is available tends to be interpreted as conveying a marked message (one which the unmarked alternative would not or could not have conveyed). In effect, what the communicative equilibrium in (13) basically says is this: the R- principle generally takes precedence until the use of a contrastive linguistic form induces a Q-implicature to the non-applicability of the pertinent R-implicature.
3.2.3 The Levinsonian model Horn’s proposal to reduce Grice’s maxims to the Q-and R-principles was called into question by Levinson (1987a, 1991, 2000). In Levinson’s opinion, Horn failed to draw
55
Neo-Gricean Pragmatics 55 a distinction between what Levinson called semantic minimization (‘Semantically general expressions are preferred to semantically specific ones’) and expression minimization (‘“Shorter” expressions are preferred to “longer” ones’).5 Consequently, inconsistency arises with Horn’s use of the Q-and R-principles. For example, in Horn’s division of pragmatic labour, the Q-principle operates primarily in terms of units of speech production whereas elsewhere, in Horn scales, for instance, it operates primarily in terms of semantic informativeness. Considerations along these lines led Levinson to argue for a clear separation between pragmatic principles governing an utterance’s surface form and pragmatic principles governing its informational content (but see Horn 2007a for a vigorous defence of his Manichaean model and Traugott 2004b for her argument that a two-pronged system is not only adequate but more explanatory in accounting for meaning shift). He proposed that the original Gricean programme (the maxim of Quality apart) be reduced to three neo-Gricean pragmatic principles, which he dubbed the Q[uantity]-, I[nformativeness]-, and M[anner]-principles. Each of the three principles has two sides: a speaker’s maxim, which specifies what the principle enjoins the speaker to say/ implicate, and a recipient’s corollary, which dictates what it allows the addressee to infer. Let me take them one by one. (14) Levinson’s Q-principle Speaker’s maxim: Do not provide a statement that is informationally weaker than your knowledge of the world allows, unless providing a stronger statement would contravene the I-principle. Recipient’s corollary: Take it that the speaker made the strongest statement consistent with what he knows, and therefore that: (i) if the speaker asserted A(W), where A is a sentence frame and W an informationally weaker expression than S, and the contrastive expressions form a Horn scale (in the prototype case, such that A(S) entails A(W)), then one can infer that the speaker knows that the stronger statement A(S) (with S substituted for W) would be false (or K~ (A(S))); (ii) if the speaker asserted A(W) and A(W) fails to entail an embedded sentence Q, which a stronger statement A(S) would entail, and form a contrast set, then one can infer the speaker does not know whether Q obtains or not (i.e. ~K(Q) or equally {P (Q), P~(Q)}). The basic idea of the metalinguistic Q-principle is that the use of a linguistic expression (especially a semantically weaker one) in a set of contrastive semantic alternates (such as a Horn scale) Q-i mplicates the negation of the interpretation 5
There is, of course, a strong tendency for the two distinct minimizations to be conflated. This general correlation, in fact, follows directly from the Zipfian theory of economy.
56
56 Yan Huang associated with the use of another linguistic expression (especially a semantically stronger one) in the same set. Seen the other way round, from the absence of a semantically stronger linguistic expression, we infer that the interpretation associated with the use of that expression does not hold. Hence, the Q-principle is essentially negative in nature. As illustrating examples, see (7) and (8) in section 3.2.2. Next, there is Levinson’s I-principle. (15) Levinson’s I-principle Speaker’s maxim: the maxim of minimization ‘Say as little as necessary’, that is, produce the minimal linguistic information sufficient to achieve your communicational ends, (bearing the Q-principle in mind). Recipient’s corollary: the rule of enrichment Amplify the informational content of the speaker’s utterance, by finding the most specific interpretation, up to what you judge to be the speaker’s m- intended point, unless the speaker has broken the maxim of minimization by using a marked or prolix expression. Specifically: (i) Assume the richest temporal, causal, and referential connections between described situations or events, consistent with what is taken for granted. (ii) Assume that stereotypical relations obtain between referents or events, unless this is inconsistent with (i). (iii) Avoid interpretations that multiply entities referred to (assume referential parsimony); specifically, prefer coreferential readings of reduced NPs (pronouns or zeros). (iv) Assume the existence or actuality of what a sentence is about if that is consistent with what is taken for granted. Mirroring the effects of his Q-principle, Levinson’s I-principle is a pragmatic law of semantic economy, the central tenet of which is that the use of a semantically general linguistic expression I-implicates a semantically specific interpretation. More accurately, in some cases, the implicature engendered by the I-principle is one that accords best with the most stereotypical and explanatory expectation given our background assumptions or real-world knowledge. (16) John pressed the spring and the drawer opened. +> John pressed the spring and then the drawer opened +> John pressed the spring and thereby caused the drawer to open +> John pressed the spring in order to make the drawer open Finally, we come to Levinson’s M-principle.
57
Neo-Gricean Pragmatics 57 (17) Levinson’s M-principle Speaker’s maxim: Indicate an abnormal, non-stereotypical situation by using marked expressions that contrast with those you would use to describe the corresponding normal, stereotypical situation. Recipient’s corollary: What is said in an abnormal way indicates an abnormal situation, or marked messages indicate marked situations. Specifically: Where S has said p containing marked expression M, and there is an unmarked alternate expression U with the same denotation D which the speaker might have employed in the same sentence frame instead, then where U would have I-implicated the stereotypical or more specific subset d of D, the marked expression M will implicate the complement of the denotation d, namely d’ of D. Unlike the Q-and I-principles, which operate primarily in terms of semantic informativeness, the metalinguistic M-principle is operative primarily in terms of a set of alternates that contrast in form. The crux of this pragmatic principle is that the use of a marked linguistic expression M-implicates the negation of the interpretation associated with the use of an alternative, unmarked linguistic expression in the same set. This is exemplified in (18b). (18) a. Mary went from the bathroom to the bedroom. +>I Mary went from the bathroom to the bedroom in the normal way b. Mary ceased to be in the bathroom and came to be in the bedroom. +>M Mary went from the bathroom to the bedroom in an unusual way, e.g. in a magic show, Mary had by magic been made to disappear from the bathroom and reappear in the bedroom Given the above tripartite classification of the neo-Gricean pragmatic principles, the question that arises next is how inconsistencies emerging from these potentially conflicting conversational implicatures can be resolved. According to Levinson (2000), they can be resolved by an ordered set of precedence, which encapsulates in part the Hornian division of pragmatic labour. (19) Levinson’s resolution schema for the interaction of the Q-, I-, and M-principles a. Level of genus: Q > M > I b. Level of species: e.g. Q-clausal > Q-scalar This is tantamount to saying that genuine Q-implicatures (where Q-clausal cancels rival Q-scalar) supersede inconsistent I-implicatures, but otherwise I-implicatures take precedence until the use of a marked linguistic expression triggers a complementary M-implicature to the negation of the applicability of the pertinent I-implicature (see e.g. Huang 2014: 64–66 for exemplification).
58
58 Yan Huang By way of summary, both Horn’s and Levinson’s neo-Gricean endeavours have put the classical Gricean pragmatic theory on a much more rigorous basis, showing that the theory can be formalized and tested or falsified, hence enhancing its predictive and explanatory adequacy.
3.3 Neo-Gricean Pragmatics and the Lexicon In the previous two sections, I outlined classical and neo-Gricean pragmatics. Starting from this section, I explore how neo-Gricean pragmatics can explain aspects of the lexicon, semantics, and syntax in linguistic theory. Let me begin with the lexicon.
3.3.1 Lexical narrowing Lexical narrowing or strengthening refers to the phenomenon whereby the use of a lexical expression implicitly conveys a meaning that is more specific than the lexical item’s lexically encoded meaning (e.g. Huang 2009). (20) a. John folded the newspaper into a rectangle. b. +> John did not fold the newspaper into a square (21) a. John had a glass of milk for breakfast this morning. b. +> John had a glass of cow’s milk for breakfast this morning Lexical narrowing can be grouped into two types. In the first, the use of the superordinate term of a hyponymic taxonomy where there is a specific hyponym denotes more narrowly the complement of the extension of the hyponym. This is the case for (20) (see (22)). (22)
Lexical narrowing of this type follows directly from Horn’s or Levinson’s Q-principle. Notice that square and rectangle form a Horn scale. Given the Q-principle, from the use of the semantically weaker rectangle, we obtain the pragmatically narrowed meaning ‘not square’. This Q-based strengthening of meaning typically gives rise to what Horn (1984) and Levinson (2000) called autohyponymy, i.e. privative polysemy. Other examples include finger +> ‘not thumb’, gay +> ‘not lesbian’, and actor +> ‘not actress’. Secondly, there is the R/I-based lexical narrowing. The basic idea here is that the use of a semantically general lexical item is R/I-implicated to a semantically more specific
59
Neo-Gricean Pragmatics 59 interpretation. This is the case for (21), where the semantically general term milk is R/I- narrowed to denote its culturally salient subset ‘cow’s milk’ (cf. goat’s milk, soy milk, almond milk, coconut milk etc.). Other examples include secretary +> ‘female secretary’, relationship +> ‘sexual/romantic relationship’, and drink +> ‘alcoholic drink’. Of these, Horn (1984) and Levinson (2000) were of the view that while drink is an autohyponym, secretary is not.6 Notice that the Q-and R/I-narrowed meanings are not part of the lexical semantics of the items under consideration. This is because, on the one hand, they can be cancelled, as in (23), (I use ‘~ +>’ to signify ‘do not conversationally implicate’), and on the other hand, they can co-occur with words such as goat, as in (24). (23) John folded the newspaper into a rectangle, if not a square. ~ +> John did not fold the newspaper into a square (24) John had a glass of goat’s milk for breakfast this morning.7
3.3.2 Lexical cloning 3.3.2.1 What is lexical cloning? Lexical cloning, formally known as ‘contrastive focus reduplication’, refers to the phenomenon whereby there is a modifier reduplication of a lexical item. The reduplicated modifier, which contains a contrastive focus accent, is utilized to single out some privileged sense, in contrast to other senses, of an ambiguous, polysemous, vague, or loose lexical expression (e.g. Huang 2009, 2015a).8 (25) a. (Dialogue between a married couple, recently separated and now living apart) A: Maybe you’d like to come in and have some coffee? B: Yeah, I’d like that. A: Just coffee-coffee, no double meanings. (Whitton 2008) b. oops, I started a really worthless thread [ … ] well, actually, I didn’t start- start it, but … (Hohenhaus 2004) c. I’m up, I’m just not up-up. (Ghomeshi et al. 2004) 6 On Horn’s (2007a: 166) view, euphemism represents a bona fide case of culturally or socially motivated R-based narrowing. 7 There is, of course, the other side of the lexical change coin, namely lexical broadening or loosening. According to Horn (2007a: 165), this process of meaning expansion can be accounted for in terms of his R-principle. 8 For an informative discussion of the morphosyntax, reduplicative phonology, and prosodic and lexical constraints on lexical cloning, see Ghomeshi et al. (2004).
60
60 Yan Huang Lexical cloning is found in a variety of English, including American, Australian, British, Canadian, New Zealand, and South African English, but it is most widely used in American English. It is a relatively recent phenomenon. Furthermore, as observed by both Ghomeshi et al. (2004) and Hohenhaus (2004) independently, the use of lexical clones is largely restricted to a certain, informal conversational register of spoken English. Even the tokens of lexical cloning that are found in written English such as scripts for plays, films, and TV programmes are largely representations of spontaneous spoken language (as a mode) in written form (as the medium) (Hohenhaus 2004). Lexical cloning is also found in a wide range of other languages in the world. These include many (Indo-)European languages such as French, German, Modern Greek, Italian, Vulgar Latin, Persian, Russian, and Spanish; African languages such as Afrikaans, Kenyang, and KiNande; Asian languages such as Malay; Australian languages such as Dyari, Dyirbal, and Western Desert language; Oceanic languages such as Maori; and (indigenous) languages of South America such as Tzeltal and Tzotzil (e.g. Huang 2009, 2015a for references therein).
3.3.2.2 Main function of lexical cloning What, then, is the main function of lexical cloning? According to Ghomeshi et al. (2004), a lexical clone is used to specify a true, real, default, salient, or prototypical denotation of the repeated lexical item. This reading is in contrast to a potentially looser or more specific interpretation. More specifically, Horn (2006b) identified three types of meaning of lexical cloning: (i) prototypical meaning, (ii) ‘value-added’ or intensifying meaning, and (iii) literal, as opposed to figurative, meaning. Furthermore, he pointed out that meaning (i) is related especially to reduplicated nouns, and meaning (ii) to reduplicated adjectives. This is also echoed by Hohenhaus (2004), who postulated two formulas, one for reduplicated nouns and the other for reduplicated adjectives/adverbs/verbs. (26) Hohenhaus’s formulas a. For reduplicated nouns An XX is a proper/prototypical/precise/just X b. For reduplicated adjectives, adverbs, and verbs An XX = really/properly/extremely X Construed thus, lexical clones function like what Lasersohn (1999) called ‘pragmatic slack regulators’—words such as exactly, precisely, and perfectly. These words serve to shrink what Lasersohn termed ‘pragmatic halos’—the set of entities associated with the denotation of a lexical item or proposition (e.g. Ghomeshi et al. 2004). If this is the case, then lexical clones can also in part be regarded as a special case of lexical narrowing, as discussed in section 3.3.1.
3.3.2.3 Context dependency of lexical clones As pointed out in Huang (2009, 2015a), the prototype intensification analysis advocated by Ghomeshi et al. (2004), Hohenhaus (2004), and Horn (2006b), however, is
61
Neo-Gricean Pragmatics 61 not without problems. What the theory predicts is at best only a general tendency. If a lexical clone were completely predictable from the isolated use of the reduplicant, then its production and comprehension would be a purely semantic issue. First, cross-linguistic evidence indicates that the pattern is not universal. There are languages in the world which display an invertible pattern. That is, while the non- reduplicated lexical item singles out the prototype category member, the lexical clone picks out its non-prototypical complement. In Dyari, whereas kintha means ‘dog’, its reduplicated form kintha-kintha-la has the meaning of ‘little dog’. Furthermore, in some of the world’s languages, the meanings of different lexical clones may move in opposite directions. A stock example of such a language is Afrikaans, in which reduplication is a highly productive kind of derivational morphology. In this language, we have bakke-bakke ‘lots of bowls’ versus ruk-ruk ‘a few times’ (e.g. Levinson 2000). Secondly, one of the most important characteristics of lexical clones is that in terms of interpretation, they are heavily context-dependent. To begin with, they typically occur in a contrastive linguistic context in the form of (27) (Hohenhaus 2004). (27) a. Not (just) X but (rather) XX b. Not (just) XX but (rather) X (28) Frank was not just my colleague; he was my colleague-colleague. We worked in the same department; in fact, we worked in the same section. Next, in order to avoid potential ambiguities and misinterpretations, as a pre- emptive measure, a user of a lexical clone often launches a self-initiated self-repair of some sort, using paraphrases, clarifications, or explanations. (29) (Colleagues are talking about a tsunami in the South Pacific Ocean, 30 September 2009) It’s Samoa Samoa rather than American Samoa. On the other hand, there are occasions where a speaker anticipates that the addressee would be able to figure out what the repeated constituent means, but in fact the addressee cannot. Consequently, the addressee will launch an other-initiated self- repair to seek clarification from the speaker. This is precisely the phenomenon displayed in (30). (30) Mike: I didn’t really bring anything. Geoff: So you didn’t bring any food!? Mike: Not food-food … Geoff: What’s ‘not food-food’, then? Mike: Got bubble gum … Geoff: How you expect to survive on gum for three days, you muppet?!? (Hohenhaus 2004) (From The Hole)
62
62 Yan Huang Thirdly, as pointed out by Whitton (2008), the same lexical clone can mean different things in different contexts. As an illustrating example, let us consider the use of drink-drink in the following contexts. (31) a. … if you must have a ‘drink, drink’ go with the hard liquor. Why is hard liquor better than beer? b. ‘Do you want a bottle of wine?’ Mac asks. ‘I think I’ll have a drink-drink,’ I say, and when the waiter comes, I order a martini. c. … and drink prices are never over the top (around 3 euros a shot and 8 euros a drink-drink).
(Whitton 2008)
Generally speaking, the use of drink-drink is considered to denote ‘alcoholic drink’. However, given the relevant context, in (31a), drink-drink includes both hard liquor and beer, while in (31b) it refers to hard liquor only, in contrast to softer drinks like wine. Finally in (31c), once again the opposition set is different; this time the contrast is between a mixed drink and a shot (Whitton 2008). In other words, in (31) the meaning of drink-drink is so indeterminate that the lexical clone is almost uninterpretable without its contrast set or context. Next consider (32). (32) a. Oh, we’re just living together living together. b. Oh, we’re not living together living together. As Horn (2006b) and Ghomeshi et al. (2004) noted, the choice of the interpretation for the lexical clone here is determined by linguistic context. Whereas living together living together has the prototypical meaning of ‘living together as roommates’ in (32a), it is taken in the ‘value-added’ sense as ‘living together as lovers’ in (32b). (33) I don’t just teach pragmatics; I teach pragmatics-pragmatics. Here, the interpretation of the lexical clone depends crucially on who said (33). If it was said by a leading neo-Gricean such as Larry Horn, the inferred reading for pragmatics-pragmatics is likely to be ‘classical and neo-Gricean pragmatics’. On the other hand, if it was uttered by a founder of relevance theory such as Deirdre Wilson, the interpretation is likely to be ‘relevance theory’. Finally, in an appropriate context, the meaning of a lexical clone can even be reversed. This is the case for (34) (34) a. (Context: two people at a fast-food restaurant sharing one meal that comes with one drink) A: What do you wanna get? B: I’ll probably just get water so if you want a drink-drink get whatever you want
63
Neo-Gricean Pragmatics 63 b. A: … Come on girls I need some drink ideas. Please—Celeste B: Are you looking for alcohol? Or just a drink drink?
(Whitton 2008)
Given that fast-food restaurants normally do not serve alcohol, the contrast in (34a) is between water and soft drinks such as coca cola. In (34b), the inferred interpretation for drink-drink is opposite of the expected, prototypical sense. It means ‘non-a lcoholic drink’. Furthermore, (34b) does not represent an isolated case, and a number of similar examples can be found in the corpus collected by Hohenhaus (2004). One such example is given in (35). (35) Oh, I had a feeling—not a feeling feeling, but a real feeling. From the above discussion, we can conclude that (i) the vast majority of lexical clones are impossible to be interpreted properly out of context, and (ii) there are dimensions of contrast other than those in prototype and intensification, and all the dimensions on which the contrast set is ordered are provided by context (see also Whitton 2008). It is largely due to this heavy context dependency that lexical clones resist lexicalization, and therefore are unlikely to enter any dictionaries. They are in essence instances of nonce word formation, which are actively produced only in performance (e.g. Hohenhaus 2004). This can be evidenced by the response to the creation of the lexical clone rape-rape in (36). (36) ‘Hollywood has rallied behind Roman Polanski after his arrest in Switzerland over the weekend, with the actor Whoopi Goldberg suggesting that whatever he was guilty of it wasn’t “rape-rape”. As a guest on The View chatshow on US television, she said: “I know it wasn’t rape-rape. It was something else but I don’t believe it was rape-rape … ”
Responses from many female and male bloggers were swift and furious…. On The Frisky Amelia McDonnell-Parry wrote: “Whoopi Goldberg, who I never expected to be a rape apologist, coins a term I’ve never heard before—‘rape-rape’”.’
(From , 29 September 2009)
3.3.2.4 A neo-Gricean lexical pragmatic analysis By Grice’s cooperative principle, in using an innovative lexical clone XX (e.g. beautiful beautiful) in English, the speaker assumes and has good reasons to believe that the addressee can compute the meaning of XX. On the other hand, XX, being like a tautology, is superficially uninformative. Confronted with this blatant infringement of Grice’s maxim of Quantity, the addressee assumes that the speaker is actually cooperative, and has to figure out why he or she has used such an apparently uninformative construction. The only way to do this is to interpret it as highly informative. Assuming next that XX is marked as opposed to both X (e.g. beautiful) and YX (e.g. very beautiful), the actual
64
64 Yan Huang meaning of XX can then be worked out in terms of the systematic interplay between Levinson’s I-and M-principles coupled with context. Since structural constraints of English allow the unmarked X or YX to be used to express the meaning Z, the speaker will use it if such an interpretation is intended. On the other hand, if unmarked X or YX is not utilized, but marked XX is employed instead, then an M-implicature is created, namely, not only meaning Z, but something more than meaning Z as well is intended. In other words, we have an M-implicated contrast set between XX and X/YX here (see also Levinson 2000). This M-implicated opposition set may involve different dimensions of contrast in different contexts. Some dimensions are those on which the contrast is between prototype versus non-prototype, a ‘value-added’/intensifying versus a ‘non-value-added’/non-intensifying use or a literal versus a metaphorical meaning, as envisaged by the prototype intensification analysts; other dimensions of contrast may involve resolving reference (as in the case of the use of proper names), disambiguating expressions (as in the case of hyponymy), and narrowing generalities (as in the case of pragmatic looseness versus strictness) (see also Whitton 2008); still others may have to do with the contrast between descriptive versus non-descriptive meaning or between different connotations of the lexical items used. Very often, the use of lexical clones gives rise to some kind of novelty for sarcastic, humorous, or other rhetoric effects. Of these different dimensions of contrast, some are clearly truth-conditional, while others are clearly not. Also, as mentioned above in section 3.3.2.3, the way in which the contrast set is selected and ordered is determined by context, perhaps along the line of what Levinson (2000) called a Hirschberg scale. A Hirschberg scale is essentially a nonce scale, that is, a contextually given ad hoc scale. Such a scale can be based on any partially ordered contrast sets in a contextually salient way. Notice finally that an M-implicature generated here is a PCI rather than a GCI in the sense of Grice and Levinson.
3.3.3 Lexical blocking Lexical blocking or pre-emption refers to the phenomenon whereby the appropriate use of a lexical expression formed by a relatively productive process is apparently prevented by the prior existence of a synonymous but distinct lexical item (e.g. Huang 2009). This process applies to both derivation and inflection. For example, glory partially blocks *gloriosity, hospitalize (v) pre-empts *hospital (v), and went fully bars *goed. Furthermore, it can also take place between morphologically unrelated stems, as in queen precluding *kingess.9 9
Blocking can also extend to syntax, where the prior existence of a word can block an entire synonymous phrase. For example, while there are this morning and this evening, there is not *this night, which is frustrated by tonight. As with the parallel case of pre-emption by synonymy, there is also pre-emption by homonymy. For instance, while we can say They summered in New Zealand, we cannot say *They falled in Canada. This is because the latter is blocked by the established, salient verb form fall, as used in Something is falling from the sky (Clark 1993). According to Horn (2007a: 175), the tendency to avoid homonyms is Q-based.
65
Neo-Gricean Pragmatics 65 Aronoff (1976) noted that the existence of a simple lexical expression can restrict the formulation of an otherwise expected affixally derived form with the identical meaning. This is the case for (37b) and (37d), where a pre-existing simple abstract nominal underlying a given -ous pre-empts its nominalization with -ity, in contrast to (37a) and (37c). (37) a. b. c. d.
curious furious tenacious fallacious
curiosity *furiosity (fury) tenacity *fallacity (fallacy)
curiousness furiousness tenaciousness fallaciousness
Aronoff’s analysis was, however, called into question by Kiparsky (1983). On Kiparsky’s view, Aronoff’s account is both too strong and too weak. On the one hand, productive derivational processes are not always prevented by the existence of a more lexicalized alternative. This is evidenced by the fact that the abstract nominals gloriousness and furiousness coexist peacefully with glory and fury. On the other hand, blocking is not limited to derivation, but extends to inflection as well. As an alternative, Kiparsky suggested that Aronoff’s blocking paradigm be reformulated as a lexical analogue of the more general and ancient Elsewhere Condition, which can be traced at least back to Panini two millennia ago. However, the existence of partial blocking like contestant/contester, informant/informer, and refrigerant/refrigerator shows that Kiparsky’s prediction (‘Special rules block general rules in their shared domain.’) is still too powerful, because partial blocking corresponds to the phenomena that ‘the special affix occurs in some restricted meaning and the general affix picks up the remaining meaning’ (Kiparsky 1983; see also Horn 1984). As an attempt to accommodate these cases, Kiparsky put forward a generalization which he dubbed ‘avoid synonymy’ (see also Clark’s 1993 principle of contrast). (38) Kiparsky’s avoid synonymy condition The output of a lexical rule may not be synonymous with an existing lexical item. What (38) basically predicts is the pre-emption of potential synonyms by established terms. As an initial illustrating example, consider cook and cooker. Given the established meaning of cook (a person who cooks), it constrains cooker with that meaning, but not with a different meaning ‘an appliance which cooks’. Here, in terms of derivational morphology, verb-to-noun (cook (v)–cook (n)) conversion takes place at level 1, whereas -er is attached at level 2. This represents partial blocking. Another case in point involves the singular and plural forms of fish. The usual plural form of fish, namely fish, blocks fishes as its plural form, but does not block it when its sense is ‘different kinds of fish’. Differences in meaning are the reason why we may find a
66
66 Yan Huang pair like childish/childlike, informant/informer, brethren/brothers, productivity/productiveness, and Israelite/Israeli. Finally, once again, (38) can provide a satisfactory denouement to why ethnicity and ethnicness can sometimes coexist peacefully. (39) a. The lanterns demonstrated the ethnicity of the restaurant. b. The lanterns demonstrated the ethnicness of the restaurant. As pointed out by Riddle (1985), the two sentences have different meanings. In (39a), the lanterns indicate to which ethnic group the restaurant belongs, but in (39b), they show that the restaurant has an ethnic appeal. Furthermore, if two words are from very different registers or have widely differing connotations, blocking will not occur, either. This is the case for begin and commence. Our final example comes from lexical change. After the Norman Conquest in 1066, English speakers at court were faced with two sets of terms for animals: one from English (calf, cattle, deer, pig, sheep … ) and the other from French (veau, boeuf, venaison, porc, mouton … ). Given (38), the French terms were prevented from becoming synonymous with its English counterparts. Eventually, English and French terms are assigned different extensions: the original English terms for animals, but the terms borrowed from French for food. Thus we have the familiar pairs: calf/veal, cattle/beef, deer/venison, pig/pork, and sheep/mutton. This has the effect that the use of a food-denoting term usually blocks the conceptual grinding mechanism with regard to the use of an animal-denoting one, as in (40). (40) John doesn’t like eating pork/?pig. But lexical blocking of this kind can be cancelled under certain conditions, resulting in what Blutner (2004) called deblocking. For example, Nunberg and Zaenen (1992) noted that the use of cow rather than beef is more appropriate in (41). The same can be said of *liver in the sense of ‘a person who lives’, as in (42).10 (41) Hindus are forbidden to eat cow/?beef. (42) a. Is life worth living? It depends on the liver. (Otto Jespersen) b. The country for easy livers, The quietest under the sun. (Oxford English Dictionary) c. This woman, Mr Allen said, was a very high Upper East Side liver, who had a precipitous drop and had to downsize radically. (International Herald Tribute, 20–1 July 2013) 10
Deblocking can also be found in syntactic blocking. For instance, *this night can be used in Why is this night different from all other nights?
67
Neo-Gricean Pragmatics 67 This shows that there is a pragmatic base for lexical blocking and deblocking. We have a neo-Gricean pragmatic explanation for the process here, along the lines of Horn (1984). By Horn’s division of pragmatic labour or Levinson’s resolution schema, the process, namely less productive/lexicalized/marked/irregular forms block more productive/lexicalized/unmarked/regular forms in the same slot, is the result of the systematic interaction between the R/I-and Q/M-principles. Given the R/I-principle, a speaker tends to use the more productive form, because it is the one which is more readily available. On the other hand, by the Q/M-principle, he or she tends to employ the less productive form to achieve some special effects. Therefore, this is a case of unmarked forms tending to be used to convey unmarked messages, and marked forms to convey marked messages, in the interface between pragmatics and morphology. Next, in a pioneering study of the role played by Gricean conversational implicature in the lexicon, McCawley (1978) discussed a number of cases of partial blocking, outside the area of derivation and inflection. One case concerns the formulation of colour terms in English. As observed, pale red is far less frequently used than, say, pale blue, pale green, and pale yellow. This is because while English has no lexical item for pale blue, pale green, and pale yellow, it has a lexical item for pale red, namely pink. Furthermore, what is of interest is the fact that pale red is found to be used occasionally. When it is used, it denotes a colour other than pink, that is, a colour that is paler than red but not as pale as pink. This indicates that pale red is partially blocked by pink. (43) a. Mary wore a pink skirt yesterday. b. Mary wore a pale red skirt yesterday. c. +> Mary wore a skirt yesterday whose colour can’t be described exactly as pink Again, the contrast shown between (43a) and (43b) falls out naturally from Horn’s division of pragmatic labour or Levinson’s resolution schema. While the use of (43a) engenders a straightforward R/I-implicated stereotypical interpretation, given that pink and pale red form an M-contrast set, the use of the marked (43b) M-implicates (43c). On the other hand, since there is no colour term to block the use of or to form an M-contrast set with pale blue, the use of pale blue (which means whitish blue) does not carry any M-implicated extra meaning. Finally, McCawley pointed out that the distribution of a productive or periphrastic causative is also affected by the existence of a corresponding lexical causative. Whereas the use of a lexical causative, as in (44a), tends to depict a stereotypical, direct causative situation via the R/I-principle, the use of a productive or periphrastic causative, as in (44b) tends to refer to a more marked, indirect causation via the M-principle, hence the M-implicature in (44c). On the other hand, when there is no unmarked lexical causative, the M-implicated contrast does not appear. Consequently, the productive or periphrastic causative has a stereotypical interpretation, as in (45) (e.g. Levinson 2000: 141).
68
68 Yan Huang (44) a. John opened the safe. b. John caused the safe to open. c. +> John opened the safe in an unusual way (45) John made Mary laugh. (cf. *John laughed Mary.)
3.3.4 Lexicalization asymmetry: Logical operators Consider the traditional Square of Opposition formulated by Boethius out of Aristotle two millennia ago. (46) Square of Opposition
Clearly, there is an asymmetry in the lexicalization of logical operators on the square. Whereas the A, E, and I vertices can all be lexicalized, the O vertex cannot. This is schematized in (47). (47) Quantifiers Adverbs Connectives Modals
A all always and must
I some sometimes or may
E none never nor must not
O not all/*nall not always/*nalways and not /*nand permit not *permitn’t
Furthermore, as pointed out by Horn (1989: 252–267), the lexicalization asymmetry seems cross-linguistic. In other words, there is a strong cross-linguistic tendency for the O corner not to be represented in the lexicon, as data drawn from Greek, Latin, French, Japanese, and Malagasy show. Rather, it is nearly always encoded by complex phrases.
69
Neo-Gricean Pragmatics 69 Why is this the case? To answer this question, we need to consider what is the relation between the subcontrary I and O vertices of the square. Put slightly differently, the question boils down to whether the I/O relation is a logical one or not. On Aristotle’s view, the relation is logical as far as the modals are concerned, but in the case of the quantifiers, it is non-logical. This was disputed by Hamilton and Jespersen, who held that the relation is logical for all the squares. By contrast, for De Morgan and J. S. Mill, it represents a non-logical relationship for all the squares (Levinson 2000: 68). Given neo-Gricean pragmatic theory, the answer to the puzzle presents itself. Notice that the vertices always form the Horn scales of and . This has the consequence that the assertion of I Q-implicates ‘~A’, which is the contradictory of A, and hence equivalent to O. Thus, the I and O corners are related by nothing but a Q-scalar implicature. Furthermore, given that conversational implicatures are cancellable, the I/O relation is a non-logical one. Let me now return to the question raised at the beginning of this subsection, namely, why only the O value resists lexicalization. The answer is straightforward within the neo-Gricean pragmatic framework: what is Q-implicated on the square is not lexicalized. But such an explanation raises a further question, namely, given that the I and O vertices have the same communicational load, why the I corner can be and indeed is lexicalized. The answer, according to both Horn (2006a) and Levinson (2000: 70–71), can be sought in the relatively complex nature or functional markedness of negation. Given a choice between a positive and a negative term with the same communicational load, the positive term is usually picked up as the basic form to be lexicalized.
3.4 Neo-Gricean Pragmatics and Semantics: Pragmatic Intrusion into What Is Said, Grice’s Circle, and the Pragmatics–Semantics Interface On a classical Gricean account, a distinction is made between what is said and what is conversationally implicated. However, as pointed out by Levinson (2000: 170), Grice’s characterization of what is said is quite complex, though it may roughly be represented as follows. (48) Grice’s concept of what is said U said that p by uttering x if and only if: a. x conventionally means p b. U speaker-meant p c. p is the conventional meaning of x minus any conventional implicature
70
70 Yan Huang Given (48), what is said is defined rather conservatively by Grice. It is in general taken to be (i) the conventional meaning of a sentence uttered with the exclusion of any conventional implicature, and (ii) the truth-conditional, propositional, semantic content of the sentence uttered (e.g. Grice 1989a: 25; Levinson 2000: 170). What is conversationally implicated is then defined in contrast to, and calculated on the basis of, what is said (and in the case of M-implicatures, together with how what is said is said). Stated in this way, what is said is supposed to provide input to what is conversationally implicated. This original, minimal Gricean conception of what is said, however, has generated a heated debate in a neo-Gricean way. First, what is said is considered ambiguous in at least three ways: (i) understood in ‘a technical and artificial strict sense’ (Salmon 1991); (ii) interpreted in an ordinary, everyday sense of ‘what is stated’ (Levinson 2000: 194); and (iii) taken to be ambiguous between ‘general content’ (i.e. i-content or what is saidmin) in one technical sense and ‘contextually enriched content’ (i.e. c-content or what is saidmax) in another everyday sense (Recanati 2010). On another dimension, the ambiguity of what is said lies in it being used in its ‘locutionary’ and ‘illocutionary’ sense (Bach 2001; see also Huang 2010c). Secondly, there are two current, opposing positions on the scope of what is said: (i) the minimalist and (ii) the maximalist. In the first, more Gricean, minimalist camp, what is said has to be ‘closely related to the conventional meaning of the … sentence … uttered’ and must correspond to ‘the elements of [the sentence], their order and their syntactic character’ (Grice 1989a: 87), which is called the ‘syntactic correlation constraint’ by Bach (2001). Consequently, no aspects of pragmatically enriched content that is not directly linked with either the conventional content or the syntactic structure of the sentence uttered can be part of what is said. This position is represented by e.g. the neo-Griceans Bach (2001, 2012) and Horn (2012a, 2012b). By contrast, neo-Griceans such as Levinson (2000) and Recanati (2004a, 2010) (and relevance theorists like Sperber and Wilson 1995 and Carston 2002) argued for a wider, more pragmatic, maximal notion of what is said, allowing pragmatics to intrude onto the truth-conditional, propositional content of what is said in the Gricean sense (see also Huang 2010c). What mainly prompts the maximalists to endorse a broader, more pragmatic concept of what is said? The answer: the widely accepted doctrine known as the linguistic underdeterminacy thesis. (49) The linguistic underdeterminacy thesis The linguistically encoded meaning of a sentence radically underdetermines the proposition a speaker expresses when he or she utters that sentence. What, then, is the evidence? First, as pointed by Grice (1989a: 25), before we work out what is said in his sense, we have (i) reference to identify, (ii) deixis to fix, and (iii) ambiguity to resolve. To (i)–(iii), Levinson (2000: 172–186) added (iv) ellipsis to unpack and (v) generalities to narrow. It turns out, however, that the determination of (i)–(v) involves pragmatically embellished meaning of some kind.
71
Neo-Gricean Pragmatics 71 Secondly, on many occasions of use, propositions associated with what is said in the classical Gricean sense contain unarticulated constituents (UCs)—propositional or conceptual elements of a sentence that is not explicitly expressed linguistically (e.g. Huang 2012b). Three stock examples are given in (50). They are the three sentences without brackets. On the other hand, the sentences with brackets contain the possible, pragmatically embellished propositional or conceptual material for the UCs. (50) a. Gentlemen prefer blonds [to brunettes]. b. Clinton made a lot of noise and had a very big impact. He has the [strong] shoulders. If he comes, the world is interested, and it brings attention to Lesotho. When I climb on the foothills of Clinton, perhaps people will notice more. (FT Magazine 19–20 August 2006) c. [The novels written by] Jane Austin is/are on the top of the shelf. Clearly the recovery of the UCs in (50) requires pragmatics. The sentence without the brackets in (50a) does not express a complete proposition. Consequently, it cannot be evaluated truth-conditionally. Therefore, it undergoes what Bach called a pragmatic process of completion and what Recanati called saturation, to provide extra propositional or conceptual material to fill in the UC, thus making it fully propositional. Completion or saturation is typically a linguistically mandated, bottom-up process. By contrast, while the proposition expressed by the uttering of the sentence without brackets in (50b) is a complete though minimal one, it falls short of what the speaker intends to convey. As a consequence, it needs to be expanded. The pragmatic process of expansion will flesh out the proposition expressed by the sentence uttered and engenders a richer proposition. Expansion is typically an optional, contextually driven, top-down process, a subtype of free enrichment, which is itself a subtype of meaning modulation (Recanati 2010). Finally, (50c) involves the pragmatic process of semantic transfer. In this type of pragmatic process, the output proposition is neither an enriched nor an impoverished version of the concept literally expressed by the input one; rather it represents a different concept, provided that there is a salient functional relation between the old and new concepts. Once again, some pragmatically enriched content intrudes into the conventional, truth-conditional content of what is said. Thirdly and finally, following in the footsteps of Cohen, Wilson, Atlas, and Gazdar, Levinson (2000) argued that contrary to Grice, conversational implicature can encroach upon truth-conditional content. In particular, he claimed that the classic Cohen–Wilson argument can be extended into logical connective constructions such as conditionals (51), comparatives (52), disjunctions (53), and because-clauses (54). (51) If her daughter gets married and has children, Mary will be happy. (52) Brushing your teeth and going to bed is better than going to bed and brushing your teeth.
72
72 Yan Huang (53) Mary’s daughter either got married and had children or had children and got married—I don’t know which. (54) Because some of her students came to her seminar, Dr Smith was disappointed. These constructions are labelled ‘intrusive’ constructions by Levinson.11 The reason is that in these constructions, ‘the truth conditions of the whole depend in part on the [conversational] implicatures of the parts’ (Levinson 2000: 198). The truth-conditional content of (51)–(53) rests crucially on the generalized I-implicature stemming from the use of and for ‘and then’. On the other hand, the quantifier some in (54) has to be Q- implicated to ‘some but not all’. Thus, there seems no avoiding the conclusion that the truth condition of the complex construction has to be calculated taking into account the conversational implicature of its part. The question that arises next is what is the pragmatic intrusion under consideration? Roughly, two current positions can be identified. The first is that the pragmatic intrusion is of a special kind, which differs from conversational implicature. Within this camp, three lines of arguments are of particular interest. According to Sperber and Wilson (1995), the pragmatic inference is an explicature, which is a development of the linguistically given logical form of the sentence uttered. Secondly, there is the position taken by Recanati (1993, 2004a, 2010) that it is the pragmatically enriched part of what is said. A third argument is due to Bach (2004), in which he proposed a third category of communicative content, intermediate between what is said and what is implicated. Bach dubbed the vehicle of such a content ‘impliciture’, because it is implicit in what is said. By contrast, the second position is represented by Levinson (2000). On Levinson’s view, pragmatic intrusion into what is said is neither an explicature, nor the pragmatically enriched said, nor an impliciture. Rather, it is the same beast as a neo-Gricean conversational implicature (see also Huang 1991, 1994, 2000a, 2006a, 2007, 2014). Levinson’s neo-Gricean proposal that pragmatics, and in particular conversational implicature, can feed into rather than just read off what is said, however, is strongly challenged by Horn (2004, 2006a, 2012a, 2012b) within the neo-Gricean camp (see also King and Stanley 2005). On Horn’s view, what Levinson has argued is inconsistent with the spirit of the original Gricean programme. In the classical Gricean ‘Golden Age of Pure Pragmatics (GAPP)’, as Horn has dubbed it, pragmatics can only read off but not feed into what is said. Conversational implicatures, by definition, cannot make any contribution to the truth-conditional content of what is said. In other words, according to Horn, the Levinsonian revisionist model is much more neo-than Gricean. Horn’s own proposal is that we should instead adopt Bach’s neo-Gricean model, in which we would allow impliciture, which is built out of what is said, to contribute to what is said, whereas the traditional Gricean semantic concept of what is said, along with a post-semantic orthodox Gricean characterization of what is conversationally implicated, is retained in a neo-classical way. 11 Some of these constructions are taken to be cases of so-c alled ‘embedded [conversational] implicature’ (see e.g. Huang 2014: 68–73).
73
Neo-Gricean Pragmatics 73 If neo-Gricean conversational implicature can intrude onto truth-conditional content, this gives rise to a problem known as Grice’s circle, namely how what is conversationally implicated can be defined in contrast to, and calculated on the basis of, what is said, given that what is said seems both to determine and to be determined by what is conversationally implicated. Levinson’s solution is that we should reject the ‘received’ view of the pragmatics–semantics interface, according to which the output of semantics provides input to pragmatics, which then maps literal meaning to speaker meaning. Rather, we should allow neo-Gricean pragmatics/conversational implicature to play a systematic role in ‘pre’-semantics, i.e. to help determine the truth-conditional content of the sentence uttered. As Levinson (2000: 242) told us: There is every reason to try and reconstrue the interaction between semantics and pragmatics as the intimate interlocking of distinct processes, rather than, as traditionally, in terms of the output of one being the input to the other.
Putting it slightly differently, in order to avoid Grice’s circle, we need both ‘pre-’ and ‘post-’semantic pragmatics, or something like what Korta and Perry (2011) called ‘near-’ and ‘far-side’ pragmatics. Such a radical proposal amounts to saying that the whole architecture of the theory of meaning needs to be radically reconstructed.
3.5 Neo-Gricean Pragmatics and Formal Syntax: The Case of Anaphora and Binding Finally, I turn to the pragmatics–syntax interface, concentrating on anaphora and binding.
3.5.1 Anaphora Anaphora can be defined as a relation between two linguistic elements, in which the interpretation of one (called an anaphoric expression) is in some way determined by the interpretation of the other (called an antecedent). Linguistic expressions that can be employed as an anaphoric expression include gaps (or empty categories), pronouns, reflexives, proper names, and definite descriptions (e.g. Huang 2000a).
3.5.2 Chomsky’s binding conditions Within principles-and-parameters theory and its minimalist descendant, Chomsky (1995b) distinguished two types of abstract feature for NPs: anaphors and pronominals.
74
74 Yan Huang An anaphor is a feature representation of an NP which must be referentially dependent and which must be bound within an appropriately defined minimal syntactic domain; a pronominal is a feature representation of an NP which may be referentially dependent but which must be free within such a domain. Interpreting anaphors and pronominals as two independent binary features, Chomsky hypothesized that we ideally expect to find four types of NP in a language—both overt and non-overt. (55) Chomsky’s typology of NPs a. b. c. d.
[+anaphor, −pronominal] [−anaphor, +pronominal] [+anaphor, +pronominal] [−anaphor, −pronominal]
Overt lexical anaphor pronoun - name
Empty NP-trace pro PRO wh-trace/variable
Of the three types of overt NP listed in (55), anaphors, pronominals, and r[eferential]- expressions are subject to binding conditions A, B, and C respectively. (56) Chomsky’s binding conditions A. An anaphor is bound in a local domain. B. A pronominal is free in a local domain. C. An r-expression is free. Binding is defined in configurational terms, appealing to purely structural concepts like c-command, government, and locality. Binding theory is supposed to account for the syntactic distribution of the three types of overt NP listed in (55). Consider, for example, (57) from English. (57) a. Newton1 admired himself1. b. Newton1 admired him2. c. Newton1 admired Newton2. In (57a), himself is an anaphor in the Chomskyan sense. As such, it falls under binding condition A, according to which, it is bound to its local antecedent Newton. Next, in (57b), him, being a pronominal, is subject to binding condition B. Given binding condition B, it cannot be bound in its local domain, and there is thus disjoint reference between it and Newton. Finally, in (57c), the second Newton is an r-repression. By binding condition C, it cannot be coindexed with the first Newton. From examples like these, Chomsky concluded that the syntactic distribution of anaphors, pronominals, and r-expressions is accounted for by binding conditions A, B, and C, respectively. However, when confronted with a wider range of languages other than English, these binding conditions run into serious difficulties (see e.g. Levinson 1987a, 1991, 2000 and Huang 1991, 1994, 2000a, 2004a, 2006b, 2007, 2014 for detailed discussion).
75
Neo-Gricean Pragmatics 75
3.5.3 Elimination of binding conditions? A current development in the Chomskyan syntactic analysis of binding is to eliminate all the conditions that are postulated specifically for binding, such as Chomsky’s binding conditions A, B, and C, discussed in section 3.5.2, and to reduce these specific conditions to elementary, general, and independent principles of the computational system of language within Chomsky’s minimalist programme. This trend is represented by Reuland (2011). Based mainly on his study of anaphora and binding in a range of Germanic and Romance languages, Reuland argued that anaphoric dependencies and their (substantial) cross-linguistic variations (in these languages) can be derived natu rally by the interaction of the lexicon, syntax, semantics, and discourse components of the language system, which reflect the cognitive capacities and design of language. This new development constitutes an important step forward in our understanding of anaphora and binding. First, from a theoretical and methodological point of view, to account for anaphora and binding in terms of some elementary, general, and independent principles of Chomsky’s minimalist programme is preferable to explaining them by means of some ad hoc, local conditions that are stipulated specifically for them. Secondly, while their focus is still on the syntax and semantics of anaphora and binding, the latest generative analyses of the kind exemplified by Reuland (2011) allow non-grammatical or pragmatic factors to play a role. On the other hand, however, accounts of this type are not without problems of their own. Again, take Reuland (2011) as an example. Theoretically and methodologically, Reuland’s analysis still suffers from containing too many rules, conditions, and principles. In other words, the theoretical machinery is not parsimonious enough. Next, from an empirical perspective, it is unclear to what extent Reuland’s analysis can extend to languages other than the Germanic and Romance ones he has examined. For example, within Reuland’s framework, the syntactic binding of a morphologically simplex anaphor (e.g. zich in Dutch, ziji in Chinese, and caki in Korea) needs to enter an A-chain that is created by movement. In order for such an anaphor to enter an A-chain, structural Case must be checked, which is achieved via the inflectional system of a language. While this analysis may work for a Germanic/Romance language, it does not apply to, for example, Chinese, because Chinese is a language which lacks any inflectional mechanism. Consequently, given that in Reuland’s system, no structural Case checking means no A-chain, binding of a morphologically simplex anaphor in a language like Chinese has to be ruled out in narrow syntax, though allowed in pragmatics.
3.5.4 The revised neo-Gricean pragmatic theory of anaphora As an alternative to various syntactic and semantic approaches, a neo-Gricean pragmatic theory of anaphora was developed by Levinson (1987a, 1991, 2000) and Huang (1991, 1994, 2000a, 2000b, 2004a, 2006b, 2007, 2014; Chiou and Huang 2010). The
76
76 Yan Huang central idea underlying the theory is that the interpretation of certain patterns of anaphora can be made utilizing pragmatically enriched meaning, such as conversational implicature, dependent on the language user’s knowledge of the range of options available in the grammar, and of the systematic use or avoidance of particular anaphoric expressions or structures on particular occasions. Applying the Q-, I-, and M-principles to the domain of anaphora, we can derive a revised neo-Gricean pragmatic apparatus for the interpretation of various types of anaphoric expressions. (58) Huang’s revised neo-Gricean pragmatic apparatus for anaphora (simplified) (i) The use of an anaphoric expression x I-implicates a local coreferential interpretation, unless (ii) or (iii). (ii) There is an anaphoric Q-scale , in which case the use of y Q-implicates the complement of the I-implicature associated with the use of x in terms of reference. (iii) There is an anaphoric M-scale {x, y}, in which case the use of y M-implicates the complement of the I-implicature associated with the use of x, in terms of either reference or expectedness. Needless to say, any interpretation generated by (58) is subject to the general consistency constraints applicable to conversational implicatures. These constraints include real-world knowledge, contextual information, and semantic entailments. Let me now return to Chomsky’s binding conditions and see how they can be reinterpreted in pragmatic terms. On the neo-Gricean pragmatic account, Chomsky’s binding conditions B and C need not be laid at the doorstep of generative syntax, and can be reduced to pragmatics. In somewhat simplified terms, this can be achieved in the following way. If binding condition A is taken to be either grammatically constructed (as in the English-type, syntactic languages) or pragmatically specified via the I-principle (as in the Chinese-type, pragmatic languages), then binding condition B can be pegged directly to the application of the Q-principle. Given a speaker’s knowledge of grammar and the I-principle, a reflexive will be chosen if coreference is intended. This has the consequence that if the reflexive is not employed but a pronoun is used instead, a Q-implicature will arise, namely no coreference is intended. In other words, we have a Horn scale here such that the use of a semantically weaker pronoun Q-implicates that the more informative, coreferential interpretation associated with the use of the reflexive cannot be truthfully entertained, as in (57b). By the same reasoning, binding condition C can also be eliminated. Wherever a reflexive could occur, the use of a semantically weaker proper name Q-implicates the non-applicability of the more informative, coreferential interpretation associated with the use of the reflexive. This is exactly what has happened in (57c). Furthermore, the revised neo-Gricean pragmatic theory can provide an elegant account of many of the anaphoric patterns that have embarrassed a generative analysis, such as the case where
77
Neo-Gricean Pragmatics 77 contra binding condition B, a pronoun is bound in its local domain. In the case of long-distance reflexivization, the concept of unexpectedness is invoked to explain why such a marked anaphoric expression is used. Examined in a more careful way, cross- linguistically, unexpectedness turns out to be mainly of three types: (i) emphaticness/ contrastiveness, (ii) logophoricity, and (iii) de se attitude/belief ascription. First, long- distance reflexives are used for emphatic or contrast marking. A second dimension of unexpectedness arising from the employment of long-distance reflexives involves logophoricity—the phenomenon whereby the ‘point of view’ of an internal protagonist of a sentence or discourse, as opposed to that of the current, external speaker, is being reported using some morphological and/or syntactic means. The term ‘point of view’ is employed here in a technical sense and is intended to encompass words, thoughts, knowledge, emotion, and perception (e.g. Huang 1994, 2000a, 2002, 2004a, 2006b, 2007, 2010b, 2013b, 2014). Thirdly and finally, long-distance reflexives can be utilized to encode a de se attitude/belief ascription, that is, a self-locating attitude/belief ascription (e.g. Huang 2013a). This use of long-distance reflexives to mark unexpectedness is accountable in terms of the M-principle. Since the grammar allows the unmarked pronoun to be employed to encode coreference, the speaker will use it if such a reading is intended. On the other hand, if the unmarked pronoun is not used, but the marked long-distance reflexive is employed instead, then an M-implicature will be licensed. The conversational implicature is that not only coreference but also logophoricity/de se attitude/belief ascription is intended by the speaker. Notice another correlation here. If relevant, the choice between pronouns on the one hand and logophoric/de se long-distance reflexives on the other is correlated with that between indicative and subjunctive mood in the embedded clause. In some languages, the use of a pronoun tends to go with that of indicative mood; the employment of a logophoric/de se long-distance reflexive tends to go with subjunctive mood. This correlation is a reflection of a semantic/pragmatic choice made by the external speaker about the responsibility he or she assumes for the truthfulness of what he or she is reporting. If a pronoun and indicative mood are used, it is indicative that the speaker asserts that the report is true. On the other hand, if a logophoric/de se long-distance reflexive and subjunctive mood are deployed, it shows that the speaker does not take responsibility for the truth of the report. This is the case for Icelandic. Since its inception, the (revised) neo-Gricean pragmatic theory of anaphora has been the impetus to a substantial amount of research and has been applied to a wide range of languages as genetically unrelated and structurally diverse as Korean, Spanish, and Turkish. In Huang (2000a) and Levinson (2000), substantial cross-linguistic evidence was presented to show that the revised neo-Gricean pragmatic theory of anaphora is more adequate than both a syntactic and a semantic approach. This indicates that pragmatics and syntax are intimately interconnected, though they are distinct levels and modes of explanation in linguistic theory. Contrary to the popular but erroneous Chomskyan view that syntax is autonomous, pragmatics plays a crucial role in explaining many of the phenomena that are thought to be at the very heart of syntax. If this is the case, then a large portion of linguistic explanation which is currently
78
78 Yan Huang sought in syntactic terms may need to be shifted to pragmatics—hence the interaction and division of labour between pragmatics and syntax. This interface and division of labour may be summarized in a Kantian apophthegm: pragmatics without syntax is empty; syntax without pragmatics is blind (Huang 1994: 259, 2000a: 213, 2014: 368). In addition, the revised neo-Gricean pragmatic theory of anaphora has important theoretical implications for universals, innateness, and learnability.
Acknowledgement This chapter is a thoroughly revised and updated version of Huang (2010a, 2015b). I am grateful to Larry Horn for his insightful comments on some of the material contained in this chapter. The usual disclaimers apply.
79
Chapter 4
R eleva nce Theory Deirdre Wilson
4.1 Introduction One of the most original features of Grice’s approach to communication was his view that meaning is primarily a psychological phenomenon and only secondarily a linguistic one: for him, speaker’s meanings are basic and sentence meanings are ultimately analysable in terms of what speakers mean (Grice 1957, 1967). Despite this reference to psychology, Grice’s goals were mainly philosophical or semantic: his analysis of speaker’s meaning was intended to shed light on traditional semantic notions such as sentence meaning and word meaning, and his accounts of the derivation of implicatures were rational reconstructions of how a speaker’s meaning might be inferred, rather than empirical hypotheses about what actually goes on in hearers’ minds. Relevance theorists have been trying to develop Grice’s insights in a different direction, by incorporating them into a psychologically plausible, empirically testable theory of overt (‘ostensive’) communication. Relevance theory, like other broadly Gricean approaches to pragmatics, takes as its starting point three of Grice’s assumptions about verbal communication. The first is that a sentence meaning is a vehicle for conveying a speaker’s meaning, where a speaker’s meaning is an overtly expressed intention that is fulfilled by being recognized. The second is that a speaker’s meaning cannot be simply perceived or decoded, but has to be inferred from her behaviour, together with contextual information. The third is that in inferring a speaker’s meaning, the hearer is guided by the expectation that communicative behaviour should meet certain standards: for Grice, a cooperative principle and conversational maxims, and for relevance theorists, a presumption of optimal relevance. However, relevance theory also differs from Grice’s framework in several important respects. One important difference has to do with the scope of pragmatics. Grice’s theoretical definition of speaker’s meaning 1 (‘meaning Nn’) was designed to apply to 1 Grice’s term was ‘utterer’s meaning’, where an ‘utterance’ is an overtly intentional attempt— whether verbal or non-verbal—to produce a certain cognitive or behavioural response in an audience.
80
80 Deirdre Wilson certain cases of non-verbal communication but exclude others. For instance, when Mary shows Peter her bandaged leg in response to his invitation to play squash, Grice (1967/1989a: 109) notes that although she overtly intended to make Peter believe both that her leg was bandaged and that she couldn’t play squash, it seems appropriate to describe her as meaning that she couldn’t play squash, but not as meaning that her leg was bandaged. He therefore added a third clause to his definition of speaker’s meaning, designed to exclude this second type of case. While Sperber and Wilson (1986/1995: 53–54) share Grice’s intuition that use of the ordinary-language term ‘meaning’ is inappropriate in certain cases, they argue that the resulting definition of speaker’s meaning does not pick out a natural class of phenomena, since the cases Grice wants to exclude fall under exactly the same generalizations as those he wants to include. They therefore propose a broader definition of ostensive communication which covers both ‘showing’ and ‘telling’, and which subsumes prototypical Gricean speaker’s meanings as a special case. This difference matters not only for non-verbal communication but also for many verbal cases where ‘telling’ and ‘showing’ combine, or where the intended effect is not easily rendered as a single proposition that the speaker can be said to have meant (as with stylistic or poetic effects). These fall squarely within the scope of a theory of ostensive communication, while in Grice’s framework they receive only partial treatment at best.2 A second difference has to do with the role of pragmatic principles or maxims in utterance comprehension. Grice was mainly concerned with pragmatic factors affecting the identification of implicatures, whereas relevance theorists (and a growing number of other broadly Gricean approaches) are equally concerned with pragmatic factors affecting the identification of explicit truth-conditional content.3 In Grice’s framework, moreover, communicators can blatantly violate a maxim in order to trigger the search for an implicature, whereas blatant violation of pragmatic principles or maxims has no comparable role in relevance theory. In particular, relevance theorists question the need for Grice’s first Quality maxim (‘Do not say what you believe to be false’)—which seems to have no other function in Grice’s framework than to be violated in figurative utterances—on two main grounds. First, there are alternative accounts of figurative utterances which involve no blatant maxim violation and fit better with current processing models.4 Second, for Grice’s account to work, the speaker must blatantly violate the first Quality maxim by saying something literally false; yet for Grice, saying involves speaker commitment, and in figurative utterances, the speaker precisely does not commit herself to the truth of the proposition literally expressed. Indeed, Grice generally describes the speaker of a 2
See Sperber and Wilson (1986/1995: 50–60); Wharton (2009); Sperber and Wilson (2015). See e.g. Wilson and Sperber (1981, 1993); Sperber and Wilson (1986/1995: 182–183); Carston (1988, 2002, this volume); Bach (1994); Levinson (2000); Recanati (2004a, 2010). 4 See e.g. Wilson and Carston (2006, 2007, 2008); Sperber and Wilson (2008); Wilson and Sperber (2012a). 3
81
Relevance Theory 81 figurative utterance as merely ‘making as if to say’ something, and in that case, it is hard to see how his first Quality maxim is violated at all.5 A third difference has to do with how far communication is cooperative in Grice’s sense. According to Grice (1967/1989a: 26), participants in a talk exchange are expected (ceteris paribus) to observe his cooperative principle (‘Make your conversational contribution such as is required, at the stage at which it occurs, by the accepted purpose or direction of the talk exchange in which you are engaged’), and pragmatic inferences— including conversational implicatures—are derivable only on the assumption that the cooperative principle is being observed. According to Sperber and Wilson (1986/ 1995: 161–162), by contrast, the only purpose a genuine communicator and a willing audience have to share is that of achieving uptake, i.e. getting the audience to recognize the communicator’s overtly expressed intention to inform them of something. While some communicative exchanges are cooperative in Grice’s sense, the presence of an accepted purpose or direction is seen in relevance theory as just one of many contextual factors that can play a role in comprehension. It may be outweighed by other factors (as when a speaker breaks off to comment on some unexpected event), or it may be entirely absent (as in casual conversation, hostile cross-examination, or subtler forms of conversational tug-of-war).6 This difference has implications for comprehension. In Grice’s framework, a speaker can conversationally implicate that she is unable to provide some required information (since in order to provide it, she would have to violate the Quality maxims) but not that she is unwilling to provide it, since in this case she would be violating the cooperative principle itself. For relevance theorists, inferences about the speaker’s abilities and preferences play an equal and parallel role in comprehension, so a speaker who fails to provide some required information may be just as well understood (in appropriate circumstances) as implicating that she is either unwilling or unable to provide it.7 Relevance theory, then, treats utterance comprehension as an inferential process which takes as input the production of an utterance by a speaker, together with contextual information, and yields as output an interpretation of the speaker’s meaning. Utterance comprehension is seen as essentially an exercise in mindreading, and the challenge for relevance theorists attempting to build a psychologically plausible, empirically testable pragmatic theory is precisely to explain how the closed formal system 5 Notice that what relevance theorists deny is the claim expressed in Grice’s first Quality maxim, that speakers are expected to say something literally true. It follows from the presumption of relevance that speakers are expected to communicate something true, since an utterance cannot be relevant without achieving some true cognitive effects (Wilson and Sperber 2002). 6 Grice (1989a: 369–370) describes these as ‘degenerate, derivative’ cases which ‘honor the cooperative principle at least to the extent of aping its application’, and adds that ‘collaboration in achieving exchange of information … may co-exist with a high degree of reserve, hostility, and chicanery and with a high degree of diversity in the motivations underlying quite meagre common objectives’. Given these substantial qualifications, it is not obvious what his insistence on the need for his cooperative principle brings to his account. 7 This difference has implications for the treatment of scalar inferences; see e.g. Sperber and Wilson (1986/1995: 272–278); Carston (1998); Noveck and Sperber (2007).
82
82 Deirdre Wilson of language provides effective pieces of evidence which, combined with contextual information, enable successful comprehension to take place. The current version of the theory is a result of many years of collective endeavour by linguists, philosophers, and cognitive scientists.8 Here I will briefly outline some of its main assumptions (while attempting to clear up some common misconceptions), and point out some recent developments and new directions for research.
4.2 Relevance and Cognition Relevance theory is grounded in a definition of relevance and two principles of relevance: a cognitive and a communicative principle. The definition of relevance was not intended to capture any of the ordinary-language senses of the word ‘relevance’, but to provide a useful theoretical concept which picks out an important psychological property and has enough in common with these ordinary-language senses to justify the name (Sperber and Wilson 1986/1995: 119). Some commentators see the resulting theoretical concept as not having enough in common with the everyday sense of the word.9 While this is a largely terminological matter, it does raise a substantive issue about the nature of relevance, and is worth a brief response. Most linguists attempting a theoretical definition of relevance start from notions such as ‘relevance to a purpose’, ‘relevance to a question’, or ‘relevance to a topic’, where the purpose, question, or topic is fixed (or sharply delimited) in advance of the comprehension process rather than identified in the course of comprehension. Sperber and Wilson (1986/1995: ch. 3) start by defining two more general notions: ‘relevance in a context’ and ‘relevance to an individual’. A context comprises mentally represented information of any type—beliefs, doubts, hopes, wishes, plans, goals, intentions, questions, etc.—and is constructed or selected in the course of the comprehension process from a range of potential contexts available to the individual. One reason for treating these more general notions as basic is that considerations of relevance play a fundamental role not only in communication but in cognition. The aim is to define relevance not only for utterances or other communicative acts, but for any external stimulus or internal mental representation which can provide an input to cognitive processes, so that not only utterances but sights, sounds, smells, thoughts, memories, or conclusions of inferences may all provide potentially relevant inputs (for an individual, at 8
Recent book-length discussions include Blakemore (2002); Carston (2002); Wharton (2009); Wilson and Sperber (2012b); Clark (2013); Ifantidou (2014); Jodłowiec (2015); Zufferey (2015). Recent encyclopaedia entries include Wilson and Sperber (2004); Sperber and Wilson (2005); Carston and Powell (2006); Carston (2012, this volume). I will try to complement, rather than repeat, those earlier discussions where possible. 9 Levinson (2000: 55) describes Sperber and Wilson’s notion of relevance as ‘a very inadequate characterization of what pretheoretically would generally be considered the nature of relevance’, and Bach (2010a: 135) refers to ‘relevance theorists’ highly idiosyncratic and misleading use of the term’.
83
Relevance Theory 83 a time). While it is not implausible that many utterances contribute to an accepted purpose, question, or topic which is fixed in advance of the comprehension process, it is quite implausible that the same is generally true of cognition. Suppose you are watching a tennis match on television. You happen to notice that one of the players repeatedly bounces the ball eight times before serving, while the other bounces it once or twice and serves straight away. Is this information relevant to you? Intuitively, it will be relevant if it interacts with some contextual information you have available to achieve a worthwhile cognitive effect: for instance, by enabling you to draw conclusions about the characters of the players, confirming your suspicion that one of them is more nervous than the other, answering a question about which of them is more likely to win, helping you decide whether to go on watching the match, and so on. According to relevance theory, other things being equal, the greater the cognitive effect achieved, and the smaller the mental effort required, the more relevant this input will be to you at the time (Sperber and Wilson 1986/1995: 260–266). In this framework, relevance to a purpose, question, or topic is a special case of relevance in a context or relevance to an individual. Typically, an input achieves relevance in a context consisting of information of many different types, none of which seems to play a privileged role in the derivation of cognitive effects. Why did you notice the players’ ball-bouncing techniques, and why did you interpret them in the way you did? At any point in your waking life, a huge variety of potential inputs are competing for your attention, and a fundamental problem for human cognition is how to allocate attention and processing resources among them. According to relevance theory, as a result of constant selection pressures towards increasing cognitive efficiency, the human cognitive system has developed a variety of mental mechanisms or biases (some innate, others acquired) which tend to allocate attention to inputs with the greatest expected relevance, and process them in the most relevance-enhancing way. This claim is expressed in the cognitive principle of relevance (‘Human cognition tends to be geared to the maximization of relevance’) (Sperber and Wilson 1986/1995: 260–266). On this approach, you happened to notice the players’ ball-bouncing techniques because, given the organization of your cognitive system, of all the potential inputs competing for your attention, this one had the greatest expected relevance for you at the time; and you interpreted this input in the context of your knowledge of tennis and tennis players, plans for the afternoon, and so on, because given the organization of your cognitive system, of all the potential contexts available to you, this was expected to enhance its relevance most.10 More generally, what makes an input relevant to an individual is that it interacts with contextual information he has available to yield worthwhile cognitive effects (e.g. warranted conclusions, warranted strengthenings or revisions 10 Of course, there is no way of guaranteeing that the cognitive system will make the right choices on every occasion, and what it expects to be relevant on a given occasion may turn out not to be. Hence the reference in the cognitive principle to a tendency to maximize relevance.
84
84 Deirdre Wilson of available information), and what makes it maximally relevant to the individual is that it yields greater effects, for less effort, than any alternative input available to him at the time. A common objection to the cognitive principle is that it is too vague and general to be falsifiable.11 However, it would be straightforwardly falsified by evidence that human attention and processing resources are systematically allocated on some other basis: for instance, to inputs which are expected to be informative without being relevant, to yield many associations but few inferential effects, to be cheap to process regardless of any expected effects, or to have many effects regardless of the processing costs incurred. The cognitive principle also makes a number of testable predictions about human perceptual, memory retrieval, and inferential mechanisms. Here, I will consider just one. It follows from the cognitive principle that human inferential mechanisms tend spontaneously to derive potentially relevant conclusions (i.e. conclusions expected to lead on to further cognitive effects), ignoring others which are logically valid but have few expected effects. This prediction was experimentally tested by Van der Henst, Sperber, and Politzer (2002). Participants were given pairs of premises such as (1a)–(1b) (a ‘determinate relational problem’) or (2a)–(2b) (an ‘indeterminate relational problem’) and simply asked ‘What follows?’ Determinate relational problem
(1) a. A is taller than B b. B is taller than C Indeterminate relational problem
(2) a. A is taller than B b. C is taller than B In principle, any set of premises yields an infinite number of logically valid conclusions, and if the goal of human inferential mechanisms were merely to derive valid conclusions, participants should be able to go on listing conclusions indefinitely. In practice, participants typically provide a single conclusion or none at all, because they interpret the question as ‘What of relevance follows?’ Thus, when presented with indeterminate relational problems such as (2), 43 per cent of participants said ‘Nothing follows’ (whereas only 8 per cent said ‘Nothing follows’ when presented with determinate relational problems such as (1), which yield a highly salient and potentially relevant conclusion). These results confirm the prediction that human
11
For recent discussion, see the papers in Burton-Roberts (2007) and Soria and Romero (2010).
85
Relevance Theory 85 inferential mechanisms do not simply derive logically valid conclusions regardless of their potential relevance. Moreover, participants who did draw conclusions from the indeterminate relational problems tended to produce ‘single-subject’ conclusions (e.g. ‘B is shorter than A and C’) rather than ‘double-subject conclusions’ (e.g. ‘A and C are taller than B’), even when, as in (2a) and (2b), this involved the extra effort of substituting one lexical item for another and altering the syntax of the premises. Van der Henst, Sperber, and Politzer (2002) argue that the extra effort is justified by the fact that single-subject conclusions have greater expected relevance than double-subject conclusions, since they are more likely to combine with available contextual information to lead on to further conclusions. This surprising result is predictable on the assumption that human inferential mechanisms are relevance-oriented, and hard to explain in other terms.
4.3 Relevance and Communication The claim that human cognition is relevance-oriented has immediate implications for pragmatics. For communication to succeed, the speaker needs the addressee’s attention. Since attention tends to go automatically to what is most relevant at the time, a prerequisite to successful communication is that the addressee must take the utterance to be relevant enough to be worth attending to. Then a speaker, by the very act of addressing someone, communicates that the utterance meets this precondition, and this is what the communicative principle of relevance states (Sperber and Wilson 1986/ 1995: 266–278): Communicative principle of relevance Every utterance communicates a presumption of its own optimal relevance. Notice that the presumption mentioned in the communicative principle is one of optimal, not maximal, relevance. Although addressees might want speakers to aim at maximal relevance, and helpful speakers may indeed try to give them what they want, what addressees are entitled to expect within this framework is something rather less. An utterance is optimally relevant under two conditions: Optimal relevance a. It is at least relevant enough to be worth the addressee’s processing effort. b. It is the most relevant one compatible with the speaker’s abilities and preferences. According to clause (a), the addressee is entitled to presume that the utterance is at least relevant enough to be worth his processing effort; otherwise, he will not attend
86
86 Deirdre Wilson to it at all.12 According to clause (b), he is also entitled to presume that the speaker will have gone beyond this minimal level of relevance—to the extent that she is both willing and able to—by reducing the processing effort required and increasing the cognitive effects achieved, thus increasing her chances of holding his attention and getting her point across (Wilson and Sperber 2002). Given this background, there is a practical heuristic that addressees can use in identifying the speaker’s meaning (i.e. the array of cognitive effects that she overtly intended to achieve): Relevance-guided comprehension heuristic (Wilson and Sperber 2002) a. Follow a path of least effort in constructing an interpretation of the utterance (and in particular in resolving ambiguities and referential indeterminacies, adjusting lexical meaning, supplying contextual assumptions, deriving implicatures, etc.). b. Stop when your expectations of relevance are satisfied. The goal is to find an overall interpretation that confirms the presumption of optimal relevance. For this, the addressee must enrich the decoded sentence meaning at the explicit level, and complement it at the implicit level, so as to yield enough cognitive effects to satisfy his expectations of relevance. The relevance-guided comprehension heuristic is an automatic procedure for achieving this goal; it is seen in relevance theory as belonging to a specialized comprehension module, a component of a broader mindreading module dedicated to attributing mental states in order to explain and predict behaviour.13 What makes it reasonable for the addressee to follow a path of least effort is that the speaker is expected (within the limits of her abilities and preferences) to have made the utterance as easy as possible for him to understand. Since relevance varies inversely with effort, the very fact that an interpretive hypothesis is easily accessible gives it an initial degree of plausibility (an epistemic advantage specific to communicated information). What makes it reasonable for the addressee to stop at the first interpretation which satisfies his expectations of relevance is that a speaker who knowingly produced an utterance with two or more significantly different interpretations, each yielding the expected level of cognitive effect, would put him to the gratuitous extra effort of choosing among them, and the resulting interpretation (if any) would not satisfy clause (b) of the presumption of optimal relevance. Thus, when a hearer following the path of least effort finds an interpretation that is relevant in the expected way, in the absence of contrary evidence, this is the best possible interpretive hypothesis. Since comprehension is a non-demonstrative inference process, this 12 How relevant the utterance has to be to meet this condition depends on what else is going on in the addressee’s cognitive environment at the time: the more relevant the other inputs competing for his attention, the more relevant the utterance has to be to be worth his processing effort. 13 For discussion, see Sperber and Wilson (2002, 2005); Wilson and Sperber (2002, 2004).
87
Relevance Theory 87 hypothesis may well be false. This can happen when the speaker formulates the utterance in a way that is inconsistent with the expectations raised, so that the normal inferential routines of comprehension fail. Failures in communication are common enough. What is remarkable and calls for explanation is that communication works at all. A common objection to the communicative principle is that it is too vague and general to be falsifiable. However, this is often based on a misconception. It has been suggested, for instance, that a hearer looking for the most relevant interpretation of an utterance can never be sure of having found it, since by spending a little more effort, it may be possible to achieve substantially greater effects, and hence more relevance (Bach 2010a: 136; Davis 2014: section 11). But, as noted above, the hearer’s goal is not to find the most relevant interpretation: it is to construct an overall interpretation on which the utterance satisfies the presumption of optimal relevance.14 Moreover, as shown above, the relevance-g uided heuristic has a clear stopping point.15 The communicative principle would be straightforwardly falsified by evidence that communicators systematically orient to some other property of utterances than optimal relevance. This could happen, for instance, if speakers systematically aim at literal truthfulness rather than optimal relevance, or produce utterances which are informative without being relevant, or prefer to save their own effort even if the result is not relevant enough to be worth processing. Here, relevance theory comes into direct conflict with Grice’s framework. For Grice, the first Quality maxim was the most important of all the maxims (Grice 1989a: 27, 371), and from this it should follow that considerations of literal truthfulness systematically outweigh those of informativeness, relevance, or perspicuity. The theoretical consequences of this difference between the two frameworks were discussed in Wilson and Sperber (2002) and experimentally tested by Van der Henst, Carles, and Sperber (2002). In an initial experiment, Van der Henst, Carles, and Sperber (2002) simply approached strangers in the street and asked ‘Do you have the time, please?’, giving no indication of why the question was being asked. The prediction was that, if the speaker’s watch showed (say) 3.13, in the absence of any indication that some crucial implications would be lost, an answer rounded to the nearest five minutes would be
14 Or, in more complex cases, on which the speaker might have thought it would satisfy, or at least seem to satisfy, the presumption of optimal relevance (Sperber 1994b; Wilson 2000). 15 In a confused and confusing exposition of relevance theory, Wayne Davis (2014: section 11) attributes to Sperber and Wilson the view that communication is governed by a ‘Principle of Maximal Relevance’, suggests that they hesitate between notions of maximal and optimal relevance (rather than simply applying them in different domains), claims that ‘neither maximal nor optimal relevance requires the speaker to minimise processing cost’, since ‘Additional effort can always be justified by an increase in informativeness’, maintains that ‘optimal relevance does not pick out a unique contribution to the conversation’, and suggests that relevance theory excludes the possibility of scalar implicatures and reminders. All these points are explicitly discussed in Sperber and Wilson (1986/1995) and many later works.
88
88 Deirdre Wilson easier for the hearer to process, and hence more likely to be produced by a speaker aiming at optimal relevance. And indeed, 97 per cent of participants with analogue watches gave a rounded answer, while 57 per cent of those with digital watches went to the extra effort of producing a rounded answer rather than simply reading off a strictly accurate (and literally truthful) answer, thus confirming the prediction that speakers systematically aim at optimal relevance rather than literal truthfulness. In a second experiment, the experimenters asked ‘Do you have the time, please? My watch has stopped’, thus explicitly indicating that a strictly accurate answer would be more relevant than a rounded one. Here, the percentage of rounders fell from 97 per cent to 49 per cent (only results for participants with analogue watches were reported in this experiment), suggesting that speakers tend to provide strictly accurate (i.e. literally truthful) answers when they expect them to be relevant. In a third experiment, the experimenters asked ‘Do you have the time, please? I have an appointment at 4.00’, at different intervals in the half hour leading up to the imaginary appointment. The results showed that speakers tended to give more strictly accurate answers as the time of the imaginary appointment approached (when some crucial implications might be lost by rounding). These results are straightforwardly predictable on the assumptions of relevance theory, and hard to explain in frameworks where a maxim of literal truthfulness is seen as the most important of all the maxims (for further tests of the cognitive and communicative principles, see Van der Henst and Sperber 2004).
4.4 The Explicit–Implicit Distinction The term ‘explicature’ was introduced in early work in relevance theory to avoid a potential problem with Grice’s notion of ‘what is said’ (Wilson and Sperber 1981; Sperber and Wilson 1986/1995: 182). For Grice, ‘what is said’ does double duty: it refers to (a) the result of combining sentence meaning with disambiguation and reference resolution (often described as the proposition literally expressed and treated as the output of semantics), and (b) part of speaker’s meaning (e.g. what is asserted rather than implicated). The problem is that (a) and (b) do not necessarily coincide. Consider Mary’s utterance in (3b): (3) a. Peter: Let’s ask Billy to see a film with us tonight. b. Mary: He has to finish a paper. Suppose that the result of combining the meaning of the sentence uttered with disambiguation and reference resolution is the proposition in (4) (where paper1 denotes academic papers, parliamentary papers, conference papers, etc., but not newspaper enterprises): (4) Billy i has to finish doing something with a Paper1 at some point in the future.
89
Relevance Theory 89 This is what Mary has ‘said’ in sense (a) above. However, if Peter understood Mary as asserting (4), her utterance would not satisfy his expectations of relevance, since it does not imply a clear response to his proposal in (3a). He is therefore likely to interpret her as asserting something richer and more pragmatically satisfactory, such as (5): (5) Billy i has to finish writing a paper1 soon. By combining (5) with the contextual information in (6a), he could derive the implicature in (6b), thus arriving at an overall interpretation that satisfies his expectations of relevance: (6) a. A good reason for not asking someone out is that they have to finish writing a paper1 soon. b. They shouldn’t ask Billyi out because hei has to finish writing a paper1 soon. Here, the two notions of ‘what is said’ come apart. Moreover, it seems clear that disambiguation and reference resolution are also geared to finding a pragmatically satisfactory overall interpretation, and therefore fall within the scope of pragmatics rather than semantics. In (3b), for instance, a hearer using the relevance-guided comprehension heuristic will interpret ‘he’ as referring to Billy i, and disambiguate ‘paper’ as paper1, because these are the most salient hypotheses which lead on to an overall interpretation that satisfies his expectations of relevance. What is left of the linguistic meaning of (3b) once disambiguation and reference resolution are removed is quite fragmentary and incomplete, and falls far short of determining a unique proposition literally expressed; Sperber and Wilson (1986/ 1995: 72–75) call this fragmentary sentence meaning a logical form. As noted above, the hearer’s goal in developing the logical form of (3b) into a fully propositional form is to find an overall interpretation that satisfies his expectations of relevance. For this, he uses the relevance-guided comprehension heuristic, following a path of least effort in using contextual information to disambiguate the ambiguous word ‘paper’, assign reference to the pronoun ‘he’, and enrich his interpretation of what Billy is doing to the paper and when he will do it, in such a way that the result will yield enough contextual implications (and other cognitive effects) to make the utterance relevant as expected. On this approach, interpreting an utterance is like solving a complex simultaneous equation, and the interpretation process is crucially seen as carried out in parallel rather than in sequence. It is not a matter of first identifying the explicit content, then supplying contextual assumptions and then deriving contextual implications (and other cognitive effects), but of mutually adjusting tentative hypotheses about explicit content, context, and cognitive effects, with each other and with the presumption of relevance, and stopping at the first overall interpretation that makes the utterance relevant in the expected way. In interpreting (3b), for instance, Peter will expect it to achieve relevance by implying a response to his proposal in (3a), and this places a strong constraint on the route he will take in developing the encoded logical form into
90
90 Deirdre Wilson a fully propositional form: it must be such that it combines with easily available contextual assumptions to imply a response of the expected type (on mutual adjustment, see Sperber and Wilson 1998; Wilson and Sperber 2002). To avoid proliferating notions of ‘what is said’, Sperber and Wilson (1986/ 1995: 182) introduced the term ‘explicature’, on the analogy of Grice’s ‘implicature’, to refer to what is explicitly communicated. An explicature has two defining features: (a) it is a communicated proposition (i.e. part of the speaker’s meaning), and (b) it is identifiable by a combination of decoding and inference (i.e. by inferentially developing an encoded logical form into a fully propositional form). Everything else communicated is an implicature. On this approach, the explicit–implicit distinction is exhaustive—a communicated proposition must be either an explicature or an implicature—but explicatures vary in the relative proportions of decoding and inference involved. Compare Mary’s utterance in (3b) with the alternative formulations in (7a)–(7c): (3) b. He has to finish a paper. (7) a. Billy has to finish a paper. b. Billy Smith has to finish writing a paper. c. Billy Smith has to finish writing an academic paper soon. Although the explicature is the same in all cases, each of (7a–c) involves more decoding and less inference than its predecessor, and intuitively makes the speaker’s meaning more explicit. Explicitness is therefore definable not only as a yes–no matter but as a matter of degree: the greater the relative contribution of decoding, and the smaller the relative contribution of pragmatic inference, the more explicit the speaker’s meaning will be (Sperber and Wilson 1986/1995: 182). ‘Explicature’ was intended as a theory-neutral term to be used in any framework, and it has been widely (though by no means universally) adopted.16 Some writers (notably Bach 1994, 2010a) object to it on terminological grounds, and since the objection raises an issue of substance, it is worth a brief mention here. Bach sees use of the term ‘explicature’ as misleading because what is communicated is not made ‘fully explicit’, and proposes the term ‘impliciture’ instead: What [relevance theorists] regard as explicit is, in general, not fully explicit but partly implicit. Indeed, this is suggested by their term ‘explicature’, which is a cognate of ‘explicate’, not ‘explicit’. To explicate something is to spell it out, and to spell out the explicature of an utterance would be to make fully explicit
16 Recanati (2004a, 2010) refers to ‘what is said’ rather than ‘explicature’; Levinson (2000) talks of implicatures ‘intruding’ into truth-conditional content, and Bach replaces ‘explicature’ with ‘impliciture’. For thorough discussion, see Carston (2002, this volume); Carston and Hall (2012); Jodłowiec (2015).
91
Relevance Theory 91 what has in fact been left partly implicit. That is why I call this partly implicit content an ‘impliciture’ (the term should not suggest that all of an impliciture is implicit). (Bach 2010a: 131)
To which a relevance theorist might reasonably respond: if ‘impliciture’ is an appropriate name for a communicated proposition that is partly explicit and partly implicit, why isn’t ‘explicature’ equally appropriate?17 How is the substitution of ‘impliciture’ for ‘explicature’ a step forward? But there is a more substantive issue behind this apparent terminological dispute. For Bach, the only way to express a thought explicitly is to encode it, and the function of linguistic meaning is precisely to enable the encoding of thoughts: My main reason for thinking that at least some sentences express propositions is very simple. If none did, then none of our thoughts would be explicitly expressible. Indeed, it is arguable that all of our thoughts are explicitly expressible, in which case for every thought there is at least one sentence that would express it explicitly. (Bach 2010a: 129)
For Bach, ‘explicit’ is an absolute term, like ‘empty’, and anything less than ‘fully explicit’ is not explicit at all. For relevance theorists, the function of linguistic meaning is not to encode the speaker’s meaning but to provide evidence of it, and the idea that any thought (let alone all of them) can be fully encoded has been rejected from the outset (see e.g. Sperber and Wilson 1986/1995: 191–193; Carston 2002: section 1.3). As Wilson and Sperber (2012b: ix) put it, There are always components of a speaker’s meaning which her words do not encode: for instance, the English word ‘he’ does not specifically refer to [e.g. Billy in (3b)]. Indeed, we would argue that the idea that for most, if not all, possible meanings that a speaker might intend to convey, there is a sentence in a natural language which has that exact meaning as its linguistic meaning is quite implausible.
For relevance theorists, ‘explicit’ is both a classificatory and a comparative concept: any communicated proposition with a linguistically encoded conceptual constituent is explicit to some degree, and the greater the proportion of decoding to inference, the more explicit it will be. On this approach, any utterance can be made more explicit, and there is no such thing as ‘full explicitness’ (what would be the ‘fully explicit’ version of Mary’s utterance in (3b)?). Thus, debates about the appropriateness of the term ‘explicature’ have their roots in a deeper disagreement about the role of linguistic meaning in communication. However, since it is generally agreed that ‘explicatures’ 17 In fact, relevance theory’s explicit–i mplicit distinction applies only to communicated propositions, and not to their constituents. One might describe the constituents of propositions as being ‘tacitly’ or ‘overtly’ (rather than ‘explicitly’ or ‘implicitly’) expressed.
92
92 Deirdre Wilson and ‘implicitures’ involve both decoding and inference, it is not legitimate to object to ‘explicature’ on the ground that it is partly explicit and partly implicit and defend ‘impliciture’, which is partly implicit and partly explicit too.
4.5 Lexical Pragmatics and the Literal–Figurative Distinction Lexical pragmatics explores the application of the semantics–pragmatics distinction at the level of the word or phrase rather than the whole utterance. A central goal is to investigate the processes by which linguistically encoded word meanings are adjusted (or ‘modulated’) in use.18 Well-studied examples of such processes include lexical narrowing (e.g. drink used to mean ‘drink alcohol’, or ‘drink substantial amounts of alcohol’), approximation (e.g. square used to mean ‘squarish’), and metaphorical extension (e.g. nightmare used to mean ‘bad experience’). A striking feature of much existing research in this area is that narrowing, approximation, and metaphorical extension tend to be seen as distinct processes which lack a common explanation. Relevance theorists have been trying to develop a more unitary account based on two main claims. First, there is no presumption of literalness: linguistically specified word meanings are typically adjusted in the course of pragmatic interpretation, using available contextual information. Second, there is a continuum of cases of broadening, from approximation through to ‘figurative’ uses such as hyperbole and metaphor, which all involve the same interpretive mechanisms and can be explained in the same way. Here I will briefly compare relevance theory’s approach to lexical narrowing and broadening with two alternative accounts. Lexical narrowing involves the use of a word or phrase to convey a more specific concept (with a narrower denotation) than the linguistically encoded meaning. For instance, red is typically narrowed in different directions in common adjective–noun combinations (e.g. red eyes, red apple, red hair, red stamp, etc.), picking out a different shade, distributed in different ways across the surface of the object, in different combinations. One approach which fits well with the stereotypical nature of much lexical narrowing is to treat it as a variety of default inference.19 For instance, Levinson (2000: 37–38, 112–134) analyses narrowing as involving a default inference governed by
18 See e.g. Recanati (1995, 2004a, 2010); Carston (1997, 2002); Blutner (1998, 2004); Lascarides and Copestake (1998); Sperber and Wilson (1998, 2008); Glucksberg (2001, 2003); Fauconnier and Turner (2002); Wilson and Sperber (2002); Horn (2004, 2012c); Recanati (2004a, 2010); Wilson and Carston (2006, 2007, 2008); Huang (2009). 19 The notion of a default inference has been developed in many different ways; see e.g. Levinson (2000: section 1.5); Geurts (2009); Jaszczolt (2014).
93
Relevance Theory 93 an Informativeness heuristic (‘What is expressed simply is stereotypically exemplified’), itself backed by a more general I-principle instructing the hearer to: Amplify the informational content of the speaker’s utterance, by finding the most specific interpretation, up to what you judge to be the speaker’s m-intended point ( … ). (ibid.: 114)
On this approach, hearers are seen as automatically constructing a stereotypical (or otherwise enriched) interpretation and accepting it in the absence of contextual counter-indications.20 The alternative view, developed in relevance theory, is that lexical narrowing is a far more creative and flexible process, involving the construction of ad hoc, occasion-specific concepts influenced by a much wider range of cognitive and contextual factors than default approaches take into account. Thus, in order to satisfy expectations of relevance, the interpretation of red eyes might be narrowed to different degrees, and in different directions, in different contexts, yielding a range of occasion- specific (‘ad hoc’) concepts, e.g. [red eyes]*, [red eyes]**, and so on.21 How might one choose between these two accounts? For Levinson (2000), default narrowings are generalized conversational implicatures, to be dealt with in a theory of utterance-t ype meaning designed to explain how sentences are systematically paired with preferred interpretations regardless of the contexts in which they are used. Levinson contrasts this with a theory of utterance- token meaning, or speaker’s meaning, such as relevance theory, which is designed to take context and speaker’s intentions into account. It should follow that on Levinson’s approach, information about the wider discourse context cannot affect the outcome of lexical narrowing, and the same default interpretation (specifying a certain shade and degree of redness, distributed over certain parts of the eye) must be automatically assigned to every occurrence of red eyes, and accepted in the absence of contextual counter-indications. As Noveck and Sperber (2007) point out, on the assumption that communicative systems tend to favour least-effort principles and to evolve in the direction of increasing efficiency, the value of a default-based approach will depend heavily on the distributional frequencies of interpretations on which the default interpretation proves acceptable and those in which it has to be overridden or cancelled for contextual reasons. To provide some evidence, Kolaiti and Wilson (2014) took the phrase red eyes—which does 20 Notice, though, that the I-Principle does not explain how the hearer identifies the speaker’s intended meaning, but presupposes that he has some independent means of judging what this is. To put it slightly unkindly, the I-Principle says ‘Choose a more specific interpretation if you think this is what the speaker intended.’ But the goal of a pragmatic theory is to explain how hearers decide that a certain meaning was intended, and given that lexical broadening is just as common as lexical narrowing, the I-Principle does not get us any closer to this goal. 21 I will follow the usual practice of representing linguistically encoded meanings (‘lexical concepts’) in small capitals (red) and occasion-specific meanings (‘ad hoc concepts’) in small capitals followed by one or more asterisks (red*, red** … ).
94
94 Deirdre Wilson not seem obviously to favour either a default or a relevance-theoretic approach—and examined its occurrences in the Bank of English (a 56 million word corpus). They found that Levinson’s default-based approach would guide the hearer in the right direction— and therefore help with processing costs—in roughly 50 per cent of cases (i.e. those involving e.g. crying, fatigue, flu/cold, eye damage, eczema, heat/sand, and sore eyes in humans), but would be positively misleading and incur the costs of cancellation in the remaining 50 per cent of cases (i.e. those involving e.g. flash photography, animals, insects, supernatural beings, fictional entities, and inconclusive cases with no obvious justification for narrowing at all). A more flexible inferential approach such as relevance theory would involve context-sensitive—and therefore relatively costly—fine-tuning of the encoded lexical meaning in the full range of cases, but without the costs of default derivation followed by cancellation and reinterpretation in 50 per cent of the cases. As Kolaiti and Wilson point out, it is far from obvious that the statistical tendencies revealed by their corpus justify a default rather than an inferential account of lexical narrowing on grounds of economy of processing, yet this was the main rationale for the default approach proposed in Levinson (2000: section 1.3). A further claimed advantage of default-based approaches to narrowing is that they explain the ready accessibility of ‘normal’, or ‘stereotypical’, narrowings in the absence of special contextual factors. However, there are other ways of explaining this ready accessibility without appeal to defaults, as in Horn’s approach based on his R principle (‘Say no more than you must’) or relevance theory’s approach, which predicts that ‘normal’ or ‘stereotypical’ interpretations will be less costly to construct in most circumstances, and will therefore be selected by the relevance-guided comprehension heuristic as long as they yield enough implications to satisfy the audience’s expectations of relevance. Moreover, neo-Griceans such as Levinson, Horn, and Blutner, who have been primarily concerned with lexical narrowings of a fairly stereotypical sort, have said little or nothing about how they would treat loose, hyperbolic or metaphorical uses of language, which are heavily context-dependent and are standardly treated as violating Grice’s first Quality maxim. As noted above, relevance theorists have consistently argued against this maxim and defended the view that there is a continuum between literal, loose, and metaphorical uses rather than a set of clearly definable theoretical categories which play distinct roles in communication and comprehension. How might one assess these alternative approaches? Lexical broadening involves the use of a word or phrase to convey a more general concept (with a broader denotation) than the linguistically encoded meaning. As noted earlier, a striking feature of much research in this area is that different interpretive procedures have been proposed for a range of phenomena which could all be seen as varieties of broadening. Thus, approximation is often treated as a case of pragmatic vagueness involving different contextually determined standards of precision (Lewis 1979; Lasersohn 1999). Metaphor and hyperbole are still widely seen as involving blatant violation of Grice’s first Quality maxim, with the use of metaphor implicating a related simile or comparison and the use of hyperbole implicating a related weaker proposition (Grice 1967/1989a). Typically, these accounts do not generalize: metaphors
95
Relevance Theory 95 are not analysable as rough approximations, approximations are not analysable as blatant violations of a maxim of truthfulness, and so on. Relevance theorists have been exploring the hypothesis that there is no clear cut-off point between literal use, approximation, hyperbole, and metaphor, but merely a continuum of cases of broadening, which are all understood in the same way, using the same relevance-guided comprehension heuristic.22 On this approach, approximation, metaphor, and hyperbole are not natural kinds, which are dealt with by different mechanisms, and there is no fact of the matter about what is ‘really’ a metaphor or hyperbole and what is not. In what follows, I will use these terms as handy descriptive labels rather than theoretical concepts. It is worth highlighting two important differences between the Gricean and relevance-theoretic approaches to broadening. First, Grice retains a sharp distinction between literal and figurative uses, and like many philosophers of language, he appears to see loose talk and rough approximations as falling on the literal rather than the figurative side (to be analysed as involving contextually determined standards of precision rather than blatant violation of a maxim of truthfulness). Second, he sees figurative uses such as metaphor and hyperbole as not contributing to truth- conditional content or ‘what is said’, but merely to what is implicated. By contrast, relevance theorists deny that there is a clear theoretical distinction between literal and figurative uses, and treat the ad hoc concepts derived via lexical-pragmatic processes as contributing to truth-conditional content (explicatures) across the whole ‘literal– figurative’ continuum. In the light of this, consider the use of painless in (8): (8) Dentist to patient: The injection will be painless. On a relevance-theoretic approach, painless in (8) might be taken to convey either its literal meaning, painless (‘with no pain’) or an approximation, painless* (‘with almost no pain’). But the presence of the small amount of pain that would justify classifying painless in (8) as an approximation shades off imperceptibly into the amount of pain that would justify classifying it as a hyperbole, painless**, (‘with less pain than expected or feared’). The Gricean framework predicts that this imperceptible shading off gives rise to a dramatic difference in processing on either side of the approximation/hyperbole divide: on the one side, the speaker is making a genuine assertion, albeit under reduced standards of precision, whereas on the other side, she is merely implicating that the injection won’t hurt too much. As far as I know, there is no experimental evidence whatsoever of such a dramatic processing difference between different degrees of broadening. In the relevance-theoretic framework, by contrast, where both approximation and hyperbole contribute to truth-conditional content or 22 See e.g. Carston (2002, 2010b), Wilson and Sperber (2002), Wilson and Carston (2006, 2007, 2008), Vega Moreno (2007), Sperber and Wilson (2008), Carston and Wearing (2011), and Wilson (2011a).
96
96 Deirdre Wilson explicatures, this imperceptible shading off between approximation and hyperbole is both predicted and explained. All this suggests that the goal of an adequate pragmatic theory should be to provide a unitary account of the full range of lexical-pragmatic processes. However, largely as a result of historical accident (perhaps combined with differences in intellectual taste), the only explicit attempts so far at developing such an account have been made by relevance theorists. Neo-Griceans working on lexical narrowing have shown little interest in extending their account to cover metaphor or hyperbole; philosophers and literary scholars working on metaphor and hyperbole have shown no interest in extending their account to approximation or narrowing, and so on, and semanticists and logicians working on approximation have shown little interest in metaphor or hyperbole. This is not, of course, to claim that relevance theory offers the only possible unitary account: the challenge is to propose a better one.
4.6 The Conceptual–Procedural Distinction The conceptual–procedural distinction was introduced into relevance theory by Diane Blakemore (1987, 2002) to account for differences between regular ‘content’ words such as dog, or red, which are standardly seen as encoding concepts that contribute to the truth-conditional (assertive) content of utterances, and discourse connectives such as so, or after all, which are standardly seen as non-truth-conditional. Blakemore suggested an original rationale for non-truth-conditional meaning, arguing that the function of discourse connectives is to guide the inferential comprehension process by imposing procedural constraints on the construction of intended contexts and cognitive effects. On this approach, so in (9a) does not affect the assertive content of the utterance, but is used to indicate that what follows it is a contextual implication of the fact that it is raining, while after all in (9b) is used to indicate that what follows it is intended to strengthen the preceding claim that the grass is wet: (9) a. It’s raining, so the grass is wet. b. The grass is wet. After all, it’s raining. This approach has been insightfully applied to a wide range of discourse connectives in many languages.23 In an initial phase of research, the conceptual–procedural distinction was seen as coinciding with the distinction between truth-conditional and non-truth-conditional 23
For recent discussions, see e.g. Iten (2005); Hall (2007); Unger (2007); Escandell-Vidal et al. (2011).
97
Relevance Theory 97 meaning. However, it soon became clear that this parallel breaks down in several ways. In the first place, illocutionary adverbials such as frankly in (10a), which are standardly seen as non-truth-conditional, have synonymous manner-adverbial counterparts which contribute to truth-conditional content in regular ways, as in (10b) (Bach and Harnish 1979): (10) a. Frankly, Bill should resign. b. John spoke frankly to Anne. The simplest solution is to treat both uses of frankly as encoding the same concept, which contributes to truth-conditional content in (10b), but not in (10a). Wilson and Sperber (1993) analyse illocutionary adverbials as contributing to so-called ‘higher- order explicatures’, which carry information about the speaker’s propositional or affective attitude, or the type of speech act she intends to perform, rather than contributing directly to truth-conditional content; thus, frankly in (10a) would be taken to indicate that the speaker is telling Bill frankly that he should resign. On this approach, illocutionary adverbials are both conceptual and non-truth-conditional. In the second place, the parallel between procedural and non-truth-conditional meaning breaks down for a range of referential expressions such as I, she, now, and then. These clearly contribute to truth-conditional content rather than implicatures, but are not plausibly seen as encoding full-fledged concepts, since their referents vary from context to context and have to be pragmatically inferred. Wilson and Sperber (1993) analyse them as encoding procedural constraints on reference resolution, so that she, for instance, restricts the set of potential referents to those appropriately picked out by use of a feminine pronoun. On this account, she is both procedural and truth-conditional.24 Finally, a variety of non-truth-conditional items such as mood indicators, sentence and discourse particles, interjections, and intonation have been analysed as encoding a still further type of procedural constraint, this time on the construction of higher- order explicatures.25 For instance, the addition of an interrogative particle, question intonation, or interrogative word order to the utterance in (11) might trigger construction of the higher-order explicature in (12a), and use of the interjection alas or certain types of affective intonation in (11) might trigger the higher-order explicature in (12b): (11) Bill was at the party. (12) a. The speaker is asking whether Bill was at the party. b. The speaker is expressing regret that Bill was at the party.
24 For procedural approaches along these lines, see e.g. Hedley (2007); Powell (2010); EscandellVidal et al. (2011); Scott (2013a,b); Sasamoto and Wilson (2016). 25 See e.g. Wilson and Sperber (1993); Fretheim (1998); Wilson (2000, 2011b); Wharton (2003, 2009); Wilson and Wharton (2006); Escandell-Vidal et al. (2011); Sasamoto and Wilson (2016).
98
98 Deirdre Wilson Along these lines, the conceptual–procedural distinction might contribute in interesting ways to current debates on the distinction between descriptive and expressive meaning.26 The conceptual–procedural distinction as proposed in relevance theory raises several questions. For instance, given that procedural meaning need not contribute to truth conditions, in what sense is it properly semantic, and what explains the disparate nature of the procedural expressions described above?27 I will respond briefly to both questions here. According to Sperber and Wilson (1986/1995: 172–173), a linguistic expression is semantically interpreted by being put into systematic correspondence with other objects: e.g., with the formulas of another language, with possible states of the world, or with states of the user of the language. Regular ‘content’ words in natural language are widely seen as semantically interpreted in the first of these ways, by being put into systematic correspondence with constituents of a conceptual representation system or ‘language of thought’; and conceptual representations are standardly seen as semantically interpreted in the second way, by being put into systematic correspondence with states of the world. Wilson (2011b) suggests that procedural expressions might be seen as semantically interpreted in the third way, by being put into systematic correspondence with states of the user of the language. The argument goes as follows. According to recent increasingly modular approaches to human cognition (Sperber 2005; Carruthers 2006), the human cognitive system comprises a large array of domain-specific procedures with distinct developmental trajectories and breakdown patterns, which may be more or less highly activated in different circumstances, and are likely to alter their level of activation in response to different cues. Among the possible states of the user of a language will be those in which a certain cognitive mechanism or procedure is highly activated. Wilson (2011b) suggests that the function of the procedural expressions in a language may be to put the user of the language into a state in which some of these domain-specific cognitive procedures are highly activated (and hence more likely to be selected by a hearer using the relevance-guided comprehension heuristic). One consequence of this proposal is that we might expect to find clusters of procedural items linked to different domain-specific capacities, with different developmental trajectories and breakdown patterns, and this seems to be just what we find. For instance, most languages have a cluster of procedural items (e.g. affective intonation, interjections, attitudinal particles) associated with mechanisms for emotion reading. The capacity to read emotions from facial and vocal cues is known to be present very early, and its outputs are particularly hard to analyse in conceptual terms (Wharton 2003, 2009); expressions of this type are therefore particularly suitable for procedural treatment. Most languages also have a cluster of procedural items (e.g. mood indicators, intonation, various types of discourse particle) linked to mindreading mechanisms. 26 27
See e.g. Blakemore (2011, 2014); Sasamoto and Wilson (2016); Carston (this volume). For discussion, see Bezuidenhout 2004; Escandell-Vidal et al. (2011).
99
Relevance Theory 99 A naive capacity for attributing mental states to others is also thought to be present very early, although its outputs may not be available to introspection or general inference until much later. Languages with grammaticalized honorific systems contain a further cluster of procedural expressions which might be seen as linked to the capacity for social cognition. Finally, most languages also have a cluster of procedural items (e.g. punctuation, prosody and various types of discourse particle) whose function is to guide the inferential comprehension process in one direction or another. As noted above, the standard relevance-theoretic account of procedural meaning treats procedural expressions as guiding the inferential comprehension process by constraining the construction of contexts and/or the derivation of cognitive effects. On the approach just outlined, there is more to be said about them than this. Notice that the capacities for mindreading, emotion reading, and social cognition are not intrinsically linked to ostensive communication: for instance, we attribute mental states to others whether or not they are communicating with us. So whereas some procedural expressions (e.g. pronouns) activate mechanisms which are properly pragmatic, others activate mechanisms with a wide range of disparate functions, and this idea might be worth exploring further.
4.7 Concluding Remarks A speaker producing an utterance has two distinct goals: to get the addressee to understand her meaning, and to persuade him to believe it. The addressee has two corresponding tasks: to understand the speaker’s meaning, and to decide whether to believe it. The first task involves the pragmatic ability to infer the speaker’s meaning from linguistic and contextual cues, by identifying the overtly intended cognitive effects. In (3b), for instance, Peter will understand Mary’s meaning if he recognizes that she intends him to believe they shouldn’t ask Billy out that evening because he has to finish writing a paper soon. However, having understood her, he may not believe her, since he may suspect she is mistaken or lying. The second task involves what Sperber et al. (2010) call a capacity for ‘epistemic vigilance’, which helps hearers avoid being accidentally or intentionally misinformed. There is a growing body of evidence suggesting that even at a very early age, children do not treat all communicated information as equally reliable, and that the capacity for epistemic vigilance develops alongside the capacity for inferential communication. Indeed, Sperber et al. (2010) suggest that comprehension, the search for relevance, and epistemic vigilance may be seen as interconnected aspects of a single overall process whose goal is to make the best of communicated information. One possible direction for future research would be to investigate points at which the capacities for mindreading, communication, and epistemic vigilance might interact. For instance, there are well-k nown parallels between irony comprehension and the ability to cope with lies and deliberate deception; both correlate with success in
100
100 Deirdre Wilson standard second-order false-belief tasks, and Grice’s account of irony sheds no light on why this should be so (Mascaro and Sperber 2009; Wilson 2009). Relevance theorists have long argued that irony requires a higher order of mindreading ability than ordinary literal or metaphorical utterances;28 and exploring possible interactions between the mindreading, communicative, and epistemic vigilance capacities in irony comprehension might well yield fruitful results. Moreover, in light of the arguments in section 4.6, given that the capacity for epis temic vigilance is distinct from those for mindreading and communication, we might expect to find clusters of procedural expressions specifically linked to epistemic vigilance mechanisms. According to Sperber et al. (2010) and Wilson (2011b), logical and discourse connectives, on the one hand, and grammaticalized indicators of epistemic modality and evidentiality, on the other, might well fall into this category, and this possibility would be worth exploring further.
Acknowledgements Many thanks to Robyn Carston, Dan Sperber, and Yan Huang for valuable comments on an earlier version.
28
On the current state of the art in irony studies, see Wilson and Sperber (2012a); Wilson (2013, 2014).
101
Chapter 5
For m a l Pr agm atics Reinhard Blutner
5.1 Introduction As prominently demonstrated in theoretical physics, the formal language of mathematics may be very useful for describing aspects of reality. That does not mean that the mathematical instruments are intended to capture a precise picture of reality. Instead, processes of abstraction and idealization are omnipresent (Stokhof and van Lambalgen 2011), generating an apparently very close fit between pre-existing, flawless mathematical structures and an idealized/abstracted picture of reality that is studied in science. In the field of natural language pragmatics, some researchers working on relevance theory (RT) or optimality-theoretic (OT) pragmatics take a similar naturalistic stance and claim that basic principles of cognitive psychology can be applied for grounding the basic mechanisms of natural language interpretation. The view of placing natural language pragmatics within the scope of a naturalistic (explanatory) approach is not without problems. This has to do with the normative character that is attributed to the Gricean setting. Speakers, as Grice (1975: 45) put it, must ‘make their contribution such as is required, at the stage at which it occurs, by the accepted purpose or direction of the talk exchange in which (they) are engaged’. If a person acts in a particular situation in a particular way we can ask why she did it the way she did; alternatively, we can ask if what the person did was reasonable, and if other options were possibly more reasonable in the given situation. Good Griceans are expected to ask the second type of questions (to take the normative stance), whereas the first question is expected to be asked by cognitive scientists (typically taking a naturalistic stance). Obviously, the Gricean principle of cooperation as stated above is normative, and so are Grice’s conversational maxims. Even though the normative and the naturalistic aspects of understanding human actions can be clearly separated from each other, in most cases it does not follow that they predict different action patterns. The idea of a rational world is not so irrational
102
102 Reinhard Blutner to be excluded from ordinary affairs. Evolutionary game theory has presented us with many examples demonstrating that the reasonable is naturally arising (Axelrod 1984). In other words, though there is a philosophical gap between Gricean pragmatics as a normative theory and pragmatic frameworks such as relevance theory as a scientific, explanatory theory of natural language, there is not a deep empirical conflict between an interpretation-oriented pragmatics and a speaker ethics. It seems that the speaker would do better to be cooperative or pretend to be cooperative if they want to use language to bring about effects on hearers. In the present article, I will discuss three recent approaches which are crucially based on formal mathematical instruments to perform their analyses, sometimes being strictly naturalistic, sometimes being based on the normative stance. The theoretic frameworks I will discuss are optimality-theoretic pragmatics, game-theoretic pragmatics, and decision-theoretic pragmatics. I will highlight both the similarities and the essential differences between these frameworks.
5.2 Optimality-Theoretic Pragmatics Optimality theory is an integrated approach to cognition that combines the advantages of symbolic, constraint-based models with the advantages of subsymbolic, neuron- style models of cognition (cf. Smolensky and Legendre 2006). In the study of natural language, OT was successfully applied to the main linguistic disciplines including phonology, morphology, and syntax, and also to the explanation of natural language acquisition and other performance traits. OT pragmatics is an application of the integrated approach to the domain of Gricean pragmatics. It has its origin in the attempt to explain certain phenomena of lexical pragmatics (Blutner 1998) and is inspired by the optimal interpretation approach proposed by Hendriks and de Hoop (2001). The view of seeing OT pragmatics within the scope of a naturalistic (explanatory) approach to cognition (as represented by the main proponents of OT) brings it close to relevance theory (RT) which likewise takes the naturalistic stance (Sperber and Wilson 1986/1995). There is another point of agreement that brings OT pragmatics and RT closely together. This point concerns the fact that both theories view the division of labour between semantics and pragmatics in a similar way. Both follow the tradition of radical pragmatics and accept these three claims (e.g. Jaszczolt 2010): 1. There is a level of logical form or semantic representation. The representations of this level do not necessarily provide truth conditions. Rather, they underspecify truth-conditional content in a number of ways. 2. There is a mechanism of enriching underspecified representations; sometimes this mechanism is called development of logical form. The result of this
103
Formal Pragmatics 103 development is propositional content. It expresses the utterance meaning of the expression under discussion. 3. There is a level of implicatures proper, understood as separate thoughts implied by the utterance. It is implicit propositional content that can be inferred from the explicit content mentioned in 2. Obviously, the consensus is about rejecting the Gricean doctrine of literal meaning (logical form conforms to literal meaning), accepting the role of underspecification (logical forms are underspecified with regard to the expressed semantic content), and acknowledging that implicature is a graded category (some implicatures are closer to logical form than others).
5.2.1 Three Variations on Grice In this subsection I will discuss three variants of Gricean pragmatics: (i) RT (Sperber and Wilson 1986/1995), (ii) Levinson’s (2000) theory of presumptive meaning, (iii) the Neo-Gricean approach (Atlas and Levinson 1981; Horn 1984; Huang 2009; see also Huang 2007). I will show how OT can formalize these three approaches and systematically relate them. In this connection it is useful to introduce the distinction between global and local approaches to conversational implicatures (cf. Chierchia 2004). According to the global (neo-Gricean) view one first computes the (plain) meaning of the sentences; then, taking into account the relevant alternatives, one strengthens that meaning by adding in the implicature This contrasts with the local view, which first introduces pragmatic assumptions locally and then projects them upwards in a strictly compositional way where certain filter conditions apply. Representatives of the global view are Gazdar (1979), Atlas and Levinson (1981), Soames (1982), (Horn 1984), Krifka (1995), Blutner (1998), Sauerland (2004), Sæbø (2004), and Geurts (2010); the local view is taken by RT (Sperber and Wilson 1986/1995, Carston 2002), Levinson (2000), and Chierchia (2004). Usually, the globalists argue against the local view and the localists against the global view. I will argue, instead, that proper variants of both views are justified if a different status is assigned to the two views: global theories provide the standards of rational discourse and correspond to a diachronic, evolutionary scenario; local theories account for the shape of actual, online processing, including the peculiarities of incremental interpretation. In this way, I will argue that seemingly conflicting approaches such as relevance theory and the neo-Gricean approach are much more closely related than the adherents of one side or the other might expect. OT will prove its power of unification in giving hints on how to relate these different frameworks in a systematic way. RT assumes the representational/computational view of the mind, and, on this basis, gives a naturalization of pragmatics adopting Jerry Fodor’s language of thought
104
104 Reinhard Blutner hypothesis (Fodor 1975). The central thesis of RT is the communicative principle of relevance, according to which utterances convey a presumption of their own optimal relevance. In other words, any given utterance can be presumed: • to be at least relevant enough to warrant the addressee’s processing effort; • to be the most relevant one compatible with the speaker’s current state of knowledge and her personal preferences and goals. From these two assumptions relevance theorists derive the following general procedure that the cognitive system follows in comprehending an utterance (cf. Sperber, Cara, and Girotto 1995: 95): (a) test possible interpretations in their order of accessibility, and (b) stop once the expectation of (optimal) relevance is satisfied (i.e. a certain context-dependent threshold value of relevance is reached). The procedure makes sure that the wanted effect (a certain value of relevance) is reached with the minimal cognitive effort. Levinson’s (2000) theory of presumptive meaning is a chameleon that in a certain sense adapts general assumptions of RT and in another sense crucially conflicts with RT, for instance in assuming more than one basic principle (maxim) for formulating the interpretational mechanism. In short, these are the general assumptions: • Differing from both RT and the standard neo-Gricean view, Levinson assumes three levels of meaning corresponding to sentence(-type) meaning, utterance- type meaning, and utterance-token meaning • utterance-type meanings are in correspondence with Grice’s generalized conversational implicatures. They are a matter of preferred interpretation calculated by a particular default mechanism. Basically, there are three such defaults or heuristics: – Q-heuristic: What isn’t said is not the case – I-heuristic: What is expressed simply is stereotypically exemplified – M-heuristic: What’s said in an abnormal way isn’t normal • In contrast to Grice’s generalized conversational implicatures, which are calculated in a global manner, presumptive meanings are local, i.e. they arise at the point at which they are triggered (for instance, the word some triggers the default interpretation NOT ALL via the Q-heuristic). The feature of local pragmatics is essential to artificial intelligence pragmatics (e.g. Hobbs and Martin 1987) and likewise to RT. Presumptive meanings are very useful for understanding natural language interpretation, especially for explaining the predominantly incremental character of utterance comprehension. Neo-Griceans (Atlas and Levinson 1981; Horn 1984, 2005a; Blutner 1998; Atlas 2005; Huang 2009) are assuming two countervailing optimization principles: the
105
Formal Pragmatics 105 Q-principle and the R-principle.1 The first is oriented to the interests of the hearer and looks for optimal interpretations; the second is oriented to the interests of the speaker and looks for expressive optimization. Here is a standard presentation of the two principles (cf. Horn 1984, 1989, 2004, 2005a): The Q-Principle (hearer-based) Make your contribution sufficient! Say as much as you can! (modulo R) (Grice’s first quantity maxim and the first two manner maxims) The R-Principle (speaker-based) Make your contribution necessary! Say not more than you must! (modulo Q) (Grice’s second quantity maxim, relation maxim and the second two manner maxims) It is tempting to identify the Q-principle with Levinson’s Q-heuristic and the R- principle with the I-heuristic. However, they are not identical though there is a correspondence between them. The difference has to do with the different status of principles in the global, neo-Gricean pragmatics on the one hand and heuristics (defaults) in Levinson’s local pragmatics on the other hand. According to the neo- Gricean picture the principles constitute a kind of communication game—either between real speakers and hearers or between fictive speakers and hearers in the mind of a language user. In this game both principles are applied in a recursive way (corresponding to the modulo clause in the formulation of the principles). In Levinson’s theory, no such interaction between real or fictive speakers/hearers takes place. Instead, presumptive meanings are default interpretations and they are processed in a nearly automatic way. No ‘mind-reading’ facilities or other mechanisms of controlled processing are required.2 The difference will become quite clear in the following subsection when I give formalization in terms of bidirectional OT.
5.2.2 Bidirectional OT Bidirectional optimality theory falls within the family of linguistic models that are based on the optimization of linguistic output against a system of ranked constraints
1
In OT, these ‘principles’ correspond to different directions of optimization where the content of the optimization procedure is expressed by particular OT constraints. This will be pointed out in more detail in the following section. 2 However, presumptive meanings can demand a lot of effort as soon ‘conflicts’ arise and the corresponding assumption has to be cancelled. Conflict resolution can be very resource-demanding. Hence, for the overall mechanism we have to take into account the peculiarities of controlled processing. Of course, this does not refer to any mind-reading facilities.
106
106 Reinhard Blutner (Blutner 2000; Blutner and Zeevat 2004; Blutner, de Hoop, and Hendriks 2005; Benz and Mattausch 2011). This theory provides a general procedure of optimization of the relation between form and meaning, simultaneously optimizing in both directions, from meaning to form, and from form to meaning. This distinguishes bidirectional optimality theory from unidirectional optimality- theoretic semantics (Hendriks and de Hoop 2001)—optimizing from form to meaning—and from unidirectional optimality-theoretic syntax (Grimshaw 1997)—optimizing from meaning to form. To put it in a nutshell, bidirectional optimality theory evaluates form–meaning pairs. As described in Blutner (2000), there are two ways of defining optimality in a bidirectional setting, a strong way and a weak way. The strong version is based on the standard definition of optimality, applying this to candidate pairs instead of output elements. In the following I will define an OT system for a set F of forms and a set M of meanings as a pair 〈Gen, ≻〉 consisting of a generator Gen ⊆ F × M that gives us the set of all potential form–meaning pairs and an ordering on elements of ≻ Gen. Informally, the relational statement f ' ≻m f says that the pair 〈 f ', m〉 satisfies the system of (ranked) constraints better than the pair 〈f, m〉; the statement m' ≻f m says that the pair 〈f, m'〉 satisfies the system of (ranked) constraints better than the pair 〈f, m〉 (borrowing the notation used by Franke 2009). In the strong version of bidirectional OT, a form– interpretation pair 〈f, m〉 ∈ Gen is considered to be (strongly) optimal iff • Interpretive Optimization: there is no pair 〈f, m'〉 ∈ Gen such that m' ≻f m • Expressive Optimization: there is no pair 〈f ', m〉 ∈ Gen such that f ' ≻m f. Informally, the first clause says that m is the optimal interpretation of f, and the second clause says that f is an optimal expression for m. The weak version of bidirectional optimality is less restrictive than the strong one and normally allows for more solution pairs. The original formulation (Blutner 2000) is close to the (recursive) formulation of Horn’s Q-and R-principle and it allows us to derive Horn’s division of pragmatic labour (Horn 1984, 1989, 2004, 2005a)—i.e., the general propensity that ‘unmarked forms tend to be used for unmarked situations and marked forms for marked situations’ (Horn 1984: 26). In the following I adopt Jäger’s reformulation of the original definition (Jäger 2002). A form–interpretation pair 〈f, m〉 ∈ Gen is considered to be superoptimal (or weakly optimal) iff • Interpretive Optimization: there is no superoptimal pair 〈f, m'〉 ∈ Gen such that m' ≻f m • Expressive Optimization: there is no superoptimal pair 〈f ', m〉 ∈ Gen such that f ' ≻m f. This formulation looks like a circular definition, but Jäger (2002) has shown that this is a sound recursive definition under very general conditions (well-foundedness of the
107
Formal Pragmatics 107 ordering relation). This recursive definition is our expression of the communication game constituted by the neo-Gricean picture as described at the end of section 5.2.1. A simple example should illustrate the difference between the two optimization concepts. The example I will use goes back to McCawley (1978) who observed that the distribution of productive causatives (in English, Japanese, German, and other languages) is restricted by the existence of a corresponding lexical causative. Whereas lexical causatives such as in (1a) tend to be restricted in their distribution to the stereotypical causative situation (direct, unmediated causation through physical action), productive (periphrastic) causatives as in (1b) tend to pick up more marked situations of mediated, indirect causation. For example, (1b) could have been used appropriately when Black Bart caused the sheriff’s gun to backfire by stuffing it with cotton. (1) a. Black Bart killed the sheriff b. Black Bart caused the sheriff to die The example presents a scenario with two forms, kill and cause to die, and two interpretations, dir and indir, referring to direct (stereotypic) causation and indirect causation, respectively. Assuming that the semantics for kill and cause to die admits the same range of interpretations, we get four form–meaning pairs described by Gen: 〈kill, dir〉, 〈kill, indir〉, 〈cause to die, dir〉, 〈cause to die, indir〉. Table 5.1 shows these four pairs together with two markedness constraints, called F and M. The constraint F (for forms) marks complex forms; in the present case it marks the cause to die construction. The other constraint is M (for meanings) and it marks the complex interpretations; in the present case it marks the indirect interpretation. The effect of the F-constraints results in the relation kill ≻x cause to die for any interpretation x, and the effect of the M-constraint results in the relation dir ≻y indir for any form y. The left part (a) of Table 5.1 illustrates the strong version of bidirectionality. Since for both forms, the direct interpretation gives the optimal interpretation and for both interpretations, the form kill gives the optimal expression, the pair 〈kill, dir〉, is the only strongly optimal pair (marked with the symbol ✌ in Table 5.1a). As a consequence, the form cause to die is blocked in each potential interpretation.3 Unfortunately, the prediction of total blocking is intuitively not correct in the present example; instead, the blocking is partial—a llowing 〈cause to die, indir〉 as a second solution pair as predicted by Horn’s division of pragmatic labour. Table 5.1b shows that the weak version of bidirectionality can explain the effects of partial blocking without the stipulation of extra constraints that link forms and meanings directly; in particular, it can explain why the marked form cause to die gets the marked interpretation indir. This is a consequence of the recursion4 implemented 3
Such cases of total blocking are attested in the literature. For example, forms such as *furiosity, *fallacity do not exist because others (fury, fallacy) do. For more examples and discussion, see Blutner (1998). 4 In the original formulation given in section 5.2.1, the recursion is indicated by the modulo clause in the Q-and R-principle.
108
108 Reinhard Blutner Table 5.1 Strong and weak bidirectionality using markedness constraints: (a) shows strong bidirectionality, (b) shows weak bidirectionality (superoptimality) (a) F
✌
M
〈kill, dir〉 〈kill, indir〉
*
〈cause to die, dir〉
*
〈cause to die, indir〉
*
*
F
M
(b)
✌
〈kill, dir〉 〈kill, indir〉
✌
*
〈cause to die, dir〉
*
〈cause to die, indir〉
*
*
in weak bidirectionality: the pairs 〈kill, indir〉 and 〈cause to die, dir〉 are not superoptimal. Hence, they cannot block the pair 〈cause to die, indir〉, and it comes out as a new superoptimal pair (likewise marked with the symbol ✌ in the table). In this way, the weak version accounts for Horn’s pattern of the division of pragmatic labour. In the literature several algorithms have been proposed for formulating explicit recursive mechanisms for calculating superoptimal pairs (Jäger 2002; Beaver and Lee 2004; Franke and Jäger 2012). Unfortunately, there are some doubts about the psychological reality of such mechanisms as models of online natural language interpretation (e.g. Blutner 2010). Instead, it has been proposed to take the diachronic perspective into account, as clearly expressed by Horn (1984). Hence, in the framework of optimality-theoretic pragmatics it is very natural to take weak bidirectionality as expressing a basic principle of natural language change. As a consequence, bidirectional optimization has nothing to do with online processes that run during normal language interpretation/production. Rather, the results of bidirectional optimization are routinized or fossilized—a phenomenon that takes place on an evolutionary timescale. According to this evolutionary view of bidirectionality, form–meaning pairs that have been determined by bidirectional optimization constitute fixed relations to a learner who sets out to acquire the language.
109
Formal Pragmatics 109 Table 5.2 Strong and weak bidirectionality using linking constraints F→M
✌
〈kill, dir〉 〈kill, indir〉 〈cause to die, dir〉
✌
*F→*M
〈cause to die, indir〉
>>
F→*M
F*→M
* * * *
No learner, indeed no user of the language, needs to perform a bidirectional computation for any form–meaning pair she encounters. Let us come back to our simple example in order to get an idea of what ‘fossilization’ could mean. Rather than considering markedness constraints, Table 5.2 presents so-called linking constraints that connect the form level with the interpretational level. In the present example there are precisely four independent linking constraints. The linking constraint F→M says that simple (unmarked) forms express simple interpretations. Hence, this is a straightforward formalization of Levinson’s (2000) I-heuristic as an OT constraint. The constraint *F→*M says that complex forms express complex interpretations, and this is an expression of Levinson’s M-heuristic.5 The two remaining linking constraints express the opposite restrictions. In the present case linking constraints can be seen as lexical stipulations that fix a form–interpretation relation in an instance-based way. Assuming that a general learning mechanism ensures that the two latter linking constraints are finally ranked lower than the former two, then the result of strong bidirectional optimization is the same as the result of weak bidirectional optimization discussed before. In addition, it can be seen from Table 5.2 that unidirectional optimization (taking the hearer’s or the speaker’s perspective) is sufficient already and gives exactly the same results. It is not the place here to discuss real candidate mechanisms for the fossilization process. Such processes can be best understood when related to an offline mechanism that is based on bidirectional learning (Blutner, Borra, Lentz, Uijlings, and Zevenhuijzen 2002; Benz 2003, 2006; Jäger 2004; van Rooij 2004c). In these approaches the solution concept of weak bidirectionality is considered as a principle describing the results of language change: superoptimal pairs emerge over time in language change. This confirms the age-old theory that synchronic structure is significantly informed by diachronic forces.
5 Levinson’s M-principle should not be confused with the markedness constraint M introduced in Table 5.1.
110
110 Reinhard Blutner Let me come back now to the earlier goal of giving an OT reconstruction of the three variations on Grice. For reconstructing Levinson’s (2000) presumptive meaning theory, unidirectional optimization is sufficient where a system of OT constraints has to be formulated conforming to his I, Q, and M heuristics and Levinson’s putative ranking Q > M > I. The unidirectional optimization procedure (interpretive optimization) is to conform with a local approach to conversational implicatures, one which satisfies the requirements of incremental interpretation. The neo-Gricean approach, on the other hand, is globalist in nature. Hence, the idea of (weak) bidirectional optimization fits best with this theory and can be used for a straightforward formalization. Unsurprisingly, this conception can be seen best from a diachronic perspective, as long as we take a naturalistic stance towards Gricean pragmatics. As a model of actual language interpretation (or production) this approach does not make real sense and was never designed for this purpose. Like Levinson’s (2000) approach, RT conforms to the localist approach and can be formulated in terms of unidirectional optimization. Let us stipulate a constraint Effect for describing the wanted effect (a certain value of relevance) and a constraint Effort for describing the cognitive effort. Then the stipulation Effect > Effort makes sure that the wanted effect is reached with the minimal cognitive effort. Obviously, there are many questions left concerning the concrete content of the constraints Effect and Effort, and the RT literature contains a number of specifications. These specifications typically have the character of linking constraints. It might be interesting to investigate recent OT models of pragmatics (see section 5.4) in the light of the general structure of RT—a task that goes beyond what can be done in the present paper. I have mentioned already that there is a relation between diachronic and synchronic systems, and I have introduced the term fossilization for describing the relevant transfer. Given the existence of this transfer, it can be demonstrated that the three variations on Grice discussed here are much more closely related than the occasional polemics led us to expect. Bidirectional OT has been used for describing a series of phenomena and observation in the domain of natural language pragmatics. In the following I give an overview of some of these applications without going into any technical or empirical details. • Disambiguation. Gärtner (2004a, b) analyses Icelandic object shift and differential marking of (in)definites in Tagalog, addressing the issue of disambiguation and partial iconicity in natural language. • Binding theory. Mattausch (2004a, b) introduces the influential work of Levinson on the origin and typology of binding theory (summarized in Levinson 2000; see also Huang 1994/2007, 2000a) and reformulates the different historical stages assumed by Levinson in bidirectional optimality theory. Mattausch’s work is of essential importance as one of the first in-depth studies showing the importance of the diachronic view for bidirectional OT.
111
Formal Pragmatics 111 • Discourse particles and presupposition. Zeevat (2002, 2004) treats discourse particles within an extended OT reconstruction of presupposition theory. In another paper, Zeevat (2007) provides a full solution to the projection problem for presuppositions. • Complex implicatures. Blutner (2007) gives an OT account of implicature projection and explains the relevant theoretic distinction between implicatures and explicatures in terms of a neo-Gricean framework. • Interpretation of stress and focus. Several articles deal with a bidirectional perspective for stress on anaphoric pronouns and the interpretation of focus (Beaver 2004; de Hoop 2004; Hendriks 2004; Aloni, Butler, and Hindsill 2007). • Marking and Interpretation of negation. Henriëtte de Swart (2004) provides a bidirectional OT approach to the syntax and pragmatics of negation and negative indefinites (see also de Swart 2010). • Permission sentences. A series of other articles deals with the interpretations of permission sentences and the analysis of the particular conditions which constitute a so-called free choice interpretation (Sæbø 2004; Aloni 2005a, 2005b; Blutner 2006). • Aspectual interpretation of the Dutch past tenses. Van Hout (2007) applied bidirectional reasoning about tense forms and their aspectual meanings. • Lexical pragmatics: Lexical Pragmatics investigates the processes by which linguistically specified (‘literal’) word meanings are modified in use. Prototypical applications include the pragmatics of dimensional adjectives (Blutner and Solstad 2000), the analysis of Dutch om/rond (Zwarts 2006), the pragmatics of negated antonyms (Blutner 2004; Krifka 2007), gender opposition of animate nouns (Zwarts, Hogeweg, Lestrade, and Malchukov 2009), and several examples of semantic change (Eckardt 2002). • Language acquisition and learning: There are several studies that test the role of weak bidirectionality in developing interpretation and production preferences in connection with (in)definite NPs (de Hoop and Kramer 2005/2006; van Hout, Harrigan, and de Villiers 2010) and pronominal anaphors (Hendriks and Spenader 2005/2006; Hendriks, van Rijn, and Valkenier 2007; Mattausch and Gülzow 2007; Hendriks, de Hoop, Krämer, Swart, and Zwarts 2010; van Rij, van Rijn, and Hendriks 2010).
5.3 Game-Theoretic Pragmatics Wittgenstein is widely acknowledged as the founding father who connected games with natural language interpretation. ‘In the 1950s, the later Wittgenstein famously moved away from the crystalline logical structure of the Tractatus to a paradigm of rule-generating “language games”’ (van Benthem 2008: 198). Another prominent
112
112 Reinhard Blutner researcher applying game theory to natural language interpretation is Hintikka, who introduced evaluation games in order to specify the truth-conditional semantics of certain fragments of natural language. In 1969, David Lewis’s book Convention was published, which introduced the important idea of signalling games to the domain of language (Lewis 1969). These games refer to the phenomena of ‘cultural evolution’ (Hurford 1998; Steels 1998) and aim to give a mathematical justification of the formation of stable meanings (Nash equilibriums of signalling games). In this section, I will discuss the close connection between OT pragmatics and game-theoretic pragmatics. In the literature, we find two different kinds of games capturing the strong and the weak solution concept in bidirectional OT: strategic games (Dekker and van Rooij 2000) and signalling games (Benz, Jäger, and van Rooij 2005a; Franke 2009; Franke and Jäger 2010). The interesting point is that the game-theoretic approach provides a richer system of solution concepts than the optimality-theoretic one. Further, it proposes an impressive spectrum of possible ways to approximate the intended solutions—as defined by a certain solution concept—by means of iterative reasoning protocols (Franke and Jäger 2012). Hence, we can expect from the game- theoretic approach not only a sound explication of optimality-theoretic solution concepts, but also new solution concepts and algorithms that challenge the standard optimality-theoretic ones. Strategic games are games where players move simultaneously. This contrasts with sequential games where players move in sequence. Let us assume a strategic game with two players (called speaker and hearer). Speaker’s possible actions are given by the set of possible forms, kill and cause to die in the present example; the hearer’s possible actions are given by the set of possible meanings, dir and indir in our example. Generally, in (two-person) games, pairs of the two players’ actions are called profiles. Hence, in our example profiles are the four possible form–meaning pairs indicated by the four small circles in Figure 5.1. In game theory, solution concepts are formal specifications of certain optimality concepts. They are rational/normative concepts relating to the reasonable choices which players may make. A famous solution concept is that of a ‘Nash equilibrium’. A Nash equilibrium for a strategic game (with two players S and H) is an action profile 〈aS, aH〉, such that each player’s action is an optimal response to the choices of the other kill
cause to die dir
indir
Figure 5.1 Nash equilibrium (indicated by •) in a concrete example
113
Formal Pragmatics 113 players in that profile, i.e. for the speaker there is no action xS such that 〈xS, aH〉 ≻S 〈aS, aH〉 and for the hearer there is no action yH such that 〈aS, yH〉 ≻H 〈aS, aH〉. In Figure 5.1 there exists exactly one Nash equilibrium indicated by the black circle. The horizontal arrows indicate the strict preferences for the hearer (the arrow directs to the stronger pair) and the vertical arrows show the strict preferences for the speaker. Informally, a pair is a Nash equilibrium if no arrow leads away from this pair. Franke (2009) pointed out that it is conceptually not very plausible to use strategic games for modelling language use. Literally taken, the two players in a strategic game (say speaker and hearer) make their choices independently from each other. The speaker chooses the preferred (lightest) form and the hearer chooses the preferred (simplest) interpretation. Even when the resulting form–meaning pair is realized in our natural language system, the underlying solution concepts for strategic games are far from providing a plausible (causal) argument why the selected form is connected with the selected meaning. Franke (2009) argued that OT systems would better be translated into some kind of sequential game with imperfect information where for a given meaning the speaker chooses a corresponding form to express this meaning, and the hearer subsequently tries to guess at this meaning on the basis of the uttered form. ‘A natural idea is to consult signaling games, a class of games which are widely used for the study of strategic communication not just in linguistics, but also in biology, economics, and the philosophy of language (c.f. Lewis 1969, Spence 1973, Grafen 1990)’ (Franke and Jäger 2012: 5). This is not the place to give a detailed introduction into the idea of signalling games and its use in formal pragmatics as done in some recent monographs (Lewis 1969; Parikh 2001; Benz, Jäger, and van Rooij 2006; Franke 2009). Instead, I will develop a simplified basic picture and outline some relations to OT pragmatics. My exposition will follow the presentation of Jäger (2007a). The basic idea of a signalling game is rather simple. In the most straightforward case we have two players called S (the speaker) and H (the hearer). The game begins with a randomly chosen meaning m that is presented to S (but not to H). Next, the speaker is requested to choose a signal f that is transmitted to H. On the basis of this signal, the hearer H is asked to guess the meaning of f. If the guess is correct, i.e. H selects the meaning m, both S and H score one point, if not both get nothing. It is easy to see that in signalling games the interests of the two players completely coincide. Further, these games are asymmetric games because the two roles of speaker and hearer are not interchangeable. What are the possible strategies of the two players? In the simplest case of deterministic (pure) strategies, a possible strategy of the speaker is a function from meanings to forms and a possible strategy of the hearer is a function from forms to meanings. I will write S(m) for applying a speaker strategy to a meaning m and H(f ) for applying a hearer strategy to a form f. Further, a similarity function sim(m, m’) is used which gives the value 1 if m and m’ completely agree and gives the value 0 if they are maximally different.6 With these prerequisites at hand, the utility function of the game is 6
For a short characterization of similarity, the reader is referred to Jäger (2007a). For a detailed treatment, see Tversky (1977).
114
114 Reinhard Blutner given by the following equation assuming the prior probabilities P(m) are common knowledge between the two players:
(
(
))
(2) u (S, H ) = ∑ m P (m) sim m, H S (m) There is no distinction made between the utility of the speaker and the hearer. Because the games in question are real partnership games, both players always obtain the same utility. Equation (2) expresses our basic intuition that communication is successful if the players of the signalling game understand each other. At the moment, the costs for interpreting the signal and generating the signal are ignored. However, it is not difficult to subtract a corresponding cost value c(f, m) from the left-hand side of equation (2) in order to take the costs for signalling and interpretation into account. This can be done in correspondence to a given OT system by respecting the following relations (e.g. Franke 2009; Franke and Jäger 2012): (3) m , f > m , f 1 1 2 2
iff c(m1 , f1 ) < c(m2 , f 2 ).
Intuitively, in signalling games the two players try to find strategies that maximize the value of the utility function u(S,H). There are different ways in which this could be realized. First consider the evolutionary interpretation of game theory (cf. Jäger 2007a). In this case the utility of a strategy is to be interpreted as the expected number of offspring of a player adopting this strategy. Technically, this is described by so-called replicator dynamics—a deterministic continuous time dynamics for sufficiently large populations. Since it is extremely difficult to solve the differential equations determining the replicator dynamics in an analytic way, Maynard-Smith (1982) developed a way to characterize the qualitative behaviour of the replicator dynamics. The central conception is that of an evolutionary stable system. Informally, an evolutionarily stable strategy is characterized by the configuration of a population that is stable in the sense that the population does not leave its state due to its inherent dynamics, and is protected against small amounts of mutation. Under certain conditions it can be proven that Nash equilibriums are evolutionarily stable. As a consequence, the superoptimal solution pairs discussed above come out as evolutionarily stable (Benz 2003; van Rooij 2004c; Benz 2006; Lentz and Blutner 2009; Franke and Jäger 2012). A second mechanism is iterated learning theory where the utility function (2), or a modification of it, is optimized by an iterated learning process where speaker and hearer learn from each other (cf. Kirby and Hurford 1997, 2002; Jäger 2004; Mattausch 2004a, b; Benz 2006; Mattausch and Gülzow 2007). Using straightforward OT learning models, the results are similar to those found in the evolutionary interpretation of game theory. Similar results are found by investigating reinforcement learning for signalling games (Franke and Jäger 2010). A third idea is realized by the iterated best response model as proposed by Matsui (1992). It was recently applied to pragmatics (Jäger 2007b; Franke 2009; Franke and
115
Formal Pragmatics 115 Jäger 2012). This model proposes a particular evolutionary interpretation of signalling games. Other than in Darwinian evolution where new strategies only emerge due to undirected random mutation, the present model suggests that whenever a new member enters the population they may freely choose their strategy. If it is assumed that the new members are rational enough to maximize their expected utility they will choose a strategy that is an optimal response to the average strategy of the population. By repeating the addition of new members indefinitely, an optimal response dynamics is defined which is different from the standard replicator dynamics sketched above. Interesting differences include the emergence of scalar implicatures and total blocking (Jäger 2007b; Franke 2009; Franke and Jäger 2012). In section 5.1 we discussed the normative and the naturalistic aspects of understanding human actions. Interestingly, the three approaches to signalling games discussed in this section clearly exhibit the naturalist stance. And they clearly relate to offline aspects of natural language processing (cultural evolution, language change, bidirectional learning). This idea corresponds to the understanding of weak bidirectionality which relates best to an offline mechanism that is based on bidirectional learning (Blutner, Borra, Lentz, Uijlings, and Zevenhuijzen 2002; Benz 2003; van Rooij 2004c). It suggests that the borderline between semantics and pragmatics is transparent in at least one direction: tendencies predicted from pragmatics (conversational implicatures modelled by weak bidirectionality) may become frozen or fossilized in the semantic component of knowledge representation. The details of the fossilization process are an open problem. Obviously, evolutionary game theory and variants of it may be a powerful instrument to explore different hypotheses concerning the self-organizing dynamics of language as an observationally learned and culturally transmitted communication system.
5.4 Decision-Theoretic Pragmatics Decision-theoretic pragmatics (Merin 1999) is closely related to argumentation theory (Ducrot 1972, 1973, 1980, 1984). According to this view, utterances are normally used as premises and conclusions in arguments. It is this argumentative use in language that determines the meaning of utterances in discourse. Interestingly, this conception of meaning goes far beyond what is normally described as the truth-conditional conception of meaning. For instance, utterances with the same informational content can be used as arguments for quite different things. A famous example is due to Anscombre and Ducrot (1983): (4) a. Should we buy this ring? b. It is nice but expensive. c. It is expensive but nice. Assuming that the informational (= truth-conditional) content of (4b) and (4c) is the same, Anscombre and Ducrot (1983) claim that (4b) and (4c) argue for opposite
116
116 Reinhard Blutner things when seen in the context of (4a): (4b) argues for not buying the ring; (4c) argues for buying it. Examples like this led many authors to believe that a purely truth- conditional semantics is not sufficient for an adequate meaning description and that the ‘argumentative potential’ of an utterance forms an essential part of its meaning. Merin (1999) has started the formalization of the key ingredients for an argumentative theory of pragmasemantics. His theory is based on the classical conception of probability and elements of decision theory. He gives a precise definition of concepts like informativeness and relevance, and he makes precise the idea of issue-based communication. According to Merin (1999), argumentation is a probabilistic relation over epistemic states. A proposition A is a positive argument for a hypothesis H iff accepting A increases the probability of H. It is a negative argument for H iff accepting A lowers the probability of H. The formal expression of the argumentation relation is the ‘relation of relevance’: (5) rH (A) = log(P ( A | H ) − log(P( A | ¬H ) Positive relevance means rH(A) > 0; negative relevance means rH(A) < 0. A simple consequence of the definition in (5) is that rH(A) > 0 iff P(H|A) > P(H) and rH(A) < 0 iff P(H|A) < P(H). For a proof one has simply to make use of the Bayesian formula. One of the cornerstones of argumentation theory is a semantic analysis of ‘but’. We will look at it since it involves an appealing application of the conception of relevance. What is the main phenomenon we have to describe? In a seminal paper, Lakoff (1971) distinguished two different uses of but, the ‘contrast use’ and the ‘denial of expectation use’. (6) a. John is tall but Sue is short. b. John is a Republican but he is honest. Examples like (6a) illustrate the contrast use of ‘but’. Such examples are always symmetric, i.e. if the order of the conjuncts is reversed no significant meaning changes are induced. Further, the substitution of ‘but’ by ‘and’ does not induce a significant change of meaning. Examples like (6b) illustrate the denial of expectation use. Such utterances are typically not symmetric and the substitution of ‘but’ by ‘and’ leads to significant changes of meaning: (6b) suggest that Republicans are normally not honest, whereas the reverse of the conjuncts suggests that honest persons are normally not Republicans. It is not difficult to see that the argumentative approach works pretty well for the denial of expectation use and allows us to express an important constraint stated by Anscombre and Ducrot (1983). As shown in Winterstein (2011), this constraint can be formulated using Merin’s (1999) notion of relevance: (7) For an utterance of the form p but q, there must be an H such that: a. rH ( p ) > 0 and rH (q ) < 0 (or equivalently r¬ H (q ) > 0) b. r¬ H (q ) > rH ( p )
117
Formal Pragmatics 117 Unfortunately, a description of the contrastive case is less obvious. For example, in (6a) it is not really clear what the debated hypothesis should be.7 There are different kinds of criticism concerning the argumentative framework in general and Merin’s decision-theoretic treatment in particular. For example, van Rooij (2004a) has argued against Merin’s view that the two participants of a dialogue play a zero sum game with opposite preferences (if one agent prefers H to be true the other prefers it to be false). Instead, van Rooij argues that the participants of a dialogue are cooperative and this should be reflected in the conceptual grounding. Without going into details, some groups of examples should be mentioned where Merin’s approach is not very explanatory and comes into considerable trouble: numerals, temperature expressions, disjunctions, and particularized scalar implicatures (for the details, see van Rooij 2004b, c). In an important paper, Iten (2000) has argued that several insights of the argumentative approach could be integrated into a Gricean framework. Comparing Anscombre and Ducrot’s (1977) treatment of ‘but’ and a recent relevance-theoretic analysis (Blakemore 2002) Iten comes to the conclusion that the two analyses ‘are remarkably similar and, arguably, the grounds for choosing between them lie more with their theoretical underpinnings than with the details of the particular accounts’ (Iten 2000: 665). Iten (2000) lists and criticizes several central conjectures of argumentation theory. Among them are the following assumptions: • Argumentation theory clearly takes the normative stance. It is a non-cognitive theory. • The semantics of utterances provided by argumentation theory is not truth- conditional. It aims to specify the ‘argumentative potential’ of the utterance. Further, the argumentative potential of an utterance does not depend on the recovery of some prior truth-c onditional meaning component. This contrasts with Gricean conversational implicatures depending on the recovery of ‘what is said’, i.e. the literal meaning that is expressed in a truth-c onditional way. • Anscombre and Ducrot (1983) use the term ‘pragmatique intégrée’ (integrated pragmatics) in order to indicate a uniform approach that is directed to the analysis of the non-truth-conditional aspects of utterance meanings. This term suggests that there is no semantics/pragmatics distinction in argumentation theory. • Anscombre and Ducrot’s (1983) concept of ‘comparative argumentative strength’ encounters counterexamples. Concerning the lexical entries of ‘but’, an important consequence of these assumptions is that ‘but’ has to be ambiguous since there is only one representational level 7
For details, the reader is referred to Winterstein (2011) who proposes an augmented argumentative approach for the contrast use.
118
118 Reinhard Blutner where the pragmasemantics of ‘but’ can be described, and the different uses of ‘but’ are to be assumed as its different meanings. Contrasting with the argumentative approach, Umbach (2005) does not accept any ambiguity for ‘but’; rather, she stipulates a core meaning for ‘but’. This core meaning comes close to what was described as the contrast use of ‘but’. Importantly, Umbach is able to show that the denial of expectation use can be derived via a general mechanism of contextual enrichment. A similar treatment was proposed by Sæbø (2003), who analyses the content of the derived material as a presupposition rather than an implicature proper. Winterstein (2009) provides a critical discussion of both approaches and argues that that there are examples such as (8) that they cannot handle: (8) Lemmy plays the bass, but Richie plays it too. Similar arguments are put forward by Zeevat (2011). However, these counterexamples do include some systematicity that connects the different uses of ‘but’. This makes it obvious that the underlying systematicity has to be described in a quite different way. What could such an alternative analysis look like? One possibility is to make use of the idea of underspecification that is prominently connected with the view of radical pragmatics (see section 5.2) and the idea of a mechanism of pragmatic enrichment. I fully agree with Zeevat (2011: 15) who states ‘that progress in the understanding of “but” is in being more precise about how to find the missing object: the question in Umbach’s (2005) account, the issue that is argued for and against in the argumentative tradition, the manifest inference or the statement under objection—it is fairly easy to go from one to the other’. Assuming that Zeevat (2011) and Winterstein (2011) are on the right track paves the way for an integration of ideas of radical pragmatics with argumentation theory. Concerning formal pragmatics, it could be useful then to combine the ideas put forward in sections 5.2 and 5.3 with crucial ideas of argumentation theory.
5.5 Conclusions In this article I have outlined the close relations between optimality-theoretic pragmatics and game-theoretic pragmatics. It has been suggested that the OT approach can profit from evolutionary game theory and variants of it in exploring different hypotheses of the self-organizing dynamics of natural language as an observationally learned and culturally transmitted communication system. Further, it has been claimed that it may be useful for the game-theoretic paradigm to overcome some bounds of the normative stance and to consider the realization of the evolutionary account and the implementation of the iterated best-response model within a plausible cognitive setting. Finally, I have critically discussed argumentation theory and decision-theoretic pragmatics. My criticism does not imply that the argumentative framework is obsolete and not worth a serious study. There is no doubt that a lot of excellent analytical work
119
Formal Pragmatics 119 has been done within the argumentation-theoretic framework based on interesting and new observations. Further, argumentation theory has highlighted the non-truth- conditional aspects of meaning and has made clear that some words (such as good, interesting, and lovely) are intrinsically subjective. However, some arguments were put forward suggesting that an integration of argumentative tradition with a (neo/post)- Gricean perspective is possible and useful.
Acknowledgement I am indebted to Stefan Blutner, Michael Franke, Jason Mattausch, Carla Umbach, Grégoire Winterstein, and Henk Zeevat for discussion, and to Paul Dekker and Yan Huang for encouragement.
120
Chapter 6
Con ti n en ta l Eu ropea n Perspecti v e V iew Jef Verschueren
6.1 Introduction The title of this chapter, as well as its placement in opposition to ‘the Anglo-American component view’ as two ‘main schools of thought’ in linguistic pragmatics, carries at least three presuppositions that must be addressed immediately. The first is that there are two separable traditions or schools that can be contrasted with each other. The second is that one of those schools can be characterized as Anglo-American, the other as (Continental) European. Thirdly, there is the assumption that the Anglo-American school can be associated with a component view, as distinct from the European perspective view. All three can and should be questioned. But I hope to show that—in doing so—relevant observations can be made about the shaping and development of ideas about and approaches to language use. The origin of the distinctions is to be found in Levinson’s (1983: ix) characterization of the contents of his pragmatics textbook—a classic in the true sense of the word—as belonging to ‘the largely Anglo-American linguistic and philosophical tradition that builds directly, for the most part, on philosophical approaches to language of both the logical and “ordinary language” variety’, in contrast to ‘the continental tradition [which] is altogether broader, and would include much that also goes under the rubric of sociolinguistics’. Ever since, the distinction has been echoed throughout the pragmatics literature. One of the more recent and more explicit ways of handling it, we find in Huang (2007). According to him, in the Anglo-American conception ‘pragmatics is defined as the systematic study of meaning by virtue of, or dependent on, language use’ (Huang 2007: 4), focusing on topics such as implicature, presupposition, speech acts, and deixis, and positioning pragmatics in contrast to phonetics, phonology, morphology, syntax, and semantics as an additional component of a theory of language, quite distinct from the interdisciplinary areas of anthropological linguistics, sociolinguistics,
121
CONTINENTAL EUROPEAN Perspective View 121 and the like. For the Continental European tradition, on the other hand, ‘pragmatics is defined in a far broader way, encompassing much that goes under the rubric of sociolinguistics, psycholinguistics, and discourse analysis’ (Huang 2007: 4), and representing a general functional perspective on every possible aspect of linguistic behaviour, in line with Morris’s (1938) all-inclusive definition of pragmatics as the study of whatever relations there are between signs and their users or interpreters.
6.2 Schools of Thought? As to the first of our presuppositions, then, we must say that there are indeed two contrasting ways of looking at what it is that a field of linguistic pragmatics should do or include within its scope. A qualification as a tradition or school, however, would be more appropriate for the Anglo-American pole in the opposition for the simple reason that it displays stronger continuity and coherence. In fact, representatives of the so-called Anglo-American tradition have often argued against an alternative, more inclusive, approach on the basis of its predictable fragmentation and its likely elusiveness. To use Davis’s words: The problem with this broad view of pragmatics is that it is too inclusive to be of much use. Using this definition, pragmatics has as its domain any human activity involving language, and thus includes almost all of human activity, from baseball to the stock market. [ … ] What groups various activities and theories together in one field of study is that they share a set of questions or a methodology. But there is no common methodology or set of questions that groups together in a natural class the full range of human sciences in which language is involved. [ … ] For ‘pragmatics’ to be a useful term, its domain must be restricted. (Davis 1991: 3–4)
In a similar vein, Morris’s claim that Since most, if not all, signs have as their interpreters living organisms, it is a sufficiently accurate characterization of pragmatics to say that it deals with the biotic aspects of semiosis, that is, with all the psychological, biological, and sociological phenomena which occur in the functioning of signs. (Morris 1938: 30)
is paraphrased by Huang (2007: 5) as ‘the study of “everything”’, adding that this ‘is hardly a viable academic enterprise’ and that Given the degree of overlap among the phenomena dealt with in other relatively well-established interdisciplinary fields of linguistics such as sociolinguistics, psycholinguistics, and neurolinguistics, it is rather difficult to see how a coherent research agenda of pragmatics can be made within the wider Continental tradition. (Huang 2007: 5)
122
122 Jef Verschueren Ariel (2010) introduces the metaphor of a big-tent pragmatics, not only to refer to the expansionist traits of what a Continental tradition would try to capture, but also to point at the absence of a clear definition of the field when restricted to the variety of phenomena included in the list of topics covered in the Anglo-American tradition. She then goes on, however, to develop a definition of the field in terms of inferential processes (as opposed to grammar) that could establish coherence within the Anglo- American realm, not automatically excluding sociocultural and psycholinguistic phenomena from pragmatics, but still identifying most of what goes on in the so-called hyphenated disciplines as being beyond pragmatics. Using Levinson’s wording, it is true that the so-called Continental European or truly ‘big-tent’ tendency in pragmatics looks like ‘a mere flag of convenience under which divergent interests can momentarily find common profit in academic coalition’ (Levinson 1987b: 61). To what extent it is justified, therefore, to speak of a tradition (or school, for that matter) on that side of the equation, is unclear. Probably it is not.
6.3 Anglo-A merican vs (Continental) European? Be that as it may, we must address the second question: can one of the traditions or schools, whether or not they qualify as such, be characterized as Anglo-American, the other as (Continental) European? This geographical anchoring is problematic, to say the least. Let us first have a look at historical roots. Of the main topics within the Anglo-American tradition, speech act theory and conversational implicature clearly originate in the work of two British philosophers, John Austin and Paul Grice, respectively, while studies of conversational interaction find their origin in an ethnomethodological tradition in American sociology. The latter, however, uses as a source of inspiration European phenomenological philosophy. As to deixis, another one of the favourite topics, remember the contributions made by Bühler, as well as Jakobson, himself a Russian linguist with an intellectual history that passes through Prague, Copenhagen, and Sweden before ending up in America. And the study of presupposition goes back at least to the work of the German philosopher Gottlob Frege. Conversely, the full-size big-tent version of pragmatics was first introduced by the American pioneer of semiotics, Charles Morris. Moving from history to present-day research practice, there are clearly proponents of both a narrower and a broader take on pragmatics on all continents. The distinction between Anglo-American and Continental European traditions usually overlooks influential local groups which escape from the categorization but which can be seen to lean in one direction or the other, but not following the geographical lines. Thus matters of great linguistic detail, very much akin to the typical Anglo-American topics, are being addressed in what could be called a French school of pragmatics (inspired
123
CONTINENTAL EUROPEAN Perspective View 123 by both British analytical philosophy and Bakhtin’s approach to utterances in terms of multiple voices; e.g. Ducrot 1996) and Prague school linguistics (addressing intricate issues of sentence-level information structuring and perspectivization; see Sgall 1995). Conversely, a very wide look at language as a social instrument emanates from the distinctly British tradition of Firthian linguistics (Firth 1964a) and feeds into the social semiotics behind pragmatically oriented theories of grammar such as Halliday’s (1985) systemic functional grammar. And whether or not they are inclined to officially bear the label pragmatics when joining the academic coalition gathering on the occasion of international pragmatics conferences, there are many lightly to heavily UK-and US-based trends that definitely belong to the more inclusive side of the pragmatic spectrum in the ethnography of communication (e.g. Saville-Troike 1982), in other strands of anthropological linguistics (e.g. Duranti 1997a), in interactional sociolinguistics (e.g. Gumperz 1982a), in cognitive linguistics (e.g. Evans and Green 2006; Langacker 1987), in conversation analysis (e.g. Drew and Heritage 1992), in critical discourse analysis (e.g. Fairclough 2003), to name just a few. Moreover, while most of the pragmatics textbooks written by British or American authors do indeed have a tendency to restrict their coverage to the traditional list of topics including deixis, presupposition, implicature, speech acts, certain aspects of conversational structure, and sometimes politeness phenomena (e.g. Leech 1983; Thomas 1995; Yule 1996; Robinson 2006; O’Keeffe, Clancy, and Adolphs 2011), some are distinctly multidisciplinary in orientation (notably Cummings 2005), while there are also truly Continental students of language who explicitly engage with theory formation within the boundaries set by the so-called Anglo-American approach (e.g. Recanati 2010). But even consecutive editions of the ‘same’ book may take different approaches: while Grundy (1995) did not venture outside the Anglo-American picket line, Grundy (2008) more resolutely includes matters of cognition, real-world interaction, and culture.
6.4 Component vs Perspective? As to the third questionable presupposition, can the Anglo-American variety of pragmatics be called a component view, as distinct from the European perspective view? It is certainly true that work within the Anglo-American tradition, as soon as it had moved beyond what Bar-Hillel (1971) called the waste-basket approach which dumped everything that syntax or semantics could not deal with onto an undifferentiated pile of supposedly pragmatic phenomena, was strongly preoccupied with issues of delimitation (more, in fact, than with true definition; see Ariel 2010). The dominant paradigm which pragmatics needed to fit into was one that conceptualized the field of linguistics as a set of distinguishable components, each with its own object of investigation or unit of analysis (and, in a more daring version, each corresponding to a designated module of the cognitive processing capabilities of the human mind): phonetics and phonology dealing with speech sounds, morphology and syntax with the
124
124 Jef Verschueren structuring of words and sentences, semantics with meaning at the lexical and propositional levels. In this framework, with pragmatics as an additional component of a theory of language, the main point of discussion was the cut-off between semantics and pragmatics (leading to extensive disputes on issues such as the borderline between semantic and pragmatic presuppositions). It is also true that the idea of a perspective view was launched within the Continental European tradition. Implicitly it was certainly there in Haberland and Mey’s (1977) editorial in the first issue of the Journal of Pragmatics: Linguistic pragmatics [ … ] can be said to characterize a new way of looking at things linguistic, rather than marking off clean borderlines to other disciplines. (1977: 5)
The first explicit formulation, however, followed only ten years later, in 1987. That year saw the publication of the proceedings of an International Pragmatics Conference held in 1985, with The Pragmatic Perspective at its title (Verschueren and Bertuccelli-Papi 1987), as well as a further development of the perspective idea in the first one of a short- lived series of Working Documents (Verschueren 1987) published by the International Pragmatics Association in preparation for a Handbook of Pragmatics (publication of which started in 1995; see Verschueren, Östman, and Blommaert 1995–). In addition to criticism directed at the Anglo-American view of pragmatics for its failure to provide coherent theorizing in spite of its restriction of scope, and adducing arguments for an alternative and much broader perspective view, this ‘Working Document’ already offered suggestions for a unifying framework, centred around the notion of adaptability, and tentatively linking topics of pragmatic importance to the different angles of this framework. The lists of topics were simply based on an inductively constructed index accompanying a comprehensive bibliography of pragmatics (Nuyts and Verschueren 1987), cumulatively compiled since the mid-1970s (when much smaller precursors were published), already with the big-tent vision of pragmatics in mind. The basic idea behind what came to be called the perspective view was very simple. Pragmatics can briefly be characterized as the science of language use. When using language, all levels of linguistic structure are involved. Therefore, there are pragmatic aspects, describable and explainable in terms of usage conditions and processes, to be situated in relation to every unit of linguistic analysis, no matter whether they have already been assigned to the realm of one of the traditional components of a linguistic theory. Similarly, using language involves cognition, social structures, and cultural context. Therefore, much of what finds its way into the hyphenated, interdisciplinary fields of linguistics must also be relevant to pragmatics. These considerations led to a view of pragmatics that was later summarized as ‘the study of linguistic phenomena from the point of view of their usage properties and processes’ (Verschueren 1999: 1). The strong association of the perspective view with a so-called Continental European tradition does not mean, however, that all linguists belonging to the Continental European camp, both geographically and in spirit, would want to subscribe to the
125
CONTINENTAL EUROPEAN Perspective View 125 characterization of pragmatics as a perspective rather than a component. Mey (2001), for instance, siding with Östman (1988), pleads for a field of pragmatics that could be defined both as a component of a linguistic theory and as a perspective on language, ‘so as to expand, rather than narrow, our epistemological horizon’ (Mey 2001: 9). One of Östman’s arguments for allowing for a component view that would not be incompatible with a perspective view takes issue with the suggestion that, in contrast with the traditional components, pragmatics would not have its own unit of analysis: Or is the unit of analysis in semantics simply meaning; the meaning of words, phrases, larger constructions, prosody, and so on? If so, then by the same token, the ‘unit’ of analysis for pragmatics could be said to be the functioning of language (or simply adaptation, for that matter). (Östman 1988: 29)
In Mey’s words: Thus, we could have a pragmatic component, understood as the set of whatever pragmatic functions can be assigned to language, along with a pragmatic perspective, i.e., the way these functions operate. We could either ask how users ‘mean what they say’, that is, how they communicate, using language, or how they ‘say what they mean’, employing the linguistic devices at their disposal to express themselves. (Mey 2001: 9)
Others have suggested that pragmatics would have discourse as its unit of analysis. All this means that in the end, it does not really seem to matter that much whether one adheres strictly to a perspective view. Even if one does not, the length and breadth of what the (so-called) Continental European tradition would like to include in pragmatics, can be covered.
6.5 The Search for Unity in an Interdisciplinary Field More important than the status of pragmatics in relation to other linguistic endeavours, is the attempt to find some unity in the wide range of phenomena to be covered by the big tent. In a fundamentally interdisciplinary field, one may be satisfied with the fact that differently oriented practitioners can talk to and inspire each other, as John Marshall did in a discussion on the topic of possible unity on the occasion of the 1987 International Pragmatics Conference: JOHN MARSHALL: [ … ] The pragmatic justification for even having the pragmatic label, I think, is that you never know where the next good idea that is relevant to you is going to come from. It might come from a psychologist, a sociologist,
126
126 Jef Verschueren a linguist, a cultural anthropologist, it may come from anywhere. If pragmatics provides a useful label for letting all such people talk to each other now and again, that seems to me to be the best justification you could ever have for pragmatics. (Verschueren 1987: 53)
This may sound very much like Levinson’s idea of an accidental coalition. Yet, it captures a significant aspect of the academic dynamics and spirit that has shaped pragmatics as the interdisciplinary, linguistic, cognitive, social, and cultural science of language use that it is today in its widest sense. This does not mean that everyone who is involved in the dynamics believes that pragmatics must be defined in such an inclusive way. In other words, the coalition is quite a loose one, including many adherents of a much more restrictive view of pragmatics (whether or not one would want to call it Anglo-A merican), who nevertheless assume that interaction with what they would regard as the fringes may be fruitful. Thus, Huang (2010d: 15) signals a useful ‘convergence between the two camps’. Moreover, those whose work is compatible with the more inclusive version of pragmatics are not necessarily in pursuit of a coherent overall framework. For instance Jacob Mey, perhaps its most prominent adherent, will not be satisfied with a form of pragmatics that would avoid issues of culture and society (or even socio-political engagement). Yet he does not make an attempt to provide an all-inclusive definition. From his writings (such as Mey 2001) it is clear that he is strongly in favour of the peaceful coexistence of a variety of types of pragmatics, though the climax of his personal academic concerns is summarized in the question ‘Whose language?’ (as in Mey 1985), probing the ways in which language (in particular its ownership in terms of unequal access to its resources) reflects, creates, and maintains social structures and power relationships. Attempts have also been made to introduce real definitional coherence in spite of the broad scope and without—let’s be clear about that—abandoning the goal of peaceful coexistence. Maybe the first was Östman’s (1986) Pragmatics as Implicitness. Östman opts, very explicitly, for a perspective view of pragmatics that allows for components as well and thus requires a relatively clear delimitation of its scope. For him, there are three components of a theory of language that really count and that have to be viewed as separable but intertwined: [ … ] syntax should indeed deal primarily with form, but it should do this in relation to meaning and use; similarly, semantics should deal primarily with meaning, but in relation to form and use; and pragmatics should deal primarily with use, but in relation to form and meaning. (Östman 1986: 13)
In his account, what characterizes the pragmatic perspective, then, is that (i) it starts from the point of view of the behaviour of speakers/hearers, individuals in relation to a world of other individuals, hence in a context of culture; (ii) its true province is, in contrast to truth-functional and propositional aspects of a message, the realm of
127
CONTINENTAL EUROPEAN Perspective View 127 meanings that are communicated implicitly. Implicitness hinges on linguistic choices ‘that the speaker in principle can deny that s/he has made’ (Östman 1986: 25). The expressibility of implicit meaning is then conceptualized by Östman in terms of the three parameters of coherence, politeness, and involvement, which are the angles from which the phenomenon of implicitness can and should be approached. Further clarification concerning the positioning of pragmatics is offered in the following reformulation: In the present theory I therefore keep a tripartition of the conglomerate called ‘language.’ The three parts are STRUCTURE, SEMANTICS, and PRAGMATICS. Roughly, structure is language specific and refers to the form, the means, the tool; semantics is (in the last resort) individual specific, and refers to the cognitive, psychological make up of a speaker; and pragmatics is culture specific, and refers to the social and the cultural. I talk about the Syntactic perspective and the Pragmatic perspective on language: you enter language from two different perspectives, simultaneously; and the Semantic filter, through which you abstract and conceptualize both the information from the two perspectives, and from the world around you. Thus the theory of linguistic behavior that I am here advancing I call The Perspectives— Filter View of Language. (Östman 1986: 316–317)
The applicability of this approach is illustrated with a lengthy analysis of how aspects of utterances ranging from question particles to passive constructions and strategies of persuasion function in the implicit anchoring to the parameters of coherence, politeness, and involvement (which are, of course, interdependent as well). An alternative, first tentatively formulated in Verschueren (1987) and further developed in Verschueren (1999), differs mainly from the ‘perspectives—filter view’ in that (i) the special status of pragmatics as a perspective rather than a component is more strongly emphasized, and (ii) the distinction between cognitive and psychological aspects on the one hand (placed under the semantics umbrella by Östman) and social and cultural aspects on the other is not made. The core elements of the theory, which was not originally intended as a theory at all but rather as a relatively loose framework for the discussion of the widest possible range of pragmatic phenomena, can be quickly summarized. My starting point is that using language is essentially an activity that generates meaning. It consists in the continuous making of choices, not only at various levels of linguistic structure, but also pertaining to communicative strategies and even at the level of context. Choice-making characterizes both language production and language interpretation. It can be a process or activity that takes place with varying degrees of automaticity or consciousness. While not all choices are equivalent (some may be more marked than others), they always evoke or carry along their alternatives by way of contrast. But never can choice-making be avoided, and always it is mediated by a human cognitive apparatus involving metapragmatic reflexivity and exerting a monitoring influence.
128
128 Jef Verschueren A first key notion to make sense of the process or activity of choice-making is variability, i.e. the property of language and of contexts of language use which defines the range of possibilities from which choices can be made. This range is itself not stable; it is fundamentally changing and changeable. A first pitfall to watch out for in pragmatic analysis is, therefore, the underestimation of variability. A second essential notion is negotiability, referring to the fact that choices are not made mechanically or according to strict rules or fixed form–function relationships, but rather on the basis of highly flexible principles and strategies. This property is responsible for various forms of indeterminacy of meaning, but at the same time for the vast meaning potential of limited (though always expandable) means. Its methodological implications are a major concern, as the temptation to draw conclusions mechanically from the observation of formal patterns, is always there. Finally, adaptability is what enables people to make negotiable choices from a variable range of possibilities in such a way as to approach points of satisfaction for communicative needs. This is a complex notion which can be used to link evolutionary aspects of language with the processes involved in language use (see Verschueren and Brisard 2002). Here the term refers essentially to the dynamic and negotiable interadaptability of forms and functions in the making of meaning. It is this notion that enables us to define four research angles, none of which should be ignored when approaching discourse data: contextual correlates of adaptability have to be identified; processes have to be situated with reference to different structural objects of adaptability; the dynamics of adaptability must be accounted for; and we must keep in mind the salience of the adaptation processes, i.e. their status in relation to a human cognitive apparatus. Basically, context and structure form the locus of the processes to be investigated. They are strongly interrelated anchoring points for any linguistic pragmatic research methodology, while always keeping in focus that what ultimately concerns us is the dynamics of meaning generation, which is the meaningful functioning of forms of expression in relation to human minds. Figure 6.1 summarizes the above remarks about the structure of a theory of pragmatics. Note that, as the visualization suggests, context and structure are intimately related (see e.g. Verschueren 2008). For one thing, as soon as an utterance is made (i.e. as soon as a structure is produced) it becomes part of the context. Second, structural choices (e.g. the choice of an informal form of address) may affect properties of context (in the
CONTEXT STRUCTURE
Locus Meaning generation Processes
DYNAMICS SALIENCE
Figure 6.1 The structure of a pragmatic theory.
Status
129
CONTINENTAL EUROPEAN Perspective View 129 same example, aspects of social relations). Moreover, changes in context may cause shifts in basic properties of structural choices (e.g. in terms of markedness). The further exploration of this interesting context–structure nexus is one of the major challenges for the development of an inclusive or integrated theory of pragmatics. Another substantial challenge is the paradox of implicitness: as linguists, practitioners of an empirical field of investigation, we are not allowed to merely speculate about implicit meanings; rather, we must be able to identify linguistic traces of intended meanings or triggers of interpretation (both sides being equally important in the meaning generation process); but if we can point at traces or triggers, the associated meaning is no longer fully implicit. In other words, the interaction between or gradability of explicitness and implicitness must be an important concern. This means, indeed, that both form and function play a crucial role for understanding the pragmatics of language— an observation that was foundational for Östman’s theory. There are two ways in which Verschueren (1999) illustrates why the dichotomous labelling of an Anglo-American versus a Continental European pragmatics must be abandoned. First of all, stepping completely outside the dichotomy, there is no place where its formulations are better known and more widely accepted than in China. Chinese pragmaticians (who hold their own annual pragmatics conferences and who are extremely active in the field of publication) have introduced the adaptability theory of pragmatics as a concept alongside speech act theory, neo-Gricean pragmatics, relevance theory, and the like, thus turning into a ‘theory’ what was originally meant as a loose framework for discussion. Second, the framework is used in Verschueren (2011) to develop an empirical methodology for discourse-based ideology research. Thus this approach to pragmatics clearly links up with the social and political goals of traditions such as critical discourse analysis, which is itself straddling the geographical divide with both strong Anglo-based and European representatives (see Fairclough 2003; Wodak and Meyer 2009). Such goals, as hinted at before, were also clearly shared by Mey (1985). More recently, strongly engaged versions of pragmatics show up in a reflexive endeavour flagged as a concern with ‘responsibility and ethics’ (e.g. Östman and Solin 2012). In addition to looking at responsibility, accountability, and agency in discourse practices (especially narratives and institutional interactions), the purpose is also to address responsibility and ethics with respect to the positioning of the linguistic researcher, as insider or outsider, in relation to data gathering, data interpretation, and intervention. In a way, the questions asked follow almost directly and quite naturally from a much more general question, namely ‘What is pragmatics good for?’ This is the question that links the field of pragmatics with its own object and its own practices. Though attempts at theory formation in more restricted, strongly delineated areas are no doubt very useful, many of the recent developments in pragmatics lead almost naturally to a widening of the scope in the direction of what used to be regarded as a European tradition. Speech acts, for instance, have survived as topics at pragmatics conferences and in the pragmatics literature. But when this happens, they are usually taken out of the somewhat restrictive context of orthodox speech act theory, and what
130
130 Jef Verschueren we encounter is a highly specific interest in individual types of speech acts (apologies, compliments, complaints, etc.), often approached comparatively across linguistic communities, but almost always placed in an interactive context that takes into account institutional structures, social patterns, and cultural processes. Similarly, the notion of context, which has been present in pragmatics all along (remember the rather simple opposition between semantics as meaning without context versus pragmatics as meaning in context), has been considerably enriched over the years. The main change has been growing emphasis on contextualization, a continuous process, in opposition to a static context-out-there, aspects of which were in the past sometimes mistakenly isolated for purposes of explaining patterns of use. Potentially relevant context is endless, but a language user’s active orientation towards specific aspects of this borderless potentiality defines what is actually relevant context in a given situation. But in order to empirically assess what is being oriented to, a pragmatic researcher must be open to the entire range of potentially relevant aspects. All this implies that—except for philosophical and formal linguistic purposes—armchair pragmatics is almost banned these days. Technical developments have had a profound effect on the ways in which actual language use can be investigated. On the one hand, audio-and video-recording have enabled the detailed scrutiny of real-world spoken interaction. The use of such tools has become mandatory for any scholar who wants to make empirical claims about conversation. On the other hand, computer tools have facilitated the study of large-scale corpora, often consisting of written data, but also increasingly spoken. The complexity is further increased by the growing demands of taking multi-modality into account. Somewhat ironically, the new media, which naturally attract more and more attention, favour a return to the armchair: for current research purposes it is sometimes no longer felt necessary to go out to collect data.
6.6 By Way of Conclusion Today there is a geographically anchored dividing line which is much more important, and more real, than the opposition between an Anglo-American and a Continental European pragmatics. In fact it has always been there: the gap between a way of looking at language use that is strongly rooted in specifically Western philosophies and concepts, and the possibly quite different conceptualizations and ways of handling language that can be found elsewhere in the world. Pragmatics, like so many other fields of inquiry, has been heavily dominated by a narrowly defined Western world (mostly North America and Europe). Though scepticism in relation to the universality of theories and findings, often assumed in spite of their localized origins, was already visible thirty years ago (mostly in the work of anthropological linguists), it has been steadily spreading. In this context, an ‘emancipatory pragmatics’ movement (see Hanks, Ide, and Katagiri 2009) has emerged which focuses precisely on the cultural embeddedness of analytical concepts and which, by way of thought experiment,
131
CONTINENTAL EUROPEAN Perspective View 131 consciously applies specific non-Western notions of language use in theory building and empirical research. This endeavour aims at the development of perspectives and theories that are based on languages other than the ones whose language-, culture-, or community-internal notions have determined much of pragmatic thinking in the past. Some of the accepted beliefs about concepts such as turn-taking, politeness, deixis, speech acts, and the like, are thus reflexively questioned. It goes without saying that an effort of that kind requires a perspective as broad as the one embodied in what was traditionally associated with a so-called Continental European pragmatics. But the strong involvement of Japanese and American as well as European language scholars again shows the limitations of that nomenclature.
132
Chapter 7
The Sociol ogica l Fou n dations of Pr agm atics Jacob L. Mey
7.1 Introduction: Manipulation in the agora Pragmatics, as a science, may be among the youngest developments in the study of language and human behaviour. As a practice, however, it goes back several thousands of years. Thucydides the historian tells us how the Athenian statesman Pericles addressed the mourners for the heroes of the first Peloponnesian War against Sparta. In his famous address in the winter of 431 bc, Pericles was keen to point out the merits and greatness of those who had made the ultimate sacrifice in defending the homeland (Thucydides, Hist. II: 35–46). In addition, however (as is the case for all of the public speeches that are interspersed throughout Thucydides’ account of the wars—which he simply called ‘Histories’), it is altogether evident that the speeches were part of a sociological technique, intended to keep the Athenians in line, despite the deprivations of the Spartan blockade and the losses of their beloved. What Pericles did, in his speech, was to use certain speech acts such as ‘praising’, ‘remembering’, ‘eulogizing’, ‘comforting the bereaved’, ‘expressing compassion’, ‘inciting to heroism and perseverance’, ‘appealing to eternal glory’, and so on, in order to generate and sustain enthusiasm for the disastrous adventure that was to spell the end of Athenian hegemony, after the wars finally came to a conclusion in 403 bc. In the same spirit, Thucydides had elsewhere (Hist. I: 22) coined the famous, everlasting phrases ktêmá te es aeí, ‘and a possession [acquired] for eternity’, rather than ‘a listening specimen for immediate consumption’, agónisma es tò paraxrêma akoúein, with reference to the recorded glories of the heroes, both ancient and modern, and to the services they had rendered to their city-state.
133
The Sociological Foundations of Pragmatics 133 As to Pericles’ own acts of speech, they not only deliver a linguistic/rhetorical message, but stand out as early instances of societal manipulation with the aid of language. Chiefly, his speech was intended to move the audience and encourage them to greater and continued sacrifices; in this, Pericles’ purported intentions coincided with the historian Thucydides’ own motivation for embarking on his writing project. Thus, there is a clear correlation between the author’s professed interest in describing the happenings of the war (which appeared to be of unique interest, a kind of ‘historical first’, as Thucydides remarks in his introductory chapters) and the hidden agenda of the politician, who seeks to move the wheels of history in the direction of his own preferences, and takes a value-laden stance towards those same happenings. As we will see in the sequel, what happened on that cold December day of 431 bc in the Athenian agora1 is emblematic for much of what people, millennia later, came to focus on under the label of ‘societally relevant linguistic practice’, aka. pragmatics.
7.2 Drama in Innsbruck, 1974 I have referred to the work of Thucydides, not to lecture my readers on Greek history or literature, but in order to make a point that I deem of relevance for the question of what counts as societal pragmatics today, and how it relates to earlier sociologically and sociolinguistically based traditions. Let me start by recounting a personal experience. In 1974, I was lecturing as a guest professor at various Austrian universities. During my stay at Innsbruck, I gave a talk on the aims and methods of pragmatics, which was then still a relatively young branch of the language sciences, and in which I had become increasingly interested after moving back to Europe from the USA a few years earlier. I had come to realize that the true value of linguistic studies was not in the exact description of languages—however meritorious an effort in itself—but in the way linguistic knowledge could be put to use in a social context, among other things by situating the users of language, and the texts produced by them, within a wider societal framework. At one point during the first part of my lecture, a gentleman on the second row to the left jumped to his feet, and shouted (in German, which was the language of the lecture): ‘Aber das ist keine Linguistik, sondern angewandte Soziologie!’ (‘But that is not linguistics but applied sociology’)—upon which he stormed out of the auditorium, slamming shut the door behind him. As I recall it, the reaction of the audience was mild consternation and a certain measure of embarrassment at this rather unexpected and unacademic behaviour on the part of one of their faculty. Following the
1 The Greek term agora is most often translated as ‘market place’. In keeping with the Greeks’ original intentions and practices, it seems more appropriate to translate it as ‘place of gathering, commons’ (which also reflects the etymological connection with the Greek verb ageirein ‘to gather’).
134
134 Jacob L. Mey interruption, I picked up where I had left off, and afterwards had a lively and fruitful discussion with the remaining professors and their students in a nearby pub. I have often thought back to this incident as a defining moment in my career, a watershed, one could say. I became aware of what had been my problem with linguistics all along, even when I still was being taught the trade and its tricks. What was missing was the ‘applied’ element, or more precisely, the human factor involved in language and its use. In those early days, when I was a student of the celebrated Louis Hjelmslev (1899–1965) at Copenhagen, I was given to understand that linguistics should stay away from any socially oriented interests and problems, and concentrate on the task at hand: describing languages and writing grammars, in the tradition of the great Danish philologist Rasmus Rask (1787–1832). As it was once pithily expressed by my late Swedish linguist friend Bengt Loman (1923–1993), professor at Åbo University: ‘Linguists should write grammars, and do nothing else; that’s the way they earn their keep.’ Any explanatory trends in language studies were thus kept strictly within the confines of linguistics itself (defined as an ‘immanent science’ by Hjelmslev, in the Saussurean tradition). Any ‘outside’ explanations of linguistic behaviour, such as people’s motivation for using a particular expression, or the possible influence of developing users on the evolution of the spoken language, a question that had already fascinated the French comparativist Antoine Meillet (1866–1936) in the early years of the past century (‘every time a child learns to speak, innovations are being introduced’; Meillet 1937 [1922]: 19),2 had been outlawed by most contemporary philologists and were considered to be strictly outside the pale of linguistic descriptive endeavour. As the American Romanist Robert Hall, Jr (1911–1997) remarked some fifty years later, if one has obtained a complete description of a language, covering its phonology, morphology, and syntax (with some semantics thrown in for good measure), then to ask for further explanation is not only misguided but outright ‘childish’ (in Hall’s own words; see the Language article he published in 1978, reviewing Pizzorusso 1975). In other words, my Austrian colleague’s utterance (though perhaps not his ‘extralinguistic’ behaviour) was completely rational and understandable: I wasn’t preaching to the linguistic choir, but rather inserting a dissonant motif in the discourse of my colleagues and—who knows—perverting their young students’ tender minds.
7.3 A Tale of Two Brothers: Of Cooperation and Power The split between a pure theoretical way of doing science and a more ‘applied’, even humanistic, way is of course not restricted to the language studies. And this is where the two brothers come into the picture. I was raised in a family of economists, where 2
à chaque fois qu’un enfant apprend à parler, il s’introduit donc des innovations.
135
The Sociological Foundations of Pragmatics 135 the brothers Abraham and Jacob Louis (II) not only had gained their doctorates in economics more or less at the same time, in the mid-forties of the past century, but also each had secured a chair in economics at their respective universities within years of each other. Given this parallel development, one would perhaps assume that the brothers had some properties in common, or some subjects of mutual interest they could discuss. But nothing could be farther from the truth. According to student ear- witnesses, the brother-professors spent much of their teaching time trying to undo the other’s ways of practising science; they did this principally by making defamatory remarks about what the other brother was doing in the common area of business economics. There may be more at stake here than a family feud, fired and fueled by brotherly envy. The split between a pure descriptive science, such as business economics, and a science that aims for explanations, as does its sister counterpart, the ‘dismal science’ of social economy—the expression is due to the nineteenth-century philosopher- author Thomas Carlyle (1795–1881)—is iconic for much of what has happened in the humanities in general, and in the study of language in particular. Brother Jacob’s3 attitude towards his science reminds one of Robert Hall’s: just give us the facts, don’t ask for more. In studying economic phenomena, Jacob prefers to focus on the organizational and technical aspects of the production process; ‘human’ consider ations having to do with labour (like considering the workforce as more than just a cost factor) are foreign to him, as are the consumer aspects of economy in general. In remarks made in 1963, at a conference in honour of the ninetieth birthday of the Genevan emeritus professor and renowned socialist thinker, Edgard Milhaud (1873–1964), Jacob observed that he will only deal with the theme of the conference, ‘cooperation’, in terms of production; consumption and its related problems are purposely left out (Mey 1963: 235). In his reply, the person who had been the target of Jacob’s remarks, Gerhard Weisser (1898–1989), professor in the University of Bonn, observed that ‘purely economic attitudes simply don’t exist’, and that ‘extra- economic objectives’ have to be recognized as legal (1963: 303; Weisser’s contribution was called ‘Cooperative Planning’). In contrast to this, consider brother Abraham’s4 career path. Over time, in addition to his economic and artistic interests, Abraham developed a distinctly social consciousness (something for which brother Jacob saw fit to ridicule him, as on the occasion of Abraham’s joining the Oxford Movement in the late thirties). It irked Jacob that his brother (who by now had taken the proto-biblical name of Abram) had obtained a position at the more prestigious, established university of Amsterdam, while he himself had to work hard to get the newly founded Department of Economics 3
Jacob Louis Meij (II), 1900–1965, professor of economics at various Dutch universities (i.a. Groningen and Delft) and founder of the Graduate School of Business Economics at Rotterdam. 4 Abraham (later Abram) Mey, 1890–1983, was professor of economics at the University of Amsterdam. (NB: The varied spellings of the brothers’ last names are intentional, and were the source of yet another festering quarrel between the two.)
136
136 Jacob L. Mey at Groningen University off the ground, and later to create what was to become the Dutch Graduate School of Economics at Rotterdam. The brothers’ diverging activities were reflected also in the books they wrote. While Abram co-published a three-volume treatise on ‘Man and Society’ (Heymans and Mey 1946), Jacob at about the same time became the author of a successful textbook (Mey 1946), again illustrating the difference between his brother’s approach to their common science and his own. As to Abram’s work, this was carried out in cooperation with an expert on organizational theory and practice, Ernst Heymans (sometimes spelled Hymans or Hijmans), a person who, despite his technical-economic background and formation (he had a doctoral degree in engineering), was likewise interested in the problems surrounding the human use of technology and economics. Abram’s ‘unscientific’ approach to science was the subject of much scorn and ridicule from Jacob, who opined that any such interests only reflected his brother’s insufficient grasp of economic theory and its applications. Contrary to this, Abram wrote in his introduction that he had been taught by the events of the recent years (he was writing during the last months of WWII) that a purely economic approach to economic problems would not be able to solve the human problems involved in economic behaviour, in particular those involving cooperation (which also was to be the subject of Jacob’s somewhat bellicose intervention at the 1963 conference, while commenting on its general theme). Cooperation is a theme that is well known to pragmaticists, beginning with the work of the Oxford/Berkeley philosopher/linguist Paul Grice (1913–1988) in the sixties and seventies (Grice 1989a). Another main issue that is key in current pragmatic thinking (as opposed to purely descriptive approaches to linguistic issues) is that of power as a socially meaningful feature. As Abram Mey sagaciously (and much ahead of his time) opined in January 1945: ‘Power is a socially useful factor, but its function is apt to become societally damaging when it is maintained unchanged under changing social conditions and the evolving infrastructures of society’ (Heymans and Mey 1946: 145; my trans.). In this view, which is also the present author’s, the study of human activities, as it is practised in the social sciences (economics, linguistics, and all the others), needs to be grounded in the realities of human social life. Just as social power loses its meaning when it is not framed within the human condition, so too does a description of social phenomena that is not firmly anchored in our everyday life-world. In the following sections, I will reflect on the general implications of the two notions, cooperative behaviour and societal power, in a wider sociological and economic context.
7.4 Cooperation and its Discontents From the very beginning of the human race, cooperation has been a condition sine qua non for our existence. The weaver can only consume so much of his own product (and in addition, since linen isn’t edible, except to certain animals), he cannot survive directly off the output of his labour. Cooperation is necessary, so he finds somebody
137
The Sociological Foundations of Pragmatics 137 who will trade him some bread for a certain quantity of broadcloth (Karl Marx’s classic example from Capital, Bk. I, 1; Marx 1946). However, even this simple cooperation is not without its problems. For one thing, it is difficult to arrive at a fair evaluation of the commodities offered for exchange. Here, Marx (re)introduced a much refined version of the classic ‘labour theory of value’, which had been around for many centuries (embryonic forms of the theory have even been ascribed to Aristotle and St Thomas Aquinas in early formulations). In this perspective, the value of the thing produced is directly related to the human producer and the efforts he or she has expended in its production; but subsequently, the product is turned into a commodity, destined to be exchanged in the market place, where “the value of the thing/is the price it will bring” (Marx, Capital, Bk. I, 1, fn. 16). As I have observed elsewhere, this ‘commodification’ of human labour (Mey 1985: 321–322) is not without serious consequences, both for the producer and for the eventual consumer of the product. When ‘value’ is uniquely considered as ‘value-in-the- market-place’ (‘exchange value’, in Marx’s terminology), the users of the product tend to establish a direct relation between ‘value’ and ‘price’, by endowing the commodity with an instant, almost natural, evaluation procedure: ‘How much is it worth?’ (meaning ‘How much can it be bought or sold for?’). Such ‘commodified’ thinking subsequently extends to other spheres of the human activity as well; thus, ‘How much is s/he worth?’ is a common query, intended to establish a person’s financial and general reliability. In our thoughts, we tend to attach a numerical, ‘marketable’ quality to everything we are dealing with, including our social, educational, and personal relationships. We are asked to judge our successes or failures in terms of five-or ten-point scales, all the way from kindergarten to university, in questionnaires and applications, and everywhere we are liable to be compared with others, publicly or privately. The pragmatic impact of all this is considerable. If words express the ‘practical consciousness’ of a society (Marx and Engels 1974: 51), it should be clear that our thinking about the world in terms of saleable commodities, to be exchanged in some kind of marketplace, will affect our speaking, our use of words, our language. As my old comrade-in-arms, the Yale University emeritus professor of the philosophy of religion, Louis Dupré once wrote: ‘consciousness is not an independent factor; it is determined by language, and language arises out of social relations, which themselves depend on the material production’ (Dupré 1966: 155). Again, the dangers inherent in splitting between a quantitative, ‘description only’ approach and a holistic, qualitative attitude to human production (material or spiritual) become evident; I will come back to this aspect of a sociological concern with human activity in the next section.
7.5 The End(s) of Sociology A common sociological approach to describing human linguistic activity is to establish a relationship between social class as defined by, say, income and varying uses of languages or dialects. To do this properly, one has to undertake a thorough analysis
138
138 Jacob L. Mey of what actually transpires when humans interact; to this end, researchers like the late Japanese Takesi Sibata (1918–2007) or the American William Labov charted their subjects’ behaviour in the minutest of fashions, using all available recording and analysing devices and methods. Well known is Sibata’s ‘24-hour’ method, in which he observed language users in their natural habitat for lengthy stretches of time and subsequently analysed the film and audio recordings of the interaction to establish patterns of speech, choices of words, use of dialects and specialized terminology, and so on (Sibata 1951). Labov, using his famous ‘fourth floor’ method, elicited responses from workers in a New York department store in order to figure out to what degree and under what conditions they would ‘sound’ their [r]s in a word like ‘floor’ (pronounced [floah] vs [floar]; Labov 1972). Here, too, the idea was to have the observed be ‘unobserved’, so as to guarantee the authenticity of the observations, and to be able to deduce valid, objective conclusions about the social differentiation of certain pronunciations. In these and similar experiments, the emphasis is on an accurate and meticulous description of the phenomena; the analysis will go no further than pointing out the correlations (if any) that can be found between, say, a person’s social standing (low, middle, upper class, as measured by family income) and his or her ways of speaking. What is left out in this kind of approach is the perception of the speaking subject as a social being. Sociality here is a matter of registration, not of understanding or interpretation; especially the latter was considered taboo by many of the classic sociologists, against whom the German Max Weber (1864–1920) had to defend his ‘interpretive’ sociological method (on which below). How the subjects would handle a situation where their speech could possibly ‘betray’ them, or where their command of the spoken word turned out to be insufficient (due to the speakers’ low social stance and/or lacking education), and other related problems are not even touched upon in this purely descriptive sociology, or in a sociolinguistics that takes its cues from there. As in the case of social economics referred to earlier, it is not enough to describe the phenomena, one must also ‘save’ them (to borrow an expression dear to Greek and other philosophers since Plato); that is to say, the phenomena must be given an i nterpretation that makes them make sense in the context in which they occur—in our case, where speech is being exercised. My use of the term ‘interpretation’ undoubtedly will remind many readers of the work of Max Weber, one of the most insightful sociologists of the two centuries his lifetime bridged. Weber’s notion of ‘understanding’ (Verstehen) social phenomena rejects an appeal to intuition as a way of interpreting; on the other hand, Weber also shuns the automatism by which certain theoreticians concluded from societal categories to individual social phenomena such as language use. Fortunately for Weber, he did not live to see the extremes to which this latter line of thinking could lead, as in the case of early Soviet Marxism; he was lucky, too, not to have to witness the infamous thirties’ and forties’ Soviet debates on the ‘homespun’ Marxist theories preached by the linguist Nikolaj Ja. Marr (1865–1934), according to whom the only acceptable social ‘interpretation’ of language was as a superstructure of society—an error which no one
139
The Sociological Foundations of Pragmatics 139 less than Josef Stalin himself had to put down in his well-k nown diatribes on language and society, published in the Moscow newspaper Pravda a few years before the dictator’s death in 1953. For Weber, in order to understand a cultural artifact (e.g. language), one must be able to connect it—not just to the objective societal context of production, but also to the subjective context of the individual users: to their intentions, beliefs, purposes, and desires. Weber’s notion of ‘understanding’ (Verstehen, as the term is mostly quoted, in the original German) implies a ‘seeing things from the perspective of the person’ (Bainbridge 1998: 915), and thus contains the implicit answer to a question often raised in connection with debates on the societal status of language: ‘How to include the subjective nature of the humans’ condition in a consideration of the same humans’ objective positions in society, and their being limited, indeed to some degree determined, by the societal conditions of their life-world?’ Irrespective of how one answers this question, the influence of Weber’s thoughts has put the matter on the agenda and redefined sociology’s goals, specifying a new ‘end’ of sociology, while at the same time signalling the ‘end’ of classical, purely descriptive social science (Weber 1949). Years later, the eminent linguistic anthropologist Dell Hymes (1927–2009) argued similarly, and successfully, against the sterile descriptivism in linguistic matters preached by Noam Chomsky and his followers (Hymes 1984); Hymes underscored the need to include the social functions—the same functions that Weber originally had defined as essential for his view of understanding, and by the same token saving, the social phenomena.
7.6 Sociology and Sociolinguistics It is customary to partition the intersecting areas of social and linguistic theories according to whether they consider themselves as linguistic theories informed by a social point of view, or as theories of social phenomena, envisioned under a linguistic angle. In the first case, we usually refer to these theories as belonging to sociolinguistics, while in the second case, we talk about the sociology of language. I think that the distinction, while practically motivated as a division of labour, does not hold water, theoretically: the social and linguistic phenomena can be distinguished, but not partitioned; that is, they cannot be separated or defined, and certainly cannot be studied, in abstract isolation. All language presupposes a social formation both for its origin and for its use; conversely, human social formations cannot be imagined without the use of language (at least not as far as we are able to ascertain with the aid of history). As we have seen, the early sociologists of language concentrated on the descriptive aspects. In the words of the American sociologist Joshua Fishman, what is needed is ‘a reliable and insightful description of any existing patterns of social organization in language use and behavior toward language’ (Fishman 1972: 47; emphasis original); such patterns are displayed in attitudes and policies towards phenomena such as (stable or
140
140 Jacob L. Mey unstable) bilingualism (Fishman 1972: 52–53), and in debates on the need to govern language use through language policies of various sorts, including aid to be given to endangered varieties or languages (such as Fishman’s own Yiddish). Similarly, in an early article, the British linguist John R. Firth (1890–1960) stressed the need to study what he termed the ‘context of situation’ (Firth 1964b: 66)—a term that originally goes back to the Polish-British anthropologist Bronislaw Malinowski (1884–1942), and was to echo in the work of sociologists, sociolinguists, and pragmaticists throughout the decades to come; the British-born Australian Michael A. K. Halliday comes to mind as a prime instance (see Halliday 1989). Firth himself, however, did not follow up on this early conceptualization; still, his notion of ‘serial contextualization’ was a prelude to much of the later work in what Fishman came to call ‘the dynamic sociology’ of language (1972: 51). One researcher who devoted his entire life to creating a synthesis of the two aspects mentioned above was the Frenchman Pierre Bourdieu (1930–2002). Starting out from his personal experiences in Algeria during the independence wars, he gradually embraced a comprehensive view of human practice. According to Bourdieu, one should be careful not to take human activity as a deterministic reaction of i ndividuals to pre-established conditions and emerging stimuli: ‘it is necessary to abandon all theories which explicitly or implicitly treat practice as a mechanical reaction, directly determined by the antecedent conditions and entirely reducible to the mechanical functioning of preestablished assemblies [or] models’ (1977: 73). Individual activity does not, by itself, lead to societal organization; the fact that people act in some kind of collectivity does not automatically index the presence of interaction. In order to coordinate the activities involved in social practice, humans have to communicate; the development of language is related to this practice, in particular the tool-making and tool-using processes that are specific for human activity (Mey 1985: ch. 3.3). What is needed is communicative interaction: individuals acting with (or against) one another and communicating against the backdrop of the ‘equalities and inequalities’ that are the primordial parameters of any society, as stated by the French philosopher Jacques Rancière (1995: 19). Again, we see how the purely descriptivist model of studying human activity is superseded by an interpretive way of looking at the ongoing action. But how does one get from action to societally oriented interaction? For interaction to happen, and to be intelligible both to the interactants and the observers, one has to suppose something more than just activity. Conversely, ‘the truth of the interaction is never entirely contained in the interaction’ (Bourdieu 1977: 81), but points to the conditions of society that vouchsafe and sanction the ongoing action, which always occurs in a climate of opposing tendencies (cf. Rancière’s ‘equalities and inequalities’). When these oppositions are resolved in human interaction, a ‘common- sensical’ system of values, ‘a commonsense world’ is established. This system is ‘taken for granted’ by all, ‘endowed with the objectivity secured by consensus on the meaning of practices and the world’ (Bourdieu 1977: 80).
141
The Sociological Foundations of Pragmatics 141 Bourdieu is here up against the same dilemma that had plagued Weber: how to establish an objective way of interpreting social phenomena that are the ‘property’ of individuals? For Bourdieu, objectivity is not the objective quality of experiments in the natural sciences, which may be repeated ad infinitum with the same invariable results. Human ‘experiments’ are founded in human experience, and as a result, they can never be the same. Because humans learn from experience, their past experiences colour that which is experienced at a later date, and thus never replicate, but either reinforce or diminish the experience of the original experiment. The objectivity that Bourdieu talks about is located in the ‘objective intentions’ of the interactive process, and should not be confused with the subjective intentions of the interactants; what he calls the habitus is conceived of as a ‘picking up’ of the objective intentions without necessarily reactivating the lived intention of the individual human agents (1977: 80). Neither should the habitus-forming processes be linked exclusively to the ontogeny of the subjects, although it is in a person’s upbringing that the habitus is formed; the habitus transcends the individual’s personal history as an unwritten, inborn law, without which any upbringing would be inefficient and indeed gratuitous. Upbringing, in Bourdieu’s view, is nothing but the ‘internalization of … objective structures as dispositions’ (ibid.) which, because they are not bound to a particular place or time or individual, are called ‘transposable’ (‘portable’, as one would say, in the parlance of the computer, of a programme that is not bound to any particular configuration or machine). As a principle governing societal interaction, the habitus is dialectically placed between the objective conditions encountered as ‘nature’ or ‘world’, and the subjective categories through which we interpret the world. The human activity aims at overcoming contrary societal tendencies such as: fact vs representation (or ‘view’), personal preference vs the common good, equality vs inequality, immigrants as threatening aliens vs immigrants as indispensable workforce, and so on and so forth. The point to focus on here is that these oppositions are neither objective (in the sense that one can ‘prove’ them experimentally, as we discussed earlier), nor are they created purely in the mind of the beholder, as it is often argued in today’s public debates when it comes to discussing problems of integration and assimilation with reference to the immigrant population. By stressing the role of activity and interaction in the production and reproduction of society, Bourdieu has laid the groundwork for an objective evaluation of the societal formation, whose ‘sexual division of labour, domestic morality, cares, strife, tastes, etc. produces the structures of the habitus which become in turn the basis of perception and appreciation of all subsequent experiences’ (1977: 78). Consequently, the habitus is needed to guarantee an effective practice of communication through language, using the linguistic structures that are given us by our nature and culture. In Bourdieu’s pithy formulation, ‘[habitus is] structured structures turning into structuring structures’ (1977: 73), ‘history turned into nature’ (ibid.: 78), our natura secunda, to borrow a term originally due to Aristotle and St Augustine. In the next section, I will show how this view reflects itself in contemporary studies of human linguistic behaviour.
142
142 Jacob L. Mey
7.7 Ethnography and Context In contrast to ‘armchair’ sociologists such as Émile Durkheim (1858–1917) and Weber, Bourdieu was thrust into the middle of events through his forced presence at the upheavals surrounding Algeria’s transition in the 1950s from French colony to an independent Islamic state. His burgeoning ethnographic interest was later solidified and anchored in the sociological traditions still lingering in Paris after the great Durkheim’s much too early demise—traditions which Bourdieu vigorously and successfully set out to revitalize. At about the same time, as we have seen earlier, the American linguistic anthropologist Hymes started publishing his seminal works on the ‘ethnography of speaking’ (1962), later expanded by himself and his collaborator, the late John Gumperz, into the ‘ethnography of communication’ (Gumperz and Hymes 1964, 1986 [1972]). It was especially John J. Gumperz (1922–2012) who built a bridge from ethnographic observation to sociological interpretation to sociolinguistic explanation of the phenomena in the borderland of language, culture, ‘habitus’, and society. And this is also how the label ‘speech’ came to be replaced by that of ‘communication’ in Gumperz’s (as well as in Hymes’s later) works. Central to Gumperz’s thinking is the notion of contextualization, where ‘context’ is not just seen as some textual material surrounding a given word or phrase, but rather comprises the entire gamut of possible features that influence the ‘speaking’ (including the use of gestures, gaze, mouth and other body movements, and so on). Contrary to the methods of certain sociologists and sociolinguists who construed a possible sentence and then inquired about the ways it could be spoken, or be given meaning, Gumperz, in the descriptive tradition of ethnography, set out to observe what was going on in real interaction, and only then took up residence in the famous armchair to distill and concentrate his findings in written form, as in the celebrated 1972 article, co-authored with the Norwegian sociologist Jan-Petter Blom, on the use of socially determined linguistic forms in the Northern Norwegian dialect of the village of Hemnes (Gumperz and Blom 1986 [1972]). Even though the conclusions drawn in this early article have been partially impugned since then (see Mælum 1996), the fact remains that we are dealing with a landmark in the history of sociolinguistics (and indirectly, pragmatics). For the first time, one was witnessing an ad oculos demonstration of the influence of social factors on language choice—as an objectively established phenomenon which, in the Bourdieuian tradition, could even be thought of as ‘automatic’ (Bourdieu 1991). But speech is not the only (or even sometimes the major) component of the communicative situation. As Gumperz observes, ‘communicating is not just a question of individuals translating their ideas into lexically and grammatically meaningful utterances’ (Gumperz 1996: 379). Rather, the communicative agents must learn to interpret their utterances in accordance with the context in which they are speaking. This interpretive contextualization depends on ‘online assessments’ made by the
143
The Sociological Foundations of Pragmatics 143 speakers, trying to adjust their expectations to what they have heard so far, and to what they might anticipate as a result of that hearing. Here, Gumperz says, ‘we are always faced with an array of potentially situated interpretations such that the significance of what happens at any one point can only be understood in relation to what precedes and what follows’ (Gumperz 1996: 375). To choose intelligently from an array of situations, we need what Gumperz calls ‘contextualization cues’, verbal and non-verbal signs that tell us how to interpret the utterances we are confronted with; such cues are indispensable elements in our communicative interaction (Gumperz 1992a). Among the contextualization cues that Gumperz discusses are mainly those of the prosodic type (Gumperz 1996: 366): for instance, they can be loudness or pitch, when used to mark off a segment of speech rhetorically, or emphasize its content. One may also think of the rules for taking turns, as they have been described in conversation analysis (on which see section 7.8); other cues may be so-called ‘parentheticals’ or discourse markers of the type ‘you know’. Here, again, there are huge differences from culture to culture, language to language; one need only think of the ordered sequences of turns in British or American conversation, as compared to the ‘polyphonic’ (or should we say caco phonic?) mêlées that are common in other cultures. The US linguistic anthropologist John B. Haviland has provided us with some delightful examples from Zinacantan, a Mayan language spoken in the highlands of Chiapas, Mexico (Haviland 1997: 548), where speakers claim their turns simultaneously; at the end of several bouts of such a conversational ‘tug-of-war’ one speaker emerges at the top of the roost (see Mey 2001: 270–271). In order to understand the mechanisms governing such ‘unruly’ behaviour, one should expand the notion of context of situation, referred to earlier, to also comprise the ‘unknown unknowns’, the invisible invisibles present in every human encounter. Erving Goffman (1922–1982), the American sociologist, tried to do this by replacing ‘situation’ with ‘setting’, a concept that incorporates those invisible elements that steer and determine the conduct of the interactants, even without their knowing it; their ‘tacit’ knowledge of those features is incorporated in their behaviour, and does not always lead to what Goffman called ‘writable statements’ (1972: 65)—in fact, many of the properties defining or changing a particular setting are manifested as ‘extralinguistic acts’ (such as changing one’s body posture or voice intonation; cf. Goffman 1972: 66). Among the invisibles that are most often overlooked in interaction are those that can be attributed to what I have called the context’s ‘unseen, but by no means sleeping partner’, viz., the ‘invisible hand’ of society (Mey 1985: 336). Using a current (perhaps more adequate) formulation, one could speak of the interaction’s contextual embedding in a particular ‘social field’ (a term originally due to Bourdieu). As the contemporary American anthropologist and linguist William F. Hanks remarks, ‘to study context is to study embedding; … context is embedding relations’ (2006a: 120–121; emphasis original). Context is not to be regarded as an ‘add-on’ feature; rather, it is there all the time, but as a truly invisible embedding. (See further, section 7.10.)
144
144 Jacob L. Mey As to the social field itself, it determines the scope and relevance of what our interaction can be about, and how we structure and realize it; compare such social fields as the hospital, the university, a professional office (a doctor’s, a lawyer’s, etc.), the courtroom, the religious gathering (such as a Quaker Friends’ meeting), and so on. There, embedding is at work socially, but it is not recognized as such until a breach in the interaction occurs and sanctions are imposed, whose seriousness depends on the nature of the contravention and on the (il)legitimacy of the participant(s) in question (whether they are ‘ratified’ or not, to use Goffman’s terms). To put it as a slogan (pace Hamlet): there is more to context than is thought of in your context.
7.8 Embedding in Action: Conversation Analysis It has often been remarked (e.g. by Levinson in his 1983 magisterial treatise Pragmatics) that conversation is the matrix of human linguistic endeavour, ‘the prototypical kind of language usage’ (1983: 284). To fathom the importance of that remark, one has to situate the beginnings of conversation analysis (CA) where they historically belong: in the late 1960s and early 1970s. Those were the heady times of Chomsky’s super-structuralism, when language was thought of as a system of rules, installed in the native speaker and manifested through the ‘performance’ of his linguistic ‘competence’. Little thought was given to the speaker (or hearer) himself: he (never a she!) was merely an idealized, mostly male, embodiment of the language’s structural system. Questions of context were thought to disturb the process of analysing the ‘language’ (which itself was a pale reflection of how people really spoke in their daily encounters, aka. conversation). Having its roots in sociology, CA has always been true to its sociological origins (cf. Hutchby and Wooffitt 1997: ch. 1), even though in many ways, and in the work of many researchers, it has come close to being identified with either applied linguistics, pragmatics, or even communication theory. The truth of the matter is that conversation analysts, in the wake of the groundbreaking work done by Harvey Sacks (1935–1975) and his collaborators, are working with real linguistic data, in contrast to (and often conflicting with) the armchair theories of the official linguists and the sanitized pieces of linguistic material that were their mainstay. Sacks had been inspired by the work of the American sociologist Harold Garfinkel (1917–2011), the founder of the ‘ethnomethodological method’ in social sciences (Garfinkel 1967; Heritage 1984), by which social phenomena are supposed to be explained based on the actions, decisions, and viewpoints of the ‘members’ rather than by ‘grand theories’ and universal social categories. Sacks’ own work on telephone conversations brought him to the idea that humans organize their social world through ‘talk’, that is, by establishing the speech categorizations they need on an ad hoc but motivated basis. Applied to ordinary conversation, Sacks found that ordinary talk never was just that, but served to delimit and categorize the social world in which we live (Sacks 1995).
145
The Sociological Foundations of Pragmatics 145 Conversation analysis is best known for its meticulous rendering of ‘talk- in- interaction’, by making use of a sophisticated transcription technique, chiefly due to Gail Jefferson (1938–2008), one of Sacks’ earliest co-workers, and for laying down the ground rules for an ordered conversation, which always takes place in well-defined turns. As a corollary, the value of a conversational contribution cannot be determined on the basis of the given ‘input’ alone: what is important is the outcome, the way the turn is taken up by the co-conversationalists. An important element in this thinking is the concept of adjacency, by which a turn’s first part ‘expects’ a suitable second part, or if such a part is not available, immediately will construe the available half-turn as being suitable. This great flexibility in interpreting human linguistic behaviour is a feature that ethnomethodologists and conversation analysts share with a pragmatic approach to interaction. Still, it would be wrong to overlook the conversation analysts’ theoretical disdain for the influence that the ‘organized segment’ of society in which interaction takes place, exerts on the interaction, in favour of the members’ own ‘construction of reality’—a term originally due to the German-US sociologist Alfred Schütz (1899–1959), in works such as The Meaningful Construction of the Social World (1932/1967). Typically, recent trends in CA tend to adopt a more pragmatically oriented view by emphasizing the institutionalization of conversation in the form of ‘talk-at-work’ (Drew and Heritage 1992), where both the interaction and the institutional setting are respected as dialectical partners in such a construction.
7.9 Discourse Analysis Early on, the Chomskyan postulate that human language use is best described in terms of well-formed sentences had been discredited both by sociologists and by linguists working in the sociolinguistic tradition. Similarly, the notion of a ‘grammar’ as the unique criterion for people’s appropriate use of language fell into disrepute in the wake of the critique that was voiced against the strict codification of linguistic structure due to Chomsky and his followers. The Russian sociologist and semiotician Mikhail Bakhtin (1895–1975) argued that the proper unit of study for linguists and semioticians had to be the situated utterance, not the sentence: utterances express a language user’s situatedness in the world, whereas sentences are mere abstract and incomplete representations of the same. Moreover, utterances are unique, while sentences may be repeated ad infinitum, as in the exemplifications of many linguists of the ‘armchair’ kind. ‘Sentences are repeatable. Sentences are repeatable’, as Bakhtin himself illustrates his point (1994: 108). By contrast, utterances are not repeatable: they always depend for their use and meaning on who says them and in what circumstances; and this is where discourse comes into the picture (Mey 2001: 199). As pointed out by the Austrian sociolinguist and pragmaticist Martin Reisigl in a recent survey article (Reisigl 2011), the term ‘discourse analysis’ itself can be ‘traced back to the American structuralist and distributionalist Zellig Harris’ (1909–1992).
146
146 Jacob L. Mey In his influential work Methods in Structural Linguistics (Harris 1951), followed by an eponymous article in Language (Harris 1952), Harris distinguished between discourse analysis as dealing with linguistic description beyond the limits of a single sentence at a time, and discourse analysis as dealing with the relation between ‘non-linguistic and linguistic behavior’, or ‘between “culture” and “language”’ (Harris 1951: 1; Reisigl 2011: 11). As Reisigl further observes, Harris stuck to the first part of the alternative; as a result, discourse analysis was for a long time confined to the domain of ‘text linguistics’, whereas the second part did not gain proper traction until the 1970s and 1980s, when people like Bourdieu and Foucault managed to turn the discussion around. The term ‘discourse’ gained in popularity as the emphasis shifted from the individual utterances to the (con)text in which the utterances were situated. Any analysis of discourse in this sense had to take into account that entire context, ‘the ensemble of phenomena in and through which social production of meaning takes place, an ensemble which constitutes society as such’ (Mumby and Stohl 1991: 315). Discourse is therefore not just a concatenation of sentences beyond the level of the individual sentence, a ‘macro-sentence’, to be analysed using the well-k nown methods of linguistics, where the text is seen as a structured complex of interdependent units, much like the sentence is built up as a connected whole of noun phrases, verb phrases, adverbs, prepositions, and so on (a conception defended by early text grammarians such as van Dijk 1977). Neither can discourse be conceived of as a kind of ‘macro-speech act’, or a sequence of such entities, where the governing principle is the intention of the speaker as expressed in speech acts in the Austin–Searle tradition (Austin 1962a; Searle 1969). If discourse, in the conceptual framework due to Michel Foucault (1926–1984), is the practice of making sense of signs (linguistic or otherwise), discourse is concerned with the active creation of meaning: it represents ‘practices that systematically form the objects of which they speak’ (Foucault 1972: 44). The practice of discourse is the practice of society; in discourse, we ‘word’ societal matters, and conversely society ‘words’ itself in our common, social discourse. We saw how this view of discourse articulated itself in the very discourses that I referred to at the beginning of this chapter, viz., the type of public speeches that the Greek historian Thucydides puts in the mouths of his protagonists: their discourses were the voices of society, pronounced and diffused by its emblematic orators (in our case, Pericles). Consider also that the French term discours is eminently used for exactly this type of linguistic activity (see Roman 2006 on the typical Thucydidean discours).
7.10 Critical Pragmatics and Critical Discourse Analysis Looking at the broader, societal context of discourse, we see that an analysis of discourse cannot be undertaken in the narrow context of the individual speakers/hearers.
147
The Sociological Foundations of Pragmatics 147 What is needed is an understanding not only of the motives that lie behind the single speaker’s or hearer’s activities, but first of all of the societal conditions that motivate, and make possible, the particular stretch of language that we are looking at in our analysis (or hearing in conversation, as the case may be). This is why discourse analysis, in the sense outlined here, naturally blends into what is often called ‘critical’ linguistics, or more aptly in the present connection, ‘critical discourse analysis’. The way this ‘critique’ is exercised has much to do with what Kant already in the eighteenth century established as the right way to investigate our conditions of knowing and forming judgment: one has to look behind the phenomena and try to discover what allows us to make a statement, pronounce a judgment, express a view, and so on. Such ‘phenomenal’ activities have to be related to the categories that we both establish and are bound by; and we establish this relation by applying a critical perspective (as also Foucault was to teach us two centuries later). The term ‘critical’ indicates a reflexive, examining stance towards the phenomena of everyday life. Building on the neo-Kantian tradition, the concept was reintroduced and expanded in the intellectual ambiance of the ‘Frankfurt School’ of sociology, which included not just sociologists like Max Horkheimer, Theodor Adorno, and Walter Benjamin, but also philosophers like Jürgen Habermas. These thinkers, despite their different starting points and perspectives (some were psychoanalytically oriented, others Marxian-inspired), all had this in common that they considered the world and its phenomena (including human activities) from a point of view that I will call a ‘considered subjectivity’: the facts as such are not interesting unless they are ‘filtered’ through a subjective, participant-bound lens of interest. For Habermas, this interest was grounded in the (academically inspired) notion of ‘communicative action’, and based on the concept of an ‘implicitly shared and immanent rationality’, common to all humans (Habermas 1981). This rationality is the basis of the civic ‘public sphere’ (Habermas 1989: 27), in which egalitarian discussions take place in a democratic milieu that is participatory (where discussion is encouraged and decisions are consensus-based), rather than representative (where we leave the discussions and decisions to our elected parliamentarians). Habermas’s influence was, and still is, significant, especially since he, in later years, has reached out from what some have called the ‘ivory tower’ of Frankfurt, by engaging in debates with people as different as Jacques Derrida and (since emeritus Pope Benedict XVI) Joseph Cardinal Ratzinger. By contrast, his direct influence on the development of pragmatic thinking has not been equally significant. Habermas has always been more oriented towards the philosophical aspects of the questions under debate (even though he was one of the first among European intellectuals to publicly denounce George W. Bush’s ‘war on terror’ and the invasion of Iraq). With regard to critical pragmatics, and critical discourse analysis in particular, an early beginning was the publication of my essay ‘Toward a critical theory of language’ (in German; Mey 1979). Here, a theoretical Marxian-oriented approach to a variety of social problems, in particular with relation to language as an educational medium, was tentatively set forth (to be followed some years later by a full-volume treatment
148
148 Jacob L. Mey of these issues; Mey 1985). At about the same time, the late Roger Fowler (1939–1999) and his colleagues at the University of East Anglia (among these Robert Hodge and Gunther Kress) launched their own project of a ‘critical linguistics’, in which they likewise incorporated a Marxian approach to sociological and pragmatic issues, among other things with the intention of demonstrating the oppressive role of the socially powerful in establishing the ‘ruling language’, and in order to emancipate people from such oppression (Kress and Hodge 1993). In a further development, critical linguistics and its successors, critical discourse analysis and critical pragmatics, examine the societally given conditions of use that determine, by affording as well as restricting, the opportunities that are available to the individual user. We never use language in a vacuum, but always in a social situation; this premier sociological observation turns out to be of the utmost importance for a correct reception and evaluation of what is uttered, by situating it as the product of a live user, rather than as an abstract statement proffered in an academic environment (as I noted earlier when comparing the two economist-brothers, Abraham and Jacob, and their diverging views on how to deal scientifically with economic activities). In important works by Teun van Dijk (e.g. 2008) and in contributions by his associates, published in the influential journal Discourse and Society, as well as in recent publications by Norman Fairclough (e.g. 1989, 1995) and his followers of the ‘Lancaster School’, these questions are intimately bound up with the (partially Habermas- inspired) debates on ‘power’ in society: those who are in power have the right to determine what is appropriate discourse by defining the societal contexts in which words may be properly used. In a thoughtful study of a different setting, viz., the sixteenth-century Spanish conquista of Mexico and its profound effects on the indigenous Mesoamerican communities of the Maya, the earlier mentioned Berkeley linguistic anthropologist William Hanks has shown how the diffusion and repartition of power between the conquerors and the conquered has led to an obligatory reanalysis of both the Spanish and Mayan societal discourses (Hanks 2010). The process of converting the Indians to the Catholic faith and their concomitant or subsequent subjugation to Spain was made possible via the discursive interpretation of powerful ‘converting words’, as Hanks calls it. The discourse of the Franciscan missionaries circumvented and penetrated the discourse of the Maya, to the degree that not only were the Maya Christianized, but their very language was, uncritically, ‘converted’ (not to say subverted) to a ‘Maya reducido’, a r educed Maya—‘reduced’ not only quantitatively speaking, but also taken in the qualitative, sociological sense of ‘belonging to the reducciones’, the settlements (‘reductions’) where the natives were brought together under the aegis of the Church, but otherwise kept to their own Christianized, yet still partly authentic, pre-Columbian conditions of life. As Hanks expresses it, ‘the new lengua reducida made its way to the deepest corners of the Indian communities—even, I will argue, those that were unsubjugated…. Over time, and across the main genres of colonial discourse, the doctrinal roots of Maya reducido would contribute to a process of semantic and grammatical reanalysis of Maya. It is this process that I refer to as the linguistic conversion … the
149
The Sociological Foundations of Pragmatics 149 joining of two languages already turned toward one another, adapted to the task of producing the semantic universe of conquista pacífica’ (Hanks 2010: 16). And, it needs to be added, this subjugation was not just semantic, but exercised in living discourse, as it informed everyday usage and practice. Unfortunately, neither the Maya nor the Spanish at the time were in the possession of the tools necessary to deconstruct this conversion—which in the end proved to be much more than just semantic.
7.11 Towards an Emancipatory Pragmatics When we speak of the sociological ‘bases’ of pragmatics, it behooves us to consider its sociological ‘ends’ as well. The connecting link in the chain is the social subject par excellence, the human user of society’s communicative resources, first of all of language. In recent times, while linguists were busy defining language ‘immanently’ from the descriptive point of view, trying to model and formalize its grammatical structures, others (among these the philosophers, sociologists, sociolinguists, and linguistic anthropologists that we discussed earlier) were interested in what one could use those resources for in a societal context. In Oxford, the philosopher John L. Austin (1911–1960) hoisted the pragmatic banner and, in his famous posthumous treatise How to Do Things with Words (1962a), raised the perennial Shakespearean question of how we may effect world changes by our use of ‘words, words, words’ and ‘pregnant replies’ (Hamlet Act II, Scene ii). At the same time, sociologists and sociolinguists, in the wake of people like Erving Goffman, were broaching the issue of what settings were allowed for a particular use or, more generally, which constraints govern the human use of language. Ethnolinguists and linguistic anthropologists responded by assigning functions to utterances ‘on the basis of the social situations that the task is conducted within’, as the influential British anthropologist and linguist Stephen C. Levinson had described it (Levinson 1983: 279); see also his earlier 1979 seminal work on ‘activity types’, now reprinted in Drew and Heritage (1992). Pragmaticists discussed in particular what role language should be understood as playing in human intercourse; Austin’s theory of speech acts, later expanded to comprise indirect speech acts and pragmatic acts, incorporated these findings in the ‘new’ science of pragmatics (Mey 2001: ch. 8; 2010). In all these discussions, one main theme that keeps cropping up is that of power. Who has the right and the power to define such a thing as a ‘correct’ or, as Fairclough (1992) calls it, ‘appropriate’ use of language? Clearly, one (often the only) way to access society’s goods is by using language, by ‘doing things with words’, using speech acts such as offering or asking for help, requesting permission or declining a request, offering and returning a favour or a greeting, identifying wrongdoing and evildoers while defending oneself in the case of real or perceived impingements on one’s territory,
150
150 Jacob L. Mey establishing social compacts as in marriage or business—a ll of these acts are subject to the ‘power of words’, a power that is not inherent in the words themselves by some kind of poorly understood magic, but can be referred to the respective societal positions of the speaker and the audience. Using language can furthermore be a means of defining oneself societally; in Japanese society, the appropriate polite use of language is one of the distinctive features of class; for instance, the wrong use of honorifics would place the speaker in an undesired category of people who do not use what the Japanese sociolinguist Sachiko Ide has called ‘discernment’ (Ide 1989). In our own society, that is, in Western cultures, the use of dialects and the presence of (foreign or domestic) accents are often decisive for one’s access to society’s goods. When it comes to command of the spoken and especially the written word, we distinguish sharply between qualified, ‘literate’ users and those who are deemed ‘illiterate’, either due to lack of access to, or by forceful exclusion from, the ‘hegemonic culture’ (the expression is the Italian sociologist and linguist Antonio Gramsci’s (1891–1937; Gramsci 1928). Historically, many cultures have denied the lower classes access to literacy, lest they become uppity; and many religions teach the poor to remain happily in their lowly societal positions (‘O let us love our occupations/Bless the squire and his relations/Live upon our daily rations/And always know our proper stations’, as a popular nineteenth-century English ditty had it; Dickens 1844: ‘Second Quarter’). Even the Jesuit order had a ban on developing literacy and bookish learning when it came to its coadiutores temporales (the lay brothers, literally called ‘wordly helpers’, as opposed to the majority of the Order’s priests, who were called ‘spiritual helpers’); the idea was that the lay brothers would be content with performing the menial functions associated with the lot of the biblical Martha, who was busy making dinner, while her sister Mary indulged in higher things, spiritually communing with the Lord (Luke 10:38). The flip side of this somewhat suspect coin is of course that by denying people the right to better themselves by using language, we also thereby condemn them to perpetual economic and social slavery. In contrast to this, by critically examining the possible uses of language and what makes such uses effective for, and accessible to, people, we may be able to remove some of the barriers that are responsible for the division of our world into its two main strata of well-educated, well-trained, well-paid, and well-fed denizens vs those who just perform their daily chores without much hope for personal betterment other than the infamous ‘pie in the sky’, as promised to the slaves of yore. However, speech acts, taken by themselves, will not do the job, just as a ‘good’ accent will not land you employment if you cannot document the educational basis of that accent and the cultural stance that goes with it (pronouncing English beautifully didn’t help Eliza Doolittle much at the Ascot races, either, as George Bernard Shaw has amply demonstrated; 1914). But how to realize this emancipatory aim of pragmatics? By way of illustration, let me take an example from a sphere of activities that has become rather controversial lately. To qualify for citizenship by naturalization, certain countries have started instituting a language requirement that places the burden of proof on the would-be immigrant: he or she must convince the authorities that s/he
151
The Sociological Foundations of Pragmatics 151 is a bona fide prospective citizen, and to confirm this good faith, the immigrant must document linguistic abilities of a certain standard. The problem is that such abilities can only be achieved through some kind of schooling—something that most immigrants, due to their limited social stance and economic status, are not able to obtain, or had been unable to obtain even in their native countries. As a result, many of these potentially capable workers spend their lives in low-qualification jobs, even though their professional capabilities would allow them to rise on the social ladder, if only they had the necessary wherewithal in the form of what Bourdieu has called ‘cultural capital’ (Bourdieu 1991), in this case, command of the language. To sum up and conclude: what a pragmatic approach could contribute in cases like these is to ‘deconstruct’ the societal power structure that prohibits access to education, e.g. by establishing free training centres for immigrants, where language is taught in ways that are commensurate with the immigrants’ needs. That is to say, rather than force-feed the students anecdotes and tidbits about life in middle-class suburbia, we should take care to instruct workers in safety routines; employees should be given access to the tools for negotiating wages and working conditions; access to higher education should be furthered by offering specific training courses for particular professions that can serve as entry steps to the world of better employment; and so on (see Mey 1985: ch. 2.2, ‘Immigrant language education’, for more detailed suggestions). Immigrant workers should be taught the art of ‘native (or native-like) speech acting’ in situations where it matters most for them (encounters with authorities such as the judiciary, the police, the township administration, hospital and health workers, the school authorities, and so on); in these courses, emphasis should be placed on the ‘awareness raising’ through language teaching and linguistic training that is a necessary (albeit, unfortunately, not a sufficient) condition for a conscious and effective use of one’s linguistic repertoire. In the final analysis, the aim of our societal concern with language, its end, should be to bring about the end of linguistic and other bondage, by promoting the emancipation of the language user.
152
153
Pa rt I I
C E N T R A L TOP IC S
154
155
Chapter 8
Implicat u re Yan Huang
The concept of implicature (both conversational and conventional) has its origin in the work of the late English philosopher H. P. Grice, though some proto-Gricean ideas can be traced back at least to the first-century bc rhetorician Dionysius and the fourth- century rhetoricians Servius and Donatus. These ideas were later reiterated by the nineteenth-century English philosophers John Stuart Mill and Augustus De Morgan. Much more recently in the 1950s, Grice was also influenced by similar notions put forward by his colleagues within the tradition of ordinary language philosophy at Oxford University (e.g. Horn 2006a, 2012a, b). Since its inception, Grice’s classical concept of conversational implicature has revolutionized pragmatic theorizing, nourishing and producing a number of new ideas, such as explicature, the ‘pragmatically enriched said’, and impliciture in various neo-and post-Gricean enterprises. In this chapter, I shall examine implicature. The chapter is divided into two parts. Part I is concerned with conversational implicature. Section 8.1.1 defines conversational implicature. The essential properties of conversational implicature are discussed in section 8.1.2. Section 8.1.3 surveys two classical and neo-Gricean dichotomies of conversational implicature: conversational implicatureO versus conversational implicatureF and generalized versus particularized conversational implicature. In section 8.1.4, I take a look at two neo-Gricean typologies of conversational implicature. Section 8.1.5 presents an overview of some current debates on generalized conversational implicature, in particular Q-scalar implicature. Next, embedded implicature is the topic of section 8.1.6. Finally, this part ends with a discussion of whether pragmatic intrusion into what is said is explicature, part of the pragmatically enriched said, impliciture, or conversational implicature (section 8.1.7). Part II covers conventional implicature. Section 8.2.1 is concerned with the question of what is conventional implicature. Next, in section 8.2.2, I discuss the essential properties of conventional implicature, comparing and contrasting them with those of conversational implicature. Finally, section 8.2.3 surveys the latest developments in the study of conventional implicature.
156
156 Yan Huang
8.1 Conversational Implicature The notion of conversational implicature is one of the single most important ideas of pragmatics. Stephen Levinson (1983: 97)
8.1.1 What is conversational implicature? Conversational implicature is definable as any meaning or proposition expressed implicitly by a speaker in his or her utterance of a sentence which is meant without being part of what is said in the strict sense. Derived via Grice’s (1975, 1989a) cooperative principle and its attendant maxims of conversation, it provides an augmented meaning, and sometimes in ‘flouts’ cases alters the meaning, of what is said (e.g. Huang 2014: 31). For example, when a speaker utters the sentence in (1a), he or she (ceteris paribus) conversationally implicates (1b).1 (I use +> to stand for ‘conversationally implicate’.) (1) a. The soup is warm. b. +> The soup is not hot Construed thus, a conversational implicature is what is communicated/conveyed/ meant minus what is said. It is beyond and above and sometimes divergent from what is said. It represents a relation between a speaker and a proposition produced by that speaker on the basis of the logic of conversation proposed by Grice (1975, 1989a). It constitutes, therefore, a component of speaker meaning.2 In other words, a conversational implicature is part of what a speaker means, though not part of what a sentence means. How, then, can the addressee compute a conversational implicature generated by the speaker? The answer provided by Grice is that an utterance automatically raises certain expectations, and these speaker expectations guide the addressee to infer what is conversationally implicated by the speaker, given that both the speaker and the addressee are rational and cooperative agents. Of course, Grice formulated the speaker 1 Strictly speaking, ‘not hot’ is what is conversationally implicated. ‘The soup is not hot’ is what is communicated/conveyed/meant, that is, the sum of what is said and what is conversationally implicated. In what follows, however, I shall not make such a difference. 2 Currently, there is a split on the issue of whether a conversational implicature is a component of speaker meaning or a pragmatic inference. Bach (2001, 2006, 2012), Saul (2002), and (following them and biting the bullet) Horn (2009, 2012a, b) held that a conversational implicature is part of speaker meaning rather than a pragmatic inference. By contrast, Sperber and Wilson (1986, 1995), Levinson (2000), Atlas (2005), Geurts (2010), and Chierchia (2013), for example, are still treating a conversational implicature as a pragmatic inference. Furthermore, Recanati (2010: 143–144) drew a distinction between an inference in the broad sense and an inference in the narrow sense and considered a conversational implcature as an inference in the narrow sense.
157
Implicature 157 expectations in terms of his cooperative principle and its constituent maxims. The cooperative principle and its attendent maxims are essentially principles of language use based on the rational nature of human communication, and indeed any shared-goal human activity (Grice 1989a: 28). Put in a slightly different way, they are general communicative norms recognized jointly, though tacitly by both the speaker and the addressee in order to communicate effectively and efficiently (Huang 2012a, 2014). By way of summary, a speaker conversationally implicates, the addressee infers, but a conversational implicature itself is not an inference. The addressee may or may not succeed in figuring out the speaker’s m-intended conversational implicature as an inference.3 Nevertheless, it is the speaker’s expectations about the appropriate inferences the addressee can reasonably be expected to draw that make the production and comprehension of a conversational implicature a rational, shared-goal activity.
8.1.2 Properties of conversational implicature Conversational implicatures are characterized by a number of distinctive properties (Grice 1975, 1989a; Levinson 1983, 2000; Huang 2007: 32–35, 2014). In the first place, there is the property of defeasibility or cancellability, i.e. conversational implicatures can vanish in certain linguistic or non-linguistic contexts. How? They are suspended if they are inconsistent with (i) semantic entailments, (ii) background assumptions or real-world knowledge, (iii) contexts, and (iv) priority conversational implicatures.4 Notice that defeasibility is a necessary but not a sufficient condition for conversational implicature.5 A second property exhibited by conversational implicatures is non- detachability, i.e. any linguistic expression with the same semantic content tends to carry the same conversational implicature. (A principled exception is those conversational implicatures that arise via the maxim of Manner.) This is because conversational implicatures are attached to the semantic content rather than the linguistic form of what is said. Therefore, they cannot be detached from an utterance simply by replacing the relevant linguistic expressions with their synonyms. Thirdly, we have the property of calculability, i.e. conversational implicatures can transparently be derived via the cooperative principle and its component maxims. A fourth property is non-conventionality, 3 When there are meaning discrepancies between utterance production (conversational implicature) and utterance comprehension (inference), we have what Saul (2002) called an ‘audience-implicature’— an ‘unauthorized’ conversational implicature that is recognized by the addressee but not m-intended by the speaker. An audience implicature contrasts with an ‘utterer-implicature’, that is, an ‘authorized’ conversational implicature that is m-intended by the speaker but not recognized by the addressee. Together, an audience implicature and an utterer implicature are labelled a ‘near-implicature’. 4 In addition, Q-implicatures can also be defeated by metalinguistic negation (e.g. Huang 2014: 54–57). 5 While defeasibility is still widely considered to be one of the hallmarks of conversational implicature, there are scholars who question it as a necessary feature of conversational implicature (e.g. Weiner 2006; Capone 2009). I shall point out some non-defeasible conversational implicatures in sections 8.1.6 and 8.1.7.
158
158 Yan Huang i.e. conversational implicatures, though dependent on the saying of what is coded, are non-coded in nature. In other words, they rely on the saying of what is said but they are not part of what is said. Fifthly, there is the property of reinforceability, i.e. conversational implicatures can be made explicit without producing too much of a sense of redundancy. This is because conversational implicatures are not part of the conventional import of an utterance. The sixth property of conversational implicatures is universality, i.e. conversational implicatures tend to be universal, because they are motivated rather than arbitrary. Finally, we have the property of indeterminacy, i.e. some conversational implicatures may be indeterminate. They can be taken as conveying an open-ended range of conversational implicatures relating to matters in hand (see e.g. Huang 2014: 39–43 for illustration).
8.1.3 Two classical and neo-Gricean dichotomies 8.1.3.1 How is a conversational implicature generated? Conversational implicatureO versus conversational implicatureF A conversational implicature can be engendered in two distinct ways. On the one hand, it can arise from strictly observing the maxims of conversation. In Huang (2003, 2007: 27, 2014: 33), I called conversational implicatures thus induced conversational implicaturesO. This is the case for (1) above. On the other hand, a conversational implicature can be generated by way of a speaker’s ostentatiously flouting the maxims. In Huang (2003, 2007: 29, 2014: 35), I dubbed conversational implicatures thus engendered conversational implicaturesF. This is the case with the generation of the conversational implicature in (2), which deliberately exploits Grice’s maxim of Quality. (2) a. The British Foreign Office is in Washington. b. +> The UK follows America’s foreign policies too closely
8.1.3.2 Generalized conversational implicature (GCI) versus particularized conversational implicature (PCI) A second Gricean dichotomy, independent of the first, is between those conversational implicatures which arise without requiring any particular contextual conditions and those which do require such conditions. Grice called the first kind generalized conversational implicatures (GCIs) and the second kind particularized conversational implicatures (PCIs). By way of illustration, consider the two conversational implicatures in Mary’s utterance in (3). (3) John: How did yesterday’s guest lecture go? Mary: Some of the faculty left before it ended. +> a. Not many/most/all of the faculty left before the lecture ended +> b. The lecture didn’t go well
159
Implicature 159 The conversational implicature in (3a) has a very general currency. Any utterance of the form ‘Some x are Y’ will have the ‘default’ meaning ‘Not many/most/all x are Y’. This reading will go through without needing any particular context, hence (3a) is a GCI. By contrast, the conversational implicature in (3b) depends crucially on context of some kind. Mary’s reply points to a possible connection, namely, if some of the faculty left a lecture before it ended, the lecture might not have gone well. Without such a specific connection, we will not have the relevant conversational implicature, thus (3b) is a PCI. Needless to say, the distinction between GCIs and PCIs is a graded concept. This is because a conversational implicature is more or less generalized/conventional or particularized/non-conventional depending on the amount of information provided by the context that is needed to recover it.6 The theoretical importance of this Gricean dichotomy has been subject to heated debates. Carston (2002), for example, doubted whether such a distinction can be maintained. On the other hand, Levinson (2000) put forward a rigorous defence of it. See Grice (1975, 1989a: 37–38) for further discussion. See also Nicolle and Clark’s (1999) experimental evidence that when speakers are provided with consistent criteria, they can distinguish between GCIs and PCIs.
8.1.4 Two neo-Gricean typologies of conversational implicature 8.1.4.1 Horn: Q- and R-implicatures Horn (1984, 2012a, b) put forward a bipartite neo-Gricean model of conversational implicature. In this model, all of Grice’s maxims (except the maxim of Quality) are replaced with two fundamental and antithetical neo-Gricean pragmatic principles: the Q[uantity]- and R[elation]-principles. (4) Horn’s Q-and R-principles
a. The Q-principle (Addressee/hearer-based) Make your contribution sufficient; Say as much as you can (modulo the R-principle).
b. The R-principle (Speaker-based) Make your contribution necessary; Say no more than you must (modulo the Q-principle). Consequently, in his system, there are two genera of neo-Gricean conversational implicatures: Q[uantity]-and R[elation]- implicatures, the former being derived via 6 In this connection, Recanati (2010: 146–152) was of the view that GCIs should be divided into two types: what he dubbed ‘“Gricean” generalized implicature (GGIs)’ and ‘default implicatures (DIs)’. While the former go through a macropragmatic process, the latter undergo a micropragmatic process in the sense of Campbell (1981). I shall return to DIs in sections 8.1.5.3 and 8.1.7.
160
160 Yan Huang the operation of his Q-principle and the latter, of his R-principle. An example of Q- implicature is given in (1), and his R-implicature is exemplified in (5). (5) Sam broken a leg yesterday. +> The leg is his own
8.1.4.2 Levinson: Q-, I-, and M-implicatures Levinson (1987a, 2000) proposed a tripartite neo- Gricean model of conversational implicature. In this model, the original Gricean programme is reduced to three neo-Gricean pragmatic principles: the Q[uantity]-, I[nformativeness]-, and M[anner]-principles. (6) Levinson’s Q-, I-, and M-principles a. The Q-principle (simplified) Speaker: Do not say less than is required (bearing the I-principle in mind). Addressee: What is not said is not the case. b. The I-principle Speaker: Do not say more than is required (bearing the Q-principle in mind). Addressee: What is generally said is stereotypically and specifically exemplified. c. The M-principle Speaker: Do not use a marked expression without reason. Addressee: What is said in a marked way conveys a marked message. As a result, in this model, three genera of conversational implicature are put forward. Q-implicatures are due to the operation of the Q-principle, I-implcature, the operation of the I-principle, and M-implicatures, the operation of the M-principle.
8.1.4.2.1 Q-implicatures Three types of Q-implicature can be identified: (i) Q-scalar implicatures, as in (1) above; (ii) Q-clausal implicatures and (iii) what I dubbed Q-alternate implicatures in Huang (2007: 42, 2014: 51). Q-scalar implicatures are derived from Horn scales (e.g. Huang 2010f). Next, Q-clausal implicatures are pragmatically enriched meanings of epistemic uncertainty. Like Q- scalar implicatures, Q-clausal implicatures also rest on a set of contrastive semantic alternates, but in this case, of a constructional kind. Wherever there is a construction Y(p), where p is not entailed by Y(p), and there is an alternative construction X(p) semantically similar and of roughly equal brevity to Y(p) except that X(p) does entail p, then the use of the semantically weaker Y(p) Q-implicates that the speaker does not know whether p obtains or not (Gazdar 1979: 59–62; Grice 1989a: 8; Levinson 2000: 108–111). An example of Q-clausal implicatures is given in (7). Other central cases include ‘disjunction’ (), ‘conditional’ (), and ‘modal’ ().
161
Implicature 161 (7) (Verbal doublet)
Mary believes that 2015 is the 800th anniversary of the publication of Magna Carta. +> 2015 may be the 800th anniversary of the publication of Magna Carta, it may not be We move finally to Q-alternate implicatures. Roughly, we have two subtypes here. First, the lexical expressions in a set are informationally ranked, as in (8), (9), and (10). Following Huang (2007), let me call Q-implicatures deriving from such a set Q-ordered alternate implicatures. By contrast, in the second subtype, the lexical expressions in the set are of equal semantic strength, as in (11) and (12). Let me term Q-implicatures thus induced Q-unordered alternate implicatures.
(8)
Horn-like scales
John tried to give up smoking. +> John did not succeed in giving up smoking
(9)
Rank orders
John’s son is a colonel. +> John’s son is not a general
(10) Hirschberg scales
John: Have they got divorced? Mary: They have got separated. +> They haven’t got divorced yet (11) Please stir-fry the vegetables. +> Please don’t e.g. bake, boil, grill, and steam the vegetables (12) We don’t teach e.g. Russian here In (8), succeed is informationally stronger than try, but it does not entail try, as is attested by John gave up smoking without even trying. In other words, succeed and try form a contrast set, but the set is non-entailing. However, as in the case of Q-scalar implicatures arising from a Horn scale, the use of the infomationally weaker try in (8) gives rise to a similar, Q-ordered alternate implicature. Next, (9) forms a rank order, in which the semantically stronger linguistic expression entails unilaterally the negation of the semantically weaker one (Horn 2007a, 2009). In a similar vein, the use of the semantically weaker colonel
162
162 Yan Huang yields the Q-ordered alternate implicature ‘not general’. Finally, more or less the same analysis can be applied to (10). But notice that there is a difference between the scales in (8) and (9) and the scale in (10). While the scales in (8) and (9) are given by the lexicon without requiring any specific context, the scale in (10) is provided by some general assumption about the world, context, and other pragmatic factors. Put in a slightly different way, whereas the former are a semantic or lexical scale that can be based on the different structures of the lexicon such as taxonomies, metonymies, and helices, the latter is essentially a nonce, pragmatic scale, that is, a contextually given ad hoc scale. Such a scale can be based on any partially ordered contrastive sets in a contextually salient way. Levinson (2000) called this kind of pragmatically defined set a Hirschberg scale. Furthermore, Horn (2007a: 168– 170) distinguished two kinds of pragmatic strengthening: informative and rhetorical. While I-implicature (to be discussed in ‘I-implicatures’ section) increases both informative and rhetorical strength, Q-implicature is informatively but not rhetorically stronger than the sentence uttered without the implicature (see e.g. Huang 2010e for further discussion). See also Verbuk (2012) for her arguments against the distinction between Horn and pragmatic scales and for her context-based question-under-discussion (QUD) account. Next, the lexical expressions in (11) are of equal informational strength, hence forming an unordered semantic or lexical scale. On the other hand, the scale in (12) constitutes an ad hoc pragmatic scale that is also based on an informational symmetry. In both cases, the use of any lexical expression in the set engenders a weak Q-unordered alternate implicature, namely, the speaker is not in a position to use any other lexical expressions in this same scale. Finally, whereas Q-implicatures derivable from a Horn or Horn-like scale are GCIs, those acting on a Hirschberg scale are PCIs.
8.1.4.2.2 I-implicatures The class of I-implicatures is heterogeneous, ranging from ‘conjunction buttressing’ via ‘inference to stereotype’ to ‘adjective interpretation’. (13)
Conjunction buttressing p and q +> p and then q (temporal sequence) +> p therefore q (causal connectedness) +> p in order to cause (teleology, intentionality) John pressed the button and the bell rang. +> John pressed the button and then the bell rang +> John pressed the button and thereby caused the bell to ring +> John pressed the button in order to make the bell ring
(14) Conditional perfection if p then q +> iff p then q If you give me a free Beethoven, I’ll buy five Mozarts. +> If and only if you give me a free Beethoven will I buy five Mozarts (15) Lexical narrowing John doesn’t drink. +> John doesn’t drink alcohol
163
Implicature 163 Other cases of I-implicature include ‘membership categorization’, ‘mirror maxim’, ‘frame- based inference’, ‘bridging cross-reference’, ‘indirect speech act’, ‘definite reference’, ‘negative raising’, ‘negative strengthening’, ‘computation of noun–noun compound’, ‘interpretation of spatial term’, ‘interpretation of possessive construction’, ‘coreferential interpretation’, ‘proper name narrowing’, and ‘systematic ambiguity’ (see e.g. Huang 2014: 58–62 for exemplification). While the category has been regarded as a heterogeneous or even motley collection of distinct and different phenomena, the I-implicatures listed above do share a number of properties: notably, first, they are semantically more specific than the utterances that engender them. In other words, the utterances have been strengthened by the conversational implicatures. Secondly, unlike Q-implicatures, I-implicatures are positive in nature. Thirdly, in some cases, they are characteristically guided by socially or culturally stereotypical assumptions. Fourthly, they are generally non-metalinguistic, in the sense that they make no reference to something that might have been said but was not actually said (Levinson 2000: 119). Finally, they normally cannot be cancelled by metalinguistic negation (Huang 2014: 58–61).
8.1.4.2.3 M-implicatures Unlike Q-and I-implicatures, which are derived on the basis of contrast in semantic informativeness, M-implicatures are derived in terms of a set of alternates that contrast primarily in linguistic form such as ‘double negation’, ‘verbal periphrasis’ or ‘causative construction’, and ‘repeated verb conjunct’. The use of the marked (b) sentence in (16), for example, gives rise to an M-implicature. (16) a. John stopped the car. +> John stopped the car in the usual manner b. John caused the car to stop. +> John stopped the car in an unusual manner, e.g. by deliberately bumping it into a wall Finally it should be mentioned that inconsistencies arising from the three types of potentially conflicting conversational implicatures can be resolved by an ordered set of precedence in the following way. Genuine Q-implicatures (where Q-clausal cancels rival Q-scalar) supersede inconsistent I-implicatures, but otherwise I-implicatures take precedence until the use of a marked linguistic expression triggers a complementary M-implicature to the negation of the applicability of the pertinent I-implicature (see e.g. Huang 2006a, 2014: 64–66 for exemplification and further discussion).
8.1.5 Some current debates on GCI, in particular Q-scalar implicature At this point, it is useful to discuss some of the current debates on GCIs, in particular Q-scalar implicatures, given that the latter represent the most heavily trodden area of the original Gricean terrain of conversational implicatures.
164
164 Yan Huang
8.1.5.1 Cardinals: Scalar expressions generating Q-scalar implicatures? In the first place, the treatment of cardinals as scalar expressions/predicates engendering Q-scalar implicatures is highly controversial. It has been noticed at least since Aristotle that a sentence like (17) has two systematically distinct interpretations: a one-sided, lower-bounded reading, as in (17a) and a two-sided, upper-and lower-bounded reading, as in (17b). (17) Some stores were crowded with shoppers. a. At least some stores were crowded with shoppers. b. Some but not many/most/a ll stores were crowded with shoppers. How can a semanticist deal with sentences like (17)? He or she has to treat these sentences as lexically or logically ambiguous. However, there is a serious problem at the very heart of this ambiguity analysis: namely, it runs directly against the spirit of ‘Occam’s razor’, the doctrine that entities are not to be multiplied beyond necessity. This has the consequence that all things being equal, an account which has to suggest two lexical items is to be rejected in favour of an analysis which does not. This is where neo-Gricean pragmatics comes in. As proposed in Horn (1972, 1989) and formalized in Gazdar (1979), the alternative analysis, which Horn (2006a) dubbed ‘the Golden Age of Pure Pragmatics (GAPP)’, is to obtain the one-sided, lower-bounded interpretation from semantics, but to derive the one-sided, upper-bounded reading via Q-scalar implicature. In other words, on this account, a sentence like (17) asserts or entails its one-sided, lower-bounded reading, Q-implicates its one-sided, upper-bounded reading, and the conjunction of the assertion and the implicature results in the corresponding two-sided, upper-and lower-bounded communicated understanding. This analysis applies to all types of scalar expressions/predicates including logical operators such as some, ordinary expressions such as warm, and cardinals such as nine. But recently, this GAPP-style analysis of cardinals has been called into question. According to König (1991), Horn (1992b, 2006c, 2009), Geurts (1998), Carston (2002), Papafragou and Musolino (2003), Ariel (2004), Bultinck (2005), and Hurewitz et al. (2006), there has been substantial linguistic and experimental evidence that such a Q-scalar implicature account is untenable. As an alternative, it has been argued that cardinals be treated as semantically underspecified, but not as being assigned the weak scalar value ‘at least n’ by semantics. The propositional content of the sentence that contains the cardinal will then be determined only in the context in which the cardinal is used. However, this revisionist position is rejected by Levinson (2000: 87–90), who argued strenuously that cardinals continue to be treated on a par with other scalar expressions so that the original neo-Gricean analysis be retained (see also Landman 2000, Barner and Bachrach 2010, and Cummins et al. 2012).
8.1.5.2 Epistemic strength of Q-scalar implicature Secondly, there is the issue of the epistemic strength of Q-scalar implicatures. Epistemic strength is concerned with the question of what it is a speaker Q-scalar implicates against.
165
Implicature 165 Two neo-Gricean pragmatic positions can be identified here: the weak epistemic one represented by e.g. Hirschberg (1991), Sauerland (2004), Geurts (2009), and Horn (2009), and the strong epistemic one advocated by e.g. Gazdar (1979) and Levinson (2000: 77–79).
8.1.5.3 Q-scalar implicature: Default, contextual, or grammatical Thirdly and more importantly, a heated debate has been going on for the last two decades or so, focusing on the nature of neo-Gricean GCIs in general and Q-scalar implicatures in particular. One view is that Q-scalar implicatures convey default (presumptive) meanings—meanings that language users have, in some sense, pre-calculated—sans a conscious inferential process, though they can nevertheless be cancelled by semantic content and context when they don’t fit. This is called the default inference theory or defaultism. On the authority of this approach, Q-scalar implicatures are generated by default, that is, blindly and automatically, as soon as the linguistic expressions that serve as triggers of the implicatures are encountered (Levinson 2000). Furthermore, they are not only generated but to a certain extent conventionalized as well in the sense that they are associated with certain linguistic expressions that trigger the blind and automatic process of generating the implicatures. The convention involved here belongs to the category of ‘convention of usage/use’ rather than ‘convention of language/meaning’ distinguished by Morgan (1978). Therefore, Q-scalar implicatures are DIs in Recanati’s terminology (Recanati 2010: 148–152). Furthermore, extending the Gricean insight of the GCI–PCI dichotomy, Levinson (2000) developed a theory of presumptive meaning through a theory of three levels of meaning. On a traditional, standard view, there are only two levels of meaning to a theory of communication: a level of sentence-type meaning versus a level of utterance-token meaning. The study of the former figures in semantics, and the investigation of the latter belongs to pragmatics. But Levinson argued that such a view ‘is surely inadequate, indeed potentially pernicious, because it underestimates the regularity, recurrence, and systematicity of many kinds of pragmatic inferences’ (Levinson 2000: 22). He proposed to add a third level—utterance-type meaning—to the two generally accepted levels of meaning. This third layer is the level of generalized, preferred, or default interpretation, which is not dependent upon direct computations about speaker intentions but rather upon expectations about how language is characteristically used (ibid.). GCIs, in particular Q-scalar implicatures, argued Levinson, should be included on this layer, as these pragmatic inferences have an expectable, stable, and even conventional interpretation. In order to account for this kind of conversational implicature, as we have already seen, Levinson has isolated a set of three default inferential heuristics—the Q-, I-, and M-principles, which is associated with a set of three default utterance-type conversational implicatures. Stated in this way, a neo-Gricean pragmatic theory of conversational implicature, which is largely concerned with GCIs rather than PCIs, is essentially a theory of utterance-type meaning on a level intermediate between sentence-type meaning on the one hand and utterance-token meaning on the other. In other words, it is a theory of presumptive meaning—pragmatic inference that is generalized, default, and
166
166 Yan Huang presumed.7 Defaultism in the weak form (weak defaultism) is represented by Gazdar (1979) and Horn (1989)8 and defaultism in the strong form (strong defaultism), by Levinson (2000) and Chierchia (2004, 2006, 2013), about which later. In counterpoint to defaultism, a second, relevance-theoretic position is that conversational implicatures are essentially inferred contextually on a case-by-case basis (e.g. Sperber and Wilson 1986, 1995). This is labelled the contextual inference theory. In addition, a more recent, third view holds that the derivation of GCIs, in particular Q-scalar implicatures, relies heavily on structural factors (e.g. Landman 2000; Chierchia 2004, 2006, 2013; Chierchia et al. 2012; Margi 2009; see also Crain and Pietroski 2002). According to Chierchia and his associates, while a standard upper-bounding Q-scalar implicature arises from a positive Horn scale, it is quite weak and even blocked in a negative Horn scale and other downward-entailing environments. On the basis of this observation, they endorse Levinson’s default analysis of Q-scalar implicatures and argue that Q-scalar implicatures be computed compositionally. Contrary to what Landman (2000) called ‘the Gricean root’, namely, the orthodox Gricean doctrine that the derivation of conversational implicatures is deferred to the output of grammar at the utterance level, Chierchia and their associates are of the view that Q-scalar implicatures are not derived after the truth conditions of the (matrix) sentence are worked out. Rather, they are introduced locally, calculated algorithmically phrase by phrase in accord with the truth conditions starting with the most deeply embedded clause, and projected in the syntactic tree diagram of the matrix sentence in a bottom-up way, and are integrated in the semantics where they occur. This has the consequence that they are an integrated part of compositional semantics, and the computation of Q-scalar implicatures falls under the computational system of grammar (see also Recanati 2010: 152–157). This is known as the structural inference theory.9 Furthermore, all three views have recently been subject to the vast swath of empirical work in experimental pragmatics. 7 Note that, as pointed by Levinson (2000), this middle layer of utterance-type meaning has constantly been subject to attempts to reduce it on the one hand, to the upper layer of sentence-type meaning, as in e.g. discourse representation theory and the structural inference theory (see below, this section), and on the other hand, to the lower level of utterance-token meaning, as in e.g. Sperber and Wilson’s (1986, 1995) relevance theory. In my view (Huang 2003, 2004a, 2007, 2014), such reductionist efforts, though methodologically highly desirable given the Occamistic principle discussed above, cannot be successful. The reason they will fail is this: on the one hand, GCIs are defeasible, that is, they can be cancelled in certain linguistic and/or non-linguistic contexts. This makes it difficult for them to be semanticized. On the other hand, other things being equal, a theory about types is better than a theory about tokens in that the former enjoys more predictive and explanatory power. Therefore, any attempts to reduce GCIs to nonce or once-off inferences should be resisted. If these arguments are correct, a three-tiered theory of communication with a layer of default interpretation sitting midway is in principle preferred over a two-levelled one without such an intermediate layer. 8 But given Horn (2006c), it is unclear whether he still holds the weak defaultism view. To quote him: ‘An implicature may arise in a default context without thereby constituting a default or automatic inference’ (Horn: 2006c: 82). 9 Another current attempt to reduce Q -scalar implicature to generative grammar and thus to the human innate linguistic mechanism is optimality-theoretic pragmatics and game-and decision- theoretic pragmatics (Blutner and Zeevat 2004; Benz et al. 2006).
167
Implicature 167 While much of the relevance-theoretically oriented experimental work undertaken by Noveck, Sperber, Breheny, and their associates (e.g. Noveck and Sperber 2007; Noveck and Reboul 2008; Breheny et al. 2006) rejects the default inference theory in favour of the contextual inference approach, there is also experimental evidence in support of the default inference theory (e.g. Grodner et al. 2007) and the structural inference view (e.g. Panizza and Chierchia 2008). See also Garrett and Harnish (2007, 2009) and Doran et al. (2009, 2012) for a summary of experimental pragmatic work on GCIs. Some of the observations made by Chierchia and his associates are not entirely novel. The projection properties of Q-scalar implicatures have long been a concern of neo- Griceans. As early as in 1979, Gazdar claimed that Q-scalar implicatures are suspended by logical operators (and) in embedded contexts. But, as pointed out by Hirschberg (1991), Gazdar’s generalization prevents too many Q-scalar implicatures. Hirschberg’s own view is that Q-scalar implicatures are barred only under overt negation. But, according to Horn (2006a), such an analysis blocks too few Q-scalar implicatures. Horn argued, contra Chierchia, that Q-scalar implicatures arising from negative Horn scales are not less robust than those which are derived from their positive counterparts. In fact, as pointed out by both Levinson (2000: 82, 254–255) and Horn (2006a), the alleged blockage of Q-scalar implicatures is due to the fact that a positive Horn scale is reversed under negation and other downward-entailing operators, and consequently a different Q-scalar implicature is derived from the inverse scale (see also Huang 2007, 2011, 2014). This is illustrated in (18). The general conceptual problem of the Chierchia-type, grammatical theories of ‘blind mandatory scalar implicature’ (Margi 2009) is, as Horn (2006c) pointed out, that implicatures have in essence to be stipulated rather than derived from general principles of cooperation and rationality such as Grice’s cooperative principle and its constitutive maxims of conversation. (18)
Negative Horn scales
Not all stores were crowded with shoppers. +> most/many/some stores were crowded with shoppers
Finally, in recent years, there has also been another intense debate on (i) whether or not pragmatically enriched or inferred content can ‘intrude’ upon or enter the conventional, truth-conditional content of what is said, and (ii) if the answer to (i) is positive, then what the pragmatic intrusion under consideration is. Is it an explicature, part of the pragmatically said, impliciture, or conversational implicature? I shall postpone the discussion of these issues to section 8.1.7.
8.1.6 Embedded implicature The conversational implicatures I have discussed so far are all generated at the utterance/sentential level. In other words, they are all derived on the basis of a full locutionay
168
168 Yan Huang act—the act of producing a meaningful linguistic expression. In this section, I turn to embedded implicatures. Recently, embedded implicatures and especially embedded Q- scalar implicatures have attracted a growing amount of attention from both philosophers of language and linguistic semanticists and pragmaticists.
8.1.6.1 What is an embedded implicature? An embedded implicature is a seeming or pseudo conversational implicature that is engendered locally at the subsentential level, typically occurring in a clause that is embedded under a logical operator, such as a propositional attitude verb, a conditional, and a comparative. (19) a. John believes that the soup is warm. b. +> John believes that the soup is not hot (strong ‘implicature’) c. +> John does not believe that the soup is hot (weak ‘implicature’) (20) Parents are less happy if their children go to bed and brush their teeth. +> Parents are less happy if their children go to bed first and then brush their teeth (21) John knows that the train timetable is not unreliable. +> John knows that the train timetable is less reliable than the uttering of the sentence ‘John knows that the train timetable is reliable’ suggests In the belief report (19), the use of warm seems to create a Q-scalar implicature arising locally in the clause embedded under the propositional attitude verb believe. In (20), the apparent I-implicature stemming from the use of and occurs within the scope of the conditional. Note further that this conversational implicature cannot normally be cancelled. Finally, (21) contains the factive verb know, embedded under which is the clause where the use of not unreliable generates a seeming M-implicature. In all these cases, the conversational implicatures are derived from a non-asserted subconstituent of the utterance/sentence.
8.1.6.2 The main problem Cohen (1971) was perhaps the first to bring the phenomenon of embedded implicature to the attention of philosophers and linguists. His examples are given in (22) and (23). (22) a. The old king has died of a heart attack and a republic has been declared. b. +> The old king has died of a heart attack first and then a republic has been declared (23) a. A republic has been declared and the old king has died of a heart attack. b. +> A republic has been declared first and then the old king has died of a heart attack
169
Implicature 169 On a standard Gricean account, the sentences in (22a) and (23a) have exactly the same truth-conditional content, even though they differ in the temporal sequence of events. This difference in conveyed meaning is captured in terms of a GCI, namely, a neo- Gricean I-implicature, as indicated by the relevant conversational implicatures in (22b) and (23b). However, according to Cohen, this analysis of Grice’s is untenable. If the sentences in (22a) and (23a) really have the same truth-conditional content and differ only in conversational implicatures, then when they are embedded in the antecedent of a conditional, as in (24) and (25), given Grice’s analysis of if, they should also have the same truth-conditional content. (24) If the old king has died of a heart attack and a republic has been declared, then Tom will be quite content. (25) (Adapted from Cohen 1971) If a republic has been declared and the old king has died of a heart attack, then Tom will not be content at all. But (24) and (25) are not truth-conditionally equivalent. The pragmatically enriched temporal relation, which holds between the conjuncts in (22b) and (23b), is an integral part of the antecedents of the conditionals in (24) and (25). With one temporal sequence of events, Tom will be happy; with the other temporal sequence of events, Tom will be unhappy (see also Cohen 1977). This indicates that the pragmatically augmented temporal relation falls within the scope of the conditional—a logical operator. However, given the classical Gricean mechanism, this is simply not possible. The reason is that in the original Gricean paradigm, conversational implicatures can only be derived at the utterance/sentential level or on the basis of a speaker’s performance of a full locutionary act, that is, of an S saying that p. Conversational implicatures thus derived are global and post-propositional (Recanati) or post-compositional (Chierchia). Consequently, they cannot arise at the sub-utterance/subsentential or the sub-locutionary act level. In other words, they cannot be embedded and pre-propositional or pre-compositional (e.g. Bach 2012, but see Walker 1975 for a defence of the orthodox Gricean analysis, and see also Geurts 2010 for other problems in relation to Q-scalar implicature).10
8.1.6.3 Analyses Currently, the study of embedded implicatures is largely restricted to a subtype of conversational implicature, namely, Q-scalar implicatures. With regard to embedded Q-scalar implicatures, there are roughly two approaches: grammatical and pragmatic. 10
According to Recanati (2010: 147), this argument was also made independently by Ducrot (1972, 1973). But there are scholars like Green (1998), who are of the view that a GCI is eo ipso embeddable. His embedded implicature hypothesis runs like this: ‘If assertion of a sentence S conveys the implicatu[re] that p with universal regularity, then when S is embedded the content that is usually understood to be embedded for semantic purposes is the proposition (S & p)’.
170
170 Yan Huang Within the first, grammatical camp, which is called ‘conventionalism’ by Geurts (2010), there are two varieties: lexicalist conventionalism and syntax-based/driven conventionalism.11 The central idea underlying both versions of conventionalism is that an embedded Q-scalar implicature is part of the lexico-grammatical content of a sentence. In other words, it is part of the conventional, truth-conditional meaning rather than a pragmatic enrichment of a sentence. More specifically, according to conventionalism, embedded Q-scalar implicatures ‘are not implicatures at all, but rather soft entailments projected from the lexicon or the grammar’ (Geurts 2010: 134). They are soft entailments because they can be defeated. With regard to lexicalist conventionalism, an embedded Q-scalar implicature is ‘hard-wired into the lexical entries’ of an implicature trigger (Geurts 2010: 145). Given that it is defeasible, it is not part of the lexical meaning of the linguistic expression involved. Nevertheless, it is part of the lexical content or lexical inference of that expression. Consequently, an embedded Q-scalar implicature can be accounted for in terms of a DI. This approach is represented by Cohen (1971, 1977), Landman (2000), Chierchia (2004), and especially Levinson (2000). Next, according to the syntax-based conventionalism, the grammar of a language is equipped with a covert or hidden exhaustivity (EXH) operator, whose meaning is akin to ‘only’ in English. This syntactic operator can be freely inserted into the tree diagram of a sentence. In other words, under the syntactic approach, an embedded Q-scalar implicature is generated by the computational system of grammar. It is calculated and integrated locally in a compositional way (see e.g. Chierchia 2006, 2013; Fox 2007; Chierchia et al. 2012; and see also Gajewski and Sharvit 2012). Construed thus, this localist or localistic model is too powerful in that the covert EXH operator engenders more readings than what is desired for any given sentence. This overgeneration has to be constrained in one way or another. The most popular proposal within the syntax-based camp is to adopt (a version of) the Power Principle, namely, the idea that out of two readings x and y, where x is stronger than y, x should be the preferred reading. From an experimental point of view, conventionalism has received support from Clifton Jr and Dube (2010) and Chemla and Spector (2011) (but see Geurts and van Tiel 2013 for a critique). We move next to the pragmatic approach. One variety is to defend the Gricean globalist account of Q-scalar implicature. This is represented by Sauerland (2004), van Rooij and Schulz (2004), Horn (2006a, 2009), Spector (2006), Russell (2006), and especially Geurts (2009, 2010). On this view, contrary to the conventionalist continuity hypothesis or stance that the upper-bounded reading of a Q-scalar implicature occurs across the board, be it at the sentential level (unembedded) or at the subsentential level (embedded) and that it ‘occurs systematically and freely in arbitrarily embedded positions’ (Chierchia et al. 2012), an embedded Q-scalar implicature requires special linguistic marking, such as a contrastive stress. It is marginal and rare and sometimes the upper- bounded reading has to be forced. In other words, an embedded Q-scalar implicature constitutes an exceptional and marked case (see also Ippolito 2010). Furthermore, the
11
Conventionalism is also called the ‘semanticization’ of embedded implicatures by Recanati (2010).
171
Implicature 171 use of embedded and unembedded scalar expressions is computed differently. While the use of unembedded scalar expressions invites Q-scalar implicatures, the use of embedded ones does not. Moreover, embedded scalar expressions are frequently dealt with on a case-by-case basis. In the case of the belief report in (19), for example, (19b) is not treated as a locally engendered Q-scalar implicature. Rather, it is analysed as following from a Q-scalar implicature licensed by (19a) (Geurts and van Tiel 2013). This version of the Gricean globalist account is experimentally supported by Geurts and Pouscoulous (2009) and Geurts and van Tiel (2013) (but see Clifton Jr and Dube 2010 for criticism, and see also Sauerland 2010). Another pragmatic account is developed by Recanati (2003, 2010) (see also Bach 1994). Under this truth-conditional pragmatic approach, an embedded Q-scalar implicature is explained in terms of a primary pragmatic process of modulation, in particular free enrichment. By modulation is meant a primary pragmatic process that is pragmatically controlled or contextually driven (top-down), optional, and produced locally. It intersperses with the compositional determination of the truth-conditional content of a sentence uttered. In other words, modulation takes the meaning of a linguistic expression as input and yields as output a pragmatically enriched meaning, which functions as a compositional value. Free enrichment, according to Recanati, is the most typical and pervasive subtype of modulation. Under free enrichment, the meaning of a linguistic expression can be contextually strengthened, that is, it can be contextually given a more specific reading than what is literally encoded by that expression. This is the case for embedded Q-scalar implicatures. For example, the meaning of the scalar expression warm in (19a) is freely enriched to ‘warm but not hot’. Finally, Recanati is of the view that the grammatical and pragmatic approaches to embedded Q-scalar implicatures may complement each other. The heated debate is still going on. According to Geurts and van Tiel (2013), the broad consensus reached between the grammatical and pragmatic camps is that there are two mechanisms that underwrite the upper-bounded reading (or upper-bounded construal (UBC) in their terminology) of the use of scalar expressions: Q-scalar implicatures and truth-conditional narrowing. What is at stake here is the issue of how to make the division of labour between the two mechanisms. I shall continue the discussion of some cases of embedded implicature in the next section.
8.1.7 Pragmatic intrusion into what is said: Explicature, the pragmatically enriched said, impliciture, or conversational implicature? As mentioned in section 8.1.5.3, recently there has also been a heated debate on (i) whether or not pragmatically enriched or inferred content can ‘intrude’ upon or enter the conventional, truth-conditional content of what is said; and (ii) if the answer to (i) is positive, then what the pragmatic intrusion under consideration is.
172
172 Yan Huang Concerning the first question, a dividing line can be drawn between pragmaticists like the relevance theorists (Sperber and Wilson 1986, 1995; Carston 2002), Recanati (2004a, 2010), and Levinson (2000), who argue for pragmatic intrusion (but see e.g. Horn 2006a, c, 2009, for serious reservations, about which later), and semanticists like Cappelen and Lepore (2005) and King and Stanley (2005), who argue against it. What, then, is the main empirical evidence for pragmatic intrusion? Firstly, as pointed by Grice (1989a: 25), before we work out what is said in his sense, we have (i) reference to identify, (ii) deixis to fix, and (iii) ambiguity to resolve. To (i)–(iii), Levinson (2000: 172–186) added (iv) ellipsis to unpack and (v) generalities to narrow. It turns out, however, that contrary to the classical Gricean position, the determination of (i)–(v) involves pragmatically embellished meaning of some kind. Secondly, on many occasions of use, propositions associated with what is said in the classical Gricean sense contain unarticulated constituents (UCs)—propositional or conceptual elements of a sentence that is not explicitly expressed linguistically (e.g. Huang 2012b). Three stock examples are given in (26)–(28). They are the three sentences without brackets. On the other hand, the sentences with brackets contain the possible, pragmatically enriched propositional or conceptual material for the UCs. (26) It is snowing [in Beijing]. (27) Some people are a bit surprised when they found out that I’ve got a [good] brain. (Catherine McQueen) (28) [The books of] Confucius is/are on the top of the shelf. The recovery of the UCs in (26)–(28) requires pragmatic enrichment of some kind. The sentence without the brackets in (26) does not express a complete proposition. Consequently, it cannot be evaluated truth-conditionally. Therefore, it undergoes a pragmatic process of completion (Bach) or saturation (Recanati) to provide extra propositional or conceptual material to fill in the UC, thus making it become fully propositional. Completion/saturation (which involves value assignment) is typically a linguistically mandated, local, and bottom-up process. By contrast, while the proposition expressed by the uttering of the sentence without brackets in (27) is a complete, though minimal one, it falls short of what the speaker m-intends to convey. As a consequence, it needs to be expanded. The pragmatic process of expansion will flesh out the proposition expressed by the sentence uttered and engenders a richer proposition. Expansion is typically an optional, contextually driven, and top-down process. It is a subtype of free enrichment, which is itself a subtype of modulation (Recanati 2010). Finally, (28) involves the pragmatic process of semantic/predicate transfer. In this type of pragmatic process, the output proposition is neither an enriched nor an impoverished version of the concept literally expressed by the input one; rather, it represents a different concept, provided that there is a salient functional relation between the old and new concepts. Once again, there is some pragmatically expanded content into the conventional, truth-conditional content of what is said.
173
Implicature 173 Thirdly and finally, following in the footsteps of Cohen, Wilson, Atlas, and Gazdar, Levinson (2000) argued that contrary to Grice, conversational implicature can encroach upon truth-conditional content. In particular, he claimed that the classic Cohen–Wilson argument can be extended into logical connective constructions, such as conditionals (29), comparatives (30), disjunctions (31), and because-clauses (32). (29) If her grandchildren get married and have children, Mary will be happy. (30) Brushing your teeth and going to bed is better than going to bed and brushing your teeth. (31) Mary’s grandchildren either got married and had children or had children and got married—I don’t know which. (32) Because some of her students came to her seminar, Dr Garman was disappointed. These constructions are labelled ‘intrusive’ constructions by Levinson. The reason is that in these constructions, ‘the truth conditions of the whole depend in part on the [conversational] implicatures of the parts’ (Levinson 2000: 198). The truth-conditional content of (29)–(31) rests crucially on the generalized I-implicature stemming from the use of and to mean ‘and then’. On the other hand, the quantifier some in (32) has to be Q-implicated to ‘some but not all’. Thus, there seems no avoiding the conclusion that the truth condition of the complex construction has to be calculated taking into account the conversational implicature of its part. Notice further that these truth- condition-contributing conversational implicatures cannot normally be defeated. This is because they have become an integrated part of the truth condition of the whole construction at the pre-semantic level (e.g. Huang 2012b). Next, regarding the second question of what the pragmatic intrusion, two current positions can roughly be identified: the something other than conversational implicature approach and the conversational implicature one. Within the first camp, three analyses are of particular interest. First, in relevance theory, pragmatic intrusion into what is said is refashioned as explicature—a proposition that is an inferential development of one of the linguistically given incomplete conceptual representations or logical forms of a sentence uttered (e.g. Sperber and Wilson 1986, 1995; Carston 2002). Defined thus, explicature corresponds roughly to the intuitive notion of what is said, though the Gricean notion of what is said is abandoned in relevance theory.12 Secondly, somewhat similar to the relevance-theoretic view is the position taken by Recanati (2004a). According to him, pragmatic intrusion under discussion is part of what he called the ‘pragmatically enriched said’. Finally, a third approach is due to Bach (1994, 2004, 2012). On Bach’s view, there is no pragmatic 12
Consequently, the majority of classical and neo-Gricean GCIs are reanalysed as explicature in relevance theory. Conversational implicature in the relevance-theoretic sense, called r-implicature in Huang (2007: 195, 2014: 281), is largely a PCI in the classical and neo-Gricean sense.
174
174 Yan Huang intrusion into what is said. This is because certain communicative content does not need to be recognized as either part of what is said or part of what is conversationally implicated. Rather, it constitutes a middle ground between what is said and what is conversationally implicated. Bach dubbed this middle level of speaker meaning conversational impliciture or impliciture for short (see also Huang 2010e). Furthermore, Bach (2012) has recently pointed out explicitly that impliciture comes in two forms: completion and expansion. By contrast, the second approach is championed by Levinson (2000). Within the neo-Gricean framework, Levinson argued that pragmatic intrusion into what is said is neither explicature, nor part of the pragmatically enriched said, nor impliciture. Rather, it is the same beast as a neo-Gricean conversational implicature.13 The question that arises next is whether explicature, the pragmatically enriched said, and impliciture on the one hand and conversational implicature on the other can conceptually be distinguished in a systematic way? As pointed out in Huang (2007, 2014), the answer might be no. The reason is threefold. First of all, so-called explicature, the pragmatically enriched said, or impliciture is (intuitively felt to be) engendered by the same Gricean pragmatic mechanism that yields a conversational implicature, though the former seem to contribute to what is said, sometimes helping to determine its truth conditions. Secondly, Recanati (1993, 2010)14 postulated two theoretical tests, namely, the availability principle and the scope principle to separate explicature, the pragmatically enriched said, or impliciture from conversational implicature, but as I argued in Huang (2007, 2010a, 2015b), neither of which seems to work. As a consequence, currently there is no reliable test that can be employed to theoretically distinguish alleged explicature, the pragmatically enriched said, or impliciture from conversational implicature. This is also the case with the plethora of work carried out in experimental pragmatics. I do not think that there is any experiment that can differentiate explicature, the pragmatically enriched said, and/or impliciture from conversational implicature. Therefore currently, there is no failsafe test (both conceptual and experimental) that can be employed to distinguish alleged explicature, the pragmatically enriched said, or impliciture from conversational implicature on a principled basis. Thirdly, other things being equal, given the metatheoretical principle known as ‘modified Occam’s razor’, namely, the Gricean doctrine that theoretical entities are not to be multiplied beyond necessity, the conversational implicature analysis is theoretically and methodologically preferable, because it postulates fewer 13 See my neo-Gricean and revised neo-Gricean analyses of anaphora in e.g. Huang (1991, 1994/2007, 2000a, b, 2004a, 2007, and 2014), which in effect argue that the pre-semantic, pragmatic enrichment involved in the interpretation of anaphora is a conversational implicature. 14 They constitute what Recanati (2010: 165–169) called a ‘primary pragmatic process’ in contrast to a ‘secondary pragmatic process’, which is a post-propositional inference à la Grice (1975, 1989a). Primary pragmatic processes include completion/saturation, expansion, modulation, and free enrichment. As we will see momentarily, they are taken as ‘pre’-semantic rather than ‘post’-semantic conversational implicatures in the neo-Gricean framework or in Korta and Perry’s (2011) terminology ‘near-’ rather than ‘far-side’ pragmatics.
175
Implicature 175 theoretical categories in the interpretation of an utterance than the explicature, pragmatically enriched said, or impliciture account.15 Levinson’s neo-Gricean proposal that conversational implicature can feed into rather than just read off what is said, is, however, strongly challenged by Horn (2004, 2006a, 2012a, b) within the neo-Gricean camp (see also King and Stanley 2005). On Horn’s view, what Levinson has argued is inconsistent with the spirit of the original Gricean programme. In the classical ‘GAPP’, pragmatics can only read off but not feed into what is said. Conversational implicatures, by definition, cannot make any contribution to the truth-conditional content of what is said. In other words, according to Horn, the Levinsonian revisionist model is more neo-than Gricean. Horn’s own proposal is that we should instead adopt Bach’s neo-Gricean model, in which we would allow impliciture, which is built out of what is said, to contribute to what is said, whereas the traditional Gricean semantic concept of what is said, along with a post-semantic orthodox Gricean characterization of what is conversationally implicated, is retained in a neo-classical way. If neo-Gricean conversational implicature can affect truth-conditional content, this gives rise to a problem known as Grice’s circle, namely, how what is conversationally implicated can be defined in contrast to, and calculated on the basis of, what is said, given that what is said seems to both determine and to be determined by what is conversationally implicated. Levinson’s solution is that we should reject the ‘received’ view of the pragmatics–semantics interface, according to which the output of semantics provides input to pragmatics, which then maps literal meaning to speaker meaning. Rather, we should allow neo-Gricean pragmatics/conversational implicature to play a systematic role in ‘pre’-semantics, i.e. to help determine the truth-conditional content of the sentence uttered. As Levinson (2000: 242) put it: There is every reason to try and reconstrue the interaction between semantics and pragmatics as the intimate interlocking of distinct processes, rather than, as traditionally, in terms of the output of one being the input to the other.
8.2 Conventional Implicature In Part I, I discussed conversational implicature. In this part, I examine the second category of implicature put forward by Grice, namely, conventional implicature. 15 Even if the dispute were entirely of a terminological rather than a substantive nature, the force of my argument seems to remain. This is because to have fewer technical terms is normally better than having more. In this connection, see e.g. Horn (2006a, c), Bach (2010), and Carston (2012: 170–171) for a debate on the virtues of the concepts/terms ‘explicature’ versus ‘impliciture’.
176
176 Yan Huang
8.2.1 What is conventional implicature? The German mathematician, logician, and philosopher Gottlob Frege was perhaps the first modern scholar to take note of conventional implicature. His analysis of the Andeutung relation can be taken as a direct precursor of Grice’s concept of conventional implicature (see e.g. Frege 1892 and 1918–1919). By conventional implicature is meant a non-truth-conditional meaning which is not derivable in any general considerations of cooperation and rationality from the saying of what is said, but arises solely because of the conventional features attached to particular lexical items and/or linguistic constructions. A number of standard examples follow (I use ‘+>>’ to stand for ‘conventionally implicate’). (33) p therefore q +>> q follows from p Taroo is Japanese; he, therefore, knows how to use chopsticks. (34) p but q +>> p contrasts with q John is poor but he is honest. (35) Even p +>> contrary to expectation Even many high-ranking communist officials don’t believe in communism. (36) p moreover q +>> q is in addition to p John can read Swahili. Moreover, he can write poems in the language. (37) p so q +>> p provides an explanation for q Mary is taking Chinese cookery lessons. So her husband has bought her a wok. In (33), the conventional implicature triggered by the use of therefore is that being Japanese provides some good reason for knowing how to use chopsticks. In (34), there is a conventional implicature of contrast between the information contained in p and that contained in q licensed by the use of the discourse particle but (Grice 1989a: 25, 88). In (35), even (a scalar particle), being epistemic in nature, conventionally implicates some sort of unexpectedness, surprise, or unlikeness. In (36), the use of moreover brings in the conventional implicature that the statement made in q is additional to the statement made in p (Grice 1989a: 121). Finally in (37), the conventional implicature contributed by so is that the fact that Mary is learning how to cook Chinese food explains why her husband has bought her a wok. Other representative, conventional-implicature-licensing linguistic expressions in English include actually, also, anyway, barely, besides, however, manage to, nevertheless, on the other hand, only, still, though, too, yet, and in spite of the fact. More recently, it has been suggested that speaker/subject-oriented sentence adverbs, epithets, slurs, and other ‘loaded’ words, prosodic features, evidential markers, affected pronouns, definite descriptions, T/V
177
Implicature 177 pronouns, and word order effects be added to the list (e.g. Horn 2012a, b; Williamson 2009; see also McCready 2010).
8.2.2 Properties of conventional implicature Properties of conventional implicatures can best be characterized in contrast to those of conversational implicatures, discussed in section 8.1.2. The main similarity between conventional and conversational implicature is that both are truth-conditionally transparent. In other words, neither makes any contribution to truth conditions. (38) a. We want peace and they want war. b. We want peace but they want war. Here, (38a) and (38b) share exactly the same truth condition, though (38b) contains the conventional implicature of contrast triggered by the use of the connective but. This indicates that like a conversational implicature (but see section 8.1.7), a conventional implicature does not contribute to the truth condition of its corresponding sentence. A second similarity is that both conventional and conversational implicatures are associated with speaker or utterance rather than sentence. On the other hand, there are a number of important differences between conventional and conversational implicatures. First, unlike conversational implicatures, conventional implicatures are not derived from Grice’s cooperative principle and its component maxims, but are attached by convention to particular lexical items or linguistic constructions. They are therefore an arbitrary part of meaning, and must be learned ad hoc. Secondly, unlike conversational implicatures, conventional implicatures are not calculable via any natural procedure, but are rather given by convention, thus they must be stipulated. Thirdly, unlike conversational implicatures, conventional implicatures are not cancellable, that is, they cannot be defeated. Fourthly, unlike conversational implicatures, conventional implicatures are detachable, because they depend on the particular linguistic items used.16 Fifthly, conventional implicatures tend not to be universal. In addition, they tend to project out of embedded contexts and to be immune to certain kinds of objections. They are also characterized by what Potts (2007) called ‘descriptive ineffability’ or contextual variability (Horn 2012a, b). For example, sometimes it is rather difficult to pinpoint the exact semantic contribution made by but; is it symmetric contrast, asymmetric denial of expectation, or correction (e.g. Huang 2014: 74)?17 16
Cf. We want peace. However/nevertheless/on the other hand, they want war. The same can be said of at least some cases of conversational implicature. For instance, the conversational implicatures generated by the use of and are sometimes very difficult to pin down: is it temporal sequence, causal connectiveness, intensionality, contrast, simultaneity, or containment (e.g. Huang 2014: 58–59)? 17
178
178 Yan Huang
8.2.3 Recent analyses It should now be pointed out that unlike the notion of conversational implicature, the notion of conventional implicature, being essentially a recalcitrant residue of the Gricean what is said versus what is conversationally implicated (Horn 2012a, b), may not be a very coherent one. Even Grice himself warned that ‘the nature of conventional implicature needs to be examined before any free use of it, for explanatory purposes, can be indulged in’ (1989a: 46). Horn, himself a leading neo-Gricean pragmaticist, went a step further by claiming that ‘the role played by conventional implicature within the general theory of meaning is increasingly shaky’ (Horn 2004: 6). Since its inception, conventional implicature has been subject to numerous attempts to reduce it to semantic entailment, conversational implicature, and presupposition (Levinson 1983: 128), and more recently, to part of what is said (Bach 1999, who also consigned it to the dustbin of mythology, but see Barker 2003 for a different view), part of tacit performatives (Rieber 1997), vehicles for performing second-order speech acts (Bach 1999), procedural meaning in relevance theory (Blakemore 2004) and view-on-content devices (VCDs) (Vallée 2008, who provided an alternative analysis in Perry’s 2001 reflexive- referential semantics). But recently, Potts (2005) has made a brave attempt to resurrect the concept of conventional implicature. He ‘retain[ed] Grice’s brand name but alter[ed] the product’ (Horn 2007b) by focusing on some non-at-issue contents such as expressive expressions like epithets, attributive adjectives, honorifics, and supplements like non-restrictive relatives, parentheticals, and appositives rather than lexical items such as but, therefore, and even. He isolated four essential properties of conventional implicature. The first of these properties is conventionality—conventional implicatures are part of the conventional meaning of the linguistic expressions involved. The second property is commitment—conventional implicatures are commitments, and thus they engender entailments. The third property is speaker orientation—the commitments are made by the speaker of an utterance. The final property is independence—conventional implicatures are logically and compositionally independent of what is said (see also von Heusinger and Turner 2006). Taking the view that conventional implicature is semantic in nature, Potts developed a logic of the notion by modelling it with a type-driven multidimensional semantic translation language (see Horn 2007b and Feng 2010 for criticisms of this analysis). Feng (2010) presented another development of Grice’s concept of conventional implicature. The properties extracted by him for conventional implicatures are (i) non-truth-conditionality, (ii) speaker orientation, (iii) infallibility, (iv) occurrency, (v) dependency, and (vi) context-sensitivity. He further argued that properties (i)–(iv) are intimately associated with subjectivity. Finally, contrary to Potts, Horn (2007b) maintained that conventional implicature—‘a non-cancellable but truth- conditionally transparent component of encoded content’ (Horn 2012a: 83)—has both a semantic and pragmatic character. This is why a conventional implicature is so named by Grice. According to Horn (2007b: 50),
179
Implicature 179 [conventional implicature] is semantic insofar as it involves an aspect of the conventional meaning of a given expression rather than being computable from general principles of rational behaviour or communicative competence, but it is pragmatic insofar as it involves considerations of appropriateness rather than truth of the sentence in which it appears.18
Whether belonging to semantics or balancing on the edge between semantics and pragmatics, Potts’, Horn’s, and Feng’s recent works have shown that the Fregeo-Gricean concept of Andeutung/conventional implicature is, after all, not that incoherent.
Acknowledgements My thanks go to Bart Geurts for his comments on embedded implicature. For conversational implicature in developmental, experimental, clinic, neuro-, computational, socio-, cross-and intercultural, and interlanguage pragmatics, see the relevant chapters in this volume.
18
As pointed out by Horn (2012a, b), in the retrospective epilogue of Grice (1989a), concepts like conventional implicature are redefined in terms of formality and dictiveness. By the former is meant part of the conventional meaning of a linguistic expression; the latter refers to part of what the linguistic expression says. Construed in this way, formality without dictiveness gives rise to conventional implicature. On the other hand, dictiveness without formality engenders something like explicature (Sperber and Wilson), impliciture (Bach), or truth-conditional pragmatics (Recanati).
180
Chapter 9
Presu pp osi t i on and Giv e nne s s Bart Geurts
9.1 Introduction Presuppositions are pieces of information associated with certain lexical items and syntactic constructions. There are many such ‘presupposition triggers’, as they are called, and the following is just a small selection: (1) Factives Jack regrets that he cheated at the test. (⇝ Jack cheated at the test.) (2) Aspectual verbs Jill has stopped writing poetry. (⇝ Jill has been writing poetry.) (3) It-clefts It was in November that we left for Kiev. (⇝ We left for Kiev.) (4) Wh-clefts What Bruce ate was kelp. (⇝ Bruce ate something.) (5) Focus particles Jill ate kelp, too. (⇝ Somebody else ate kelp.) (6) Quantifiers Jill spoke with all the applicants. (⇝ There were applicants.) (7) Definites The pizzeria in the Vatican is closed on Sundays. (⇝ There is a pizzeria in the Vatican.) A speaker who utters one of the sentences in (1)–(7) commits himself to the truth of the corresponding sentence in brackets. Of course, this does nothing to distinguish
181
Presupposition and Givenness 181 presuppositions from ordinary entailments. An utterance of (6), for example, commits the speaker to the assumption that Jill spoke with all the female applicants, but this is an entailment, not a presupposition. The difference becomes apparent, however, as soon as we embed presupposition triggers in non-entailing contexts: (8) a. Jill hasn’t stopped writing poetry. b. Maybe Jill has stopped writing poetry. c. If Jill has stopped writing poetry, her husband will be relieved. Here (2) is embedded in the scope of a negation operator, a modal operator, and in the antecedent of a conditional, and, remarkably, it appears that the sentences in (8) commit a speaker to the truth of ‘Jill has been writing poetry’ just as much as (2) does. This behaviour sets apart presuppositions from entailments, as the following illustrate: (9) a. Jill didn’t speak with all the applicants. b. Maybe Jill spoke with all the applicants. c. If Jill spoke with all the applicants, she must have a sore throat. Here (6) is embedded in the same environments as in the previous example, and as in (8), the presupposition that there were applicants is part of our understanding of the sentences in (9) just as it is with (6). In contrast, the inference that Jill spoke with all female applicants, which is an entailment of (6), is not licensed by any of (9a–c). Generally speaking, presuppositions tend to escape from any embedded position in the sense that, whenever a sentence ϕ contains an expression triggering the presupposition that χ, an utterance of ϕ will imply that χ is true. This is only generally speaking, because this rule, though correct in the majority of cases, does not hold without exceptions: (10) a. It is possible that Jack thought the matter over, and that he regrets that he cheated at the test. b. It is possible that Jack cheated at the test, and that he regrets that he cheated at the test. (11) a. If Jack had spinach, what Bruce ate was kelp. b. If Bruce ate anything, what he ate was kelp. (12) a. Jill is hungry, and she believes that the pizzeria in the Vatican is closed on Sundays. b. Jill believes that there is a pizzeria in the Vatican, and she believes that the pizzeria in the Vatican is closed on Sundays. A factive verb like ‘regret’ or ‘know’ triggers the presupposition that its complement clause is true, and so (10a) implies that Jack cheated at the test (cf. (1)), which is in line with the observation that presuppositions typically escape from embedded positions. However, (10b) shows that this rule does not always hold. Structurally, (10b) is analogous to (10a), but whereas the latter presupposes that Jack cheated at the test, the former does not. Similarly, it is the case with the sentence pairs in (11) and (12). In (11) the wh-cleft triggers
182
182 Bart Geurts the presupposition that Bruce ate something (cf. (4)), but unlike the structurally identical construction in (11a), (11b) does not seem to preserve this presupposition. Finally, by uttering (12a) a speaker would normally commit himself to the assumption that there is a pizzeria in the Vatican (cf. (7)), but an utterance of (12b) would not entail this commitment. It appears from these observations that presuppositions are normally though not invariably inherited by the sentences in which they occur. This is the so-called ‘projection problem’ for presuppositions: how do the presuppositions of a sentence depend on the presuppositions of its parts? This formulation is a bit rough, though, because strictly speaking presuppositions are not carried by sentences. Presuppositions are made by speakers or, perhaps, utterances. As long as proper care is exerted, it may be convenient to speak of presuppositional expressions as inducing inferences which often are ‘inherited’ by the sentences in which they occur. But it should be kept in mind that this is just a manner of speaking: whenever it is said that sentence ϕ presupposes that χ, what is actually meant is that, normally speaking, a speaker who uttered ϕ would thereby commit himself to the presupposition that χ is true.
9.2 Presuppositions as Given Information To presuppose something is to take it as given: in some sense or other, the presuppositions of an utterance are prior to the main speech act it serves to perform, which might be an assertion, a question, or what have you. Put otherwise, presuppositions are part of the context in which a speech act is performed. But what exactly is a context, and how does this way of characterizing presuppositions help to solve the projection problem, if indeed it helps at all? And come to think of it, isn’t it evident that presuppositions aren’t always contextually given? These are the questions that will engage us in the following.
9.2.1 What’s in a context? The standard answer is that presuppositions are part of the common ground between speaker and hearer: by using an expression which triggers the presupposition that χ, the speaker signals (or acknowledges) that χ is already part of the common ground (Stalnaker 1973, 1974). At any given moment in the discourse, the common ground consists of the information all participants accept as true at that point.1 The common
1
It is often assumed that the relevant notion of acceptance is ‘mutual’ in the following sense: if χ is in the common ground, then:
(a) all participants accept χ;
(b) all participants accept (a);
(c) all participants accept (b);
and so on.
183
Presupposition and Givenness 183 ground waxes and wanes as the discourse proceeds (though, normally speaking, the waxing should predominate), and everything that is common ground, at any given point, can be presupposed at that point. This highly influential view is not without its problems: (13) I know you don’t believe General Amigo has any supporters left, but I tell you he has, and many of his followers are quite influential, too. In this example, the definite NP ‘his followers’ triggers the presupposition that General Amigo has followers, but according to the speaker, the addressee refuses to accept this as true. Nevertheless, despite the fact that this presupposition is expressly stated to be outside the common ground, the speaker’s utterance is perfectly coherent. Examples like (13) suggest that, in the general case, it is better to say that presuppositions are not necessarily given in the sense that they are part of the common ground, but in the sense that they are part of the speaker’s commitment slate: the claims, opinions, wishes, etc. the speaker has committed himself to by virtue of his linguistic and paralinguistic acts (Hamblin 1971; Asher and Lascarides 2008; Schlenker 2010). It may be that examples like (13) are somewhat exceptional, and that in practice, presuppositions are usually contained not only in the speaker’s commitment slate, but in the common ground, too. But that doesn’t alter the fact that there are quite fundamental differences, technically as well as conceptually, between a commitment-based notion of presupposition and the more standard view, according to which presuppositions are common ground, even if the two views are related; for, to a first approximation at least, the common ground is just the sum of all the individual speakers’ commitments. However, in the remainder of this article, we will bypass this issue, and use the term ‘context’ non-committally.
9.2.2 Accommodation On any sensible construal of ‘context’, it is not quite right to say that presuppositions have to be in the context of utterance when a trigger is used, for in many cases presupposition triggers serve to introduce information that is plainly new. Karttunen (1974: 191) gives the following examples: (14) a. We regret that children cannot accompany their parents to commencement exercises. (⇝ Children cannot accompany their parents to commencement exercises.) b. It has been pointed out that there are counterexamples to my theory. (⇝ There are counterexamples to my theory.) c. There are almost no misprints in this book. (⇝ There are misprints in this book.) The italicized expressions trigger presuppositions that may well be new. In (14a), for instance, the factive verb ‘regret’ triggers the presupposition that children cannot
184
184 Bart Geurts accompany their parents to commencement exercises, and it is clear that this sentence may be used felicitously in a context in which the presupposition is not given yet. In fact, a construction like this will often be preferred to a straightforward assertion like (15), which may sound too blunt: (15) Children cannot accompany their parents to commencement exercises. Hence, presuppositions need not be given information. It is more accurate to say that presupposed information is presented as given: a speaker who presupposes something represents himself as supposing that the presupposition is already in the context. Just as one can present old stories as hot news, a speaker can present new information as if it were given. In such a case, the speaker dispenses new information by pretending that his audience already knows. Normally, the pretence will be transparent, and it will be recognized by everyone that this is what is going on: ‘In some cases, it is just that it would be indiscreet, or insulting, or tedious, or unnecessarily blunt, or rhetorically less effective to openly assert a proposition that one wants to communicate’ (Stalnaker 1974: 202). For example, if I arrive late for a meeting, I may say by way of excuse: (16) I’m sorry I’m late, my bicycle broke down. By saying this I presuppose that I have a bicycle. It may be that no one in the audience knew this beforehand, but they will let me get away with my presupposition because it is innocent enough (in my country, it is quite common for people to own bicycles), and because they appreciate that in order to avoid presupposing that I have a bicycle I would have had to use a prolix formula such as: (17) I’m sorry I’m late, I own a bicycle which I used to get here, and it broke down. There are at least two ways of conveying to my audience that I have a bicycle: by using the definite NP ‘my bicycle’ or by stating that I have ‘a bicycle’. In both cases, the information that I have a bicycle may be new, but whereas in the latter case it is presented as such (that’s what the indefinite article is for), in the former case it is presented as given (that’s what the definite article is for). Hence, new information can, and often will, be conveyed by way of presupposition, but it is important to realize that this is an instance of what Grice (1975) calls ‘exploitation’: the speaker exploits the rules of communication by breaking them. This form of exploitation has come to be known as ‘accommodation’ (Lewis 1979; Beaver and Zeevat 2007). From the hearer’s perspective, accommodation is a form of repair. If the speaker presupposes something that is not yet in the context, the hearer may be prepared to go along with the speaker’s pretence that his presupposition is already given, and revise his representation of the context accordingly. In general, however, hearers will only be prepared to accommodate presupposed information that is not particularly newsworthy. For example, whereas in my part of the world, it is common for people to own
185
Presupposition and Givenness 185 bicycles, chariot owners are few and far between, and therefore one would not expect to be told that: (18) I’m sorry I’m late, my chariot broke down. If it was a fact that I own a chariot, it would be a remarkable one, and remarkable facts are better stated than presupposed.
9.3 Satisfaction If we view presuppositions as pieces of information that are presented as given, it is less than obvious how that view might contribute to solving the projection problem. In order to explain how, let us consider the following examples: (19) a. It was Jack who strangled the maid, and Jill shot the bellman. b. If it was Jack who strangled the maid, then Jill shot the bellman. (20) a. Jill shot the bellman, and it was Jack who strangled the maid. b. If Jill shot the bellman, then it was Jack who strangled the maid. (21) a. Someone strangled the maid, and it was Jack who strangled the maid. b. If someone strangled the maid, then it was Jack who strangled the maid. Each of these sentences contains a cleft sentence which triggers the presupposition that someone strangled the maid, and while a speaker who uttered any of the sentences in (19) or (20) would be required to accept this presupposition, he would not be taken to accept it if he uttered either of the sentences in (21). In the case of (21a), we would say that he had asserted rather than presupposed that someone strangled the maid, and (21b) doesn’t imply in any way that the maid was strangled. Intuitively, it is obvious why the presupposition triggered by the wh-cleft is blocked in (21a,b), while it goes through in all the examples in (19) and (20). Apparently, if the second half of a conjunction or conditional is of the form ϕχ (where χ is a presupposition triggered in ϕ), and χ is entailed by the first half, then the presupposition will be blocked; otherwise it will go through. So the following generalizations seem to hold: (22) a. ‘ϕχ and ψ’ presupposes that χ ‘ϕ and ψχ’ presupposes that χ, unless χ is entailed by ϕ b. ‘If ϕχ then ψ’ presupposes that χ ‘If ϕ then ψχ’ presupposes that χ, unless χ is entailed by ϕ We might say that these clauses describe the ‘projection profiles’ of ‘and’ and ‘if … then’, and as it turns out, they are the same. The idea we will develop in the following
186
186 Bart Geurts is that these projection profiles can be explained in terms of the dynamics of discourse contexts.
9.3.1 Context change and satisfaction An utterance changes the context in which it is made. If I say, (23) The Lord is my shepherd. the context will change in a number of ways. For one thing, I have committed myself to the claim that (23) is true. For another, if my audience indicates that they accept my statement (by nodding their consent, for example, or by tacit acquiesence), then it becomes part of the common ground that (23) is true. The crucial insight is that such changes may take place midway in an utterance, and that these intermediate context shifts may be the key to the solution of the projection problem (Stalnaker 1973, 1974; Karttunen 1974). A speaker who utters a sentence of the form ‘ϕ and ψ’ in a context c, changes c in two steps: first c is enlarged to a context (call it ‘c + ϕ’) in which ϕ is true, and then c + ϕ is enlarged to a context in which ψ is true. Hence, the presuppositions required by ϕ and ψ arise in different contexts: presuppositions required by ϕ are taken to be given in c, while presuppositions required by ψ are taken to be given in c + ϕ. Presuppositions triggered in ψ may still impose restrictions on c, but only indirectly. Since ϕ is evaluated in c, any presuppositions triggered by ϕ will have to hold in c. However, ψ is evaluated not in c but in c + ϕ, and therefore a presupposition triggered in ψ that is entailed by ϕ will impose no restrictions on c. All of which is just a roundabout way of saying that the presuppositional asymmetry of ‘and’ is due to the fact that a speaker who is in the process of uttering a conjunction will take for granted the truth of the first conjunct when he starts uttering the second. This account extends to conditional sentences as follows. It is plausible to assume that the antecedent of a conditional statement of the form ‘If ϕ then ψ’, as uttered in a context c, serves to temporarily extend c to c + ϕ; for this is what it means to suppose that ϕ is true, which is what the if-clause does. Therefore, the local context for ψ is c + ϕ rather than c, and any presuppositions required by ψ that are entailed by ϕ will not be ‘inherited’ by the conditional statement as a whole. Let’s say that if a sentence of the form ‘ϕ and ψ’ or ‘If ϕ then ψ’ is uttered in context c, ϕ’s local context is c and ψ’s local context is c + ϕ. Now the foregoing observations may be summed up as follows: (S) Satisfaction If a speaker utters a sentence which contains an expression triggering a presupposition χ, then χ must be entailed by its local context. (If this condition is not met, accommodation may be called upon to ensure that χ is entailed by its local context after all.)
On this view, a presupposition must be given in the sense that it is entailed by the local context in which it is triggered. In order to distinguish it from other notions of
187
Presupposition and Givenness 187 givenness to be considered below, we will henceforth say, following common usage, that a presupposition must be satisfied by its local context.
9.3.2 Issues concerning accommodation and satisfaction We have arrived at a pragmatic account of presupposition projection which hinges on the idea that presuppositions are given information and the assumption that the context of interpretation may change, not only as the result of, but also in the course of, an utterance. However, what we have so far is little more than a first step in the direction of a full-fledged theory of presupposition. Let us have a brief look at some of the issues that arise when we start going in that direction. Our working hypothesis is that a presupposition must be satisfied in its local context, and if it isn’t, it must be accommodated. To a first approximation, this is just to say that the presupposition must be added to the context before we proceed to interpret the remainder of the utterance. For example, if (24) is uttered in a context c, and c doesn’t satisfy the presupposition that Jack has a lawn, then the hearer is expected to interpret the asserted content of (24) in c + ‘Jack has a lawn’. (24) Jack is mowing his lawn. This idea is simple enough as long as we proceed from the idea that there is just one context per utterance, but in the meantime we have helped ourselves to an enriched notion of context, as a consequence of which different parts of an utterance may be interpreted in different contexts. Thus we are faced with the question of where presuppositions must be accommodated. To see that this is a non-trivial problem, consider the following sentence: (25) If he isn’t with Jill, Jack is mowing his lawn. Suppose that (25) is uttered in context c and that c doesn’t satisfy the presupposition that Jack has a lawn. Then the local context in which this presupposition is triggered is c + ‘Jack isn’t with Jill’; call this local context c′. Now, in which context is the presupposition to be accommodated: c or c′? If it is added to c, it is accommodated globally, and the interpretation we obtain for (25) can be paraphrased as follows: (26) Jack has a lawn. If he isn’t with Jill, Jack is mowing his lawn. Not exactly a paragon of style, but it captures the most natural interpretation of (25) well enough. The second option is to accommodate the presupposition locally, in c′, which results in a reading along the following lines: (27) If he isn’t with Jill, Jack has a lawn and he is mowing his lawn.
188
188 Bart Geurts This isn’t pretty, either, but more to the point: the implication that Jack has a lawn is lost, and (25) as a whole is construed without this presupposition. In this case, global accommodation gives us the interpretation we want, but in other cases, local accommodation may be called for. For example, if someone calls me on the phone, and introduces himself as the Pope, I may utter (28): (28) Sure, and if you’re the Pope, I’m the Chinese Empress. (29) a. If you’re the Pope, China has an Empress and I’m the Chinese Empress. b. China has an Empress. If you’re the Pope, I’m the Chinese Empress. In this case, local accommodation gives us the reading in (29a), which seems correct, and since it is a well-known fact that China is a republic, it is clearly preferred to the global-accommodation reading in (29b). Since, in general, presuppositions can be accommodated in more than one context, we need a way of deciding when to accommodate globally and when to accommodate locally.2 Since this issue was first discussed (by Heim 1982), there has been a consensus that global accommodation is the norm, and that local accommodation is resorted to only if global accommodation would yield an unlikely reading. This proposal leads us to expect that (28) is a special case, which it clearly is. So far, so good. Things are about to become murkier, though, for the notion of accommodation is more problematic than one might think. In our discussion of (25), we said that the presupposition that Jack has a lawn is preferably accommodated in the global context c rather than its local context c′ = c + ‘Jack isn’t with Jill’. Sure enough, this gives us the reading we would like to have. Unfortunately, however, this does not follow from the view on presupposition expressed in (S). On that view, the presupposition that Jack has a lawn (call it χ) must be satisfied in c′, and if it isn’t, c must be adjusted (we’re opting for global accommodation) so as to guarantee that χ is satisfied in c′. The problem is that, in order to achieve this, we don’t have to add χ to c; for it suffices that we increment c with a weaker proposition, to wit: (30) If Jack isn’t with Jill, then he has a lawn. This is the weakest proposition that, when added to c, suffices to ensure that c′ satisfies χ. Which is to say that the global-accommodation reading predicted for (25) is not (26) but rather: (31) If Jack isn’t with Jill, he has a lawn. If he isn’t with Jill (and therefore has a lawn), Jack is mowing his lawn. 2
A third option is ‘intermediate accommodation’, which targets a non-global context between the global context and the local context of the presupposition. This option is contentious, because not all theories of presupposition projection allow for this possibility, which has prompted some authors to claim that intermediate accommodation doesn’t exist (Beaver and Zeevat 2007).
189
Presupposition and Givenness 189 This may sound potty, but actually it is merely too weak: instead of presupposing that Jack has a lawn, we end up with a conditionalized, and therefore weaker, version of that presupposition. This is known as the ‘proviso problem’ (Geurts 1996, 1999a). It may not be entirely clear, at first, that the proviso problem is much of a problem. For if χ is not satisfied in its local context, isn’t it obvious that we want to accommodate it in the global context? Yes, that’s what we want. But the point is that, given the satisfaction view on presupposition, it would be ad hoc to do so, for it follows from (S) that a weaker presupposition should suffice, which it doesn’t. Therefore, the search for a solution to the proviso problem continues (Beaver 2001; Singh 2007; van Rooij 2007; Schlenker 2011).
9.4 Binding According to the view outlined in the foregoing, presuppositions must be given in the sense that they must be satisfied by their local contexts. In this section, we will see that there is another notion of givenness that may be implicated in the behaviour of at least some presupposition triggers. Around 1990, it was observed that there are striking similarities between two parts of the pragmatic forest that had escaped notice for two decades: presupposition projection and anaphoric binding (van der Sandt 1989, 1992; van der Sandt and Geurts 1991; Kripke 2009). Note, to begin with, that many of the examples that are routinely used to explain the projection problem sound somewhat redundant, and are markedly improved by deploying anaphoric expressions: (32) a. Jill looks good, and she knows {it/that she looks good}. b. Having smoked for nearly 25 years, Jack suddenly decided to quit {∅/smoking} last month. Clearly, the longer versions of (32a,b) sound more cumbersome than the shorter ones, but more importantly, their meanings are the same. These observations already suggest that presupposition and anaphora are kindred phenomena. The following examples will serve to tease out the similarities a bit further: (33) a. Bruce wore lenses for years, but he doesn’t wear them anymore. b. Jack bought a melon, but left it in his car. (34) a. If Bruce ever wore lenses, he doesn’t wear them anymore. b. If Jack bought a melon, he left it in his car. In (33a), the particle ‘anymore’ triggers the presupposition that Bruce used to wear lenses, but the sentence as a whole doesn’t presuppose that he did, because the first
190
190 Bart Geurts conjunct explicitly asserts what the second conjunct presupposes; the presupposition triggered in the second half of the sentence is satisfied by the first half. Now consider (33b). In this example, the pronoun ‘it’ in the second conjunct enjoins the hearer to find a suitable referent, which as it happens was introduced in the first conjunct, by way of the indefinite ‘a melon’: the discourse entity associated with ‘it’ is bound to (or equated with) that introduced by the indefinite NP (Karttunen 1971; Kamp 1981; Heim 1982). The examples in (34) are similar. On the one hand, one of the key elements of the projection problem is to explain why a presupposition triggered in the consequent of a conditional is ‘blocked’ if its content is made explicit in the antecedent: in (34a), this is what happens with the presupposition that Bruce used to wear lenses, which is triggered by the particle ‘anymore’. On the other hand, a long-standing issue for theories of anaphora is how a pronoun in the consequent of a conditional can be bound by an indefinite NP in the antecedent; this is what (34b) illustrates (Geach 1962). These and other observations have inspired the idea that anaphora is a special case of presupposition (van der Sandt 1992; Geurts 1999a). On this view, presupposition triggers want to be bound in the same sense that anaphoric pronouns want to be bound: they introduce discourse entities that must be given in the local context in which they occur. Anaphoric pronouns are special only in that, in general, they are difficult to interpret by way of accommodation, but even this is a difference of degree rather than kind: whereas (35a) is infelicitous unless the context proffers a suitable referent for the pronoun, it may be argued that the pronoun in (35b) is interpreted by way of accommodation (Geurts 2012): (35) a. I thought it was Vernon. b. When the doorbell rang I thought it was Vernon. In (35b), the pronoun refers to the person who rang the doorbell, who wasn’t mentioned before, so this qualifies as an instance of accommodation. If this view is correct, then the reason why the presuppositions in (33a) and (34a) are not ‘inherited’ by the sentences in which they occur is that they are bound sentence- internally, just like the pronouns in (33b) and (34b) are bound. On this account, presuppositions are still given information (or at least presented as such), but ‘givenness’ now has a stricter meaning than before: it doesn’t suffice that presuppositions are entailed by the local context; rather, presupposition triggers want to be linked to discourse entities that are available in the local context. This difference has consequences for the notion of accommodation, too. For example, suppose that (36) is uttered in a context in which it is not given yet that Jack has a lawn, so that ‘his lawn’ has to be interpreted by way of accommodation: (36) Jack is mowing his lawn. (= (24))
191
Presupposition and Givenness 191 On the view expounded in the last section, accommodation requires that the context be revised so that it comes to entail that Jack is a lawn owner. On the view currently under consideration, accommodation necessitates the introduction of a new discourse entity to represent Jack’s lawn. Which is not the same thing, though it may be that the latter notion of accommodation implies the former. Apart from the fact that it offers a unified treatment of presupposition and anaphora, another appealing feature of this view is that it doesn’t run into the proviso problem. If presuppositions are discourse entities of some sort or other, then, given that the pronoun refers to Jack, ‘his lawn’ in (37) requires that Jack’s lawn (or, more accurately, some discourse entity representing Jack’s lawn) be contextually given: (37) If he isn’t with Jill, Jack is mowing his lawn. (= (25)) If Jack’s lawn is not given, there is, again, more than one context in which it might be accommodated, but there is a crucial difference with the satisfaction approach: we are not merely trying to ensure that the local context of ‘his lawn’ entails that Jack has a lawn; rather, we want to ensure that Jack’s lawn is available in that context. Therefore, on the present account, it is fully expected that accommodation involves incrementing a context with the integral presupposition as triggered by (say) a definite NP. Hence, if the presupposition triggered by ‘his lawn’ in (37) is accommodated globally, we obtain the most natural reading of this sentence without further ado: (38) Jack has a lawn. If he isn’t with Jill, Jack is mowing his lawn. (= (26)) On the view outlined in the foregoing, presuppositions and pronouns are the same kettle of fish, the only difference being that the descriptive content of the latter tends to be poorer than that of the former; but that is merely a matter of degree (van der Sandt 1992; Geurts 1999a). This view seems especially apt for definite NPs, like ‘his lawn’ or ‘the pizzeria in the Vatican’, which are widely agreed to be closely related to referential pronouns; and the difference between the pronoun ‘it’ and a attenuated description like ‘the thing’ seems very slight indeed. But how plausible is the more sweeping claim that, in general, anaphora is just a special case of presupposition? There are at least some instances of presupposition that, prima facie, seem to belie this view. Consider, for example, the well-worn ‘bachelor’ case: (39) Bruce is a bachelor. It is standardly assumed that (39) merely asserts that Bruce is not married, and that it presupposes that Bruce is an adult male. If this is a presupposition, it wouldn’t seem to be the kind of thing that wants to be bound in the same way a pronoun wants to be bound (Geurts 1999b). In this case, at least, the satisfaction view seems more natural.
192
192 Bart Geurts What about other presupposition triggers? Let us consider some of the variety, starting with: (40) The judge examined every document. (⇝ There were documents.) Universal quantifiers have been counted amongst the presupposition-inducing expressions since Strawson (1952). Clearly, (40) will tend to be read as implying that there were documents, and this inference projects as any bona fide presupposition would (as we saw in the introduction). However, it is also clear that a mere existential inference doesn’t quite capture what (40) seems to presuppose, for this sentence requires that some set of documents be contextually given, or otherwise it will sound odd. It appears, therefore, that this presupposition needs to be bound in much the same way as a pronoun needs to be bound. The same applies, mutatis mutandis, to the presuppositions triggered by focus particles like ‘too’: (41) 17 is prime, too. (⇝ Some number other than 17 is prime.) With focus on the subject term, (41) implies that 17 is not the only prime number. But again, thus formulated, the presupposition is too weak, and indeed it will be a truism in many contexts. An utterance of (41) would normally require that at least one prime number is contextually salient, and if this condition isn’t met, the sentence will sound odd. In this case, too, it seems that the presupposition is not merely existential and wants to be bound. In contradistinction to the last two examples, the presupposition triggered by a factive verb does not require a contextual antecedent: (42) is a perfectly acceptable way of announcing, inter alia, that something was wrong: (42) Jill knew that something was wrong. (⇝ Something was wrong.) Prima facie, at least, it seems that factive presuppositions are just existential, and need not be given in any stronger sense. Let’s take stock. In the foregoing, we introduced two notions of givenness. One is purely existential, and merely requires that some proposition be entailed by the context. The other might be called ‘referential’, and requires that some discourse entity be contextually available. We surveyed preliminary evidence that, of the many inferences that pass the standard projection test for presuppositionhood, some need to be given in the second sense, whereas for others it suffices if they are given in the first sense. It must be stressed, though, that the evidence is preliminary, and that the classification of presupposition triggers is a hairy matter if anything is. There seems to be a widespread feeling, in the more recent literature, that presupposition is not a unified phenomenon, but there isn’t anything like a consensus on how to proceed from there (Zeevat 1992, 2002; Geurts 1999b; Beaver and Geurts 2010; Abrusán 2010; Simons et al. 2010). Still, for definites the referential model seems to be the most appropriate, and in the following we will assume that it is.
193
Presupposition and Givenness 193
9.5 Specifics One of the hallmarks of presuppositions is that, if they aren’t satisfied or bound, they tend to take scope over anything else in the sentence.3 In general, this property makes for a stark contrast between definite and indefinite NPs, with the latter having a strong preference for being interpreted in situ. However, indefinite NPs can have a so-called ‘specific’ reading, and on this reading they behave very much as if they were definite (Givón 1978; Fodor and Sag 1982; Geurts 1999b; Matthewson 1999). The first indefinite in (43) is a case in point: (43) Last week, several students were harassed by a 43-year-old man posing as a casting agent. Though in principle it may be possible to read this sentence as being non-committal as to whether all the students were harassed by the same man, normally speaking it would be understood that they were. In general, the distinction between definite and indefinite NPs is clear enough: while the purpose of definite NPs is to refer to entities that are given (in some sense or other), indefinite NPs introduce new discourse entities; and while definite NPs routinely escape from the scope of any operator, indefinite NPs usually stay put. Specific indefinites blur this pattern. In particular, a specific indefinite seems to have much the same interpretative effects as a definite that is construed by way of accommodation. This is the reason why I propose to dwell on specific indefinites for a while, even if this may seem to take us away from our main topic. Let’s have a closer look at the commonalities between definites and specific indefinites. In many languages, these two categories are bracketed together in some respect or other. In English, for instance, the nominal constituent of a partitive PP must be definite or specific; non-specific indefinites and quantified NPs are not allowed in this position (Ladusaw 1982): (44) Jack is one of {the/several/*most/*all/*sm/*∅} employees who will be fired.
Here ‘sm’ represents unstressed ‘some’, which has a distinct preference for a non-specific reading, like the bare plural, indicated by ‘∅’. Perhaps the most telling evidence for the kinship between definiteness and specificity is that in language after language they are lumped together into the same morphosyntactic rubric. I will give a handful of more or less arbitrarily chosen examples. In Bemba, a Bantu language, there is a class of nominal prefixes of the form ‘consonant– vowel’, and another class of the form ‘vowel–consonant–vowel’. The former are used 3
I am using the notion of ‘scope’ merely to characterize the sort of readings that tend be associated with certain expressions, leaving it open whether or not these readings can be accounted for in terms of scope-taking, in the grammatical sense of the word.
194
194 Bart Geurts to mark non-specific indefinites, while the latter alternatively convey definiteness or specificity. The following examples are from Givón (1978); here and in the following glosses are reproduced from the original source: (45) a. Umu-ana a-a-fwaaya vcv-child he-past-want
ci-tabo. cv-book
‘The child wanted a book (be it any).’ b. Umu-ana t-a-somene ci-tabo. vcv-child neg-he-past-readcv-book ‘The child didn’t read a/any book.’ c. Umu-ana a-a-fwaaya vcv-child he-past-want
ici-tabo. vcv-book
‘The child wanted the book’ or ‘The child wanted a specific book.’ St’át’imcets (Lillooet Salish) features an indefinite article which can only occur within the scope of a negative expression, a question, a modal, and so on. In the absence of such operators another article must be used, which has a specific-definite function. The following examples are from Matthewson (1999): (46) a. Cw7aoz Neg
kw-s Det-Nom
áts’x-en-as see-Tr-3Erg
ku Det
sqaycw. man
‘S/he didn’t see any men.’ b. * Áts’x-en-as see-Tr-3Erg
ku Det
sqaycw. man
‘S/he saw a man.’ (47) a. Húy-lhkan going.to-1sg.Subj
ptakwlh, tell.story
ptákwlh-min tell.story-Appl
lts7a here
ti Det
smém’lhats-a … woman.Dimin-Det
‘I am going to tell a legend, a legend about a girl …’ b. Wa7 Prog
ku7 Quot
ílal cry
láti7 Deic
ti Det
smém’lhats-a woman.Dimin-Det
‘The girl was crying there.’ The foregoing observations should suffice to show that many languages treat definiteness and specificity as related notions, which together stand in opposition to non-specific
195
Presupposition and Givenness 195 indefiniteness. How are we to interpret this pattern in the light of our discussion of presupposition and givenness? We have assumed that, de jure, the definite article signals givenness: by using an expression of the form ‘the N’, the speaker presents an individual as given. However, we have seen that, de facto, the referent of a definite NP is often new, in which case the hearer is expected to accommodate a suitable discourse referent before he proceeds to interpret the speaker’s utterance. This goes some way to account for the kinship between definites and specific indefinites: both types of expression will serve to introduce new referents into the context, though in the case of definite NPs that is not their proper function. However, this can only be part of the story, for it doesn’t explain why specific indefinites seem to project in much the same way as definites do. In the following section, we will see that this may be just part of a much bigger issue, and gesture in the general direction where a solution might be found.
9.6 Not at Issue Ever since they were introduced by Frege (1892), presuppositions have been viewed as ancillary to another type of content, which Frege identified as the ‘thought’ expressed by the sentence (in modern parlance this would be the ‘proposition’). In the recent literature, more pragmatic dichotomies are commonly preferred, and presupposition is variously opposed to ‘foregrounded information’ (Geurts 1999b), ‘assertion’ (Abbott 2000; Horn 2002c), the ‘main point’ of an utterance (Abrusán 2010), what is ‘at issue’ (Simons et al. 2010), and so on. Although the finer points vary from case to case, it seems to me that the same core intuition underlies all these proposals, and I will focus on that, mixing terminology ad libitum. By uttering a sentence a speaker usually conveys a considerable amount of information, only a small portion of which is central to his concerns. The remainder is less important vis-à-vis the speaker’s main point: it merely serves to anchor the foregrounded information to the context, or is entered en passant. Backgrounded information is not necessarily unimportant, but it is of secondary interest in relation to what is foregrounded. Thus the notion of background is primarily a negative one: backgrounded information is what is left when we abstract out the speaker’s main point. It may well be, therefore, that it is impossible to provide a single positive description covering all ways of backgrounding. What does this have to do with projection? The answer is seemingly straightforward (Geurts 1999b; Simons et al. 2010): Buoyancy Backgrounded information projects. Actually, this simplicity is quite deceptive, because all sorts of issues are lurking in the wings. Unfortunately, however, we’ll have to keep them lurking, and confine our attention to the essentials. While there is a broad agreement that the behaviour of at least some presupposition triggers can be explained in terms of Buoyancy, or something very much like it, that is where the consensus ends, and different authors would paint
196
196 Bart Geurts the general picture in rather different ways. In the following I will paint my own version. In my view, a crucial characteristic of the notion of background is that it doesn’t entail givenness in either of the senses discussed above; only the converse is true. Backgrounded information may be given, or presented as given, but new information is not necessarily foregrounded. For example, enclosing new information in (intonational or orthographic) parentheses often serves to indicate that it is of secondary importance, which is to say that it is backgrounded, but not that it is presented as given in any strong sense. If backgrounded information need not be given, there is no reason why it couldn’t be marked as new. My suggestion is that this is not just a logical possibility: backgrounded information sometimes is marked as new, for this is precisely what specificity comes down to. Intuitively, specific indefinites carry information that is used to set the stage for the utterance’s main point. Like presuppositions, specific indefinites are separate units of communication, to be integrated into the context before the remainder of the utterance is processed; this is why semantically incorporated indefinites cannot be specific (van Geenhoven 1998). If this view is on the right track, then the projection behaviour of the first indefinite in (48) is due to the fact that ‘a 43-year-old man posing as a casting agent’ is interpreted as separate information which is of secondary importance to the sentence’s main point, and will therefore project out, courtesy of the Buoyancy principle: (48) Last week, several students were harassed by a 43-year-old man posing as a casting agent. ( = (43)) On this view, specificity is a pragmatic phenomenon. Returning to the topic of presupposition, one of the main attractions of the Buoyancy principle is that it may help to explain why presuppositions are triggered. Thus far, we have taken it for granted that certain expressions trigger presuppositions while others do not, and in some cases it is arguable that that is all the explanation we are going to get. For example, it seems quite likely that the definite article is a presupposition trigger because that’s its job; likewise for focus particles like ‘too’. In many other cases, however, it seems that presupposition triggering is not just a matter of convention. Consider, for example, lexical inferences like the following, which have often been said to be presuppositional in nature: (49) Bruce is a bachelor. (⇝ Bruce is a man.) ( = (39)) A naive account of this inference would be to suppose that the lexical content of ‘bachelor’ falls into two parts: an assertional part which specifies that ‘bachelor’ is truthfully predicated only of unmarried individuals, and a presuppositional part which says, among other things, that a bachelor is a man; of course, it is the second half of the content of ‘bachelor’ that triggers the presupposition in (49). There are several problems with this naive account. First, as it stands, this analysis implies that every occurrence of
197
Presupposition and Givenness 197 ‘bachelor’ gives rise to the presupposition that the individual it is being applied to is a man, and therefore it predicts, for instance, that (50) Jill is allergic to bachelors. means something like, ‘Jill is allergic to unmarried individuals who are presupposed to be men’, which is not what we want. The solution to this problem is fairly obvious: the word ‘bachelor’ should only be allowed to trigger its presupposition when it is being used predicatively. But this seems to entail that ‘bachelor’ is ambiguous between a presupposing and a non-presupposing reading, which is not exactly an appealing consequence. The second problem, which is related to the first, is the following. Suppose that it is encoded in the lexicon that predicating ‘bachelor’ of some individual a carries with it the presupposition that a is a man. Consider now how the words ‘bachelor’ and ‘man’ are related to each other: the former is a hyponym of the latter, and the only distinctive feature of the word ‘bachelor’ is that it applies to unmarried individuals. But at the same time that is all we are saying, as opposed to presupposing, when we call somebody a bachelor. Could this be a coincidence? I think it is pretty clear that it is not. For one thing, other hyponyms behave alike: ‘spinster’ presupposes ‘female’, ‘woodpecker’ presupposes ‘bird’, and so on. For another, an intuitively plausible story about this phenomenon is readily available. If a speaker utters (49), it is likely that the essential bit of information he intends to convey is that Bruce is married, not that Bruce is an adult male. Therefore, the information that Bruce is a man is backgrounded, which implies, courtesy of the Buoyancy principle, that it will project.
9.7 Where Are We Now? It is not very clear where the foregoing discussion leaves us. We started, as is standard, by operationalizing the concept of presupposition in terms of projection: the hallmark of presuppositions, we said, is that they exhibit projection behaviour. But in the meantime we have seen that all that projects is not a presupposition; for example, specific indefinites give rise to projection effects too, but we wouldn’t want to say that they are presupposition triggers. Furthermore, we have seen that distinctions have to be made within the class of canonical presupposition triggers, too. For example: • Some presuppositions are easier to interpret by way of accommodation than others (section 9.3). • Whereas some presuppositional expressions invite a binding analysis, which brackets them with anaphors, others resist such a treatment, and are more naturally construed in terms of satisfaction (section 9.5). • Whereas some presuppositions are triggered lexically, others may be induced non-conventionally, e.g. by something along the lines of the Buoyancy principle (section 9.5).
198
198 Bart Geurts Apparently, there is quite a bit of taxonomizing to do. In my view, this is one of the main problems that theories of presupposition are currently facing. A further issue is partly contingent on the first: while for most of this article, we have been supposing that presupposition is to be analysed in terms of givenness, the observations of the last two sections undermine that assumption. If it is true that a principle like Buoyancy explains the projection behaviour of at least some of the traditional presupposition triggers, then in these cases there is no essential connection between presupposition and givenness. Of course, there may be a non-essential connection: it may be that backgrounding is most naturally construed as givenness. But since not all backgrounded material is given (or even just presented as given), givenness cannot be called upon to explain why information projects. Which leaves us with the last issue, the most fundamental of them all: Why do presuppositions and their kin project? To the best of my knowledge, we don’t have a principled answer to this question, yet. We have not succeeded in answering all our problems—indeed we sometimes feel we have not completely answered any of them. The answers we have found have only served to raise a whole set of new questions. In some ways we feel that we are as confused as ever, but we think we are confused on a higher level and about more important things. (Kelley 1951: 2)4
Acknowledgement This research was supported by a grant from the Netherlands Organization for Scientific Research (NWO), which is gratefully acknowledged.
4 According to quoteinvestigator.com, this is the earliest attested occurrence of a phrase that, over the years, has been attributed to a variety of sources, the most famous of which is Enrico Fermi, .
199
Chapter 10
Speech Ac ts Stephen C. Levinson
10.1 Introduction The concept of speech act is one of the most important notions in pragmatics. The term denotes the sense in which utterances are not mere meaning-bearers, but rather in a very real sense do things, that is, perform actions. This is clear from a number of simple observations: (i) utterances in conversation (and that is the only kind considered in this article) respond not to the shape or meaning of what was said, but to the underlying ‘point’ or action performed by the prior turn at talk, which might have been expressed in any number of ways; (ii) utterances often have non-verbal counterparts (cf. waving to saying hello; bidding at auction by hand or voice); (iii) utterances interdigitate with non-verbal actions in action sequences (cf. ordering a sandwich in a service encounter); (iv) utterances have real-world consequences just like non-verbal actions (a $1,000 bid at an auction commits you to paying; saying you have nothing to declare in an airport can get you a big fine). These actions are on a different ontological plane than the actions of the vocal organs in speech, which of course activate the motor cortex just as much as reaching for a glass—speech acts are more like moves in chess, whose meanings are circumscribed by rules and expectations. Trying to understand how utterances can have these abstract action-like properties, how they are coded linguistically, and how we recognize them are some of the core issues in this domain. Despite the fact that speech acts are clearly central to an understanding of language use, they have been largely off the linguistics agenda since the 1980s. As is often the case in science, research on speech acts boomed for a little over a decade (in the 1970s
200
200 Stephen C. Levinson and 1980s), and then went out of fashion without the most fundamental issues being resolved at all. Amongst these unanswered questions are: How many types are there, and are they universal or culturally specific? How are they expressed in language? And how are they recognized or attributed in actual language use? These questions are addressed in sections 10.3–10.9 below.
10.2 A Brief History of the Concepts Leading to the Current State of the Art In philosophy of language during the 1930s and 1940s the picture theory of meaning, and the broader correspondence theory of truth, began to be challenged by theories of language use being developed by the later Ludwig Wittgenstein at Cambridge and the ‘ordinary language’ philosophers like Gilbert Ryle and J. L. Austin in Oxford. It is Austin who is usually credited with the first developed theory of speech acts, although his influential lectures ‘How to do things with words’ were not published until 1962 after his death (Austin 1962a). Austin took the view that philosophy of language had wrongly concentrated on statements, or even just propositions, and in doing so had lost track of what language is mostly used for. Rather, he claimed, utterances attempt to do things, and just like other actions can fail for a range of reasons. He catalogued the kinds of actions performed, by noting that most speech acts (however colloquially expressed) can be paraphrased in the normal form ‘I hereby Vperformative’ where a delimited set of verbs like order, promise, warn, congratulate could appear. He also classified the reasons for success or failure of speech acts, dubbed ‘felicity conditions’, noting that they often require appropriate subjective states (later called ‘sincerity conditions’ by Searle) as well as appropriate circumstances (Searle’s ‘preparatory conditions’). In this sort of way all the reasons for my bid at Christie’s for a Picasso not succeeding (I am not a registered bidder, lack the funds, don’t succeed in getting the attention of the auctioneer, etc.) can be spelled out. Speech acts can be understood on the analogy of ceremonies, like marriage or toasting the monarch’s health—in the same sort of way they are conventional arrangements for creating new states of affairs, and consequently are in principle open-ended in kind. Austin went on to notice that these success conditions not only parallel truth conditions, but actually subsume them; statements are therefore just a special class of speech acts with sincerity conditions of belief and presuppositions or preparatory conditions that must also be met. He also went to some pains to clarify all the different senses in which actions could be said to be performed by utterances: the ‘locutionary act’ is the saying of the words with the intended meanings, the ‘illocutionary act (or force)’ is the speech act proper (ordering, advising, warning, etc.), and the ‘perlocutionary act’ is the further act or consequences that are context-specific and not part of the specific conventions invoked (e.g. by asking your advice I might flatter you).
201
Speech Acts 201 Austin also developed a number of notions whose importance was not immediately realized—for example, the concept of ‘uptake’ (the ratified receipt and recognition by a recipient). Austin’s work was influentially systematized by John Searle (1969), who connected the theory to sociology and jurisprudence on the one hand (speech acts are built as constitutive rules, whereby doing X counts as constituting a new state of affairs, like scoring a goal, or being guilty of a specific crime), and to linguistics on the other hand. Noting, following Hare (1952), that the same propositional content could occur across speech acts (as in ‘Pass the exam’, ‘Did you pass the exam?’, ‘Good luck with the exam’), he added a ‘propositional content condition’, so that the felicity conditions together now effectively defined the speech act. He went on to suggest that an exhaustive typology of speech acts could be arrived at by clustering types of felicity conditions, so that there can be seen to be just five main types: representatives (statements and the like), directives (questions, requests, orders), commissives (threats, promises, offers), expressives (thanking, apologizing, congratulating, etc.), and declarations (like christening, declaring war, firing, etc. which rely on elaborate institutional backgrounds). Searle’s theory was well articulated and proved attractive to linguists, as recounted below. Meanwhile, other philosophers took a more psychological view of language use, chief among them Grice and Strawson, who both thought that speech acts should be thought about as specific classes of intention, e.g. intentions to cause beliefs in addressees, or intentions to get them to do things. Grice (1957, 1975) reconstructed the notion of meaning along these lines, and characterized the use of language in conversation as guided by rational action between partners. Although he never laid this out in print, it is clear that he thought that felicity conditions simply follow from the specific classes of intention: if I want to get you to pass the water by saying ‘Could you pass the water?’, it would simply be irrational if I didn’t want the water, if the water is not in your reach, if you are deaf or otherwise preoccupied. This intentional perspective was followed up by work in natural-language processing that related speech act recognition to plan recognition (see section 10.7). During the period of generative semantics, linguists became increasingly interested in language usage and how sentences might encode aspects of the contexts in which they are used. Searle and other theorists had not concentrated on the actualities of speech act coding, presuming instead that illocutionary force is coded in the major sentence types (imperatives, interrogatives, and declaratives) and in the explicit performative verbs when so used—these would be the ‘literal illocutionary forces’ of utterances. But as any practical grammarian of English or other languages knows, in fact one has to learn idiomatic means of expressing speech acts. Gordon and Lakoff (1971) noted for the first time that ‘indirect speech acts’ could also routinely be expressed by querying or stating a felicity condition: ‘Do you need that pencil?’, ‘Could I have that pencil’, ‘Is that your pencil?’, ‘I’d like that pencil’ all query or state a precondition on requesting. They also noted that adverbials like please or frankly might force a particular speech act reading (as in ‘Please could we begin on time?’). There followed a large literature on indirect speech acts, investigating the forms used especially for requests across cultures,
202
202 Stephen C. Levinson the psychological processing (indirect speech acts seemed to be processed without any complex detour through a literal meaning), and the politeness reasons for the mismatch between direct and indirect speech act coding. By the end of the 1980s, however, linguistic interests had moved largely elsewhere. Meanwhile, a completely different approach, unrelated to the linguistic and philosophical traditions, was being taken in sociology, where the empirical study of conversation was being born in the late 1960s and early 1970s. Unencumbered by theory, the conversation analysts (Harvey Sacks, Manny Schegloff, and Gail Jefferson especially) were observing all sorts of fundamental organizations for interactive language use: turn-taking, repair, and sequence organization (see e.g. Schegloff and Sacks 1973, Schegloff 2007a, and this volume). In doing so, they were finding speech acts that had no vernacular names, no associated performative verbs or (it seems) special markings, for example pre-closings (e.g. the exchange of wells before goodbyes in phone calls), assessments (evaluations of shared events or things), repair initiators (like excuse me?), pre-invitations (What are you doing on Friday night?), and so forth. Such actions (as the conversational analysts call them, treated here as equivalent to the notion of speech act) can only be understood against the background of sequential position—that is, where they come with relation to prior or following turns. Despite the fact that many observations have now accrued about the sets of actions and their sequential placement, little systematic theory about actions has emerged from this work (for a survey see Levinson 2013a; Drew 2013). Although this brief review cannot do justice to the extensive work that has been done in the different disciplines interested in speech acts (linguistics, psychology, conversation analysis) (see Levinson 1983, 2013a), it serves as a pointer to the state of the art. There is general acceptance of the importance of the subject, but little recent research that advances our understanding of the fundamental questions.
10.3 The Essential Insight and the Leading Issues In contrast to the emphasis in modern linguistics on language as a device for an endless sound–meaning correspondence, J. L. Austin’s core insight was that the central function of language is not to deliver meanings but to deliver speech acts. For the core ecological niche for language, and still its primary use and the locus of its acquisition, is conversation. Each of us produces on average perhaps 16 000 words and 1200 turns at talk a day—and each turn delivers a speech act: all in all we are participating in exchanges with something like 5000 speech act moves a day. In order to respond on time (within the c.200 ms allowed by the turn-taking system; Stivers et al. 2009) we need to decode or attribute speech acts at lightning speed, because it is the illocutionary force, not the meaning, that we primarily respond to. One of the central puzzles is that speech
203
Speech Acts 203 acts are not for the most part simply or directly coded in the linguistic form: for example, Where are you going? could be an idle question, or a challenge, or a reprimand, or a prelude (a pre-) to a request for a ride or to an offer to give you a ride, and the relevant response depends on the correct attribution. How then are speech acts recognized in the tight time-frame allowed? Is there a finite list of possible action types, or can they be created de novo? Further, as just illustrated, an utterance or turn can perform more than one action simultaneously: in asking a question (Where are you going?) the speaker could also be transparently performing a pre-request in such a way that the addressee can make an offer in next turn (Downtown, would you like a ride?). How many acts can be performed at once? These then are the central puzzles in this area, to be taken up below. Faced with these difficulties, to which current research yields no definitive answers, it is tempting for linguistic theory to simply hand over the can of worms to some other discipline (conversation analysis, for example) as e.g. Bierwisch (1980) recommended. However, as discussed in section 10.8, there is a substantial intersection of speech acts and linguistic structure, which makes the topic of central importance for e.g. the study of syntax. Usage and structure in fact go hand in hand.
10.4 The Nature of the Beast: Identifying Speech Acts In this section we consider the problem of identifying and cataloguing speech acts, given some problematic properties, like their implicit character and non-one-to-one mapping onto utterances. There are four (three basic and one related) approaches to identifying or characterizing speech acts. First, one could rely on natural metalanguage, as in English offer, request, invitation, greeting, and so on. Austin’s own tack here, recollect, was to do the lexicography of performative verbs (I hereby declare/choose/delegate/promise/undertake/bequeath … ). But there are many reasons to distrust natural metalanguage. Many speech acts have no vernacular names (such actions as pre-invitations, continuers, repair initiators, and the like), as discovered by the conversational analysts. In addition, while written languages often have large metalanguage resources of this kind, unwritten ones often do not, and they may have speech acts alien to us. So natural language terms are a poor guide. A second approach is the use of felicity conditions to characterize speech acts, as in classical speech act theory. A problem here is that taken as necessary conditions jointly sufficient to define speech acts, it is hard to specify them right. Thus the conditions for genuine information-seeking questions, exam questions, questions checking facts, and questions used in repair will all be subtly different—they form a loose family of speech act types not easily captured by a definitive checklist of conditions.
204
204 Stephen C. Levinson A third approach favoured by conversation analysts is to use the character of responses to identify prior actions. For example, if a range of utterances X–Y–Z are all immediately responded to by fellow interactants passing the speaker something, then prima facie X–Y–Z are requests. The observation is that many speech acts come in pairs (‘adjacency pairs’), with an initiating action having a characteristic response, as in greetings followed by greetings, offers by acceptances (or declinings), questions by answers, and so forth (Schegloff 2007a; Stivers 2013). Thus if one can independently characterize the responding action, one can type the eliciting action. Conversation analysts argue that this is how we check that we are understood—we expect a response of a certain type. Consider, the following example, where the response marked by thanks and excuses suggests that for B, A’s turn appears to have been an offer, though that is not obvious from its structure or content: (1) A: She says you might want that dress I bought, I don’t know whether you doB: Oh thanks, well, let me see I really have lots of dresses (Levinson 1983: 335) A fourth, related approach is to appreciate that an utterance gets parts of its identity from the sequential position it occupies. Consider the following tokens of the utterance Okay, each doing entirely different things (labelled here with the action codings used in conversation analysis—see Schegloff 2007a): (2) a. N: Y’wanna drink? C: Yeah N: Okay (← Sequence-closing third) b. C: Okay (← Pre-closing) R: Okay [Bye C: Bye c. B: How are you? A: Okay (← Answer) One aspect of speech acts thus highlighted is that they are necessarily interactional in character. Consider a proposal (say about going for a walk together)—for success, the action depends on the uptake: it takes two to tango. This is a fundamental aspect of speech acts neglected in Searlian analysis—almost all speech acts are joint actions (Clark 1996).1 Most analysis actually makes use of all four of these different kinds of identifying properties, trading on our vernacular terminology, trying to tighten it up by defining criteria, considering how participants themselves respond to utterances, and noting how utterances play different roles depending on their positioning vis-à-vis other speech acts. 1
A possible exception are ‘outlouds’ or ‘response cries’ like private exclamations (Goffman 1978), which may be produced with or without an audience, but by definition without an addressee.
205
Speech Acts 205
10.5 The Inventory and its Universality A natural question is how many kinds of speech acts are there? The question presumes a level of abstraction away from the specific propositional content, which may of course be unique: it’s a question about how many types of illocutionary force exist. Austin suggested an open-ended list, convention-based, so cultural in nature. In contrast Grice (in unpublished work: Grice 1973; see also Schiffer 1972) had suggested that complex speech act types could be built up from the two propositional attitudes of wanting and judging. His target was the ‘moods’ expressed in the major sentence types, namely declaratives, imperatives, and interrogatives. Most languages grammatically code at least two of these, which could be taken as a hint of a cross-cultural core of basic speech acts. However it is moot whether these forms really code speech acts since they are in practice used for diverse action types, while other minor sentence types like English expressives more directly code for force (see section 10.6). But the idea that speech acts fall into classes of intention is persistent (see e.g. Tomasello 2008). Searle, taking an intermediate position, has argued that there are in fact just five large classes of things one can do with language—five major speech act types. The classification uses three parameters: the ‘essential conditions’ (Searle’s term for the intentional goal), the sincerity conditions, and ‘direction of fit’ (whether the words copy the world as in statements or the world copies the words as in promises). Searle’s classes are representatives (assertion-like), directives (questioning, requesting, etc.), commissives (promising, threatening, offering), expressives (thanking, apologizing, etc.), and declarations (blessing, christening, etc., which rely on special institutional backgrounds). Searle’s classification cannot however be exhaustive. First it fails to accommodate many of the actions noted by the conversation analysts (e.g. the continuer hmhm, the pre-s, the repair initiators and the repair responses, and so forth). Second, it is culture- bound. Consider the following exchange simplified and in translation from the language Yélî Dnye (Levinson 2005): (3) A: He’s yelling into a bit of bush-knife B: He’s yelling under a mangrove tree This is an adjacency pair of a special kind peculiar to this matrilineal Papuan culture, in which men make jokes by alluding to some unfortunate accident or event that befell the other man’s father-in-law, to which the response must be immediate and in kind (B’s father-in-law killed his wife and then himself with a bush-knife, while A’s father-in-law died falling from a mangrove tree; they are ostensibly commenting on a man yelling down a megaphone). These utterances are paired father-in-law jokes and they don’t describe states of affairs or express the feelings of the speakers or otherwise fall within Searle’s taxonomy. In addition, Searle’s classification is of course a higher-order grouping of types, so it will not help us understand the specifics of action and response in conversation.
206
206 Stephen C. Levinson Austin or Searle’s armchair classifications are based on intuitions about salient types of speech acts. These are nearly always first parts of (base) adjacency pairs (see Schegloff 2007a, this volume)—that is, the initiating actions (like questions, offers, invitations) to which responses are due (even then, many such initiatory actions have proved relatively unavailable to intuition, like repair initiators, continuers, assessments, and the like). But the actions that lead in to these initiators (e.g. pre-announcements, pre-closings) or the responses themselves (e.g. answers, agreements, continuers, counter-offers), or the actions that interpose between first part and second (e.g. clarification questions) escape proper treatment in classical speech act theory. Consider (with arrowed action labelling): (4) 1. D: Didju hear the terrible news? 2. R: No. What 3. D: Y’know your Grandpa Bill’s brother Dan? 4. R: He died. 5. D: Yeah
← Pre-pre-announcement ← Answer + Go-ahead ← Pre-announcement ← Guess ← Confirmation (Terasaki 1976)
Describing line (1) as a question would miss its basic function, namely to check whether a news announcement should be made; line (2) makes clear it should (note the what); line (3) sets up the topic of the announcement in such a way than no announcement proves necessary, for the recipient guesses in line (4). Thus although (1) and (2) could be said to be questions that is not their main function, which is as preliminaries to an announcement (see Levinson 1983: 345–364; and Schegloff 2007a for more on pre-s). Recollect as mentioned above that conversation analysts have emphasized that it is the character of the response, or the locus in a sequence, that plays a major role in giving speech acts their identities. To return to the central questions of this section: Is there a finite set of speech act types, and if so how big is it? The answers are that we really don’t know. Is the set universal in character? Not in the sense that all speech acts are pan-cultural (witness Yélî Dnye father-in-law jokes, or any of the institutionally circumscribed acts like finding guilty, proposing toasts, declaring war, etc.), but it is an open question as to whether there is a pan-cultural core with such plausibly general functions as telling, questioning, requesting, greeting, agreeing, or initiating repair.
10.6 The Multiple Action Problem One particularly troubling feature of the mapping of speech acts onto utterances is that such a mapping is not necessarily, or even mostly, 1:1. Sometimes turns at talk have
207
Speech Acts 207 more than one constructional component, and each part can perform an action, as in the previous example (4) above and in (5): (5) A: How are you= B: =Fine. How are you? ← Answer and Question But often a single constructional unit (whether or not it exhausts the turn) can do more than one action (as in (4) where Didju hear the terrible news? might be said to be a question, but carries with it the obligation to tell the news, conditional on the answer ‘no’). Consider the following example from a verbal tussle between a mother and her 14-year-old daughter Virginia wanting more allowance or pocket money: (6) Virginia VIR: MOM: VIR: MOM: VIR:
But—you know, you have to have enough mo:ney¿ I think ten dollars’ud be good. ← Proposal (0.4) .hhh Ten dollahs a week? ← Repair-I, Q, Pre-challenge Mm hm. ← Repair, A, Go-ahead Just to throw away? ← Repair-I, Q, Challenge and Pre-rejection (0.5) Not to throw away, to spe:nd. ← Repair, A, Defence
Viriginia’s proposal is responded to by a question-like response, which has the form of an other-initiator of repair or OIR (i.e. is initiated by the responder, seeking repair on the prior turn). But it is a prosodically incredulous OIR, adumbrating an upcoming challenge (call it a pre-challenge), which after a go-ahead, is duly delivered (Just to throw away?) but again in the form of a question inviting repair. That extreme formulation of the question in turn prefigures a rejection (call the turn then a pre-rejection), and gets a defence. And so forth. But now notice we have multiple layers of function for each turn—up to four actions packed into the one subclausal turn in Just to throw away! The question that arises is whether there is any limit to the number of actions that a single turn can bear. Notice that some of these might merely be a matter of granularity of description, e.g. a special kind of question is often used to ask for repair. But that is not the kind of relation between the question and, say, the challenge: notice how the response deals with both. The literature acknowledges the existence of turns performing two actions: on one account, a ‘literal speech act’ is used to deliver an ‘indirect speech act’ (Searle 1975), and conversation analysts talk about one action being the vehicle for one other action (Schegloff 2007a). But there is no explanation for turns that perform three or more actions (see, however, the suggestions in terms of plan reconstruction at the end of the next section).
208
208 Stephen C. Levinson
10.7 Bottom-Up and Top-D own Inference in Speech Act Recognition and Attribution Speech acts, it has been suggested, are not easy to individuate or identify, are not known to come from a finite or universal set, and can be laminated one on top of another. These are problematic properties. But an even greater problem is how they are recognized (more properly attributed2) under the extraordinary time pressures of spoken conversation (or any other interactional use of language). Here we concentrate on the comprehension problem. As already mentioned, on average across languages the gaps between turns are on the order of 200–300 ms (Stivers et al. 2009; Levinson and Torreira 2015). Given that the fastest response from conception to word takes 600 + ms (Levelt 1989; responses of any complexity, e.g. three or more words, take 900–1500 ms or more to prepare), it is clear that speakers in conversation predict the end of the incoming turn in order to launch their own response on time. But that response must ‘type’ the incoming turn, as e.g. a question, request, statement, before it has finished in order to compose the relevant response and launch it so it comes out on time. Probably this is done on average about halfway through the incoming turn (see Magyari et al. 2014). This makes the speed at which speech acts are attributed appear quite miraculous. For, as already made clear, the coding of speech acts is for the most part not directly marked: most syntactic forms, even whole constructions like Why don’t you …, are multi-duty (why don’t you turns out to code proposals, advice, invitations, and complaints, while Do you want codes requests, invitations, offers, and so forth; Couper-Kuhlen 2010). Speech act recognition is similar to any perception problem, where pattern has to be discerned and categorized out of noise. Both ‘bottom-up’ information (in the signal) and ‘top-down’ information (expected categories) are usually involved, and the noisier the channel the greater the role for ‘top-down’ factors. Let us consider them in turn. Bottom-up information is whatever clues to speech act type can be found directly coded or cued in the signal, by lexical choice, construction, or prosody. Given the turn-taking facts, it is clear that signals early on in a turn are going to be more important than signals at the end of turns, since by then the choice of response must have already been made. This suggests that effective cues will be ‘front-loaded’, coming early in the turn (see Levinson 2013a). Here the cross-linguistic facts are curious. Take the grammar of interrogatives, associated (though not exclusively) with the illocutionary force of questioning. First, wh-or content interrogatives are only grammatically initial in about
2 ‘Recognition’ presupposes correct attribution that matches speaker intent, but since we are interested in the comprehension process which will include occasional misattributions, ‘attribution’ is the more accurate term.
209
Speech Acts 209 one third of languages (Dryer 2011b); however, this is the dominant single strategy since the alternative positions are various, and Dryer notes that only ‘a few languages exhibit at least a weak tendency to place interrogative phrases at the end of sentences’ (he mentions two out of a sample of 900 languages). These facts are in line with the ‘front-loading’ prediction from the psycholinguistic facts, but only as a tendency. The prediction would be that languages with late (right-located) wh-words would have developed compensatory cues like prosody or particles positioned earlier in the clause. Second, take polar (yes/no) questions (Dryer 2011a). The commonest coding strategy (60 per cent of languages) is by particle, and of these about 30 per cent are in initial or second position; however the commonest position of particles is final (50 per cent of all particle types). It is worth noting, however, that 30 per cent of languages have no lexical or morphosyntactic coding at all for polar questions, relying solely on intonation or prosody. These facts do not seem to be in line with the ‘front-loading’ expectation. Further light is thrown on these issues by studies of usage in corpora. In a study of ten languages, we found that those sentence-final particles are omitted or absent 40 per cent of the time in Lao and 70 per cent in Korean (Enfield et al. 2010); two of the languages lacked any coding (including prosodic); and morphosyntactic coding as in English inversion is also mostly omitted. One can conclude that polar-question marking must carry low functional load, wherever it is located. These usage studies also showed that interrogatives (whether content or polar) only perform the function of seeking new information about 30 per cent of the time; around 40 per cent of them are involved in repair or checking or confirming just-given information, and the remaining 30 per cent perform many different functions, including offers, requests, and so on. To summarize so far: there is no one-to-one match of form to function. Even where apparently dedicated morphosyntactic machinery exists to code speech acts (as in interrogatives), the coding may be omitted: about 60–70 per cent (in various corpora) of English polar questions are unmarked declaratives in form, and do not carry rising intonation (Geluykens 1988). Cross-linguistically, the tendency is for two thirds or more of all questions (in a broad sense) to be polar questions (unpublished data from Stivers et al. 2009). Even though wh-or content questions would seem to require a wh-form, this is not necessarily true; many languages have indefinite quantifiers that double as interrogative words, and many allow gaps to code the variable (as in John is going to _? instead of Where is John going?). There are then distinct limits to the bottom-up coding and inference of speech act force. Nevertheless, some detailed studies suggest that underlying the apparent many- to-many correspondences between utterance forms and speech acts there might be a clockwork system. For example, in a study of requests in English telephone calls, it was found that the Can you/Could you/Would you … forms are used for requests where the speaker has clear rights or entitlements and knows what the request would involve; where the entitlements are low and the contingencies involved less clear, the I wonder if form is preferred (Drew and Curl 2008). This suggests that where multiple forms are available, they may each carry subtly different presuppositions about background conditions.
210
210 Stephen C. Levinson Nevertheless, it is more likely that the cues to illocutionary force are multiple and probabilistic in character. Indeed, there is now considerable work in natural-language processing (NLP) that seems to show this. This work takes speech corpora, usually from task-oriented dialogues, and tags them by hand with a very constrained set of speech act categories that seem to reflect the functions in each particular corpus. Machine-learning algorithms are then trained on a subcorpus, inducing the association between surface cues—lexical items, phrases, or intonation—and the pre-coded tags. The algorithm is then let loose on the rest of the corpus to see how well it emulates the human tagging. So, for example, it was found that ‘assessments’ (value judgements like ‘That was great’ that usually call for a response in kind) have quite restricted elements (Goodwin 1996): that as subject in 80 per cent of cases, intensifiers really or pretty, and adjectives drawn from a short list including great, good, nice, wonderful … etc. (Jurafsky 2004). So a combination (an unstructured list) of surface cues may be a crude but very effective trigger for speech act categorization: the chances of being an assessment given just one cue like really might be low, but in combination with that and great may be greatly increased. This would be just the kind of low-level associative process that could rapidly deliver probabilities of speech act assignment in comprehension, and since these cues are distributed throughout the turn, an incoming turn could be incrementally classified with increasing certainty. Turning to top-down information, this includes all the accumulated contextual and sequential information that forms the niche for the incoming turn. For example, in service encounters, the goals for speaker and addressee will be largely pre-set, so that an utterance like Do you have coffee to go? can be understood directly as a request. In free conversation, though, the context is usually more local. One factor of constant relevance is the current state of the common ground between participants. We noted earlier that polar questions in English and many other languages are typically unmarked, and thus have the shape and often the prosody of declaratives. How then can they be understood as questions? As Labov and Fanshel (1977) pointed out, the recognition is done on the basis of knowledge asymmetry: thus You’re hungry is likely to be understood as a question, while You’re smart is likely to be interpreted as a compliment. Statements about what the other knows best are candidate questions, and this explains how a fifth of languages can do without any lexical or morphosyntactic marking of polar questions (prosody may often help of course, but in some languages it seems never to play this role; see e.g. Levinson (2010) on Yélî Dnye or Dryer (2011a) on Chalcatongo Mixtec). Epistemic asymmetry or symmetry is such a strong indicator that it can overrule interrogative marking: thus Isn’t it a beautiful day is not likely to be interpreted as a question, since we can all be presumed to have access to the weather. Heritage (2012) argues that epistemic status trumps question marking in all cases. A second always relevant factor is sequential location in the sequence of turns. The power of sequential location to map illocutionary force onto utterances can be appreciated from a number of angles. Consider as a limiting case silence, where there is literally no signal, yet the silence can imply a response, as in the following example where the two-second silence is taken to imply ‘no’ and functions to block a forthcoming request:
211
Speech Acts 211 (7) C: I was wondering would you be in your office on Monday (.) by any chance? (2.0) ← (Pre-request won’t go through) C: Probably not (Levinson 1983: 320) The inference relies on the ‘conditional relevance’ of a second pair part and on the principle that dispreferred responses are typically delayed or mitigated. Another way to appreciate the power of sequence to attribute speech act force is to consider cases where ambiguities arise, as in the following example (8) where the arrowed turn is ambiguous (Schegloff 1988). It could be a straight question, or it could be a pre- announcement—that is, an offer to tell conditional on the recipient indicating that he doesn’t know the indicated news. Note that the question force is not the ‘literal force’ (a question about knowledge), but a question about who is going. Pre-announcements often have this form (cf. Do you know the joke about the plumber?) and the pre- announcement reading is encouraged by the context, where Russ had produced a pre- announcement just before in the first line, and Mom could be reciprocating in kind. The ambiguity comes about because both readings are salient in the context. (8) Russ: I know where you’re goin’,Mom: Where? Russ: To the uh (eighth grade)= Mom: = Yeah. Right. Mom: Do you know who’s going to that meeting? ← (speech act ambiguous turn) Russ: Who? Mom: I don’t kno:w! Russ: O::h probably Missiz McOwen en … A related type of high-level information can also be brought to bear on the interpretation of a turn, namely an assessment of how the turn fits into the likely goal structure or plan of the speaker. For this is the inference schema we use to understand any sequence of actions: if you are sitting opposite and grasp your mug and lift it up, I’ll expect you to put it to your mouth and take a drink. The sub-actions I see (grasping the mug, lifting it) are preconditions to the action I infer (taking a sip), and seeing the initial parts I can make the metonymic inference to the whole. Interestingly, the same pattern of inference works for speech acts. Consider the following service encounter in example (9), where a precondition to buying pecan Danish pastries is queried, and the seller responds both to the question and the underlying request. (9) C: Do you have pecan Danish today? S: Yes we do. Would you like one of those?
← Q + (Pre-)request ← Answer ← deals with request
(Merritt 1976)
Notice however that no request has been issued, so how exactly does this work? Consider the analysis sketched in (10), in terms of customer C’s plans and the seller
212
212 Stephen C. Levinson S’s reconstruction of them from the first utterance in the sequence. From Do you have pecan Danish today the seller can infer that this is a precondition on asking for some, therefore the request is likely to follow—given which the seller can truncate the sequence as she does, by responding to the foreseeable forthcoming request (in the dotted box in the figure in (10)). It is this projected request that gives Do you have pecan Danish today its pre-request flavour; in this way speech acts can acquire multiple actions mapped onto one turn by virtue of projectable next actions. (10) Plans underlying speech acts in example (9) C: Request Danish preconditions
Check availability
Request
C: “Do you have pecan Danish today?”
+ Pre-Request
S: “Yes we do.
C: “I’d like one of those”
Would you like one of those”
Notice this account explains why mentioning a felicity condition on a speech act is one way of performing that speech act (this is the classical theory of ‘indirect speech acts’, as in Searle 1975). But it has much wider application. Consider the telephone exchange in (11): the caller C in line 3 queries what the recipient is doing, which is a potential prequel to an invitation. The response in line 4 not only answers the query but at the same makes clear that there is no impediment to an invitation, thus projecting an acceptance. The lamination of actions throughout this sequence is straightforwardly explicable in terms of current action plus foreseeable next action, as sketched in the figure in (12). (11) 1. C: 2. R: 3. C: 4. R: 5. C: 6. R: 7. C:
Hi Hi Whatcha doin’. Not much. Y’wanna drink? Yeah Okay.
← ← ← ←
Q + Pre-invitation A + Go-ahead for invitation Q + Invitation A + Acceptance
213
Speech Acts 213 (12) Plans underlying speech acts in example (11) Go out with Clara
Invite
Check availability
“Whatja doin’?” ←Q
+ Pre-Invite
“Not much” ←A
“Y’wanna drink?” + Go-Ahead
Be available
“Yeah” Accept
Go out with Nelson
The virtues of this mode of analysis become especially clear when one considers cases like the following where the main actions are projected, but never actually performed. (13) D:
M:
‘hh My ca:r is sta::lled ← Announcement of problem ((5 lines omitted)) I don’ know if it’s po:ssible, but (0.2 hhh) ← Unvoiced Request for ride see I haveta open up the ba:nk.hh (0.3) a:t uh: Brentwood?hh= =Yeah:-en I know you want-(.) ← Unvoiced Rejection en I whoa-(.) en I would, but I’ve gotta leave in about five min(h)utes. (hheh)
Here there is no feasible ‘indirect speech act’ in terms of classical felicity conditions: there is rather an indication of a predicament which would have an obvious solution, while the recipient produces an account for why the obvious solution cannot be performed. In the same sort of way, in example (6), Mom’s Just to throw away? performs four actions, as question, repair initiator, challenge, and pre-rejection because it is transparent that Mom intends to resist Virginia’s claim for more weekly pocket money by countering Virginia’s every move. Neither indirect speech act theory nor the conversation analyst’s notion of one action being the ‘vehicle’ for another (as in Schegloff 2007a) can explain this kind of quadruple depth of speech act lamination on a single turn.
214
214 Stephen C. Levinson Plan reconstruction as an account of speech act comprehension was first advanced by Allen (1979) and Cohen and Perrault (1979) and applied to the problem of indirect speech acts by Allen and Perrault (1980) (see also Clark 1979, Levinson 1981). These approaches in classical Artificial Intelligence style make use of the heavily intentional approach favoured by Grice and reviewed in section 10.1, cranking through a calculus of desire and belief to arrive at a final ‘indirect speech act’ (Cohen et al. 1990). The insights can be understood, however, in a slightly different way, in terms of an utterance being designed to reveal, variously, the whole or part of the iceberg of underlying interactional goals, where projectable next turns serve to laminate one or more ‘indirect speech acts’ onto the current turn. Both bottom-up cues, which may be just probabilistic associations of linguistic features and speech acts, together with top-down factors like the role of sequence, epistemic asymmetries, and plan attribution, almost certainly play a role together in speech act comprehension. Curiously, cases where interlocutors misunderstand one another as in (8) are vanishingly rare. But there is no complete model of how these various kinds of information come together in action attribution.
10.8 Syntax, Sentence Types, and the Grammar of Speech Acts We return now to the grammar of speech acts. We’ve noted that in general there is no one-to-one mapping between form and function. This is especially true of the ‘big three’ sentence types, declarative, interrogative, and imperative, which are probably best seen as carrying a very general semantics (e.g. a wh-interrogative expresses an open proposition with a blank constituent, which is why the same form may double as an indefinite expression in many languages). However, as discussed above under the rubric of cues, there can be many surface elements that will help to narrow down an illocutionary force. There are for example adverbs like please that unambiguously mark requests or pleadings, adverbs like obviously or frankly that mark statements (Gordon and Lakoff 1971), and interjections like Wow, My God that mark exclamations. In addition there are minor sentence types that are indeed specialized for illocutionary force (Sadock and Zwicky 1985). A classic case are exclamatives, where English has rich specialized constructional resources as in What a beautiful day!, That it should come to this!, Why, if it isn’t the trouble maker!, You and your linguistics!, Of all the stupid things to do!, To think I nearly won a medal! (well described in grammars like Quirk et al. 1989). Exclamatives are a category of some typological interest (see Michaelis 2001, who defines them semantically and finds them often coded in quasi- interrogative or topic constructions or NP complements). Similarly, English codes wishes as optatives (If only I’d done it, May the best man win, Oh to be in England), and suggestions or proposals in special forms (How about joining us?, What if you came earlier?, Let’s go, Why not have a drink?). Many other languages have their own specialized forms for warnings, blessings, and the like. Unfortunately, studies of the usages of these forms are still few and far between, so we cannot be sure they are as specialized in usage as the grammars suggest—but it is an important subject for future research.
215
Speech Acts 215
10.9 Conclusions— The Centrality of Speech Acts The central function of language, it has been argued, is to deliver speech acts (Searle 1972). The rest of the linguistic apparatus, with all of its complex syntax and propositional structure, is there to serve this purpose. For speech acts are the coin of conversation, and conversation the core niche for language use and acquisition. A retort might be that the central function of missiles is to target explosives, but this doesn’t help one understand much about the inner complex engineering of a missile—the outer function can be remote from design details, partly because there may be innumerable different engineering solutions that would answer the same function. Linguistics then would be effectively autonomous from the study of speech acts. What has been argued here, however, is that such a disjunction is unlikely to be tenable. First, language design has to accommodate to the tight constraints of conversation, so that speech acts have to be decoded early partly from bottom-up aspects of the signal—hence constructions of many different kinds serve this purpose, if often in a non-deterministic way. Second, the very clausal structure of language is almost certainly due to the tight turn constraints into which sentences must fit, where each turn must deliver at least one speech act. Third, whatever one’s views on the origin of language, short turns delivering speech acts were almost certainly a design feature of protolanguage—languages have evolved within this ecological niche, spinning complexity in the tight confines of the turn. Another way to appreciate the centrality of speech acts in language design is to appreciate how many of the features we think of as most intimately connected to language structure are actually also exhibited in the sequential organization of speech acts. Consider recursion, argued by Chomsky (2007, 2010) to be the most central design feature exclusive to language. Now consider that the clearest type of recursion, namely centre-embedding, is restricted in language to just two, occasionally three, levels of nesting. Karlsson (2007) found no examples of triple embedding in huge corpora, and just 13 in the whole history of Western literature; for spoken language, the limit is two. Since small numbers of centre-embeddings can easily be modelled with a finite state device, there is poor evidence for the need for phrase structure grammars here. Yet centre-embedding within discourse shows none of these limits, and is sufficiently multiple and routine to provide a much better basis for escalation to phrase structure grammars. Here is a simple example of one-degree centre-embedding: (14) Merritt (1976) A:
May I have a bottle of Mich?
B:
Are you twenty one?
A:
0.1
B:
No
No
216
216 Stephen C. Levinson Since this can be recursively elaborated, we could express the indefinite recursion by the rule: Q&A →Q (Q&A) A (Levinson 1981, 2006; Koschmann 2010). The following shows an example with degree-three internal embedding (each level numbered), a level exceeding all syntactic embedding in spoken languages (the speech acts, or adjacency pairs, here relevant are request + compliance, question + answer, and two repair initiator + repairs). (15) Merritt (1976) S: Next 0C: Roast beef on rye 1S: Mustard or mayonnaise? 2C: Excuse me? 3S: What? 2 3 3C:1 Excuse me? 2: I didn’t hear what you said 1S: Do you want mustard or mayonnaise/ 1C: Mustard please. 0S: ((provides))
← Request to order ← Order ← Q1 ← Repair Initiator (RI1) ← Repair on RI ← RI2 ← Q1 = Repair ← A1 ← Compliance with order
It is easy to show that degree-six or more centre-embedding occurs in spoken dialogue (see Levinson 2013b). When one finds a domain where a capacity is more evolved than in another domain, there is reason to assume that it has a longer evolutionary history. While short-term memory constraints are often invoked to explain our failure to produce centre-embedding in syntax, these do not seem to be a constraint in the interactive domain. This would suggest that linguistic recursion at least partly originates from this type of push-down stack in action sequencing, which as far as we know is universal in dialogue. Incidentally, it is also possible to show that cross-serial dependencies can be found in the sequential structure of speech acts, showing once again that complexity attributed to syntax may be more easily found in dialogue structure. All in all, a better case can be made for the need to climb the Chomsky hierarchy of grammars based on speech acts in dialogue than on syntactic structure. For all the reasons outlined in this article, speech acts are a fundamentally important area of study in the language sciences. Work in this domain has been relatively, and inexplicably, neglected since the 1970s and 1980s, and it is time for a renaissance of work on speech acts and their use in dialogue.3
3
My thanks to Penelope Brown and Kobin Kendrick for helpful comments on the manuscript.
217
Chapter 11
Deixis and t h e Interact i ona l Fou ndati ons of Referenc e Jack Sidnell and N. J. Enfield
11.1 Introduction All reference involves directing the attention of some other person to something. The something to which attention is directed may or may not be present in the immediate context of interaction. Whether the referent is a hilltop in plain view, a bird’s singing, Gottlob Frege, sorrow, the ideas of Augustine, or the concept of liberty, making reference requires bringing the recipient’s attention in line with that of the speaker. If human cognition is fundamentally intentional in the sense of being about or directed towards something, reference is a form of shared intentionality in which the cognitive focus of two or more persons is aligned and jointly focused. In deictic reference, this directing of attention is accomplished by relating an object of reference to some aspect of the event of speaking—the indexical origo (Bühler 1982 [1934])—via a ground. So for instance when I point to a book and say ‘this one’ in response to the question ‘Which are you reading?’, my recipient’s attention is directed to the book by relating it to my location, and specifying the relation as one of relative proximity (or immediacy of access— see Fillmore 1982; Hanks 1990, 2005). In this chapter we develop an account of deixis that builds from its simplest manifestation in acts of gaze-following. For humans, gaze-following results from a basic propensity to attend to the attention of others. Because co-present others are able to control their own gaze and other visible signs of attention they can actively manipulate another’s attention such that what was a cue becomes a signal (see Krebs and Dawkins 1984). Pointing and all other forms of deixis (indeed all forms of reference) exploit this
218
218 Jack Sidnell and N. J. Enfield propensity by actively directing others’ attention. Of special importance to our account is so-called lip-pointing in which a meta-communicative facial expression (conveyed by a configuration of lip and or head; Sherzer 1972, Enfield 2009: ch. 3) indicates that a participant’s gaze direction is, at that moment, to be understood as an intentional, communicative signal. With shared intentionality as a foundation, all languages have developed systems of deictic markers: for example, demonstratives such as English that and this. These systems display a defining semiotic property of human communication, namely the use of signs that not only have meanings in themselves, but whose meanings are enriched through relations of opposition and contrast with other elements of the system, such that each element has a composite meaning, a combination of what it is and what it is not. Simple systems in the domain of deixis feature a semantically marked form in opposition to an unmarked one. More complex systems involve multiple dimensions of contrast. A further way in which the meanings of elements of a deictic system may be enriched is through their mapping onto the local socioculturally constituted worlds of their users. Speakers use deictic forms to refer to locally relevant features of the environment and deictic systems are interwoven with the sociocultural world in complex and sometimes counter-intuitive ways. An overarching question to be addressed is, ‘What’s special about deixis as a form of reference? How does it differ if at all from reference accomplished by non-deictic means, and what consequence does this difference have for its function or use in actual situations of social interaction?’ In order to address this question, we begin by developing an account of deixis that is rooted in basic, instinctive human propensities for (a) intentional, goal-directed behaviour and (b) the capacity for two or more individuals to share attention. Together, these human capacities provide a basis for the collective or shared intentionality that underwrites all forms reference, including reference accomplished via the use of deixis. We then turn to briefly sketch the semantic domain describing the essential elements of deictic reference and some of the documented typological variation. Much of the literature in this area focuses on just these issues and so here we do little more than provide a thumbnail sketch and point to relevant landmarks. We then consider demonstrative reference in which the recipient’s attention is directed either by talk, gesture, or gaze to some enumerable thing. Here we show that deixis is a low-cost, high-efficiency, minimally characterizing way to accomplish reference. These features surely account for many of its uses in interaction. But we suggest that referrers select deixis for reference for reasons other than efficiency. First, the semantically general character of deictic forms makes them well-suited for reference to hard-to-describe and/or nameless objects. In such a situation a deictic form can exploit features of the artifactual environment including the presence of the thing being referred to. Second, the semantically non-specific, minimally characterizing features of deixis allow speakers to avoid description where such description may be counterproductive to some interactional goal. Third, because these forms require for their interpretation the application of knowledge in common ground (shared knowledge), successful reference via such a form can be
219
Deixis and the Interactional Foundations of Reference 219 a demonstration of social proximity—an informational enactment of intimacy (see Enfield 2006).
11.2 Directing Attention in Deixis At about nine months of age human infants begin to engage in a suite of joint attentional behaviours such as gaze-following and joint engagement with objects. These behaviours differ markedly from those of younger infants which are primarily dyadic. At about this age, ‘infants for the first time begin to “tune in” to the attention and behavior of adults toward outside entities …’ (Tomasello 1999: 62). We can think of gaze-following schematically as in Figure 11.1. In following the gaze of another, a human infant is attending to that other’s attentional state. Essentially the infant is treating the other’s gaze direction as a sign and their own gaze redirection is an interpretant of that sign (see Kockelman 2005). Importantly, however, gaze-following of this kind occurs at least partially independently of whether the other intended their own gaze to function as a communicative signal. The initial gaze redirection then may function to prompt an infant’s gaze redirection either as a signal or a cue. As Tomasello notes, it is at around this same age—nine months—that infants also begin to direct adult attention to things using deictic gestures such as pointing, or by holding up an object to show it to someone. So at least ontogenetically there seems to be a correlation between the emergence of gaze-following and the emergence of deictic pointing and showing. There is also a clear conceptual connection between gaze-following and deictic pointing. Milgram and colleagues (1969) showed, somewhat inadvertently, that gaze- following in adults was sensitive to the character of the stimulus. The study showed that larger crowds of gazing individuals were more likely to promote gaze-following than smaller crowds. The basic, apparently instinctive, propensity of humans to follow the gaze of others is then available for manipulation—altering aspects of the stimulus/sign will make gaze-following by others more likely. This, then, allows us to see the connection between gaze-following and ‘true’ pointing, one version of which, often referred Infant
Parent
Object
Figure 11.1 Basic structure of gaze-following
220
220 Jack Sidnell and N. J. Enfield to as ‘lip-pointing’, is done with gaze—indeed, it is essentially ‘gaze-pointing’. Figure 11.2 is a still image from video taken by Niclas Burenhult of a Jahai speaker in Malaysia lip-pointing. As Enfield (2001: 186) writes, in relation to a study of lip-pointing among speakers of Lao, the term ‘lip-pointing’ ‘should not be taken to suggest that only the lips are involved…. Additional actions of chin-raise/head-lift, gaze direction, and eyebrow raise are usually involved.’ Key for our purposes is the fact that the vector of pointing is defined by gaze while the ‘lips’ actually serve a meta-communicative purpose, signalling that the gaze is being used as a point. Enfield (2001: 185) thus writes, ‘the “vector” of lip-pointing is in fact defined by gaze, and the lip-pointing action itself (like other kinds of “pointing” involving the head area) is a “gaze-switch”, i.e. it indicates that the speaker is now pointing out something with his or her gaze.’ The example of lip-pointing thus illustrates the way that humans can accomplish intentional reference (i.e. non-natural meaning in Grice’s 1957 sense) through small manipulations of naturally meaningful behaviours (gaze direction) which exploit the human propensity to follow another’s gaze. The introduction of a meta-communicative overlay (chin/lip/head) on gaze direction transforms a cue into something another can recognize as a true intentional signal—‘He’s referring to that thing/person/area over there.’
Figure 11.2 Still image from video of a Jahai speaker lip-pointing, provided by Niclas Burenhult
221
Deixis and the Interactional Foundations of Reference 221 Table 11.1 Current and projected focus of attention in deixis Gaze-following
Lip-pointing
Finger-pointing
CFA only
CFA=PFA
CFA=PFA or CFA≠PFA
We are now in a position to describe one distinctive feature of finger-pointing relative to the other forms of primitive deixis so far described. Specifically, in finger-pointing it is possible to separate the speaker’s current focus of attention from the focus that they are proposing for a recipient. We can describe the basic elements and their combinatorial possibilities by means of the following table and figures. In Table 11.1 ‘current focus of attention’ is annotated CFA and ‘proposed focus of attention’ is annotated PFA. This can be easily seen in the frame-grabs in Figure 11.3, taken from a video-recorded interaction among speakers of Bequia creole. In the first frame Viv (in the foreground) is telling Baga (in the background) about a man she thinks he might know but whose name he does not recognize. When the description given allows Baga to identify the person Viv is talking about, he points up the hill to his right (Figure 11.3b). Notice that when Baga initially points, his own gaze is directed to the place he is indicating with his finger (i.e. CFA = PFA). In Figure 11.3c, he maintains the pointing gesture but now gazes toward Viv apparently to check whether his reference has been successful—checking, that is, on his recipient’s focus of attention (i.e. CFA≠PFA). He finds Viv pointing to the same place and the two engage in a moment of mutual gaze. Here then we can see, in the visible behaviour of the participants, how reference involves joint attention such that two persons are not only publicly projecting their attention to the same referent, but where they are, in addition, mutually aware of the current alignment, and thus sharedness, of their two lines of attention. So we can see why this possibility of separating the speaker’s/gesturer’s directing signal from the speaker’s gaze is important since a joint attentional frame crucially involves the speaker monitoring the recipient’s attention to some third object (Carpenter et al. 1998; Tomasello 1999; Liszkowski et al. 2004; Tomasello et al. 2005). This monitoring of the recipient transforms common attention to a THIRD into true joint attention—a basic form of shared intentionality (see also Gilbert 1989; Searle 2010). It is relevant to note here that ‘lip-pointing’—which crucially involves gaze as noted above—seems specialized among Lao speakers in two ways: First, ‘lip-pointing is apparently restricted to cases when the addressee is looking at the speaker’ (Enfield 2001: 192) and second ‘to acts of direct ostension in which the location or identity of a referent in the physical environment is in focus’ (Enfield 2001: 196, emphasis added). The prior establishment of recipiency along with the already ‘in focus’ character of the referent, it can be supposed, obviates or at least alleviates the need to monitor the recipient. Finger-pointing (Kita 2003) would also seem to allow for a higher informational load than do the forms of ‘lip/gaze-pointing’ we have considered. Thus, researchers have noted various functional contrasts here in little versus big points (Enfield et al.
222
222 Jack Sidnell and N. J. Enfield (a)
(b)
(c)
(d)
Figure 11.3 Finger-pointing, Bequia, St. Vincent (see Sidnell 2005) (Still image from video recording).
2007), those that are accompanied by gaze versus those without (Streeck 1993), as well as informational possibilities associated with different hand and finger configurations (Wilkins 2003; Kendon and Versante 2003). Finger-pointing also makes ‘path descriptions’ possible, as well as illustrative combinations. At the far end of the informational scale are diagrammatic representations in which pointing gestures are used to identify positions within a virtual drawing (see Enfield 2005). We can see many of the basic features of deictic reference in another form of behaviour among infants which Kidwell and Zimmerman (2007) as well as Tomasello (1999) and Clark (2003) describe as ‘showing’. In a typical showing sequence, a young child will approach another (typically an adult) with an outstretched arm and an object in hand (see Figure 11.4), the other might produce a response which identifies the object (‘Watermelon’), expresses a social-relational feature of the object (‘Your shoe’), or appreciates it in some way (‘Oh wow, a pretty hat’). The showing child then withdraws the object from view and/or moves out of the recipient’s line of vision, either returning to the activity she was engaged in before the showing or initiating some new activity. Such ‘showings’ are arguably one of the most basic forms which exhibit the triadic, joint attentional interaction configuration that constitutes the very foundation of reference in all its various forms (see Tomasello 1999, 2003). Clark (2003) has explicated
223
Deixis and the Interactional Foundations of Reference 223
Figure 11.4 Human infant showing object to camera person
the parallels and the key difference: in pointing, the other’s attention is made to move toward the current location of a thing, while in showing, a thing is moved into the current line of the other’s attention—either way, the other’s attention ends up directed towards the thing. In the current context, showings can be understood as an early form of demonstrative, or better, ‘presentational’ deixis akin to adult uses of French ‘voilà’ or English ‘look at this’. Instructional activities build upon the human propensity for attending to the attention of another, and showings play an important role in their organization. Rembrandt’s Anatomy Lesson of Dr Nicolaes Tulp provides a stunning illustration (Figure 11.5). Here Tulp is presenting a part of the cadaver for the consideration of his students, some of whom look attentively at that which is being shown. Through such presentations or showings, novices are socialized into new ways of ‘seeing’ the world around them, ways of seeing that are appropriate to some particular status or role (see Goodwin 1994; Kockelman 2007). In showings, then, we see not only the roots of reference in
224
224 Jack Sidnell and N. J. Enfield
Figure 11.5 The Anatomy Lesson of Dr Nicolaes Tulp, Rembrandt Harmenszoon van Rijn (1632)
human action but moreover the interactional foundations of human teaching, learning, the transmission of knowledge across generations, and thus, ultimately, of culture (see Tomasello 1999).
11.3 Demonstrative Systems In order to achieve joint attention on something, that thing must be somehow picked out from the range of possible things that one might be attending to. Often there are many possible things a person might be looking at or pointing to, and there are various ways to solve the problem of figuring out just which thing is the focus of attention. In the joint-attentional behaviours described in the previous section, details of body comportment such as eye gaze, pointing, and showing constitute relatively straightforward ways to narrow another’s attention on something to the exclusion of other possible referents in a context.1 But when the deictic function is supplied purely by the selection 1
Of course, as Wittgenstein (1953) and others pointed out, all reference involves a certain degree of indeterminacy. So, for instance, in the example from Rembrandt a recipient must infer, on the basis of
225
Deixis and the Interactional Foundations of Reference 225 of a word, there is little of inherent value in the word form itself that helps to solve this narrowing-in function. This is why demonstratives like that and this are often accompanied by some form of deictic bodily behaviour (or descriptive lexical content—e.g. ‘That blue one’ etc.). At the same time, such linguistic forms are also able to rely on the special salience of potential referents as determined by the current common ground of interlocutors; for instance, one might say ‘My brother has a car like that one’ while there are numerous cars in view, but where the car has a special salience in the scene—for example, it just drove past us, or it is painted a garish colour, or is particularly expensive- looking (Clark, Schreuder, and Buttrick 1983). Take examples like I heard that, Take this, or Were you at that party? These are semantically very general forms of expression, and a listener can only make sense of them by connecting the speech to something semantically much more specific such as a physical object or something in the spoken discourse or other shared knowledge, in other words, in the common ground (Clark 1992, 1996). The salience required for the successful connecting of a demonstrative to a referent may come from different sources. Certain things might be salient already because they are large, bright, central, or otherwise prominent in their surroundings (Clark et al. 1983). And one can render something salient in various ways (e.g. by pointing at it, looking at it, using a laser pointer, shining a light, holding the thing up). Ultimately, however, even where many sources of information converge to suggest a single referent, recipients of deictic expressions must infer what is being indicated. Syntactically, demonstratives may serve a range of different functions. For example, in English that may occur as an independent noun phrase (e.g. I saw that) or as a modifier within a noun phrase (e.g. I saw that car). Some demonstratives are ‘adverbial’ in function, in that they can be seen to relate to or modify events and actions (e.g. there in I went there). Depending on which language system we consider, demonstratives show different distributions (thus, in English I saw that/*there, I went *that/there, I saw that car/*there car). The details of such distinctions are subtle and complex and are particular to each language system (see Anderson and Keenan 1985, Diessel 1999, Dixon 2003 for reviews). One common function of demonstratives in spoken language is ‘exophoric’. In exophoric uses, reference is made to physical things and places that can be seen and pointed to in the context of the speech event. Alongside these exophoric functions, there are also endophoric referential uses of demonstratives (Halliday and Hasan 1976). In endophoric uses, reference is made not to things that can be physically pointed to and shown, but to things in the discourse context, which often includes things that have been said (e.g. anaphoric use of that in He said it was good and I agreed with that), but could also refer to things that will be said next (e.g. cataphoric use of this in What I want to say is this: I agree). Another kind of endophoric reference points to whatever evidence is available, whether the doctor intends to draw attention to the arm, the tendon, the flesh, or the entire body of the cadaver. Or again whether it is the colour, the size, or shape of some or all of the cadaver that is being indicated.
226
226 Jack Sidnell and N. J. Enfield things in the shared common ground, sometimes referred to as a ‘recognitional’ usage (Himmelmann 1996, after Sacks and Schegloff 2007b [1979], see later); this is found in cases like He reminds me of that boyfriend of Jane’s, where in order to resolve the reference of ‘that’, the listener consults neither the physical setting nor the current discourse, but rather the interpersonally shared common ground of the dyad. The endophoric uses of demonstratives are often regarded as secondary or derived from exophoric uses, based on arguments from both ontogeny (infants acquire exophoric functions first) and diachrony (endophoric functions often develop from exophoric ones; see Diessel 1999 for a statement of this position). However, it is not clear that from a synchronic perspective either function is subordinate to the other. Hanks (1990) and Enfield (2003) have argued that the core meanings of demonstratives do not semantically specify an exophoric versus endophoric distinction, rather that these are simply distinct (and sometimes not-so-distinct) pragmatic contexts of use of the semantically general terms. Typological work on demonstratives indicates that there is significant and subtly complex variation across languages in terms of the semantic dimensions that are encoded, the number of distinctions, and the grammatical properties of the various elements of the systems. Here we do not attempt to give an overview of the typological properties of demonstratives and demonstrative systems (for that, see e.g. Himmelmann 1996; Diessel 1999; Dixon 2003; Huang 2014). We will simply introduce a few of the known ‘realms of possibility’, concentrating specifically on the number and semantic types of possible distinctions found in systems of ‘demonstrative adjectives’ (i.e. words like that and this as modifiers in expressions with nominal referents; e.g. that car or this book). A demonstrative system can be extremely simple in terms of the number of distinctions it makes along a given dimension such as distance from speaker. Colloquial German, for example, has essentially a one-term system of demonstrative adjectives. While grammars of German state that a noun may be modified by one of three distinct terms: der ‘that/the’, dieser ‘this’, and jener ‘yon’, in fact only der (and its variants die and das, depending on the gender of the head noun) tends to be used. So, for instance, a German speaker would be more likely to say das Buch hier for ‘this book’ (proximal to the speaker, literally ‘that/the book here’) to distinguish from das Buch ‘that/the book’. A more common and still very simple type of system features a two-way distinction. In English, for example, a ‘proximal’ term this stands in opposition to a ‘distal’ term that. There is an archaic term yon ‘far distal, over there’, but it is almost never used. A similar situation is found in the non Pama-Nyungan language Kayardild, with ‘distal’ dathin- and ‘proximal’ dan-, and a ‘rarely used’ form nganikin- meaning ‘that, beyond the field of vision’ (Evans 1995: 206–210). It is surprisingly difficult to determine precisely what is the semantic distinction between the terms in such a system, though the most common characterization is ‘proximate’ versus ‘distal’. This captures the fact that, in general, things that one refers to with the word this tend to be spatially closer to the speaker than things one would refer to with that. However, there are problems with this suggestion. For one thing, these words are used in endophoric, non-spatial domains where the application of an analysis in terms of ‘proximate’ and ‘distal’ is metaphorical at best. A more parsimonious analysis would then not specify spatial distance as the
227
Deixis and the Interactional Foundations of Reference 227 operative factor (Enfield 2003; Hanks 1990; Kirsner 1979). For another thing, there is no objective measure of what would count as ‘proximal’ versus ‘distal’, yet these terms imply some kind of specifiable distance. When we observe actual usage, it turns out that spatial distance between speaker and referent does not predict which term will be used. This was demonstrated in an analysis of situated usage of a two-term system in Lao (a Southwestern Tai language of Laos; Enfield 2003). The account that best captures the observed data posits a semantic asymmetry in the system: one of the terms is semantically specified as ‘external’, ‘distal’, or more accurately ‘not here’, while the other term has no specification for ‘externality’ or ‘distance’. This is a basic ‘informativeness scale’ (Horn 1989; Levinson 2000), by which the unmarked member of a paradigm can readily pick up extra pragmatic meaning by virtue of its opposition to the other members. In the Lao case, the semantically general form tends to imply ‘proximal’, not because it semantically specifies proximal but because it is being chosen when ‘distal/ external’ could have been chosen instead. A similar solution has been implied in analyses of the English that/this opposition, though without consensus as to which term is the semantically unmarked one (Halliday and Hasan 1976: 59 say that that is basic, while Wierzbicka 1980: 27 and Dixon 2003: 81 say that this is the basic form). Many languages have three-term systems, often described in terms of the familiar ‘proximate’ versus ‘distal’ distinction, but where there are two ‘proximate’ terms: one refers to things that are proximate to the speaker, the other to things that are proximate to the addressee. For example, in Yimas, a Lower Sepik language of Papua New Guinea, there are three deictic stems: -k ‘this (near me)’, m-‘that (near you)’, and -n ‘that yonder (near neither you nor me)’ (Foley 1991: 112). Or in Manambu, also spoken in the East Sepik, there are the forms k-‘close to speaker’, wa-‘close to hearer’, and a-‘far from both’ (Aikhenvald 2008: 201). Other three-term systems operate on different semantic principles. In Turkish, alongside a contrast between ‘proximal’ (bu) and ‘distal’ (o), there is a term (şu) that encodes ‘the absence of the addressee’s visual attention’ on the thing being referred to (Küntay and Özyürek 2006: 304). There are also many languages with demonstrative systems that have more than three terms. Often the extra terms mark spatial contrasts associated with living in a particular kind of physical environment and lifestyle. For example, in Kri, a Vietic language of Laos (Enfield and Diffloth 2009), there is a five-term system of exophoric demonstratives, featuring a familiar-looking proximal versus distal distinction, in addition to semantic distinctions of ‘across’, ‘up’, and ‘down’, motivated by the Kri speakers’ riverine up–down environment (this system is also used with reference to small-scale or ‘table top’ space; see further discussion of the Kri system in section 11.4): (1) a. b. c. d. e.
nìì naaq seeh cồồh lêêh
general (‘this’, proximal) external (‘that’, distal) external, across (‘yon’, far distal) external, down below external, up above
228
228 Jack Sidnell and N. J. Enfield A similar system is found in Lezgian, a Nakho-Daghestanian language of the Eastern Caucasus (Haspelmath 1993: 190; note that according to Haspelmath, in ‘modern standard’ Lezgian, only the two forms glossed as ‘that’ and ‘this’ are commonly used). In the Lezgian system, yet another term (ha) is added, which has a dedicated ‘discourse anaphoric’ function: (2) a. b. c. d. e. f.
this that yonder the aforementioned that up there that down there
‘i’ a at’a ha wini aǧa
These few examples can only hint towards the complexity and subtlety of different demonstrative systems in languages of the world. The list of possible semantic distinctions is long. In his typological survey of demonstrative systems, Diessel (1999: 52) summarizes all of the semantic features that are attested. These divide into ‘deixis’ and ‘quality’, subcategorized in Table 11.2. Adding to the complexity and richness of the possibility space for demonstratives, the various terms may be enlisted in many different ways for endophoric Table 11.2 Diessel’s summary of semantic distinctions attested in demonstrative systems (A) Semantic distinctions in demonstratives of the type ‘deixis’: (i) distance (ii) visibility (iii) elevation (iv) geography (v) movement (or direction) (B) Semantic distinctions in demonstratives of the type ‘quality’: (i) ontology (ii) animacy (iii) humanness (iv) sex (v) number (vi) boundedness
229
Deixis and the Interactional Foundations of Reference 229 usages, and in other syntactic functions (e.g. as demonstrative adverbs like English there and here). The most important future line of research is to test the proposed semantics of these systems in the context of their usage in everyday life. Since the understanding of demonstratives are so heavily context-dependent, they cannot be meaningfully studied without looking at a corpus of usage. This issue is discussed in section 11.4.
11.4 Demonstratives in the Context of Common Ground We began this survey of deictic reference with the simplest kinds of joint-attentional scenes, the kinds that allow a 9-month-old to get started on his or her long journey of socialization. It is a years-long path of countless moments of joint attention, countless instances of learning and guidance, of gradual convergence in knowledge and stance with elders and peers, first through simple gestures and shared participation frames, and soon within the increasingly rich matrices of language, kinship, ritual, livelihood, and material culture. These aspects of the sociocultural world all form the basis of a community’s common ground, and thus are naturally caught up in the elements of demonstrative systems, dependent as they are on whatever sources of ‘mutual salience’ happen to be at hand. Most previous work on deixis, such as the research on demonstrative systems outlined in the previous section, has approached the task as a search for the right ‘gloss’ of each form’s meaning. However, deictic terms like demonstratives are especially hard to gloss in the abstract since interpreters are so heavily dependent on context in figuring out what they refer to on any given occasion. Research such as that by Hanks (1990) and Enfield (2003) has shown that the situated dynamics—b oth spatial and social- relational—of social interaction bears directly on how a simple demonstrative distinction, e.g. between that and this in English, is to be interpreted. The key to interpreting deictic expressions is the common ground that pertains between interlocutors (Clark 1996; cf. Hanks 2006b). In a study of Lao, Enfield (2003) shows how the rapidly changing common ground arising from fluidly evolving participation frames in marketplace interactions can affect the differential selection of demonstratives for picking out referents that are all proximate and in common view. In other kinds of context, we see how common ground of the more enduring kind—that is, cultural common ground (Clark 1992)—also has a bearing on the selection and interpretation of demonstratives. Let us consider an example from research on speakers of Kri, an Austroasiatic language of Laos (Enfield and Diffloth 2009).
230
230 Jack Sidnell and N. J. Enfield
Figure 11.6 Kri house
In the Kri-speaking community of Mrka village in upland central Laos, houses are built to a precise plan, by which the physical layout of the house is a diagram of certain social-relational asymmetries, on two axes (see Enfield 2009 for detailed discussion). Running laterally across the house is an ‘in–out’ axis, where ‘in’ maps onto ‘private, family, women, children, storage room, food preparation’ and ‘out’ maps onto ‘public, non-kin, men, adults, guests, drinking, public ritual’. Orthogonal to this is an axis that runs from what we would call in English the ‘front’ of the house, where one enters, to the ‘back’ of the house. In Kri, this is referred to as a ‘below–above’ axis, where ‘below’ maps onto socially lower rank, and ‘above’ to socially higher rank, where relative ‘height’ is determined primarily by relative age, often attenuated by classificatory kinship. See Figures 11.6 and 11.7. The Kri house is therefore conceptualized spatially as a mini-version of the larger geographical environment, as coded in the demonstrative system. Recall that in that system (see (1)), beyond the ‘proximal’ and ‘distal’ forms, there are three forms in addition: ‘the one up/above/upstream/uphill’, ‘the one down/below/downstream/downhill’, and ‘the one across’ (i.e. away but neither up or down). While the house floor is normally perfectly level, the ‘up/down/across’ scheme is nevertheless readily mapped onto
231
Deixis and the Interactional Foundations of Reference 231 5 m approx. prùng kùùjh ‘fire pit’
upper roong ‘upper corner’
sùàmq ‘inner room’
sùàmq
sùàmq
tkoolq ‘giant mortar’ khraa ‘storage and work room’ prùng kùùjh ‘fire pit’
cààr ‘verandah (covered)’
krcààngq ‘ladder’
cààr ‘verandah (open)’
lower
outer
inner
Figure 11.7 Kri house floor plan
it, thanks to its diagrammatic relation to the socialcultural dimensions represented as ‘in–out’ (family versus non-family) and ‘up–down’ (senior versus junior). Now consider an example from a video-recorded interaction between a group of Kri-speaking women sitting on a front verandah, in which this socioculturally motivated mapping provides the solution for a simple referential problem of locating an object. Figure 11.8 shows a still image from the video recording.
232
232 Jack Sidnell and N. J. Enfield
Figure 11.8 Image of the speakers (Still image from video recording).
The scene is in the house of E, the elderly woman at the right of frame. We focus on an exchange between her and B, the young woman second from left, visible in the door frame, who does not live in this house. (3) Kri interaction B: piin sulaaq Give leaf Pass some leaf. E: sulaaq quu kuloong lêêh, sulaaq, quu khraa seeh Leaf LOC inside DEM.UP leaf loc store DEM.ACROSS The leaf is inside up there, the leaf, in the store. môôc cariit hanq one backpack 3SG (There’s) a (whole) backpack. (5s; B walks inside) Here, B makes a request to be given some ‘leaf ’ (actually, corn husk) with which she can roll a cigarette. In E’s reply, she uses a complex combination of referential
233
Deixis and the Interactional Foundations of Reference 233 expressions to inform B of the location of the ‘leaf ’ so that she can go and get some herself. First, an intrinsic spatial reference (kuloong ‘inside’) is combined with the ‘up’ demonstrative lêêh, in alignment with the up–down axis of the house. From their perspective sitting on the verandah, the ‘lowest’ part of the house, the inside area of the house is ‘up’, and, accordingly, this is coded in the demonstrative chosen. E then narrows in further on the spatial location; where they are currently sitting is the ‘outer’ edge of the house, and the ‘leaf ’ in question is located inside the khraa ‘storage room’ at the ‘innermost’ side of the house: once one has entered the house going ‘up’ from where the speakers are sitting, one would then have to go ‘across’; this is specified with use of the relevant demonstrative seeh ‘the one across there’. This example has illustrated one way in which the interpretation of demonstratives depends crucially on shared background knowledge, as relevant to the context of speaking. In the case of Kri, the selection and interpretation of demonstratives draws directly on a conventional mapping of the sociocultural domain of kinship and other personal relations onto the 2D spatial array of the house floor plan.
11.5 What’s Special about Deixis as a Form of Reference? In this final section we address the central question with which we began: what is special about deixis as a form of reference? Another way to ask the same question is: where both deictic and non-deictic formulations of a referent are possible, why might a speaker choose the deictic one? Consider the following case from the second presidential debate between John McCain and Barack Obama in 2008. Here the moderator has asked McCain the following: ‘Should we fund a Manhattan-like project that develops a nuclear bomb to deal with global energy and alternative energy or should we fund 100,000 garages across America, the kind of industry and innovation that developed Silicon Valley?’ McCain has already begun to respond when he produces the following segment: (4) McCain2 01 JM: By the way my friends: I-I know you grow a little wea:ry 02 with this back-and-forth. 03 (.) 04 It was an energy bill on the floor of the Senate loaded down 05 with goodies. billions for the oil companies. An’ it was 06 sponsored by-Bush and Cheney. 2
English examples are presented using the transcription conventions originally developed by Gail Jefferson. For present purposes, the most important symbols are the period (‘.’) which indicates falling and final intonation, the question mark (‘?’) indicating rising intonation, and brackets (‘[’ and ‘]’) marking the onset and resolution of overlapping talk between two speakers. Equal signs, which come
234
234 Jack Sidnell and N. J. Enfield 07 08 09 10 11 12 13
(0.2) You know who voted for it, might never know, That one. You know who voted against it? Me. I have fought time after time against these pork barrel—these-these bills that come to the floor and they have all kinds of goodies an’ all kinds of things in them for everybody and they buy off the votes,
Notice then that McCain selected the deictic formulation ‘that one’ in referring to Obama who was sitting close by at the time (see Figure 11.9). This is clearly a marked usage in contrast to ‘Obama’ or ‘Senator Obama’ and it was noted in the press, with many ordinary people as well as political pundits weighing on what the formulation might ‘mean’. For instance, the Huffington Post reported: During a discussion about energy, McCain punctuates a contrast with Obama by referring to him as “that one,” while once again not looking in his opponent’s direction (merely jabbing a finger across his chest). That’s not going to win McCain any Miss Congeniality points. Nor will it reassure any voters who believe McCain is improperly trying to capitalize on Obama’s “otherness.”
David Axelrod—an Obama strategist—was reported as saying: ‘Senator Obama has a name. You’d expect your opponent to use that name.’—clearly drawing attention to the marked character of ‘that one’. Other commentators suggested that the usage was disrespectful, rude, or even racist. Defenders of McCain, in contrast, argued that the press and others were making something out of nothing. Drawing on the basic principles sketched in this chapter we can develop an analysis of how people were able to arrive at these diverse interpretations. First, the reference is accompanied by a pointing gesture in the direction of Obama (Figure 11.9), indeed there is prior point at Obama produced over ‘you know who voted for it?’ Second, while producing the reference (the deictic formulation ‘that one’ with point in Obama’s direction), McCain was gazing at the studio audience. Third, the reference combines the deictic ‘that’ with the characterizing ‘one’—a usage which denotes any enumerable person or thing. The combination is roughly equivalent to ‘him’ in denoting a third person, non-participant in the immediately available speech situation, i.e. not a speaker, not an addressee; and note that it is compatible with the referent being an inanimate object. Fourth, McCain can be seen to have selected ‘that’ from the pair of contrasting terms ‘this/that’—‘that’ is what we gloss as the distal member of the pair and, in contrast to ‘this’, conveys distance from speaker (see Stivers 2007). in pairs—one at the end of a line and another at the start of the next line or one shortly thereafter—are used to indicate that the second line followed the first with no discernable silence between them, i.e. it was ‘latched’ to it. Numbers in parentheses (e.g. (0.5)) indicate silence, represented in tenths of a second. Finally, colons are used to indicate prolongation or stretching of the sound preceding them. The more colons, the longer the stretching. For an explanation of other symbols, see Enfield and Stivers (2007).
235
(a)
(b)
Figure 11.9 McCain and Obama, ‘That one’
236
236 Jack Sidnell and N. J. Enfield We can see that this reference positions Obama as a non-participant in a speech event comprised of McCain and the audience to whom his talk is directed. In addition, the use of ‘one’ and ‘that’ (rather than ‘this’) conveys distance. These effects, along with McCain’s use of ‘my friends’ to address and align the audience, thus work together to construe an interactional rift that divides himself and the audience on the one side from Obama on the other. At the same time, of course, these meanings are defeasible—from another perspective, McCain was simply using a highly efficient, minimally characterizing referring expression to identify who he was talking about. The availability of seemingly incompatible, even opposed interpretations is surely an outcome of the fact that so much of the meaning of these forms is inferred rather than encoded. We are now in a position to summarize at least some the features of deixis that distinguish it from other forms of reference and to see how these might shape a speaker’s selection of a deictic over non-deictic formulation. 1. Deictic reference is a low-cost, highly efficient, minimally characterizing way to accomplish reference. Many of the examples we have so far discussed exemplify just this point. Simply put, there are many situations in which a deictic formulation is the most efficient way to accomplish reference. Where the intended referent is already available in the common ground and perhaps even co-present, a deictic formulation constitutes the most straightforward way of referring to it. Notice that this likely explains the universal occurrence of deictic words in the world’s languages—a language without them would be unnecessarily cumbersome. It should be noted however that there are some (perhaps many) situations in which sociocultural norms override any pressure towards efficiency. So for instance in Vietnamese, in many situations, speakers avoid minimally characterizing deictic formulations in referring to speaker and hearer (tôi/ta ‘I’, mày ‘you’) in favour of kin terms which explicitly characterize the social relationship between speaker and hearer (Luong 1990; Sidnell and Shohet 2013). So while matters of efficiency are clearly at play, their relevance may not always be paramount. 2. The semantically general character of deictic forms makes them well-suited for reference to hard-to-describe and/or nameless objects. In such a situation a deictic form can exploit features of the artifactual environment, including the presence of the thing being referred to. For instance, in the following case something hanging on the door of the small room where three children are playing is initially referred to by ‘it’. However, when the recipient initiates repair of the reference with ‘move what?’, a deictic formulation is used which locates the referent relative to landmarks in the physical environment rather than characterizing or describing it.
237
Deixis and the Interactional Foundations of Reference 237 (5) Kids_11_24_05(2of2)T7 @11:33 01 02 03 04
A: C: -> A: C:
((looks at door)) Maybe R---, maybe you can move it, °Move what?° Move that thing that’s in the lock Okay.
3. The semantically non-specific, minimally characterizing features of deixis allow speakers to avoid description where such description may be counter-productive to some interactional goal. There are situations in which a speaker may wish to avoid characterizing the thing referred to and here deictic formulations are particularly well-fitted. Sacks (1995) discussed this issue in his consideration of ‘indicator terms’ (the term used by analytic philosophers such as Russell and Goodman to talk about deictics). Sacks observed that in the context of group therapy one patient may wish to avoid saying ‘why are you in therapy?’ and prefer instead ‘why are you here?’—these questions having quite different implications. The first invites an answer that makes reference to the real or supposed psychological issues with which the recipient is struggling. The second, in contrast, can be answered with something like ‘my father sent me’ or ‘it’s a condition of my parole’ etc.—i.e. practical circumstances. This points to some of the ways sociocultural rules or norms may come into play in the selection of deictic or non-deictic forms. Levinson (2005, 2007) has discussed data from Rossel Island that is also relevant here. The Rossel Islanders observe taboos on name use when the bearer of the name is recently deceased. In their attempts to observe these taboos, speakers of Yélî Dnye sometimes resort to highly circumspect reference often involving elaborate deictic gestures or linguistic formulations—eyebrow flashes to distant locales, points, or expressions like ‘that girl’ and so on. 4. Because these forms require for their interpretation the application of knowledge in common ground (shared knowledge), successful reference via such a form can be a demonstration of social proximity—an informational enactment of intimacy (see Enfield 2006). Schegloff (2007b) discusses how this works via the indexical meaning of a person’s voice. In the following example, Clara picks up the phone and says hello (line 6b), to which the caller, Agnes, responds with ‘Hi’ (line 6c). From this one-syllable voice sample, Clara knows it is Agnes, and demonstrates this knowledge in her subsequent utterance, by using Agnes’s name. (6) a. b. Clara c. Agnes d. Clara
((Ring)) Hello Hi Oh hi, how are you Agnes
238
238 Jack Sidnell and N. J. Enfield This indexically-based understanding is a way of making a genuine demonstration of shared knowledge between a particular dyad. Had the caller been someone who Clara did not know, or knew less well, she would have been simply unable to make this demonstration of knowing who it was, and thus would have made explicit the greater social distance between the two. This example relates to the indexical meaning that allows us to recognize a person just from their voice, and so is not in the realm of linguistic deixis; however, we see exactly the same effect in the domain of grammatical deixis. In this example from Lao (see Enfield 2006 for more information), a man is talking about a riverine environment near his village, where villagers were once able to collect large amounts of a certain herbal medicine. (7) 6 tè-kii4 before paj3 go
haak5 vang2-phêêng2 pcl vp
nanø tpc.nonprox
tèø-kii4 before
khaw3 3pl.b
tèq2-tòòng4 touch
‘Before, in Vang Phêêng weir, before, for them (the villagers) to go and touch it 7 bòø daj4, neg can
paa1-dong3 forest
man2 3.nonresp
lèwø prf
dêj2 fac.news
was impossible, it was the forest of itnon-respect, you know.’ The deictic element in line 7—man2 ‘it’—has no local antecedent, and so the speaker is evidently assuming that his listeners will know how or what ‘it’ is. A couple of lines later, a woman who is listening to the man’s story asks: (8) 8
FW
khuam2 reason
phen1 3.p
haaj4 angry
niø tpc
naø tpc.periph
‘Owing to itsrespect being angry?’ She uses a different pronoun, this time marking respect, however the referent is still entirely inexplicit. In the next line, the man does make the referent explicit: (9) 9
FM qee5 —bòò1 mèèn2 yeah neg be phii3 spirit
lin5 play
vang2-phêêng2 V
lin5 play
dêj2, fac.news
niø pcl
‘Yeah—It’s not playing around you know, the spirit of Vang Phêêng.’
239
Deixis and the Interactional Foundations of Reference 239 The deictic expressions man2 and phen1, both third person pronouns, were first used in this sequence in such a way as to assume certain cultural common ground; namely that ‘weirs’ and similar deep water environments have spirit owners that protect the aquatic resources and that are feared and respected. The fact that these interlocutors were able to successfully refer to these spirits with only the use of these semantically very general demonstrative expressions is a demonstration of their common membership in a particular sociocultural world, and not only in a common ‘speech community’. In this chapter we have sketched the interactional foundations of deixis (and reference in general) in the joint attentional scenes and associated action trajectories of ordinary social life. We then discussed two ways in which the basic features of deictic reference are elaborated—in semantically complex systems of linguistic opposition and in the way they map onto the rich, conventionally meaningful cultural systems that make up the life-world. Finally, we have tried to address the fundamental question of why any given speaker on any given occasion would select a deictic over a non-deictic expression.
Abbreviations Used Orthography used for Lao in this book follows Enfield (2007). Orthography used for Kri follows Enfield and Diffloth (2009). Following are the conventions used for interlinear morphemic glossing: 1 1st person 2 2nd person 3 3rd person B bare dem demonstrative dir directional dist distal fac factive loc locative neg negation news news marker nonprox
non proximal
pcl particle pl plural prf perfect tpc topic
240
Chapter 12
Refere nc e Barbara Abbott
12.1 Introduction Even taking a very constrained view, words like refer and reference have two distinct but related uses. As used in, e.g., a common translation of Gottlob Frege’s famous 1892 paper ‘Über Sinn und Bedeutung’ as ‘On sense and reference’, they are associated with the relation between linguistic expressions and the entities that they apply to. In this sense they tag a semantic relation. Using the words this way we might say that the noun phrase (NP) the star of Evita refers to the singer Madonna.1,2 However, 58 years later P. F. Strawson argued strenuously that it is wrong to speak of expressions as referring. Rather, referring is something speakers of a language use expressions to do. Under this construal (which, it must be admitted, is closer to everyday usage), ‘refer’ and ‘reference’ tag elements within the realm of pragmatics. As this is a pragmatics handbook, the latter construal will be the focus of the current chapter. The next question is which kinds of expressions can be used by speakers to refer to things? Keeping within our relatively constrained view, overwhelmingly the assumption has been that NPs are the primary tools of reference. Although it would not be outrageous to say something like Mary was using the words deceive and underhanded to refer to your activities, suggesting that verbs and adjectives might be said to be used in referring, nevertheless it is almost entirely the case that scholars have focused on NPs in this area. A related fact is the tendency of philosophers to focus on what they call singular reference, or reference to individuals, in their discussions of reference of either the semantic or the pragmatic type, where the singular terms which may participate in singular referring relations are all NPs. 1
Many linguists today use ‘determiner phrase’ or ‘DP’ the way ‘NP’ will be used in this chapter; those individuals often use ‘NP’ for the constituent containing the head noun and restrictive modifiers, but not the determiner. I regret the potential for confusion. 2 The terms denote and denotation are often used now for such semantic relations; not just the one which holds between NPs and the entities they apply to, but also the relations between general terms (common nouns, verbs, adjectives, etc.) and the sets of entities or properties that they apply to.
241
Reference 241 A brief clarificatory note: the terms ‘singular reference’ and ‘singular term’ are misleading in their suggestion that only grammatically singular NPs are eligible for referential status. This would drive a wedge between, e.g. The cat is on the mat and The cats are on the mat. However, since important work by Sharvy (1980), Link (1983), and Landman (1989a,b), among others, which argues for the possibility of unitary referents (groups or mereological sums of atomic entities) for plural NPs (as well as mass terms), we need not confine ourselves arbitrarily to only grammatically singular NPs. This proviso should be borne in mind throughout—discussion of reference to ‘an individual’, for instance, should be understood to include plural entities as possible individuals (for more on plurals, see Landman 1996). The rest of this chapter will address some basic and not-so-basic issues arising in the area of reference. We’ll start by subdividing NPs into those that can properly be said to be used to refer (the singular terms), and those that (it has been argued) cannot. The reader must be warned that there are major differences of opinion floating around in the literature on this topic; generally speaking, philosophers are more conservative about granting referring rights while linguists are more liberal, although of course this does not always hold. We look next at what speakers are using these expressions to refer to: complications are created by the difference between discourse reference and ‘real- world’ reference, and by various kinds of what we might call indirect reference. Section 12.4 will concern itself with the various possibilities a speaker must choose among: in referring to a person, for instance, a speaker may typically have available a proper name, any number of definite or demonstratives descriptions, and a personal or demonstrative pronoun. The factors governing such choices are many and varied. Finally we take the perspective of the addressee in trying to determine what is being referred to; determining the referents of third-person pronouns will be an important topic here.
12.2 Which NPs May Be Used to Refer? One of the most conservative views on reference was held by Bertrand Russell. As a solid materialist, he thought that understanding implied reducibility to sensory inputs. Referents, as constituents of propositions, thus had to be fundamentally sensory— either properties (expressed by predicates) or sensible entities. So ultimately he concluded that singular reference was confined to use of a demonstrative pronoun (this in English), used demonstratively (cf. Russell 1918). While not toeing the strict Russellian line on understanding, a number of philosophers have more recently accepted the idea that propositions may have entities (e.g. people, chairs) as constituents. Such propositions are frequently referred to as singular propositions.3 As was the case with Russell’s reasoning, this has consequences for 3
Such propositions may also be called ‘Russellian Singular Propositions’ (or RSPs) to distinguish them from propositions about individuals which contain constant individual concepts instead of individuals themselves (cf. Abbott 2010, 2011).
242
242 Barbara Abbott reference: a proposition containing an entity such as a person or a chair cannot be said to be understood by someone who does not have some kind of acquaintance with the person or chair in question. Exactly what kind of acquaintance is deemed necessary varies from scholar to scholar, of course. Kent Bach, a notable proponent of this line, suggests the following: To be in a position to refer to something (or to understand a reference to it) requires being able to have singular thoughts about it, and that requires perceiving it, being informed of it, or (having perceived or been informed of it) remembering it. (Bach 2006b: 518)
This position is considerably more liberal than Russell’s, which required current perceptual acquaintance. Bach does not even require past perceptual connection, allowing that information about the entity may suffice.
12.2.1 Demonstratives and pronouns It is clear that on Bach’s view demonstratives—both demonstrative pronouns (e.g. this and that in English) and NPs with demonstrative determiners, called ‘complex demonstratives’ (e.g. those red horses)—when successfully used demonstratively, may be used to refer to entities with which both speaker and addressee are sufficiently acquainted, and thus to express and convey singular propositions. This is because in such situations, both speaker and addressee are in direct perceptual contact with the referents in question. This would also be true for first-and second-person pronouns, since they are used for conversational participants. Demonstratives also have anaphoric uses, as in (1). (1) a. I told him he was crazy and that made him furious. b. Do you know the Blimcocks? Those people are extremely wealthy. In these cases we would expect to find variation in whether or not Bach’s criteria are satisfied, that is, whether it would be correct to say that a demonstrative has been used to refer to something. In the case of the examples in (1), the underlined demonstratives are intended to tag the event described in the first sentence of (1a) and the Blimcocks in (1b). Most likely the speaker, for these particular examples, would have the required degree of acquaintance, but whether the addressee would, and whether that would be required for referring, are other issues. Similar remarks go for third-person personal pronouns. These may be used demonstratively, and such uses would typically be referential. They may be used anaphorically, and in these cases whether or not speaker reference has occurred would depend on whether or not the requisite degree of acquaintance was manifest. We discuss pronouns at greater length below in section 12.5.
243
Reference 243 Both demonstratives and pronouns also have various kinds of generic uses; these will be touched on below.
12.2.2 Proper names Proper names (e.g. Madonna, Barack Obama) are commonly used to refer to entities that are familiar to both speaker and addressee. In fact, as noted by Prince (1992), it is mildly infelicitous to use a proper name which you believe your addressee to be unfamiliar with, unless you include some additional clarifying information, e.g. as in (2). (2) Lyn Motlow—that’s my cousin who lives just down the road—is going to be stopping by later. Would the clarification given in (2) be sufficient to meet Bach’s criteria for acquaintance? Certainly the identification which the speaker has supplied in (2) (that’s my cousin who lives just down the road) provides information about Lyn Motlow, and so we might consider the addressee to have been informed about Lyn. On the other hand Bach is careful to distinguish singular reference from mere use of a uniquely identifying description: ‘We cannot form a singular thought about an individual we can “think of ” only under a description’ (Bach 2006b: 522). Of course the addressee of (2) would now have both a proper name and a description, and could use these to derive additional pieces of information (e.g. that Lyn Motlow lives on North Lake Leelanau Drive), and so examples like this one might pass muster. (See Bach 2010b for further discussion of these issues.)
12.2.3 Definite descriptions This brings us to definite descriptions and possessive NPs. The two categories are frequently merged, but it’s useful to keep them separate. Definite descriptions in English begin with the. The determiner of a possessive NP, on the other hand, is a genitive NP—e.g. my in my cousin, or Smith’s in Smith’s murderer. Such possessive NPs are often regarded as definite descriptions, and frequently can be paraphrased as such (e.g. the murderer of Smith). However Haspelmath (1999) has argued persuasively that possessive NPs are different, and indeed, some are obviously not ‘definite’ at all, e.g. somebody’s book, which may appear in focus position in an existential (e.g. There’s somebody’s book on the table). We’ll concentrate here on definite descriptions, which have probably engendered more philosophical and linguistic discussion than any other type of NP. Possessive NPs with definite determiners (possessive pronouns, proper names, or possessive definite descriptions) would be expected to be similarly classified. Russell (1905) analysed definite descriptions as quantificational expressions, like e.g. every hammock and no pieces. His famous analysis of (3a) made it roughly equivalent to (3b).
244
244 Barbara Abbott (3) a. The king of France is bald. b. There is one and only one king of France, and he is bald. Russell argued that this analysis (which in effect eliminates definite descriptions as semantic constituents) solved several difficult problems of reference—problems to do with non-existent entities, and sentences about people’s beliefs and desires, which still today do not have universally agreed-upon solutions. Nevertheless this analysis is hardly intuitive; definite descriptions don’t seem like quantificational NPs, and in the paper referred to in our opening paragraph, Strawson (1950) argued against Russell’s analysis and indeed, against his whole narrow view of reference. In Strawson’s view, (3a) would not be used to make the assertion in (3b); rather the speaker would be presupposing the existence (and uniqueness) of a king of France, and using the definite description the king of France to refer to that individual and predicate baldness of them. If there isn’t any king of France (as there hasn’t been for some time), such an assertion cannot be successfully made—the speaker cannot succeed in making either a true or false statement, according to Strawson. In another major contribution to this literature, Keith Donnellan (1966) took aim at both Strawson and Russell, arguing that definite descriptions are ambiguous in having both referential and attributive uses. Used referentially, the description is just a device to get one’s addressee to recognize whom or what one is speaking of, and any other description serving that purpose would do as well. On this use the speaker presupposes of some individual that they meet the description in question. On the other hand, when a description is used attributively, the content of the description is crucial to what is being expressed, and it is not the case that any other description would do as well. It is also typically (although not necessarily) the case that the speaker has no acquaintance with the individual matching the description—they presuppose merely that there is such an entity. One of Donnellan’s examples is given in (4). (4) Who is the man who is drinking a Martini? For a referential use, imagine a bar scene where the man being described is in plain view and the speaker wants more information about him. Here any other description that served to identify the man to the addressee would do as well—e.g. the man in the plaid shirt, the tall guy with the red hair. Note that this situation would satisfy Bach’s criteria for referential use as well, since both participants are in perceptual contact with the intended referent. On the other hand, for an attributive use of the description in (4), imagine a different scene—say a Southern Baptist church picnic where absolutely no alcohol was to be consumed, but where the minister has just learned that somebody has smuggled in a Martini and uses (4) to ask about their identity. This use would not satisfy Bach’s criteria for reference, since the speaker knows only the one description. There are other analyses of definite descriptions which view them as referential. Heim (1982, 1983), following Christophersen (1939), treated definite descriptions as
245
Reference 245 referential devices whose referents must be familiar—that is, assumed to be known to the addressee—rather than the unique satisfiers of the descriptive content (as on Russell’s analysis), or uniquely identifiable (as many assume today, cf. e.g. Birner and Ward 1998). (Cf. also Stalnaker 2002 for another familiarity view.) Such approaches run into problems with frequent uses of definite descriptions to introduce novel entities into a discourse—possibly as many as half of all occurrences, as found in examinations of naturally occurring data (cf. e.g. Fraurud 1990; Poesio and Vieira 1998). Familiarity theories are forced to treat these as cases of accommodation (Lewis 1979), which reduces the predictive power of such theories substantially (see Abbott 2008 for discussion and references).
12.2.4 Indefinite descriptions This brings us to indefinite descriptions—NPs beginning with the indefinite article a/ an in the singular, and with some as determiner, or no determiner at all, for the plural or with a mass head noun. Bach (2006b), like Ludlow and Neale (1991), holds that only when an item is the current focus of attention of both speaker and addressee can reference be made to it with an indefinite. An example from Ludlow and Neale is in (5). (5) Look! A man is uprooting your turnips. (Ludlow and Neale 1991, ex. 3) We are to imagine an utterance of (5) where both speaker and addressee are in visual contact with the man outside; in such a case, according to Bach and Ludlow and Neale, we might say that the speaker has used a man to refer to the person in question. If the utterance is successful, the speaker has conveyed to the addressee a proposition containing the man they are looking at. Note that here, though, speaker’s reference (the pragmatic sort of reference that is our primary concern in this chapter) comes apart from any semantic relation. Semantically, the assertion in (5) must be viewed as purely quantificational—asserting the existence of a man uprooting the turnips. This is because, should the man that the speaker and addressee are observing not actually be uprooting turnips, but instead some other man is so engaged, then the assertion of (5) would still be true. In other words the statement in (5) is true if and only if there is a man who is uprooting the addressee’s turnips (see King 1988 for discussion of this point). When we come to other uses of indefinite descriptions, we find some differences of opinion. Chastain (1975), for example, argued that indefinite descriptions which are used to introduce referential chains should also be considered to be referential singular terms. So consider (6), as uttered to the homeowner over the phone. (6) A man1 is uprooting your turnips. He1 looks angry. Chastain argued that such an occurrence of a man is just as referential as the following ‘coreferential’ pronoun. (Cf. also Donnellan 1978 and Fodor and Sag 1982. Also the
246
246 Barbara Abbott analysis of Heim 1982, mentioned above, which sees definite descriptions as denoting familiar entities, was paired with a referential analysis of indefinites on which they denote novel entities.) Similar remarks would presumably hold for other weak NPs which may introduce chains of reference—e.g. several children, seventeen squirrels (see e.g. Milsark 1977, Barwise and Cooper 1981, for the weak/strong distinction). However Bach and Ludlow and Neale hold that specific uses of indefinite descriptions such as that in (6), where there is no mutual perceptual contact, are not referential. In this case the speaker could not be said to have referred to anybody; in Bach’s terms, the speaker has only alluded to the man in question (cf. Bach 2006b: 532). The reasoning behind this conclusion depends on the assumption that singular propositions contain actual entities. Since the addressee is perfectly well able to understand the proposition the speaker wants to convey with their utterance, it must not contain an entity that the addressee is unfamiliar with. Of course if singular propositions do not actually contain entities of this sort, the argument would be neutralized. Chastain specifically denied referentiality in the case of uses of indefinites in hypothetical or imaginary contexts such as (7) (Chastain 1975: 202–203). (7) Imagine that a man is uprooting your turnips, and that he looks angry. Here there is no intended connection to something external to the discourse, and probably most philosophers would agree with Chastain that it would not be correct to describe the speaker of (7) as intending to refer to anything. But for a linguist, to put such a gulf between (6) and (7) just does not feel right. Karttunen (1969) introduced the important notion of discourse reference, as distinct from reference to real-world entities, which enables us to group (6) and (7) together, as they intuitively belong. (It should be noted that many linguists view language in general as relating, not to the outside world, but rather to mental representations; cf. e.g. Lakoff 1987; Jackendoff 1990; Chomsky 1995a.)
12.2.5 Quantificational NPs We will return to discourse referents and their relation to the actual world in the following section, but first, we need to complete our catalogue of NPs. One large remaining category consists of strong quantificational NPs like most singers or everyone who came to my party. In the philosophical literature, singular reference is strictly distinguished from quantification, and so none of these kinds of NPs would traditionally have been held to be referential. And it is true that many times these NPs are used to make general kinds of statements, and are intuitively not referential. Some examples are given in (8). (8) a. Every student at ESU will receive a diploma if and when they graduate. b. Most modern vehicles contain catalytic converters.
247
Reference 247 Similarly pronouns which are bound by such quantificational NPs—e.g. they, in (8a)—are not intuitively referential, and seem correctly analysed as similar to the bound variables of logic. Nevertheless, from the linguistic point of view the strict separation between quantificational and referential NPs might seem arbitrary in some cases. Compare the examples in (9): (9) a. The committee members went out to lunch. b. All of the committee members went out to lunch. The propositions expressed in the two sentences of (9) are definitely similar, and the main difference seems to be whether exhaustiveness is made explicit. That is, (9a) more readily allows for an exception or two—one or two committee members who stayed in and ate a bag lunch. (Cf. Horn and Abbott 2012 and the works cited there.) On the other hand, as Sebastian Löbner (2000) has pointed out, the two underlined NPs in (9) do differ in coherence—a property revealed when the examples are negated: (10) a. The committee members didn’t go out to lunch. b. All of the committee members didn’t go out to lunch. (10a) implies that none of the committee members went to lunch, while (10b) allows a reading on which some did and some didn’t.
12.2.6 Generic NPs Our final category is generically interpreted NPs. All but one of the categories of NP that we have looked at so far (the exception being proper names) can be used generically, as in the examples in (11). (11) a. Those/the contestants who finish first always receive a prize. b. If somebody wins a prize they get cited in the local paper. c. A rolling stone gathers no moss. d. All men are mortal. Here even linguists might be less likely to speak unhesitatingly of reference. Intuitively no particular individuals are being spoken of. Instead the statements are quite general, and intended to apply to any and all who fit the characterization spelled out in the subject NP. Yet it is not so clear that singular reference is completely out of the question for generics; another kind of NP with a generic interpretation is the bare NP—either a bare singular with a mass noun as head, or a bare plural. Examples are given in (12).
248
248 Barbara Abbott (12) a. Water is the most plentiful substance on earth. b. Pencils with sharp points are best for writing. The uses exemplified in (12) should be distinguished from uses like those in (13), which are not generic but simply indefinite. (13) a. We put water on the griddle to test for heat. b. There are pencils with sharp points in that drawer. Greg Carlson (1977) argued that a unified analysis of bare NPs could be achieved by analysing them as proper names of kinds of things. On this view the difference between (12) and (13) would lie in the predication rather than the NP—the predications in (12) applying to the kinds, and those in (13) to spatio-temporal stages of individuals realizing those kinds. If this analysis is correct, and we add both kinds and stages to our ontology as possible referents, then generics cannot be automatically excluded from the category of referential NPs. However Carlson’s analysis has not been without challenges; see Carlson and Pelletier (1995). An important subcategory of generic NP contains pronouns whose apparent binding conditions present interesting problems. Classic examples (after Geach 1962: 117) are given in (14). (14) a. Every farmer who owns a donkey beats it. b. If a farmer owns a donkey, he beats it. Gareth Evans (1977a,b, 1980) argued that such pronouns are different from all previous uses we have mentioned—demonstrative, anaphoric, or bound variable. He called such uses e-type; commonly now they may also be called donkey pronouns, because of these classic examples.
12.3 What Are Speakers Referring to? Our next bundle of issues concerns what it is that speakers are referring to, when they are using various kinds of expressions to refer. As we have already observed, this question is by no means independent of the one addressed in the preceding section, concerning which NPs can be said to be usable in acts of referring. We’ll review first our natural language ontology, adding a new kind of concealed referent. We then look at discourse referents, and how they compare with simple things in the actual world; and then finish with a brief look at some cases of reference achieved indirectly.
12.3.1 Natural language ontology We have already seen that, in addition to ordinary concrete physical objects like people and chairs, our linguistic ontology must include a lot of other kinds of things: groups and
249
Reference 249 sums, if Landman and Link are right about the reference of plural and mass terms; and stages and kinds, if Carlson is right about the analysis of bare NPs. It appears that properties may also be referred to (in addition to their usual use in predication), as illustrated in (15): (15) a. The car was bright red. b. Red is my favorite colour. See Chierchia and Turner (1988) for further arguments. There is some evidence for including events as well, as argued most notably by Donald Davidson (1975). On traditional analyses of simple sentences, the verb buttered as it occurs in (16a) below would need to be distinguished from the verb buttered as it occurs in (16b)—the examples are from Davidson. (16) a. Jones buttered the toast. b. Jones buttered the toast with a knife. This is because the verb in (16a) is a two-place verb (i.e. one taking just two arguments), while the verb in (16b) is a three-place verb. But then we cannot account for the very strong intuition that sentence (16b) entails (16a). Davidson’s solution to this problem was to propose that the logical form of such sentences includes implicit reference to an event—in this case an event of buttering whose agent is Jones, whose theme is the toast, and whose instrument is a knife. Space precludes further consideration of events, but see Bach (1986) for an early discussion of types of events and their similarities to concrete objects.
12.3.2 Discourse referents As noted briefly above, Karttunen (1969) first brought to the attention of American linguists the important concept of discourse referents. He was concerned with which occurrences of indefinite NPs could serve as the beginning of a referential chain. As he noted, this was an issue which could not be solved within the boundaries of the sentence, but is rather an issue of the larger context in which an NP finds itself. Karttunen noted contrasts like that in (17). (17) a. She has a car. It/the car is blue. b. She doesn’t have a car. #It/the car is blue. This suggests that failure to entail the existence of an entity is sufficient to block a referential chain. Yet we have already seen above, following Karttunen, that the situation is not as simple as that—that referential chains can be established without real-world existence. Some of his examples are given in (18) (= Karttunen 1969, exx. 11–14).
250
250 Barbara Abbott (18) a. John wants to catch a fish and eat it/the fish for dinner. b. If she had a car, I would have seen it. c. I wish she had a car. She would give me a ride in it. I would drive the car too. d. You have to write a letter to your parents and mail it right away. The letter must be two pages long. Examples like this fall under the heading of what is now termed modal subordination (see, e.g., Roberts 1989). The idea is that, within a discourse, a hypothetical situation may be set up with various hypothetical entities, and these may be spoken of with the usual referential apparatus as long as the reference is within that hypothetical context. Note that if we try to escape the bounds of such contexts, anomaly ensues, as in (19): (19) I wish she had a car. #She gave me a ride in it. The second sentence of (19) is only possible with the (distinctly less preferred) specific interpretation for a car in the first sentence; in that case we are speaking of something which exists in the actual world, and not just in the hypothetical desire-fulfilment worlds under consideration. A different kind of example showing the need to distinguish discourse reference from real-world reference is given in (20) (from Heim 1983, ex. 4). (20) John came, and so did Mary. One of them brought a cake. Here we may imagine that the speaker intends to speak of the actual world, so it is not a question of introducing entities within the scope of imaginings. Nevertheless we have a discourse entity which we cannot match up to a real-world one. Somehow our semantic rules must ensure that the discourse in (20) is true only if either John or Mary brought a cake. Examples such as these, and others— e.g. the donkey pronouns noted above—inspired new ‘dynamic’ approaches to semantics which treat sentences, not as isolated statements with truth conditions, but instead in terms of what they contribute to a discourse—t heir context change potential (see, e.g., Kamp 1981; Heim 1982; Groenendijk and Stokhof 1991; Kamp and Reyle 1993).
12.3.3 Indirect referents4 In this subsection we look briefly at some ways in which NPs and other expressions find referents which are other than the normal straightforward ways. So, consider various 4 It’s important to distinguish this term, in the way I use it here, from a use found in (translations of) Frege (1892). By ‘ungerade Bedeutung’ Frege meant the shift in reference he postulated to occur in propositional attitude contexts, whereby expressions referred to their customary sense. My present use of ‘indirect’ has nothing to do with that analysis of Frege’s.
251
Reference 251 figures of speech such as metonymy, synecdoche, and so forth. Some examples are given in (21). (21) a. Today the White House issued a Presidential Proclamation designating October as National Cybersecurity Awareness Month. (from , 10/03/11) b. The ham sandwich is sitting at table 7. (Nunberg 1993, ex. 96) c. The cat is out of the bag. The White House is a building in Washington DC where the President lives, but the NP the White House is commonly used to refer, not to this building, but instead to the President and/or his or her operatives. Similarly it’s possible for the waitron in a restaurant to refer to someone by what they have ordered. And an utterer of (21c) would most likely be using the cat and out of the bag metaphorically, to speak of some piece of information and its no longer being held in secret, respectively. Geoffrey Nunberg has been responsible for making linguists aware of a particularly interesting indirect use of indexicals. Consider the examples in (22). (22) a. Condemned prisoner: I am traditionally allowed to order whatever I like for my last meal. (Nunberg 1993; ex. 32, underlining added) b. Pointing at the Pope: He is usually an Italian. (Elbourne 2008; ex. 40) As the quantificational adverbs (traditionally, usually) make clear, the intended subjects of the propositions expressed by means of the sentences (22) are not particular individuals but rather classes of individuals sharing a property—in (22a) the property of being condemned to death, and in (22b) the property of being pope. Nunberg (1993) proposed that indexicals have a more complex interpretation than had usually been assumed. Kaplan (1977) had proposed that indexicals contribute their referents to propositions expressed using them. However Nunberg, using examples like those in (22), argued that what an indexical immediately picks out is something that serves as an index, and that what winds up in the relevant proposition may be that item or it may be something different which is related to the index in some way, depending on what is being predicated by the rest of the sentence. (Cf. also Elbourne 2008.)
12.4 Choosing Expressions We turn now to questions surrounding a speaker’s choice of which expression to use to refer to something or someone. Thus for the remainder of this section we will be taking the perspective of the speaker (turning to the addressee in the next), but we must start with an important proviso: it has been clear for some time that achieving reference in conversation often involves both parties. For instance, Clark and Wilkes-Gibbs (1986)
252
252 Barbara Abbott pointed out a number of ways in which reference may be negotiated between speaker and addressee, involving repairs, suggestions on the part of the addressee, instances where speakers continue to provide information until the addressee signals that they have identified the referent, and so forth. This proviso should be kept in mind in the following discussions. We’ll start with some relevant distinctions drawn by Ellen Prince (1981b, 1992), and proceed to the approaches of Gundel, Hedberg, and Zacharski (1993, 2001) and Mira Ariel (1990, 2001), finishing with a cross-linguistic look at reference specifically to persons.
12.4.1 Prince—kinds of new and old information Noting that terms like ‘given’ and ‘new’, as applied to information in discourse, are used in more than one way, Prince (1981b) proposed a taxonomy of referent types with respect to the degree to which they are assumed to be novel to the addressee. Referents which are brand-new are those the speaker assumes are being introduced to the addressee for the first time—these may be either anchored (that is, tied in some way to a known entity) or unanchored. Referents which are new to the discourse but which the speaker assumes that the addressee already knows about from other contexts Prince labelled unused. The category of inferrable (sic) entities contains those whose existence could be expected to be inferred from the existence of prior discourse referents (e.g. the ears on a dog). And finally two subcategories of evoked entities are those which are present textually (i.e. they have been explicitly referred to in the preceding discourse), and those which are present situationally (such as the speaker and addressee, and anything else in the immediate environment). In subsequent work Prince (1992) revised her previous approach with two cross- cutting distinctions—that between discourse-new and -old entities, and that between hearer-new and -old entities. Thus her former category of Brand-New entities are those which are assumed to be new to both the discourse and the addressee, while the Unused entities are new to the discourse but presumed not to be to the addressee. The Evoked entities are old in both senses, while the Inferrables present an interesting intermediate category. Prince’s work with these categories focused primarily on their correlation with grammatical positions. She found that by and large sentential subjects tend to be NPs with Discourse-Old referents—interestingly, it was the discourse factor rather than assumptions about the independent acquaintance of the addressee which is active here. The earlier work also uncovered some interesting differences between written and spoken texts; the former tended to include more reference to entities whose existence can be inferred, but also in many cases it was unclear whether the author of the text would be assuming that their readers were already familiar with the entity in question or would in fact be inferring its existence on the spot. The spoken text was more clear-cut on this issue. (We will return to the difference between written and spoken texts below.)
253
Reference 253
12.4.2 Gundel et al.—familiarity and expression type In some ways the work of Gundel, Hedberg, and Zacharski (1993, 2001) is a natural continuation of Prince’s. What’s new in this work is that specific expressions are associated with specific degrees of familiarity or acquaintance that are assumed by the speaker to hold of the addressee. Gundel et al. propose the givenness hierarchy given below in (23). (23)
in focus
activated
familiar
uniquely identifiable
that N
the N
this pronouns
that this N
referential
indefinite this N
type identifiable a/an N
Each of the cognitive statuses in this chart outranks the one to its right, with ‘type identifiable’ being the weakest. But whereas Prince’s taxonomy resulted in mutually exclusive categories, the categories in the table in (23) are successively absorbing. That is, if an item is in focus, then (according to Gundel et al.) it is also activated, familiar, uniquely identifiable, and type identifiable.5 One of the interesting claims of Gundel et al. is that use of an NP from any point lower in the hierarchy than the extreme left (the ‘in focus’ slot) would be expected to convey a conversational implicature to the effect that no higher relation is assumed to hold. (Prince 1981b: 245 had made a similar suggestion.) So for example, an indefinite description ought to be able to replace any of the NPs in the chart in (23), but doing so would result in an implicature to the effect that the speaker was assuming only that the addressee was familiar with the kind in question and would not be expected even to construct a mental model of the entity being referred to, much less identify it. (See Kehler and Ward 2006 for a critique of this aspect of Gundel et al.’s proposals.) Another claim of Gundel et al. is more problematic—that the expressions listed in the chart in (23) actually encode the cognitive status in question. Thus the abstract of their 1993 paper states: ‘These statuses are the conventional meanings signaled by determiners and pronouns … ’ (Gundel et al. 1993: 274, emphasis added). This claim has been criticized by Kent Bach (1998), who argues that the cognitive statuses are a result of the conventional meanings of the expressions—roughly speaking the more informative an 5
There are some use-mention problems here. Some of these classifications apply to entities (or mental representations of entities) being referred to: ‘in focus’, for example, or ‘activated’; while others apply to the expressions referring to the entities, e.g. ‘referential’. ‘Type identifiable’ and ‘uniquely identifiable’ are categories whose defining properties cross the expression/referent boundary. Thus ‘uniquely identifiable’ is characterized by Gundel et al. (1993: 277) as follows: ‘The addressee can identify the speaker’s intended referent on the basis of the nominal alone.’
254
254 Barbara Abbott expression the weaker the implied cognitive status. For the most part this generalization seems very apt (though see (25) below), although there are some distinctions that do not seem fully explained by it; e.g. the difference between the indefinite article in English (which may be non-specific) and indefinite this (which is necessarily specific; see Prince 1981a), or the difference between that N and the N. Compare the examples in (24) (based on Gundel et al. 1993; exx. 4 and 5). (24) a. I couldn’t sleep last night. The dog next door kept me awake. b. I couldn’t sleep last night. That dog next door kept me awake. For felicitous use (24b) requires an assumption that the addressee has heard about the dog before, but this is not required for (24a).
12.4.3 Ariel—the accessibility hierarchy Bach’s criticisms apply also to the proposals of Ariel (1990, 2001), which are similar to those we have been looking at, although much more highly articulated. Thus Ariel proposes a hierarchy of no fewer than eighteen slots. For example, use of a full name is distinguished on the one hand from use of a full name plus a modifier (e.g. former president Bill Clinton), and on the other from use of just the last name or just the first; long definite descriptions are distinguished from short definite descriptions; and stressed pronouns with gestures are distinguished from stressed pronouns without gestures and from unstressed pronouns. Like Gundel et al., Ariel claims that ‘each referring expression codes a specific (and different) degree of mental accessibility’ (2001: 30, emphasis added); this claim seems even more improbable given Ariel’s highly articulated hierarchy of accessibility. Proposals like these of Gundel et al. and Ariel fail to recognize other aspects of discourse that may play a role in choice of NP. For instance, the fact that definite descriptions encode information about an intended referent makes them useful for helping an addressee identify a referent which has not yet been introduced into the conversation, but they also provide information that may be useful in addition to referent identification. So consider the following, from a gossip column paragraph about Gene Hackman. (25) Hackman also has been writing: His newest novel, Payback at Morning Peak, was released in June. There is at least one man who wants to get the star of Hoosiers and The French Connection back on the big screen: Alexander Payne. The director of the comedy/drama The Descendants, starring George Clooney and now in theaters, is keen on casting Hackman in his upcoming movie Nebraska. (USA Weekend, 11/25–27/11, p. 2) The two underlined definite descriptions are used here to refer to individuals who have not only been introduced into the discourse but who are, at that point, in focus—first
255
Reference 255 Hackman, and then Payne. The reason is obvious—the author of the column wants to get as much information about these referents and their current Hollywood doings into the column as possible, and definite descriptions are a compact way to do that.
12.4.4 Enfield and Stivers—reference to persons The just-cited work of both Gundel et al. and Ariel involved not only examinations of NP types in English, but also other languages—Gundel et al. (1993) examined data from Japanese, Mandarin Chinese, Russian, and Spanish; and Ariel (2001) has many examples from Hebrew, as well as citing data from Chinese, Dutch, Plains Cree, and Nayaka, among others. The papers in Enfield and Stivers (2007) focus on references specifically to persons, but from a cross-linguistic and cross-cultural perspective. There are in principle quite a variety of expressions possible for the initial reference to someone in a conversation: a proper name, a kin term plus or minus an indication of possessor (e.g. in English grandpa vs Bill’s uncle), some other kind of relational expression (e.g. my next-door neighbour), or a definite description not expressing any particular relation (e.g. the actor who plays the Neandertal in all those Geico commercials). The emerging picture from the papers in this volume indicates both commonalities and differences across languages and cultures: one universal generalization seems to be that the primary motivating factor in choice of a referring expression is making it easy for the addressee to recognize who is being referred to. Beyond that languages differ in (a) what the unmarked choice of NP is for an initial reference in a conversation—either a proper name or kinship term plus possessor (e.g. her aunt), and (b) whether economy of expression outranks the need to relate the referent to the conversational participants.
12.5 Determining Referents We turn now to the perspective of the language interpreter, and in particular what kinds of factors go into determining referents. We’ll spend the most time with third-person pronouns, because of the interesting problems they present, but we’ll start with some results from the psycholinguistic literature on definite and indefinite descriptions.
12.5.1 Definite vs indefinite descriptions Recently fine-grained accounts of the process of reference determination by hearers have been made possible with the use of tools such as light-weight head-mounted eye trackers, which allow researchers to monitor the eye movements of subjects as they search a ‘Visual World’ consisting of potential referents. For example, Chambers et al. (2002) examined subjects’ responses to commands involving definite or indefinite
256
256 Barbara Abbott descriptions—e.g. Put the cube inside the/a can, when they were viewing a display of items. Recall that definite descriptions require a unique referent within the relevant universe of discourse; as predicted, subjects hesitated in their processing of definite descriptions when faced with a display containing two cans big enough to contain the cube in question. However, when one of the two cans in the display was too small to accommodate the cube, the subjects’ eyes went without hesitation to the one that was big enough—showing that the domain of possible referents was automatically modified given the exigencies of the task situation. Interestingly, Tanenhaus and Brown-Schmidt (2008) report that a subsequent experiment asking subjects simply Can you put the cube inside the can? found that once again, subjects were hesitant when faced with two cans, even though only one would accommodate the cube. (See also Tanenhaus 2007 for reviews of this kind of research.)
12.5.2 Third-person pronouns Third-person pronouns are notable for the lack of descriptive information they encode. In English this is confined to number and gender, the latter only for singulars. (Pronouns also encode information about their grammatical relations, but that is not directly relevant to determining referents.) Given this lack of descriptive information, other factors must be involved in figuring out who or what a speaker wants to refer to in using a pronoun. One kind of factor has to do with the syntactic location of the pronoun, especially with respect to potential antecedents. As an integral part of his (1981) binding theory, Chomsky proposed two categories for pronouns: anaphors are pronouns which require an antecedent within a narrow range of their occurrence (in English, these are reflexives like themselves and reciprocals like each other); and pronominals are the other pronouns, which cannot have an antecedent within that narrow range. (See Levinson 1991 for a reworking of this theory.) This is a step in the direction of a solution to the reference determination problem, but does not really tell us a lot. We know, for example, that herself in (26a) below (26) a. Louise pays a lot of attention to herself. b. Louise pays a lot of attention to her. must have the same referent as Louise, and that the plain pronoun in (26b) cannot have that referent (at least not as presupposed), but out of context we have no other information about who that referent might be—other than the fact that it is a singular female entity (or a boat). There appear to be at least three additional kinds of factors that go into determining pronominal referents. One has to do with discourse factors such as topic and focus; the second concerns what is being predicated of the referent in question; and the third involves more general issues of discourse coherence. We’ll review the three in turn.
257
Reference 257 A number of scholars have pointed out the relevance of salience in assigning a referent to a pronoun. Thus Craige Roberts suggests that the referent of a pronoun must be ‘the most salient discourse referent’ which meets the number and gender requirements of the pronoun (Roberts 2004: 517). (In order to include demonstrative uses, we must assume that the discourse referents include prominent entities in the extralinguistic context, even if they have not been explicitly mentioned yet.) Salience, in turn, is determined in part, by the question currently under discussion in the conversation, but also by various grammatical properties (Roberts 2003: 331–334). The latter elements are also explored in centering theory—an approach to pronoun interpretation which concentrates on intersentential relations of topichood. In any given sentence a number of entities may be referred to; these can be ordered in terms of prominence depending on their grammatical relations, with the subject being most prominent, direct or indirect objects less so, and adjunct NPs being the least prominent. One of these NPs will typically refer to an entity which was the topic of the preceding sentence—the backward-looking center. This approach ranks various ways in which topics may shift from one sentence to the next, in order of their coherence. So consider the following mini discourse (from Beaver 2004, ex. 1; matching subscripts indicate intended coreference in (27a, b)). (27) a. Janei likes Maryj. b. Shei often brings herj flowers. c. She chats with her for ages. Since Jane is the topic of (27a), she is the backward-looking centre in (27b). Centring theory predicts that she in (27c) will also be construed as referring to Jane, since maintaining a topic from one sentence to the next is a smoother transition than switching it would be, and her will again be taken to refer to Mary. (See the papers in Walker et al. 1998 and Beaver 2004 for more details about this approach.) Although the kind of approach just sketched accounts for many generalizations about pronoun interpretation, there are some that have not been mentioned yet; as noted above, the predications involved in the sentence can have an influence on interpretation. Probably the most frequently cited piece of evidence for the relevance of predication factors is the pair of examples in (28), from Winograd (1972: 33): (28) a. The city councilmen refused the demonstrators a permit because they feared violence. b. The city councilmen refused the demonstrators a permit because they advocated revolution. The pronoun they in (28a) is most naturally taken to refer to the city councilmen, while in (28b) it is most naturally taken to refer to the demonstrators. The reason for these preferences obviously lies in what is being predicated of the referents in the two cases: city councilmen are more likely to fear violence, while demonstrators are more likely to advocate revolution. The approaches just reviewed do not directly account for
258
258 Barbara Abbott this difference in interpretation, although centring theory does make the prediction that (28b) will be slightly more difficult to process, since the demonstrators in the first sentence has a lower status than the city councilmen. Finally, Andrew Kehler (2004) has drawn attention to the third factor alluded to above—discourse coherence. Based on a suggestion of Hume (1748) (as well as work by Hobbs 1990a), Kehler proposes just three basic kinds of intersentential coherence relations—cause–effect relations (where adjacent sentences either assert or deny such a relation), resemblance relations (such as parallel or contrast), and contiguity relations (sequences of events). Consider an example like (29) (Kehler’s ex. 31): (29) a. Carl is talking to Tom in the Lab. b. Terry wants to talk to him too. Despite the fact that, as subject, Carl ranks higher in prominence than Tom, there is a definite preference for construing him as referring to Tom rather than Carl. The predication relations don’t really help here either, since they don’t rule out either referent. Kehler argues that the coherence relation which is operative in this mini discourse is the resemblance relation of parallel, which accounts for the preferred interpretation. Kehler points out that example (30) (his (33)) is ambiguous. (30) Colin Powell defied Dick Cheney, and George W. Bush punished him. Here the semantics of the predications, plus the general factors of discourse coherence, can account for whether him is construed as Colin Powell or Dick Cheney. If the coherence relation is parallel, then him will be understood to refer to Cheney. However given our understanding of the verbs defy and punish, we will most likely infer a coherence relation of cause and effect, with Powell being the punishee.
12.6 Conclusions It was hinted at the outset that we would be taking a constrained view of reference. That was necessary given not only the space constraints of this chapter, but also the existence of other chapters in this volume. Viewed in an unconstrained way, ‘reference’ could be held to subsume the entire domain of linguistic meaning and use. Nevertheless even on our constrained view, reference is a massive topic and one which is actively studied not only by linguists but also by philosophers, psychologists, sociologists, and computer scientists. And there are still many areas that we have not had space here to explore, such as the problems presented by reference to abstract objects (cf. e.g. Asher 1993), temporal variables in reference (cf. e.g. Musan 1999), mass vs count terms (cf. e.g. Allan 1980; Rothstein 2010), vagueness (cf. e.g. Alxatib and Pelletier 2011). The reader is invited to pursue these and other topics on their own.
259
Chapter 13
C ontext Anita Fetzer
13.1 Introduction Context has not only become a major field of research in the arts and humanities, and in the social sciences, but also in information technology and engineering. While the impact of context and contextual features has been acknowledged explicitly in the latter, context itself has become an object of investigation in the arts and humanities and social sciences, where diverse, but not mutually exclusive conceptualizations of context have been implemented and applied to research designs in natural-and non- natural language communication, computer-mediated communication, information technology and robotics, and social-action-based analyses. The multifaceted nature of context and the context dependence of the concept itself have made it almost impossible for the scientific community to agree upon one commonly shared definition of context or one commonly accepted theoretical perspective, and frequently, only a minute aspect of context is described, analysed, or formalized (cf. the interdisciplinary conferences on context: Akman et al. 2001; Blackburn et al. 2003; Bouquet et al. 1999). Context is seen as an interactional achievement in ethnomethodological conversation analysis and interactional linguistics (Heritage 1984; Goodwin and Duranti 1992; Schegloff 1992a); it is considered a relational construct in sociopragmatics (Fetzer 2004, 2010); it is described as a psychological construct in relevance theory (Sperber and Wilson 1986); and it is referred to as ‘other minds’ in functional grammar (Givón 2005). Within these research paradigms, context is further refined as a dynamic construct, which relates interlocutors and the language that they use in a dialectical manner. To capture the dialectics of the dynamic processes, communication has been described as both context-creating and context-dependent (Bateson 1972). In a similar vein, context is seen as imported into communication and as invoked in communication (Levinson 2003a), and in interactional-sociolinguistic terms, context is brought into the communicative exchange and it is brought out in the communicative exchange (Gumperz 1992a).
260
260 Anita Fetzer The interaction-based conceptualization of context is anchored to the premise of indexicality of communicative action, relating an exclusively product-oriented conception of context-as-given which is external to a conversational contribution to the inherently dynamic process of contextualization which is interdependent on a conversational contribution and its surroundings. Context is thus no longer solely a social construct but rather a dynamic sociocognitive construal feeding on the contextualization of communicative action in general, and on the contextualization of communicative acts in particular. Consequently, contextualization has been assigned the status of a universal in human communication, which manifests itself locally with respect to the negotiation and co-construction of meaning in context (Gumperz 1996). The traditional distinction between semantics and pragmatics is based on the relationship between meaning and context: the former is described as the study of context-independent meaning, and the latter is referred to as the study of context-dependent meaning. More recently, the mutually exclusive description of the two fields of research has been blurred by the analysis of meaning in default contexts, accommodating pragmatic principles in the determination of truth conditions (see, e.g., Jaszczolt 2005; Meibauer 2012a). Determining meaning needs to go beyond the level of what is said, considering not only the differentiation between what is said and what is meant, but also between what is said and what is meant in default and non-default contexts. Linguistic pragmatics thus examines the question of how interlocutors do things with (and without) words in context, viz. how they communicate felicitously in context as regards both the production and interpretation of ‘words in context’. It examines the influence of linguistic context (or co-text) and extralinguistic context (or social context) on the production and interpretation of utterances, identifying regularities and non-regularities of language use with respect to the research questions of how interlocutors use language in particular situations, e.g. institutional and non-institutional settings, what purpose they may have at a particular stage in a communicative exchange, and how they may achieve particular goals in communication. The field of experimental pragmatics shares these goals but has refined their research methodology by the explicit accommodation of contextual parameters in order to control the impact of context on the production and interpretation of ‘words in context’, considering in particular generalized conversational implicature and presupposition. On a more general level, pragmatics is fundamentally concerned with communicative action and its felicity in context, investigating action with respect to the questions of what is action, what may count as action, what action is composed of, what conditions need to be satisfied for action to be felicitous, and how action is related to context. These research questions and the object of research require action in general and communicative action in particular to be conceived of as relational concepts, relating action and context, relating action and communicative action, relating communicative action and interlocutors, and relating interlocutors with the things they do with (and without) words in context. The goal of this chapter is to present prominent approaches to context, accommodating (1) a social and sociocultural perspective, conceiving context as social and sociocultural constraints which channel the production and interpretation of conversational contributions accordingly; (2) a text-anchored perspective, conceiving context
261
Context 261 as linguistic material surrounded by other linguistic material and surrounding other linguistic material; (3) a cognitive perspective, conceiving context as a psychological construct; and (4) a relational perspective, conceiving context as relating the constitutive part of a communication by providing the ‘glue’ that makes the parts cohere. The following section contextualizes the concept of context and presents a typology of the most prominent approaches to context. Section 13.3 summarizes possible bridging points between the diverging, but not mutually exclusive frames of reference.
13.2 Contextualizing Context ‘What is a context?’ is the title of a very recent publication of an edited volume on linguistic approaches and challenges to context (Finkbeiner, Meibauer, and Schumacher 2012). Implicit in the question is the presupposition that context exists, while the wh- question implies that there is an answer to that question, and that that answer is valid across different contexts. Finding appropriate answers to that question, however, is not such a straightforward endeavour as the volume informs us. This is not only because of the context dependence of the concept itself but also because of different research designs, different research goals, and different perspectives towards the object of analysis. That is, context may be examined from the participant’s perspective as is reflected in the interactant’s implicit and explicit references to context and in her/his negotiation of the appropriateness of the references in order to co-construct local and global contexts, which are, of course, decontextualized and recontextualized by the analyst in her/his research. Context can also be examined from the analyst’s perspective, abstracting away the participants by concentrating on the conversational contribution and its constitutive parts. While the former captures the dynamics of context, the latter favours a more stable conceptualization of context. To capture the dynamics of context, however, its conceptualization would need to go beyond a simple description or classification of various types of context. It would explicitly need to account for the administration of context as regards the updating of incoming and outgoing contextual information, which may include restructuring operations or deletions. In the following, diverse perspectives on context are presented which conceive context as a multilayered construct. All of them are—more and less explicitly—based on the premises of (1) intentionality of communicative action, entailing conscious participants endowed with rationality who are accountable for their communicative acts; (2) cooperation; and (3) contextualization and indexicality of communicative action.
13.2.1 Participant’s construal In communication, context is not only of relevance to the analyst—for instance, the linguist, philosopher, or sociologist—but also to the participants who produce and
262
262 Anita Fetzer interpret their utterances in context, relating them to external surroundings, to what has just been said and to what is going to be said next. A participant’s construal of context is reflected in negotiation-of-meaning sequences, in particular in her/his acceptance, rejection, or acknowledgement of prior utterances and their constitutive parts. This may be done both explicitly and implicitly—for instance, by saying ‘yeah’, ‘hm’, or ‘that cannot be the case’. Context is thus construed and negotiated by participants in communication. Construals of context can be anchored to the production format, viz. speakers and the footings they may adopt (Goffman 1981; Levinson 1988) and to the reception format, viz. addressee, hearer, audience or overhearers and the footings they may adopt. Dialogue-based conceptualizations of context differentiate between speakers and hearers as individual entities and their individual construals of context, and the set of participants with collective construals, if not a collective construal of context. Collective construals of context need to overlap with the respective individual construals. Sometimes, the individual construals of context and collective construals may conflate. Participant-anchored construals of context lie in the local discourse context. They share the premise that communicative action in general and communicative acts in particular can never be fully explicit, as has been spelled out by Ariel: ‘Underdeterminacy is an inherent characteristic of human language, since no natural-language sentence can encode interlocutors’ intended statements fully’ (Ariel 2008: 265). In a similar vein, but more explicitly, Levinson argues that intentionality is a fundamental premise of natural-language communication: human interaction, and thus communication, depends on intention-ascription. Achieving this is a computational miracle: inferences must be made way beyond the available data. It is an abductive process of hypothesis formation, yet it appears subjectively as fast and certain—the inferences seem determinate, though we are happy to revise them when forced to do so. (Levinson 1995: 241)
Underdeterminacy and the necessary processes of inferencing are connected with the Gricean paradigm and its differentiation between what is said and what is meant, which both refer to utterances produced in context (Grice 1975); and they are also connected with relevance theory’s concept of pragmatic enrichment (Sperber and Wilson 1986). In both frames of reference, communication is seen as a context-dependent endeavour, in which communicative meaning may go beyond the level of what has been said. Hence, what is said cannot be equated with pure linguistic meaning but rather is ‘closely related to the conventional meaning of the words (the sentence) ( … ) uttered’ (Grice 1975: 44). Unlike the rather controversial status of ‘what is said’ in semantics and pragmatics, ‘what is meant’, that is, the sum of ‘what is said’ and ‘what is implicated’ (e.g. Huang 2014: 31–32), has always been equated with non-natural meaning. Grice (1975: 43–44) differentiates between implicate and the related nouns implicature (cf. implying) and implicatum (cf. what is implied). He distinguishes between two basic types of implicature: conventional implicature and conversational implicature.
263
Context 263 The latter is subdivided into generalized conversational implicature and particularized conversational implicature. Generalized conversational implicature is also referred to as default implicature or pragmatic regularities (Bach 2006a, 2007). This is in line with Levinson’s claim that ‘utterance-types carry generalized implicatures ( … ): rational speakers meannn both what they say (except in non-literal uses of language) and what that saying implicates; different layers of meaning all come under the umbrella of meaningnn’ (Levinson 2000: 373). While conventional implicature is connected closely with linguistic form, for instance with connectives (e.g. but), implicative verbs (e.g. manage, forget to), honorifics, or non-restrictive relative clauses, conversational implicatures are essentially connected with certain general features of discourse, viz. dovetailedness and the maxims, the turn-taking mechanism, and sequentiality. To communicate felicitously, speakers need to presuppose certain brute facts about the world and of the world they communicate about, and they anchor their communicative acts to discourse, or, to employ Gricean terminology, they make their conversational contribution ‘such as is required, at the stage at which it occurs, by the accepted purpose or direction of the talk exchange in which you are engaged’ (Grice 1975: 45). From a construal-of-context perspective, this means that they make their contribution as explicit ‘as is required’ and that they import the appropriate amount of context, which is required to secure felicitous communication. Speakers may import context into the discourse through conventional means, such as deictic expressions, referring to person (and their social positioning), location, time, and discourse. They may import context in a more direct manner through quotations, which generally require a lower effort on the hearer’s side as regards inferencing. This is because quotations contain a higher degree of entextualized objects and therefore are more explicit than the use of an indexical expression only. The lower effort on the hearer’s side also holds for context importation through generalized conversational implicatures. Context may also be imported into discourse through more particularized means, e.g. interactional-sociolinguistic-based contextualization cues (Gumperz 1996) or particularized conversational implicatures. While the latter rely on the Gricean cooperative principle and its maxims, in particular relation, as well as on context to retrieve the speaker’s communicative intention, the former are meta-linguistic indexicals, whose communicative meaning is calculated against the background of their co-occurrence with other linguistic and semiotic devices. Contextualization cues can be realized phonologically, such as intonational contours, stress, and pauses; they can be realized lexically, for instance as particles and meta-communicative comments; and they can be realized non-verbally. They ‘serve to retrieve the contextual presuppositions conversationalists rely on making sense of what they see and hear in interactive encounters. They ( … ) have no propositional content. That is, ( … ) they signal only relationally and cannot be assigned context-free lexical meanings’ (Gumperz 2003: 9). Regarding their function in discourse, they import context into the discourse and they bring context out in the speech activity by channelling ‘inferential processes that make available for interpretation knowledge of social and physical worlds’ (Gumperz 1996: 383).
264
264 Anita Fetzer
13.2.1.1 Production-based construal Speaker-anchored context is a participant-anchored and production-based construct par excellence, whose prime focus lies in the local discourse context. Adopting a bird’s eye perspective on the world rather than on discourse and its participants, Penco (1999) transcends the speaker-centred outlook on context, differentiating between subjective context and individual context. The former refers to a subjective, viz. cognitive (or epistemic), representation of the world, and the latter refers to an individual representation of the world. Both may be identical but need not be. That is because subjective context refers to an individual set of beliefs, which may belong to an individual participant or to a community, and individual context refers to an individual representation of the world, which is functionally synonymous with the set of beliefs of an individual participant, which may be quite idiosyncratic. The differentiation between subjective context and individual context is of great importance to the analysis of a participant-centred construal of context as it allows for an explicit distinction between the sociocognitive construal of individual context anchored to a single participant only, and the participants’ construal of—in Penco’s terms—subjective context, which is negotiated by the participants in and through the process of communication and thus is shared—to some extent—by the participants. In an ethnomethodological frame of reference, Penco’s conceptualization of ‘subjective context’ would be referred to as the social construct of context. From a language-usage perspective, participants construe context in the process of communication, and they co-construct context through the process of communication. They construe local context against the background of prior conversational contributions, and they co-construct local context for upcoming contributions. In Heritage’s terms, the production of talk is doubly contextual (Heritage 1984: 242), viz. an utterance relies upon the existing context for its production and interpretation, and it is, in its own right, an event that shapes a new context for the action that will follow. This is because conversational contributions contain context and they are at the same time contained in context. However, ‘doubly contextual’ may not only refer to external context—that is, linguistic context (or co-text) and social context. ‘Doubly contextual’ may also refer to the dynamics of context construals, indexing individual context—in Penco’s terms—and collective context, viz. a social construct of context, which is negotiated and co-constructed in communication. More recently, the focus on the production format as the sole creator of meaning has shifted to analysing the role of the reception format in natural-language communication, and to the conceptualization of natural-language communication as a dyadic endeavour (Linell 1998) or a joint undertaking (Clark 1996). Reception-based construals of context, in particular hearer-and collective-centred context, which are co- constructed contexts par excellence, are presented in the following.
13.2.1.2 Reception-based construal The importation of context and its invocation in discourse are complementary if one subscribes to communication as a dialogic (or dyadic) process. Against this
265
Context 265 background, context importation can only be felicitous if the hearer (or audience) takes up the speaker-intended importation of context and invokes the intended context, as is reflected in the universal of contextualization in interactional sociolinguistics and in pragmatic enrichment in cognitive pragmatics. Both inferencing processes are hearer- and collective-centred concepts. In communication, speakers intend hearers to construe certain contexts. While they may import particular types of context through local conventional and non- conventional means—for instance, deictic expressions, quotations, generalized and particularized implicatures, and contextualization cues—the invocation of context is done by the employment of more globally oriented conventional and non-conventional means, such as style, register, social deixis, and generalized and particularized implicatures exploiting deviations from how things are typically done. This is reflected, for instance, in the use of informal expressions or non-standard phonetic realizations in political discourse or academic discourse, or in formal expressions and standard phonetic realizations in vernacular-based interactions. The importation and invocation of context is based on the premise that language is a socially situated form and that language variation and alteration are not random or arbitrary, but communicatively functional and meaningful. In interactional sociolinguistics, language use is always embedded in the delimiting frame of a speech activity, and against that background, the contextualization of utterances is performed through the local and global cognitive operations of conversational inferencing: It is useful to distinguish between two levels of inference in analyses of interpretive processes: (a) global inferences of what the exchange is about and what mutual rights and obligations apply, what topics can be brought up, what is wanted by way of a reply, as well as what can be put into words and what is to be implied, and (b) local inferences concerning what is intended with my one move and what is required by way of a response. (Gumperz 2003: 14)
The distinction between different types of meaning inferred through deductive and non-monotonic reasoning, viz. between the discrete operators of a formal language and their natural-language counterparts, is further refined by Levinson, pointing out their subjectively determinate nature: ‘Conversational inferences have a number of very special properties: they are speedy, they are non-monotonic (the same premises can give different conclusions in different contexts), they are ampliative (you get more information out than went in) and they are subjectively determinate’ (Levinson 1995: 238). Conversational inference and contextualization are connected closely. Both require particular inference triggers to initiate a process of context-dependent cognitive operations, e.g. inductive, deductive, and abductive reasoning. A device, which is of key importance in that respect, is the interactional-sociolinguistic contextualization cue. Context, contextualization, and contextualization cue are relational concepts. To
266
266 Anita Fetzer account for the micro–macro interface, the explicit accommodation of context is a necessary condition, as Gumperz points out: With respect to context, psychologists, cognitive scientists, and many linguists who pay attention to context tend to define it almost entirely in extra-communicative terms. I argue that, while these factors are, of course, significant, contextual information is imported into the interpretative process primarily via indexical contextualization cues, in the form of presuppositions of what the activity is and what is communicatively intended. (Gumperz 2003: 119)
From the perspective of participants, context is construed in and through the process of communication, it is negotiated and co-constructed and thus an interactional achievement, as is examined below.
13.2.1.3 Context as an interactional achievement In ethnomethodological research (Garfinkel 1994), context is negotiated, interactionally organized, and co-constructed, as is illustrated with the following excerpt from a political interview recorded and transcribed from the programme ‘On the Record’ (BBC1). To keep the scenario as minimal as possible, let us consider three participants: interviewer (Jonathan Dimbleby), interviewee (Tony Blair as the then leader of the opposition), and an overhearing audience, viz. the media audience, who is not directly involved in the interaction, but at whom the conversational contributions are directed in the first place. The interviewer makes the following contribution, which contains the italicized yes/no- question: ‘In broad terms, as a result of the policy that you have drawn up, which you believe to be more fair. Will unions have more muscle?’ In the context of a mediated political interview, both the interviewer and the audience do not expect the interviewee to simply produce a minimal agreement or disagreement by saying ‘yes’ or ‘no’. Rather, they expect him to argue his case, which he complies with by replying ‘no I don’t think it’s a question of that at all. I think what most people believe about trade unions and this in a sense is relevant within the document is that trade unions are effective in representing their members. They should be allowed to be effective in representing their members. They should also be accountable under the law, and so what we have done throughout this document is where we have provided rights we’ve also provided for responsibilities.’ During that exchange, the interviewer, interviewee, and audience construe a—more or less—common context, which emerges in the interviewer’s conversational contribution and his reference to a policy document drafted by himself and the Labour Party making explicit their position about the rights and obligations of trade unions. At the same time, the reference to that document indexes contextual frames about (British) trade unions, ballots, sympathy action, and strikes. The construal of that kind of shared context amongst the face-to-face interactants is reflected in the interviewee’s references to (1) trade unions and the effective representation of their members, and (2) the provision of rights and responsibilities which is to ensure that possible action be accountable under the law. On a more general level, the construal of context is manifest in the interlocutors’ contributions which accept or reject a prior contribution-as-a-whole, or which accept
267
Context 267 or reject one or more of the constitutive parts of a contribution, including their presuppositions. In the Searlean parlance, the contributions are recognized, with recognition going ‘all the way from enthusiastic endorsement to grudging acknowledgments, even the acknowledgment that one is simply helpless to do anything about, or reject, the institutions in which one finds oneself ’ (Searle 2010: 8). Context is negotiated in so far as the interactants refer to context, import it, and invoke it through their contributions. Their agreeing or disagreeing with prior conversational contributions and their constitutive parts makes manifest whether they accept a particular construal of context or whether they reject it. The interactional-achievement perspective on context contains a number of possible construals of context anchored to the participants and their interactional roles: the speaker’s construal, the hearer’s construal, and the audience’s construal, or rather the individual audience-members’ construals of context. While the direct face-to-face interactants may negotiate both their individual and collective construals of context by agreeing in a more or less explicit manner with what has been construed in their representation of context thereby co-constructing a shared context, the overhearing audience—unless they engage in a process of meta-communication—does not generally negotiate their construal of context in that manner. In the following, a typology of context based on the connectedness between language and language use, language use and society, and language use and the cognitive system is presented, as is reflected in linguistic context, social and sociocultural context, and cognitive context.
13.2.2 Analyst’s construal Construals of contexts by the analyst are based on stretches of discourse—or stories, in Meibauer’s terms (2012b)—which count as pragmatic evidence. The following typology of context does not focus on how the participants in the interaction construe context and on how they negotiate and co-construct context, but rather on what that thing called context contains. From a language use perspective, context contains linguistic material referred to as linguistic context (or co-text). Linguistic context comprises grammatical constructions (or parts) embedded in adjacent grammatical constructions (or further parts), composing a whole clause, sentence, utterance, turn, or text. Social and sociocultural material is referred to as social and sociocultural context. Both comprise the context of a communicative exchange and are defined by deducting linguistic context and cognitive context from a holistic conception of context. Constituents of social context are, for instance, participants, the immediate concrete, physical surroundings including time and location, as well as macro contextual institutional and non-institutional domains. Sociocultural context represents a particularization of social context, coloured by culture-specific variables. Cognitive material is referred to as cognitive context. Cognitive context is the foundation on which inference and other forms of reasoning are based, and thus is indispensable for the interpretation of language and language use, and other semiotic codes. Constitutive elements of cognitive context are mental representations,
268
268 Anita Fetzer propositions, contextual assumptions, and factual assumptions. Since cognitive contexts are anchored to an individual but are also required for a cognitively based outlook on discourse and communication, they need to contain assumptions about mutual cognitive environments.
13.2.2.1 Linguistic context Linguistic context comprises the actual language use delimited by a clause, sentence, turn, or text. From a parts–whole perspective, linguistic context (or co-text; cf. de Beaugrande and Dressler 1981; Janney 2002) denotes a relational construct composed of local and not-so-local adjacency relations. The connectedness amongst a grammatical construction (or a part) and other grammatical constructions constituting a text (or the whole) is looked upon analogously to Searle’s conception of regulative rules and constitutive rules (Searle 1969). That is to say, the rule-governed realization of grammatical constructions in context constitutes an utterance act, which counts as a move within the game of producing and interpreting utterance acts. At the same time, the utterance act counts as a move within the game of producing and interpreting speech acts in context. While the rule-governed realization of grammatical constructions is constrained by the rules of grammar, the production and interpretation of speech acts are constrained by felicity conditions. The production and interpretation of an utterance act is anchored to the constitutive parts of language: syntax, morphology, phonology, semantics, and pragmatics. While syntax is composed of structural units—for instance, constituents in traditional grammar, phrases in functional grammar and generative grammar, groups in systemic functional grammar, or constructions in construction grammar—it is the linear ordering of the individual parts within a hierarchically structured sequence which constitutes their grammatical function. The adverb really, for instance, realizes the grammatical function of a sentence adverbial with wide scope if positioned initially or finally, as is the case in the utterance Really, Tom is a dedicated linguist. If the adverb really is positioned medially, it is assigned the grammatical function of the adverbial of subjunct with narrow scope, as in Tom is a really dedicated linguist. Or, the proper noun Steve can realize the grammatical function of object in Sally insulted Steve, and it can realize the grammatical function of subject in Steve insulted Sally. Thus, it is not the grammatical construction as such, which is assigned a grammatical function. Rather, it is the positioning of a grammatical construction within a hierarchically structured sequence, which assigns it a grammatical function. The relational nature of linguistic context is also reflected in a sentence’s topological units of pre-field, middle-field, and post-field, and their respective subfields, which are also conceived of in relational terms. For instance, a change in the canonical word order SVO in English in the utterance Stephen criticized Peter to a non-canonical OSV Peter Stephen criticized with stress on initial O does not change the propositional meaning of the utterance. From a discursive viewpoint, however, the fronting of the object signifies a contrastive set. That is, the speaker intends the message that Stephen criticized Peter while at the same time implicating that Stephen did not criticize other not-named but
269
Context 269 presupposed members of the contrastive set, for instance Tom, Larry, or Mark. The investigation of syntactic structure from a context-anchored parts–whole perspective has demonstrated that the whole, viz. the whole utterance, is more than the sum of its parts. This is because the linear ordering of the constitutive parts gives off additional discursive meaning. An investigation of morphology from a context-based perspective sheds further insights into the morphological processes of inflection, derivation, and compounding. In a context-anchored analysis of morphological processes, inflection is seen as a context- construing device, which may signal the grammatical status of words and their potential status in a clause or sentence. The inflectional morpheme [s] in the word form [[drink][s]], for instance, signifies the type of connectedness between the constitutive parts of a lexical form by making explicit the grammatical function of plural in the NP five drinks as in the utterance he had five drinks. It may also signify the grammatical function of tense in the VP drinks as in Sue drinks more than Mark. Again, it is not the positioning of the morpheme within a lexical form but rather the positioning of the lexical form within a lexical phrase, which contributes to the lexical or grammatical status of the morpheme under investigation, and to its particularization as a marker of tense or number. Derivation and compounding may also be analysed against the background of making explicit the connectedness between morphemes in the context of lexical expressions, such as the prepositional verb [[take] [over]] and the compound [[over] [take]], or the derivational affix and free morpheme [ism] in [[contextual][ism]] and in There’s been too much ism lately. As has been shown for syntactic structure, an analysis of morphology from a context-anchored parts–whole perspective leads to more refined results as regards inflection, derivation, and compounding. An explicit accommodation of context from a parts–whole perspective in phonology also leads to stimulating new insights. Here, assimilation is looked upon as the adaptation of a part to its phonological context (or whole). The following sequence, ten pencils, consists of two parts, [ten] and [penslz]. When realized as a whole, the alveolar nasal [n]is adapted to its local phonological context [p] and realized as another bilabial sound [m] in the phonological sequence [tempenslz]. The context-anchored parts– whole perspective is further manifest in the realization of a phonological form as a full or reduced form. For instance, the preposition to in the sequence I cycled to university can be realized as a reduced form with a schwa-sound stressing the location to which the speaker went, namely university, or it can be realized with the monophthong [u] as a full form stressing the direction of the movement. A context-anchored perspective is also of importance in the field of supra-segmental phonology, and here in particular in intonation and intonational phrases. In systemic- functional grammar, intonation is seen as a signalling system (Halliday 1994), and in interactional sociolinguistics, an intonational phrase is assigned the status of a contextualization device (Gumperz 1996). For instance, the intonational contour of a fall signifies the illocutionary force of directive in English, and the intonational contour of rise signifies the illocutionary force of offer. The one-word utterance beer realized with
270
270 Anita Fetzer a fall contextualizes the utterance as a directive, e.g. I want beer or take that beer, while its realization with a rise contextualizes it as an offer, have some beer or are you really offering me beer. Again, a context-anchored parts–whole perspective leads to exciting insights by showing that the phonological realization of an utterance contextualizes a speaker’s communicative intention (Pierrehumbert and Hirschberg 1990). Semantics has been traditionally defined as the investigation of context-independent meaning while pragmatics has been promoted as the investigation of context-dependent meaning. From a parts–whole perspective, truth-conditional semantics examines the meaning of a whole proposition by identifying its constitutive parts of reference and predication. Whenever all of the constitutive parts are true, the meaning of the whole proposition is true. In that frame of reference, the propositions Mary does not do anything to change her life and Mary does not do nothing to change her life do not share the same truth conditions and therefore are not identical. From a pragmatics-based outlook, however, they may share the same communicative status in communication, for instance, as a realization of the speech act of rejection. Possible-worlds-anchored semantics restricts the investigation of a proposition’s meaning from truth conditions which are valid in any context to that of one of its subsets, a possible-world scenario. Here, the meaning of a proposition is true in a specified scenario only. Discourse semantics focuses on anaphora resolution, cohesion and coherence, and lexical semantics examines the semantic meaning of lexical expressions, such as large. Again, the explicit accommodation of a context-anchored parts–whole perspective may lead to more refined results, as has been pointed out by Akman and Alpaslan (1999: 10) in their examination of the meaning of the adjective large. In the utterance Sally built a large snowman the lexical meaning of the adjective large is interdependent on the size of the discourse identity of Sally. If Sally is a toddler, ‘large’ denotes a size of about 1.20 meters, and if Sally is an adult, ‘large’ denotes a size of about 1.80 meters. In the following, cognitive context which is a necessary condition for a cognition- based theory of language and language use is examined.
13.2.2.2 Cognitive context Cognitive context is not only of relevance to cognitive linguistics and cognitive pragmatics, but also to the field of psychology, and here in particular to the psychology of communication. Bateson (1972) conceives context along the lines of the gestalt- psychological distinction between figure and ground and the related concepts of frame and framing. Frame is seen as a delimiting device which ‘is (or delimits) a class or set of messages (or meaningful actions)’ (Bateson 1972: 187). Because of its delimiting function, ‘psychological frames are exclusive, i.e. by including certain messages (or meaningful actions) within a frame, certain other messages are excluded’ and they are ‘inclusive, i.e. by excluding certain messages certain others are included’ (ibid.). The apparent contradiction is eradicated by the introduction of set theory’s differentiation between set and non-set, which—like figure and ground—are not symmetrically related. To use Bateson’s own words: ‘[p]erception of the ground must be positively inhibited
271
Context 271 and perception of the figure ( … ) must be positively enhanced’ (Bateson 1972: 187). This leads him to the conclusion that the concept of frame is meta-communicative, which also holds for context. Or in his words: ‘the hypothesis depends upon the idea that this structured context also occurs within a wider context—a metacontext if you will—and that this sequence of contexts is an open, and conceivably infinite, series’ (Bateson 1972: 245). Bateson explicitly connects set and non-set, frame and meta-frame, and context and meta-context with a parts–whole perspective: ‘whenever this contrast appears in the realm of communication, [it] is simply a contrast in logical typing. The whole is always in a metarelationship with its parts. As in logic the proposition can never determine the metaproposition, so also in matters of control the smaller context can never determine the larger’ (Bateson 1972: 267). The concept of frame is fundamental to the construction of meaning: ‘In general, then, the assumptions that cut an activity off from the external surround also mark the ways in which this activity is inevitably bound to the surrounding world’ (Goffman 1986: 249). While the connectedness between frame and framing needs to be based on meta-representation, framing also needs to be recursive. Cognitive context is not only of key importance to the psychology of communication, but also for language processing and the corresponding inference processes involved. Relevance theory (Sperber and Wilson 1986) differentiates between cognitive environment and cognitive context: the former refers to a set of facts, while the latter refers to a set of premises, namely, true or possibly true mental representations. Constitutive elements of cognitive context are mental representations, propositions, contextual assumptions which may vary in strength, and factual assumptions. Assumptions are read, written, and deleted. In the meantime, contextual implications are raised in strength, lowered in strength, or erased from memory. Since cognitive contexts are anchored to an individual but are also required for a cognitively based outlook on communication, they must contain assumptions about mutual cognitive environments. Thus, cognitive context is not only defined by representations but also by meta-representations. To describe multilayered cognitive context, relevance theory employs the onion metaphor and represents context as an onion with its constitutive layers. What is of importance for language processing and inferencing is the premise that the order of inclusion corresponds to the order of accessibility. This ensures that both processes are ordered, and that their order is based on meta-representations, meta-layers, and meta-contexts. In functional grammar, context also denotes a psychological construct, which Givón (2005: 91) explicates in Context as Other Minds as follows: First, we noted that context is not an objective entity but rather a mental construct, the construed relevant ground vis-à-vis which tokens of experience achieve relatively stable mental representation as salient figures. Whatever stability mental representations possess is due, in large measure, to the classification of tokens of experience into generic categories or types.
272
272 Anita Fetzer What is important to the investigation of cognitive context is the differentiation between types of experience and tokens of experience. While the former are of prime relevance to language processing and inferencing, the latter are intrinsically connected with practical reasoning and abduction, in and through which tokens are categorized into types. Contexts are not objective or deterministic constraints of society or culture, but subjective participant interpretations, which are negotiated in communicative exchanges and thus assigned the status of social constructs. In Givón’s terms, the negotiation of construals of contexts requires the individual construals to be categorized as types of context, which display a higher degree of stability. The token-type differentiation is not only of relevance to the micro domain. It has been extended to the meso domain of genre, channelling and filtering the production and interpretation of conversational contributions, as is made explicit by Thibault (2003: 44): Rather, genres are types. But they are types in a rather peculiar way. Genres do not specify the lexicogrammatical resources of word, phrase, clause, and so on. Instead, they specify the typical [emphasis is original, A.F.] ways in which these are combined and deployed so as to enact the typical semiotic action formations of a given community.
Cognitive context is a structured and multilayered whole, which is indispensable for language production, language processing, and inferencing. The nature of the connectedness between its constitutive layers and subsystems is meta-communicative and meta-systemic. In the following, social and sociocultural context is examined more closely.
13.2.2.3 Social and sociocultural context Social context is often considered to comprise the context of a communicative exchange and is defined by deducting linguistic context and cognitive context from a holistic conception of context. Constituents of social context are, for instance, participants, the immediate concrete, physical surroundings including time and location, and the macro contextual institutional and non-institutional domains. Frequently, language use in social contexts has been allocated to communicative performance (or parole) which has been assigned the status of an individual and momentary product. This has not only been denied by ethnographic and speech-act-theoretic studies, but also by artificial- intelligence-anchored analyses (Gumperz and Levinson 1996; Recanati 1998; Sbisà 2002). The non-individualistic use of language is also manifest in the contextual phenomenon of deixis and its realization as deictic expressions, viz. temporal deixis, local deixis, participant deixis, discourse deixis, and social deixis. Unlike anaphora resolution, which requires linguistic context, discourse deixis is informed by both linguistic and social contexts. In sociocultural context, which is a particularization of social context coloured by cultural variables, e.g. mono-and polychronic conceptualizations of time, or mono-and
273
Context 273 multidimensional conceptualizations of participants, the categories of speaker, hearer, and audience are no longer seen as analytic primes. Rather, they denote interactional categories and have been refined by Goffman (1981) with respect to their footing anchored to the participation framework. In an actual speech situation, however, the interactional roles do more than simply produce and interpret conversational contributions. In a sociocultural context, they subcategorize into social roles and their gendered and ethnic identities, to name but the most prominent ones. In institutional communication, the participants’ institutional roles embody institutional power as is reflected in their context-dependent rights and obligations. The research paradigm of ethnomethodology investigates the interactional organization of society. It represents a micro sociological perspective based on the premise of indexicality of social action. Ethnomethodology focuses on intersubjectivity and examines the questions of how separate individuals are able to know or act within a common world, and of how members (or participants) negotiate or achieve a common context: ‘in an interaction’s moment-to-moment development, the parties, singly and together, select and display in their conduct which of the indefinitely many aspects of context they are making relevant, or are invoking, for the immediate moment’ (Schegloff 1987a: 219). Here, common context is synonymous with sociocultural context, whose relevance is spelled out by Hanks (1996) as follows: Hence it is not that people must share a grammar, but that they must share, to a degree, ways of orienting themselves in social context. This kind of sharing—partial, orientational and socially distributed—may be attributed to the habitus, or relatively stable schemes of perception to which actors are inculcated. (Hanks 1996: 235)
Another culture-dependent outlook on communication has been promoted by ethnography of communication (Saville-Troike 1989), in particular by the concept of speaking grid (Hymes 1974). Hymes systematizes the embeddedness of communication with respect to its constitutive components of situation (the physical setting and the psychological scene), participants (speaker, hearer, and audience, and their statuses in the participation framework), ends (the goal and the purpose of the speech event from a sociocultural viewpoint), act sequence (how something is said with regard to message form and what is said with regard to message content), key (mock or serious), instrumentalities (channels, i.e. spoken, written, e-mail, multimodal), norms of interpretation and forms of speech (vernacular, dialect, standard), and genre. The concept of speaking grid has been refined by Gumperz (1992a) who explicitly connects the cognitive operation of inference with the sociocultural activity of conversation. His conception of conversational inference represents a context-bound process of interpretation in which others’ intentions are assessed, and in which “self ” illustrates their understanding and comprehension through their response. Gumperz assigns language usage the status of actual language practice, and he interprets evaluation as a social activity. Saville-Troike summarizes his contribution to the ethnography of communication as follows: ‘Gumperz builds on this in proposing the outline of a theory of
274
274 Anita Fetzer how social knowledge is stored in the mind, retrieved from memory, and integrated with grammatical knowledge in the act of conversing’ (1989: 131). Because of its cultural base, the meaning that emerges in conversation is different for participants if they are not members of the same speech community. Gumperz’s original contribution to the field of ethnography of communication bridges the gap between linguistic context on the one hand, and sociocultural and social contexts on the other, and between linguistic, social, and sociocultural contexts on the one hand, and cognitive context on the other. This is because (1) language is seen as a socially situated form, (2) inference is given a context-dependent interpretation, and (3) its unit of investigation, the speech activity, is a meso category going beyond the micro domain of isolated speech acts, utterances, or turns. Speech activities (or genres) provide some kind of blueprint which embed them in local context while at the same time delimiting them from more global context.
13.3 Context: Dynamic and Relational The theoretical construct of context has been described in different research paradigms, and depending on their goals, various aspects are highlighted, such as the importation and invocation of context in sociopragmatics, context as a psychological construct in relevance theory and in cognitive grammar, or context as a set of antecedent premises which are required for a speech act or discourse act to be felicitous. Furthermore, context is conceptualized along the distinction between context as type and context as token, and thus assigned the status of a more generalized and a more particularized context. The classification of context into micro and macro linguistic, cognitive, social, and sociocultural context types, and the accommodation of the meta-linguistic device of a contextualization cue, which expresses relational meaning by signifying the nature of the connectedness between objects-in-context and context-as-a- whole, is a first move towards a theory of context as a dynamic construct. In order to account for the dynamics of context in an appropriate manner, a further distinction needs to be introduced, namely one between default and non-default context (Recanati 1998). Against this background, the different types of context analysed above require a further differentiation into default contexts, or unmarked contexts in functional-grammar terminology (Givón 1993), and non-default (or marked) contexts. In functional-grammar terms, the marked category is more complex regarding structure, less frequent regarding distribution, and harder to process, while the unmarked category is less complex regarding structure, more frequent regarding distribution, and less hard to process (Givón 1993: 178). The classification of context as default and non-default is performed through conversational inferencing. Analogously to the Gricean implicature ‘inferring also involves a two-step process in which the contextual
275
Context 275 ground, in terms of which an assessment of what is perceived is made, must be first retrieved and related to stored memories before an interpretation is arrived at’ (Gumperz 1996: 383). For the connectedness between contextualization and default/non-default context this means that unmarked contexts are less complex regarding structure, that is to say the process of contextualization does not need to accommodate ‘more’ contextual information ‘than is required’ (Grice 1975: 45) for the current purpose of the talk exchange. Furthermore, unmarked contexts are more frequent and thus defined by a higher degree of conventionalization, and they are less hard to process as both their structural configuration and the higher degree of conventionalization are internally documented in culture-dependent default frames, or stored memories in Gumperz’s terminology. Marked contexts, by contrast, deviate from the default configuration by being more complex regarding structure. That is to say, the process of contextualization needs to accommodate more contextual information, and thus needs to go beyond the default frame. Marked contexts are less frequent and therefore defined by a lower degree of conventionalization, and they are harder to process as all of the surplus contextual information, which is encoded in more tokens, needs to be administered and attributed to the appropriate types so that all of the contextual information may be stored in the default frames. The dynamics of context is implicit in the relevance-theoretic framework, and it is explicit in cognitive grammar and pragmatics. Another necessary differentiation of context is provided by the interactional-sociolinguistic conception of context as given and reconstructed. Against this background, context can no longer be seen as an analytic prime. Rather, context is dynamic and relational, it is subjective and individual (Penco 1999), it is social and institutional (Goodwin and Duranti 1992), and it is a parts–whole configuration. The multilayered outlook on context requires an analytic frame of reference based on methodological compositionality informed by linguistics, psychology, sociology, linguistic anthropology, and cultural studies. Only then is it possible to cross and transcend disciplinary boundaries and account for inherently unbounded theoretical constructs, which may become bounded when instantiated. The most appropriate delimitation thus seems to be a functional one: context is conceived of as a frame of reference whose job it is to frame content by delimiting the content while at the same time being framed and delimited by less immediate adjacent frames. To account for multilayered, dynamic context in pragmatics, communication needs to be anchored to the basic pragmatic premise of intentionality of communicative action (Austin 1976; Brandom 1994; Fetzer and Akman 2002; Searle 2010). Not only are context and intentionality of relevance to the investigation of speech activities, but so is the contextualization of a conversational contribution and of its constitutive parts. To use Gumperz’s own words: ‘Since all interpretation is always context-bound and rooted in collaborative exchanges that rest on shifting contextual presuppositions, contextualization must be a universal of human communication’ (1996: 403). Contextualization is connected intrinsically with enriching a conversational contribution or some of its
276
276 Anita Fetzer parts, and at the same time it is connected with decontextualization and entextualization, thus bridging the gap between internal and external contexts, micro and macro contexts, and context importation and context invocation. In the process of contextualization, indexical tokens are assigned values. In the process of recontextualization, the values assigned to tokens in a particular context are adapted to the constraints and requirements of a different context and thus re-evaluated, and in decontextualization, an indexical token is extracted from its context, re-evaluated, and assigned a default value, which obtains in a default context. In the process of entextualization, the referential domain of an indexical expression is made explicit and an unbounded object is assigned the status of a bounded object (Fetzer 2011). In a dynamic theory of context, boundedness and unboundedness are conceptualized as scalar notions which are delimited by the meso concept of genre. It provides a frame of reference in which the relevant information for the production and interpretation of context-dependent information is stored, and which functions as some kind of blueprint for appropriate utterance production and utterance interpretation, channelling the necessary cognitive operations accordingly. A dynamic theory of context needs to be anchored to the basic pragmatic premises of intentionality, indexicality, and contextualization. This requires the accommodation of cooperation and collaboration on the one hand, and micro, meso, and macro referential domains on the other. In such a meta-systemic framework, context can be delimited in spite of the fact that ‘this sequence of contexts is an open and conceivably infinite series’ (Bateson 1972: 245).
277
Pa rt I I I
M AC RO - P R AG M AT IC S A N D C O G N I T ION
278
279
Chapter 14
C o gnitive Pr ag mat i c s Bruno G. Bara
14.1 Why ‘Cognitive’? Communication is a social activity of a combined effort of at least two participants, who consciously and intentionally cooperate to construct together the meaning of their interaction. Cognitive pragmatics is the study of the mental states of people who are engaged in communication. The adjective ‘cognitive’ is added here because our goal is to investigate what hap pens in participants’ minds rather than the message ‘per se’. The focus is on participants’ mental processes rather than on the formal structure of the message. Human communication is completely different from communication in other spe cies (see Hauser 1996 and Tomasello 2008 for its evolutionary reasons). It is also differ ent from communication between two machines and between a human and an artifact. Obviously there may be resemblances, but the differences are striking: human com munication is based upon two powerful minds/brains, and if one of the two or both are missing, one cannot speak properly of communication.
14.2 State of the Art It is a cliché to mention the innumerable definitions of pragmatics, and the situation becomes even more complex if one has to consider cognition together with commu nication. For the sake of simplification, one could ask: which books deal with cog nitive pragmatics? We cannot consider authors like Wittgenstein, Austin, Grice, and Searle: they have been and are immensely influential in the field, but are not concerned with mental processes. Philosophers discuss the steps of communication processes and describe them using logical terms. There is no claim that the process as analysed by them is the same as the one that occurs in humans. It is plain, for instance, that Searle’s
280
280 Bruno G. Bara (1979a) description of how an indirect speech act may be comprehended has little to do with how human beings understand indirect speech acts. The logical steps assumed by Searle to force the conclusion that the meaning of indirect speech acts cannot be derived directly, are complex to a degree that would hamper the most trivial table con versation. As Gibbs (1994) and Bucciarelli, Colle, and Bara (2003) have shown, young children comprehend indirect speech acts, but are unable to understand the logical passages that Searle assumes a person should follow in order to comprehend them. From the perspective of linguistics and philosophy, pragmatics deals with the logic of producing and comprehending communication acts. I use the term ‘communication act’, in assonance with ‘speech act’, in order to consider both linguistic and extralingui stic types of communication. The turning point in cognitive pragmatics is to take into consideration not only communicative competence, but also communicative perfor mance (Bara 1995).
14.2.1 Competence Chomsky (1957) introduced the term ‘competence’ in his authoritative work Syntactic Structures. By this term he intended to mean the set of abstract capacities of a system, independently of how these capacities are put to actual use. In linguistic terms this means dealing with the structures that make the generation and comprehension of sentences possible, and neglecting the problem of establishing whether humans really do possess precisely those hypothesized structures and whether these explain their lin guistic behaviour. On this view, language is seen as being completely independent of the mental functions, such as memory and thought, which enable humans to use it.
14.2.2 Performance Contrasting with competence is performance, which may be defined as the set of abili ties actually exhibited by a system in action, that is, its behaviour in real situations. The concept of performance is a familiar one in psychology, since this is a science con cerned with behaviour occurring in real life rather than with potential capacities. The methodological difficulty presented by performance is that it varies in accordance with individual differences, while competence is invariant across individuals. It is indissolu bly linked not to the abstract average subject, whose existence is of a purely statistical nature, but on the contrary to the irreducible individuality of every performance that really takes place. What is easy for one subject may be difficult for another. Everyone is influenced in a different way by the particular circumstances of the situation. Hence all individual differences must be included and explained. These differences include age, culture, personality, pathologies, and so forth. Philosophers of mind, of language, and of action (such as Wittgenstein, Austin, Grice, and Searle) have never considered performance in their analysis of communication. The
281
Cognitive Pragmatics 281 first scholar to study how language is used in actual context by different people is the ex perimental psychologist Herbert Clark. In many papers and in one influential book (Using Language, 1996), Clark explored language use as a form of joint action, both theoretically and experimentally. Another psychologist, Raymond Gibbs, conducted a series of experi ments to explore specific questions about pragmatic performance (e.g. the comprehension of figurative language; Gibbs 1994). Gibbs offered brilliant answers to specific questions within his fields of interest, but has never bothered to construct a comprehensive theory of pragmatics. Dan Sperber and Deirdre Wilson’s Relevance (1986) was the first major attempt to provide a broad theoretical account of performance. Their work generated a series of ex perimental projects, directed to find empirical evidence in favour of their tenets. The other seminal book on cognitive pragmatics does not present a single theory, but is a collection of specific insights into different basic problems regarding the processes underlying the pro duction and comprehension of communication. Three scholars of Artificial Intelligence, Philip Cohen, Jerry Morgan, and Martha Pollack edited Intentions in Communication (1990), comprising papers by experts of human and man–machine communication.
14.3 Fundamental Concepts The analysis of mental processes involved in human communication must be based on three fundamental concepts: cooperation, sharedness, and communicative intention. All three were originally proposed by Grice (1989b [1975]), although each has since been refined by other scholars.
14.3.1 Cooperation The cooperative nature of human communication was a perspective originally opened by Grice (1975), who threw light on the cooperative reasoning that allows a conversa tion to be successful. Adopting the same perspective, Tomasello (2008) analysed co operation from an evolutionary point of view and proposed a theory based essentially on this notion, which is outlined as follows. If human communication is cooperatively structured in ways that communication among other primates is not, the question arises of how it could have evolved. The issue is that in modern evolutionary theory the emergence of cooperation is regarded as problematic. But if the infrastructure of human cooperative communication is basically the same as that of all other collabo rative activities, the possibility is that it evolved as part of a larger human adaptation for cooperation and cultural life in general. The main difference in comparison with other primates is that great apes have non-cooperative forms of both group activities and intentional communication, underlain by skills for understanding individual in tentionality. While other primates live in an agonistic social context, humans have built a cooperative context. This all began with mentalistic activities in which an individual
282
282 Bruno G. Bara who helped her partner was simultaneously helping herself. But then there was a gen eralization to more altruistic situations in which individuals simply informed or shared things with others freely, possibly as a way to cultivate reciprocity and a reputation for cooperation within the cultural group. Only later still did humans begin to communi cate in this new cooperative way outside of cooperative contexts for non-cooperative purposes, leading to the possibility of deception by lying (Tomasello 2008). Figure 14.1 depicts all the different components of Tomasello’s cooperation model of human communication, and something of their interrelations. Beginning at the top left and following the arrows: I as communicator have many goals and values that I pursue in my life: my individual goals. For whatever reason, I feel that you can help me on this occasion with one or more of them, by helping me or accepting my offer of informa tion (which I want to make for my own reasons) or sharing attitudes with me: my social intention. The best way for me to get your help, or to help you, or to share with you in this situation, is through communication, and so I decide to make mutually manifest to us (in our current joint attentional frame) a communication act; this is my communicative intention (perhaps indicated by ‘for you’ signals such as eye contact or with some expression of motive). I draw your attention to some referential situation in the exter nal world—my referential intention—which is designed (along with some expression of motive) to lead you to infer my social intention via processes of cooperative reasoning, since you are naturally motivated to find out why I want to communicate with you (based on mutual assumptions or norms of cooperation).
Common Ground/Joint Attention: we know together: Individual Goals – many levels
Action comply with what C wants
Social Intention/Motive – want R do X – want R know Y – want R share Z
– express whine – express ø – express smile
Comprehension know what C wants (social intention)
Communicative Intention want R know social intention
“for you” signals [eye contact, etc.]
Cooperative reasoning => Relevance
Referential Intention – want R attend to §
point to §
Reference identify C’s referent
C
R Norms of Cooperation and Cooperative Reasoning
Figure 14.1 Summary of cooperative model of human communication (C = communicator; R = recipient) Source: Tomasello (2008)
283
Cognitive Pragmatics 283 Thus, you attempt to identify my referent first, typically within the space of our common ground, and from there attempt to infer my underlying social intention, also typically by relating it to our common ground. Then, assuming that you have com prehended my social intention, you decide whether or not to cooperate as expected (Tomasello 2008). This fundamentally cooperative process makes human communication utterly dif ferent from the communicative activities of all other species on the planet. The evi dence in favour of Tomasello’s model is based on his extensive comparisons between baby humans and chimpanzees, analysed in a variety of experimental situations.
14.3.2 Sharedness The motivation to cooperate, however, is not sufficient per se to grant communica tion: human beings go a step further, because they are both able and willing to share mental states, from beliefs to emotions, and from intentions to desires. In the pragmatics literature, the concept of belief is standardly formalized as a predicate or modal operator, and is employed as a primitive. To introduce shared beliefs, I must first differentiate between three types of belief: individual, common (also called mutual), and shared. Agents may believe a certain thing, or believe that other agents believe a certain thing, but in a totally autonomous fashion, with no connection existing between the agents themselves. We will call this type of belief individual belief. Often, however, in a given context, all agents have the same indi vidual beliefs: all agents generally share knowledge of their surrounding environ ment, or a certain amount of knowledge that is culturally transmitted. For instance, A may share with B a love of opera, and with all pacifists, the opinion that all atomic weapons should be banned, and with all humans, the evidence that we are born of a mother and a father. Much human interaction is based on this type of belief, which is spread over more or less a wide group of people, and which we will call common belief or mutual belief. Clark (1996) speaks of common ground, meaning the sum of knowledge, beliefs, and suppositions that two or more people share. However, having common beliefs is not a sufficient condition for enabling communication to take place. In order to communicate, each participant, in addition to possessing a common set of beliefs, must also be aware of the fact that all the other participants possess those very same common beliefs. I define a shared belief as a belief which is not only common to all the participants engaged in a speech event, but also which each participant is aware is possessed by all the other participants. From a psychological standpoint, shared belief has a crucial feature: it is subjective but not objective, as is common belief. In actual fact, no one can ever be certain that another person has knowledge of a certain type: she may at most assume that he has that knowledge, and may be convinced that they share it. A relevant combination of the motivation to cooperate and of a shared structure of knowledge is represented by behaviour games (see section 14.4.1).
284
284 Bruno G. Bara
14.3.3 Communicative intention Making information achieve the status of sharedness does not mean that one has communicated it. The necessary condition for real communication to take place is that such information be intentionally and explicitly proposed to the interlocutor. Communicative intention has been defined by Grice (1989b) as the intention to com municate something, plus the intention that that intention to communicate that par ticular something be recognized as such. To be more precise, A has a communicative intention that p, with regard to B—that is to say: A intends to communicate that p to B—when A intends the following two facts to be shared by both A and B: (i) that p, and (ii) that A intends to communicate that p to B. Grice points out that communicating includes not only the speaker’s first-order intention I1, that of achieving a certain effect on the interlocutor, but also the second-order intention I2, namely, that the first-order intention I1 be recognized as such by the interlocutor. Strawson (1964) has, however, drawn attention to the fact that if we wish to speak of open communication, not even the second-order intention I2 is sufficient; a third-order intention I3 is also required. I3 ensures that I2 is recognized for what it is. Though logically sound, Strawson’s examples are rather complicated, so much so that they even irritated Grice, and we need not discuss them in detail here. But at this point, Airenti, Bara, and Colombetti (1993) have demonstrated that if an nth-order intention In is required in the definition of com munication, then the actor might not have the nth+1-order intention In+1. In this case, the interactive situation would not be fully open, because part of the situation would not be intended by the actor as having to be recognized, but would be a part she intends to keep private. From a technical standpoint, this sets up two alternatives: either an infinite hierar chy of intentions is postulated, or a circular definition of communication is furnished by employing the notion of shared belief that was introduced earlier in this chapter. Formally, communicative intention may thus be defined as follows: (1) CINTAB p ≡ INTA SH BA ( p ∧ CINTAB p) What formula (1) means is that A has the communicative intention that p towards B (in symbols, CINTAB p) when A intends (INTA) that the following two facts be shared by B and herself (SHBA): that p, and that she intended to communicate to B that p (CINTAB p). All of this may be translated into more acceptable English as follows. A intends to communicate a certain thing to B. A concurrently wants B to take as shared between the two not only the specific content she wishes to convey, but also the fact that she actually did wish to convey that content to him. Similar to shared belief, communicative intention is also a primitive in pragmatics. This means that it implies, but is not reducible to, an infinite number of finite embeddings of
285
Cognitive Pragmatics 285 intentions and shared beliefs. The following logical implications may be derived from formula (1): (2) CINTAB p ⊃ INTA SH BA p CINTAB p ⊃ INTA SH BA INTA SH BA p CINTAB p ⊃ INTA SH BA INTA SH BA INTA SH BA p CINTAB p ⊃ . . . Formula (2) means that given the fact that A intends to communicate a certain thing to B, we may infer that A also intends her original intention to communicate that par ticular thing to be recognized. If needed, this includes the further inference that A wishes B to recognize her intention of letting B know that she really did intend him to become aware of her intention to communicate that particular message to him. And so on, until the cognitive resources possessed by both manage to make sense of the sequence of embedding.
14.4 Cognitive Pragmatics Theory In my definition, communication is an activity consisting of a combined effort of at least two agents, who consciously and intentionally cooperate to construct together in their shared mental space the meaning of their interaction. In this section I shall pre sent an outline of my Cognitive Pragmatics theory (Bara 2010). The nature of the analytical tools introduced in section 14.3 is such that they may be applied to any form of communication, whether linguistic or extralinguistic. I shall abandon the usual but self-contradictory difference assumed between verbal and non-verbal behaviour, where verbal behaviour is just the spoken language in input (Hinde 1972). I propose an alternative to the distinction based on input that is framed instead on the way data are processed: linguistic communication is based on the communicative use of a system of symbols, whereas extralinguistic communication consists of the use of a set of symbols (Bara and Tirassa 2010). Intuitively, the essential difference lies in the principle of compositionality: language may be subdivided into smaller constituent components with an autonomous meaning, for example, words, whereas extralinguistic communi cation comes about through the use of components that cannot be decomposed into smaller, autonomous units: a smiling face is a smiling face, and not the sum of many small parts each of which is smiling. I shall present brain imaging evidence to support my distinction in section 14.4.2, after introducing the relevant theoretical concepts. Linguistic communication: Compositionality. Linguistic communication may be de fined as the communicative use of a system of symbols. This means that language is an
286
286 Bruno G. Bara entity based on compositionality: language is built up recursively from separate compo nents that may be joined together (Chomsky 1957). Some linguistic expressions have an atomic structure, that is, they cannot be subdivided into smaller, constituent units. Other expressions have a molecular structure, that is, they are composed of smaller constituents which may in their turn be either atomic or molecular. The semantic con tent of an expression—be it atomic or molecular—depends on both its global structure and the semantic content of its constituents. Extralinguistic communication: Associative. Moving now to extralinguistic commu nication, this form of communication may be considered as the communicative use of a set of symbols. This mode is essentially non-compositional: that is, it consists of parts and not of constituents. Extralinguistic signals are molecular blocks that cannot be decom posed any further, inasmuch as they are equipped with intrinsic, global significance. The parts do not have atomic meanings into which they may be further subdivided. The pirouette performed by a ballerina is a pirouette performed by the entire body, and not a pirouette that is performed by the left leg plus a pirouette that is performed by the right leg plus the torsion of the trunk and so forth. To be more precise, gestures may be inserted into a meaningful sequence as well. In fact, associativity (and hence extralin guistic communication) possesses a zero-order syntax, that is, a syntax based solely on the consecutivity of meanings.
14.4.1 Behaviour games In his Philosophical Investigations (1953, part I, remark 1) Wittgenstein advanced the idea that ‘the entire process of using words’ may be seen as a kind of game, which he called language game. His revolutionary proposal was that one should focus on lan guage use instead of language form. In conversation we must make a clear distinction between communicative compe tence and stereotypical interaction schemas. Communicative competence is a general characteristic of the mind, whereas stereotypical interaction schemas are culture- bound. Indeed, the latter may pertain to a small group of individuals or even two people. The idea that will be developed in this chapter is that communicative compe tence may be viewed in formal terms as a metalevel property that controls first-level inferences; such inferences are carried out on shared representations of stereotypical interaction schemas. Consider (3): (3) A: Tomorrow’s Thursday. Will you coordinate the exam invigilation? B: Actually, the Vice Chancellor has fixed a meeting for 9 a.m. In any standard context, B’s reply would be taken as a justification for refusing to carry out A’s request. As stated earlier, B cooperates conversationally but not behaviourally. The intuitive concept of a behaviour game allows us to explain (3), for it enables us to assert that through her request, A is proposing that B and she play the behaviour game in (4).
287
Cognitive Pragmatics 287 (4) Pedagogical duties A is responsible for running departmental activities from Monday to Wednesday; B is responsible for running departmental activities from Thursday to Saturday. With his reply, B rejects A’s proposal, justifying himself by explaining that he has an other duty that takes priority over the exam invigilation. A thus takes B’s reply as con currently constituting a counter-request on B’s part to ‘take over’ his exam invigilation. The point is that in order to cooperate, at least at the level of conversation, both agents must share the behaviour game Pedagogical duties. In real terms, mutual knowledge of game (4) is exploited to achieve conversational cooperation, even if B fails to execute the moves foreseen by A and therefore behavioural cooperation is not achieved. The reason for introducing behaviour games is that the literal meaning of an ut terance is only the departure point for its comprehension. ‘Why is she saying this to me?’ and ‘What does she want from me?’ are the real questions requiring an answer. A behaviour game is that structure which enables actors to coordinate their interper sonal actions, and which actors employ to select the real meaning of an utterance from among the many meanings that utterance might in theory convey. For two actors to cooperate at the level of behaviour, they must operate on the basis of a plan that is shared at least in part. Following Airenti, Bara, and Colombetti (1993), I shall call a behaviour game between A and B an action plan that is shared by A and B. The shared knowledge required for two actors to be able to interact in the same game may be a combination of implicit and explicit. As we shall see later, the two actors may have an explicit representation of the game, or they may have a tacit representation that is sufficient to enable them to direct their actions. Stated differently, for a game to be playable, it must be represented in the actor’s memory. In addition to actions, behaviour games include validity conditions that specify the conditions under which the game may be played. Validity conditions may be viewed as an extension of the felicity conditions that Austin (1962a) invoked to guarantee the success of performatives. The essential features of validity conditions, which not only apply to performatives but to any move of the game, are time and place. However, some games may impose other conditions connected to the mental states of the participants, or constrained by the actions to be executed. Finally, a game is only playable if the relationship between the participants allows it. In some cases, if the game has wide social applicability, as is the case with asking some one the time, then the participants need not even be acquainted. In other cases, par ticipants must be members of the same group. For example, only two freemasons who have both recognized their common status as freemasons may discuss topics reserved to those belonging to that lodge. At the extreme of this type of case we have games that can only be played by two specific players and by no one else. This is the case, for exam ple, with games played by parent and child or by a married couple. The relationship is therefore the set of behaviour games that two people may play together. The structure of a behaviour game is shown in Figure 14.2. The first thing to note is that each agent has her own subjective, rather than objective, vision of the game, since
288
288 Bruno G. Bara [NAME OF THE GAME] Relationship between the players (A, B, C) Validity conditions time ……………………………………………………………………. ……………………………. place …………………………………………………………………… ……………………………. others ………………………………………………………………….. ……………………………. Moves in the game A does something B does something C does something …
Figure 14.2 Structure of the behaviour game
the entire process is based on the notion of shared knowledge, this latter notion also being subjective and not objective. Thus, in our notation, the perspective of each player should appear beside the name of each game: G (A, B) represents game G viewed from A’s standpoint, while G (B, A), instead represents the same game, G, viewed from B’s standpoint.
14.4.2 Conversation games Cognitive Pragmatics views the global structure of dialogues as deriving from the shared knowledge of an action plan. Consequently, the global structure of a dialogue does not derive from linguistic rules, but from behaviour games. In sum, the behaviour game governs the interaction as a whole, whereas the conversation game is responsible for the harmonious local development of the dialogue. A conversation game may be
289
Cognitive Pragmatics 289 defined as a set of tasks that each participant in the conversation has to fulfill in a given sequence. Formally, it consists in a set of metarules that define both the task to be car ried out in each phase and which task is to be activated next. I do not intend to analyse an elementary exchange in a dialogue in formal terms, but to explain the scheme of conversation in its entirety, since the latter is a special and fun damental case of communication among humans. Conversation is not so much a game of table tennis, in which the agents alternatively exchange information, as a communal and simultaneous effort to build something together. The general scheme is as follows: the actor produces an utterance; the partner builds a representation of its meaning. The partner’s mental states pertaining to the topic of the conversation may be modified by the comprehension of the utterance(s). He then plans the next move in the conversation, which he then generates. The rules proposed comprise a dyadic model of communication acts that range from comprehension to reaction, that is, from the reconstruction of the meaning intended by the speaker to the establishment of the high-level intentions required to generate the response. Assuming that actor A produces an utterance addressed to partner B, we can distin guish five logically connected steps in B’s mental processes: Stage 1. Expression act, where A’s mental state is reconstructed by B starting from the locutionary act. Stage 2. Speaker meaning, where B reconstructs A’s communicative intentions, in cluding the case of indirect speech. Stage 3. Communicative effect, which consists of two processes: (a) Attribution, where B attributes to A private mental states such as beliefs and intentions; and (b) Adjustment, where B’s mental states concerning the topic of the conver sation may be altered as a result of A’s utterance. Stage 4. Reaction, where B produces the intentions he will communicate in his response. Stage 5. Response, in which B produces an overt communicative response. The linking together of these five stages is managed by the conversation game: stated more formally, it is the equivalent of a set of metarules. The standard sequence is that described from stage 1 through to stage 5. However, if any one of the first three stages fails to complete its task, the normal chain is interrupted and the process moves directly on to the response stage. This is due to the fact that the conversation game lays down the rule that the partner will react to the actor’s utterance, even when he does not un derstand it. This he may do, for instance, by asking for clarification. The global outline of these five stages is shown in Figure 14.3. The execution of each task is governed by a set of base-level rules that defines which dominion-dependent inferences are to be employed to carry out that task. These rules have different roles in different processes. I must stress that such rules are a convenient
290
290 Bruno G. Bara Stage
Stage task
Partner B SH B,A Shared by partner B and actor A Recognition of expression act
Recognizing actor’s expression act
Comprehension of speaker meaning
Inferring actor’s communicative intentions
Communicative effect
Attributing private mental states to the actor Adjusting his own mental states with respect to the topic
Reaction
Generating communicative intentions
Response generation
Planning the communicative response
SH B,A Shared by partner B and actor A
Figure 14.3 The five stages of comprehension and generation of a communication act Source: Airenti, Bara, and Colombetti (1993)
means for describing communicative interaction. This does not imply that they actually exist in the mind or brain. Comprehending the expression act (stage 1) and speaker’s meaning (stage 2) is managed by a limited number of specialized rules. The reason for this is that compre hension is a process that is shared and achieved by two people, with the result that the actor must, in principle, be able to predict how her partner will reconstruct the mean ing of her utterance. Stated differently, since comprehension rules are constitutive of meaning, meaning construction is shared by all those who are taking part in the inter action. In contrast to the two initial stages, the effect of the utterance on the partner is a question of private mental processing. In this case, individual motivation and general intelligence prevail over shared social norms. This means that it is impossible to for mulate an exhaustive set of rules for stage 3. Reaction (stage 4) is again different. The task here consists in planning a communication act of which the private motivations activated by the flow of the dialogue form the starting point. It should therefore be pos sible to identify a set of norms that describes cooperative interaction. Such norms are neither universal nor logically necessary. They depend on the specific culture to which the agents belong and the specific circumstances in which the dialogue takes place: in our terms, on the behaviour game that is being played. This being the case, I shall not introduce base-level response rules, but simply present a few paradigmatic examples. The task of defining cultural and situational taxonomies of the rules for stage 4 must
291
Cognitive Pragmatics 291 be left to the scholars of ethnicity. Finally, response generation (stage 5) is based on a highly specialized type of planning and a set of shared and constitutive linguistic and extralinguistic rules.
14.4.3 Non-standard communication The theory outlined in the preceding section describes the process of comprehension and construction of the response in situations that may be classified as standard, that is, situations that trigger default rules. There are, however, a large number of interest ing cases that go beyond the bounds of normality, which may be classified under four headings: (i) Non-expressive interaction: the use of an utterance without there being any in tention to express the mental state associated with that utterance. (ii) Exploitation: the special use of a communication rule to obtain a communica tive effect that is different from that normally associated with that rule (e.g. to create irony). (iii) Deception: the attempt to convey a mental state that is not in fact possessed. (iv) Failure: an unsuccessful attempt to achieve the desired communicative effect. The analysis of these cases is important not only in itself, but also because it pro vides independent evidence in favour of Cognitive Pragmatics, in the sense that the structural features typical of possible interactions that are non-expressive, that exploit, that deceive, and that fail, overlap perfectly with the different representations on which the cognitive process of communication is based. An interruption in the standard inferential chain may be ascribed to one of two different reasons: either the actor wanted the interruption to come about, or else it came about in the partner’s mind without the actor’s having wished it. Intentional interruption means that the actor intended to employ a form of exploitation or of non-expressive interaction. Alternatively, failure comes about either because the partner does not follow the inferential chain when he was meant to, or conversely, because he follows the chain when he was not supposed to, since the actor had proposed a non-standard mode. Finally, an attempt at deception takes place when the actor employs false shared knowledge in order to achieve her objectives. As I have already explained, it is the conversation game that governs the succession of the five stages in communication. Stated differently, the conversation game works at a metalevel—employing metarules—that ensures that at the base level all the standard inferences follow on smoothly without any blocks occurring, simply by applying the base rules. The purpose of the conversation game is, on the one hand, to guarantee that each stage accomplishes its task in an adequate fashion, thereby enabling the next stage to receive the information it requires to proceed, and on the other hand, to inter vene if a given stage fails to achieve its objectives. Indeed, if a stage does not realize its
292
292 Bruno G. Bara predetermined goals, then the conversation game intervenes at the metalevel in order to specify what has to be done about the problem that has occurred at the base level. In this case, the metalevel blocks the default rules that are specific to that stage, activating an alternative inferential process that does not employ that particular type of automatic rule. Just as I termed the communicative process utilizing default rules standard, so I shall call non-standard those processes that must have recourse to classic inferencing procedures, since they cannot apply standard default rules because the latter are inap propriate to the context. Each non-standard situation has a logical place in the framework of the comprehen sion process I have outlined. Non-expressive interaction, exploitation, and deception fall naturally into the first three stages of my model. In the first stage—understanding the expression act—the only non-standard path the actor may have followed, and which the partner must therefore identify, is that of non-expressive interaction. In the second stage—understanding the speaker’s meaning—all non-standard inferences are cases of exploitation. In both cases, the actor tries to ensure that the partner will identify the non-standard path and follow it correctly. Should this not happen, the outcome will be failure. The third stage—communicative effect—is where deception occurs. Deception cannot be found in the comprehension stages because it is not achieved by any special form of communication. It concerns the relationship between what the actor commu nicates and her private mental states. For Cognitive Pragmatics, deception is a conscious violation of a shared behaviour game. Though A knows that she should act in a certain way in order to respect the behaviour game being played by B and herself, she carries out a communicative behavioural act that is premeditated to make B believe that it is a game move even if she knows perfectly that it is a violation of that game. Finally, one has to analyse failures, which may occur at any stage. In order to recog nize and possibly to repair a failure, one cannot use the default rules of communication; one has to fall back on classic inferential procedures. Even though it might at first seem paradoxical, from the standpoint of cooperation, even a failure must be agreed on by the actors. In order to remain consistent with all the literature on the subject, I shall continue to use the term failure in a generic sense to indicate all those cases that exhibit communicative failure at whatever level this might have taken place. I shall introduce the term shared failure to indicate particular cases in which both the agents are aware of what has happened and accept the definition of the occurrence as constituting a failure in the proper sense of the term. What I wish to underline is the fact that to define an occurrence as a shared failure both the agents must consciously recognize that there has been a non-negotiable refusal, and that this refusal is to be considered as shared knowledge. If one of the two agents, either A or B, does not believe that this is what has happened, then the case cannot be classified as shared failure, despite the difficulties that have occurred. We are now in a position to define a communicative failure as an abortive attempt to produce a given communicative effect on one’s partner. From A’s standpoint, communi cative failure may come about in any of the first three stages I have hypothesized: at the level of the expression act (non-comprehension), at the level of the speaker’s meaning
293
Cognitive Pragmatics 293 (misunderstanding), or at the level of the communicative effect (refusal). An interrup tion at any point along the communicative chain will produce a failure to achieve the speaker’s goal, that is, the speaker will fail to generate in her partner’s mind the mental state she intended to generate. None of the existing pragmatic theories offers a global account of successful and failed communication. A noteworthy exception is relevance theory, which establishes a continuum between the idealizations of success and failure. Sperber and Wilson (1995 [1986]) measured the efficacy of communication in terms of attempted relevance, as compared with achieved relevance. The introduction of strong and weak implicatures emphasizes the idea that each communicative instance conveys core meaning and per petual implications. The notion of failure is spread over a wide set of implicatures, both those that are attempted and those that can be possibly achieved. Still, relevance theory has never generated systematic hypotheses for explaining communicative failures. Following the assumptions of Cognitive Pragmatics, Bosco, Bucciarelli, and Bara (2006) proposed a taxonomy of different sorts of failure, grounded on the mental representations and cognitive processes involved: failure of the expression act, failure of the speaker’s meaning, and failure of the communicative effect. When failure of the expression act occurs, the partner fails to comprehend the literal value of the utter ance; when failure of the speaker’s meaning occurs, the partner fails to comprehend the speaker’s communicative intention; and finally, when failure of the communicative effect occurs, the partner does not modify his mental states in the way the speaker de sires, that is, he refuses to adhere to the speaker’s goal. Depending on the sort of failure that occurs, the speaker might enact a different kind of repair. The taxonomy allowed us to generate hypotheses about the relative difficulty in recognizing and repairing dif ferent kinds of failure. An experiment on eighty children aged 3 to 8 years confirmed the predictions based on the taxonomy, both in the recognition and repair tasks.
14.5 Experimental Evidence We can now take a look at the empirical evidence for or against the Cognitive Pragmatics theory along four different dimensions: comparative studies, especially in primates (Tomasello 2008, 2014); developmental pragmatics, investigating both normally and atypically developing children; clinical pragmatics; and neuropragmatics, encompassing both lesion studies and brain imaging studies.
14.5.1 Developmental pragmatics Within the framework of the Cognitive Pragmatics theory, we investigated the develop ment of children’s linguistic and extralinguistic abilities with the aim of (i) systematically describing the emergence of the two different modalities reunified under a single theoretical
294
294 Bruno G. Bara paradigm, and (ii) offering a baseline against which to compare the outcomes of atypical development in relation to linguistic and extralinguistic components. Atypically developing children frequently have difficulties with the comprehension and production of communication acts, and the extent of the deficit depends on the type of pathology, namely head injury, hydrocephalus, focal brain damage, and autism (Bara, Bosco, and Bucciarelli 1999). But the key point is that it is hard to understand the nature of the deficits when one does not know how normal development takes place. Our studies aimed to clarify the factors involved in the comprehension and production of different kinds of communication acts, following their emergence during typical de velopment in order to gain a deeper understanding of deficits in case of pathological conditions. Bucciarelli, Colle, and Bara (2003) claimed that three factors determine the complexity of the mental representation involved in the comprehension of a pragmatic phenomenon. The factors involved may be referred to as (i) conflicting representations, (ii) exploitation of sharedness, and (iii) inferential load.
14.5.1.1 Conflicting representations Conflicting representations involve a discrepancy between what is communicated and what is privately entertained by the actor. In the case of no conflict, we are dealing with standard communication, where the actor merely produces an utterance that is in line with her private belief and with the behavioural game she shares with the part ner. In the case of conflict, we are dealing with non-standard communication, which involves the violation of default rules and the occurrence of more sophisticated mental representations. Direct communication acts, conventional indirect communication acts, and non-conventional indirect communication acts are all examples of standard communication: in terms of mental representations, the partner has merely to refer the actor’s utterance to a valid behaviour game. Deceit and irony are instead examples of non-standard communication, where the partner has to deal with the difference between the mental states expressed by the actor and those she privately entertains. It may consequently be predicted that standard communicative phenomena are easier to deal with than non-standard ones.
14.5.1.2 Representations where shared beliefs are exploited Representations involving a belief expressed by an actor that is in contrast with a belief shared with the partner are more difficult to handle than representations that do not involve such a contrast. When comprehending deceit, the partner recognizes the discrepancy between the mental states that are expressed and those that are privately entertained by the actor. An utterance becomes ironic when, along with this discrepancy, the partner also has to recognize the contrast between the expressed mental states and the scenario provided by the knowledge shared with the actor. The concurrent activation of the representa tion of the actor’s utterance (p) and of the contrasting shared belief (non-p) makes ironic statements difficult for a child to manage. It follows that deceits should be easier to deal with than ironies.
295
Cognitive Pragmatics 295
14.5.1.3 Inferential load The need to build a long chain of inferences is what discriminates between simple and complex communication acts. Direct and conventional indirect speech acts may be re ferred to as simple communication acts because they immediately make reference to a behavioural game. Non-conventional indirect speech acts may be referred to as com plex communication acts because they do not make direct reference to a behavioural game and require a more complex inferential process. The length of the inferential chain necessary to connect the communication act to the behavioural game shared between the interlocutors is the factor determining the different levels of difficulty chil dren find in comprehending simple and complex communication acts. In Bucciarelli, Colle, and Bara (2003) we investigated these theoretical points. We administered an experimental protocol structured in both linguistic and extra linguistic modalities and comprising simple and complex pragmatic phenomena to 160 children—ranging in age from 2.5 to 6 years. Half of the children were ran domly assigned to the linguistic protocol, and half to the extralinguistic one. The results of the study confirmed all of our predictions. Standard communication was easier to comprehend than non-standard communication in both the linguistic and extralinguistic modalities. Moreover, direct and simple indirect communication acts were equally easy to comprehend, and they were easier than simple deceits, which were in turn easier than simple ironies. The results also revealed that a prag matic phenomenon is equally difficult to understand, whether in the linguistic or extralinguistic modality. In particular, we found no difference between children’s performance in the two protocols, neither for the phenomena across the board, nor for the standard and non-standard phenomena, or for the single phenomena considered separately. We compared performance by typically developing children in both the production and comprehension of different pragmatic phenomena expressed in different com municative modalities, using the linguistic, extralinguistic, and paralinguistic scales of the ABaCo battery, which will be described in section 14.5.2. The results concern ing the comprehension and production of standard and non-standard communication acts followed the same pattern on both the linguistic and extralinguistic scales. Such data support a unified theoretical framework in which linguistic and extralinguistic modalities are different aspects of a unique communicative competence. Our data also showed an increase in children’s paralinguistic ability, in particular in their ability to focus on paralinguistic cues in the case of contradiction with the expressed linguistic content. Within the same framework, we investigated the ability to comprehend direct, indirect, deceitful, and ironic communication acts performed through linguistic and extralinguistic means in deaf children (De Marco, Colle, and Bucciarelli 2007). The assumption underlying the experiment was that the construction of the mean ing of a communication act is independent of the input modalities, such as sign language, spoken language, or gestures. Therefore, we expected the same gradation of difficulty in the comprehension of linguistic and extralinguistic communication
296
296 Bruno G. Bara acts in deaf children, as in previous studies of normally developing children. The results with 46 deaf children (aged 2–4 years and 4,6–7 years) confirmed our expectation.
14.5.2 Clinical pragmatics We have also developed a new clinical tool designed to evaluate a patient’s pragmatic ability following brain damage or other acquired/ congenital neuropsychological disorder (from head injury to autism and schizophrenia). We had previously exam ined individuals suffering from traumatic brain injuries (Bara, Tirassa, and Zettin 1997; Bara, Cutica, and Tirassa 2001; Cutica, Bucciarelli, and Bara 2006). Essentially, in these studies we detected a trend of comprehension and production of commu nication acts similar to that exhibited by children. The ABaCo (Assessment Battery for Communication; Sacco et al. 2008) measures a wide range of communicative skills on five evaluation scales: linguistic, extralinguistic, paralinguistic, context, and conversational. It can be used to identify specific levels of impairment and provides a comprehensive assessment of the patient’s pragmatic abilities and deficits in order to guide individualized rehabilitation programs. In the ABaCo normative data study (Angeleri et al. 2012), we provided norms for a sample of 300 healthy and cognitively intact individuals of different ages and educa tional levels. Age and education both affected participants’ scores; consequently, the set of norms were stratified across age and educational level. Such norms will be useful for obtaining clinically indicative and reliable information, by making it possible to differentiate between real deficits and those attributable to the normal ageing process or educational background, and will thus be helpful for designing specific treatment plans. We constructed two parallel forms of the ABaCo, so that the patients can be assessed and reassessed before and after a rehabilitation program, thus obtaining a reli able measure of treatment efficacy. The ABaCo battery was utilized in a study aimed at assessing the pragmatic per formance of traumatic brain injury (TBI) patients (Angeleri et al. 2008). The results showed that although the patients’ comprehension was impaired, it was neverthe less linguistically valid; on the other hand, they performed worse than controls in the extralinguistic comprehension of deceit and irony. The TBI patients showed im paired production in both the linguistic and extralinguistic modality. Moreover, the difficulty in manipulating mental representations had a great impact on the pa tients’ performance: we found an increasing trend of difficulty for different kinds of pragmatic phenomena that involve dealing with embedded mental representations. Specifically, patients found both comprehension and production of standard com munication acts easier than deceits, which in turn were easier than ironies. They also showed a pronounced impairment in managing paralinguistic aspects of com munication, neglecting the emotional meaning expressed through facial expression or prosody. Finally, TBI patients were impaired in grasping subtler conversational
297
Cognitive Pragmatics 297 violations. This study represents the first detailed investigation of pragmatic abili ties in TBI patients, and shows how Cognitive Pragmatics theory can be used to formulate specific predictions about the communicative impairment resulting from pathology.
14.5.3 Neuropragmatics A less traditional way of seeking experimental evidence consists of investigating how theoretical concepts are embodied through neuroimaging techniques. The first issue to put under test was my assumption that a common communica tive competence — independent of the linguistic or extralinguistic gestural means — is instantiated at the level in which a communicative intention is inferred and compre hended within a specific social context, that is, at the pragmatic level. Consequently, I hypothesized that there is no difference in brain activity between the recognition of a communicative intention involving observed linguistic behaviour and that of the same communicative intention involving observed gestural behaviour: the dynamic inten tional network should be recruited independently of the modality used to convey the different intentions. In an fMRI study of twenty-four healthy participants, Enrici et al. (2011) showed that linguistic and gestural modalities share a common communicative competence at the pragmatic level, viz. at the level at which an actor’s communicative intention has to be reconstructed by the partner. The second issue we wished to address was whether any additional brain areas other than those involved at the pragmatic level, that is, the intentionality network areas, are specifically recruited depending on the modality used. Perisylvian language areas are recruited by linguistic modality, whereas sensorimotor and premotor areas are recruited by extralinguistic modality. In conclu sion, we proved that the intentionality network is modality-independent, while the expressive means involved activate the brain regions corresponding to language or to gestures. Our conclusion supported my hypothesis that communicative competence is modality-independent. I performed a second series of brain imaging experiments to start to investigate the recursive nature of communicative intention. We used fMRI to test three basic types of prior intention (Searle 1983): private, social, and communicative. Private intentions (PInt) require only an actor to perform actions adequate to reach her goal (e.g. A in tends to drink a glass of water). On the contrary, a social intention requires at least one partner, who is necessary for the actor to reach her goal (e.g. A intends to play tennis with B). In my approach, the prototypical case of a social intention shared in the pre sent is Communicative intention (CInt). However, the social goal of certain intentions lies in the future. For instance, Alice may intend to visit her father next Sunday. This kind of social intention involves the representation of a social goal when A and B are not actually interacting but B is part of A’s goal. We define this kind of intention as Prospective social intention (PSInt). In the case of PSInt, the desired social interaction
298
298 Bruno G. Bara is not present at the moment, but if the PSInt will become shared in the future, then it will cause the social interaction to occur. A first series of experiments (Walter et al. 2004) allowed us to propose a dynamic intentionality network; a deeper level of analysis based on signal time courses for the four regions of interest extended and confirmed our earlier results (Ciaramidaro et al. 2007). In sum, our results showed that whereas the right TPJ and the precuneus are necessary for processing all types of prior intentions, the left TPJ and the MPFC are specifically involved in the understanding of social prior intention. More specifically, the left TPJ is activated only when social prior intentions occur in the present: this is the case of communicative intention (Bara et al. 2011). A subsequent step was to test the intentional network not only in healthy indi viduals, but also in people with schizophrenia and autism. We hypothesized that paranoid schizophrenics could show hyperintentionality, i.e. the tendency to over attribute intention not only to persons but also to objects. The thesis was confirmed in a group of twelve patients with paranoid schizophrenia: their intention detector became hyperactive in the paranoid interpretation of the physical world (Walter et al. 2009). Adopting a line of reasoning similar to that of Crespi and Badcock (2008), we claimed that the impairment in understanding others’ intentions ex hibited by paranoid patients and autistic patients can be considered as the two ex tremes of a continuum. People with autism are hypo-intentional, as they tend not to attribute the features of sociality to actors, neither interacting with them (CInt), nor preparing a social interaction (PSInt). Autistic comprehension of social inter action is quite similar to autistic interpretation of the physical world (Ciaramidaro et al. 2015).
14.6 Final Remarks The possibility of studying human performance in communication has opened new perspectives in cognitive pragmatics. The comparison between evolutionary and de velopmental investigation, à la Tomasello, has rooted the concept of cooperation in both the phylogenetic and ontogenetic dimensions. The integration between neuropsy chological and neuroimaging studies as achieved by my research group made it pos sible to embody philosophical concepts like sharedness and communicative intentions. Different research methodologies find their source of inspiration in the seminal works of the philosophers of mind, introducing renewed vitality with respect to papers based only on logics and intuitions. The empirical study of performance is increasing the theoretical interest in competence, in a creative atmosphere unknown to single disci plines. The dance between philosophy, cognitive science, and neuroscience may prove extremely powerful for both partners, and full of promise for the growth of pragmatics (not only cognitive).
299
Cognitive Pragmatics 299
Acknowledgements This research was supported by the Regione Piemonte Project: Institutions, Behaviour and Markets in Local and Global Settings (Project IIINBEMA). I am grateful to the col leagues who generously commented on my work: Romina Angeleri, Monica Bucciarelli, Livia Colle, and Ivan Enrici.
300
Chapter 15
Devel op me nta l Pragm at i c s Pamela R. Rollins
15.1 Introduction Developmental pragmatics is a heterogeneous field on a range of topics asso ciated with the study of how young children develop the skills to use language effectively and appropriately in social interaction. Researchers have studied the development of communicative intentions, conversation skills, discourse rules, politeness rules, as well as the role of culture and caregiver behaviours needed for children to gain communicative competencies for successful interpersonal rela tions (see Ninio and Snow 1996 for a review). Others have focused on the relation ship among pragmatic skills and more formal areas of language such as children’s early lexicon (Carpenter, Nagell, and Tomasello 1998) and syntax (Rollins and Snow 1998). Developmental pragmatics has gained importance over the last sev eral years in studies of disordered populations. This is especially true for children with Autism Spectrum Disorders (ASD) for whom pragmatic impairment is the core feature of the disorder (Rollins 1999, 2014; Landa 2000; Young et al. 2005; see also Cummings this volume). Understanding the developmental pathway of social pragmatic skills in typical children aids our ability to assess pragmatic skills in disordered populations and to forge developmentally appropriate intervention (Prizant, Wetherby, and Rydell 2000; Rollins 2014). In this chapter, I focus on the development of pragmatic skills in early interaction with predominately Western caregivers. Developmental pragmatics can be seen as the intersection of social cognitive and communicative development. It is well accepted that infants are socially precocious, orienting towards socially salient information from very early in life. Newborns prefer to listen to mothers’ voices (DeCasper and Fifer 1980; DeCasper and Spence 1986) and
301
Developmental Pragmatics 301 are able to imitate adults of their own volition (Meltzoff and Moore 1983, 1989; Kugiumutzakis 1999). They begin to engage in vocal turn- taking interactions (Trevarthen 1979), respond differentially to persons and objects (Trevarthen 1979; Legerstee 1991) and selectively attend to human faces (Maurer 1985; Johnson and Morton 1991). Over the first year of life, typical infants undergo several qualitative changes in how they monitor, control, and predict the behaviour of others, culmi nating in the ability for mutual understanding and cooperation with people around them. These gradual qualitative changes have been quantified as movement from ‘sharing emotions’, to ‘sharing perceptions and pursuing goals’, to ‘sharing attention’ and intention’ (Tomasello et al. 2005; Rollins 2014). With each new way of thinking comes new pragmatic abilities.
15.2 Sharing Emotions Around two months of age, infants across cultures become increasingly alert and begin to smile in response to social stimuli (Spitz 1965; Wolff 1987). The onset of social smiling coupled with an increase in gazing at the partner’s face is highly significant to Western caregivers, launching dyads into a new quality of shared experiences (Stern 1977, 1985; Rochat and Striano 1999). These dyadic, face-to-face interactions reflect well-balanced, reciprocal, and rhythmic exchanges of affect and emotions (Brazelton, Koslowski, and Main 1974; Trevarthen 1977, 1979; Stern 1985; Rollins and Greenwald 2013). The sensitive caregiver responds to the infant as a communicative partner, and these interactions take on a conversational-like qual ity (Snow 1977), so much so that they have been referred to as ‘protoconversations’ (Bateson 1975; Trevarthen 1979). Protoconversations are thought to be a universal feature of caregiver–infant interaction (Keller, Schölmerich, and Eibl-Eibesfeldt 1988; Trevarthen 1993; Tomasello et al. 2005) and involve a range of affect, emo tions, social expectations, and rounds of vocal turn-taking (Trevarthen 1980; Murray and Trevarthen 1985; Rollins and Greenwald 2013). The infant’s initial pat tern of social responsiveness propels the caregiver and child into a ‘complex joint anticipatory system’ (Bruner 1983). The young infant responds differentially to fa miliar and unfamiliar persons (Dunn 1982), and by four months the infant becomes more sensitive to the timing and organization of the protoconversational envelope (Rochat, Querido, and Striano 1999). Caregivers learn to adjust to the fledgling responsiveness of the child, and over the next few months, protoconversations become livelier with predictable routines, rhymes, tickling, body movements, and opportunities for the partners to exchange roles (Bruner 1978, 1983). It is within the protoconversational envelope that the young infant learns to initiate and terminate protoconversational exchanges and learns the rules for conversational turn taking (Bateson 1975).
302
302 Pamela R. Rollins
15.3 Sharing Perceptions and Pursuing Goals Intuitively, some caregivers begin to include objects in their play (Adamson and Russell 1999) and may exploit the infant’s fascination with the appearance and disappearance of interesting toys (Bruner 1983). By 6 months of age, infants have mastered complex upper-body motor skills and are able to focus on distal objects within the immediate environment. A broad array of attentional options is now available and infants spend increasing amounts of time focused on objects with no indication that they want to share the objects with the caregiver. Nonetheless, the infant–caregiver collaboration can continue, expanding towards the end of the first year to include triadic interactions that incorporate the object (Trevarthen and Hubley 1978; Bakeman and Adamson 1984; Adamson, Bakeman, and Deckner 2004). Bakeman and Adamson (1984) described these early triadic interactions as passive joint engagement. Infants are thought to be passive because they do not explicitly acknowledge their caregiver’s contribution to the interaction by looking back at the caregiver and smiling. Caregivers may actively follow the child’s focus of attention, thereby supporting social communication by ex panding the child’s solitary focus to include caregiver verbal and non-verbal informa tion about the attentional target. Thus, from 6 to 10 months, triadic interactions involve the infant and caregiver jointly perceiving an object or an event towards which they both direct their actions and in so doing share a goal (Bakeman and Adamson 1984; Adamson, Bakeman, and Deckner 2004; Tomasello et al. 2005; Moll and Tomasello 2007). The young child begins to understand that their caregivers have intentional actions. He or she can now monitor his or her caregiver’s actions and can make predic tions about what comes next in the interaction exchange (Tomasello et al. 2005; Moll and Tomasello 2007). For example, the caregiver and infant may both be looking at a block. The caregiver may start to build a block tower. The 6–10-month-old infant begins to understand that the caregiver has a goal in mind of building a tower. She can now share in that goal by monitoring the caregiver’s behaviour and join in an alternating sequence of placing blocks on the tower. After the caregiver places a block on the tower, the child predicts it is his or her turn and does the same. While they are both perceiving the blocks and experiencing the same activity of building the tower, the child does not yet look back at the caregiver and coordinate his or her attention between the object and adult. The child does not yet have the understanding that they are sharing their at tention and intention to build the tower, the intersubjective awareness that it is a shared and cooperative experience (Moll and Tomasello 2007). Bruner (1978, 1983) described these early triadic interaction exchanges as evolving in concert with the changing interests and abilities of the child. Responsive caregivers tailor their behaviour to mesh with the social and cognitive advancement of the child (Bruner 1978, 1983). For example, variations of peek-a-boo games exploit the child’s growing awareness of the permanence of unseen objects and people and have been
303
Developmental Pragmatics 303 shown to change over time (Bruner 1983), allowing the infant to take increasing control of the management of the interaction exchanges. As predictable sequences occur, rou tines are established, and each participant anticipates the actions of the other (Bruner 1983; Moll and Tomasello 2007; Rollins 2014). Within learned routines, roles are ex changed, and the infant learns to initiate and regulate social conventions that will later serve as the platform for conventional language (Bruner 1978, 1983; Rollins 2014). Somewhere around 9 months of age, the infant’s gestures and vocalizations to initiate and respond in these interactions become intentional (Bates, Camaioni, and Volterra 1975; Bates et al. 1979; Tomasello et al. 2005). Elizabeth Bates’ seminal work on the early pragmatic development (Bates, Camaioni, and Volterra 1975; Bates 1976) suggested that the earliest preverbal intentions served both proto-imperative and proto-declarative communicative acts. There is now ample evidence to suggest proto-imperatives appear earlier in development then proto-declaratives and are related to the infant’s nascent social-cognitive abilities (Camaioni 1993; Rollins and Snow 1998). Nine-month-old in fants who are endowed with the social-cognitive ability to share perceptions and pursue goals can regulate or influence the behaviour of others but cannot yet influence their mental states (Camaioni 1993; Rollins 2014). Influencing another’s behaviour requires little more than the attribution of agency to the interlocutor and the ability to share perceptions of intended goal (Camaioni 1993; Tomasello et al. 2005). While some au thors would suggest that we should ‘not conflate intentionality in communication with instrumentality or agency’ (Camaioni 1993: 94), early emerging communicative acts that regulate behaviour such as requesting and protesting can be considered intentional because the child has the will to affect the caregiver by some purposeful behaviour (Ninio and Snow 1996; Moll and Tomasello 2007; Rollins 2014).
15.4 Sharing Attention and Intention As the infant transitions to sharing attention and intentions (around 9 to 10 months), he or she is capable of not only monitoring the caregiver’s behaviour but is able to ac tively monitor the caregiver’s attentional focus (Hubley and Trevarthen 1979; Bakeman and Adamson 1984; Tomasello 1995; Carpenter, Nagell, and Tomasello 1998). This milestone is sometimes referred to as responding to joint attention (RJA) (Mundy and Thorp 2008) and marks the child’s recognition that the caregiver’s attention is different from his or her own (Trevarthen and Hubley 1978; Tomasello 1995). Soon (10–12 months), the infant becomes socially motivated and checks for the caregiver’s focus of attention by actively looking back and forth between the caregiver and the object of attention (Bakeman and Adamson 1984; Carpenter, Nagell, and Tomasello 1998). During these episodes, known as coordinated joint engagement (Bakeman and Adamson 1984; Carpenter, Nagell, and Tomasello 1998), the infant coordinates their attention between the caregiver and an object of mutual interest. This newly acquired social competency is a form of cooperative intersubjectivity as it includes the active
304
304 Pamela R. Rollins sharing of thoughts and emotions about an outside entity (Trevarthen and Aitken 2001; Tomasello et al. 2005; Moll and Tomasello 2007). Tomasello et al. (2005) refer to this development as shared intention, reflecting the understanding that other persons have unique attentions, intentions, and goals. This new level of social-cognitive skill, which emerges around the first birthday, is a monumental achievement not observed in non- human primates (Gomez 1990; Tomasello and Call 1997; Tomasello et al. 2005) and found to be extraordinarily difficult for children with ASD (Camaioni 1993; Rollins and Snow 1998; Rollins et al. 1998; Charman 2003; Tomasello et al. 2005). Many young children with ASD develop the communication skills for instrumental purposes, but their development of joint intention and shared cooperative activities is deficient and severely protracted (Mundy, Kasari, and Sigman 1992; Rollins 1994, 1999, 2014; Rollins and Snow 1998; Rollins et al. 1998; Charman 2003). The development of joint intention requires that both the adult and the child have mutual knowledge that they are doing something together in relationship, marking the emergence of a mutual cooperation (Tomasello et al. 2005; Moll and Tomasello 2007). When the preverbal child is capa ble of understanding that others have attentions and intentions different from his or her own, true intentionality in communication emerges (Camaioni 1993; Rollins 2014). Children endowed with shared intention are capable of directing the caregiver’s atten tion with gestures by showing an object or pointing to an object for the purpose of sharing interest (Bates, Camaioni, and Volterra 1975; Ninio and Snow 1996; Tomasello, Carpenter, and Liszkowski 2007). As the young child learns words, their early com municative repertoire continues to reflect the unfolding of shared intentionality and the mutual understanding that they are communicating together (Adamson and Bakeman 2006; Adamson, Bakeman, Deckner, and Nelson 2013). Directing the other’s attention (which has also been referred to as initiating joint attention; see Mundy and Thorp 2008) and a new skill of discussing a joint focus of attention continue into the second year of life. These early discussions of the here-and-now often take the form of naming object or events in the immediate environment while interacting around toys or look ing at picture books together (Adamson et al. 2013). As children’s early vocabularies grow and they can use more words to convey intentions, these discussions increase in sophistication. Soon, naming games take on a different quality as caregivers begin to question the young child about the knowledge he or she has about events, objects, and pictures with which the dyad is engaged (Adamson and Bakeman 2006; Adamson et al. 2013). Caregivers may request information about animal sounds or some shared past event. With the adult assistance, the child is now engaging in communicative in tentions that represent a decrease in context embeddedness (Ninio and Snow 1996; Adamson and Bakeman 2006; Adamson et al. 2013). Thus, at 20 months of age, most children discuss objects and events that are not in the environment but are somehow related to objects, events, and pictures that are present (Snow et al. 1996; Adamson et al. 2013). The trend of incorporating discussions less embedded in the immediate context continues to 32 months when children begin to discuss objects and events that have no perceivable reference in the environment (i.e. discuss non-present) as well as talk about thoughts and feelings (Snow et al. 1996; Adamson et al. 2013). The developing child’s
305
Developmental Pragmatics 305 pragmatic understanding is now reflective of underlying motivations for cooperation and shared intentionality (Ninio and Snow 1996; Snow 1999; Tomasello et al. 2007).
15.5 Conversational Skills As described in section 15.2 on sharing emotions, young infants are remarkably preco cious in the development of turn taking when they are just a few months old. These early ‘protoconversations’ are filled with nascent vocal turn taking and non-verbal signals (Bateson 1975; Stern 1977; Trevarthen 1977; Trevarthen and Hubley 1978) and are considered the underpinning of later reciprocal conversations (Bloom, Russell, and Wassenberg 1987). Late in the first year of life, the infant and his or her caregiv ers engage in conversational-like exchanges within the context of routines and script- like interactions. These exchanges may be characterized as being mutually supportive where both partners are committed to the joint activity (Adamson and Bakeman 2006; Adamson et al. 2013), but they lack the Gricean conversational maxim of being inform ative (Grice 1975; see also Huang, this volume). Instead, the sensitive caregiver provides opportunities for the infant to contribute a turn within routine activities such as peek- a-boo, reading books, or naming objects (Ninio and Bruner 1978; Snow, Perlman, and Nathan 1987; Ninio and Snow 1996; Adamson and Bakeman 2006). The predictable patterns of these socially constructed routines make it possible for the young child to participate in conversational turn-taking exchanges well before they can add new information to the interaction (Ninio and Snow 1996). These early social routines alert the child to the information that should be attended to and define what can be presup posed (Bruner 1995). Children learn to anticipate routines, and to spontaneously take their turn by supplying the frame-appropriate speech. The attributes of true adult conversation unfold in the developing child as she learns to maintain topic and add new information to the conversation. As the child begins to share attention and interest with the caregiver, she begins to initiate conversational exchanges by pointing, showing, or commenting on something in the immediate en vironment. These early object-mediated displays of sharing interest do not require the child to maintain the conversation, a skill that begins to develop over the next few months. Snow et al. (1996) found that between 12–20 months, the proportion of speech acts children used to direct the caregiver’s attention decreased, while those used in discussions about a joint focus of attention increased. They interpreted this to mean that young children and their caregivers begin to join in each other’s topics of conversa tion, and the children begin to maintain their own topics of conversation once initiated (Ninio and Snow 1996). Many of these early conversations require minimal response on the part of the child, who may simply provide a point, shrug, yes/no, or one-word answer to an adult question (Eisenberg 1981; Uccelli et al. 2005). This is commensurate with Bloom, Rocissano, and Hood’s (1976) findings that children maintain the topic of conversation by adding new information between 19–32 months as their language
306
306 Pamela R. Rollins becomes increasingly more proficient. They found that the children maintained topic about half of the time at 21 months, and this early cohesion expanded to three quarters of the time at 36 months. Two primary strategies for topic maintenance were observed at all ages. Children either imitated part of the adult’s utterance, or they added new and relevant information. The later strategy, which satisfies two of the conversational maxims under Grice’s cooperative principle (see Huang, this volume), increased from 21 per cent to 46 per cent during this time span. The young child’s fledgling conversational skills are enhanced by a responsive and cooperative adult partner, resulting in the appearance of a more advanced level of conversational skills (Ninio and Snow 1996). Adults are forgiving of the child’s viola tions of conversational principles, and they use several devices like asking questions or giving choices to facilitate conversational exchanges from their young partners (Bloom, Rocissano, and Hood 1976; Ervin-Trip 1979). In contrast, peers are less accommodat ing. Preschool children either fail to maintain topic or use the developmentally earlier strategy of imitation in interaction with their peers (Blank and Franklin 1980; Benoit 1982). Brinton and Fujiki (1984) found considerable individual variation in the conver sational strategies of 5-and 9-year-old children when talking to peers. Nonetheless, they found 5 year olds continue to use the imitative strategy of repeating the previous utterances of other young children while 9 year olds created novel utterances with new information. Both 5-and 9-year-old children initiate many more topics of conversation with their peers than adults do in adult–adult interaction (Brinton and Fujiki 1984). Furthermore, from 5 to 9 years, school-aged children’s ability to maintain topic im proves while the number of different topics discussed decreases. Understanding the developmental trajectory of cohesive conversation is important as individual differ ences play a role in children being accepted by peers, making friends, and how adults view them (Schley and Snow 1992; Ninio and Snow 1996). In addition, it sheds light on developmental delays and differences in children who have pragmatic impairments, such as children with ASD. Ultimately, this helps us forge developmentally appropriate interventions.
15.6 Narratives One of the most important aspects of language after 5 is organizing language across utterances such as in the form of narratives (Karmiloff-Smith 1986). The ability to tell a narrative requires the ability to (a) organize one’s thoughts and experiences using causal-explanatory relationships (Peterson and McCabe 1983); (b) understand one’s own social, cultural, and emotional perspectives and those of others in a variety of contexts (Bruner 1986; Nelson and Fivush 2004); (c) evaluate what information is needed by the listener to form a shared frame of reference; (d) monitor and adjust the style and content of the interaction based on the listener’s reactions; and (e) in tegrate all of these tasks to produce a cohesive and compelling account or exchange
307
Developmental Pragmatics 307 (Peterson and McCabe 1983; Karmiloff-Smith 1985). While there are many different types of narratives (e.g. picture book narratives, fantasy narratives) the focus here is on the social support for personal past-tense event narratives. Personal event narratives are the genre frequently composed in the early conversational exchanges between young children and their partners (McCabe and Rollins 1994). They are culturally and socially mediated but are a universal means to bring the past into the present, helping make sense of experiences and form relationships (McCabe and Peterson 1991; McCabe 1996; Nelson and Fivush 2004). In this way, narratives serve both a referential and evalua tive function (Labov and Walentzky 1967). European American1 children begin to tell personal narratives at 2 years of age (Sachs 1982; Nelson 1989). These early narratives begin as reports of just one event at a time (Peterson and McCabe 1983; McCabe and Peterson 1991), lacking the temporal progression identified by Labov (1972) as an impor tant component of a ‘true’ narrative. Peterson and McCabe (1983; McCabe and Peterson 1991) extended Labov’s analytic approach to the analysis of personal event narratives of typically developing children aged 3 1/2 to 9 years. By 31 months, children begin to chain two past-tense events together. Around ages 3–4, they begin to chain more than two events together, although the sequence of events may not be coherent and informa tion may be omitted. By age 5, children’s narratives are well-ordered but often end at the emotional climax of the event, neglecting to discuss the resolution of the story. ‘By 6 years of age, children are able to tell a well-formed story, one that orients the listener to who, what, and where something happened; that retells the sequence of events that builds to some sort of climax or high point; and that then goes on to resolve itself by telling how things turned out’ (Rollins, McCabe, and Bliss 2000: 224). Narrative ability continues to improve through adulthood, including more complex structures, story telling devices, and development of personal style (Reese et al. 2011). As with other areas of pragmatic skill, children’s narratives are co-constructed with caregivers long before they are skilled at telling an autonomous narrative. In addi tion, individual differences in caregiver’s style of structuring conversations about non-present and past experiences with their young children relates to how narrative skills develop in young children during the preschool years and beyond (McCabe and Peterson 1991; Nelson and Fivush 2004; Uccelli et al. 2005). Ninio (1988) observed that caregivers begin to alert their infants about which imme diate past events to pay attention to as early as 10 months during routine interactions. For example, the infant may knock over a block tower, and the caregiver may respond ‘uh oh’. Ninio found that this type of marking of immediate past events is often ac companied by a description of what happened. So in this example, rather than simply exclaiming ‘uh oh’ upon seeing the block tower collapse, the caregiver may expand with the description ‘fell down’. These early discussions about immediate past events were also observed with 18-month-old children but were rarely in conversations with 26-month-old children. By the time the children were 26 months, caregivers spent
1
European American children refers to primarily white middle-class children of European descent.
308
308 Pamela R. Rollins more time talking about non-present events. Ninio (1988) suggested that these early markings help the young child understand that past events are something to pay atten tion to and talk about. Around 20 months of age, the early routines and naming games played by young children and their caregivers expand to include (i) conversations about objects and events that are rooted in the here-and-now and (ii) conversations about non-present topics related to the joint focus of attention (i.e. discussing a related present). For ex ample, a caregiver may point to a picture of a lion in a book and say ‘a lion, did we see a lion?’ the child nods and the caregiver continues ‘that’s right, at the zoo’. The child repeats ‘zoo’, and the caregiver asks, ‘Did the lion say roar?’ Uccelli et al. (2005) found that a child’s opportunity to discuss a related present at 20 months accounted for 23 per cent of the variation in their autonomous narratives at age 5. Further, they suggested that the shift in conversations from discussing a joint focus to discussing a related pre sent provides the child with a rudimentary opportunity to remember past events and bring them into the present. As depicted in the ‘lion’ example above, some caregivers elaborate and extend the child’s contribution in talking about past experiences. They ask additional questions about who, what, where and when an event occurred and supply additional information and details about the child’s experience. In contrast, other caregivers ask fewer questions about past events, repeat what their child says, or let the topic drop if the child does not elaborate (Fivush and Fromhoff 1988; McCabe and Peterson 1991; Reese, Haden, and Fivush 1993). These differences in caregiver style have been observed in caregiver con versations with children from 18 months to 3 years and found to be stable over time (see Nelson and Fivush 2004 for review). Further, children whose caregivers elaborate more tend to have better autonomous narratives during the preschool years (McCabe and Peterson 1991; Nelson and Fivush 2004). Interestingly, the caregiver’s style of talking about past events with their young children is not reflected in the caregiver’s verbosity (Nelson and Fivush 2004). This suggests that questioning young children about shared experiences and elaborating on their verbal and non-verbal comments plays a more important role in the emergence of the structural components of preschool children’s narratives than the sheer amount of talk addressed to the child.
15.7 A Brief Discussion This chapter has focused on a description of phenomena related to the use of social communication in early interaction of predominately Western caregiver–child dyads. Not all infants, even Western infants, are exposed to an abundance of dyadic in teractions with a caregiver (Ochs and Shieffelin 1984; Lieven 1994). Ethnographic re search reveals variation in the socialization of infants across diverse cultural groups (e.g. Ochs and Schieffelin 1984; Bavin 1995; Akhtar 2005), and cultural attitudes govern caregiver beliefs about whether it is appropriate to engage in conversation and other
309
Developmental Pragmatics 309 language-based activities with infants (Bavin 1995; van Kleeck 2006). Nevertheless, much of what is known about the developmentof pragmatic skills in young children comes from studies of European American caregivers and their infants, predominately within dyadic interactions. While the timing may differ, there is evidence to suggest that the developmental trajectory of communicative intention may be similar across cultures as shown from several longitudinal studies in Israel, the USA, China, and Japan (Ninio 1983; Ninio and Wheeler 1984; Ninio and Goren 1993; Snow et al. 1996; Tsuji 2002; Zhou 2002). The developmental trajectories of conversational and narrative skill are culturally and socially mediated, and therefore, culture has a greater impact on how these do mains unfold. For example, the more elaborative conversational style associated with better narrative development is less prevalent among low-income families; nonethe less Peterson, Jesso, and McCabe (1999) found that intervention efforts to increase the mother’s elaboration when discussing past events led to improvements for low- income children’s language and narrative skills. McCabe (1996) documents cultural variation in the narratives of children from diverse groups. For example, Japanese and Latino children may not provide as much narrative detail as European American chil dren, and they may not necessarily elaborate on one event (see McCabe 1996 for a review). Similarly, young African-American girls produce a chaining strategy where narratives consists of ‘a series of implicitly associated personal anecdotes’ (Michaels 1981: 429) not centred on a specific theme or topic. Michaels (1981) points out that when a child comes to school with a culturally non-dominant narrative style, she is less well integrated into literacy activities which adversely affect school success. Within the social interactionalist framework outlined here, pragmatics is the study of the use of language in context for the purpose of communication. The work de scribed here furthers our understanding of the developmental trajectory of pragmatic skills while joint intention is emerging, and Western children begin to engage in shared cooperative activities. As these children begin to understand others’ mental states, they can take others’ perspectives and understand what knowledge is shared and with whom (Baron-Cohen 1989; Ninio and Snow 1996). They move from joint perceptual focus to less context-embedded interactions to finally more decontextualized communication, which does not rely on the listener for inference (Ninio and Snow 1996; Adamson et al. 2013).
310
Chapter 16
Experim e nta l Pragm at i c s Raymond W. Gibbs, Jr
16.1 Introduction Experimental pragmatics has had a long, rich history. Starting back in the mid-1970s, a few psychologists turned away from the dominant interest in isolated sentence com prehension and began studies on the ways people used language in context for social, communicative purposes. This research effort was inspired by both the rejection of classic generative, Chomskyan accounts of linguistic structure and behaviour, and the emergence of philosophical and linguistic proposals on utterance interpretation in context. Linguistic pragmatics explicitly focused on what speakers meant by what they said, and how listeners interpreted these messages. As the writings of philosophers such as John Searle and H. Paul Grice became more prominent in the 1970s, psycholo gists devised experimental tests of the psychological implications of these pragmatic theories. The initial interest here concerned whether philosophical accounts of how people inferred what speakers meant really reflected the underlying mental processes by which ordinary listeners interpreted language in discourse. Eventually, psycholo gists also commenced study of the interactional dynamics of speaking and listening to explore the coordinated, collaborative nature of everyday language use. This chapter describes some of the important research in experimental pragmat ics, most notably studies related to recovering speakers’ intentions, inferring conver sational implicatures, and the role of common ground in discourse understanding. My aim is to demonstrate the utility of different experimental methods for studying pragmatics, and how research findings in the field are relevant to traditional concerns within the linguistic pragmatics community. But I will also argue that the emerging trends in experimental pragmatics mirror some of the division in studies of linguistic pragmatics. On the one hand, many pragmatists seek broad principles that shape all people’s context-sensitive language use, while others focus on individual and cultural
311
Experimental Pragmatics 311 differences in people’s pragmatic actions. The experimental literature demonstrates a similar divide, with studies showing great regularities and significant variation, both within and across individuals, in the ways people speak and understand language. My alternative view claims that dynamical, self-organizing processes form the critical background from which meaningful pragmatic actions emerge. (Gibbs 2006a). The implications of this position for interdisciplinary pragmatic research will be discussed.
16.2 Understanding Speakers’ Intentions Consider a conversation between two colleagues who have both just returned from a holiday break. (1) Mary: ‘Hi, Bob, did you enjoy your trip to Paris? Bob: ‘Well, everything was way too expensive.’ Understanding Bob’s response requires listeners to draw an inference about what he implied given what he said. Most listeners would interpret Bob’s answer as suggesting a negative reply to Mary’s question. This inference assumes that what Bob said un derspecifies his intended meaning. Recognizing a speaker’s intention is a special kind of meaning, which Grice (1957) called an ‘m-intention’. An m-intention is a speaker’s intention to produce an effect in the listener by means of the hearer’s recognition of that intention. If the listener successfully recognizes this intention, then she will have drawn the ‘authorized’ inference. Any other inferences drawn are ‘unauthorized’ or not m-intended. Not surprisingly, there is a tremendous body research from experimental pragmat ics supporting the idea that people regularly seek what others intend to communicate when understanding language. For example, people typically remember a pragmatic implication of an utterance rather than the utterance itself or what it directly asserts (Bransford 1979). One set of studies showed that when people heard sentences such as ‘He dropped the delicate glass pitcher’ and ‘The housewife spoke to the manager about the increase in meat prices’ they often remember them according to their prag matic force, as in ‘He broke the delicate glass pitcher’ and ‘The housewife complained to the manager about the increase in meat prices’ (Schweller, Brewer, and Dahl 1976). Listeners encoded inferences that went beyond the explicit information contained in the sentences, but which were in accord with speakers’ probable intentions in making these utterances. The intention behind a speaker’s utterance is encoded and represented in memory, not the sentence or utterance meaning. The empirical evidence that people often interpret utterances according to speakers’ presumed intentions does not necessarily indicate whether this is an automatic psycho logical process. One attempt to investigate automatic intention recognition during lin guistic processing examined whether people immediately inferred the specific speech
312
312 Raymond W. Gibbs, Jr act motivating a speaker’s utterance (Holtgraves 2008). For example, a wife who says to her husband ‘Don’t forget to go to your dentist appointment today’, aims to remind him of the appointment, and the husband should implicitly recognize the speech act remind as part of her understanding the speaker’s intention. Experiments testing this hypothesis involved participants reading scenarios ending with statements like the above, after which they made a speeded judgement as to whether or not a word like ‘remind’ described the speaker’s action in saying his or her utterance. Across a series of experiments, people quickly agreed that the probe words were relevant to the speakers’ aims significantly more often than when people read control contexts that changed the speech act performed (e.g., when Jenny said, ‘I’ll bet you forgot to go to your dentist appointment today’). This result suggests that people may automatically understand speakers’ remarks by inferring something about the specific illocutionary forces moti vating these utterances, consistent with the proposals of speech act theory (Searle 1969). The strong impulse that people have towards inferring speakers’ communicative inten tions is nicely illustrated in a large body of research on figurative language understand ing (Gibbs 1994; Gibbs and Colston 2012). Many experimental studies demonstrate that people are able to quickly infer the intended pragmatic meanings of many verbal meta phors (e.g. ‘Surgeons are butchers’), idioms (e.g. ‘John spilled the beans about his marriage to Mary’), different types of irony (e.g. ‘A fine friend you are’ implying that you are not a good friend), and proverbs (e.g. ‘Don’t put all your eggs in one basket’). In fact, people often understand figurative meanings for these expressions in contexts far more quickly than they do so-called literal uses of these expressions or non-figurative equivalent state ments (e.g. ‘John revealed the secret about his marriage to Mary’). The speed with which people can infer figurative and indirect meanings demonstrates how they are readily dis posed to understand what speakers intended to communicate by their speech acts, and are not immediately biased towards what speakers’ utterances literally mean apart from their pragmatic interpretations. People’s strong orientation to infer intentional meanings can sometimes lead them to make interpretive errors. For instance, people often misunderstand speakers’ inten tions because of superficial processing, such as answering the question ‘How many of each type of animal did Moses take on the ark?’ by quickly responding ‘two’, because they fail to notice that it was Noah, and not Moses, who rescued animals from the great flood (Erickson and Mattson 1981). Studies also show that people often fail to under stand garden-path sentences, such as ‘While Mary bathed the baby played in the crib’. When people read this sentence and were then asked, ‘Did Mary bathe the baby?’ most replied, ‘yes’, even though the sentence does not say that Mary bathed the baby (i.e. she only bathed herself) (Ferreira, Christianson, and Hollingworth 2001). Ordinary language comprehension may therefore be structured around the use of fast and frugal heuristics to infer speakers’ intentions which occasionally leads listeners to make in terpretive errors. Listeners may sometimes recognize speakers’ intentions without necessarily re sponding to their implied messages. Imagine a debate between political candidates where they were asked a question about their views on the ‘war on terror’ or ‘abortion
313
Experimental Pragmatics 313 rights’. This sensitive topic may be one that a candidate hoped to avoid. Studies show that people better recognize that a speaker is dodging someone else’s question if they are oriented towards the relevance of the response than when they are asked to make a social evaluation of the speaker (e.g. ‘Do you like the person?’) (Rogers and Norton 2011). People even rated speakers as positively when they dodged the question as when they directly answered the question. Of course, there are situations in which dodging a direct question is polite (e.g. when answering ‘Do you like the dress I am wearing?’), and people often ignore questions, or their implications, in many conversations that touch on a wide range of topics. Still, it is clear that people engage in various strategies in conversation in regard to whether or not they focus directly on speakers’ apparent communicative intentions. This conclusion does not imply that people are unable to infer intentional meanings, only that their responses need not always directly address the import of speakers’ communicative intentions. A common argument, especially within literary circles, is that intentional meanings are essentially tied to face-to-face conversations, but are less relevant when under standing written texts (see Gibbs 1999). But experimental research demonstrates that people’s assumptions of authorial intentions are critical to their understanding of writ ten discourse. Consider one set of studies that showed how readers’ assumptions about authors determine the amount of cognitive effort they put into understanding writ ten figurative statements (Gibbs, Kushner, and Mills 1991). Participants were presented with various comparison statements (e.g. ‘Cigarettes are time bombs’) and were told that these were written either by famous twentieth-century poets or randomly con structed by a computer programme lacking intentional agency. The participants’ task in one study was to rate the ‘meaningfulness’ of the different comparisons and in another study simply to read and push a button when they had comprehended these statements. Readers found metaphorical comparisons, such as ‘Cigarettes are time bombs’, to be more meaningful when supposedly written by famous twentieth-century poets than when these same metaphors were seen as random constructions of a computer pro gramme. People also took much less time to comprehend these comparisons when they were told the poets wrote the statements. But people took longer to reject anomalous utterances (e.g. ‘A scalpel is a horseshoe’) as ‘meaningful’ when the poets supposedly wrote these statements. Readers assume that poets have specific communicative inten tions in designing their utterances, an assumption that does not hold for the output of an unintelligent computer programme. Consequently, people put more effort into trying to understand anomalous phrases when they were supposedly written by poets. Thus, people’s assumptions about the basic fact of human intentionality (i.e. hypotheti cal intentionalism) shape their immediate processing, and even in some cases their ultimate interpretations, of written language. Overall, there is a significant body of research showing that people are typically fo cused on recovering speakers’ intentions when they interpret oral and written language. One implication of this claim is that people must strip away the specific language to infer privately held intentional meanings located in the minds of speakers or writers. Yet the idea that intentions are private mental states in the minds of individual people
314
314 Raymond W. Gibbs, Jr can be criticized in various ways. For example, some philosophers take issue with the standard Gricean view and suggest that linguistic understanding is analogous to direct perception rather than inferential reasoning (Millikan 2005). Under this perspective, linguistic understanding is not a matter of inferring speakers’ mental states, but is more focused on determining the functions that an utterance has in context. We conse quently ascribe meaningful content to linguistic utterances via social and environmen tal structures (e.g. conventions, functions, practices) that guide a speaker’s or writer’s linguistic actions. Speakers can be conceived as behaving purposefully in producing linguistic utterances, just as hearts and kidneys behave purposefully, without having to make assumptions about listeners’ underlying intentional beliefs (Gregoromichelaki et al. 2011). Part of the difficulty in attributing specific intentions to speakers is that many utter ances in discourse are incomplete, with other speakers quickly chiming in with their own contributions, some of which take the talk in a new direction. Consider this brief excerpt from a discussion among people attending a ‘Friends of the Earth’ club meeting (Gregoromichelaki et al. 2011: 208). (2)
A: ‘So what is that? Is that er … booklet or something?’ B: ‘It’s a book.’C: ‘Book.’ B: ‘Just &talking about al you know alternative’ D: ‘Om erm, renewable yeah’ B: ‘energy really I think’ A: ‘Yeah.’
This set of exchanges illustrates how people move in and out of the parsing and produc tion roles so that it is not possible to ascribe specific intentions to particular speakers at particular times. Figuring out who intended what meaning in discourse is even more complex in cases of interpreting written texts authored by many people. In some aca demic fields, like physics, scholarly papers have several hundred authors, only some of whom communicate with one another about the writing of the text, and in many cases the authors do not know one another or have maybe never met before (Gibbs 1999). Interpreting what an ‘author’ intends in these cases makes little sense because there is no single, or even generalized, mind from which the intention actually emerged. Similarly, many face-to-face situations also do not come with single, prespecified intentional meanings that originate in the minds of speakers (Gibbs 1999). Consider a situation where a father attempts to get his son to practise his trumpet playing by saying ‘You have a big concert next week. You better start practising.’ Although the father’s primary goal may be to get his son to practise in order to be prepared for the upcoming concert, he may also be concerned with other things as well, such as motivating his son to take greater personal responsibility for his musical career, his obligations to others, or to be more in charge of organizing his own life. What specific intentions must we infer to understand the father’s utterance to his son? Sperber and Wilson (1995) argued, contrary to the Gricean view, that a fundamental
315
Experimental Pragmatics 315 mistake is to suppose that pragmatics ‘should be concerned purely with the recovery of an enumerable set of assumptions, some explicitly expressed, others implicitly con veyed, but all individually intended by the speaker…. There is a continuum of cases, from implicatures where the hearer was specifically intended to recover implicatures which were merely manifest, and to further modification of the mutual cognitive en vironment of speaker and hearer that the speaker only intended in the sense that she intended her utterance to be relevant and hence to have rich, and not entirely foresee able cognitive effects’ (Sperber and Wilson 1995: 201). For example, the father may have said what he did to make manifest a wide array of meanings that he didn’t explicitly have in mind when speaking. Finally, some cultural communities appear to be less concerned than others with in ferring individual speaker intentions. Studies of Samoan discourse, for example, show that a speaker’s understanding of an event, and his or her personal motivations and intentions in saying what is stated are deemed irrelevant (Duranti 1988). What an ut terance means depends on what others take it to mean. Samoans ignore the orator’s alleged intentions and concentrate on the social consequences of a speaker’s words. Consequently, a Samoan speaker will never reclaim the meaning of his words by saying ‘I didn’t mean it’. The audience is more likely to be asked to say more about what some thing means than is the original speaker. Samoans practise interpretation as a way of publicly controlling social relationships rather than as a way of figuring out what a given person intended to communicate. These various observations pose complications for any theory that assumes listeners and readers always infer specific intentional meanings when understanding language. Sufficient experimental research indicates that people aim to infer context-sensitive messages and rarely focus on linguistic meaning apart from what speakers and writ ers try to accomplish in discourse. As such, experimental research is consistent with some aspects of the claims of linguistic pragmatics. Yet there exists significant variabil ity in the extent to which speaker intentions can be clearly identified or automatically inferred.
16.3 Conversational Implicatures How people infer what speakers mean has been the topic of greatest interest to re searchers in experimental pragmatics. Recall the conversation between Mary and Bob. Understanding that Bob’s comment is meant as a particular answer to Mary’s question requires that listeners go through a chain of reasoning because Bob’s answer does not logically follow from Mary’s original question. Grice (1975) called the intended mes sage behind Bob’s utterance a ‘conversational implicature’ which is a natural outcome of speakers’ and listeners’ tacit adherence to the ‘cooperative principle’. This states that a speaker must ‘make your conversational contribution such as is required, at the stage at which it occurs, by the accepted purpose or direction of the talk exchange in which
316
316 Raymond W. Gibbs, Jr you are engaged’ (Grice 1975: 45). The cooperative principle carries with it four maxims in (3): (3) Maxims of conversation Maxim of Quantity: Make your contribution as informative as is required, but not more so, for the current purposes of the exchange. Maxim of Quality: Do not say anything you believe to be false or for which you lack adequate evidence. Maxim of Relation: Say only what is relevant for the current purposes of the conversation. Maxim of Manner: Be brief, but avoid ambiguity and obscurity of expression. Grice noted that speakers do not always uphold these maxims. As long as speakers gen erally adhere to the overall cooperative principle, they can ‘flout’ any of these maxims to produce certain implicatures. For example, Bob’s response to Mary’s question flouts the maxim of quantity. According to Grice’s analysis, Mary would not consider Bob’s response to be uncooperative. Instead, Mary would continue to assume that Bob’s re sponse was cooperative and would therefore derive an acceptable and ‘authorized’ in terpretation (e.g. that Bob did not enjoy his holiday because of the expense). One area in which Grice’s theory of conversational implicature has attracted much atten tion is figurative language understanding. For example, consider the following exchange between two old friends who are catching up with each other after some time apart. (4) Peter: ‘Are you happy being married? ’Mary: ‘My marriage is an icebox.’ How do we interpret Mary’s metaphoric meaning by her statement? Grice’s theory specifically claimed that figurative language (e.g. metaphor, irony, metonymy, idioms) is understood in a series of steps (Grice 1975). First, listeners analyse the literal meaning of the entire expression. Second, they assess whether this literal interpretation is appro priate for the specific context. Third, if the literal meaning is contextually inappropriate, as is the case for figurative language, listeners must then derive the intended figura tive (e.g. metaphorical, ironic) meaning via the cooperative principle (Grice 1975). This view suggests, then, that figurative language should be more difficult to comprehend than corresponding literal speech, because figurative speech requires an additional processing step in which their literal meanings are rejected and their intended figura tive meanings are subsequently inferred (also see Searle 1979b). Many reading time studies dramatically showed, however, that people do not nec essarily take additional time to comprehend the figurative meanings of metaphors, ironies, idioms, proverbs, and so on, if these are presented in realistic social and lin guistic contexts (Gibbs 1994; Gibbs and Colston 2012). There are situations in which people may take quite a bit of time to infer, for example, metaphoric meanings (e.g. interpreting very novel metaphors), and people may require additional effort to infer
317
Experimental Pragmatics 317 context-sensitive messages as when what a speaker says (e.g. ‘I usually sleep with ear plugs’) also implies some further meaning (e.g. ‘My neighbourhood is noisy at night’) (Hamblin and Gibbs 2003). Furthermore, people surely engage in processing of word meanings along the course of discourse understanding. But the classic Gricean model that people first analyse literal or semantic meanings of entire phrases or statements before pragmatics comes into play is simply not an accurate reflection of what people ordinarily do when interpreting figurative language. A different topic where Grice’s theory of conversational implicature has been tested involves scalar implicatures. For example, the statement ‘You got some of the gumballs’ pragmatically implies that you have some, but not all, of the gumballs. How do people derive these pragmatic messages? One theory posits that scalar implicatures can be quickly understood, regardless of context, because they reflect ‘default’ readings (i.e. generalized conversational implicatures) (Levinson 2000). However, many experimen tal studies indicate that scalar inferences are cognitively costly, requiring extra time to be understood (Breheny, Katsos, and Williams 2006; Huang and Snedeker 2009a; Bott, Bailey, and Grodner 2012). These results may be explained by a different theory in which people first analyse the semantic content of each expression before prag matic knowledge is brought in to derive the implicature reading (Huang and Snedeker 2009a), an account that is similar to the one rejected by experimental research on figurative language understanding. One difficulty with most studies is that they examine how people interpret scalar implicatures for sentence without context (but see Breheny et al. 2006). For example, many studies use sentence verification paradigms in which participants are presented with sentences such as ‘Some elephants are mammals’ and are asked to judge whether the sentence is true or false. If the sentence is interpreted semantically as ‘At least one elephant is a mammal,’ it is true. If it is interpreted pragmatically as ‘At least one, but not all, elephants are mammals,’ it is false. Most studies indicate that pragmatic responses are generally slower than semantic responses. This is taken as evidence that scalar infer ences are slow and costly. But a different theory, called the constraint-based view, assumes that listeners proba bilistically weigh alternative interpretations, which include, in these cases, ‘some and possibly all’ and ‘some but not all’ (Degen and Tanenhaus 2014). For example, the exact number is more natural to state than ‘some’ when referring to small sets (e.g. 1, 2, and to a lesser degree 3), while ‘some’ is more appropriate to use when referring to larger num bers. These claims were tested in a series of experiments using a ‘gumball paradigm’. On each trial, participants saw an image of a gumball machine with an upper chamber and a lower chamber. Initially, all 13 gumballs were in the upper chamber. Gumballs then dropped to the lower chamber and participants evaluated statements, such as ‘You got some of the gumballs’. The results showed that ‘some’ is, indeed, less natural for refer ence to small sets (1, 2, and 3 of 13 gumballs) and unpartitioned sets (all 13 gumballs) compared to intermediate sets (6–8). These findings suggest that some of the delays found for scalar implicature processing in earlier results may be due to interference from more natural and more available lexical alternatives to ‘some’ such as number
318
318 Raymond W. Gibbs, Jr terms rather than to a two-stage process in which semantic meanings are determined prior to pragmatic ones. More generally, the speed and robustness with which a scalar implicature is generated depends on the amount of probabilistic contextual support for the implicature. Both the experimental findings on figurative language and scalar implicatures imply that people do not invariably compute literal, semantic, default, or context-free mean ings before pragmatic knowledge is accessed to infer context-sensitive meanings. Some psychologists, including myself, have further interpreted this body of research to ques tion the very existence of literal meaning and the classic distinction between semantics and pragmatics (Gibbs 1994; Gibbs and Colston 2012). Most dramatically, the Grice and Searle views on conversational implicature and indirect speech act understand ing cannot be viewed as psychological theories capable of accounting for the diverse experimental findings. There are other theories, however, that may be better able to account for the ex perimental research on pragmatic inference processes. Relevance theory maintains that speakers aim to be optimally relevant in saying what they do (Sperber and Wilson 1995; Wilson and Sperber 2012b). Under this view, every act of ostensive behaviour commu nicates a presumption of its own optimal relevance, that is, a presumption that it will be relevant enough to warrant the addressee’s attention and as relevant as compatible with the communicator’s own goals and preferences (the communicative principle of relevance). Speakers design their utterances to maximize the number of cognitive ef fects listeners infer while minimizing the amount of cognitive effort to do so. Listeners understand speakers’ communicative intentions via the ‘relevance-theoretic compre hension procedure’ (Sperber and Wilson 2002), by following a path of least effort in computing cognitive effects. They do this by testing interpretive hypotheses (e.g. dis ambiguations, reference resolutions, implicatures) in order of accessibility, and then stopping when their expectations of relevance are satisfied. Consider one detailed experimental test of the relevance theory view. Imagine a situ ation in which a stranger approaches you on the street and says ‘Excuse me, do you have the time?’ If you were wearing a watch, how would you interpret and respond to this person’s request? Some possible replies include the following in (5): (5) a. ‘It’s about 4.’ b. ‘It’s 3 minutes before 4.’ c. ‘It is um … 3:57.’ All these responses provide a reasonable answer to the person’s request. But the three re sponses differ in the exactness of their time given, the form in which it is given (minute– hour vs hour–minute), and whether the answer was given directly or included other paralinguistic information (pauses and filled pauses). Although statement (5b) provides the same cognitive effect as does (5c), it requires more cognitive effort to comprehend than (5c), given the extra mental computation needed to derive the exact time of 3:57 from the statement ‘It is 3 minutes before 4’. Statement (5b) is therefore less optimally
319
Experimental Pragmatics 319 relevant because greater effort is expended than what is required to understand state ment (5c). At the same time, the filled pause in (5c) may work to signal that an answer is forthcoming which is indeed worth the addressee’s continued attention. In this manner, statement (5c) may convey an additional cognitive effect over that seen in (5b), namely that a highly relevant answer is forthcoming, which clearly benefits the addressee and may facilitate her understanding of the speaker’s communicative intention. Of course, statement (5a), ‘It’s about 4’, may provide sufficient cognitive effects with little cognitive effort, unless the questioner first mentioned the fact that he needed to reset his stopped watch. In that case, the approximator ‘almost’ would supply a highly relevant cognitive effect that the following numerical answer is just good enough (e.g. ‘It is almost about 4’). Complicating the pragmatics of the time-answering situation is that some people may wear digital watches and others analogue watches. Although it may be ideal to answer any ‘Do you have the time?’ question with an exact answer, doing so when wear ing an analogue watch may require more effort than when wearing a digital watch. How do people respond to time requests given some of these pragmatic complica tions? Research shows that when people are asked ‘Do you have the time?’ they typi cally provide rounded answers, even when wearing digital watches (van der Henst, Carles, and Sperber 2002; Gibbs and Bryant 2008). The fact that respondents tend to round their answers to time questions, even when wearing digital watches that provide exact times, suggests that conversational exchanges are not guided by an egocentric bias to state what is easiest, or to follow a maxim to always speak truthfully (cf. Grice 1975), both of which would predict that digital watch wearer should invariably give the exact time. Rather, people aim to provide answers to questions that are optimally relevant for the circumstances. In other research, people were approached and asked ‘Do you have the time?’ and their answers secretly tape-recorded (Gibbs and Bryant 2008). An analysis of the re sponses showed that speakers plan their answers to time questions in specific ways by often including acknowledgements (‘Yeah, it is 10 till 4’), approximators (‘It is about 3:30’), and filled pauses (‘It is um 10 till 4’). These linguistic and paralinguistic cues do not simply indicate that the speaker is experiencing production problems, but func tion as a green light for the addressee to continue with the process of deriving relevant cognitive effects. Furthermore, people who wear digital watches and gave exact replies took longer to plan these than the same group of people who gave rounded answers. Thus, people with digital watches who are looking at their watches with the exact time actually put more cognitive effort into producing the exact time than if they had pro duced a rounded answer! But digital watch wearers did not take longer to produce exact replies in a context where the original speaker asked ‘Excuse me, my watch has stopped. Do you have the time?’ This pattern of findings suggests that respondents most easily understood that giving an exact time was optimally relevant in the case where it appeared that the questioner wanted the time in order to reset his watch. On the other hand, respondents were less sure that an exact time was relevant when the question only stated ‘Do you have the time?’ despite the fact that the exact time was easiest to retrieve for digital watch wearers.
320
320 Raymond W. Gibbs, Jr This research on answering time questions is unusual in that it explored people’s pragmatic responses in a real-world context, while still measuring response latencies as is done in typical laboratory psycholinguistic experiments. The results indicate that people appear to be striving for optimal relevance when formulating their pragmatic responses to people’s indirect time requests. Being optimally relevant requires that people not aim for the greatest efficiency in an abstract sense, but take pragmatic con siderations into their immediate evaluation of what to say and their interpretation of speakers’ contextual messages. Still, it is important to note that not all participants in the Gibbs and Bryant studies acted consistently within each experimental condition. There are important, statistically reliable patterns in the data, but also various areas of variability that must also be accounted for, as will be argued in sections 16.4 and 16.5.
16.4 Common Ground One of the long-standing debates in linguistic pragmatics concerns the role that people’s common-ground knowledge plays in context-sensitive speaking and understanding. Psychologists and sociologists have long argued that speakers design each utterance so that their addressees can figure out what they intend by considering the utterance against their current common ground (Clark 1996). Common-ground information is personal (i.e. information uniquely shared by two or more speakers), cultural (i.e. in formation broadly shared by members of a community), and related to what is physi cally co-present. When people converse, they typically design their utterances to take into account the perspective of the listener, which facilitates addressees understanding speakers’ communicative intentions. A simple demonstration of this is seen in a study looking at people’s assumptions about mutually known beliefs and knowledge when speaking with others (Krauss 1987). This study had an experimenter stopping people on the street in downtown Boston, Massachusetts where he asked for directions to Jordan Marsh, a large department store about six blocks away. To a third of the people, the experimenter asked ‘Can you tell me how to get to Jordan Marsh?’ To another third, the experimenter said, ‘I’m from out of town. Can you tell me how to get to Jordan Marsh?’ To the remaining third, the experimenter asked, ‘Can you tell me how to get to Jordan Marsh?’ but did so employ ing a rural Missouri accent, representative of a speech style in a different part of the United States. The addressees’ responses were secretly tape-recorded and analysed for the number of words spoken and the number of places en route that were referred to. When the experimenter prefaced his question with ‘I’m from out of town’, people responded with significantly more words and more place names than when asked this question without the preface. Alerting the addressee to the fact that the speaker does not share the same community knowledge clearly gets respondents to design their answers differently. But the respondents also gave longer, more detailed, answers when the experimenter asked
321
Experimental Pragmatics 321 his question without the ‘I’m from out of town’ preface, but spoke with a Missouri accent. Again, people designed their answers given their assumptions about how well their addressee may most easily infer their communicative intentions. The accrual of common ground enables speakers and listeners to more readily coor dinate their intentional meanings in discourse. One set of studies had two people, who could not see each other, collaborate over about the arrangement of Tangram figures (geometric shapes that are vaguely suggestive of silhouettes of people and other ob jects) (Clark and Wilkes-Gibbs 1986). One person (the director) had an ordered array of these and had to explain their arrangement to the other (the matcher) so that the other person could reproduce the arrangement. Each director–matcher pair did this six times. The main hypothesis is that as common ground is established between the director and matcher during the conversation, it should be easier for them to mutually determine where each figure should go. As expected, the number of words used per Tangram figure fell from forty on the first trial to around ten on the last. For instance, a speaker referred to one figure on Trial 1 by saying ‘All right, the next one looks like a person who’s ice skating, except they’re sticking two arms out in front’, while on Trial 6 the speaker said ‘The ice skater’. A similar decline was observed in the number of turns required to complete the arrangement task, showing that the interchange became more economical as common ground was established. Other studies using this experimental paradigm indicate that speakers and listeners can also coordinate to hide information from overhearers without damaging their own understanding of each other’s commu nicative meanings (Clark and Schaefer 1987). These data demonstrated that the assess ment of common ground has an integral part in determining what speakers specifically say and in facilitating listeners’ recovery of speakers’ intentions. One implication of these findings is that utterance interpretation is a ‘joint activity’ of both speakers and listeners, and not solely the responsibility of listeners. Not all psychologists, however, agree that speakers and listeners always aim to be cooperative in conversation by taking the other person’s perspective into account during speaking and listening. Some psycholinguistic research has questioned the necessity of common ground for production and comprehension. Multiple stud ies suggest that under some circumstances, such as stress or high levels of cogni tive burden, speakers are more egocentric in their productions than the traditional common-ground view would predict. People frequently misjudge the effectiveness of their own communication precisely because they do not correctly understand what is, and is not, part of their common ground with others. Speakers who have learned the meaning of opaque phrases, for instance, sometimes overestimate the likelihood that other people know those meanings (Keysar and Bly 1995). Speakers also think their own utterances are less ambiguous and more effective than they actually are (Keysar and Henly 2002). Listeners do not consistently consider common ground in their comprehension (Barr and Keysar 2005). One set of studies specifically examined whether people took others’ perspectives into account when interpreting sarcastic remarks (Keysar 1994). These experiments tested if readers could withstand their own privileged knowledge
322
322 Raymond W. Gibbs, Jr when judging whether a speaker meant something sarcastic or not. Participants read stories ending with comments (e.g. ‘Oh yeah, Professor Jones is a real nice guy’) that only they knew were really intended as sarcastic by the speakers (i.e. participants knew that the speaker believed the professor was rude). The participants were, how ever, asked to make the sarcasm judgements from the perspective of a story character that did not have this privileged information most relevant to the speaker’s sarcastic intention. Overall, people generally were far more likely to attribute a sarcastic in terpretation on the addressees’ parts than should have been the case had the par ticipants recognized the presumed common ground existing between speakers and listeners in the story context. People apparently cannot suppress their own privileged knowledge about a speaker’s sarcastic intent when trying to assess the likely interpre tation of other listeners. This conclusion calls into question the idea that an evalu ation of common-ground information is an automatic part of figurative language interpretation. These various studies do not contest the view that common ground exists and may constrain language production and comprehension. Instead, the argument is that initial stages of speech production and understanding are inherently egocentric, particular in moments when speakers experience cognitive stress in some manner. Common-ground information may be used at a later, corrective stage of linguistic production and understanding. Nonetheless, later research argues that the meth ods and theoretical conclusions in Keysar (1994) were flawed, primarily because of a critical missing control group. Gerrig, Ohaeri, and Brennen (2000) showed that readers were just as likely to judge a speaker’s remark as sarcastic given either nega tive privileged information or no privilege at all. Thus, experimental participants can track story characters’ knowledge and intentions as part of their ordinary recourse to common ground during linguistic processing. People may not always correctly assess what is presumably in their common ground with others, which will lead to misun derstandings of speakers’ communicative intentions. One development in experimental studies is that various low-level linguistic, social, and cognitive mechanisms provide explanations of how people appear to achieve audience-designed productions without the need for constructing explicit models of interlocutors (Horton and Gerrig 2005), or the need to rationalize speakers’ actions in terms of mental state ascriptions (Pickering and Garrod 2004). Computational models also demonstrate how it is possible to achieve ‘common-ground’ understandings within a community without each individual necessarily having that knowledge be part of their own private cognition (Barr 2004). Most generally, experimental work shows that people often coordinate and collabo rate when speaking and listening. Although this ability may appear to require people to rely on ‘common-ground’ assessments, speakers and listeners may not always ac curately assess what may be their common ground with others, and in some cases may react more egocentrically. Listeners may also anticipate what may be in their common ground with others, but fail to properly integrate this information during immediate utterance interpretation (Barr 2008).
323
Experimental Pragmatics 323
16.5 Dealing with Variability in Pragmatic Behaviour The empirical findings from experimental pragmatics paint a complex picture of peo ple’s linguistic pragmatic abilities. There is good evidence to suggest that people are sensitive to what others are trying to communicate with language, and actively coor dinate and collaborate with others as well to achieve both individual and social goals. Many studies also demonstrate that drawing pragmatic inferences may often be done quite quickly, given the salient presence of contextual information and background knowledge. But the experimental literature also reveals some complexities about these general conclusions, especially in regard to whether people seek communicative inten tions, easily draw pragmatic inferences, and can really rely on common-ground knowl edge during speaking and understanding. A closer look at the experimental studies also indicates significant variability in people’s pragmatic behaviours. Not only do some people fail to adhere to standard pragmatic norms, many people in the context of a single experiment vary considerably from trial to trial in their performance. For ex ample, particular individuals will sometimes act collaboratively in a given context, but sometimes not. People will sometimes appear to easily comprehend figurative utter ances in context, and sometimes they take considerable effort to infer speakers’ mes sages. The difficulty is that experimental studies typically report average performances, which hide the deeper, non-linear complexities in linguistic pragmatics. None of this should be terribly surprising, in my view. As noted earlier, linguistic pragmatic studies have long differed in the extent to which they affirmed regularities in pragmatic behaviours, presumably driven by implicit pragmatic norms and principles (e.g. of the sort advanced by Grice, Searle, and Sperber and Wilson), or emphasized variations in pragmatic actions, depending on individual motivations, specific social contexts, and cultural norms (e.g. as seen in many sociolinguistic and anthropological studies). Overall, some of the conflicts evident in the experimental pragmatics litera ture closely mirror ongoing debates in linguistic pragmatics. How shall we deal with this state of affairs? Is it at all possible to create a comprehen sive theory of pragmatics that is sensitive to both emergent regularities in pragmatic behaviours as well as different individual and contextual variations? I think that such a comprehensive theory is indeed possible to create. But doing so requires that we think of pragmatics in a different manner than is traditionally done. The key point in this claim is that there is no individuated set of pragmatic modules which drive our linguis tic and real-world actions. Instead, regularities in experience, including pragmatic be haviours, are emergent properties of our self-organizing nature. Biological and physical scientists now recognize that nature is composed of many interacting subsystems that exhibit a strong tendency to self-assemble or self-organize (Kelso 1995; Bak 1996; Spivey 2007). Any system whose structure is not imposed from outside forces or from inter nal blueprints (e.g. internal mental representations) alone can be said to self-organize.
324
324 Raymond W. Gibbs, Jr Self-organizing systems are capable of creating new structures because their dynamics are dominated by these interactions instead of by the activity of isolated components. Emergent mechanisms are temporary, or ‘soft-assembled’, because they do not endure as passively stored representations within the system’s dynamics. Soft-assembly pro cesses operate in highly context-sensitive ways within particular environmental niches to create the very specific physical patterns and behaviours within each system. A wide variety of physical, biological, and human behaviours have now been de scribed as emergent products of self-organizational processes, including the formation of galaxies, termite nests, snowflakes, the foraging patterns of bees and ants, the dy namic shapes of flocks of birds, the symmetrical patterns on butterfly wings, the regular spots on a leopard’s skin, the formation of whirlpools in rivers, the formation of bacte rial cultures, dynamics of traffic jams on freeways, the performance of stock markets, and neuronal activity in the human brain. The self-organizational view suggests that there are multiple constraints affecting linguistic production and understanding. People’s choices of words, and how they are expressed, and listeners’ interpretations of linguistic utterances unfold from dynamical processes that should not be reduced to some specialized set of pragmatic norms, rules, or even principles which are presumably encoded in minds of individuals. It may very well be true that people often act cooperatively in linguistic interactions, and appear to be interested in what others may know and are intending to do. Yet these regulari ties are, once more, emergent from the human system overall and not the output of some specialized ‘pragmatics’ parts of mind. Under this view, people’s in-the-moment behaviours are shaped by a wide range of constraints, from evolutionary forces to fast- acting neural processes that together produce context-sensitive adaptive behaviours. Not surprisingly, different people in different circumstances will self-organize in highly nuanced ways that manifest in varying pragmatic behaviours both in the real world and in the context of laboratory-based experiments. Understanding these complex dynam ics, at the very least, demands that we do not seek averages in human performance, but carefully study how pragmatic behaviours change given differences in who the people are, the languages spoken, the specific task people are engaged in, and the various methods for assessing people’s in-the-moment experiences (Gibbs and Colston 2012). A major advantage of this dynamic, self-organizing approach to pragmatics is that it can explain both the stabilities, and regularities, that are sometimes evident in lin guistic behaviours, as well as the instabilities, or variations, in the way people prag matically behave. Different sets of constraints will ‘push’ the system toward areas of regularities in behaviour and, in other cases, toward moments of instability in people’s pragmatic actions. There is no need to insist that everyone necessarily adheres to puta tive abstract pragmatic norms or rules, nor is there is a need to dismiss variability in pragmatic behavior as mere noise or outliers in a theory of linguistic pragmatics. Both stabilities and variation in experience arise as part of the non-linear structure of self- organization processes. Furthermore, a dynamic, self-organizing perspective enables us to account for pragmatic behaviours as full-bodied, context-sensitive actions that include not only language, but a wide range of non-linguistic factors such as prosody,
325
Experimental Pragmatics 325 disfluencies, gesture, posture, and facial expressions, all of which together are part of human communication. At present, pragmatic theories are unable to account for the ways that speech, gesture, and facial expressions are dynamically linked, and indeed how the very specific bodily behaviours of individuals in conversation, for example, also become coupled to form their own in-the-moment self-organizing systems. The important implication of this dynamical view on linguistic performance, for both experimental pragmatics and more traditional linguistic pragmatic studies, is a cautionary note on how we explain any instance of pragmatic behaviour. We must be careful not to assume that certain regularities in pragmatics necessarily are caused by specialized mechanisms of mind, with all other deviant behaviours simply being seen as irrelevant to theoretical explanation. Traditional pragmatic norms, rules, or principles should be recognized as emergent products of pragmatic behaviour, and not the causal roots of why people behave as they do pragmatically. We should focus on the diverse interactions that arise given differences in the people being studied, their motivations and goals, the specific language and language forms they use, and their explicit and implicit tasks. Finally, we need to pay greater attention to the full-bodied, interactional nature of pragmatic behaviour so that we can better understand how communicative actions are comprised of integrated systems of speech and bodily actions.
326
Chapter 17
C om pu tat i ona l Pragm at i c s Harry Bunt
17.1 General and Computational Pragmatics Pragmatics has been characterized as the study of the relations between linguistic properties of utterances on the one hand, and aspects of the context in which a given utterance is used on the other hand (Bunt and Black 2000). Computational pragmatics is pragmatics with computational means, which include corpus data, context models, and algorithms for context-dependent utterance generation and interpretation. Relations between linguistic form and context of use are apparent, for example, in the definiteness of a noun phrase, encoding a speaker’s or writer’s assumption about be liefs shared with the hearer or reader (presuppositions); in the way the sentence struc ture reflects the speaker’s intention to structure information into a part believed to be new for the hearer and a part assumed to be known (given-new structuring); or in a particular word order indicating which information the speaker wants the hearer to focus on (topic-focus marking). The linguistic side of the relations that are studied in pragmatics is formed by utter ances in a conversation or sentences in a written text. The context side consists in the case of written text of the surrounding text and the setting in which the text is meant to function. In dialogue, the context of an utterance is likewise formed by what has been said before and the interactive setting, but additionally also by perceptual informa tion, e.g. which objects and events are visible for the participants; by social obligations, such as the pressure to return a greeting; and by the epistemic context, e.g. what do the participants know about the topic of the conversation and about each other. Much of this information is dynamic, as it changes during a dialogue and, more importantly, as a result of the dialogue, since the participants in a conversation influence each other’s
327
Computational Pragmatics 327 state of information. Dialogue contexts are thus updated continuously as an effect of communication. The use of large corpora, supported by software tools for searching and querying the data, has the effect that studies in pragmatics can be informed more systematically by empirical observations. This is evidently beneficial for developing pragmatic theories with better coverage and validity. Such developments may well go beyond minor ad justments to existing theories; they may lead to a rethinking of pragmatic concepts and theoretical frameworks. An example is the development of the theory of dialogue acts, originating from speech act theory, under the influence of spoken dialogue analysis through corpus collection and computational modelling. Modern dialogue act theory deals with very different and richer sets of action types then traditional speech act theory, and uses a different, more complex and articulate notion of communicative action. Pragmatics has also been characterized as being concerned with making explicit what a speaker leaves implicit. Conversational implicatures and indirect speech acts are two well-known cases where the speaker means something else or something more than what he says. For example, suppose I am meeting a student to discuss an essay that she had handed in, and that I praise the student for the attractive layout, for the catchy title, for its length, for the pretty diagrams that it contains, for the nice colours in the screen shots, for the good quality of the paper on which it is printed, … and the student starts to worry: by praising peripheral aspects of the essay, an implicature emerges (from the apparent violation of the Gricean maxim of relevance) that I am not very positive about the contents. Note that this implicature is specific for the context in which conversation takes place; if I would say similar things to an assistant in a copyshop, then no such implicature would arise. Indirect speech acts, as illustrated by the classical example ‘Can you pass me the salt?’, also illustrate how a speaker can mean more (‘Please pass me the salt’) than he says, by relying on the addressee to infer the intended meaning in the given context. Again, this indirect interpretation arises only in certain contexts; in the context of a rehabilitation clinic where passing the salt is a common physical exercise, that interpretation would be unlikely to occur. Conversational implicatures and indirect speech acts illustrate that speakers can convey something more or something else than what they say by relying on the ad dressee’s ability to infer the intended interpretation by combining information from the utterance with information from the context, using Grice’s general cooperative princi ple and the more specific maxims to guide the inference process. Context-based inference is needed not only to understand an indirect speech act or a conversational implicature, but in general to understand any utterance in a conver sation. A simple illustration is formed by utterances of ‘Yes’, which can be a positive answer to a propositional question, or a positive feedback act, or the acceptance of an offer, or an expression of agreement, or several other things. Without taking the pre ceding dialogue into account there is no way of knowing the meaning of this utterance.
328
328 Harry Bunt An interesting, more intricate example can be found in the following fragment of a dia logue between the information service at Amsterdam Airport (I) and a client (C). The utterances (1c) and (1d) have different meanings in spite of being identical; their correct understanding clearly depends crucially on information which is not in the utterances themselves but in the context. (1) a. C: what departure times do you have for flights to Munich early in the morning? b. I: I have Lufthansa at 07:15, KLM at 07.25, again KLM at 07:50, and another Lufthansa at 08:20 c. C: and that’s on Saturday too d. I: and that’s on Saturday too Utterance (1c) is an example of what has been called a declarative question, i.e. a ques tion expressed by a declarative sentence. Beun (1989) found that in a corpus of spoken information-seeking dialogues about 20 per cent of the questions had this form, and that the participants, as well as subjects participating in recognition experiments, rec ognized these utterances perfectly well as questions (more specifically, as check ques tions), even though they look like statements. The understanding of an utterance like (1c) as a question can be explained as follows. 1. A participant in an information-seeking dialogue may have several reasons for performing a communicative action with content p: he wants to know whether p; or he wants the other participant to know that p; or he believes that the other participant would like to know whether p. 2. In a dialogue setting of this kind, the client believes the information agent to be an expert concerning the domain, so for every proposition p about the discourse domain D, C believes that I knows whether p (and that I knows this ‘better’ than C). It therefore makes no sense to interpret C’s utterance as a statement. 3. Likewise, it cannot be the case that C contributes the utterance because he be lieves that I wants to know whether p, hence it makes no sense to interpret C’s utterance as an answer. 4. Looking for an interpretation that does make sense (in view of the maxim of relevance), a remaining possibility is that C contributes this utterance because he wants to know whether p. This means that the essential condition for C to ask a question about p is fulfilled. 5. As already noted in 2, C believes that I knows whether p. The most important supporting condition for C to ask I whether p is thus also fulfilled. Therefore the utterance can be interpreted as the question whether p. Note that interpretation not only of utterance (1c) as a (check) question but also of utter ance (1d) as a confirmation requires inferencing, even though none of these utterances
329
Computational Pragmatics 329 exemplify indirect speech acts, conversational implicatures, or other phenomena that are well known to require inferencing. In fact, the interpretation of utterances in a con versation or in a running text in general involves inferencing; it is the rule, rather than the exception, that interpretation and inferencing are interlocking processes.
17.2 Inferencing in Computational Pragmatics Inferencing can take a variety of forms, which can be divided into deductive, abductive, and inductive reasoning. Deductive reasoning is the form of inferencing that has been defined by logicians for combining facts and hypotheses in order to construct proofs of their logical consequences. This form of reasoning is important in science, but is not used much in daily life. Abduction works in the reverse direction: given an observation, abductive inference yields hypotheses that could explain the observation. For example, given the observation that the street is wet and the general fact that streets get wet when it rains, abduction yields as a possible explanation the hypothesis that it rains. Different from the consequences derived by deduction, the hypotheses derived by abduction are not logically valid—they may be wrong. For example, the street might be wet because a water pipe has burst, or because the fire brigade has held an exercise. Abduction is a form of inferencing that people seem to apply all the time in order to interpret and explain what they see and hear. Induction is inferencing by generalizing from examples. For instance, from encoun tering lots of examples of white swans, and never seeing a swan which is not white, one may induce that swans are white. Of course, one will revise one’s opinion upon seeing a black swan; like the results of abductive inference, the results of inductive inference may be wrong; they are not logically valid. And like abduction, induction is a form of inferencing that people commonly use. Induction is especially important for construct ing models of the world, while abduction is important for the interpretation of observa tions, in particular for the understanding of communicative behaviour. Human inference in everyday situations is shallow rather than deep, employs vast amounts of general as well as situation-specific knowledge, and aims at plausible and useful rather than logically valid results. A fundamental issue in the construction of plausible interpretations is the question what makes one interpretation more plausible than another. Consider again the observation that the street is wet (O1), which would be entailed both by the circumstance that it is raining (P1) and by the circumstance that a water pipe has burst (P2). According to abductive inference, both P1 and P2 are possible explanations of O1. Most people would say that P1 is more plausible, due to the fact that it rains more often than a water pipe bursts. Note that this is only the case in the ab sence of further information; if the observation O1 occurred in a context where it only rarely rains, but where bursting water pipes are a notorious problem, then P2 could be
330
330 Harry Bunt the more plausible explanation. So frequency of occurrence apparently has an influence on the plausibility of interpretations. Stickel (1988) implemented a form of abduction called weighted abduction, which uses numerical plausibilities built into the inference process. Propositions involved in inferences are given a cost. For example, when a rule is applied of the form P ⇒ Q then Q will cost more than P, so shorter proofs cost less than longer ones. Proofs with lower cost are considered to provide more plausible explanations than proofs with higher cost. The framework called ‘Interpretation as Abduction’, developed by Hobbs and asso ciates (Hobbs 1990b; Hobbs et al. 1993), applies weighted abduction to explain a variety of context-dependent semantic and pragmatic phenomena. As an illustration of what this form of abduction can do, Hobbs et al. (1993) use the example (2), which displays three phenomena that require reasoning with context in formation in order to arrive at a good understanding: (1) the use of a definite article; (2) the unspecified relation in the nominal compound ‘Boston office’; and (3) the meto nymical relation connecting an office and the act of making a phone call. (2) The Boston office called. Concerning the interpretation of the nominal compound, compositional semantic analysis may produce a logical form of the following kind, where the predicate NN rep resents the unknown relation between Boston and the office, and the variables x and y should be understood as existentially quantified: (3) boston’(x) & office’(y) & NN(y,x) The knowledge base against which this interpretation takes place contains among other facts the existence of a certain office B1, located in Boston, i.e. it contains the following fact: (4) office’(B1), located-in(B1,B), boston’(B) It also contains the general knowledge that location is a possible implicit relation be tween the elements of a nominal compound, i.e. it contains the fact located-in(x,y) ⇒ NN(x,y). From these facts, abductive inference does indeed construct the interpreta tion that the office mentioned in (3) can be understood as the office B1. There may of course be other offices in Boston than B1, and other known offices with other relations to Boston, which would provide alternative interpretations of (3). This is where the weights of different possible interpretations come into play, and determine which in terpretation will come out as most plausible. This example illustrates that the inferencing in language-understanding systems typi cally combines situation-specific and general world knowledge. Ovchinnikova et al. (2014) have presented a system based on weighted abduction, called mini-TACITUS (re implementing Hobbs’ TACITUS system; Hobbs 1986), which incorporates a knowledge
331
Computational Pragmatics 331 base extracted from the large-scale resources WordNet (Fellbaum 1998) and FrameNet (see Ruppenhofer et al. 2006). The ‘synsets’ of which WordNet is made up correspond to word senses; a lexeme can participate in several synsets. For every word sense, its frequency in the WordNet annotated corpora is indicated. Ovchinnikova et al. use the lexeme-synset mapping for generating axioms with the frequencies of word senses con verted into axiom weights. For example, the verb compose is mapped into its sense synset- X (in WordNet version 3.0), which represents one of its senses, as shown in (5). (5) synset-X(s,e) ⇒ compose(e, x1, x2) FrameNet represents the lexical meaning of predicates in terms of frames that describe prototypical situations. Every frame contains a set of roles corresponding to the partici pants in the situations that may be described. Moreover, syntactic patterns show the surface realization of verbs and their arguments and contain information about their frequency in the FrameNet annotated corpora. From these patterns and their frequencies, weighted axioms are derived, which together with the axioms derived from WordNet form a large base of common-sense knowledge about the world as we speak about it. Preliminary ex periments with this approach show promising possibilities for making clear exactly how knowledge about the world and abductive inference can explain such phenomena as con versational implicatures, understanding of nominal compounds, and metonymy. The mini-TACITUS system and its knowledge base illustrate the three most impor tant kinds of tools that computational pragmatics brings to general pragmatics: (1) al gorithms, in this case for making inferences; (2) data, notably digital corpora with utterances annotated with pragmatically relevant properties, in this case WordNet and FrameNet corpora; and (3) computational representations of knowledge and context, in this case the general knowledge base derived from these corpora.
17.3 Language as Action in Context 17.3.1 Speech acts and dialogue acts A fundamental contribution from the study of spoken language is the insight that ut terances are best viewed as actions. When we talk, we greet, ask questions, apologize, answer, make promises, and so on—we perform speech acts. While specific syntactic structures, lexical items, and prosodic forms can be used to encode some of these action types, in general the communicative functions of an utterance cannot unambiguously be derived from its linguistic form, but require taking the context into account. This point was illustrated by utterances (c) and (d) in example (1) above. In computational pragmatics the notion of a dialogue act (Bunt 1969, 1989) has become popular for modelling the use of language as the performance of actions. While the term ‘dialogue act’ is sometimes understood in the sense of ‘speech act used in
332
332 Harry Bunt dialogue’, the distinction between the concepts of speech act and dialogue act is more fundamental. Where speech act theory is an approach to meaning within the philoso phy of language (Austin 1962a; Searle 1969), the theory of dialogue acts is an empiri cally based approach to the modelling of linguistic, non-verbal, and multimodal com municative behaviour, and considers dialogue acts as acts with an articulate internal structure, related to their functions in a dialogue. A formal definition of the dialogue act concept, as used in ISO standard 24617-2 for dialogue annotation, is as follows: (6) A dialogue act is a stretch of communicative activity of a dialogue participant, interpreted as having a certain communicative function and a semantic content, and which may additionally have certain functional dependence relations, rhetorical relations, and feedback dependence relations. (ISO 24617-2:2012; cf. Bunt et al. 2010)
Dialogue acts offer a way of characterizing the intended meaning of communicative behav iour computationally in terms of update operations on the information states of dialogue participants; this approach is known as the ‘information-state update’ or ‘context-change’ approach (Bunt 1989, 2000; Traum and Larsson 2003). For example, when an addressee understands the utterance ‘Do you know what time it is?’ as a question about the time, then the addressee’s information state is updated to contain (among other things) the informa tion that the speaker does not know what time it is and would like to know it. If, by con trast, it is understood that the speaker is reproaching the addressee for being late, then the addressee’s information state is updated to include (among other things) the information that the speaker does know what time it is. Distinctions such as that between a question and a reproach concern the communicative function of a dialogue act, which is one of its two main components. The other main component is its semantic content, which describes the objects, properties, relations, situations, actions, or events that the dialogue act is about. The communicative function of a dialogue act specifies how an addressee updates his informa tion state with the information expressed in the semantic content, when he understands the dialogue act. The major differences between speech act theory and dialogue act theory are the following: a. speech act theory is concerned with verbal behaviour; dialogue act theory applies also to non-verbal and multimodal behaviour; b. speech act theory assumes that every utterance encodes one speech act; dialogue act theory assumes utterances to be multifunctional; c. dialogue acts have a computational semantics defined in terms of update opera tions on dialogue participants’ information states; d. dialogue acts are not considered in semantic isolation, but as dependent on other dialogue acts, to which they mave have semantic and pragmatic relations. These differences are discussed in the remainder of this section.
333
Computational Pragmatics 333
17.3.2 Dialogue segmentation A spoken dialogue is naturally segmented into turns, defined as stretches of speech produced by one participant and bounded by periods of silence of that participant. The ‘turn’ notion is closely related to that of having ‘the floor’ (Sacks et al. 1974), or the ‘speaker role’, which has been defined as role of a participant who has temporary control over the dialogue and speaks for some time (DAMSL Revised Manual 1997). Turns can have a complex structure, as the following example illustrates (Allwood 1992): (8) A: Yes! Come tomorrow. Go to the church. Bill will be there. OK? B: The church, OK. A’s turn contains sequentially the five functions feedback giving, request, request, statement, and response elicitation. This shows that dialogue acts often correspond to seg ments that are smaller than turns. Utterances are mostly understood to be contiguous stretches of linguistic behav iour which satisfy some well-formedness constraints, e.g. being a grammatical unit such as a sentence, clause, or phrase.1 Syntactic and prosodic features are often used as indicators of utterance endings, but the detection of utterance boundaries is very hard in general (see e.g. Shriberg et al. 1998; Stolcke et al. 2000; Nöth et al. 2002). Using grammatically and/or prosodically defined units of segmentation runs into the problems that (a) not every grammatical unit expresses a dialogue act, and (b) not every functionally relevant segment forms a grammatical unit, since such segments are not always grammatically well-formed, may stretch over more than one turn, are not always contiguous, and may contain parts contributed by different speakers, as illustrated below. In order to deal with these phenomena, the notion of a functional segment has been introduced (Geertzen et al. 2007), which is defined as follows: (9) A functional segment is a minimal stretch of communicative behaviour that has a communicative function (and possibly more than one); it is minimal in the sense of not having parts that are irrelevant for expressing its communicative function(s).
Example (10) shows that a functional segment may be discontinuous: (10) A: Do you know what time the next train leaves? B: The next train is … let me see … at 7.48.
1
Confusingly, in the literature the term ‘utterance’ is sometimes used to refer to everything said within a single turn, and sometimes to refer to smaller segments that have a communicative function.
334
334 Harry Bunt The discontinuous stretch ‘The next train is [ … ] at 7.48’ is a functional segment with the communicative function Answer; due to the minimality condition in (9), this func tional segment does not include the stretch ‘ … let me see …’, which is itself a functional segment with the communicative function Stalling. Example (11) shows that a functional segment may stretch over more than one turn: (11) a. b. c. d. e. f.
A: and what departure times do you have on Saturday? B: on Saturday I have a Lufthansa flight in the morning leaving at 8:15, A: yes, B: and a KLM flight at 08:50, A: yes, B: and a Garoeda flight at 10:30.
The utterances (b), (d), and (f) in (11) together constitute a multi-turn answer to the question in utterance (a). The intervening ‘yes’ utterances are functional segments with two communicative functions: they provide positive feedback, indicating that the pre ceding utterance was well understood, and they give the speaking turn back to the previous speaker. Example (12) shows that a functional segment may contain parts contributed by dif ferent speakers: (12) A: and then, what is the, ehm, the branch office admin, ehm, … B: booking code—the admin booking code for this office is 14 2600 In (12), participant A is struggling to formulate a question; B jumps in to help and com pletes the question (and subsequently answers it). The functional segment expressing A’s question is the discontinuous stretch ‘what is (…) the branch office admin (…) booking code’, of which B contributed the last two words. Example (13) shows that a functional segment is not always grammatically well-formed: (13) A: what time did he say the meeting will resume? B: uhm, two-thirty I think In (13), B’s contribution ‘two-thirty I think’ is not a well-formed sentence, clause, phrase, or other grammatical unit; yet it is clearly a functional segment, expressing an answer to A’s question. The utterance-initial ‘uhm’ is a separate functional segment with the communicative functions of taking the turn and stalling for time. The use of functional segments, as defined in (9), solves many problems in the seg mentation of dialogue into meaningful units, taking into account the occurrence of interruptions, of multiple speakers, of overlapping talk, and of simultaneous verbal and non-verbal dialogue acts.
335
Computational Pragmatics 335
17.3.3 Multifunctionality in utterances As noted, classical speech act theory assumes that speakers perform one speech act with each utterance. In a critical examination of this view, Allwood (1992) distinguished two cases where a speaker performs multiple speech acts, sequential and simultaneous multifunctionality, both illustrated in dialogue fragment (8), which is repeated here: (8) A: Yes! Come tomorrow. Go to the church. Bill will be there. OK? B: The church, OK. The sequence of five dialogue acts performed by A illustrates the phenomenon of se quential multifunctionality; moreover, A’s contribution illustrates simultaneous multi functionality in that ‘Bill will be there’ can be taken to be both a statement and a promise. Sequential multifunctionality disappears when sufficiently small units of segmenta tion are used; this evidently happens when we divide A’s turn in (8) into five segments. However, the size of segments is not what really matters. Bunt (2010) notes that multifunc tionality also occurs in more complex forms, and additionally distinguishes overlapping, discontinuous, and interleaved forms of multifunctionality. The latter is illustrated in (14): (14) I think twenty five euros for a remote … is that locally something like fifteen pounds … is too much money to buy an extra remote or a replacement one … or is it even more?
This stretch of speech cannot be cut up into a sequence of functionally meaningful contiguous subutterances, since the part ‘I think twenty five euros for a remote’ does not express a dialogue act, and neither does ‘is too much money to buy an extra remote or a replacement one’. By contrast, using functional segments works perfectly well; it divides (14) into the discontinuous segments ‘I think twenty five euros for a remote ( … ) is too much money to buy an extra remote or a replacement one’ and ‘is that locally something like fifteen pounds ( … ) or is it even more?’. Similarly, the occurrence of nested discontinuities in segments corresponding to dialogue acts can be handled well by means of functional segments, as example (15) shows.2 Four functional segments can be identified here: (1) the turn-initial ‘and’, which is a Turn-Keeping act, indicating that the speaker wants to keep the speaker role; (2) the discontinuous segment ‘and so ( … ) we started from an empty lot’, expressing an Inform act; (3) the embedded discontinuous segment ‘we started in ( … ) we started from’, which expresses a Self-Correction; and (4) the embedded segment ‘uh, …,’ which forms a Stalling as well as a Turn-Keeping act. (15) A: and so [we started in [uh, … ,] we started from] an empty lot
2
From the Switchboard corpus, .
336
336 Harry Bunt The use of functional segments has important advantages for accurately analysing dia logue in terms of its functional units, doing justice to the many forms of multifunction ality. Functional segments by definition have no sequential, interleaved, overlapping, or discontinuous multifunctionality, but they do allow simultaneous multifunctionality: a single functional segment may express more than one dialogue act.
17.3.4 Dialogue act interpretation Two important differences between dialogue acts and the acts of traditional speech act theory is that dialogue acts have a more articulate internal structure, described below, and a computational interpretation in terms of information state updates. The details of such a semantics depend on the precise definition of information states. Poesio and Traum (1998) describe an axiomatic semantics for dialogue acts using an enriched form of discourse representation structures (Kamp and Reyle 1993) to model information states. Other proposals include the use of Constructive Type Theory (CTT; see Ahn 2001), of Type Theory with Records (RTT; see Cooper 1998, 2000), of minimal partial models (Bunt 2000), and of typed feature structures (Petukhova 2011). The definition of DiAML, the Dialogue Act Markup Language which is part of the ISO annotation standard 24617-2, includes a semantics for dialogue acts that makes no assumptions about the representation formalism used in information state modelling. The only assumption of this semantics is that an information state has a part called the ‘dialogue history’, which records the contributions that have been made to the dia logue, and a part called the ‘pending context’ where update information is buffered that needs to be checked for consistency before being added to the consolidated part of the information state (Bunt 2011, 2014). The semantics exploits the hierarchical structure of taxonomies of dialogue act types defined in annotation schemes such as DIT++ and the ISO 24617-2 scheme by defining the update operations of dialogue acts as joins of certain elementary update operations. The following example illustrates this. The semantics of the communicative function Propositional Question is defined as (16), where Va is the evaluation function for dialogue act annotation structures in DiAML: (16) Va(Propositional Question) = λ X. λY. λYi. λp. U10(X,Y,Yp, p) ∪ U11(X,Y,Yp, p) The variables in the right-hand side have the following significance; X and Y are the sender and addressee of the dialogue act; Yp is the addressee’s pending context; and p is the propositional content of the question. The elementary update operations U10 and U11 are defined as follows: (17) U10(X,Y,Yi, p): add to Y’s pending context the information that participant X wants to know whether p U11(X,Y,Yi, p): add to Y’s pending context the information that participant X assumes participant Y to know whether p
337
Computational Pragmatics 337 Applied to two participants and a proposition, the join (16) is a function that updates the addressee’s pending context. This approach to dialogue act semantics accounts for inferences among dialogue acts, such as a Confirm act entailing an Answer act, because the interpretation of a confirmation is an update operation which differs from that of an answer in that it causes the additional update saying that the sender believes the addressee expected the information supplied in the confirmation. Similarly, a Threat can be shown to entail an Inform act, and the acceptance of an offer to entail a request. See Bunt (2014) for details.
17.3.5 Dialogue acts and grounding Stalnaker (2002) introduced the notion ‘common ground’ as ‘what the participants in a conversation treat as their common or mutual knowledge. Two people’s common ground is the sum of their mutual, common, or joint knowledge, beliefs, and supposi tions’. The participants in a conversation must have a certain common ground at the start of the interaction, in order to be able to understand each other and to contrib ute utterances that can be understood by others. Clark (1996) situates the notion of common ground squarely within pragmatics when he says: ‘Common ground is im portant to any account of language that appeals to “context”’. Grounding in dialogue is the phenomenon that the participants in a conversation update their common ground, in particular adding elements to the perceived common ground.3 In Clark and Schaefer’s classical model of grounding (Clark and Schaefer 1989), participants in a dialogue try to establish for each utterance the mutual belief that the addressees have understood what the speaker meant. This is accomplished by the col lective actions of the current contributor and the partner, in units called ‘contributions’. Contributions are divided into an acceptance and a presentation phase, giving every con tribution the role of accepting the previous contribution, except for those that express negative evidence. Computational studies based on this model include its extension to human–computer interaction by Brennan and collaborators (Brennan 1998; Cahn and Brennan 1999) and the formal theory of grounding by Paek and Horvitz (1999). Traum (1994) provides a computational model of how conversants reach a state of mutual understanding of what was intended by the speaker of an utterance. His model relies on the distinction of so-called ‘grounding acts’, and the use of protocols which can determine, for any sequence of grounding acts, whether the content expressed by the utterances com prising the acts is grounded. Matheson et al. (2000) use elements of Traum’s model in their treatment of grounding from the Information State Update perspective. They repre sent grounded and ungrounded discourse units in the information state, and change their status from ungrounded to grounded through grounding acts. Acknowledgement is the only type of grounding act that is implemented; its main effect is to merge the information 3 The analysis of the relation between symbols and the entities that they refer to involves the notion of ‘symbol grounding’, which is not directly related to ‘grounding’ in the sense of common-ground construction.
338
338 Harry Bunt in the acknowledged discourse unit into the grounded information. The model keeps only the last two utterances in the information state, so it is not quite clear what would happen if the utterance to be grounded occurs more than two utterances back. Bunt et al. (2007) describe a computational model of grounding based on the DIT theory of dialogue acts. In this theory, whenever a speaker performs a dialogue act in tending to make the addressee aware that a certain proposition p holds, in the absence of information to the contrary the speaker expects the addressee to become aware that p. Moreover, the addressee knows that speakers operate this way. And the speaker knows this. And addressees know that speakers have this knowledge. And so on, and so on. This infinite iteration can be summed up in a finite way as ‘speaker and addressee mutually believe that the speaker expects that the addressee becomes aware that p’. On this approach, the performance of a single dialogue act creates an element in their common ground, albeit a belief of an uncertain nature, an ‘expectation’, which would need to be strengthened to become a firm belief of the kind that is usually thought to form the common ground. It is argued on empirical grounds that such a strengthening tends to occur after two rounds of positive feedback, making use of the ‘Feedback Chaining Principle’: (18) If you receive positive feedback on your last contribution to the dialogue, then that is evidence for you that the current speaker believes that you successfully processed his preceding contribution.
The following dialogue fragment illustrates this principle. (19) a. b. c. d. e.
U: Where should I insert the paper? S: In the feeder. U: Should I put it in the bottom front tray? S: No, in the open tray on top. U: OK thanks.
In utterance (e) in (19), participant U gives positive feedback on S’s utterance (d), indi cating that S’s answer was well understood and was a useful answer to the question in (c). From this, S may infer that he correctly understood that question. In other words, positive feedback on the last contribution implies positive feedback on the contribution before that. Bunt et al. (2007) show that this model is backed up by evidence from dialogue corpora, and that it can be effectively implemented in the Information State Update approach.
17.4 Relations in Dialogue and Discourse Some dialogue acts are inherently dependent for their meaning on one or more dia logue acts that occurred earlier in the dialogue. This is for example the case for answers,
339
Computational Pragmatics 339 whose meaning is partly determined by the question that is responded to, and for the acceptance or rejection of offers, suggestions, requests, and apologies. The following example illustrates this, where the meaning of (20a) clearly depends on whether it is a response to the question (20b) or to the question (20c). (20) a. A: I’m expecting Jan, Alex, Claudia, and David, and maybe Olga and Andrei. b. B: Do you know who’s coming tonight? c. B: Which of the people from the project will be there tonight? As an answer to (20b), it says that no other people are expected to come than the ones that are mentioned, but as an answer to (20c) it leaves open the possibility that others will come, who are not ‘from the project’. For dialogue acts which have such a dependence on other dialogue acts, due to their responsive nature, the marking up of the links to their ‘antecedent’ dialogue acts allows the annotation to express not just that an utterance is an answer, for example, but also to which question it is an answer. This type of relation is called a functional dependence relation. Feedback acts also relate to what happened earlier in the dialogue, but in a different way; they signal something about the processing of what was said before—such as its perception or its interpretation. The following example illustrates this. (21) a. A: Is this flight also available on Thursday? b. B: On Thursday you said? With his utterance, B checks whether he heard correctly what A said. This is a response to A’s utterance, rather than to the dialogue act that the utterance expresses. This type of dependence relation is called a feedback dependence relation. Note that positive feedback, signalling correct hearing, understanding, or agreement with what was said, is often expressed non-verbally, for instance by nodding, or in a multimodal way by head movements in combination with vocal backchannels like ‘mm’. Nodding can also be used to perform responsive dialogue acts with a positive semantic content, such as a positive answer to a propositional question, or an accept ance of an offer. These are some examples of the interpretation of non-linguistic com municative behaviour in terms of dialogue acts. Dialogue acts may also be semantically related through other relations, as in (22):4 (22) a. A: it ties you on in terms of the technology and the complexity that you want b. A: like for example voice recognition c. A: because you might need to power a microphone and other things Of the three dialogue acts that are contributed sequentially by the same speaker, the one expressed in (22b) is related to the one in (22a) through an Exemplification relation,
4
From the AMI corpus.
340
340 Harry Bunt and (22c) is related to the one in (22a) through an Explanation relation. Such relations, known alternatively as ‘rhetorical relations’, ‘coherence relations’, or ‘discourse relations’, have been studied most as relations between units of written text. A wide diversity of taxonomies and classifications of such relations have been proposed and discussed in the literature; e.g. by Mann and Thompson in the framework of Rhetorical Structure Theory (RST; Mann and Thompson 1988); Sanders et al. (1992); Hovy and Maier (1993); Penn Discourse Treebank (PDTB; Prasad et al. 2008). These studies are all focused on discourse relations in written discourse; Lascarides and Asher (2007), Tonelli et al. (2010), and Petukhova et al. (2011) study such relations in dialogue, where they may occur either between the events or propositions that form the semantic contents of dialogue acts, or between the dialogue acts themselves, as illustrated in (23) and (24), respectively. (23) a. A: I can never find these remote controls. b. B: That’s because they don’t have a fixed location. (24) a. A: How much would people be willing to pay for a remote, max? b. A: I’m afraid we tend to forget that when we consider all these features. In (23) the semantic content of the dialogue act contributed by B is rhetorically related through a Cause relation to that of A’s contribution; in (24), by contrast, the dialogue act expressed in the second utterance provides a Motivation for asking the question expressed in the first utterance. The three types of relations considered here, functional dependence relations, feed back dependence relations, and rhetorical relations, all contribute to the meanings of the related dialogue acts, in particular of the ‘dependent’ acts. They are therefore taken into account in the theory of dialogue acts and in the more advanced dialogue act an notation schemes, such as DIT++ and ISO 24717-2.
17.5 Computational Resources for Discourse and Dialogue Pragmatics 17.5.1 Corpora, standards, and other tools Large-scale annotated corpora, especially those with (morpho)syntactic, semantic, and pragmatic annotations are the basis of modern research in linguistics, both in develop ing empirically valid theories and in empirical testing of linguistic theories. And not only that: such corpora are also the basis for developing modules in natural-language- processing systems, such as syntactic and semantic parsers, through the application
341
Computational Pragmatics 341 of machine-learning techniques, as well as for evaluating such modules by means of quantitative measurements of their performance. Large-scale annotated corpora of written, spoken, and multimodal discourse in which pragmatic phenomena have been marked up, are still scarce, however. Typically, they either contain annotations of communicative functions (in transcribed spoken di alogue) or annotations of coherence relations in written text. The scarcity of pragmatic resources is partly due to the lack of sufficiently well-developed and generally accepted theoretical accounts of pragmatic phenomena, which would be needed as a basis for widely accepted annotation schemes. Generally speaking, the creation and use of large annotated corpora depend on the availability of two kinds of computational and methodological resources: • well-founded and widely accepted annotation schemes, ideally in the form of an notation standards, including comprehensive sets of carefully defined annotation concepts and interoperable representation formats; • software tools to support the use of such schemes in the creation of consistently annotated corpora and in their querying. Existing resources of these kinds are discussed in the rest of this section.
17.5.2 Annotation schemes Of the pragmatic phenomena that have received attention in the construction of anno tated corpora, the use of language to express dialogue acts in spoken interaction stands out. In the 1980s and 1990s a variety of dialogue act annotation schemes was developed, including those of the TRAINS project in the US (Allen et al. 1994), the Map Task studies in the UK (Carletta et al. 1996), and the Verbmobil project in Germany (Alexandersson et al. 1998). These schemes were all designed for a particular purpose and a specific ap plication domain; they made use of different but overlapping sets of dialogue act types, and used often mutually inconsistent terminologies. In the 1990s a group of researchers gathered in the ‘Discourse Research Initiative’, and drafted a general-purpose schema for multidimensional dialogue act annotation called DAMSL: Dialogue Act Markup using Several Layers (Allen and Core 1997; Core and Allen 1997). With its focus on multidimensionality and domain independence, DAMSL represented an important step forward compared to earlier dialogue annotation schemes, even though its design was left in an unfinished state. Several variations and extensions of DAMSL have been designed for specific annotation tasks, such as COCONUT (Di Eugenio et al. 1998) and Switchboard-DAMSL (Jurafsky et al. 1997). While more advanced than most other schemes at the time, DAMSL and its deriva tives have serious shortcomings. The communicative functions in DAMSL lack precise definitions; the choice of its dimensions is not well motivated; and its inventory of com municative functions is incomplete. Although intended to support multidimensional
342
342 Harry Bunt annotation, DAMSL is not based on an analysis of multidimensionality; the term ‘di mension’ is used informally to denote a cluster of intuitively similar functions. The design of the DIT++ annotation scheme is based on the observation that par ticipants in a dialogue do not just act in order to achieve a certain goal or perform a certain activity, but also provide and elicit feedback; pause and stall for time in order to avoid unexplained silences; take turns; help each other in expressing themselves; cor rect themselves and each other when detecting a speech error, and so on. A dimension is defined in DIT++ as follows: (25) A dimension is a class of dialogue acts that are concerned with a particular aspect of communication, corresponding to a particular category of semantic content.
Aspects of communication include advancing the task or activity that motivates the dialogue; monitoring attention and understanding; taking turns; managing the use of time; editing one’s own speech or that of another speaker; opening and closing topics; and dealing with social obligations like thanking and apologizing. The categories of semantic content that correspond with these activities are task-related information; the success of processing previous utterances; the allocation of the speaker role; the time requirements of contributing to the dialogue; speech disfluencies; topic progression; and social obligations in conversation. Using this notion of dimension, the DIT++ taxonomy was developed by establish ing criteria for distinguishing dimensions and communicative functions, and incorpo rating communicative functions defined in various schemes (including DAMSL and its derivatives, AMI, DIT, ICSI-MRDA, and Vermobil) into a single comprehensive scheme with precise definitions. This scheme served as the basis for the ISO 24617-2 standard for dialogue act annotation, developed in a collaborative effort involving an international team of experts (see Bunt et al. 2010). The ISO 24617-2 standard includes the definition of the Dialogue Act Markup Language (DiAML), with the information state update semantics mentioned above (see Bunt 2009, 2014). The research involved in constructing the Penn Discourse Treebank (Prasad et al. 2008), where discourse relations have been added to the syntactic trees in the Penn Treebank, has recently been taken as the starting point for an ISO effort aiming to es tablish a standard for the annotation of spoken and written discourses with coherence relations (ISO 24617-8; see also Bunt and Prasad 2016). This effort combines forces with a European initiative (TextLink; see ) to define an annotation scheme for discourse relations applicable to all the languages of Europe, including languages of immigrant groups such as Arabic and Chinese.
17.5.3 Software tools Software tools that have been developed for supporting the annotation of dialogue data include DialogueView (Heeman et al. 2002), ACT (Yang et al. 2002), the GATE tools
343
Computational Pragmatics 343 (Cunningham 2002), the NITE XML toolkit (Carletta et al. 2009), ANVIL (Kipp 2001), ELAN (Wittenburg et al. 2006), and MMAX2 (Müller and Strube 2003). The ANVIL tool, for example, offers a graphical user interface for creating annota tion elements on as many user-defined tiers as desired for a particular purpose or an notation scheme. Tiered representations (also used in some other tools) are convenient for annotating multimodal dialogue, allowing the use of different tiers for different modalities (e.g. one tier for speech, one for gaze direction, one for body posture, one for facial expression … ), and also convenient for multidimensional annotation, using dif ferent tiers for different dimensions. ANVIL allows different tag sets to be imported, as well as annotations to be exported in a variety of formats including the DiAML format of the ISO 24617-2 standard (Bunt et al. 2012). For the annotation of discourse relations, RSTTool (O’Donnell 2000) is a popular tool for the segmentation of written text and the construction of tree structures accord ing to Rhetorical Structure Theory. The Conano tool (Stede and Heintze 2004) sup ports PDTB-style annotation of local coherence relations, detecting words that might function as discourse connectives and using syntactic information to guess the argu ments of the relation. Software tools for corpus searching and querying tend to be corpus-specific, applica ble only to treebanks, for example. Moreover, most search engines are sentence-based, which makes them less useful for discourse phenomena that stretch over multiple ut terances. The generalization of such tools in order to be able to handle discourse-level patterns across corpora and theoretical frameworks is an important task for specialists in computational pragmatics. Another kind of software tool to support pragmatic studies is exemplified by the machine-learning-based automatic dialogue act annotator developed by Petukhova (2011). This annotator has been applied successfully in the multidimensional annota tion of spoken dialogues with DIT++ or ISO 24617-2 tags. The programme performs both the identification of functional segments in raw speech and the assignment of communicative functions to each functional segment. Applied to data from the AMI corpus and the HCRC Map Task corpus, accuracies have been achieved of up to 96 per cent, indicating that it may be an excellent basis for semi-automatic annotation (Petukhova and Bunt 2011).
17.5.4 Annotated corpora As mentioned above, existing pragmatically annotated corpora are concerned mainly with discourse relations in written text or dialogue act occurrences in spoken dialogue. The Penn Discourse Treebank (PDTB) is a useful, fairly large corpus with discourse relation annotations added to the Penn Treebank corpus of syntactically annotated English texts from the Wall Street Journal. The annotation of discourse relations fol lows a lexically grounded approach (see Prasad et al. 2008) and aims to be theory- neutral with respect to the nature of higher-level representation of discourse structure
344
344 Harry Bunt (as in RST, for example), in order to allow the corpus to be usable within different theoretical frameworks. Corpora with (manually constructed) annotations for the oc currence of discourse relations exist for a range of languages, including Chinese, Czech, Danish, Dutch, French, German, Hindi, and Turkish, but these are all of modest size. The Potsdamer Commentary Corpus (version 2.0; see Stede and Neumann 2014) is a collection of 175 newspaper commentaries, annotated with nominal coreference rela tions, discourse connectives (similar to those in the PDTB), and rhetorical structures according to RST. Useful dialogue corpora, annotated with dialogue act information, include the Switchboard corpus (Jurafsky et al. 1997), the ICSI-MRDA corpus (Shriberg et al. 2004), the AMI corpus (Ashby et al. 2005) and the related AMIDA corpus,5 and the HCRC Map Task corpus (Carletta et al. 1996). Unfortunately, the annotations in each of these corpora have been made using corpus-specific annotation schemes and frame works, which make them hard to use in other theoretical frameworks. The creation of large-scale corpora with dialogue act annotation according to the ISO 24617-2 standard remains an important goal for the near future; initial steps in this direction have been made by Fang et al. (2011) in studies of the possibility of semi-automatically converting the annotations in the Switchboard corpus (which uses the SWBD-DAMSL variant of the DAMSL scheme) to ISO-24617-2 annotations and by Bunt et al. (2016) in develop ing the DialogBank, a resource that contains dialogues with gold standard annotations according to the ISO 24617-2 standard.
17.6 Conclusions and Perspectives In summary, computational pragmatics offers in the first place a number of tools and resources in support of research in pragmatics. The use of corpora, annotated for prag matic phenomena, is especially beneficial for the coverage and empirical validity of ac counts of these phenomena. Large annotated corpora are useful not only for linguistic research but also for building components of language-processing systems through the application of machine-learning techniques. The construction of such corpora depends crucially on widely agreed annotation schemes; efforts like the establishment of ISO annotation standards are therefore important. Other computational resources, such as automatic reasoning programmes like the abductive prover of the mini-Tacitus system, are interesting for pragmatic research since they allow us to investigate in detail how inference processes that combine lin guistic information with world knowledge can explain pragmatic phenomena such as conversational implicatures, and context-based interpretation more generally. The fundamental challenge of pragmatics is to understand how language interacts with context and how inference interacts with the interpretation and generation of 5 .
345
Computational Pragmatics 345 language. Computational pragmatics makes important contributions to the efforts to meet this challenge by providing computational models of interpretation, generation, inferencing, and learning. Something which is still missing, however, is the construc tion and use of powerful context models. Much of the work that takes context informa tion into account considers only the linguistic context, i.e. the preceding discourse. This is the only kind of context information that is available in corpora, and therefore for applying machine-learning techniques. This means that only a fraction of the relevant context information is taken into consideration. Richer context models have been in troduced for describing the semantics of dialogue acts, and should also be considered for use in annotating dialogue and discourse corpora. Manual addition of this infor mation to corpus annotations hardly seems feasible in view of its complexity, therefore new computational methods will have to be developed to make such information avail able in annotations.
346
Chapter 18
Clinical Pr ag mat i c s Louise Cummings
18.1 Introduction For nearly forty years, clinical investigators have been characterizing breakdown in the pragmatics of language in children and adults. During this time, a number of develop ments have been notable. Many of these developments have seen pragmatics ascend to a position of prominence in a clinical context (Cummings 2010). Pragmatics is no longer in the shadows of phonology, syntax, and semantics, but is an area of clini cal assessment and treatment in its own right. In fact, pragmatic insights are serving to transform how clinicians assess and treat language at all of its levels. The growing clinical significance of pragmatics reflects to a large extent the increased prominence of pragmatics within linguistics in general (Cummings 2005). But there is a further reason why pragmatic skills are increasingly high on the agenda of clinicians. These skills form an important interface between language and cognition. Cognitive abilities such as theory of mind (ToM) and executive functions are now known to be impaired in many child and adult clients with pragmatic disorders. These cognitive skills can no more be overlooked in the assessment and treatment of clients than pragmatic skills themselves. The clinician who is versed in pragmatics is well placed to understand the cognitive character of many of the communication problems that are experienced by children with autism spectrum disorder as well as adults with schizophrenia, a traumatic brain injury or right-hemisphere damage. Pragmatics, it emerges, is a dual-facing area of enquiry that looks to language in one direction and to cognition in another direction. To the extent that clinicians must address impairments of pragmatics, they must be equally prepared to countenance both dimensions of this discipline. Certainly, this will be a guiding assumption of the discussion of clinical pragmatics in this chapter. In the sections to follow, the aim will be to provide the reader with an overview of the state of the art in clinical pragmatics. Historical developments that led to the emer gence of clinical pragmatics are interesting both in their own terms and for what they can reveal about the present-day study of pragmatic disorders, but will not be pursued
347
Clinical Pragmatics 347 in this context.1 Rather, the discussion will seek to convey to the reader the types of clinical populations studied by investigators in clinical pragmatics as well as the range of pragmatic impairments found in these populations.2 This will include developmen tal and acquired pragmatic disorders which will be discussed in sections 18.2 and 18.3, respectively. The pragmatic turn in the study of communication disorders has had its most profound impact on the assessment and treatment of these disorders. Pragmatic disorders are no exception in this regard with their assessment and treatment em bodying many of the insights brought about by this pragmatic reorientation in clini cal thinking. How clinicians assess and treat pragmatic disorders will be examined in section 18.4. Not all studies that form the large body of empirical research in clinical pragmatics have succeeded in throwing light on the nature of pragmatic disorders. In some cases, it is doubtful that these studies are even addressing behaviours that are truly pragmatic in character (Cummings 2007a, 2007b). One of the reasons so many empirical investigations of pragmatic disorders have failed to be particularly revealing is that these studies are often not theoretically motivated. The lack of a clear theoretical rationale for many investigations is steadily being addressed through the increased use of pragmatic and cognitive frameworks in clinical pragmatic studies. Several of these frameworks will be discussed in section 18.5.
18.2 Developmental Pragmatic Disorders Individuals with developmental pragmatic disorders constitute a large and clini cally diverse population. Pragmatic disorders, which can be linked to events in the developmental period, may be found in children and adults with autism spectrum disorders (ASDs). This neurodevelopmental disorder has severe and wide-ranging consequences for the acquisition of language in general and pragmatic skills in par ticular. Even in those ASD cases where language does emerge, impaired pragmatics can pose a significant, lifelong barrier to effective communication. Pragmatic im pairments may also be found in children with developmental language disorders. In some cases, these impairments may be related to deficits in structural language skills. But it is also clear that other pragmatic impairments are less readily explained by deficits in syntax and semantics. In these latter cases, the pragmatic disorder is pri mary in nature. Children and adults who exhibit intellectual disability may also pre sent with pragmatic disorders. These disorders may be commensurate with deficits
1 The so-called ‘pragmatics revolution’ in the study of communication disorders is well documented by Judith Felson Duchan. The reader is referred to Duchan (1984, 2010) for discussion of the sweeping changes that were introduced by this revolution. 2 A complete review of clinical studies of pragmatics is not possible or desirable in this chapter. For recent, extensive discussion of the literature in this area, the reader is referred to Cummings (2009, 2012a).
348
348 Louise Cummings in other areas of language (e.g. syntax and semantics). Alternatively, pragmatics may be more or less impaired than structural levels of language. Finally, clinical investiga tors are increasingly attempting to characterize the pragmatic skills of children with emotional and behavioural disorders (EBDs). These disorders, which include atten tion deficit hyperactivity disorder (ADHD), conduct disorder, and selective mutism, are slowly revealing themselves to have significant pragmatic impairments (Cummings 2014). This section will examine each of these clinical populations and the pragmatic disorders which occur within them. Approximately 50 percent of individuals with autistic disorder do not develop func tional speech (O’Brien and Pearson 2004). For those individuals with autism who do become verbal communicators, pragmatics is often more deviant than other aspects of language.3 An extensive literature exists on pragmatic impairments in ASD. These chil dren and adults have been found to have difficulty comprehending irony and metaphor (Martin and McDonald 2004; Gold et al. 2010), detecting violations of Grice’s maxims (Surian et al. 1996), using features of context in utterance interpretation (Loukusa et al. 2007b), and synchronizing gestures with speech (de Marchena and Eigsti 2010) (see Cummings 2012b for a discussion of context in clinical pragmatics). Conversational and discourse problems are also commonplace. Jones and Schwartz (2009) found that children with autism initiated fewer bids for interactions, commented less often, used fewer conversational turns to continue ongoing interactions, and responded less often to communication bids than typically developing children during dinner conversa tions. Colle et al. (2008) found that adults with high-functioning autism or Asperger syndrome used fewer personal pronouns, temporal expressions and referential expres sions than control subjects during narrative production. Some areas of intact pragmatic functioning have also been reported in individuals with ASD. There is clear evidence, for example, that subjects with ASD are just as likely as control subjects to derive scalar inferences or implicatures (Chevallier et al. 2010; Pijnacker et al. 2009). Pexman et al. (2011) found that children with high-functioning ASD were as accurate as typically developing children in judging speaker intent for ironic criticisms. However, differ ences in judgement latencies, eye gaze, and humour evaluations revealed that children with ASD displayed less accurate appreciation of the social functions of irony. Pragmatic skills have also been examined in children with developmental language disorders of unknown aetiology, a diagnostic category which is today known by the label ‘specific language impairment (SLI)’. This clinical population includes children in whom pragmatic problems are secondary to deficits in structural language, specifically syntax and semantics. However, it has become increasingly clear that this population also includes a subset of children in whom pragmatic impairments are not so readily explained by structural language deficits (i.e. the pragmatic deficits of these children are primary in nature). In a study of the conversational responsiveness of children with SLI, Bishop et al. (2000: 177) make this same point as follows: ‘this study lends support 3
Volden and Phillips (2010) found that the children with ASD in their study had age-appropriate structural language skills while their pragmatic skills were impaired.
349
Clinical Pragmatics 349 to the notion that there is a subset of the language-impaired population who have broader communicative impairments, extending beyond basic difficulties in mastering language form, reflecting difficulty in responding to and expressing communicative intents’. Following Bishop (2000), these children are described as having pragmatic language impairment (PLI). Several pragmatic deficits have been reported in children with SLI and PLI. These deficits include problems using context to understand implied meanings (Rinaldi 2000), difficulty employing the maxim of informativeness and com prehending metaphor (Katsos et al. 2011; Norbury 2005), and the use of inappropri ate conversational responses (Bishop et al. 2000). Furthermore, a substantial number of studies have reported problems in the generation of inferences by children with SLI, many of which play a role in pragmatic interpretation (Botting and Adams 2005; Spanoudis et al. 2007; Ryder et al. 2008; Adams et al. 2009; Holck et al. 2010). Pragmatic impairments have been extensively documented in children and adults with intellectual disability. General statements about these impairments are difficult given the range of presenting symptoms and diverse organic aetiologies (e.g. genetic syndromes) that are associated with intellectual disability. It is for this reason that clini cal investigators have sought to characterize pragmatic impairments in this population on a syndrome-by-syndrome basis. In this way, individuals with fragile X syndrome produce tangential language during conversation, engage in topic repetition, and have difficulty signalling non-comprehension of language to a speaker (Sudhalter and Belser 2001; Murphy and Abbeduto 2007; Abbeduto et al. 2008). Subjects with Williams syn drome have difficulty with the comprehension of irony and metaphor, as well as with referential communication, the latter in the context of communicating to a speaker that a message is inadequate (Annaz et al. 2009; John et al. 2009; Sullivan et al. 2003). There is evidence that individuals with Down’s syndrome have less impaired pragmatic skills than subjects with these other genetic syndromes (Laws and Bishop 2004). Even in in dividuals with Down’s syndrome, however, there are problems with referential commu nication as well as metaphor and idiom comprehension (Abbeduto et al. 2006; Papagno and Vallar 2001). Other pragmatic findings in this clinical population include problems decoding the flouting of the Gricean relevance maxim in children with intellectual dis ability caused by perinatal hypoxia (Tényi et al. 2008), poorer pragmatic function than other facets of language in a child with cri-du-chat syndrome (Piérart and Remacle 1996) and poorly developed pragmatic skills in a child with FG syndrome with callosal agenesis (McCardle and Wilson 1993). There has been little detailed documentation of the pragmatic skills of children with emotional and behavioural disorders. Such studies as have been conducted reveal that pragmatics is an area of considerable impairment in these children and, in many cases, is more impaired than structural language (Geurts and Embrechts 2008). This is con firmed in a study by Benner et al. (2002), who reviewed twenty-six studies of language skills in children with EBD. These investigators found that across these studies, which included 2,358 children with EBD, pragmatic deficits formed the largest language im pairment (71 per cent) followed by expressive deficits (64 per cent) and receptive defi cits (56 per cent). In an investigation of pragmatic skills in ADHD, Bishop and Baird
350
350 Louise Cummings (2001) reported that 73 per cent of their child subjects attained a score below the 132 cut-off point indicative of pragmatic impairment on the Children’s Communication Checklist (Bishop 1998). The scale measuring inappropriate initiation of conversation revealed particularly poor scores. Adachi et al. (2004) studied twenty-nine children with ADHD and found that their comprehension of metaphor and sarcasm was lower than in a control group of normal schoolchildren. Problems with the comprehension of figurative language have also been reported by Bignell and Cain (2007). There is evi dence that children with ADHD have difficulty drawing inferences when listening to spoken texts (McInnes et al. 2003; Berthiaume et al. 2010). Two-thirds of the children with conduct disorder studied by Gilmour et al. (2004) displayed pragmatic impair ments and behavioural features similar to those found in autism. Donno et al. (2010) reported poorer pragmatic language skills in twenty-six persistently disruptive chil dren than in children who formed a comparison group.4
18.3 Acquired Pragmatic Disorders For the large majority of people, pragmatic skills are normally acquired during the de velopmental period. However, these skills may then be disrupted by disease, illness, and injury which occur in adolescence, adulthood, and later life. An adult may have a cere brovascular accident (or stroke) which causes damage in the left or right hemisphere of the brain. If a lesion occurs in the left hemisphere, pragmatic deficits may occur as part of a wider aphasia (for most people, the left hemisphere contains the language centres that are damaged in aphasia). Alternatively, a lesion in the right hemisphere of the brain may result in a marked impairment of pragmatics in the presence of relatively intact structural language skills. The focal brain damage that occurs in a stroke is quite unlike the pattern of brain pathology which is found in traumatic brain injury (TBI). Multifocal brain damage in the patient with TBI is associated with cognitive communication impairments with pragmatics and discourse skills most adversely affected. Previously intact pragmatic skills may become disrupted in the adult who develops mental ill nesses such as schizophrenia and bipolar disorder. Finally, pragmatic disorders may also result from the onset of a number of neurodegenerative disorders. Chief amongst these disorders is the dementias, several of which are only beginning to have their prag matic impairments characterized. However, it is now clear that pragmatics is disrupted in a number of other neurodegenerative conditions including Parkinson’s disease and
4 In each of these studies, the children have a primary diagnosis of behavioural problems. However, there is also evidence that children with a primary diagnosis of pragmatic disorder display behavioural problems. In this way, Ketelaars et al. (2010) found that pragmatic competence was highly correlated with behavioural problems in a sample of 1,364 children aged 4 years who have pragmatic language impairment. These behavioural problems included most prominently hyperactivity and a lack of pro- social behaviour.
351
Clinical Pragmatics 351 Huntington’s disease. This section will examine what is known about acquired prag matic disorders in each of these clinical populations. Subjects with left-hemisphere damage (LHD) were traditionally assumed to have pragmatic disorders as a consequence of their impairments in structural language (i.e. aphasia). The adult with aphasia, it was argued, could be expected to have dif ficulty generating certain indirect speech acts, for example, not through any deficit in pragmatics as such, but because he or she cannot undertake the syntactic inversion of subject pronoun and auxiliary verb that is the standard way of performing such an act in English (e.g. ‘Can you open the window?’). However, certain findings suggest that not all pragmatic disorders in this population are the result of structural lan guage deficits. Not only is there evidence that pragmatic disorders can persist in adults with LHD even as structural language skills improve, but these adults can also present with impairments of non-verbal pragmatic behaviours (Coelho and Flewellyn 2003; Cutica et al. 2006). Among the pragmatic deficits reported in the LHD population are problems with the comprehension of implicatures and proverbs (Chapman et al. 1997; Kasher et al. 1999), difficulty with verbal pragmatic aspects of discourse produc tion (Bloom et al. 1993; Borod et al. 2000) and, in patients with left prefrontal lesions, impairment of pragmatic inferences (Ferstl et al. 2002). However, studies have also revealed some areas of preserved pragmatic functioning in adults with LHD. In this way, Ulatowska and Olness (2007) found that adults with aphasia were able to achieve discourse coherence during the production of personal narratives through the use of a tightly structured temporal-causal event line, development of a theme, and evalu ation of information.5 This somewhat mixed pattern of pragmatic skills and deficits suggests that investigators still have some way to go before the true status of pragmatic disorders in the LHD population is understood. Although patients with right-hemisphere damage (RHD) can experience aphasic language impairments,6 the primary communication disorder in the RHD population tends to involve pragmatics and discourse. Among the pragmatic deficits in this popu lation, investigators have reported impaired comprehension of non-literal language in idioms, proverbs, and humour (Brundage 1996; Cheang and Pell 2006; Papagno et al. 2006). Other pragmatic findings include reduced sensitivity to violations of Gricean maxims and difficulty varying the production of requests in accordance with the in terpersonal and situational features of an interaction (Brownell and Stringfellow 1999; Surian and Siegal 2001). Champagne et al. (2003) examined the processing of non-literal speech acts in twenty subjects with RHD. These investigators found that subjects with RHD were able to process ironic statements and statements that violated the maxims of
5 The emotional content of the personal narratives had a facilitative effect on discourse production by the aphasic adults in this study. The facilitative role of emotional content on discourse production in LHD is confirmed by Bloom et al. (1993). 6 Aphasia does occur, although rarely, as a result of right-hemisphere stroke. See Dewarrat et al. (2009) for discussion of the language performance of sixteen patients with aphasia after a single first- ever ischaemic right-hemisphere stroke.
352
352 Louise Cummings relation and quantity less easily than their explicit counterparts. Also, indirect requests were processed as easily as direct requests, a finding that was attributed to the conven tional nature of the former speech act.7 Patients with RHD display significant discourse deficits.8 The discourse of adults with RHD has been described as exhibiting tangenti ality, egocentrism, and extremes of quantity, i.e. verbosity or paucity of speech (Blake 2006). The subjects with RHD studied by Marini et al. (2005) produced narratives that had poor information content and lacked cohesion and coherence. Pragmatic inferences have also been found to be impaired in subjects with RHD. The subjects with RHD stud ied by Saldert and Ahlsén (2007) displayed difficulty with inferences about the attitudes or motives of characters in a task (see discussion of theory of mind in section 18.5). Clients who sustain a traumatic brain injury (TBI) can pass standardized language batteries and yet present with significant pragmatic and discourse deficits (Coelho 2007). MacLennan et al. (2002) found pragmatic impairments in 86 per cent of 144 patients with TBI in their study. On a pragmatic rating scale, components that exam ined cohesion, repair, elaboration, initiation, and relevance displayed the highest fre quency of impairment. Subjects with TBI have difficulties in inferencing and inten tionality, the latter related to the mental states and intentions involved in pragmatic skills such as the production of speech acts and the understanding of irony (Dennis and Barnes 2001). These subjects also have difficulty meeting the informational needs of their listeners. In this way, McDonald (1993) reported attempts by two closed- head-injured subjects to explain a novel procedure to a blindfolded listener. These subjects’ productions were rated as disorganized, confusing, and ineffective, with one subject considered to be overly repetitive while the other produced too little detail. Angeleri et al. (2008) found that subjects with TBI performed worse than controls on all scales—linguistic, extralinguistic, paralinguistic, context, and conversational—of an assessment protocol that examines the main pragmatic elements involved in a communicative exchange. Other pragmatic impairments reported in subjects with TBI include violations of Gricean maxims (quantity, relation, manner) and problems with topic management and the use of politeness markers in conversation (Togher and Hand 1998; Coelho et al. 2002; Douglas 2010). Pragmatic and discourse deficits in subjects with TBI have been linked to poor social and occupational outcomes in these clients (Cummings 2011).
7 Some theorists have argued that an indirect speech act such as ‘Can you pass the salt?’ has been used so often to make a request that this standard, conventional form for making a request is now processed as if it were a direct speech act. This conventional form is called a ‘convention of usage’ by Morgan (1991: 250): ‘One can readily see how the expression [“Can you pass the salt?”] could have, via Grice’s maxims, the implicature of a request. In fact it has become conventional to use the expression in this way. Thus speakers know not only that Can you … has a certain literal meaning (a convention of language); they know also that using Can you … is a standard way of indirectly making a request (a convention of usage)’. 8 The first formal study of the discourse deficits of patients with RHD was undertaken by Penelope Myers in 1979. For discussion of Myers’s characterization of these deficits, the reader is referred to section 3.3 in Cummings (2009).
353
Clinical Pragmatics 353 Schizophrenia is a serious mental illness which has a prevalence among adults in the range 0.5 per cent to 1.5 per cent (American Psychiatric Association 2000). The disorder is characterized by severe language and communication difficulties, the most marked of which occur at the level of pragmatics (Cummings 2008). Pragmatic and discourse impairments in this population are wide-ranging in nature. They include dif ficulties in the decoding of Gricean conversational maxims (Tényi et al. 2002). The use of linguistic context during language processing is impaired as is the interpretation of non-literal forms such as irony, metaphor and idiom (Kuperberg et al. 2000; Langdon et al. 2002a; Mazza et al. 2008; Tavano et al. 2008). Subjects with schizophrenia display a strong tendency towards literal interpretation of non-literal language forms. However, this tendency can be suppressed under certain conditions such as when a literal inter pretation represents an implausible scenario. In this way, Titone et al. (2002) found that patients with schizophrenia misinterpreted idioms that are literally plausible (e.g. kick the bucket) and correctly interpreted idioms that are literally implausible (e.g. be on cloud nine). During narrative production, patients with schizophrenia contribute ir relevant information and engage in derailments (Marini et al. 2008). Pragmatic deficits have also been reported in children and adolescents with schizophrenia. Baltaxe and Simmons (1995) studied forty-seven child and adolescent subjects with schizophrenia who ranged in age from 6.9 to 17.2 years (mean age, 13 years 4 months). Pragmatics was the area of language impaired in the greatest number of subjects—83 per cent displayed pragmatic impairments. Pragmatic language skills have been shown to be related to overall functioning in patients with schizophrenia (Byrne et al. 1998). The population of adults with neurodegenerative disorders is clinically diverse. It includes first and foremost clients with dementia, but also adults with Parkinson’s disease, Huntington’s disease, and motor neuron disease (also known as amyotrophic lateral sclerosis). Increasingly, clinical studies are revealing a range of pragmatic im pairments in subjects with these conditions. The best characterized of these impair ments are found in dementia related to Alzheimer’s disease (AD). Investigators have reported that subjects with AD have difficulty with the comprehension of figurative language, the use of cohesion devices, and with referential communication (Ripich et al. 2000; Papagno 2001; Carlomagno et al. 2005; Feyereisen et al. 2007). In recent years, more has become known about the pragmatic skills of clients with non- Alzheimer’s dementias. Kertesz et al. (2010) found significant pragmatic disturbance in patients with semantic dementia. Some 75.7 per cent of these patients displayed problems in pragmatics. Rousseaux et al. (2010) examined verbal and non-verbal com munication in patients with AD and patients with behavioural variant frontotemporal dementia (FTD) and dementia with Lewy bodies (DLB). Least pragmatic impairment was found in the patients with DLB. Greeting behaviour, understanding deictics, and using gestures were impaired in the patients with AD. Pragmatic impairments, which included problems in logically organizing discourse and adapting to interlocutor knowledge, were most severe in the patients with FTD in the study. A growing number of studies are reporting pragmatic deficits in adults with Parkinson’s disease, includ ing difficulty with conversational appropriateness and turn-taking and problems with
354
354 Louise Cummings the comprehension of speech acts, irony, and metaphor (McNamara and Durso 2003; Monetta and Pell 2007; Monetta et al. 2009; Holtgraves and McNamara 2010). Performance on complex comprehension tasks which draw upon pragmatic and dis course skills has been found to be impaired in patients with Huntington’s disease (Saldert et al. 2010).
18.4 Assessment and Treatment of Pragmatic Disorders The pragmatic turn in the field of communication disorders has had its most profound impact on how these disorders are assessed and treated. For the first time it was at least as important that clinicians have a clear understanding of the range of speech acts at a client’s disposal, for example, as it was for them to know the syntactic and semantic structures within a client’s linguistic repertoire. Gradually, the word-and sentence- level testing formats that had dominated clinical assessment began to be displaced by techniques that examined how clients used and interpreted utterances across a range of contexts. Many of these techniques were based upon insights from discourse analysis and conversation analysis. Certainly, all of them were united by a common concern to place emphasis on a range of pragmatic phenomena—conversational partners, speaker meaning, context—that had all but been excluded from more traditional methods of language assessment. Having developed a means of characterizing their clients’ prag matic deficits, clinicians then needed to devise a set of intervention techniques that could achieve the remediation of these deficits. Here again, treatment techniques that target structural language skills were poorly suited to this purpose. Even a well-formed sentence could do little to advance a client’s communicative goals if that sentence was used without regard for politeness constraints or the knowledge state of an addressee in a particular context. To match the proliferation of pragmatically oriented clinical assessments, clinicians increasingly found themselves developing intervention tech niques that gave prominence to the remediation of pragmatics. Today, clinicians have access to several different methods for the assessment and treatment of pragmatic dis orders. In this section, the most prominent of these methods will be briefly examined. The reader is referred to chapter 6 in Cummings (2009) for further discussion. Pragmatics is assessed using one or more of the following methods: (1) pragmatics pro files and checklists, (2) pragmatics tests, (3) assessments based on conversation analysis, (4) assessments based on discourse analysis. Profiles and checklists typically take the form of an inventory of verbal and non-verbal pragmatic behaviours. On the basis of observa tion of communication skills and interviews with relatives and carers, an assessor (usually a speech and language pathologist) decides if a particular behaviour is part of a client’s communicative repertoire. Three such profiles and checklists are the Pragmatic Protocol (Prutting and Kirchner 1987), the Pragmatics Profile (Dewart and Summers 1995) and the
355
Clinical Pragmatics 355 Children’s Communication Checklist (CCC-2; Bishop 2003).9 These assessments differ along a number of parameters including administration procedures, age range of clients, number and type of items used, and diagnostic implications. In this way, the CCC-2 takes the form of a seventy-item questionnaire which is administered by a caregiver in approxi mately five to fifteen minutes. Standard scores and percentiles are provided for the follow ing ten scales: speech, syntax, semantics, coherence, inappropriate initiation, stereotyped language, use of context, non-verbal communication, social relations, and interests. Based on a child’s performance across these areas, two composites are obtained. The General Communication Composite is used to identify children who are likely to have clinically significant communication problems. The Social Interaction Deviance Composite may be used to identify those children who have a communicative profile that is characteristic of autism. Increasingly, investigators are examining the psychometric properties of profiles and checklists. In a validation study of the CCC-2, Norbury et al. (2004) report good in terrater agreement (r=0.79) on the Social Interaction Deviance Composite. Formal tests are less frequently used to assess pragmatic language skills. This is due to a widespread perception that tests are not the best way to reveal impairments of at least certain pragmatic skills.10 The Test of Pragmatic Language-2 (TOPL-2; Phelps-Terasaki and Phelps-Gunn 2007) is certainly the best-known pragmatics test currently available. This test is designed to be used with subjects between the ages of 6 years and 18 years 11 months. It is administered by speech and language pathologists and can be completed in 45 to 60 minutes. The TOPL-2 is norm-referenced and examines pragmatic function ing in the following six areas: (1) physical setting, (2) audience, (3) topic, (4) purpose (speech acts), (5) visual-gestural cues, and (6) abstraction. A question of some interest to investigators is how tests of pragmatics perform, both on their own terms and in com parison with pragmatics checklists, in identifying clients with pragmatic impairment. On this issue, the findings are varied. Young et al. (2005) found that the first edition of the TOPL was effective in differentiating pragmatic language disorders in children with ASD from controls matched on verbal IQ and language fundamentals. In a study of children with ASD with age-appropriate structural language skills, Volden and Phillips (2010) reported that the TOPL identified only nine of sixteen of these children as having pragmatic impairment. This compared to thirteen of sixteen identified as pragmatically impaired by the CCC-2. Neither assessment identified any of the typically developing children in a control group as having pragmatic impairment. Geurts and Embrechts 9
The Children’s Communication Checklist is the most prominent of these assessments. Adams (2002: 976) states that the CCC has ‘rapidly become the instrument of choice for the identification of pragmatic language impairment’. Some indication of the prominence of the CCC can be gleaned from the fact that it has been adapted for use in other languages (Helland et al. 2012; Ketelaars et al. 2009). The CCC has also been employed to examine pragmatic impairment in a range of clinical populations including children with epilepsy (Broeders et al. 2010), children with ASD, ADHD, and SLI (Geurts and Embrechts 2008), and youths with mood or anxiety disorders (Pine et al. 2008). 10 Adams (2002: 973) states that ‘[f]ormal testing of pragmatics has limited potential to reveal the typical pragmatic abnormalities in interaction but has a significant role to play in the assessment of comprehension of pragmatic intent’.
356
356 Louise Cummings (2010) found that the Nijmegen Pragmatics Test and the CCC-2 were able to distinguish between preschoolers with and without pragmatic language impairments. However, the high specificity and relatively low sensitivity values obtained for both assessments indi cated that better cut-off scores for pragmatic impairment are needed. Other pragmatic assessments have embodied the principles and techniques of con versation analysis (CA) and discourse analysis. CA in particular is a valuable means of assessing interactional aspects of pragmatics such as conversational turn-taking and the repair of conversational breakdown. It can also locate the source of certain prag matic difficulties in the conversational style of a client’s communicative partner (Beeke et al. 2007). One CA-based assessment is the Conversation Analysis Profile for People with Aphasia (CAPPA; Whitworth et al. 1997). CAPPA is designed to be used with adults with aphasia and their conversational partners. It includes a structured inter view with the client with aphasia and his or her key conversational partner, an analysis of a ten-minute sample of conversation between the person with aphasia and his or her partner, and a summary profile that brings information from both these elements together. The analysis component of the profile examines initiation and turn-taking, repair, and topic management.11 A range of pragmatic skills, including the use of cohe sive devices, the tailoring of information to a hearer’s knowledge state, and the adher ence to Gricean maxims, can be examined through the elicitation of narrative, proce dural, and descriptive discourse from clients. Clients with TBI in particular often pass standardized language batteries and yet exhibit significant discourse deficits.12 These clients’ deficits, Coelho (2007) has demonstrated, can be characterized at microlin guistic, microstructural, macrostructural, and superstructural levels of discourse using measures such as productivity (e.g. words per T(erminable)-unit), a range of cohesive ties, measures of local and global coherence, and story grammar, respectively. The gains in ecological validity which are possible though the use of these techniques are often offset by the time required to record, transcribe, and analyse even small amounts of conversation and discourse.13 Treatments that target pragmatic language skills are particularly diverse in nature.14 Notwithstanding wide variation in methods, techniques, and goals, four main 11
It should be emphasized that although conversation analysis has been used most often in the assessment of patients with aphasia, CA techniques have also been used to examine the conversational and social interactions of clients with TBI and progressive neurological diseases such as motor neurone disease (Bloch 2011; Denman and Wilkinson 2011). Also, Barnes and Armstrong (2010) proposed the use of CA to examine pragmatic skills in clients with RHD. 12 Coelho (2007: 123) remarked of speakers with TBI that ‘examining performance by means of such batteries may give the impression that communicative skills are intact. However, when individuals with disordered pragmatics are engaged in interactions, the listener has the impression that they are off target, disorganized, or tangential. Thus, the communicative behaviour of interest lies beyond the level of single words or sentences, which such individuals have little difficulty with, but rather involves longer units of language such as discourse’. 13 Coelho (2007: 126) states that ‘to transcribe and analyze a 15-minute sample of conversation may require 3 hours’. 14 Adams (2001: 301) remarks that ‘approaches to pragmatic therapy currently in use tend to be eclectic and a “method” of intervention would currently be difficult to identify’.
357
Clinical Pragmatics 357 approaches to pragmatic language intervention are discernible: (1) remediation of con versation skills, (2) social communication skills training, (3) pragmatic skills training, and (4) teaching theory of mind. Conversation skills are targets of treatment in most pragmatic interventions. Some of these interventions draw upon CA techniques and principles; many others engage with conversation skills in the absence of any particular framework or methodology. Two CA-based interventions are CAPPA and Supporting Partners of People with Aphasia in Relationships and Conversation (SPPARC; Lock et al. 2001). CAPPA has been used successfully to treat a man (‘J.B.’) with fluent aphasia as a result of an intracerebral bleed that occurred when he was 59 years old (Whitworth et al. 1997; Booth and Perkins 1999). In conversation with his brother, J.B. was often put through lengthy repair sequences that had an adverse impact on his participation in conversation. A reduction in these sequences became a target of intervention. CAPPA- based intervention has also been used in a group setting with adults with aphasia (Booth and Swabey 1999). Gains in collaborative repair management were reported by the conversational partners of these adults and were observed in a post-intervention repair analysis. SPPARC contains a Conversation Training Programme that addresses behaviours which are problematic to interaction by means of three stages. Participants are first encouraged to gain insight into conversational patterns (e.g. overlap), then reflect upon these patterns, and finally identify and actively experiment with options for change. Supported Conversation for Adults with Aphasia is a non-CA-based con versational intervention which has been shown in efficacy studies to produce good treatment outcomes (Kagan et al. 2001; Rayner and Marshall 2003).15 Pragmatic language skills may also be targeted for treatment as part of a social com munication intervention. While most clinicians and researchers agree that pragmatics plays a role in social communication, investigators differ on the extent of that role. According to Adams (2005), pragmatics is one of four areas that collectively constitute social communication.16 Adams (2005: 184) employs a metapragmatic approach to the remediation of pragmatics in which ‘intervention focuses on direct work on the formal aspects of pragmatics at a reflective level, explicitly talking about rules and conventions and putting these into practice’. The ‘aspects’ of pragmatics that Adams addresses are conversational conventions, topic management, speech acts, turn-taking, linguistic co hesion, and matching style to context (e.g. politeness). Pragmatic skills may also be di rectly trained apart from any wider emphasis on social communication. Such training typically occurs in groups and involves role-playing activities. Pragmatic skills training was used by Hyter at al. (2001) to treat children diagnosed with emotional and behav ioural disorders and by Wiseman-Hakes et al. (1998) in the treatment of adolescents 15
Although each of these conversation interventions involved adults with aphasia, conversation skills training has been used with other clinical groups, principally individuals with autism and clients with schizophrenia. For discussion of these interventions, the reader is referred to section 6.3.1 in Cummings (2009). 16 Adams (2005: 182) states that ‘social communication development is founded on the synergistic emergence of social interaction, social cognition, pragmatics (verbal and nonverbal aspects), and language processing (receptive and expressive)’.
358
358 Louise Cummings with acquired brain injury. Both treatments produced statistically significant gains in performance as measured on pragmatics tests and scales. Finally, investigators are in creasingly recognizing the role of a range of cognitive deficits in pragmatic disorders. One cognitive deficit in particular, impairment in theory of mind, is increasingly being linked to these disorders (Cummings 2012c, 2013). Clinicians teach ToM skills to children and adults with ASD with a view to achieving gains in the social and commu nication skills of these clients (see Swettenham 2000 for an excellent review of work in this area). ToM-based treatments still have some way to go, however, before they can demonstrate the type of gains achieved through other pragmatic interventions.
18.5 Theoretical Accounts of Pragmatic Disorders As the discussion in sections 18.2 and 18.3 demonstrates, there is now a well-developed clinical literature on pragmatic disorders. Much of this literature is concerned with the characterization of pragmatic disorders in children and adults. We now know, for exam ple, the specific speech acts that are lacking from the pragmatic repertoire of the child with ASD and the non-literal language forms which pose the greatest difficulty for the adult with RHD. Yet even as investigators were undertaking increasingly detailed descriptions of the pragmatic deficits encountered by a range of clients, a worrying gulf was opening up between the large number of studies that characterized pragmatic disorders and an altogether smaller number of studies that attempted to explain those disorders. The result has been a preponderance of findings about the presence and extent of pragmatic disor ders to the almost complete neglect of any theoretical account of those disorders. In recent years, this situation has started to change largely on account of theoretical developments in pragmatics itself and in other disciplines (e.g. developmental psychopathology). These developments are increasingly influencing the types of studies undertaken in clinical prag matics with the theoretical import of studies now a central concern of investigators. In this section, two theoretical frameworks which have had a significant influence on clinical studies of pragmatics will be examined. These frameworks are relevance theory (Sperber and Wilson 1995) and a set of concepts known as theory of mind which have been cham pioned by Simon Baron-Cohen, amongst others. These theoretical positions will be de scribed and their influence on clinical pragmatic research discussed. A critical evaluation of both positions is undertaken elsewhere (Cummings 2005, 2009, 2012c, 2013). In relevance theory, Sperber and Wilson (1995) set out to explain pragmatic inter pretation in terms of an information-processing model of communication that has its roots in cognitive psychology.17 Their position is a distinctly Gricean one in that they subscribe to Grice’s view that the process of communication involves an exchange of 17 Such is the cognitive psychological nature of relevance theory that Kempson (1988: 16) has described Sperber and Wilson’s theory as ‘unrepentant cognitive psychology’.
359
Clinical Pragmatics 359 communicative intentions. However, unlike Grice, who proposed the cooperative prin ciple and submaxims of quality, quantity, relation, and manner as a means of explaining the inferential processes that make this exchange of intentions possible, Sperber and Wilson propose a principle of relevance to account for a hearer’s ability to recover a speaker’s intention in producing an utterance. This principle operates in a cost-benefit fashion with the relevance processing of an utterance ceasing as soon as the cognitive cost of processing that utterance exceeds any contextual implications which can be de rived from it. In demonstration of Sperber and Wilson’s views, consider how relevance theory would account for the implicature of Fred’s utterance in the exchange below: Sally: Would you like to join me for dinner later? Fred: I’m going to the cinema this evening. Fred may be taken to implicate that he does not want to join Sally for dinner later. This particular implicature is the least effortful one for Sally to derive and, as such, is the impli cature that Sally is likely to attribute to Fred. Sally is able to derive this implicature because she possesses certain knowledge and beliefs about the world and about the conversational exchange in which she is engaged. She knows, for example, that a person cannot simulta neously be in two places, so if Fred is at the cinema, he cannot also be with Sally for dinner. Sally also knows that Fred will want to observe politeness constraints in conversation by declining her invitation to dinner indirectly by way of an implicature and by providing some account of why he cannot go to dinner. For Sally to derive any other implicature from Fred’s utterance would involve a much greater level of cognitive processing which, in turn, would incur a higher level of cognitive cost. Imagine if Sally had derived the impli cature that Fred will be coming to dinner. For this implicature to come about, Sally would need to draw upon a larger number of background propositions including her beliefs that Fred likes to have his evening meal before going to the cinema, that he can easily walk the short distance between Sally’s house and the cinema, and that he always attends the late showing of a film. The greater cognitive cost of processing this set of propositions is what precludes Sally from deriving this particular implicature of Fred’s utterance. Relevance theory has a number of features which lend it to an explanation of prag matic disorders. Firstly, the framework is explicit in that Sperber and Wilson are making a clear set of claims about how linguistic utterances are processed by language users. This level of explicitness allows theorists to make predictions about the types of utterances that will be more or less easily processed, about the conditions under which processing will occur or will be terminated, and much else besides. Once validated in language-intact subjects,18 relevance-theoretic claims can then become the basis of hypotheses about the utterances that will cause difficulty for pragmatically impaired clients. Secondly, relevance theory casts pragmatic interpretation in cognitive terms. This is a sine qua non of any pragmatic theory that is to succeed in capturing the various ways in which pragmatics is disordered in children and adults. Increasingly, 18
A number of relevance-theoretic claims have already been validated in language-intact subjects (Noveck and Posada 2003; Ryder and Leinonen 2003; De Neys and Schaeken 2007).
360
360 Louise Cummings clinicians and researchers are recognizing the involvement of a range of cognitive deficits in pragmatic disorders (Cummings 2009). By locating the processes by means of which we interpret utterances alongside cognitive resources such as memory, rea soning, and attention, relevance theory makes the links between utterance interpreta tion and cognition explicit. This situation is advantageous to investigators who are attempting to explain the role of cognitive factors in pragmatic disorders. Thirdly, Sperber and Wilson’s central principle of relevance is characterized in cost-benefit terms. It is likely that some pragmatically impaired clients, such as those who sustain TBI, have fewer cognitive resources to direct to utterance interpretation. These re stricted resources are likely to distort the type of cost-benefit analysis that is the basis of the principle of relevance. So characterized, the principle of relevance becomes a useful device in explaining certain types of pragmatic disorder. Given these features of relevance theory, it is unsurprising that an increasing number of investigators are turning to this framework to explain pragmatic disorders. Loukusa et al. (2007a, b) found a failure to cease relevance processing of utterances in two groups of children (7-to 9-year-olds and 10-to 12-year olds) with Asperger’s syndrome or high-functioning autism. Both groups of children with AS/HFA were more likely than control children to produce a relevant response to a question, but then continue by drifting away from the original answer. Loukusa et al. (2007: 372) concluded that ‘some children with AS/HFA have difficulties in being optimally relevant and in stopping processing after they have given a correct answer’. Other clinical pragmatic studies of children that have drawn upon relevance theory include those undertaken by Leinonen and Kerbel (1999) and Schelletter and Leinonen (2003). Schelletter and Leinonen used the assumption of optimal relevance in relevance theory to explain specification of referents by children with specific language impairment. Leinonen and Kerbel used relevance theory to explain data obtained from three children with reported pragmatic difficulties. Episodes of communicative ‘oddness’ were assessed by both authors and were accounted for in terms of breakdown of key relevance-theoretic notions such as explicature. Clinical pragmatic studies of adults have also drawn upon relevance theory. Langdon et al. (2002b) based predictions about the understanding of metaphor and irony in patients with schizophrenia on the relevance-theoretic distinction between metaphor as a descriptive use of language and irony as an interpretive use of language. Specifically, these investigators predicted that patients with schizophrenia would show greater impairment in the understanding of irony than in the understanding of meta phor because only the former required complex (second-order) mind-reading skills. This relevance-theoretic prediction was confirmed. In recent years, theory of mind (ToM) has had an increasing influence on the study of pragmatic disorders. ToM describes the ability to attribute mental states both to one’s own mind and to the minds of others. One mental state of particular significance to pragmatic interpretation is communicative intentions. The hearer who cannot attrib ute such intentions to the mind of a speaker is unlikely to grasp the ironic intent with which certain utterances are produced, establish the illocutionary force of speech acts, or undertake any of a number of other forms of pragmatic interpretation. Although
361
Clinical Pragmatics 361 different theoretical accounts of this ToM ability exist,19 an underlying assumption of all such accounts as applied to the study of pragmatic disorders is that impairments of ToM are causally related to these disorders. This assumption is certainly intuitively plausible—some capacity for mental state attribution would appear to be fundamen tal to pragmatic interpretation and any impairment of this capacity would seem to have adverse implications for the interpretation of utterances. Moreover, there is now clear evidence of ToM deficits in the types of clinical subjects we described as having pragmatic disorders in sections 18.2 and 18.3. These subjects include individuals with autism spectrum disorder, emotional and behavioural disorders, intellectual disability, right-hemisphere damage, schizophrenia, traumatic brain injury, and neurodegenera tive disorders such as Alzheimer’s-related dementia (see Cummings 2012c, 2013 for discussion of ToM deficits in these subjects). Yet, empirical support for a direct causal relation between ToM deficits and pragmatic disorders is still somewhat tenuous. We conclude this section by considering the findings of several studies that have examined this relation. Typically, studies that examine the relation between ToM deficits and pragmatic disorders attempt to relate some aspect of pragmatic functioning to performance on false-belief tests. These tests are the standard means of assessing ToM skills and can be used to reveal first-and second-order ToM deficits. Most pragmatic interpretation involves second-order ToM reasoning, in that the hearer of an utterance must be able to attribute mental states about another person’s thoughts to the mind of a speaker in order to achieve utterance interpretation. Measures of second-order ToM perfor mance have been variously related to pragmatic skills in clinical subjects. Martin and McDonald (2004) found that second-order ToM reasoning was significantly associ ated with the ability to interpret ironic jokes in individuals with Asperger’s syndrome. A similar finding is reported by Winner et al. (1998) who found that the ability to distinguish lies from ironic jokes in subjects with RHD correlated strongly with two measures of second-order belief attribution. Brüne and Bodenstein (2005) found that approximately 39 per cent of the variance in proverb comprehension among the pa tients with schizophrenia in their study was predicted by ToM performance. McDonald and Flanagan (2004) reported that second-order ToM judgements were related to the ability to understand conversational inference in adults with TBI. Cuerva et al. (2001) administered tests of indirect requests and conversational implications to thirty-four subjects with probable Alzheimer’s disease. These investigators found a significant association between performance on a second-order false-belief task and these tests of pragmatic skills in their subjects. Although these studies provide tentative support for the claim that ToM deficits play a causal role in pragmatic disorders, further research is clearly needed to address this issue in a definitive way. 19 The three contenders to a theoretical account of ToM are: (1) ToM as a cognitive module (modular theory of ToM), (2) ToM as theory construction (theory theory account), and (3) ToM as imaginative projection (simulation theory). See Cummings (2012c) for a discussion of each of these ToM theories in relation to pragmatic disorders.
362
Chapter 19
Neu roprag mat i c s Brigitte Stemmer
19.1 Introduction Our brain, mind, and body, in short our organism constantly interacts with the environment, resulting in human behaviour. The way the organism interacts with the environment depends on many variables such as whether we are asleep or awake (arousal state), whether we are in a good or bad mood (emotional state), what our experiences and goals are, and the environmental configuration in which we inter act. This interaction is reflected in brain activity that ultimately enables language and pragmatic behaviour to emerge (Stemmer 1999a). The way we use linguistic and non-linguistic signs in interaction for the purpose of communication has been referred to as pragmatics, and the underlying brain correlates and processes that bring about pragmatic behaviour have been coined neuropragmatics. (For a dis cussion of the term ‘pragmatics’ see, for example, Levinson 1983 and Huang 2007. See Stemmer 1999a, Perkins 2005, and van Berkum 2010 for the term ‘neuroprag matics’.) As pragmatic behaviour is closely tied to interactive processes that emerge from an organism and operate in the environment, the underlying brain correlates are influenced by both the external (environmental) as well as internal (organism) factors or mechanisms. Early investigations that aimed at elucidating brain correlates and processes underly ing pragmatic behaviour were mainly based on behavioural techniques (such as neu ropsychological testing, reaction time measures, visual half-field stimulation, etc.) and studies with patients having suffered brain damage (so-called lesion studies). These investigations have provided invaluable information and advanced testable hypotheses on the brain substrates and the processes implicated. With the advent of neuroimaging techniques the knowledge gained with the classical methods has been expanded, modi fied, or questioned, as will be outlined later. Regardless of the method used for investi gation, studies in neuropragmatics have served multiple goals such as investigating the breakdown of aspects of pragmatic behaviour in specific patient populations, testing or
363
Neuropragmatics 363 refining psycholinguistic theories, or elucidating the role that each hemisphere plays in language and pragmatic processing. The goal of this contribution is to summarize the findings of studies investigating aspects of linguistic pragmatic behaviour and its relation to brain substrates and/or processes. Considering that linguistic pragmatics operates on words and sentences, it is sensible to first briefly investigate what we know about the neural networks involved in lower-level language (e.g. words and sentences) as they inevitably will also play a role in higher-level language (e.g. text and discourse) processing. This is subsequently followed by a discussion of the neural systems underlying pragmatic behaviour, with a focus on structural discourse and figurative and non-literal language.
19.2 The Default Language Network Language comprehension starts with verbal information entering our brain via audi tory pathways (from the inner ear via the auditory nerve to the primary auditory cortex in the temporal lobe) or visual pathways (from the retina of the eye via the optic nerve to the primary visual areas in the occipital lobe). At this basic level the input is pro cessed in so-called primary sensory brain areas (such as the auditory or visual cortex) and interpreted primarily at a perceptual level. We know, for example, that there is a sound but not that it is the sound of a phone ringing. At this level we are unaware of what the input really means. Further processing of the input occurs in non-primary brain regions (secondary sensory systems and the association cortices). Similarly, when producing language, a thought emerges from (mostly still unspecified) neural systems and for it to be externally expressed we need either to utter it out loud or write it down and thus involve our motor system (which creates coordinated movement of our ar ticulators or writing hand). It is thus not surprising that the sensory and motor systems of the brain are involved in language comprehension and production. The comprehen sion of words, sentences, and discourse involves a number of processes and processing stages (specified in different word and sentence comprehension models) that run in a serial or parallel manner. While we are not yet able to specify in detail the underlying neural circuitry involved in those different processing stages, a rough picture of the neural substrates involved and their possible connections underlying the more general process of language comprehension and production seems to emerge. Those regions in the brain that have consistently been implicated across various studies targeting similar functional tasks at the word or sentence level have been re ferred to as the default language network. It needs to be understood that the default lan guage network represents a rough approximation at a very general level of the neural substrates implicated in word-and sentence-level processing. It is not very well defined and the focus has been on anatomical regions rather than anatomical networks. Only recently have anatomical and functional connections between these anatomical regions been discussed in more detail. The network subsumes the classical language regions
364
364 Brigitte Stemmer such as Broca’s speech production area (left inferior frontal gyrus) and Wernicke’s com prehension area (the left posterior part of the superior temporal gyrus) as well as other brain regions including prefrontal, frontal, temporal, and inferior parietal cortex. At a functional level it has been suggested that this network (and to a lesser extent its right- hemisphere homologue) is recruited to perform semantic unification (Hagoort et al. 2009). For this purpose, specific connecting fiber pathways between these regions and subregions within these regions have been suggested (for summaries, see Hickok and Poeppel 2004, 2007; Vigneau et al. 2006; Hickok 2009; Price 2010; Friederici 2011). Mapping sound to meaning and supporting higher-level syntactic and semantic lan guage processes has been associated with fiber pathways travelling via dorsal (superior) brain regions and connecting the superior temporal lobe with the frontal regions (more precisely the posterior part of the superior temporal gyrus with the premotor region and the inferior frontal gyri). Language comprehension and aspects of speech produc tion have been related to pathways travelling via ventral (inferior) brain regions and linking superior temporal with inferior frontal regions. (For a summary on language pathways see Catani et al. 2005; Saur et al. 2008, 2010; Friederici 2009, 2011.) Within the subregions of the language network, acoustic analysis is first performed in the auditory cortex bilaterally from where information is then distributed to other brain regions. Generally, the recognition of words seems to implicate neural substrates in the superior temporal lobe of both hemispheres. The picture is less clear for the neural organization of conceptual semantic information. While some researchers sug gest a widely distributed network for representing and processing semantic concepts (involving superior/middle/inferior temporal and inferior frontal regions), others pro pose a more focally organized semantic ‘hub’ in the anterior temporal regions sup porting syntactic and/or semantic integration processes. (For a detailed discussion of spoken word recognition and conceptual semantic systems and their neural substrates see Hickok 2009.) The precise role of Broca’s area (in the left inferior frontal lobe) has been discussed extensively (Grodzinsky and Santi 2008; Willems and Hagoort 2009). While its in volvement in aspects of language production is agreed upon, its role in comprehension is still controversial. Although Broca’s area in the frontal lobe has been the classical brain region associated with syntactic processing, non-imaging and neuroimaging findings suggest that regions in the temporal lobe also play a role in sentence-level pro cessing such as the anterior temporal lobe bilaterally and posterior temporal regions (for a detailed discussion, see Hickok 2009; Friederici 2011; Fedorenko et al. 2011). In addition, brain areas lying more anteriorly to Bora’s region in the inferior frontal cortex have been associated with complexity effects such as complex syntactic or complex semantic processing demands (Friederici 2011). Generally, however, there is still a lack of convincing evidence that associates specific syntactic processing with specific neural circuitry (Fedorenko et al. 2011). A widely debated issue is the question where syntactic and semantic information is integrated to achieve interpretation. One view suggests that the left inferior part of the frontal cortex is part of a semantic unification network in which semantic information
365
Neuropragmatics 365 from different modalities is unified. Other regions that have been associated with se mantic unification are the left superior/middle temporal cortex, the left inferior pa rietal cortex, and, to a lesser degree, the right-hemisphere homologues of these areas (Hagoort et al. 2009; see Friederici 2011 for different views). The anterior temporal lobe has also been suggested as a region supporting the integration of semantics with syntax (Friederici 2011). Speech production seems to be guided by sensory-motor systems and is supported by neural circuits that lie predominantly in regions adjacent to the Sylvian fissure (peri-Sylvian regions) such as left fronto-parietal and temporal brain regions (Hickok 2009). For example, feedback of the auditory system is important in modulating our speech output and auditory and motor information must be integrated when learning the sound pattern of new words and their articulation. It seems that the integration of sensory-motor information implicates predominantly regions in the left hemisphere, that is the parietal-temporal junction and the frontal lobe (Hickok 2009). In sum, specific neural systems have been identified that seem to be involved rather consistently in language processing at the word or sentence level (lower language levels). The question thus arises what happens at higher language levels. Are the same regions implicated that have been identified as the default language network? And if so what does this mean? One interpretation would simply be that the same system re sponsible for lower-level language is also responsible for higher-level language process ing. Another one would be that although similar brain regions are involved, they are involved in different ways. Alternatively, brain regions other than the default language network may play a role in higher-level language processing. These issues will be inves tigated in the following sections.
19.3 The neural systems underlying linguistic pragmatic behaviour Historically, the neural substrates underlying pragmatic behaviour have been investi gated in patients who show problems with different aspects of pragmatics. Investigating neural systems based on brain-damaged populations is challenging considering the heterogeneity and variability of these patients, the diversity of the methods used, and the problems related to studying pathological populations for the purpose of gaining insights into the workings of neural systems. With the advent of neuroimaging tech niques, the focus has shifted towards the study of healthy populations. While some findings based on the study of clinical populations were confirmed or modified, others were not and many issues are still unresolved. Although studying neural systems with neuroimaging techniques and a healthy population circumvents some of the challenges mentioned, other challenges remain (such as individual variability, comparisons across groups of different age, gender, and education, and comparisons of findings based
366
366 Brigitte Stemmer on different experimental tasks and procedures), and new challenges inherent to the method or based on the researcher’s decisions arise (such as scanning methods, differ ences in statistical threshold levels, method of analyses or reporting anatomical results). These issues limit the feasibility and thus the significance of the results of qualitative comparisons and quantitative meta-analyses of neuroimaging studies. Certainly, non- imaging methods have their caveats as well. It is thus not easy to decide who or what to trust. A rather non-scientific approach but nevertheless a good rule of thumb seems to be the following: be aware if findings based on different methods do not converge. Unless there are good reasons for the discrepancy, there is room for doubt. Similarly, if findings have not been replicated (and ideally by different research groups), don’t yet bet all your money on it. The findings reported in the following sections should thus be perceived with an open and critical mind.
19.4 Discourse Discourse is more than just linking sentences or utterances to produce a text. It is also a form of social action and interaction of our physical being, the organism, with the environment that surrounds us. As such, discourse can be viewed as a vehicle through which mental events manifest themselves (Kintsch 1998; Stemmer 1999a). Discourse has many facets and one view is that its dimensions include verbal structures (verbal use), communication of knowledge, beliefs, opinions, and action and interaction in social situations (van Dijk 1997). Investigations into the neural systems underlying the verbal use of discourse started with the study of patients with focal lesions in the right hemisphere (right- hemisphere-damaged patients, RHD) usually due to a stroke. RHD patients show a dissociation between well-preserved linguistic abilities and impaired communi cative abilities. This is reflected in difficulties with topic maintenance, identifica tion and extraction of relevant themes, the organization of discourse structure and discourse cohesion and coherence (for summaries, see Brownell and Martino 1998; Myers 2005; Johns et al. 2008). Several suggestions have been made to explain the observed discourse problems in RHD patients such as difficulties with processes that require inferencing and integrating abilities, problems in forming representations of other peoples’ mental states (theory of mind hypothesis), or, as an overarching theory, problems with manipulating mental models (mental model hypothesis) (for summaries see Martin and McDonald 2003; Stemmer 2008). An impairment of spe cific brain mechanisms (such as attention, emotion, inhibition/control mechanisms, working memory functions, or, more generally, executive functions) as underlying causes for the previously mentioned difficulties have also been discussed as contrib uting factors (see, for example, Martin and McDonald 2003, 2006; McDonald 2008). Based on behavioural and lesion studies it was assumed that the neural systems re sponsible for the impaired aspects of discourse reside in the frontal part of the right hemisphere. This view, however, was undermined when it was reported that not all
367
Neuropragmatics 367 RHD patients show discourse problems (Brady et al. 2005) and similar problems were described in other patient populations (such as patients with left-hemisphere frontal lobe damage and subsequent aphasia, patients with temporal lobe lesions de veloping Alzheimer disease, patients with damage to the frontal and temporal lobes developing fronto-temporal dementia, patients with frontal lobe damage due to head trauma, or patients with schizophrenia). These observations showed that discourse ‘symptoms’ are not specific to one patient population (see also Cummings forthcom ing). At the same time, the hypothesis that neural systems outside the frontal cortex and/or the right hemispheres may be implicated in discourse processing gained pop ularity.1 These hypotheses have mostly been investigated with healthy populations using neuroimaging techniques.2
19.4.1 Brain correlates and mechanisms associated with text comprehension The neuroimaging technique most frequently used to investigate the brain substrates underlying language and communication is the functional magnetic resonance imag ing (fMRI) technique. This technique is an excellent tool when asking questions about where in the brain something happens. In other words, this technique has a very good spatial resolution. fMRI studies investigating the neural substrates underlying text comprehension will be summarized first before turning to figurative and non-literal language. FMRI studies of structural discourse have primarily focused on specific aspects of text comprehension such as cohesion and coherence phenomena and inferenc ing processes (for reviews, see Mason and Just 2006; Bornkessel-Schlesewsky and Friederici 2007; Ferstl 2007, 2010; Ferstl et al. 2008; Stemmer 2008). As described pre viously, lesion studies have frequently (although not always) associated specific aspects of text processing with the right hemisphere while remaining vague on specific brain substrates. Neuroimaging studies have thus attempted to clarify the issues surrounding the left versus right hemisphere dichotomy, and, in addition, to identify specific brain substrates involved in text processing. In a meta-analysis of twenty-three fMRI studies addressing text comprehension, Ferstl et al. (2008)3 first compared the brain regions that were activated when language 1 It should be noted that despite different disease aetiology and pathophysiology, it is still possible that the same or similar neural circuitry implicated in specific aspects of discourse processing is affected in various disease processes albeit in different ways. 2 It is beyond the scope of the essay to introduce the basics of neuroimaging and its advantages and limitations. For a brief introduction addressing the lay person see Frank A. Rodden and Brigitte Stemmer, ‘A brief introduction to common neuroimaging techniques’, in Brigitte Stemmer and Harry A. Whitaker (eds), Handbook of the Neuroscience of Language (Amsterdam; San Diego: Academic Press/Elsevier, 2008), 57–70. 3 Note that due to the methodological differences of neurological imaging studies (see section 19.3) numerous studies could not be considered in the meta-analysis, and the results of some of these studies do not fit the suggested pattern.
368
368 Brigitte Stemmer tasks were performed with non-language perceptual tasks. The authors reported ac tivation of brain regions that are typically reported when language is investigated at the word and sentence level (the default language network, see section 19.2), such as the bilateral anterior temporal lobe, left inferior frontal regions, and left-dominant middle and posterior temporal regions. In the next step the authors compared the comprehension of coherent language with incoherent language and reported activa tions again in the typical language regions, however, with dominant activations iden tified in the anterior temporal lobe in both hemispheres and subnetworks in the left frontal and left temporal lobes consisting of the left inferior frontal gyrus and the left superior temporal sulcus. In addition to these regions, coherence building activated left-sided regions of the medial side of the frontal lobe, more specifically the poste rior cingulate cortex, the dorsomedial prefrontal cortex, and ventromedial prefrontal cortex. A region in the inferior parietal cortex (the inferior precuneus) was also acti vated. After identifying specific brain regions that activate during the comprehension of coherent language, the next question to address is why these regions activate, that is, what does this mean? Activation of the anterior temporal lobe has been a consistent finding in discourse comprehension studies and it is also part of the default language network. Some regions of the temporal lobe (the middle temporal gyri) have been viewed as critical gateways for integrating and accessing distributed relevant information and have therefore been characterized as multimodal and transmodal high-order association areas (Mesulam 2000). The anterior extreme of the temporal lobe (the so-called temporal pole) is part of the paralimbic system which plays a role in linking cognition with visceral states and has functionally been related to memory and emotion behaviours (Mesulam 2000). The activation of anterior parts of the temporal lobe in language and discourse com prehension (specifically here coherence building) has thus been interpreted as reflect ing the integration of syntactic, semantic, and episodic memory information to derive semantic representations (Ferstl et al. 2008). Activation of the dorsomedial prefrontal cortex and the border to the parietal cortex (precuneus, posterior cingulate cortex) occurred when the task required active partici pation of the participant, such as making explicit judgements, and was thus associated with inferencing processes (Ferstl 2010). No activation in the medial prefrontal regions was reported in two studies involving predictive inferences (Virtue et al. 2006; Jin et al. 2009) leading Ferstl (Ferstl and Siebörger 2007; Ferstl 2010) to suggest that the dorso medial prefrontal cortex supports conscious strategic processes. Note, however, that studies investigating inferencing frequently use material that requires the comprehen sion or production of different types of inferences which makes these studies difficult to compare. Psycholinguistic research has suggested that different processes underlie different types of inferences (Frederiksen et al. 1990; Frederiksen and Stemmer 1993; Graesser et al. 1994; Mason and Just 2004) and individuals may process inferences dif ferently (St. George et al. 1997). The neural substrates that support inferencing may thus be different for different types of inferencing and in individuals. Finally, it should be noted that the assumption (based on lesion studies) of right-hemisphere involvement
369
Neuropragmatics 369 in inference processes was not supported by the meta-analysis of neuroimaging studies (Ferstl et al. 2008). Another issue that has been addressed in several neuroimaging studies and reviewed by Ferstl (2010) is the way we integrate text content with our background knowledge to build and process situation and mental models (van Dijk and Kintsch 1983; Johnson- Laird 1989). It seems that most studies converge on the finding that regions in the medial parietal lobe (the posterior cingulate cortex and the adjacent precuneus) play an important role in situation model building during the processing of continuous text while the integration of text information is supported by the anterior temporal lobe (Ferstl 2010). Similarly to inferencing studies, no clear picture has emerged concerning the role of the right hemisphere in building and processing situation models. While the fMRI technique provides excellent spatial resolution it is less suitable when the focus is on the timing of processes, that is, when something happens. In those cases techniques based on electrophysiological measures (such as electroencephalo gram, EEG, and event-related potentials, ERPs, derived from the EEG) are used. EEG/ ERP studies have primarily investigated the time course of integration processes, that is, when during processing the information from various sources (such as context, words, sentences, discourse, and non-linguistic sources) is integrated and comprehen sion emerges. This research indicates that information about the phonology, syntax, and semantics of single words and sentences as well as discourse information, world knowledge, and non-linguistic context information immediately converge to support interpretation (Hagoort and Berkum 2007; see, however, Friederici 2002). Electrophysiological studies have also indicated that language users anticipate up coming communication when making sense of what is said and that such predictions are supported by long-term memory (van Berkum 2010). Similarly, language users in corporate valence systems (something is ‘good’ or ‘bad’) when communicating, thus relying on non-language systems. Neuropragmatics is thus not confined to identifying the neural substrates that support communication but also looks at how communica tion evolves over time and at the interaction of non-linguistic systems that influence interaction. To sum up, lower-level language studies have pointed to brain regions that are rela tively consistently activated when language is involved, the so-called default language network. This network is also activated during the production and comprehension of specific verbal structures in discourse but may be extended by additional brain regions (coined the ‘extended language network’ by Ferstl 2010), such as the dorsomedial pre frontal cortex or anterior temporal lobe bilaterally. It was suggested that the dorso medial prefrontal cortex may support conscious strategic processes and the anterior temporal lobe the integration of syntactic, semantic, and episodic memory information (Ferstl 2010). While there is overlap there is no agreement amongst researchers con cerning the functional interplay of these regions. Similarly, there is no consensus as to the specific role of the right hemisphere. Apart from investigating the neural substrates involved, there has also been active research concerning the timing of the processes underlying production and comprehension of verbal discourse structures. While some
370
370 Brigitte Stemmer research indicates immediate interaction between verbal discourse structures (includ ing lower-level linguistic structures), non-linguistic systems, and external environmen tal variables, others have argued that syntactic processes of structure building precede and are independent of semantic processes at early processing stages.
19.5 Brain correlates and mechanisms of figurative and non-literal language Another area of research that has received broad attention is the way we communi cate the ‘unsaid’ as reflected in metaphors (‘Choices are crossroads’), other figurative language (irony, sarcasm, idioms, proverbs), or non-literal language such as indirect requests (‘My car is in the garage’ intended as a request to get a lift home). The neural systems implicated in such computations have been addressed directly or indirectly with behavioural or neuroimaging techniques. Traditionally non-figurative and figurative language comprehension was viewed as being different from literal language comprehension in that conversational norms were violated. It was assumed that in literal language speakers say what they mean (based on semantic knowledge) while in non-literal and figurative language speakers do not di rectly express what they mean and the hearer has to compute meaning by incorporating the context (based on pragmatic knowledge). Understanding literal language was thus assumed to require one processing step while additional processing steps were necessary to comprehend metaphoric language. This has been referred to as the standard pragmatic model and was derived from work in pragmatics (Grice 1975; Searle 1975). Many early studies investigating the communicative abilities of pathological populations based the interpretation of their findings on the standard pragmatic model. When it was reported that patients with focal damage to the right hemisphere showed difficulties comprehend ing non-literal and figurative language, the right-hemisphere hypothesis gained promi nence. This hypothesis advocated the view that the right hemisphere is responsible for the processing of non-literal and figurative language (thus pragmatics), while the left hemisphere was concerned with literal language (thus syntax and semantics) (for a sum mary, see Joanette et al. 1990). Subsequent psycholinguistic evidence with healthy popu lations did not always support the standard pragmatic model and opponents of this view claimed that there was no difference between what speakers say and what they mean in context and hence no additional processing steps were necessary. These issues were hotly debated and alternative views emerged including the view that metaphors are conceptual and reflect embodied experiences (for summaries, see Gibbs 2001, 2006b, 2008; Giora 2003). Despite the lively controversies, many studies in neuropragmatics have used or still use the standard pragmatic model as their starting point or interpretative frame. A variant of the standard pragmatic model is the graded salience hypothesis which suggests that salient meanings are processed initially, regardless of either literal or
371
Neuropragmatics 371 contextual fit (Giora 2003). Saliency is determined by factors such as conventionality, frequency, familiarity, and prototypicality. The graded salience hypothesis predicts a se lective right-hemisphere recruitment in the processing of novel, non-salient meanings. Another view that has gained popularity is that both hemispheres are involved in metaphor comprehension but in different ways. The coarse semantic coding theory suggests that the hemispheres compute semantic information differently and that these differences are reflected in the microcircuitry of the cortex (Beeman 1998; Jung- Beeman 2005). More specifically, while the right hemisphere codes information in a rather coarse way, maintains broader meaning activations, and recognizes distant re lations, the left hemisphere codes fine (close) semantic relationships (Beeman 1998; Jung-Beeman 2005). While separate but highly interactive brain areas in both hemi spheres support semantic processing (that is, the inferior parietal lobe for semantic activation, the inferior frontal lobe for semantic selection, and the temporal lobe for semantic integration), the components of semantic processes are computed in each hemisphere, albeit differently. The coarse coding theory has been proposed to underlie non-literal and figurative language interpretation and discourse integration. As we will see below, there is ample evidence from neuroimaging studies speaking against the standard pragmatic model while evidence for or against the coarse coding or graded salience view is still equivocal (see, however, Giora 2007).
19.5.1 Metaphor comprehension 19.5.1.1 Processing mechanisms and the left–right hemisphere dichotomy Studies of metaphor comprehension have mostly focused on processing mechanisms using time-sensitive electrophysiological techniques (e.g. EEG/ERPs) or on the contri butions of each hemisphere using spatially sensitive techniques (e.g. fMRI). While few neuroimaging studies reported left-hemisphere activation only, most studies showed either right or bilateral activation (for summaries, see Schmidt and Seger 2009; Ferstl 2010; Bohrn 2012). Reasons for these discrepancies are most likely due to the hetero geneity of the experimental paradigms (such as stimuli and task characteristics), indi vidual variability, and analysis methods. Despite this heterogeneity, Schmidt and Seger (2009) suggested a consistent pattern: right-hemisphere activation was reported when novel metaphors were used (Mashal et al. 2005, 2007; Stringaris et al. 2006; Ahrens et al. 2007) and left-hemisphere activation when frozen, conventional, or familiar met aphors were used (Eviatar and Just 2006; Lee and Dapretto 2006; Rapp et al. 2007; Stringaris et al. 2007; Mashal and Faust 2008). However, not all studies fit this pattern. Shibata et al. (2007) and Yang et al. (2009) reported left-dominant activations for novel metaphors, thus leaving room for debate. Processing mechanisms and the time course of meaning activation in literal and metaphoric words/utterances have been investigated in numerous electrophysiological
372
372 Brigitte Stemmer studies. The findings indicate that metaphoric meanings are activated very early in pro cessing and most evidence supports the view that metaphor comprehension and literal language share qualitatively similar processing mechanisms. The evidence thus speaks against a literal-figurative dichotomy (for a summary, see Coulson 2008) and is also dif ficult to reconcile with the classical right-hemisphere hypothesis which was based on such a dichotomy. This is not to say that the right hemisphere is not involved and one possibility is that certain features of the metaphor (or other figurative language) play a role such as the degree of familiarity with the metaphor, the level of figurativeness, the frequency of occurrence, or the complexity of the metaphor. On the basis of the coarse semantic coding theory one would argue that the right hemisphere is involved in pro cessing metaphors because the semantic relationships are more distant in metaphors than in literal language (figurativeness) or in familiar metaphors (familiarity). Schmidt and Seger (2009) made an attempt to separate the effects of figurativeness, familiar ity, and difficulty on the recruitment of neural systems involved in right-hemisphere mechanisms using the fMRI technique. Difficulty was operationalized as requiring more cognitive processing (that is, being more effortful) as reflected in longer reaction times in a task. The authors reported that neither difficulty nor familiarity alone could account for right-hemisphere recruitment during metaphor processing. It was both task difficulty and familiarity conjointly that activated right-middle frontal regions. This partially overlaps with findings from another fMRI study which also reported that task difficulty modulated right-hemisphere activation rather than novelty or figurative ness (Yang et al. 2009). No figurativeness effect in the right hemisphere was found in yet another fMRI study (Diaz et al. 2011). Figurativeness was also viewed as less impor tant for the involvement of the right hemisphere in metaphor comprehension than the ‘complexity of the underlying mapping and integration operations’ by Coulson (2008) who reviewed findings from a series of studies using different methodologies. Finally, no general consensus emerged speaking for or against the coarse coding theory or the graded salience hypothesis. In sum, there is little evidence from behavioural, electrophysiological, and fMRI studies that supports the classical right-hemisphere hypothesis or the standard prag matic model. Concerning the coarse coding theory, some studies seem to fit the theo retical framework, while others do not or do so only partially (Tompkins et al. 2008; Diaz et al. 2011; Gouldthorp and Coney 2011; for a summary, see Coulson 2008). What seems to emerge is that task difficulty or complexity plays a role. There are, however, still too few studies that have systematically investigated the relation between specific metaphor characteristics and task difficulty to reach consensus. In addition, the interaction between traits of the individual (attentional resources, memory capacity, affect), various metaphor features, and task difficulty are currently unknown. So far we have focused on processing mechanisms and the left-versus right- hemisphere dichotomy in metaphor comprehension and bypassed the specific brain regions reported in the studies. As we will see below, reaching consensus concerning the specific neural substrates that are implicated proves even more difficult than reach ing agreement on the more general left–right hemisphere dichotomy.
373
Neuropragmatics 373
19.5.1.2 Brain correlates and metaphor comprehension The brain regions activated across fMRI studies on metaphor comprehension are mostly in temporal and frontal brain regions (for a summary see Schmidt and Seger 2009). Numerous areas within these regions have been reported to activate but these vary widely across studies. There are only a few areas on which studies concur. Several studies associated the figurativeness of metaphors (comparing metaphors to literal sentences) with activations in the right insula and the left temporal pole (Mashal et al. 2005; Lee and Dapretto 2006; Schmidt and Seger 2009). Easy and familiar metaphors (compared to literal sentences) activated regions in the right frontal lobe (inferior fron tal gyrus) while difficult metaphors (compared to easy metaphors) were associated with regions in the left frontal lobe (inferior frontal gyrus) (Rapp et al. 2004; Mashal et al. 2005, 2007; Lee and Dapretto 2006; Ahrens et al. 2007; Stringaris et al. 2007). Although these metaphor studies report similar activation areas in association with specific metaphor features, there is currently no agreement on what these associations mean. Further, there are metaphor studies that show discrepant findings (see Schmidt and Seger 2009 for a summary). While most fMRI metaphor studies have focused on the contribution of each hemi sphere or of specific brain regions to metaphor processing, a different approach has been taken by Bambini et al. (2011). These authors focused on the cognitive demands underlying metaphor comprehension when the context is kept constant. They hypoth esized that the process of metaphor comprehension engages a vast array of cognitive subfunctions in distributed brain networks. Comparing metaphor to literal sentence comprehension the authors reported activation of a diffuse neural network in fron tal (bilateral inferior frontal gyrus; anterior cingulate), temporal (left superior tem poral gyrus), and parietal lobes (left angular gyrus). Within these global systems the authors identified the bilateral inferior frontal gyri as reflecting operations of the conceptual system (integration of information with world knowledge stored in long- term memory), regions along the superior temporal sulcus for familiarity-sensitive mind-reading operations, and the anterior cingulate cortex and prefrontal cortex for attentional control processes. These findings thus suggested that compared to literal processing, metaphorical processing uses a more extended brain network implicating both hemispheres and frontal, temporal, and parietal regions. It has further been sug gested that this extended brain network may not be specific to metaphorical processing but may apply to the processing of other pragmatic phenomena as well, such as other figurative or non-literal language, discourse, or humour processing.
19.5.1.3 Metaphors grounded in bodily experiences While most behavioural and neuroimaging studies consider metaphors a figure of speech, the view that metaphor influences how people think and reason in everyday life and that they are grounded in bodily experience has gained prominence in cognitive linguistics and psychology (for summaries, see Lakoff and Johnson 1999; Gibbs et al. 2004; Gibbs 2006b). In cognitive neuroscience a discussion has ensued around the idea that there is
374
374 Brigitte Stemmer a functional link between sensory-motor systems and conceptual systems of the brain and that semantic mechanisms are grounded in action-perception systems of the brain (e.g. Pulvermüller et al. 2001; Pulvermüller 2005; Aziz-Zadeh and Damasio 2008; Barsalou 2008, 2010; Pulvermüller and Fadiga 2010). It has been suggested that metaphorical mean ings may recruit those brain areas/networks that are involved in the particular source and target domains of the metaphor (Coulson 2008). For example, action metaphors would activate brain areas associated with the comprehension of action and spatial metaphors would activate brain areas associated with spatial cognition (for details see Coulson 2008). There is currently no consensus from the few neuroimaging studies investigating the relation between figurative language and action systems. Some neuroimaging studies report activation of motor brain regions by figurative action language (Boulenger et al. 2009; Desai et al. 2011) while others do not (Aziz-Zadeh et al. 2006; Raposo et al. 2009). One study associated secondary motor regions with metaphoric action sentences (regard less of the level of metaphor familiarity) while primary motor regions were negatively correlated with metaphor familiarity (Desai et al. 2011). This was interpreted as indicating that unfamiliar action metaphors are understood by relatively detailed sensory-motor simulations and that these simulations become less detailed as familiarity increases. It was further suggested that metaphor understanding is not only based on sensory-motor systems but also on abstract systems as metaphoric action sentences and sentences con taining abstract verbs activated left temporal brain regions which, according to Desai et al. (2011), may be associated with the computation of abstract meaning.
19.5.2 Other figurative and non-literal language comprehension 19.5.2.1 Idioms Besides metaphors, idioms are another form of figurative language frequently studied in the neuroscience of language. Idioms are formulaic expressions that are character ized by their stereotyped form, conventionalized meaning, and contextual conditions. Compared to novel expressions (lexical items assembled by grammatical rules), idioms are processed and stored, comprehended and produced differently (van Lancker Sidtis 2012). Similar to metaphors, idioms present as a heterogeneous category with individ ual characteristics that may influence how they are stored and processed. For example, decomposable idioms (‘button your lips’) seem to be processed differently compared to non-decomposable idioms (‘kick the bucket’) which may be stored as a single unit in the mental lexicon (for a discussion on characteristics of idioms see Gibbs 1993). While idioms share some features with metaphor (e.g. figurativeness), they also have their own individual characteristics (e.g. parsing operations, analysability of their parts) (Hillert 2008; see also the collection of articles in Cacciari and Tabossi 1993). As with metaphors, it has traditionally been assumed that idiom comprehension relies on the right hemisphere (van Lancker and Kempler 1987; Kempler et al. 1999) and,
375
Neuropragmatics 375 as with metaphors, this view was challenged when patients who suffered a left-sided lesion (such as aphasic patients) showed problems with idiom comprehension and studies with right-hemisphere-damaged patients provided ambiguous findings (for reviews see Papagno and Lauro 2010; van Lancker Sidtis 2012). Reviewing neuropsy chological studies on idiom comprehension, Papagno and Lauro (2010) suggested that the left hemisphere is involved in the interpretation of single elements in the idiomatic expression based on the observation that idiom comprehension improved in aphasic patients if contextual cues were present while it was impaired in right-hemisphere- damaged patients with prefrontal lesions. A more detailed picture evolved from studies using the fMRI technique. A common network of cortical activation implicating the default language network for idiomatic or literal sentences was reported when com paring the literal and non-literal interpretation of idioms (Zempleni et al. 2007; Lauro et al. 2008). Non-literal interpretation of idioms generally elicited stronger and spa tially more extended activation in both hemispheres including temporal (right supe rior and middle temporal gyri, temporal pole) and frontal regions (left superior frontal gyrus, left and right inferior frontal gyrus) (Zempleni et al. 2007; Lauro et al. 2008; Mashal and Faust 2008). The magnitude of the activation was higher in areas of the left hemisphere (left inferior frontal gyrus) when participants processed the salient, idi omatic meanings of ambiguous idioms compared to their non-salient, literal meanings (Mashal and Faust 2008). Literal interpretations of idioms increased activation in right lateralized brain regions (precuneus, middle frontal gyrus, and posterior middle tem poral gyrus) (Mashal and Faust 2008). These findings were interpreted as support for the graded salience hypothesis with the right hemisphere being sensitive to non-salient linguistic interpretations and involved in ambiguity resolution as shown in faster access of the literal than idiomatic meanings of the idiomatic expressions. The left hemisphere was viewed as supporting salient meanings (the idiomatic meaning of idioms and the literal interpretation of literal sentences). Right anterior temporal lobe activation was associated with integration processes, right-middle temporal gyrus with attention, and right precuneus with the use of mental imagery (Mashal and Faust 2008). The question whether metaphor and idiom processing share brain correlates or whether there are brain regions specific to metaphor or idiom processing has not been addressed directly. In most general terms, what has evolved is that metaphor as well as idiom processing relies on both hemispheres, thus refuting the standard pragmatic model and the classical right-hemisphere hypothesis. Similar to metaphor process ing, idiom processing recruits the default language network and seems to extend this network spatially and in magnitude rather than recruiting new brain networks. There are, however, still too few studies and replications of studies to make generalizations concerning the specific role of the hemispheres and the specific brain correlates and networks activated in idiom processing.
19.5.2.2 Irony and Sarcasm Like other figurative language, irony has traditionally been conceptualized as lan guage in which what is meant is different from what is said (Grice 1975; Searle 1979a).
376
376 Brigitte Stemmer Alternative views are echoic theories of irony in which literal meaning echoes an ex pectation that has been violated (Sperber and Wilson 1995). In these accounts mean ing is attributed to an ironic utterance by alluding to explicit comments or implicit expectations. One form of irony is sarcasm which is frequently used to express im plicit criticism about a situation or a person. Lesion studies have shown that damage to the prefrontal lobe of the right hemisphere impairs the ability to interpret sarcasm to some degree (for a review see McDonald 2000; Shamay-Tsoory et al. 2005). Problems with picking up prosodic cues, inferring the emotional state or what is on other peo ple’s minds, have been discussed as contributing to these difficulties (McDonald 2000; Shamay-Tsoory et al. 2005). As with other neuroimaging studies on figurative language, both hemispheres con tribute to irony and sarcasm processing but the exact role of the right hemisphere still needs clarification. Concerning the contribution of brain correlates within the hemispheres, the findings of neuroimaging studies are difficult to reconcile. Although most studies report activation of the medial prefrontal cortex (Uchiyama et al. 2006; Wakusawa et al. 2007; Shibata et al. 2010; for different results, see Eviatar and Just 2006), the regions within the medial prefrontal cortex do not always coincide. For ex ample, Wakusawa et al. (2007) reported activation of the orbitofrontal part, Uchiyama et al. (2006), more dorsally situated regions, and Shibata et al. (2010) rostral regions. Interpretation of these activations is equally diverse (reflection of complex cognitive processes by Shibata et al. 2010; theory of mind mechanisms by Wakusawa et al. 2007; inductive reasoning in the context of mentalizing by Uchiyama et al. 2006). Activation of temporal regions is reported by three out of four studies, without, however, con verging on similar areas within these regions (superior and middle temporal gyrus by Eviatar and Just 2006, the right temporal pole by Wakusawa et al. 2007, and the left temporal pole by Uchiyama et al. 2006). Two of the studies investigated, at the same time, metaphors and ironic utterances (Eviatar and Just 2006; Wakusawa et al. 2007). While one study reported activation of similar areas but differences in magnitude of the activation for metaphor and irony processing (Eviatar and Just 2006), the other study (using conventional metaphors/ idiomatic phrases and proverbs) did not report such similarities. In general, it is dif ficult to reconcile or make any generalizations based on the findings of these studies. As with other figurative language, the brain correlates implicated depend on the type of task and procedure used, the cognitive and non-cognitive processes implicated, and the individual’s mental state and experience.
19.6 Summary, Conclusion, and Outlook Linguistic pragmatic behaviour operates on words and sentences and it is thus not sur prising that brain correlates involved in word and sentence processing are also activated in higher-level language (see section 19.2 on the default language network). Research
377
Neuropragmatics 377 on the brain correlates implicated in linguistic pragmatic behaviour has particularly focused on the left–right hemisphere dichotomy and the standard pragmatic model. Studies of structural discourse comprehension and figurative and non-literal language comprehension support the implication of both hemispheres although there is still no consensus on the specific contribution of each hemisphere. The default language network is implicated in structural discourse as well as in figura tive and non-literal language. In structural discourse comprehension (especially coher ence building) and situation model building it has been suggested that additional brain correlates are implicated such as the anterior temporal lobe reflecting semantic and episodic memory integration processes, the dorsomedial prefrontal cortex reflecting conscious strategic processes, and the anterior temporal lobe with the parietal cortex supporting the integration of text information (see section 19.4.1). Although several neuroimaging studies support the implication of the roles of the anterior temporal lobe and dorsomedial prefrontal cortex in structural discourse and text comprehension, viewing these findings as preliminary seems appropriate until more evidence has been gathered. From a chronological perspective, information from lower-and higher-level language conjointly with world knowledge and non-linguistic context information seem to converge immediately to support the interpretation of discourse. There is general agreement that the classical standard pragmatic model and the right- hemisphere hypothesis are not supported by findings from studies on figurative and non-literal language processing. Support for the coarse coding theory and the gradient salience hypothesis is ambiguous and awaits further clarification. From a chronological point of view, metaphoric meanings are activated very early in processing, suggesting that metaphor comprehension and literal language share qualitatively similar process ing mechanisms, thus speaking against a literal–figurative dichotomy. From a processing perspective, different types of figurative and non-literal language seem to share some common ground; the question, however, whether they also share common brain correlates or in which aspects they may differ remains currently unan swered. The attempt to clarify the role of the right hemisphere in figurative language comprehension has shown that numerous studies focusing on metaphors have associ ated the right hemisphere with the comprehension of novel metaphors and the left hemisphere with frozen, conventional, or familiar metaphors. Not all studies, however, support this view. It has further been suggested that right-hemisphere involvement in metaphor comprehension is due to task difficulty rather than specific metaphor fea tures. Whether this is also true for other figurative or non-literal concepts remains currently unanswered. Reaching consensus concerning specific brain regions associated with metaphor and other figurative language comprehension has, so far, proven difficult. It may very well be that—apart from the default language network—no common or ‘standard’ brain correlates can be identified as much may depend on how the individual variables inter act. In addition to the specific characteristics of metaphor or other figurative language concepts, the cognitive systems, mental state, and experience specific to the individ ual, the environment in which these concepts occur and the way these variables are
378
378 Brigitte Stemmer weighted may ultimately determine the brain correlates implicated in their processing. Instead of one specific brain activation network underlying figurative and non-literal language, it seems more likely that multiple interacting subsystems converge on impor tant hubs of integration. Cognitive and psycholinguistic studies have grounded figurative language in bodily experience and linked them to the sensory-motor system. Although some neuroimag ing studies provide support for this idea (especially when the concept contains some sort of action component), for metaphor comprehension there is some evidence that it relies on both sensory-motor and abstract systems. The theoretical bases for studies investigating the brain correlates implicated in lin guistic pragmatics have primarily been adopted from work in psycholinguistics, cogni tive psychology, and philosophy. Considering that we are interested in understanding how the brain deals with pragmatic issues, it seems sensible also to explore these issues from a brain perspective, that is, within a theoretical framework of overall brain func tioning. From such a perspective, body and mind are inextricably intertwined and our thoughts emerge through the interaction of body and mind with the environment. Internal, personal needs and desires are related to external, environmental reality by concepts that are created by neural networks of the brain. These neural networks are organized at multiple levels. The contribution that multiple levels of neural organiza tion make to language comprehension at the word and sentence level has been in vestigated by Tucker and colleagues (Tucker 2007; Tucker et al. 2008). These authors have adopted a microgenesis framework that argues that each mental process traverses levels of the neuraxis (nerve pathway) in the same direction as those structures were laid down in evolution (Brown 2002). It has been suggested that networks in the brain that underlie cognitive systems (including language) and non-cognitive systems (such as arousal, attention, emotion, motivation, motor systems) are organized along various dimensions: left–right (such as left-and right-hemispheric specialization); front–back (such as action in the frontal networks of the brain and perception in the posterior ones); and inside–out (such as from the inner visceral, limbic core of each hemisphere out to the somatic, sensorimotor shell that links the brain with the world) (for details, see Tucker 2007). Support for the microgenesis theory comes from a series of studies reviewed by Tucker et al. (2008) that show, for example, a progression from general semantic processing in medial, limbic networks to specific semantic processing in cor tical regions (i.e. the inside–out dimension). The consistent finding of temporal lobe activation in studies on structural discourse may reflect the link between visceral and somatic states (inside–out dimension) and the involvement of frontal and parietal re gions may reflect action and perception processing (front–back dimension). The claim of the coarse coding theory that each hemisphere contributes differently to semantics would be in line with a microgenetic framework (the left–right dimension) (Beeman 1998; Jung-Beeman 2005; Tucker et al. 2008), although it must be noted that support for this theory is still ambiguous (see sections 19.5.1 and 19.5.2). The view that figurative/ non-literal language is grounded in bodily experience would also be accommodated by microgenetic theory which advances that the organization of linguistic acts is embodied
379
Neuropragmatics 379 in the neural architecture of the brain. The proposal that pragmatics is an emergent phenomenon also fits the microgenetic view that actions and thoughts evolve through vertical integration of coordinated processing at multiple levels of brain structure. Currently, these thoughts must remain speculative as the applicability of a microge netic framework to neuropragmatics has not been systematically explored. This may, however, prove a fruitful endeavour in the advancement of neuropragmatics.
Acknowledgements Thanks goes to Harry A. Whitaker for his witty comments on an earlier version.
380
381
Pa rt I V
M AC RO - P R AG M AT IC S A N D S O C I E T Y /C U LT U R E
382
383
Chapter 20
P olitenes s a nd Imp olitene s s Penelope Brown
20.1 Introduction Politeness means many different things to different people. In lay terms it refers to socially correct or appropriate speech and behaviour. Its core sense is perhaps a matter of attention to interactional sensibilities, of speech and behaviour that attends to the feelings and expectations of those one is interacting with so that social interaction proceeds smoothly. Politeness is the oil that keeps the interactional hinges from creaking, and prevents offence by pre-emptively anticipating the possibilities for offence and offsetting them. Many folk notions capture different aspects of these kinds of attention to feelings, expressed in terms like manners, courtesy, tact, deference, sensibility, poise, rapport, urbanity, civility, graciousness, as well as terms for the contrasting behaviours—interactional offences such as rudeness, gaucheness, social gaffes, insults—and their consequences, from embarrassment or humiliation to conflict and even warfare. Such terms attest both to the pervasiveness of notions of politeness and to their cultural and situational framing. Politeness in its varying forms is an issue in every culture, and—at least latently—in every social interaction. The written history of attention to politeness goes way back, with for example prescriptive etiquette books dating back to ancient Egyptian times (Terkourafi 2011) and extending up into modern times with prescriptions in the popular media (e.g. the American Ann Landers and Dear Abby advice columns), and to Henri Bergson’s (1885) philosophical discourse on three kinds or senses of politeness: politeness of manners as ‘some art of testifying to each by his attitude and words, the esteem and consideration to which he is entitled’, politeness of mind as the intellectual flexibility involved in ‘the faculty of putting oneself in the place of others’, and politeness of the heart, as the
384
384 Penelope Brown ‘grace of the spirit’ exemplified in words and movements detached from usefulness and aiming only to please. Linguists too have puzzled over how to analyse routinized formulaic utterances, including stereotypical polite formulae, as part of a linguistic system (e.g. Ferguson 1976; Coulmas 1981, 1991), and since the 1970s, politeness has been a major focus for work in linguistic pragmatics. Clearly, such a wide range of notions covers much ground, from etiquette to morality. But what these notions have in common is this: generally, such attention to interactional expectations and feelings requires work, the production of some kind of evidence that one is attending to interlocutors’ concerns, and hence deviations from the direct efficient expression of Gricean communicative intentions (Grice 1975). This observation was the motivation behind the scientific study of politeness as a linguistic phenomenon, which began in the early 1970s with a paper by Robin Lakoff (1973) using a Gricean framework to understand linguistic politeness. A broader view of politeness considers it to be an intrinsic aspect of social interaction, crucial to the construction and maintenance of social relationships (Goffman 1967a,b; Gumperz 1982b; Goody 1995) and hence bearing on human cooperation and universals in human interaction (Carrithers 1992; Enfield and Levinson 2006; Tomasello 2008, 2009). In this perspective, politeness in communication goes right to the heart of social life and interaction; indeed it is probably a precondition for human cooperation in general. Language use is a crucial arena for expressing and negotiating such cooperation, and politeness is the feature of language use that most clearly reveals the nature of human sociality as expressed in speech. Politeness phenomena have therefore attracted interest in a wide range of social sciences, particularly linguistics, anthropology, social psychology, sociology, and communication. In this article I survey work in these disciplines that focuses on linguistic politeness. I first present—in section 20.2—the theories of the ‘founding parents’ of the study of linguistic politeness: the linguists Robin Lakoff and Geoffrey Leech and the linguistic anthropologists Penelope Brown and Stephen Levinson, all of whom attempted to capture general principles of politeness with some claim to universality. In section 20.3, I discuss a range of critiques of these universalist approaches and the corresponding alternative proposals. Section 20.4 surveys some empirical findings of research on politeness in different cultural settings, focusing mainly on work since 1990 and especially on recent attention to impoliteness and interactional conflict. The final section offers some conclusions and suggestions for future research.
20.2 Different Approaches to Defining and Analysing Linguistic Politeness The first steps in an analytical approach to politeness were taken in the 1970s and ’80s. Three distinct approaches can be identified.
385
Politeness and Impoliteness 385 (a) Politeness as social rules or norms. To the layman, politeness is a concept designating ‘proper’ social conduct, rules for speech and behaviour stemming generally from high-status individuals or groups. In literate societies such rules are often formulated in etiquette books. These ‘emic’ notions range from polite formulae like please and thank you, the forms of greetings and farewells, etc., to more elaborate routines for table manners, or the protocol for formal events. Politeness in this view is conventionally attached to certain linguistic forms and formulaic expressions, which may be explicitly taught to children and may be very different in different languages and cultures. This is how the ‘person on the street’ tends to think about politeness, as inhering in particular forms of words. Some analytical approaches to politeness are formulated in terms of the same sorts of culture-specific rules for doing what is socially acceptable: for example, the work by Ide and others on Japanese politeness as social indexing or ‘discernment’ (Ide 1989; see also Watts et al. 1992). In these approaches, politeness is a matter of social norms, and inheres in particular linguistic forms when used appropriately as markers of pre-given social categories. This approach is most appropriate for fixed aspects of language use— the more or less obligatory social marking of relatively unalterable social categories and social actions. The two other approaches were influenced by Generative Semantics, and share an interest in developing a general theory of the construction of polite utterances. (b) Politeness as adherence to Politeness Maxims. A second rule-based approach derives politeness as a set of social conventions coordinate with Grice’s Cooperative Principle for maximally efficient information transmission (‘Make your contribution such as required by the purposes of the conversation at the moment’), with its four ‘Maxims’ of Quality, Quantity, Relevance, and Manner (Grice 1975; see Huang this volume). Lakoff (1973) suggested that three ‘rules of rapport’ underlie the choice of linguistic expression, rules which can account for how speakers deviate from directly expressing meanings. Choice among these three pragmatic rules (‘Don’t impose’, ‘Give options’, ‘Be friendly’) gives rise to distinct communicative styles. Lakoff (1973) aims to identify general rules guiding interaction, arguing that ‘[p]oliteness is a system of interpersonal relations designed to facilitate interaction by minimizing the potential for conflict and confrontation inherent in all human interchange’ (Lakoff 1990: 34). Leech’s more detailed proposal (1983) is in the same vein. Complementary to Grice’s Cooperative Principle, Leech proposes a Politeness Principle—‘Minimize the expression of impolite beliefs,’ with the six Maxims of Tact, Generosity, Approbation, Modesty, Agreement, Sympathy. As with Grice’s Maxims, deviations from what is expected give rise to inferences. Cross-cultural differences, in Leech’s theory, derive from the different importance attached to particular maxims, which provide pragmatic scales that are ‘very widespread in human societies, but their interpretation differs from society to society, just as their encoding differs from language to language’ (Leech 2007: 200).
386
386 Penelope Brown The conversational maxim approach shares with the social norm approach the emphasis on codified social rules for minimizing friction between interactors and the view that deviations from expected levels or forms of politeness carry specific messages. (c) Politeness as strategic face management. A more sociological perspective places ‘face work’ at the core of politeness. The sociologist Erving Goffman (1967a,b) considered politeness as an aspect of interpersonal ritual, central to public order. He defined face as an individual’s publicly manifest self-esteem, and proposed that social members have two kinds of face requirements: positive face, or the want for approval from others, and negative face, or the want not to offend others. Attention to these face requirements is a matter of orientation to Goffman’s ‘diplomatic fiction of the virtual offense, or worst possible reading’ (Goffman 1971: 138ff.), the working assumption that face is always potentially at risk, so that any interactional act with a social–relational dimension is inherently face-threatening and needs to be modified by appropriate forms of politeness. Brown and Levinson (1978, 1987) drew on Goffman’s analysis but introduced a new, comparative perspective by drawing attention to the detailed parallels in the construction of polite utterances across widely differing languages and cultures, and argued that universal principles underlie the construction of polite utterances. The cross-linguistic parallels they noted are of two sorts: how the polite expression of utterances is modified in relation to social characteristics of the interlocutors and the situation, and how polite utterances are linguistically constructed. At least three social factors are involved in deciding how to be polite: (a) one tends to be more polite to social superiors; (b) one tends to be more polite to people one doesn’t know. In the first case, politeness tends to go one way upwards (the superior is less polite to an inferior); in the second, politeness tends to be symmetrically exchanged. (This regular pattern of language use was first described by R. Brown and Gilman (1960) for the ‘T/V’ (tu/vous) pronouns of ‘power’ and ‘solidarity’.) In addition, (c) in any culture there are norms and values affecting the degree of imposition or unwelcomeness of an utterance, and one tends to be more polite for more serious impositions. In language there are also detailed parallels, with the linguistic structures for realizing particular kinds of politeness displaying remarkable similarities across languages. The politeness of solidarity is characterized, for example, by the use of intensifiers, in- group identity markers and address forms, exaggerated intonation patterns, and forms for seeking or emphasizing agreement and avoiding disagreement. Avoidance-based politeness is characterized by self-effacement, formality, restraint, deference, with the use of honorifics, hedges, indirect speech acts, and impersonalizing mechanisms like pluralization of pronouns, nominalization, and passive constructions. The question motivating Brown and Levinson’s approach to politeness is this: Why are these kinds of detailed parallels across languages and cultures to be found in the minutiae of linguistic expression in socially analogous contexts? Explanations in terms of social norms or rules can account for politeness in a particular social group, but not the cross-cultural patterns, which seem to require a strategic account in terms of
387
Politeness and Impoliteness 387 what people generally are trying to do when they are being polite. Brown and Levinson (1978, 1987) proposed an abstract model of politeness wherein human actors are endowed with two essential attributes: face and rationality. Face consists of two specific kinds of wants: positive face (i.e. the desire to be approved of, admired, liked, validated), and negative face (the desire to be unimposed upon, unimpeded in one’s actions). The second ingredient in the model—rationality—provides for the ability to reason from communicative goals to linguistic means that would achieve these goals. From these two assumptions—face and rationality—and the assumption that speakers mutually know that all interlocutors have these attributes, Brown and Levinson developed a model of how speakers construct polite utterances in different contexts on the basis of assessments of three social factors: the relative power (P) of speaker and addressee in the context, their social distance (D), and the intrinsic ranking (R) of the face-threateningness of an imposition. P, D, and R are seen as abstract social dimensions indexing kinds of social relationship (P and D) and cultural values and definitions of impositions or threats to face (R). Brown and Levinson distinguish five general types of strategies of politeness, ranging from avoiding a face-threatening act (FTA) altogether, to carrying it out but ‘off-record’ (indirectly). On-record realization of an FTA can be done without any redressive action at all (‘baldly’). It may be carried out with positive redress, which is essentially approach-based, addressing the hearer’s positive face wants by emphasizing closeness and solidarity. Politeness may also be carried out with negative redress, which is essentially avoidance-based, addressing negative face wants for distance, deference, freedom from unexpectable impositions. Speakers are assumed to choose the linguistic framing of their utterance from this set of strategic possibilities according to the weightiness of the FTA, which is assessed with reference to the three contextually dependent social factors P, D, and R. For low levels of FTA threat, bald on-record or positive politeness is most appropriate and cost-effective; for higher levels, negative politeness is required; for the highest threats, indirectness is the safe option. In short, the argument is that there are universal dimensions to cultural values and social structures, which can be abstracted from the variety of individual societies and compared—that underlying the variety, in all societies people recognize degrees of social distance, degrees of (vertical) social hierarchy, and degrees of impositions which can be made to their universally recognized desires to maintain ‘face’, and that universal pragmatic principles produce cross-linguistic parallels in the ways in which people linguistically encode their speech acts in different contexts. Cross-cultural variability in politeness is attributable to facts of social structure, cultural meaning, and cultural value (how hierarchical/egalitarian is the society, how much value do people place on respect and social distance vs brotherhood and conviviality, what kinds of social relationships do they have, how do they define situations and activities and what kinds do they find especially threatening, etc.), but across diverse societies, the same principles are at work producing analogous ways of putting things— communicative styles—in relatively analogous situations (e.g. intimate vs formal). However culturally and situationally variable the kinds of social relationship and
388
388 Penelope Brown kinds of face threat might be, underlying them are pan-cultural social dimensions (relative power, social distance, ranking of face threateningness) which universally go into the reckoning—and the interpretation—of strategic language choice, and hence one can derive the cross-cultural similarities in choice of linguistic realizations of politeness strategies that empirically seem to be in evidence. Brown and Levinson claimed further that stable social relationships are characterized in part by stable patterns of language use, which may distinguish particular societies or social groups, and that therefore their model of politeness universals could be applied in particular cultural settings as an ethnographic tool for analysing the quality of social relationships in any society. This model of politeness provides a set of analytical tools for studying linguistic politeness, and makes detailed predictions about what one will—and won’t—find when looking at patterns of politeness in different cultures. It stimulated a flurry of research on politeness, as well as a barrage of critiques from many directions, to which we now turn.
20.3 Critiques of Politeness Theory The goal of Brown and Levinson was ambitious: to formulate a universally applicable ‘etic’ set of concepts in terms of which politeness can be analysed in ‘emic’ terms for any particular society. Etic analytical concepts are drawn from a universal set, defined from outside a particular culture and used to compare behavioural or linguistic systems across different cultural groups; emic ones are within-culture meaningful ones used to describe a system in its own terms. The etic/emic distinction is derived from the linguists’ distinction between phonetic and phonemic analyses, and is often conflated with cross-cultural vs within-cultural analysis (Jahoda 1995). Critiques of the Brown and Levinson model reveal several major points of contention about what politeness is, what an emic vs etic kind of analysis is, whether it is legitimate to generalize across cultural systems, and how a theory of politeness should be formulated. I review these in turn. The universality of face and of politeness scales. Many critics have challenged Brown and Levinson’s formulation (via Goffman and Durkheim) of positive and negative face wants, as a valid way of conceptualizing the universal underpinnings of politeness. Negative face, in particular, considered as wants for freedom from imposition, appears entirely too embedded in Western individualism to sit well with conceptions of face in some other (e.g. East Asian) cultures. In part, this objection is due to a misconstrual: face wants in the Brown and Levinson model are abstract; they do not necessarily correspond clearly to conscious emic notions in a particular cultural setting. What Brown and Levinson claimed is that underlying very diverse folk notions is a cross- culturally applicable core of two interactionally relevant wants (for ratification and for freedom from impositions), desires concerning one’s public self-image in the context of
389
Politeness and Impoliteness 389 the moment of articulating a communicative intention which are assumptions oriented to in interaction. Other theorists (e.g. O’Driscoll 1996; Arundale 1999) have argued for notions of positive/negative face that are even more abstract, in terms of merging/individuation or closeness/separation, as the universal heart of politeness. In this highly abstract sense, politeness has universal aspects, and these universals are consistent with the many different cultural ideologies concerning polite speech that can be found in different social groups. Challenges to the universality of the Brown and Levinson model also extend to the proposed hierarchy of increasing politeness (from bald on-record to positive to negative to indirectness) with increasing threat to face. Assessments of the P, D, R factors are both situationally and culturally very variable, it is possible to accumulate different strategies in one utterance and to balance elements of negative politeness with positive politeness in one communicative act, and indirectness is not always seen as the most polite option; indeed, there are many uses of indirectness that can be construed as impolite (Blum-Kulka 1987, 1989). These observations have led some researchers to argue against the possibility of identifying any kind of universal basis for polite behaviour; politeness is simply incommensurate across societies. Those who take this extreme relativistic line can have no explanation for the observable cross-cultural parallels in patterns of language use, for how people manage (sometimes) to understand others from culturally different backgrounds, or for cross-linguistic parallels in the diachronic sources of particular linguistic features—for example, honorifics—from politeness strategies. Politeness as communicated or taken for granted. In contrast with rule-and norm- based approaches, Brown and Levinson insist that politeness inheres not in words or in sentences per se, but in utterances uttered in a context, by virtue of the successful communication of a polite attitude or intention at that moment. Politeness is an implicature, an inference that may be conveyed by utterances spoken in context, by virtue of successful communication of a polite attitude or intention. Polite utterances are not necessarily communicating ‘real’ feelings about an interlocutor’s social persona, but expressing contextually expected concern for face. Politeness is ascribed to a speech act, or to an interactional move (if you prefer), not to a strategy or its linguistic realization per se. Even the use of apparently inherently polite formulae—like ‘please’, ‘thank you’, or honorifics and other ‘markers’ of social status—does not guarantee politeness in any given utterance, as they may be overridden by intonation or other contextual cues suggesting a lack of sincere polite attitude or intention. Politeness cannot be automatically ‘read off ’ of a linguistic form—it is the use of the form in a specific context, along with the associated prosodic and kinesic features, that together produce a ‘meaning-in-context’ that is the basis of the assessment of the utterance as polite or not (Brown 1995). In other approaches to the analysis of polite phenomena (for example, Fraser’s (1990) ‘conversational contract’, Watts et al.’s (1992) ‘politic behavior’), politeness is taken to be the expected background to interaction; it is normally not communicated but consists in following expectations as to appropriate behaviour.
390
390 Penelope Brown Broad vs narrow scopes for politeness theory. Another bone of contention is the scope of phenomena considered under the rubric of ‘politeness’. The narrower view taken by Brown and Levinson takes politeness to be strategic orientation to potential face threats. Many motivations other than politeness guide human behaviour: there are some situations (e.g. task-oriented ones, highly urgent ones, confrontational ones) where politeness may be subsumed to other goals, and there are many reasons for being indirect in speech other than politeness (e.g. humour, irony, rhetorical force). A more inclusive view, favoured by many scholars in the fields of social psychology and communication, sees politeness as orientation to the social- relationship dimension of every interaction, with attention to face taken to be an omnipresent necessity. The whole continuum, from extreme politeness through a quite neutral level of politeness (maintaining the status quo, ‘discernment’) to rudeness to face attack (outright intentional face threat), then needs to be brought into the theory. Politeness from the point of view of the individual, the dyad, or the social group. The Brown and Levinson model takes the interacting dyad as its unit; it models how interlocutors pre-emptively foresee the possibilities for offending each other and adjust their utterances to display this attention, how they make inferences of politeness from one another’s strategic deviations from Gricean efficient communication, and how stable patterns of strategies that characterize particular dyadic interactions provide an index to the quality of the social relationship. Many of the Brown and Levinson politeness strategies are quintessential examples of ‘intersubjective perspective-taking’— putting oneself in the others’ position, which presumes a dyad as the minimal unit of analysis. Indeed, a major goal of Brown and Levinson was to insist on the centrality of social interaction as a significant level of social life, intermediate between the individual and society, where social/cultural facts (status, role, values, norms, rights, and obligations) are integrated with individual ones (goals, plans, strategies, communicative intentions). Yet the Gricean foundation of the theory and the speech-act-based formulation of the strategies have made many critics see the Brown and Levinson model as purely psychological (how a speaker calculates how to frame an utterance). Arundale (1999, 2006), for example, argues for a theory of how face is jointly constituted in ongoing interaction: it is not a property of individuals. In fact, I believe we need both perspectives: face is indisputably interactionally created and manipulated. Nevertheless, it can be considered from the point of view of the individual speaker or hearer (as Brown and Levinson do in their production/comprehension model), or of the society or social group (as Arundale’s ‘face constituting theory’ and most sociolinguists do). A more recent attempt at a broader theory of interpersonal communication incorporating politeness is the ‘rapport management theory’ of Spencer-Oatey (2000, 2005, 2007, 2008). In both of these approaches, relations, not the individual or the dyad, are the focus, and both are more appropriate to the study of politeness in a specific cultural setting (Watts’ 2003 ‘pol1’), rather than to assessing cross-cultural patterns and parallels in interaction.
391
Politeness and Impoliteness 391 A major reason for misconstruing the Brown and Levinson model as a psychological model of the individual is that, as formulated, it is too reliant on analysis at the utterance level, which places limitations on how applicable it is to analysing naturally occurring empirical data, as opposed to peoples’ judgements of relative politeness. To understand politeness phenomena in any situated context, we need to look at how politeness strategies evolve and are co-constructed through interactional time (Kasper 1990; Haugh 2007). The indeterminacy problem. A further shortcoming of the Brown and Levinson model as an ethnographic tool for analysing the quality of social relationships in any society is that its usefulness in analysing naturally occurring interactions is undermined by the contextual dependency of ratings of the social factors. The problems of the mixing and ordering of strategies in ongoing discourse are compounded by the indeterminacy of context-dependent P, D, R assessments, which make it hard to reliably code levels of politeness in any concrete situation. This is a problem with any theory in terms of actors’ intentions when applied to empirical data; as both interactors and conversation analysts know, it is not always possible to be certain what interlocutors’ intentions are at a particular point in natural interaction. First-order politeness (Pol1) vs second-order politeness (Pol2) and the nature of the individual/society relation. A final critique is more fundamental; this is the postmodernist critique derived from Bourdieu (1977), which rejects entirely the enterprise undertaken by Brown and Levinson. This is cast by Eelen (2001), perhaps its most coherent advocate in the politeness literature, as follows: the ‘Parsonian’ models of linguistic politeness (a set to which the Brown and Levinson model is said to belong) share a flawed world view about the nature of language and social reality. They collapse emic/etic concepts and treat politeness as a characteristic of behaviour, related to social norms that are independent of speakers and hearers, with an implicit view of culture as an ‘ideal consensus’ and with politeness as the regulatory force geared to maintaining social order. Eelen prefers Bourdieu’s approach through ‘habitas’, with its three crucial properties of argumentativity, historicity, and discursiveness, which takes politeness to be a ‘moral argumentative social tool’, essentially contested (like ‘beauty’ or ‘democracy’). This perspective, says Eelen (2001: 247) ‘provides a more dynamic, bi-directional view of the social-individual relationship’. Watts (2003) takes a comparable line, distinguishing pol1 (a common-sense layman’s view of normative politeness) from pol2 (a scientific, comparative view of politeness principles), and preferring to look at pol1 (based on perceptions of the participants involved), in contrast to pol2 (based on researchers’ interpretations and on pragmatic theory). This distinction tends to go hand in hand with a focus on the interpretation and assessment of utterances (an evaluative hearer’s perspective, hence necessarily pol1) rather than on how polite utterances are constructed by speakers. This argument, like all postmodernist posturing, draws a line in the sand: study a phenomenon my way or not at all. Comparative work is out of bounds, as there are no
392
392 Penelope Brown etic concepts applicable at the emic (culturally meaningful) level; all cultural phenomena are incommensurate. Recent examples of this ‘discursive’ approach can be found in Mills (2003), Locher (2004, 2006), Locher and Watts (2005), and in a collection (2011) edited by the Linguistic Politeness Research Group. In short, a major source of disagreement in research on politeness is whether or not it is legitimate to investigate cross-linguistic, cross-cultural commonalities in how people construct utterances in ways that display attention to their interlocutors’ social persona or ‘face’. Indeed, overviews of the Brown and Levinson programme from a ‘discursive’ perspective (e.g. Locher 2012) often ignore its main motivation: the universal patterns of distribution of politeness strategies across disparate languages and cultures. But this is the crucial claim of the Brown and Levinson theory—that there are analogous patterns of language use across contexts in widely differing cultural settings—and it has not (to my knowledge) been addressed in any alternative theory. The one kind of evidence which would really disprove the Brown/Levinson framework is if someone were to find a culture/language where the proposed parallels didn’t hold—for example, where small face-threats (as culturally defined) were handled with a lot of mitigating face-attention and large ones with less (other things being held constant), or where high-status people and strangers (as culturally defined) got imposed upon with less face redress than low-status people and intimates (other things held constant). No one has shown anything like this. Even if we retain an interest in accounting for such universal aspects of politeness phenomena, the Brown and Levinson model of politeness as originally formulated clearly needs revision. The intellectual climate of research has changed radically in the four decades since the early 1970s, when the model was fostered at the University of California, Berkeley by the confluence of scholars there at the time—including the philosophers of language Paul Grice and John Searle, the linguistic anthropologists John Gumperz, Brent Berlin, and Paul Kay, the linguists George and Robin Lakoff and Charles Fillmore, and the psychologist Dan Slobin—all of whom were constructing innovative theories of language usage. The approach of Brown and Levinson, as anthropologists, was both empirically founded and comparative, based on language usage data drawn from recordings of naturally occurring interactions in two field sites—the Mayan community of Tenejapa in southern Mexico and Tamilnadu in South India— supplemented by American and British English recordings of interaction and by much published work on usage in other languages. The cross-linguistic, cross-cultural parallels are patently observable; the problem is to account for them in a way consonant with modern linguistic and anthropological theorizing. One promising attempt is Claudia Strauss’s proposal (2004) to incorporate a model of ‘cultural stance’-taking into a politeness model, to accommodate the observation that assessments of politeness are negotiable. Strauss’s claim is that a speaker expressing an opinion on a topic should indicate the cultural standing of that view in the relevant ‘opinion community’ (Strauss 2004: 161). She argues that Brown and Levinson’s politeness model and her cultural-standing model are complementary, in that neither is complete without the
393
Politeness and Impoliteness 393 other: ‘Cultural standing considerations affect speakers’ judgments about what would be considered a possible FTA in the expression of opinions, and negative and positive politeness strategies for mitigating FTAs, while politeness considerations help explain why cultural standing is marked in discourse’ (Strauss 2004: 166). Turning now from theoretical debates to empirical research, I will briefly summarize some of the major themes in work on politeness, broadly construed, over the past twenty-five years.
20.4 Empirical Results of Work on Politeness Empirical studies of politeness phenomena have greatly increased the amount of information we have about language use and social interactional styles in different contexts and different societies. Many books and edited collections are devoted to describing politeness phenomena in particular settings (e.g. Coulmas 1981, 1991, Katriel 1986, Blum-Kulka et al. 1989, Bayraktaroglu 1991, Sifianou 1992, Watts, Ide, and Ehlich 1992, and more recently, Bayraktaroglu and Sifianou 2001, Hickey and Stewart 2005, Lakoff and Ide 2005, Placencia 2006, Bargiela-Chiappini and Kádár 2011, Fernández-Amaya et al. 2012, and Kádár and Haugh 2013). There was a special issue devoted to politeness of the International Journal of the Sociology of Language edited by Walters (1981), and the journal Multilingua has featured many special issues on politeness, with the most recent one on honorific usage edited by Pizziconi in 2013. There have also been at least three special issues of the Journal of Pragmatics (vol. 14, 1990; vol. 21, 1994; and vol. 28, 1997). More recent are cross-cultural comparisons of impoliteness, as in the special issue of the journal Intercultural Pragmatics (Garcés-Conejos Blitvich 2010c). Midstream in this era, a large bibliography of politeness work in linguistic pragmatics appeared in the Journal of Pragmatics (Dufon, Kasper, Takahashi, and Yoshinaga 1994). Since 2005, the study of politeness has had its own journal, The Journal of Politeness Research, published by Mouton; although intended to be an interdisciplinary forum (Christie 2005) this is primarily a forum for advocates of discursive approaches to politeness. Despite this burgeoning literature, there are unevennesses and some glaring gaps in the data that has appeared. There is a large and rich literature on politeness in Japan. Indeed, Japanese linguists (e.g. Matsumoto 1988, 1989; Ide 1989), partly inspired by the Brown and Levinson model and partly reacting to its perceived shortcomings, have since been very active participants in the ongoing development of politeness theory, making major contributions with their emphasis on cultural differences in assumptions about face and propriety in speech and offering alternative views of politeness in Japanese, as well as detailed studies of particular aspects of Japanese language usage
394
394 Penelope Brown (e.g. honorifics, Okamoto 1997, 1999; conversational turn-taking, Tanaka 1999, 2000, Hayashi and Mori 1998; face, reputation, and self-esteem expressed in conversation, Hayano 2013; see also Nwoye 1989, Wetzel 1994, 2004). A Japanese translation of Brown and Levinson 1987 was published in 2012. In contrast, there is almost no politeness-related research at all from certain parts of the world—e.g. New Guinea, or Aboriginal Australia (one exception is Wilkins 1986). There is also very little from Oceania (exceptions are Duranti 1992; Keating 1998) or South Asia (except Bickel et al. 1999), from Africa (exceptions include Irvine 1974, 1985; Nwoye 1992), from South America (except for Hardman 1981; Wolfowitz 1991; Bolin 2006), or from Central or North American Amerindia (exceptions include Scollon and Scollon 1983, 1995; Haviland 1987, 2005, 2010; Rhodes 1988; Brody 1991; Cowell 2007; Basso 2007; Reynolds 2008). Another limitation derives from the fact that researchers from the different disciplines studying politeness phenomena (sociolinguists, social psychologists, linguistic pragmaticists, linguistic anthropologists, conversational analysts) are often quite unaware of one another’s work.
20.4.1 Research topics in politeness research Here it is possible just to sketch the range of phenomena and cite some exemplary studies. Topics investigated include the following: Speech acts. An ongoing interest has been in how particular kinds of potentially threatening speech acts—requests, offers, compliments, thanks, apologies, disagreements, criticisms, complaints, etc.—are formulated in different cultural settings, and how strategies for expressing them are shifted in relation to contextual variables (e.g. Blum-Kulka, House, and Kasper 1989; Gass and Neu 1996; Fukushima 2000). Some more recent contributions can be found in Byon (2006) and Fernández-Amaya et al. (2012). Cross-cultural pragmatics and misunderstandings. Closely connected to the cross- cultural study of particular speech acts is the huge literature on miscommunication due to lack of shared understandings of how particular speech acts are culturally expressed. See for example Gumperz, Jupp, and Roberts (1987), Blum-Kulka et al. (1989), Gumperz and Roberts (1991), Kasper and Blum-Kulka (1993a), Scollon and Scollon (1994), and Gumperz (2001) (see also the references in Kotthoff and Spencer-Oatey 2007). The cultural construal of face. Views of face and politeness in different cultures and contexts have regularly appeared, along with proposals for reconceptualizing face as a core element in theories of politeness (e.g. Katriel 1986; Matsumoto 1988, 1989; Chen 1990–1; Gu 1990; Nwoye 1992; Mao 1994; Ervin-Tripp, Nakamura, and Guo 1995). These empirical descriptions contribute to the ongoing debates about the nature of face and its role in social interaction (e.g. O’Driscoll 1996, 2011; Bargiela-Chiappini 2003; Terkourafi 2007; Sifianou 2011).
395
Politeness and Impoliteness 395 Style. Politeness strategies as underlying the stylistic coherence of particular types of interaction have been an ongoing research focus, for example, in studies of how speakers convey affiliation with social categories such as gender, age, or ethnicity (e.g. Brown 1980, 1990; Tannen 1981; Brody 1991; Wolfowitz 1991; Rundquist 1992; Ochs 1992; Holmes 1995; Mills 2003). Honorific usage as a style is another focus (Irvine 1985, 1998). Recent related work can be found in the ‘genre’ approach to politeness and impoliteness of Garcés-Conejos Blitvich (2010a,b). Politeness in child language. How children are inducted into appropriate speech styles has long been a preoccupation of sociolinguists and scholars in the field of language socialization (e.g. Mitchell-Kernan and Ervin-Tripp 1977; Gordon and Ervin-Tripp 1984; Ervin-Tripp, Guo, and Lampert 1990; Snow, Perlmann, Gleason, and Hooshyar 1990; Bolin 2006; Hamo and Blum-Kulka (2007); Kampf and BlumKulka (2007); Reynolds 2008; Burdelski 2010, 2013a,b; Ehrlich and Blum-Kulka 2010; Burdelski (2011); Schöll 2011; Cekaite, Blum-Kulka, Grǿver, and Teubal 2014 provides a recent overview of this literature. For its application to autistic children, often said to have deficits in social cognition that might impact on their ability to produce and understand indirectness (and by implication, politeness), see Sirota (2004). The sequential development of face- oriented actions in conversational interaction. There is an increasing amount of work on interactional practices with politeness implications from a conversation-analytic perspective (e.g. Bayraktaroglu 1991; Lerner 1996; Okamoto 1999; Hayashi and Mori 1998; Heritage and Raymond 2005; Heinemann 2006; Curl and Drew 2008; Heritage 2013; Hayano 2013). Some of this pursues politeness-related themes in interactions in special contexts: for example, doctor– patient interaction (Aronsson 1991), email interactions (Haugh 2010), and online news (Neurauter-Kessels 2011). The social psychology of face management. Social psychologists have long been interested in interpersonal perception and in the expression of selfhood and identity in interaction, topics with implications for politeness (e.g. Tracy 1990; Penman 1990; Holtgraves 1992, 2005; Ting-Toomey 1994; Wetzel 1994; Tracy and Tracy 1998; see also Spencer-Oatey 2007; Garcés-Conejos Blitvich 2013). Anthropologists have looked into related themes in examining politeness and the expression of identity (see, for example, Brody 1991; Bucholtz and Hall 2005; Hall and Bucholtz 2010). Politeness as a functional motivation for linguistic structure. The structure of politeness formulae and honorific systems, and the historical development of these and of other linguistic structures with politeness implications, has been explored intermittently by linguists and linguistic anthropologists (e.g. Ferguson 1976; Wilkins 1986; Agha 1994; Bickel, Bisang, and Yadava 1999). Politeness theory and culture as ‘rhetoric’. Politeness theory has been adapted by anthropologists and applied to the analysis of formal ritual and to a view of culture as founded in rhetoric (Strecker 1988, 2010; Strecker and Lydell 2006; Tyler and Strecker 2009; Gudeman 2009; Carrithers 2009; Meyer and Girke 2011). See also the related
396
396 Penelope Brown arguments from an evolutionary perspective, for example Carrithers (1992) and Dunbar (1998). Politeness in Shakespearean interactions. Perhaps not unrelated are occasional efforts to analyse interactions portrayed in literature from the perspective of politeness theory. Brown and Gilman (1989) and Rudanko (2002) are two notable examples. Impoliteness and interactional conflict. Research on conflict talk has been a concern of sociolinguists and anthropological linguists since the 1970s—see, for example, the studies of Afro-American ritual insults (Mitchell-Kernan 1971; Labov 1973), Goodwin’s studies (1991) of black girls’ street talk and of girls’ recess talk (2006), Haviland’s studies of Tzotzil Mayan and Mexican Spanish insulting talk (1997, 2005, 2010, 2011), and Brown’s on Tzeltal Mayan confrontational talk (1990). A recent special issue of the Journal of Linguistic Anthropology (2010, 2001) which focused on Performing Disputes: Cooperation and Conflict in Argumentative Language, demonstrates that this is an ongoing concern. The occasional linguist has also turned his attention to conflict talk (e.g. Lachenicht 1980). But within the past twenty years, with the recognition that sometimes interactors actually want to be aggressive or insulting, examination of conflict talk and other forms of ‘impolite’ interaction has proliferated across various disciplines, along with attempts to integrate aspects of impoliteness into theories of politeness and of interactional cooperation. Among these are Blum-Kulka (1987); Culpeper (1996, 2011); Culpeper, Bousfield, and Wichmann (2003); Locher (2004, 2012); Bousfield (2008); Bousfield and Locher (2008); Hutchby (2008); Stewart (2008); Tracy (2008); Terkourafi (2008); Garcés-Conejos Blitvich (2009); Lascarides and Asher (2009); Bousfield and Grainger (2010). In this work, the distinction between the omission of expectable levels of polite redress (i.e. failing to be as polite as required by the situation, hence ‘rude’) vs commission of overt attacks on face (insults, threats, verbal aggression not contextualized as joking, where ‘impoliteness’ or ‘rudeness’ seems not to be the point) recalls the distinction we raised above between conventional norms of politeness and strategic attention to face. Failure to match up to the conventional expectations of politeness (omissions of redress) in a situation can be readily incorporated into politeness theories (e.g. Culpeper 1996). It is much harder to create a unified theory of social interaction that takes into account the entire spectrum from smooth cooperative interaction through the range of forms of ‘aggravated impoliteness’ (Rudanko 2002), that is, intentional aggressive attacks on an interlocutor’s face. Again, the difficult problem of establishing speakers’ intentions raises its head—much apparently aggressive and insulting behaviour occurs between friends, as a form of teasing (Collier 1991). One wonders whether the burst of interest in documenting and understanding interactional nastiness is related to the notable cultural shift in the popularity—at least in the Anglo-American world—of TV shows featuring overtly aggressive or
397
Politeness and Impoliteness 397 insulting interactions (e.g. Ann Robinson’s The Weakest Link, or American Idol). Certainly, these kinds of media performances make it possible to study overt interactional aggression in a readily available form. Yet acted-out aggression (the display of face-threatening actions for an audience) is not the same as naturally occurring face aggression, which is notoriously hard to capture on video and has scarcely been studied. In any case, it is now recognized that studying the various forms of impoliteness can shed light on, and make a necessary contribution to, our understanding of politeness phenomena.
20.5 Conclusion Politeness has attracted an enormous amount of research attention since the 1970s, and continues to be a major focus for research in disciplines concerned with social interaction. The study of politeness phenomena can provide insight into widely differing issues; foci of interest correspondingly differ widely, as reflected in the wide spread of phenomena examined in empirical research. There continue to be lively theoretical and methodological debates about the nature of politeness and of face, and about the kind of research approach appropriate for their study. Despite the large amount of empirical work, however, in my view research on politeness has been much weaker on the theoretical front. The fact that politeness research has been a major arena for ideological positioning of scholars on the ‘scientific’ vs ‘postmodern’ spectrum has not improved the level of theoretical debate. Another major limitation is the kinds of data used in analyses. A large proportion of studies take as their data people’s conscious evaluations of politeness expressed in sentences, judgements which tend to be both prescriptive and stereotypical. Many others use role-play, interviews, elicited conversations, or readily accessible sources of staged interactional data: TV shows, radio plays, films, YouTube videos. Far fewer studies use as data recordings of situated conversational exchanges to explore how politeness is achieved sequentially in naturally occurring discourse (as exemplified in conversation-analytic approaches), and only a tiny handful provide the crucial kind of evidence necessary to test the universality of any theory of politeness: for a particular society, an ‘ethnography of speaking’ providing evidence of naturally occurring language usage across different kinds of contexts to show how politeness is modulated in relation to local social factors (P, D, R, and others) in that society. The emphasis in research has been largely on cross-cultural differences, with insufficient attention addressed to the cross-linguistic/cross-cultural parallels which tend to be taken for granted when they are not disputed. Researchers, impressed by different cultural views of face, propriety, conviviality, the individual, and the self, have generally preferred to study culture-specific patterns of language use. Yet the significance of politeness lies far beyond the culture-specific rules of appropriate behaviour and
398
398 Penelope Brown speech that seem so salient to members of the culture. This wider significance lies in the fact that through their regular patterns of language choice humans interactively construct their social relationships. Work in this area therefore needs to be anchored in a theory of social interaction that takes into account both our common human nature—including our uniquely human capacities for cooperation and our ability to communicate cross-culturally—and the cultural differences which can sometimes lead us to misunderstand one another. In the new millennium, research agendas and priorities have changed. The study of social interaction is turning to focus on the interactional foundations of language, its cognitive underpinnings, and its instantiation in the brain. Research in the West continues on the sociocultural contexts of language use, with a major emphasis on face as co-constructed in interaction and on the sequential development of this interactional process, and with impoliteness as well as politeness as objects of enquiry. A corresponding emphasis on the sequentially developing contexts for face and related politeness issues (entitlement, ‘ownership’ of knowledge, ‘fairness’) appears in the field of conversation analysis; students of politeness have much to learn from a CA approach about how to demonstrate interactors’ orientations to issues of face in specific contexts (Haugh 2007; Hutchby 2008; see also Haugh 2015). There is also an increasing trend towards the comparative study of interaction from a conversation- analytic perspective (Sidnell 2009), as well as new theoretical approaches to the study of social relations (Agha 2007), both of which can be expected to feed into politeness theory. Another new direction is the emerging emphasis on the cooperative basis of human sociality and on the interactional underpinnings to language, its evolution, and its ethological base; see, for example, Goody (1995), Enfield and Levinson (2006), Tomasello (2008, 2009), Stivers et al. (2009), Stivers, Enfield, and Levinson (2010). There are recent investigations into the cognitive and brain underpinnings to interaction (e.g. Noordzij et al. 2009, 2010; De Ruiter et al. 2010). This work forms a new context within which the search for universal bases for politeness can be pursued, exploring the cognitive prerequisites, such as recursive theory of mind, without which the elaborate demonstration of mutual regard would not be possible. This context suggests that the inhibition of aggression, the soothing character of politeness work, and the elaborate interpersonal ritual that politeness represents, may have played not only a crucial role in the evolution of the specifically human forms of social life and elaborated culture, but may have also been a fundamental factor driving the evolution of human cognition. From this perspective the early theoretical focus on the maintenance of social harmony—rather than the more recent corrective emphasizing conflict and interactive aggression—seems motivated by an evolutionary perspective on how such constraints on language use could have evolved. The hugely broad sweep of work on politeness and impoliteness, coming from different academic disciplines, with different methods, different theoretical presuppositions and priorities, and vastly different research goals, has meant that often the work of one group is quite inaccessible to another. A truly interdisciplinary approach, with
399
Politeness and Impoliteness 399 researchers informed about each other’s disciplines and with the intellectual generosity to accommodate to each other’s presuppositions—as exemplified, for example, in the volume edited by Enfield and Levinson (2006)—would certainly improve the level of discourse, and of discovery, in what remains an area of critical importance in the study of human social interaction.
400
Chapter 21
Cross-C u lt u ra l a nd Interc u lt u ra l Pragm at i c s Istvan Kecskes
21.1 Introduction The terms ‘cross-cultural pragmatics’ and ‘intercultural pragmatics’ are often used interchangeably although they refer to two different disciplines. While cross-cultural pragmatics compares different cultures, based on the investigation of certain aspects of language use, such as speech acts, behaviour patterns, and language behaviour, intercultural pragmatics focuses on interactions among people from different cultures, speaking different languages. Cross-cultural pragmatics considers each language and culture separately and analyses the differences and similarities between various entities. Intercultural pragmatics focuses on the communicative process. It investigates the speech production and comprehension of interlocutors who represent different cultures and languages, and use a common language (lingua franca) for communication. Cross-cultural pragmatics represents the positivist research endeavours of the 1980s and 1990s with a motto of ‘when you are in Rome, do as the Romans do’. In order for someone to do that, s/he has to be familiar with the differences and similarities of language use and language behaviour in the given culture. This is why the major goal of the discipline has been to investigate and highlight aspects of language behaviour in which speakers coming from various cultures have differences and similarities. The fundamental tenet of cross-cultural pragmatics was best summarized by Wierzbicka in the following way: ‘In different societies and different communities, people speak differently; these differences in ways of speaking are profound and systematic, they reflect different cultural values, or at least different hierarchies of values; different ways of speaking, different communicative styles, can be explained and made sense of in terms of independently established different cultural values and cultural priorities’ (Wierzbicka 1991: 69).
401
Cross-Cultural and Intercultural Pragmatics 401 Intercultural pragmatics is a relatively new discipline which is the result of the constructivist endeavours of the 2000s. Intercultural pragmatics was defined by Kecskes (2004, 2011a, 2011b, 2013) as an inquiry that is about how the language system is put to use in social encounters between human beings who have different first languages, communicate in a common language, and usually, represent different cultures. The communicative process in these encounters is synergistic in the sense that it is a blend in which pragmatic norms of each participant are represented to some extent, and blended with the elements co-constructed by the interlocutors in the process of interaction. Intercultural pragmatics represents a sociocognitive perspective in which individual prior experience and actual situational experience are equally important in meaning construction and comprehension. Research in intercultural pragmatics has four main foci: (1) interaction between native speakers and non-native speakers of a language, (2) lingua franca communication in which none of the interlocutors has the same L1, (3) multilingual discourse, and (4) language use and development of individuals who speak more than one language. The new field of inquiry received a journal Intercultural Pragmatics in 2004, which, with its biannual conferences has become the main driving force of research in the discipline.
21.2 Merging, Crossing, and Blending Cultures Before discussing the major tenets of cross-cultural pragmatics and intercultural pragmatics we should look at current approaches to culture and explain what crossing cultures, merging cultures, and blending cultures means. Research in the last three decades has focused mainly on crossing and merging cultures according to the positivist (crossing) and constructivist (merging) approaches (Kecskes 2013). Blending, however, is a relatively new development, as we will see later. In the sociocognitive approach, language is defined as a system of signs resting upon a conceptual system that is relatively unique to each culture (Kecskes 2010b, 2013). Culture is seen as a socially constituted set of various kinds of knowledge structures that individuals turn to as relevant situations permit, enable, and usually encourage. It is a system of shared beliefs, norms, values, customs, behaviours, and artifacts that members of society use to cope with their world and with one another (Bates and Plog 1980: 6). Culture cannot be seen as something that is ‘carved’ in every member of a particular society or community. It can be made, changed, manipulated, and dropped in the course of interaction. (In fact, it is not culture that can be changed, manipulated, and dropped in talk but its manifestation.) Culture has fuzzy boundaries, and is characterized by dynamic changes both synchronically and diachronically. The conceptual system of individuals directly influenced by culture includes encyclopedic knowledge
402
402 Istvan Kecskes that refers to the knowledge of the world as distinguished from knowledge of the language system. This conceptual system plays a profound role in how human beings make sense in communication. In our globalized world most people speak more than one language and are affiliated with and/or affected by more than one culture. Crossing cultures is an essential part of our everyday life. Technology and people’s global mobility bring together representatives of cultures from all over the world. The frequency and variety of such encounters have changed our understanding of culture and communication. Gumperz (1982a) and Gumperz and Roberts (1991) called our attention to the fact that ‘culture’ is not present in communication in the ‘old’ sense of a transcendent identity which is composed of values and norms and linearly related to forms of behaviour. Cultural phenomena in speech are contingent, situational, and emergent in nature. This means that cultures are merged rather than crossed in the communicative process. Blommaert (1998: 4) claimed that what we can observe and analyse in intercultural communication are ‘different conventions of communication, different speech styles, narrative patterns, in short, the deployment of different communicative repertoires. As far as “identity” is concerned (cultural, ethnic identity), it can be an inference of these speech styles: people can identify selves or others on the basis of such speech styles. But in actual fact, not “culture” is deployed, but communicative repertoires.’ The main argument of research represented by Gumperz, Hymes, Blommaert, Rampton and others is that there is no single language, culture, or communicative style (see Kecskes 2013). What we have is language, culture, and communicative style instantiated in several group and individual varieties. In intercultural communication, speakers have a ‘repertoire’ of varieties of styles and a combination of styles which are deployed according to communicative needs in the changing context. The nationality or ethnic membership of people may suggest the possibility of ethnic or cultural marking in communicative behaviour. However, the interplay of several different factors affects the emergence of ‘ethnically’ or ‘culturally marked’ aspects of communicative behaviour which is most frequently dominated by other than just cultural factors. Frustration and anger, powerlessness, or a feeling of threat may trigger ethnic style (cf. Giles and Johnson 1986; Blommaert 1998, 2001). Rampton’s data (1995) showed that, depending on who is addressed, when, and in what particular type of activity (e.g. playing, discussing, listening to music), the role and function of ethnically marked communication styles may change. Based on their data, Blommaert and Rampton argued that ‘culture’ is situational in all its meanings and with all its affiliated concepts, and depends on the context in which concrete interactions occur. This argument has close connections with the claim of theorists of meaning who say that meaning is a situated, fluid, and online phenomenon (e.g. Evans 2009; Wittgenstein 2001). Thus, researchers in two different paradigms seem to have come to similar conclusions. Blommaert (1998), following the line of argument developed by Gumperz (1982a), Hymes (1996), and others, argued that culture is rarely unified, and new contexts generate new cultures and new forms of intercultural communication. Rampton’s research (1995) provided empirical substance for the
403
Cross-Cultural and Intercultural Pragmatics 403 old Sapirian claim that one society can hide many societies, one culture can hide many cultures, and one language can hide many others. This, however, should not mean that there is nothing relatively stable and unifying in culture (see Kecskes 2013). Existing ‘communicative repertoires’ have been developed from prior experience and communicative encounters. What the online creation of culture means is similar to what online meaning construction means: the bringing about of something needed online by blending the existing ‘repertoires’ and newly emergent elements. Culture, just like meaning, is characterized by both regularity and variety. It certainly is more than just an online created and co-constructed phenomenon. In communication, interlocutors can rely on two types of repository of prior experience and encounters: lexical units and communicative styles. Like lexical items, cultural patterns (often expressed in different communicative styles) code prior experience and encounters, i.e. relatively standard cultural behaviour models and expectations which are activated in the given actual situational context. In the course of interaction these existing models are modified and blended with situationally emergent new elements. This process of blending that relies both on existing and emerging factors constitutes the communicative encounter. Blending means smoothly joining these elements and/or factors into new intercultures. This is where cross-cultural pragmatics and intercultural pragmatics differ. While cross-cultural pragmatics is based on crossing cultures as described above, the driving force in intercultural pragmatics is blending as will be discussed below.
21.3 Cross-Cultural Pragmatics 21.3.1 History Cross-cultural pragmatics grew out of sociopragmatics which was created as a new line of inquiry within pragmatics after Leech (1983) and Thomas (1983) divided pragmatics into two components: pragmalinguistics and sociopragmatics. Pragmalinguistics refers to the resources for conveying communicative acts and relational or interpersonal meanings. These resources include pragmatic strategies such as directness and indirectness, routines, and a great variety of linguistic forms which can intensify or soften communicative acts. For example, compare these two versions of request: (1) Police officer to a driver: Can I see your driver’s licence? (2) Alessandro to his American friend, Bill: Would you mind showing me your driver’s licence? In both cases, the speaker chooses from among a great variety of available pragmalinguistic resources of the English language which can function as a request. However,
404
404 Istvan Kecskes each of these two expressions indexes a very different attitude and social relationship. This is why sociopragmatics is important in speech analysis. Leech (1983: 10) defined sociopragmatics as ‘the sociological interface of pragmatics’. He referred to the social perceptions underlying participants’ interpretation and performance of their communicative action. Speech communities differ in their assessment of speakers’ and hearers’ social distance and social power, their rights and obligations, and the degree of imposition involved in particular communicative acts (Kasper and Rose 2001). According to Thomas (1983), while pragmalinguistics is, in a sense, akin to grammar in that it consists of linguistic forms and their respective functions, sociopragmatics is about appropriate social behaviour. Speakers must be aware of the consequences of making pragmatic choices. Sociopragmatics was further developed in Gumperz’s, Tannen’s, and Scollon and Scollon’s works. Gumperz (1982a) founded interactional sociolinguistics with his work which demonstrated that systematically different ways of using language to create and interpret meaning contributed to employment discrimination against London residents who were from Pakistan, India, and the West Indies. Tannen’s (1985, 2005) focus is not just on language, but on how communication styles either facilitate or hinder personal interactions. For instance, according to her, men and women are products of different cultures. They possess different, but equally valid, communicative styles. Scollon and Scollon (2001, 2003) located meaning in the richness and complexity of the lived world rather than just in the language itself. They consider communicative action as a form of selection that positions the interlocutor as a particular kind of person who chooses among different meaning potentials a subset of pathway (Scollon and Scollon 2003: 205). Sociopragmatics has served as a basis for the development of several subfields of pragmatics, including interlanguage pragmatics, cross-cultural pragmatics, and, to some extent, intercultural pragmatics.
21.3.2 Interlanguage pragmatics and cross-cultural pragmatics Difference should be made between interlanguage pragmatics and cross-cultural pragmatics. Although these terms are often used interchangeably, they do not refer to the same inquiry. Interlanguage pragmatics focuses on the acquisition and use of pragmatic norms in L2: how L2 learners produce and comprehend speech acts and how their pragmatic competence develops over time (e.g. Blum-Kulka and Kasper 1993; Kasper 1998). Boxer (2002) argued that interlanguage pragmatics focuses on the language learner’s appropriation and/or acquisition of pragmatic norms represented in the host language community. To date, many cross-sectional, longitudinal, and theoretical studies have been conducted mainly with focus on L2 classroom interactions, which has resulted in a special tie between interlanguage pragmatics and second-language acquisition research.
405
Cross-Cultural and Intercultural Pragmatics 405 In a way, interlanguage pragmatics incorporates cross-cultural pragmatics, although there is some difference between the two lines of research. Cross-cultural pragmatics ‘takes the view that individuals from two societies or communities carry out their interactions (whether spoken or written) according to their own rules or norms, often resulting in a clash in expectations and, ultimately, misperceptions about the other group’ (Boxer 2002: 151). Cross-cultural studies focus mainly on speech act realizations in different cultures, cultural breakdowns, and pragmatic failures, such as the way some linguistic behaviours considered polite in one language may not be polite in another language. A significant number of these studies use a comparative approach to different cultural norms reflected in language use (e.g. Thomas 1983; House 2000; Spencer-Oatey 2000). Interlanguage pragmatics and cross-cultural pragmatics are based primarily on three theoretical approaches: Gricean pragmatics, Brown and Levinson’s politeness theory, the so-called “interlanguage hypothesis” (Selinker 1972), and an understanding of pragmatic competence in L2 which was developed within the interlanguage paradigm. Recently, attempts have been made to integrate relevance theory (e.g. Escandell-Vidal 1996; Jary 1998) and conversation analysis (e.g. Markee 2000; Kasper 2004) into interlanguage pragmatics although the main foci of research have remained pragmatic competence, speech acts, politeness, and pragmatic transfer. Kecskes (2011a, 2013) argued that the problem with interlanguage pragmatics is that it represents a monolingual and cross-cultural rather than a multilingual and intercultural view, inasmuch as all of its theoretical resources (Gricean theory, politeness theory, and the interlanguage hypothesis) advocate the relative independence rather than the interdependence of language systems and cultures, and proclaim the universality of principles such as those of cooperation and politeness. Wierzbicka (1985), Goddard and Wierzbicka (1997), and Meier (1997), among others, have questioned the claims made for the universality of Grice’s cooperative principle (Grice 1961), and Brown and Levinson’s theory of politeness (Brown and Levinson 1987). For instance, they have made a case for the cultural relativity of definitions of sincerity and relevance in a given speech community or the ranking of imposition when a request is made.
21.4 Theoretical Foundation of Intercultural Pragmatics 21.4.1 General remarks The study of intercultural pragmatics supports a less idealized, more down-to-earth approach to communication than current pragmatic theories usually have. Whilst not denying the decisive role of cooperation, context, and politeness in communication, intercultural pragmatics also gives equal importance to egocentrism, chaos, aggression,
406
406 Istvan Kecskes trial and error, and salience in the analysis of language production. Intercultural pragmatics adopts a sociocognitive approach (SCA) to pragmatics that takes into account both the societal and individual factors including cooperation and egocentrism that, as claimed here, are not antagonistic phenomena in interaction (Kecskes 2008, 2010a). SCA is considered an alternative to current theories of pragmatics that may not give an adequate account of what really happens in the communicative process. Current theories consider communication an idealistic, cooperation-based, context-dependent process in which speakers are supposed to carefully construct their utterances for the hearer, taking into account all contextual factors, and hearers do their best to figure out the intentions of the speakers. This approach relies mainly on the positive features of communication including cooperation, context, rapport, and politeness while almost completely ignoring the untidy, trial-and-error nature of the process and the importance of prior and emerging contexts captured in the individual use of linguistic expressions. The overemphasis on cooperative, societal, politeness-related, and contextual factors has led to the neglect of individual factors such as egocentrism and salience that are as important contributors to the communicative process as cooperation, context, and rapport. The sociocognitive approach serves as a theoretical framework for intercultural pragmatics to incorporate and reconcile two seemingly antagonistic sides of the communicative process and explain the dynamic interplay of prior and actual situational contexts. Before describing the main tenets of SCA we have to make a clear distinction between the sociocognitive approach proposed by Kecskes (2010a, 2011a, 2013) and van Dijk’s understanding of the sociocognitive view in language use. Van Dijk (2008: x) said that in his theory it is not the social situation that influences (or is influenced by) discourse, but the way the participants define the situation. He goes further and claims that ‘contexts are not some kind of objective conditions or direct cause, but rather (inter)subjective constructs designed and ongoingly updated in interaction by participants as members of groups and communities (van Dijk 2008: x)’. SCA adopts a dialectical perspective by considering communication a dynamic process in which individuals are not only constrained by societal conditions but also shape them at the same time. Speakers and hearers are equal participants in the communicative process. They both produce and comprehend speech relying on their most accessible and salient knowledge expressed in their private contexts in production and comprehension. Blending, which is the main driving force of intercultural interactions is more than just the co-constructing. It is combining the interlocutors’ prior experience with the actual situational experience which creates a blend that is more than just a merger. In blending the constituent parts are both distinguishable and indistinguishable from one another when needed. Blending incorporates the dynamic interplay of crossing (parts are distinguishable) and merging (parts are indistinguishable). Depending of the dynamic moves in the communicative process either crossing or merging becomes dominant to some extent. From the perspective of blending it is very important to realize that there are social conditions and constraints (contexts) which have some objectivity from the perspective
407
Cross-Cultural and Intercultural Pragmatics 407 of individuals. Of course, there can always be slight differences in how individuals process those relatively objective societal factors based on their prior experience. But it would be a mistake to deny the presence of any objectivity in social contexts. When language is used, its unique property is activated in two ways. When people speak or write, they craft what they need to express to fit the situation or context in which they are communicating. But, at the same time, the way people speak or write the words, expressions, and utterances they use creates that very situation, context, sociocultural frame in which the given communication occurs. Consequently, two things seem to happen simultaneously: people attempt to fit their language to a situation or context that their language, in turn, helped to create in the first place (e.g. Gee 1999). Social and cultural routines result in recurring activities and institutions. However, these institutions and routinized activities have to be rebuilt continuously in the here and now. The question is whether these cultural models, institutions, and frames exist outside language or not. The social constructivists insist that models and frames have to be rebuilt again and again so it is just our impression that they exist outside language (see van Dijk 2008). However, the sociocognitive approach that is the theoretical frame for intercultural pragmatics argues that these cultural mental models have some kind of psychological reality in the individual mind, and when a concrete situation occurs the appropriate model is recalled, which supports the appropriate verbalization of triggered thoughts and activities. Of course, building and rebuilding our world occurs not merely through language but through the interaction of language with other real-life phenomena such as non-linguistic symbol systems, objects, tools, technologies, etc.
21.4.2 Sociocognitive approach to intercultural pragmatics The sociocognitive approach to intercultural pragmatics (Kecskes 2008, 2010a, 2013; Kecskes and Zhang 2009) emphasizes not only the role of co-construction but also the importance of prior knowledge in the interaction. SCA points out the complex role of cultural models and private mental models, and how these are applied categorically and/or reflectively by individuals in response to socio-cultural environmental feedback mechanisms, and how this leads to and explains different meaning outcomes and knowledge transfer. In meaning construction and comprehension individuals rely both on pre-existing encyclopaedic knowledge based on their prior experience and current knowledge created in the process of interaction. In the sociocognitive paradigm, communication is driven by the interplay of cooperation required by societal conditions and egocentrism rooted in prior experience of the individual. Egocentrism should be considered part of human rationality as much as cooperation is. Human beings are inherently cooperative and inherently egocentric. Consequently, egocentrism and cooperation are not mutually exclusive phenomena. They are both present in all stages of communication to a different extent because they represent the individual and societal traits of the dynamic process of communication (Kecskes 2010a, 2010b, 2013). On the one hand, speakers and hearers are constrained by
408
408 Istvan Kecskes societal conditions but as individuals they all have their own goals, intentions, desires, etc. that are freely expressed and recognized in the flow of interaction. This is not the denial of the pragmatic theories that have grown out of the cooperation- centred Gricean approach. Recognizing the important of egocentrism of speaker-hearers, the sociocognitive approach is more like a synthesis of the cooperation-centred view of communication and the egocentrism-based cognitive psychological approach. Several researchers (e.g. Keysar and Bly 1995; Giora 2003; Barr and Keysar 2004; Keysar and Barr 2005) have indicated that speakers and hearers are egocentric to a surprising degree, and individual, egocentric endeavours of interlocutors play a much more decisive role in the initial stages of production and comprehension than current pragmatic theories envision. Egocentrism here does not refer to anything negative. Egocentric behaviour is rooted in the interlocutors’ greater reliance on their own knowledge instead of mutual knowledge. Recent research (e.g. House 2003; Kecskes 2007) in intercultural communication also affiliates with cognitive dynamism. Kecskes (2007) argued that especially in the first phase of the communicative process, instead of looking for common ground, which is absent to a great extent, lingua franca speakers articulated their own thoughts with linguistic means that they could easily use. It should be emphasized, however, that this egocentrism of interlocutors is not a negative phenomenon. As it is the result of prior experience that strongly affects memory and how the mind operates, individuals just cannot help that whatever they do (including communication) is subconsciously, automatically influenced by what they have already done. So what we are speaking about here is not a conscious, deliberate egotistic behaviour but the effect of cumulative experience that can hardly be ignored. In the sociocognitive approach framed by the dynamic model of meaning (Kecskes 2008; Kecskes and Zhang 2009), communication is characterized by the interplay of two traits that are inseparable, mutually supportive, and interactive: Individual traits: attention private (prior) experience egocentrism salience
Social traits: intention actual situational experience cooperation relevance
Communication is the result of the interplay of intention and attention motivated by sociocultural background that is privatized individually by interlocutors. The sociocultural background is composed of environment (context in which the communication occurs), encyclopaedic knowledge of interlocutors deriving from their ‘prior experience’, tied to the linguistic expressions they use, and their ‘actual situational experience’, in which those expressions create and convey meaning (Kecskes 2013). A central element of SCA is privatalization, a process through which the individual blends his prior experience with the actual situational (current) experience, and makes an individual understanding of collective experience. The following example illustrates this process.
409
Cross-Cultural and Intercultural Pragmatics 409 (3)
Two friends are talking. Jill: I met someone yesterday. Jane: Good for you. Jill: He is a police officer. Jane: Are you in trouble?
The utterance ‘I met someone yesterday’ sets the scene; it creates the actual situational context. Jill wants to talk about some kind of romantic involvement. She says that she met a police officer. The expression ‘police officer’ usually has a highly conventionalized negative prior context attached to it, which the actual situational context cannot override as Jane’s question demonstrates (‘are you in trouble?’). Jill’s public context (collective experience) is changed (privatized) as a consequence of her positive experience with a police officer. But, because she did not have that positive experience Jane continues interpreting ‘police officer’ according to the public context attached to the word.
21.5 Interculturality Interculturality is a central notion for intercultural pragmatics. We need to define interculturality in communication and separate it from intraculturality. There have been several attempts (e.g. Nishizaka 1995; Ting-Toomey 1999; Samovar and Porter 2001; Gudykunst and Mody 2002) to explain the difference between the two terms. According to Samovar and Porter (2001) ‘intracultural communication’ is ‘the type of communication that takes place between members of the same dominant culture, but with slightly different values’, as opposed to ‘intercultural communication’ which is the communication between two or more distinct cultures. This approach has led to a common mistake that several researchers have committed. They have considered interculturality as the main reason for miscommunication (e.g. Thomas 1983; Hinnenkamp 1995; Ting- Toomey 1999). In fact, some researchers’ findings show the opposite (e.g. House 2003; Kecskes 2007). The use of semantically transparent language by non-native speakers results in fewer misunderstandings and communication breakdowns than expected. The insecurity experienced by lingua franca speakers makes them establish a unique set of rules for interaction which may be referred to as an ‘interculture’, according to Koole and ten Thije (1994: 69) a ‘culture constructed in cultural contact’. Blum-Kulka et al. (2008: 164) defined interculturality as ‘a contingent interactional accomplishment’ from a discoursive-constructivist perspective. Several researchers have argued that a growing literature explores interculturality as a participant concern (e.g. Mori 2003; Markee and Kasper 2004; Higgins 2007). Nishizaka (1995) pointed out that interculturality is a situationally emergent rather than a normatively fixed phenomenon. The sociocognitive approach goes one step further and defines interculturality as a phenomenon that is not only interactionally and socially constructed in the course of communication but also relies on relatively definable cultural models and norms that
410
410 Istvan Kecskes represent the speech communities to which the interlocutors belong. Consequently, interculturality has both relatively normative and emergent components. In order for us to understand the dynamism and ever-changing nature of intercultural encounters, we need to approach interculturality dialectically. Cultural constructs and models change diachronically while cultural representation and speech production by individuals changes synchronically. Interculturality is a situationally emergent and co-constructed phenomenon that relies both on relatively definable cultural norms and models as well as situationally evolving features (Kecskes 2010a, 2011a, 2011b). Intercultures are ad hoc creations. They are created in a communicative process in which cultural norms and models brought into the interaction from prior experience of interlocutors blend with features created ad hoc in the interaction in a synergetic way. The result is intercultural discourse in which there is mutual transformation of knowledge and communicative behaviour rather than transmission. The emphasis is on transformation rather than on transmission. Interculturality has both an a priori side and an emergent side that occur and act simultaneously in the communicative process. Consequently, intercultures are not fixed phenomena but are created in the course of communication in which participants belong to different L1 speech communities, speak a common language, and represent different cultural norms and models that are defined by their respective L1 speech community. The following conversation (source: Albany English Lingua Franca Dataset collected by PhD students) between a Brazilian girl and a Polish woman illustrates this point well. (4) Brazilian: And what do you do? Pole: I work at the university as a cleaner. B: As a janitor? P: No, not yet. Janitor is after the cleaner. B: You want to be a janitor? P: Of course. This interaction shows clearly how blending takes place in intercultural communication. In this exchange interlocutors represent two different languages and cultures (Brazilian and Polish), and use English as a lingua franca. This is the prior knowledge and experience that participants bring to the interaction. They create an interculture, which belongs to none of them but emerges in the course of conversation. Within this interculture the two speakers have a smooth conversation about the job of the Polish woman. Neither of them is sure what the right term is for the job the Polish woman has. They try to apply what their cultural models dictate but cannot be sure that the English words they have chosen describe what their culture ‘recommends’. However, there are no misunderstandings in the interaction because each participant is careful to use semantically transparent language in order to be as clear as possible. The Polish woman sets up a ‘hierarchy’ that may not exist in the target language culture (‘cleaner’ versus ‘janitor’). However, this is an emergent element of the interculture the interlocutors have been constructing.
411
Cross-Cultural and Intercultural Pragmatics 411 Intercultures come and go, so they are neither stable nor permanent. They just occur. They are both synergetic and blended. Interculturality is constituted on the spot by interlocutors who participate in the conversation. But isn’t this a phenomenon that also occurs in intracultural communication? Why and how should we distinguish intercultural communication from intracultural communication? Basically the currently dominant approach to this issue is that there is no principled difference between intracultural and intercultural communication (e.g. Winch 1997; Wittgenstein 2001). This is true as far as the mechanism of the communicative process is concerned. However, there is a qualitative difference in the nature and content of an intracultural interaction and an intercultural interaction. Speakers in intracultural communication rely on prior knowledge and culture of a relatively definable speech community, which is privatized by individuals belonging to that speech community. No language boundaries are crossed; however, subcultures are relied upon and representations are individualized. What is created in the course of communication, to some extent, enriches the given culture, contributes to it, and remains within the fuzzy but still recognizable confines of that language and culture. In the case of intercultural communication, however, prior knowledge and experience that are brought into and privatized in the communicative process belong to different cultures and languages, and what participants create during the communicative process will disappear and may not become an enrichment and/or addition to any particular culture or language. Intercultures are ad hoc creations that may enhance the individual and the globalization process but can hardly be said to contribute to any particular culture. This is exactly what we see in example (4) above. Speakers created a hierarchy between ‘cleaner’ and ‘janitor’ just to create common ground and assure their own mutual private understanding of a given situation. This interculture usually disappears when participants stop talking. However, intercultures can also be reoccurring for a while in certain cases such as international negotiating teams, international classrooms, international tourist groups, etc.
21.6 Data Analysis in Intercultural Pragmatics In intercultural pragmatics, focus is on the discourse segment rather than just on the utterance as is the case in pragmatics proper. Analysis in pragmatic theories is directed on communicative actions (speech acts, pragmatic action, utterance) while in intercultural pragmatics discourse segment is the main concern. Pragmatic theories are utterance-centred, which means that in these theories the most significant difference between a sentence and an utterance is that sentences are judged according to how well they make sense grammatically, while utterances are judged according to their communicative validity (Habermas 1979: 31). Austin’s (1976) work is widely associated with the concept of the speech act and the idea that speech is itself a form
412
412 Istvan Kecskes of action. In his opinion language is not just a passive practice of describing a given reality, but a particular practice to invent and affect those realities. What can we do/ achieve with words and utterances? This question has been one of the main driving forces of contemporary pragmatics research. So it is not surprising that the main focus of pragmaticists since Austin has been meaning conveyed by an utterance in its actual situational context in intracultural communication. However, utterance analysis in intercultural pragmatics may be problematic for two reasons. On the one hand, utterances in intercultural communication are often not quite properly formed because of language proficiency issues. On the other hand, in an experimental study Kecskes (2007) demonstrated that the creativity of lingua franca speakers is observable on the discourse level rather than utterance level. A similar claim was made in Prodromou (2008). Consequently, in intercultural communication, it is more practical to analyse discourse segments rather than just utterances. This makes sense if we think about the fact that the criticism of intercultural interaction as a phenomenon characterized by miscommunication, lack of systematic coherence, and a low level of creativity becomes invalid when we analyse interactions on a discourse segment rather than an utterance level. The following discourse segment demonstrates that coherence and interpretability in intercultural communication is more like a discourse-level phenomenon than an utterance-level phenomenon (see also Kecskes 2013). (5) A Polish woman is speaking about housing and English study with a man from Hong Kong and a Bolivian girl. HKM: How about you? … circumstances … Where do you live? You rent … or? PAF: Mmm … I am live … not so far from the university. HKM: Hmm. PAF: It is a college … you don’t know … It is not so far. It’s Albany. HKM: Hmm. PAF: But I have 3 minutes … errh … to go to my work. HKM: Hmm. PAF: And this house is … errh … on our own so my husband and I … HKM: Hmm. BIF: Doesn’t have any kids? PAF: No. don’t have … BIF: No? PAF: Yeah. BIF: Errh … Are your husband American or … ? PAF: Errh … Actually he is … he is Polish. He is American but … BIF: OK. PAF: … because he came into the United States when he was a child. HKM: Hmm. PAF: He was something like twelve. BIF: All right.
413
Cross-Cultural and Intercultural Pragmatics 413 The whole interaction is quite segmented. If we analyse the utterances that the speakers produce what we see is that they are full of mistakes, and occasionally do not seem to be directly relevant to the previous utterance. They are also quite short and do not reflect much creativity. However, on the discourse level the segment appears to be coherent. The dialogue as a whole makes sense, and the speakers understand each other perfectly. Creativity is clearly detectable at discourse level. This segmented and occasionally ungrammatical nature of intercultural communication requires us to revise our understanding the role of contextualization cues, which represented the central innovation in Gumperz’s analysis of discourse and have been so important and influential in any kind of oral and written text, dialogue and discourse analysis. Using an interactional sociolinguistics approach, Gumperz (1982a) analysed gatekeeping encounters (such as job interviews) to shed light on sources of misunderstanding and miscommunication. Contextualization cues have played a pivotal role in his analysis. These cues are any linguistic or paralinguistic signals that give meaning to an utterance and help its interpretation. They are present in the surface structure of messages so they are empirically detectable (Gumperz 1982a, 1992b). Gumperz (1982a) underlined that suprasegmental features are crucial to the process of conversational inference. Levinson (2003b) further developed the understanding of contextualization cues, arguing that cues are first of all prosodic or paralinguistic in nature. When contextualization cues are lexical or grammatical they are a matter of fine-tuned distinctions not readily observable without particular analysis, such as word choice or register and/or minor grammatical structures like, for instance, particles. Levinson argued that contextualization cues are ‘non-propositional content, e.g. affectual, rhetorical, or metalinguistic’ and ‘reliant on a large dose of inferencing’ (Levinson 2003b: 37), so cues are context-dependent. What cues depend upon is co-occurring expectation. Speakers develop these expectations through their prior experience usually rooted in their own culture, and through their previous interactions (usually in their L1). So the cues help speakers make hypotheses about an interaction (contextual presuppositions), and then interpret meaning as the conversation moves forward. This is where the Gumperzian idea meets with one of the major tenets of the sociocognitive approach: the message can carry with it or project the context. But it is not necessarily the message content that does that but the message background, the attached sociocultural load whose presence can be triggered by conversational cues. The goal of these cues is to project the context in which the meaning of the message should be interpreted. Kecskes (2013) argued that the reason why semantically strong messages are occasionally interpreted out of the actual situational context is that the cues that could project the right interpretation are missing. If we have a contextualization cue, the semantic content of the message can be directed to the right interpretation. For instance, in the film Survivors Robin Williams says the following: ‘I had to sleep with the dogs. Platonically, of course … . ’ The contextualization cue ‘platonically’ directs the hearers to the right interpretation of the utterance. This is what Levinson said about the nature of contextualization cues: ‘This is what I take to be the Gumperzian notion, in which the term ‘cue’ denotes an encoded or
414
414 Istvan Kecskes conventional reminder, like a knot in a handkerchief, where the content of the memo is inferentially determined. Thus the ‘cue’ cannot be said to encode or directly invoke the interpretative background, it’s simply a nudge to the inferential process. Moreover, the interpretative process is guided more by a series of nudges now in one direction and now in another—thus ‘cues’ come as complex assemblages where the result of the whole assemblage cannot be equated with the inferential results that each part alone might have. The interpretive process may be guided by general pragmatic principles of a Gricean sort, and thus be in many ways universal in character; but the ‘cues’ are anything but universal, indeed tending toward subcultural differentiation. Hence the Gumperzian perspective on communication: at once potentially possible across cultural divides and inevitably thwarted by cultural nuances’ (Levinson 1997: 27). As far as the use of contextualization cues in intercultural interactions are concerned, the main issue is that, according to the Gumperzian approach, most of these cues are culture-specific so ‘ … they can only be learnt by rich exposure to a communicative tradition, a deep immersion in social networks’ (Levinson 1997: 29). This leads us to the question: how can interlocutors in intercultural interactions manage with or without these cues? Will their use or no-use lead to misunderstandings or miscommunication? Well, in certain cases yes, in other cases no. For instance, example (6) demonstrates a case of ‘yes’. (6) An American male student and a Korean female student are talking. NS: But I’ve never °skiied before. Have you skiied before?° NNS: (Oh yeah, in) Korea. NS: Really? (0.7) NS: Are there a lot of places to ski in Korea? (1.5) NNS: Actually in in winter I don’t like go to the gym. NS: Ye[ah: NNS: [So. And I don’t like (0.7) to walk, (.) because (0.5) It’s ↑too cold. NS: Yeah. (0.7) NS: >So you ski?< 11:00 NNS: ((laughs)) It kind’ve not–doesn’t make sense but–(0.7) ski::s (0.5) I like ski:s. NS: It’s exercise, right? NNS: Yeah, exercise. There appears to be a contradiction between two statements of the Korean speaker. First she said that she does not like walking in winter because it is cold. Then she said that she goes skiing instead. The NS notices that contradiction and refers to it in a
415
Cross-Cultural and Intercultural Pragmatics 415 very subtle way ‘so you ski?’. Both the intonation and the question can be considered as contextualization cues. The reaction of the Korean speaker shows that she perfectly understood what the American student has referred to by the cues. In the sociocognitive approach contextualization cues are considered not always culture-specific. This is a significant difference from the Gumperzian approach. There are individual ways of using contextualization cues in an interaction, and they do not always derive from collective experience and standard prior context that is tied to the L1 culture or any other culture of the speaker. Sometimes these contextualization cues are ad hoc creations of the individual in response to actual situational context, as demonstrated in examples (5) and (6) where the speakers representing different cultures act according to neither of those cultures. They select their wording and contextualization cues as they find appropriate in the actual situational context. Utterances are co-constructed and contextualization cues are created from scratch. However, this is only one side of the process. We cannot generalize that this is always what is happening. In other cases or in other parts of the same discourse sequence existing, culture-specific contextualization cues are used. So the point is that intercultural interaction is a dynamic phenomenon characterized by both a priori and emergent features that blend in the course of interaction. The process in these encounters is synergistic in the sense that it is a blend in which pragmatic norms of each participant are represented to some extent, and blended with the elements co-constructed by the interlocutors in the process of interaction. This is what intercultural pragmatics is deemed to investigate and describe.
416
Chapter 22
Interl ang uag e Pragm at i c s J. César Félix-B rasdefer
22.1 Introduction Interlanguage pragmatics (ILP) lies at the intersection of pragmatics and the study of second language acquisition (SLA). ILP, like SLA and pragmatics, is an interdisciplinary field that has been studied from various theoretical, analytical, and methodological perspectives. SLA examines how learners (children or adults) of a second language (or third or subsequent languages) with (limited) exposure to that second language (L2) sequentially learn (or do not learn) its linguistic system: namely, the sound system (phonology), vocabulary, and word formation (morphology); how meaning is expressed in words or sentences (semantics); and how sentences are formed (syntax). Researchers in SLA often refer to three linguistic systems: (i) the language produced by L2 learners is referred to as the interlanguage (IL); (ii) the learner’s native language (NL), also called the mother tongue, or the first language (L1); and, (iii) the target language (TL), L2, or the language being learned (Selinker 1972). Unlike ILP, which focuses on the use of the TL by non-native speakers (NNSs), Bardovi-Harlig (2013) introduced the term ‘L2 pragmatics’ to refer to learning that occurs in the developing system of one or more learners over time, and in different learning environments. These may include a foreign language (FL) classroom; a naturalistic context in an L2, in which the TL is the official language; or an immersion setting in the learner’s culture, such as learning French in Canada, Spanish in one-way or dual immersion schools in the US, or as a result of study abroad. This chapter focuses on the L2 pragmatic system—specifically, on the learner’s pragmatic competence, broadly defined as the learner’s ability to produce and comprehend linguistic (and non-verbal) action in socially appropriate communicative settings. This chapter is organized as follows. First, I present the scope of ILP, including the dimensions of pragmatic competence. Then, I provide a selective account of the most
417
Interlanguage Pragmatics 417 influential theoretical and methodological models that examine learning and use within L2 pragmatics. Then, I offer an overview of the concept of pragmatic transfer, followed by key concepts and theories used to examine pragmatic development in different learning contexts. I also provide a selective account of data collection methods in ILP pragmatics research. In the last section, I outline future directions that researchers in ILP should focus on in order to broaden our understanding of learners’ pragmatic competence.
22.2 Defining the Pragmatics of Interlanguage Pragmatics Pragmatics can be analysed from two perspectives: the Cognitive- Philosophical view (or Anglo-American pragmatics) and the Sociocultural-Interactional view (or European-Continental pragmatics). Under the first view, pragmatics is defined as ‘the systematic study of meaning by virtue of, or dependent on, the use of language’ (Huang 2014: 4). Within this view, pragmatics is considered to be one of the components of the mental grammar (similar to phonology, morphology, semantics, or syntax) and includes topics such as deixis, speech acts, reference, presupposition, and implicature. In contrast, under the second view, the Sociocultural-Interactional perspective, pragmatics adopts ‘a functional perspective on all core components and ‘hyphenated’ areas of linguistics and beyond’ (Huang 2014: 5). The field of ILP falls under the second area since it examines linguistic action taking into account cognitive and sociocultural aspects for the production and understanding of social action in an L2 or FL setting, such as degrees of impoliteness, interpretation of implicature, and directness or indirectness. It examines functional knowledge—specifically, the pragmatic meaning of linguistic and non-linguistic action produced and interpreted by L2 learners in institutional and non-institutional settings. Pragmatics has been defined from various perspectives. For example, Levinson (1983) adopts an inferential view of pragmatic meaning. For this author, pragmatics is concerned with ‘detailed inferences about the nature of the assumptions participants are making, and the purposes for which utterances are being used. In order to participate in ordinary language usage, one must be able to make such calculations, both in production and in interpretation’ (Levinson 1983: 53). This definition is adopted by researchers who examine the learning of implicature among L2 learners (e.g. Bouton 1994). Next, Yule’s definition of pragmatics is concerned with ‘meaning as communicated by a speaker (or writer) and interpreted by a listener (or reader)’ (Yule 1996: 3). Crystal (1997: 301) defines pragmatics as ‘the study of language from the point of view of users, especially of the choices they make, the constraints they encounter in using language in social interaction and the effects their use of language has on other participants in the act of communication’. This definition encompasses both the speaker’s production
418
418 J. César Félix-Brasdefer of meaning and the hearer’s interpretation of that meaning in communicative settings. It includes both the illocutionary force of a speech act as well as the perlocutionary effects of what is said on the feelings, attitudes, and emotions of the interlocutor during the realization of joint action (Clark 1996). On the basis of these definitions, efforts to define ILP adopted a narrow perspective of pragmatics. In their detailed study of data collection methods in ILP, Kasper and Dahl (1991: 271) defined pragmatics as ‘referring to nonnative speakers’ comprehension and production of speech acts, and how their L2-related speech act knowledge is acquired’. Kasper and Schmidt (1996: 150) add the learning component to ILP with respect to ‘acquisitional patterns of interlanguage knowledge over time’ and define ILP as ‘the study of the development and use of strategies for linguistic action by nonnative speakers’. While ILP includes the comparison and use of linguistic action by L2 learners, the learning component of pragmatics is referred to as acquisitional pragmatics (Bardovi-Harlig 1999), or L2 pragmatics which focuses ‘on the development of the L2 pragmatic system’ (Bardovi-Harlig 2013: 69). Thus, research on pragmatic development and use can be included under the field of ILP, but not all studies of ILP belong to L2 pragmatics, as the latter focuses on acquisitional patterns and the developing system of L2 pragmatics. The targets of ILP should include the production and comprehension of pragmatic meaning in an L2: speech acts, deixis, reference, speech act sequences in written and spoken discourse, conversational interaction, implicature, discourse management, politeness and impoliteness, and sociolinguistic aspects of language use, such as pragmatic regional variation among learners of different target varieties of the same language. Studies in ILP often distinguish between two types of pragmatic knowledge, namely, pragmalinguistic and sociopragmatic knowledge. Pragmatic knowledge, according to Leech (1983) and Thomas (1983), is comprised of two components: pragmalinguistic competence—knowledge about and performance of the conventions of language use or the linguistic resources available in a given language that convey ‘particular illocutions’ in contextually appropriate situations (Leech 1983: 11). It includes knowledge of strategies (e.g. directness, conventional indirectness) and the linguistic and non- linguistic resources (e.g. prosody) used to convey pragmatic meaning, such as the various linguistic resources employed to express an apology: ‘I’m (so, very, awfully, terribly) sorry’; ‘I apologize, please forgive me’, etc. In contrast, sociopragmatic competence refers to knowledge about and performance consistent with the social norms in specific situations in a given society, as well as familiarity with assessments of (im)politeness and variables of social power and social distance. For example, the form ‘I can’t’ (literal form) can be used in English to express a refusal to an invitation. However, when refusing an invitation to a birthday party, a speaker can select from a variety of other linguistic forms to convey the refusal, for example: ‘Thanks, but I really can’t’; ‘I’m sorry, but I have plans’; ‘I don’t know, I have to think about it’. Sociopragmatic knowledge includes knowledge of what expressions are appropriate (or are not appropriate) to use in an L2 when refusing a professor’s advice to take a class or apologizing to a friend for crashing his/her car over the weekend.
419
Interlanguage Pragmatics 419 In light of this distinction, pragmatic competence refers to the learner’s ability to produce and comprehend linguistic and non-verbal action at both the micro-level (speech acts) and macro-level (speech act sequences/discourse). According to a recent conceptualization proposed by Laughlin, Wain, and Schmidgall (2015), pragmatic competence encompasses five dimensions of knowledge. These include sociocultural knowledge, pragmatic functional knowledge, grammatical knowledge, discourse knowledge, and strategic knowledge. Accordingly, the challenge for the learner is to develop an understanding and control of these types of knowledge when producing and negotiating joint action. As will be shown in section 22.5.4, language instructors should be familiar with these concepts when developing materials for teaching and assessing pragmatics in the classroom.
22.3 Theoretical and Methodological Models in ILP ILP has been influenced by theoretical and methodological approaches that examine the production and comprehension of pragmatic meaning: speech act theory, conversational implicature, linguistic politeness, and cross-cultural pragmatics. These models have been used to examine various aspects of the learners’ pragmatic competence.
22.3.1 L2 speech acts Most research in ILP takes the speech act as the unit of analysis to examine one (or more) aspect(s) of the learners’ pragmatic knowledge, whether that comprises pragmalinguistic or sociopragmatic ability. Speech act theory (Austin 1962a; Searle 1969) represents the predominant framework in ILP. The focus of analysis in L2 pragmatics has been what Searle called the ‘illocutionary force’ of the speech act (i.e. the intended or unintended communicative force expressed by an act such as requesting or promising), including indirect speech acts (Searle 1975) (e.g. ‘I was wondering whether you would have to write me a letter of recommendation’). The majority of studies in ILP use Searle’s five-way classification of speech acts (1976): representatives (informing, stating), directives (asking, requesting), commissives (promising, refusing), expressives (complimenting, congratulating) and declarations (appointing, declaring) (see Kasper and Dahl 1991; Wierzbicka 2003; Márquez Reiter and Placencia 2005: ch. 2, Félix-Brasdefer 2007, and Bardovi-Harlig 2010 for an overview of empirical speech act research in different languages). Of these, requests, apologies, refusals, expressions of gratitude, and compliments represent the most widely studied speech acts in ILP. Different taxonomies have been proposed to analyse the pragmalinguistic structure of speech acts (e.g. Blum-Kulka 1982 [requests]; Blum-Kulka et al. 1989 [requests and
420
420 J. César Félix-Brasdefer apologies]; Beebe et al. 1990 [refusals]; Olshtain and Cohen 1983 [apologies]). For example, Koike (1994, 1995) proposed a taxonomy of pragmalinguistic expressions to examine suggestions, Valdés and Pino (1981) used a revised classification of compliment responses (based on Pomerantz 1978), and Nelson and Hall (1999) employed a revised classification of compliment formulas in Spanish (adapted from Manes and Wolfson 1981). As shown in sections 22.3.1 and 22.3.3, speech act research is conducted within two main theoretical frameworks: namely, speech act theory (Searle 1969, 1975) and linguistic politeness (Brown and Levinson 1987). Most work on ILP focuses on the linguistic realization of strategies to produce direct or indirect speech acts; the focus has generally been on conventional indirectness.
22.3.2 L2 implicature The concept of conversational implicature was introduced by British philosopher Paul Grice (1975) to show how speaker meaning, not directly encoded in the words expressed, can be inferred (recognized) by the hearer (see Huang’s chapter ‘Implicature’ in this volume). For example, if speaker A says ‘Has John arrived?’, and speaker B responds ‘There is a blue car in the driveway’, one can infer, under the appropriate circumstances and based on shared assumptions between the interlocutors, that John has arrived. Research in L2 implicature shows that beginning learners are less able to understand and use implicatures in the L2. Instead, they comprehend and use utterances that are direct in illocutionary force. Implicatures require more knowledge and linguistic skill from the L2 learner on all levels (vocabulary, syntax, semantics, pragmatics, etc.), so an increase in L2 proficiency allows greater understanding and use of L2 implicatures (Koike 1989, 1992). Research also shows that over time some types of implicature are acquired by learners of English. Roever (2013) examined the use of implicature for the purposes of diagnostic assessment among high-proficiency learners of English and NSs of English, using a ten-item multiple-choice task to test two types of implicature, idiosyncratic (general conversational implicature) and formulaic (indirect criticism, irony, scalar implicature). As expected, results showed that NSs scored significantly higher than the learners. In a series of studies, Bouton (1992, 1994, 1999) examined the ability of advanced ESL learners to interpret various types of implicatures in English using a cross-sectional design. The studies focused on pragmatic comprehension and used a multiple-choice questionnaire in fully contextualized situations to elicit the data. Learners with different L1 backgrounds were divided into three groups according to their length of residence (LR) in the United States: seventeen months, thirty-three months, and four to seven years. When the mean scores between the seventeen-month group (18.06) and the thirty-three-month group (18.80) were compared, no significant differences were found, although the mean scores were slightly higher in the group with a longer LR. Thus, it was concluded that learners’ ability to interpret implicatures seemed to have been achieved during the first seventeen months of their stay; after seventeen months
421
Interlanguage Pragmatics 421 in the L2 culture, the learners’ ability to interpret most implicatures stabilized, but progress was slow. However, due to the large spread between groups, it is difficult to know at precisely what point pragmatic development took place, or whether pragmatic development increased over the seventeen months and by the end of this period learners’ pragmatic ability to comprehend most implicatures had stabilized. More research on L2 implicature is needed to further examine the processes by which learners comprehend pragmatic meaning through implicature, including the ability to recognize irony, metaphor, and sarcasm.
22.3.3 L2 politeness and impoliteness Brown and Levinson’s (1987) universal model of linguistic politeness has been the predominant framework for the analysis of linguistic politeness in the realization of speech acts in ILP since the early 1980s. Researchers in ILP examine whether NSs of a given culture and L2 learners employ linguistic strategies that are oriented towards satisfying the hearer’s positive face through positive politeness strategies (e.g. agreeing with the interlocutor, use of in-group identity markers, as in ‘Come here, honey’) or negative face realized through negative politeness strategies (e.g. being conventionally indirect, as in ‘Can you please pass the salt?’ or using hedges, as in ‘I think that John is coming too’). The notion of a ‘face-threatening act’ (FTA) (e.g. requests, refusals, suggestions) is often used to examine instances of negative politeness. These authors argued that during social interaction a speaker must rationally assess the nature of an FTA by means of three independent factors that are culturally sensitive: the social distance (D) and social power (P) between a speaker and a hearer, and the absolute ranking (R) of impositions in a particular culture. Most researchers in ILP adopt the positive–negative politeness distinction with respect to the speaker’s intentional meaning. However, to avoid the confusión of positive and negative politeness, other researchers have used Scollon and Scollon’s (2001) dichotomy of involvement and independence face, and their systems of politeness to express solidity, deference, and hierarchical politeness (e.g. Félix-Brasdefer 2004, 2007). Likewise, Kecskes (2013) proposed the sociocognitive approach to examine instances of intercultural (im)politeness, taking into account cultural models, norms, and conventions in intercultural encounters. Kecskes’ model was used to examine the negotiation of facework strategies and power inequality in non-trial hearings between US judges (NS of English) and Spanish-speaking defendants (Lavin 2015); specifically, Lavin’s study looked at the pragmalinguistic strategies used by defendants when responding to the judge’s direct and imposing questions. Intercultural pragmatics includes NS– NNS interactions using English as the language of communication, and lingua franca encounters, where the target language is neither the native language of the speaker nor the hearer (e.g. a conversation between a French speaker and a Chinese speaker in English). For example, Johns and Félix-Brasdefer (2015) looked at the realization of linguistic politeness in direct and indirect requests in Dakar, Senegal. In this multilingual
422
422 J. César Félix-Brasdefer African context, French functions as the lingua franca (and the official language). It is mainly spoken in formal settings as an L2, whereas Wolof is the NL of Senegalese speakers, and is predominantly spoken in informal and familial settings. According to Kecskes, in intercultural interactions, the co-construction of (im)politeness is realized by means of shared cultural assumptions, emergent common ground, and assessments of (im)politeness in socially appropriate sociocultural contexts. Learners’ ability to express politeness in an L2 is a frequent topic of study in ILP research. For example, Scarcella’s (1979) study analysed the emergence of polite features when performing English (role-play) requests among two learner groups: one group of beginners and one advanced group. The author found that the learners’ use of internal modifiers to express politeness in their requests (e.g. please, maybe, kind of, I think) improved with increasing proficiency. Using role-play data, other studies have shown that direct requests decrease and a preference for conventional indirectness (using higher levels of polite requests) emerges with higher proficiency levels (e.g. Achiba 2003; Hassall 2003; Félix-Brasdefer 2007). Unlike the aforementioned research on the learning of polite practices, research on the learning of impolite behaviour is scarce. For instance, in her analysis of honorifics in Korean, Brown (2013) introduced learners to features of impolite behaviour in the classroom via input from a Korean TV drama; pragmalinguistic features of the interaction were discussed in class, followed by communicative practice. The metapragmatic comments revealed that learners were willing to learn about impolite practices in the classroom. Schepers (2014) also used a TV drama show to teach polite and impolite practices in the classroom. Similarly, Mugford (2008) showed that FL teachers of English are willing to teach impolite practices in the classroom. Finally, using a production questionnaire, Beebe and Waring (2005) conducted a cross-sectional study with two groups of learners (lower and higher proficiency) who responded to rude situations. Results showed proficiency effects, with the higher- proficiency group using more aggressing strategies (e.g. insult, threat, criticism) and persisting strategies (e.g. arguing than the lower-proficiency group), which showed a preference for acquiescing strategies (e.g. apology, thanks). This study shows evidence of pragmatic development in the learners’ ability to react to rude behaviour, with the higher-proficiency group sounding more aggressive and more assertive in their ability to react to rude behaviour. Additional studies on L2 impoliteness and the teaching of impolite behaviour in the classroom is a fruitful area for future investigation in ILP.
22.3.4 L2 conversations and interactional competence ILP has used the analytical tools proposed by conversation analysis (CA) to examine L2 encounters in social interaction. CA offers a rigorous methodological framework for the analysis of talk-in-interaction in both formal and non-formal contexts. It is concerned with sequential organization (e.g. openings, closings, speech act sequences,
423
Interlanguage Pragmatics 423 pre-/post-sequences) as well as the organization of turn-taking and preference or dispreference, the concept of repair, and the organization of laughter and topic development (Sacks, Schegloff, and Jefferson 1974; Schegloff 2007a, see also Schegloff this volume). The goal of CA is to explain how sequences are organized in spoken discourse. Kasper (2006) uses concepts from CA to explain the sequential development of social action from a pragmatic-discursive perspective. Gardner and Wagner (2004) also use the CA framework to analyse learners’ interactional competence by looking at the co-construction of repair, collaborative interaction, turn-taking, delay, and restarts, and the co-construction of sequences in L2 spoken interaction. These are NS–NNS interactions in naturalistic settings. In their edited volume on the assessment of L2 pragmatics, Ross and Kasper (2013) outlined various issues related to interactional competence in oral proficiency interviews including management of task uptake, development of extended turns, and assessment of L2 pragmatic ability. The authors adopted the CA framework to examine various aspects of learners’ interactional competence when responding to interviewers’ questions. Combining CA notions and interactional competence in the context of SA in Spain, Dings (2012) observed the participation and growth in conversational skills by one Spanish learner over the course of a year in Spain, detailing longitudinal changes in patterns of correction. The female learner conversed with the same NS in six encounters. In a different study with learners in an FL context, Félix-Brasdefer and Lavin (2009) looked at how Spanish learners used grammatical resources, including ‘increment initiators’ (e.g. y ‘and’, pero ‘but’), epistemic expressions (e.g. creo ‘I believe’), and prosodic cues (intonation, syllable length) to illustrate how grammatical, lexical, and pragmatic resources interact with the discourse to convey the effect intended by the speaker. Most of the studies on L2 interaction follow Young’s (2011) framework of interactional competence to trace development over time. His framework has implications for language learning, teaching, and testing.
22.3.5 Cross-Cultural Speech Act Realization Project (CCSARP) The CCSARP (Blum-Kulka et al. 1989) motivated empirical speech act research in cross-cultural and ILP research. It is an analytical framework for speech act analysis that is predominantly based on politeness theory (Brown and Levinson 1987) and a revised version of speech act theory (Bierwisch 1980), which situated the study of speech acts in linguistic communication. Blum-Kulka et al. (1989) investigated cross-cultural (i.e. interlingual) and intralingual variation in the realization of requests and apologies in seven countries. The data were collected using a Discourse Completion Test (DCT) initially used by Blum-Kulka (1982) in L2 requests with learners of Hebrew and later by researchers in L2 Spanish. The participants are asked to respond in writing to a series of hypothetical situations as they would in real life. The unit of analysis for these speech
424
424 J. César Félix-Brasdefer acts is the ʻsequence of utterances used to complete the missing lines in the discourse completion testʼ (Blum-Kulka et al. 1989: 20). The CCSARP proposed a classification of pragmalinguistic strategies for the analysis of requests and apologies (head acts) and it has been adapted to fit the Spanish data in contrastive (e.g. Márquez Reiter 2000) and ILP studies in L2 Spanish (e.g. Pearson 2006; Félix-Brasdefer 2007; Shively 2011). Internal modifiers consist of lexical (e.g. diminutive, ‘please’, ‘Do you think …?’) and syntactic elements (e.g. conditional, imperfect) that are often used to soften or mitigate the illocutionary force of a request. The CCSARP framework has influenced research in ILP with regard to pragmalinguistic (i.e. head acts and internal and external modification) and sociopragmatic variation (i.e. perceptions of social distance and power).
22.4 Pragmatic Transfer Pragmatic transfer plays a significant role in ILP research. Kasper (1998) distinguished two types of pragmatic transfer. Positive transfer occurs ‘when learners’ production of a pragmatic feature is the same (structurally, functionally, distributionally) as a feature used by target language speakers in the same context and when this feature is paralleled by a feature in learners’ L1’ (p. 193). Negative transfer, on the other hand, is observed ‘when a pragmatic feature in the interlanguage is (structurally, functionally, distributionally) the same as in L1 but different from L2’ (p. 194). A requirement for positive transfer is that the L2 group shows speech act behaviour similar to that of the NS group (Maeshiba et al. 1996). Of the two types of transfer, negative transfer has been a major research interest in ILP. Transfer occurs at both the pragmalinguistic and the sociopragmatic levels. First, pragmalinguistic transfer may occur when learners transfer specific expressions from their L1 to communicate linguistic action and these features coincide in form and function with those forms used by NSs of the TL (positive transfer). On the other hand, negative transfer (or interference from the L1) occurs when pragmatic features from the L1 are transferred inappropriately to the L2. For example, US learners of Spanish may use conventional indirectness (e.g. ‘Can I have …?’) to make a request for service in Spain or Mexico, instead of the expected direct request for service (e.g. ‘Give me a summer red [wine] please’) or elliptical requests (e.g. ‘a waffle with chocolate and cream’) (Félix-Brasdefer 2015; Shively 2011). In the context of service encounters, conventional indirectness represents an instance of negative pragmatic transfer (or an instance of pragmalinguistic failure), as conventional indirect requests do not represent a sociocultural expectation in service encounters in Spain, where the appropriate request for service would be through a direct request. Second, sociopragmatic transfer occurs when learners transfer the sociocultural norms or perceptions with regard to their understanding of social distance or social power from their L1 to the L2, although there may be differences between the L1 and L2. For instance, an insistence after an offer or following a refusal to an invitation represents a sociocultural expectation in
425
Interlanguage Pragmatics 425 many cultures (e.g. Peru, Mexico, China), while in other societies an insistence may be viewed as an imposition or inappropriate behaviour (e.g. United States). Nevertheless, before making generalizations with respect to pragmatic transfer, one must also consider the role that universal pragmatic competence plays in learners’ production and comprehension of linguistic action. That is, adult learners come to the classroom equipped with different types of implicit (pragmatic) knowledge of and ability to do the following: the rules of politeness, the form and use of communicative acts (e.g. refusing, requesting), the sequential structure of conversation (e.g. turn-taking, repair, openings and closings), and formal and informal registers (e.g. talking to a professor vs talking to a close friend). For additional information on universal pragmatic competence, see Kasper and Rose (2002). Speech act research has revealed that learners transfer their L1 norms into inappropriate L2 contexts. Negative transfer from the learners’ L1 has been found in the frequency, order, and content of semantic formulas (Beebe et al. 1990; Gass and Houck 1999; Félix-Brasdefer 2004); the transfer of L1 social perceptions to L2 contexts (Robinson 1992; Félix-Brasdefer 2003, 2007), and in the use of indirect request strategies in the L2 (Takahashi 1996). Instances of negative transfer in NNSs’ performance, however, must be carefully examined since a lack of sufficient L2 grammatical knowledge (i.e. vocabulary, morphology, phonology, and syntax) may not allow the learner to effectively perform appropriate language functions, or to comprehend the interlocutor’s intended message. Language instructors need to be aware of the conditions that govern pragmatic transfer to avoid instances of pragmalinguistic and sociopragmatic failure, and ultimately, to raise awareness of these concepts in the FL classroom.
22.5 Pragmatic Development in a Second Language Since research on ILP focuses on language use—and to a lesser degree on learning— Bardovi-Harlig (2013: 69) proposed the term ‘L2 pragmatics’ to denote the ʻdevelopment of the L2 pragmatic systemʼ, and as such ILP comprises the field of L2 acquisition of pragmatics. The term ‘L2 pragmatics’ is used exclusively to refer to studies that examine development or acquisition of L2 pragmatic knowledge over time as a result of incidental learning or pedagogical intervention. It is important to mention that pragmatic development is generally studied through longitudinal and cross-sectional studies. Longitudinal studies focus on direct observation of the same participants over an extended period, whereas cross-sectional studies look at development by comparing different learners at various proficiency levels. Each of these will be examined in this section, followed by an overview of pragmatic development in SA contexts and as a result of pedagogical intervention.
426
426 J. César Félix-Brasdefer
22.5.1 Longitudinal studies Research on pragmatic development as a result of exposure to input in naturalistic settings began in the early 1980s. Different theoretical and methodological approaches have been used to examine pragmatic development of the same learners over time. The acculturation model (Schumann 1978) represents one of the first theoretical approaches that looked at pragmatic development over time among adult learners. This model claims that variables such as social and affective distance with speakers of the target language influence the learner’s communicative linguistic ability. Schmidt (1983) employed this model to look at the development of interactional and discourse practices over time. The author showed that after a three-year observation period, his adult language learner, Wes, made little progress in his grammatical competence, while his pragmatic and discourse ability improved. He showed various examples of sophisticated interactional competence, despite insufficient pragmalinguistic knowledge of linguistic (polite) expressions to soften direct requests. Using the model of language socialization (Ochs 1996), Shively (2011) examined the pragmatic development of requests in the context of service encounters among seven US students learning Spanish in Spain (Toledo) over the course of one semester. Language socialization was defined as ‘the process whereby children and other novices are socialized through language, part of such socialization being a socialization to use language meaningfully, appropriately, and effectively’ (Ochs 1996: 408). The data in Shively’s study were collected from face-to-face service encounter interactions with NSs of the target language in stores, markets, and public institutions. Data were collected three times over one semester. The author found that as a result of direct socialization in the target culture, learners improved their ability to make appropriate requests (direct and elliptical/verbless requests) in intercultural encounters. The study emphasized the importance of implicit socialization that appeared to be conducive to pragmatic development over time. Pragmatic development has also been examined through Vygotsky’s notion of the Zone of Proximal development. In his qualitative study of pragmatic development, van Compernolle (2013) examined learners’ interactional competence in L2 French as part of a six-week research programme. This model looks at pragmatic development in both children and adult learners. It focuses on the learner’s developmental trajectory with assistance from the interlocutor in social interaction. Van Compernolle provided an analysis of the production of eight intermediate-level learners of French with regard to the sequential organization of opening and closing mediation sequences, as well as the interim discussion during the negotiation of the task. This study has pedagogical implications for learners’ interactional competence in the dynamic assessment of L2 pragmatic abilities. Pragmatic development has also been investigated from an L2 learning perspective. In their study of L2 advising sessions, Bardovi-Harlig and Hartford (1993) found that while NSs used lexical and syntactic mitigators, almost at a 100 per cent rate when performing suggestions, the advanced L2 learners employed a narrow range of
427
Interlanguage Pragmatics 427 mitigators, at only a 50 per cent rate over the course of one semester. Later, Bardovi- Harlig and Salsbury (2004) analysed the development of oppositional talk in L2 English conversation—specifically, turn organization in NS–NNS interactions. It was found that even though most learners began with direct disagreements, all of them elaborated their disagreement over time in the following ways: they increased the amount of talk, included both agreement and disagreement components, postponed their disagreements in later positions in the initial turn, and employed multiple turn structure to avoid disagreement. These studies provide strong evidence for the importance of natural conversation for the development of interactional competence as a result of incidental learning. With regard to longitudinal studies of children’s requests, various stages of development have been identified. For example, Ellis (1992) examined the opportunities that a classroom setting afforded two ESL boys (aged 10 and 11) when performing requests over a period of two years. He found that although the range of the learners’ request types extended slightly over time, direct requests predominated throughout, mostly in the form of imperatives. In addition, use of internal or external modification was limited. Three developmental stages were identified: (i) minimal request realizations in which the pragmatic intent was highly context-dependent; (ii) unanalysed routines and direct requests used as formulas (imperatives and query preparatory requests); and (iii) unpacking of routine formulas with a productive use of various request types and slow emergence of conventional indirectness even after two years of exposure in the host environment. In a more recent study, Achiba (2003) reports on the pragmatic development of her 7-year-old Japanese child (Yao) during her seventeen-month residence in Australia. Unlike the boys in Ellis’s study, Yao showed a fast decline in the proportion of imperatives and a shift to conventional indirectness. Four phases of development were observed: (i) frequent use of formulaic utterances (Ellis’s second stage); (ii) a shift from formulaic to non-formulaic language use and a substantial increase in some of the linguistic forms (Ellis’s third stage); (iii) pragmatic expansion: an increase and a wider variety of pragmalinguistic forms to express requestive intention (productive use of modals) and an increase in the use of conventional indirectness; and (iv) fine-tuning of requests: an increase of syntactically mitigated forms to express indirect requests. Although both studies examined the request behaviour of children in the host environment (England and Australia), Ellis’s requests are representative of classroom discourse, whereas Achiba’s data were collected at home during play interactions. The fact that Achiba’s phases 3 and 4 were not observed in Ellis’s study seems to indicate that Achiba’s learner may have had a more advanced level of linguistic proficiency than Ellis’s learners. Results from both studies, however, are limited to the developmental path of one or two children who were exposed to pragmatic input in the host environment in academic (Ellis 1992) and social (Achiba 2003) contexts. Taken together, five stages of pragmatic development of requests arose from Ellis’s (1992) and Achiba’s (2003) studies: (i) pre-basic (context-dependent; lexical requests with no functional inflection) (Klein et al. 1995); (ii) formulaic (unanalysed formulas); (iii) unpacking (productive language); (iv) pragmatic expansion (adding new forms); and (v) fine-tuning of
428
428 J. César Félix-Brasdefer requestive force and context (Kasper and Rose 2002). These stages of development have also been observed in cross-sectional studies with natural and elicited data.
22.5.2 Cross-sectional studies Cross-sectional studies that examine pragmatic development abound in the ILP literature. These studies look at development among learners (two or more groups) with different proficiency levels. One of the earliest cross-sectional studies on the development of L2 requests is Scarcella’s (1979) analysis of the emergence of polite features when performing English (role-play) requests. Development was analysed in two learner groups, one group of beginners and one advanced group. It was found, among other things, that learners’ use of internal modifiers (e.g. please, maybe, kind of, think) increased with increasing proficiency. In addition, it was observed that some of these politeness features emerged quite early in adult L2 acquisition, but that imperatives were invariably employed by the beginners. Hassall (2003) elicited requests through role-plays among lower-to upper-intermediate-proficiency Australian learners of Indonesian in the host environment. In Hassall’s study, direct requests decreased with increasing proficiency and a preference for conventional indirectness was observed with higher levels of proficiency. In the FL classroom context, Félix-Brasdefer (2007) showed that beginners and intermediate learners of Spanish appear to be impolite due to the lack of pragmalinguistic resources to mitigate a request in learner–NS role-play interactions. In this study, four stages of development were identified among beginners (i. pre-basic stage and ii. basic stage) and intermediate and advanced learners of FL Spanish (iii. unpacking of formulaic use and iv. pragmatic expansion). Bardovi-Harlig and Bastos (2011) analysed the interaction of proficiency, length of stay, and intensity of interaction with regard to recognition and use of conventional expressions by host-environment learners in four different levels. Results showed that both proficiency and intensity of interaction were significant factors that influenced the production of conventional expressions, whereas length of stay did not have a significant effect on either recognition or production. Flores-Salgado’s (2016) study looked at the effect of intensity of interaction and length of stay in two groups of Mexican students, one with study-abroad experience at a US summer camp; the other with no study-abroad experience in an English-speaking country. Effects were found for intensity of interaction but not for length of stay. Cross-sectional studies abound in the ILP literature because they allow the researcher to compare large amounts of data from two or more proficiency levels at once. Different variables can also be controlled, such as proficiency level, social distance and social power, individual differences, the register such as formal or informal style, and the setting (e.g. classroom or non-institutional contexts). They also give the researcher the opportunity to add a pedagogical treatment followed by one or more post-tests to measure whether explicit or implicit instruction (or both) has an effect on the learners’ developmental system. However, although cross-sectional studies yield predictions
429
Interlanguage Pragmatics 429 regarding pragmatic development with learners at different proficiency levels, they fall short in permitting the observation of the development of pragmatic patterns among the same learners over time. Thus, generalizations from cross-sectional studies should be interpreted with care.
22.5.3 Pragmatic development in study-abroad contexts Pragmatic development has also been documented among learners who study abroad for different lengths of time. Unlike learners who study an L2 in the FL classroom with limited access to natural input, learners in study-abroad contexts may improve their pragmatic knowledge as a result of exposure to authentic input from NSs of the TL and frequent participation and varied opportunities in the target culture. To observe pragmatic development in study-abroad contexts, researchers often compare the development of learners who receive instruction at home (AH) in the FL classroom with that of learners in SA contexts. The learner data are usually collected at various times (prior to departure, during, and at the end) to examine whether learners who are exposed to the TL during an SA experience improve their knowledge of various aspects of pragmatics at both the production and, to a lesser degree, the comprehension levels in communicative acts such as requesting, apologizing, advising, and conversational activities, including greetings, repetition, self-repairs, and accuracy checks. Pragmatic development is also examined in SA contexts where learners spend from four to seven weeks during the summer to one or two semesters studying Spanish abroad. In most of these studies there are control (the FL classroom [AH]) and experimental groups (SA). Some of the issues often addressed include the length of stay, learner characteristics, contact with the L2/TL or exposure, and the nature of the input in naturalistic settings, as well as intensity of interaction with the target culture. Usually the only difference between the AH and SA groups is exposure to input in the host environment for the SA group (often referred to as the treatment [natural input]), while the learners in the AH context are exposed to traditional instructional classroom input during the same period as the SA learners. Some of the speech acts analysed in study- abroad contexts include refusals (Félix-Brasdefer 2013), compliments (Félix-Brasdefer and Hasler-Barker 2015), requests (Bataller 2010; Shively 2011; Alcón 2013), requests and apologies (Cohen and Shively 2007; Shively and Cohen 2008). In these studies the length of stay abroad ranged from eight weeks (e.g. Félix-Brasdefer 2013) to one semester (Cohen and Shively 2007; Shively and Cohen 2008; Bataller 2010; Shively 2011). Of these, two studies added a pedagogical component to the SA groups. Shively (2011) provided metapragmatic instruction upon arrival in Spain prior to the beginning of service encounters with clerks in commercial and non-commercial settings. In their pedagogical intervention, Cohen and Shively (2007) provided the SA group with instruction on learning strategies for maximizing their experience. Other studies have shown evidence of incidental learning (e.g. Bataller 2010). Finally, in an examination of communication strategies by learners of Spanish in Alicante over the course of one
430
430 J. César Félix-Brasdefer semester, Lafford (2004) found that although both groups decreased their reliance on communication strategies (e.g. self-repair, circumlocution, asking for confirmation), learners in the SA group used significantly fewer communication strategies on the post-test than the AH group. The data consisted of extracts and role-plays taken from OPI interviews both before (pre-test) and after (post-test) SA. For additional information on SA research and L2 pragmatic development, see Kinginger (2013, 2015) and Ren (2015).
22.5.4 Pragmatic development as a result of pedagogical intervention The topic of instruction in pragmatics has received significant attention in ILP research. Research on ILP shows that learners who receive instruction in pragmatics develop different pragmatic systems from learners who do not receive instruction (Bardovi-Harlig 2001). Pragmatic development as a result of pedagogical intervention indicates that a consciousness-raising approach (Schmidt 1993) followed by metapragmatic instruction is more efficient than input alone (Kasper and Rose 2002). Moreover, the type of instruction—implicit, explicit, or both—influences the type of learning of the pragmatic target. Other factors that may influence pragmatic instruction include the modality in which input is provided, production vs comprehension (Koike and Pearson 2005), or whether the type of the task (e.g. role-plays or oral DCTs)—combined with metapragmatic instruction—impacts the development of specific pragmatic targets, such as compliments and compliment responses (Hasler-Barker 2013). Although learners who are exposed to pragmatic input in naturalistic settings acquire certain aspects of pragmatics over long periods, adding an instructional component facilitates learners’ ability to produce and interpret linguistic action. That is, learners who receive instruction in pragmatics through explicit or implicit teaching, and participate in activities which raise their awareness of communicative actions, outperform those who do not receive instruction (Bardovi-Harlig 2001; Hasler-Barker 2013). Further, many aspects of pragmatics can be taught in the classroom, but teachers need to incorporate elements of pragmatics that are appropriate to the students’ proficiency level. For instance, certain aspects of linguistic action that are not linguistically complex (e.g. greetings, compliments, openings and closings) can be taught at the beginning levels of instruction in the classroom by means of making input salient (e.g. bolding [in written materials] certain elements necessary to make a request), followed by communicative activities that help the learner notice both the form and function of particular expressions used in a specific context. Other concepts in pragmatics that require sophisticated linguistic knowledge can be taught at more advanced levels of proficiency at the discourse level, such as issuing and declining invitations, which require negotiation across multiple turns. Some studies that reviewed the effects of instruction in L2 pragmatics include Bardovi-Harlig (2001), Rose and Kasper (2001), and Rose (2005).
431
Interlanguage Pragmatics 431 Félix-Brasdefer and Cohen (2012) focused on variation in L2 speech act production, assessment of pragmatic knowledge, pragmatic variation, and the teaching of grammar as a communicative resource. Further, Félix-Brasdefer (2006) examined the teaching of speech acts at the discourse level by analysing the organization of turns, speech act sequences (refusals) and mitigation (see also Félix-Brasdefer 2008a, 2008b and Hasler- Barker 2016), and a consciousness-raising approach. Shively (2011) proposed a pedagogical model for teaching how to negotiate requests for service between L2 learners of Spanish and L1 clerks in Spain. Since the ultimate goal of communication is interaction, one alternative for the language teacher is to administer online role-plays in which the learner is presented with a script and an online interlocutor who guides the interaction, as in Félix-Brasdefer and Hasler-Barker’s (2012) pedagogical proposal for examining compliments and compliment responses (http://www.indiana.edu/~discprag/practice_ compliments.html). Overall, ILP should go beyond speech acts as single units of analysis, and instead focus on the analysis of communicative action at the discourse level using theoretically and pedagogically informed models. In this way, we will advance our understanding of both the learning process and the pedagogical value of L2 pragmatics in the classroom.
22.6 Data Collection Methods in Interlanguage Pragmatics Methods in ILP research range from natural data to elicited tasks. Kasper (2000) described three types of methods of data collection: (i) spoken interaction (e.g. authentic discourse, elicited interaction, and role-plays); (ii) methods to obtain written responses (e.g. production questionnaires, multiple choice, and scale response instruments); and (iii) verbal reports (e.g. think-aloud protocols and verbal protocols). Metapragmatic judgement tasks are also used to measure sociopragmatic knowledge. As mentioned in previous research (e.g. Bardovi-Harlig 2012), the selected task to gather elicited data is often motivated by the aim of the research question. With regard to natural data, Wolfson’s (1981) seminal work on naturally occurring speech acts (i.e. compliments and invitations) was among the early studies in the 1980s that showed that speech act data need to be gathered ‘through [direct] observation and participation in a great variety of spontaneously occurring speech situations’ (Wolfson 1981: 9) and speech acts ‘must be observed in naturalistic settings in order for any analysis to be valid’ (p. 7). However, natural data are not error-free, especially when data are collected for the sake of research. Collecting speech act data in naturally occurring settings poses some challenges for the researcher (e.g. Kasper and Dahl 1991; Beebe and Cummings 1996; Kasper 2000; Cohen 2004, 2012; Félix- Brasdefer 2010). These include: (i) the difficulty of controlling sociolinguistic variables such as gender, age, educational level, ethnic group, and social class in comparable situations; (ii) a series of
432
432 J. César Félix-Brasdefer speech acts may be difficult to observe outside of the classroom in learner–NS interactions; (iii) a particular speech act cannot be captured with high enough frequencies for analysis; and (iv) the inability to find NSs of the target language to interact with NNSs in a wide variety of settings. Three types of data collection methods predominate in the ILP literature: production questionnaires (i.e. Written Discourse Completion Tests [WDCTs]), role-plays, and verbal reports. WDCTs represent the most frequently used instrument in ILP and cross-cultural pragmatics research. The instrument elicits experimental (simulated) speech act data under controlled conditions so as to measure offline pragmalinguistic or sociopragmatic knowledge in a non-interactive format. That is, WDCTs measure what the participants know, rather than how they use their ability to interact with an interlocutor. In her study of L2 pragmalinguistic and sociopragmatic development, Flores-Salgado (2011) adapted Rose’s (2000) cartoon oral production task (COPT) to elicit requests and apologies among adult learners of English as an FL in Mexico. One advantage of the COPT instrument is that it provides a visual context for the situation (i.e. pictures representing different social situations) and the data are recorded orally (as an oral DCT). Finally, in order to increase the degree of validity of the WDCT, and by providing rich audiovisual and contextual information in the situation prompt, Schauer (2004) designed the computer-based multimedia elicitation task (MET) to examine ILP development in requests in sixteen scenarios among German learners of English during a year-long SA programme in Great Britain. The MET controls the time and the nature of the audio and visual input, guarantees equal conditions for every participant, elicits oral data, and is delivered by means of a computerized presentation format with visual (photographic images) and audio input (description of the situation). Role-plays offer simulations of communicative encounters in formal and informal settings that elicit interactional data in which two interlocutors assume roles under predefined experimental conditions. Role-plays can be of two types: closed and open (Kasper and Dahl 1991). In closed role-play, the participant responds to a role-play situation without a reply from an interlocutor (as in the oral DCT described in the previous subsection). Open role-play, on the other hand, specifies the actor’s roles, but the course and outcome of the conversation is not predetermined. During a role-play task, participants are often asked to read a situational description and to respond orally as they would in a real situation with an interlocutor in face-to-face interaction. With role-plays, researchers can control for a series of contextual parameters: the situation, the degree of social distance and social power between the interlocutors, the weight of imposition, gender, and age of the participants, learning environment (FL vs SL), and proficiency level. Unlike the WDCT, the (open) role-play technique has the advantage of including interaction in a face-to-face format with another participant. Role-plays are generally audio-or video-recorded. Finally, verbal reporting ‘is a special type of introspection and consists of gathering protocols, or reports, by asking individuals what is going through their minds as they are solving or completing a task’ (Mackey and Gass 2005: 77). This method was grounded in psychology research within the Processing Model that interprets human
433
Interlanguage Pragmatics 433 cognition as information processing with limited capacity, namely, short-term (STM) and long-term memory (LTM) (Ericsson and Simon 1993). Under this framework, the authors indicate that ‘it is assumed that information recently acquired (attended to or heeded) by the central processor is kept in STM, and is directly accessible for further processing (e.g., for producing verbal reports), whereas information from LTM must be first transferred (transferred to STM) before it can be reported’ (Ericsson and Simon 1993: 11). In think-aloud protocols, participants are instructed to think aloud or to voice their thoughts while performing a task (e.g. WDCTs). On the other hand, retrospective reporting consists of verbal reports obtained from the participants immediately after the completion of a task (e.g. role-plays) while much information is still in STM and can be directly reported. Both types of verbal reports have been used to draw inferences during production or after task completion. Verbal reports are instrumental in understanding speech act performance because ‘one may learn what the respondents actually perceived about each situation (e.g., what they perceived about the relative role status of the interlocutors) and how their perceptions influenced their responses’ (Cohen 2004: 321). Verbal reports have been used in ILP to examine pragmatic competence in order to complement data from production questionnaires (e.g. Robinson 1992) and role-plays (e.g. Cohen and Olshtain 1993; Widjaja 1997; Félix- Brasdefer 2004, 2007, 2008a; Woodfield 2012). Since the quantity, quality, frequency, and modality (e.g. oral vs written) of pragmatic input that FL learners are afforded inside and outside of the classroom differs considerably from what learners are exposed to in naturalistic contexts, the design of the instrument needs to be improved to fit the needs of learners in the particular learning environment. Ethically, it should be noted that collecting simulated data in FL and L2 contexts may require approval of the Institutional Review Board (or other group) from the institution where the data are collected for the protection of the rights of human subjects and this information should be reported in the study.
22.7 Future Directions The field of ILP, originally grounded in cross-cultural and intercultural pragmatics, has developed to cover not only language use (NNSs as users) but also learning over time (L2 pragmatics) as a result of incidental learning or pedagogical intervention. Researchers should further our understanding of ILP to cover various dimensions of learners’ pragmatic competence, including not only pragmatic functional knowledge and pragmalinguistic knowledge, but also other areas such as sociocultural knowledge, discourse knowledge, and strategic knowledge (Laughlin, Wain, and Schmidgall 2015). Researchers in ILP need to move forward to an analysis of pragmatic meaning in social interaction with particular attention to speech act sequences (Félix-Brasdefer 2014) in spoken learner–learner or learner–NS interactions. Other topics under-represented in ILP include L2 impoliteness, assessment for the development of interactional
434
434 J. César Félix-Brasdefer competence (Ross and Kasper 2013), the assessment of implicature in the FL classroom and in the target culture, the negotiation of meaning in service encounter interactions in commercial and non-commercial settings (Félix-Brasdefer 2015), and pragmatic variation in ILP (Félix-Brasdefer and Koike 2012). Researchers should also refine methods of data collection to promote the analysis of speaking and social interaction in institutional and non-institutional contexts. The analysis of natural data should be the priority. However, researchers should also consider other methods of data collection to fit the aims of the research question, such as interactive role-plays, metapragmatic judgement tasks, or computer-delivered tasks for the production and comprehension of conventionalized expressions and other pragmatic targets. And given the acquisitional focus of L2 pragmatics, researchers should further our understanding of pragmatic development with both longitudinal and cross-sectional studies to examine development over time, including the effects of pedagogical intervention. Furthermore, researchers in ILP should pay close attention to other factors such as frequency of input, language variety, individual differences (e.g. age, gender, motivation), and delivery of the input in the classroom and/or in the target culture. Finally, researchers focusing on use and learning should adopt theoretical and methodological approaches from an interdisciplinary perspective, using a mixed-method approach, to further our understanding of interactive dimensions of pragmatic competence in multilingual contexts and in a variety of institutional and non-institutional contexts.
435
Chapter 23
C onversation A na lysi s Emanuel A. Schegloff
23.1 Introduction Arguably, conversation analysis (CA) began with observations, claims, and analyses whose proper analytic locus is action formation/recognition.1 Indeed, in one of its earliest published papers, ‘Opening up closings’ (Schegloff and Sacks 1973, but drafted in the summer of 1969), we wrote in the second paragraph (289–290, emphasis supplied): This project is part of a program of work undertaken several years ago to explore the possibility of achieving a naturalistic observational discipline that could deal with the details of social action(s) rigorously, empirically, and formally. For a variety of reasons that need not be spelled out here, our attention has focused on conversational materials; suffice it to say that this is not because of a special interest in language, or any theoretical primacy we accord conversation. Nonetheless, the character of our materials as conversational has attracted our attention to the study of conversation as an activity in its own right, and thereby to the ways in which any actions accomplished in conversation require reference to the properties and organization of conversation for their understanding and analysis, both by participants and by professional investigators.
What sorts of observations had engendered such a ‘program of work’? Here are a few exemplars. On Sacks’ part, there was the observation that (what we would now call) repair initiation targeting a phone-answerer’s self-identification could serve as a device for bypassing the turn slot in which the caller might ‘owe’ a reciprocal self-identification. In the mid-1960s, Sacks was participating in a research programme centred on an emergency ‘hotline’ for suicidal persons (or persons calling about suicidals). One focus of attention 1
In what follows, I have drawn generously on an earlier publication, Schegloff (2006).
436
436 Emanuel A. Schegloff was the issue of getting the names of the callers—needed as evidence for the funding agency that supported the hotline. Call-takers reported that if they could not get the caller’s name at the very beginning of the call, it was highly unlikely that they would get it at all (e.g. at the end of the call). One recurrent form the openings took is exemplified in (1a); the call-taker would self-identify, and that opening line made it appropriate for callers to reciprocate in the very next turn. (1a) Sacks 1992: I.3 (1964) 01 SPC: 02 CAL:
This is Mr Smith may I help you Yes, this is Mr Brown
But, Sacks noted, in calls that made getting an identification problematic, there was often a claim of ‘trouble on the line’, as exemplified in (1b)—in many cases in calls which seemed in a later review to be acoustically quite clear. (1b) Sacks 1992: I.3 (1964) 01 02 03 04
SPC: CAL:→ SPC: CAL:
This is Mr Smith may I help you I can’t hear you This is Mr Smith Smith
Dealing with the trouble (what would much later come to be termed ‘repair initiation’), then, took over the place where a reciprocal self-introduction would ordinarily occur, and subsequent overt efforts to secure the caller’s identity were doomed to failure. This appeared, then, to be a practice for avoiding reciprocal self-identification. And another example. Sacks was permitted to audio-record a series of group therapy sessions for teenagers. In one of these, a new patient named Jim has just arrived, some 45 minutes into the session, and the therapist, Dan, has introduced him to the three other male teenagers in the group (the one female has not come to this session)—Ken, Al, and Roger. After the round of introductions and greetings, the talk goes as in (2): (2) Sacks 1992: I.144 (1965) 01 Ken:→ 02 Rog: 03 Al:
((cough)) We were in an automobile discussion, discussing the psychological motives for drag racing on the streets.
On the face of it, Ken’s turn is (one might say) ‘restarting the talk that was in progress when Jim arrived’—that is what it is doing. Pressed to say more, we might say it is ‘orienting the newcomer to what had been going on, and what this same turn is
437
Conversation Analysis 437 restarting’—that is also something it is doing … or might be doing. Sacks proposed that what we have here is ‘a possible invitation’. Whereas the first two descriptions fit comfortably with what we think a restarting would be or look or sound like, or what orienting a latecomer would be, it does NOT look like what we would ordinarily think an invitation would be like. Here in a nutshell is how Sacks developed his point, in my rendering of it: • a newcomer arrives at an already ongoing interaction; • the pre-present parties can either welcome and absorb the newcomer or exclude or marginalize him/her (both in posture and in the talk); • in the talk, the pre-present parties may abandon the pre-arrival topic in favour of something recipient-designed for the newcomer; or they can continue the talk-in- progress with no special regard to the newcomer; or they can formulate the talk in a fashion designed to convey that it is ill-suited to the newcomer, or in a fashion designed to convey that it is of potential interest and access to the newcomer, thereby in effect inviting him/her to join in; • here the new arrival is known or thought to be new to psychotherapy and therefore is uncertain of how he will be understood; by formulating the talk in which they were engaged as ‘an automobile discussion’ (which in the lingo of the day was quite different from ‘talking about cars’; it was about being ‘a hot-rodder’), Ken in effect is reassuring him that (if he is a hot-rodder), he can talk here as he talks elsewhere and be at ease; hence an invitation; • as it happens, Ken is the ‘poor little rich boy’ in the group, an isolate relative to the other two guys, and he can be understood as here looking for an ally; • that something like this is understood by the others to be going on can be seen in what they do to this turn-so-far by extending it in what Sacks termed a collaborative construction: Roger characterizes the talk in which they are engaged as discussing psychological motives—just what the newcomer may be anxious about; and Al extends this anxiety into the very safe haven that Ken had suggested when he finishes the now-collaborative utterance by applying it to drag racing on the streets. By recalling this very early piece of analysis, I want to alert readers to a later tack in our discussion which will be: (i) discouraging our undertaking from starting with category terms for actions that particular cases ‘transparently’ exemplify, (ii) encouraging our asking about any given target turn or turn-constructional unit (henceforth ‘TCU’) or component of a TCU what IT is doing HERE, at this juncture of this interaction (which must itself therefore be characterized), and (iii) actually, what it is POSSIBLY doing here. These are three of the topics I mean to touch on as prolegomena to analysing action(s). (As for starting points in Sacks’ work, there are, of course, two thick volumes of his Lectures on Conversation, containing many others subsequently.)
438
438 Emanuel A. Schegloff On my part there was the treatment of the first utterance in telephone calls to the police (as in (3))—as ‘self-identifications’. (3) Schegloff 1968: 1079
01 Dis:→ Police Desk
But then I heard myself referring to them as ‘answering the phone’ in that way, and I realized that while ‘self-identifications’ was not wrong, it was not quite right either; if they were answering, what were they answering? And that led to seeing that while ‘Police Desk’ may have been the first utterance, it was not the first action in these conversations; the first actions were the ‘summonses’ conveyed by the mechanically expressed telephone rings, and self-identification was just one way of ‘answering’ them. Another way of answering—overwhelmingly common in the USA at that time outside organizational contexts—was ‘hello’, which, it turned out, was in these circumstances NOT the ‘greeting term’ it is commonly taken to be. Whatever they might be called, what they were actually doing was not a greeting, but instead was a response to a summons. Or, to mention one last starting point: there was the demonstration that what appeared to be a question, indeed one of the forms that we now term an ‘other-initiation of repair’ (Schegloff, Jefferson, and Sacks 1977), would more correctly be analysed as a display of alignment with its recipient. This exchange in (4) is taken from a radio call-in programme in New York City in the 1960s. The caller (‘Cal’) is a high-school student who is complaining to the Host (‘Hst’) about his History teacher, with whom he had a difference of opinion about the Vietnam War—a ‘difference’ which had escalated to a disagreement about the proper function of government. (4) Schegloff 1984: 28 (1976) 01 02 03 04 05 06 07 08 09 10 11 12 13 14 15 16
Cal:
Hst: Cal: Hst:→ Cal:→ Hst:→ Cal:→
An’s-an ( ) we were discussing, it tur- it comes down, he s-he says, I-I-you’ve talked with thi-si-i-about this many times. I said, it come down t’this:= =Our main difference: I feel that a government, i-the main thing, is-th-the purpose a’the government, is, what is best for the country. Mmhmm He says, governments, an’ you know he keeps-he talks about governments, they sh-the thing that they sh’d do is what’s right or wrong. For whom. Well he says-[he- [By what standard That’s what-that’s exactly what I mean. He s- but he says …
439
Conversation Analysis 439 The Caller initially understands the Host’s intervention at line 12 to be in search of a fuller account of the teacher’s position (and potentially beginning to align with him—the teacher); but when the Host interrupts to do it again but with a different ‘topic inquiry’ at line 14, the Caller recognizes that these questions are designed to be challenges to the (just articulated) teacher’s position, and thereby alignments with the Caller. What these four pieces of work shared as hints on how to proceed were several features, of which I’ll mention just four: 1. An ‘obvious’ vernacular (or even traditional academic) understanding was either demonstrably incorrect, or, at the very least, partial; 2. A proper understanding of what the object of analysis was doing involved attention not only to its composition or construction, but also to its position or location, which itself required analysis to locate and formulate the terms of its relevant characterization; 3. Such analysis required coming to terms with the various considerations to which parties to the interaction were demonstrably oriented, involving at least position in the conversation, the course of action in which any next turn would figure and to which it would be taken to respond, the turn as the home for spates of talk which could be either understood or misunderstood, respected or violated, and so forth. And 4. That a lot of work would be required to get at least a sketch of the several organizations of conduct that appeared to constitute the host environment for the actions we had tried to understand. A fair amount of time has been—and is still being—spent on this last requirement, and I think we have reasonable initial understandings of how turns at talk are constructed and understood, and how opportunities to construct them get distributed among parties to various settings of talk-in-interaction; how troubles in speaking, hearing, and understanding the talk are dealt with; how sequences of these turns are organized to effect recognizable courses of action; how these sequences aggregate to compose occasions of interaction—sometimes bounded and limited from the outset and requiring pretty much continuously sustained talking, sometimes being less so and composed by what we called ‘continuing states of incipient talk’, and finally, and to a very uneven degree, how the components of these turns that constitute these sequences in these occasions of interaction get selected, combined, and deployed to embody the actions that had launched us on this expedition in the first place. So now that we have some understanding of the resources out of which ‘doings’ are fashioned, we return to the actions that got us into this in the first place, only to realize that the terms have now changed—that we need to make explicit the bearing of what we think we have learned on going to work on action and actions, that is, not single or singular actions but simultaneously multiple actions.
440
440 Emanuel A. Schegloff
23.2 Generic Problems for Talk-in- Interaction and Practised Solutions for them Although it is almost certainly the case that many important organizational problems of talk-in-interaction and their solutions are as yet unknown, let alone understood, it appears that the following ones will have a continuing claim on researchers’ attention.
23.2.1 The ‘turn-taking’ problem Who should talk or move/act next and when should they do so? How does this affect the construction and understanding of the turns or acts themselves? So far it seems to be the case that wherever investigators have looked carefully, talk- in-interaction is organized to be done one speaker at a time.2 Achieving and maintaining such a state of talk may prompt the invocation of conventionalized arrangements like a Chair to allocate the turns, or mapping the order of turn allocation onto ordered features of the candidate participants such as relative status (Albert 1964). But the first of these marks the setting as institutionally or ceremonially distinct from ‘ordinary talk’, and the latter engenders a range of problems that make it virtually unsustainable as a general organization of interaction. What is most fundamentally at stake in ‘turn- taking’ is not politeness or civility, but the very possibility of coordinated courses of action between the participants (allowing, for example, for initiative and response)— very high stakes indeed. Even with just two participants, achieving one at a time poses a problem of coordination if the talk is to be without recurrent substantial silences and overlaps: how to coordinate the ending of one speaker and the starting up by another. If there are more than two ‘ratified participants’ (Goffman 1963), there is the additional issue of having at least one of the current non-speakers, and not more than one of the current non- speakers, start up on completion of the current speaker’s turn. One can imagine quite
2 Two sorts of exception should be mentioned here. One involves the claim that there is a place where talk-in-interaction is not so organized, as in Reisman’s (1974) claim for ‘contrapuntal conversation’ in Antigua; Sidnell (2001) casts considerable doubt on Reisman’s account. The other involves specifications of where in conversation the ‘one at a time’ claim does not hold, e.g. Lerner (2003) on ‘choral co-production’ or Duranti (1997b) on ‘polyphonic discourse’; here the phenomenon being described is virtually defined as an object of interest by its departure from the otherwise default organization of talk. Work on ‘overlapping talk’ (e.g. Jefferson 1984, 1986, 2004; Schegloff 1987b, 2000, 2001) locates the topic by reference to its problematic relation to the default one-at-a-time organization.
441
Conversation Analysis 441 a variety of putative solutions to these problems of coordination, but none of them can be reconciled with the data of actual, naturally occurring ordinary conversation (Schegloff 2000). The ‘Simplest systematics for turn-taking’ paper (Sacks et al. 1974) sketches an organization of practices that, on the whole, seems to work. It describes units and practices for constructing turns at talk, practices for allocating turns at talk, and a set of practices which integrates the two. So far, this account has appeared to work across quite a wide range of settings, languages, and cultures, and departures from interactional formats familiar to Western industrialized nations appear to involve what might be termed ‘differences in the values of variables’, rather than differences in the underlying organization of practices. To give one example briefly, there may be differences between cultures or subcultures in what the unmarked value of a silence between the end of one turn and the start of a next should be. Leaving less than the normative ‘beat’ of silence or more than that can engender inferences among parties to the conversation; starting a next turn ‘early’ or starting a next turn ‘late’ are ways of doing things in interaction, and conversation between people from different cultural settings can find themselves misfiring with one another. For example, one difference often remarked on by urban, metropolitan people about rural or indigenous people is that the latter seem to be dull-witted and somewhat hostile; comments range from Marx on the ‘idiocy of the rural classes’ to Ron and Suzanne Scollon’s work (1981) on the relation between migrants from the ‘lower 48’ states in the US and the indigenous peoples of Alaska. Having asked them a question, the urbanites—or should I say urbane-ites— find themselves not getting a timely reply and sense resistance, non-understanding, non-forthcomingness, etc. Often they break what they perceive as ‘the silence’ that greeted their question with a follow-up question, which may be taken by their interlocutor to exemplify the high-pressure aggressiveness of ‘city slickers’. But what differs between them is not that their turn-taking practices are different or differently organized, but the way they ‘reckon’ the invisible, normative beat between one turn and the next. I have, of course, just pointed at the organization of turn-taking; an account of what that organization is, and how it works, will have to be sought out in the by-now substantial literature addressed to those matters (cf. especially Lerner 2003).
23.2.2 The ‘sequence-organizational’ problem How are successive turns or actions formed up to be ‘coherent’ with the prior one (or some prior one) and constitute a ‘course of action’, and what is the nature of that coherence? The most common way researchers have addressed actual spates of talk has been to ask what it is about, and how movement from one ‘topic’ to another occurs, and what it reveals about the intentions and meanings being conveyed by the speaker
442
442 Emanuel A. Schegloff or the several participants. Talking about things—‘doing topic talk’—is surely one observable feature of talk-in-interaction. But having framed it as something participants do should trigger the further observation that that is only one of the things people do in talk-in-interaction. We would do well to open inquiry to the full range of things that people do in their talking in interaction—asking, requesting, inviting, offering, complaining, reporting, answering, agreeing, disagreeing, accepting, rejecting, assessing, etc. Indeed, doing topic talk is itself largely composed of such doings—telling, agreeing, disagreeing, assessing, rejecting, etc. Proceeding in this way treats action and courses of action as the more general tack, and doing topic talk as one of its varieties. If we ask how actions and courses of action get organized in talk-in-interaction, it turns out that there are a few kernel forms of organization that appear to supply the formal framework within which the context-specific actual actions and trajectories of action are shaped. By far the most common and consequential is the one we call ‘adjacency-pair-based’ (Schegloff and Sacks 1973; Sacks 1995: II.521–569; Schegloff 2007). The simplest and minimal form of a sequence is two turns long: the first initiating some kind of action trajectory—such as requesting, complaining, announcing, and the like, and the second responding to that action in either a compliant or aligning way (granting, remedying, assessing, and the like, respectively) or in a disaligning or non-compliant way (rejecting, disagreeing, claiming prior knowledge, and the like, respectively). Around and inside such ‘simple’ pairs of actions, quite elaborate expansions can be fashioned by the participants. There are, for example, expansions before the first part of such a pair, such as ‘pre-announcements’ (‘Didju hear who’s coming?’), ‘pre- invitations’ (‘Are you doing anything this weekend?’), and the like. Or, to cite actual data of a pre-invitation: (5) CG,1 (Nelson is the caller; Clara is called to the phone) 01 02 03 04 05 06 07 08
Cla: Nel: Cla: Nel:-> Cla:-> Nel: Cla: Nel:
Hello Hi. Hi. Whatcha doin’. Not much. Y’wanna drink? Yeah. Okay.
And of a pre-announcement: (6) Terasaki 2004: 195 (1976: 43) 01 Jim:-> 02 Gin:-> 03 Jim:
Y’wanna know who I got stoned with a few(hh) weeks ago? hh! Who. Mary Carter ‘n her boy(hh)frie(hh)nd. hh.
443
Conversation Analysis 443 Notice that these themselves make a response relevant, and so themselves constitute an adjacency pair, and can therefore themselves be expanded (for example, ‘Hey Steve’, ‘Yeah?’, ‘Didju hear who pulled out of the conference?’, ‘No, who?’). And there can be expansions after the first action/turn in an adjacency pair and before the responding second part—an inserted sequence. For example: (7) Schegloff et al. 1977: 368 01 02 03 04 05
Bel:->Fb Mar:->Fi Bel:->Si Mar:->Sb
Was last night the first time you met Missiz Kelly? (1.0) Met whom? Missiz Kelly. Yes.
Again, notice that if a first-pair part is followed not by an action/turn which could be its second-pair part, then what occurs in its place is itself a first-pair part and requires a response, so it too is an adjacency pair and it too can get expanded. And after the response to the initiating action/turn there can be further talk that clearly is extending that trajectory of action. Sometimes that can be a single turn which does not make a response to it relevant next, as at lines 3 and 8 in the following specimen, which has two such sequences. (8) HG, 16:25–33 01 02 03 04 05 06 07 08 09
Nan: Hyl: Nan:-> Nan: Hyl: Nan:->
=˙hhh Dz he av iz own apa:rt[mint?] [˙hhhh] Yea:h,= =Oh:, (1.0) How didju git iz number, (·) I(h) (·) c(h)alled infermation'n San Fr’ncissc(h)[uh! [Oh::::. (·)
But it can also be something that does make a response to it relevant next; so it too is itself an adjacency pair and can take the kinds of expansions I have been sketching here. (9) Connie and Dee, 9 01 02 03 04 05 06
Dee: Con: Dee:-> Con:-> Dee:->> Dee:->>
Well who’r you workin for. ˙hhh Well I’m working through the Amfat Corporation. The who? Amfah Corpora[tion. T’s a holding company. [Oh Yeah
444
444 Emanuel A. Schegloff Note here that the question/answer sequence at lines 1–2 is expanded after the answer by another at lines 3–4 (addressing a hearing/understanding problem), and that the latter is expanded by a single turn expansion, first at line 5 (where it is caught in overlap) and then again at line 6. I hope that it is clear that what started as a simple two-turn/action sequence can be a framework which ‘carries’ an extensive stretch of talk.3 There are some deep connections between what are nonetheless largely autonomous organizations of practice—the organization of turn-taking and the organization of action sequences. Just as interaction cannot do without practices for allocating opportunities to participate and practices for constraining the size of those opportunities—i.e. an organization of turn-taking—so it cannot do without an organization of practices for using those opportunities to fashion coherent and sustained trajectories or courses of action–sequence organization.
23.2.3 The ‘trouble’ problem How to deal with trouble in speaking, hearing, and/or understanding the talk or other conduct such that the interaction does not freeze in place, that intersubjectivity is maintained or restored, and that the turn and sequence and activity can progress to possible completion. If the organization of talk-in-interaction supplies the basic infrastructure through which the institutions and social organization of quotidian life are implemented, it had better to be pretty reliable, and to have ways of getting righted if beset by trouble. And so it is. Talk-in-interaction is as prone as any organization is to transient problems of integration and execution; speakers cannot find the word they want, find that they have started telling about something that needs something else to be told first, hear that they articulated just the opposite word from the one they are after, find that another is talking at the same time as they are, etc. And talk-in-interaction is as vulnerable as any activity is to interference from altogether unrelated events in its environment—overflight by airplanes, an outburst of traffic noise, or other ambient noise that interferes with their recipient’s ability to hear, etc. For such inescapable contingencies there is an organization of practices for dealing with trouble or problems in speaking, hearing, and understanding the talk. It turns out that this organization—which we term an organization of repair—is extraordinarily effective at allowing the parties to locate and diagnose the trouble and, in virtually all cases, to deal with it quickly and successfully. The organization of repair differentiates between repair initiated and carried through by the speaker of the trouble source on the one hand, and other participants in the interaction on the other. The practices of repair are focused in a sharply defined window of opportunity in which virtually all repair that is initiated is launched (Schegloff et al. 3
For an analysis of quite an elaborate sequence—125 lines of transcript composing a single sequence—cf. Schegloff (1990).
445
Conversation Analysis 445 1977). This ‘repair initiation opportunity space’ begins in the same turn—indeed, in the same turn-constructional unit—in which the trouble source occurred and extends to the next turn by that speaker.4 The consequence is that the initial opportunity to initiate repair falls to the speaker of the trouble source, and a very large proportion of repairs are addressed and resolved in the same turn, and same turn-constructional unit, in which the trouble source occurred (‘same-turn repair’), or in its immediate aftermath (‘transition space repair’). These largely involve troubles in speaking, but can also be directed to anticipatable problems for recipients—problems of hearing and/or understanding. The ‘preferences for self-initiation of repair and self-repair’ have as one of their manifestations that recipients of talk which is for them problematic regularly withhold initiating repair in next turn to allow the trouble-source speakers an additional opportunity to themselves initiate repair. If they do not do so, the next opportunity for addressing the trouble falls to recipients—ordinarily in the next turn. Finally (for our purposes), a speaker may have produced a turn at talk and had a recipient reply to it with no indication of trouble, only to find that the reply displayed what is to the prior speaker a problematic understanding of that turn. Then, in the turn following the one which has displayed the problematic understanding, the speaker of what now turns out to have been a trouble-source turn may take the next turn to address that problematic understanding (the canonical form being ‘No, I didn’t mean X, I meant Y’; cf. Schegloff 1992). As the talk develops through the repair space, there are fewer and fewer troubles or repairables that get addressed. Most are dealt with in the same or next turn, and these range from production problems (such as word selection, word retrieval, articulation, management of prosody, etc.) and reception problems (hearing and understanding of inappropriately selected usages, such as person reference terms, technical terms, complicated syntax, etc.) to issues of intersubjectivity and strategic issues of delicateness. It is hard to say which are more important: without virtually immediate resolution of the production and reception problems, the interaction can be stalled indefinitely with unpredictable consequences; without ways of spotting departures from intersubjectivity and restoring it, the shared reality of the moment is lost, again with unpredictable consequences. It is hard to imagine a society or culture whose organization of interaction does not include a repair component, and one that works more or less like the one I have sketched. We know that details may vary in ways linked to the linguistic structure of the language spoken—either its grammatical structure (cf., for example, Fox, Hayashi, and Jasperson 1996) or its phonological inventory (cf., for example, Schegloff 1987b). But the structure of the repair space and the terms of its differentiation between same and other repair are likely not to vary. For, among its other virtues, it is the availability of the practices of repair that allows us to make do with the natural languages that 4 In fact, the way repair is organized can have the consequence that it is sometimes initiated at a greater ‘distance’ from the trouble while still being within the boundaries that can here be only roughly characterized. For an account of this, cf. Schegloff (1992).
446
446 Emanuel A. Schegloff philosophers and logicians have long shown to be so inadequate as to require the invention of artificial, formal ones. It is repair that allows our language use not only to allow, but to exploit many of the features that have been treated as its faults—ambiguity, polysemy, contradiction, etc. Designed not for automatic parsers but for sentient beings, should these usages not be transparently solvable, the practices of repair are available to get solutions (Schegloff 1989).
23.2.4 The word selection problem How do the components that get selected as the elements of a turn get selected, and how does that selection inform and shape the understanding achieved by the turn’s recipients? Turns are composed of turn- constructional units (TCUs)— sentential, clausal, phrasal, and lexical, in English and a great many other languages.5 But of what are turn- constructional units composed? I have waffled on this question in referring to this generic organization as ‘word selection’. That is a vernacular way of putting it, or perhaps a linguistic or psycholinguistic one for some varieties of those disciplines. And sometimes it is a relevant way of putting it in conversation-analytic work. But here I want to call attention to the interactional practices which are only incidentally lexical or about words. These are practices of referring, or describing, or—perhaps most generally— practices of formulating. In talk-in-interaction, participants formulate/refer to persons (Sacks 1972a, 1972b; Sacks and Schegloff 1979; Schegloff 1996b), places (Schegloff 1972), time, actions, and so on. It turns out that understanding how talk gets to be composed the way it is ill-served by treating the usages that are employed as having been employed because they are correct. The person writing this (and that is one formulation already) is not only a sociologist; he is also (as the pronoun inescapably revealed) male, Californian, Jewish, etc. The place I am writing is not only my office, it is in Haines Hall, at UCLA, in Los Angeles, on the west side, in the USA, etc. And although I already formulated my current activity as ‘writing this’, it is also typing, rushing to finish before a student arrives, etc. That is, ‘correctness’ won’t do as the grounds for populating utterances with this or that formulation, because there are always other formulations that are equally correct. What is central is relevance (not, obviously, in the sense of Sperber and Wilson 1986)—what action or actions the speaker is designing the utterance to embody. Consider, for example, this bit of interaction. Hyla has invited Nancy (the two of them were college juniors in the early 1970s) earlier in the day to go to the theatre that night to see a performance of the play ‘The Dark at the Top of the Stairs’, and they are 5 To conserve time and space, I have omitted the treatment of practices of turn construction as a generic organization in talk-in-interaction, though it has a key role in the organization of turn-taking, on the one hand, and the organization of sequences, on the other (cf. Schegloff 1996a). The final section of this chapter briefly addresses this decision.
447
Conversation Analysis 447 talking on the phone in the late afternoon about that upcoming event (among other things). After a brief exchange about when they will meet, Nancy asks: (10) Hyla & Nancy, 5:07–39 01 02 03 04 05 06 07 08 09 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33
Nan: Hyl: Hyl: Hyl: Nan: Hyl: Nan: Hyl: Nan: Nan: Hyl: Nan: Hyl: Nan: Hyl: Nan: Hyl: Hyl: Nan: Hyl: Nan: Hyl: Nan:
[How did]ju hear about it from the pape[r? [˙hhhhh I sa:w- (0.4) A'right when was:(it,)/(this,) (0.3) The week before my birthda:[y,] [Ye] a[:h, [I wz looking in the Calendar section en there was u:n, (·) un a:d yihknow a liddle:: u- thi:ng, ˙hh[hh [Uh hu:h,= =At-th'-th'theater's called the Met Theater it's on Point[setta.] [The Me]:t, (·) I never heard of i[t. [I hadn't either.˙hhh But anyways, .- e n theh the moo-thing wz th'[Dark e'th' [Top a'th' [Stai[:rs.] [Mm-h]m[:, [En I nearly wen'chhrazy cz I[: I : lo:ve ] that] mo:vie.] [y:Yeah I kn]ow y]ou lo:ve] tha::t.= =s:So::, ˙hh an' like the first sho:w,= =M[m hmm, ] [wz g'nna] be:, (·) on my birthday.= =Uh hu[h, ] [I'm] go'[n awhh whould hI love- [(So-) (·) yihknow fer Sim tuh [take me tuh that.] [Y a y u : : h , ]
I want to call attention here to only two bits of Hyla’s responsive talk starting at line 02: the time formulation ‘the week before my birthday’ (line 06), and (at lines 08–09) the activity formulation ‘I was looking in the Calendar section’ (an ethnographic note: the ‘Calendar’ section of the Los Angeles Times is the ‘culture and entertainment’ section).
448
448 Emanuel A. Schegloff First note that Hyla conducts an out-loud search for ‘when it was’; (lines 04–06) she is taking care with this time formulation. There are, of course, many other ways of referring to the time in question: how many weeks ago; which week of the month; the date; etc. She chooses ‘the week before my birthday’. And now (at lines 08–09) ‘I was looking in the Calendar section’: not ‘reading the paper’; not ‘looking at the Calendar section’; not the ‘I saw’ with which she had initially begun (at line 02), etc. Putting the two together, she is describing, she is ‘doing’—‘I was looking for what to do on my birthday’. There is not the room to expand on this here, other than to register the theme I mean to be putting before you. In turns at talk that make up sequences of actions, the elements of the talk are selected and deployed to accomplish actions and to do so recognizably; and recipients attend the talk to find what the speaker is doing by saying it in those words, in that way. Using ‘words’ or ‘usages’ or ‘formulations’ is a generic organization of practices for talk-in-interaction because that talk is designed to do things, things which fit with other things in the talk—most often the just preceding ones. Talk- in-interaction is about constructing actions and trajectories of actions—which is why it does not reduce to language, and why a pragmatics that does not attend to the sequential organization of actions is at risk of aridity.
23.2.5 The overall structural organization problem How does the overall structural organization of an occasion of interaction get structured, what are those structures, and how does placement in the overall structural organization inform the construction and understanding of the talk and other conduct as turns, as sequences of actions, etc.? Some actions are positioned not with respect to turns or sequences (though they are done in turns and sequences) or the repair space, but by reference to the occasion of interaction as a unit with its own organization. Greetings and ‘goodbyes’ are the most obvious exemplars, being positioned at the beginning and ending of interactional occasions, respectively. Less obvious, perhaps, is that greetings are just one of a number of action sequence types that may compose an opening phase of an interaction (Schegloff 1986), and ‘goodbyes’ are the last of a number of components that make up a closing section of an interaction. What happens in between can take either of two forms (as far as we know now)—a state of continuously sustained talk and what we can call ‘a continuing state of incipient talk’ (Schegloff and Sacks 1973). The latter term is meant to refer to settings in which the parties talk for a while and then lapse into silence (silence which does not prompt a closing of the interactional occasion), at any point in which the talk may start up again. Characteristic settings in contemporary industrial societies might be families or roommates in the living room in the evening, occupants of a car in a carpool or a long journey, seatmates on an airplane, diners at table, co-workers at a workbench, etc. In some societies, this may be the default organization of virtually all talk-in-interaction.
449
Conversation Analysis 449 Although greetings and ‘goodbyes’ are pretty much tied to their positions in the overall structural organization, other types of action may take on a distinctive character depending on where in the overall structural organization of a conversation they occur. Some types of action are commonly withheld from occurrence early in a conversation; ‘requests’ are a case in point. Doing a request early in the organization of an interaction can be a way of marking its urgency, or some other feature known to be recognizable to the recipient(s). On the other hand, many kinds of ‘noticings’ are ordinarily meant to occur as soon as possible after detectability. Withholding them from early enactment can be taken as failing to notice them or as treating the noticeable as negatively valenced. The generic character of the overall structural organization of the unit ‘a single conversation’ consists straightforwardly in its provision of the practices for launching and closing episodes of interaction with the commitments of attention that they place on their participants. If talk-in-interaction is going on, the parties will find themselves to be someplace in it by reference to this order of organization.
23.3 Upshot If we understand ‘pragmatics’ to be addressed to what a bit of conduct is doing (in contrast to semantics’ address to what it is ‘meaning’), then it will not do to examine the bit of conduct alone—shorn of what preceded it and what it projects as possible ‘next’s. It will not do for serious inquiry because it is not so done by the participants in interaction. For recipients of an utterance register what it is doing by reference to what has preceded (including silence), and what it might be possibly projecting for the moments to come. Indeed, the utterance itself most often gives evidence in this respect, and has been designed to do so. The consequence is that what an utterance (or any part of it) is doing is woven into—and shaped by—the other orders of organization I have meant to call attention to here: spates of turns at talking, sequences of actions, dealing with trouble in production or reception, selection among alternative forms of reference and description, and place in the structure of the interactional occasion. So far, the stance taken up some forty years ago and reproduced in the first paragraph of this chapter appears to be robust and worthy of being taken seriously by those seriously committed to developing an empirically grounded account of humans’ action in interaction.
450
451
Pa rt V
I N T E R FAC E S
452
453
Chapter 24
Pragmati c s a nd Sem ant i c s Robyn Carston
24.1 Introduction: Pragmatics First There was a time when pragmatics was viewed as the ‘wastebasket’ of linguistics, a bin for dumping whatever recalcitrant bits of utterance meaning could not be accommodated by the formal methods of syntax and semantics.1 Happily, things have moved on since then and a more systematic approach has been taken to pragmatic phenomena (the contents of the wastebasket); for instance, a distinction between two kinds of speaker meaning has been made (that is, between what is explicitly communicated and what is implicitly communicated) and different kinds of pragmatic contribution to each of these levels have been investigated (e.g. disambiguation, indexical reference and other saturation processes, modulation of word meanings in context, scalar implicatures and other kinds of implicature, descriptive versus interpretive uses of language, ad hoc concept construction, the derivation of attitudinal meaning, and more). Nevertheless, the assumption that semantics (context-invariant sentence meaning) is somehow primary and that we only turn to pragmatics when it seems that a semantic account cannot be given still tends to prevail. In this chapter, I will take a different approach and, following the (somewhat unusual) order of the conjuncts in the title of this chapter, will attempt to take a resolutely ‘pragmatics first’ approach, arguing for the logical and temporal priority of pragmatics from three perspectives: communicative, developmental, and evolutionary.2 1
It was Bar-Hillel (1971) who coined the term ‘pragmatics wastebasket’, reflecting the attitude of the time. More than 25 years later, Bach (1997: 36) wrote: ‘In linguistics, the category of pragmatics has served mainly as a bin for disposing of phenomena that would otherwise be the business of semantics (as part of grammar) to explain. Relegating such phenomena to pragmatics freed linguistic theory, already becoming more and more complex, of numerous additional complications.’ 2 For a wide-ranging survey of how the relation between pragmatics and semantics has been construed from the 1930s to the present day, see Recanati (2004b).
454
454 Robyn Carston As discussed here, pragmatics is the study of the human capacity for ostensive- inferential communication: that is, the ability to produce and comprehend acts of overt communication, in particular verbal utterances, but also certain non-verbal behaviours, including pointing, miming, and other ostensive bodily gestures. What distinguishes this sort of communicative behaviour from other kinds of intentional behaviour is the type of intention that it manifests: a complex higher-level intention to make evident to an addressee the intention to make some thought(s) manifest to him. I will refer to this as the ‘communicative intention’ throughout.3 The domain of pragmatics is often given as ‘speaker meaning’, but we can talk of it more broadly as ‘communicator’s meaning’, thus taking in non-verbal and multimodal communication too.4 Considered in relation to this view of pragmatics, there are two quite distinct domains that are called ‘semantics’: (i) the meaning contributed by the linguistic expressions used by someone communicating verbally; (ii) the content of the thoughts communicated. The relations among the three components (linguistic semantics, pragmatics, and thought semantics) will emerge as we progress, but, in brief, the picture is as follows. Thoughts (or propositions) are the primary carriers of truth-conditional content (that is, they are truth-evaluable representations of the world); what communicators intend to convey by their ostensive acts (e.g. utterances) are thoughts, and when they use language, the meaning of the linguistic expressions they employ provides rich evidence of the thoughts they intend to communicate but inevitably falls short of encoding those thoughts (fully truth-conditional contents).5 This gap between the linguistically encoded meaning (‘semantics’, in the first sense) and the thoughts communicated arises not only for implicitly communicated content (implicatures), but also at the level of what is explicitly communicated (explicature), as in the following example: (1) She’s had enough.
3
In relevance theory, a communicative intention is defined as an intention ‘to make it mutually manifest to audience and communicator that the communicator has a particular informative intention’, and an informative intention is defined as an intention ‘to make manifest or more manifest to the audience a set of assumptions [thoughts (RC)]’ (Sperber and Wilson 1986: 54–64). This is a fairly major modification of the original pioneering account of speaker meaning given by Grice (1957, 1989a) in terms of reflexive intentions (see also Bach and Harnish 1979). Other theorists focus more on the communicator’s manifesting an intention that the addressee should jointly attend with her to some information (Tomasello 2008). For the purposes of this chapter what these definitions have in common is more important than their differences: they all aim to capture the overtness of the kind of communication at issue and they all involve higher-order theory of mind (that is, several levels of meta-representation of mental state representations). 4 For discussion of the communicative role of facial expressions, affective tones of voice, and manual gestures, when used alone or together with language, see Wharton (2009) and Kendon (2014). 5 The cognitively based view of pragmatics and semantics that I present here is largely derived from work in relevance theory (Sperber and Wilson 1986, 2002; Carston 1999, 2002, 2004; Wilson and Sperber 2004, 2012b; Wilson this volume), but its broad outlines could be compatible with alternative views on the specifics of the pragmatic principles or maxims that guide utterance interpretation.
455
Pragmatics and Semantics 455 We can easily envisage a range of different discourses and extralinguistic contexts in which this sentence might be uttered by a speaker intending to communicate not only any number of very different implicatures (e.g. that she needn’t eat any more; that she’s going to leave her husband; that she should get some rest, and so on), but also any number of very different explicatures, that is, explicitly communicated thoughts that are built out of the linguistically encoded meaning: (2) a. Mary has eaten enough of her dinner to satisfy her mother’s wishes. b. Karen has drunk enough fruit juice to reach the limit of her daily sugar intake. c. Jane has endured enough bad treatment from her husband to the point of being unable to take any more. d. Rachel has worked hard enough (and long enough) to be ready now to retire. e. … The key pragmatic contributions here are the assignment of a referent to the pronoun ‘she’, the specific interpretation of the very general verb ‘have’, and the completion of ‘enough’ (of what? to what end?). This is a very ordinary conversational use of a sentence which, I hope, is sufficient to provide some initial grounding to the claim that, as interpreters of utterances addressed to us, we inevitably employ pragmatic inference (constrained by decoded linguistic semantic meaning) in identifying the proposition the speaker is explicitly communicating. The emphasis in this chapter is on pragmatics as a human cognitive capacity and so on the study of pragmatics as falling within the cognitive sciences, from which it follows that pragmatic theories should be responsive to relevant empirical research on human cognition: its architecture and its evolution, the nature and time course of the processes responsible for understanding acts of ostensive communication, and the development of the relevant cognitive capacities in young children. So, in the next section, I discuss the human pragmatic and linguistic semantic capacities from the following three perspectives: first, the architecture of the cognitive systems responsible for utterance interpretation; second, the evolutionary emergence of pragmatics and language (hence of linguistic semantics) in human cognition; third, the development of communicative competence (pragmatic and semantic) in the child. However, much work in modern pragmatics has its origins in the philosophy of language, where it was inevitably viewed in relation to semantics construed as concerning truth-conditional contents. So, in section 24.3, I present some of this work and its relation to a more cognitively oriented approach to pragmatics and semantics. I conclude, in section 24.4, with a short discussion of the more fundamental kind of semantics, the semantics of thoughts (whether communicated or not), which is entirely independent of pragmatics, and is the proper domain of truth-conditional content.
456
456 Robyn Carston
24.2 Pragmatics and Semantics in Cognitive Science 24.2.1 Pragmatics, semantics, and cognitive architecture From the perspective of human cognitive architecture, the pragmatics–semantics interface is the point where the meaning or content provided by our linguistic knowledge interacts with the inferential capacities that we bring to bear on the interpretation of ostensive communicative acts. There are a range of possibilities about how the two systems at this interface interact: they might be rigidly separated and sequential in their operation; they might be distinct systems in so far as they deploy their own specific procedures but nevertheless operate in parallel; they might not be distinguishable in any interesting sense and instead constitute a single general interpretive system that uses information from any source (perception, language, memory) as it becomes available. The idea that our cognitive capacities are significantly modular in their architecture was brought to prominence by Jerry Fodor (1983). On his construal, a modular system is one which applies to a limited specific stimulus domain and is encapsulated from top-down expectations and utilities of the overall cognitive system, thus being both rigid in its operations and also having a kind of objectivity in representing the input stimulus (uncontaminated by beliefs, desires, hopes). In his view, only the peripheral systems, that is, those systems that present aspects of the outside world to our central thought processes, are modular. So, while our various perceptual systems and the linguistic parser are modules, the central conceptual systems, whose job it is to reach well-founded decisions about what to believe and what to do, have to be highly context- sensitive (responsive to relevant top-down information) and so cannot be modular. Interpreting utterances and other communicative acts is clearly one such context- sensitive process and so, on this Fodorian view, our pragmatic capacities cannot but be non-modular. However, pragmatic processes seem to have at least some of the characteristics of an autonomous mental system. Ever since Grice, it has been widely assumed that overtly communicative behaviour is interpreted in accordance with maxims or principles that do not apply more generally (to the interpretation of other kinds of purposive behaviour). For instance, Gricean maxims of quantity and quality, and their neo-Gricean modifications (e.g. Horn 1984; Levinson 2000), apply only to the domain of rational communicative behaviour, specifically linguistic utterances; no one would expect the information we might gain from watching someone’s intentional but non- communicative behaviour (e.g. cooking a meal, packing a suitcase) to meet the standards of informativeness, relevance, and truthfulness we expect when someone overtly requires us to direct our attention to the stimulus he has produced as is the case in overt
457
Pragmatics and Semantics 457 communication.6 Furthermore, the processes of utterance understanding are spontaneous, automatic, and fast—we can’t help but infer a speaker’s meaning when an utterance is directed at us (and often even when it is not). Sperber and Wilson (2002) have taken these observations seriously and proposed that pragmatics is indeed a modular mental system, albeit not quite in the Fodorian sense in which input systems (perception and language) are modular. The claim that pragmatics is modular has to be situated within a broader view of the whole mind as massively modular (Sperber 1994a, 2005; Carruthers 2006). A driving consideration here comes from ideas in evolutionary psychology (see, for instance, Cosmides and Tooby 1994): just as natural selection favours the development of specific organs (eyes, ears, hearts, livers) to carry out specific survival-enhancing biological processes, so it favours specialized solutions to the specific cognitive problems the organism encounters in its environment—that is, mechanisms that are dedicated to processing a circumscribed input domain whose regular properties can be exploited by specific cognitive procedures. On this basis, it is far more reasonable to suppose that our central cognitive system consists of multiple specialized subsystems rather than a single general-purpose system for interpreting and responding to the myriad issues the world presents us with. Of course, once we move beyond the peripheral ‘input’ systems that Fodor focused on to the central conceptual systems, the concept of a ‘module’ has to be modified somewhat, in particular with regard to the issue of the information available to the system. Sperber (1994a, 2000) develops an account in which conceptual modules are multiply interconnected, so that the procedures of each of the individual modules can be informationally encapsulated, while ‘chains of inference can take a conceptual premise from one module to the next, and therefore integrate the contribution of each in some final conclusion’ (Sperber 1994a: 133). It is a defining property of pragmatic processes that they are context-sensitive, so, whatever the principles and procedures of the pragmatics module, they must have access to contextual assumptions from a wide range of sources (current perception, earlier discourse, general and cultural knowledge stored in memory). The account of pragmatics given in relevance theory provides a solution to this tension between responsiveness to broad context and the informational encapsulation of a modular system, in that the general comprehension procedure which is the engine of the system 6
It has been pointed out to me (by both Deirdre Wilson and Yan Huang, independently) that these claims are somewhat controversial. Indeed, I should acknowledge that Grice himself maintained that the conversational maxims were special cases of principles governing more general cooperative (not just communicative) behaviour, e.g. ‘Quantity. If you are assisting me to mend a car, I expect your contribution to be neither more nor less than is required; … Relation. I expect a partner’s contribution to be appropriate to immediate needs at each stage of the transaction; … ’ (Grice 1975: 47–48). However, Grice’s examples (of cooperative interactions not involving ‘talk’, as he puts it) raise further questions for me, in that although they are clearly not cases involving speaker meaning, I think some of them may fall within the broader domain of ostensive communication. No doubt, there are issues here that call for further thought.
458
458 Robyn Carston applies not only to the derivation of hypotheses about the communicated content (explicatures and implicatures) but also to the accessing of contextual assumptions. Here is the relevance-based comprehension procedure: (3) a. Follow a path of least effort in constructing an interpretation (of the ostensive stimulus)—that is, in resolving ambiguities and referential indeterminacies, enriching encoded meaning, supplying contextual assumptions, computing implicatures, etc. b. Stop when your expectations of relevance are satisfied. It is not my concern here to motivate this procedure or explain the general cognitive underpinnings of our context-specific ‘expectations of relevance’ which play the crucial role in terminating the process and settling on an interpretation (for detailed accounts, see Wilson and Sperber 2004, 2012b; Wilson this volume). The key point here is that, in effect, the pragmatics module itself regulates its accessing of information stored in other conceptual modules and imposes a degree of encapsulation which ensures that the interpretive process can be both swift and accurate. A further important feature of Sperber and Wilson’s view of the pragmatics module is that it is like the more general ‘theory of mind’ module in being inherently meta- representational—that is, it explains specific kinds of human behaviour in terms of mental representations in the mind of the person who produced the behaviour. So, while the outputs of our perceptual systems consist of base-level descriptive representations like ‘X is red’ or ‘X is screaming’, the outputs of the theory of mind module are higher-order representations in that they are attributions of lower-order mental representations to others, e.g. ‘X believes there are chocolates in the box’ and/or ‘X desires to eat some chocolates’, and so too those of the pragmatics module, e.g. ‘X intends me to recognize that she is informing me that she is hungry and wants to eat lunch now’. Thus the pragmatics module can be viewed as a special submodule of the more general theory of mind module: that is, it has its own dedicated procedures (such as (3) above and perhaps other more specific heuristics and inferential shortcuts based on it) which would not work if applied to other kinds of theory of mind tasks (for detailed argument along these lines, see Sperber and Wilson 2002; Wilson 2005).7 How are we to construe the pragmatics-semantics interface within this modular picture of the mind? Assuming with Fodor (and Sperber and Wilson) that language is a modular system,8 then the output of that system is a representation of the
7
See Siegal and Surian (2006) for a critical assessment of Sperber and Wilson’s view of pragmatics as a specialized submodule of theory of mind and Fodor (2000) for criticism of the massively modular view of the mind. 8 See Fodor (1983), Pinker (1994), and Carston (1996) for discussion of evidence in support of the position that both initial lexical access and syntactic parsing are encapsulated from top-down expectations. For dissenting views, see Marslen-Wilson and Tyler (1987), Elsabbagh and Karmiloff- Smith (2004), and Prinz (2006).
459
Pragmatics and Semantics 459 context-invariant meaning provided by the sentence (that is, the meaning encoded in its lexical and syntactic components). We can call this a ‘semantic representation’.9 What form it takes is part of the subject matter of linguistics and will depend on the semantic analyses of a great many distinct linguistic phenomena and how they interact, e.g. quantification, tense and aspect, negation, mood and modality, as well as resolution of more general issues about the nature of the lexical meaning of open-class words (nouns, verbs, and adjectives), closed classes including coordinating connectives like ‘and’, ‘or’, ‘if ’, ‘but’, and ‘for’, subordinating connectives like ‘although’, ‘because’, ‘when’, and ‘while’, prepositions and derivational affixes like ‘-ize’, ‘-er’, ‘-al’. According to relevance theorists and to ‘contextualists’ more generally, this semantic representation is not the kind of entity to which truth conditions can be assigned, that is, it is neither fully propositional nor in the right format to constitute a thought, so is not susceptible of truth evaluation. However, there is a long tradition of work in the philosophy of language according to which sentence meaning is truth- conditional. I will consider some versions of this view in section 24.3, but for now the focus is on the interface of this semantic representation, whether propositional or not, with pragmatics. Consider again the example in (1), repeated here: (1’) She’s had enough. Without attempting anything like a proper semantic analysis, we can describe the context-independent meaning of this sentence along the following lines: the pronoun ‘she’ provides a variable and a constraint on the specific content that can be pragmatically assigned to it, namely, that of being a particular female; the verb ‘have’ provides a very general meaning (or, if it is ambiguous, perhaps several somewhat more specific meanings: one that is similar to ‘consume’, another similar to ‘experience’); in addition to its meaning, the adjective ‘enough’ may specify some open slots corresponding to the questions ‘enough of what?’, ‘enough for what purpose?’; the syntactic structure of the sentence indicates that certain relations (agent, patient, etc.) hold between the referent of ‘she’ and that which she has had enough of. As uttered by a particular speaker to a particular addressee, this decoded meaning (i.e. the ‘semantics’ of the sentence type) 9 Chomskyan linguists are more apt to talk of this representation as ‘LF’ or ‘logical form’, as does Fodor (1983: 88–90) in his discussion of the ‘shallow output’ of the language module. I avoid this terminology here for two reasons: (i) what is meant by ‘logical form’ is notoriously unstable across linguists and philosophers, many of the latter using it for a much richer representation of sentence meaning unconcerned with the different sources of this meaning (context-invariant/linguistic or context-variable/pragmatic), and (ii) use of the term ‘semantic representation’ is more consonant with the main focus of this chapter, which is on the interface of pragmatics and ‘semantics’. However, I find appealing the view sometimes expressed by Chomsky that, from the internalist representational perspective, there is only syntax and pragmatics, so what I’m calling a ‘semantic’ representation here can be thought of as simply the one among the various syntactic representations of a sentence that interfaces most directly with the internal conceptual system (Chomsky 2000).
460
460 Robyn Carston is crucial evidence constraining the inferences of the addressee’s pragmatics system whose goal is the recovery of the thoughts the speaker is intent on communicating. On the account as outlined so far, pragmatics and language (including semantics) constitute two distinct modular systems of the mind. The pragmatics module is activated whenever an ostensive stimulus (whether linguistic or non-linguistic) is registered, thus triggering the interpretation process. Consider a case of non-verbal ostension: your friend Mary is smiling fixedly at you and gesturing meaningfully with quick movements of her head and eyes in the direction of someone, Y, who has just entered the room. Your pragmatics module is activated and infers from the evidence indicated by Mary’s ostensive behaviour that she is communicating that Y is her new boyfriend (whom she has excitedly told you about the previous night). In the case where Mary produces a verbal utterance, e.g. ‘That’s him’, some of the evidence provided by the communicative act would come from linguistic semantics (the output of the language module) and this would strongly constrain your interpretation of the communicative act. In many instances, of course, linguistic evidence is far more detailed and fine-grained than this and provides immeasurably richer clues to the content the speaker wants to communicate than any non-verbal ostensive stimulus ever could. So there is a sense in which semantics is temporally prior in the processing stream (in that it provides input to pragmatics). However, when an utterance is produced (a linguistic ostensive stimulus), both systems are activated in parallel: the pragmatics module is triggered by the ostensive character of the stimulus and the language module by its linguistic character (whether speech or text), and the pragmatics may start working on non-linguistic clues even before it receives any semantic input. There is an open question here about what unit of semantic representation is passed from the language system to the pragmatics system, but what is clear is that it is not a fully sentential semantic representation which is first formed and then handed over as a whole. Pragmatic hypotheses about the intended referent of an utterance-initial ‘she’ or ‘that’, for instance, are usually made well in advance of the whole sentence having been linguistically decoded and, in the case of multimodal ostensive stimuli (the usual case in face-to-face speech), those hypotheses may be informed by eye gaze, pointing, or the perceptual salience of individuals in the external context.
24.2.2 Pragmatics and semantics from an evolutionary perspective An interesting issue in the context of the ‘pragmatics first’ approach that I am taking here is the question of the evolutionary emergence of ostensive communication (hence pragmatics) and language (hence semantics) in the human species. Assuming that the picture outlined in the last section is correct, according to which language and pragmatics constitute two distinct but constantly interacting cognitive modules, they must surely have to some extent coevolved, each fine-tuning the other. However, as Sperber (2000: 121) suggests, ‘it still makes sense to ask which of these two, the linguistic or
461
Pragmatics and Semantics 461 the metarepresentational [= the mind-reading/pragmatic (RC)], might have developed first to a degree sufficient to bootstrap the co-evolutionary process’. There seem to be three possibilities: (a) language first, then communication/pragmatics, with the former enabling the latter; (b) communication/pragmatics first, then language, with the former creating a niche for the emergence of the latter; (c) independent, possibly parallel, emergence of the two systems, with subsequent recruitment of one by the other and coevolution. The first view is quite widely held,10 but Sperber (2000) argues in favour of the second—that is, that our capacity for metarepresentation emerged first and enabled a rudimentary kind of (non-verbal) ostensive-inferential communication. The idea is that the metapsychological ability to attribute contentful mental states to each other (beliefs, desires, intentions, etc.) evolved as an adaptation to the pressures of living in social groups, greatly enhancing our ancestors’ ability to predict and explain each other’s behaviour and so to both compete and cooperate better with each other. Sperber outlines how this awareness of the mental states of others opens up possibilities for various kinds of intentionally informative behaviour, including ultimately ‘ostensive’ communication, through which, as communicators, we can overtly indicate our intention to alter the mental states (beliefs and desires) of others, and, as addressees, we can gain access to a huge amount of information, the understanding and acceptance/rejection of which are both based on our assessment of the speaker’s mental states (her beliefs and intentions). Ostensive-inferential communication, Sperber suggests, may have initially been a side effect of our theory-of-mind ability, but its highly beneficial character would have led to its becoming a specialized function of that system, with its own dedicated computational procedures, hence the pragmatics submodule (Sperber 2000: 121–127). Sperber further claims that ostensive communicative behaviour creates a particularly favourable environment for the emergence of a new adaptation, namely, the linguistic ability (hence linguistic semantics). He argues that language as manifest in public utterances requires that a pragmatic capacity already be in place, given that the linguistic expressions employed (phrases or sentences) do not encode the speaker’s meaning (what she intends to communicate), even at the level of what is explicitly communicated (let alone implicature), where pragmatic inference is typically required for disambiguation, identification of entities referred to by use of pronouns and demonstratives, conceptual completions (recall the use of ‘enough’ in (1) above), and other kinds of meaning adjustment.11 10
See, for instance, Millikan (1984) and Dennett (1991). A leading thought on the side of the ‘language first’ position is that verbal utterances introduce public, hence perceptible, representations into the environment and so may pave the way for the development of a cognitive ability to grasp and represent representations qua representations—that is, a meta-representational ability, which (perhaps) could extend from linguistic representations to mental representations and so underpin a theory- of-mind capacity. However, Sperber (2000: 121–122) argues, convincingly in my view, against the plausibility of this scenario. 11 As well as arguments in favour of theory of mind/pragmatics first, Sperber (2000) gives interesting arguments against the language-first position, based on the untenability of the code model of communication. See also Origgi and Sperber (2000) and Sperber and Origgi (2010).
462
462 Robyn Carston However, Sperber does not address the question of the nature of the linguistic code that emerged or why it should have the properties that linguists have shown it to have: unbounded clausal embedding, long-distance dependencies, so-called ‘island’ constraints (e.g. ‘Who did John call Mary and?’ is glaringly ungrammatical but easily interpretable), seemingly redundant subject–verb inversions (e.g. ‘Why did John go?’ rather than ‘Why John went?’), and a variety of other complexities that are not inherently communication-enhancing. This seems to me to leave an opening for the third possibility mooted above, namely, that the human language faculty emerged independently of the theory of mind/pragmatics capacity, perhaps in parallel with it (or even before it), and that the ostensive communicative environment that favoured the employment of a system of coded meanings led to the co-opting of this pre-existent system by pragmatics. Noam Chomsky has always disputed the view that the primary function or purpose of language as he construes it (I-language, a computational procedure) is communication. Furthermore, he argues, on the basis of what we can discern from the archaeological record, language emerged quite suddenly in the mind/brain, such that it appears to be the result of some fortuitous and sudden mutation rather than of gradual processes of natural selection (see, e.g., Hauser, Chomsky, and Fitch 2002; Chomsky 2010). On this view, the use of language for communication would have been a subsequent development, a matter of linking the core linguistic capacity to sensorimotor systems required for its externalization in verbal production and for its perceptual registration in the first stage of verbal comprehension. From then on, we can envisage the two systems, pragmatics and language, coevolving in ways that would facilitate the hugely beneficial function of verbal communication in human life, perhaps involving, on the one side, the establishment of pragmatic inferential shortcuts based on patterns of language usage, and, on the other side, the coining of new lexical items to further constrain pragmatic inferences and so speed up comprehension.12 A tentative conclusion, then, is that language construed narrowly, as a recursive computational procedure, may have arisen in the human mind/brain independently of theory of mind or pragmatics, but that language construed more broadly, as a system for making public bits of coded information, would have both depended on and hugely enhanced the functioning of a pre-existing pragmatics system. 12 An interesting speculation here is that certain words whose encoded meanings do not contribute to truth-conditional content but seem rather to function as pragmatic inference indicators might have arisen as a result of the deployment of the I-language system for communication. Candidates are ‘discourse connectives’ (e.g. ‘however’, ‘moreover’, ‘after all’, ‘thus’, ‘anyway’), discussed in these procedural terms by Blakemore (1987, 2002), and various other apparently ‘pragmatic’ lexical items like politeness markers (e.g. ‘please’), honorifics, and illocutionary indicators. (See Carston 1999, 2008a for further discussion bearing on this idea about the origin of linguistic devices with ‘procedural’ meaning, and Wilson 2011b for a comprehensive overview of the conceptual/procedural distinction in linguistic semantics.)
463
Pragmatics and Semantics 463
24.2.3 Pragmatic-semantic development in children The domain of pragmatics is ostensive communication, an activity which, as far as we know, only humans are capable of.13 Children manifest this kind of communicative capacity before they produce their first words. For instance, from the age of 12 months, infants use the ostensive gesture of pointing for a number of purposes, not only when they want the addressee to give them something (proto-imperative pointing), but also when they want to share an experience with their caregiver (e.g. so they can jointly attend to a dog in the park) and even simply in order to be helpful to an interlocutor (e.g. to direct her attention to the keys she has dropped). At around the same age, they begin to respond to acts of pointing by others, directing their own attention to the apparent target of the point. The evidence for the latter two kinds of pointing, for sharing attention and experience, and for informing others helpfully, is a strong indication of the emergence of what Michael Tomasello and his colleagues term ‘shared intentionality’ (Liszkowski 2006; Tomasello 2008), the basis for a full-fledged theory-of-mind and the pragmatic capacity. Early communication by pointing (and eye gaze) seems to be a human universal. Using a specially developed method (the ‘decorated room’ context) for eliciting spontaneous pointing, Liszkowski et al. (2012) observed and recorded the pointing behaviour of 10–14-month-old infants with their caregivers across seven very different cultural settings, including Bali, Japan, Peru, Mexico, and Papua New Guinea. They found no influence of cultural differences on pointing: all the infants spontaneously used pointing to communicate with their caregivers, and the behaviour emerged at the same age across all the cultures and with the same frequency of occurrence. Liszkowski et al. conclude that their findings ‘support the existence of a gestural, language-independent universal of human communication that forms a culturally shared, prelinguistic basis for diversified linguistic communication’ (Liszkowski et al. 2012: 698). Children start producing words with appropriate meanings (often accompanied by pointing) in their second year of life, indicating an embryonic lexical semantic competence, whose growth escalates over the next few years (Bloom 2000: 25–47). This requires that they learn to make links between particular chunks of linguistic form and particular objects and activities in the world (or percepts/concepts of those objects and activities). Experimental work by Baldwin (1993), Bloom (2000), and Tomasello (2001) shows that in the process of acquiring these form–meaning links, children are making inferences about the referential intention of the speaker who has produced the word form. Consider, for instance, an experiment with 18-month-olds, which involved two new toys in the room, for neither of which the child had a name. When
13
For discussion of the distinguishing characteristics of ostensive communication, including its dependence on a highly developed meta-representational capacity, see Sperber (1994b, 2000), and for discussion of the absence of these characteristics in the communicative behaviour of other primates, see Tomasello (2006, 2008) and Levinson and Holler (2014).
464
464 Robyn Carston the experimenter introduced a new word by saying, e.g., ‘It’s a modi’, the child did not link the word to the toy he was playing with, but rather looked away from that toy to the speaker/experimenter and then redirected his gaze to the object that she was looking at (the second toy, which was inside a bucket). When later asked to find the ‘modi’ he picked up the object the experimenter had been looking at in the bucket when she first uttered the word. Bloom (2000) maintains that what is involved here is the child’s use of theory-of-mind (that is, the general capacity to attribute beliefs and intentions to others), but a stronger claim may well be warranted: given that these children are already in the business of ostensive communication—that is, they both have and interpret communicative intentions as manifest in pointing behaviours—it seems plausible that it is this pragmatic ability (rather than more general theory-of-mind abilities) that is doing the work here. There are perceptual and conceptual constraints on the word meanings children acquire; for instance, they tend to interpret a new word form as a name for a whole object, e.g. ‘hamster’ for the whole animal, rather than for a subset of its properties, such as being small, furry, and active (see Bloom 2000: ch. 4, on this ‘whole object bias’). There seem also to be pragmatic constraints on the learning of word meanings— that is, constraints that are best accounted for in communicative terms. So, for instance, it is widely noted that children are strongly biased towards assuming that words do not have overlapping reference: that each object can be named by only one word (Clark 2003). In a study in which 3-year-old children were introduced to a new noun together with an object, they interpreted the word as a label for the object only if the object was unfamiliar to them (that is, they did not already have a word for it); if the object was familiar (something for which they already had a word), they interpreted the new word as referring to some specific part or property of the object (Markman and Wachtel 1988). Similarly, when presented with two actions, one familiar and one unfamiliar, together with a new verb, e.g. ‘The girl is torping’, they assumed the new verb referred to the action for which they did not already have a name. These results have since been replicated in studies of much younger children (18–24 months old). According to the pragmatic account of this phenomenon, children’s ‘avoidance of lexical overlap’ is explained by their presumptions about rational communicative behaviour. Specifically, the claim is that children expect speakers to use familiar (presumed shared) words for objects or actions on the basis that this facilitates hearers’ understanding, so when a speaker/experimenter uses a novel word, they infer that she intends to refer to some other object/action than the one for which there is a familiar word (for details of the pragmatic reasoning attributed to the children, see Diesendruck and Markson 2001; Grassmann et al. 2009). Of course, there is vastly much more to be investigated in children’s pragmatic development, especially as it interacts with their increasing semantic sophistication; for instance, the development of their referential abilities (the use of pronouns vs demonstratives vs descriptions of various sorts), their ability to infer various unarticulated constituents of explicatures, e.g. quantifier domains (as in ‘Some students finished early’) and implicit arguments (as in ‘We’ve eaten’), quantifier scope relations (as in
465
Pragmatics and Semantics 465 ‘Everyone saw a famous actor’), the topic–comment distinction, indirect communication (via implicatures of different kinds), and the full range of non-literal uses of language (metonymy, metaphor, irony). My aim here has been just to advance the position that children are ostensive communicators before they are language users and that this communicative (pragmatic) ability plays a key role in their subsequent acquisition of word meanings. For more extensive accounts of developmental pragmatics and the pragmatic foundations of much semantic development,14 see Pouscoulous (2013), Matthews (2014), Zufferey (2015), and Rollins (this volume).
24.3 Semantics and Pragmatics: Philosophical Perspectives While the cognitive perspective taken in the previous section favours an account in which the pragmatic inferential capacity takes precedence over (linguistic) semantics, communicatively, evolutionarily, and developmentally, when we turn to work on semantics and pragmatics in the philosophy of language, the perspective, at least until quite recently, has been quite different. A key concept in traditional philosophy of language was that of the proposition semantically expressed by a sentence, that is, the truth-conditional content of the sentence, which was often equated with ‘what is said’ when the sentence is uttered. The centrality of this view of sentence semantics goes back to Frege, Russell, and Carnap, who were first and foremost logicians, interested in the semantic properties of formal languages, such as the predicate calculus. However, they extrapolated from these artificial languages to human (natural) languages, which they assumed would have the same fundamental properties. So, just as the semantics of logical formulae was taken to be a matter of how the external world must be for them to be true (that is, their truth conditions) and the semantics of logical connectives such as ‘&’, ‘∨’, ‘¬’ was fully captured by truth tables, it was assumed that natural-language sentences also have truth conditions and natural-language connectives such as ‘and’, ‘or’, ‘if … then’ are truth-functional. The presence within natural languages of indexical elements (e.g. ‘I’, ‘today’), which depend on a context of use for their ‘semantic value’, was seen as an interesting extra issue to be dealt with but no threat to the overall picture. The programme of giving a formal
14 I take it that, while learning the meaning of open-class content words (nouns, verbs, adjectives) is pragmatically mediated, as discussed here, there are closed-class functional words, including quantifiers, determiners, sentence operators (e.g. negation, disjunction, conjunction), and complementizers (e.g. ‘that’, ‘to’, ‘whether’), whose semantic properties are more likely to come online as a consequence of stages of maturation of the grammar (for some discussion, see Crain and Thornton 2006).
466
466 Robyn Carston truth-theoretic account of the semantics of sentences was accordingly modified so that the truth of a sentence was made relative to a small number of contextual parameters (speaker, addressee, time, and place of utterance). This ‘ideal language’ approach to the semantics of human language was challenged by Austin, Grice, Strawson, and the later Wittgenstein, who developed the ‘ordinary language’ approach, aimed at describing natural-language phenomena rather than forcing them into the logical mould. They rejected the equation of sentence meaning with truth conditions and maintained that it is speech acts or utterances that express propositions and so have truth conditions; that is, it is the statement made that has truth conditions; the sentence per se does not. This ‘speech act’ view of truth-conditional content has had a huge impact on the study of linguistic meaning, including pragmatic-semantic accounts developed within the cognitive perspective discussed in the previous section. However, a less positive tendency among some practitioners of the ordinary language approach was to blur the distinction between the linguistic meaning of a sentence and its appropriateness in particular communicative contexts. This led to the positing of multiple ambiguities or rich complex meanings that seemed to be infected with features that arise from assumptions about their use by rational speakers. For instance, the connective ‘and’ was taken to have as one of its meanings a cause–consequence component in order to accommodate cases like (4a), and the meaning of perception verbs (e.g. ‘look’, ‘feel’, ‘sound’, ‘smell’), as in (4b), was taken to include the implication that it is doubtful that X is F. (4) a. Mary insulted John and he walked out. b. X looks F to me. One of Grice’s important contributions was to find a way to keep separate the intrinsic meaning of expression types and the meaning that arises from regularities of use, and to thereby reconcile the logical/ideal and the ordinary language accounts of linguistic content. He showed how extralinguistic components of utterance meaning could be accounted for by a system of conversational maxims which regulate the rational communicative use of language and function as premises in the (non-demonstrative) inferential processes by which conversational implicatures are derived, e.g. ‘John walked out because Mary insulted him’ for (4a) and ‘X is not F’ for (4b). In this way, Grice maintained, the central semantic content of these utterances (that is, what is said by a speaker in uttering the sentence) could be kept apart from these usage effects and the truth or falsity of the utterance assessed on properly semantic grounds. For instance, an utterance of the sentence in (4a) is true if and only if two events took place (at particular times): (a) Mary insulted John, (b) John walked out. This was a groundbreaking move and the beginning of modern pragmatics. However, in equating semantics with ‘what is said’ and pragmatics with ‘what is conversationally implicated’, it did not yet do justice to the full extent of the role of human pragmatic capacities in linguistic communication. The key construct that needed to be unpacked was Grice’s notion of ‘what is said’, which was required to be both a semantic and a
467
Pragmatics and Semantics 467 pragmatic entity. It was a semantic notion in that it was closely tied to the conventional meaning of the words and syntax of the uttered sentence, with only a very minimal context-dependent component, restricted to choosing between the senses of ambiguous words and supplying values for indexicals, both apparently achieved on the basis of best contextual fit (Grice 1975: 44). However, it was pragmatic in that, like the act of implicating, the act of saying something comes with a communicative intention (an m-intention, in Grice’s terms), so that what is said and what is implicated were taken together to constitute what the speaker meant by her utterance (for discussion, see Neale 1992; Recanati 2004a: ch. 1). The problem is that no single level of meaning can do double duty as both sentence semantics and speaker-meant primary (explicitly communicated) meaning. What a speaker says and means is often something different (sometimes more specific, sometimes looser) than the semantic content of the sentence employed, even when any indexical elements have been assigned context-specific reference. Consider the following examples: (5) a. It was snowing. b. The children formed a circle. Arguably, what a speaker of the sentence in (5a) explicitly means/communicates is that it was snowing in a particular place (say, London) on a particular day—her utterance would not be made true by snow falling anywhere else on that day. However, there is no location constituent in the sentence uttered so this cannot be a component of the meaning of the sentence. Turning to (5b), the sentence contains the word ‘circle’, which denotes a closed curved line whose every part is equidistant from a fixed point, so that is the meaning that the word contributes to the sentence. However, it is very likely that the concept communicated by the speaker’s use of the word is not that of the perfect geometric shape but rather a more general one that allows for quite a range of irregularities. What these examples indicate is that it is just not generally right that what a speaker says (and means) is as close to the conventional meaning of the sentence uttered as Grice’s definition of ‘what is said’ requires (Grice 1975: 44). The construct has to be split into two distinct entities, one semantic and one pragmatic: the first is the meaning of the sentence (which is seldom, if ever, the same as the speaker’s meaning) and the second is a fully pragmatic (speaker-meant) notion of ‘what is said’.15 The upshot of this splitting of the Gricean concept of ‘what is said’ is a three-way distinction between sentence meaning (linguistic semantics), what is said (or explicitly communicated), and what is implicated (or implicitly communicated). This division has been quite widely accepted and adopted, albeit in different variants, by many philosophers of language (e.g. Recanati 1989, 2004a; Bach 1994; Stainton 2004; Neale 2004; Borg 2004; Cappelen and Lepore 2005). However, one of the several issues that 15
For more detailed discussion of these issues, see Carston (2004), Recanati (2004a), and Carston and Hall (2012).
468
468 Robyn Carston remain unresolved concerns the nature of the semantic content that provides input to pragmatic inferential processes. While the two kinds of communicated contents, the pragmatic ‘what is said’ (or explicature) and conversational implicatures, are propositional and so truth-evaluable (that is, by and large, we communicate thoughts to each other), we can question the traditional philosophical view according to which sentence semantics is propositional. The idea that the semantics of natural-language sentences is and must be truth- conditional continues to have a strong hold in current philosophy of language (see Borg 2004, and Cappelen and Lepore 2005 for two recent manifestations of this view). Nowadays, this approach is known as ‘minimal semantics’ and its central claims are (a) that sentences semantically express propositions, and (b) the involvement of context in the identification of that propositional content is ‘minimal’, at most involving the fixing of specific values for a small class of inherently indexical words. I will focus here on the version of this position taken by Emma Borg because she shares more of the assumptions that animate the cognitive-scientific approach taken in this chapter than do other semantic minimalists, thus enabling a more direct comparison of the positions. She adopts the Fodorian view of language as a modular input system and insists that the proposition semantically expressed by a sentence is an algorithmic function of the lexical and syntactic components of the sentence alone—that is, it is informationally encapsulated from extralinguistic context and the hearer/addressee’s pragmatic capacities.16 An apparent obstacle to this formally driven kind of propositional semantics is the issue of indexical/demonstrative content. Most truth-conditional semanticists have resorted to notions like ‘demonstrated object’, ‘contextually salient object’, or ‘intended object’, which they include, by stipulation, in the set of objective contextual parameters, despite their being clearly pragmatic concepts. Borg, however, holds firm to the constraints of mental modularity and accepts that, in utterance comprehension, identifying the referents of demonstatives like ‘this’, ‘that’, ‘she’, ‘he’, ‘it’, ‘then’, ‘there’, etc., is a post-semantic (pragmatic) matter, requiring consideration of speaker intentions. Nevertheless, she maintains, the modular semantic analysis of a demonstrative-containing sentence is a propositional form. Her idea is that each tokening of a demonstrative or indexical syntactically triggers the creation of a singular concept which is its semantic content, although figuring out what object that concept 16
Cappelen and Lepore (2005) seem happy to allow pragmatic considerations to enter into the identification of the intended senses of ambiguous words and the context-specific content of indexicals and demonstratives (147–149) and do not see this as problematic for their conception of the ‘semantically expressed’ proposition. It is difficult to make sense of this in the context of their claim that the semantically expressed proposition is the only component of utterance meaning that carries over from one context to another and which is, therefore, the one component of meaning that speaker and hearer can be fully confident of sharing with each other. (For further discussion of this point, see Recanati 2004a, Borg 2007, and Carston 2008a.) In any case, whatever Cappelen and Lepore intend by their semantically expressed proposition, it is clearly a very different kind of entity from the one that Borg and other language modularists intend.
469
Pragmatics and Semantics 469 refers to requires consideration of speaker intentions and so is a task that lies beyond the remit of the formal semantic processor. According to this account, what we grasp when we understand an utterance of ‘That’s mine’ is a proposition or thought of the shape [α is β’s] where ‘α’ and ‘β’ are singular concepts. Each comes with a further bit of information along the lines that ‘α’ is a that concept while ‘β’ is a speaker concept, information provided by the semantic character of the linguistic expressions, which functions as a constraint on how the singular concepts are subsequently integrated with other language-module-external information (from perception or memory). As I see it, there are two problems with this propositional approach to linguistic semantics, the first quite specific, the second a more foundational issue. First, it is not at all clear that this modular semantic account of indexicals/demonstratives does result in a full-fledged propositional representation, that is, one with a truth-evaluable content. Take an utterance of ‘That is a butterfly’. The semantics (i.e. the truth-conditional content) of this utterance, according to Borg, is [μ is a butterfly], where butterfly is a general concept and μ is a singular concept, but where the object that constitutes the referent of that singular concept has not been identified (that being a post-semantic, properly pragmatic, matter). The problem is the following: if we, grasping just this formally supplied content, were presented with an exhaustive array of the butterflies in existence, we would not be able to judge whether this utterance was true or false. It is difficult, then, to see in what sense we can be said to have recovered the truth- conditional content of the utterance; the constituent μ, allegedly a singular concept, does not seem (prior to the pragmatic identification of a referent) to make a truth- conditional contribution. The second, more general, question is why we should expect or want the semantics of a natural-language sentence to be a truth-evaluable entity. It seems right to expect the output of successful communication to be truth-evaluable thoughts (propositions)—it is these that we agree or disagree with, believe or doubt, hold people to, act on the basis of, etc.—but why should we expect them from a theory of the meanings encoded by sentences, meanings which function as multiply reusable tools in communication and which are virtually always supplemented, enriched, or otherwise adjusted when so used. The best argument for the output of semantic processing being propositional would be that this property plays a key role in the overall account of verbal communication and comprehension in which it is lodged, but, as far as I can see, there just is no such argument. (For more detailed discussion of this issue, see Carston 2008a, 2008b.) The account of pragmatics and semantics as cognitive capacities, presented in section 24.2, strongly suggests that what the semantics (the output of the language processor) provides is a representation which is not fully propositional. That is, given a theory of mind/pragmatics capacity already in place, so already recognizing ostensive stimuli and figuring out the content of communicative intentions, what the linguistic- semantic input is required to provide is reliable evidence that facilitates and constrains the processes of that system in its goal of recovering the speaker’s intended meaning. Even some recent philosophically oriented accounts that do not sign up to the specifics of this cognitive approach favour a non-propositional account of sentence semantics.
470
470 Robyn Carston For instance, both Bach (1994, 2006c) and Recanati (2001, 2010), who do not adopt the modularity picture or express any views on the evolutionary or developmental priority of pragmatics and semantics, eschew a propositional account of sentence meaning.17 Recanati advocates what he calls ‘truth-conditional pragmatics’—that is, the position that linguistic semantics does not deliver truth conditions but rather ‘conditional truth- conditions, or schemata, or characters, or whatever. To get full-blown truth-conditional content, pragmatics will be needed’ (Recanati 2010: 3). Bach roundly criticizes positions on sentence semantics that ‘accept Propositionalism, the fancy version of the old grammar school dictum that every complete sentence expresses a complete thought’ (Bach 2006c: 436). He maintains that many syntactically complete sentences are ‘semantically incomplete’, lacking at least one constituent needed for them to be evaluable as true or false. For instance, the process of compositionally assembling the content of sentences like ‘John is ready’, ‘Mary is too tall’, and ‘Fred has drunk enough’ from the meaning of their lexical and syntactic parts, does not result in a proposition but in what he calls a ‘propositional radical’ (ibid.: 437). To sum up, the specific cognitive-scientific approach adopted in relevance theory and the views of those philosophers who take seriously the role of pragmatics in identifying the communicated/speaker-meant content of linguistic utterances converge on a non-propositional construal of the meaning that sentences of the language provide to those pragmatic processes. Sentence semantics is merely a template or blueprint or schema, on the basis of which the fully truth-conditional explicature (‘what is said’ in Recanati’s terms, ‘impliciture’ in Bach’s terms) is pragmatically constructed.
24.4 Conclusion: Pragmatics, Thoughts, and Truth-Conditional Semantics The pragmatics–semantics interface, as discussed in this chapter, is the point of contact of two cognitive systems: the pragmatics system, or ostensive stimulus processor (which is ‘modular’ in Sperber’s sense of being an evolved special-purpose mechanism attuned to the regularities of a particular input domain) and the language processor (which is modular in both Fodor’s and Sperber’s senses). The ‘semantic representations’ that are the output of the language module provide the pragmatics system with key evidence of the speaker’s intended content, evidence which places strong constraints on 17
This position originated with the ordinary language philosophers: ‘ … if you just take a bunch of sentences … impeccably formulated in some language or other, there can be no question of sorting them into those that are true and those that are false; for the question of truth and falsehood does not turn only on what a sentence is, nor yet on what it means, but on, speaking very broadly, the circumstances in which it is uttered’ (Austin 1962b: 110–111, his emphases), and is held by many who identify as ‘contextualists’, that is, who take it that components of the truth-conditional content of an utterance come from extralinguistic context.
471
Pragmatics and Semantics 471 the interpretive hypotheses this system infers. As argued in section 24.3, it is very unlikely that these representations are fully propositional, nor is there any reason to think they should be, since it is the role of the pragmatics system, rather than the language system, to recover the (propositional) thoughts the speaker is trying to communicate. On this basis, I have suggested that the domain of a truth-conditional semantics, a semantics whose primary goal is to explicate the relation between representations and the world which they represent, is thoughts (or sentences of ‘mentalese’). So it is the output of pragmatics—explicatures and implicatures—rather than the linguistic semantic input (sentence meanings) that falls within the domain of a truth-conditional semantics. The ‘semantic representation’ of a sentence is simply a translation from one kind of representation (lexical-syntactic) into another (a conceptual schema or template, with an array of open slots). Only once it has been pragmatically completed and enriched does it constitute a propositional (hence truth-evaluable) representation which is susceptible of a ‘real’ (= truth-conditional) semantics, that is, a semantics that matches representations with states of the world and thereby captures the ‘aboutness’ of our thoughts (and, derivatively, of our utterances). Plainly, this kind of semantics does not interface with pragmatics (a system for interpreting communicative behaviour) but exists entirely independently (and ‘first’). Although we may occasionally reflect on one of our thoughts, perhaps recontextualize it and infer new implications from it, still we do not interpret it in the sense of figuring out its (intended) content. As Fodor (2001: 14) puts it: ‘ … whereas the content of a sentence may be inexplicit with respect to the content of the thought it expresses, a thought can’t be inexplicit with respect to its own content; there can’t be more—or less—to a thought than there is to its content because a thought just is its content.’ It is thought rather than language that has semantic (propositional, truth-conditional) content in the first instance. We can talk, derivatively, of the propositional content of an utterance because what we express with our utterances are thoughts, while sentence meanings are simply a means by which we can get our interlocutors on track towards grasping those thoughts. The history of truth-conditional semantics (as a theory of natural-language semantics) is replete with clever strategies for avoiding or sidelining an array of problems which are all, essentially, matters for pragmatics: ambiguity, vagueness, indexicality, and incompleteness. The approach has also had to simply leave out of its account the many linguistic expressions that do not fit into the truth-conditional mould, e.g. discourse connectives (e.g. ‘however’, ‘anyway’, ‘moreover’, ‘well’), various attitudinal and illocutionary particles, expressives, and interjections (e.g. ‘alas’, ‘hey’, ‘ouch’, ‘ugh’), non- canonical sentence structures (such as clefts, e.g. ‘It was John who paid the price’). Each of these expressions encodes a meaning but that meaning is arguably some kind of constraint on how pragmatic inferencing is to proceed and does not contribute to truth-conditional content.18 18
For an updated and more nuanced account of the various kinds of encoded procedural meaning, see Wilson (2011b, this volume).
472
472 Robyn Carston Once we recognize the distinction between a semantics for linguistic expressions (words, phrases, sentences) and a semantics for thoughts (sentences of mentalese), the difference between the tasks involved in the two enterprises becomes clear. On the one hand, linguistic semantics must address a range of quite distinct ‘meaning’ phenomena: (a) the polysemy of ordinary ‘conceptual’ words like ‘school’, ‘cold’, ‘open’; (b) pronouns, demonstratives, and many other words with an indexical component (e.g. ‘local’, ‘national’, ‘near’, ‘distant’); (c) function words like quantifiers, determiners, and logical connectives, and the semantic-structural constraints they can impose on a sentence; (d) discourse connectives, attitudinal particles and other words with a ‘procedural’ (pragmatically oriented) meaning. On the other hand, a semantics of thoughts (propositional contents) need not concern itself with polysemy, or indexicality, or incomplete or non-canonical sentence structures, or with those communicative devices, like discourse connectives, whose function is to constrain pragmatic interpretive processes (and so, arguably, do not occur in the language of thought). Perhaps, then, truth-conditional semantics, as applied to thoughts, can proceed pretty much as was originally envisaged by the ideal language philosophers, when they tried to carry it over from logical languages to natural languages.
Acknowledgements I am grateful to Catherine Wearing, Deirdre Wilson, and Vladimir Žegarac for helpful and encouraging comments on an earlier version of this chapter, and to Yan Huang for his excellent editorial support. During the period of writing this chapter, I have been supported by funding from the Centre for the Study of Mind in Nature, Oslo.
473
Chapter 25
Pragmati c s a nd Gram ma r More Pragmatics or More Grammar Mira Ariel
It takes both grammar and pragmatics, two quite different cognitive competencies, to explain natural language use and interpretation.1 This chapter proposes that the only viable definition for a grammar/pragmatics division of labour is the code versus inference distinction (1). I attempt to explain why it is that richer, multiple-criteria divisions of labour between grammar and pragmatics cannot work in principle (2). I then show that having a clear code versus inference definition for grammar versus pragmatics does not guarantee unified linguistic analyses (3). I end by addressing the question of whether grammar shrinks or gets extended in size vis-à-vis pragmatics under a code versus inference definition (4). My main point is that we cannot offer a single answer to this question. With respect to sentence anaphora, the role of pragmatics must be increased. For what I term ‘ugly facts’, we need a larger grammar at the expense of pragmatics, while scalar quantifers may require that we partially swap the roles of grammar (semantics) and pragmatics. When applying the code versus inference distinction, a case-by-case empirical approach must be adopted.
25.1 Grammar vs Pragmatics: The Demise of the Ideal Division of Labour Most philosophers and linguists assume a grammar versus pragmatics division of labour. Each, it is agreed by all, stands for a different competence, although both are 1
Funding for this research was received from the Israel Science Foundation, grant no. 161/09.
474
474 Mira Ariel essential for successful linguistic communication. But what is the difference between grammatically determined linguistic use and pragmatically determined linguistic use? Is there a single, consensual grammar/pragmatics division of labour? There’s both agreement and disagreement in the field about which criteria should serve to draw the distinction. And there’s both agreement and disagreement about how to apply the distinction(s), namely, which phenomena fall on the grammar side of the border and which on the pragmatics side. Thus, while most researchers adopt some grammar versus pragmatics division of labour, these do not converge with one another. Ideally, pragmatics and grammar should be defined based on a set of dichotomous criteria, each distinguishing between a clearly grammatical set of phenomena and an equally clear set of pragmatic phenomena. Note that the more criteria we can offer, the more informative the definition for each of the fields, and the more distinct the two competencies are shown to be. But, of course, the different criteria must converge with one another: they must draw the division of labour along the very same lines, so that the same set of phenomena are considered either grammatical or pragmatic by each criterion. Moreover, since linguists have had a pre-theoretical conception about which phenomena are grammatical and which are pragmatic, the set of criteria should converge with our pre-theoretical intuitions regarding grammatical and pragmatic phenomena. When speaking of grammatical/pragmatic phenomena, linguists commonly refer to rather general topics (such as presupposition, agreement) as unified wholes, which means that the grammar/pragmatics divide should place each such high-level topic on one side of the fence. Each such topic must be either totally pragmatic or else totally grammatical (this is what I have called the monolithic assumption in Ariel 2010). Finally, many linguists (beginning with the Radical Pragmatists) prefer to maximize pragmatics at the expense of grammar. The guideline is that whenever we can choose whether to account for some phenomenon as a piece of grammar or as a piece of pragmatics (because either account is descriptively adequate) we should prefer the pragmatic account.2 Following Grice’s (1989a) Occam’s Razor principle, pragmatic accounts are considered more economical and hence theoretically more desirable. A grammatical account is presumed to ‘burden’ the speaker (because pieces of grammar are each mentally represented and hence tax our limited memory). A pragmatic account does not, for it relies on a general cognitive (rational) capability we always put to use anyway (for computing relevance, etc.). No specific memorized body of knowledge is required if we can account for some linguistic use by reference to rational reasoning. All in all, then, the ideal division of labour between grammar and pragmatics must meet a number of conditions: it should rely on multiple criteria (the more of them the merrier); the application of these criteria must not clash, so that all converge on predicting the very same set of grammatical versus pragmatic phenomena; and the defined 2 The motivation here is that ‘with heavy doses of Gricean pragmatics, a very great deal of grammar can be completely done away with by making supposedly arbitrary lexical and syntactic facts follow from a few general principles of conversation’ (Sadock 1978: 285, emphasis added).
475
Pragmatics and Grammar: More Pragmatics or More Grammar 475 sets of phenomena should be precisely those considered canonical grammatical and pragmatic phenomena respectively, which means that whole topics are en bloc either pragmatic or grammatical. Table 25.1 lists some of the main criteria for distinguishing grammar and pragmatics discussed in the literature. Pragmatics makes an essential use of contextual background, but grammar is independent of context (1). In order for a pragmatic account to work, interlocutors must consider the specific context in which the utterance was made. Not so for grammatical phenomena, which are used in the same way regardless of the specific context. Similarly, but not totally overlapping with this definition is the linguistic/extralinguistic criterion, whereby grammatical phenomena only concern linguistic forms or units (2). No reference can be made to extralinguistic concepts, such as given information. The hallmark of pragmatics is that it does associate linguistic forms with extralinguistic factors. Next, grammatical phenomena are conventional (3). They are encoded. Pragmatic phenomena are the complement set. They are not conventional, but rather inferred ad hoc. If grammatical phenomena are encoded, but pragmatic phenomena are inferred, then grammatical (= semantic) meanings are uncancellable (i.e. they are necessarily triggered), whereas pragmatic meanings are cancellable (i.e. they only arise in supportive contexts). Another distinction between grammar and pragmatics is that grammatical phenomena are restricted to sentence scope (4). No reference need be made to material outside the specific sentence where the grammatical form occurs. But accounting for pragmatic phenomena may well require crossing the sentence boundary, relying on material outside the sentence (e.g. reference resolution).
Table 25.1 Some commonly assumed criteria for drawing the grammar/pragmatics divide Criterion
Grammar
Pragmatics
1. Context
Independent
Dependent
2. Linguistic
+
-
3. Conventionality/ Cancelability
+ (Encoded)/-
- (Inferred)/+
4. Scope
Sentence
Discourse
5. Explicit, primary
+
-
6. Violation
>*
>??
7. Motivation
Arbitrary
Functional
8. Hemisphere
Left
Right
9. Truth-conditional
+
-
476
476 Mira Ariel Grammatically specified meanings are explicit and therefore primary (5). They have definitely been conveyed and can directly be targeted by the interlocutor (who may assess the information as true or not, as interesting or not, etc.). Pragmatically derived meanings are implicit and indirect. As such they are secondary to the communication and cannot so easily be targeted by an interlocutor’s direct response. Next, grammatical and pragmatic violations are thought to carry different consequences (6). Grammatical violations are more severe. Pragmatic violations create dispreferred (but grammatical) utterances. As is well known, grammar is arbitrary, its raison d’être derived from its functionality alone (7). Thus, there is no particular reason why adjectives precede the nouns they modify in English, but the opposite is true for Hebrew. Pragmatic phenomena are said to be motivated, by which we mean that the form/function correlation involved is not arbitrary. It is no accident, for example, that temporal expressions (such as since, after) are routinely mobilized for conveying very specific non-temporal uses (such as ‘because’). The shift from temporal succession to causal succession is transparent and inferable, and hence interactionally motivated. Next, different cognitive processes characterize grammar and pragmatics (8). Grammar constitutes a cognitive linguistic competence, often associated with the left hemisphere in the brain, while pragmatics, some have suggested, constitutes a general performance system, often associated with the right hemisphere of the brain. By far the most popular criterion for distinguishing between grammatical (semantic) and pragmatic meanings is the criterion of truth-conditionality (9). According to this criterion, semantics accounts for all and only truth-conditional meanings, and pragmatics complements it by accounting for meanings which are not relevant in determining the truth conditions of the proposition expressed. Combining all the above criteria, a pragmatic phenomenon is expected to be context-dependent, non-truth- conditional, implicit, and secondary; it should involve an inference (and hence be cancellable); it should be accounted for by some extralinguistic factor; the account should be extra-sentential (discourse level) and handled by the right hemisphere. Since pragmatics and grammar complement each other, a grammatical phenomenon is expected to be context-independent, truth-conditional, explicit, and primary; it should be accounted for by some specifically linguistic factor; the account should be sentential; it should involve a convention (uncancellable) and should be handled by the left hemisphere. Some phenomena seem quite amenable to this very rich grammar/pragmatics delimitation: particularized conversational implicatures, for example. Consider the following pair of examples, where the impositive speech acts (‘peel that’ and ‘check in … ’) clearly differ in degree of politeness: (1) a. A:You always put us to work when we come here Ted there must be something I can do to help B: Yeah peel that (LSAC) b. Mhm. Would you be willing to check in and see what, what their policy is?(LSAC)
477
Pragmatics and Grammar: More Pragmatics or More Grammar 477 Would you be willing … conveys a high degree of politeness, and this interpretation meets quite a few of the pragmatic criteria: it is motivated, and inferred by taking into account the profuseness of expression, as well as the specific context; it is neither conventional nor truth-conditional; it is implicit, analysed at the discourse level, involves extralinguistic factors, and violating the use of that form may (but does not have to) lead to a dispreferred (but not ungrammatical) utterance. But not all classical pragmatic phenomena fall squarely on the pragmatics side of the fence. English well and Hebrew harey ‘after all’ only meet some of the pragmatic criteria. On other criteria they come out grammatical. Consider: (2) a. DARRYL: PAMELA: b …. yesh There.is
What does that have to do with heaven and hell in the book. … Well, … I’m just sort of reiterating. (SBC: 005) harey after.all (a)
divuax sofi (Lotan: 8). report final (= final report).
Well here indicates that the information it modifies is a dispreferred response, and harey ‘after all’ marks the information it modifies as ‘given’. As such, the uses of these expressions meet the following pragmatic criteria: non-truth-conditionality, implicitness (we can’t directly deny these interpretations), discourse-level analysis (for well only), reference to extralinguistic concepts (‘givenness’, ‘dispreferred response’). Violating their use conditions (when the information is not ‘dispreferred’ or ‘given’, respectively) would lead to a non-optimal utterance, but not one that is ungrammatical. However, other criteria show these very expressions to be grammatical, most notably conventionality. No inference mediates between the expressions and the interpretations mentioned above. They are encoded for well and harey. Context is no more involved here than it is when (semantically) ambiguous words are concerned. Finally, at least harey is sentential in scope (it modifies the sentence it occurs in). In general, the findings are that the ideal picture presented in Table 25.1 is unrealistic. The grammar/pragmatics division cannot be supported by criteria consistently lining up properly, so that single-topic headed phenomena (such as reference, presupposition, topicality, agreement) are totally grammatical or totally pragmatic. Unified topics are not well behaved (Ariel 2010: ch. 3). Not all aspects of speech acts or information structure are pragmatic, nor are all agreement and sentential anaphora grammatical, for example. Here is an example which demonstrates that even agreement is pragmatically mediated sometimes. Note that the couple is considered singular twice, but plural once: 3. For 12 years the couple has refused to pay federal income taxes … But now the Internal Revenue Service wants the $23,478.31 … the couple owes the government … The couple stopped paying federal income taxes in 1977, and have remained self employed … Corner works as … architect and Kehler is a public policy researcher … (B. J. Roche, The Boston Sunday Globe; Reid 1991: 215, ex. 8a,b)
478
478 Mira Ariel This variation is not random. Pragmatic inferences can explain it: When the couple acts as a single unit (refusing to pay taxes) they are treated as singular, triggering singular agreement. But when they are seen as a collection of two individuals, each with his or her own line of work, the verb is marked plural. Reflexive pronouns, equally considered grammatical, are not necessarily restricted to sentence scope: 4. Now Kittyi could see Mauricej, his red pullover the only colour in the gloom. There was no one but themselvesi+j in the huge building. (Anita Brookner, Providence, London: Thiad, 1983: 124) And here is an example where a classical pragmatic phenomenon (information structure functions associated with syntactic constructions) seems arbitrary in that the form/function correlation is different in English and in Yiddish, both Germanic languages, as Prince (1988) emphasized: (5) a. English: Matrix + embedded clause: ~It was they who found Eichmann (Prince’s 2b) b. Yiddish: A single matrix: Dos hobn zey gefunen aykhmanen This have they found Eichmann (Prince’s 5b, an attested example) While the discourse function associated with each of the constructions in (5) is the same, the structures themselves are quite distinct, undermining a motivated (and certainly iconic) form/function relation. Pragmatic interpretations must not invariably be motivated, or the same grammatical coding would have been used in both languages. All phenomena originally analysed as conventional implicatures, such as those associated with but, are problematic for the classical analysis. On the one hand, such interpretations are conventional, rather than inferred, which is a major feature of grammar. On the other hand, they are implicit and do not contribute to the truth conditions of the proposition, characteristic of pragmatics. Conventional implicatures are not the only case which straddles both sides of the grammar/pragmatics fence on the classical division of labour. So too does the temporal inference associated with and. While no doubt pragmatically derived, note that the woman (‘I’) insists that it is truth-conditional. ‘He’ is the policeman with whom the woman is filing her complaint about her ex-husband: 6. I: He broke open the door with a kick. Started to act madly. To yell … He: (reads) My husband (his name) entered the house and broke open the door … I: Excuse me. I correct the report. He first of all broke open the door and then entered …
479
Pragmatics and Grammar: More Pragmatics or More Grammar 479
He: Madam, don’t interrupt me and don’t teach me how to fill out a complaint … I: Try to understand, Sir, it’s impossible that he first of all entered and then broke open the door … He: It’s not important. That’s how you fill out the complaint. I: But it’s not true, and I won’t sign this complaint unless … (Originally Hebrew, Haaretz, 26 March 1982)
And here is another case where pragmatically inferred material affects the truth conditions of the proposition: 7. M: You don’t build by yourselves S: No, we don’t have laborers, we have 10, 15 laborers (Lotan: 7). How can ‘no laborers’ be compatible with ‘10, 15 laborers’? It is not. It is only the pragmatically enriched ‘We don’t have (enough) laborers (to build by ourselves)’ that is compatible with having 10, 15 laborers, which is why (7) is not a contradiction. Given that the various criteria clash with each other (Levinson 1983: ch. 1; Huang 2007; Ariel 2010), we have no choice but to give up the ideal, multiple-criteria distinction between grammar and pragmatics. One solution is to give up on an absolute division of labour. On one such view, grammar prototypically complies with the grammatical criteria and pragmatics prototypically manifests pragmatic features in accordance with the multiple criteria, but not all of grammar and not all of pragmatics do, which explains why no absolute boundary can be established (Recanati 2004b). Huang (2007) then cites Lyons’ (1987) pessimistic view that it is in fact futile to insist on a motivated and consistent boundary between semantics and pragmatics. Others have taken a different stand. If we don’t see ourselves as bound by the classical list of topics deemed pragmatic (and grammatical) and if we’re willing to make do with a single criterion, we can in fact draw an absolute and consistent grammar/ pragmatics boundary. But which criterion should we opt for? When conventionality and truth- conditionality clashed, most notably in the case of conventional implicatures, Grice chose truth-conditionality over conventionality: Despite the fact that conventional implicatures constitute encoded, uncancelled meanings, they count as pragmatic, because they are not truth-conditional (and they are implicit). While most pragmatists followed Grice on this classification, not all did. Prince (1988), Sperber and Wilson (1986/1995), and Ariel (1999, 2008, 2010) chose conventionality over truth-conditionality. This decision does not only rely on cases such as (7), which show that clearly pragmatic meanings may very well contribute truth-conditional interpretations. First and foremost the decision rests on the unquestionable assumption that whatever else grammar may be, its most basic feature is that it associates linguistic forms with their meanings and use conditions. This set of conventions must count as grammar, and if other criteria clash with this one, they must be given up. Since expressions analysed as triggering conventional implicatures are conventionally associated with specific meanings, they must
480
480 Mira Ariel fall on the grammar side of the fence, despite the fact that they do not affect the truth conditions of the proposition. In fact, Bach (1999) argued that conventional implicatures may be truth conditional, and the only reason why their falsity does not seem to affect the truth value assigned to the whole proposition is that they provide subsidiary, rather than primary interpretations. Bach’s point is that the contrast interpretation for but or although, for example, may be ignored provided that the two conjuncts (the primary meanings) are true, since it is only a subsidiary interpretation. My findings for although cases where there is no contrast between two true clauses are especially telling. Asked to judge the truth of (8), only 38 per cent of my subjects thought it was true, despite the fact that the false contrast interpretation is only pragmatic. Another 23.5 per cent thought it was ‘close to true’, and as many as 38 per cent thought it was either false or close to false: (8) Ehud Olmert is the Prime Minister although he is balding. In fact, there was no significant difference between the four judgements, which means that the contrast interpretation is not necessarily ignored when truth conditions are concerned. In conclusion, grammar must pair off all linguistic forms with all the conventions which specify their distributions and interpretations. It would have been nice if in addition, it had been truth-conditional (always and only), explicit, etc. But it is not. If so, encoding is not just an essential aspect of grammar, it must be its defining feature. Similarly, inferences must account for whatever rule-governed patterns cannot account for. Pragmatics must then account for the inferences accompanying semantic meanings. Again, it would have been nice if, in addition, such interpretations had been non- truth-conditional (always and only), implicit, etc. But they are not. If so, inference is not only essential to pragmatics, it must be its defining feature. Grammar and pragmatics are distinct. Grammar consists of a set of codes; pragmatics accounts for inferences accompanying those codes.
25.2 No Grand Design behind Grammar vs Pragmatics The upshot of section 25.1 is that the grammar/pragmatics divide must be based on a single code-versus-inference distinction. But we may want to ask why this is so. Why do the quite reasonable criteria in Table 25.1 not converge? Why can we not predict ahead of time and for all languages which topics/aspects of language use belong in grammar and which in pragmatics? Above all, what does that teach us about grammar and pragmatics? The simple answer is that there is no ‘grand design’ behind grammar, nor behind pragmatics. Today’s grammar is yesterday’s pragmatics. Grammar evolves out of pragmatically motivated discourse patterns: initially inferred interpretations may
481
Pragmatics and Grammar: More Pragmatics or More Grammar 481 conventionalize (gradually, over time). If so, the very same phenomena are pragmatic at one point in time and/or for one language, but grammatical at another point in time and/or for another language. Crucially, aspects of language use cannot be ‘inherently pragmatic’ if they are potentially grammaticized. Still, not every pragmatic pattern grammaticizes. Why is it not the case, then, that what potentially grammaticizes is distinctive of ‘grammar’, and what does not undergo grammaticization is ‘inherently pragmatic’, in which case, at least grammar can be given some essential content? And if this weaker claim cannot be supported, does this mean that grammaticization is arbitrary? I would say no. Grammaticization is motivated, despite the fact that it ruins the ideal division of labour commonly assumed for the grammatical vs the pragmatic (as shown in Table 25.1). It is just that the particular distinctions philosophers and linguists were so keen to attribute to grammar and to pragmatics respectively are orthogonal to grammaticization processes (and to grammar and to pragmatics). There is no conceptually defined overarching principle guiding what gets into grammar and what is left for inference. Grammar codes (best) what speakers do most (Du Bois 1987: 851). The current grammar includes those patterns which previously constituted salient and frequent discourse patterns. Salient and frequent discourse patterns are not accidental facts. The messages consistently expressed by the same linguistic means (the source for grammaticization) are those which are near and dear to speakers. So, what ends up in grammar is not at all arbitrary. Still, there is no reason for meaning and use aspects associated with a specific topic to all form highly consistent and salient discourse patterns. Similarly, there is no reason for all meaning and use aspects associated with a specific topic to constitute ad hoc inferred uses. This is why not all Gricean implicatures are pragmatic. Some are conventional and hence grammatical. Agreement, on the other hand, may be governed by pragmatic factors, such as contextual considerations whether a couple counts as a single unit or as two individuals (see ex. (3) again). In fact, why should all aspects relevant to deixis (anaphora, presupposition, speech acts, etc.), for example, uniformly manifest a consistent discourse pattern which would render them potential grammar material? Or, why should they all be less than perfectly consistent, i.e. ad hoc interpretations which do not form any consistently enough salient discourse pattern, in which case they must remain pragmatic? There is no reason to expect a uniform grammar or pragmatic patterning for whole topics.3 Take they, for example. Of course, some of the interpretations associated with they are encoded: more than one third-person entity is intended; the mental representation standing for these entities is highly accessible to the addressee. But they are partly inferred: most probably the entities are human; most probably the entities are not a combination of human and inanimate entities. They’s association with plurality and high accessibility for the entities is automatic. Its association with humanity and ‘non- mixture’ is not as strong. But there’s nothing inherently pragmatic or inherently grammatical about e.g. the humanness of personal pronouns. Such ‘inconsistent’ states of affairs are the norm for linguistic expressions, rather than the exception. If there’s no 3 See Ariel (2008: section 2.1) for an analysis of presuppositions that combines code and inference, and Ariel (2010: chs 6–8) for such non-unitary analyses for practically all the classical pragmatic topics.
482
482 Mira Ariel inherent or necessary connection between form/function correlations and either codes or inferences, it is not surprising that neither grammatical nor pragmatic associations are fixed for all times or for all languages. Hebrew hen ‘they-fem’, for example, denotes human referents much more often than the counterpart masculine form (Gafter 2008). It is not impossible that the association between hen and female, specifically human, third persons will be entrenched into grammatical code. Case marking provides another such example. Proto-Indo-European had eight grammatical cases, Old English showed three cases, and French and Spanish have no case marking. How do speakers figure out, then, who does what to whom? Word order may help, but pragmatics is always there too. We can easily draw on our general knowledge to infer who was most likely to do what to whom. But why was a piece of grammar lost to pragmatics? Again, no grand design lies behind this change. Just phonetic erosion (of the case markers). Grammatical codes are here today, gone tomorrow. And they may be back again! It does not take a grand design to determine what is in the grammar and what remains pragmatic. All it takes for something to become grammatical is repeated experience with some task. Indeed, fMRI studies show different patterns of brain activation for performing the same task (such as proposing appropriate verbs for nouns), depending on whether the task is novel or practised (Raichle 1998; McCrone 1999). Moreover, the difference is not only quantitative: it is not just that the task is performed more rapidly due to practice. The difference is qualitative, involving different brain regions. Pragmatic patterns can and do cross over to the grammatical territory when they become entrenched. Their ‘essence’ doesn’t change in the process. The nature of the form/function association does. A special Hebrew construction, maca et moto ‘found one’s death/died’, very clearly demonstrates how one and the same expression encodes some of its functions, but leaves to (rather frequent) inference another function. Consider the first example, (9), where the journalist doesn’t deny that the monk died, but rather, that the circumstances of his death (who killed him) are not clear: (9) Moderator: We’re hearing now about a monk who found his death (= died). Journalist1: He didn’t find his death but rather, he was killed by IDF forces (Hebrew, Reshet Bet, 4.2 2002) Obviously, for (9) to be acceptable, ‘found his death’ must somehow convey that it is not clear how the person died. What else does the expression convey? We can test it by examining other potential responses to the announcement in (10):4 (10) ~ Journalist2: No, he only got injured. ~ Journalist3: No, he committed suicide. ~ Journalist4: No, he passed away at a ripe old age. Based on all the responses above, Hebrew ‘found his death’ must convey the following four interpretations: 4
~ indicates an invented example.
483
Pragmatics and Grammar: More Pragmatics or More Grammar 483 (11) a. Died (Journalist2) b. It is unknown who is responsible for the death (Journalist1). c. The person who died bears no responsibility for their death (Journalist3). d. Death was unnatural, unexpected at the point in which it occurred (Journalist4). Now, what is the status of these meanings? Is it unified? The cancellability criterion shows that interpretations (12a,b,c) cannot be pragmatic, for they are uncancellable: (12) a. ?? He found his death and didn’t die. b. ?? He shot himself and found his death. c. ?? He found his death at a ripe old age. Note, however, that interpretation (b) is cancellable: (13) Florianus, 276, found his death at the hands of his soldiers. (wikipedia.org/ wiki) If so, some of the interpretations associated with the very same Hebrew ‘found his death’ are semantic (‘died’, ‘no responsibility of the dead’, and ‘death is unexpected’), but an additional common interpretation is only inferred (‘it is unknown who/what is responsible for the death).5 All in all, we can now better understand why the last decades have seen the descent of the ‘miraculous’ semantics view, according to which it is the precisely compositional meaning, no more no less, that gives rise to the truth-conditional meaning of the proposition. Note that this descent was accompanied by a parallel ascent of the thesis of the underdeterminacy of grammar, whereby pragmatics is a major contributor to propositional content (see especially Carston 2002). Grammatically and pragmatically governed phenomena are not essentially different from each other. It is the form/function associations they maintain that are different.
25.3 Clear Definitions Do Not Necessarily Lead to Unequivocal Boundaries Now, one would think that equipped with the very clear code-versus-inference definition outlined in the previous section, it would be straightforward to apply the grammar/ 5 Should we posit a further (orthogonal) distinction between two types of encoded meanings, that is, between the truth-conditional ‘died’ and the non-truth-conditional ‘no responsibility’ and ‘unexpected death’? See Ariel (2010: section 9.6) for discussion.
484
484 Mira Ariel pragmatics division of labour to specific cases. But that is not invariably so. Consider reference resolution according to relevance theory (Kempson 1984; Wilson 1992), neo- Gricean theory (Levinson 1987b; Huang 2000a), givenness theory (Gundel et al. 1993), and accessibility theory (Ariel 1990, 2001). All these approaches accept that grammar alone cannot account for all aspects of the phenomenon. Pragmatics is heavily implicated. But what kind of pragmatic work is involved exactly? Relevance theoreticians emphasize the ad hoc nature of the inferences required in reference resolution, which, they propose, are derived based on the principle of relevance. While the other three approaches do not at all underestimate the major role of context-specific inferences guiding reference resolutions, they have each argued for some general pragmatic procedure mediating such inferences: the Q-, I-, and M-principles for the neo-Griceans, givenness status for Gundel et al., and degree of mental accessibility for Ariel. In addition, while neo-Gricean and relevance-theoretic approaches assume the conventional semantic meanings apply to referring expressions, givenness and accessibility theories have proposed to enrich the grammatical code of referring expressions. Referring expressions each encode some givenness status according to Gundel et al., and they each encode a relative accessibility status according to Ariel. But the neo-Griceans too have extended the role of the extragrammatical. Levinson and Huang, and following them, accessibility proponents have argued that grammatical aspects of sentential anaphora (the binding conditions), although grammatically specified, are externally motivated (by their respective theories). Finally, the neo-Griceans, Gundel et al., and Ariel all view referring expressions as constituting an ordered set based on extra-grammatical principles. The scale is arranged according to informativity and markedness for Levinson and Huang. Gundel et al.’s scale is arranged according to degrees of givenness. Ariel’s scale is based on the degree of mental accessibility indicated by the referring expression, which is for the most part a function of how informative, rigid (uniquely referring), and phonetically attenuated the specific form is. The neo-Gricean scale is rather limited (Ø, reflexive pronoun, personal pronoun, and definite lexical NP are the forms most often discussed). The givenness scale is richer (incorporating demonstratives in addition, and most notably, indefinite NPs as well). The accessibility scale is in principle infinite, for it is based on three universal form/function correlations (plus some language-specific grammatical arbitrariness in addition). Relevance theoreticians have no use for any particular ordering of referential forms. Each referring expressions is used according to its semantics, while at the same time taking into account the specific context and the relevant pragmatic inferences it gives rise to. It therefore treats as epiphenomenal the ordered scales of referring expressions argued for by each of the other approaches. Consider my accessibility theory account (Ariel 1988 and onwards). On my account, in addition to the obviously grammatically encoded content of referring expressions, they also encode a specific degree of accessibility. This degree of accessibility, I have argued, while not arbitrary, cannot simply be inferred from formal properties of referring expressions. This is why zero pronouns, for example, do not encode the same degree of accessibility in different languages. Chinese zero pronouns encode a wider
485
Pragmatics and Grammar: More Pragmatics or More Grammar 485 range of high accessibility, wider than the range of zero pronouns acceptable in Hebrew and certainly in English (see Ariel 1990; Huang 1994). Here are some of the expressions listed on my accessibility scale: (14) Full name+modifier > full name > long definite description > short definite description > last name > first name > distal demonstrative+modifier > proximate demonstrative+modifier > distal demonstrative + NP > proximate demonstrative + NP > distal demonstrative (− NP) > proximate demonstrative (− NP) > stressed pronoun+gesture > stressed pronoun > unstressed pronoun > cliticized pronoun > verbal person inflections > zero Now, what’s the nature of this scale? How does it relate to determining whether coreference or disjointness was intended? How does it help addressees determine which NP was the intended antecedent? According to Reboul (1997), relevance theory is in no need of such a scale, for the use conditions associated with these expressions can rationally be derived from the semantic content of the expressions. In other words, even if the scale correctly accounts for reference resolution, it is an epiphenomenon. According to neo-Griceans (who restrict themselves to a rather shorter list of items on the scale), the scale is based on a generalized conversational set of principles responsible for linguistic choices in general. Their idea is that coreference is a preferred choice, and tends to be associated with uninformative forms. Disjointness is associated with informative forms. I have proposed that for the most part the ordering on the scale is not arbitrary. For example, the difference between definite descriptions and pronouns and pronouns and zeros is easily pragmatically derived. In fact, I have proposed three criteria for determining the degree of accessibility associated with referring expressions: informativity, rigidity, and attenuation. All of these converge to predict the three-way difference between definite descriptions, pronouns, and zeros. Definite descriptions are more informative than pronouns, which are in turn more informative than zeros. Definite descriptions are also more rigid than pronouns, which are more rigid than zeros, in that the expressions restrict the choice of antecedent more: the antecedent has to meet the description specified by the definite description, so if it is the teacher, all non-teachers are excluded. For (gendered) pronouns the antecedent is restricted to the specific person (and gender). Not so for zeros. Finally, phonetically, definite descriptions are least attenuated, pronouns have a small phonetic size, while zeros are maximally attenuated. So, many of the relative distinctions can be pragmatically predicted on the accessibility account. But not all of them can. For example, Hebrew hu ‘he/it’ and ze ‘this one+/-human’ are equally informative, rigid, and attenuated, but the former encodes a higher degree of accessibility. So this is a grammatically specified difference. Moreover, had the scale been completely pragmatic we would not have expected there to be differences between the use of various expressions cross-linguistically. Given that, e.g., Chinese, Hebrew, and English all have both pronouns and zeros, and assuming the difference in their use is pragmatically motivated, one would assume that they are used quite similarly. After all, pragmatic considerations cannot be language-specific. But the
486
486 Mira Ariel conditions under which the three expressions are used in the three languages do differ, such that zeros occupy the largest niche in Chinese, an intermediate one in Hebrew, and a smaller niche in English.6 For further discussion of the grammar/pragmatics divide with regard to reference, see Ariel (1990, 2001: ch. 2, 2012).
25.4 Do We Now Have More Grammar or More Pragmatics? Traditionally, the grammar/pragmatics border skirmishes were such that researchers were trying to maximize pragmatics at the expense of grammar. The neo-Gricean proposal to partly reduce the binding conditions to pragmatic principles is a classic example. But I discuss here two other cases: what I call ‘ugly facts’, where seemingly pragmatic phenomena must be relegated to grammar (4.1) and quantifier most, where I have proposed a reshuffling between the grammatical and the pragmatic (4.2).
25.4.1 Ugly facts: More grammar, less pragmatics Ugly facts are facts that ruin a beautiful picture. The beautiful picture in our case is an analysis where anything that can in principle be left to inference IS in fact inferred, following Grice’s Modified Occam’s Razor Principle of economy (see section 25.1). If so, semantic monosemy, accompanied by pragmatically derived polysemy is beautiful. Semantic polysemy is ugly. Let’s consider English after all, as in: (15) MONTOYA:
… how can Doctor ~Montoya say that Latinos have no no real power, After all, they have two people that are cabinet members, (SBC: 012)
After all1 presents given information and supports another discourse proposition (Ariel 1985; Blakemore 1987). It favours non-final, unstressed position. Next, here’s after all2: (16) REED: DARREN: 6
I mean he’s gonna paint for a year, and just see whether he can do well enough with that to— Cause he doesn’t want to be a doctor after all? (SBC: 046)
Note that it is not true that English lacks zero anaphora. Consider: i. And Stragwaysi’s friends at hisi club say hei was perfectly normal. Øi left in the middle of a rubber bridge—Øi always did when … (Dr. No, Ian Fleming 1958: 26)
487
Pragmatics and Grammar: More Pragmatics or More Grammar 487 After all2, which presents an unexpected conclusion, favours final position and focal stress. Now, are these interpretations two semantic senses, or can we analyse after all as encoding one semantic meaning, deriving the other (or both) via pragmatic inference? I have opted for a semantic ambiguity analysis, but Fretheim (2001) has proposed an ingenious monosemic analysis. On his proposal, the semantics of after all instructs the addressee to construe a premise–conclusion relation between the after all proposition and some other contextually relevant proposition. Both after alls are said to participate in an inferential chain. It is pragmatics that determines whether the after all proposition functions as a premise (the given after all1) or as a conclusion (the unexpected conclusion after all2). In (15) the after all proposition serves as a premise, justifying the speaker’s rhetorical question in the previous utterance (the conclusion). In (16), the after all proposition is the conclusion from the previous speaker’s utterance. No doubt these relations are pragmatically derived, by reference to the specific context. But there is one use condition which we cannot actually derive pragmatically, namely, the givenness of the information under after all1. Note that we cannot derive it from the fact that it is a justification, because justifications are not necessarily given: (17) MARILYN: But you still have to clean off that table cause it’s grody. ROY: Which table. This table here? (SBC: 003) While ‘it’s grody’ serves as an appropriate justification for Marilyn’s request that Roy clean the table, the fact that the table is grody is definitely not given here (see Roy’s response). Indeed, given Roy’s state of knowledge, Marilyn could not utter: (18) ~?? You still have to clean off that table. Ø/??after all, it’s grody. Givenness cannot always be inferred ‘for free’ here. If pragmatics cannot deliver the givenness condition, semantics must, and the conclusion is that a larger grammar is needed for after all (two lexical meanings, rather than monosemy + inferences). After all is not alone. Ugly facts are everywhere. Consider (19) versus (20) and (21): 19. ~Waiter: a. Would you like anything else? b. Will there be anything else? 20. ~Host: a. Would you like anything else? b.??Will there be anything else? 21. ~WaiterHebrew: a. Would you like anything else? b. ??Will there be anything else?
488
488 Mira Ariel Note that there is a difference between what the waiter and the host can say to a guest they are offering to serve some more food to. (20b) sounds too businesslike for the host, but then, for some reason, its Hebrew equivalent is not appropriate for the waiter speaking Hebrew either. The association of a service encounter with different linguistic constructions constitutes an ugly fact. It is one that would in principle seem to be derived based on the form and the circumstances under which it was used, but it cannot in reality. The same goes for the seemingly transparent expression clothes line, which is rendered by the equivalent of ‘laundry line’ in Hebrew. While each expression is well motivated, and may initially have been pragmatically derived, it no longer is. Convention has set in to associate the two expressions with their meaning now. So, not everything that could in theory be inferred is in fact inferred in real-time interactions. All these ugly facts require a larger grammar (semantics) than the optimistic Gricean would like to assume. In this respect, the grammar/pragmatics division of labour here supported calls for more grammar and less pragmatics.
25.4.2 Most: A grammar/pragmatics swap The last option for a redrawing of the grammar/pragmatics division of labour involves no change in the size of the territory covered by each, but a swap in their roles: What is commonly assumed to be pragmatic is reanalysed as a piece of grammar, and what is assumed to be grammatical is reanalysed as pragmatic. This is what I have proposed for the analysis of the quantifier most (Ariel 2004, 2008: 3.2), as in: (22) Poll: Most Israelis, Palestinians support Geneva Accord. (Haaretz, English edition, 24 November 2003) Proponents of the classical division of labour for most (and other scalar quantifiers) have argued that the semantic meaning of most is ‘more than 50 per cent’, which means that ‘100 per cent’ is one of the values it can denote (Horn 1972, 1989). Here’s an example that shows that most is indeed sometimes compatible with ‘all’: (23) The target date for the meeting is Jan. 17 in Los Angeles, provided most of the Hall of Famers can make it. (International Herald Tribune, 24–5 December 2002, p. 16) Surely the speaker here commits to holding the meeting should all the Hall of Famers be able to attend. But of course, the typical use of most is as ‘less than 100 per cent’, and the upper bound is assumed to be pragmatically derived. It is a generalized conversational implicature for neo-Griceans, based on a comparison with the stronger, more informative all that the speaker could have used, but chose not to. If she didn’t choose all, she must not have intended ‘all’. According to relevance theory, the upper bound is
489
Pragmatics and Grammar: More Pragmatics or More Grammar 489 an explicated interpretation, the result of a more general narrowing process of the very broad 51–100 per cent semantic meaning. My own circumbounded analysis posits a lower as well as an upper-bound semantic analysis. On my analysis, most Palestinians support peace means that there is a proper subset of Palestinians, larger than 50 per cent who support peace (the set being a proper subset assures that the speaker only commits to less than 100 per cent). Examples such as (22) are then straightforwardly accounted for, with no need for a pragmatic upper bound. Such cases, I should emphasize, are the great majority of most cases. The ‘possibly all’ reading of (23) actually poses no problem, for any expression can sometimes be interpreted with an ‘at least’ interpretation, numbers included, as in (24), where 75 is understood as ‘at least 75’: (24) ~The target date for the meeting is Jan. 17 in Los Angeles, provided 75 of the Hall of Famers can make it. On the circumbounded analysis, there are however (rare) cases where an ‘all exclusion’ pragmatic inference is derived in addition, as in (25): (25) Michaeli: We’ll talk about EVERYTHING. Gaon: About most things. (Channel 2, 4 April 2002, originally Hebrew) I distinguish between the prototypical cases (as in (22)), where only a proper (majority) subset is profiled, and (25), where the context sets up a contrast between ‘all’ and ‘most’, thereby also profiling the complement set to the majority reference set. Here I agree with the classical analysis that ‘all exclusion’ is actually conveyed by the speaker. Note that ‘all exclusion’ entails ‘proper subset’, which is how the classical analysis derives the upper bound, but ‘proper subset’ (the semantic analysis on the circumbounded theory) does not entail ‘all exclusion’. To see that we need to make a distinction between an upper-bounded reading (as in (22), lexical on my account) and an ‘all exclusion’ reading (as in (25), pragmatically derived on my account), see (26): (26) The majority decided for peace. Me too. (bumper sticker, originally Hebrew, spotted April 2002) According to the classical analysis, since the majority in (26) is interpreted as ‘less than all’ a ‘not all’ implicature must have been generated. But could the speaker have really intended ‘most but not all’ in (26)? Not very likely, for while ‘a majority’ is seen as a solid argument for following the subset who decided for peace, ‘not all’ most certainly is not. Consider the less acceptable: (27) ~? The majority, but not all, decided for peace. Me too. Most utterances are overwhelmingly used to support speakers’ arguments (Ariel 2004) based on the majority reference set. The complement minority set actually poses
490
490 Mira Ariel Table 25.2 The semantic/pragmatic swap for most Semantic
Pragmatic
Classical
‘more than half ’ (23)
‘not all’ (22, 25)
Circumbounded
‘A proper subset, larger than 50 per cent’(22)
‘not all’ (25) ‘more than half ’ (23)
a counterexample to the speaker’s point, and cannot then be a speaker-intended message. Table 25.2 shows the semantic and pragmatic contributions under the two analyses. I indicate in parentheses the relevant example numbers. Note that while (25) requires a pragmatic ‘all exclusion’ interpretation on both analyses, (22) is pragmatically derived on the classical analysis, but is semantically accounted for on the circumbounded analysis. (23), on the other hand, is handled semantically by the classical analysis, but requires pragmatic inferencing on my analysis. Here are three questionnaire results which support the circumbounded analysis (I indicate in parentheses how often the response was confirmed by my subjects): (28) Most high-school students drink alcohol. Which of the following cases could the speaker mean?
A. 80% of high-school students (90.6%) B. 50% of high-school students (10.9%) C. 100% of high-school students (6.25%) D. 28% of high-school students (0%) E. None of the above (12.5%).
Note that the subjects responding to this question saw no difference between the lower bound and the upper bound. They equally rejected violations of both boundaries. This is surprising on the classical analysis, where 100 per cent is semantically supposed to be as likely as 80 per cent is as the value intended by the speaker. At the very least, the 100 per cent value should have been rejected much less often than the 50 per cent value, where the semantic meaning is violated.7 In other words, the semantically anomalous 50 per cent and the pragmatically anomalous 100 per cent are equally rejected, which is surprising, given that only semantic violations are supposed to lead to non-acceptability. Next, I tried to force subjects to accept the 100 per cent value by not providing any other semantically viable option: 7 Note that in all the questions above, subjects were strongly encouraged to choose as many possible responses as they could. They did not have to avoid the 100% option just because they also circled the 80% option, and indeed, very often they did accept multiple options (see ex. (30), and see Ariel 2004 for many more results).
491
Pragmatics and Grammar: More Pragmatics or More Grammar 491 (29)
Most of the students in the class were born in 1970. How many students could the speaker mean? A. 100% of the students (9.4%) B. 20% of the students (6.25%) C. 50% of the students (15.6%) D. 49% of the students (9.3%) E. None of the above (81.25%).
Again, the supposedly pragmatically banned 100 per cent value was seen to be as anomalous as the semantically anomalous values. Such findings show that the upper bound is not just a pragmatically derived interpretation, which is therefore cancellable. My subjects ‘refused to cancel’ the supposedly ‘not all’ implicature even when that was the only legitimate way to justify the use of most according to the classical analysis. Finally, the results from the following question also argue against the relevance- theoretic proposal that there is nothing special about the upper-bound interpretation, for it is derived as part of a routine process of narrowing of semantic meanings. First, note that indeed, such a narrowing is relevant to most. There is a difference between the majority we envision for (22) (subjects like 80–85 per cent) and the majority we envision when we hear that most babies born in California are Hispanic (just above 50 per cent would be most people’s choice here). But what would motivate subjects to rule out 100 per cent in the following question? Why is 99 per cent so very different from 100 per cent, if both are included under the concept of ‘majority’? 30.
An overwhelming majority of the students passed the test. What percentage of students may have passed the test according to the speaker’s sentence? A. 97% (100%) B. 98% (95.8%) C. 99% (95.8%) D. 100% (4.2%)
It is strange that narrowing down the value of ‘an overwhelming majority’ should distinguish so very sharply between 99 per cent and 100 per cent, which are as different from each other as 97 per cent, 98 per cent, and 99 per cent are from each other. There is, then, something special about ‘less than all’ for most. I’ve proposed that this interpretation is semantic, rather than pragmatic. In sum, on the circumbounded analysis scalar quantifiers involve more grammar regarding the prototypical upper bound (a pragmatic > semantic shift). More pragmatics, however, is necessary to derive the less common reading of ‘possibly all’ (23). There is no difference with respect to the equally rare ‘all exclusion’ reading (25).8
8
Note that both accounts require a pragmatic inference on top of the semantic analysis here.
492
492 Mira Ariel
25.5 Conclusions: The Grammar/ Pragmatics Division of Labour I have proposed that the grammar/pragmatics division of labour must boil down to a single code-versus-inference distinction. On this view, grammar specifies a set of codes, while pragmatics provides a set of context-dependent inferences. Such pragmatic inferences may very well contribute towards the truth-conditional meaning of the proposition. However, despite this very clear grammar/pragmatics division, I have tried to emphasize that what should be analysed as code and what as inference is often not self-evident. Such decisions must be based on a case-by-case analysis, with the result that interpretations, as well as use conditions, previously analysed as part of grammar may be reanalysed as pragmatics; and vice versa, aspects of use and interpretation previously analysed as pragmatic may turn out to be grammatical.
493
Chapter 26
Pragmati c s a nd Morph ol o gy Morphopragmatics Wolfgang U. Dressler and Lavinia Merlini-Barbaresi
26.1 Pragmatics and Morphology An old tradition among linguists is that of contrasting pragmatics and grammar (including morphology) in a dichotomous way, even claiming that right-handers process grammar in the left hemisphere of the brain, pragmatics in the right hemisphere (cf. Grodzinsky 2000; but see Stemmer 1999b; Perkins 2007). Another time-honoured tradition is that of deriving pragmatic from semantic meanings as secondary nuances, even in morphopragmatics (e.g. Kiefer 2004). In contrast, our contribution assumes a direct interface between pragmatics and morphology, and following the semiotics of Morris (1971), a priority of pragmatics over semantics.
26.1.1 A brief history of research previous
to the establishment of morphopragmatics
An interface between pragmatics and morphology has been assumed from the beginning of pragmatic research on politeness (cf. Leech 1983; Levinson 1983; Braun 1988). This will be discussed later in the section on inflection. As to word formation, pragmatic studies have concentrated on diminutives. Two important precursors to a pragmatic approach in the study of diminutives in Romance languages were Leo Spitzer (1921) and Amado Alonso (1933/1961). To Spitzer (1921: 201– 202), we owe various important observations on the ludic, emotional character of
494
494 Wolfgang U. Dressler and Lavinia Merlini-Barbaresi diminutives and, more importantly, the fundamental conception of ‘sentence diminutive’, i.e. a diminutivized word which extends its meaning scope to the entire sentence, a notion taken up by Alonso and later by the present authors (cf. Dressler and Merlini Barbaresi’s 1994 concept of the ‘landing site’ of diminutive formation in a speech act). Such a meaning is clearly non-semantic. Alonso (1933/1961) is a pioneer in the pragmatic analysis of diminutives, having produced the richest and most pragmatically oriented study, at a time when pragmatics was not yet a discipline. He downgrades the denotative meaning of smallness, in favour of emotional values whose meanings and effects depend on the context, the participants’ attitudes and the type of speech act (again ante litteram). He left a rich inheritance, especially for his accurate survey of emotions, which was the source for various ‘emotionalists’, as for example Gaarder (1966) on Mexican diminutives and augmentatives and later Volek (1987), and many others. Grabiaś (1981), in his study of Polish diminutives and augmentatives, relies on important intuitions about speaker’s evaluation and emotional meanings. Klimaszewska, in her contrastive study on Dutch, German, and Polish diminutives, consciously adopts a pragmatic perspective (1983: 2) but she co-identifies them with expressive connotations (1983: 6, 27, 30). Nieuwenhuis’s study (1985) in his unpublished doctoral dissertation provides a wealth of cross-linguistic data, but he is not specifically focused on pragmatics, although a large number of his examples would deserve a pragmatic analysis. Volek, in what she calls a ‘pragmatic analysis’ (1987: 149–175) of Russian emotive signs (diminutives), makes various relevant observations. Most importantly, she distinguishes between the ‘emotive attitude’ towards the phenomenon named in the base vs phenomena not named in the base, but rather inherent in the addressee or the speech situation. She interprets the first type of attitude as due to a connotative meaning feature of the diminutive word base, and classifies the second type as aspects statically pertaining to the speech situation, with a prejudice to a notion of pragmatic meaning as dynamically obtained in the course of the speech event via the application of the morphological rule (MR) of diminutive formation. Wierzbicka (1984, 1991) carries out many important analyses of the illocutionary meanings of diminutives and other morphological devices, but in her theoretical configuration she doubts the rational existence of clear borders between referential/denotative and pragmatic/attitudinal meanings and preferably handles pragmatic meanings within her semantic framework. More pertinent observations, although not fully developed, on the pragmatic roles of diminutives and other morphological elements are found in Bazzanella, Caffi, and Sbisà (1991). For emotive signs in general, cf. Ochs and Schieffelin (1989).
26.2 Morphopragmatics The theory of morphopragmatics was pioneered by the present authors in successive steps (Dressler and Merlini Barbaresi 1986, 1987, 1989), and expanded into a full-fledged
495
Pragmatics and Morphology: Morphopragmatics 495 model in 1994 which precisely covers the area of intersection between morphology and pragmatics. Other early applications are Dressler and Kiefer (1990) on German and Hungarian excessives, Kilani-Schoch and Dressler (1999) on the French -o suffix, Crocco Galéas (1992) on the pragmatic difference between learned (originally humanistic latinized) and usual Italian ethnics such as Patavino vs Padov-ano ‘from Padova’, Merlini Barbaresi (2001) on the English -y/ie suffix, and Cantero (2001) on various Spanish morphopragmatic elements. Morphopragmatics is an area of integration of grammar and pragmatics, where morphological rules interact with pragmatic conditions. More precisely, it describes grammatical morphological operations, capable of systematically contributing autonomous pragmatic meanings to the speech act, i.e. regular pragmatic changes that take place when moving from the input to the output of a morphological operation, be it derivational or inflectional. It can be paralleled with the following well-established sub-areas, such as morphosemantics, lexical semantics of morphology, lexical pragmatics of morphology, and pragmatics of syntactic patterns and textual strategies, but it is to be carefully distinguished from them: (a) Morphosemantics studies the semantic meanings of morphological rules, i.e. the regular denotational and connotative modifications obtained by derivational or inflectional rules. Within a morphosemantic investigation, pragmatic variables connected with speech situations become irrelevant. Reference to a pure denotative meaning of smallness added by a diminutive suffix, as in flat-let, belongs here. (b) Lexical semantics of morphology deals with the denotative and connotative meanings of single, morphologically complex words, such as lexicalized star-let. (c) Lexical pragmatics deals with the pragmatic meanings idiosyncratically conveyed by single complex words, like lexicalized bunn-y ‘rabbit’, selecting a child environment and obtaining a meaning of endearment. (d) Syntactic patterns and textual strategies may convey pragmatic meanings of their own and interfere with those obtained by single constituents of the text. Morphopragmatics deals, instead, with pragmatic meanings that can be regularly obtained through the sole application of morphological rules, given certain sets of contextual conditions. The authors characterize the subfield of morphopragmatics as the end-point of a diachronic and synchronic process of grammaticalization of pragmatic phenomena. Specifically, they configure a level of morphologized pragmatics, which is meant to cover the area of the general pragmatic meanings of morphological rules. A main goal of this theory is to demonstrate their autonomy in conveying pragmatic meaning. The privileged objects of a morphopragmatic description are evaluative suffixes, such as diminutives and augmentatives, on which the authors based their main
496
496 Wolfgang U. Dressler and Lavinia Merlini-Barbaresi argumentation, but also elatives, (It. -issimo), reduplicatives, excessives (as Ge. das Aller-schlecht-este exc-bad-superl ‘the very worst of all’), and, within inflection, personal pronouns of address and Japanese honorifics, which basically interact with the same factors relevant for evaluatives. A morphological rule is relevant for morphopragmatics if it contains a pragmatic variable that is necessary for the description of its meaning. The premise is in line with the authors’ more general theoretical claim that pragmatics is a superordinate of semantics (following Morris 1971). This implies that the pragmatic meaning involved cannot be subsumed by the semantic meaning of the expression. In Italian diminutives, for example, the denotative meaning [small], with its allosemes [unimportant] and [young], even combined with connotations of the type [emotional, pleasant, etc.], would only characterize the word base, but would be unfit for expanding to the entire utterance and speech act and for characterizing the communicative situation pragmatically. It is proposed, instead, that morphopragmatic effects of diminutives and augmentatives be based on a general feature [fictive], which implies immediate reference to the component of the speaker’s attitude in the speech event. Fictiveness directly derives from the characterization of these morphological means as evaluative, that is, from the notion of the speaker’s transition from the real to an imaginary world where the conventional norms on objects and states are suspended in favour of a subjective, negotiable judgement. Fictiveness confers dynamicity to the theory because it is capable of immediately generating pragmatic meaning, given a favourable set of situational circumstances. In diminutives, fictiveness is further specified as a character [non-serious], which is responsible for the majority of their meanings in discourse (e.g. imprecision, attenuation, but also irony, meiosis, etc.) and which, in general, indexes a speaker’s lowered responsibility and entails lower distance between speaker and addressee. The component of the speaker attitude is interestingly developed and modelled in later studies, for example in Fradin’s (2003) notion of the ‘interlocutor’s pole’ in the meaning of diminutives, which he correctly states to be unimportant for French diminutives (cf. Dressler 2010). This emphasis on the relation to the addressee is clearly more relevant for the understanding of morphopragmatic phenomena than Reynoso Noverón’s (2005) focus on subjective attitudes of the speaker.
26.3
Word Formation
In this section we are going to describe the word formation mechanisms that are most eligible for expressing pragmatic meanings. Thus, we will privilege those rules that manipulate meaning and form in a regular, predictable way, i.e. that operate within grammar. Extra-grammatical operations will be more briefly described in section 26.3.4.
497
Pragmatics and Morphology: Morphopragmatics 497
26.3.1 Evaluative suffixes Among the morphological objects of this description, evaluative suffixes (also called alteratives), such as diminutives, as in Sp. hasta lueg-ito, ‘good-bye-DIM’, Ge. Papp-erl ‘meal-DIM’, augmentatives, as in Port. animal-aço ‘huge animal’ and in It. grass-one ‘fat-AUG’ and pejoratives, as in It. om-accio ‘man-PEJ’, are the best representatives, as they appear capable of holding a direct and exclusive relationship with pragmatics. Morphopragmatics, the theoretical model presented here, is a framework specifically intended to shed light on cases of grammaticalized (morphologized) pragmatics, i.e. pragmatic meanings/effects produced by grammatical operations using morphological means. Evaluative suffixes are capable of modifying the denotative semantic meaning of their bases in terms of dimension (diminutives and augmentatives) or in terms of quality (pejoratives). They may be totally responsible for the added utterance meanings, with the word base being either neutral (book-let) or contributory (dear-ie, It. piccol-ino ‘small- DIM’) or even contrary (Sp. gord-ita or It. grass-ina ‘big-DIM’) to the effect pursued. Such suffixes are considered evaluative because they imply a type of judgement or attitude of the speaker. More often than expected and recognized, the semantic meaning they convey only accounts for a limited part of the overall meaning modification they obtain. Evaluatives may also confer to their bases (and the entire utterance) a vast range of connotative and pragmatic meanings, exclusively or in addition to the semantic modification, and these co-vary with contextual and discursive variables. Pragmatic meanings are especially conveyed during informal interactional discourse (in-group or intimate), in which participants’ attitudes, emotions, and beliefs are foregrounded. The type and intensity of these meanings/effects are regulated by the participants’ epistemic commitment and evaluative judgements.
26.3.1.1 Diminutives Among evaluative suffixes, diminutives are the unmarked evaluative category, i.e. the presence of a productive category of augmentatives in a language implies the presence of diminutives (cf. Grandi 2011). They are almost universally represented cross- linguistically, at least all languages possess the subcategory of hypocoristics, which are productively used at least in the discourse types of child-directed, pet-directed, and lover-directed discourse and related discourse types. And the main function of hypocoristics as derived from names (e.g. Liz, Tomm-ie) or from name-like uses of nouns denoting close relatives and friends (e.g. mum(m-y)) is pragmatic. Moreover there is a large overlap between markers of hypocoristics and of diminutives. Still we can distinguish hypocoristics and proper, non-hypocoristic diminutives in some languages formally too. For example, in German, the diminutives of the feminine noun die Mutter ‘the mother’ are the neuters das Mütter-chen/lein, Austrian Ge. Mutter-l, but the hypocoristic derivation referring only to the speaker’s own mother does not transform the feminine into a neuter and truncates the unstressed final syllable: die
498
498 Wolfgang U. Dressler and Lavinia Merlini-Barbaresi Mutt-i. Similarly, from the adjective lieb ‘dear’ or its syntactic, gendered nominalization der/die Lieb-e ‘the dear one’ is derived the neuter diminutive das Lieb-chen ‘the beloved girl/woman’ and the hypocoristic referring only to the speaker’s boyfriend: der Lieb-i. This allows us to propose a second implication: if a language has diminutives, it also has hypocoristics, which implies also the basicness of pragmatic meanings. Diminutives are the morphological mechanism that best exemplifies the variety of relevant morphopragmatic meanings. They allow the largest number of strategic uses in speech acts and speech situations, in which they obtain a range of predictable pragmatic effects, given certain circumstances. Emotionality in child/pet/lover-centred speech situations, but also the ludic character of playfulness among intimates, familiarity, and informality in general, sympathy and empathy and also understatement, euphemism, false modesty, irony, and sarcasm, are the circumstantial factors that favour a pragmatic use of diminutives. Diminutives are also effective devices for politely hedging or downgrading requests, mitigating orders and assertions, for increasing attractiveness in offers and invitations, etc. Here follows a limited sample of such uses (in a variety of European languages, recorded by the authors, or taken from the Web or from books, but all certified): 1. Emotion/tenderness: Italian (to a little child) (1) Adesso la mamma prepara la vasch-ett-ina e fa un bel bagn-etto alla sua piccol-ina. ‘Now mum gets our nice little basin (basin-DIM1-DIM2) and gives a nice little bath (bath-DIM) to her little baby (little-DIM)’. 2. Emotion/tenderness: Viennese German (to a little child) (2) Nein, rühr das Wass-erl nicht an ‘ No, don’t touch the water-DIM’. 3. Jocular irony: English (3) He’s got a wife and a couple of wif-ie-s ‘girl friends’ 4. Euphemism and irony: German (4) Er hat ein Gläs-chen über den Durst getrunken ‘ He has just drunk one little glass too many (for ‘he is totally drunk’)’. 5. Understatement: Italian (5) Ormai cammini per una strad-ina di campagna e anche lì ti becchi il tuo bel piomb-ino
499
Pragmatics and Morphology: Morphopragmatics 499 ‘You just stroll around a country lane-DIM and you get yourself a nice dose of lead (lead-DIM)’. 6. Jocular understatement: English (6) It’s a major story –well, a sort of major-ette story 7. Sarcasm: Italian (to a thief) (7) Il suo è un mestier-ino che rende! ‘Yours is a pretty lucrative job (job-DIM)!’ 8. Empathy: Italian (8) Vuoi il tuo whisk-ino serale? ‘What about [Do you want] your nice little whisky (whisky-DIM)?’ 9. Ludic character+ irony: English (9) You eat like a pigg-y 10. Polite request (which mitigates the obligation imposed on the addressee and perlocutionary sequels in case the speaker returns only much later): German (10a) and Czech (10b) (10a) Bitte warte noch ein Viertelstünden-chen! ‘Please, wait for still a quarter-of-an-hour-DIM’. (10b) Počkej na mĕ hodin-ku /minut-ku /vteřin-ku ‘Wait for me (an) hour-DIM /minute-DIM /second-DIM (Nekula 2003: 171)’ 11. Hedged request: Spanish (11) ¿Me haces un favorc-illo? ‘Can you do (are you doing) me a little favour?’
12. Customer-attracting offer: Italian (12) Perché non prende due spaghett-ini alla sorrentina, con pomodor-ini e mozzarella? ‘Why don’t you take some spaghetti-DIM Sorrento recipe, with miniature tomatoes (tomato-DIM-es) and mozzarella?’
These are just a few of the numerous uses and nuances of pragmatic meanings/effects that diminutives can convey, especially in languages having a rich alterative paradigm
500
500 Wolfgang U. Dressler and Lavinia Merlini-Barbaresi (such as Dutch, Lithuanian, Italian, Spanish, Portuguese, Slavic languages, Modern Greek, Hungarian, Arabic, some of which are acquired very early, cf. Savickiene and Dressler 2007). Among the various formations seen above, only two diminutives, strad- ina in (5) and pomodor-ini in (12) convey also a denotative semantic meaning of smallness (quantitative diminution), and while favorc-illo in (11) exploits the semantic meaning of smallness pragmatically (secondary pragmatics) to downgrade the load of the request, all the remaining diminutives are engaged in producing exclusive morphopragmatic effects. These modifications belong to the microstructure of the text, but, as we can see above, they extend their effects to the entire speech event. For example, the ironical meiosis in (5), with piombino, mirrors the argumentative point of the speaker, set against a background of global criticism, clearly shared by the addressee. The notion of smallness is metaphorically transposed to the rhetorical plane of meiosis but does not affect the semantics of the word piombo ‘lead’. Thus the pragmatic meanings of the diminutive suffixes do not change the meaning of the nouns to which they are attached, but refer to the speech-act as a whole. The nominal bases to which they attach are only the most appropriate landing sites among the formal word structures of the utterance.
26.3.1.2 Augmentatives and pejoratives These word formation types are much less widespread cross-linguistically (Dahl 2006; Grandi 2011) and in terms of pragmatic effects less efficient than diminutives. Their markedness does not only show in their cross-linguistic implicational relation to diminutives (section 26.3.1.1), but also in the fact that if they occur in the same language, then the paradigm of diminutives is richer than the paradigms of augmentatives and pejoratives (type frequency) and that diminutives are used more often than augmentatives and pejoratives (token frequency). Their pertinence to pragmatics is less direct and exclusive, in the sense that, at most, they confer on their bases a combination of semantic and pragmatic meaning (Merlini Barbaresi 2004: 288). When pragmatics is predominant, both augmentatives and pejoratives may be very close to diminutives in their effects, i.e. they may actually be alternative marks for signalling morphopragmatic meanings. In Italian, for example, augmentatives or pejoratives can convey, in addition to their semantic meanings [big] and [bad], respectively, tenderness and jocular closeness, as in: (13) Il mio fratell-one/ino!
‘ My brother-AUG/DIM!’
or in a famous epithet pronounced by the comedian Roberto Benigni: (14) Ah, Wojtył-accio
‘Oh (pope) Wojtyła-PEJ’.
which combines Tuscan ludic irreverence and affection. Augmentatives may also be used to modify the strength of speech acts, e.g. politely downgrade a request, as in:
501
Pragmatics and Morphology: Morphopragmatics 501 (15) Devo chiederti un piacer-one
‘I must ask you a big favour’
where recognition of the load of the request makes a possible refusal by the addressee less sanctionable socially.
26.3.1.3 Evaluatives: Discussion Many other studies are mainly on language-specific analyses (Stefanescu 1992; Rainer 1993; De Marco 1998; Mutz 2000; Gràcia and Turón 2000; Cantero 2001; Schneider 2003; Nekula 2003; Bardaneh 2010; among many others). Santibáñez Sáenz (1999) represents another attempt, within the framework of cognitive linguistics, to derive the pragmatics of diminutives from small size, but with no parallel account for the similar effects of augmentatives (see section 26.3.1.2). A formal model is elaborated in Fortin’s (2011) thesis. Wierzbicka’s (1999, 2009) and Inchaurralde’s (1997) work on emotions and Grandi’s (2002) more general cross-linguistic work have contributed to keep high the interest in evaluative morphology. In Dressler and Merlini Barbaresi (1994; cf. Bardaneh 2010) we have argued that the most favouring speech situation for diminutives is one which centres on very young children, and then, derived from it, pet-centred and lover-centred speech situations. Another derived speech situation is represented in edifying ecclesiastic texts (at least German and Russian baroque), where the priest, as the author, assumes the role of a spiritual father towards his flock, whom he admonishes and consoles (cf. Tkachov 2011; Resch and Dressler in press). Challenges to the morphopragmatic model come from Jurafsky, whose radial semantic model (1993, 1996) is especially opposed in a major theoretical premise, concerning the priority relationship between semantics and pragmatics. He derives the meaning of diminutives from a basic meaning [child], but gives no clue what the basic meaning of augmentatives or pejoratives should be (the original meaning of augmentatives may be ‘elephant’ in some African languages). Jurafsky’s approach is further open to critique because of the insufficient distinction between synchrony and diachrony and the many lacunae in the neighbourhood chains of his radial model. More important is what follows from the next section.
26.3.1.4 Priority of pragmatics in diminutives Arguments for the priority of pragmatic over semantic meaning of diminutives come from various sources. The cross-linguistic distribution of diminutives does not provide indications for that, because several languages, such as present-day French (cf. Dal 1997; Fradin 2003; Dressler 2010), the North Germanic languages, Finnish, and Estonian, in adult-directed adult language, use diminutives either only or nearly only semantically, when they use them at all. But, importantly, all languages appear to have hypocoristics, which are productively used at least in the discourse types of
502
502 Wolfgang U. Dressler and Lavinia Merlini-Barbaresi child-directed, pet-directed and lover-directed discourse and related discourse types. And the main function of hypocoristics is pragmatic. Given the implication: ‘If a language has diminutives, it also has hypocoristics’ (section 26.3.1.1), we may add that those languages that employ diminutives to form hypocoristics give evidence of the basicness of pragmatic meanings of diminutives. Another argument is the meaning closeness between diminutives and augmentatives, which must be of a pragmatic nature, since their semantic meanings are opposed and contradictory (‘small’ vs ‘big’). Both diminutives and augmentatives can have very similar pragmatic effects in the modification of the illocutionary force of the speech act, e.g. in Italian: (17) Ah! niente è come il mio lett-ino/lett-one!
‘Oh, nothing (is) like my bed-DIM/bed-AUG!’
The speaker has clearly undergone a tiring period away from home and expresses all her satisfaction and pleasure in being back. There is just a tiny semantic difference: when uttering the diminutive lett-ino, the referent, the bed, may be of any size, whereas, when using the augmentative lett-one, the bed cannot be very small (on the greater importance of size for augmentatives, cf. also Ricca 1998 in his review of Dressler and Merlini Barbaresi 1994). Therefore, in the translation of Italian augmentatives into English or German, languages which have diminutives, but no augmentatives, sometimes use diminutives as equivalents of the augmentatives of the source language as in It. grass-one → E. fatt-y, or Sp. angel-ote ‘angel-AUG’ → E. sweet-ie. This pragmatic closeness of diminutives and augmentatives also explains some diachronic developments of alteratives in Romance languages, where the etymologically identical suffix may have an augmentative meaning in one Romance language, but a diminutive meaning in others, e.g. the Spanish augmentative suffix -ote is cognate with the diminutive suffixes Fr. -ot, It. -otto; the augmentative It. -one, Sp. -on is cognate with diminutive Fr. -on. Work on first-language acquisition of diminutives (Savickiene and Dressler 2007; cf. Dressler and Merlini Barbaresi 2001) in several languages has shown that at least some pragmatic meanings emerge before the semantic meaning of smallness, which first children express with adjectives meaning ‘small’. This is true, and not only for hypocoristics for ‘mumm-y’, which children use when they are in a friendly communication with their mother, whereas they tend to use the base ‘mother’ when they are angry with her. Or, expression of empathy is signalled via diminutives by the Lithuanian girl Rūta in Savickiene and Dressler (2007: 35): (18) Kengūr-ytei skauda, skauda kak(t)-ytę ‘Kangaroo- DIM:DAT hurts, hurts forehead- DIM:ACC‚ The kangaroo’s forehead hurts’
503
Pragmatics and Morphology: Morphopragmatics 503 As can be expected from the markedness relation between diminutives and augmentatives, augmentatives are acquired later than diminutives. And, in Italian, the acquisition of augmentatives coincides with the acquisition of the semantic meaning of diminutives, i.e. when the semantic opposition between the two evaluative categories emerges. Finally, the distribution of diminutives varies very much with style and genres, and at least one type of style variation provides further evidence for the pragmatic nature of diminutives, namely the contrastive distribution of diminutives in Italian literary texts with love as the central topic (cf. Dressler and Merlini Barbaresi 1994: 380–394), e.g. in Mozart’s operas La finta giardiniera (‘The false gardeneress’, text by Giuseppe Petrosellini), Così Fan Tutte ossia la Scuola degli Amanti, dramma giocoso (‘So all women do or the school of lovers, playful comedy’, text by Lorenzo da Ponte), and Il dissoluto punito ossia Don Giovanni (‘The punished lecher or Don G.’, text by da Ponte): in the first and second opera love is dealt with in a playful, jocular way, in the last in a dramatic way. As a result, there are many diminutives in the first two operas, few in the third. And in Mozart’s Lucio Silla, due to its heroic style, no diminutives occur, similarly in several Handel operas (e.g. Agrippina). Or, to cite a case which does not have to do with love, in Thomas Mann’s German novel Das Wunderkind (‘The infant prodigy’) many diminutives are used when the boy is described as such (e.g. Bein+chen, Kinder+händ+chen ‘leg-DIM, child-hand-DIM’), but when the author’s perspective is about the prodigy’s astonishing skill and great success, then there are no diminutives, although, of course, the size of the boy’s body parts remains unmodified.
26.3.2 Other areas of morphopragmatics in derivational morphology In all areas of derivational morphology, morphopragmatic phenomena may occur, with one motivated exception. Morphopragmatics is excluded from conversion, because conversion is defined as word class change without formal morphotactic change, i.e. conversion lacks a formal index which could trigger pragmatic inferences. In contrast, a mere morphotactic modification of the base would be sufficient to trigger pragmatic inferences, as shown by Basque diminutive formation, which consists in palatalization of base consonants, as in tonto ‘raindrop’ → diminutive t´ont´o. Other selected phenomena of derivational morphology dealt with in morphopragmatic terms (beyond ethnic formation already cited in section 26.2 and the Quebecois prefix ti < petit, as in ti-Jean ‘Johnn-y’) are:
26.3.2.1 French interactional -o formation Interactional -o suffixation of advanced French introduces a relation of interpersonal closeness and familiarity with the interlocutor(s), sometimes similar to the abovementioned effects of diminutives, as in the case of hypocoristic Cathérine → Cath-o, Paul → Paul-o. Another area of usage is among extreme leftist or otherwise politically marginal social clubs, to which Brown and Levinson’s (1987: 111) definition of an in-group marker is applicable, in their words: ‘By referring to an object with a slang term, S (speaker)
504
504 Wolfgang U. Dressler and Lavinia Merlini-Barbaresi may evoke all the shared associations and attitudes that he and H (hearer) both have toward that object’. An example of typically Parisian usage of such circles is: (19) D’après ce qu’on m’a dit de lui je crois que c’est un intell-o‘After what I’ve been told about him, I believe that he is an intellectual’ where the pragmatic difference between intell-o and intellectuel lies in a pejorative evaluation of the former, which the speaker believes to share with the addressee. A possible negative reaction to (19) is (19a) but not (19b): (19a) Non, ce n’est pas du tout un intello (19b) *Non, ce n’est pas du tout un intellectuel ‘ No, this is not at all an intellectual’ Also coordination is impossible between -o forms and non-o forms, as in: (20) Pas les socialos, les stalos /pas les socialistes, les staliniens /*pas les socialos, les staliniens ‘Not the socialist ones, the Stalinist ones!’ -o forms are clearly excluded from purely transactional contexts, as in: (21) Est-ce que tu peux me vendre l’affiche socialiste /*socialo? ‘Can you sell me the socialist poster?’ This can be contrasted with the interactional minimization of the illocutionary imposition on the addressee (which recalls meiosis of diminutives) in: (22) Est-ce que tu peux coller l’affiche socialo? ‘Can you paste the socialist poster on?’ Meiosis can also be found in friendly irony or reproach, as in: (23) Tu es un véritable stal-o /alcool-o ‘You are a real Stalinist /alcoholic’ A final example illustrates the use of the interactional suffix with adverbs: (24) Vas-y tranquill-o / tranquillement! ‘Go on calmly!’ where the slow rhythm suggested is not the speaker’s imposition but rather a reflexion of the interactional minimization entailed by the choice of the -o formation.
505
Pragmatics and Morphology: Morphopragmatics 505
26.3.2.2 Pragmatic conditions for the use of feminine motional suffixes Pragmatics plays a regulatory role in recent changes in the use of feminine motional suffixes in many languages for referring to women in reports, advertisements, and legal documents (cf. Bussmann and Hellinger 2001; Jobin 2004; Doleschal 2005). In many periods of these languages, masculine, but not feminine, designations of occupations, ethnics, etc. have been used as generics, i.e. in reference to both men and women. Only in emphasizing reference to women, derivational or inflectional suffixes have been used, with an analogical extension to animals. The inverse of this asymmetry is the fact that feminine motional suffixes are usually productive and frequently used, whereas masculine motional suffixes are unproductive and rarely used. For example, in German there are only four common instances of the latter: (25) Hex-erich, Gäns-erich, Ent-erich, Täub-erich vs. Lehrer-in, Student-in, Professor-in, etc. etc. ‘Male witch, goose, duck, dove vs. female teacher, student, professor’. Clearly, there are sociopragmatic reasons for these asymmetries. With the rise of feminism, the use of such feminine motional suffixes has highly increased, but the choice between such suffixed forms and generic masculines depends to a large extent on the pragmatic attitude of the user, of the addressee(s), and the pragmatic conditions of the situation. For example, Savić (1985: 12) reported that in the eighties (female) Serbian students would use the feminine terms, instead of the masculine bases, (pro-)dekan-ica ‘(pro)dean-FEM’ only when joking about the person named, to express positive dealings towards her; or that they would use šef-ica (‘chief/ head-FEM’) if angry or mocking’ (cf. Doleschal 2002). Such attitudes are often also a matter of personal choice: for example, the first female president of the Italian chamber of deputees, the communist Nilde Iotti, energetically refused to be called president-essa instead of the generic presidente.
26.3.3 German intensifying adjective compounding Within compounding, a candidate for a morphopragmatic study is German intensifying adjective compounding with a partially demotivated first element (Sachs 1963; Pittner 1996; Schmitt 1998; Klara 2009). The second element, the head component, is an adjective, the first element (nearly always) a noun with the intensifying meaning ‘very’, which, however, has originally a very diverging lexical meaning, as in blut+jung ‘lit. blood-young’, stock+dumm ‘lit. stick-stupid’, stein+reich ‘lit. stone- rich’, sau+kalt/+heiss ‘sow-cold/hot’. The first element may still carry some connotation derived from its original meaning, but, in any case, these compound adjectives are always (either positively or negatively) evaluative (cf. E. stone dead, dead serious, stone cold, pitch black, etc., whose exaggerated meaning can easily be exploited pragmatically).
506
506 Wolfgang U. Dressler and Lavinia Merlini-Barbaresi
26.3.4 On the pragmatics of extragrammatical word formation We define extragrammatical morphology as a set of heterogeneous operations and forms separated from the module of grammatical morphology which is part of grammatical competence (Dressler 2000; Kilani-Schoch and Dressler forthcoming). We use the term ‘extragrammatical morphology’ and not Zwicky and Pullum’s (1987) term ‘expressive morphology’ because extragrammatical morphology is not always expressive and vice versa. Extragrammatical is then a more precise term. Since these operations, such as reduplicatives discussed just below, are outside grammar and thus outside morphosemantics and, as morphological elements, outside the lexicon in the narrow sense and therefore outside lexical semantics, their meanings are necessarily pragmatic in a broad sense. However, since we are primarily interested in the interface between pragmatics and morphological grammar, we will deal with extragrammatical morphological operations only briefly. French colloquial speech offers two examples of extragrammatical morphological reduplication. Non-hypocoristic reduplication is mainly used in child-directed speech, but not only, e.g. yé-yé, coco < communiste, colloquial baba-cool ‘new hippie’ (Morin 1972; Scullen 2002; Kilani-Schoch and Dressler forthcoming). The meaning of reduplication appears to be parallel to the meaning of diminutives in several languages (cf. Morin 1972; Dressler and Merlini Barbaresi 1994), but it is more restricted as regards the addressee. That is, it can express empathy, endearment, in child-centred situations but also be derogatory in other situations, depending on the relation with the addressee— for example, le chien-chien à sa mémère (‘mummy’s doggy-woggy’ but figuratively ‘yes- man’) may express either of the two aforementioned meanings. French hypocoristic reduplication, applied to proper names, e.g. Vivi (Viviane), Titof (Christophe), Zizou (Zinédine Zidane), is probably more irregular than non- hypocoristic reduplication: it occurs generally in several variants, e.g. Sophie→ Soso, Fifi; Dominique → Dodo, Mimi; Valérie → Vévé, Lil. It is normally accompanied by clipping and substantial sound modifications. It does not change meaning and only expresses connotations such as endearment and interpersonal proximity.
26.3.4.1 English reduplicatives English, unlike other Western European languages, widely and productively exploits extragrammatical reduplication as a word formation mechanism for enriching the lexicon. But reduplicatives can also cover areas of morphopragmatic use that in other languages are normally covered by evaluative suffixes (Merlini Barbaresi 2008). In (26a) Don’t be silly-billy! the same type of jocularity and mild criticism is obtained as in Italian (26b) Non fare la sciocch-ina!
‘Don’t make the silly-DIM’
507
Pragmatics and Morphology: Morphopragmatics 507 (27a) Who’s my boobsy-woobsy? said by a tender mother to her child parallels Italian (27b) Chi è il mio bimb-ol-ino piccol-ino? ‘Who is my baby-INTERFIX-DIM little-DIM’ We distinguish various patterns: (1) ablaut reduplicatives (also called apophonic), exhibiting vowel apophony, that is a systematic alternation of the stressed vowel, such as chit-chat, dilly- dally, flip-flop, knick-knack, see-saw, zigzag, ping pong; (2) rhyming reduplicatives, exhibiting rhyming constituents and apophony of the initial consonant, as in boogie-woogie, bow-wow, fuzzy-wuzzy; (3) rhyming compounds, in which both bases are meaningful, for example, artsy-craftsy, creepie-peepie, fag-hag, willy-nilly, walkie-talkie, nit-wit; (4) copy reduplicatives, in which the second member is the exact copy of the first, as in bye-bye, ack-ack, gale-gale, go- go, ga-ga. All types have morphopragmatic applications. For example, rhyming reduplicatives are especially used for hypocoristics, as in Georgie-Porgie, Humpty-Dumpty, Lizzy-Wizzy, and in Ruskin’s appellatives used in some intimate letters to his mother: grammie-wammie- mammie, Poos-Moos, Poosky-Woosky, Puss-Moss, where he also affectionately refers to an allegorical figure of ‘Logic’ in a painting as Lodgie-Podgie. In a large majority of cases, the mechanism of reduplication also involves the adding of the evaluative/familiarizing/nursery suffix –y/ie, which contributes to the same pragmatic meaning.
26.4 Inflection Among inflection categories, pragmatic meaning may be either primary or secondary or constitute the basis of morphosemantic meanings.
26.4.1 Japanese honorifics A case of primary pragmatic meanings is represented by Japanese honorifics (cf. Dressler and Merlini Barbaresi 1994: 65–67, 72–80, Ide 2003). First of all, variants of personal pronouns differ only or primarily in pragmatics. An extreme case is the variant of the first singular tin which can be used only by the Japanese emperor, but which the present emperor Akihito does not use anymore, favouring instead the most formal common variant watakusi, but not the variant watasi, which is more informal, etc. Much more common are several Japanese humbling devices, which have no other meaning of their own, such as the suffix -masu, as in Harada’s (1976: 553–554) triplet of plain (28a), polite (28b), and superpolite (28c) versions of the sentence ‘here is a book’: (28a) Koko ni hon ga aru (28b) Koko ni hon ga ari-masu (28c) Koko ni hon ga gozai-masu.
508
508 Wolfgang U. Dressler and Lavinia Merlini-Barbaresi The suffix -masu is attached in the polite form (28b), whereas in addition a lexical suppletion of the verb takes place in the superpolite form. The relevant elements of the speech situation which determine the choice among these variants, are the speaker, the addressee, the participants, the occasion (place and time), and the topic.
26.4.2 Pragmatic uses of plurals Cases of secondary pragmatic meanings of inflectional categories are much more frequent, especially (or maybe only) with non-prototypical or inherent inflectional categories, i.e. categories which do not depend prototypically on the syntactic context, such as case, a category of prototypical or contextual inflection (cf. Dressler 1989; Booij 1996). Thus, generally, the speaker has a free choice whether to select a plural or singular, in contrast to choosing the accusative or the genitive. Examples of secondary metaphoric or secondary indexical uses of the plural are the pluralis maiestatis introduced in Late Latin, employed by an emperor or bishop speaking/writing not only in his own name but also in those of fellow-emperors or fellow-ecclesiastics, and which later became a pragmatic device for stressing the importance and venerability of the speaker. But the plural may be also used as a defocusing device (cf. Shibatani 1990), as in the case of pluralis modestiae or author plural. Recently Dressler and Mörth (2012) have found that German plural doublets, where an -s plural competes with another plural marker, can have a virtual or actual connotation of strangeness or foreignness, as in the doublets Rikscha-s referring to rikshaws in Asia vs Riksch-en referring to rikshaws in Germany or Datscha-s referring in West Germany and Austria to country houses in the Soviet Union vs Datsch-en referring to indigenous country houses in former East Germany (sg. die Datscha from Russ. dača). Or, the rival -s plural may refer pejoratively to an outsider group, as in the case of Schmock-s referring to journalists of Jewish origin, used by anti-Semites and especially Nazis (e.g. Joseph Goebbels), vs Schmöck-e (sg. der Schmock ‘shmock, hack writer’ after the literary figure of a bad journalist). This is a secondary development of pragmatic meanings which renders plural doublets distinct in meaning. This is different from semantically distinct plural doublets, such as G. Wört-er ‘(single) words’ (as in Wörter+buch ‘dictionary’, sg. Wort) vs Wort-e ‘(coherent) words’, as in die letzten Worte Goethes ‘Goethe’s last words’.
26.4.3 Excessive Another non-prototypical (inherent) inflectional category is adjective (and adverb, etc.) gradation. A less common subcategory of gradation is the excessive, first theoretically described for Hungarian by Hjelmslev (1972: 83–84, 89). It compares with the Danish and Swedish excessives, formed (as in Norwegian, German, and Dutch) with the prefix aller-‘of all’ attached to the superlative and ascribing to it the meaning
509
Pragmatics and Morphology: Morphopragmatics 509 of absolute intensification. The Hungarian agglutinating formation is szép ‘beautiful’, comparative szép-ebb, superlative leg-szép-ebb, excessive leg-es-leg-szép-ebb, ‘the very/ absolutely most beautiful’ (where és means literally ‘and’). In their morphopragmatic analysis Dressler and Kiefer (1990; cf. Dressler and Merlini Barbaresi 1994: 558–573; Kiefer 2004: 328–332) have shown how this semantic meaning of absolute intensification can be pragmatically exploited for deferentially insisting on the holiness of an absolute monarch or feudal souvereign, as in: (29) Der aller-gnädig-ste Herr = Dan. allernaadigst = Lat. gratios-issimus the exc.gracious
lord exc.gracious
gracious-superl
‘The Most Gracious Lord’
In discourse, as an instance of Jakobson’s principle of equivalence (cf. Holenstein 1976: 109), this very last stage on a paradigmatic scale of intensification can be projected into the very last instance of a succession of instances within a coherent text chunk or within semantically/pragmatically connected parts of a text or related texts in discourse, in order to insist that what is said represents the very last word. This occurs already in the German children’s rhyme when pointing to all the individuals of a row of girls: (30) Du bist schön und du bist schön und du die allerschönste
‘You are beautiful, and you are beautiful, and you the very most beautiful of all’
Other pertinent examples are to be found in Dressler and Merlini Barbaresi (1994).
26.4.4 Pragmatic bases of inflectional patterns Pragmatics is often the background for formal inflectional properties. For example, in many languages, the second person plural is formally the plural of the singular, whereas there is strong suppletion in the first person (cf. Dressler and Merlini Barbaresi 1994: 60–64), e.g. in: (31) Modern Greek 1. egò—pl. emìs, 2. esì—pl. esì-s This is an iconic reflection of the fact that the second-person plural you (Fr. vous) refers very often to a plurality of addressees, i.e. to a plurality of thou (Fr. toi), whereas it is extremely rare that we refers to a strict plurality of I, i.e. it only applies when singing or chanting in a choir. Subdivisions of the animacy scale are based in pragmatics, e.g. in Slavic languages, animate masculines have a different declension from inanimates, and different Slavic
510
510 Wolfgang U. Dressler and Lavinia Merlini-Barbaresi languages decide in different ways whether a human corpse is to be assigned to the animate or inanimate class. In Polish, the animate scale has been subdivided for differences in the declension of masculine nouns (especially in the nominative plural). The lowest category are inanimate nouns and plants. Next are animals, as for example a cardinal bird or a tomcat: sg. kardynał, kot, pl. kardynał-y, kot-y. The next higher category is normal human males, such as sg. student, aktor, pl. studenc-i, aktorz-y. The highest category is represented by male dignitaries and close relatives, such as cardinals, professors, sons, Mr, bosses, etc., and words like kardynał, profesor, syn, pan, boss have the longest plural suffix: nom. pl. kardynał-owie, profesor-owie, syn-owie, pan-owie, boss- owie. But if one speaks derogatorily of humans, then they get an animal plural, e.g. a cardinal gets the plural of the cardinal bird. Inversely, when in a children’s book, tomcats are personified, they get the highest and longest plural kot-owie (cf. Dressler and Mörth 2012).
26.5 Conclusion Among the studies on pragmatic meanings of morphological constructions, work on morphopragmatics typically assumes the possibility of a direct interaction between pragmatics and morphology, instead of an indirect mediation via morphosemantics. Thus we can establish at least four types of interaction between pragmatics and morphology. The first abovementioned variant implies primary pragmatic meanings of evaluative morphology, as with diminutives, augmentatives, and pejoratives, French interactional -o, hypocoristics, and the Japanese honorific suffix -masu; their pragmatic effects are not limited to the lexical base to which these suffixes are attached (i.e. to their ‘landing site’ in Dressler and Merlini Barbaresi 1994), but extend to the whole speech act. In contrast, the secondary pragmatic meanings that exploit the morphosemantics of morphological patterns, such as the Germanic or Hungarian excessives, of noun plurals, of feminine motional suffixes, etc., focus on the respective word, an adjective in case of the excessive degree, or the noun to which they refer if attributive or predicative adjectives. Closely related is the third type of conditioning or regulatory pragmatic factors for the use of morphological patterns, as in the case of feminine motional suffixes. A fourth type of interaction between pragmatics and morphology refers to the pragmatic foundation of morphosemantic categories, as in the relation between I and we and the asymmetries in feminine vs masculine motional suffixes.
511
Chapter 27
Pragmatics a nd t h e Lexic on Laurence R. Horn
La parole est moitié à celuy qui parle, moitié à celuy qui l’escoute. (Michel de Montaigne [1580], ‘De l’expérience’, in Essais de Montaigne, IV, p. 288. Paris: Charpentier, 1854) Maximize meaning, minimize means. (Maxim of graphic designer Abram Games, 1914–1996)
27.1 Lexical Pragmatics: Minding one’s Qs and Rs Within the study of word meaning, an explicit domain of lexical pragmatics, in which cooperation and rationality-driven Gricean pragmatics drive aspects of word meaning, word choice, and meaning change, was not recognized until the publication in 1978 of McCawley’s ‘Conversational implicature and the lexicon’. The term itself traces back to Blutner (1998): Lexical pragmatics is a research field that tries to give a systematic and explanatory account of pragmatic phenomena that are connective with the semantic underspecification of lexical items. Cases in point are the pragmatics of adjectives, systematic polysemy, the distribution of lexical and productive causatives, blocking phenomena, [and] the interpretation of compounds. (Blutner 1998: 115; cf. also Blutner 2004)
While these phenomena have been fruitfully explored through the lens of lexical pragmatics (as I have undertaken to do in earlier works: cf. Horn 1972, 1978, 1984, 1989,
512
512 Laurence R. Horn 1993, 2006b, 2007a, along with related work by McCawley 1978, Atlas and Levinson 1981, and Levinson 2000), the stream of work surveyed by Blutner has overflowed these banks. Thus, Wilson and Carston (2007) characterize lexical pragmatics as a ‘relatively new field’ that applies the methodology and results of Relevance Theory at the word and phrase level to explore the inferential basis for the construction of ad hoc concepts and their role in broadening, narrowing, and metaphorical transfer. Closer to the spirit of the current presentation is Huang (2009), which considers the role of neo-Gricean pragmatics in constraining word formation and meaning in a variety of constructions. Allan’s (2011) handbook chapter offers a formalization of rules dealing with pragmatic subclasses of encyclopedic knowledge and non-monotonic inferences arising therefrom and provides a more formally grounded introduction to a number of the relevant issues. A brief note on part of what I will not be covering: In concentrating on the role of conversational implicature in the lexicon, I may seem to be slighting Grice’s other major component of what is meant without being said. In fact, conventional implicature—the relation responsible for, inter alia, the difference between p but q and p and q, between Tu as raison ‘You (sg. fam.) are right’ and Vous avez tort ‘You (sg. formal) are right’, or between I need a new shotgun and I need me a new shotgun— involves semantic (encoded) aspects of meaning, while not affecting the truth conditions of the sentences in which they occur, and hence is not strictly germane to a discussion of lexical pragmatics. See Horn (2007b, 2008) for elaboration and additional candidates for this status and Horn (2002b) for the related category of ‘assertorically inert’ entailments. My focus in the present study will be on the role of conversational implicature in the formation and distribution of lexical items and in meaning change, as well as the functioning of such contextually dependent notions as Roschian prototype effects and Aristotelian privation. Given the breadth limits inherent in a handbook entry, it is difficult to cover more than one theoretical model while still covering a substantial empirical range. The particular notion of implicature I will assume is based on the Manichaean model (Horn 2007a) invoking countervailing Q and R principles. The inspiration for this dualist approach is as much neo-Paulist as it is neo-Gricean. In his Prinzipien der Sprachgeschichte—whose English translation appeared in 1889, exactly a century before the appearance of Grice’s posthumous collection—Paul surveys a range of phenomena whose form and distribution reflect the interplay of two functional principles, the tendency to reduce expression and the contextually determined communicative requirements on sufficiency of information: The more economical or more abundant use of linguistic means of expressing a thought is determined by the need … Everywhere we find modes of expression forced into existence which contain only just so much as is requisite to their being understood. The amount of linguistic material employed varies in each case with the situation, with the previous conversation, with the relative approximation of the speakers to a common state of mind. (Paul 1889: 351)
513
Pragmatics and the Lexicon 513 The descendants of Paul’s dualism include the two opposed communicative economies of Zipf and Martinet and the interdefined halves of Grice’s Maxim of Quantity. While most linguists associate G. K. Zipf ’s name with the principle of least effort, the Zipfian framework (1935, 1949) in fact distinguishes the speaker’s economy, which would tend toward ‘a vocabulary of one word which will refer to all the m distinct meanings’, from the countervailing auditor’s economy, tending toward ‘a vocabulary of m different words with one distinct meaning for each word’. The Speaker’s Economy places an upper bound on the form of the message, while the Hearer’s Economy places a lower bound on its informational content. By Zipf ’s law of abbreviation, the relative frequency of a word is inversely correlated with its length; the more frequent a word’s tokens, the shorter its form. Frequency, and its effect on utterance length and phonological reduction, is relativized to the speaker’s assumptions about the hearer and their shared common ground: High frequency is the cause of small magnitude … A longer word may be truncated if it enjoys a high relative frequency [either] throughout the entire speech community [or] if its use is frequent within any special group. (Zipf 1935: 31–32)
Zipf ’s two mutually constraining mirror-image forces are periodically invoked (or rediscovered) in the diachronic and psycholinguistic literature: The linguist must keep in mind two ever-present and antinomic factors: first, the requirements of communication, the need for the speaker to convey his message, and second, the principle of least effort, which makes him restrict his output of energy, both mental and physical, to the minimum compatible with achieving his ends. (Martinet 1962: 139) The speaker always tries to optimally minimize the surface complexity of his utterances while maximizing the amount of information he effectively communicates to the listener. (Carroll and Tanenhaus 1975: 51)
This minimax of effort or complexity on the one hand and informative content or distinctness on the other is directly reflected in the tension between articulatory economy and perceptual distinctness, as detailed in work by phoneticians and phonologists from Lindblom and MacNeilage to Hayes and Flemming, and in particular in the optimality-theoretic dialectic of faithfulness and markedness; cf. Horn (2006b) for references and discussion. It has been clear for some time that the Zipfian parameter of familiarity is also a major player in the phonological field; segmental and prosodic reduction and simplification mark (or unmark) familiar or frequent items, while unfamiliar or unpredictable words are assigned (or retain) extra stress or pitch (cf. e.g. Fidelholtz 1975; and the work of Bybee 2007 and her associates). In Bybee’s refinement of the Zipfian model (2007: 12), ‘high-frequency words undergo reductive changes at a faster rate that low- frequency words … [T]he major mechanism is gradual phonetic reduction brought about by the reduction and overlapping of articulatory gestures.’ Following Paul and
514
514 Laurence R. Horn Zipf, I have referred to this correlation (Horn 1993) under the rubric ‘Familiarity breeds cntnt’: the more the assumed familiarity, the less the expression of content.1 Reduction of unfamiliar, unrecoverable material can impair the message. The trade- off between brevity and clarity was recognized by classical rhetoricians, as captured in Horace’s dictum Brevis esse laboro; obscurus fio (‘I strive to be brief; I become obscure’—Ars Poetica, line 25). The resolution typically takes the form of the Golden Mean (or Goldilocks?) principle: ‘If it is prolix, it will not be clear, nor if it is too brief. It is plain that the middle way is appropriate …, saying just enough to make the facts plain’ (Aristotle, Rhetoric, 3.12–3.16). The speaker aims to maximize ease and brevity, correlates of least effort; well before Zipf, the ‘principle of economy’, also known as ‘the principle of least effort’, or ‘laziness’, figures prominently in reflections by Sweet (1874), Sayce (1880), Paul (1889), and Passy (1890) on sound change, synonymy avoidance, word meaning, and meaning change. On the other hand, the hearer requires sufficiency of content and discriminability of form. Speaker and hearer are aware of their own and each other’s goals, a mutual awareness that generates a variety of effects based on what was said and what was not. It is this interaction that makes it possible to fold the maxims of conversation (Grice 1989a: 26–37) into two general mirror-image principles that I have dubbed Q and R in deference to Grice’s (first) Quantity submaxim and Relation maxim respectively. On this view, implicatures may be generated by either the Q Principle (essentially ‘Say enough’, a generalization of Grice’s first submaxim of Quantity) or the R Principle (‘Don’t say too much’, subsuming the second Quantity submaxim, Relation, and Brevity). The hearer-oriented Q Principle is a lower-bounding guarantee of the sufficiency of informative content; collecting the first Quantity submaxim along with the first two ‘clarity’ submaxims of Manner, it is systematically exploited to generate upper-bounding (typically scalar) implicata. The R Principle, by contrast, is an upper- bounding correlate of Zipf ’s principle of least effort dictating minimization of form; it collects the Relation maxim, the second Quantity submaxim, and the last two submaxims of Manner, and is exploited to induce strengthening implicata. The application of Q-based upper-bounding scalar implicature allows for a systematic and economical treatment of sets of logical operators and ordinary non- proposition-embedding predicates that can be positioned on a scale as defined by unilateral entailment: (1) Q-scales: logical operators < some, many, most, all > < sometimes, often, usually, always> < or, and > < possible, likely, certain > < a, the >
Q-scales: ‘ordinary’ values < warm, hot > < lukewarm, cool, cold > < OK, good, excellent > < like, love, adore > < finger, thumb >
1 Lest this seem obscure, I am alluding here to the second-order maxim ‘familiarity breeds content’, itself a variation on ‘familiarity breeds contempt’. But in this case, it’s not the adjectival conTENT that familiarity brings about but reduced CONtent, hence cntnt.
515
Pragmatics and the Lexicon 515 In each case, given an informationally weaker value W and a stronger value S plotted on the same scale < W, S > as determined canonically by unilateral entailment,2 in asserting [ … W … ] I Q-implicate, ceteris paribus, that I was not in an epistemic position to have asserted [ … S … ] salva veritate, i.e. that I don’t know that S, and hence, all things being equal, that I know that ¬[ … S … ] holds. Scalar values are lower-bounded by their literal meaning (‘what is said’) and upper- bounded by quantity-based implicature. Thus the ‘one-sided’ meanings delivered by the linguistic semantics may be pragmatically enriched to yield the ‘two-sided’ understandings typically communicated: (2)
what is said a. You ate some of the cake. b. It’s possible she’ll win. c. He’s a knave or a fool.
‘some if not all’ ‘at least possible’ ‘… and perhaps both’
d. It’s warm.
‘… at least warm’
⇒what is communicated ‘some but not all’ ‘possible but not certain’ ‘knave or fool but not both’ ‘warm but not hot’
This accounts for the role of context in the cancellation and reinforcement of the upper bound of scalar predications and allows for generalizations across operator types (quantifiers, binary connectives, deontic and epistemic modals, simple non-embedding predicates), without invoking any lexical ambiguity for the relevant operators (e.g. inclusive vs exclusive disjunction), a move that would violate the principle that Grice ([1967]1989a: 46) dubs Modified Occam’s Razor: ‘Do not multiply senses beyond necessity’.3 This model also allows us to provide a satisfactory denouement to the story of *O, the non-occurrence of values corresponding to the O or south-east vertex of the Square of Opposition (Horn 1972, 1989, 2012c): (3) A: all F are G
I: some F are G 2
E: no F are G
O: some F aren’t G; not all F are G
As stressed by Hirschberg, entailment-defined scales represent a proper subset of semantic relations motivating quantity implicatures. Thus in referring to you as my friend, I will in some contexts Q-implicate that you are not my spouse or lover, despite the absence of a unilateral entailment, since lovers and spouses are often friends but need not be (Hirschberg 1991: 98; cf. also Horn 1989: 546 on the robustness of the scale in the face of the musical observation ‘I don’t like you but I love you’). (See also discussion below on rank orders.) In earlier work on scales and scalar implicature, beginning with Horn (1972), scales are represented in the format rather than format; the notational majority has swung to the latter mode, which I follow here. 3 For more on the so-called ambiguity of inclusive vs exclusive disjunction in English and other languages—and on the pragmatic reconstruction of that ‘ambiguity’—see Horn (1989: §4.2) and Jennings (1994: chapter 3).
516
516 Laurence R. Horn Thus, alongside the possible quantificational determiners all, some, no, we never find (in any language) an O determiner *nall; corresponding to the quantificational adverbs always, sometimes, never, we have no *nalways (= ‘not always’, ‘sometimes not’). We may have equivalents for both (of them), one (of them), and neither (of them), but never for *noth (of them) (= ‘not both’, ‘at least one … not’ = the Sheffer stroke); we find connectives corresponding to and, or, and sometimes nor (= ‘and not’), but never to *nand (= ‘or not’, ‘not … and’). This asymmetry extends to modal operators (e.g. can’t = ¬CAN; mustn’t = MUST¬) and other values that can be plotted on the Square, as well as to O→E drift, the tendency for sequences that might be expected to yield O meanings get E interpretations, as when Dutch nimmer (lit., ¬IMMER, i.e. ‘nalways’) can only be understood as ‘never’, or when French Il ne faut pas que tu meures (lit. = ¬[You must die]) = ‘You must [not die]’. On the neo-Gricean approach to this asymmetry, the systematic, cross-linguistically attested restriction on the lexicalization or direct expression of values mapping onto the O vertex is attributable to the mutual Q-based implicature relation obtaining between the two subcontraries I and O and to the marked status of negation dictating the preference for I over O forms (cf. Horn 2012c for elaboration).
27.2 R in the Lexicon: Narrowing and Strengthening Unlike the upper-bounding associated with Q-based implicature, R-based effects involve articulatory reduction and pragmatic strengthening: the speaker makes her contribution relatively brief or uninformative and counts on the hearer to recognize the operation of least effort and fill in the missing material. Where Q enjoins the speaker to ‘Provide necessary specification’, R dictates ‘Avoid unnecessary specification’. One obvious place to look for R-based effects in the lexicon is in processes like acronymy and initialism (radar, sonar, AIDS; US, TV, FAQ), blending (smog, motel, netiquette), and clipping (phone, bus, math), in which relatively long or complex descriptors for frequently invoked referents undergo reduction to simpler forms that tend to become lexicalized. These processes, direct reflexes of Zipf ’s Law of Abbreviation, do not operate unchecked: ‘The demands of the speech functions must set a limit to the economic tendency’ (Stern 1931: 257)—i.e. Q constrains R. Zipf (1935: 31–32) recognizes both local and global factors in reduction: ‘A longer word may be truncated if it enjoys a high relative frequency [either] throughout the entire speech community [or] if its use is frequent within any special group.’ Thus, he notes, moving pictures are abbreviated throughout the English-speaking world into movies (more on which below), while gas may truncate natural gas, gasoline, nitrous oxide, or flatulence in contexts of energy options, petrol stations, dentists, and beans, respectively. In our own era, PC may be a personal computer, politically correct, Providence College, or personal communication, CD a compact disc, certificate of deposit, or (in discussing the Prague Linguistic Circle) communicative dynamism.
517
Pragmatics and the Lexicon 517 And OSU will refer to whichever of the three state universities of Ohio, Oklahoma, or Oregon that happens to be most salient in a particular discourse context. R-based effects are a robust factor in semantic change as well, but here it is largely information rather than articulatory complexity that is economized. In the well-known process of sense narrowing, we can distinguish Q-based narrowing, which is linguistically motivated and results from the hearer-oriented tendency to avoid ambiguity, from R-based narrowing, the socially motivated restriction of a set-denoting term to its culturally salient subset or member. In Q-based narrowing, where the existence of a specific hyponym of a general term licenses the online (or occasionally standardized) use of the general term for the complement of that hyponym’s extension: (4) animal cat finger lion rectangle
(including or excluding humans, birds, fish) (including or excluding kittens) (including or excluding thumbs) (including or excluding lionesses) (including or excluding squares)
In R-based narrowing, a general term denoting a set narrows to pick out a culturally or socially salient subset, allowing the speaker to avoid overtly specifying the subdomain via the assumption that the hearer will fill in the intended meaning. Unlike the cases in (4), this process does not depend on the hearer’s inference from what was not said, but counts instead on his recognition of what need not be said. The result may be a complete referential shift as in (5) or the development of autohyponymy—in which the original broader sense persists alongside the narrowed meaning—as in (6): (5) corn deer hound liquor poison wife
(‘wheat’ [England], ‘oats’ [Scotland], ‘maize’ [US]) (originally ‘(wild) animal’, as in Ger. Tier) (originally ‘dog’, as in Ger. Hund) (originally ‘liquid substance’) (originally ‘potion, drink’) (originally ‘woman’, as in (4) above)
(6) drink (in particular [+alcoholic]) friend (in the sense of ‘friend-plus’ or ‘umfriend’4) man (orig. ‘human’, now chiefly ‘male adult human’) number (in particular ‘integer’) smell (as intransitive, = ‘stink’) temperature (in particular, one in the ‘fever’ range) Ger. Frau, Fr. femme, Span. mujer (‘woman’, ‘wife’) 4 A relatively new lexical item that has arisen from a social dilemma of having to introduce one’s girlfriend/boyfriend/lover to one’s older relative: ‘Grandma, this is my … um … FRIEND, Chris.’
518
518 Laurence R. Horn The locus classicus of socially motivated R-based narrowing is the euphemism, ranging from sex (sleep with, make love, lover) and excretion (toilet, go to the bathroom) to death and illness (disease, accident, undertaker). The pragmatically motivated nature of such shifts explains the tendency of critics to find such shifts illogical, as noted by the coiner of the term semantics: Restriction of meaning has at all times been a cause of astonishment to etymologists. We know the observations of Quintilian on the subject of homo: ‘Are we to believe that homo comes from humus, because man is born of the earth, as if all animals had not the same origin?’ Yet it is most certain that homines did signify ‘the inhabitants of the earth.’ (Bréal 1897: 114)
The same R-based narrowing is similarly illustrated by our own earthling (rarely applied to non-human animals that share our planet) or in the history of Adam, not a proper name in Hebrew but a generic term for ‘human’, from ‘adam = lit. ‘from the ground/earth’, ‘adamah. In other semantic fields, Quintilian’s misplaced scepticism is echoed by those resistant to the truncation of compounds to what originated as a qualifier of an understood head noun (as in general < general officer, private < private soldier). One such newfangled clipping was eloquently (if pointlessly) ridiculed by McQuade (1915): The coinage of ‘movie’ was most assuredly childish. It stands for ‘moving picture.’ The coined word, please note, is not taken from the name of the thing itself, but from the qualifying word ‘moving.’ It is not at all unreasonable, therefore, to call everything which is not at rest a ‘movie,’ including the sun, moon and stars, the earth, an automobile, an airplane and the city garbage cart. Even man himself when in motion is a ‘movie,’ and so is a fly, and so is that other pestiferous insect with a name nearly alike. Is this childish word ‘movie,’ on the ground of etymology, a correct word to represent ‘moving picture’ in our dictionaries? Is it a correct word from the common sense point of view? Is it a correct word for grown-ups to use, unless they are still fit for the nursery in mind and accomplishments? By all means let the children use ‘movie’ to their little hearts’ content; but in the name of all that is logical and customary in the making and adoption of the words of a language, let us, grown-ups, put it tenderly away.
The Manichaean model of pragmatics is criticized by Carston (2002: chapter 3, 2005: 314–315) on the grounds that both Q-based and R-based implicature involve ‘a strengthening of communicated content’; I have defended the model (Horn 2007a, 2009) by distinguishing informative and rhetorical strength. But note also that the dualistic model makes it possible to tease apart the two conflicting motivations for lexical narrowing, and to see why we can predict that it is only the latter species, where the hearer typically lacks access to the information required to reconstruct the shift,
519
Pragmatics and the Lexicon 519 that regularly leads to conventionalized semantic shifts (as opposed to online meaning restriction). Related to R-based lexical narrowing, and again motivated by social considerations— in particular, those relating to respect for negative face—is the strengthening of negative expression. As Bosanquet (1888: 306) puts it, ‘The essence of formal negation is to invest the contrary with the character of the contradictory.’ Across a wide range of languages we find a tendency for the speaker to weaken the force of expressed negative judgements, counting on the hearer to fill in an intended stronger negation. In English, the resultant contrary negatives in contradictory clothing include affixal negation, ‘neg-raising’, and simple litotes, as illustrated in (7a–c) respectively (cf. Horn 1989: chapter 5). (7) R-based negative strengthening a. contrary readings for affixal negation He is unhappy I disliked the movie
(stronger than ¬ [He is happy]) (stronger than ¬ [I liked the movie])
b. ‘neg-raising’ effects across clause boundaries I don’t believe it’ll snow He doesn’t want you to go
(= I believe it won’t) (= He wants you not to go)
c. litotes (understatement) in simple denials She’s not happy with it I don’t like ouzo
(stronger than ¬ [She’s happy with it]) (stronger than ¬ [I like ouzo])
In each case a general formally contradictory negation is strengthened to a specific, contrary understanding; where the constructions differ is in the degree of conventionalization of this strengthening. I say I don’t like ouzo precisely to avoid directly acknowledging my antipathy; at the same time, I count on your willingness to fill in the intended R-strengthened (contrary) interpretation rather than simply settling for the contradictory negation literally expressed. In an embedding environment, this same practice is responsible for the ‘neg-raising’ effect seen in (7b), where a negation outside the scope of certain predicates of opinion, desire, or likelihood is taken to have lower- clause scope. Here again, the contrary meaning (‘x believes that not-p’) is sufficient but not logically necessary to establish the truth of the contradictory (‘x does not believe that p’), yet it is treated as if it were necessary—not surprisingly, both because it represents the inductively salient case that makes the contradictory true and because there may be social and cultural constraints against the direct expression of the stronger contrary. (See Horn 2000a for a general account of the R-strengthening of sufficient to necessary-and-sufficient conditions, focused on so-called conditional perfection, but with applications to euphemism and related lexical processes.)
520
520 Laurence R. Horn
27.3 Q and R in (Inter)action: The Division of Pragmatic Labour In his study of the penetration of Gricean reasoning into word formation and word use, McCawley (1978) observes that in forgoing an unmarked communication like John went from the kitchen to the living room in favour of, say, John ceased to be in the kitchen and came to be in the living room, a speaker would be understood as conversationally implicating teleportation, magic, or some other unusual circumstances, although such implicatures are by definition cancellable: Reference to one of the more marked situations requires comment, and if the speaker does not provide the addressee with warning that the marked situation is intended, the addressee is justified in assuming that one of the less marked situations is intended … (McCawley 1978: 255)
More generally, periphrasis indicates that something is being conveyed beyond the choice of words used, a fact that follows directly from Gricean reasoning: [T]he lack of interchangeability between the lexical item and its periphrastic equivalent are due not to idiosyncratic restrictions that must be incorporated into the relevant dictionary entries, but rather are consequences of general principles of cooperative behavior. (McCawley 1978: 257–258)
The principle that I dub the Division of Pragmatic Labour (Horn 1984, 1993) is designed to extend this insight by generalizing it to a range of examples beyond those involving simple periphrasis.5 While the inevitable clash between Q and R in simple cases results in the principled indeterminacy of what is implicated that is typically resolvable through a variety of contextual factors, the interaction of the two functional drives is dynamically resolved into an equilibrium: given a pair of a priori coextensive linguistic expressions, the relatively unmarked (briefer and/or more lexicalized) expression will tend to become R-associated (pragmatically or conventionally) with a particular unmarked, stereotypical meaning, use, or situation, while the use of its less lexicalized counterpart (typically more complex or prolix) will tend to be Q-restricted to those situations outside the stereotype, for which the unmarked expression could not have been used appropriately. Thus consider: (8)
a. She got the car to stop. b. She stopped the car.
5 My development of the division of pragmatic labour was strongly influenced by McCawley’s invocation of a least-effort hypothesis in his work on implicature and the lexicon, and problems arising therefrom; cf. Horn (1978).
521
Pragmatics and the Lexicon 521 (9)
a. The slacks are pale red. b. The slacks are pink.
(10)
a. She wants her to win. b. She wants PRO to win.
(11)
a. I am going to marry you. b. I will marry you.
(12)
a. My brother went to {the church/the jail/the school}. b. My brother went to {church/jail/school}.
(13)
a. It’s not impossible that the Mets will win. b. It’s possible that that the Mets will win.
(14)
a. That’s my father’s wife. b. That’s my mother.
S is aware (and H is aware that S is aware) that H knows that S will be attempting to reduce her effort. Based on this knowledge, H recognizes that S’s choice of a relatively complex or marked utterance in each (a) example implicates that she was not in a position to have used the simpler or less marked alternative in (b) (see Horn 1993: 40–43 for elaboration). The crucial assumption is that if S has expended what appears to be unnecessary effort in her utterance, she must have had a sufficient reason to do so, although just what reason may be indeterminate (see Horn 1991 for detailed elaboration with respect to the double negation case in (13)). When the speaker appears to have gone out of her way to provide additional material (as in the form of modification), the hearer will assume—given the R Principle—that (ceteris paribus) the extra material is relevant. This is not a new observation. Ducrot, for example, recognizes that to posit a restriction will often suggest by ‘a kind of law of economy’ that the general predication holds not just when, but only when, this restriction is satisfied (Ducrot 1969: 22, emphasis and translation mine): Le locuteur observe, dans le choix de son énoncé, une espèce de loi d’économie. Si on dit d’une personne qu’elle aime les romans policiers, l’auditeur est tenté de conclure, pour s’expliquer la précision apportée par le mot ‘policiers’, qu’elle aime peu, ou moins, les autres romans. Car, si elle aimait tous les romans, à quoi bon ajouter cette determination? The speaker observes, in the choice of her utterance, a kind of law of economy. If it is said of someone that he likes detective novels, the hearer is tempted to conclude, in order to explain the specificity provided by the word ‘detective’, that he is less fond of other novels. For, if he liked all novels, what’s the point of adding this qualification?
Formalizations of the division of pragmatic labour have been undertaken within bidirectional Optimality Theory and game-theoretic pragmatics, although the results of
522
522 Laurence R. Horn such work are a matter of debate; cf. e.g. Blutner (2004, this volume), van Rooij (2009) and references cited therein. Within his framework for deriving implicatures that posits an interaction of Q, I (≈ R), and M (for Manner) heuristics, Levinson’s reconstruction of the division of pragmatic labour (2000: §2.4) involves not Q but the M heuristic, given that while some differs from all in informative content, kill differs from cause to die in complexity of production or processing. As Levinson acknowledges, however, the Q and M patterns are closely related, since each is negatively defined and linguistically motivated: S tacitly knows that H will infer from S’s failure to use a more informative and/or briefer form that S was not in a position to have used that form. Unlike the essentially negative Q implicature, whose calculation is based on what has not been but could have been said, R- or I-based implicature is often motivated socially rather than linguistically, typically invoking a culturally salient stereotype (Atlas and Levinson 1981, Levinson 2000; cf. also Huang 2007 and Chapman 2011 for useful overviews). For relevance theorists (e.g. Sperber and Wilson 1986; Carston 2002), only one all- encompassing basic pragmatic principle need be invoked—that of Relevance (defined in non-Gricean terms). Arguably, however, the RT programme is itself ultimately Manichaean, given that Relevance itself is calibrated as a minimax of effort and effect. As Carston (1995: 231) puts it, ‘Human cognitive activity is driven by the goal of maximizing relevance: that is … to derive as great a range of contextual effects as possible for the least expenditure of effort.’ Carston (esp. 2002: chapter 3) has criticized the dualist position on the grounds that R-based implicature and Q-based implicature are essentially alike, in that ‘there is a strengthening of communicated content from “at least some” to “just some”’ (Carston 2005: 314–315) that is entirely parallel to the negative strengthening seen in examples like those in (9). But does the upper-bounding effect of Q-based, and in particular scalar, implicature amount to strengthening? In fact, while R-based implicature increases both the informative content and rhetorical strength (positive or negative) of the assertion, what is communicated as a result of Q-based upper-bounding, while more specific and hence informatively stronger than the unbounded utterance, is not rhetorically stronger than the utterance sans implicature: Some but not all the students passed, while unilaterally entailing some F are G, yields a more specific but not a stronger positive assertion. Another argument for separating these two notions of strength comes from the distribution of rank orders (Horn 2000b) such as those in (15) involving army and (US) academic ranks respectively: (15) private < corporal < sergeant < … < lieutenant < colonel < general lecturer < assistant professor < associate professor < full professor In a scale , … Y … unilaterally entails … X … : if it’s hot, it’s warm; but in a rank order X < Y, … Y … unilaterally entails … ¬X … : If he’s a general, he’s not a
523
Pragmatics and the Lexicon 523 lieutenant and if she’s a full professor, it’s false that she’s an assistant professor—although it’s true that she’s at least an assistant professor. Rank-ordered items essentially build in the upper bound: Doolittle is a full professor and Dr Doolittle is an associate professor are equally informative, in that neither entails the other. Yet the former is assertorically stronger in asserting that the higher rank holds. Once again, we see that rhetorical strength is distinct from informative strength.
27.4 Nonce Word Formation as Lexical Pragmatics: Clones and Un-words Within Relevance Theory (Sperber and Wilson 1986; Carston 2002), effort is conceived of as a property of the hearer’s processing tasks rather than the speaker’s production, but surely both are relevant in the shaping of linguistic form. In particular, Carston (2005: 314) challenges the status of arguments from Brevity (and the R principle subsuming it), whose role as a component of speaker’s effort she rejects. One area worth exploring in this light is that of repetition in discourse and reduplicative constructions in the lexicon and syntax. There is a widespread and widely recognized iconically motivated tendency for repetition, particularly of verbal or clausal units, to be used to represent increased salience along one or more physical dimensions, as determined by aspectual considerations. If I say ‘We walked and walked and walked’ or ‘They kissed and they kissed and they kissed’ or ‘Tomorrow, and tomorrow, and tomorrow, creeps in this petty pace’, the greater locutionary effort required by your iteration implies a commensurately greater distance traversed, intensity achieved, or desperation felt. Our focus here, however, is on a lexical correlate of repetition instantiated in English and many other languages in the lexical clone construction, also known as the double (Dray 1987), contrastive focus reduplication (Ghomeshi et al. 2004), or (somewhat misleadingly) identical constituent compounding (Hohenhaus 2004). Cross-linguistic research has revealed a wide range of meaning effects associated with clones and other uses of partial or complete lexical reduplication in both spoken and signed languages (Hurch 2005), but we focus here on the effects in English. In many contexts, the reduplicant singles out an element or subset of the extension of the noun corresponding to a true, real, default, or (in the sense of Rosch 1978) prototype category member: a dog dog may be an actual canine (excluding hot dogs or unattractive people) or it may be a golden retriever or collie (excluding Chihuahuas and toy poodles), a salad salad (in Western culture) is based on lettuce, not tuna, potatoes, or squid, a DOCTOR doctor is an MD, not a PhD in linguistics, and so on. On the other hand, my request for a drink drink is likely to ask for the ‘real’ [+ alcoholic] thing: not a prototype potable but the culturally salient one, with the clone functioning as a quasi-euphemism. This is what we can call the value-added or intensifying
524
524 Laurence R. Horn use of the clone construction. Such an understanding is especially prevalent in the case of cloned adjectives: ‘It’s HOT hot’ will tend to convey ‘very hot’, ‘hot-plus’, with the appropriate scalar intensification in whichever sense (temperature or spiciness) the context favours. Why should the clone exist, given its apparent communicative redundancy? As first observed in Dray (1987), the lexical clone offers a natural laboratory for observing the interplay of the Q and R Principles. A clone XX is more effortful (to produce and presumably to process) than the simple nominal X. On the other hand, the clone XX is less informative, and arguably less effortful (for speaker and/or hearer) than a phrase or compound YX, where the modifier Y is distinct from the modified X. XX must be both necessary (as against X), given the R Principle, and sufficient (as against YX), given the Q Principle, to narrow the domain appropriately.6 (As Dray observes, the speaker might find it harder to characterize the narrowed domain by spelling out the default overtly than by invoking it via the clone.) Dray also stresses the essential role of context in establishing and inferring the appropriate interpretation, as illustrated by the elegant minimal pair in (27):7 (16) a. Oh, we’re just `LIVing together living together. b. Oh, we’re not vliving together living together.
[` = simple fall contour] [v = fall-rise]
With the prototype clone in (16a), the speakers present themselves as just room- mates, not romantically or sexually involved, while the negated clone in (16b) must be interpreted in the opposite (value-added) sense, with the result that both affirmative and negative sentences assert a platonic status. At least three distinct functions can be isolated for the clone construction in English: (17) (i) singling out prototype category members (especially with nouns) (ii) assigning a value-added or intensifying use (especially with adjectives) (iii) picking out a literal, as opposed to figurative, use Notice that the same functions and contextual dependence can be realized in non- clone constructions, as recognized by Austin (1962b: 70):
6
This analysis, developed by Dray (1987) and Horn (2006b), remains somewhat oversimplified; see Huang (2009: 133–141) for a valuable inventory of additional empirical and theoretical considerations and an alternative neo-Gricean/neo-Levinsonian derivation of the operation of lexical cloning. 7 As this example shows, not just simple lexical items can be cloned, but verb phrases like living together; cf. the semi-conventionalized ‘Do you LIKE him like him?’ (= ‘Do you like him romantically?’) or productive cases like this one, from host Jon Stewart’s television interview on The Daily Show (28 January 2014) of Louis C. K., whose movie Tomorrow Night had just been released, years after it was made; note the completive feature added by the clone: Louis C. K.: ‘It was delayed.’ Jon Stewart: ‘When did you FINish it finish it? ‘Cause I remember you were editing it, and this was ‘96.’
525
Pragmatics and the Lexicon 525 [A]definite sense attaches to the assertion that something is real, a real such-and- such, only in the light of a specific way in which it might be, or might have been, not real. ‘A real duck’ differs from the simple ‘a duck’ only in that it is used to exclude various ways of being not a real duck—but a dummy, a toy, a picture, a decoy, &c.; and moreover I don’t know just how to take the assertion that it’s a real duck unless I know just what on that particular occasion, the speaker has it in mind to exclude … [T]he function of ‘real’ is not to contribute positively to the characterization of anything, but to exclude possible ways of being not real.
We might add that the real duck—perhaps imprinting on the DUCK duck—may also be a healthy (rather than diseased or wounded) specimen, or perhaps a normal (rather than an albino or dwarf) duck.8 As noted in the studies cited above, the meaning effects noted here are found quite generally, although not universally or exclusively, with clones, but some or all of the same pragmatic effects may also be associated with prefixes or dedicated modifiers; as Huang (2009: 136) points out, we are dealing here with the semantic narrowing effects of what Lasersohn (1999) calls pragmatic slack regulators. Poser (1991) analyses Japanese ma as a Roschian function: maX ‘restricts the denotation of the base form to [X’s] canonical point’, and with natural kind nouns to the core or prototype category members: (18)
siro ‘white’ kuro ‘black’ kita ‘north’ natu ‘summer’ hadaka ‘naked’ iruka ‘dolphin’
massiro ‘snow white’ makkuro ‘pitch black’ makita ‘due north’ manatu ‘midsummer’ mappadaka ‘stark naked’ mairuka ‘the common dolphin’
In each case, Poser argues, the relevant salient exemplar is picked out, but salience is locally determined. Thus ma-represents a lexicalized counterpart of the deictic or ‘online’ narrowing associated with the lexical clone construction in English. (See Hohenhaus 2004, Ghomeshi et al. 2004, and Huang 2009 for useful cross-linguistic surveys of lexical clone constructions and their morphological cousins.) The speaker can count on the hearer to be implicitly aware of the lexical and cultural prototype meaning normally picked out by the use of underspecified modifiers (clones, prefixes, adjectives), while at the same time recognizing that the hearer will be able to override the default interpretation if salient aspects of the context render another understanding more plausible. Given the role of general properties of meaning and 8
Note too that in employing a lexical clone XX, a speaker implicates that the category X is not (perhaps contrary to received opinion) monolithic. Whoopi Goldberg notoriously challenged such an assumption in her defence of Roman Polanski during a segment on The View, broadcast on 28 September 2009. According to Goldberg, while it may have been reprehensible for a 43-year-old man to have plied a 13 year old with drugs and champagne before having sex with her, Polanski’s act did not rise to the level of ‘rape rape’. According to those rejecting Goldberg’s parsing, rape is rape, allowing no subcategories for ‘rape rape’ to single out or exclude.
526
526 Laurence R. Horn context in deriving the narrowing signalled in these uses, we are dealing here with lexical pragmatics rather than lexical semantics, although conventionalization always remains a possibility. English offers a constructional opposite of the prototype clone in the form of the un- noun. In accordance with the principles governing these productive formations (cf. Horn 2002a, 2005b), a class A un-noun of the form un-X is not an X, but is situated just outside a given category with whose members it shares a salient function—the uncola is a carbonated soft drink just outside the cola domain, i.e. 7-Up, an unturkey is a large vegan treat designed for the holiday table, an unpublication is a scholarly article that was circulated but never published. Class A un-nouns exemplify Aristotle’s privation, a fundamentally pragmatic notion of opposition defined by disappointed expectation; privatives are marked exceptions lacking a property one would expect to find instantiated at the species or genus level. Just as not just any creature without teeth or eyesight can be called toothless or blind, as Aristotle notes (Categories 12a28–33), so too not just anything that isn’t a cola (e.g. a glass of milk, a hamburger, a T-shirt) could plausibly qualify as an uncola. While an A-class un-noun unX like uncola is Almost but not quite a member of the category X, a B-class un-noun unX is just Barely an X, representing a peripheral member of a given category. Thus, an uncollege (in fact, the US Naval Academy) may lack fraternities, cheerleaders, and binge drinking, an unpolitician is someone seeking office who is not a traditional office-seeker—not a POLITICIAN politician; an unhotel is technically a hotel but one too homey or funky to qualify as a good exemplar of the class—not a HOTEL hotel. If the lexical clone marks the core of category membership, the B-class un-noun represents the periphery; to cite attested examples, an indie unmovie is the opposite of the echt Hollywood movie-movie (Cast Away, as it happens). But what counts as a default instance of the unX or the XX must be grounded inextricably in the speaker’s assumptions (and her assumptions about the hearer’s assumptions) concerning cultural norms and utterance context. Another locus of pragmatically governed word formation is the un-verb. Since Whorf (1936), many linguists have tried their hand at corralling the restrictions on the formation of ‘reversative’ un-verbs; cf. e.g. Marchand (1969), Dowty (1979), Horn (1988b), and Clark et al. (1995). Why can you unwrap a sandwich but not unrecognize its contents or unremember to toss it in the trash? Why can a snake uncoil while a painting can’t unhang? If unfreeze is the opposite of freeze, why is unthaw a synonym of thaw? The standard approach to the constraints on un-verb formation invokes Whorf ’s notion of a cryptotype—a covert category encompassing transitive verbs of covering and enclosing that rules out a wide range of possible bases and outputs of the relevant rule. Following Whorf, Pullum (1999) reckons that there are ‘about a dozen verbs’ that allow un-prefixation, citing undo (a good deed) and unknow as examples of formations we know ‘intuitively’ are impossible because they reverse what it is irreversible. Similarly, other analysts (Clark et al. 1995; Kemmerer and Wright 2002) rule out unbury, unbend, unboil, and undecorate. Yet hundreds of the verbs depicted in the literature as impossible, non-occurring, or (as Whorf labels unsay and unmake) ‘semi-archaic’ are freely attested, even if the actions they denote may be physically irreversible. The Pullum– Whorf hypothesis incorrectly limits the productivity of the process by conflating the
527
Pragmatics and the Lexicon 527 semantic (aspectual) restrictions on un-verb formation with the pragmatic conditions on the way the world (normally) works; verbs like unsay, unknow, unboil, and unhappen are motivated precisely by the need to describe those (typically counterfactual) situations in which the tape of reality is reversed. The pragmatic nature of the restrictions on un-verb formation is supported by a survey of contexts that favour the emergence of innovative un-verbs: advances in science and technology (as in the unerase and undelete commands, or the unfriend or unlike of social networking), science fiction (as in time- travel scenarios), advertising copy (as in KFC’s and Smart cars’ unthink campaigns), and the imagination of poets from Shakespeare (whose ‘un-king’d’ Richard II is the unchallenged monarch of this realm) to pop songsters (‘How can I unlove you?’, ‘Un- break my heart’), whether the implausibility of a given reversal is conceded, mourned, or overridden. (See Horn 2002a and Zimmer et al. 2011 for elaboration.) The semi-productive or nonce formation of lexical clones, un-nouns, and un-verbs demonstrates the role of context in licensing new forms. Two related processes are denominal verbs (e.g. to Houdini one’s way out of a tight situation: Clark and Clark 1979) and deictic compounds (Downing 1977). The latter construction is canonically illustrated by the apple juice seat (for a place setting at which a glass of apple juice had been placed as opposed to the others set with orange juice) or Ferrari woman (in the headline of a news piece referring to a recently deceased woman whose will stipulated that she be buried in her Ferrari, as earlier stories in the same newspaper had revealed). The local context or common ground, assumed to be shared between speaker/ writer and hearers/readers, is typically required to pin down the reference, as when the mother and principal suspect in the disappearance and probable murder of 2-year-old Caylee Anthony in Florida in 2008 could be regularly depicted as the tot mom—for as long as viewers and readers are assumed to be following the case. One underappreciated factor in the possibility of nonce applications of semi- productive word formation processes is the role of priming. This comes in two flavours, of which the more obvious is syntagmatic priming: if the base form has previously occurred (or is just about to occur) in a given local discourse context, it will serve to ease the way for the creative formation. This is obvious in the case of lexical clones, where XX is frequently a response to an earlier speaker’s assertion that or question whether … X …, but it is also a salient property of -ee nominals (cf. Barker 1998) and of both un- verbs and un-nouns, as in (19a) (from the collection in Horn 2002a) or the related attested examples in (19b): (19) a. When a mother knows [that her child has died in an accident], she does everything to unknow. (Edna O’Brien) All those things did happen. Keeping them secret isn’t going to unhappen them. (Iris Murdoch) The English noses in their shapes and unshapes. (Thomas Carlyle) b. I’m telling you this because now that I’ve worked it all out, I can’t unwork it again. (Debby Holt) She pinches and unpinches a crease in her lower lip. (James Hynes) I saved your life; I’d be happy to unsave it. (from House, TV medical show)
528
528 Laurence R. Horn In paradigmatic priming, it’s not the discourse context that does the catalysis but the mental lexicon. When a speaker proposes to unbury the hatchet, or wishes she could unspill the milk, or acknowledges having bridges to unburn, or complains that striking inadmissible testimony won’t unring the bell, the invocation of the corresponding positive idiom or collocation is clear. And the two forms of priming can be symbiotic: (20) I have written books and I cannot unwrite them. I cannot unbelieve what I believe. I cannot unlive my life. —(Hilary Mantel, Wolf Hall, 2009; gratia Annie Zaenen)
27.5 Avoiding Synonymy—and Homonymy I close with a brief survey of one more arena in which the agonistic drama of Q and R plays out: the avoidance of synonymy and homonymy in the lexicon and lexical acquisition. For Sweet (1874) and Passy (1890), a ‘principle of economy’ figures in the avoidance of synonymy as well as in sound change, while Sayce (1880: 192) cites the role of the principle of least effort or ‘laziness’ in word meaning, syntax, and phonology, while conceding that ‘Laziness will not explain everything in speech any more than it will in the ordinary dealings of mankind’. Paul (1889), remarking that ‘Language abhors superfluity’ (1889: chapter XIV), explores the resolution of potential synonymy by the disappearance of one of the two synonymous forms or their gradual differentiation in meaning (cf. Bréal 1897 for a similar account). Typically, the older form becomes restricted to a specialized or secondary meaning: brethren (vs brothers), deer (vs animal), fowl (vs bird). Paul attributes children’s overgeneralization of word meaning to the same underlying principle. Paul’s observations appear elsewhere in a variety of incarnations, including the Elsewhere Condition, which traces back to Pāṇini (cf. Anderson 1992; Giegerich 2001) and its descendants, including Blocking (Aronoff 1976), Pre-Emption by Synonymy (‘If a potential innovative word-form would be precisely synonymous with a well- established word, the innovative word is pre-empted by the well-established word and is therefore considered unacceptable’—Clark and Clark 1979), the Avoid Synonymy Condition (Kiparsky 1983), and within language acquisition the principles of Contrast and Mutual Exclusivity. Contrast (Clark 1987) stipulates that any two distinct forms contrast in meaning and that innovative forms fill lexical gaps but don’t displace established lexicalizations. Mutual Exclusivity (cf. Markman and Wachtel 1988 and much subsequent work) addresses children’s predisposition to reject lexical overlap, in which an unfamiliar label is applied to a familiar object with a known name. If synonymy avoidance is a reflex of the R Principle, as suggested by its characterization in terms of least effort, laziness, or paradigmatic economy, why do we not find
529
Pragmatics and the Lexicon 529 analogous principles correlated with the Q Principle, Zipf ’s hearer’s economy? Or do we? Note first the functional parallel between the two tendencies: AVOID SYNONYMY 2 forms → 1 meaning R-based (Costly for Speaker) Result: 1 form grows a new meaning
AVOID HOMONYMY 2 meanings → 1 form Q-based (Hard for Hearer) Result: 1 meaning grows a new form
Whence then the asymmetry? In fact, earlier sources do find a symmetry between the two principles, with the proviso (e.g. in Paul 1889) that apparent homonymy is tolerated because the context determines the intended or salient meaning, as in the differential interpretations of sheet at a clothing store, on a yacht, or at a printer’s shop. Pernicious homonymy arises only when the hearer is generally unable to recover the speaker’s intended meaning because of an overlap not just in the phonology but in the grammatical category, subcategory, sense domain, and register of each homophone: ‘Only when the words are alike in sound, when they are in common use in the same social and intellectual circles, when they perform the same syntactical functions in the language, within the same sphere of ideas, do they become subject to mutual confusion and conflict’ (Williams 1944: 5). Thus no problem is posed by pairs like fly (N/V), red/read, limp (A/V), cape, pound (weight/currency); but homonymy destroys the weaker member of pairs like gat (