142 40 23MB
English Pages [843] Year 2017
The Cambridge Handbook of Cognitive Linguistics The best survey of cognitive linguistics available, this handbook provides an in-depth explanation of its rich methodology, key results, and interdisciplinary context. With thorough coverage of the research questions, basic concepts, and various theoretical approaches, the handbook addresses newly emerging subfields and shows their contribution to the discipline. The handbook introduces fields of study that have become central to cognitive linguistics, such as conceptual mappings and construction grammar. It explains all the main areas of linguistic analysis, traditionally expected in a full linguistics framework, and includes fields of study such as language acquisition, sociolinguistics, diachronic studies and corpus linguistics. Setting linguistic facts within the context of many other disciplines, the handbook will be welcomed by researchers and students in a broad range of disciplines, including linguistics, cognitive science, neuroscience, gesture studies, computational linguistics and multimodal studies. B A R B A R A D A N C Y G I E R is Professor in the Department of English, University of British Columbia, Canada. Her many books include The Language of Stories (2012), Figurative Language (2014, co-authored with Eve Sweetser) and Viewpoint in Language (2012, co-edited with Eve Sweetser), all published with Cambridge University Press.
c a m b r i d g e ha n d b o o k s i n l a n g ua g e an d l i n g u i s t i cs
Genuinely broad in scope, each handbook in this series provides a complete state-of-the-field overview of a major subdiscipline within language study and research. Grouped into broad thematic areas, the chapters in each volume encompass the most important issues and topics within each subject, offering a coherent picture of the latest theories and findings. Together, the volumes will build into an integrated overview of the discipline in its entirety.
Published titles The Cambridge Handbook of Phonology, edited by Paul de Lacy The Cambridge Handbook of Linguistic Code-switching, edited by Barbara E. Bullock and Almeida Jacqueline Toribio The Cambridge Handbook of Child Language, Second Edition, edited by Edith L. Bavin and Letitia Naigles The Cambridge Handbook of Endangered Languages, edited by Peter K. Austin and Julia Sallabank The Cambridge Handbook of Sociolinguistics, edited by Rajend Mesthrie The Cambridge Handbook of Pragmatics, edited by Keith Allan and Kasia M. Jaszczolt The Cambridge Handbook of Language Policy, edited by Bernard Spolsky The Cambridge Handbook of Second Language Acquisition, edited by Julia Herschensohn and Martha Young-Scholten The Cambridge Handbook of Biolinguistics, edited by Cedric Boeckx and Kleanthes K. Grohmann The Cambridge Handbook of Generative Syntax, edited by Marcel den Dikken The Cambridge Handbook of Communication Disorders, edited by Louise Cummings The Cambridge Handbook of Stylistics, edited by Peter Stockwell and Sara Whiteley The Cambridge Handbook of Linguistic Anthropology, edited by N.J. Enfield, Paul Kockelman and Jack Sidnell The Cambridge Handbook of English Corpus Linguistics, edited by Douglas Biber and Randi Reppen The Cambridge Handbook of Bilingual Processing, edited by John W. Schwieter The Cambridge Handbook of Learner Corpus Research, edited by Sylviane Granger, Gae¨tanelle Gilquin and Fanny Meunier The Cambridge Handbook of English Historical Linguistics, edited by Merja Kyto¨ and Pa¨ivi Pahta The Cambridge Handbook of Linguistic Multicompetence, edited by Li Wei and Vivian Cook The Cambridge Handbook of Formal Semantics, edited by Maria Aloni and Paul Dekker The Cambridge Handbook of Morphology, edited by Andrew Hippisley and Greg Stump The Cambridge Handbook of Historical Syntax, edited by Adam Ledgeway and Ian Roberts The Cambridge Handbook of Linguistic Typology, edited by Alexandra Y. Aikhenvald and R. M. W. Dixon The Cambridge Handbook of Areal Linguistics, edited by Raymond Hickey The Cambridge Handbook of Cognitive Linguistics, edited by Barbara Dancygier
The Cambridge Handbook of Cognitive Linguistics Edited by Barbara Dancygier
University Printing House, Cambridge CB2 8BS, United Kingdom One Liberty Plaza, 20th Floor, New York, NY 10006, USA 477 Williamstown Road, Port Melbourne, VIC 3207, Australia 4843/24, 2nd Floor, Ansari Road, Daryaganj, Delhi – 110002, India 79 Anson Road, #06–04/06, Singapore 079906 Cambridge University Press is part of the University of Cambridge. It furthers the University’s mission by disseminating knowledge in the pursuit of education, learning, and research at the highest international levels of excellence. www.cambridge.org Information on this title: www.cambridge.org/9781107118447 DOI: 10.1017/9781316339732 © Cambridge University Press 2017 This publication is in copyright. Subject to statutory exception and to the provisions of relevant collective licensing agreements, no reproduction of any part may take place without the written permission of Cambridge University Press. First published 2017 Printed in the United Kingdom by TJ International Ltd. Padstow Cornwall A catalogue record for this publication is available from the British Library. ISBN 978-1-107-11844-7 Hardback Cambridge University Press has no responsibility for the persistence or accuracy of URLs for external or third-party Internet Web sites referred to in this publication and does not guarantee that any content on such Web sites is, or will remain, accurate or appropriate.
To Jacek and Szymek
Contents
List of Figures List of Tables List of Contributors Acknowledgments
page x xiii xiv xvii
Introduction Barbara Dancygier
1
Part I Language in Cognition and Culture 1 Opening Commentary: Language in Cognition and Culture N. J. Enfield 2 Relationships between Language and Cognition
Daniel Casasanto 3 The Study of Indigenous Languages Sally Rice 4 First Language Acquisition Laura E. de Ruiter
and Anna L. Theakston 5 Second Language Acquisition
Andrea Tyler
Part II Language, Body, and Multimodal Communication 6 Opening Commentary: Polytropos and Communication in the Wild Mark Turner 7 Signed Languages Sherman Wilcox and Corinne Occhino 8 Gesture, Language, and Cognition Kensy Cooperrider
and Susan Goldin-Meadow 9 Multimodality in Interaction
11
13 19 38 59 73 91
93 99 118
Kurt Feyaerts, Geert Broˆne,
and Bert Oben 10 Viewpoint Lieven Vandelanotte 11 Embodied Intersubjectivity Jordan Zlatev
135 157 172
viii
Contents
12 Intersubjectivity and Grammar
Ronny Boogaart
and Alex Reuneker
188
Part III Aspects of Linguistic Analysis 13 Opening Commentary: Linguistic Analysis John Newman 14 Phonology Geoffrey S. Nathan 15 The Construction of Words Geert Booij 16 Lexical Semantics John R. Taylor 17 Cognitive Grammar Ronald W. Langacker 18 From Constructions to Construction Grammars
207 209 214 229 246 262
Thomas Hoffmann Construction Grammars Thomas Hoffmann Cognitive Linguistics and Pragmatics Kerstin Fischer Fictive Interaction Esther Pascual and Todd Oakley Diachronic Approaches Alexander Bergs
284 310 330 347 361
19 20 21 22
Part IV Conceptual Mappings 23 Opening Commentary: Conceptual Mappings Eve Sweetser 24 Conceptual Metaphor Karen Sullivan 25 Metonymy Jeannette Littlemore 26 Conceptual Blending Theory Todd Oakley and Esther Pascual 27 Embodiment Raymond W. Gibbs, Jr. 28 Corpus Linguistics and Metaphor Elena Semino 29 Metaphor, Simulation, and Fictive Motion Teenie Matlock
377 379 385 407 423 449 463 477
Part V Methodological Approaches 30 Opening Commentary: Getting the Measure of Meaning
491
Chris Sinha The Quantitative Turn Laura A. Janda Language and the Brain Seana Coulson Cognitive Sociolinguistics Willem B. Hollmann Computational Resources: FrameNet and Constructicon Hans C. Boas 35 Computational Approaches to Metaphor: The Case of MetaNet Oana A. David 36 Corpus Approaches Stefan Th. Gries 37 Cognitive Linguistics and the Study of Textual Meaning Barbara Dancygier
493 498 515 533
Part VI Concepts and Approaches: Space and Time 38 Linguistic Patterns of Space and Time Vocabulary
623
31 32 33 34
Eve Sweetser and Alice Gaby
549 574 590 607
625
Contents
39 Space–time Mappings beyond Language Alice Gaby and Eve
Sweetser 40 Conceptualizing Time in Terms of Space: Experimental Evidence Tom Gijssels and Daniel Casasanto 41 Discovering Spatiotemporal Concepts in Discourse
Thora Tenbrink References Index
635 651 669 684 817
ix
Figures
7.1 7.2 10.1 10.2 10.3 11.1
12.1 12.2 12.3 12.4 12.5 12.6 17.1 17.2 17.3 17.4 17.5 17.6 17.7 17.8 17.9 17.10 19.1 19.2 20.1 22.1
MI N D I S A CO N T A I N E R T O B E UN L O C K E D page 107 Modal forms in LSC 113 Decompression for viewpoint 167 Viewpoint compression from Discourse Space to Story-Viewpoint Space 167 Embedded viewpoints with cross-input projection 169 Pictures eliciting non-actual motion descriptions, according to the two parameters Affordance and Perspective: (a) [+afford, 3pp], (b) [+afford, 1pp], (c) [-afford, 3pp], (d) [-afford, 1pp]. Reprinted from Blomberg (2014), with explicit permission from the author 185 Subjective and objective construal 190 Basic scheme for the meanings of expressions 191 Construal configuration and its basic elements 191 Construal configuration for a non-perspectivized utterance197 Construal configuration for first-person perspective 198 Construal configuration for third-person perspective 198 Viewing arrangement 264 Vantage point 264 Profiling of things 265 Profiling of relationships 266 A symbolic assembly 270 Connection and grouping 273 Semantic vs. prosodic grouping 274 Seriality and constituency 274 Moving window metaphor 280 Packaging and ellipsis 281 Partial taxonomic construction network for the Ditransitve construction family 314 Selected multiple inheritance relationships in (1) 316 A tutor introduces a lamp to the iCub robot 342 The S-curve of diffusion 368
List of Figures
24.1 Angry man with steaming ears and exploding head 24.2 A CMT diagram of U N D E R S T A N D I N G I S S E E I N G 24.3 When two containers hold the same volume of fluid, children will judge the taller container (here, the one on the left) to contain more fluid 24.4 The verb exercise evokes the Exercising frame 24.5 The item exercise evokes the Exercising frame and the B O D Y domain 25.1 The ‘food and drink’ ICM 25.2 Key metonymy types in Radden and Ko¨vecses (1999) taxonomy (adapted from Littlemore 2015) 26.1 Example 1 (Classic mental spaces analysis) 26.2 Example 1 (ersatz neighbors blend) 26.3 Examples 2 and 3 (material anchors) 26.4 Image: Douwe Egberts ‘Return-Ticket Coffee’ 26.5 Mirror network blend (Douwe Egberts) 26.6 Example 6 (single-scope network) 26.7 Example 7 (double-scope network) 26.8 Example 8 (grounded blend: Michelin) 26.9 Example 10 (Ikea cartoon, used with permission from the artist) 26.10 Example 11 (‘I love you’) 26.11 Symbicort ad: grandfather and grandson 26.12 Symbicort ad: wolf with COPD 26.13 Symbicort ad: wolf and she-wolf 26.14 Symbicort ad: party 26.15 Three Little Pigs fable 26.16 Big Bad Wolf with COPD (composition and completion) 26.17 Big Bad Wolf with COPD (elaboration) 26.18 Symbicort ad: grounding spaces 31.1 Percentage of articles presenting quantitative studies published in Cognitive Linguistics 1990–2015 31.2 CART tree for Russian gruzit’ ‘load’ from Baayen et al. 2013 32.1 The left hemisphere of the human cerebral cortex with major landmarks indicating its division into four lobes. (Brain image was visualized with BrainNet Viewer [Xia, Wang, and He, 2013]) 32.2 Image of the left hemisphere with approximate locations of the two classical language areas. (Brain image was visualized with BrainNet Viewer [Xia, Wang, and He, 2013].) 34.1 Subcorpora of avenge for annotation in the FN Desktop 34.2 Description of Revenge frame 34.3 First part of lexical entry report of avenge 34.4 Second part of lexical entry report of avenge 34.5 Partial FrameGrapher representation of Revenge frame 34.6 First part of Way_manner construction entry
390 394
396 402 404 410 411 424 425 429 431 432 433 434 436 437 437 441 441 442 442 443 444 445 446 499 509
517
521 554 556 558 559 563 568
xi
xii
List of Figures
34.7 34.8 35.1 35.2 35.3 35.4 35.5 35.6 35.7
Second part of Way_manner construction entry Third part of Way_manner construction entry Schematic conceptual network MetaNet integration of three systems – repository, identification, and annotated data Schematic representation of frames and metaphors Relations among frames Relations among metaphors, and relationship of frames to metaphors The metaphor identification pipeline Sample metaphor identification run identifies pillars of democracy as metaphoric
569 569 580 582 583 584 584 586 586
Tables
3.1 Some of the indigenous or minority languages studied from a CL or CL-compatible perspective by authors who have conducted their own original fieldwork or who published dissertations or multiple studies on these languages page 39 3.2 Some cross-domain referential and relational extensions of Tarascan spatial morphemes based on body-part terms (adapted from Friedrich 1969) 43 7.1 Mind–body dualism applied to language, sign, and gesture 102 11.1 The mimetic hierarchy of semiotic development (adapted from Zlatev 2013) 177 19.1 Summary of properties of Construction Grammars 328 20.1 Frame elements of the Lamp frame 343 20.2 Average number of frame elements of the Lamp frame mentioned and results of a univariate ANOVA for human–robot interaction with contingent response (HRI-cont), random response (HRI-rand) and child-directed speech (CDS) 345 22.1 I-Umlaut and analogy 363 34.1 Syntactic variability of FEs with verbal LUs evoking the Revenge frame 560 34.2 Lexical and constructional description and annotation compared 567
Contributors
Alexander Bergs Institute for English and American Studies, Osnabru¨ck University, Germany Hans C. Boas Department of Linguistics/Department of Germanic Studies, The University of Texas at Austin, USA Ronny Boogaart Leiden University Center for Linguistics, Leiden University, The Netherlands Geert Booij Leiden University Center for Linguistics, Leiden University, The Netherlands Geert Broˆne Department of Language and Communication, KU Leuven @ Antwerp, Belgium Daniel Casasanto Department of Psychology and Grossman Institute for Neuroscience, University of Chicago, USA Kensy Cooperrider Department of Psychology, University of Chicago, USA Seana Coulson Cognitive Science Department, University of California, San Diego, USA Barbara Dancygier Department of English, University of British Columbia, Canada Oana A. David Cognitive and Information Sciences Department, University of California, Merced, USA Laura E. de Ruiter ESRC International Centre for Language and Communicative Development (LuCiD), University of Manchester, UK N. J. Enfield Department of Linguistics, The University of Sydney, Australia Kurt Feyaerts Department of Linguistics, KU Leuven, Belgium Kerstin Fischer Department of Design and Communication, University of Southern Denmark, Denmark Alice Gaby School of Languages, Literatures, Cultures and Linguistics, Monash University, Australia Raymond W. Gibbs, Jr. Department of Psychology, University of California, Santa Cruz, USA
List of Contributors
Tom Gijssels Department of Psychology, University of Chicago, USA; Department of Linguistics, Free University of Brussels, Belgium Susan Goldin-Meadow Department of Psychology, University of Chicago, USA Stefan Th. Gries Department of Linguistics, University of California, Santa Barbara, USA Thomas Hoffmann English Language and Linguistics, Department of English and American Studies, Katholische Universita¨t Eichsta¨ttIngolstadt, Germany Willem B. Hollmann Department of Linguistics and English Language, Lancaster University, UK Laura A. Janda Department of Language and Culture, UiT – The Arctic University of Norway, Tromsø, Norway Ronald W. Langacker University of California, San Diego, USA Jeannette Littlemore Department of English Language and Applied Linguistics, University of Birmingham, UK Teenie Matlock Cognitive and Information Sciences Department, University of California, Merced, USA Geoffrey S. Nathan Department of English, Wayne State University, USA John Newman Department of Linguistics, University of Alberta, Canada Todd Oakley Department of Cognitive Science, Case Western Reserve University, USA Bert Oben Department of Linguistics, KU Leuven, Belgium Corinne Occhino Department of Linguistics, University of New Mexico, USA Esther Pascual Department of Spanish and Department of Linguistics, School of International Studies, Zhejiang University, China Alex Reuneker Leiden University Center for Linguistics, Leiden University, The Netherlands Sally Rice Department of Linguistics, University of Alberta, Canada Elena Semino Department of Linguistics and English Language, Lancaster University, UK Chris Sinha School of Foreign Languages, Hunan University, China; Centre for Language and Literature, Lund University, Sweden Karen Sullivan School of Languages and Cultures, University of Queensland, Brisbane, Australia Eve Sweetser Department of Linguistics, University of California, Berkeley, USA John R. Taylor Department of Linguistics, University of Otago, New Zealand Thora Tenbrink School of Linguistics and English Language, Bangor University, UK Anna L. Theakston ESRC International Centre for Language and Communicative Development (LuCiD), University of Manchester, UK Mark Turner Department of Cognitive Science, Case Western Reserve University, USA
xv
xvi
List of Contributors
Andrea Tyler Department of Linguistics, Georgetown University, Washington, DC, USA Lieven Vandelanotte English/Linguistics, University of Namur and KU Leuven, Belgium Sherman Wilcox Department of Linguistics, University of New Mexico, USA Jordan Zlatev Centre for Language and Literature, Lund University, Sweden
Acknowledgments
A handbook like this one is, by definition, a collaborative piece of scholarship. I want to thank all the contributors for agreeing to write, helping me through various stages and for giving their time, enthusiasm, and wisdom to the project. While there are surely things I have missed and mistakes I have made, whatever I did right was thanks to the support I received in difficult moments. I want to give special thanks here to John Newman, Arie Verhagen, Alice Gaby, Nick Enfield, Daniel Casasanto, Eve Sweetser, and Lieven Vandelanotte, who helped solve issues, offered advice, and gave me the courage to go on. I am grateful to Cambridge University Press, for their unfailing professionalism and support, and especially to Andrew Winnard, for his vision, advice, and understanding. My thanks also go the Institute for Advanced Study, Durham University, UK, for providing an excellent environment to focus on the project. Special thanks are due to Jasmine Spencer, for her expert and friendly help with copy-editing. I am also grateful, as always, to Jacek and Szymek. Without them, nothing is possible.
Introduction Barbara Dancygier
Cognitive Linguistics is a relatively new but fast-growing discipline of language study. It started out in the 1980s, with the work of George Lakoff, Ronald Langacker, Charles Fillmore, and Len Talmy, and has progressed since then into a rich and multifaceted research paradigm, offering new tools to many branches of linguistic enquiry and giving new coherence to a range of linguistic interests. At the same time, it is a framework which both listens to and speaks to outlying disciplines of cognitive science, psychology, and philosophy of mind. It has also inspired new interests in areas such as linguistic anthropology, semiotics, and stylistics, to name just a few. This handbook takes stock of what cognitive linguistics does, how it approaches the study of language, and how it contributes to the study of cognition and communication.
0.1 Cognitive Linguistics and Its Theoretical Priorities Cognitive linguistic enquiry builds on the assumption that meaning is an inseparable part of linguistic study. Much of the research described in this handbook attempts to uncover correlations between form and meaning, rather than focus on these aspects of language independently of each other. The assumption has created two increasingly popular approaches to grammatical form – construction grammar and cognitive grammar – but is also re-defining possible theoretical takes on central components of linguistic analysis such as morphology and phonology. Specific models, especially of construction grammars, now propose a number of ways to approach the issue of the relations between form and meaning, but the overall commitment remains the same. Cognitive linguistic enquiry further assumes that language study needs to be usage-based. Originally, this meant calling for an end to ‘armchair linguistics’ and a thorough account of the facts of language, in its use and
2
DANCYGIER
context, in its various instantiations, and in its connection to culture on the one hand and to the mind on the other. It also means that language as such emerges from usage. Actual usage is at the core of cognitive linguistic study, which has not only opened up many areas of interest, but has also redefined the scope of what is considered central to linguistic enquiry. For many cognitive linguists, engaging with broadly construed and varied methods of data collection has also become important. There are no artificially drawn dividing lines – attested and responsibly gathered linguistic data are all subjects of study. The usage-based approach means not only that theoretical concepts represent the data well, but also that lines of investigation can be postulated on the basis of what usage suggests. Furthermore, explanations of major processes, such as first language acquisition, are also embedded in the collected usage facts and support the argument that language emerges in use, to satisfy communicative needs, rather than being driven by innate grammatical mechanisms. This general approach of theoretical responsiveness to actual facts of language communication (rather than abstract grammaticality judgments) is what makes cognitive linguistics a particularly appealing approach to language study. There are many ways in which the commitment is addressed – computational approaches to language flourish within the cognitive linguistic framework, and corpus work and other quantitative methods are on the rise. Also, pragmatic considerations and cognitive discourse analysis enhance our understanding of how meaning and form operate in discourse contexts, while cognitive sociolinguistics puts language in a broader frame of social and typological differences. The relationship between language and cognition is one of the important tenets of cognitive linguistics. It has often been talked about as the cognitive commitment – a promise to build linguistic descriptions and postulate theoretical concepts which are at least informed, if not fully justified, by what is now known about the human brain and human cognition. This approach has several effects. In keeping with the cognitive commitment, cognitive linguists can be expected to learn from the research of neuroscientists and cognitive psychologists and seek various forms of cooperation with them. It is also hoped that what linguists uncover in their interpretation of linguistic facts will be interesting and useful to those on the other side. This explains why so many of the contributors to this handbook have their intellectual homes in psychology or cognitive science departments. However, the precise nature of the connection between language and conceptualization is a question which needs constant re-evaluation. Initially, the assumption was that the study of linguistic meaning would help uncover the conceptual nature of meaning. Introspective study of meaning has indeed postulated a number of important ideas and established a broad range of terms which still constitute the core of cognitive linguistic study – mapping, construal, frame, mental space, etc.
Introduction
However, more and more questions were being asked about how to confirm the validity of these claims. There are essentially three ways to address the issue. One is to look at other communicative facts – cospeech gesture and multimodal interaction – to see if similar conceptual structures emerge from the facts established. The second one is to seek confirmation, inspiration, and more specific knowledge by taking advantage of available language databases (corpora) and collecting data through the searches. The third way is to turn to an experimental context – a linguist’s, psycholinguist’s or neuroscientist’s lab. All three approaches have brought out significant clarifications and help move the discipline forward. The interdisciplinary nature of such collaborations has made Cognitive Linguistics the rich and multifaceted discipline it has grown into. Importantly, these engagements have led to important cross-checks: introspective analysis of examples can be tested in the environment of discourse analysis and broad computer searches, examined from the perspective of cross-linguistic validity and gesture/multimodality representation, and investigated via psycholinguistic experiments. The final picture may confirm some of the original hypotheses, while adding new questions and highlighting new connections. The multifaceted approach and its ability to view language phenomena from many perspectives is the most striking quality of the cognitive linguistic approach. Part VI of this handbook shows the potential of these ‘checks and balances’ by showcasing various approaches to the issue of linguistic expressions of space and time. The interests mentioned above create a very complex picture. On the one hand, cognitive linguists are deeply invested in what has typically been considered the ‘core’ of linguistic analysis – phonology, morphology, and syntax. On the other hand, all the work done is focused on meaning, in all its aspects. This means that cognitive linguists do not specifically reference the fields of linguistic semantics and pragmatics as such – if all linguistic facts are about linguistic choices made in usage, there seems to be no need to speak separately about context, inferences, etc. This is often an unspoken assumption, taken up afresh by one of the chapters in the handbook. Looking beyond its central assumptions, cognitive linguistics is also engaging with areas of study not typically considered to belong to the core – sociolinguistics, discourse analysis, corpus linguistics, stylistics, diachronic studies, etc. This cross-pollination between cognitive linguistics on the one hand and traditionally less central areas of language study on the other is creating a new situation – a framework whose basic principles not only serve various linguistic interests, but also bring once-impossible cohesion to our broad understanding of how language works. This does not mean that cognitive linguistics is just a loose coalition of approaches. On the contrary,
3
4
DANCYGIER
language is being looked at in all its complexity from the coherent perspective of meaning-emergence processes and their embeddedness in conceptual patterns.
0.2 Analytical Directions There are some analytical concepts, richly represented in this handbook, which have achieved significant prominence. An area of research which is an exclusively cognitive linguistic contribution to the study of language is the area of conceptual mappings. Initiated over thirty years ago by George Lakoff and Mark Johnson, in their linguistic and philosophical enquiry, the study of mappings has advanced work on lexical usage, while offering new and in-depth explanations of phenomena such as polysemy, which until recently have been treated in a less explanatory way. Mappings are now studied extensively and in many ways (analytically, through language corpora, cross-linguistically, in a gesture lab and a psycholinguist’s lab). They also begin to form the basis of a clearer understanding of the overall roles that figuration and mappings play in the very nature of language, beyond lexical semantics. One of the central contributions of cognitive linguistic methodology to the study of language is the study of constructions – form-meaning pairings linking syntactic forms to their meaning. The original bold claim that syntax has meaning may have felt somewhat counterintuitive at the time of its emergence, but this claim has given rise to a range of approaches to syntax which are changing the way linguistic form is studied. Several chapters in the handbook discuss Construction Grammar – its history, current approaches, computational applications, intersubjective grounding, etc. Another broad concept routinely used in cognitive linguistic study is embodiment – the understanding that linguistic meaning, including abstract meaning, is rooted in the role of the human body in shaping human cognition, and thus also language. The range of work relying on embodiment in one way or another is impressive – from reliance on cognitive scientists’ treatment of the grounding problem to basic assumptions about conceptual mappings or the formulation of the principles of Embodied Construction Grammar. But there is also a renewed and exciting line of research which attempts to clarify the role of the human body in interaction and communication, not just in the formation of concepts; in other words, this is a sustained line of research attempting to see language beyond what it does in the brain. There are in fact several research frameworks emerging from these interests. The most prominent is the study of spontaneous co-speech gesture, which also extends into communicative role of eye-gaze, facial expression, and other embodied aspects of interaction. Multimodality – the approach which looks at spontaneous
Introduction
communication in all its embodied and linguistic complexity – is making possible the development of powerful research tools and is taking a very prominent role in shaping debates about the nature of language. It is also moving forward the issue of usage and reliable data: while cognitive linguists increasingly use language corpora, the methodology still relies primarily on usage recorded, transcribed, and tagged. Looking at multimodal interaction involves expanding the nature of relevant data: looking at language in its visual, aural, oral, and sensory-motor embodied dimensions. For the first time, we are building a complete understanding of how humans communicate. Clearly, cognitive linguistics has developed into a full linguistic paradigm, offering methodologies to study traditionally central areas of linguistics (such as phonology, morphology, or syntax), while also enriching them and bringing them into a cohesive framework with what has often been considered the periphery – lexical meaning, discourse, or pragmatics. It has taken language out of a predetermined cell, and allowed it to be seen in the realistic context in which it is used. Cognitive linguistics also developed methods, concepts, terminologies, and consistent frameworks to cover the complexity of meaning emergence, in all its multiplicity and variety. It is important here to note how the discipline has been changing through time. It started out looking at word meaning (especially from the perspective of conceptual mappings such as metaphor, as defined by George Lakoff and Mark Johnson), and at the approach to grammar which describes the nature of how speakers make linguistic choices to represent best their construal of the situation, as proposed by Ronald Langacker. Much of the work was introspective, though increasingly based on attested examples. The idea of formmeaning pairings was then extended to prompt the concept of a construction, as introduced in the work of Charles Fillmore and Adele Goldberg. These core interests – mappings, cognitive grammar, and constructions – remain central to much of the work cognitive linguists do, but several important trends have emerged meanwhile. First of all, the analyses are increasingly based on corpora (collections of data which can be searched for a range of forms) or experiments. The range and sophistication of such methods are increasing. Second, the facts gathered are now worked on through computational methods – frames, constructions, and metaphors have all been approached from that perspective. Third, the study of multimodality in interaction is developing new methods, labbased, but also relying on video corpora, documenting all aspects of communicative events. And finally, improved knowledge of the flexibility of the brain’s functioning and more sophisticated experimental methods used in cognitive science create a much better fit between the linguist’s work and the cognitive scientist’s endeavors. To sum up, cognitive linguistics started out by opening a range of questions that scholars of language can approach in a disciplined way; soon, it
5
6
DANCYGIER
was able to offer research tools and theories which shape the enquiry and build a sophisticated and rich linguistic paradigm. It opened its door to existing subdisciplines of language study, and created new subdisciplines, such as the study of multimodality. It increased its reliability through sustained exchanges with cognitive scientists, and, last but not least, expanded its methodological reach by the incorporation of corpus linguistics methods and increasing the use of quantitative and computational approaches. It has become the most versatile and descriptively adequate linguistics paradigm on offer. It continues improving its methods and the accuracy of its claims. It is poised to become the approach whose explanatory and descriptive power will lead language study in all its forms. Language is the core of what it is to be human. It does not reside in a welldefined area in the individual brain; instead, it inhabits all that we do and how we interact with the world and each other. It is complex – as complex as the human capacity to create and express meanings. It calls for an approach which opens itself to the complexity, with its attendant challenges. The cognitive linguistic approach has prepared the groundwork for understanding that complexity and meeting the significant challenges that the approach entails.
0.3 The Structure of This Handbook This handbook documents the range of ways in which cognitive linguists do their job. It covers the main concepts, approaches and methodologies used in cognitive linguistics and shows the results of the interaction between cognitive linguists and other scholars researching language. It consists of six major parts, each of which covers a range of theoretical or methodological questions. Each part (except Part VI) starts with an opening commentary – an introductory reflection providing a bird’s-eye view of the issue that the chapters in that section address in detail. The reader interested in any of the specific chapters might first look at the opening commentary to be able to contextualize the claims of the chapter in question within a broader context. Each chapter is a free-standing text, but they all connect, in various ways, with many other chapters in the handbook. In some cases it is because the chapter focuses on one methodological solution or concept, while the general background to the question has been given elsewhere, or because two chapters give two different takes on the issue. Following cross-chapter connections, via the index and in-text links, will help the reader get an in-depth perspective on a question of interest. Part I (Language in Cognition and Culture) includes chapters which jointly consider the way that language participates in human understanding of the world and of culture. Language does not exist in a vacuum – it gives every individual speaker a way to participate in a linguistic
Introduction
community. Linguistic communities, in turn, rely on shared access to language. Naturally, participating in a culture or cultures requires using one’s cognitive and linguistic abilities. The aim of the chapters in Part I is thus to spell out broader assumptions about the cognitive linguistic treatment of cross-linguistic evidence, about the role of culture, and about the cognitive processes responsible for language acquisition and learning. Specifically, Chapter 1 introduces the range of questions raised by the dual role of language – its participation both in individual cognition and in shared culture or cultures. Chapter 2 contextualizes the goals of cognitive linguistics in the current work on cognition and embodiment, showing what the ‘cognitive’ in ‘cognitive linguistics’ means, in view of current research in cognitive science. Chapter 3 presents the role that cognitive linguistics can play in the study of indigenous languages, but also, more specifically, in the issues of linguistic diversity, language revitalization, and the treatment of endangered languages; after that, Chapters 4 and 5 show the methods and results of the cognitive linguistic study of language acquisition – Chapter 4 considers first-language acquisition, while Chapter 5 presents the work on second-language acquisition. Both chapters review a range of studies, showing what the usage-based approach means in addressing fundamental issues of language emergence. In Part II (Language, Body, and Multimodal Communication), six chapters show how cognitive linguistics uses ideas from psychology and cognitive psychology which relate mental abilities of humans to the human body and the multimodal nature of interaction. They propose specific theoretical solutions to issues of language use in the context of embodied behavior and interaction. Chapter 6 shows why the idea of multimodality of communication is changing the way we view language, and, consequently, linguistics work. Chapter 7 considers the issue of mind–body dualism, to clarify the nature of signed languages, and the application of various linguistic approaches to signedlanguage research; it also addresses the connection between the use of co-speech gesture and signed languages. Chapter 8 explains the role gesture plays in communication and also in providing evidence of conceptualization patterns. Chapter 9 expands the theme, arguing for the need to focus the study of linguistic usage on the context of multimodal interaction. This chapter provides the big picture by reviewing theoretical claims and methods and presenting instructive case studies. The theoretical approach to multimodality and meaning emergence in general also needs to uncover general mental phenomena, such as profiling of viewpoint – in grammatical choices, lexical choices, and textual forms; this is shown in Chapter 10. Finally, Chapters 11 and 12 present two approaches to the question of intersubjectivity – explaining the embodied grounding of communication and showing its role in linguistic form, including intersubjective constructions. While Chapter 11 gives a broader, philosophically based understanding of
7
8
DANCYGIER
intersubjectivity, also showing its role in the emergence of figurative forms such as metaphor, Chapter 12 explains the nature of the intersubjective construal and cognitive coordination across speakers, showing the ways in which they prompt grammatical choices. Part III shows the ‘linguistic’ approach proposed within cognitive linguistics. Cognitive linguistics is primarily focused on representation of meaning, at various levels of linguistic structure. However, recent research applies cognitive linguistic methodology to all aspects of language study. This part presents the theoretical foundations of cognitive linguistics methodology in all traditionally recognized areas of language study and presents standard theories that deal with such phenomena: Chapter 13 provides a theoretical overview, while Chapters 14, 15, and 16 deal with phonology, morphology and lexical semantics. Importantly, Chapter 14 offers a sound comparison of the cognitive approach to phonology with the generative one. Chapter 15 discusses morphology in terms of constructional schemas, thus linking the structure of words to the structure of sentences. Chapter 16 gives a broad view of lexical semantics and approaches to word meaning, before showing some of the specific contributions of cognitive linguistic theories. Chapter 17 presents a rich overview of one of the best established cognitive approaches to language analysis – Langacker’s Cognitive Grammar. Looking at lexicon, morphology, and syntax as a continuum of symbolic structures, the chapter shows how grammar is shaped by the interplay of interactive, descriptive, discursive, and processing factors, rather than being seen as a selfcontained mechanism. The next two chapters, 18 and 19, present Construction Grammar; while Chapter 18 presents the re-emergence of the concept of a ‘construction’ in cognitive linguistics, Chapter 19 goes into much theoretical detail, showing what various construction grammars share and how they are different. Later in Part III, Chapter 20 clarifies what cognitive linguistic work shares with issues raised within pragmatics, and how the two disciplines might approach the same questions, while Chapter 21 examines the structure of fictive interaction – an important phenomenon in tracing the connections between morphology and syntax on the one hand and discourse patterns on the other; finally, Chapter 22 looks at the historical perspective on language; it first describes the general effects of analogy, iconicity, and frequency on linguistic change, and then illustrates the cognitive linguistics contribution to major questions regarding grammaticalization, lexicalization, and constructional change. Part IV presents an overview of types of conceptual mappings studied by cognitive linguists and the ways that they have effectively been studied within various fields of linguistics, psychology, and cognitive science. Chapter 23 gives a broad overview of the types of questions mappings raise; then Chapter 24 describes the mapping which has received the most attention – conceptual metaphor. Chapter 25 discusses metonymy,
Introduction
illustrating the mapping with a range of examples, while Chapter 26 shows the linguistic and discourse applications of conceptual blending – a general approach to meaning emergence, here described from the perspective of its role in linguistic enquiry. The remaining three chapters in Part IV contextualize the study of metaphor: Chapter 27 looks at the connections between work in linguistics and a range of studies in psychology and cognitive neuroscience, clarifying the nature of the claims regarding the ‘embodied nature of meaning.’ Chapter 28 shows how mappings are studied within corpus linguistics, exemplifying several approaches and showing practical examples. Finally, Chapter 29 considers fictive motion, a figurative form which raises important questions about the nature of conceptual mappings and the cognitive mechanisms behind them. Part V gives a sampler of various methodological approaches. The overview chapter (Chapter 30) discusses the impact of the methodological shift that the discipline is undergoing now. The remaining chapters explain and exemplify a range of approaches to the cognitive study of language. Chapter 31 gives a useful overview of quantitative methods, while Chapter 32 presents the current view from neuroscience, which stresses the malleability of the neural substrate of language. The approach gives an in-depth justification to the view of language proposed by cognitive linguistics. Chapter 33 discusses the emergence of a relatively new approach – cognitive sociolinguistics. The approach described builds on the interdisciplinary temperament of both cognitive linguistics and sociolinguistics and suggests paths that could lead to further methodological developments. The next two chapters, 34 and 35, describe computational approaches which create new research tools: Framenet, Constructicon, and MetaNet. These approaches not only provide computerized access to resources, but also propose theoretical solutions which enhance the clarity of analyses and understanding of the data. In Chapter 36, the reader will find a survey of corpus-based methods in cognitive linguistics, along with clear explanations of how corpora can be used in cognitive linguistic analyses. Chapter 37 shows a very different approach, where a careful analytical look at how language is used in creative forms of expression (though not necessarily in literature) helps propose research questions and refinements to analyses arrived at through other methods. The focus of the handbook is to show how cognitive linguistics approaches language and what results can be obtained. It also stresses repeatedly that due to the growth in the methodological toolkit cognitive linguists can use, the same problem can in fact be analyzed in more than one way. The results obtained through different methods, whether fully compatible or not, can give a fuller view of a language phenomenon. But at the same time, this suggests that it might be necessary to look at a phenomenon through various methodological lenses to see its full
9
10
DANCYGIER
nature. Part VI shows what that ‘disciplinary interdisciplinarity’ yields and what we can expect to find out when different methods are used. The concepts focused on here are space and time – two staples of cognitive linguistic work. Chapter 38 discusses a range of data documenting the claim generally accepted by cognitive linguists, that expressions of time are often metaphorical, relying on a range of models of space. Then Chapter 39 expands the view of space/time relationships, by looking at the cultural context, gesture, cross-linguistic studies, and lab evidence. Chapter 40 continues the experimental exploration of time and space, while acknowledging the way in which observations of usage facts may suggest hypotheses regarding conceptualization patterns. The chapter also shows that experimental work is necessary for us to understand the actual mental models used by speakers of various languages. Finally, Chapter 41 shows a new methodology (CODA) which allows one to consider phenomena in the context of actual discourse; the approach broadens the range of questions, and adds the discourse component to the discussion. Jointly, the chapters in Part VI suggest that the methodological variety is necessary for researchers truly to appreciate the meaning and usage potential of a concept or concepts. Analyzing examples and proposing hypotheses is important, it is the necessary first step, but the proposed claims should then be tested in several ways: through crosslinguistic comparisons, a view of gesture and other multimodal forms of expression, lab experiments, and the study of discourse. Relying on one approach alone simply does not suffice. Overall, the handbook shows how cognitive linguistics deals with the challenge of offering reliable analyses of language phenomena. The breadth of its perspective, the methodological richness, the interest in all aspects of language, the cognitive commitment, and the primary attention to usage are factors which make cognitive linguistics an extremely appealing approach. It is varied and multifaceted, as is language. It is flexible and always open to new methodological solutions. It faces up to challenges, rather than skirts around them. It is interdisciplinary, and it seeks new engagements. It will keep on asking the right questions, and proposing better answers.
Part I
Language in Cognition and Culture
1 Opening Commentary: Language in Cognition and Culture N. J. Enfield Linguists want to know: What is language like? Why is it like that? Why do only humans have it? When cognitive linguists ask these questions, two core commitments are implied. The first is that our answers to these questions should not only appeal to human cognitive capacities, but also strive to account for language in terms of more general cognition before they posit language-dedicated cognitive capacities. The second is that our answers should both explain and appeal to facts of language as it occurs in usage, as captured by the adages of a usage-based approach: Grammar is Meaning, Meaning is Use, Structure Emerges from Use. The key to understanding language through a cognitive linguistics lens is to see that these two commitments are intimately related.
1.1 More General Cognition Langacker (1987: 13) uses the phrase “more general cognition” in contrast to the kinds of cognition implied by language-dedicated faculties or modules that nativist accounts of language propose (Chomsky 1965: 25; cf. Hauser, Chomsky, and Fitch 2002, Chomsky 2011). A parsimonious account of language would be in terms of cognitive abilities that humans are known to possess for reasons independent from language. For example, there is “the ability to compare two events and register a discrepancy between them” (Langacker 1987: 6). These are aspects of our general intelligence for interpreting and reasoning about physical domains like space, quantities, and causality. Are such abilities necessary for language? Are they sufficient? Our quest to answer these questions must be guided by the knowledge that while other species may have some of what is necessary for language, they do not have what is sufficient. Cognitive linguistics research has explored ways in which aspects of more general cognition can support the learning and processing of
14
ENFIELD
language (Lakoff 1987, Croft 2001, Tomasello 2003, Goldberg 2006, inter alia). This has led to the discovery of principles of conceptual structure which provide people with ways to represent or construe the things, events, and states we wish to talk about. These principles also allow people to productively elaborate those representations in creative and expressive ways. They provide generative resources for setting up conceptual correspondences, typically between target ideas (concepts to be communicated) and source ideas (concepts used as means for communicating). The principles are amply illustrated in this handbook, and many references herein. They include analogy, metaphor, metonymy, gestalt thinking, image schemas, conceptual blends, idealized cognitive models, and more. We can note two key properties of the aspects of more general cognition that have been most widely relied on in cognitive linguistic research. First, they are primarily relational. Second, they are primarily non-social.
1.1.1 Primarily Relational What does it mean to say that the elements of more general cognition relied on in cognitive linguistics are primarily relational? It means that they provide ways of describing relations between concepts, whether the scope of conceptual relation is an isolated linguistic expression or an entire semiotic system. This makes these aspects of more general cognition especially suitable for capturing conceptual relations within an atemporal/synchronic frame. There are of course other frames, dynamic temporal-causal frames including not only the diachronic frame, but also the microgenetic, ontogenetic, phylogenetic, and enchronic frames (Enfield 2014: 9–19; cf. Bybee 2010, see below). Research approaches related to cognitive linguistics, such as psycholinguistics, tend to work within dynamic frames; for example, focusing on language production or comprehension (in a microgenetic frame) or language learning (in an ontogenetic frame). Cognitive linguistics is increasingly well linked to fields like psycholinguistics thanks to the efforts of interdisciplinaryminded researchers in both psychology (e.g. Tomasello 2003) and linguistics (e.g. Goldberg 2016; see also Hurford 2007, 2011). These collaborations are promising to extend the boundaries of what we understand “more general cognition” to mean. And there are further aspects of more general cognition that have important connections to language, but are yet to be explored within the purview of cognitive linguistics proper. A particularly promising area is bounded rationality, the toolkit of fast and frugal heuristics that balances simplicity and economy with functional efficacy (Gigerenzer, Hertwig, and Pachur 2011). Cognitive scientists have begun to explore ways in which this aspect of more general cognition sheds new light on the pragmatics of language
Language in Cognition and Culture
(Barr and Keysar 2004). Such work suggests that cognitive linguistics will enrich its account of imagistic thinking by looking at it in the light of heuristic thinking.
1.1.2 Primarily Non-social What does it mean to say that the aspects of more general cognition widely studied in cognitive linguistics are primarily non-social? It means that they focus more on how we interpret, conceive, and reason about physical phenomena such as space, quantities, and causality, than on interpersonal phenomena in the social domain. Our species is the only one with language. What makes this possible? A challenge for cognitive linguistics, given its emphasis on more general cognition, is that so much of our general cognition is shared with other species. Why do they not have language too? To answer this, we must pinpoint what it is about our specific forms and combinations of more general cognition that other species lack. To be clear: Proposing that the cognition involved for language is unique to our species does not entail that this cognition is specifically linguistic. No other species should be capable of the same kind, or perhaps degree, of cognitive capacity in the relevant forms of thinking. Is this because we have unique capacities for analogy, imagery, metaphor, metonymy, and pragmatic inference, among other principles? There is good reason to think that what really makes language possible is our social cognition (Enfield and Levinson 2006a). A recent comparative study of cognition in the great apes argues that general intelligence – as measured using tests in physical domains of space, quantities, and causality – does not greatly distinguish humans from our closest relatives such as chimpanzees. “Supporting the cultural intelligence hypothesis and contradicting the hypothesis that humans simply have more ‘general intelligence,’ we found that the children and chimpanzees had very similar cognitive skills for dealing with the physical world but that the children had more sophisticated cognitive skills than either of the ape species for dealing with the social world” (Herrmann et al. 2007: 1360). The conclusion? Sociocultural cognition makes the difference for language. Humans are especially attuned to other minds, and to the cultural construction of group-specific, conventional systems of meaning and practice as shared frameworks for communication and joint action. This is what makes it possible for human populations to foster the historical development of complex systems of shared cultural tradition, of which language is one form. This does not detract from the demonstrated importance for language of non-social aspects of more general cognition, including those that are clearly shared by other species. Hurford (2007, 2011) has argued that the core principle of predicate-argument organization in the syntactic
15
16
ENFIELD
organization of language – any human language – is based on properties of brain function and anatomy that are shared with many other species. Many species display these same basic properties of neural organization and cognitive processing (specifically, the integration of a ‘where’ system with a ‘what’ system, supplying the essential ingredients of argumentpredicate relations). So why do other species not have language? Hurford’s answer is that they do, it is just that they don’t make it public. This in turn means that conceptual structure does not enter the public domain, and so it is impossible for cultural processes of symbolic community-wide conventionalization, and subsequent grammaticalization, to get started (Bybee 2010, Hurford 2011, Enfield 2015).
1.2 Language Usage As one of a set of functionalist approaches to language, cognitive linguistics does not just analyze linguistic structure, “it also analyzes the entire communicative situation: the purpose of the speech event, its participants, its discourse context”; it maintains “that the communicative situation motivates, constrains, explains, or otherwise determines grammatical structure” (Nichols 1984: 97). This orientation is well grounded in insights dating back to Wittgenstein (1953), Zipf (1949), and beyond. Embracing the idea that language is a tool for thought and action, cognitive linguistics is usage-based (Barlow and Kemmer 2000). In this way a strong focus on conceptual representation is increasingly often complemented by close attention to the dynamic, causal, utilitarian underpinnings of language and its structure. In a usage-based model, “the process of language use influences the structure of the representation” (Croft and Cruse 2004: 326–27). Taken together, the three key concepts invoked here – use, influence, and structure – imply a causal conception of language. It is not enough to describe a piece of language structure, a linguistic (sub)system, or a pattern of variance in language. We must ask why it is that way. One way to answer this is to find what has shaped it. “Everything is the way it is because it got that way,” as biologist D’Arcy Thompson is supposed to have said (cf. Thompson 1917). Bybee echoes the sentiment in relation to language: “a theory of language could reasonably be focused on the dynamic processes that create languages and give them both their structure and their variance.” Seen this way, linguistic structure is “emergent from the repeated application of underlying processes” (Bybee 2010: 1). The aim is to explain structure by asking how structure is created through use. The goal of unpacking the key concepts of use, influence, and structure – and the relations between these – points to new horizons in cognitive linguistics. Many of these horizons are explored in this handbook, as the reader will find. Here I will only remark that if we are going to map those
Language in Cognition and Culture
horizons systematically and with clarity, a central conceptual task is to define the temporal-causal frames within which we articulate our usagebased accounts (see Enfield 2014: 9–21). Some of those frames are well established: In a microgenetic frame, sub-second dynamics of psychological processing, including heuristics of economy and efficiency, provide biases in the emergence of structure in utterances; in a diachronic frame, population-level dynamics of variation and social diffusion provide biases in a community’s conventionalization of structure; and in an ontogenetic frame, principles of learning, whether social, statistical, or otherwise, provide biases in the individual’s construction of a repertoire of linguistic competence in the lifespan. Then there is the phylogenetic frame, through which our evolved capacities provide the defining affordances for our species’ capacity for language. If there is less charted territory, it is in the enchronic frame, the moveby-move flow of interlocking, action-driven, forward-going sequences of linguistic action and response in social interaction (Schegloff 1968, Clark 1996, Schegloff 2007). By orienting to the enchronic frame, recent work in descriptive linguistics has begun to analyze linguistic structures in terms of their distribution in relation not only to morphosyntactic units, or units of discourse, but also to structural units that can only be observed and defined in data taken from dialogue (Enfield 2013; cf. also Du Bois 2014). Gipper (2011) sheds new light on the analysis of multifunctionality in evidential marking by comparing the functions of Yurakaré evidentials in differently positioned utterances in conversation; she finds that evidentials can have quite distinct functions depending on whether they occur in initiating utterances (e.g. questions, new assertions) versus responsive utterances (e.g. answers to questions, expressions of agreement). A different kind of outcome from orienting to the enchronic frame in research on language and cognition is that it requires us to confront and explain phenomena that are clearly linguistic but that have hardly been on the map in any form of linguistics until now; key examples include repair (Schegloff et al. 1977, Hayashi, Raymond, and Sidnell 2013) and turntaking (Sacks et al. 1974, Roberts, Torreira, and Levinson 2015), both of which have significant implications for our understandings of the language-cognition relationship (Dingemanse et al. 2015, Levinson and Torreira 2015). Without the usage-based approach that cognitive linguistics advocates, these implications would remain out of sight.
1.3 Conclusion The challenge now is to further enrich our understanding of the causal influence of use on structure in language, and thus to see better how it is that only human cognition supports language. A first move is to broaden the scope of the key ideas – influence, use, and structure – with a concerted
17
18
ENFIELD
and systematic approach to discovering how the multiple causal-temporal frames of language use operate, both in themselves and in relation to each other. Like the rest of cognitive linguistics, this work is as much about culture as it is about language, for language is not only a form of culture – being a local and historical cumulation of social practice – but it is our main tool for constructing culture itself.
2 Relationships between Language and Cognition Daniel Casasanto
2.1 Language, Cognition, and the Goals of Cognitive Linguistics Understanding relationships between language and thought is fundamental to cognitive linguistics. Since the 1970s, cognitive linguists have sought to build cognitively informed theories of language. Such theories are exemplified by Ronald Langacker’s (2008a) Cognitive Grammar or Charles Fillmore’s (1988) Construction Grammar. With the rise of Conceptual Metaphor Theory (Lakoff and Johnson 1980a), however, a second, complementary goal began to drive research in cognitive linguistics: the goal of building linguistically informed theories of cognition. According to George Lakoff, “the locus of metaphor is not in language at all, but in the way we conceptualize one mental domain in terms of another” (1993: 203). Analyzing metaphorical language can be an end in itself, but it is a means to an end with respect to the central focus of metaphor theory and related research programs (e.g. Blending Theory; Fauconnier and Turner 2002) that seek to understand “the nature of the human conceptual system” (Lakoff and Johnson 1980b: 195). These two goals are synergistic, and ultimately one cannot be achieved independently of the other. Yet, distinguishing them is crucial for defining research questions and strategies effectively. When we build cognitively informed theories of language, discoveries about the brain and mind serve as sources of hypotheses and potential constraints, but the data that are critical for evaluating these theories are, for the most part, linguistic data (e.g. descriptive analyses, corpus analyses, studies of language production or comprehension). By contrast, in order to avoid circular reasoning (Pullum 1991), when we build linguistically informed theories of cognition, discoveries about language serve as sources of hypotheses and potential constraints, but the data that are critical for evaluating these theories are, for the most part, nonlinguistic data (e.g. studies of how people think, perceive, act, feel, decide, remember, or imagine).
20
C AS A S A N TO
One thing that unites work toward both of these goals is the belief that “language offers a window into cognitive function, providing insights into the nature, structure and organisation of thoughts” (Evans and Green 2006: 5). This belief is supported by innumerable points of data: often, the way people talk reflects the way they think. Yet, importantly, this is only one of the possible relationships between language and thought. For cognitive linguistics to continue to grow, the idea that language reflects thinking must change from a dogma to a hypothesis: a hypothesis that has been validated in many cases, but starkly disconfirmed in others. This chapter will demonstrate five different ways in which language and nonlinguistic mental representations can be related, using metaphorical language and thought as a testbed. These different relationships arise from patterns of interaction that people have with their linguistic, cultural, and physical environments: 1. Linguistic metaphors can reflect mental metaphors (i.e. nonlinguistic metaphorical mappings). 2. Linguistic metaphors can determine which mental metaphors people use. 3. Linguistic metaphors can create new mental metaphors. 4. People can think in mental metaphors that do not correspond to any linguistic metaphors. 5. People can think in mental metaphors that directly contradict their linguistic metaphors. I will illustrate each of these relationships via spatial metaphors in language and/or the mind. First, however, I will give a brief introduction to Conceptual Metaphor Theory for the uninitiated reader, and then make an argument for some much-needed theoretical and terminological shifts, for the initiated reader.
2.2 From conceptual metaphor to mental metaphor According to the main claim of Conceptual Metaphor Theory (Lakoff and Johnson 1980a), people often conceptualize one domain of knowledge or experience (the Target domain) using mental representations from another domain as a scaffold (the Source domain). The Target domain is typically abstract: something that we can experience, but can never see or touch (e.g. Time, Happiness, Value). The Source domain is typically (though not always) something relatively concrete that we can experience directly through the senses (e.g. Space, Force, Motion). Grafting a Target domain onto a Source domain may make the abstract domain effectively more concrete, and therefore easier to think about. Constructing abstract domains metaphorically may also allow people to ‘recycle’ neural resources: taking structures that evolved to support perception and action
Relationships between Language and Cognition
in the physical word and reusing them to support more abstract cognitive capacities that may be uniquely human (see Pinker 1997, Casasanto 2010, Merritt, Casasanto, and Brannon 2010). Linguistic metaphors are hypothesized to signal nonlinguistic Source–Target mappings. Prices can be high or low; romances can be hot or cool; arguments can be weak or forceful. In each case, abstract entities (prices, romances, arguments) are described in terms of a concrete Source domain that can be apprehended directly through the senses (height, temperature, force). The Source–Target relationship is asymmetric: People do not typically use expressions like that skyscraper is pricey to describe a building’s height, or say that today’s weather will be passionate to indicate that the temperature will be warm. This asymmetry in the linguistic roles that words from Source and Target domains play is taken as an indication of an asymmetric dependence between nonlinguistic representations in these domains (for discussion see Gijssels and Casasanto this volume Ch. 40). In addition to being asymmetric, Source–Target mappings in language are systematic: In many cases a continuum of values in the Target domain is mapped onto a continuum of values in the Source domain; this systematic mapping gives metaphors inferential power. The inferences that can be made about the relative height of any two points in space get imported into the Target domain, and support inferences about the relative height of any two prices, etc. Finally, Source–Target mappings in language are productive. New instances of a given Source–Target mapping can be created on the fly, and easily understood. Even if you had never heard the expression their romance was boiling, you could understand this phrase as an instance of the p a s s i o n i s h e a t mapping. Such asymmetric, systematic, productive mappings between nonlinguistic representations in Source and Target domains are what Lakoff and Johnson (1980a) named a conceptual metaphor. The term conceptual metaphor, however, is often used ambiguously: It can refer to metaphorical expressions in language, or to nonlinguistic Source–Target mappings, or to both linguistic and nonlinguistic mappings (or to the link between them).1 This widespread ambiguity has been tolerated, and perhaps unnoticed, because of the dogma that language reflects thinking. A standard assumption is that when people are using linguistic metaphors they are also using the corresponding nonlinguistic Source–Target mappings. Yet, in light of data like those I will review below, it is not possible to assume that metaphorical language reflects metaphorical thinking. This assumption is an impediment to a more complete understanding of relationships between language and cognition. 1
A standard phrasing of Conceptual Metaphor Theory’s main claim contributes to confusion about whether metaphors under discussion are in language or thought. When we say that people ‘conceptualize one domain in terms of another’ we are using a THINKING IS TALKING metaphor to characterize the relationship between nonlinguistic domains.
21
22
C AS A S A N TO
To disambiguate between metaphors in language and in thought, I will use the term linguistic metaphor to refer to metaphoric expressions in language and mental metaphor to refer to mappings between nonlinguistic representations in Source and Target domains (see Casasanto 2009, 2013 for discussion). This distinction becomes crucial here as we explore multiple dissociations between linguistic metaphors and mental metaphors.
2.3 Linguistic Metaphors That Reflect Mental Metaphors People use space to talk about abstract concepts (cf. Whorf 1956). Linear spatial schemas provide particularly productive Source domains. In linguistic metaphors, Time flows along a sagittal (front–back) axis; Political Views are organized along a lateral (left–right) axis; Number and Emotional Valence (i.e. positive versus negative emotions) are among the domains that are organized according to a vertical (up–down) axis. Beyond talking about these abstract Target domains in linear spatial metaphors, do people think about them metaphorically? That is, when people think about these Target domains, do they implicitly activate the particular Source domain representations that conventional metaphors in their language suggest they should? For these Source–Target pairs and several others, experimental evidence suggests the answer is often ‘Yes.’
2.3.1 Sagittal Mappings of Time in Language and Thought In many languages, the future is ahead and the past is behind: do people think about time this way? Some seminal experimental studies suggested that they do, but the inferences they supported were limited, either because the stimuli and responses were linguistic, or because the sagittal mappings were represented laterally on a two-dimensional page or computer screen. Subsequent studies have overcome these limitations. In one study participants were assigned to think about the past or the future, while wearing motion trackers. Participants assigned to think about the past tended to lean backwards, whereas participants assigned to think about the future tended to lean forwards (Miles, Nind, and Macrae 2010; see also Boroditsky 2000, Miles et al. 2010, Torralbo, Santiago, and Lupia´n˜ez 2006). These results suggest that, even when people are not using sagittal space–time metaphors in language, they may spontaneously conceptualize the past as ahead and the future as behind them.
2.3.2 Vertical Mappings of Number in Language and Thought Similarly, experiments suggest that people implicitly associate numbers with vertical space, consistent with linguistic metaphors designating some numbers as high and others as low. In one study, participants were asked to
Relationships between Language and Cognition
judge numbers as odd or even, responding by moving their eyes to a target location in vertical space. Making parity (odd/even) judgments ensured that the numbers’ cardinal magnitudes and ordinal positions were both irrelevant to the task. Still, the results showed that participants were faster to judge larger numbers’ parity with upward eye movements, and smaller numbers’ parity with downward eye movements (Schwarz and Keus 2004). Another study revealed how automatically these implicit number–space mappings are activated in people’s minds. Participants were asked to generate random numbers between 1 and 30 each time a metronome ticked, while the experimenters tracked their eye movements. Participants tended to make upward eye movements spontaneously before producing a higher number, and downward eye movements before producing a lower number (Loetscher et al. 2010). The magnitude of the eye movements was correlated with the magnitude of the numbers produced (e.g. the largest upward movements preceded the highest numbers), suggesting that a continuum of numbers was mapped systematically to a vertical spatial continuum.
2.3.3 Vertical Mappings of Valence in Language and Thought Finally, in many languages metaphorical expressions link positive and negative emotional valence with the top and bottom of a vertical spatial continuum (Lakoff and Johnson 1980a): A happy person is high on life, but a sad person is down in the dumps. A variety of experiments suggest that such linguistic expressions correspond to nonlinguistic mental metaphors. For example, people are faster to judge the valence of positive words like ‘loyal’ when they appear at the top of a screen, and negative words like ‘cruel’ when they appear at the bottom (Meier and Robinson 2004). When asked to recall the locations on a map where positive and negative incidents occurred (e.g. winning a prize versus having an accident), the locations of positive events tend to be shifted upward in people’s memories, and the locations of negative events shifted downward (Brunye´ et al. 2012). Upward- and downward-directed bodily actions can influence the retrieval of emotional memories. In one experiment, participants were assigned to move marbles either upward or downward, from one cardboard box to another, while retrieving and retelling stories from their past. They recounted more positive memories during upward movements, and more negative memories during downward movements (Casasanto and Dijkstra 2010).
2.3.4 When Language Reflects Thinking Together, these studies (and others like them) indicate that when people think about the Target domains of Time, Number, and Emotional Valence, they tend to activate spatial Source domain representations with a high
23
24
C AS A S A N TO
degree of automaticity, and activating them has measurable consequences in a variety of behavioral tasks. For the present discussion, it is notable that people activate the particular spatial schemas that conventional metaphors in their language suggest they should. Therefore, these results support the hypothesis that language reflects thinking. The next four sections challenge this hypothesis, first by showing how language can do more than simply reflect mental metaphors (sections 2.4–2.5), and then by showing how linguistic metaphors can dissociate from mental metaphors (sections 2.6–2.7).
2.4 Linguistic Metaphors Can Determine Which Pre-existing Mental Metaphors People Use Could the linguistic metaphors people use influence how they think, causing people who use different linguistic metaphors to think in different mental metaphors? In order to support this claim, it would be necessary: (a) to show that different groups of people use different linguistic metaphors for the same Target domain; (b) to show that people who talk differently also think differently, in corresponding ways; (c) to avoid logical circularity by showing that between-group differences can be observed at the level of nonlinguistic mental representations; (d) to show that, beyond a language-thought correlation, language plays a causal role in determining how people think; and (e) to determine whether using different linguistic metaphors leads people to create different mental metaphors, or whether experience with language changes the way people use pre-existing mental metaphors. A set of experiments by Sarah Dolscheid and colleagues on mental metaphors for musical pitch meets all of these requirements.
2.4.1 Different Ways of Talking and Thinking about Pitch In many languages, pitch is metaphorized in terms of vertical space: pitches are high or low. But this is not the only possible spatial metaphor for pitch. In languages like Farsi, Turkish, and Zapotec, high-frequency pitches are thin and low-frequency pitches are thick (Shayan, Ozturk, and Sicoli 2011). Beyond talking about pitch using spatial words, do people think about pitch using spatial representations? Several studies suggest that speakers of height languages like English activate vertical Space-Pitch mappings when judging pitches (e.g. Pratt 1930, Roffler and Butler 1968). In one set of experiments, Dolscheid et al. (2013) investigated whether speakers of height languages and thickness languages tend to use the same mental metaphors for pitch, or whether their mental metaphors are shaped by their experience of using linguistic metaphors.
Relationships between Language and Cognition
Like English, Dutch describes pitches as high (hoog) or low (laag), but in Farsi high pitches are thin (na¯zok) and low pitches are thick (koloft). Dolscheid et al. (2013) tested Dutch and Farsi speakers on a pair of nonlinguistic pitch reproduction tasks. Participants were asked to reproduce the pitches of tones that they heard in the presence of irrelevant spatial information: either lines that varied in their height (height interference task) or their thickness (thickness interference task). Dutch speakers’ pitch estimates were strongly affected by irrelevant spatial height information. On average, a given tone was sung back higher when it had been accompanied by a line that was high on the computer screen, and lower when it had been accompanied by a line that appeared low on the screen. By contrast, lines of various thicknesses had no measurable effect on Dutch participants’ pitch estimates. Farsi speakers showed the opposite pattern of results. Lines of varying heights had no measurable effect on Farsi speakers’ pitch estimates, but tones accompanied by thin lines were sung back higher, and tones accompanied by thick lines were sung back lower.
2.4.2
Do Dutch and Farsi Speakers’ Nonlinguistic Representations Differ? This pattern of spatial interference on people’s pitch reproduction performance reflected the Space–Pitch metaphors in their native languages. This pattern cannot be explained by differences in overall accuracy of pitch reproduction, or in differences in musical training between groups. Importantly, this pattern also cannot be explained by participants using language covertly, during the task: labeling the pitches they needed to reproduce as high/low or thick/thin. This explanation was ruled out by the experimental design, in which there was no correlation between space and pitch in the stimuli. As such, covertly labeling high pitches as high (or thin) and labeling low pitches as low (or thick) could not produce the observed effects of space on pitch reproduction; on the contrary, labeling pitches with the spatial metaphors in one’s native language during the task could only work against the observed effects. Rather than an effect of using language online during the task, these experiments show an effect of people’s previous experience using either one linguistic metaphor or the other, and thereby strengthening either the nonlinguistic Height–Pitch or Thickness–Pitch mapping in memory. To confirm that the observed effects did not depend on participants covertly labeling pitches during the task, Dolscheid et al. (2013) repeated the height interference task in Dutch speakers with the addition of a verbal suppression task. On each trial of the task, participants had to rehearse a novel string of digits while perceiving and reproducing the pitches. As predicted, verbal suppression had no effect on the results of the pitch reproduction task. Dutch speakers still showed strong
25
26
C AS A S A N TO
height–pitch interference, consistent with an offline effect of participants’ previous linguistic experience on their subsequent nonlinguistic pitch representations (see also Casasanto 2008a).
2.4.3
Does Language Play a Causal Role in Shaping Mental Metaphors? The results reviewed so far show a correlation between people’s linguistic metaphors and their nonlinguistic mental metaphors, but they do not provide any evidence that language causes Dutch and Farsi speakers to mentally represent pitch differently. Dolscheid et al. (2013) reasoned that if using Thickness–Pitch metaphors in language is what causes Farsi speakers to activate Thickness–Pitch mappings implicitly when reproducing pitches, then giving Dutch speakers experience using similar Thickness–Pitch metaphors in language should cause them to reproduce pitches like Farsi speakers. A new sample of Dutch speakers were assigned to one of two training conditions: Participants in the thickness training group learned to describe pitches using Farsi-like metaphors (e.g. a tuba sounds thicker than a flute), but participants in the height training group (i.e. the control group) described pitches using standard Dutch metaphors (e.g. a tuba sounds lower than a flute). After about twenty minutes of this linguistic training, participants in both groups performed the nonlinguistic thickness interference task described above. Whereas height trained participants showed no effect of irrelevant thickness information on their pitch estimates, thickness trained participants showed a thickness interference effect that was statistically indistinguishable from the effect found in native Farsi speakers. Even a brief (but concentrated) dose of thickness metaphors in language was sufficient to influence Dutch speakers’ mental metaphors, demonstrating that linguistic experience can give rise to the differences in nonlinguistic pitch representations that were found across natural language groups.
2.4.4
What Role Does Language Play in Shaping Space–Pitch Metaphors? Do Space–Pitch metaphors in language cause people to develop the corresponding nonlinguistic Space–Pitch mappings, or does using linguistic metaphors change how likely people are to use a pre-existing mental metaphor? To evaluate these possibilities, Dolscheid et al. (2014) tested four-month-old infants on a pair of Space–Pitch congruity tasks. Infants heard pitches alternately rising and falling while they saw a ball rising and falling on a screen (height congruity task) or a cylinder growing thicker and thinner (thickness congruity task). For half of the trials, changes in pitch and space were congruent with the Height–Pitch and Thickness–Pitch mappings encoded in Dutch and Farsi, respectively, and for the other half of the trials they were incongruent with these
Relationships between Language and Cognition
Space–Pitch mappings. The data showed that infants looked longer at congruent Space–Pitch displays than at incongruent displays. This was true both in the height congruity condition and in the thickness congruity condition. There was no difference in the magnitude of the congruity effect between conditions, suggesting that there was no difference in the strength of the Height–Pitch and Thickness–Pitch mappings in the infants’ minds. Four-month-olds are completely unable to produce Space–Pitch metaphors in language and are also, presumably, unable to understand them. Yet, they already exhibit two of the mental metaphors for pitch found in adults, suggesting that language is not responsible for creating these mappings – only for determining which mapping people tend to use.
2.4.5 Hierarchical Construction of Mental Metaphors How could infants who are sensitive to both Height–Pitch and Thickness–Pitch mappings turn into adults who appear to activate only one of these mappings when they represent pitch? This process can be understood in terms of hierarchical mental metaphors theory (HMMT; Casasanto and Bottini 2014). According to this proposal, the implicit, nonlinguistic mental metaphors that people tend to use most often are specific members of a more general family of mental metaphors. The development of language-specific mental metaphors occurs over two stages. First, a superordinate family of mappings is established, which in the case of Space and Pitch includes both the Height–Pitch and Thickness–Pitch mappings. These mappings may be constructed, over either ontogenetic or phylogenetic time, on the basis of observable correlations between space and pitch in the natural world. The Height–Pitch mapping reflects the fact that people involuntarily raise their larynxes, chins, and sometimes other body parts (e.g. their eyebrows) when they produce higher pitches, and lower them when they produce lower pitches. It also reflects a statistical tendency for higher pitches to originate from higher locations, and lower pitches from lower locations (Parise, Knorre, and Ernst 2014). The Thickness–Pitch mapping reflects a pervasive correlation between pitches and the size of the objects or creatures that produce them. Consider the different pitches produced by: strumming thin versus thick strings on a guitar; banging on a large steel oil drum versus a small steel can; barking by a small dog versus a big dog; etc. Although Dolscheid et al.’s (2014) data confirm that both the Height–Pitch and Thickness–Pitch mappings are present in infants’ minds, they leave open the question of exactly how and when these mappings become established initially. Whatever the ultimate origin of Space–Pitch mappings in prelinguistic children may be, when children learn metaphors in language, a second process begins. Dolscheid et al.’s (2013) findings in adults
27
28
C AS A S A N TO
suggest that each time people use a linguistic metaphor like a high pitch they activate the corresponding mental metaphor, strengthening this mapping at the expense of competing mappings in the same family of Space–Pitch associations. As a consequence, speakers of height languages like Dutch and English come to rely on vertical spatial schemas to scaffold their pitch representations more strongly than multidimensional spatial schemas, whereas speakers of thickness languages like Farsi come to rely on multidimensional spatial schemas more strongly than vertical spatial schemas. According to HMMT, strengthening certain mental metaphors results in the weakening of other members of the family of mappings (consistent with the dynamics of many memory networks) – but this does not cause these dispreferred mappings to be extinguished. This aspect of the theory may explain the representational flexibility demonstrated by Dolscheid et al.’s (2013) training experiment. Dutch speakers could be induced to use a nonlinguistic Thickness–Pitch mapping after only a brief training intervention because no spatial mappings had to be created or destroyed; rather, the new pattern of language experience boosted the strength of the Thickness–Pitch mapping that had presumably been present in the Dutch speakers’ minds since infancy, causing participants to think about pitch in a way that was not new, just rarely used. HMMT provides an account of how mental metaphors can be fundamental to our understanding of Target-domain concepts, and yet surprisingly context-dependent.
2.5 Linguistic Metaphors Can Create New Mental Metaphors Beyond selecting among pre-existing mental metaphors, can learning conventional metaphors in language cause people to create new mental metaphors? To date, one clear instance of a language-based spatial metaphor has been described: the left–right mapping of Political Views. In the late eighteenth century, the conservative members of the French Legislative Assembly sat on the right side of the room, and the liberal members on the left (Oppenheimer and Trail 2010). This arrangement had enduring consequences. More than two centuries later, liberal and conservative values are metaphorized on a left–right continuum, across many languages and cultures, as evidenced by English expressions like the liberal left, centrist politics, and right-wing conservatives. In order to establish that a mental metaphor arises from experience using linguistic metaphors, it is necessary to show that: (a) beyond talking about the Target domain metaphorically, people use the corresponding Source domain representations to think about the Target domain; (b) other plausible origins of the mental metaphor have been
Relationships between Language and Cognition
ruled out; (c) beyond a language-thought correlation, metaphorical language plays a causal role in establishing the mental metaphor. A series of studies described by Casasanto (2013) satisfies the first two of these three requirements for the l i b e r a l i s l e f t / c o n s e r v a t i v e i s r i g h t mapping.
2.5.1 Do People Think about Politics Spatially? Left–right linguistic metaphors for politics appear to correspond to active mental metaphors. In one experiment, US students were asked to sit in an ostensibly broken office chair while completing a political attitudes survey. Unbeknownst to participants, a wheel had been removed strategically from one side or the other, causing the chair to lean leftward or rightward. Responses showed that, on average, participants who had been assigned to sit in the left-leaning chair expressed more agreement with Democrats (traditionally the more liberal party), whereas participants assigned to sit in the right-leaning chair tended to agree more strongly with Republicans (Oppenheimer and Trail 2010). The automaticity with which people activate an implicit left–right mapping of politics was confirmed in a series of reaction time studies in Dutch participants. Although The Netherlands has many political parties, which differ along multiple dimensions, the parties’ liberality or conservativism is often described using left–right metaphors (Bienfait and van Beek 2001). Accordingly, when presented with parties’ acronyms, Dutch participants were faster to make judgments about more liberal parties with their left hand (or when the acronym appeared on the left of the screen), and faster to make judgments about more conservative parties with their right hand (or when the acronym appeared on the right; van Elk, van Schie, and Bekkering 2010).
2.5.2 Ruling Out Nonlinguistic Origins of the Mental Metaphor Where does this mental metaphor come from? Pointing to its historical roots does not answer the question of how this mental metaphor arises in individuals’ minds: That is, the arrangement of eighteenth-century French politicians does not explain how individuals come to intuit a mapping between politics and space today. One potential origin of this mental metaphor is correlations in linguistic experience. Using the words right and left in both literal contexts (e.g. the can is on the left of the shelf) and metaphorical contexts (e.g. the candidate is on the left of the political spectrum) could cause people to implicitly associate politics with left–right space, building upon our general propensity to spatialize abstract concepts. Before accepting the conclusion that linguistic experience instills a nonlinguistic Space–Politics mapping in individuals’ minds, it is important to consider other possibilities. Specifically it is important to
29
30
C AS A S A N TO
rule out the possibilities that: (1) this mental metaphor could be innate; (2) it could arise from embodied experience; (3) it could arise from nonlinguistic cultural experiences. First, in principle, the mapping could be innate (and this implicit mental metaphor could have biased the arrangement of the French Legislative Assembly). In practice, this proposal is dubious. Mental metaphors like the mapping of Temporal Duration onto Spatial Distance (see Casasanto and Boroditsky 2008, Gijssels and Casasanto this volume Ch. 40) are good candidates for an innate origin, since space and time are ubiquitous in our physical experience, and the need to compute Space–Time relationships has been important throughout our evolutionary history. It is unlikely, however, that liberal and conservative political ideologies, or even the concepts of left and right (which are absent from some modern languages and cultures; Levinson and Brown 1994), arose early enough in human history to have been encoded in our genes and neurally hardwired. Second, there is no basis for the left–right mapping of politics in our physical interactions with the natural environment. This metaphor appears to function much like other orientational metaphors, which Lakoff and Johnson (1980a) have suggested arise from “correlations in our embodied experience.” And yet, this mapping in language and in thought cannot be acquired through incidental learning of correlations between Politics and Space in the natural world. It is not credible, for example, that as we encounter other people in our everyday environment we tend see people with liberal views on our left and people with conservative views on our right – with such regularity that politics becomes implicitly mapped onto left–right space. Third, the left–right mapping of politics in individuals’ minds could arise from the spatialization of politics in nonlinguistic cultural conventions (e.g. graphic representations of politics and politicians in the media). This suggestion, which deserves consideration, assumes that liberal and conservative politicians or symbols are, in fact, commonly represented on the left and right, respectively. But this does not appear to be the case. For example, videos of televised debates between US presidential candidates for the most recent election years (2008, 2012) show no systematic link between the candidates’ parties and their left–right positions on the stage. The US Senate and Congress are arranged similarly to the eighteenthcentury French Legislative Assembly, but most people are not members of the Senate or Congress, and are therefore unlikely to be exposed to this spatialization of the political left and right with sufficient frequency to give rise to an implicit mental metaphor. When the Senate and Congress are viewed on TV, the camera’s viewpoint varies between photographs and videos, sometimes showing the Democrats and Republicans on metaphorcongruent and sometimes on metaphor-incongruent sides of viewercentered space.
Relationships between Language and Cognition
The spatial positions of people may be subject to numerous functional constraints, but the spatialization of symbols in graphics is largely at the discretion of the artist. In the United States, donkeys and elephants symbolize the Democrat and Republican parties, respectively. Often these animals are depicted side by side, presumably to indicate opposition between the parties, or to represent the voters’ two main alternatives. Is the donkey usually depicted on the left and the elephant on the right? To address this question, one study used Google Images (www.google.com) to a survey donkey–elephant images (Casasanto 2013). Of the images sampled, 51 percent were metaphor-congruent (donkey on the left, elephant on the right) and 49 percent were metaphor-incongruent (elephant on the left, donkey on the right); this difference did not approach statistical significance. This survey provided no evidence that the symbols of the liberal and conservative parties are typically presented on the left and right sides, respectively. This may be because right and left depend on perspective. In many images, the donkey and elephant are facing out of the page, toward the viewer. Artists who wished to make the animals’ locations congruent with the l i b e r a l i s l e f t metaphor might sometimes place the donkey on the viewer’s left, and other times on the elephant’s left (i.e. the viewer’s right), leading to the appearance of randomness.
2.5.3 Linguistic Metaphors Can Create Mental Metaphors Whatever the reason for the apparent lack of any systematic use of left–right space in the placement of politicians before the TV camera, or in political graphics, the implications for the present question are clear. If politicians or their symbols are not systematically spatialized in nonlinguistic cultural conventions, these conventions cannot be responsible for establishing the spatial mapping of politics in people’s minds. It appears that talking about liberal and conservative political attitudes in terms of left–right space is what causes people to think about them that way. This conclusion awaits further experimental validation: Only an experimental intervention, like Dolscheid et al’s (2013) training study (section 2.4.3), can establish a causal relationship between language and thought. But since the origins of this mental metaphor do not appear to lie in innate hardwiring, correlations in the natural world, or correlations in cultural conventions, linguistic metaphors seem to be the only plausible source of the spatial metaphors for politics in people’s minds. If so, then beyond just reflecting or modifying our mental metaphors, linguistic metaphors can also create new ways of thinking about abstract concepts.
31
32
C AS A S A N TO
2.6 Mental Metaphors That Do Not Correspond to Any Linguistic Metaphors The preceding sections support and extend the thesis that the way people talk reflects (and sometimes shapes) the way they think. But this is not always the case. To find dissociations between metaphorical language and thought, we need look no farther than two conceptual domains already discussed above: Time and Number.
2.6.1 Lateral Mappings of Time Linguistic metaphors for Time in English indicate that events unfold along the sagittal (front–back) axis, not the lateral (left–right) axis: The future is ahead, not to the right. Lateral Space–Time metaphors are not conventionalized in any known spoken language. And yet, there is abundant evidence that people conceptualize sequences of events according to a left–right spatial mapping. Evidence of a lateral mental timeline comes from dozens of experiments using diagram tasks (Tversky, Kugelmass, and Winter 1991), reaction time tasks (e.g. Torralbo, Santiago, and Lupia´n˜ez 2006, Weger and Pratt 2008), and studies of spontaneous gesture (Cienki 1998, Cooperrider and Nun˜ez 2009, Casasanto and Jasmin 2012). People exposed to languages that are written left to right tend to conceptualize earlier events on the left and later events on the right; people exposed to languages with right-to-left orthographies like Arabic and Hebrew show the opposite tendency, conceptualizing earlier events on the right and later events on the left (Tversky et al. 1991, Fuhrman and Boroditsky 2010, Ouellet et al. 2010). An experiment in which left-to-right readers were exposed to mirror-reversed text confirmed that experience reading in one direction or another is sufficient to determine the direction of the lateral mental timeline (even though participants read the same words and phrases in both orthography conditions, ruling out influences of language, per se; Casasanto and Bottini 2014). In a study of spontaneous gestures, English speakers produced far more gestures that were congruent with the e a r l i e r i s l e f t / l at e r i s r i g h t mapping than with the f u t u r e i s a h e ad / p a s t i s b e h i n d mapping, suggesting that the lateral timeline may be activated more strongly or automatically than the sagittal timeline encoded in language (Casasanto and Jasmin 2012). Participants gestured laterally even when using sagittal language (e.g. gesturing left when saying “farther back [in the past]”; ibid.: 661); they were thinking left even when saying back, underscoring the dissociation between their linguistic metaphors and their mental metaphors for time (see Gijssels and Casasanto this volume Ch. 40 for further discussion).
Relationships between Language and Cognition
2.6.2 Lateral Mappings of Number A similar dissociation between language and thought is observed for spatial conceptions of number. In addition to spatializing numbers vertically, as linguistic metaphors suggest, people also implicitly spatialize them along a lateral mental number line. In Western cultures, smaller numbers are associated with the left side of space and larger numbers with the right, despite the complete absence of corresponding linguistic metaphors in any known language: nine is higher than eight, not righter than eight. The vertical space–number studies reviewed in section 2.3.2 also had horizontal conditions. When asked to judge the parity of numbers (odd versus even) and respond by moving their eyes to a target, participants were faster to indicate larger numbers’ parity with rightward eye movements, and smaller numbers’ parity with leftward eye movements (Schwarz and Keus 2004). When asked to generate random numbers, participants tended to make rightward eye movements spontaneously before producing a higher number, and leftward eye movements before producing a lower number (Loetscher et al. 2010). The most common test of the lateral mental number line uses manual responses. Typically, when Western participants indicate the parity of numbers by pressing buttons with the left or right hands, they are faster to classify smaller numbers with the left hand and larger numbers with the right. This reaction time effect was given a memorable name, the SNARC effect (Spatial-Numerical Association of Response Codes; Dehaene, Bossini, and Giraux 1993), although experiments like Loetscher et al.’s show that this is a misnomer; the effect does not depend on an association between numbers and explicit responses, but rather on a spatial conception of number that can be observed whether or not spatialized responses are required. The SNARC effect, which has been replicated in hundreds of experiments (for review see Wood et al. 2008), is reversed or extinguished in some non-Western cultures, leading many researchers to conclude that, like the mental timeline, the direction of the mental number line is determined by experience with a left-to-right versus right-to-left orthography. However, this conclusion is not well supported by experimental data: the only direct tests of the effect of reading experience on the mental number line have produced null results (Dehaene, Bossini, and Giraux 1993, Pitt and Casasanto 2016). The direction of the mental number line does appear to depend on other culture-specific practices, however, such as the typical arrangement of numbers in graphic displays (Fischer, Mills, and Shaki 2010) and finger-counting routines (Fischer 2008, Pitt and Casasanto 2014). For the present discussion, it is sufficient to note that, according to hundreds of studies, people implicitly organize numbers along a culturespecific mental number line: a mental metaphor that does not correspond to conventional metaphors in any known language.
33
34
C AS A S A N TO
2.7 Mental Metaphors That Contradict Linguistic Metaphors To date two mental metaphors have been documented that directly contradict conventional metaphors and idioms in their users’ languages, in the domains of Time and Emotional Valence.
2.7.1 Conceptualizing the Past as Ahead, Despite Linguistic Metaphors That Say the Future is Ahead Speakers of Darija, a Moroccan dialect of Modern Standard Arabic, use sagittal metaphors for time that are similar to those found in English and many other languages: The past is ahead and the future is behind, in their language. Yet, when their implicit mental metaphors for Time were tested, Darija speakers showed the opposite Space–Time mapping. In two experiments, they showed a strong tendency to conceptualize the past as ahead and the future as behind (de la Fuente et al. 2014). Since no explanation for this reversed Space–Time mapping could be found in participants’ language (cf., Nun˜ez and Sweetser 2006b) or in their bodily interactions with the environment (cf. Clark 1973), de la Fuente and colleagues turned to aspects of culture. They found that, compared to Europeans, Moroccans tend to place more value on ancient rituals and traditional values, focusing on the past, whereas Europeans tend to focus on the future. Further experiments showed that the culture-specific tendency to focus on the past or the future predicted where people tended to place the past and future in their mental models of Time – irrespective of their linguistic metaphors. When induced to think about the past, Spaniards (who normally conceptualize the future as ahead) were more likely to place the past ahead in their mental models, contra the conventional metaphors in their language. Together, de la Fuente et al.’s (2014) findings suggest that, at any moment, people may be spatializing the past and future in their mental metaphors exactly the opposite of the way it is spatialized in their linguistic metaphors (see Gijssels and Casasanto this volume Ch. 40 for further discussion).
2.7.2 Associating Good with Left, Despite Linguistic Expressions Associating Good with Right Across many languages, good things are associated with the right and bad things with the left (e.g. my right-hand man; the right answer versus sinister intent; two left feet). These patterns in language are reinforced by nonlingusitic cultural conventions. In Western cultures, these customs include shaking hands with the right hand, and raising the right hand to swear an oath. In other cultures, they include prohibitions against pointing or eating with the left hand, which is reserved for dirty jobs (Kita and Essegby 2001).
Relationships between Language and Cognition
Unconsciously, people also conceptualize good and bad in terms of left–right space, but not always in the way linguistic and cultural conventions suggest. Rather, people’s implicit associations between space and valence are body specific: Right-handers associate good with right, but lefthanders associate good with left, implicitly, in spite of conventions in language and culture. When asked to decide which of two products to buy, which of two job applicants to hire, or which of two alien creatures looks more honest, right- and left-handers respond differently. Righthanders tend to prefer the product, person, or creature presented on their right side but left-handers tend to prefer the one on their left (Casasanto 2009). This pattern persists even when people make judgments orally, without using their hands to respond. Children as young as five years old already make evaluations according to handedness and spatial location, judging animals shown on their dominant side to be nicer and smarter than animals on their non-dominant side (Casasanto and Henetz 2012). Reaction time tasks show that the body-specific association between valence and space is activated highly automatically. When judging the valence of words or faces, right- and left-handers are faster to classify stimuli as positive when responding with their dominant hand, and faster to classify them as negative when responding with their non-dominant hand (de la Vega et al. 2012, de la Vega et al. 2013, Kong 2013). Beyond the laboratory, the association of good with the dominant side can be seen in left- and right-handers’ spontaneous speech and gestures. In the final debates of the 2004 and 2008 US presidential elections, positive speech was more strongly associated with right-hand gestures and negative speech with left-hand gestures in the two right-handed candidates (George W. Bush, John Kerry), but the opposite association was found in the two left-handed candidates (John McCain, Barack Obama; Casasanto and Jasmin 2010). In a simulated election, left-handers were fifteen percentage points more likely to vote for a candidate whose name appeared on the left of the ballot, compared to right-handers, suggesting that this implicit mental metaphor may have measurable real-world consequences (Kim, Krosnick, and Casasanto 2015). Three lines of evidence show that this mental metaphor is not shaped by experience with language or culture, but rather by bodily experience – the experience of acting more fluently with the dominant hand, and therefore associating good with the side of space where we act more fluently and bad with the side where we act more clumsily. First, across experiments, the g o o d i s l e f t mapping in left-handers appears to be stronger than the g o o d i s r i g h t mapping in right-handers; the opposite would be expected if there were additive effects of linguistic/cultural and bodily experiences (Casasanto 2009). Second, similar implicit mental metaphors are seen across cultures with different explicit Space–Valence associations. Whereas left is associated with being clumsy or rude in many European cultures, it is associated with being filthy or evil in Morocco,
35
36
C AS A S A N TO
a predominantly Islamic nation. Yet, experiments showed no significant difference in the strength of the Space–Valence mapping across Spaniards and Moroccans (de la Fuente et al. 2015). Finally, changes in bodily experience can change people’s Space–Valence mappings, independent of any change in language or culture. In one experiment, right-handers performed a fine motor task, arranging dominoes on a table, while wearing a cumbersome glove on either their left hand (which preserved their natural right-handedness), or on their right hand (which turned them temporarily into left-handers, in the relevant regard). After about twelve minutes of lopsided motor experience, participants removed the glove and performed a test of Space–Valence associations, which they believed to be unrelated. Participants who had worn the left glove still thought right was good, but participants who had worn the right glove showed the opposite g o o d i s l e f t bias, like natural left-handers (Casasanto and Chrysikou 2011). The rapidity of this change can be understood in terms of hierarchical mental metaphors theory (see Casasanto 2016a, 2016b). Together, these studies reveal a second mental metaphor that stands in direct opposition to the most relevant linguistic and cultural conventions. Left-handers (and impaired right-handers) manifest an implicit g o o d i s l e f t mapping, despite conventions in language and culture indicating that g o o d i s r i g h t .
2.8 Different Relationships between Talking and Thinking: Roles of Language, Culture, and Body According to a foundational assumption in cognitive linguistics, the language we use reflects the way we think. Often this is true, but not always. There are multiple relationships between talking and thinking, which are conditioned by the particulars of our linguistic, cultural, and bodily experiences. The range of these language-thought relationships can be illustrated by the different relationships that have been discovered between linguistic metaphors and mental metaphors. Often our spatial metaphors in language reflect nonlinguistic mental metaphors. For example, people may unconsciously activate sagittal spatial schemas to think about the past and future, and vertical spatial schemas to think about emotional positivity and negativity, or high and low numbers (section 2.3). Beyond reflecting our thoughts, spatial metaphors in language can also shape them: causing people, for example, to activate schematic representations of either height or thickness when conceptualizing musical pitch (section 2.4). Linguistic metaphors can even cause us to create new mental metaphors, as appears to be the case for lateral mappings of politics (section 2.5). Yet, in other cases people think in mental metaphors that do not correspond to any conventional linguistic metaphors, as when we conceptualize earlier and later events
Relationships between Language and Cognition
according to a left–right mental timeline, or lower and higher numbers according to a left–right mental number line. These lateral mappings of time and of number arise from experience with culture-specific activities like reading (a recent cultural overlay on natural language), using calendars and graphs, and counting on our fingers: activities that reinforce particular space–time or space–number associations in our minds (section 2.6). Finally, people can think in mental metaphors that directly contradict their linguistic metaphors; for example, when Moroccans conceptualize the past as in front of them, or when left-handers implicitly map good and bad to the left and right sides of space. The Moroccans’ p a s t i s ah e a d mapping appears to arise from a culture-specific tendency to focus on the past (and therefore to place the past in front of the eyes, where one could focus on it literally, if events in time were objects in space). The body-specific g o o d i s l e f t mapping arises from patterns of fluent and disfluent manual interactions with the physical environment (section 2.7). The discovery of multiple relationships between language and cognition presents both challenges and opportunities for cognitive linguistics. The challenge lies in recognizing that, in many circumstances, analysing patterns in language constitutes one step in the process of elucidating relationships between language and cognition. Specifically, patterns in language provide a rich source of hypotheses about nonlinguistic mental representations. And therein lies the opportunity. Testing these hypotheses requires diverse methods, and encourages collaboration between linguists and researchers in allied fields. Even when predictions from language-based hypotheses are not upheld, testing them may be illuminating. In the studies reviewed here, for example, the hypothesis that people would conceptualize domains like Time, Number, and Valence in the same ways they that talk about them was repeatedly disconfirmed. Yet, more broadly, the hypothesis that people use spatial schemas to conceptualize these non-spatial domains received consistent and convergent support: people think metaphorically, in more ways than linguistic analyses, alone, can uncover. The project of exploring language-based hypotheses about the mind, using a combination of linguistic and nonlinguistic methods, opens the door for cognitive linguistics to have a broad influence on allied fields, and to become increasingly integrated with the other neural and cognitive sciences.
37
3 The Study of Indigenous Languages Sally Rice
3.1 What Has the Study of Indigenous Languages (IL) Done for Cognitive Linguistics (CL)? The decades immediately following the early 1960s saw the rise of independent linguistics departments in major universities across North America, but these departments were usually characterized by a strong generative theoretical orientation in terms of research and training. Whether cognitive linguistic approaches were already incubating in some quarters post-mid century or were not yet manifest, by the late 1960s and early 1970s frustration at the theoretical abstractness, Angloand Euro-centrism, and descriptive indifference of generative grammars was reaching a peak among field linguists, developmentalists, psychologists, and other language empiricists for whom data were sacrosanct. In this context, syntacticians looking at data from non-Indo-European languages found some professional camaraderie and a modicum of theoretical shelter in the nascent and diverse strands of scholarship that eventually jelled under the cognitive linguistics banner in the 1980s. Before that time, however, a number of linguists with scientific interest in non-traditional speech communities (oral languages in which speaker variation was pronounced, with complex and polysynthetic morphosyntaxes, highly conventionalized expressions, and often unfamiliar semantic systems) either developed their own theories of language or they looked outside their own departments and drew inspiration from research coming out of missionary work (e.g. the Summer Institute of Linguistics, or SIL) as well as from psychology, anthropology, sociology, and communication departments to help make sense of data which strained to fit universalist and supposedly context-independent claims. What these linguists brought or brought back to the study of language proper helped set the agenda for the often-amorphous theoretical movement that has come to be called cognitive linguistics (CL).
The Study of Indigenous Languages
Table 3.1 Some of the indigenous or minority languages studied from a CL or CL-compatible perspective by authors who have conducted their own original fieldwork or who published dissertations or multiple studies on these languages* Region
Family
North America i r o q u o i a n
hokan uto-aztecan
algonquian athapaskan
Language
Linguist
Seneca
Wallace Chafe, Marianne Mithun Wallace Chafe Marianne Mithun
Onondaga Mohawk, Cayuga, Tuscarora, Oneida Atsugewi Luiseño Cora Náhuatl Classical Náhuatl (Aztec) East Cree Michif Dene Sųłiné Hupa Koyukon pan-Athapaskan
Mesoamerica
caddoan eskimo-aleut pomoan salishan
Caddo Central Alaskan Yup’ik Central Pomo Coeur d’Alene
wakashan
various Salishan various Wakashan
isolate oto-manguean
chibchan mayan
Tarascan Ayoquesco Zapotec San Marcos Tlapazola Zapotec Chalcotongo Mixtec Copala Trique Rama Jacaltec Yukatec Tzeltal Mopan Maya Zinacanteco Tzotzil
South America q u e c h u m a r a n t u p i - g u a r a n í
Wanca Quechua Guaraní
Africa
Ik various Khoisan Yoruba various Niger-Congo
nilo-saharan khoisan niger-congo
Leonard Talmy Ronald Langacker Eugene Casad David Tuggy Ronald Langacker Marie-Odile Junker Lindsay Morcom Sally Rice, Martin Thiering Jocelyn Ahlers Chad Thompson, Melissa Axelrod Sally Rice, Conor Snoek Wallace Chafe Marianne Mithun Marianne Mithun Gary Palmer, Roy Ogawa Lindsay Morcom Lindsay Morcom Paul Friedrich Robert MacLaury Kristine Jensen de López Claudia Brugman, Monica Macaulay Barbara Hollenbach Colette Craig (Grinevald) Colette Craig (Grinevald) John Lucy, Jürgen Bohnemeyer Penelope Brown Eve Danziger John Haviland Rick Floyd Maura VelázquezCastillo Bernd Heine Bernd Heine Mark Dingemanse Bernd Heine
39
40
RICE
Table 3.1 (cont.) Region
Family
Language
Linguist
bantu
Chaga Chindali, various Bantu Wolof Shona Ewe, Akan Siwu
Michelle Emanation Robert Botne Gary Palmer Felix Ameka Felix Ameka Mark Dingemanse
Palauan Kilivila Longgu To’aba’ita, Manam, Vangunu various Oceanic Samoan, Hawa’ian
Michael Smith Gunter Senft Deborah Hill Frantisek Lichtenberk
pama–nyungan
Yélî Dnye Kalam Nen, various Papuan Guugu Yimithirr
gunwinyguan
Mparntwe Arrernte Kuuk Thaayorre Dalabon
Stephen Levinson Andrew Pawley Nicholas Evans Stephen Levinson, John Haviland David Wilkins Alice Gaby Nicholas Evans
Tamil Lao Tai Khamti
Eric Pederson Nick Enfield Douglas Inglis
kwa Pacific
malayo-polynesian melanesian oceanic
polynesian Australasia
Asia
p a p u a n (sic)
dravidian tai-kadai
John Bowden Kenneth Cook
* For limits of space, most of this research is not cited in the reference section, but the interested reader should have enough information to find the associated studies.
There has been a long tradition within the nearly four-decades-old CL movement of research that champions linguistic diversity, constructional specificity, and meaning as the starting point in any analysis of language. Much of this work has been based on data from unfamiliar and minority languages, a significant number of which have represented exclusively languages spoken by those regarded as indigenous peoples (see Table 3.1). Because most indigenous languages (ILs) are oral and endangered, and the speech communities have been small, linguists working on these types of languages have been able to bring fresh data with a minimum of analytical preconception to the cognitive/functional/typological table. Just possibly, work on these types of languages served as an inspiration to the entire movement in the first place (cf. section 3.1.1). The upshot of cognitively inspired IL analyses and analyses of languages other than ‘Standard Average European’ (SAE) has been a direct challenge to generative assumptions about universality and the separation of language and cognition, culture, the environment, and social structure. Furthermore, engagement with data and speakers of ILs (and other minority languages matching the IL prototype of lacking a written literary tradition, a purportedly
The Study of Indigenous Languages
transparent lexicon, or an analytic syntax, while relying heavily on stock or idiomatic expression and extensive evidential marking) has supplied a further assault on the purported lexical/grammatical divide, supported a re-conceptualization about claims of linguistic relativism, led to an integration of phenomena such as polysemy, heterosemy, and grammaticalization within linguistic theory, and provided additional demonstration of the ubiquity of metaphor and metonymy – all themes prevalent in the ‘mainstream’ CL literature. This chapter traces the impact of ILs on the origins of CL and its continued development as a school of linguistics, in part because CL has embraced linguistic phenomena that have long been marginalized by other theoretical approaches.
3.1.1 Progenitors (1965–1975) In a very real sense, field exposure with several Native American languages by a few key individuals who inspired, originated, or eventually became allied with cognitive linguistics informed early CL theorizing about the nature of language and the inseparability of meaning and form. These languages were the Mexican isolate Tarascan (P’orhe´pecha), in the case of Paul Friedrich, Seneca and Onondaga [i r o q u o i a n ] in the case of Wallace Chafe, Atsugewi [s h a s t a –p a l a i h n i h a n , possibly h o k a n ] in the case of Leonard Talmy, and various Uto-Aztecan languages in the case of Ronald Langacker – a language family, by the way, that includes Hopi, which had more than a little influence on Benjamin Lee Whorf and the linguistic relativity hypothesis (LRH) long associated with CL. Of course, the LRH is also called the Sapir–Whorf Hypothesis and not coincidentally linked to two prominent field linguists who examined much IL data. The influence of this contact with speakers of oral and morphologically rich languages on the development of themes central to CL is most widely recognized in Talmy’s 1972 dissertation from the University of California at Berkeley, which compared the distinct lexicalization patterns found in Atsugewi and English motion verbs and the different kinds of conceptual packaging associated with verb (phrase) morphology. But Talmy was neither the first nor the only linguist whose scholarship, at least partly inspired by contact with IL data, was admired and emulated by what we might call the first bona fide CL generation in the 1980s (see section 3.1.2). Indeed, Friedrich, Chafe, Talmy, and Langacker were largely working independently of each other in the 1960s and, except for Langacker, independently of themes of concern to the generative zeitgeist. What they had in common was exposure to ILs that had little to no prior documentation to which they were able to bring original insights. Their insights converged, especially, on the theme of the priority of meaning over form or, in Chafe’s terms, semanticity over syntacticity. In the case of Friedrich and Chafe, they came of age linguistically before the Chomskyan
41
42
RICE
revolution and their field experience insulated them from the more abstract, formal, universalist, and synchronic fixations of transformational generative grammar then sweeping the United States. In the case of Talmy, his linguistic training coincided with the rise of generative thought, but he did his undergraduate and graduate training at UC Berkeley, long associated with the study of American Indian languages, as well as with Brent Berlin and Paul Kay’s cross-cultural work on color naming, psychologist Eleanor Rosch’s hugely consequential work on graded categorization and prototype theory (Rosch, herself, conducted fieldwork with speakers of the indigenous language of Western New Guinea, Dani [p a p u a n ]), and a linguistics department that is recognized as one of the North American birthplaces of CL. Finally, in the case of Langacker, he was the only natively trained generativist of the four, but his move to southern California and the University of California at San Diego (UCSD) in the mid-1960s into a linguistics department in which Margaret Langdon – a student of Mary Haas’, herself a student of Edward Sapir’s – was also a colleague, and his work on various Uto-Aztecan languages, saw a profound shift in his analytical focus and theoretical perspective. These four scholars, with their deep exposure to ILs and each seminal to the development of CL, are discussed in turn. Paul Friedrich (PhD 1957, Yale), a linguistic anthropologist who spent most of his career at the University of Chicago, is by no means considered to be a cognitive linguist, yet his 1969 and 1971 studies on Tarascan suffixes of space are two of the most compelling and widely cited examples of cross-domain metaphorical mapping in the CL literature. He detailed how the referential use of animal and human body-part terms had extended lexically to new domains of reference (parts of artifacts, houses, etc.), four of which are shown in Table 3.2. Friedrich’s fine-grained analysis of the Tarascan locative system had a direct influence on a number of similar studies of Mesoamerican ILs by a group of linguistics graduate students at Berkeley in the 1980s, detailed below. It was also given significant coverage in Svorou’s 1994 influential CL monograph on the origin and grammaticalization of spatial terms cross-linguistically, which contained much IL data. Collectively, these analyses cemented the centrality of spatial semantics in the CL canon, which saw an explosion of research in the 1980s and 1990s (see section 3.2). A quote by Friedrich in a recent retrospective interview attests to the wonderment experienced by the analyst trying to crack the semantic code of a very unfamiliar language: “I think that what really fired me was not just the book by Sapir but the essays . . . I love[d] the idea of exploring and understanding an American Indian language, and that was part of the reward of going to Mexico . . . the second time I went, I worked exclusively on the language. That was very rewarding – deeply rewarding – because I got excited by these incredible words that they build with their structure” (Friedrich and Pesmen 2014: 17).
Table 3.2 Some cross-domain referential and relational extensions of Tarascan spatial morphemes based on body-part terms [first row] (Friedrich 1969: 36–37) d o m a i n \ suffix
-ŋaṛi
-mu
-ča
-ɕ̉i
body
whole body, eye cheekbone, face, lap, shin exterior surface of female genitalia social ‘face’, memory, imagination, emotions inside wall
lip, mouth area labia, vulva, vagina
outer neck, inner throat, inside of mouth; upper calf penis
outside of head, hair, brains, toothtip, kneecap head of penis, semen
speaking, talking
desire, memory
anal-genital psycho-social house part tree/ corn part pottery part other s i g n i f i c a t i o n (schema?)
conversation topic, profit in trade, social head door area intersection of roof and wall (la gable, altar barda) cob beneath husk, inner face of any bud, leaves around top of neck where root or shoot joins top of corn above cobs, tree worked log young cob tuber or fruit tops, root tips above ground inside of pot belly lip, part attached to neck of pot curve between mouth and belly cover of pot of pot cliff, hillside, liquid surface cuff, hem, edge of town, hole, various neck-like, neck-related hilltop, liquid surface river flat, interior surface edge, usually of an orifice narrowing, usually of a longish top, above, exterior of upper object at intersection surface
44
RICE
This sentiment was echoed by Wallace Chafe (PhD 1958, Yale) in his own memoir: Ever since my early fieldwork I have been convinced of the value of working in depth with one or more languages radically different from one’s own. Getting one’s hands dirty observing real speech, even in a familiar language but especially in an unfamiliar one, brings insights obtainable in no other way. I learned, too, that uncovering the mysteries of a language through fieldwork is the purest application of the scientific method. Observing a language in use, imagining explanations for why it has the shape it has, and seeing those provisional explanations confirmed, disconfirmed, or modified by further observations brings unequaled thrills of discovery, an experience that becomes less available each time a language dies. (Chafe 2002 [emphasis mine])
Like Friedrich, Chafe started out as a field linguist. His dissertation and descriptive analyses on Seneca [i r o q u o i an ] soon gave way to a series of well-placed publications in the 1960s on very philosophical topics about the nature of language that both prefigure and are fully consonant with classical CL themes. These titles include ‘Phonetics, semantics, and language’ (1962), ‘Meaning in language’ (1965), ‘Language as symbolization’ (1967), and ‘Idiomaticity as an anomaly in the Chomskyan paradigm’ (1968). These ideas came together in the very original theory of language articulated in his 1970 monograph, Meaning and the Structure of Language (MSL) (1970a), and applied in the article ‘A semantically based sketch of Onondaga’ (1970b), another Iroquoian language. The simultaneity of these two publications suggests that both were taking form and informing each other at the same time. Indeed, Chafe even included a chapter on Onondaga in MSL, a book that had a profound effect on many linguists who eventually identified with CL. Therein, he explored a number of themes that, today, are taken as given by most cognitive linguists, such as the centrality of semantics, that constructional meaning is more than the sum of constituent meaning alone, that language is made up of symbolic units of essentially a phonetic-semantic nature, that idiomaticization and information structure play an important role in the shape of predications, and that conceptual categories are graded, as is categorization as a whole. He has gone on to publish extensively over the next four decades in areas as diverse – and as central to CL – as language and memory, language and consciousness, discourse structure, verbalization, schemas, information flow, evidentiality, prosody and intonation units, epistemic modality, and grammaticalization, all the while producing analyses of various Iroquoian and Caddoan languages. His long career has seen description intertwined with a theoretical perspective all his own, but one which has run parallel to and influenced a wide range of CL work. Sustained engagement with ILs and intense scrutiny of their patterning, meaning, and use in conversation frequently fuels a point of view about human language that places semantics at the center and recognizes the role of diversity. This has clearly been the case for Chafe.
The Study of Indigenous Languages
Unlike Friedrich and Chafe, Talmy and Langacker are wholly aligned with CL and are two of its most central proponents. Chafe’s influence on them both cannot be overstated, nor can their own encounters with the intricacies and idiosyncracies of ILs (idiosyncracies only from the perspective of a more analytic or SAE mindset). Talmy’s first and warmest acknowledgment in his 1972 dissertation goes to Chafe, then chair of the Berkeley linguistics department. Also in 1972, Langacker published an article-length review of MSL in Language in which he concludes, MSL is . . . valuable because it focuses needed attention on several matters of central importance for future progress in linguistic theory and analysis. The importance of semantic structure requires no comment. Idiomaticity has vast implications for the relation between lexical and syntactic structure. Syntactic and semantic change is a neglected area that promises to shed much needed light on certain theoretical questions. Finally, it would be difficult to exaggerate the present importance of linguistic studies that go beyond the bounds of a single language family. (Langacker 1972a: 160)
In Talmy’s dissertation, just as in MSL, semantic analysis takes precedence over syntactic with regard to the differential encoding of participant structure, manner, direction, aspect, and evidence within the verb complex (the verb and any satellites) for motion predications across the two languages. Moreover, concerns for language generalities yield to interest in language particulars, although he wrote, “in comparing Atsugewi and English – as different from each other as two of the world’s languages might be – one discovers and can characterize a common core to both” (Talmy 1972: 5). Indeed, much of the exposition in the dissertation is occupied by an analysis of exceptional granularity surrounding ‘translation equivalents’ between the two languages for expressions of motion, like those shown in (1) and (2) (note the discontinuous morpheme,’-w- -a, which he glosses as ‘third person [f i g u r e , subject], factual [m o d e ]): (1)
(2)
w̉ oqʰputı´c̉ ta yes/no question marking > topic marking Wilcox (2014) has expanded on this grammaticalization path, proposing in addition that conditionals have as their source this generalized question gesture, which he suggests originally derives from a pan-species indication of surprise. Signed languages following the first route appear to show the same grammaticalization paths as those reported for spoken language (Sweetser 1990). However, as Pfau and Steinbach (2006) point out, there do seem to be modality-specific properties of grammaticalization. First is the obvious difference that signed languages grammaticize visual-manual gestures in a way that acoustic languages cannot. Second, signed languages appear not to exhibit grammaticalization from free to bound grammatical morphemes.1 Pfau and Steinback’s two explanations for this second finding, however, suggest that more research is needed. First, they hypothesize that the effect may be due to the comparatively young age of signed languages. However, as we have seen, we cannot assume that all signed languages have a shallow historical depth. Second, they suggest that the effect is due to the general scarcity of affixational, sequential morphology in signed languages. This begs the question of why signed languages should show such a typological tendency. The second route proposed has yet to be attributed in spoken language grammaticization, although Haiman (1998) makes an intriguing case for the ritualization of intonation and prosody. Ritualization is commonly invoked as a way to describe grammaticization. 1
At least one case, the grammaticalization of FINISH as a completive and perfective marker in ASL, seems to contradict this claim. In certain forms, such as SEE-FINISH, the marker is produced with one hand and appears to have become an affix to the one-handed form SEE.
Signed Languages
7.6.2 Signs as Language-gesture Fusions Many sign linguists now propose that some signs are simultaneous composites of linguistic and gestural material. This view originated with Liddell’s (2003) claim that certain forms are blends of conventional signs with gestural locations. For example, Liddell suggests that in the construction UPRIGHT-PERSON-WALK-TO, the handshape representing the person is fixed, encoding what he called lexical meaning, while the starting and ending locations he describes as variable and gradient. On this basis, Liddell argues that the former are listable in the grammar, but the latter are uncountable and unlistable, and therefore they are gesture. Wilbur (2013: 223) summarizes Liddell’s argument: “if the locations in space that are used for indexic and referential pointing are not listable, they cannot be part of the grammar, and therefore must be external to it, that is, part of ‘gesture’.” The current metaphor for this view is that certain signs constitute fusions of linguistic and gestural material (Liddell 2003, McKee et al. 2011, Cormier, Schembri, and Woll 2013, Meier and Lillo-Martin 2013). To support the claims of the fusion metaphor, sign linguists have proposed objective criteria by which to distinguish language from gesture, such as gradience (gesture) versus discreteness (linguistic) (Liddell 2003); idiosyncratic (gesture) versus conventional (linguistic) (Ferrara and Johnston 2014); contextsensitive (gesture) versus less context-dependent (linguistic); and holistic (gesture) versus dually patterned (linguistic) (Sandler 2009). The claim that certain signs, or, more correctly, certain signed constructions, are language-gesture fusions runs counter to several elements of cognitive linguistics. We outline these problems and discuss a usage-based alternative in the following section.
7.6.3 Sign and Gesture: Categorizing Usage Events The claim that some signs are language-gesture fusions assumes the linguist can objectively categorize certain elements in a construction as ‘language’ and others as ‘gesture.’ However, the criteria proposed by which to make this categorization, such as discreteness versus gradience, conventionalized versus unconventionalized, or context-sensitive versus less context-dependent, are continua rather than classical categories with clearly defined boundaries. Even more problematic, while these features may prove convenient for the analyst, they may not reflect the true nature of the phenomena under study. Langacker (2008a), for example, concludes that discreteness is more often than not imposed on language, and Bybee (2010) shows from a range of data that all types of linguistic units show gradience. A usage-based, cognitive linguistic approach offers a different approach (Wilcox and Occhino 2016). Rather than imposing certain categories on
115
116
WILCOX AND OCCHINO
language by theoretical assumption, the task of the linguist is to discover the categories in the minds of language users. For hearing people, who perceive both acoustic and optical signals, their experience is that multimodal usage events exhibit two commonalities: some elements of the event are heard, and some are seen. These two perceptual experiences are naturally categorized as language (speech) and gesture. The issue is thus to understand how people categorize usage events. Categorization is the interpretation of experience with respect to previously existing structures. When making a categorization judgment, people compare two structures and classify one as an example or instance of another (Langacker 2000: 14–17). For a deaf person, a visible usage event is a categorization problem to be solved. Deaf language users have developed schemas which serve to categorize usage events. The outcome of the categorization task will depend on the discourse context, the person’s prior experience, and the individual’s embodied knowledge – knowledge which, from the cognitive linguistic perspective, is crucially shaped by one’s unique perceptual and sensorimotor capabilities. Rather than seeing this task as one of determining which elements of a usage event are ‘language’ and which are ‘gesture,’ the usage-based, cognitive linguistic approach treats it as a grammaticality judgment. Certain expressions, which hearing users may judge to be gestures, are abstracted from multiple usage events and become conventionalized, entrenched, and eventually categorized as language. Brain studies of language and gesture are beginning to offer evidence for this analysis. For example, Newman et al. (2015) investigated whether the neural organization of language is influenced by language experience. Using fMRI, they tested deaf signers and hearing non-signers who viewed verbs of motion and gesture and found that lifelong experience with a visual language alters the neural network. They concluded that “For native signers . . . rather than sign language being processed like gesture, gesture is processed more like language. Signers recruited primarily language-processing areas, suggesting that lifelong sign language experience leads to imposing a language-like analysis even for gestures that are immediately recognized as nonlinguistic” (Newman et al. 2015: 11688).
7.7 Summary Signed languages are full, natural human languages, unique only in that they are expressed in the optical rather than the acoustic modality. Signed language linguists have described the traditional levels of structure that all languages exhibit – phonology, morphology, syntax, semantics, pragmatics, and discourse. Signed languages have been found to take unique advantage of their modality, including the use of space to express nominal and pronominal elements, verbal arguments, and narrative structure.
Signed Languages
Grammatical structure is often expressed by non-manual or facial markers. Signed languages have been shown to exhibit rich and productive iconicity, metaphor, and metonymy at all levels, which deaf signers use not only in creative genres but also in their natural and everyday discourse. The relation of signed languages and gesture has long been a contentious issue. Scholars first assumed that signed languages were merely depictive gestures with no linguistic structure. In the modern era of linguistics, sign linguists sought to demonstrate the linguistic status of signed languages by distinguishing them entirely from gesture. A later period of rapprochement ensued in which sign linguists explored historical and synchronic connections between gesture and sign. Gestures are often the source for signs in the lexicons of signed languages. Gesture also has been shown to be the source for grammatical elements, emerging through the well-known linguistic process of grammaticalization. The study of the relation between signed language and gesture continues to be a topic of research, much of it framed within a cognitive linguistic framework. A full understanding of the implications of a usage-based, cognitive linguistic approach leads linguists to attend to the lived, embodied, perceptual-motor experiences of language users. This has brought renewed attention to the significance of modality, of language by ear versus language by eye, and of the perceptual, cultural, and experiential histories of deaf versus hearing language users. Cognitive linguistic theories provide sign linguists with the analytical tools to move beyond imposing theoretical constructs based on structuralist/formalist analyses of spoken languages and to discover what is unique about signed languages. In this way, research that is informed by usage-based models offers a way to expand our knowledge of all human language.
117
8 Gesture, Language, and Cognition Kensy Cooperrider and Susan Goldin-Meadow
8.1 Introduction Sounds in the air, words in a transcript, rules in the head. These are the phenomena that researchers in linguistics have always focused on. As a result, we now know a great deal about the remarkable formal properties of human language. But this is only one side of language, and another side is easy enough to see. Next time you find yourself in a lively public place, tune out what the people around you are saying and instead watch what they are doing with their bodies. Where you are does not really matter: it could be a quad in California, a coffee shop in Istanbul, a village in Papua New Guinea; the people could be speaking Japanese, Quechua, Russian, or any other language. If you cannot make out what the people are saying or do not understand the local language, all the better. With the words tuned out, another side of language will come to life. Speakers will point, shrug, dismiss, affirm, and implore; they will use their hands to re-enact events, or to depict the shape and size of invisible objects; they will slice, punch, poke, and pull at empty air. Granted, it may not be obvious what these conversations are about. Perhaps a complicated moral quandary? A remodeling project? A close call? But the take-away should be clear: Whenever and wherever people talk, no matter the language or the topic, they gesture. Language and gesture go hand in hand from a very young age, and many speculate that they have gone hand in hand since the earliest stirrings of communication in our species. How does gesture’s pervasiveness change our understanding of language and its cognitive underpinnings? One of the guiding ideas of cognitive linguistics is that language is bound up with cognitive systems for acting, perceiving, attending, reasoning, and so on. Early evidence for this idea was found in the structures of language, how those structures are put to use, and how they change over time (Lakoff and Johnson 1980a, Langacker 1987, Sweetser 1990, Talmy 2000a). More
Gesture, Language, and Cognition
recent evidence has come from psycholinguistic experiments that shed light on what is going on in the mind when people produce or understand language (Bergen 2012). Our focus in this chapter is on another source of evidence for the interplay of language and cognition: the gestures – like those described above – that universally, ubiquitously, and spontaneously accompany speech. When we look closely at how gesture partners with words, it becomes clear that it is much more than some bodily version or ‘translation equivalent’ of spoken language. Gesture is a distinctive medium of expression. In what follows, we show that examining gesture can provide insights into language and cognition. In section 8.3, we focus on gesture’s intimate relations with spoken language, as well as gesture’s capacity to reveal aspects of conceptualization that are not evident from the words in a transcript. In section 8.4, we describe the communicative functions of gestures – that is, how gestures serve to convey important layers of meaning to listeners. In section 8.5, we describe the cognitive functions of gestures – that is, how gesturing helps speakers talk and think, and how it shapes reasoning. Before we can consider gesture’s relations to language and cognition, however, we must first consider some preliminaries: what counts as a gesture, how we know when a gesture starts and stops, and how gestures can be classified once they have been identified.
8.2 Defining, Individuating, and Classifying Gestures In popular parlance, the term gesture is used loosely and unevenly. In the academic study of gesture, however, the term has a restricted meaning: Gesture refers to movements of the body that are produced in the course of communication, as part of utterances (often called ‘co-speech gestures’) or even as utterances in themselves (Kendon 2004). This definition excludes practical, non-communicative movements, like taking a sip of coffee or fixing a ponytail. These are not integrated with speech in the same way and lack other properties of gesture. Research on gesture has focused on gestures of the hands, though people do gesture with other body parts, especially the head and face. Headshakes and head nods are commonplace if not universal (Jakobson 1972, Kendon 2002) and, in some cultures, the head and face play a prominent role in pointing (e.g. Sherzer 1972). When we say gesture in what follows, we mean gestures of the hands unless otherwise specified. Gesture can sometimes look like a blur of movement. But despite appearances, the gestural stream can be segmented into individual units and subparts, much like the speech stream. Here we cover only the very basics of gesture segmentation (more detailed discussions can be found in McNeill 1992, and Kita, Van Gijn, and Van der Hulst 1998). The heart of a gesture is called the stroke – this is the part of the gesture that is most
119
120
COOPERRIDER AND GOLDIN-MEADOW
meaningful and effortful. Other movements play important supporting roles too. In order to produce a gesture stroke, a speaker will often first have to get the hand or hands into better position. This phase is called the preparation. For some gestures, once the stroke has been produced, the speaker’s hand freezes in place. This phase is called a hold. Finally, after a gesture stroke is completed, the hands return to a resting position. This phase is called the retraction. In an unbroken sequence of gesturing, the stroke of one gesture will give way to the preparation phase of the next, with no retraction until the end of the entire sequence. Note that the stroke of the gesture is really what we focus on when we classify gestures, and when we consider the relationship between a gesture and its accompanying speech. A number of different schemes have been used to classify gestures into different types (a history of these schemes is reviewed in Kendon 2004). Many of the schemes are grounded in a basic distinction that the philosopher Charles S. Peirce made between the three different ways that a sign can carry meaning. Here a sign is any communicative unit, whether a gesture, a word, a graphical symbol, or the like (for discussion, see Clark 1996, Enfield 2009). Peirce’s categories are useful not only for classifying gesture, but also for thinking about how gestures relate to spoken and signed language, as well as to other communicative forms. First, gestures can mean by creating a direct connection to a referent in space and time – this meaning relation is sometimes called indexicality or simply indicating. The canonical indicating gesture involves extending the arm and index finger, but people also point with a jerk of the head or in variety of other ways across cultures (see contributions in Kita 2003). Tapping or holding up an object that one is talking about are also ways of indicating (Clark 1996). Second, gestures can mean by looking like what they refer to – that is, by depicting a referent in at least some respects. This meaning relation is sometimes called iconicity, and gestures excel at it. Because gestures involve actions of the body, they readily depict human actions. Similarly, because they unfold in space, they excel at depicting motion through space and spatial relationships (Emmorey, Tversky, and Taylor 2000). Gestures are not limited to representing actual actions, concrete images, and real spatial relationships, however. They are equally good at embodying metaphors in which actions, images, and spatial relations represent more abstract concepts. (This abstractness is one of the reasons that, as noted earlier, it can be hard to tell what exactly someone is talking about based on her gestures alone.) Third, and finally, gestures can mean because of a convention, much the way that the English word ‘chair’ means because of a rule that links a particular sound and meaning. This meaning relation is known as conventionality. Conventional gestures – like the ‘thumbs up,’ or the ‘shhh gesture’ – are called ‘emblems.’ Like words, they have standards of form.
Gesture, Language, and Cognition
You cannot change a feature of an emblem’s form and expect it to mean the same thing, just as you cannot insert a new syllable at the start of a word and expect people to know what you are talking about. Emblems are what first comes to mind when lay people think of gesture, but they are not necessarily the best studied (but see Ekman and Friesen 1969, Morris et al. 1979). Researchers interested in cognitive aspects of gesture have been drawn to gestures that are idiosyncratic rather than rule-bound in nature, especially iconic gestures and, to a lesser extent, pointing gestures. It is these gestures that are taken to offer a “window onto the mind” (McNeill 1992, Goldin-Meadow 2003). In some cases, a gesture ‘means’ in one and only one of the above three ways, which has led the literature to talk about a few main gesture types: ‘deictic gestures’ (those that involve indicating), ‘iconic gestures’ (those that look like their referents in some way), ‘metaphoric gestures’ (those that iconically represent abstract ideas), and ‘emblems’ (those that are frozen conventions). However, gestures regularly combine indexicality, iconicity, and conventionality to various degrees (McNeill 2005, Enfield 2009). Consider the beckoning gesture you might make when you catch a friend’s eye across a room and want her to approach. The gesture is indexical in that it ‘points’ toward the intended recipient; it is iconic in that it represents the path between the gesturer and the recipient; and, finally, it is conventional to the extent that, in some cultures, people beckon with the palm up (e.g. in the US) and, in other cultures, they beckon with the palm down (e.g. in Mexico). Idiosyncratic gestures lack a conventional component, by definition, but they still often combine iconicity and indexicality (for discussions, see Goodwin 2007, Cooperrider 2014). If you pay close attention to gestures in everyday conversation, you will notice that a gesture’s meaning is sometimes murky – it may not crisply point or depict, and it may not have a clear conventional meaning either. For instance, speakers commonly make movements that are time-locked with speech, but that do not mean in any of the three ways described above. These movements are often called ‘beat gestures’ because they keep the rhythm of language but do little else (McNeill 1992). The communicative and cognitive functions of beat gestures remain poorly studied and thus somewhat mysterious, but they cannot be swept under the rug. Nor can so-called ‘interactive gestures,’ another class of gestures that are common but function differently from the gestures just described. Several researchers have proposed a basic split between gestures that operate at the level of what the speaker is talking about – called ‘content’ gestures – and gestures that operate at the level of the conversational interaction itself – called ‘interactive gestures’ (Bavelas et al. 1992; see, also, Kendon 2004 for a related class termed ‘pragmatic gestures’). Speakers use the latter type of gestures to clarify who has the floor, to engage the audience, to carry out speech acts like requesting, or to
121
122
COOPERRIDER AND GOLDIN-MEADOW
provide commentary on their own utterances. Here we follow the field’s focus, which has been on idiosyncratic gestures that relate to the content of what the speaker is talking about – the events, characters, relationships, and ideas that form the substance of our stories and explanations. However, we also note that gestures on the fringes of this traditional focus – including emblems, beat gestures, and interactive gestures – are promising ground for future researchers to chart.
8.3 Gesture and Language One of the foundational observations of modern research on gesture is that it is tightly linked to language – so tightly that it is sometimes argued that gesture should be considered part of language. The tight linkage between gesture and language is evident across different levels of analysis and across timescales. We first review some of the evidence for this integration, with a focus on how speech and gesture go hand in hand in the moment-to-moment flow of discourse (thus leaving aside interesting issues about how gesture and language develop together in children, e.g. Goldin-Meadow 2014, and about how they may have evolved, e.g. McNeill 2012). We then turn to the ways in which gestural meaning is able to pick up where linguistic meaning leaves off.
8.3.1 Gesture and Language, Hand in Hand A first way that gesture and language go hand in hand is in their timing. Gestures do sometimes stand alone as utterances in their own right – consider the beckoning gesture described above, or a headshake in response to a yes–no question. And gestures also sometimes fill gaps in the speech (for discussion of these cases, which are sometimes called ‘component gestures,’ see Clark 1996). More commonly, though, gestures and speech overlap and are finely coordinated in how they overlap. Usually, the stroke of the gesture will line up in time with whatever part of the speech stream it is most closely related to in meaning. This synchrony is referred to as co-timing, and it has an interesting signature: Gesture strokes often precede the relevant unit of speech, and sometimes line up with it exactly, but almost never follow it (Schegloff 1984, McNeill 1992, Morrel-Samuels and Krauss 1992). A second major way that gesture and language go hand in hand is that they co-express meaning. This synchrony is seen, first, in how the spoken and gestural streams together serve to highlight the parts of an utterance that are important. Speakers are selective in choosing what part of the speech stream they stress, and they are also selective in choosing where in an utterance to place a gesture. Interestingly, these choices often coincide: Gesture will most often go along with whatever is stressed in speech
Gesture, Language, and Cognition
(Kendon 2004). This timing characteristic applies to beat gestures as well as iconic gestures (McClave 1994). In the case of iconics, the part of the speech stream that a gesture relates to in meaning (which is also the part it aligns with in timing) is sometimes called the ‘lexical affiliate’ (Schegloff 1984, Morrels-Samuels and Krauss 1992). Thus a speaker might say cut while depicting the action of using scissors, circle while tracing a circular shape with the fingers, or above while moving the hand upward. In other cases, the gesture lines up, not with a single word, but with a larger phrase. And, in cases of the ‘interactive gestures’ mentioned earlier, how speech and gesture align may be governed by other principles (e.g. Kendon 1995). Finally, as we describe in more detail later, there are deviations from this tidy co-expressivity and these deviations turn out to have important cognitive implications. Relationships between gesture and speech in timing and meaning are strikingly robust. For example, when stutterers start to stutter, they pause in the gesture too, thus preserving the lockstep timing between the two modalities (Mayberry and Jacques 2000). More generally, when speakers experience disfluencies and decide to repair their speech, they tend to reproduce their gesture as well (Seyfeddinipur and Kita 2001). Findings like these lend support to the idea that the co-expressivity of gesture and speech is not an accident – rather, in McNeill’s (2005) terms, speech and gesture exhibit an ‘unbreakable bond.’
8.3.2 Gestural Meaning Where Linguistic Meaning Leaves Off Now we come to an important wrinkle. Though gesture and speech may exhibit an unbreakable bond and work together to co-express meaning, they rarely mean in precisely the same way. Then again, how could they? They are two different modes of expression with different properties. Words are fully conventional and are strung together in linear order according to well-defined rules. Gestures are usually created on the spot and communicate meaning in global chunks rather than in an analytic fashion (see McNeill 1992 for discussion). Moreover, words are categorical bits of meaning, whereas gestures are analogue three-dimensional images that unfold in time. Consider an analogy to another case where word and image are called upon to co-express meaning: The titles of works of art (discussed in Enfield 2009). There are cases when a painting has a clear point of focus (e.g. the mountain called Denali) and the title seems to capture it perfectly (e.g. ‘A View of Denali’). But even in such seemingly unproblematic cases, the image involves particulars not captured by the title: The perspective, the fall of light, the time of year, and so on. Observations like these are, of course, what inspired the saying that a picture is worth 1,000 words. And so it is with gesture. The point is not that gestures are hyper-realistic and language is impoverished; in fact, gestures are often highly schematic. The point is that a number of choices
123
124
COOPERRIDER AND GOLDIN-MEADOW
about perspective, orientation, handshape, and so on, may be evident in gesture even when they are not evident in the accompanying speech. To make this point concrete, consider the utterance Then I opened the door. Co-timed with opened, the speaker reaches out a hand, turns an imaginary knob, and pulls the door toward him. No problem – gesture and speech seem to tidily co-express the idea of opening a door. But this tidiness is a bit of an illusion. Think about all of the different ways this gesture could have been produced, each expressing a variant on the idea of opening a door. It could have been a swinging door with no knob, in which case the gesture may have depicted a pushing motion; it could have been a door so heavy that two hands were needed to open it; and so on. This range of possibilities is not just a peculiarity of the word door in English – consider aspects of meaning left out in sentences like The cafe´ is next to the dry cleaners (on which side?), or It was a chocolate cake (what size or shape?). Over the years, cognitive linguists have taken much interest in how subtle choices of linguistic form reflect differing construals of the same underlying event or situation (e.g. Croft and Cruse 2004, Bergen 2012). This observation can be extended further: Even utterances that are identical in form may involve different underlying construals, and one way we could detect a difference in construal is by attending closely to subtleties of the co-occurring gestures. In the next sections, we examine the insights gesture offers into how people are construing events or ideas by considering several different dimensions of meaning. Perspective. When you tell a friend that A rat scurried across the room, you probably have some specific image in mind. But it is hard to tell just what that image is from your speech alone. For starters, the words alone tell us nothing about the perspective you are taking on the scene. Are you taking the view of the rat, perhaps imagining the quick arm movements that might be involved in scurrying? Or are you taking the vantage point of an observer, imagining the path of the rat’s movement across your field of view? These two different possibilities have been observed in gestures and are known as ‘character viewpoint’ and ‘observer viewpoint,’ respectively (McNeill 1992, Parrill 2010). Whether someone takes a perspective inside or outside of a scene is also evident in the gestures (and, interestingly, the signs) produced when describing spatial configurations. In the case of spatial descriptions, the different viewpoints are called ‘route’ (inside perspective) and ‘survey’ (outside perspective) (Emmorey, Tversky, and Taylor 2000). Note that, in both cases, the language used in the broader discourse may tend to resonate with the perspective adopted in gesture. The crucial point is that, for any given utterance, this perspective may be evident only in gesture. Spatial orientation. Let us consider the scurrying rat in a bit more detail. Another aspect of the scene, one that is easy enough to leave out of speech, is which direction the rat was scurrying. The word ‘scurry’ on its own says nothing about direction – it does not specify whether the rat scurried left
Gesture, Language, and Cognition
or right, up or down, east or west. But a scurry gesture has to move in one direction or another. Kita and Oyzurek (2003) found that speakers’ gestures reliably preserved the direction of a motion event in a recently observed cartoon, even though speakers’ words rarely mentioned it (interestingly, in other respects, the same gestures and speech were tightly coexpressive). Similar observations apply to the representation of static scenes, where a direction-free word like ‘next’ might be co-produced by a direction-specifying gesture (Tutton 2013). In these studies, speakers’ gestures reliably capture spatial direction or orientation in terms of the bodily coordinates of the observer; for example, in terms of their left–right or front–back axis. It has been suggested that, in cultures where cardinal direction terms are used pervasively, speakers produce gestures that preserve the original cardinal orientation of previously observed events or configurations in space, rather than the orientation with respect to the viewer’s body. Haviland (1993) illustrates examples of this phenomenon in Guugu Yimitthirr, an Aboriginal language of Australia. A storyteller describing how a boat capsized recreated the event – in gesture, but not in speech – as happening from east to west. This orientation is clear because the same cardinal orientation was preserved across two different tellings of the story in which the storyteller faced different cardinal directions. Consistent use of cardinal orientation in gesture has also been documented in speakers of Yucatec Mayan when describing the spatial relationship between two static landmarks (Le Guen 2011). Embodied action. When speakers gesture about actions like opening a door, wrapping a present, or tying a knot, their gestures often incorporate details about how the action was performed – for instance, what kinds of grips or movements were involved. Cook and Tanenhaus (2009) had people solve a classic reasoning puzzle, the Tower of Hanoi. Solving the puzzle involves moving a set of differently sized disks from one side of an apparatus to another, while also obeying certain rules. One group of participants solved the puzzle using a physical apparatus with heavy disks and high pegs. The other group solved a digital version of the puzzle by clicking and dragging with a mouse. The question was whether people in these different conditions would gesture differently when later explaining to another participant how to solve the puzzle. Those in the physical apparatus group, compared to those in the digital group, used more grasping handshapes (like the handshapes you would use when actually picking up the disks) and produced gestures with higher trajectories (like the trajectories you would use when having to lift the disks over the high pegs). Importantly, the researchers found no differences in the language used by the two groups. Problem-solving strategies. From the preceding discussion, it is clear that gesture offers insights – over and above speech – into how people are thinking about events, actions, and spatial relationships in the physical world. But gesture can also offer insights into how people are approaching
125
126
COOPERRIDER AND GOLDIN-MEADOW
much more abstract problems. An example is children’s understanding of a tricky concept: conservation of quantity. If you pour all of the water from a tall, skinny glass into a short, wide glass and ask children whether the amount of water has changed, children under the age of seven or eight will often say that it has. These children (called ‘non-conservers’) are focusing on the fact that the height (or the width) of the water has changed, and not on the crucial fact that these changes are complementary and compensate for one another. But the gestures of these non-conserving children often reveal something more interesting. Some children produce speech that highlights the height of the two glasses (e.g. This one is taller than that one) while, at the same time, producing gestures that highlight their width (e.g. showing a short distance between the thumb and index finger near the tall glass while saying taller and a bigger distance near the short glass when saying shorter). Children who produce these so-called ‘mismatches’ between gesture and speech seem to be entertaining two ideas at once. And, interestingly, it is the mismatching children who are particularly likely to learn from subsequent instruction (Church and Goldin-Meadow 1986). The predictive power of children’s mismatching gestures has also been found in reasoning about mathematical equivalence (Perry, Church, and Goldin-Meadow 1988) and the balance scale (Pine, Lufkin, and Messer 2004). In general, gesture speech-mismatches seem to occur at moments of cognitive instability. This phenomenon is seen not only in children, but also in adults learning a task. For example, when explaining how they solved the Tower of Hanoi puzzle – the same disk-and-peg task described earlier – both children and adults sometimes produce gestures that suggest a problem-solving step that is different from the step described in speech. Interestingly, such mismatches are not randomly sprinkled throughout people’s explanations, but crop up at precisely those points in the problem where a decision needs to be made between two alternative steps (Garber and Goldin-Meadow 2002). In another study, Alibali et al. (1999) examined adults’ gestures as they described algebra problems that they would later have to solve. Some problems involved discrete change. When people’s speech and gestures together represented this discreteness, they tended to use a corresponding strategy to solve the problem later. Other problems involved continuous change. Again, when people’s speech and gestures both represented this continuousness, they went on to use a corresponding solution strategy. When people only represented the problem as discrete or continuous in speech, with no gestures, this explanation was not as strong a predictor of their eventual solution strategy. And, more interestingly, when they represented different strategies in speech and gesture (continuous in one, discrete in the other), they were just as likely to use the gestured strategy as the spoken strategy. Speech alone is thus not as powerful a predictor of how someone is thinking about a problem as are speech and gesture together.
Gesture, Language, and Cognition
Metaphor. Reasoning about abstract ideas is not confined to cases where we are presented with difficult problems in the classroom or laboratory. In everyday conversation, people talk about notions like progress, order, and the passage of time; about intricate social scenarios and moral issues; about status and power; and about many other slippery but important abstractions. As cognitive linguists have insightfully shown, discourse surrounding topics of this sort is packed with metaphor (Lakoff and Johnson 1980a, and Sullivan this volume Ch. 24). Less appreciated is the fact that such discourse is also packed with gestures (Cienki and Mu¨ller 2008a). As described earlier, some researchers distinguish a separate type of gesture called ‘metaphorical gesture’ (McNeill 1992), though ‘metaphoricity’ may also be thought of as a dimension of gesture (McNeill 2005). Pointing gestures can be metaphorical, and emblematic gestures can have metaphorical aspects. Research on metaphorical gestures to date has yielded at least two important insights about metaphorical thinking. First, just as gestures about concrete topics can provide vivid insights into visuo-spatial and motor aspects of construal that are not part of speech, gestures about abstract topics can provide vivid insights into spatial and motor aspects of metaphorical construal that are not part of speech. Second, as observed initially by Cienki (1998), gesture often lets us know that metaphors are afoot when there would be no hint of the metaphor in a verbal transcript. Insights from gesture have been especially important in deepening our understanding of how people think about time (see Gaby and Sweetser this volume Ch. 39; see also Cooperrider, Nu´n˜ez, and Sweetser 2014 for a review). Time has long been a favored model system for understanding metaphorical thinking because it is both ineffable and indispensable (Casasanto 2008b). Consider the statement The meeting took forever. In an accompanying gesture, the speaker might depict a large swath of time between her hands, as though conceptualizing the duration of the meeting as an elongated object oriented across her body. Such a gesture tells us, first, that the speaker is construing time as linear and oriented left to right. Just as a gesture depicting a scurrying rat must be oriented in space in some way, a gesture depicting a stretch of time must be oriented too. Second, look back at the speech – there is no trace of verbal metaphor. At a very basic level, the gesture thus tells us that the speaker is reasoning metaphorically when we would not know this otherwise. Speakers do not spatialize time willy-nilly in their gestures. They tend to gesture systematically in ways that fit with a culturally shared model. The most common pattern for gesturing about time in English, for instance, is to show a timeline oriented from left to right. This orientation is surprising given that verbal metaphors in English (The week ahead or Back in the eighties) exclusively mention the front–back axis (Casasanto and Jasmin 2012). Speakers do gesture about time along the front–back axis, albeit less frequently than along the left–right axis. And, interestingly,
127
128
COOPERRIDER AND GOLDIN-MEADOW
they also sometimes mix together these two metaphors in their gestures, producing, for instance, a gesture with ‘tomorrow’ that unfolds forward and rightward and is thus simultaneously consistent with both front–back and left–right metaphors (Walker and Cooperrider 2016). Because time gestures are systematic and culturally shared, they can also give us insights into cross-cultural variation in the conceptualization of time, even when clues to this variation are either ambiguous or elusive in the language alone (Nu´n˜ez and Sweetser 2006b). Time has been a major focus of research on metaphorical gestures to date, but related insights have also been gleaned from other domains. Enfield (2009) describes a gestural system used by Lao speakers for construing kinship relations spatially. Winter, Perlman, and Matlock (2013) describe some of the metaphorical gestures speakers produce when talking about numbers, including gestures that construe numbers as big, high, or further rightward along an imagined number line. And Marghetis (2015) describes how gestures reflect different ways of metaphorically construing arithmetic, such as collecting together objects or moving along a path.
8.4 Gesture and Communication In the previous section, we considered how researchers can glean important insights into subtleties of what speakers are thinking by attending to subtleties of how they gesture. But, clearly, speakers are not gesturing as a service to curious researchers. So who is all this gesturing for anyway? Is it for the listener’s benefit? Is it helpful for the speaker? The ‘who is gesture for’ question is of long-standing interest, and it remains hotly contested. To preview, there is not one answer to the question: Gesture serves both communicative functions (i.e. for the listener) and cognitive functions (i.e. for the speaker). In this section, we consider some of gesture’s communicative functions. Do listeners glean information from gestures and, if so, what kinds of information? Which aspects of a gesture – such as its form or its relationship to the accompanying language – play into the comprehension process? Despite some early skepticism in the field (e.g. Krauss, MorrelSamuels, and Colasante 1991), it has now been demonstrated beyond doubt that listeners extract substantive information from gesture (see Kendon 1994 for a review of earlier studies and Hostetter 2011 for a review of recent experimental findings). A compelling kind of evidence for this claim comes from cases where listeners glean information that is altogether absent from the speech. Recall the Tower of Hanoi study by Cook and Tanenhaus (2009) described earlier, in which speakers produced gestures of varying heights to represent moving the disks from one peg to another. Again, this variation in height was not part of what speakers said – it showed up exclusively in their hands. In a later phase of the
Gesture, Language, and Cognition
same experiment, the researchers looked at what listeners took away from these gestures. They asked the listeners to solve the Tower of Hanoi puzzle themselves afterward on a computer, so that they could precisely track the listener’s mouse movements. What they found was that people who saw higher-arcing gestures during the explanations produced higher-arcing mouse movements during their own solutions. Listeners thus extracted information about a gradient – and task-irrelevant – feature of gesture and incorporated it into their subsequent actions. Another convincing demonstration of this kind of integration is that listeners incorporate the gestures they see into their own subsequent talk and gesture. Cassell, McNeill, and McCullough (1998) had participants view stories told by an actor, who was trained to produce particular gestures–speech combinations. At one point in the story, the actor says And Granny whacked him one with an accompanying gesture. There were two versions of this key moment: in one, the accompanying gesture was a punch and, in the other, it was a slap. When retelling the story, a listener who saw the version with the punch gesture reported verbally that Granny punches him; a listener who saw the version with the slap used the word whack and demonstrated the slap in gesture. These findings show clearly that listeners can, and do, extract meaning from gesture. Recent studies have gone a bit farther, showing that listeners integrate gesture even when they are instructed to ignore it. Kelly, Ozyu¨rek, and Maris (2010) had people watch ‘action primes’ – clips of simple actions like chopping vegetables. Immediately after viewing the prime, they were presented with a target video in which a speaker made a gesture and said a word, and the participants had to judge whether the new video related to the earlier one. People were fastest to respond when both speech and gesture related to the earlier video (e.g. the word chop with a gesture representing chopping). Interestingly, this effect was found even when participants were told to pay attention only to the speech in the target – they could not help but process the gesture as well as the speech. Automatic processing of this kind has also been found for pointing gestures (Langton and Bruce 2000). A new frontier of research is also examining the cognitive and neural processes that underlie gesture–speech integration. For instance, to what extent does understanding gesture engage neural circuits that support spatial reasoning or motor imagery (Ping, Goldin-Meadow, and Beilock 2014, Wu and Coulson 2015)? To what extent does gesture understanding engage classic language circuits, such as Broca’s area (Skipper et al. 2009)? Listeners glean information from gesture, but does their doing so have real-world consequences? Yes, it does. The best evidence comes from studies of gesture’s role in the classroom. Consider cases such as those described above where children produce gesture–speech ‘mismatches’ when struggling to explain a new concept like conservation of quantity or mathematical equivalence. This information is not
129
130
COOPERRIDER AND GOLDIN-MEADOW
just useful for researchers trying to predict who will learn, but is also accessible to teachers and ordinary listeners, and information found uniquely in teachers’ gestures is accessible to their pupils (see GoldinMeadow and Alibali 2013 for a recent review). Gestures that learners see during instruction can thus play a role in helping them form new concepts. Lessons about mathematical equivalence (Church, AymanNolley, and Mahootian 2004), symmetry (Valenzeno, Alibali, and Klatzky 2003), and conservation of quantity (Ping and Goldin-Meadow 2008) have been found to be more effective when they include gesture than when they do not. In fact, lessons may be particularly effective when they include gestures that provide information over and above the information that is in speech. For instance, Singer and GoldinMeadow (2005) instructed children in how to solve mathematical equivalence problems using explanations in which either (a) the same strategy was presented in speech and gesture, or (b) two different strategies were presented – one in speech and one in gesture. Children receiving two different strategies outperformed those receiving one. They even outperformed children receiving the same two strategies but conveyed entirely in speech. This result is striking in that not only were children able to make sense of the mismatch between speech and gesture, but they also benefited from it. The kinds of evidence described in the preceding paragraphs show clearly that gesture shapes what listeners understand, and that this shaping has consequences later on. An important related question is whether speakers intend their gestures to communicate in the first place. After all, gestures could be largely involuntary, perhaps in the same way that facial expressions have been argued to be involuntary (Ekman and Friesen 1969). In some cases, it is clear that speakers are tailoring their gestures to communicate. Pointing gestures co-produced with this or that, or iconic gestures co-produced with like this, are good examples of such tailoring. And, in fact, when speakers cannot see their listeners, they do not call attention to their gestures verbally (Emmorey and Casey 2001, Bangerter 2004). But even gestures that are not explicitly marked in this way may be intended to communicate. Evidence for this possibility comes primarily from studies in which speakers cannot see their listeners – perhaps because they are talking through a wall, on the telephone, or over an intercom. Studies of this sort consistently find that speakers gesture less when their listeners cannot see them (though, interestingly, production of beat gestures is not affected) (Alibali, Heath, and Myers 2001, Bavelas et al. 2008). Notice, however, that these findings have an intriguing flipside – they make it clear that communication cannot be the whole story, as speakers continue to gesture at non-negligible rates even when their listeners cannot see them. This is a first hint that speakers may have ‘selfish’ motives after all.
Gesture, Language, and Cognition
8.5 Gesture and Cognition We may not have needed an experiment to convince us that people gesture when there is no listener to see them – this behavior is easy enough to observe in ourselves when we talk on the phone. Other cases of ‘gesture for self’ may be a bit more surprising. People continue to gesture in completely non-communicative contexts, such as when alone solving problems involving mental rotation (e.g. Chu and Kita 2011) or counting objects (Carlson et al. 2007). Congenitally blind people gesture even though they have never seen a gesture in their lives, and are talking to other blind people who have not, either (Iverson and GoldinMeadow 1998). Even in cases where certain kinds of gesture are taboo – as is the case of pointing with the left hand in Ghana – people simply cannot help themselves (Kita and Essegeby 2001). In short, we all seem to experience what Kita and Essegeby (2001) describe as a ‘cognitive urge’ to gesture. One reason for this urge may be that gesture helps us talk and think. For example, gesturing can help us access the right words (Krauss 1998), particularly in talk about spatial topics. Speakers produce gestures at higher rates when talking about space than when talking about non-spatial topics. Further, when people are prevented from gesturing by physically restraining their hands, this manipulation selectively disrupts their ability to talk fluently about space (Rauscher, Krauss, and Chen 1996). Another idea, not incompatible with the possibility that gesturing aids lexical access, is that gestures help speakers organize their thinking in a way that is suitable for verbal expression (Kita 2000). Gesture may also help thinking by lightening ‘cognitive load.’ People can only handle so many cognitive tasks at once – handling several at the same time leads to greater cognitive load. Gesturing while speaking could be thought of as a cognitive task that increases cognitive load. If so, gesturing should hinder our ability to handle other tasks performed at the same time. But if gesturing forms a single integrated system with speaking, then gesturing could actually free up cognitive resources that can be used to perform other tasks. To examine these possibilities, GoldinMeadow et al. (2001) gave people two tasks to perform simultaneously – one was a memory task, the other an explanation task. Would telling people to gesture during the explanation task help or hinder performance on the memory task? As it turned out, people did better on the memory task when they gestured than when they did not gesture. In this experiment, participants’ gestures naturally related to what they were explaining. It has since been found that ‘meaningless movements’ – rhythmic movements that do not relate to the content of the speech – do not lighten the speakers’ cognitive load in the same way (Cook, Yip, and Goldin-Meadow
131
132
COOPERRIDER AND GOLDIN-MEADOW
2012). Movements need to be meaningfully integrated with speech in order for them to lighten cognitive load. In addition to helping speakers talk and lightening their cognitive load, gesture can also play a role in reasoning. Alibali et al. (2011) showed that people’s gestures may lead them down particular reasoning paths when solving gear problems. In this type of problem, people are shown a configuration of multiple interlocking gears, and are asked how moving one gear would affect another gear in the configuration. People typically use one of two strategies: (1) a ‘depictive strategy’ in which they simulate the physical movement of each gear involved; or (2) an ‘analytic strategy’ in which they determine whether the number of interlocking gears is odd or even. Encouraging people to gesture in this task nudges them toward using the depictive strategy, presumably because gesture lends itself to visuo-spatial simulation. Relatedly, Beilock and Goldin-Meadow (2010) have found, using the Tower of Hanoi puzzle discussed earlier, that the gestures speakers produce when explaining how they solve the puzzle have repercussions when they try to solve similar versions of the puzzle later on. Some of the best evidence for gesture’s cognitive functions comes from studies of the role of gesture in learning. These studies also make it clear that gesture can have important real-world consequences for the gesturer. As we have already seen, the gestures that learners produce when encountering a tricky new concept can often tell us something about their cognitive state. But, it turns out, those gestures can also be instrumental in changing that state. The mere act of encouraging children to gesture – without telling them how – appears to activate their unspoken ideas, which, in turn, prepares them to benefit from later instruction. This kind of benefit has been found in mathematical equivalence (Broaders et al. 2007), as well as in moral reasoning (Beaudoin-Ryan and GoldinMeadow 2014). Moreover, teaching children to produce particular types of gestures can give them totally new ideas that they are able to transfer to new problems (Goldin-Meadow, Cook, and Mitchell 2009). Even subtle properties of the gestures learners are asked to produce can make a big difference. One recent study compared the learning benefits of two kinds of gestures. One kind represented concrete actions, such as picking up and moving number magnets in order to solve an equivalence problem; the other kind of gesture abstracted away from these concrete actions, instead highlighting an abstract operation – grouping – involved in solving the problem. Children taught to use the abstract gesture developed a more abstract understanding of equivalence, as evidenced by the fact that they were better able to generalize what they learned to new types of equivalence problems than children taught to use the concrete gesture (Novack et al. 2014). In this and the previous section, we have reviewed evidence that gesture serves both communicative and cognitive functions – that it is
Gesture, Language, and Cognition
both for the listener and for the speaker. We have also described how, even though it may not be obvious in the moment, the gestures we produce have consequences for reasoning and understanding. Importantly, the fact that gesture can serve all these functions does not mean that it is always serving those functions in each and every instance. A challenge for future research is to work out in better detail when gesture serves communication and when it serves cognition, as well as when it has important cognitive consequences. Not all gestures are alike, and different gestures may turn out to have different functions and consequences.
8.6 New Directions and Conclusion Our focus has been on relationships between spoken language and gesture as they are co-produced in the moment-to-moment flow of discourse. A number of additional aspects of the language–gesture relationship would no doubt also be of interest to cognitive linguists. Several of these have already received substantial attention, including how gesture and language relate during language development (see Goldin-Meadow 2014, and Goldin-Meadow and Alibali, 2013 for a review); how gesture relates to language when it is produced in the manual modality (i.e. sign language; see Goldin-Meadow et al. 2012, Goldin-Meadow and Brentari in press); and how gesture varies across cultures, in ways that do and do not parallel variation in language (see Kita 2009). Other topics remain to be explored in fuller detail. For example, does gesture lead to ‘framing effects’ like those that have been described for spoken language (Thibodeau and Boroditsky 2011)? To what extent are gestural forms conventionalized, and how does such conventionalization occur? How do the communicative and cognitive functions of gesture compare to those of its close cousins – for example, the words of spoken language, diagrams, facial expressions, and the signs of sign language? What is the role of gesture in the creation of new communication systems, and what can this tell us about its possible role in language evolution (Fay et al. 2014)? And many more. For centuries, language has been a topic of singular interest among researchers working on human communication and cognition. Much of this work has focused on the formal side of language – its properties as a complex system of words and rules. But there is another side of language hiding in plain sight – gesture. Gestures show that, when people talk, they are doing more than using the words and rules of a formal system. Gestures and speech are intimately interwoven even as, at the same time, they are distinct – gestures are imagistic and often idiosyncratic, whereas words are categorical and conventional. This synchronization does not mean, however, that gesture and speech are translation
133
134
COOPERRIDER AND GOLDIN-MEADOW
equivalents in different modalities. Gestural meaning often picks up where linguistic meaning leaves off, affording the researcher vivid insights into how speakers are construing what they are talking about. Gesture is much more than just a tool for researchers, however. It also serves the listener – whether learner, teacher, or lay observer – by adding layers of meaning to spoken messages, and it serves the speaker by aiding talking and thinking and by shaping reasoning.
9 Multimodality in Interaction Kurt Feyaerts, Geert Brône, and Bert Oben
This contribution stands at the crossroads of two research paths, namely interactional linguistics and multimodal meaning analysis, which until a few years ago did not figure all too prominently on the road map of Cognitive Linguistics (CL). Yet each of these lines is part of the avenue of empirically validated research, which from the 1990s onwards CL has taken. This chapter draws attention to recent and ongoing research in the rapidly growing field of multimodal interaction analysis and explains its relevance to CL. It is structured as follows. In the first part, we present an overview of CL-inspired studies, which have the analysis of (some aspects of) face-to-face interaction as their major research objective. After a general introduction concerning the intersubjective process of meaning coordination, we focus our attention on three research domains, interactional humor (9.1.1), interactional alignment (9.1.2), and construction grammar (9.1.3), where a number of studies demonstrate the empirical shift being made. Section 9.2 focuses on the growing interest in multimodality in the sense of getting all semiotic dimensions involved when analyzing the complexity of a usage event. In this context, our specific attention concerns the issue of whether or not the framework of Construction Grammar can adequately account for the identification and analysis of multimodal constructions (9.2.2). In section 9.3, we stand at the crossroads of these two new empirical developments in CL and consider the extent to which studies in CL have already integrated both empirical perspectives in their project design. Again, we focus our attention on three domains in which recent empirical studies reveal an increasing activity in this respect: multimodal interaction management (9.3.1), multimodal alignment (9.3.2), and multimodal stance taking expressing obviousness (9.3.3). Finally, in section 9.4, we zoom out again, claiming that in the integration of these new empirical challenges, CL is well underway to achieve yet another recontextualization: systematically
136
FEYAERTS, BRÔNE, AND OBEN
including the analysis of interactive usage events as well as adopting a multimodal research perspective. It would be erroneous to claim that cognitive linguistics does not have the analysis of face-to-face interaction and multimodality on its current research agenda. As a matter of fact, both areas of interest are central to a steadily increasing number of CL-inspired or CL-related studies. It would lead us too far to try to present an all-encompassing overview of those studies. We do want to address the most debated topics in current research in both fields.
9.1 Interaction One of the most recent empirical shifts in CL-inspired research concerns the extension of its more traditional focus on speaker-centered conceptualization and construal mechanisms to a genuinely interactive and therefore sociocognitive account of meaning. This perspective starts from the basic assumption that dialogue is the most basic form of language use and, accordingly, takes the coordination of meaning between different interlocutors as its unit of analysis (Barlow and Kemmer 2000: ix). Langacker’s (2001) programmatic contribution about the relationship between grammar and discourse is one of the first CL-studies to show the way to this new field of research. In his Current Discourse Space model, in which he characterizes “discourse as a series of interactive events” (2008a: 460), Langacker (2005a, 2008a) elaborates a dynamic account of language study, according to which meaning emerges as part of the interaction between interlocutors in a conversation. At any point in time, interlocutors are situated in a usage event, in which they are tuned toward each other, constantly engaged in a mutual process of reaction and anticipation. Even in situations where no other interlocutors are directly involved, as in writing a newspaper article, producing or interpreting language can always be seen from an interactional perspective, of which the inclusion of previous, current, and future assumed expectations, attitudes, emotions and/or knowledge of one or more other, (un)known interlocutors constitutes an essential part. This view is in line with Harder (1996), who defines linguistic structures as instructions directed at ‘other’ language users, but it is also inspired by H. Clark’s seminal book Using Language (1996), in which he articulates his joint action hypothesis, according to which any language use is essentially a joint activity, in which interlocutors constantly coordinate whatever they produce and interpret in communication. This coordination process, compared by Clark to the anticipatory and reactive moves of two dance-partners, relies on the exploitation of different types of ‘common ground’ (Clark 1996), which are made up of socially or culturally established structures of shared knowledge, experiences, beliefs, attitudes, emotions, etc. and which individual interlocutors
Multimodality in Interaction
assume to be present in each other’s minds. This intersubjective account of language and, more specifically, meaning-making hinges on our unique cognitive ability to take other people’s perspective and to model the mental states of our interlocutors (Tomasello and Racoczy 2003 speak of ‘self-other equivalence’). As such, this view clearly relates to the theory of mind (Whiten 1991, Givo´n 2005), which revolves around our ability to conceptualize thoughts, ideas, emotions, attitudes, beliefs, etc. in other people’s minds (Broˆne 2010: 91–92). In his most recent account of intersubjectivity and in line with Tomasello’s earlier work on the interpersonal alignment of the self-concept, Verhagen (2015: 240) observes that “the ideas of ‘self’ and subjectivity only make sense in the context of some awareness of others and their intentional stances – realizing that one can be the object of another’s attention, and that there are different points of view, one of which is one’s own.” Verhagen (2015: 238) also adds that in coordinating their activities in a communicative situation, interlocutors primarily act as a group, thus constituting a sociocognitive unit in its own right. In Clark’s terms, two dance partners do not merely engage in their artistic act as two separate individuals, bound by their ‘joint attention’ (Zlatev 2008: 227), but as a dance couple instead. Over the last ten to fifteen years, this intersubjective view on language use has increasingly established itself in CL-inspired studies, thus profiling the social dimension of cognitive meaning, which was less pronounced before (however, see Croft 2009b). Compare, for example, Verhagen (2005, 2008b) on intersubjective syntactic constructions, Zlatev (2008) on bodily mimesis, Ka¨rkka¨inen (2006), Haddington (2007) and Du Bois (2007) on stance-taking in interaction, Zima (2013) on heckles in parliamentary debates, Feyaerts (2013a) on the inherent intersubjective nature of creativity, Oben (2015) and Oben and Broˆne (2015) on patterns of interactive alignment, and Verhagen (2015) on the semantic structures of cooperative communication. As such, within the cognitive linguistic framework, one of whose basic tenets concerns the identification of construal mechanisms, a focus shift toward intersubjective processes of meaning coordination does not really represent a reorientation of the paradigm’s research program. It basically concerns a logical elaboration of the long-standing issue about the difference between the subjective and objective construal of a usage event (see, e.g., Brisard 2006, De Smet and Verstraete 2006, Nuyts 2014). Although this intersubjective account of meaning coordination seems to have established itself as the dominant semantic construal mechanism in most of the current CL studies, this conceptual shift does not seem to have triggered an immediate empirical shift toward a systematic analysis of interactional data of face-to-face conversations. Until its empirical turn toward the analysis of usage-based data (Barlow and Kemmer 2000, Bybee and Hopper 2001, Bybee 2006, 2010), CL was mainly concerned with the corpus-based analysis of naturally occurring, written non-interactional data. Quite telling in this respect is the
137
138
FEYAERTS, BRÔNE, AND OBEN
recent appeal addressed by Zima and Broˆne (2015) to finally embrace spoken dialogue as a new area of CL study. This is not to say that until now hardly any study in CL has focused on the analysis of interactional data: far from it. Apart from individual studies that appeared over the last years and which cannot all be listed here, there are some systematic onsets within CL, sometimes well-elaborated already, expressing and pursuing the need to investigate the mechanisms underlying different types of dialogic exchange within the contours of a usage-based paradigm like CL.
9.1.1 Interactional Humor A first illustration concerns a series of cognitive studies focusing on the identification of construal mechanisms in the complex process of meaning coordination in different types of creativity and humor in interaction (compare Veale, Feyaerts, and Broˆne 2006, Broˆne 2008, 2010, 2012, Broˆne and Oben 2013, Feyaerts 2013b, Feyaerts and Oben 2014, Broˆne, Feyaerts, and Veale 2015). These studies use key concepts from the Joint Action Hypothesis as elaborated by Clark (1996) (e.g. common ground and intersubjective meaning coordination, cf. supra) as an analytical starting point. Clark (1996: 96) observes that “when [we] act on the basis of our common ground, we are in fact acting on our individual beliefs or assumptions about what is in our common ground.” In his analysis of so-called staged communicative acts, comprising utterances like irony, lying, friendly teasing, joint comical hypotheticals etc., Clark (1996: 368) further identifies the notion of ‘layered meaning’ as a key concept of his theory of verbal interaction. Accordingly, the primary or basic layer of meaning pertains to the actual communicative situation (the ‘ground’ in Langacker’s terms) between participants in a conversation. On the basis of mutually assumed elements of common ground, interlocutors may, however, decide to establish a secondary layer of meaning, which can only operate relative to and therefore dependent on the primary layer of interpretation. This may be the case, for example, among close friends, estimating their friendship to be strong enough to support a teasing sequence as in (1). (1)
a: b: a,c: b:
How was your driver’s licence exam? I failed Oh come on! And you only scratched two parked cars. . . That’s so unjust! Thanks for your great support, guys!
On the basis of their long-standing friendship, A and C feel confident to mock B’s misfortune, thus producing a staged communicative act, which involves an additional layer of ironic teasing. The irony in the final response by B indicates the success of the move by A and C, which expresses and in the end maybe even strengthens the social relationship
Multimodality in Interaction
among all of them (Feyaerts and Oben 2014). This semantic aspect of layered meaning, which is closely related to or even dependent on a particular type of intersubjectivity, only surfaces through studying naturally occurring interactional data, thus profiling the need to further intensify the investigation of face-to-face interaction. Another illustration of what the study of interactional language use may reveal is the identification of a dynamic semantic mechanism called trumping (Veale, Feyaerts, and Broˆne 2006) or hyperunderstanding (Broˆne 2008), which serves the strategic purpose of successfully coping with the facethreatening effects of interactional adversarial humor. In a dialogue, communicative success hinges on the functional reinterpretation by a second, responding speaker of a so-called key element (Broˆne 2008), which can be any verbal element used by a preceding interlocutor, in order to gain the upper-hand in a vicious-humorous exchange. Consider the exchange in (2), taken from Veale, Feyaerts, and Broˆne (2006: 317ff.), between a company’s chairman and its under-performing CEO, who opens up bragging about himself as he uses the expression two men in its generic meaning to evoke the expected positive concept of a capable and hardworking employee. However, in assigning two degenerate referents to this expression, the chairman skilfully undermines this move and manages to gain verbal and intellectual superiority over his opponent. (2)
ceo: chairman:
I do the work of two men for this company! Yes, Laurel and Hardy. . .
The powerful impact of this response thrives on the chairman’s perfect estimate of both interpersonal and cultural common ground with his interlocutor, thus assuming that both men will acknowledge the chairman’s victory in this duel. The example clearly shows that in the full semantic analysis of interactional sequences, both layered meaning and common ground as aspects of intersubjective meaning coordination play a crucial role.
9.1.2 Interactional Alignment Studies on alignment form a second illustration of a line of research into cognitive mechanisms underpinning conversational language use. The study of this topic originated in psycholinguistics and can be described as the phenomenon by which speakers use the same words and constructions (of whatever type) as their conversational partners. Alignment thus involves copying behavior in which (parts of) the verbal contributions by one speaker are re-used by another. Copying behavior can be very intentional and serve specific communicative goals such as trumping your conversational partner (Broˆne and Oben 2013, Zima 2013) or displaying analogy or contrast (Du Bois 2014); however, most of the time speakers copy their conversational partners automatically and unintentionally.
139
140
FEYAERTS, BRÔNE, AND OBEN
Both psycholinguistic experiments (a.o. Garrod and Anderson 1987, Bock and Griffin 2000, Branigan, Pickering, and Cleland 2000) and corpus linguistic studies (a.o. Gries 2005, Szmrecsanyi 2005, Howes, Healey, and Purver 2010) have repeatedly and convincingly demonstrated that speakers effortlessly and constantly draw on their partners’ verbal production to formulate their own utterances. Interestingly, there is quite some debate on which process(es) underlie the phenomenon of alignment. First, according to Brennan and Clark (1996), alignment can be seen as resulting from interactive grounding: conversational partners temporarily agree on the construal of the topic of discussion. They reach conceptual pacts that are not directly transferable to other conversations. For example, if speakers are discussing the latest car model that Ferrari released, they each have an enormous amount of lexical and grammatical possibilities to refer to that car (the car, the red car, the car that’s red, the sports wagon, the convertible, etc.). Within a conversation, Brennan and Clark argue, speakers reach a temporary agreement on how they label the referent. When engaged in a different conversation, these conceptual pacts have to be re-established, which makes it well possible that one speaker systematically uses ‘the red convertible’ in one conversation, but ‘the car that’s red’ when talking to a different conversational partner. Second, and opposed to Brennan and Clark (1996), Pickering and Garrod in their seminal work on interactive alignment (2004, 2006) argue that alignment does not presuppose shared conceptualization. They claim that alignment is automatic and driven by the mechanistic process of priming. For example, if a speaker uses ‘the red convertible’ during a conversation, this will lead the addressee to activate the lexical items ‘red’ and ‘convertible’ during the comprehension of the speaker’s utterance. Because of this rise in activation, addressees will be more likely to use ‘red convertible’ themselves in subsequent language production; that is, they are primed to do so. It is important to highlight that this priming operation does not involve any other-modeling (as is key in Brennan and Clark’s view). Because of the dynamics of face-to-face conversation, there is a lack of cognitive bandwidth to constantly model your interlocutor’s conceptualizations and beliefs, and therefore the process of priming and activation is “automatic and does not involve a conceptual pact between the interlocutors” (Pickering and Garrod 2006: 221). Third, and more recently, the temporal dimension of synchronization has gained relevance in the study of alignment. Interlocutors not only copy each other’s words and constructions, but they also appear to synchronize their alignment (Dale and Spivey 2006, Fusaroli, Konvalinka, and Wallot 2014, Oben 2015). For example, by using cross-recurrence analyses, Louwerse et al. (2012) show not only that speakers copy each other in terms of discourse markers such as ‘alright’ or ‘well,’ but also that those markers occur with systematic time lags between them. More specifically
Multimodality in Interaction
they found that typically 1.5 seconds after a speaker uses ‘alright,’ the other speaker in the dyad would use the same marker ‘alright.’ In a series of related studies in which computational linguistics and cognitive sciences meet, conversation is claimed to be shaped in a dynamic, selforganizing network (Fusaroli et al. 2012, Fusaroli, Konvalinka, and Wallot 2014, Dale et al. 2014, Paxton et al. 2014, Fusaroli and Tyle´n 2015). From this perspective, face-to-face conversation is perceived as a complex adaptive system of interlocutors that mutually constrain each other’s interactional contributions (Beckner et al. 2009, Tollefsen and Dale 2012, Fusaroli, Konvalinka, and Wallot 2014). Alignment is one of the factors that drives linguistic choices. We use words and constructions partly because our conversational partners do so. That alignment significantly shapes language use can only be analyzed and demonstrated by drawing on empirical conversational data. In accounting for cognitive processes that shape language use, this interactional aspect should be part of the analysis.
9.1.3 Construction Grammar A final illustration of the empirical shift toward the analysis of situated interactional language lies at the very heart of CL and concerns the growing number of studies in Construction Grammar (CxG), which aim at analyzing constructional patterns in spoken language (Fischer 2000a, ¨ stman 2005, Imo 2007) or on the level of face-to-face interacFried and O tion. As such, identifying and analyzing form-meaning pairings in spoken language does not represent a truly innovative extension of CxG as it may result in a usage-based bottom-up description of yet another cognitively motivated language system, a constructicon (or repertoire of entrenched form-meaning pairings), alongside the network of constructions of written language. However, focusing attention on recurrent patterns in interaction takes Construction Grammar beyond the traditional description of symbolic pairings on the level of morphological, lexical, or syntactic structures into the pragmatic realm of socially situated communicative (re-)action. This relatively recent new interest of Construction Grammar reflects a growing consensus in and around CL. This consensus is that the prototypically dialogic nature of language-use, as a coordination process between two or, more interlocutors, ought to be the focus of a usage-based cognitive language theory (compare a.o. Chafe 1994, Clark 1996, Langacker 2001, 2008a, Du Bois 2007, 2014, Broˆne and Oben 2013, Zima 2013, Broˆne and Zima 2014, Nikiforidou, Marmaridou, and Mikros 2014, Zima and Broˆne 2015, Verhagen 2015). Looking at this development from the perspective of CL, this step toward the analysis of form-meaning patterns in ongoing interaction inevitably takes CxG close or maybe even right onto the playground of neighboring Interactional Linguistics (Interaktionale Linguistik), which defines itself
141
142
FEYAERTS, BRÔNE, AND OBEN
through the application of conversation analytic methods to linguistic structures (Couper-Kuhlen and Selting 1996, 2001, Hakulinen and Selting 2005, Deppermann 2006, 2011). As indicated by Deppermann (2011: 208), an increasing number of scholars with a background in CA or Interactional Linguistics have carefully adopted certain aspects of CxG as part of their theoretical framework for the description of grammatical structures. Among them are quite a few European linguists like Peter Auer, Karin Birkner, Arnulf Deppermann, Susanne Gu¨nthner and Wolfgang Imo (for an overview, see Gu¨nthner and Imo 2006, Gu¨nthner and Bu¨cker 2009), along with American scholars like Cecilia Ford (1993, 2004), Barbara Fox (2007), Paul Hopper (1998, 2004), and Sandra Thompson (2002), who feel at home in the discourse-functional paradigm (see also Ford, Fox, and Thompson 2003).1 Broˆne and Zima (2014) present a study which combines a construction grammar approach to interactional language use with the above-mentioned phenomenon of alignment in dialogue. More specifically, they take as a starting point John Du Bois’s recently developed framework of dialogic syntax (Du Bois 2014), which focuses on the processes by which language users in interaction reproduce (parts of) preceding utterances, thus establishing structural parallelisms and conceptual resonances across turns, as in (3), an example from the Santa Barbara Corpus, taken from an early manuscript of Du Bois’s dialogic syntax. One important effect of these structural mapping relations (or alignment patterns, if you will) is the emergence of local routines in dialogue: within the boundaries of ongoing conversations, speakers establish locally entrenched form-meaning pairings or ad hoc constructions. In the case of (3), one of the speakers creatively exploits the conventionalized formula the X of Y as a reference to a suit of cards (as in the king of spades or the queen of hearts) to construe an ad hoc instantiation the king of puppy dogs’ feet. Within the specific dialogic context, language users do not seem to have any trouble arriving at the intended meaning, precisely because the parallel structure between turns forces a mapping between puppy dogs’ feet and a particular suit of cards. Taking insights from Construction Grammar and Cognitive Grammar, Broˆne and Zima argue that the emergence of these grammatical patterns at the micro-level of a single conversation (in this case the semi-fixed phrase the king of) in fact involves the same cognitive mechanisms described for the abstraction of conventional grammatical constructions from usage patterns. (3)
1
(Risk SBC024: 1452.231–1461.613) j e n n i f e r : We’re gonna pa:ss, the king of (Hx) spades.
From the perspective of Interactional Linguistics, as helpful as insights from CxG may be, the reification of interactionally situated patterns as contextually isolated, type-like constructions (on the basis of processes of extension and schematization) remains a problematic reduction of the socially situated communicative action.
Multimodality in Interaction
dan:
King of (0.7) puppy-dogs’ feet. (0.6) @ (1.2) (H) How come you don’t pass the king of: clubs.
9.2 Multimodality 9.2.1 Introduction Along with the increased interest in the investigation of face-to-face interaction, CL has also started to widen its research scope with regard to the different types of information considered to be worthwhile investigating for a usage-based account of language analysis. These information types pertain to the different semiotic channels, both verbal and non-verbal, which are involved in any usage event and whose complex interplay contributes to the intersubjective process of meaning coordination. These semiotic channels range from lexical, morphological, and syntactic expressions over prosodic features to gestures, gaze behavior, facial expression, and other non- or paraverbal stimuli. Taken together, these stimuli provide the essential components of a so-called multimodal account of meaning.2 Ever since the inclusion of discourse as a new and challenging research dimension for Cognitive Grammar, Langacker has repeatedly stated the importance of any aspect of a usage event as a potentially relevant element in the complex process of meaning coordination among interlocutors. Indeed, it does make a big difference, whether an expression like Oh my God! is being uttered without any specific gesture or striking intonation, or along with a gesture of slapping the right hand’s palm to one’s forehead and maybe simultaneously shaking one’s head, or bringing both hands up to one’s face, and vertically closing them around one’s mouth and nose with both eyes wide open, or to pronounce the utterance with clear pauses between each word, or to merely articulate the words without making a sound thus pretending to shout, etc. Some of these multimodal constellations may fit a usage event of expressing disbelief and profound disappointment (our colleague missing a crucial penalty in a soccer game); others might better fit the expression of disbelief and profound joy (we have won the lottery); still others may express a meaning of true worship and so forth. Verbal
2
The notion of ‘multimodality’ is generally applied to two types of interaction between different semiotic modes of representation. On the one hand, accounts like Forceville and Urios-Aparisi (2009) and Veale, Feyaerts, and Forceville (2013) cover multimodality as the interaction between text and image as in adverts, comics, animation films etc. On the other hand, there is an increasing interest in the account of multimodality as the multiplicity of aspects of face-to-face interaction. In this contribution we restrict our focus to the latter interpretation.
143
144
FEYAERTS, BRÔNE, AND OBEN
expression, gestures, posture, facial expression, intonation, and prosody all play a role in the realization of the intended meaning because, as Langacker states, “Cognitive Grammar takes the straightforward position that any aspect of a usage event, or even a sequence of usage events in a discourse, is capable of emerging as a linguistic unit, should it be a recurrent commonality” (Langacker 2001: 146, original emphasis). Also in his later work, Langacker explicitly advocates a multimodal view of the required analysis of the inherent complexity of a usage event: A discourse comprises a series of usage events: instances of language use in all their complexity and specificity . . . An event is bipolar, consisting on both conceptualization and means of expression. On the expressive side, it includes the full phonetic detail of an utterance, as well as any other kinds of signals, such as gestures and body language (conceivably even pheromones). Conceptually, a usage event includes the expression’s full contextual understanding – not only what is said explicitly but also what is inferred, as well as everything evoked as the basis for its apprehension. (Langacker 2008a: 457ff., original emphasis)
Langacker further specifies his point indicating that “not any gesture (or pheromone. . .) constituting the usage event automatically qualifies as part of the linguistic structure” (quoted in Zima 2014: 3), adding: It does not however follow that every gesture (that accompanies speech) has linguistic status . . . A structure per se qualifies as an element of a language just to the extent that it is entrenched in the minds of the speakers and conventional in the speech community . . . [However,] if a gesture is both familiar and conventional in a speech community, bearing a systematic relation to the expressions it occurs in, its exclusion from the language would be arbitrary. (Langacker 2008b: 250ff.)3
Compared to neighboring fields such as psychology-based gesture studies (McNeill 1992, 2005, Tversky et al. 2009, Parrill, Bergen, and Lichtenstein 2013, Quinto-Pozos and Parrill 2015), communication studies and anthropology (Streeck 2009, who also includes a praxeological and conversation analytic approach), or Conversation Analysis, where already in 2013 a special issue about ‘Conversation Analytic Studies of Multimodal Interaction’ (Deppermann 2013) was released, Langacker’s call to step up to the empirical and methodological challenge of a multimodal analysis has not been answered wholeheartedly by CL yet. It is clear, however, that along with the growing interest in spoken and dialogic interaction, a steadily growing number of scholars in and around CL include acoustic
3
To illustrate his point, he describes another scene: “When a baseball umpire yells Safe! and simultaneously gives the standard gestural signal to this effect (raising both arms together to shoulder level and then sweeping the hands outward, palms down), why should only the former be analyzed as part of the linguistic symbol? Why should a pointing gesture not be considered an optional component of a demonstrative’s linguistic form?” (Langacker 2008b: 250).
Multimodality in Interaction
and/or visual cues of the usage event in their analysis of the process of meaning coordination. This trend is also catalyzed by technological and methodological innovations such as lightweight, unobtrusive, and highly performative camera systems. Among the earliest multimodal studies in CL are Mittelberg (2007) on methodological issues of integrating gesture data with the analysis of speech, Sweetser (2007) and Hougaard and Oakley (2008) on the use of cospeech gestures in marking mental spaces, and Cienki and Mu¨ller (2008a) on the relation between metaphor and gesture. A scan through the issues of the journal Cognitive Linguistics from 2005 onwards generates a somewhat limited series of studies, most of which primarily investigate the use of co-speech gestures as markers of conceptual and grammatical aspects in different types and settings of language use. These types and settings of language use include syntactic subjects (Parrill 2008), temporal aspects (Casasanto and Jasmin 2012), manner (Brown and Chen 2013), conceptual metaphor (Chui 2011), and event representations (Parill, Bergen, and Lichtenstein 2013). In a recent contribution, Kok and Cienki (2016) present an elaborated study on convergences and common challenges, which strongly unite Cognitive Grammar and gesture studies in establishing a multimodal research perspective on the complex nature of language use. In the field of language acquisition, Gullberg and Narasimhan (2010) investigate the way in which gestures mark semantic distinctions in Dutch children’s placement verbs. Finally, Ibbotson, Lieven, and Tomasello (2013) do not focus on gesture, but analyze eye gaze as a cueing event in the attention-grammar interface. Beyond the limits of this journal, several other studies have appeared as well, in which a multimodal approach to the analysis of meaning was adopted. The vast majority of these focus on the role of co-speech gestures in the realization of different (aspects of) construal mechanisms in noninteractive usage events.4 In the context of this chapter, our particular interest concerns studies on the role of gestures and other semiotic markers in the complex process of meaning coordination in face-to-face interactions (cf. Bavelas et al. 1992, Wagner, Malisz, and Kopp 2014, Lu¨cking et al. 2013, Cienki 2015 among others).
9.2.2 Multimodal Construction Grammar In a very recent development, researchers have inquired into the potential extension of Construction Grammar and Cognitive Grammar beyond purely verbal form-meaning pairings. Research in gesture studies has 4
There is also a growing number of studies on the use of co-speech gestures, which mainly originate from the field of gesture studies aiming at a bottom-up description of what may be categorized as a gesture grammar. Many of the scholars involved in this endeavor also take part in the CL paradigm: Alan Cienki, Irene Mittelberg, and Cornelia Müller just to name a few. For a more encompassing account of studies on the interplay between speech and gesture, see elsewhere in this volume.
145
146
FEYAERTS, BRÔNE, AND OBEN
pointed at relatively entrenched associations of verbal and kinesic structures with conventionalized semantics, which raises the question of to what extent such multimodal pairings might be considered as multimodal constructions. In other words, can recurrent combinations of verbal and kinesic patterns be viewed as multimodal constructions, or are they simply to be treated as multimodal manifestations of essentially monomodal (i.e. verbal) constructions? A possible answer to this question can be found in the very foundations of symbolic accounts of grammar, viz. that “any aspect of a usage event, or even a sequence of usage events in a discourse, is capable of emerging as a linguistic unit, should it be a recurrent commonality” (Langacker 2001: 146). If recurrence is the essential criterion for acquiring unit status, then units beyond the traditional scope of grammatical analysis could or should be part of language users’ constructicon. In previous research, this idea was applied to larger discourse patterns, ¨ stman (2005) and O ¨ stman and Trousdale (2013). These including O accounts use insights from Construction Grammar to describe how knowledge of genre structures and grammatical knowledge interact to shape the ¨ stman 2005: 131). In another “constructional peculiarities of discourse” (O line of research, more directly relevant to the present chapter, the form side of the construction is not extended in terms of larger discourse chunks, but rather in terms of the semiotic channels. Again, this extension is motivated by some of the root principles from Langacker: Less commonly recognized as part of language is the gestural channel, including manual gestures, facial expression, and bodily posture. These are, however, subject to conventionalization and coordination with other linguistic processes. (Langacker 2008a: 462)5
Explorations of these conventionalization and coordination processes across semiotic modes have revealed some interesting candidates for what could be called multimodal constructions or conventionalized form meaning-pairings across semiotic modes. Among the phenomena studied in what is to date still a small body of studies (most of which are discussed in more detail in Zima 2014), are modal particles in German (which tend to co-occur with specific hand and head movements, Schoonjans 2014a, 2014b, Schoonjans et al. 2016); motion constructions in English, and more specifically the [V(motion) in circles] construction and the [all the way from X PREP Y] construction, which exhibit strong co-occurrence patterns with specific hand gestures in spoken language (Zima 2014); existential constructions in German (the es gibt-construction, Mittelberg 2014); constructions combining past tense with a proximal deictic, as in It was now possible to . . ., which essentially revolves around a blend of viewpoints (narrated versus experienced, Nikiforidou 2012; and Steen and Turner 2013); markers of obviousness in Dutch and German (counterparts of the 5
In a similar vein Langacker (2005a: 104) argues that “the form in a form-meaning pairing is a specifically phonological structure. I would of course generalize this to include other symbolizing media, notably gesture and writing.”
Multimodality in Interaction
English of course, well, just), which frequently co-occur with headshakes, tilts, and shoulder shrugs (Mangelschots et al. 2017, in prep). Depictions of instrumental actions in spoken language (e.g. actions of cutting and breaking), which according to Sambre and Broˆne (as reported in Schoonjans et al. 2016) exhibit an interesting relationship between the degree of specificity of the verbal predication and the frequency of co-speech gesture: the more schematic the verbal prediction (e.g. when speakers use a generic predicate to do to refer to an instrumental action), the more likely it will be accompanied by a representational gesture. In almost all of the above-mentioned studies, the authors stress the systematicity of the multimodal co-occurrences in corpus data as an argument for their construction status. This, however, raises an important issue, as the frequency argument is a notoriously complex one for Construction Grammar in general, and not only for multimodal co-occurrences (Schoonjans and Zima 2014, Schoonjans, Broˆne, and Feyaerts 2015). How do we determine the minimum frequency of occurrence for constructions to emerge? What is the base rate, both for monomodal and multimodal constructions? In the reported studies, the co-occurrence rates of verbal and gestural patterns range in frequency, depending on the perspective one takes, but they (obviously) never reach a 100 percent match. Where one needs to draw the line, however, still is very much a point of debate and is considered by some to be impossible to operationalize (cf. also Traugott and Trousdale 2013: 11).
9.3 Standing at the Crossroads: Multimodal Interaction Analysis In the previous sections, we have described the increasing interest of scholars in CL for both multimodal and interactive dimensions of a usage event. In this section, we present a few case studies in which both perspectives are linked to each other, thus demonstrating the complex trade-off between different types of semiotic markers in the dynamic process of interactive meaning coordination.
9.3.1 Multimodal Interaction Management One of the key features of interactional language use is its sequential organization in successive speaker turns, a process which requires a high degree of synchronized behavior across interlocutors. Early work in conversation analysis (e.g. Schegloff 1972) already pointed at the intrinsic multimodal nature of dialogue management, but nevertheless, the focus has primarily been on (para)verbal signals guiding the speech exchange system (cf. Schmitt 2005 for a review). Only recently has the role of other semiotic channels (gesture, posture, gaze) been studied more systematically. Most
147
148
FEYAERTS, BRÔNE, AND OBEN
notably, there is a growing interest in the emergence of composite signals (Clark 1996, McNeill 2006) involving multiple channels jointly contributing to dialogue management. A key feature is the sequential organization of the different channels, as illustrated, for example, in Mondada (2007) for pointing-gestures as a resource in establishing speakership. In a recent cognitively inspired study, Broˆne et al. (2017) focus on the role of eye-gaze patterns in relation to other signal systems in dialogue management. Using a corpus of two-party and three-party face-to-face conversations, in which the gaze behavior of all participants was recorded with mobile eye-tracking technology (the InSight Interaction Corpus, Broˆne and Oben 2015), this study focused on patterns of gaze distribution across speakers and hearers. More specifically, the exploratory analysis yielded two significant results. First, interlocutors seem to exhibit specific patterns of gaze distribution, related to the temporal organization of dialogue acts (such as turn-holding, turn-giving). Different dialogue acts typically display specific gaze events at crucial points in time, as, for example, in the case of brief gaze aversion (200–400 ms) associated with turn-holding, and shared gaze between interlocutors at the critical point of turn-taking. Second, the analysis shows a strong correlation and temporal synchronization between eye gaze, speech and gesture in the realization of specific dialogue acts, as shown by means of a series of cross-recurrence analyses for specific turn-holding mechanisms (e.g. filled pauses such as uh and uhm cooccurring with brief gaze aversion). In general, the results of this study illustrate (i) the interactive function of eye gaze as a communicative instrument (next to its relevance as an index of comprehension processes), and (ii) recurrent patterns in the physical co-occurrence, functional trade-off and interactive set-up of multiple modes of representation (in this case verbal markers and gaze events) in face-to-face settings. Linking back to section 9.1.3, these patterns may again be indicative of the existence of multimodal discourse constructions in a more broadly defined Construction Grammar.
9.3.2 Multimodal Alignment That lexical (Brennan and Clark 1996, Pickering and Garrod 2004, Manson et al. 2013) and syntactical (Branigan, Pickering, and Cleland 2000, Gries 2005, Reitter, Moore, and Keller 2006) alignment frequently, and more often than can be conditioned only by chance, occur has already been pointed out in section 9.1.2. Interlocutors also align their conversational behavior at non-verbal or para-verbal levels. For example, when people interact, they copy each other’s prosodic features (among many others, see: Webb 1972, Gregory and Hoyt 1982, Giles, Coupland, and Coupland 1991, Stanford and Webster 1996, Collins 1998, Szczepek Reed 2010, Lewandowski 2012). This was demonstrated both within very strict
Multimodality in Interaction
experimental settings of shadowing tasks6 (e.g. Fowler et al. 2003), in more natural settings of targeted collaborative tasks (e.g. Pardo 2006) and spontaneous face-to-face conversations (Kousidis et al. 2009, Vaughan 2011, De Looze et al. 2014, Oben 2015). Interlocutors also align in terms of the (hand) gestures they use. This happens especially during explicit grounding or meaning negotiation and co-construction (Kimbara 2008), for the gesture features of representation technique, handshape, palm orientation, and handedness (Bergmann and Kopp 2012), when interlocutors have to perform a more difficult task (Louwerse et al. 2012) or simply talk longer to each other (Oben 2015). Even though alignment has been studied at different multimodal levels, only very recently have researchers begun to integrate these levels into analyses that cover several levels at the same time. Louwerse et al. (2012) combine alignment at the different levels discussed so far: lexical, syntactic, and gestural. They show that synchronization increases over time and over task difficulty. The longer participants talk to each other and the more cognitively demanding their task, the more they synchronize their aligned behavior. Moreover, synchronization also increases over a social dimension they controlled for. In the experimental set-up, participants took turns in either giving or receiving instructions to navigate through a complex map. For most of the behaviors under scrutiny, the instructionfollower copied the instruction-giver significantly more often than vice versa. This hints at the importance of conversational role in the directionality of alignment. Better informed or dominant conversational partners align less than their less informed, less dominant partners. Also Oben (2015) and Oben and Broˆne (2015) present a number of case studies in which conversational behavior at different semiotic channels is analyzed at the same time. In a first study they show how gaze has an effect on alignment. For lexical alignment they found that if a speaker utters a word while looking at the addressee, that addressee is significantly more likely to use that same word in a subsequent utterance (compared to when the speaker does not look at the addressee while speaking). The gaze behavior of the addressee did not appear to play a role in explaining lexical alignment. For gestural alignment, no such gaze effect of speaker or hearer looking at each other was found. What did matter was whether or not the gestures themselves were fixated: if an addressee fixated a hand gesture (i.e. focusing on it for 200 ms or more), he was significantly more likely to use that same gesture in subsequent language production, compared to if he had not fixated the gesture. Together these case studies show that gaze has a different effect on lexical than on gestural alignment. That different factors explain alignment at different levels was confirmed in a third case study in which a number of factors were checked for their ability to 6
In shadowing tasks, participants typically get to hear a syllable, word, or word group, and are then asked to repeat as fast as possible. Researchers then study the phonetic similarity between prime and response.
149
150
FEYAERTS, BRÔNE, AND OBEN
account for alignment. For lexical alignment, cumulative priming appeared to be the only significant factor: the more often a speaker uses a word, the higher the likelihood that the other speaker will use that word as well. For gestural alignment the factors distance and block proved to be significant: if two gestures are performed at the same time it is very likely those two gestures are identical; and the longer two interlocutors are engaged in interaction, the higher the chances of gestural alignment are. What these case studies demonstrate is that a multimodal approach is a valid if not an essential gateway into a better understanding of conversational phenomena such as alignment. The multimodality here does reside not only in the fact that multiple semiotic channels are considered in the analyses, but also in the fact that the interplay and relation between behavior at those multiple levels is the explicit topic of investigation.
9.3.3 Stance Taking: Expressing Obviousness In this final case study, we focus on the interplay between verbal and gestural markers as reported in empirical studies on the expression of obviousness. More specifically, our description concerns the role of cospeech gestures in utterances with a particular German modal particle (einfach) expressing obviousness. The use of modal particles such as ja, denn, doch, mal, halt, and eben counts as a typical characteristic of spoken German in an interactive setting.7 These lexical elements are “used to add an (inter)subjective nuance to the utterance, in that they indicate the speaker’s position vis-a`-vis the content of the utterance, the expected hearer reaction, or the relation of the utterance to the context” (Schoonjans 2013a: 1).8 In its downtoning function, the particle einfach, which roughly paraphrases as ‘simply’ or ‘just,’ is often used in assessments, where it marks the speaker’s evaluation that they consider the content of the proposition as obvious or inevitable (Thurmair 1989, Spreckels 2009, Schoonjans 2013a) as in (4) and (5). (4) (5)
Es ist einfach unfassbar, was hier passiert. [It is simply incomprehensible what is happening here.] Ich habe das einfach fu¨r mich gemacht, weil ich Spaß dran hatte. [I just/simply did it for myself, because it was good fun.]
As far as their formal characteristics are concerned, modal particles are inflective, are unstressed, cannot be questioned, intensified, coordinated or negated and, prosodically and syntactically, are well integrated in the surrounding utterance. Formally they make an almost punctual
7
For an extensive and recent overview of this phenomenon, including its multimodal realizations, we refer to Diewald 2007 and Schoonjans 2014a.
8
The stance of nuancing is referred to as “downtoning” by Waltereit (2006).
Multimodality in Interaction
appearance, but quite importantly, their functional scope or, for that matter, their ‘meaning’ stretches over the turn-constructional unit (TCUs) or even the entire turn (Schoonjans 2014a, Schoonjans 2014b: 91–92). With regard to a multimodal study of utterances with einfach, Schoonjans, Broˆne, and Feyaerts 2015 analyzed about seventeen hours of face-to-face communication as recorded from talk shows and sports interviews on television. Based on the observation that some TCUs with einfach co-occur with a headshake by the speaker, a systematic scan through the entire data set9 revealed an increased co-occurrence of einfach and non-emblematic headshakes within the same TCU.10 With their distinction between the rather pronounced emblematic headshaking, which traditionally in Western cultures signals some sort of negation, on the one hand, and more subtle non-emblematic, but pragmatic headshakes on the other, Schoonjans, Broˆne, and Feyaerts (2015) align with Kendon (2002) and McClave (2000) who already made a similar distinction for English. In the study reported here, of the 115 relevant tokens of einfach in assertions,11 twenty-seven (23.48 percent) were accompanied by a pragmatic headshake (Schoonjans, Broˆne, and Feyaerts 2015: 297). If one narrows the collection of tokens down to only those cases in which einfach is accompanied by a gesture, the frequency rises to 62.79 percent (twenty-three of forty-three tokens), which means that if a TCU with einfach is accompanied by a co-speech gesture, it is likely that this gesture will be a headshake (Schoonjans, Broˆne, and Feyaerts 2015: 297). As for the question of why the modal particle einfach tends to co-occur with a headshake and not, say, a nod, the pragmatic, non-emblematic meaning of the headshake, which may be paraphrased in terms of assessment, intensification, and exceptionlessness, clearly involves overlapping features with the meaning of obviousness of the modal particle einfach (Schoonjans 2014a, Schoonjans, Broˆne, and Feyaerts 2015: 299).12 Accordingly, the meaning of the headshakes in the context of this modal particle pertains to the illocutionary level of the utterance and can be paraphrased as ‘I don’t know what else could be said or argued’ or ‘I don’t see any other way of putting it’ etc. (Schoonjans, Broˆne, and Feyaerts 2015: 299).
9
This procedure consisted of two rounds. First, all three authors separately annotated all TCUs with einfach on the presence of a headshake, which resulted in an initial inter-coder agreement of slightly more than 80 percent. After that, all coders looked at the remaining cases for which no agreement was reached. In this second round, it turned out that almost all disagreements were due to highly subtle headshakes, which had passed unnoticed to at least one of the coders. In the end, however, after careful analysis among all three coders, almost all of these cases were agreed upon.
10
Schoonjans, Brône, and Feyaerts (2015) side with Kendon (2002: 149), who describes headshaking as “whenever the actor rotates the head horizontally, either to the left or to the right, and back again, one or more times, the head always returning finally to the position it was in at the start of the movement.”
11
The modal particle einfach also occurs in questions and requests; in this study, we only report on occurrences in
12
In this respect, see Schoonjans (2013b) on the grammaticalization of gestural meaning.
assertions.
151
152
FEYAERTS, BRÔNE, AND OBEN
Example (6) is taken from Schoonjans, Broˆne, and Feyaerts (2015: 298) and offers a sequence produced by a former biathlon champion who is admiring the first major victory of one of his young countrymen. Within these four TCUs, the expert uses einfach no less than three times, each time accompanied by a headshake, the beginning and ending of which being marked by ‘//’. (6)
1192 1193 1194 1195 [1192 1193 1194 1195
und a¨h christoph stephan hAt //das einfach nur verDIENT und;// //h. einfach KLAS//se. mir fa¨llt s jetzt wirklich SCHWER da gefasst zu sein und einfach zu analysIEren. also war //war einfach PHA¨nomenal.// and a¨h christoph stephan simply has deserved it and simply great I find it really difficult now to be composed and simply analyse well was was simply phenomenal.]
With regard to the meaning expressed by both the verbal and the gestural markers, the transcription clearly shows that the gesture sets on somewhat earlier than the occurrence of einfach. According to (McNeill 2005: 35), this is mostly the case when comparing ‘equivalent’ verbal and gestural markers. Most interesting, however, is the observation that the scope of the headshake also stretches beyond the realization of the modal particle covering large parts of the intonation unit. This observation on the extended scope of the gesture links its functionality (or meaning) to the meaning of a modal particle, whose scope typically covers (parts of) the entire utterance (Diewald 2007: 127). In this respect, Schoonjans, Broˆne, and Feyaerts (2015) also report a number of TCUs without einfach, but with a pragmatic headshake. This indicates that in expressing the complexity of a usage-event, multiple trade-offs between verbal and gestural markers seem to be possible, including: a) a combined and therefore truly multimodal expression of verbal and gestural markers, b) merely verbal markers; or c) merely gestural markers. Future research may further uncover the peculiarities and conditions of these onomasiological options. Even if at first sight, the frequencies presented here may not look too impressive, they do give us an empirically well-founded indication of a multimodal cooccurrence pattern involving both (a lexical equivalent of) this modal particle einfach and a pragmatic headshake. This co-occurrence is a good candidate for qualifying as a multimodal construction (see section 9.2.2).
9.4 Another Recontextualization? In this contribution, we have made our case for including the analysis of interactive usage events as well as a multimodal research perspective in the CL-research program. In the case studies presented here, we have
Multimodality in Interaction
shown that both perspectives, the intersubjective process of meaning coordination in interaction on the one hand, and the complex trade-off between different semiotic channels constituting a usage event on the other, strengthen each other. There can be hardly an adequate analysis of an interactive usage event without including the multimodal range of information channels, while a multimodal perspective on the process of meaning coordination almost automatically deals with (face-to-face) interaction. With respect to the increasing interest in CL to analyze interactional data, we side with Croft’s (2009b: 395) programmatic appeal that “[i]n order to be successful, cognitive linguistics must go ‘outside the head’ and incorporate a social-interactional perspective on the nature of language.” Regarding the aspect of multimodality, CL-inspired studies are coming into the fore. Following the early lead taken by gesture studies, most of these studies focus on the exploration of co-speech gesturing, whereas more recently, the multimodal research attention increasingly shifts beyond the realm of gestures into the semiotic complexity of eye gaze, facial expression, body posture and physiological effects. The view that the focus on multimodal interactional data may qualify as a new and ongoing recontextualization in CL is supported by the observation that most of the reference works of CL, which appeared during the last decade and which present a broad overview of the paradigm’s past, present, and future research lines, do not always include a section on (aspects of) multimodal interaction. This nicely illustrates that this empirical and perspectival shift represents a rather young research development in CL (see, for example, Evans, Bergen, and Zinken (2007), Geeraerts and Cuyckens (2010), Brdar, Gries, and Zˇic Fuchs (2011), Littlemore and Taylor (2014), and Da˛browska and Divjak (2015), in which the contributions by Turner (2015) and Verhagen (2015) do draw attention to interactional aspects of meaning and conceptualization). As one of the earliest volumes to report on these new research lines in CL, we refer to Gonzalez-Marquez, Becker, and Cutting (2007) on the use of different methods in CL, where Wilcox and Morford (2007), Sweetser (2007), Mittelberg (2007) and Richardson, Dale, and Spivey (2007) discuss gaze, gesture, and sign. More recently, Borkent, Dancygier, and Hinnell (2013)13 have edited a volume on, among other topics, multimodal communication in relation to different aspects of human creativity. In another recent overview volume celebrating Dirk Geeraerts’s 60th birthday, Zenner et al. (2015: 3ff.) identify five recontextualization tendencies which have characterized both Dirk’s and the community’s study of language and linguistics in CL since its earliest days: “the usage-based hypothesis still stands very strong . . . After a long time span in which language was primarily studied as an isolated phenomenon, 13
Blending or Conceptual Integration Theory does offer a model to analyze the realization of conceptual structures across interactional and multimodal settings.
153
154
FEYAERTS, BRÔNE, AND OBEN
contemporary linguistic research has in recent decades eventually been characterized by several attempts to recontextualize language” (Zenner et al. 2015: 3).14 The authors proceed to list the following five tendencies: 1) the study of language in relation to general principles of cognition, 2) the application of the usage-based hypothesis to the study of lexical semantics and lexical change, 3) the study of semantic processes with respect to grammar, 4) the study of cognitive sociolinguistics, and 5) the methodological requirements and empirical challenges that come with each of these recontextualizations. The study of the interactional nature or the multimodal complexity of language use does not figure among these recontextualizations, but in the same volume, Divjak (2015) looks ahead and raises four methodological challenges for a usage-based account of linguistics, which are expected to set the research agenda of CL in the near future. Two of these challenges deserve closer scrutiny as they pertain to the type of data (annotation) discussed in this contribution.15 In one of the challenges, Divjak (2015: 304ff.) draws attention to the methodological difference as well as to the issue of the ecological validity of studying language in a controlled experimental lab setting versus studying language in a naturalistic usage event, thus warning against overgeneralization in either way. Accordingly, an adequate study of language use requires methodological care from both sides by staying as close as possible to naturalistic (settings of) language use while also attempting to keep control over certain variables. With respect to capturing and understanding the process of interactional meaning-coordination in all its multimodal detail, this challenge directly appeals to scholars of multimodal interaction analysis, inviting them to adopt insights and methods from corpus linguistics to highly naturalistic interactional data. The second challenge concerns a reflection on the possibly worn-off and introspective nature of (most of) the linguistic categories by which we annotate and schematize our data. Divjak (2015: 300) wonders about the benefits of staying closer to the raw data and sides with Wa¨lchli and Cysouw (2012: 703) in stating that “An approach that stays closer to the raw data and captures every possible clue comes with the added benefit that keeping as much detailed information as possible – even throughout advanced stages of analysis – is crucial because we never know if what we believe to be the relevant features really are the only essential ones.” In line with this challenge, we advocate the capturing of multimodal data as possible essential clues in the ‘linguistic’ analysis of an interactional usage event. As we have seen above, a lot of effort is already underway, documented in many case studies, to map the dynamic trade-off between verbal and gestural meaning markers. However, very much in 14
These recontextualizations and the corresponding contributions in the volume are said to be “indicative of the way in which cognitive linguistics has developed and diversified over the past decades” (Zenner et al. 2015: 13).
15
The two other challenges concern a) the problematic overgeneralization of frequentist probabilities in the statistical analysis of our data and b) the issue of testing our analytic models against human behavior (Divjak 2015).
Multimodality in Interaction
line with Langacker (2008a: 457), who suggests that we take into account “any other kinds of signals such as gestures and body language (conceivably even pheromones)”, we interpret this challenge to include, when considered functionally relevant, other non-discrete semiotic markers in the analysis, such as gaze behavior, facial expressions and biopsychological parameters such as heart rate, breathing, etc. A broad, Langackarian perspective on the potential semiotic relevance of any aspect constituting an interactional usage-event is comparable to the empirical grounding which the introduction of ERP and FMRi methods bring to the interdisciplinary investigation of multiple aspects of embodied cognition (see, e.g., Coulson and Van Petten 2002 or Coulson 2008, 2012). This kind of approach will contribute to a sociocognitive account of language. In this vein, any focus on the dialogic and multimodal nature of language use will require new empirical methods and settings, thus widening the parameters of research conducted in CL. As a consequence of this shift in perspective toward including different types of verbal and non-verbal data in empirical research or as empirical objects worthy of research, studying language in use must increasingly become an interdisciplinary and technological matter, involving expertise from psychology, neuroscience, physiology, engineering, statistics and computer science, etc.16 We come to our conclusion. In the very first issue of the journal Cognitive Linguistics, George Lakoff (1990: 40) identifies two commitments, the generalization commitment and the cognitive commitment, by which all cognitive linguists in their endeavor to explore the system and dynamics of language use are bound, or as Divjak (2015: 297) puts it:17 All cognitive linguists are committed (or assumed to be committed) to providing a characterization of the general principles governing all aspects of human language in a way that is informed by and accords with what is known about the brain and mind from other disciplines.
When looking at quite a number of these other (sub)disciplines in the research fields surrounding cognitive linguistics, including conversation analysis, interactional linguistics, psychology, joint action theory, and also interaction studies in biological and artificial systems, cognitive science, and social neuroscience, we believe that social interaction must become the primary object of study because the “experimental quarantine” approach . . . of studying isolated individuals that has guided the majority of cognitive (neuro-)science research has been giving way to studies of social interaction that try to do justice to the complexity of interpersonal cognition and action. (Obhi and Sebanz 2011: 329)
16
Compare, in this respect, Divjak (2015: 305), who “conclude[s] that studying language in use is a discipline at the intersection of linguistics and psychology.”
17
See also Evans, Bergen and Zinken (2007: 4–5).
155
156
FEYAERTS, BRÔNE, AND OBEN
Along with the semiotic dimension of multimodality, advocated so effectively by Langacker, and in the light of both cognitive commitments, CL’s methodological challenges for facing the future, its firm appeals for empirical rigor and the growing number of studies reported here and elsewhere in this volume, we make a strong plea for yet another recontextualization of cognitive linguistics into what may be categorized as multimodal interaction analysis.
10 Viewpoint Lieven Vandelanotte
In the humanities, viewpoint has long been studied in the context of narrative, with a focus on such questions as focalization (e.g. Genette 1980, Bal 2009) and speech and thought representation (e.g. Banfield 1982, Fludernik 1993). In formal linguistics, analyses of logophoric pronouns (e.g. Cantrall 1974, Culy 1997) and so-called empathetic syntax (Kuno 1987) provide early foci of interest, but arguably cognitive linguistics, with its principled consideration of construal choices and viewing arrangements, has been well-placed since its inception to elaborate the notion. However, viewpoint has recently started to take center stage, as witnessed by edited volumes such as Dancygier and Sweetser (2012), Dancygier, Lu, and Verhagen (2016), Dancygier and Vandelanotte (2017a), and Vandelanotte and Dancygier (in prep). This focusing of efforts is owed in part to recent developments in cognitive science and psychology, where the embodied simulation hypothesis was formulated and tested (see, e.g., Bergen 2012), fostering further viewpoint studies across a range of modalities and research paradigms, whether experimental or more theoretical. The pervasive importance of viewpoint offers a challenge and an opportunity to demonstrate the usefulness of tools such as image schemas, frames, mental spaces, and blends to coherently formulate and analyze viewpoint complexities in grammar, discourse, and multimodal artifacts. This chapter proceeds as follows. In its first section, general discussion shows that viewpoint is not only ubiquitous, but also intersubjectively and multimodally construed. The second section looks at specific viewpoint phenomena at different levels of structure in different modes of expression, and suggests ways of integrating multiple local viewpoints into a global viewpoint network. Based on this exploration, the third section proposes to conceive of types of viewpoint along a deictic–cognitive divide.
158
VAN DEL ANOT TE
10.1 General Properties of Viewpoint: Embodied, Multimodal, Intersubjective The potential application of the notion of viewpoint can seem bewilderingly broad, as ultimately any aspect of construal reveals something about the views of the language user: choosing one word over another, one focus over another, one temporal or aspectual form over another, and the list goes on (cf. Langacker 2008a: 55–89). At the same time, the term ‘viewpoint’ etymologically gives pride of place to the visual modality, although not all aspects of viewpoint can be or need to be visually imagined. This reflects the way in which vision is prioritized among the senses (Bergen 2012: 49), and its links with conceptualization and understanding are deeply rooted (Sweetser 1990, Langacker 1995). By way of initial definition, viewpoint will be understood here as “a discourse participant’s alignment with an aspect of a frame or situation” (Dancygier and Vandelanotte 2016: 14). In other words, to speak of viewpoint requires a conceptualizer in a discourse event assuming a position (for instance, in terms of perception and spatiotemporal location, likelihood and knowledge, attitude and feeling, or solidarity/power dynamics) toward an element within a described situation or knowledge structure. Viewpoint is, in this view, a fairly comprehensive notion moving well beyond the spatial and, by extension, temporal, location toward or within a described scene; the advantage of such an inclusive understanding is that it points to a unified conception of low-level phenomena all the way up to viewpoint flows in discourses of different sizes, from Internet memes to whole novels. In analyzing viewpoint, we are trying to get at a discourse participant’s conceptualization via its observable realizations in forms of language and in various forms of embodied behavior. Viewpoint is ubiquitous and inescapable: in Sweetser’s formulation, “we never have experience of the world except as a viewpoint-equipped, embodied self among other viewpointed, embodied selves” (2012: 1). This summing-up focuses attention on both the embodied nature of viewpoint and the fact that it is construed intersubjectively, in a negotiation with other participants in a given speech event. Both concepts are addressed more fully in other chapters of this handbook, but their specific relation to and relevance for viewpoint warrants discussion here. An immediate sense in which viewpoint is embodied pertains to the body articulators’ involvement in expressing viewpoints, as a string of studies on co-speech gesture have been making clear at least since the seminal work by McNeill (1992, 2005), who distinguished between observer viewpoint and character viewpoint in so-called ‘iconic’ gestures (“bear[ing] a close formal relationship to the semantic content of speech,” McNeill 1992: 12). For example, a cartoon prompt involving a skunk hopping up and down in an experiment led by Parrill (2009, 2010) leads some
Viewpoint
participants retelling the cartoon to use their finger to trace the pathway traversed by the skunk (observer viewpoint), whereas others ‘embody’ the skunk, imitating its motion and possibly other (e.g. facial) features (character viewpoint). As many researchers have observed, ignoring co-speech gesture in analyzing spoken discourse leads to impoverished or misleading results, also in terms of viewpoint shifts (e.g. Narayan 2012). Conversely, comparing results may prove to be mutually reinforcing, as in Parrill’s (2012) study showing that given versus new discourse status, whose relevance to linguistic referring expressions has long been established, similarly impacts on gesture choice: in the absence of shared knowledge, speakers produce significantly more gestures, and significantly more character viewpoint gestures. Another such correlation is that between transitivity and gestural viewpoint, with transitive and intransitive utterances showing a significant preference for character and observer viewpoint utterances respectively (Beattie and Shovelton 2002, Parrill 2010). Unlike the linear processing of spoken or written linguistic structure, the visual presence of a whole body in spoken interaction affords a simultaneous presence of different viewpoints in what is known as body partitioning (Dudis 2004), whereby different sections of the body can express different viewpoints. For instance, Sweetser and Stec (2016) describe cases where gaze and facial expression represent one viewpoint, and the speaker’s hands and body simultaneously represent another. While “single-viewpoint gestures are the most frequent” (Stec 2012: 333), multiple viewpoint gestures have been recorded and discussed for both cospeech gesture (e.g. Parrill 2009, Sweetser 2013) and sign language (e.g. Janzen 2004). As the example of multiple viewpoint gestures shows, viewpoint embodiment involves more than mere manual gesturing, and detailed analyses of interactions need also to consider such parameters as body posture, gaze and facial expressions, head tilts and nods and, at the aural level, “vocal gestures such as pitch, intonation and accent” (Stec 2012: 332), all of which can help to effect shifts in viewpoint (see Stec 2012 for a review). Current work is starting to address the finer points of these other parameters, for instance in studying the role of head nods and tilts and of facial expressions in sarcasm (Tabacaru 2014) or in describing the ways in which gaze performs viewpoint functions such as enacting characters, visually ‘checking’ that interlocutors are engaged and ‘on board,’ or signalling that speakers are ‘consulting’ their memories (Sweetser and Stec 2016). While sign languages such as ASL or BSL use broadly similar viewpoint strategies to co-speech gesture (Stec 2012, Quinto-Pozos and Parrill 2015), embodied viewpoint in them arguably is more structured, and “clearly more conventionalised” (Earis and Cormier 2013: 340). As an example, consider types of viewpoint shift in ASL. The default type shows some combination of body shifting, pointing and/or eye gaze directed toward
159
160
VAN DEL ANOT TE
some static location to mark whose viewpoint is in focus when comparing characteristics of different referents (Janzen 2012b). In the special case of mentally rotated space (Janzen 2004, 2012b), on the other hand, the signer rotates the other viewpoint mentally to their own position. In one example discussed by Janzen, a signer, in explaining how she was signalled by the police to pull over, switches, across two consecutive constructions, from a clear leftward moving gesture (adopting the police officer’s viewpoint) to a rightward move (adopting her own again). In addition to ‘outwardly perceivable’ viewpoint embodiment, there is now a rich literature on embodied cognition focusing on phenomena in the brain, whose effects are demonstrated in experimental work. An accessible introduction covering this field is Bergen (2012); here I only highlight some of the research showing that and how “in understanding language, we use our perceptual and motor systems to run embodied simulations” (Bergen 2012: 195). Various picture-sentence matching tests have shown that people respond faster when an entity’s orientation or shape in the picture matches the implied orientation or shape in the sentence prompt in examples such as The carpenter hammered the nail into the floor versus into the wall or The ranger saw the eagle in the nest vs. in the sky (Stanfield and Zwaan 2001, Zwaan, Stanfield, and Yaxley 2002). This suggests that we visually simulate objects we read or hear about, and there is good evidence that this simulation is viewpointed. More importantly, the viewpoint adopted in simulations can be manipulated by language. For instance, in one experiment (Yaxley and Zwaan 2007), the sentence prompts were Through the clean goggles, the skier could easily identify the moose versus Through the fogged goggles, the skier could hardly identify the moose; remarkably, people were quicker to identify a high-resolution picture of a moose following the ‘clean’ sentence and a low resolution picture in the ‘fogged’ case. Understanding language really does seem to be “almost like being there,” in Bergen’s phrase, with simulations occurring in language comprehension adopting the viewpoint of an immersed experiencer (Bergen 2012: 66). In addition to words, grammatical structures such as argument structure constructions and grammatical aspect influence different facets of simulations (Bergen 2012: ch. 5). Of particular interest is the influence of grammatical person: second-person sentences have been shown to prompt internal, participant viewpoint; third-person sentences external, observer viewpoint (Brunye´ et al. 2009). For instance, in an experiment run by Bergen and his lab (2012: 110–13), reaction times were quicker when an ‘away’ visual prompt (showing two images of a baseball of which the second is smaller) followed a second-person sentence (you rather than the pitcher threw the baseball to the catcher). To describe viewpoint as fundamentally embodied implies describing it as fundamentally multimodal as well: combining spoken language with
Viewpoint
co-speech gesture, for instance, naturally relies on both the vocal/auditory modality and the kinesic/visual one (Green 2014: 9). An impressive case study of multimodality in this strict sense is Green’s (2014) analysis of Central Australian sand stories, in which spoken language and song are supported by co-speech gesture, sign language, and sand drawings, the latter using both graphic schemas and ‘props’ such as leaves representing story characters. In a looser sense, however, a form of multimodality can also be posited for communicative artifacts that use different semiotic resources, even if, for instance, they are all visuo-spatial, as in combining text and image in comics or advertising. There is a rich tradition of communication research centered on this broader conception of multimodality (e.g. Kress and Van Leeuwen 2001), and in cognitive linguistics the question has been focused on multimodal metaphor (Forceville and UriosAparisi 2009) and multimodal interaction (Feyaerts, Broˆne and Oben this volume Ch. 9). One recent study in this vein shows how a variety of Internet memes serve constantly to reconstrue evoked viewpoints, making effective use of the (semi-)constructional status which linguistic and visual elements alike acquire (Dancygier and Vandelanotte 2017b). A final general property of viewpoint to be highlighted here is its intersubjective nature. At first glance, this is common sense: suppose you are teaching in a classroom, for instance, you take into account the fact that what you see in front of you on your computer screen students see projected behind you. We are constantly aware of other viewpoints. Less commonsensically, perhaps, we appear to be hard-wired to monitor and, in an attenuated sense, even simulate other humans’ actions around us, in the sense that some of the brain structures used to perform a given action are activated when we “recognize, predict and understand” (Bergen 2012: 77) someone else carrying out that action. Research on monkeys’ so-called mirror neurons (e.g. Rizzolatti, Fogassi, and Gallese 2001) made scientists alive to this reality; more indirect brain imaging evidence suggests humans too operate with mirror systems (e.g. Aziz-Zadeh et al. 2006). Converging trends in a string of related fields of inquiry have demonstrated the pervasive significance of intersubjectivity. In both phylogeny and ontogeny, prelinguistic intersubjective skills such as imitation of actions, joint attention and pointing ‘ground’ language, to use Zlatev’s (2008) term, and more complex, linguistic intersubjective skills continue to develop as language evolves. For instance, Tomasello (1999: 173) lists disagreements, repairs/explanations and meta-discourse as types of discourse in which children need to be able to model another’s viewpoint; at the more basic level of selecting lexical items (e.g. cat versus animal), children appear to take into account which viewpoint is pragmatically being cued for (Clark 1997). Children’s acquisition of advanced ‘theory of mind’ skills (such as false-belief understanding or pretence) further illustrates that being able to share and understand the experiences of others underlies much of what we do.
161
162
VAN DEL ANOT TE
In linguistics, intersubjectivity has inspired work in various fields. In diachrony, for instance, much work has focused on changes from less or non-subjective, to subjective and possibly intersubjective, taking into account the addressee’s viewpoint (e.g. Traugott 2010); in linguistic typology, a wide range of multiple viewpoint constructions are being analyzed (e.g. Evans 1995); and in construction grammar, constructions such as complementation, connectives, and sentential negation have been shown to have particular intersubjective orientations, in the sense of serving to coordinate mental states between conceptualizers (Verhagen 2005). It is to viewpoint constructions and how best to analyze them that the next section turns.
10.2 Tools for Analyzing Viewpoint Networks This section focuses on new contributions that Cognitive Linguistics has made in the study of viewpoint in terms of range of phenomena studied and of analytical tools used. Cognitive linguists have contributed to existing research programs on narratorship and focalization (e.g. Bal 2009, Hu¨hn, Schmid, and Scho¨nert 2009b), speech and thought representation (e.g. Banfield 1982, Fludernik 1993), and deixis and evidentiality (e.g. Hanks 1992, Aikhenvald 2004, Evans 1995). At the same time, they have also revealed the viewpoint potential of hitherto neglected constructions; they have brought into focus the manifold interactions between viewpoint and image schemas, frames and mental spaces, on which metaphorical and metonymic mappings are built; and they have ventured beyond sentence-level interpretations to analyze the management of viewpoints within smaller and larger discourses. Speech and thought representation was interpreted in terms of mental space embedding in Sanders and Redeker (1996), with the embedded space being accessed directly, as a new Base Space, in direct speech, and only indirectly, with varying degrees of ‘narrator’ influence, in (free) indirect speech/thought; they also distinguish forms of implicit viewpoint (cf. Sanders 2010). The specific kind of blend involved in the free indirect mode was characterized by Vandelanotte (2004, 2009, 2012) as locating viewpoint with the represented speaker (or character), in contrast to a separate type, distancing indirect speech/thought, in which the current versus represented speaker discourse blend locates viewpoint with the current speaker and is decompressed to construe a higher (attitudinally ‘distanced’) space from which the blend is ultimately viewed. An example is (1), in which the I-narrator Max is questioned by Carlo Grace about how he and his parents spend their vacation in a chalet; the we corresponds to an original addressed you, drawn into the deictic viewpoint of the current speaker, who incorporates Carlo’s discourse without shifting the deictic centre.
Viewpoint (1)
Now and then from the other end of the table Carlo Grace, chewing vigorously, would bend a lively gaze on me. What was life like at the chalet, he wanted to know. What did we cook on? A Primus stove, I told him. “Ha!” he cried. “Primus inter pares!” (Banville 2005: 208; italics original)
This work on discourse representation phenomena broadly assumes a notion of “discoursally motivated” (Nikiforidou 2012: 193) construction. The reliance particularly of blended forms such as free and distancing indirect speech/thought on the narrative context clearly sets them apart from more readily identifiable constructions such as morphemes or argument structure constructions. Recent suggestions concerning the question of what counts as a construction promise a better handle on this problem by focusing on smaller-scale component constructions which may framemetonymically evoke the larger constructional frame. This is what Dancygier and Sweetser (2005) refer to as constructional compositionality (see also Dancygier 2009, 2011), and it is in this spirit that Nikiforidou (2010, 2012, 2015) analyzes the was + now pattern as one construction which, in marking a viewpoint shift away from the current speaker, may prompt the higher-level narrative construction of free indirect discourse (as in Banfield’s example How my heart was beating now!, 1982: 99). Along similar lines other typical ‘features’ of free indirect discourse, such as logophoric or ‘viewpoint’ reflexives (e.g. Zribi-Hertz 1989, Van Hoek 1997) or viewpoint-shifting connectives (Sotirova 2004), could be reinterpreted as such subconstructions. Further scrutiny by Dancygier (2012a: 190–93) has shown that now in past tense contexts acquires a resumptive function, marking a return to the narrative events currently in focus, as in example (2), in which the narrator uses now to shift away from what the library stacks have always represented to him in the (pre-story) past, back to the specific moment being narrated: (2)
Realizing I was stuck awhile, I began to see the place differently. The stacks had always been a purely functional means to an end. But now, I lived there. A long night ahead, and the third-biggest collection in the country to pass it on. (Richard Powers, The Gold Bug Variations; quoted in Dancygier, 2017b)
Now is only one among various examples where the viewpoint potential of ‘small’ constructions is reconsidered or even discovered in recent work, also including, for example, genitives, determiners, and negative markers (Dancygier 2009, 2011, 2012b). Even a widely studied language such as English still holds constructions with unexpected viewpoint function in certain contexts, though they are not necessarily grammaticalized, ‘dedicated’ viewpoint constructions like was + now or free indirect speech (cf. Dancygier 2016a: 283).
163
164
VAN DEL ANOT TE
Some uses of genitives premodifying proper names – which naturally call up rich frames of knowledge – were shown by Dancygier (2009, 2011) to function as markers of viewpoint. In talking about “one Hillary who has not yet reached the summit of her political Everest,” for instance (quoted in Dancygier 2009: 171), the frame of Mt. Everest and Edmund Hillary’s climbing of it is selected to reflect the viewpoint of Hillary Clinton as seeing the presidency as the ultimate achievement. ‘Her’ thus means something like ‘as she sees it,’ which is not standardly the case in speaking of, for instance ‘Hillary’s success.’ A similarly viewpointed use of genitives is found in the pattern one person’s trash is another person’s treasure, in which one and the same referent is evaluated negatively by some (trash) and positively by others (they view it as a treasure). Note again the difference with ordinary uses as in I stumbled over my neighbor’s trash, in which my neighbor is not the ‘viewer’ of the trash. A striking example of the pattern is (3), commenting on the question of whether you should respect gay people if you find homosexuality ‘immoral.’ The value-laden expressions here are sacred cow and double cheeseburger, with the latter (unlike treasure) receiving its evaluative charge by virtue of the construction, and receiving ample metaphorical elaboration in the next sentence. (3)
Your sacred cow is somebody else’s double cheeseburger. If you will simply look the other way when they are eating, not make rude noises, and don’t try to legislate against their lunch, that’d be plenty respect right there. (www.slate.com/blogs/outward/2014/01/06/should_you_re spect_gay_people_if_you_find_homosexuality_immoral.html)
As a final example of recent extensions of the repertoire of viewpointed uses of constructions, consider the case of intersubjective negation described by Dancygier (2012b) and exemplified in (4), in which I don’t hope refutes the hopeful stance which the interviewer attributes to the actor being interviewed, thereby invalidating the question itself: (4)
– What do you hope the mainstream audiences will learn about this artistic/bohemian culture. . . – I don’t hope they learn anything . . . I don’t know who goes to see films and why. (Internet example quoted in Dancygier 2012b: 85)
The role of both established and more discoursally based constructions is one important strand in current viewpoint research, and suggests that starting from low-level, small-scale constructions, which may or may not form part of larger constructions, is a fruitful approach. Another strand, represented in Dancygier and Sweetser (2014), highlights the role of viewpoint in figurative language. Even relatively mundane examples of metonymic and metaphoric usage turn out to subtly imply specific viewpoints: a metonymic example such as Table 5 needs more water, for instance, relies on the Restaurant frame from the viewpoint of the waiting staff, not the customer (Dancygier and Sweetser 2014: 21); or the oft-analyzed metaphor
Viewpoint
That surgeon is a butcher only really works when the viewpoint adopted is that of the patient in the Surgery frame (Dancygier and Sweetser 2014: 216–17). Recent work by Dancygier (2016b) has drawn attention to an even more fundamental level at which viewpoint comes in, and on which subsequent framing, metaphor, and blending operates, and that is the level of image-schematic structure as defined in the work of, for example, Johnson (1987), Lakoff and Johnson (1999), and Mandler and Paga´n Ca´novas (2014). Using the example of a barrier, involved in infants’ primary scenes such as being enclosed in a play pen, Dancygier (2016b) shows a range of increasingly elaborate construals in which barriers are crossed, removed, or reconstrued in textual and visual artifacts involving walls and borders, whether real or imaginary. The essential viewpointed experience of a barrier – being on either side of it, causing the experiencer to be restricted in terms of motion and vision, or having an aerial view of both sides – provides the scaffolding for these creative elaborations. To cite one visual example Dancygier analyzes, Banksy’s West Bank wall art creates different viewpointed experiences of ‘unwalling’ the wall, for instance by smashing it open with a wrecking ball, peeling it off like wallpaper or flying over it by hanging onto a bunch of balloons. In addition to image schemas and frames, also mental spaces and blending have been used productively to analyze viewpoint structures, and this across different modes and modalities. In his analysis of a visual poem by bpNichol, Borkent (2010) for instance unpacks a series of blends (including the ‘pond’ frame and fictive motion, among others) involved in ‘viewing’ a poem consisting of three handwritten lines containing the graphemes ‘fr,’ ‘pond,’ and ‘glop,’ with a curvy line connecting the ‘r’ and ‘g’ via the ‘o’ of ‘pond.’ As Borkent shows, the poem construes, with minimal formal means, the viewpoint of the agent (the frog), lying on the muddy edge of a pond, jumping in, producing a ‘glop’ sound and leaving behind it only o-shaped ripples on the water. In his study of gesture in political rhetoric, Guilbeault (2017) demonstrates how politicians often paraphrase their opponents’ viewpoints verbally, while criticizing and contextualizing these viewpoints in the more implicit gestural modality, thereby reducing cognitive load. In analyses of co-speech gesture and sign language more generally, the actual physical environment (including the body) of a language user serves as an input (the ‘Real Space,’ Liddell 1998) to viewpoint blends of varying complexity. Returning to written narratives, Dancygier (2005) first proposed the mechanism of viewpoint compression which allows lower-level viewpoints to be integrated with higher-level (narrative) viewpoint. In (5), for instance, a travel writer, Jonathan Raban, recounts his experience of seeing a TV report covering his own departure; the lower viewpoint is that of
165
166
VAN DEL ANOT TE
the writer-as-TV viewer, enriching the higher-level viewpoint of the narrator of Old glory: (5)
The TV news went local. An Englishman had left Minneapolis that day in a small motor boat . . . In the picture on the screen his face had a cheesy pallor . . . He looked to me like a clowning greenhorn. (Jonathan Raban, Old glory; quoted in Dancygier 2005: 109)
This is one case of what Dancygier (2012a: 100–02) calls decompression for viewpoint: for the duration of the description of the TV report, the narrative viewpoint is compressed with the ‘TV viewer’ persona as a decompressed subidentity of Raban, and the traveller persona moves temporarily to the background. A more complex case is (6), in which the ‘insider’ viewpoint, compressed with the narrative viewpoint, is contrasted with the ‘outsider,’ tourist’s viewpoint on Cairo, but both remain accessible: (6)
My fellow-diners and I had come at Cairo from different angles, and we’d arrived at different places. They’d flown from Gatwick to the land of the Pharaohs, while I had made a homecoming of sorts from Sana’a. (Jonathan Raban, Arabia; quoted in Dancygier 2005: 110)
These and related phenomena were integrated into a cognitive model of how readers construct stories from narrative text in Dancygier (2012a), which distinguishes a Main Narrative Space (MN) containing various embedded Narrative Spaces (NS), with all of this structure ultimately supervised by the Story-Viewpoint Space (SV) which comprises the features of narratorship; in this model, readers construct stories by blending elements from the different spaces available. The approach has been applied to fragmented contemporary novels (Dancygier 2007, Vandelanotte 2015) as well as news narratives (Van Krieken, Sanders, and Hoeken 2016). As can be expected, the construals can be very complex across multiple embedded spaces, but by way of illustration, let us reconsider two examples previously discussed. Figure 10.1 represents the decompression of Raban in example (5) above into ‘traveller’ and ‘TV viewer,’ and the subsequent compression of the ‘TV viewer’ with the narrative viewpoint. Figure 10.2 revisits example (1), in which the character Carlo Grace’s questions ‘What is life like at the chalet?’ and ‘What do you cook on?’ get to be represented as ‘What was life like at the chalet? What did we cook on?’ from the deictic viewpoint of the I-narrator Max. As Figure 10.2 shows, the pronoun we, referring to the I-narrator Max and his parents, functions at the level of the embedded Discourse Space, where it has the role of addressee, while simultaneously representing the I-narrator in the top-level StoryViewpoint Space (for similar analyses, see Dancygier 2012a: 187–88, 2017b). The notion of Story-Viewpoint Space was generalized in Dancygier and Vandelanotte (2016) into that of Discourse Viewpoint Space, regulating and supervising a networked configuration of viewpoints. This approach
Viewpoint
Story-Viewpoint Space Narrator I : first person
Viewpoint compression Narrative Space X
Narrative Space Y
Raban-as-TV viewer → Narrator I
Narrative Space 1
Raban-as-traveller → he Raban as traveller travelling down the Mississippi
Decompression Raban-as-traveller
Embedded Space 1
Raban-as-TV viewer
TV report – Raban as TV viewer An Englishman... his face had a cheesy pallor He looked to me like a clowning greenhorn
Figure 10.1 Decompression for viewpoint
Story-Viewpoint Space Narrator I : first person
Narrative Space X
Narrative Space Y
Narrative Space 1
Viewpoint compression Addressee you → Narrator I Discourse present → Narrative past
Discourse Space 1 Speaker: character (Carlo Grace): I Addressee: narrator (Max) + parents: you What is life like at the chalet? What do you cook on?
Figure 10.2 Viewpoint compression from Discourse Space to Story-Viewpoint Space
can help explain why and how, as a function of its place in the network, one and the same linguistic form (e.g. a first-person pronoun, an indefinite determiner) can have a different meaning. Across example sources such as film discourse, Internet memes, and video art, the analysis drives home the point that these often condensed discourse forms can show
167
168
VAN DEL ANOT TE
considerable viewpoint complexity, which is clearly not the sole preserve of long forms such as novels. Consider, for instance, the popular said no one ever meme: while naturally understood to ridicule unlikely or untenable viewpoints (as in Can I wear your Crocs? said no one ever), it requires a zoomout from its lowest level Discourse Space, in which the person depicted in the meme ostensibly asks someone (by default, the meme’s reader) a straightforward question, to a distanced, higher-level space prompted by the ‘said no one ever’ part, usually presented visually as bottom text, from which the lower-level viewpoint is viewed and understood as ‘nonquotative’ direct speech, actually attributable to no one. The resulting Discourse Viewpoint arrived at is that Crocs are too ugly to be desirable. The notion of zoom-out to achieve a view of a viewpoint was first proposed for the analysis of irony proposed by Tobin and Israel (2012). Irony on their account involves a layered configuration of mental spaces, where a perceived incongruity prompts a shift in attention from an inner to an outer layer (zooming out) resulting in a dynamic blended construal of an event from two incompatible viewpoints, one of which is rejected and ‘looked down on.’ Interestingly, sometimes the zoom-out does not have an unambiguous and stable nature, leaving the overall Discourse Viewpoint deliberately unclear. One fascinating example they discuss is the short story ‘Borges and I,’ written by Jose´ Luis Borges, where a kind of infinite regress of possibilities makes it impossible to pin down whether the overarching Discourse Viewpoint is that of Borges as public figure, or Borges as ‘real’ person. Work in cognitive linguistic humor studies more generally incorporates an important viewpoint component (e.g. Ritchie 2006). To take one example, Broˆne’s (2008) analysis of interactional humor in the British sitcom Blackadder relies on complex embedded viewpoints. In one scene involving misunderstanding on the part of dim-witted Lieutenant George, the soldiers are in no-man’s land and George is looking at a map, guessing that the ‘area marked with mushrooms’ which he sees on the map means ‘that we’re in a field of mushrooms.’ The following exchange between Captain Blackadder and Lieutenant George ensues: (7)
– Lieutenant, that is a military map, it is unlikely to list interesting flora and fungi. Look at the key and you’ll discover that those mushrooms aren’t for picking. – Good Lord, you’re quite right sir, it says “mine.” So, these mushrooms must belong to the man who made the map. – Either that, or we’re in the middle of a mine-field. . . – So, he owns the field as well? (Curtis et al. 1998: 362)
Figure 10.3 represents just one small aspect of the misunderstanding, that pertaining to the meaning of the mushroom symbol (for a more detailed analysis, see Broˆne 2008: 2036–42). As Figure 10.3 suggests, the viewer derives pleasure from seeing the whole picture from the Discourse Viewpoint Space, particularly the counterpart relations established across
Viewpoint
Discourse Viewpoint Space Narrator: 0 (offstage)
Narrative Space X
Narrative Space Y
Narrative Space 1 Soldiers in no man’s land Belief Space 1
Belief Space 2
Blackadder
George
mushroom (mine)
mushroom (plant) Belief Space 3
Cross-input projection
Blackadder
counterpart relations mushroom (mine) – (plant)
mushroom (plant)
Figure 10.3 Embedded viewpoints with cross-input projection
the different Belief Space inputs: Blackadder’s actual belief, George’s mistaken belief, and, embedded within the latter, the belief which George wrongly attributes to Blackadder (given his lack of uptake following Blackadder’s sarcastic follow-ups, he clearly believes Blackadder takes the mushrooms to represent mushrooms, not mines). Irony and interactional humor are just some of the genres in which viewpoint mechanisms have been analyzed in mental space terms. Dancygier and Sweetser (2014: ch. 8), for instance, analyze discourses on illness or political discourse in which frames and blends are driven by viewpoint. The general idea of seeing discourse viewpoint as a matter of networked configurations of lower-level viewpoints, prompted by constructions, frames and blends, aims to capture the complexity of the phenomena by reconciling local multiplicity with the global coherence achieved at the level of the Discourse Viewpoint Space. In the closing section which follows, terminological and classificatory questions adhering to viewpoint are addressed.
10.3 Viewpoint Types and Relations to Other Concepts The terminological diversity in viewpoint research can feel overwhelming and unhelpful. The following passage illustrates the problem:
169
170
VAN DEL ANOT TE
Every point of view belongs to a character, hence every focalization happens according to a character’s perspective; similarly, every character’s voice presupposes the existence of a focalizing perspective in which that character is vocalized. (Mey 1999: 149–50)
The potential for confusion is understandable, given the number and nature of linguistic and embodied construal choices which reveal a conceptualizer’s viewpoint on smaller or larger aspects of a frame or situation. Certainly the narratological distinction between narrative voice (‘who speaks?’) and focalization (‘who sees?’ and, more broadly, ‘who perceives?’) articulated by Genette (1980, 1988) captures an important intuition, though as Niederhoff (2009a, 2009b) shows, it has spawned reifications (for instance by Bal 2009) capturing different aspects than those intended, and the issues remain hotly contested. For Genette, focalization replaces notions such as perspective or point of view, and pertains to “selections of narrative information” (Niederhoff 2009a: 122), distinguishing between unrestricted access to the storyworld (zero focalization), restriction to the experience of a character (internal focalization), or a further restriction to observable information only without access to characters’ minds (external focalization). In cognitive narratology, Herman (2009) has attempted to overcome disagreements by proposing a theory of narratorial mediation in which narrating and perspective-taking (focalization) are intricately interconnected, given that “all storytelling acts are grounded in the perceptual-conceptual abilities of embodied human minds” (2009: 128). Even if full narratological consensus on these issues is likely to remain elusive, a basic distinction along the lines suggested by Genette should be attempted. Given the choice to use viewpoint as the overarching concept, the distinction could be expressed in terms of deictic viewpoint, referring to the deictic center (conceptualizer, time, and place) from which some aspect of a situation is viewed, as opposed to various kinds of cognitive viewpoint, referring to whose cognitive states (thoughts, knowledge states, experiences, beliefs, attitudes, etc.) are being viewed, recognizing both the basic observation that various conceptualizers’ cognitive states can be accessed from a constant deictic viewpoint and the possibility for the deictic viewpoint itself to shift (cf. Galbraith 1995), as for instance in direct speech reports. In example (8) below, for instance, the deictic viewpoint remains with the I-narrator (who is trying to block a woman from the view of her hawk, to avoid upsetting it), but the cognitive viewpoint shifts from the I-narrator to the woman’s viewpoint of her (‘a weirdo in a tattered jacket’), and subsequently blends with that of the hawk, a process completed in the final two clauses which compress the hawk’s perception with the narrative viewpoint. (8)
I hold her close to my chest and turn in a slow circle to block the woman from view. The woman doesn’t see the hawk. What she sees is a weirdo in a tattered jacket and baggy corduroy trousers revolving on the spot for no
Viewpoint good reason . . . I’ve been with the hawk so long, just her and me, that I’m seeing my city through her eyes. She watches a woman throwing a ball to her dog on the grass, and I watch too, as baffled by what she’s doing as the hawk is. I stare at traffic lights before I remember what they are. Bicycles are spinning mysteries of glittering metal. The buses going past are walls with wheels. (Macdonald 2014: 100–02)
We know of course that buses are not walls with wheels, but having assimilated the hawk’s visual perception, the narrator conceptualizes them as precisely that (compare the notion of ‘fictive vision’ discussed in Dancygier 2012a: 102–06). While arguably a useful distinction to make, the deictic–cognitive divide is not impermeable, and some forms straddle the boundary. Social deixis is a case in point: choice of address or reference form for people (for instance Dirk or Professor Delabastita) includes a deictic component (third person, not the speaker) but also a cognitive, social one, pertaining to the degree of familiarity to the speaker but also, as Sweetser (2013: 242) points out, intersubjectively, to the hearer (as I might use ‘Professor Delabastita’ in the presence of an undergraduate student, but ‘Dirk’ in private correspondence with my close colleague). Similarly, as suggested above, deictic elements such as now may be recruited to reflect an ‘immersed,’ present cognitive viewpoint in contexts where the overall deictic viewpoint may be past. Things are thus rarely simple, and cognitive viewpoint in particular is by its very nature multifaceted, covering, for instance, phenomena that have been explicated in terms of experiential viewpoint (as in the viewpoint genitives discussed in section 10.2), grammatical aspect (which modulates conceptualization; Parrill, Bergen, and Lichtenstein 2013), and epistemic and emotional viewpoint. These latter notions have often been analyzed under the rubric of stance-taking (e.g. Du Bois 2007), and in Cognitive Linguistics have featured prominently in analyses of conditional constructions (e.g. Fillmore 1990, Dancygier and Sweetser 2005). Broadly defined, along a general deictic–cognitive divide, as a conceptualizer’s alignment with an aspect of a frame or situation, viewpoint is reflected in many linguistic and gestural constructions expressive of underlying conceptual viewpoint (Parrill 2012). This chapter has not sought to draw up an inventory of such constructions, since any such list would likely be incomplete, and unlikely to yield the kinds of generalizations sought here. However, descriptive work on such forms across different languages remains needed, while their applications across different modalities, domains, and genres are increasingly being tested and described using methodologies from across the cognitive humanities and sciences.
171
11 Embodied Intersubjectivity Jordan Zlatev
11.1 Introduction Is language primarily grounded in the body or in sociality? The easy answer would be ‘in both,’ but if forced to make a choice, most cognitive linguists would probably opt for the first. This would be unsurprising given the paradigmatic status of the notion of embodiment in the tradition (Lakoff and Johnson 1999, Gibbs 2005a, Rohrer 2007), reflected in terms like embodied realism (Johnson and Lakoff 2002) and embodied meaning (Tyler and Evans 2003, Feldman and Narayanan 2004). But we all have our individual bodies, while language is fundamentally social and intersubjective (Wittgenstein 1953). Criticism from interactional linguistics (Linell 2009b), functional linguistics (Harder 2010), and the philosophy of linguistics (Itkonen 2003) has zeroed in on this lacuna, arguing that until it is resolved, cognitive linguistics will be committed to an individualist, solipsist model of meaning. There have, of course, been attempts to combine embodiment and intersubjectivity in a unified account of the foundations of language (Zlatev 1997, Verhagen 2005, Zlatev et al. 2008), but it is fair to say that the tension remains. One way to resolve this tension is with the help of phenomenology, the philosophical tradition inaugurated by Edmund Husserl over a century ago, which analyzes the nature of conscious experience, in its multitude of aspects – subjective and intersubjective, perceptual and representational, etc. – and describes how meaning emerges from it. As part of the project of rethinking the ontological and epistemological foundations of cognitive linguistics (Zlatev 2010), I here show how phenomenology seamlessly fuses embodiment and intersubjectivity as complementary aspects of the same phenomenon, reflected in the expression intercorpore´ite´ or embodied intersubjectivity (Merleau-Ponty 1962). In section 11.2, I explicate why Merleau-Ponty and other phenomenologists attributed a central role to the sentient and active human body for relating to others and for the creation of joint meaning. However, as pointed out by Merleau-Ponty,
Embodied Intersubjectivity
following (Husserl 1936 [1970]), while this may ‘ground’ the social and normative meanings of language and science, there remains a gap between these layers, and it can only be bridged through several stages of historicity. In section 11.3, we will see that such a model ties in well with research from developmental psychology, thus framing the discussion in a more empirical manner. Subsequently, we turn to areas that have been of explicit concern for cognitive linguistics, and show how the adopted perspective helps reframe some key concepts with respect to image/mimetic schemas (section 11.4), conceptual metaphors (section 11.5) and construal operations, with focus on non-actual motion (section 11.6). In all these cases, we will see structures of bodily intersubjectivity serving to help establish and motivate linguistic structures and meanings, but, crucially, without determining them, as “it is experience that proposes, but convention that disposes” (Blomberg and Zlatev 2014: 412). Finally, we summarize the main points of the argument, highlighting the benefits of a phenomenological cognitive linguistics.
11.2 Embodied Intersubjectivity in Phenomenology Phenomenology is becoming increasingly influential in cognitive studies (Gallagher and Zahavi 2008, Zlatev 2010, Koch et al. 2012, Blomberg and Zlatev 2014), along with an appreciation of its sensitive analyses of the lived body (Leib for Husserl; corps ve´cu for Merleau-Ponty) and their potential for resolving problems of mind/body dualism, ‘anything goes’ relativism, the existence of ‘other minds’ and other trappings of (post)modernity. Most relevant for our current concerns is, however, that embodied intersubjectivity can help ground the intersubjective nature of language as “phenomenologists have often endeavoured to unearth pre- or extra-linguistic forms of intersubjectivity, be it in simple perception, tool-use, emotions, drives, or bodily awareness” (Zahavi 2001: 166). The functions of these ‘pre- and extra-linguistic forms’ for motivating linguistic meaning will become more explicit in subsequent sections, and I ask the reader here simply to reflect on the phenomena described. While there are multiple aspects of embodied intersubjectivity, the most fundamental is the double aspect nature of the body: “My body is given to me as an interiority, as a volitional structure and a dimension of sensing, but it is also given as a tactually and appearing exteriority” (Zahavi 2001: 168). The compound term Leibko¨rper utilizes the fact that German has two different terms that profile the ‘internal’ (Leib) and the ‘external’ (Ko¨rper) aspects, respectively. As profiling does not imply different (ontological) entities but different perspectives, my Leib and my Ko¨rper ultimately coincide. This is clearly shown in the experience of double sensation:
173
174
Z L AT E V
when I touch my right hand with my left, my right hand, as an object, has the strange property of being able to feel too . . . the two hands are never simultaneously in the relationship of touched and touching to each other. When I press my two hands together, it is not a matter of two sensations felt together as one perceives two objects placed side by side, but of an ambiguous set-up in which both hands can alternate the rôles of “touching” and being “touched.” (Merleau-Ponty 1962: 106)
In fact, I could touch your right hand instead of my own, and the sensation would not be completely different (though not completely the same either). As Merleau-Ponty (1962: 410) expresses it: “The other can be evident to me because I am not transparent for myself, and because my subjectivity draws its body in its wake.” And Zahavi (2003: 104) states this even more forcibly: “I am experiencing myself in a manner that anticipates both the way in which an Other would experience me and the way in which I would experience an Other . . . The possibility of sociality presupposes a certain intersubjectivity of the body.” It should be noted that this analysis does not collapse self and other into an amorphous anonymity as there remains an asymmetry: “The grief and anger of another have never quite the same significance for him as they have for me. For him these situations are lived through, for me they are displayed” (Merleau-Ponty 1962: 415). Still, as lived-through and displayed are connected in a reversible relation that resembles double sensation, we can pass from one to the other, without any need for inference or simulation: “I perceive the grief or the anger of the other in his conduct, in his face or his hands . . . because grief and anger are variations of belonging to the world, undivided between the body and consciousness” (415). Using the concept of bodily resonance, Fuchs (2012) analyzes emotions as seamless blends of an internal ‘affective’ component and an outwarddirected ‘emotive’ component. In a social context, the felt affect of the Leib is displayed in its Ko¨rper’s emotive expression, which then results in an affect in the Leib of another embodied subject, giving rise to emotive expression through its Ko¨rper and so on. In such an ‘inter-bodily’ loop, one literally perceives (rather than ‘infers’ or ‘simulates’) the other’s emotion, through a kind of meaning and action-oriented process currently known as enactive perception (Gallagher and Zahavi 2008). Building on Husserl’s notion of operative intentionality (fungierende Intentionalita¨t), referring to a basic pre-conceptual, but meaningful directedness toward the world, Merleau-Ponty proposed the influential notion of a body schema (sche´ma corporel) recently explicated as “a system of sensorimotor capacities that function without awareness or the necessity of perceptual monitoring” (Gallagher 2005: 24). It is important to distinguish this from the notion of body image, which “consists of a system of perceptions, attitudes and beliefs pertaining to one’s own body” (24). The latter takes the body into focal consciousness, making it an intentional object of perception or conception, while the body schema constitutes the pre-personal
Embodied Intersubjectivity
embodied subject him- or herself, and is imbued (at most) with marginal consciousness. As both Merleau-Ponty (1962) and Gallagher (2005) argue, the body schema is preserved in some clinical cases where the body image is compromised, and should thus be seen as distinct empirically as well.1 The body schema involves learning and memory, referred to in the phenomenological tradition as body memory, which “does not represent the past, but re-enacts it through the body’s present performance” (Fuchs 2012: 11). At least two of the forms of body memory distinguished by Fuchs (2012) are inherently intersubjective. The first, deeper and least open to reflection, is called intercorporeal, emerging from infancy and resulting in “implicit relational know-how – bodily knowing of how to interact with others, how to have fun together, how to elicit attention, how to avoid rejection, etc.” (15). A second, and more reliant on conscious attention form of body memory, is called incorporative; it presupposes full self-other differentiation, and the more or less intentional adoption of postures, gestures, and styles from others, based on imitation and identification. Given the asymmetrical, power-based relationship between learner and model involved, it is unsurprising that Fuchs relates the emergence of incorporative memory to the well-known sociological concept of “the habitus – embodied history, internalized as second nature and so forgotten as history” (Bourdieu 1990, Fuchs 2012: 56). It is thanks to such intersubjective body memory that ‘body language’ is so often transparent: “The communication or comprehension of gestures comes about through the reciprocity of my intentions and the gestures of others, of my gestures and the intentions discernible in the conduct of other people. It is as if the other person’s intentions inhabited my body and mine his” (Merleau-Ponty 1962: 215). However, as the distinction between the two forms of body memory outlined above indicates, we should not assume that such transparency is universal: habitus and communicative gestures are, to a considerable extent, culture and group-specific. The forms of embodied intersubjectivity reviewed so far have profiled dyadic, subject–subject relations. However, actions and gestures of the kind mentioned above may involve objects as well (Andre´n 2010). For Husserl, the transcendent (i.e. real, ‘objective’) nature of an object like the coffee mug in front of me was of central concern (for else the worlddirectedness of intentionality would be in question), and he eventually concluded that even simple object-directed intentionality presupposes intersubjectivity. The argument, in brief, is that while I may only see the mentioned coffee mug from one perspective (at one moment), I coperceive its other sides, including its bottom and container-shape, and I synthesize these into an identity: the coffee mug itself. The foremost 1
Still, we should not dichotomize this distinction too much, since analogously to that between Leib and Körper, typical human experience allows flexible re-location of consciousness between intentional object (noema) and process (noesis), related to the cognitive linguistic notion of objective/subjective construal (Zlatev, 2010).
175
176
Z L AT E V
reason that this can be done is that I am implicitly aware that the other perspectives are available for other embodied subjects. “My perceptual objects are not exhausted in their appearance for me; rather, each object always possesses a horizon of co-extending profiles which . . . could very well be perceived by other subjects, and is for that very reason intrinsically intersubjective” (Zahavi 2001: 155). Finally, this analysis would hold even for natural objects; for example, if I had a stone on my working table rather than the proverbial coffee mug. For cultural artifacts or tools, there is an extra layer of embodied intersubjectivity, as their affordances – the possibilities for action that they invite, inscribed on structures of body memory, “contain references to other persons” (271) – those who have made them, or could use them in manners similar to mine. In sum, we have here briefly sketched a number of phenomenological structures of embodied intersubjectivity: Leibko¨rper duality, bodily resonance, body memory, intersubjective object perception and the sociality of artifacts (see Zlatev and Blomberg [2016] for more discussion). Their relevance for language, and in particular for cognitive semantic analysis, will be shown in what follows. What is important to emphasize for the time being is that (a) embodiment and intersubjectivity are complementary aspects of experience at the most fundamental levels of consciousness and (b) the meaningfulness of such structures is not private, but interpersonal – not ‘in the head’ but ‘in the world’ and (c) their meaning (e.g. of incorporated memory schemas) can be expected to underlie (‘ground’) linguistic meaning, but is not identical with it. This principle clearly applies to ontogenetic development, as shown in the next section.
11.3 From Embodied Intersubjectivity to Language in Development A phenomenology-inspired approach to intersubjectivity, and its foundational role for language, has been influential in developmental psychology (Trevarthen 1979, Stern 2000, Gallagher 2005, Braº ten 2006). One of its features is that it provides a coherent alternative to the more common ‘theory of mind’ approach to social cognition. In contrast to this, the shared mind approach (Zlatev et al. 2008) emphasizes that people are attuned to each other’s subjectivity from birth (Trevarthen 2011), that key cognitive capacities are first social and only later understood in private or representational terms (Vygotsky 1978), that experiences are shared not only on a cognitive level, but also on the level of affect, perceptual processes and conative engagements (Hobson 2004), and that intersubjectivity involves different levels and stages in development and evolution. One such model is the Mimesis Hierarchy (Zlatev 2008, 2013), which distinguishes between fives stages/layers, with subsequent ones
Embodied Intersubjectivity
Table 11.1 The mimetic hierarchy of semiotic development (adapted from Zlatev 2013) Stage 5 Language
Novel capacity Language-mediated folk psychology
Examples of cognitivesemiotic skills
Approximate age
- complex sentences - discourse - onset of narrative
30 m -
4 Protolanguage Symbols: communicative, - vocabulary spurt conventional - reorganization of gestures representations - gradual increase in utterance complexity
20–30 m
3 Triadic mimesis
Communicative intent
- declarative pointing - reciprocal (joint) attention - associative schemas
14–20 m
2 Dyadic mimesis
Volitional control and imitation
- generalized/deferred imitation - coordinated (joint) attention
9–14 m
-
0–9 m
1 Proto-mimesis Empathetic perception
neonatal imitation emotional contagion ‘proto-conversations’ synchronous (joint) attention
superimposing upon rather than superseding earlier ones. The first three layers are pre-linguistic, and provide the foundation for the highest two levels, where language emerges, with its representational, normative and systematic character (Zlatev 2007b). Metaphorically, each layer can be seen as ‘grounding’ those following it, as shown in Table 11.1. Let us summarize each briefly, with reference to some of the phenomenological notions discussed in the previous section. Proto-mimesis. The child has a minimal embodied self from birth, largely based on the sense of proprioception. For example, neonates react differently when their own hand and someone else’s hand touches their cheeks (Rochat 2011). At the same time, such newer evidence does not contradict the position that during the first months of life the embodied subject does not completely differentiate their own body schema from the perceived body of the other, paradigmatically, the mother (Piaget 1962, Werner and Kaplan 1963). This is one way to interpret the well-known (though still controversial) phenomenon of “neonatal mirroring” (Gallagher 2005). By this analysis, such mirroring is distinct from full imitation (of novel actions), which requires volitional control of the body, carefully matched to that of the other. At this early stage, interpersonal relations are heavily based on processes of bodily resonance, reflected in ‘emotional contagion’ (spontaneously picking up the feelings of other) and mutual attention (prolonged bouts of looking into the eyes of their caregivers). Body
177
178
Z L AT E V
memory is of the intercorporeal kind, realized in turn-taking ‘protoconversations’ (Trevarthen 1979) and interactional ‘formats’ like peekaboo (Bruner 1983). Dyadic mimesis. By nine months of age, infants have considerably expanded body schemas, with increased motility, volitional control, and a distinct sense of self, as well as the beginning of a body image. It is characteristic that around this age the first pointing gestures and true imitation emerge, driven largely by the need to maintain intersubjective closeness (Werner and Kaplan 1963, McCune 2008). Imitation progresses from sensory-motor imitation, through deferred imitation, to representational imitation where “the interior image precedes the exterior gesture, which is thus a copy of an ‘internal model’ that guarantees the connection between the real, but absent model, and the imitative reproduction of it” (Piaget 1962: 279). While the first two correspond to incorporative memory, the achievement of the third form during the second year of life marks the onset of the first concepts (Piaget 1962), or focally conscious mental representations, which Zlatev (2005) calls mimetic schemas: “dynamic, concrete and preverbal representations, involving the body image, accessible to consciousness and pre-reflectively shared in a community” (334), including schemas such as Kiss, Kick, Eat, and Jump. As reflected in these quotations, the idea is that bodily mimesis develops from re-enactive, nonrepresentational memory to representational thought and a form of bodybased semantic memory. These are clearly structures of embodied intersubjectivity: “The posited mimetic schemas . . . as the name suggests are schematizations over multiple mimetic acts . . . internalized from overt imitation of the actions of the other” (Mo¨tto¨nen 2016: 164). Triadic mimesis. The earlier stage is named ‘dyadic’ despite the possible involvement of objects in communicative interaction (e.g. desired ones in so-called imperative pointing) or imitated actions with objects (e.g. in feeding), since, only starting from about fourteen months, child, other, and object become fully integrated in a referential triangle. This is reflected in ‘declarative pointing’ gestures (look at that!) and gaze oscillations between object and addressee that mark full joint attention, implying an intersubjective form of ‘third level mentality’ (I see that you see that I see) (Zlatev 2008). This can be seen as an empirical manifestation of Husserl’s transcendental intersubjectivity, according to which, external objects gain their fully transcendent status only when they can be the potential focus of joint attention. Furthermore, it is only with triadic mimesis that Gricean intentional communication can take place, where one is intending both to communicate something to an audience, and for the audience to recognize this intention. This implies an embodied second-order intention, enacted through bodily actions like gaze, smile, raised eye brows, or holding of the gesture’s reply (Zlatev et al. 2013) (Proto)language. While the first words have likely appeared much earlier, these are typically restricted to specific contexts and function more like
Embodied Intersubjectivity
indexical schemas, associated with, rather than as a symbols ‘standing for,’ emotional states and external events (Bates et al. 1979). However, based on the ‘infrastructure’ of body memories, mimetic schemas, and triadic mimetic gestures, around their second birthday most children display a marked increase in the number and variety of their words. This is often described in the developmental literature as a ‘vocabulary spurt.’ In line with phenomenology, the model relates this to the emergence of reflective consciousness (Zelazo 2004), which is instrumental for the symbolic insight that ‘things have names’ and that these names are common, that is conventional (Zlatev 2013). From this point, language co-develops and interacts with the use of other semiotic resources such as gestures (McNeill 2005) and pictures (DeLoache 2004), gradually increasing in structural complexity (Tomasello 1999). This makes the distinction between ‘proto-language’ and language-proper a gradual one, and the main reason that it is worth making it an ontogeny is that a certain degree of linguistic proficiency in representing both events, and their interrelations, is required for the production of narratives, which for its part has an important effect on the construction of both self- and other representations (Hutto 2008). Unlike some language-centered philosophers (e.g. Dennett 1991), however, the present model and embodied phenomenology at large hold that narratives cannot ‘make up’ the self, as there would be (a) no story to tell, (b) no one to tell it, (c) no one to tell it to, and (d) no medium to tell it in – unless the self, others, and the world where not first cyclically coconstituted, and language built up upon this fundament. In the following three sections, we will see how this general scenario plays out in the analysis of central cognitive linguistic phenomena.
11.4 Image Schemas and Mimetic Schemas One of the most influential, and at the same time most ambiguous, concepts in cognitive linguistics has been that of image schemas (cf. Zlatev 2005). Usually these have been characterized as highly abstract, nonrepresentational structures such as Path, Container, and Verticality, underlying the meaning of closed-class, spatial terms (e.g. to, from, in, out, up, down). A commonly cited definition of an image schema is that of “a recurring dynamic pattern of our perceptual interactions and motor programs that gives coherence to our experience” (Johnson 1987: xiv). The use of the first-person plural possessive pronoun in this definition, as well as characterizations of image schemas as structures of “public shared meaning” (190), however, remain not clearly justified. Lakoff and Johnson (1999) profess to do so in terms of “shared” human anatomy and physiology, claiming that this would explain why “we all have pretty much the same embodied basic-level and spatial-semantic concepts” (107). But how can biological bodies ‘on their own’ (in both senses: without their lived
179
180
Z L AT E V
counterparts, and without others) give rise to shared intersubjective experience (Zlatev 2010)? Furthermore, ‘perceptual interactions’ differ with environments and cultural practices (habitus) so the implicit universalism of the notion is also questionable. One of the motivations behind the concept of mimetic schemas was to deal with this problem. As shown in section 11.3, they are assumed to emerge from being with others, schematizing multiple acts of overt or covert bodily mimesis. Their harmonious relation with phenomenological analysis has also been acknowledged: “With respect to the phenomenology of body memory . . . this approach is particularly promising as it allows one to highlight the role not only of habitual, but also of intercorporeal and incorporative body memory in the process of meaning formation” (Summa 2012: 38). The semantics of the majority of the ‘first verbs’ acquired by (Englishspeaking) children (McCune 2008), including intransitive verbs like eat and transitives like hit, can be seen as rooted in such schemas (Zlatev 2005). Furthermore, the analysis of the gestures of three Thai and three Swedish children between eighteen and twenty-six months has provided evidence that mimetic schemas also give rise to children’s first iconic gestures, helping to account for the close semantic and temporal integration between speech and gesture (Zlatev 2014). In particular, 85 percent of the iconic gestures were found to fall into ‘types’ on the level of specific actions like kicking, kissing, and applying lotion. Only two such types/ schemas were found to be common to both cultural groups (Kiss, Feed), while the rest were culture-specific (or at least culture-typical). Nearly all corresponded to mimetic enactments (with the body representing itself, so to speak), and those few that did not involved the use of toys: Doll-walk, Car-drive, etc. Image schemas of the Path and Container kind have been used in gesture analysis, but characteristically the considerably more abstract gestures that correspond to them are first found in older children and adults, suggesting a possible ‘division of labour’ with mimetic schemas (Cienki 2013). One possibility is that image schemas come about as generalizations from the more concrete structures of embodied intersubjectivity analyzed so far in this chapter, and may thus inherit their interpersonal aspects this way. But it also possible that image schemas are not ‘pre-verbal’ as originally claimed, but are at least co-determined by the semantics of the closedclassed items that they are supposed to ‘ground’ (Zlatev 2005). As such expressions are typically acquired later in life, this would also explain why image-schema-like gestures are absent in toddlers. Another proposal is that of a more co-temporal, synchronic development and interaction between mimetic schemas and image schemas. For example, McCune (2008) follows Piaget (1962) in distinguishing between ‘figurative’ and ‘operative’ structures of sensorimotor cognition. While mimetic schemas correspond to the figurative (concrete, specific) kind,
Embodied Intersubjectivity
children’s physical and social interactions may results in pre-linguistic schemas of Appearance-Disappearance, In-Out (which resemble image schemas) and these may ground the meaning of dynamic event words like allgone, in, and out. Piagetian embodiment has not been widely known for its intersubjectivity, but given the pivotal role of mimesis in imitation and symbolic play in this theory, it would not be implausible to view these as structures of embodied intersubjectivity as well, especially if McCune’s analysis for pre-verbal image schemas holds.
11.5 Metaphors and Their Experiential Roots Two interrelated debates characterize the recent literature on metaphors in cognitive linguistics (cf. Johnson 2010). The first concerns the fundamental level on which metaphors operate: is it that of so-called conceptual metaphors, defined as “a cross-domain mapping of structure from a source domain to a target domain, where the two domains are regarded as different in kind” (Johnson 2010: 407), commonly illustrated with the relation between the domains of Time and Space? Or is it the level of recurrent discourse metaphors such as our European home, which are intermediary in their conventionality between novel metaphors relying on analogical reasoning and literalized expressions (Zinken 2007)? Or is it perhaps the most specific, contextual level, so that “rather than conceiving of metaphors as discrete units they should be regarded as a process of meaning construal in which new metaphoric expressions dynamically emerge, are elaborated, and are selectively activated over the course of a conversations” (Kolter et al. 2012: 221). It should be emphasized that it is what is ‘most basic’ that is contested, as there is a degree of consensus that the levels of (universal) bodily experiences and cognitive processes, (conventionalized) cultural practices, and situated language use are not exclusive but rather complementary (Zlatev 2011). The second debate concerns the balance between universality and language/culture specificity of metaphors. As may be expected, the stance taken in this debate correlates with that taken in the previous one: if metaphors, and especially so-called primary metaphors (Grady 1997a) such as A F F E C T I O N I S W A R M T H “are acquired unconsciously through our bodily engagement with our environment” (Johnson 2010: 410), a considerable degree of universality may be expected. If, on the other hand, metaphors are essentially discursive constructs, cultural specificity follows almost by definition. Thus, while proponents of the universalist stance have been happy to accept a degree of cultural variation (Ko¨vecses 2000), for example, “social constructions are given bodily basis and bodily motivation is given social-cultural substance” (14), the controversy will persist as long as the deeper, definitional disagreement does.
181
182
Z L AT E V
Without pretentions to fully resolve these issues here, we can see in these debates another reflection of the familiar ‘body versus culture’ dichotomy, to which embodied intersubjectivity could provide a remedy. We may begin by observing that, while this is seldom acknowledged, suggested ‘primary metaphors’ are almost always intercorporeal. This is hardly surprising as they are hypothesized to be “acquired by children simply because of the nature of the bodily experience (in perception and bodily movement) for the kinds of the structured environments they inhabit” (Johnson 2010: 410) and as in the paradigmatic example A F F E C T I O N I S W A R M T H , these ‘environments’ are crucially interpersonal. As with image schemas, we could see Johnson (2010) using first-person pronouns (‘our’) in his description of primary metaphors. In the present case this is justified, as such ‘correlations’ (affection-warmth, etc.) correspond to felt, reciprocal qualities in the Leibko¨rper of infant and mother, that is, to intercorporeal body memory. However, is it really appropriate to analyze such structures, emerging through bodily interaction and involving (in most cases) emotion, as ‘cross-domain mappings’? As shown in section 11.2, unlike the way they are usually represented in cognitive semantics (Ko¨vecses 2000), emotions hardly constitute an ‘abstract target domain,’ and even less so one that is ‘different in kind’ (see Johnson’s definition above) from their bodily expressions. Rather, following Fuchs’s analysis, the internal and external sides of emotion are intimately connected through bodily resonance, allowing them to be perceived directly, without the need for inference or simulation. Of course, emotions can be conceptualized in many different ways, in different cultures and languages (e.g. as ‘forces’ or ‘fluids’) but these are secondary, language-mediated, and indisputably metaphorical constructions. If it is the pre-conceptual, and (largely) universal bodily roots of metaphors that concern us, then an analysis in terms of ‘cross-domain mappings’ would seem like placing the cart before the horse. Studying the ‘mapping’ between the ‘domains’ of motion and emotion in more or less related languages seems to confirm this theoretical conclusion. On the basis of native-speaker intuitions and corpus-analysis, Zlatev, Blomberg, and Magnusson (2012) performed an extensive search for conventionalized expression-types in English, Swedish, Bulgarian, and Thai where a motion verb is used to denote a state of affect/emotion, without any perceived motion, as in my heart dropped. One-hundred and fifteen such ‘motion-emotion metaphors’ (on the intermediary level of conventional metaphor types) were identified and compared. As expected, the findings showed considerable cross-language differences, especially between Thai and the three European languages, and also significant similarities. For example, in all four languages upward motion was correlated with positive affect and downward motion with negative. At the same time, this was expressed differently, with, for example, Thai requiring compounds in which caj, ‘heart-mind,’ marks the phrase as an
Embodied Intersubjectivity
emotion rather than as a motion expression. In line with what was emphasized in the previous sections, this showed the need to distinguish between the levels of (i) conventional linguistic expressions and (ii) prelinguistic motivations. What may the pre-linguistic motivations be in this case? As pointed out as early as Lakoff and Johnson (1980a) one likely motivation for the E M O T I O N I S M O T I O N ‘primary metaphor’ is that positive/negative emotion corresponds to higher/lower bodily posture. In such cases, this would be yet another expression of Leibko¨rper duality, with felt qualities being, so to speak, worn on the sleeve of the living body, rather than thinking of Leib and Ko¨rper as two distinct domains. In other cases, especially when the motion verb expresses motion through a liquid (one typically sjunka ner, ‘sinks down,’ into a depression in Swedish), a cross-domain mapping (analogy) may be a more appropriate analysis. The four languages (and others studied since then) also featured similar conventional metaphors corresponding to the mimetic schemas Stir, Shake, and Shatter. In the case of Shake there is obvious motivation in Leibko¨rper duality; for example, an electric shock affects both body and soul. For Stir and Shatter, the motivation is more likely in the analogy of the felt (inner) sensation and the observed transformations of external objects: with brews being stirred, and fragile things shattered (Zlatev, Blomberg, and Magnusson 2012). In sum, there is nothing to guarantee that the partial overlap of metaphors across unrelated languages is to be explained by the same mechanism, rather than by different intercorporeal motivations. To repeat again, the latter should not be confused with the metaphors themselves, which require expression, linguistic or otherwise. A very different kind of study also supports this view of interaction between – but non-identity of – bodily motivations and expression. An interdisciplinary team of cognitive linguists, therapists, and phenomenologists (Kolter et al. 2012) studied how patients expressed aspects of their lives first only through body movement, and then by simultaneous use of body movement and speech. One patient first enacted her feelings with two body patterns: one swinging movement from side to side, and a “spiral movement, executed with the left hand from the highest peak above the head down to the waist” (207). Later she stated that she felt that her life was as a wave, which sometimes goes up and sometimes down. At some point, however, she noticed the mismatch between the downgoing spiral of her gesture, and the verbal description, and adapted the movement to be bidirectional. The authors interpreted the (initial) body movements of the patient as “expressions of learned behavioural patterns and attitudes, which are sedimented in her body memory” (210), and the transition from non-matching to matching relation between movement and language as the ‘waking’ of a ‘sleeping metaphor,’ that is, the emergence of that particular metaphorical construal of her life into conscious awareness.
183
184
Z L AT E V
11.6 Non-actual Motion and Construal Sentences like (1) and (2) have been provided in argument for a nondenotational semantics: neither the mountain range nor the path are in motion, but are construed as such, apparently due to some underlying cognitive process. Recently, it has been especially popular to see this process as some kind of neural/mental ‘simulation’ (Matlock 2010). (1) The mountain range goes all the way from Canada to Mexico (Talmy 2000a). (2) The path rises toward the summit (Langacker 2006). Such an analysis leaves a number of problems, analogous to those discussed in the previous section. First, there are a number of different possible experiential motivations that could explain the occurrence of such expressions of non-actual motion (NAM) – a term that is neutral with respect to hypothetical cognitive processes, unlike notions of ‘fictive,’ ‘subjective,’ or ‘apparent’ motion. Linking ideas expressed by Talmy, Langackar, and Matlock to concepts in phenomenology (Blomberg and Zlatev 2014) distinguish between the following three motivations: i. the enactive, action oriented nature of perception; ii. the correlational (act-object) nature of intentionality; iii. imagination of counter-factual states, closest to a truly metaphorical, or ‘fictive’ reading of sentences like (1) and (2). While some NAM-expressions could be linked predominantly with one or another of (i–iii) there is nothing to restrict multiple motivations. Furthermore, as was the case with metaphor, pre-linguistic experiential motivations do not determine or constitute linguistic meanings, which conform to culture and language-specific conventions. For example, certain languages like Yucatec Maya severely restrict the possibility to use NAM-sentences (Bohnemeyer 2010). Finally, while many cognitive linguistic theories locate meaning in the (individual) head, neither (i) nor (ii) are, strictly speaking, either representational or individual. First, they are primarily perceptual, in the way that affordances (Gibson 1979) are: paths afford walking, as chairs afford sitting on, etc. In the case of figures that do not afford self-motion like mountain ranges, the process of ‘scanning’ their length is still mostly perceptual, or quasi-perceptual (as reflected in the terminology used by Langacker). Moreover, in line with the structures of embodied intersubjectivity discussed in section 11.2, we can affirm their intersubjective rather than private character. With respect to (i) the affordances of paths and highways are the affordances of cultural artifacts, which are not just ‘mine’ but inherently point to other embodied subjects:
Embodied Intersubjectivity From a phenomenological point of view, affordances are publically distributed, that is, taught and learned, patterns of interaction that are accompanied not only by perspectival experiences but also by the consciousness of the public nature and sharedness of these experiences. (Mo¨tto¨nen 2016: 160)
With respect to (ii), as perceiving any three-dimensional object, and not just cultural artifacts, implies the implicit awareness of other perspectives and possible co-perceivers (cf. section 11.2), we may say that in ‘scanning’ an elongated figure like a fence we are implicitly aware that others could do likewise, or perhaps differently. Scanning is thus something like a perceptual affordance. Such argumentation may appear as too theoretical for some tastes, but Blomberg (2015) has made it fully concrete in an elicitation-based study. Native-language speakers of Swedish, French, and Thai were asked to describe pictures such as those shown in Figure 11.1 in single sentences. Half of the pictures represented figures that afford self-motion (a, b), and half that do not (c, d). Crossed with this, the same objects were displayed either from a first-person perspective, 1pp (b, d) or third-person perspective, 3pp (a, c). The reasoning was that if only ‘mental simulation’ of actual motion determined the use of NAM-sentences, then perspective would not
(a)
(b)
(c)
(d)
Figure 11.1 Pictures eliciting non-actual motion descriptions, according to the two parameters Affordance and Perspective: (a) [+afford, 3pp], (b) [+afford, 1pp], (c) [-afford, 3pp], (d) [-afford, 1pp]
185
186
Z L AT E V
matter. On the other hand, if mental scanning were the most determinative factor, then most NAM-sentences would be produced in the 3pp condition. The results showed that all categories of pictures elicited NAMsentences on the scale of 40 percent for all three languages, significantly more than control pictures showing non-extended figures. All categories of target pictures were thus often described by NAM-sentences, but interestingly, for all three language groups, pictures like that in (b) – which most closely resemble a situation that affords self-motion – elicited more NAM-sentences than the others. Finally, the elicited NAM-sentences consistently relied on languagespecific conventions for expressing actual motion, but making this more ‘bleached’ by avoiding the use of Manner-verbs like run and crawl. In Swedish, this tendency was reflected in the common use of the two generic motion verbs gaº , ‘go’ and leda, ‘lead’. The French speakers used a wider range of motion verbs, including Path-verbs such as sortir, ‘exit’; Thai speakers used serial-verb constructions, but in most cases omitted the Manner-verb in the series, using only a Path-verb together with a deictic verb. In sum, the study confirmed the distinction between experiential motivations and conventional meanings, and the intersubjective-perceptual, rather than individual-representational character of the first. Mo¨tto¨nen (2016) extends a similar analysis to the key concept of construal as such, arguing that it should not be seen as an individual mental operation, but as an intersubjective one, on at least three levels: (i) on the level of perception – perspective presumes the co-existence of other perspectives and the identity of the referential object; (ii) on the pragmatic level of alignment in conversation – individual meaning-intending (referential) acts like that creature there and the dog differ in specificity, but remain co-referential; and (iii) on the level of conventionalized construals – sedimented through multiple instances of (i) and (ii) over individual and historical time. Even from this cursory summary, it can be seen how higher levels presuppose lower ones, with level (i) being essentially the level of pre-linguistic embodied intersubjectivity.
11.7 Conclusions This chapter shows the relevance of phenomenology in general, and of embodied intersubjectivity in particular, for cognitive linguistics. Some scholars from interaction studies have viewed the field as lacking in this respect: “from ‘embodied cognition’ to cognitive linguistics to microethnology: the paradigmatic importance of intercorporeality . . . has not even begun to be recognized” (Streeck 2009: 210). The analyses of the experiential grounding of mimetic and image schemas, metaphors, and dynamic construals presented here clearly show that this claim is an overstatement.
Embodied Intersubjectivity
On a more general theoretical level, our discussion has a number of important implications for the foundations of cognitive linguistics. First, from its onset (Lakoff and Johnson 1980a), the school has emphasized that language is ‘grounded in experience,’ but has not fully resolved the question whether physical or social experience is most essential. Armed with the concept of embodied intersubjectivity, we have argued that body and sociality are interlocked from the start, freeing us from the need to choose one or the other (see Zlatev 2016). Second, we have seen that some central experiential structures that underlie linguistic meaning should not be understood as ‘private’ processes (like neural simulation) that cannot be shared by definition, but as bodily intersubjective ones. To summarize: mimetic schemas involve intercorporeal and incorporative body memory; E M O T I O N I S M O T I O N , and other ‘primary’ metaphors are to be explained not so much with the help of invisible cross-domain mappings but through bodily resonance and Leibko¨rper duality; non-actual motion expressions are rooted not so much in individual representational cognition, but in perceptual intersubjectivity, involving affordances and the enactive perception. Third, as pointed out repeatedly, a clear distinction must be maintained between the pre-linguistic, intercorporeal processes and structures here described and the qualitatively different kind of meaning and intersubjectivity of language. The first are pre-predicative motivations, accounting for a degree of cross-linguistic overlap by functioning as attractors, without the need to postulate ‘linguistic universals’ (Evans and Levinson 2009a), while linguistic meanings are normative, systematic, and predicative (Blomberg and Zlatev 2014, Zlatev 2013, Zlatev and Blomberg 2016). Many a cognitive linguistic analysis has not always made this crucial distinction, conflating linguistic meaning with conceptualization. Critics of this have argued that the distance between the subjective Vorestellung and the conventional Sinn (to use the Fregean terms) is so great, that conscious experience becomes almost irrelevant for semantics (Itkonen 2003). Perhaps the greatest advantage of the phenomenology-inspired approach to language endorsed in the chapter is that it opens the door to a third way: on the one hand, pre-linguistic experience is not a matter of private Vorstellungen, but to a considerable degree of embodied intersubjectivity; on the other hand, linguistic meanings are not neutral or arbitrary conventions, since they emerge from and come to embody shared intersubjective perspectives, or construals of worldly objects and events.
187
12 Intersubjectivity and Grammar Ronny Boogaart and Alex Reuneker
12.1 Introduction From the start of cognitive linguistics, it has been one of the basic tenets that all use of language is subjective. By their choice of words and grammatical constructions, speakers unavoidably present a specific conceptualization, or construal (Langacker 1987: 487–88, Langacker 1990a: 61), of reality. However, in more recent years it has been argued that there is an additional dimension of language and communication that should not be neglected by cognitive linguists: the intersubjective dimension of ‘cognitive coordination’ between speaker and hearer (Verhagen 2005: 7). In fact, as we will show in this chapter, the function of many grammatical constructions can only be adequately understood if, in addition to the subjective dimension, the intersubjective dimension of language is taken into account. We will start out, in section 12.2, with a brief introduction to the notion of intersubjectivity and its relation to subjectivity and argumentativity. Then its relevance is shown for the analysis of negation, complementation, modality, and conditional constructions.
12.2 Subjectivity and Intersubjectivity 12.2.1 The Descriptive Dimension In most semantic traditions, specifically in formal semantics (see Heim and Kratzer 1998 for an introduction and Portner and Partee 2002 for an overview), the focus of research is on the descriptive dimension of language. Accordingly, language is analyzed as a referential tool, that is, a linguistic means for exchanging information about something. Ducrot’s (1996: 42) example below seems to be a clear case. (1)
There are seats in the room.
Intersubjectivity and Grammar
When this sentence is seen as a purely descriptive expression, its semantics can be described in terms of truth conditions: knowing the meaning of (1) equals knowing under which conditions the sentence is true, that is, knowing when there are indeed seats in the room. Such an approach sets out to “explain how linguistic expressions say things about the world” (Jackendoff 2002: 294). Whereas it seems undeniable that language may be used to describe the world, cognitive linguists have always questioned whether this descriptive dimension can provide the semantics of linguistic items. Moreover, describing the world may not be the primary function of language use. These questions are addressed in the following sections respectively.
12.2.2 The Subjective Dimension In response to an objectivist kind of semantics, cognitive linguists starting with Lakoff (1987) have pointed out that instead of expressing ‘things about the world,’ linguistic utterances tell us how the speaker conceives of, or construes, the world. One and the same situation in reality, such as that of seats being in a room, may be presented in many different ways, using different words or grammatical constructions, as in (2) and (3), presenting only two of an essentially infinite number of alternatives. (2) (3)
Seats are standing in the room. The room has seats.
It is hard to see how the alternative phrasings of (1), in (2) and (3), correspond to different truth-conditions and yet one would like to be able to represent the semantics of the presentative there-construction in (1), the effect of adding a progressive construction and a posture verb in (2), and the effect of taking the room rather than the seats as a ‘starting point’ for the sentence in (3). In the words of Langacker (2008a: 55): “Every symbolic structure construes its content in a certain fashion.” The meanings of linguistic elements, then, are to be identified with different construals of the world rather than with references to that world (Langacker 1991: 1–2). This account of linguistic meaning shifts the focus from reference and truth-conditions to construal and subjectivity, that is, from equating the meaning of a sentence with its truth-conditions (cf. Heim and Kratzer 1998: 1) to “equating meaning with conceptualization” (Langacker 1991: 1). The subjective dimension of language is captured by Langacker (2008a: 73–74) in his ‘viewing arrangement’ concept, as visualized in Figure 12.1. The diagram shows the conceptualizing subject (S), that is, the speaker, and the object conceptualized (O), that is, what the utterance ‘is about.’ Here, the bold line indicates that the object of conceptualization is profiled maximally, while the subject is not. In example (1), for instance, the fact that there are chairs in the room (O) is explicitly presented, while the speaker (S) is not mentioned. This should not, however, be taken as an
189
190
BOOGAART AND REUNEKER
S = subject of conceptualization O
O = object of conceptualization = full scope of awareness = “onstage” region = directing of attention
S Figure 12.1 Subjective and objective construal (adopted from Langacker 2008a: 260 by permission of Oxford University Press, USA)
indication that such utterances are objective expressions. There is always a subject of conceptualization directing attention and because in (1) this S is not ‘put onstage,’ the construal is in fact maximally subjective: the activity of focusing attention on the object of conceptualization is not itself explicitly addressed and thus S lacks ‘self-awareness’ (Langacker 2008a: 260). In sentences such as (4), on the other hand, the speaker is ‘onstage.’ (4)
I think that there are seats in the room.
Here, the speaker is part of the conceptualization and may thus be considered ‘objectified’: she is less subjectively construed than in (1), (2), and (3).
12.2.3 The Intersubjective Dimension The emphasis on subjectivity and conceptualization may suggest that Langacker’s approach is entirely speaker-oriented. However, the notion of ‘subject of conceptualization’ in Figure 12.1 is an abstraction. It is made more specific and concrete in Figure 12.2, illustrating that both speaker (S) and hearer (H) are the ‘primary conceptualizers.’ The ‘general locus of attention’ of an expression is captured in the maximal scope (MS), while the immediate scope (IS) includes only what is put onstage and is directly relevant (Langacker 2008a: 63). The speech situation, including the interaction between speaker and hearer – as symbolized by the arrows going from S to H and vice versa – constitutes the Ground (G) of the discourse: it is always there, but it may be more or less explicit in linguistic expressions and, therefore, less or more subjective (as in (4) versus (1)–(3)). Langacker, then, clearly acknowledges the role of both speaker and hearer and in fact ascribes a “dynamic, intersubjective, context-dependent” nature to meaning construction in actual discourse (2008a: 28).
Intersubjectivity and Grammar
MS IS
S
H
G
Figure 12.2 Basic scheme for the meanings of expressions (adopted from Langacker 2008a: 261 by permission of Oxford University Press, USA)
However, when one thinks about the reasons for spending cognitive effort on producing linguistic expressions, neither a truth-conditional nor an exclusively ‘subjective’ analysis will suffice. Why would S in (1) present her description or conceptualization of reality to H in a linguistic utterance, if not to achieve certain effects with that utterance?1 What would be the point of S exchanging information about seats in the room (O) to H? It is basically in view of such questions that Verhagen (2005) proposes a modified version of Langacker’s viewing arrangement, as visualized in Figure 12.3.
O: Object of conceptualization:
S: Subject of conceptualization (Ground):
1
2
Figure 12.3 Construal configuration and its basic elements (adopted from Verhagen 2005: 7 by permission of Oxford University Press, USA) 1
Also see Croft’s (2009b) notion of “construal for communication” and Harder’s (2010) “social turn in cognitive linguistics.”
191
192
BOOGAART AND REUNEKER
Verhagen’s construal configuration incorporates both the descriptive and the subjective dimension of language use. The latter is represented by the vertical line in the middle connecting the Ground (the subjects of conceptualization; the S-level) to the descriptive contents of the utterance (the objects of conceptualization; the O-level). However, more than in Langacker’s account, this is a shared perspective between language users: speaker and hearer engage in what is called “cognitive coordination” (Verhagen 2005: 7), as represented by the horizontal line connecting Speaker and Hearer. In this view, the goal of linguistic communication is to invite the other “to jointly attend to an object of conceptualization in some specific way, and to update the common ground by doing so” (Verhagen 2005: 7). This means that the speaker invites the hearer to change his cognitive system by drawing inferences evoked by the linguistic utterance used, and to adjust the common ground accordingly. In fact, these inferences, at the S-level, rather than the linguistically coded, descriptive content of the utterance, at the O-level, constitute the point of the utterance. When language is seen from this perspective, as a social instead of an informational tool, the focus of analysis automatically shifts from its referential properties and its subjective, perspectival properties to its intersubjective dimension: a speaker expresses (1) not to describe a room containing seats, or only subjectively to construe this situation in some way, but to invite an interlocutor to draw inferences about, for instance, the comfort provided in the room. This is not ‘merely’ a case of pragmatics, like indirect speech acts or pragmatic implicatures put ‘on top of semantics,’ but a systematic effect of language itself, as can be exemplified by the use of the conjunction but in (1a).2 (1a)
There are seats in the room but they are uncomfortable. (Ducrot 1996: 42)
The use of but here demonstrates opposing orientations of the two connected utterances; the first induces positive inferences (e.g. ‘the possibility of sitting down’), while the second cancels such inferences by inducing negative ones (e.g. ‘the impossibility of sitting down’). Because of these opposed orientations, use of the contrastive conjunction but produces a coherent discourse. Now consider (1b), in which but is replaced by and moreover. (1b)
# There are seats in the room and moreover, they are uncomfortable.
Because the connective here is not contrastive but additive, the second utterance is incompatible with the inference licensed by ‘there are seats in the room.’ This shows that (1) is not the neutral (objective), nor merely subjective utterance it seemed to be. The utterance, in and of itself, has a
2
For an elaborate discussion on this point in reply to criticism by Hinzen and Van Lambalgen (2008), see Verhagen (2008a).
Intersubjectivity and Grammar
certain ‘argumentative orientation’: it is meant to trigger specific inferences rather than others. Otherwise, it could not be explained why only the use of but makes (1a) a coherent sequence. Consequently, the reverse is true when the second utterance invites inferences which are in accordance with those of the first. (1c) (1d)
# There are seats in the room but they are comfortable. There are seats in the room and moreover, they are comfortable.
From a purely descriptive point of view, it is unclear why such systematic differences exist, but from an intersubjective perspective, the meaning of words resides in their contribution to the argumentative orientation of an utterance (Ducrot 1996: 27); these orientations are opposed in (1a), requiring a contrastive conjunction like but, while they are similar in (1d), requiring an additive connective. Intersubjectivity, in this view, thus relates to the participants in linguistic communication and consists of the mutual influence they exert on each other’s cognitive systems (Verhagen 2005: 26).3 The nature of this influence is called argumentative, since utterances are conceived of as arguments for conclusions, thus as means to invite the discourse participant to draw certain inferences. This argumentative nature of language is what characterizes the relation – or cognitive coordination – between subjects of conceptualization.
12.3 Intersubjectivity in Grammatical Constructions If, as Du Bois (1985: 363) argues, “grammars code best what speakers do most,” and cognitive coordination is always involved in language use, it is to be expected that grammatical constructions encode meaning on the level of intersubjectivity (Verhagen 2005: 4).4 Including this level of analysis sheds light on persistent problems occurring in the more traditional analysis of grammatical constructions. To illustrate this point, the following sections will discuss a number of widely studied linguistic phenomena from this perspective, starting with negation.
12.3.1 Negation Negation is an extensively studied phenomenon (see Horn 2010 for an overview) and it makes intuitive sense to attribute to negation a truth-conditional semantics. In terms of the construal configuration in Figure 12.3, this would boil down to the speaker using a negative statement to inform the hearer that
3 4
Verhagen (2015) discusses the relationship between this notion of intersubjectivity and other interpretations. The role of speaker–hearer interaction in grammar is also emphasized in Du Bois’s (2014) dialogic syntax, the notion of fictive interaction introduced by Pascual (2014) and in the framework of interactional construction grammar (see Deppermann 2006, cf. Boogaart, Colleman, and Rutten 2014: 9–11).
193
194
BOOGAART AND REUNEKER
something is not the case in ‘the world,’ thus at the level of the object of conceptualization. In principle, this is not incompatible with the general argumentative perspective on language use sketched in the previous section. After all, at the S-level, H could combine such negative information about O with shared cultural and contextual knowledge to draw all kinds of relevant inferences. However, Verhagen’s (2005: 28–77) analysis of negation and related negative expressions takes a crucial step further, as he argues that linguistic meaning itself, including that of negation, may be conventionally associated directly with the S-level. More specifically, rather than concerning the connection between language and the world, negation regulates the relation between distinct “mental spaces” (Fauconnier 1994) that, for this purpose, may be identified with conceptualizers 1 and 2 in the construal configuration of Figure 12.3 (Verhagen 2005: 30–1). By using negation, S instructs H to entertain two different cognitive representations – both the negative one and its positive counterpart – and to adopt the first while abandoning the second.5 The fact that negation (¬ p) triggers the construction of a mental space in which its positive counterpart (p) is valid, is evidenced by the observation, in (5), that both p and ¬ p are available for reference in the subsequent discourse. (5)
This time, there was no such communication [about the plans]. It’s a pity because it could have resulted in greater participation by the employers. (Verhagen 2005: 29)
Whereas the first it in the second sentence of (5) refers to the fact that there was no communication, the second it refers to the presence of communication (that could have resulted in greater participation). Apparently, both are made cognitively accessible by the first sentence. Additional evidence for this claim may be adduced by the behavior of on the contrary in (6) (Verhagen 2005: 31). (6a) (6b) (6c)
Mary is not happy. On the contrary, she is feeling really depressed. #Mary is sad. On the contrary, she is feeling really depressed. #Mary is unhappy. On the contrary, she is feeling really depressed.
In (6a), the fact that Mary is really depressed is not the ‘contrary’ of the fact that she is not happy. In fact, it is contrary to Mary being happy, that is, to the positive counterpart of the situation expressed in the first sentence. This second mental space must be activated by the use of sentential negation: it is not made available by the lexical item sad in (6b) or by the morphological negation (unhappy) in (6c). Since ‘feeling really depressed’ is not contrary to being ‘sad’ or ‘unhappy,’ and no other space is available for reference, (6b) and (6c) are incoherent. Use of sentential negation, then, is intrinsically argumentative since S explicitly – as part of the linguistic meaning of negation – activates a standpoint in order to oppose it. As Dancygier (2012b) argues, there is a general correlation between linguistic 5
Experimental evidence for this is provided by Beukeboom, Finkenauer, and Wigboldus (2010).
Intersubjectivity and Grammar
elements triggering alternative spaces, like negation, and their use as argumentative or stance device. The crucial role of the intersubjective dimension rather the descriptive dimension of language in the system of negation and related expressions is further illustrated by the sentences in (7) (adapted from Verhagen 2005: 42–47). (7a) (7b) (7c) (7d)
Our son did not pass the exam. Our son barely passed the exam. Our son almost passed the exam. Our son passed the exam.
In section 12.2, it was shown how, from an argumentative perspective on semantics, the function of linguistic items is to trigger inferences on the part of H. In (7a), then, S is not so much sharing the information about the son’s result as he is trying to make H infer the consequences thereof. The exact content of these inferences may differ, depending on contexts and cultural models, but negation as such has a ‘negative argumentative orientation.’6 Interestingly, this is true also of barely in (7b). If we look at this utterance from a descriptive perspective, it is clear that the son did pass his exam. However, the inferences S wants to trigger in uttering (7b) are actually similar to the ones in (7a), that is, they are inferences that would follow from the son not passing, albeit it in a weaker form. While (7a) and (7b) may differ in argumentative strength, they share their negative argumentative orientation. Conversely, in (7c), almost passed means that the son did not pass the exam and yet the utterance has a positive ring to it that is lacking in (7b). This is because almost has a positive argumentative orientation: S wants H to draw inferences that follow from our son passing the exam, as in (7d), but the positive argument in (7c) is of course weaker than that of the unmitigated utterance in (7d). It is important to note that positive and negative argumentative orientation are context-independent functions of ‘argumentative operators’ like almost and barely that concern the polarity of the associated inferences rather than their evaluation. Thus, the negative orientation of barely in (8a) triggers inferences associated with ‘not failing the exam’ and the positive orientation of almost in (8b) triggers inferences of the kind that ‘failing the exam’ would. (8a) (8b)
Our son barely failed the exam. Our son almost failed the exam.
From the perspective of evaluation, these inferences will probably be relatively hopeful in (8a) and somewhat disturbing in (8b), opposite to those in (7b) and (7c). This is determined by general cultural knowledge and more specific contextual information. The argumentative orientation 6
In a somewhat different but compatible way, this notion is used in experimental research on attribute framing (e.g. the difference between ‘half full’ and ‘half empty’) by. for example, Holleman and Pander Maat (2009).
195
196
BOOGAART AND REUNEKER
of the linguistic items involved is, however, constant and can thus be considered part of the linguistic system.
12.3.2 Complementation Constructions In most syntactic theories, both traditional and modern, complementation constructions, such as in (9), are basically analyzed as simple clauses with a clause instead of a noun phrase as direct object, as in (10) (Verhagen 2005: 78). (9) (10)
George said that his opponent was closing in. George said something.
Consequently, any difference between (9) and (10) is attributed to the difference between clauses and noun phrases, arguing that there are no crucial differences between simple transitive clauses and complementation constructions; as in (10), the matrix clause in (9) describes an event of saying, rendering the complement clause as subordinate to it. However, Thompson (2002) shows that complementation constructions function differently in discourse; finite complements such as ‘it’s cool’ in (11) do not have lower prominence than the matrix clause they are subordinate to. Rather, they present a ‘common object of attention’ to which the matrix clause adds epistemic stance. (11)
(talking about a photo collage on the wall) Terry: I think it’s cool. Abbie: it i = s cool. Maureen: it i = s great. (Thompson 2002: 132)
This view is corroborated by Diessel and Tomasello’s (2001) observation that the earliest uses of complementation constructions by children include marking of epistemic stance and illocutionary force. These findings suggest that, when viewed in terms of the construal configuration in Figure 12.3, complementation constructions do not represent an event (of thinking, saying) as an object of conceptualization, but they invite the hearer on the level of subjects of conceptualization to adopt the perspective (or ‘stance’) of the onstage conceptualizer (Verhagen 2005: 97). Consider the following example from Verhagen (2005: 107), which, in a referential analysis, would amount to analyzing all of B’s reactions to A’s question as references to the world: (B1) refers to the scheduled time, (B2) to the belief of the speaker, and (B3) to the speech act of John. (12)
A: Will we be in time for the launch? B1: It was scheduled for 4 p.m., so we still have lots of time. B2: I think it was scheduled for 4 p.m., so we still have lots of time. B3: John said it was scheduled for 4 p.m., so we still have lots of time.
Intersubjectivity and Grammar
O: Object of conceptualization:
S: Subject of conceptualization (Ground):
1
2
Figure 12.4 Construal configuration for a non-perspectivized utterance (adopted from Verhagen 2005: 106 by permission of Oxford University Press, USA)
From an intersubjective point of view, the differences between B’s expressions cannot be adequately expressed in terms of references to the world. Rather, all of B’s expressions are considered means to invite the same inference (i.e. ‘we still have lots of time’). They have the same argumentative orientation, while they differ in argumentative strength, that is, the force with which the hearer is invited to draw the inference. The non-embedded clause in B1’s response presents the strongest argument, because the information profiled at the level of objects of conceptualization is shared directly between conceptualizers 1 and 2.7 The speaker is not put onstage as subject of conceptualization (see section 2.2), which is represented by dotted lines in Figure 12.4, while the object of conceptualization is profiled, represented by bold lines. In B2, the speaker is put onstage by the expression of the matrix clause (I think), presenting a perspective on the object of conceptualization in the complement clause. The indirect introduction of the object of conceptualization into the Ground opens up the possibility of a difference between the speaker’s perspective and reality, decreasing the force with which the hearer is invited to draw the intended inference. This is visualized in Figure 12.5, in which both the object of conceptualization and the firstperson perspective are profiled. In B3, the relation of the complement to the Ground is even more indirect. The speaker temporarily adopts a third person’s perspective, as represented by the arrow in Figure 12.6 (for a more elaborate intersubjective model of third-person perspectives, see Van Duijn and Verhagen in press). It is crucial here that this perspective is not analyzed on the level of the object of conceptualization, but on the level of subjects of conceptualization; that is, the object of conceptualization in B3 is shared between 7
In Langacker’s terms, both B1 and B3 would be ‘maximally subjective,’ as the subjects are not linguistically referred to. One benefit of the intersubjective approach is that it is able to explain the difference between B1 and B3 on the level of subjects of conceptualization.
197
BOOGAART AND REUNEKER
think
O: Object of conceptualization:
S: Subject of conceptualization (Ground):
1
2
Figure 12.5 Construal configuration for first-person perspective (adopted from Verhagen 2005: 106 by permission of Oxford University Press, USA)
O: Object of conceptualization:
said
198
S: Subject of conceptualization (Ground):
1←3
2
Figure 12.6 Construal configuration for third-person perspective (adopted from Verhagen 2005: 106 by permission of Oxford University Press, USA)
conceptualizers 1 and 2 through the temporary adoption of the perspective of onstage conceptualizer 3. Consequently, the lower argumentative strength results from the possibility of a difference between the perspective of the speaker and the onstage conceptualizer. By including both the intersubjective and the objective dimension in the analysis of B’s reactions in (12), their grammatical differences can be explained in terms of argumentativity: the simple clause and complementation constructions share the same orientation, while they differ in strength. What this shows is that the parallel between simple clauses and complementation constructions stems from a theoretically motivated desire to describe constructions syntactically in terms of general rules, and semantically in terms of references to the world. The linguistic expression of viewpoint by means of complementation constructions can be analyzed more adequately in terms of negotiation between Speaker and Hearer (cf. Sweetser 2012: 6, Dancygier 2012b). It shows that complements operate on
Intersubjectivity and Grammar
the level of objects of conceptualization, while the matrix clauses present the speaker’s stance toward it (sometimes indirectly, through another point of view). The intersubjective approach has also been successful in analyzing the more general phenomenon of speech and thought representation. Vandelanotte (this volume Ch. 10) remarks that viewpoint “is construed intersubjectively, in a negotiation with other participants in a given speech event.” Third-party perspective in narrative texts, for instance, is analyzed in terms of the mutual coordination of perspectives (see Vandelanotte 2009, Dancygier 2012a, 2012b, Dancygier and Vandelanotte 2016, Van Duijn and Verhagen in press, and several contributions to Dancygier, Lu, and Verhagen 2016). When we finally return to the difference between (9) and (10) at the start of this section, we see that (10) is not ‘just’ a case of a direct-object slot filled by a clause instead of a noun phrase, but a non-perspectivized invitation to adopt the claim made by the speaker, resulting in maximal argumentative strength, while (9) is an invitation to adopt George’s perspective temporarily adopted by the speaker, resulting in lower argumentative strength. Consequently, the function of complementation constructions is to link the intersubjective dimension of communication, linguistically expressed in a matrix clause, to the objective dimension of communication, expressed as a complement.
12.3.3 Modality Modality is crucially concerned with the speaker’s perspective on reality and, thus, with construal and subjectivity. (See Boogaart and Fortuin 2016 for an overview of mood and modality in cognitive linguistics.) More specifically, within Cognitive Grammar, Langacker (1991: ch. 6) has argued that English modal verbs are grounding devices connecting the object of conceptualization to the Ground of the discourse. As such, they may be called subjective since they express the speaker’s assessment of the world, without the speaker and the Ground being profiled. Furthermore, different uses of modal verbs are assumed to differ in the extent to which they subjectively construe elements of the Ground. For instance, the use of must in (13a) is considered less subjective than the use of must in (13b). (13a) (13b)
He must be home by 6, so he should really go now. He must be home since the lights are on.
The deontic use of must in (13a) refers to an obligation the subject referent has in the world, whereas the function of epistemic must in (13b) is to attribute a certain degree of probability to the situation described. Since the latter is entirely a matter of reasoning and evaluation by the speaker, the interpretation of must in (13b) is more dependent on the speaker than is the interpretation of must in (13a). Synchronically, the different uses of must exemplified in (13) can thus be described in terms of their different
199
200
BOOGAART AND REUNEKER
degree of subjectivity and, diachronically, the development of epistemic uses of modals from non-epistemic uses is an instance of subjectification (Traugott 1989, 1997). Given the characterization of (13a) and (13b), that is in fact more generally applicable to the difference between non-epistemic and epistemic modality, it will be clear how these different meanings may be related to Verhagen’s construal configuration in Figure 12.3: non-epistemic modals (also) profile an element at the O-level, while epistemic modals are concerned with the construal relation and the relationship between S and H at the S-level. However, in contrast to the subjective dimension, the latter, intersubjective dimension of modality has so far not received much attention in the literature. We want to mention two, interrelated ways in which this framework could further contribute to our understanding of modals. First, taking Verhagen’s claim on the argumentative nature of language use seriously, we should treat utterances as a means S uses to trigger specific inferences on the part of H. In the general sense outlined in section 12.2, this is true for utterances with or without modals, but it seems that modal verbs constitute a conventional linguistic system, like the system of negation, that speakers use to provide arguments for conclusions with greater or lesser argumentative strength (see Rocci 2008 for a compatible perspective in argumentation theory). In (13a), it was already shown how a non-epistemic modalized utterance may be used as an argument for a conclusion: S mentions the obligation to be home on time to motivate the urgency to leave. Of course, epistemic utterances such as (13b) are used to trigger inferences in a highly similar way. For instance, depending on the context, a relevant inference may be the one made explicit in (14). (14a) (14b)
He must be home now, so this is a good time to try and talk to him. He could be home now, so this is a good time to try and talk to him.
As the difference between (14a) and (14b) shows, the epistemic modal system comprises both necessity modals and possibility modals and this provides S with the possibility of distinguishing between stronger and weaker arguments for basically the same conclusion. This parallels, in fact, the difference between sentential negation and words like barely, discussed in section 12.3.1, and between the various kinds of matrices in the complementation construction from section 12.3.2. A further point of comparison between modality and negation concerns the fact that by using a modal, S explicitly introduces different possible scenarios, or mental spaces. From the perspective taken here, these can be regarded as competing standpoints and, like in the case of negation, this makes the use of modals inherently argumentative. Another way in which the intersubjective perspective on language may be helpful in the domain of modality relates to the well-known polysemy of modals, as was already illustrated in (13), and the fact that it is often very hard to distinguish between ‘objective’ (non-epistemic) and ‘subjective’
Intersubjectivity and Grammar
(epistemic) uses of these verbs. This is true also for the two uses of the Dutch verb beloven (‘promise’), illustrated in (15) (Verhagen 2000, 2005: 19– 24; cf. Traugott 1997 and Cornillie 2004 on its English and Spanish equivalents respectively). (15a) (15b)
Het debat belooft spannend te worden. ‘The debate promises to be exciting.’ Hij belooft de grondwet te verdedigen. ‘He promises to defend the constitution.’
In (15b) the verb describes an actual act of promising at the O-level, whereas in (15a) the use of beloven is epistemic in the same sense as must in (13b): it is confined to the S-level of intersubjective coordination. In addition to such clear cases of ‘objective’ and ‘subjective’ beloven, however, there are many intermediate cases that allow for both readings and, when asked about them, speakers of Dutch do not really agree on where to draw the line (Verhagen 2005: 21). This situation, then, is very much like the distinction between epistemic and non-epistemic uses of the core modals, which are often very hard to distinguish as well. (The problems are discussed at length by Boogaart 2009.) Now, with respect to beloven, Verhagen makes the interesting point that, in practice, the distinction between its subjective and objective use does not make much difference for communication to be successful, since the different readings share their argumentative orientation at the S-level. Just like, in (15b), the act of promising counts as an argument strengthening the expectation that the constitution will be defended, the contribution of beloven in (15a) consists exclusively of this argumentative orientation. Now, if linguistic elements are primarily meant to trigger certain inferences at the S-level, it is clear that the problem of the polysemy of modals does not really have to be a problem for communication. The different uses of modal verbs may occupy different positions on the scale from O-level to S-level,8 but if their contribution consists of their argumentative orientation and strength, there is no need to determine their exact position on this scale and language users do not have to agree on this for communication to be successful.
12.3.4 Conditional Constructions In formal-semantic traditions, conditionals are analyzed in terms of truth conditions (for an overview, see Bennet 2003, Von Fintel 2011 and Kratzer 2012). The main distinction made is that between indicative and subjunctive conditionals, as in (16) and (17) respectively (Von Fintel 2011: 1517). (16) (17)
8
If Grijpstra played his drums, de Gier played his flute. If Grijpstra had played his drums, de Gier would have played his flute.
Modal verbs used as hedging device or politeness strategy are clear examples of the most intersubjective use (Traugott 2003, cf. Sweetser’s 1990 notion of speech-act modality).
201
202
BOOGAART AND REUNEKER
The difference between (16) and (17) is that the latter carries an implicature of counterfactuality (i.e. it is at least suggested that Grijpstra did not play his drums; cf. Comrie 1986), while the former leaves either possibility open. The similarity between (16) and (17) is that the situation expressed in the main clause or consequent is caused or enabled by the situation expressed in the if-clause or antecedent; that is, the truth of the consequent depends on that of the antecedent. This focus on the descriptive dimension works for hypothetical conditionals, as in (16), (17), and (18) below, but it encounters problems when other uses of conditional constructions are considered, as exemplified in (19) to (21). (18) (19) (20) (21)
If Mary goes, John will go. (Sweetser 1990: 114) If she’s divorced, (then) she’s been married. (Sweetser 1990: 116) If you need any help, my name is Ann. (Dancygier and Sweetser 2005: 113) My husband, if I can still call him that, hates onion soup. (Dancygier 1998: 106)
While in (18) the situation expressed in the antecedent (‘Mary goes’) causes the situation expressed in the consequent (‘John will go’), in (19) to (21) there is no such direct causal relation; that is, being divorced does not cause being married, needing help does not cause someone to go by the name Ann, and being able to call someone your husband does not cause him to hate onion soup. Consequently, it cannot be argued that the antecedent and consequent of conditionals in general are connected by direct causality. In cognitive linguistics therefore, the causal character of predictive conditionals, as exemplified in (18), is taken to be the prototype for other, less central (i.e. non-predictive) types of conditionals (cf. Dancygier 1998: ch. 7; also see Athanasiadou and Dirven 1997). Dancygier and Sweetser (2005), building on earlier work by Sweetser (1990) and Dancygier (1998), provide an analysis of different conditionals as ‘mental space builders’ (cf. Fauconnier 1994) in different domains. Contrary to the predictive nature of (18), in (19) there is a less direct connection between antecedent and consequent. Knowledge of the truth expressed in the antecedent enables the conclusion expressed in the consequent and accordingly, these inferential conditionals (cf. Dancygier 1998) operate in the epistemic domain of reasoning rather than in the domain of real-world causality. In (20), the relation between antecedent and consequent is even more indirect: the former addresses a felicity condition (cf. Austin 1962) for uttering the latter and as such, it functions in the domain of speech acts and is consequently known as a speech-act conditional. The last type of non-predictive conditional discussed by Dancygier and Sweetser (2005) is the metalinguistic conditional exemplified in (21), in which the antecedent comments on the appropriateness of the linguistic form of the consequent, thus operating in the domain of metalinguistic communication. What can be observed, then, is a decreasing directness in the relation between antecedent and consequent in (18) to (21). Many classifications
Intersubjectivity and Grammar
essentially describe this relation exclusively in terms associated with the level of objects of conceptualization within the construal configuration (e.g. necessity, sufficiency, recurrence, fulfilment). However, the domain approach by Dancygier and Sweetser is compatible with the intersubjective approach to grammar. The degree of directness from predictive to metalinguistic conditionals can be said to be inversely proportional to the degree of intersubjectivity. In predictive conditionals, there is not only an intersubjective component, as in all utterances, to the relation between antecedent and consequent, but this relation clearly also resides on the level of objects of conceptualization in the construal configuration. There is a real-world causal link between antecedent and consequent and the degree of intersubjectivity is relatively low. In epistemic conditionals, the relation between antecedent and consequent is primarily construed at the level of subjects of conceptualization, that is, the speaker construes one object of conceptualization as an argument for another, based on a realworld causal connection and therefore the degree of intersubjectivity is higher. In speech-act and metalinguistic conditionals, the relation resides solely on the intersubjective level, that is, relating a felicity condition in the antecedent to a speech act in the consequent or commenting on the linguistic form of an utterance. The added value of the intersubjective approach is that conditionals can be analyzed on more than one level at the same time. This makes it possible to explain why different constructions are used to express conditional relations. The function of various constructions can differ from if on the level of both objects and subjects of conceptualization. For instance, paratactic conditionals, as in (22), may be paraphrased by means of if while maintaining their relation on the truth-conditional and the domain level (i.e. a predictive relation between antecedent and consequent). (22) (23)
Break that vase and I will break your neck. (Fortuin and Boogaart 2009: 642) If you break that vase, I will break your neck.
However, (22) and (23) clearly differ on the intersubjective level: the former is a stronger threat than the latter is. The grammatical form used (a combination of a directive imperative and parataxis with and) functions as a grounding element in the same way as wording might reflect the stance of the speaker toward what is expressed (e.g. commie versus communist; Langacker 2008a: 262). It thus has the same argumentative orientation as the if-conditional in (23), but it differs in strength, that is, the speaker directs the hearer more strongly toward the intended inference not to break the vase, making the paratactic construction particularly suitable for threats. In contrast to focusing on antecedents, consequents, and their relations, this approach enables the analysis of a conditional construction as a whole, showing its functions in discourse. It contributes a more
203
204
BOOGAART AND REUNEKER
grammatical perspective to the growing number of studies on the use of conditionals as threats and advice (e.g. Ohm and Thompson 2004, Evans 2005, Haigh et al. 2011) by explaining the function of conditionals in terms of argumentative orientation (positive and negative) and argumentative strength.9 In line with Dancygier’s (2008a) suggestion to relate different conditional constructions to partly overlapping construal configurations, it is expected that differences between prototypical if and other conditionals, including for instance the conditional use of prepositional phrases as exemplified in (24), might be explained in a similar way. (24)
That course is mandatory: without a license, the couple will not be permitted to marry. (Reuneker 2016: 126)
Notions from more traditional analyses of conditionals reside mainly on the level of objects of conceptualization. In the construal configurations central to the approach presented here, this does not make them incompatible with more pragmatic analyses of conditionals in terms of, for instance, desirability (Akatsuka 1999) and control over the consequent (Ohm and Thompson 2004). These notions function mainly on the level of subjects of conceptualization, and by combining both levels, the intersubjective approach to language may enable a next step in the analysis of conditional constructions in language use.
12.4 Conclusion In this chapter, the notion of intersubjectivity was used both in a general sense and in a more specific, linguistic sense. In general, the term describes communication as cognitive coordination between two subjects of conceptualization. A speaker invites a hearer to construe an object of conceptualization in a certain way (Langacker) and to update the common ground with the inferences that follow from this specific perspective on reality. This intersubjective dimension (the relation between Speaker and Hearer) is mostly neglected in accounts that focus on either the descriptive dimension of language (the object of conceptualization) or on the subjective dimension (the relation between Speaker/Hearer and the object of conceptualization). Following Anscombre and Ducrot (1983), the intersubjective relation may be regarded as argumentative since, in this view, utterances are meant primarily to invite the hearer to draw certain conclusions. In a specific, linguistic sense – and this is the main contribution of Verhagen’s (2005) work – the meaning of grammatical elements may operate directly on the intersubjective dimension: many grammatical 9
On a more speculative note, Mercier and Sperber (2011) argue that, in the evolution of language, the argumentative use of conditionals may even have preceded their use in reasoning.
Intersubjectivity and Grammar
constructions exhibit an argumentative orientation restricting the inferences the hearer is supposed to make, and an argumentative strength providing weaker or stronger arguments for these conclusions. By way of illustration, we demonstrated how Verhagen applies this perspective to the study of negation and complementation, and we explored how it could be extended to research on modality and conditional constructions. All grammatical constructions discussed show that there is an intimate connection between alternative spaces, viewpoint, and argumentativity. What we have presented in this chapter illustrates that moving beyond the descriptive and the subjective dimension of language to the intersubjective dimension may be fruitful in both solving some long-standing problems in the study of grammar and understanding the very essence of human communication.
205
Part III
Aspects of Linguistic Analysis
13 Opening Commentary: Linguistic Analysis John Newman
A cognitive linguistic orientation has proven to be applicable and revealing in describing, and theorizing about, language at all levels, from phonetic detail to discourse, including the interaction of spoken language and gesture. Cognitive linguistics is especially suited to linguists who seek a comprehensive account of language in all its semantic and pragmatic richness, as used in all contexts (conversation, narrative writing, poetry, advertising, etc.). From its inception, cognitive linguistics attracted scholars who were eager to explore language in its larger setting (as opposed to decontextualized samples of language) and in its actual use (as opposed to constructed examples drawn solely from introspection), and who were prepared to consider a broad range of cognitive, functional, and external factors that might influence language behavior. At the time when the cognitive linguistics movement came into being, culminating in the establishment of the International Cognitive Linguistics Association in 1989, the field of linguistics was not so accustomed to viewing language in its communicative entirety, and cognitive linguistics seemed very much out of the ordinary. I believe things are altogether different now, and an openness to exploring the full richness of language (actual usage, poetic language, mental processing of language, the dynamics of interaction of speech participants, etc.) is more readily accepted by linguists. Linguistic analysis within cognitive linguistics is permeated by the same trends affecting the movement as a whole. Typical preferences evident in this research include: an interest in a broad spectrum of language facts, as opposed to focusing on only a ‘core’ set of language phenomena or focusing exclusively on clause-level phenomena like argument structure (even if one might still choose to specialize in some more narrowly defined phenomenon); seeking explanations in general cognitive principles rather than positing, say, syntax-specific principles that have no cognitive counterpart; acknowledging a key role for usage facts when it comes to describing language phenomena; grammar understood as meaningful and, as
210
NEWMAN
such, comparable to the lexicon, without recourse to so-called “empty” categories; meaning understood as a dynamic process, rather than something fixed and invariable; an interest in experimental research that sheds light on mental processes, rather than carrying out formal analysis for the sake of formal analysis; a larger place in linguistic analysis for a more quantitative style of analysis, an inevitable outcome when corpora and experimental findings are incorporated into analyses; the desirability of multiple methodological approaches in attempts to fully describe language phenomena. These traits are evident in a number of the chapters that follow. Even so, there is a place for more conceptual approaches; for example, the work of Ronald Langacker, in developing broad theory. The pairing of form and meaning as an integrated whole – which I will refer to as a ‘constructionist‘ approach – is evident in almost all linguistic analysis in cognitive linguistics. In its most literal interpretation, “form” in the form-meaning pairing so central to constructionist approaches applies to phonetic form. This could take the shape of a stretch of phonetic transcription such as [tri] representing the phonetic component of a word tree. The phonetic form is susceptible to study in ways similar to that used for semantic analysis. One can, for example, consider the asymmetry of autonomous versus dependent components of a syllable, figure-ground auditory perception, more prototypical versus less prototypical members of a phoneme category, etc. Attention is now being paid to more subtle form-meaning pairings in the form of correlations between classes of sounds (obstruents, nasals, etc.) and word classes that are grammatically or semantically defined (gender classes, parts of speech, etc.), relying on quantitative analyses to establish the statistical significance of such correlations. Cognitive linguists would be naturally inclined to seek out connections between such phonetic detail and subtle nuances in meaning, grammatical classes of words, cues in turn-taking in conversation, etc. A constructionist orientation encourages an interest in the details of how the stem meaning and affix meaning are integrated into the meaning of a whole word and how new meaning is created in the process, especially the role of conceptual blending within the whole word. An important finding that has emerged from psycholinguistic studies on morphology is that morphological structure – specifically, decomposability of words into stems and affixes – is inherently a gradient phenomenon (Hay and Baayen 2005). A word such as invulnerable, for example, is judged by speakers to be relatively “more complex” than incongruous, even if linguists would consider both words to consist of a prefix in- + stem. A key to understanding the basis for the different evaluations by the experimental participants lies in the critical difference in the frequency of usage of the prefixed form versus the unprefixed form. The frequency of the stem vulnerable is greater than the frequency of the prefixed form invulnerable, whereas with incongruous, the opposite holds: incongruous is considerably
Linguistic Analysis
more frequent than congruous. These differences in usage of the unprefixed versus the prefixed forms leads to different psychological realities when it comes to recognizing morphological parts of the affixed word. To the extent that one can generalize from this result, the implications are highly significant for any approach to morphological decomposition that gives weight to speakers’ psycholinguistic realities as opposed to the theoretical constructs of linguists that are built solely on linguists’ intuitions. Lexical semantics has always been prominent within cognitive linguistics, offering a much richer view of meaning than traditional dichotomies such as sense versus reference, or connotation versus denotation, allowed. A distinctively cognitive linguistic approach to semantics is the explication and elaboration of polysemy, the individuation of the sub-senses of a meaning, and the overall architecture of the polysemy network. Polysemy is most prototypically associated with the word level, but it is applicable to the morpheme level and the larger constructional level, where one speaks of constructional polysemy. The polysemy associated with derivational morphemes or case marking is no less elaborate than the polysemy we find in words. Thus, one can explore the range of senses of a case morpheme in a language or an affix such as English -er (baker, ruler, locker, diner, etc.), identifying prototypical and schematic meanings and paths of extension of meaning, either as attested diachronically or, in a more speculative way, inferred from synchronic patterning. Even a very familiar semantic relation such as synonymy has been injected with new interest through experimental and corpus-based methods that seek to explore the degrees of semantic relatedness between multiple approximate synonyms, rather than just assuming pairs of synonyms. The usage-based approach making use of corpora allows for a much richer understanding of semantic relations between words than was possible with traditional context-independent notions such as polysemy or antonymy. Cognitive Grammar, as articulated by Ronald Langacker, continues to be of central importance in the field of cognitive linguistics. As has been emphasized by Langacker, Cognitive Grammar should be seen as a conceptual framework rather than a formal theory. Even so, Cognitive Grammar has proven to be a remarkably inspiring and illuminating framework within the cognitive linguistic community, regardless of whether one chooses to employ its ‘imagistic’ notational devices. It is sometimes said that a highly formal theoretical model of language (with language equated with sentences, rather than communicative activity) is desirable, even necessary, in order to test claims made by a model. Arguably, however, it is the pre-theoretical stage where decisions are made about the very nature of language and the concepts that are to play a role in explaining language that is deserving of more attention. The fact that Cognitive Grammar is not a formal theory per se, but rather a rich collection of cognitively inspired
211
212
NEWMAN
notional tools for the analysis of language, proves to be an advantage when one seeks to apply this framework to a more complete range of linguistic phenomena such as sign language, gesture accompanying speech, poetry, or narrative structure. So, for example, construal, figure-ground, perspective, mental scanning, fictivity – concepts introduced initially by Langacker to support the linguistic analysis of the sentence structure of mundane spoken and written language as used in an everyday context – may be applied in a natural, even self-evident, manner to the study of literary texts, giving rise to a still-developing field of cognitive poetics. The dynamic nature of language structure (e.g. meaning understood as conceptualization, rather than simply a concept) is central to Langacker’s approach, as is the integrity of form and meaning at each level of analysis, and the meaningfulness of grammar (eschewing the recognition of meaningless forms). The fundamentals of Cognitive Grammar have remained unchanged since its inception in the 1970s, though Langacker and others have further refined some key notions, such as reference point constructions, dominion, and the existential core of a sentence. Form-meaning pairings are key to Construction Grammar which, in one version or another, is the grammatical theory par excellence for most cognitive linguists. One can now speak of a family of Construction Grammars that would include Radical Construction Grammar, Embodied Construction Grammar, Fluid Construction Grammar, etc., including Cognitive Grammar, even if it stands apart in its rejection of the more familiar types of formalism seen in Construction Grammars. Construction Grammar is a highly active and successful field of research, with the different flavors showing varying degrees of interest in the larger cognitive questions that drive cognitive linguistics. Obviously, the more explicitly usage-based or cognitively oriented the approach to Construction Grammar is, the more naturally it fits into a cognitive linguistic framework. An area of increasing interest among construction grammarians is so-called negative evidence in grammar whereby speakers are able to make judgments about the appropriateness of words inserted into constructions even when such word-construction combinations may never actually occur in usage. Accounting for the different judgments speakers can make about non-occurring data poses an interesting challenge which needs to be addressed by linguists of any stripe. The adoption of a usagebased approach to constructions has lead to increased attention to the lexical preferences associated with a construction, not just as an incidental footnote to the analysis of a construction but rather as a fundamental aspect of it. Linguistic analysis in such usage-based approaches typically incorporates a good deal of probabilistic patterning supported by tools of statistical analysis. A Construction Grammar framework has been successfully applied, too, in the area of historical linguistics, giving rise to Diachronic Construction Grammar, where the paths of historical
Linguistic Analysis
evolution are now viewed in terms of how an individual construction is itself the locus of change. Any constructionist approach that can be extended to the discourse level will be of special interest to linguists who view discourse as a more default setting of language use, rather than the sentence level. The larger spoken or written discourse surrounding an utterance or sentence will motivate many choices observable in the smaller units within discourse. Just as a semantic frame at a clausal level functions as a satisfying organizing principle within the clause, so too the surrounding discourse functions as a high-level frame motivating clausal frames. The complementarity of sentence-level and discourse-level phenomena is more naturally accommodated when the basis of linguistic description at both levels is a formmeaning pairing, with essentially the same descriptive concepts available at both levels. Cognitive Grammar has proven to be particularly relevant to accounting for the reader’s construction of a mental space or text world in the course of engagement with a literary text, incorporating successive layers of representation. Increasingly, the discourse associated with interaction between speakers, including accompanying gestural activity, has become an area of interest and description in cognitive linguistics and in linguistics more generally, especially with the availability of tools that allow for comfortable multimodal analysis. The study of the dynamics of interactional settings, a focus of conversational analysis and cultural linguistics, would seem to go hand in hand with a strongly cognitive approach. Many of the trends alluded to in the overview offered here can be seen as trends that are making their way through linguistics as a whole and are not unique to cognitive linguistics. Thus, one occasionally finds references to an ‘empirical turn,’ ‘a quantitative turn,’ ‘a dialogic turn,’ not to mention ‘a cognitive turn’ affecting the whole field. While there are traces of these phenomena in the field at large, it is cognitive linguistics considered as a community of scholars that would appear to be the most natural and welcoming academic setting for these ideas to take shape and play themselves out.
213
14 Phonology Geoffrey S. Nathan
14.1 Introduction While cognitive linguistics (CL) has been a major force in linguistics since the early 1980s, there have been very few treatments of phonology expressly within the CL framework. The first paper on the topic to be published (Nathan 1986) suggested a program for thinking about phonological structure based on the early tools outlined in Lakoff’s seminal book (Lakoff 1987). In particular, the author suggested that the concepts of radial prototype category structure and image schema transformation were cognitive analogs of the traditional concepts of phonemes and allophonic processes. It should be noted that, unlike syntax and, to some extent, morphology, no consensus has emerged about what a CL phonology would look like. The situation has become somewhat more complex as CL in general has moved away from the tools of the early works (prototypes, image schemas) and toward a ‘usage-based’ model that concentrates on what could be roughly termed ‘learnability,’ not in the Chomskyan sense, but in the sense of how speakers make generalizations from innumerable speech events stored at varying degrees of abstraction. Although some phonologists have suggested that such a model works appropriately for phonology (e.g. Bybee 2001, Pierrehumbert 2002, Kristiansen 2007, Nesset 2008, Bybee 2010), others, including the present author (Mompean 2004, Nathan 2007b, 2008, 2009), have argued that phonology has special characteristics that impose additional constraints on phonological representations, and on the production and perception of speech sounds.
14.2 Some Thoughts on Universals Chomsky (1965: 28) introduced into linguistics the concept of formal and substantive universals. Chomsky defined substantive universals as having,
Phonology
just as one example, a “substantive acoustic-articulatory characterization independent of any language,” while formal universals would include “the abstract conditions that a generative grammar must meet,” such as a transformational component or cyclic operation of phonological rules. An argument could be made that cognitive grammar makes such a distinction too. The notion that linguistic units are categories of a certain type (with prototype structure, for example; see Lakoff 1987), or exemplar-based (Pierrehumbert 2002, 2003a, 2003b) or that learning is entirely usage-based and sensitive to frequency distributions should probably count as a ‘formal’ universal. Langacker (2015b and earlier) argues that nested figure-ground (or, as he calls it, “baseline” and “elaboration”) organization of representations is also a formal universal. On the other hand, aspects of our cognition of ‘stuff’ (objects, actions, relationships) affect our linguistic structures as well. For example, the energy flow in a prototypical active sentence seems to be reflected in the prevalence of SVO word order worldwide and the preference for subjects to be agents, and for various aspects of the animacy hierarchy to be reflected syntactically (see, e.g., Langacker 2008a: 438–39). I have argued elsewhere (e.g. Nathan 2007a, 2015) that phonological structure abounds in substantive universals (syllable nuclei are preferentially vowels, all languages have vowel height contrasts but some lack front–back contrasts, point of articulation systems cluster around LABIAL–CORONAL–DORSAL) and some of these will be discussed below. The work on phonetic grounding of phonology (e.g. Pierrehumbert 2002, 2003a, 2003b) also explores these concepts.
14.3 Phonological Units, Phonological Universals, and Phonological Acquisition It is impossible to talk about phonology without considering the concept of the phoneme. The term ‘phoneme’ was first used in a synchronic sense by Kruszewski, a student of Baudouin de Courtenay, although it was Baudouin who first described the phoneme in the mental storage sense that most contemporary linguists accept (Baudouin de Courtenay 1972). The idea that speakers stored and ‘operated on’ roughly segment-sized units became a standard way of thinking about phonology in the 1890s, and was further reinforced by Sapir’s classic description in the 1920s (Sapir 1972). Since that time there has been general acceptance of the notion that speakers store and perceive phoneme-sized units that are in some sense abstracted away from the actual articulations and/or sound waves used in speech. A few researchers over the years have questioned the concept of phoneme-sized units (see, e.g., Frith 1998, Ziegler and Goswami 2005, Port 2007, 2010), arguing that linguists’ perception of segments is the result of their lives within a literate culture.
215
216
N AT H A N
Most phonologists over the years, however, have argued that segments, varying or constant, have some cognitive reality that explains much language behavior. It is noteworthy that the concept of segment actually antedates the academic field of linguistics by almost three thousand years, in that all of the earliest phonological (as opposed to morphemic/ semantic) writing systems are alphabetical, and at least one, Phoenician, represents the origins of the modern Roman alphabet used to write this document (see Daniels and Bright 1996 and Sampson 2015 for a history of writing systems).1 This writing system, in various flavors, has been adopted by a majority of the languages of the world with strong literacy, ranging across language families and geography (from English to Indonesian to Vietnamese). Similarly, scripts derived from Brahmi (Devanagari, Bengali, Thai, Burmese) are also used by untold millions of speakers, and are also at least to some extent segment-based. There are writing systems based on larger units, although they tend in general to be semantically based (some of Sumerian cuneiform, for example, and Chinese). Some writing systems are either syllabic (Cherokee, Korean, although with segmental information packed inside the syllabary) or semisyllabic (Japanese and the phonetically based cuneiform writing systems such as Hittite, which have symbols for CV and VC pieces, as well as for simple vowels). For some neurological arguments that both segments and syllables are units of perception (acting in parallel) see Poeppel, Idsardi, and van Wassenhove (2008). In a frequently cited work, Stampe (1987), and in subsequent work by Stampe and Donegan (Donegan and Stampe 1979, 2009), extensive evidence is presented to argue that, irrespective of the size of the stored units, what is stored is not identical to what is produced (and therefore what sounds ‘strike the eardrum’). Evidence from folk poetry, spoonerisms and other speech errors, naive spellings, and the nature of established orthographies argue that language users abstract away from actual speech toward a more idealized, but nevertheless speech-like, mental storage of units that are the size of traditional segments. Note that this by no means rules out the possibility that storage is in terms of physical events (or the mental representation thereof), as opposed to lists of abstract features, either extensive or minimalized (‘underspecified’) as most contemporary generative models argue. Stampe and Donegan have also argued that features themselves are not abstract classifications of speech classes, but rather are real cognitive categories based on similarities of sounds that are treated as equivalent by their language because of identifiable articulatory and/or acoustic characteristics that speakers discover in the process of acquiring their 1
The segment is also the basic unit upon which the International Phonetic Alphabet (IPA) is based. Despite the wide availability of spectrograms, the IPA is still the representation of choice for virtually all linguists working on spoken (as opposed to signed) language.
Phonology
language. Thus features are emergent properties of speech sounds that speakers acquire as they discover their ability to speak. This view is explored extensively in Mielke (2008) and will be discussed in more detail below. The mainstream view in phonology is still probably that founded on the generative ‘enterprise,’ and that includes the notion of Universal Grammar – the view that linguistic structures are innate and part of our genetic heritage. Concepts such as features, and the relationship between ‘underlying’ and ‘surface’ phonology are not thought to be acquired through experience, but rather are tools that learners come equipped with when beginning to confront linguistic data. Cognitive linguistics has, from the beginning, rejected this view (Langacker 1987, Tomasello 2003, Langacker 2013), arguing that existing learning processes are capable of explaining language acquisition. While there is no need, within CL, for innate mental structures, the similarities among languages that are not genetically related still need an explanation. This constraint is particularly noticeable within phonology because the similarities are objectively not based on biases introduced through hundreds of years of Latin-based grammatical tradition (as might be the case with possible syntactic universals such as ‘object’ or ‘passive’), but rather reflect genuine universal tendencies. For example, all languages have vowel height contrasts (but not necessarily tongue advancement contrasts).2 That is, there are no languages that contrast only front versus back vowels, but at the same time, all languages contrast (at least) two vowel heights. And all languages make use of at least two of the three basic points of articulation categories (Labial, Coronal, Dorsal), with all but a few having all three for stops, while sounds at intermediate locations (e.g. prepalatal, uvular) are frequently absent. While these universals require explanation, they do not require Universal Grammar. As phonology is the direct implementation of physical gestures and the perception of the sounds those gestures make, we can explain universals on the basis of the physiological and even physical natures of the apparatus we use to make and hear sounds.3 To make this concrete in phonological terms, Stevens (1989, 1998) showed that there are acoustic ‘notches’ in our perceptual systems tuned to the acoustic cues associated with sounds made at labial, coronal, and velar points of articulation, such that contrasts in zones between these points are difficult to perceive, while those at the points are more clearly 2
These “implicational universals,” as Greenberg called them, are documented, for example, in Ladefoged and
3
An analogy might make the point here. There are staircases in medieval European cathedrals, the Great Wall of China
Maddieson (1996). and Mayan ruins, all of which have roughly the same shape and riser height, and were arguably built independently of each other. One might ‘explain’ this by appeal to a ‘stair acquisition device,’ or by universals of historical change, but a simpler explanation might be that human legs are approximately the same length, and so is (adult) human stride size, so that anyone making stairs is constrained by average human leg size.
217
218
N AT H A N
audible. Thus, as long as all human beings share general articulatory physiology and their auditory system includes frequency detectors and spectrum analyzers, languages will cluster their stops in the most perceptible zones. Similarly, studies of consonant systems worldwide show that in systems with a voicing contrast, if there is a gap in voiceless stops it is in the labial column (a well-known example being the lack of /p/ in Arabic, Fijian, Yoruba, and many other languages; see Ladefoged and Maddieson 1996), while the voiced stops usually have gaps in the velar column (as found, e.g., in Dutch). While the reasons for the missing /p/s are unclear, the missing /g/s are easily understood. Voicing requires airflow through the larynx, but oral stops prevent airflow (that’s what ‘stop’ means, after all). Air can be made to flow by enlarging the oral cavity, but the less space between the glottis and the obstruction, the smaller the cavity available to enlarge. Thus /b/s can be made by lowering the larynx, raising the velum and puffing out the cheeks, but production of /g/s cannot easily be enhanced in this way, and thus the effort required to make them makes them candidates for loss (Keating, Linker, and Huffman 1983, Keating 1984).
14.4 Deep, Surface, or Neither? Beginning with Chomsky (1965), linguists have argued that there are two ‘levels’ of linguistic structure. One, which we might term the WYSIWYG level, represents speech as articulated/perceived. It is important to note, however, that even this level represents somewhat of an abstraction from the actual movement of articulators, or the fine details of the sound waves. No linguist has seriously proposed, for example, that the most superficial level of linguistic structure is as concrete as which part of the alveolar ridge the tongue tip touches, or the exact value in Hertz of the Mandarin first tone. Among other reasons, these values not differ only among speakers along variables such as age, size, gender, but also differ from moment to moment, as well as being influenced by emotional factors in addition to the linguistic context. On a purely linguistic level, vowel formants are affected by the following or preceding vowel, so that, for monosyllabic words there are, in some sense, ‘different’ vowels for each combination of preceding versus following word. Put more concretely, the formant values (and, presumably, x-ray tongue placement data) for bet is different whether you are saying He bet Sam or You bet now. Assuming there could be agreement on a maximally concrete level of representation, the other level that has been assumed within phonology since the work of Baudouin is a level of representation that stands somehow ‘between’ the variant pronunciations that go beyond those mentioned in the previous paragraph.
Phonology
These are alternate ‘outputs’ governed by structural properties of the utterances in question – and are traditionally referred to as allophones, such as aspirated (versus non-aspirated) voiceless stops or nasalized (versus oral) vowels. Another less-often cited case is, for example, the syllable-governed complementary allophones [e] and [ɛ] of the Standard Northern French phoneme /e/. The lower variant occurs in closed syllables (bête [bɛt] ‘animal’); the higher one in open syllables (bébé [bebe] ‘baby’). Native speakers of French below (roughly) the age of fifty are totally unaware of this alternation, although older speakers (and those who speak some nonstandard dialects) have minimal pairs (classic examples include allé [ale] ‘went, pptl’ versus allait [alɛ] ‘went, imperfect’). Note that the spelling still codes the contrast. Younger speakers not only cannot hear the contrast, they have no control over the spelling contrast, leading older prescriptivists to make fun of younger speakers misspelling words whose spellings are trivial to the older speakers. Cases like these (and there are innumerable examples in every language4) led early phonologists such as Baudouin and Sapir to propose that speakers store speech in a somewhat more abstract representation which they called a phoneme. When Chomsky introduced the idea of deep structure, this level came to be called an underlying representation. The exact nature of this representation varies broadly among theories of phonology. One current generative view is that each phoneme is represented solely by those features5 that distinguish it from all other phonemes in that language. Moreover, in some versions of that theory, called ‘underspecification theory’ (Archangeli 1988), since each feature can have a ‘default’ value, there could be phonemes that have no feature representations at all. For example, the English schwa has been proposed to have only the feature [+syllabic], and /t/ to be only [-syllabic], with all other features being filled in by default. On the other hand, some current versions of Optimality Theory (McCarthy 2004) require that all phonemes have a complete set of specified features, with allophonic variation being created through the effects of assorted constraints on the lexical representations. Natural Phonology has argued that phonemes are stored as actual, fully represented, but mentally imaged sounds. Thus, all their ‘features’ would be completely specified.6 Allophones, therefore, are altered phonemes, adjusted by allophonic (and other) processes, and there are not exactly levels (in the sense of abstraction from some putative ‘surface’), but rather 4
To cite just one anecdotal example, while teaching introductory phonetics I introduced the voiceless velar fricative [x], mentioning that it occurred in Mandarin as an allophone of /h/. A native speaker of Chinese argued with me that I was mistaken. His name was He, pronounced [xə], phonemically /he/, but he was unable to hear that his pronunciation of his name was quite different from my pronunciation of, say, her in my English, much to the amusement of the American English speakers in the room.
5
The concept of features will be discussed in section 14.5.
6
Natural Phonology has always argued that features are not phonological primitives, but rather refer to the mental characteristics of kinds of sounds that phonological processes refer to. The sounds are real; the processes refer to classes of sounds that share qualities.
219
220
N AT H A N
modifications of one surface form to another to fit the circumstances. An analogy would be my path from the kitchen table to the sink. I have a normal path to the sink, but might take a detour to the refrigerator, or take a larger step to avoid a water spill. Each path is real, yet we could reasonably argue that the most direct, or most convenient path is the one (if any) that is ‘stored,’ and each alternative path (‘avoid the dog,’ ‘kiss my partner’) is ‘generated’ on the fly as we make our way to the sink. Similarly, we could argue, we store a prototypical /t/ for each word that contains one, but as we speak, implement alternative ‘t’s,’ some even overlapping with competing phonemes – that is, with sounds that, under other circumstances, would coincide with the prototypical instantiation of some other sound (this issue is explored in greater depth in Mompean 2004). The issue of what level of abstraction is needed to account for the fundamental facts of phonology is, in part, the issue of categorization. What kind of thing is a phoneme? Categorization has been a crucial topic in cognitive linguistics since its inception. Lakoff (1987) introduced the concept of a radial prototype category, and I have argued in earlier work (Nathan 1986, 2007a, 2008) that the phoneme is just such a category, with prototypical sounds picked out by prototypicality effects that were first explored (under other rubrics, of course) by Trubetzkoy (1939 [1969]) and Greenberg (1966). As we know, of course, there are competing views of how categories work, and Pierrehumbert, Bybee, and others have suggested that phonemes are actually exemplar categories (see, e.g., Pierrehumbert 2002, 2003a, 2003b, Bybee 2001, 2010). The nature of concepts was thoroughly explored in a major survey work (Murphy 2002), and, in the psychological literature, prototype categories have somewhat fallen out of favor in contrast to exemplar-style categories, so some discussion is required here. Phonemes are, of course, classical categories within the generative paradigm, because they are defined as a set of distinctive features. They were first specified in Halle (1962) and Chomsky and Halle (1968) and are still found in most contemporary textbooks (Gussenhoven and Jacobs 2005, Hayes 2009). That is, a phoneme is a sound that has all and only certain features. As mentioned above, I have argued (Nathan 1986, 2008) that phonemes are prototype categories. That is, they do not have well-defined boundaries (specifically, they may overlap), and they do not have distinctive features. Instead, phonemes are defined by a central, prototypical sound (corresponding, roughly, to the principle allophone within classical structuralist phoneme theory), and the set of phonologically natural extensions of that sound, as motivated by locally sanctioned processes deforming that prototypical sound into increasingly less similar instances. These instances may well also be instances of other phonemes, which have traditionally been referred to as neutralization. And, because the successive deformations may be in widely differing directions, items on
Phonology
two different chains of deformation may very well share no phonetic features in common except that they can be traced back to a common prototypical origin. As an example consider the American English /l/ phoneme. For most speakers it is pronounced as a plain voiced alveolar lateral approximant [l] in syllable onset position (leaf ), and as a heavily velarized [ɫ] in coda position (feel). However, for many speakers, particularly in more casual speech, the velarized version loses its alveolar contact, hence becoming a vowel, best transcribed as [o̯ ]. Under these circumstances, Hal /hæl/ and how /haʊ/ may be homophonous because the glide part of the diphthong is very similar to, if not identical to, the vocalic allophone of the /l/. As mentioned above, allophones may share no features in common. For example, and more familiarly, the phoneme /t/ in most North American English dialects may be instantiated as [t̪ ʰ] – voiceless, aspirated, dental stop (‘with two for dinner’) or as [ɾ] – voiced alveolar tap (‘but all the other kids’), two sounds which share no features in common except ‘consonant.’ Since these variants are in general not perceived by native speakers,7 they are clearly categorized as ‘the same’ sound. Note also that there is never a need to teach children to be sure to spell the ‘d’ in ‘width’ the same as the ‘d’ in ‘dinner,’ despite the considerable differences between them, because the children (and probably the teacher as well) are unaware that they actually are different. This example illustrates both the concept that allophones are instances of a radial prototype category, and that they are basic-level categories, because they are virtually automatically grasped (certainly as opposed to the allophonic variants). Exemplar theory suggests that concepts such as phonemes are abstractions drawn from stored examples. Each phoneme is represented by one (or even a cloud of) representative instance(s) as determined by measures of frequency. Other instances are in a network of related cases, thus permitting similar understandings of relatedness as outlined above for prototype category structure. Unlike prototype theory, however, frequency is the only available structuring principle for how alternate realizations (‘allophones’) are arranged or stored. However, this seems to fly in the face of the fact that prototypical representations may well not be the most frequent instances. In addition, exemplar theory does not have a simple way for our embodied experience of sound-making and hearing to influence the nature of the sounds being stored. If all that is stored is a network map based on frequency of occurrence we cannot explain the articulatory and acoustic pressures that shape phoneme systems throughout the world, nor can we explain why, for example, many children acquiring American English pronounce words with voiceless alveolar 7
Flapping/tapping is perceived, because its effects result in a merger/instance of neutralization, such that /t/ ends up ‘sounding like’ /d/. As Stampe has noted, however, this perception is unidirectional – no native speakers believe that /d/ ‘sounds like a /t/’ in ‘spiders.’
221
222
N AT H A N
stops as [t]s even if their parents never pronounce them that way (see Donegan and Stampe 2009 and Smith 2010 for discussion). If we assume that children adjust their representations to what the speaker’s intentions were, rather than what simply came out of their mouths, we can understand cases like these. The notion of perceiving others’ intentions is crucial to the acquisition of language if innately programmed linguistic structures are not assumed (see Tomasello et al. 2005 and Da˛browska 2015b for discussion).
14.5 How are Sounds ‘Spelled’? Since the early twentieth century, linguists have wrestled with the nature of the units that make up words. Beginning with the phoneticians who developed the IPA, speech was ‘spelled’ using symbols representing segment-sized units (‘segments’). Originating in the European (e.g. Trubetzkoy 1939 [1969]) and American (Bloomfield 1933) structuralist phonologists and further pursued by early generative phonologists (Jakobson and Halle 1956), an alternative view was developed, in which sounds were represented not as single gestalts, but rather as lists of distinctive features. This view became popular quite early; Bloomfield, for example, in his classic Language, said that phonemes were actually “bundles of distinctive features” (Bloomfield 1933: 79). This view was widely adopted within the framework of generative phonology and persists to this day. In the mid-twentieth century, this view was mildly challenged by Firth and his colleagues (Firth 1948) who argued that some features could not be isolated to a single segment, but instead were ‘smeared’ across a stretch of speech, similar to the way tones were. The original conception of features is that they were attributes of segments, and hence ‘stacked’ in little segment-sized boxes (leaving aside the Firthian suggestion to the contrary). In the late 1970s, Goldsmith (1976) argued that this “absolute slicing hypothesis” was in error, beginning with a discussion of the representation (and behavior) of tones in a number of African languages. Similar work by Liberman (1979) showed that English intonation patterns can best be understood in the same way, as a sequence of a few specified tones spreading across much longer stretches of speech. Goldsmith (and all later proponents of autosegmental phonology), however, did not challenge the notion that lexical representations (underlying or perceived representations) were sequences of those same stacked lists of features that Bloomfield and Trubetzkoy had proposed in the 1930s. And all generative phonologists of all stripes have agreed that, in some sense, features are ‘real’ and ‘phonemes,’ in the sense of whole objects, are not, a claim that was made originally in the famous attacks on the phoneme found in Halle (1962, 1964: 336) and Chomsky and
Phonology
Halle (1965). They argued that the only reality of phonemes was as a set of features. This was based on the argument that their phonological model could be greatly simplified if both sounds and the rules that applied to them were expressed in feature terms only. Consequently, there was no room for the notion of sounds as basic level conceptual objects (to be fair, the notions of basic level units, not to mention prototype conceptual theory, was over twenty years in the future – see extensive discussion in Mompean 2014 and Taylor 2006). In addition, however, it has always been the generative commitment that linguistic theory was developed autonomously, and that the results just were cognitively real.8 Cognitive Grammar, of course, begins with the assumption that basic perceptual mechanisms, along with whatever learning mechanisms people come equipped with, determine how language is acquired, stored, and used. Developing descriptions that are maximally simple (in the sense of using a minimum of distinct symbols) is not a concern in Cognitive Grammar, because there is no evidence that cognition attempts to maximize computation and minimalize storage. From this point of view, speech should be represented as a sequence of real sounds, essentially the same phonemic segments the Phoenicians discovered four thousand years ago. However, this is not to say that sounds do not have characteristics that have real influences on how they are perceived and produced. As an analogy, consider the way humans grasp objects. Smaller objects are picked up with the forefinger and thumb, while larger objects are grasped with all fingers making a cylinder with the thumb on the other side. Contemporary understanding of how precision and power grips are acquired argues that the physical skills are not either ‘innate’ (in the Chomskyan sense) nor simply ‘learned’ (in the usage-based sense) but rather are emergent:9 The development of grasping action patterns is consistent with the notion of properties of self-organizing principles in biological systems . . . emergent features of the constraints imposed on the organism-environment synergy, rather than prescriptions for action as is typically postulated by the maturation hypothesis and the extant cognitive formulations of motor development. (Newell and McDonald 1997: 252)
14.6 What Level(s) of Representation Do We Need? There have been two ‘levels’ of phonological representation since the notion of ‘alternation’ was introduced by Baudouin (Baudouin de 8
This is, of course, Chomsky’s claim that generative grammar is cognitive psychology; and it is the psychologist’s job to figure out how it fits in with the rest of psychology.
9
For more discussion of emergence, and its relationship to innateness and learning, see Blumberg (2006).
223
224
N AT H A N
Courtenay 1972). He argued that there were, essentially, basic and derived versions of forms, with two forms of ‘derivation’ having different psychological status. Those produced by what Baudouin called physiophonetic processes were (we would now say) psychologically real, while those produced by psychophonetic processes were, metaphorically, archaeological remnants of earlier physiophonetic alternations, and might be noted by speakers as morphologically related, but did not represent articulatory and perceptual pressures on the speakers. However, American structuralists (European ones too) did not maintain this distinction, and generative phonologists originally denied there was any difference between processes governing low-level phonetic alternations and abstract alternations such as those seen in the English vowel shift pairs like divine ~ divinity (for the original argument, see Chomsky and Halle 1968). Stampe’s insistence on a cognitive distinction between rules and processes was ultimately recognized by most generativists in a related (but not identical) conceptualization: lexical versus post-lexical (Kiparsky 1982, Mohanan 1986). The lexical/post-lexical contrast is based on the effect of processes on kinds of representations. Lexical processes change one pre-existing phoneme into another (they are structure-preserving), while post-lexical processes convert phonemes into allophones.10 Lexical processes have access to morphological, and perhaps even syntactic information, but post-lexical processes apply to any string of segments of the right kind. I have argued (Nathan 2008, 2015) that the real difference between processes and rules is one of kind, not degree. Rules are speakers’ representation of recurring patterns, and are generally equivalent to analogies. They are more or less conscious, and violating them, or extending them to new instances is often perceived as jocular, or as uneducated. Examples might include regularized plurals like wolfs (instead of the irregularly voiced version wolves) and beeth (extending the goose-geese pattern to replace the regular booths). Violating processes, on the other hand, is perceived as ‘having an accent’ or a speech defect. So omitting aspiration in initial voiceless stops sounds ‘French,’ while failing to nasalize vowels before nasal consonants in the same foot sounds like you have a cold. Further, not making these kinds of alterations requires special training, either from an acting teacher or a course in phonetics, but violating rules, as in saying *nous tienons in French for nous tenons ‘we hold’ is the wrong verbal mood, not the wrong dialect or a funny accent.11 The primary reason for differentiating between rules and processes is that they have different ontologies. Rules are the expression of recurrent patterns within morphologically related paradigms. As such they may or 10
This is not exactly right, but is close enough for our purposes.
11
The /jɛ/ ~ /ə/ alternation in French verb forms (tiens vs. tenons) is a classic ‘rule’ developed early in generative phonology by Schane (1968).
Phonology
may not represent actual linguistic knowledge, and certainly could be both represented by a rule and also listed separately – the classic case of the rulelist fallacy originally discussed in Langacker (1987). Processes, however, are the speaker’s response to the inherent limitations of the physiological ‘equipment’ needed to make speech sounds, and the perceptibility of those physiological events by the hearer. They are directly due to the embodiment of speech. Bergen and Feldman express particularly clearly the notion of embodiment I have in mind here: Language understanders automatically mentally imagine or simulate the scenarios described by language. The mental simulations they perform can include motor detail at least to the level of the particular effector that would be used to perform the described actions, and perceptual information about the trajectory of motion (toward or away from the understander; up or down) as well as the shape and orientation of described objects and paths . . . these simulations involve some of the very brain mechanisms responsible for perceiving the same percepts or performing the same actions. (Bergen and Feldman 2008: 317)
In other words, processes express the internalized physiological and perceptual constraints on pronunciation that particular languages have sanctioned, but rules express constraints on what speakers ‘can’ say. As a consequence, rules may be violated playfully (as someone might say /pərsnɪsəti/ for ‘persnickety’12) but no native speaker of American English would playfully say [spʰɑt] for ‘spot’ or [pɑt] for ‘pot.’ Velar Softening is an accidental pattern inherited from an earlier stage of the language (actually, to be precise, an earlier stage of another language, Old French), but aspiration is something speakers do as they speak, as is nasalization of vowels before tautosyllabic nasals and lengthening of vowels before word-final voiced consonants. The reason I have spent considerable time talking about phonological processes is because the notion of a mental representation different from a store of actualized instances (which is the default assumption in a usagebased model) requires active justification.
14.7 Higher Level Units – Syllables, Feet and More We know that sounds are neither produced nor perceived in isolation. The fact that there are writing systems based on larger units suggests that those units are easily available to introspection by naive native speakers. And, of course, there is well over a hundred years of scholarship on the existence of the smallest phonological construction, the syllable. Variation in the kinds of syllables permitted in languages is well studied, and there are clear sets of implicational universals going back (at least) to work by 12
This violates the famous ‘Velar Softening’ rule discussed in Chomsky and Halle (1968).
225
226
N AT H A N
Greenberg (1966). All languages have CV syllables, but only a subset have syllables with greater complexity, either permitting more than one consonant before the vowel or permitting codas, again of varying complexity. In addition, there seems to be principles governing what kinds of things are permitted both in complex onsets and in codas, whether complex or simple. I have argued elsewhere Nathan (2008, 2015) that the overall shape of syllables is determined by preference principles based on how humans perceive rhythmic behavior in general (Goswami 2012, Goswami and Leong 2013). Speech, like other rhythmic behavior, is perceived as a sequence of beats. Beats are perceived most clearly when they have the shape of an abrupt onset followed by a slow fade (think of a drum beat, for example). Consonants (particularly the most prototypical consonant, a voiceless stop) provide the sharpest onset, while a vowel or vowel plus glide makes the ideal center of a beat, as vowels are the loudest speech sounds. And the well-known preference for sonorants in codas (many languages only permit sonorants in codas, for example), leads to a more gentle downslope for each beat. Additionally the extremely widespread preference for rising sonority in onset clusters (i.e. /kw-/ or /dr-/ being greatly preferred over onset plateaus such as Greek /mn-/ or Russian /zdr-/) is explained as a compromise between a pure, sharp onset and one that permits complex onsets but maintains the overall syllable shape. So here we have the perceptual explanation for the existence of the simplest of phonological constructions (i.e patterns of recurrent combination of units). Langacker (2015a) argues that the conceptual structure of syllables (minimally V, additionally CV, then CVC etc. as detailed above) is simply an instance of recursion of what he calls ‘baseline’ and ‘elaboration’ (or alternatively, ‘figure’ and ‘ground’). A vowel can serve as the baseline for a syllable, with an extension being the addition of an onset, the combination CV then being extended through the addition of a coda and so on. Further support for the fundamental importance of syllables to the storage and production of speech comes from the work of MacNeilage and his co-workers in MacNeilage (2008) and Davis and MacNeilage (2004), who argue that earliest speech emerged as an adaptation of the available movement capacity of the vocal apparatus in hominid speakers. Biomechanically simple serial vocal patterns paired with meaning in early protoword forms were thus made available for cultural transmission in service of interpersonal and community-wide communication. (Davis and MacNeilage 2004: 2)
That is, speech evolved out of the primitive sucking behavior which mammals develop in the process of learning how to eat.13 The alternating open–close gesture of sucking and chewing corresponds, both in timing 13
For evidence that sucking and chewing is actually learned, or at least emergent, in mammals, see Blumberg (2006).
Phonology
and in physiology, to the alternating open–close gestures of CV syllables, which MacNeilage argues form the framework upon which speech is built. As is well known, rhythmic activity is generally perceived as consisting of groups of beats. The most frequent groupings are either weak–strong (iambic) or strong–weak (trochaic). There tends to be a bias toward weak beats being shorter than long ones in iambic perception, and strong beats being louder than weak ones in trochaic perception (the iambic–trochaic law – Hayes 1995, but see Hay and Diehl 2007, Hay and Saffran 2012 for evidence that at least some of this perception is not limited to humans). This grouping leads to the existence of the second level of phonological grouping, namely feet. Extensive research in the past twenty years has indicated that native speakers of many languages manipulate language in various ways that require the awareness of foot-level grouping. Two wellknown cases are McCarthy (1982) for English ‘expletive infixation’ (fandarn-tastic) and Poser (1990) for Japanese hypocoristic formations. Feet, of course, occur not only in the phonology of languages (even those, such as Japanese, that appear to lack ‘stress’) but also in folk (and professional) poetry.14 Evidence for larger phonological units than feet is less clear, and even the generative tradition has not settled on a solution. Units such as the phonological word and the phonological phrase have been suggested, and it is clear that there is a need for a unit the size of the traditional notion of intonation contour, since units of that size must be part of the apparatus for speech production and perception.15 The traditional English intonation contours (fall, rise, fall–rise) are used by all speakers to produce discourse, and also for managing turn-taking in conversation. Conversational analysts have shown not only that speech can be divided into a series of these contours, but also that speakers time their ‘next turn’ to coincide with the end of an interlocutors last tone unit. Note also that choice of how long a phonological phrase will span is a semantic/pragmatic decision and, in that sense, is, strictly speaking phonemic. Consider the currently popular online mode of writing certain kinds of emphatic speech: ‘This. Will. Not. Happen.’ It is clear that this a folk representation of giving each word a separate, falling intonation contour.
14.8 Conclusions A particularly elegant expression of how Cognitive Grammar can account for both the universal and the language, or even speaker-specific aspects of phonology, comes from Tomasello’s discussion of ‘Universal Grammar’: 14
Note, for example, that nursery rhymes in English, Benkulu (an Austronesian language), Cairene Arabic, and Mandarin normally have four lines consisting of four feet (of any kind) (Burling 1966).
15
A classic generative discussion of these levels can be found in Selkirk (1984).
227
228
N AT H A N
human linguistic universals derive from the fact that all humans everywhere: (1) conceive non-linguistically of agents of actions, patients of actions, possessors, locations, and so forth; (2) read the intentions of others, including communicative intentions; (3) follow into, direct, and share attention with others; (4) imitatively learn things from others, using categorization, analogy, and statistical learning to extract hierarchically structured patterns of language use; and (5) process vocal-auditory information in specific ways. (Tomasello 2009: 470)
In short, phonology is formed, acquired, and shared through the embodied learning of sound production, shaped by the input surrounding us, but also by the experience of making sounds, and recognizing them as produced by others. We store language as we hear it, but we hear it filtered through the articulatory and auditory pressures that our individual languages have ‘chosen’ to employ. Those constraints our language has chosen to suppress (e.g. it is hard to make voiced stops, tense-lax vowel contrasts are difficult to hear in noise) shape the basic ‘alphabet’ (phoneme system) in which we ‘spell’ the sounds of our language. We produce those sounds by adjusting them to their phonological, sociolinguistic, and stylistic circumstances, and hear them by recovering the adjustments we assume our interlocutors have made (i.e. we recover their intentions).
Acknowledgments I would like to thank Jose Mompean for extensive comments on an earlier draft of this chapter. Many of those comments are now part of the chapter.
15 The Construction of Words Geert Booij
15.1 Introduction: Constructional Schemas for Complex Words An essential part of the linguistic competence of native speakers of a language is their potential awareness of the systematic correspondence between the form and meaning of complex words. Consider the following four pairs of related English words: (1)
answer bake abridge archive
answerable bakeable abridgeable archiveable
The difference in form between the words in the left column and those in the right column correlates systematically with a difference in meaning: all words in -able have the meaning ‘can be V-ed’, where V stands for the meaning of the corresponding verb on the left. There is also a difference in syntactic class: the words on the left are verbs, and those on the right are adjectives. English features hundreds of such adjectives, and this set can be easily extended. For instance, as there is a verb to skype, we may coin the adjective skypable. Therefore, users of English can conclude to a generalizing schema for the construction of adjectives ending in -able, based on their actual experience with a number of such adjectives. This schema is a constructional schema: it accounts for the systematic pairing of form and meaning in a set of words. Constructions are pairings of forms and meanings, and constructional schemas express the systematicity in this pairing of form and meaning. We may thus assume the following constructional schema for English: (2)
< [[x]Vi able]Aj ↔ [can be Patient of SEMi] PROPERTYj >
The schema is demarcated by angled brackets. The double arrow denotes the correspondence between the form on the left and the meaning on the
230
BOOIJ
right. The indices specify correlations between form parts and meaning parts. The index i of the verbal base corresponds with a meaning component of the adjectives, SEMi, where SEM stands for Semantics. In other words, the meaning of the verb is a part of the meaning of the corresponding adjective. The index j correlates the form and the meaning of these adjectives as a whole. The variable x stands for the phonological form of the base verb. Speakers of English have stored lots of such adjectives in -able in their lexical memory, but will differ in which adjectives exactly are stored. Constructional schema (2) has two functions. On the one hand, it has the function of ‘motivation’ with respect to stored exemplars of this word formation pattern. This means that the meaning of such adjectives is not completely arbitrary, unlike that of phonologically similar simplex words such as table and cable. On the other hand, this schema serves to enable one to understand the meaning of adjectives of this form that one has not come across before, and to enable a language user to coin new derivational adjectives. This type of abstract knowledge of language is usage-based: the language user has to be exposed to a sufficient number of use of different types of such deverbal adjectives before (s)he can develop the corresponding abstract schema. The use of schemas for the representation of linguistic knowledge is a characteristic feature of cognitive linguistic approaches, and in particular for the various models of Construction Grammar. This approach contrasts with the rule-based approach of classical generative grammar. An important difference between these two is that schemas are outputbased, and can express directly that linguistic knowledge is based on observed language use. Below, we will encounter more detailed arguments for why schemas are better than rules for the representation of linguistic knowledge, in particular lexical knowledge. In section 15.2, I sketch a model of the architecture of grammar in which morphology finds its proper place. Section 15.3 discusses the nature of the lexicon, and its hierarchy of schemas and subschemas. Section 15.4 argues that constructional schemas for complex words can account for mismatches between the form and meaning of complex words. In section 15.5 I show how the use of morphological schemas, both inflectional and derivational ones, may depend on syntax. In section 15.6 it is argued that constructional schemas have the optimal format for expressing generalizations concerning inflectional systems. Section 15.7 summarizes our findings concerning the proper analysis of the construction of complex words.
15.2 The Architecture of Grammar How does ‘the grammar of words’ fit into a general model of the architecture of grammar? In order to answer this question, we have
The Construction of Words
to reflect on the types of information involved in lexical representations. A word is usually a combination of phonological, morphosyntactic, and semantic information. Therefore, schemas for complex words have to contain these three types of information. Note that the form part on the left-hand side of schema (2) above is in fact a conflation of information on two different tiers: phonological information and morphosyntactic information. We can decompose it as follows: (3)
The leftmost part of (3) states that words in -able are prosodic words (symbolized by ω) ending in the phonological sequence /ǝbǝl/ preceded by a number of segments (x) which correspond with the phonological string of the base verb. The middle part of this schema states that these words are adjectives derived from verbs. Schema (3) thus expresses the correspondence between information on three different levels of representation. This is referred to as the ‘tripartite parallel architecture of grammar’ (PA) (Jackendoff 2002, Booij and Audring 2016). The model of morphological analysis within PA is referred to as Construction Morphology (abbreviated as CxM), and is outlined in detail in Booij (2010). In radical versions of Cognitive Grammar one finds the claim that schemas express a direct correspondence between phonological form and meaning, without an intervening level of morphosyntactic information, also at the word level (Taylor 2002a: ch. 14, 2015). For instance, the English morpheme /s/ is claimed to be linked directly to the meaning PLURAL. In my view this is inadequate. For instance, the morphosyntactic property [plural] does not only express “a plurality of entities” (Taylor 2002a: 267), but also requires a more complicated set of semantic interpretation rules. For example, in the question Do you have children? the answer can be Yes, one, whereas the answer No, only one is odd. That is, plural forms of nouns may receive a generic interpretation. On the other hand, the use of plural forms does not depend on the specific phonological form of the plural marking. Dutch features two competing suffixes for plural marking, -s and -en. The choice between these two suffixes is governed by phonological output conditions, and morphological conditions as well (Booij 2002: 21–32). That is, the morphosyntactic feature [plural] is the feature that mediates between phonological forms on the one hand, and a range of semantic interpretations on the other. Therefore, a grammar with three tiers, as also assumed in Croft and Cruse (2004: 254–56), is to be preferred. The level of semantics should be given a broad interpretation, also comprising the pragmatic dimension, because morphological constructions may have specific pragmatic properties. In many languages diminutive morphology may also express endearment, and this has a specific pragmatic or stylistic value. Another example of the pragmatic role of
231
232
BOOIJ
morphology is that the infinitive form of Dutch verbs can be used as the predicate of simplex sentences as an imperative, as in: (4)
Oppass-en (jij) ‘Be-I N F careful (you)!’ Opet-en die appel ‘Eat-I N F that apple’
Hence, it does not suffice to specify the morphosyntactic properties of the Dutch infinitive (form: stem + -en, non-finite, can function as neuter noun). In addition, the potential for imperative use illustrated in (4) must be specified for this inflected form of Dutch verbs. Therefore, the term ‘semantic tier’ should be replaced with the more general term ‘conceptual tier’ which comprises pragmatic and discourse properties as well (Jackendoff 2002: ch. 12). German complex verbs in -eln (-n is the infinitive ending) illustrate how semantic and pragmatic properties of morphological constructions can be intertwined. These verbs can be derived from verbal, nominal, and adjectival stems. Here are some examples: (5)
hu¨st-eln ‘cough slightly and repetitively’ fu¨ß-eln ‘play footsie’ fremd-eln ‘be scared of strangers’
< husten ‘to cough’ < Fuß ‘foot’ < fremd ‘strange’
These verbs have an expressive or evaluative, more specifically attenuative, meaning. This attenuative meaning has both semantic aspects such as low intensity and repetition, and pragmatic aspects such as contempt, affection, euphemism, or trivialization. These semantic and pragmatic properties have to be specified on the tier of conceptual structure. A detailed CxM account of these properties in terms of constructional schemas can be found in Weidhaas and Schmid (2015). A basic insight of Construction Grammar is that constructions may have holistic properties that cannot be derived from properties of their constituents. A straightforward example from the domain of morphology is full reduplication, in which a word is doubled. The doubling configuration as such has specific semantic values, depending on the language and the word class, as in, for instance, ‘plural’ for reduplicated nouns, and ‘intensity’ or ‘repetition’ for reduplicated verbs. These meanings are not derivable from one of the (identical) constituents of such compounds, but are evoked by the copying configuration as a whole. Constructional schemas can express such holistic properties. For instance, the plurality meaning of noun reduplication can be accounted for as follows (PLUR stands for the semantic operator of plurality): (6)
In the case of exocentric compounds (compounds without a head constituent) such as the Greek compound filo-sofos, “lit. love-wisdom, lover of wisdom, philosopher” (Ralli 2013), with the formal structure [V N]N, a certain meaning component (in this example the Agent of the action) is
The Construction of Words
not expressed by one of the constituents of this compounds, but evoked by the compound structure. Semantic coercion is another indication of the holistic semantic properties of morphological constructions. For instance, the deadjectival suffix -ness creates words to denote properties. Hence, when we coin a word like Britishness ‘the state or quality of being British, or of embodying British characteristics,’ the relational adjective is coerced into a qualitative adjective that denotes the prototypical or characteristic properties of British people. The same holds for the prefix un-, as in un-British, ‘not characteristic of or consistent with British customs, habits, or traditions,’ which contrasts with non-British, in which the relational meaning is preserved. Another example is the use of the English prefix un- with other base words than inchoative or causative verbs. The attachment of un- to such verbs coerces a change of the semantic class of the base: “un- can take a stative, activity or other kind of verb and force it into a causative/inchoative verb that implies a reversible result” (Bauer, Lieber, and Plag 2013: 374). Examples are the verbs un-inhabit, un-grow, un-see, un-have, and un-hit. These examples show that morphological constructions have the power to trigger semantic overrides, and this implies that the constructions themselves have specific semantic properties. A construction may have holistic properties on the phonological level as well. In other words, there is construction-specific phonology, as shown with many of examples in Inkelas (2014). For instance, in Turkish the place name-forming suffix -iye triggers lengthening of the vowel /a/ in stem-final open syllables, as in Murad (name) /murad/ – Murad-ije (place name) [mura: dije] (Inkelas 2014: 33). This vowel lengthening is the exclusive property of the morphological construction with the suffix -iye. Another example is that in English compounds it is normally the first constituent that carries the main stress (the Compound Stress Rule). This implies the existence of a stress pattern that is specific for compounds, hence a property that is specific for this morphological construction.
15.3 The Hierarchical Lexicon and the Role of Subschemas A basic idea of Construction Grammar is that the grammar of a language is to be conceived as a multidimensional network of constructions of varying degrees of abstractness. As far as morphology is concerned, the lexicon comprises a hierarchy of constructional schemas and subschemas. These schemas dominate the individual existing complex words by which they are instantiated. The necessity of subschemas, in between the most abstract schemas and the individual words, is illustrated here by a set of Dutch NN compounds. Consider the following compounds of Dutch, all found through an internet search (July 20, 2015):
233
234
BOOIJ
(7)
pracht-baan ‘great job’ pracht-dag ‘great day’ pracht-cadeau ‘great gift’ pracht-kerel ‘great guy’ pracht-professor ‘great professor’ pracht-stad ‘great city’
In all these (right-headed) compounds the noun pracht ‘splendour, beauty’ expresses a positive evaluation of the referent of the head noun.1 This use of pracht as an evaluative expression can also be observed in a phrasal evaluative construction, the Dutch equivalent of the English an X of a Y-construction: (8)
een pracht van een meid ‘a splendor of a girl, a beautiful girl’ een pracht van een baby ‘a splendor of a baby, a beautiful baby’
What we observe here is that the noun pracht has acquired a productive, more abstract meaning of very positive evaluation in the morphological construction of N + N compounds (and also in the an X of a Y-construction). Hence, compound constituents may acquire productive, yet constructionbound, meanings. At first sight, this extension of the use of the bound meaning ‘great’ for pracht may look as a case of analogy, in which this use is extended from one established compound to another, new one. A clear example of such an analogical extension is the coinage of the German compound Mundwerker (Vater 2010): (9)
Hand-werker ‘manual laborer, who uses his hands to make a living’ Mund-werker ‘oral laborer, who uses his mouth to make a living’
However, in the case of pracht, there is no specific compound that can be designated as the source of the other ones. Hence, we are justified to conclude that the grammar of Dutch has acquired the following constructional schema: (10)
This schema may give rise to new compounds of this type, without a particular established compound with pracht being the source. Construction (10) is a subcase of the [N+N]N compound construction of Dutch. Like all [N+N]N compounds, the pracht-compounds are right-headed. This generalization can be expressed in a hierarchical lexicon, in which abstract constructional schemas are instantiated by subschemas that generalize over subsets of compounds. The individual compounds are instantiations of these subschemas. Thus we get the following hierarchy for pracht-N compounds, with prachtbaan as one of the many established instantiations: 1
An additional, related use of pracht can be found in the nomenclature of biologists for particular species of plants and animals, as in pracht-anjer, lit. ‘splendor-carnation, dianthus superbus’ and prachtschildpad, lit. ‘splendor-turtle, rhinoclemys pulcherrima.’
The Construction of Words (11)
This example also serves to illustrate how language users develop morphological knowledge. The starting point is the words they encounter in language use, and which are stored in lexical memory. Recurrent patterns in the set of stored complex words can be recognized, and will lead to the construction of schemas of various degrees of abstraction, some still partially lexically specified. “Repeated analogical extensions may over time lead to the emergence of a general schema . . . which invites further additions to the range of expressions occurring in this now partly schematic idiom” (Hilpert 2013b: 471). The repeated occurrence of various types of compounds in language use will bring the language user to store such compounds, and to construct compound schemas of various degrees of abstractness which are linked in hierarchies of the type exemplified in (11). In short, there are various levels of schematicity in usage-based approaches to morphology: micro-, meso-, and macro-constructions (Traugott 2008a, Barðdal 2011). This also relates to productivity, as pointed out by Barðdal (2008, 2011) “who has claimed that the productivity of abstract constructions can be seen as an inverse correlation of type frequency and semantic coherence, with highly abstract macro-constructions only arising if the underlying meso-constructions have a high type of frequency and a high degree of variance in semantic distribution” (Hoffmann 2013a: 315). Another example of this type of morphological patterning comes from Sranan Tongo, an English creole language spoken in Surinam. In this language the English borrowing man ‘man’ has acquired the more abstract meaning of agent, as illustrated by the following examples: (12)
borgu ‘guarantee’ siki ‘illness’ strafu ‘punishment’ wroko ‘work’
borgu-man ‘money-lender’ siki-man ‘patient’ strafu-man ‘prisoner’ wroko-man ‘laborer’
The word man as part of a compound has lost its male meaning component as well, as it can also be used in words denoting females, which confirms its more abstract ‘agent’ meaning: (13)
nai ‘to sew’ was ‘to wash’ yepi ‘help’
nai-man ‘seamstress’ was-man ‘laundress’ yepi-man ‘midwife’
The hypothesized general design features of language that underlie this model of the lexicon are the following three (Beekhuizen, Bod, and Zuidema 2013):
235
236
BOOIJ
(14)
a. experience: storage effects, frequency, conventionalization; b. heterogeneity; both concrete language constructs and schemas of various degrees of abstraction; small pieces and big pieces are stored; c. redundancy: predictable information may be stored.
This model of lexical knowledge does justice to the dialogic relationship between the competence of individual language users (I-language) and the language use (E-language) that they are exposed to (Taylor 2012, Booij 2014). The use of pracht as an evaluative modifier within compounds is similar to the use of evaluative prefixes, the difference being that prefixes are always bound, whereas the noun pracht can also be used as an independent word. Words with a construction-bound meaning are referred to as affixoids, as they are functionally similar to real affixes. The historical sources of prefixes and suffixes are words with a bound meaning, occurring as the left and right constituents of compounds respectively. For instance, the use of wise as a suffix, as in money-wise, derives historically from the noun wise ‘manner’ (Dalton-Puffer and Plag 2000). This diachronic observation is important, as it is an additional criterion for the adequacy of linguistic models that they should provide a sound basis for the understanding and explanation of processes of language change (Jackendoff 2011).2 Constructional schemas as exemplified in (10) and (11) have two roles. On the one hand, they have the function of motivation. A linguistic sign is motivated to the extent that the relationship between its form and its meaning is not completely arbitrary. The meaning of the established Dutch pracht-compounds is not arbitrary, but predictable, and hence motivated, on the basis of schema (10), and the meanings of their head nouns which are listed in the lexicon. That is, these compounds inherit their formal and semantic properties from the schema that dominates them and from their head words. Similarly, schema (10) inherits properties such as right-headedness and syntactic category from the highest schema in (11). Inheritance of formal and semantic properties is therefore a basic ingredient of CxM and the notion of the ‘hierarchical lexicon’ (Booij 2017). On the other hand, schemas serve to express how new complex words can be understood and formed. However, there are degrees of productivity: not all possible words (as defined by the set of morphological schemas) can be actualized easily. Some processes are marginally or semi-productive, and this is a kind of knowledge that language users possess and have intuitions about, based on their experience with actual language use such as the occurrence of hapaxes (Baayen 1992, 2009), and it has also to do with the register used (Plag, Dalton-Puffer, and Baayen 1999). The actual use of a morphological construction may also be hampered by blocking 2
Detailed studies of affixoids, both synchronically and diachronically, can be found in Booij and Hüning (2014), Hüning and Booij (2014), Norde and Van Goethem (2014), Van Goethem (2008, 2010), Namiki (2010), and Namiki and Kageyama (2016).
The Construction of Words
effects, the existence of synonymous words in the conventional lexicon of the language that impedes the coinage of a new complex word with the same meaning (Rainer 2013).
15.4 Mismatches between Tiers Form-meaning incongruities are important test beds for the adequacy of models of the architecture of grammar (Francis and Michaelis 2003). Morphological construction schemas provide the means for accounting for mismatches between the form and meaning of complex words. Above I mentioned the Greek exocentric compound filosofos ‘philosopher.’ In this compound, the meaning component ‘actor’ (the person who loves wisdom) has no corresponding formal constituent, and this meaning component has to be specified as a property of the compound as a whole. Almost all Romance languages also feature a productive pattern of exocentric [VN]N compounding, here illustrated by Italian [VN]N compounds with an instrumental meaning (Von Heusinger and Schwarze 2013): (15)
apri-porta lit. ‘open-door, door-opener’ tergi-cristallo lit. ‘wipe-windshield, windshield wiper’ trita-carne lit. ‘grind-meat, meat-grinder’
The instrumental meaning can be specified on the conceptual tier of a constructional schema: (16)
Hence, exocentric compounds are a case of systematic mismatch between the morphosyntactic tier and the conceptual tier which finds readily expression in the schema format. Mismatches between the phonological tier and the morphosyntactic tier abound in the inflectional systems of Indo-European languages, which exhibit many cases of violation of the one form–one meaning ideal. For example, the Dutch verbal suffix -t, as in (hij) werk-t’ (he) work-s’ expresses four properties: third person, singular number, present tense, indicative mood. Inflectional case endings on nouns usually express both a specific case property and a value for number, as in Latin mens-am ‘table, accusative singular.’ These are two examples of the general phenomenon of cumulative exponence, the expression of more than one morphosyntactic or morphosemantic property by the same morpheme. In derivational morphology we find empty morphemes that do not express a meaning of their own. For instance, a number of Polish denominal adjectives have an empty morph -ow- between the nominal stem and the adjectival suffix -ski (Szymanek 2010: 253): (17)
noun kat ‘hangman’
corresponding adjective kat-ow-ski
237
238
BOOIJ
szpieg ‘spy’ tcho´rz ‘coward’ hrabi-a ‘count’
szpieg-ow-ski tcho´rz-ows-ki hrabi-ow-ski
The constituent ow is not a part of the final suffix, as -ski is also used without ow, as in amerykan-ski ‘Americ-an.’ The presence of this empty morph can be specified in the formal part of the relevant constructional schema, and there will be no semantic component co-indexed with this morpheme ow. This example underlines that it is not the morpheme, but the word that is the primary symbolic unit of form-meaning correspondence in the lexicon. Morphemes in the Bloomfieldian sense of ‘minimal meaningbearing units’ are just a subset of the morphological constituents into which complex words can be decomposed. There are various types of subconstituents of words without a meaning of their own, which only contribute information through their being embedded in a complex word. Bracketing paradoxes are another case of form-meaning mismatch. In English derived nouns such as diner-out, hanger-on, looker-on, passer-by, and picker-up the deverbal nominalizing suffix -er is not attached at the right periphery of the corresponding particle verbs dine out, hang on, look on, pass by, and pick up respectively, but to the verbs themselves. Hence, there is a mismatch between the formal structure and the semantic structure of these derived nouns, as the suffix -er has semantic scope over verb and particle together. A diner-out, for instance, is someone who dines out. This semantic interpretation pattern is specified in (18): a combination of a particle and a nominalized verb is interpreted as the nominalization of the corresponding particle verb. (18)
≈
(the symbol ≈ is used here for presentational purposes to indicate the paradigmatic relationship). The meaning of particle verbs, denoted here by SEMm is to be specified by constructional schemas for the various types of particle verbs, as discussed in detail in Los et al. (2012). In (18) we specify the relationship between two schemas by means of co-indexation. A set of schemas linked by means of co-indexation of constituents is a secondorder schema, a schema of schemas. Such second-order schemas serve to express the systematic paradigmatic relationships between constructions. The action nominalization of Dutch particle verbs provides similar evidence for a systematic paradigmatic relationship between phrasal constructions and morphological constructions. Dutch particle verbs are lexical items, but not morphological constructions (just like English particle verbs). They are phrasal constructions, because the particle and the verb are separated in main clauses (Booij 2010, Los et al. 2012). Nominalizations
The Construction of Words
of particle verbs should not be seen as derived words, but as nominal compounds consisting of a particle and a nominalized verb. For instance, the nominalization of the particle verb aan-komen ‘to arrive’ has the form aankomst ‘arrival.’ This word should be analyzed as a nominal compound, with komst as its head: [[aan]Prt [[kom]V-st]N]N. This contradicts the structural analysis one might have expected from a semantic point of view, that is, [[aan-kom]V -st]N, because aankomst means ‘event of arriving.’ The relevant generalization is that Dutch particle verbs mirror the nominalization type of their corresponding base verbs, whether the nominalization is productive or not. This is illustrated in (19), data from Booij (2015): (19)
a. default: suffixation with -ing zend ‘to send’ zend-ing ‘mission’ uit-zend ‘to broadcast’ uit-zending ‘broadcast’
b. no formal change (conversion) val ‘to fall’
val ‘a fall’
aan-val ‘to attack’
aan-val ‘an attack’
in-grijp ‘to interfere’
in-greep ‘interference’
c. with vowel change grijp ‘to seize’ greep ‘grip’
d. stem change and/or suffixation gaan ‘to go’
gang ‘going’
kom ‘to come’ kom-st ‘coming’
af-gaan‘to fail’
af-gang ‘failure’
aan-kom ‘to arrive’
aan-kom-st ‘arrival’
slaan ‘to hit’
slag ‘hit’
aan-slaan ‘to touch’
aan-slag ‘a touch’
zien ‘to see’
zich-t ‘sight’
aan-zien ‘to view’
aan-zich-t ‘view’
The compound analysis predicts this pattern. Unproductive types of deverbal nouns must be listed in the lexicon, and hence, they are available for functioning as heads in compounds with a particle as modifier constituent. Thus, it is correctly predicted that the nominalization of particle verbs features unproductive types of nominalization. This analysis implies a second-order schema that expresses the systematic relationship between particle verbs and their nominalizations in Dutch, similar to the secondorder schema in (18) (V-NOM stands for the class of verbal nominalizations): (20)
≈
The shared indexes of the two schemas express that these schemas are paradigmatically related, and form a second-order schema. More examples of the necessity of such second-order schemas are given in Booij and Masini (2015). Paradigmatic relationships between different types of complex words thus require constructional representations of the relevant sets of complex words. Second-order constructional schemas also provide the right format to account for various types of non-concatenative morphology, that is, morphology that makes use of other means than the concatenation of words
239
240
BOOIJ
and affixes for the formation of words. For instance, vowel alternations are used in Germanic languages for both inflection and word formation. In English, nominalization of quite a number of mainly non-native disyllabic verbs can take place by means of stress shift: disyllabic nouns that are trochees (Bauer, Lieber, and Plag 2013: 204–05): (21)
verb (iamb) acce´nt disco´unt esco´rt torme´nt
noun (trochee) a´ccent díscount e´scort to´rment
Such morphological relationships can be expressed in a second-order schema that paradigmatically links two constructional schemas by means of co-indexation. The second-order schema for nominalization by stress shift can be represented as follows: (22)
≈ < [[(x)σs (y)σw]Vi]Nj ↔ [Action/Result of SEMi]SEMj>
where σs and σw stand for ‘strong syllable’ and ‘weak syllable’ respectively. The necessity of schemas, which define output forms rather than rules that operate on input forms, is confirmed by the phenomenon of word truncation. In many languages, nicknames and abbreviations are formed by means of truncation of the original long forms (Downing 2006: 58–64). Whatever the phonological shape of the long form, the truncated form usually has a defined shape. For example, German truncated nicknames consist of a disyllabic foot that ends preferably in the vowel /i/: (23)
Alkoholiker ‘alcoholist’ Amerikaner ‘American’ Ost-Deutscher ‘East-German’ West-Deutscher ‘West-German’ Waldemar ‘boy’s name’ Gabriele ‘girl’s name’ Trabant ‘car make’ Gorbatsjov ‘idem’
Alki Ami Ossi Wessi Waldi Gabi Trabi Gorbi
This form of truncation can be expressed as a paradigmatic relation between two types of nouns: (24)
≈
where x, y and z are phonological variables, and EVAL stands for the semantic operator of evaluation (which may take the more specific value of endearment). In the case of Waldemar, for instance, x = wal, y = d, and z = emar. This results in the phonological form of the endearment word Waldi [(wɑl)σ (di)σ].
The Construction of Words
15.5 Construction-dependent Morphology An important observation on the interaction of morphology and syntax, with implications for what the architecture of grammar should look like, is that the use of morphology may be dependent on specific syntactic construction, and hence be construction-specific. This applies to both inflection and derivation. Let us first look at an example from the domain of inflection. Dutch used to have a case system in which dependent nouns in NPs and PPs were marked as genitives. The ending -s was used to mark genitive case on singular masculine and neuter nouns. In present-day Dutch the case system has disappeared. Yet, the genitive marking has been preserved in a number of syntactic constructions (Booij 2010: ch. 9). Examples of the use of the ‘trapped’ old genitive -s in Dutch are the following (data from an internet search, July 27, 2015): (25)
a.
Het is niet de-s dominee-s om zo met publiciteit om te gaan It is not the-s reverend-s to so with publicity deal ‘It is not proper for a reverend to deal with publicity in that manner’ b. Dat door-praten is zeer de-s vrouw-s That on-talking is very much the-s woman-s ‘That going on talking is very characteristic of a woman’
What we observe here is the use of the construction [des N-s]NP with the constructional meaning ‘characteristic of N.’ This NP is used as a predicate, and often in combination with the negative adverb niet ‘not.’ The article des is the old gen.sg form of the masculine and neuter definite articles de and het. The use of des in present-day Dutch is restricted to idioms. What is remarkable here is that the use of the genitive marker -s on masculine and neuter nouns, and the use of the inflected masculine/neuter definite article des, has been extended to feminine nouns, as shown in (25b), and that this construction is productive in modern Dutch. The genitive markers on the determiner and the noun have been reinterpreted as markers of a specific construction. This diachronic development shows that the notion ‘syntactic construction’ is essential for a proper interpretation of this change, and for a proper synchronic account of this use of inflectional morphology. The relevant constructional schema is given in (26): (26)
This schema illustrates once more that schemas may contain both variables and fixed lexical items or morphological constituents of lexical slots. Syntactic constructions also play a role in derivational morphology because they may require derived words of a particular morphological form. An example is the Dutch PP-construction [op het [A-e]N af] (examples from Booij and Audring (in prep.) which expresses the idiosyncratic meaning ‘almost A’:
241
242
BOOIJ
(27)
a.
dun op het anorectisch-e af thin on the anorexic off ‘so thin that it is almost anorexic’ b. op het briljant-e af ‘almost brilliant’ op het gemen-e af ‘almost mean’ op het knapp-e af ‘almost handsome’ op het lullig-e af ‘almost silly’ op het onbehoorlijk-e af ‘almost indecent’
The noun in this PP-construction is always a nominalized adjective, derived by means of the suffix -e. This nominalization pattern is also used outside this construction. (28)
Het gemen-e is dat. . . The mean-e is that ‘The mean thing is that. . .’ b. Ik waardeer het briljant-e van deze redenering I appreciate the brilliant-e of this reasoning ‘I appreciate the brilliance of this reasoning’
a.
The crucial observation is that only this type of derived noun can be used in the noun-slot of the [op het [A-e]N af] construction. For instance, the following PPs are ill formed: (29)
a.
*op de [[smerig]A-heid]N af on the dirti-ness off ‘almost dirty’ b. *op de smerige eigenschap af on the dirty property off ‘almost dirty’
The construction op het A-e af is the unification of two independent constructions, the syntactic construction [op het N af]PP (exemplified by op het doel af ‘to the goal off, toward the goal’) and the morphological construction [A-e]N, and hence it inherits most of its properties from these two source constructions. This unified construction has acquired the specific meaning ‘almost A,’ and has thus acquired a life of its own. The use of this construction boosts the productive use of deadjectival nominalization with the suffix -e. This makes it a case of ‘embedded productivity’: word formation processes becoming (more) productive in specific morphological or syntactic constructions (Booij 2010: 47–49). A theoretical consequence of this analysis is that the internal morphological structure of words may have to be visible to syntax, because the noun slot can only be filled by nouns derived by means of the nominalizing suffix -e. Morphological constructions and syntactic constructions may be intertwined, and thus, there is no sharp boundary between morphology and syntax. This confirms the conclusions concerning the principle of Lexical Integrity in Booij (2009). Lexical Integrity defines the criteria for the
The Construction of Words
demarcation of morphological and syntactic constructs, that is, the criteria for wordhood of a linguistic construct. Syntactic rules cannot manipulate word-internal structure, such as movement of word constituents in syntax. What it should not exclude is the concept that syntactic rules or constraints may require access to information on the morphological composition of words. This is also desirable for semantic analysis. For instance, in a noun phrase like scientific researcher the semantic scope of the adjective scientific is not the whole word researcher, but only the nominal base word research. Hence, the morphological structure of researcher should be accessible for a proper semantic interpretation of this NP.
15.6 Inflection in Construction Morphology The idea of paradigmatically related constructional schemas argued for above for the domain of word formation is also relevant for inflection. The necessity of paradigmatically related schemas for the domain of inflection is obvious in the Word-and-Paradigm approach to inflection (Blevins 2006). In this approach, the forms in the cells of an inflectional paradigm are not computed on the basis of an abstract stem to which the inflectional endings are added. Instead, these forms are computed on the basis of principal parts of the paradigm. A schoolbook example is the way in which Latin noun declinations work. The nominative plural of rex ‘king,’ for instance, is computed by starting from the genitive singular form reg-is which is the revealing form: we compute the correct plural form reg-es by replacing -is with -es. A particular inflectional form may play two different roles in accounting for the construction of inflectional forms. First, particular inflectional forms or a combination thereof may be used to identify the inflectional class to which a word belongs (Stump and Finkel 2013). For instance, the genitive singular form of the Latin noun rex, reg-is, identifies this noun as belonging to the third declension class. That is, reg-is is a principal part of the inflectional paradigm of rex. Second, an inflectional form may be used to compute the form of other cells in the same inflectional paradigm (Blevins 2006, Ackerman, Blevins, and Malouf 2009). For instance, the first declension nouns of Saami exhibit a pattern based on two principal parts: the genitive singular and the nominative singular. These noun forms are subject to gradation. If the nominative singular form is strong, and hence has a geminate, the illative singular and the essive form are also strong. In that case, the genitive singular has a weak form, with a single consonant (as in bihtta´ versus bihta ‘piece, nom.sg/gen. sg’). Conversely, if the nominative singular form is weak, the corresponding illative.sg and the essive are weak as well, whereas in that case the gen.sg form is strong (as in bargu versus barggu ‘work, nom.sg/gen.sg’ (Blevins 2006: 546)). In other words, morphological generalizations about such paradigms can
243
244
BOOIJ
only be made in terms of systematic paradigmatic relationships between cells of these paradigms. The relations between the nominative.sg, the illative.sg, and the essive can be expressed as paradigmatic correspondence relations between the form tiers of morphological schemas for Saami nouns that share the variable x: (30)
[x-a´]NOM SG ≈ [x-a´i]ILLATIVE SG ≈[x-a´n]ESSIVE
If the variable x stands for a strong stem with a geminate consonant, as in bihtta´, this geminate consonant will be predicted to recur in all three forms. Inversely, if x stands for a weak stem, as in bargu, it is predicted that this weak stem also shows up in these three inflectional forms. That is, these mutually implicative relationships between paradigm cells can be expressed straightforwardly by making use of schemas for fully specified inflectional forms and paradigmatic relationships between such schemas. Since Construction Morphology allows for both morphological schemas and their instantiations to be listed, it is possible to list inflectional forms that are completely regular, because the inflectional schemas will indicate their predictable properties. This is necessary because inflectional forms may function as principal parts, and because the form of one paradigm cell may be predicted from another. The possibility of storage of inflectional forms is also welcome from a psycholinguistic point of view. It has been shown, for instance, that regular plural nouns in Dutch and Italian exhibit a frequency effect in lexical decision tasks. This implies the lexical storage of these plural nouns (Baayen, Dijkstra, and Schreuder 1997). This sketch of the role of paradigmatic relations in inflectional morphology is in line with recent findings on the allomorphy of genitive forms of German nouns (Feheringer 2011). Some nouns have two genitive forms, for instance Ort ‘place’ (Orts or Ortes). If one of the forms is strongly preferred, this pattern recurs in the genitive form of compounds with these nouns as their heads. For instance, Spiel ‘game’ strongly prefers -s, and this strong preference recurs in compounds such as Fußballspiel ‘soccer game’: the gen.sg form Fußballspiel-s is the most frequently taken option. This effect is explained by the assumption that these genitive singular forms are stored (otherwise there cannot be a frequency effect), and that they can be paradigmatically related to corresponding heads of compounds.
15.7 Summary and Conclusions In this chapter a number of arguments have been presented for a constructionist approach to morphology: the holistic properties of morphological constructions, the need for schemas and subschemas instead of rules in order to account for regularities and subregularities in a hierarchical lexicon, the necessity of output-based schemas for various
The Construction of Words
forms of non-concatenative morphology, and the role of systematic paradigmatic relationships between words, and between morphological and phrasal constructions. Although the focus of this chapter was on word formation, the concepts introduced appeared to be relevant for a proper analysis of inflection as well. The model of Construction Morphology briefly presented starts from the assumption that morphological knowledge is based on language use, and on awareness of lexical relatedness among stored words (Spencer 2013). Thus, this model fits very well into the broader enterprise of cognitive approaches to the architecture of grammar.
245
16 Lexical Semantics John R. Taylor
Lexical semantics is the study of word meanings. The topic is not an easy one to research. Unlike pronunciations, which are public, word meanings cannot be directly observed. They are private affairs, lodged inside each speaker’s head, and possibly not even available to introspection, let alone interpersonal validation. To equate meanings with ‘concepts’ or ‘ideas’ – to say, for example, that the meaning of shrub is the concept SHRUB, or the idea of a shrub – is to skirt around the issue. For unless we are able to state the nature of concepts, we are merely playing semantics (in the nontechnical and pejorative sense of the word). Indeed, there is quite a long line of scholars (Bloomfield 1933, Quine 1960, Lyons 1968, Wittgenstein 1978) who advise against the study of meanings as things in the head. Some doubt whether it is profitable even to speak of word meanings at all (Austin 1961). In this chapter I review three approaches to the issue. The first two involve the study of observables, the third returns to the question of meanings as mental entities. Although in principle distinct, they overlap to some extent, and I shall treat them as complementary. The first approach appeals to the notion that we use words to refer to things and situations in the world. This approach leads us to examine the correlations between words and the situations in which they are used. The second approach looks at word meanings in terms of relations between words, especially as these are manifest in textual data, or can be elicited from subjects in controlled situations. Both of these approaches can be pursued ‘objectively’, in the sense that the data is public and research findings can be replicated and validated. Arguably, however, the two approaches merely address the symptoms of word meanings, not their underlying nature. We are thus led back to the quest for the elusive mental entities which we are unable to access directly. Whatever we have to say about these entities, however, will need to be consistent with observable findings (cf. Riemer 2015).
Lexical Semantics
16.1 Words and the World There is a fairly obvious sense in which to know the word shrub means being able to apply the word correctly, namely, to things that are called shrubs. One points to a shrub and says that this ‘and similar things’ (Wittgenstein 1978: §69) are shrubs. (The throw-away line ‘and similar things’ raises the far from trivial question of what is to count as ‘similar,’ and on what basis.) First-language acquisition involves quite a lot of matching of words to referents, either real or depicted (as in children’s picture books) (see, e.g., Clark 2009). Indeed, there are a substantial number of words in every language whose learning depends crucially on familiarity with their referents. At issue are not only words which refer to things which furnish the world (such as shrubs), but also those words whose understanding appeals to visual, tactile, or otherwise embodied experiences. As argued long ago by Locke (1979 [1690]), you can only understand ‘physical-quality’ words (Goddard and Wierzbicka 2014) like hot, hard, heavy, and rough by direct acquaintance with things exhibiting these qualities. Before proceeding, a few preliminary comments are in order. First, the term reference. I take this in the ordinary sense, whereby a linguistic expression is intended by the speaker, and is taken by the hearer, to pick out some entity/ies or situation(s) in the world. There are, to be sure, some well-known philosophical problems, concerning, for example, the indeterminacy of reference, famously illustrated by Quine’s (1960) story of the field linguist’s encounter with a rabbit; gavagai, uttered by an informant in the presence of a rabbit, could refer to a potentially indefinite set of things, and even a lifetime’s experimentation would not be sufficient to test each of the hypotheses. Acquisition researchers – for example, Bloom (2000) – address the issue by proposing that learners operate with a number of interpretational heuristics which severely limit the search space. There is also the question of reference to entities which exist only in imagined, fictitious, or counterfactual worlds. This need not be a major stumbling block, provided we accept, with Fauconnier (1994), that reference is not to entities in the world as such, but to elements of mental spaces set up by speakers as a discourse progresses, and that these elements may, but need not, have counterparts in the ‘real’ world. Second, we need to bear in mind that words, in themselves, do not refer. Words refer only when used referentially. Mostly, this involves their being embedded in a larger referring expression, such as a noun phrase. It is the noun phrase the shrub which (when used appropriately) refers, not the decontextualized noun shrub. On the other hand, the referential burden of the shrub is carried by the noun, not by the article (Croft 2001: 258). Provided this point is kept in mind, there is probably no great harm in
247
248
TAY LO R
talking tout court of the referents of the word shrub, a practice which I shall follow below. Third, we need to take a broad understanding of what it is that words – and linguistic expressions more generally – can refer to. The potential referents comprise not only things, but also categories (or types) of things, as well as ideas, properties, quantities, abstractions, relations, and events. Thus we will want to say that verbs, verb phrases, and the like, if used appropriately, can refer. Today is muggy can be taken to refer to the atmospheric conditions prevailing when it is uttered, and it would be an entirely legitimate exercise to enquire into just what those atmospheric conditions are that justify the use of the word muggy, and the predication be muggy. There are two perspectives to the relation between words and their referents. One perspective goes from words to the world, the other from the world to words. We can ask, for this word, what are the kinds of things (situations, etc.) that it can be used to refer to? Alternatively, we can ask, for this thing (situation, etc.), what are the linguistic expressions that can be used to refer to it? A major finding of the words-to-world approach is that the referential possibilities of a word cannot, in general, be explained in terms of a set of features, or properties, which the referents necessarily share – though some referents may well have privileged status as the word’s prototypical examples. Similarly, on the world-to-words approach, it turns out that while some names may be privileged, how a thing is named is ultimately a matter of contextual factors, such as degree of speaker expertise and the point of the exchange in progress (Murphy 2004). Perhaps the most systematic application of the twin approaches has been in the domain of color (Berlin and Kay 1969, MacLaury 1995). On the first – the semasiological, or referential approach – subjects are asked to pick out, on a color chart, those color samples that can be named by the term, as well as the best example(s) of the term. The procedure operationalizes the notion of a word’s extension; it also enables the prototypical referent(s) to be identified. On the second – the onomasiological, or naming approach – subjects are shown a color sample and are asked to name it. By enabling the researcher to collect data on the most frequently used names, the approach operationalizes the notion of basic level term (for the notions of prototype and basic level, see Rosch 1978; also Taylor 2003, Murphy 2004). For obvious logistical reasons, there are not many domains in which the twin approaches can be systematically applied. Especially worthy of note, therefore, is the ingenious study by Geeraerts, Grondelaers, and Bakema (1994) of clothing terms, the referents being garments depicted in fashion magazines, the referring terms being culled from the accompanying text. Mostly, and again for logistical reasons, the naming approach has tended to be favored over the referential approach. An early example was Labov’s
Lexical Semantics
(1973) study of the naming of drinking vessels. Subsequent studies have looked at the naming of body parts (Majid, Enfield, and van Staden 2006), spatial relations as depicted in little line drawings (Bowerman and Pederson 1992), and events of cutting and breaking, as depicted in video clips (Majid and Bowerman 2007). A particularly fruitful line of research has involved the narrating of stories depicted in cartoon-like picture sequences, where the focus has been on how subjects describe the motion events that may be inferred from the pictures (Berman and Slobin 1994, Slobin 2004). These studies are based on the assumption that the cartoons depict a non-linguistic reality, the dependent variable being the way in which the reality is linguistically encoded. Extending this approach have been comparative studies of source texts and their translations (e.g. Slobin 1996), the rationale being that the translator, on reading a text in one language, mentally pictures the described situation (words-to-world relation), then considers how this conceived situation can best be represented in another language (world-to-words relation). Using the words-and-world approach, researchers have been able to pinpoint sometimes subtle differences between languages (Malt 2015). Indeed, the methodology has played a major role in the growing field of semantic typology (Evans 2010). There are some evident limitations to the approach. There are many words whose referents cannot be picked out in the world (whether real or imagined): soul, spirit, and mind, for example. There are also many words which do not obviously have referents at all, most notably, the little words of a language, such as the articles and negative particles. But even for words that do have identifiable referents, the approach ignores connotations and affective components. More often than not, speakers are concerned with conveying their attitude toward and their evaluation of a situation, rather than simply referring to it. Even in the case of color, the methodology fails to do justice to the cultural and symbolic values of colors and color terms; color terms are about more than just colors. A further issue concerns the selection and classification of the referents. The study of color is facilitated by the fact that the possible referents of color terms can be positioned on a three-dimensional grid, defined by the continuously varying dimensions of hue, brightness, and saturation, thus enabling extensive cross-language comparisons. In other domains, the conceptual grid can only be deduced post hoc. Take the case of spatial relations, prototypically expressed (in languages which have them) by prepositions. The little line drawings of Bowerman and Pederson (1992) have been immensely valuable for pinpointing subtle differences between pairs of languages, such as the different usage ranges of English on vs. Dutch aan and op (Bowerman 1996). However, once the range of languages is increased, the data become so dense that it may be difficult to discern broader patterns. The visual stimuli need to be reduced to a smallish number of
249
250
TAY LO R
dimensions (such as degree of inclusion, occlusion, support; see Moore et al. 2015), and the selection of these dimensions is ultimately a matter of the researcher’s intuition, aided, as the case may be, by cluster analysis and other statistical techniques. Similar remarks apply to the narrating of the cartoon stories, where there are so many degrees of freedom that it might be difficult to separate out the contribution of any single feature. I have considered the words-and-world approach in terms of what words refer to. The ‘world’ to which words relate also comprises the speaker and hearer, and indeed the more general context in which words are used. As Hudson (2007) observed, words are not only things but also actions, and as such require a specification of their actants. At issue here is word use in relation to well-known sociolinguistic variables, such as the demographic profiles of speaker and hearer and features of the interaction in progress. Some words are indexical of their speakers. To know the word wee involves, inter alia, knowing that the word is characteristic of Scottish and New Zealand speakers. Some words are typical of the discourse field: compare marriage, matrimony, and wedlock (Cruse 1986: 284). Some words are closely tied to the discourse context of their use. It would be fruitless to enquire into the referents of words like hello and anyway; the meanings of these words can only be discussed in relation to their function (interactional in the first case, discoursal in the second).
16.2 Words and Other Words De Saussure (2006[1916]) famously declared that the “value” of a word is given, not by its conceptual content, but by the network of its relations to other words (cf. Lyons 1968). Two kinds of relation are at issue: syntagmatic and paradigmatic. Syntagmatic relations have to do with the (kinds of) words that a word co-occurs with – the words it keeps company with (Firth 1957: 11). Paradigmatic relations are relations in absentia, having to do with word choices which could have been made (in a given context) but which were not. It follows that words in a paradigmatic relation belong to the same word class, this being, indeed, the standard criterion for identifying word classes in the first place.
16.2.1 Paradigmatic Relations Conceptually and methodologically, paradigmatic relations are the more problematic. According to the above account, the words which are able to occur in the following slots stand in a paradigmatic relation: (1) (2)
I saw a(n) ___ in the yard. It was a(n) ___ table.
Lexical Semantics
The items which can occur in (1) comprise count nouns, more precisely, nouns designating visible things that are likely to be found in yards, while the set in (2) consists of adjectives that are applicable to tables. These are not, on the face of it, terribly useful categories, either for the semanticist or for the syntactician. For the student of word meanings, paradigmatic relations are generally restricted to relations of synonymy, hyperonomy, and antonymy, and a few more, such as meronymy. While these relations seem, in many cases, to be intuitively evident (large is the opposite of small, sofa is a synonym of settee), many researchers have felt the need for more objective criteria for identifying the relations. Two kinds of criteria may be mentioned. The first is based on the relation of entailment, whereby sentence S1 entails sentence S2 if the truth of S1 guarantees the truth of S2 (though not vice versa). Note that entailment is a relation between sentences (or, if one will, between the propositions represented by sentences), not between words. To say that dog entails animal is but a shorthand way of saying that a sentence of the form This is a dog entails This in an animal. (The choice of the containing sentence is crucial: I didn’t see a dog does not entail I didn’t see an animal.) The entailment relation may be invoked in identifying cases of synonymy (whereby This is an X and This is a Y mutually entail each other), also in disentangling different kinds of opposites (complementary, reversive, converse, etc.; see Cruse 1986). Complicating the matter, however, is the fact that entailment relations are ultimately a matter of judgment and are embedded in the “fabric” (Quine 1951: 39) of our (contingent) knowledge and beliefs. Does father entail male? It certainly used to, but arguably no longer does. Indeed, once we leave textbook examples of assassinate and kill, dog and animal, intuitions tend to become rather unsure. What, for example, is the relation between build and construct? Does I built the tool-shed entail I constructed the tool-shed, and/ or vice versa? (And where does I made the tool-shed fit in?) The second criterion for diagnosing paradigmatic relations appeals to the occurrence of words in sentence (or phrasal) frames, a procedure pioneered by Cruse (1986). Thus, the normalcy of dogs and other animals may be taken to support the notion that dog is taxonomically subordinate to animal. Diagnostic frames were championed by Hearst (1992) as a tool for automatically extracting lexical relations from text; one simply searches for occurrences of X(s) and other Ys in order to locate examples of hyponymy. The results, however, are not always reliable. From the normalcy of dogs and other pets, it would be an error to conclude that dog is taxonomically subordinate to pet; not all dogs are pets, and This is a dog does not entail This is a pet. And what are we to say about the above example of build and construct? What light do the following (and allowing for changes to part of speech) shed on the paradigmatic relations between the two words? (3) (4)
Building, and other kinds of construction activities. Construction, and other kinds of building activities.
251
252
TAY LO R
The focus on diagnostic frames may be turned on its head, as when we look at the textual properties of the diagnostic frames themselves. Intuitively, we might suppose that the phrase or more precisely could serve as a diagnostic of the hyponymy relation, where the speaker zooms in to specify the meaning of a word in greater detail, as in (5). Paradoxically, however, the phrase can also introduce a contrary, as in (6). (5) (6)
an opera, or more precisely a Singspiel the interest rate scenario – or, more precisely, the lack of it
On the basis of these and similar examples, Renouf (2001) claims that the phrase is better understood, not as signaling a semantic relation, but in terms of its discourse function, such as summarization, modification, or specification. Perhaps the most elusive and difficult of the paradigmatic relations is the relation of near-synonymy. At issue are words which appear to be similar in meaning, yet are not obviously contrastive, in the sense that This is X is incompatible with This is Y. Examples are legion: small and little, high and tall, wide and broad, below and beneath, engine and motor, begin and start, to name just a few (see Taylor 2002b for a discussion of the issues). Mostly, where one word can be used, so can the other – but not always. In fact, a useful approach to the study of these words is to focus on the preferred linguistic contexts of their use, a topic to which we now turn.
16.2.2 Syntagmatic Relations Following Firth (1957), syntagmatic relations between words may be studied in terms of collocation and colligation. Collocation refers to the tendency for a word to co-occur with another word, either immediately adjacent or within a more broadly defined window: an unmitigated disaster, the foreseeable future, a blinding light, blindingly obvious. Colligation refers to the tendency for a word to co-occur with words of a certain syntactic category. Kinship terms, for example, are likely to co-occur with possessives, whether possessive pronouns (my son), ’s genitives (the girl’s mother), or of phrases (the father of the bride). More generally, colligation may be seen in terms of the syntactic frames favored by a given word. Put, for example, is virtually restricted to occurring in the frame [___ NP PP], where [PP] designates a goal location: put the book on the shelf, put the cat out, etc. Sophisticated statistical techniques have been brought to bear in assessing the degree of attraction (or, as the case may be, repulsion) between words, whether contiguous or within a more broadly defined window, and between words and the syntactic configurations in which they occur (Gries 2014b). These techniques exploit what is perhaps the most robust finding of corpus linguistic studies, namely that in any representative sample of texts, a linguistic unit, whether word, morpheme, or syntactic construction, will be associated with a distinctive and reasonably stable frequency of
Lexical Semantics
occurrence. Non-meaning bearing elements of a language, such as phonemes and syllables, display the same property – thus putting paid to the naive notion that frequency is driven by semantic (or referential) concerns. Frequency data make it possible to compare the actual frequency with which items co-occur with the expected frequency, the latter being simply the product of the expected frequencies of the items taken singly. It is not, however, just a question of assessing the conditional probability of one word in relation to another; syntactic and semantic considerations also play a role. As an adjective, unmitigated is already likely to occur in an [Adj N] configuration; the noun, moreover, is likely to designate a bad event. Even bearing these restrictions in mind, it is noteworthy that unmitigated preferentially selects disaster as its neighbor, rather than any of the other plausible candidates. An important finding of syntagmatic studies is that distributional facts pertain to word forms, not to lexemes. Singular eye distributes in the language quite differently from plural eyes – for one thing, idiomatic phrases (keep an eye out for, etc.) tend to make use of the singular form, whereas in reference to the organ of sight the plural form dominates (Sinclair 2004: 31). One might even conclude that eye and eyes are distinct lexical items, stored independently in the mental lexicon. We noted earlier that the relation between words and the world can be studied from two perspectives. The same goes for the relation between words and their linguistic context. We can ask, for this word, what are the contexts in which it is most likely to occur? Conversely, we may ask of a linguistic configuration, which words are most likely to feature in it? The word blindingly strongly evokes its collocate obvious. The reverse relation does not hold. Obvious is only moderately associated with blindingly – according to data from the COCA (Davies 2008), it logs in as only the 34th most frequent leftmost collocate. On the other hand, the phraseology [a ___ waiting to happen] very strongly evokes the word disaster, whereas disaster hardly at all evokes the waiting to happen frame (Stefanowitsch and Gries 2003). For many researchers, collocation and colligation have been studied as purely textual phenomena, that is, as properties of “external language” (Chomsky 1986). We also need to consider the implications for ‘internal language.’ Do speakers of a language ‘know’ the distributional facts of the words in their language? There is considerable psycholinguistic evidence that this is indeed the case (Taylor 2012). Hoey (2005, 2015) proposes that a word primes the contexts (both collocational and colligational, as well as situational) of its previous uses, and that these mental traces influence a person’s current understanding of the word and their use of it on future occasions. In brief, to know a word is to know, inter alia, the kinds of constructions in which it can occur, while to know a construction is to know, inter alia, the kinds of words which are likely to feature in it. A further crucial aspect of knowing a word is knowing its frequency in the language, both its overall frequency and its frequency with respect to a range of contextual and situational variables.
253
254
TAY LO R
16.2.3 Word Associations The relations between words may be approached from a different angle, namely, through the elicitation of word associations (Deese 1965). The technique is a familiar one and has been in use at least since Galton (1879): given the cue word bread, what is the first word that comes to mind? (For most people, butter.) Amongst the commonly attested relations are the syntagmatic (bread – butter), paradigmatic (good – bad), as well as some others, such as clang relations (butter – batter), highlighting the fact that words may also be associated through their sound (and perhaps also their spelling). Chomsky (1959) was critical of the Skinnerian view that intraword associations governed creative language production, and it is no doubt for this reason – along with a more general turn against behaviorism – that the association paradigm has tended to be neglected by mainstream linguists and semanticists (whether cognitive or otherwise). On the other hand, it can be argued that word association data do provide a unique insight into the structure of the mental lexicon, uncontaminated, as it were, by speakers’ communicative and expressive needs (De Deyne and Storms 2015). Word Grammar (Hudson 2007) and Cognitive Grammar (Langacker 1987, 2008a) both take linguistic knowledge to reside in a complex network of relations. Word association data can shed some light on the structure of these networks. One of their more remarkable features is the fact that some words appear to have a privileged status as local ‘hubs,’ characterized by a large number of incoming and outgoing links. This “small world” structure (Ferrer i Cancho and Sole´ 2001, Baraba´si 2003) serves to reduce the distance between any pair of items, thereby enabling, or at least facilitating, the (usually) rapid and effortless accessing and processing of words in both comprehension and production.
16.3 Words and Concepts Without doubt, studies of the relations between words and their linguistic context have greatly increased our understanding of words and their properties. Langacker, however, cautions that intra-linguistic relations may not be the whole story: The meaning of a lexical item . . . is not revealingly described just by indicating the systemic relations it bears to others (e.g. its participation in relationships of antonymy, hyponymy, and so on). At least as important – and arguably more fundamental – is the task of providing a positive characterization in terms of conceptual content and construal. Without a specification of content a system of oppositions is just an empty shell. (Langacker 2009: 224)
The challenge, then, is to provide a positive account of the content of word meanings.
Lexical Semantics
16.3.1 Conceptual Networks As noted at the beginning of this chapter, many scholars have had reservations about the very possibility of stating the conceptual content of words. To some extent, I think, these misgivings derive from the view (still popular in some circles) that words stand for fixed, discrete packages of information (their ‘meanings’), and that the meaning of any syntagm can be compositionally computed over these packages. Leaving aside the question of how to state the contents of these packages, there are good reasons to doubt whether such an approach can be made to work at all. Consider, by way of an example, Searle’s (1980) remarks on the verb cut. Cut your finger, cut the cake, and cut the grass refer to different kinds of activities (words-to-world relation), with different sets of truth conditions. You do not cut the cake by making a small incision in its surface, nor does cutting your finger involve lopping off a part of it. While we can no doubt perceive some commonality in the different uses of the verb, it would be difficult to assign to it a fixed invariant meaning, such that we can compute the full interpretation of each of the complex phrases. Yet we might hesitate to say that cut is multiply polysemous, with distinct meanings corresponding to each of its uses; we might be equally wary of claiming that the expressions are merely non-compositional idioms. The notion of network, introduced above, suggests how we might approach the issue. Just as the words of a language are structured by an intricate network of relations, so too are the elements of conceptual structure. For Langacker (1987: 57), a language is a “structured inventory” of linguistic units (phonological, symbolic, and semantic), the structure residing in networks of relations amongst the units, while Hudson (2007) sees no qualitative difference between the network of words, the network of concepts, and the network of relations linking the two. Indeed, a paradigmatic relation between words is, in many cases, not only a lexical relation, but also a relation between concepts. To be sure, the clang relation between butter and batter is a purely lexical relation. On the other hand, the hyponymy relation between dog and animal, or the meronymy relation between wheel and car, are above all conceptual relations, preserved by the substitution of synonyms. These kinds of lexical-cumconceptual relations are the basis of the WordNet project (Fellbaum 2015). To this extent, Langacker’s strictures on the inadequacy of intra-word relations lose some of their force. It is against the background of a network approach that we can note a number of concepts which Langacker introduced for the characterization of word meanings, such as the notions of profile and base, and the domain matrix against which word meanings take shape. Rim, spoke, and hub all involve the notion of ‘wheel’ – the words profile specific components of the broader concept, which in turn presupposes its components. Or take the word window. We can point to referents of the word (word-to-world relation),
255
256
TAY LO R
but we understand the word against background knowledge of architectural design, building technologies, and human living spaces, as well as the kinds of activities we perform with respect to windows; indeed, it is this kind of real-world knowledge that goes into knowing what counts as a possible referent of the word. Different facets of this knowledge may be highlighted by different uses of the word: break the window, install the window, sit in the window, paint the window, look out of the window, etc. To ask, what does window (or, for that matter, cut) actually mean, abstracted from these uses and the background knowledge that is invoked, seems to be a singularly misplaced question. Rather than regarding word meanings as nodes in a conceptual network (Collins and Loftus 1975), whose content is essentially fixed, we should regard words as providing “paths of access . . . to open-ended domains of knowledge” (Langacker 2008a: 42), keeping in mind that the knowledge in question may also include the content of previous discourse. Word meaning is therefore encyclopedic in scope, and inherently fluid and subject to ongoing negotiation. The illusion (and it is, I believe, an illusion) that words have fixed and stable meanings may be due to the fact that some paths are very well-trodden and likely to be activated in default (or zero) contexts, or in the absence of contrary indications. Equally, different contexts of use are liable to activate sometimes subtly different aspects of background knowledge; one might even say that no two uses of a word ever have precisely the same meaning (Langacker 2008a: 50, Wray 2015). An especially important aspect of background knowledge are the various ‘background situations’ and ‘event sequences’ which words evoke (Barsalou and Wiemer-Hastings 2005). Event sequences are relevant not only for words designating concrete objects, such as window, but also for abstract concepts like freedom and truth, and even for ‘logical’ words like or and not (Barsalou 1999). The matter is particularly pressing with regard to verbs, adjectives, and other non-noun words. As a generalization, we might say that verbs refer to events or states; these, however, presuppose participants and circumstances, which are only specified in individual acts of usage. In many cases, the events/states can only be properly understood against the background of expected scenarios and frames. To be sure, buy and sell are lexical converses (X sells Y to Z entails, and is entailed by, Z buys Y from X). At the same time, the two verbs are understood against the knowledge of a commercial transaction, involving buyer, seller, goods, price, and possibly bargaining, as well as notions of possession, contractual obligations, and even monetary systems, thus invoking wider domains of law and finance (Fillmore 1975). An especially interesting set of words are risk, danger, threat(en), and safe (Sweetser 1999). These invoke a scenario in which a person’s well-being is threatened by some unnamed contingency. It is by invoking and by filling in the details of this scenario that we are able to get an interpretation of safe beach, safe house, safe medical procedure, and so on. The effects are
Lexical Semantics
sometimes paradoxical. To risk your life entails risking death (and vice versa). The German warning Lebensgefahr (‘life danger’), which one sees attached to electricity pylons, corresponds to French Danger de mort (‘danger of death’). Lakoff (1987) addressed these kinds of issues in terms of his notion of Idealized Cognitive Models (ICMs) – stereotypical scenarios by means of which we make sense of the world. Thus, bachelor and spinster are understood against the ICM of (stereo)typical marriage practices. While X lacks Y entails X does not have Y, lack is understood against a scenario in which X is supposed to have, or in the normal course of events may be expected to have, Y. Or, for another example, take the contrast between frugal and stingy. Both words refer to a person’s attempts to limit their expenditure. The former is understood against a scenario in which being careful with one’s money is evaluated positively, the latter where it is subject to censure. The two words offer different construals of the same real-world situation. The above discussion represents a development of the world-to-words relation discussed in section 16.1. The ‘world,’ in this case, is not the world of observable, perceptible things and situations, but a second-order world, comprising (mental) conceptions of (kinds of) situations or (kinds of) events, in principle independent of any specific language. Indeed, this matter was already broached with respect to motion events and their lexicalization in different languages. Take, as another example, the notion of force dynamics – the idea that an eventuality is the outcome of (invisible) opposing forces (Talmy 1988a). With respect to the observable world, it would not be possible to distinguish The ball rolled down the slope and The ball kept rolling down the slope, or between He stood there for half an hour and He kept standing there for half an hour. The sentences with keep are understood against the assumption of an interplay of forces, between, for example, the ball’s inherent tendency toward inertia and a force counteracting its inherent tendency.
16.3.2 Definitions Lexicographers have not shied away from the challenge of offering contentful accounts of word meanings, namely in the form of definitions. Definition, in the Aristotelian tradition, was per genus proximum et differentia specifica. The definition was meant to capture the essence of a word’s meaning; the definition, in fact, was the meaning. Scholastic definitions presuppose a comprehensive ontology, ideally, a Porphryian tree of everything (Taylor 2002a: 130). In practice, this kind of definition is available only for a tiny portion of a language’s vocabulary – mostly names for biological species and other natural kinds – for which a scientific taxonomy is available. John Locke proposed a more modest aim; the point of a definition is to enable a third party to grasp the idea which the speaker
257
258
TAY LO R
intended by a word, this being “the only use and end of definitions” (Locke 1979 [1690], III, iv. 6). A definition is therefore a hint to understanding, and as such needs to be couched in terms which are simpler to understand than the term being defined. Johnson’s (1755) definition of horse as “a neighing quadruped,” though appearing to conform with the Aristotelian format (‘quadruped’ being the genus, ‘neighing’ the differentiating feature), manifestly fails on this score, since the notion of neighing rests on a prior notion of what horses are (and the sounds they make). In order to avoid this kind of circularity – endemic in dictionary definitions – Wierzbicka and associates have sought to define words in terms of a small set of universal “semantic primes” which, indefinable in themselves, are presumed to enter into, and indeed to make up, the meanings of all words in all languages. There is now a very considerable body of research (Wierzbicka 1996, Goddard and Wierzbicka 2014, and references therein) offering detailed definitions (or “explications”) of hundreds of words in dozens of languages, including many cases of near synonyms, near translation equivalents, and culturally marked items. These are remarkable for their explicitness and for their sensitivity to facts of usage. Indeed, as Goddard (1998: 31) stated, one hallmark of a good definition is that it will “predict the appropriate range of use of a word.”
16.3.3 Polysemy A major challenge of lexicography is to come up with a definition which is sufficiently general that it covers all uses of a word, and yet is sufficiently specific that it differentiates the word from its paradigmatic alternatives (in particular, its near synonyms). More often than not – especially for words in frequent use – the enterprise fails. This recognition leads to the postulation of polysemy – the claim that a word is associated with two or more distinct (though related) meanings. Polysemy has been a major topic in cognitive linguistic studies of word meaning, with a focus on studying the relations between these meanings and the factors that might have facilitated their emergence. As a matter of fact, the very notion of polysemy rests on the belief which I queried above, namely, that words are associated with fixed packages of content, polysemous words being associated with two or more distinct (though in some way related) packages. Particularly problematic has been the singling out of a unique ‘central’ sense and the criteria for identifying and enumerating all the related senses. (One need only compare dictionary accounts of common words, or consult the now voluminous literature on the word over, to appreciate the extent of the problems.) Here, I suggest, the relation of a word to its linguistic context again comes into play. A clue to the role of context comes from a simple, though remarkable fact. If (as is generally acknowledged) most words are polysemous to some degree, practically any utterance ought to be multiply ambiguous. If word X has n meanings,
Lexical Semantics
and word Y has m meanings, any syntagm containing X and Y ought to have n x m interpretations. Yet ambiguities and misunderstandings (jokes, puns, and concocted textbook examples of the kind She can’t bear children and It was a colorful ball excepted) are rarely encountered. This is because the intended reading of a word is usually guaranteed by the linguistic context. Thus, a first and essential step in identifying polysemy has to be the characterization of usage contexts, or ‘contextual profiles,’ two or more distinctive contextualization patterns being symptomatic of an equivalent number of meanings (see Gries 2006 on the many senses of the verb to run). An obvious indication of different contextualization patterns is differences in grammatical status amongst a word’s uses; essentially, these are a matter of colligation. Early studies of over (the ‘prototype’ being the account in Lakoff 1987) tended to put all uses of the word into the mix, and to view them as constituting a large category of senses, underplaying syntactic (syntagmatic) differences; for example, between over as a transitive preposition (with an overt nominal complement), as an intransitive preposition (without complement), as a verb particle, and as a verb-prefix, not to mention distinctive collocations such as over again and all over. (For a recent attempt to address these issues, see Glynn 2015.) To be sure, in deciding on a word’s syntagmatic properties, ever finer distinctions can always be made, down to a listing of each attested use. This is symptomatic of a more general problem encountered in any classificatory enterprise, namely, how finely the different contextual profiles of a word should be drawn, and how to balance the virtues of lumping with the necessity of splitting. Lumpers search for broad generalizations, which may not, however, be adequate for covering niceties of usage. Splitters, on the other hand, are liable to miss the generalizations. Langacker (1987, 2008a) incorporates the best of both, by proposing the coexistence, within the mental grammar, of both high(er) and low(er) level representations. Having identified distinct senses (uses) of a word, a further concern has been to explain, or motivate, the relations amongst the senses. There is a long tradition in lexical semantics which has studied these relations. These comprise not only metaphor and metonymy (favorites with cognitive linguists), but also specialization, generalization, pejorization, and (less commonly encountered) amelioration. The history, and current usage range, of just about any mid- to high-frequency word will typically reveal a complex interplay of several of these processes. An important factor in all of this is the process of conventionalization. What might initially have started out as a speaker’s idiosyncratic and context-dependent usage may, if taken up by others in the speech community, become a conventionalized value. These matters, though prominent in cognitive linguistic studies, are by no means new; they have been bread and butter issues with traditional philologists; see, for example, the case studies in C. S. Lewis’s magisterial Studies in Words (1960).
259
260
TAY LO R
16.3.4 Metaphor Another prominent feature of the cognitive linguistic enterprise has been the study of metaphor (Lakoff 1987, 1993). What we are dealing with here, I would suggest, is a kind of third-order world-to-words relation, the ‘world’ in this case being conceptual structures which map one domain of experience onto another. The mapping (generally) goes from the domain of observables (often having to do with spatial orientation and spatial relations) to more abstract, non-observable domains. The experiential domains are taken to be pre-linguistic, and possibly universal, with different languages elaborating the mappings in language-specific (and even lexeme-specific) ways. A crucial methodological issue for the Lakovian theory is whether independent (i.e. non-linguistic) evidence is available for the reality of the conceptual mappings. As a matter of fact, quite a lot of evidence has accrued, concerning, for example, the spatial basis of temporal concepts (Matlock, Ramscar, and Boroditsky 2005), the role of up-down orientation in emotional concepts (Meier and Robinson 2004), and the importance of embodied experience in metaphorical understanding more generally (Gibbs 2014). It should be borne in mind, however, that metaphor is not always from concrete to abstract; sometimes, it may be from concrete to concrete, as with the eye of a needle, the tongue of a shoe (i.e. the flap beneath the laces), or the head (of foam) on a glass of beer. Here, the metaphors fill an onomasiological gap; specialized ‘technical’ concepts need to be named, and this is achieved by projecting from features of a more familiar, more homely, source domain (often, as in the above examples, the human body).
16.4 Concluding Remarks Lexical semantics studies the meanings of words. What, then, is ‘the meaning of a word’? For Austin (1961), this was not a good question to ask. A much better question is: ‘How do we know that a person knows the meaning of a word?’ This question is, in principle, easier to answer: A person knows the meaning of the word if s/he is able to use the word appropriately, in conformity with the norms of the language, in ways that are accepted by other speakers of the language. There are two aspects to ‘using a word appropriately.’ One concerns the use of the word in reference to states of affairs in the world. The other concerns the linguistic environment of the word’s use. These ‘behavioral’ manifestations of word meaning can be studied ‘objectively,’ and, for many researchers, this is all there is to the matter. But we can also regard the referential and linguisticcontextual uses of a word as reflexes of its conceptual content. Both kinds of evidence point to the view that the conceptual content is not a fixed invariant package. Research on prototype effects, for example, reveals that a word’s extension is typically fluid, with fuzzy and fluctuating
Lexical Semantics
boundaries, and while favored collocations and colligations can usually be identified, finer distinctions of usage typically reveal considerable variation at the margins (for an instructive case study, see Fillmore and Atkins 2000). Indeed, without this kind of variation, languages would remain essentially static, frozen in time, and insulated from future change. Is it legitimate, then, to equate the meaning of a word with a concept in the head? If by ‘concept’ we mean a fixed conceptual package, which a word contributes to the expressions in which it occurs, the answer is ‘no.’ Words provide a gateway to networks of conceptual knowledge, which is selectively activated as occasion demands. Our understanding of complex, multiword expressions emerges in interaction with our knowledge of real-world practices and eventualities. The belief that words do have a fixed content reflects the fact that some activation routes are deeply entrenched, and liable to be invoked in zero or default contexts.
261
17 Cognitive Grammar Ronald W. Langacker
17.1 Introduction Cognitive Grammar (CG) is a particular descriptive framework within the broader movement of cognitive linguistics. Over a span of four decades, it has continually been refined and elaborated but has not been fundamentally revised. At the outset it was quite radical – even ‘outrageous’ (Hudson 1992) – given the theoretical climate in linguistics. Now it lies closer to the mainstream represented by the investigation of language in multiple disciplines. The objective has always been an account of language that is natural, unified, and comprehensive. CG is natural by being based on general cognitive capacities, relying only on known or demonstrable phenomena, and seeking compatibility with basic findings in other disciplines. It is unified in that the same capacities and descriptive notions apply to the various facets of language structure, so it does not posit rigid boundaries or separate components. It is comprehensive by grounding language structure in social interaction through discourse, so that linguistic elements are characterized in terms of their interactive and discursive functions. The development of CG is roughly divisible into two broad phases, each aimed at unification. The initial phase (represented by Langacker 1987, 1991) provided a unified account of lexicon, morphology, and syntax. These are inherently meaningful, comprising a continuum of symbolic structures, that is, form-meaning pairings. The second phase (with Langacker 2008a representing the transition) envisages a unified account of structure, processing, and discourse. Language structure is dynamic, consisting in processing activity at the neural, psychological, social, and discursive levels.
Cognitive Grammar
17.2 Grammar as Symbolization CG is most closely akin to Construction Grammar (e.g. Fillmore 1988, Fillmore, Kay, and O’Connor 1988, Goldberg 1995, Croft 2001). In historical terms, they were developed largely independently but with mutual awareness of their affinities. Main points of convergence (still contentious to some degree) include the following: (i) constructions (rather than constructive rules) are the primary objects of description; (ii) lexicon and grammar represent a continuum of constructions, consisting in form-meaning pairings; (iii) constructions are characterized at different levels of specificity, lower-level patterns being at least as important as schematic ones. A main difference is the emphasis in CG on providing reasonably explicit descriptions of conceptual meaning (or at least certain aspects of it), including the semantic characterization of fundamental grammatical notions (Langacker 2005a). Meaning is the starting point for CG description. It is identified as conceptualization in the broadest sense of the term, encompassing any aspect of our experience, and as such is dynamic, embodied, and interactive. Linguistic meaning is encyclopedic in that it presupposes a vast conceptual substrate – comprising knowledge, abilities, and contextual awareness – that it draws upon in a flexible and open-ended manner (Haiman 1980, Langacker 1987: section 4.2; cf. Wierzbicka 1995). The distinction between semantics and pragmatics, or linguistic and extralinguistic knowledge, is therefore graded rather than categorical. Still, linguistic meaning is not just conceptualization but represents its exploitation and adaptation for linguistic purposes (Levinson 1997, Langacker 2008a: section 2.1.3). An expression’s meaning depends not only on the conceptual content it invokes but also on construal, our capacity to conceive and portray the same situation in alternate ways (Langacker 2008a: ch. 3, 2015a). Both lexical and grammatical meanings consist in content construed in a certain manner. The difference – one of degree – is that lexical elements are typically rich in conceptual content, whereas grammatical meaning is largely a matter of the construal imposed on lexical content (Talmy 1988b). The many aspects of construal can be treated under the non-exclusive headings of selection, perspective, prominence, and imagination. Every element is selective in the sense of being limited in what it expresses. In describing someone as a woman, one is not describing her in terms of occupation, health, nationality, or disposition. With an intransitive clause like (1a), a change of state is selected from a complex event of causation, as in (1b). An essential aspect of selection is degree of specificity (conversely, schematicity): the precision and detail at which a situation is characterized. Compared to break, the lexeme shatter is more specific, while change is quite schematic. In the grammar of nominalizations, by is
263
264
LANGACKER
more specific than of in regard to participant roles; (1c) makes it clear that the teachers do the criticizing, but in (1d) their role is indeterminate. (1)
a. b. c. d.
The window broke. I broke the window with a hammer. She welcomes criticism by teachers. She resents the criticism of teachers.
Perspective pertains to the viewing arrangement (Figure 17.1), that is, the relation between the subject of conception (the conceptualizer) and its object (the entity conceived). This relation is asymmetrical: the object is salient as the ‘onstage’ focus of attention, whereas the subject (as such) is not attended to. Being the primary subjects of conception, the speaker and the hearer both have an essential role in linguistic meaning but remain implicit unless they are also part of the situation being described. Onstage facets of that situation – those directly relevant for apprehending the focused object – comprise an expression’s immediate scope (a portion of its maximal scope); for example, the immediate scope for Tuesday includes the conception of a week (the direct basis for its characterization) but not its relation to a month or a year. Basic aspects of the viewing arrangement are vantage point and orientation. For instance, yesterday and tomorrow both describe a day from the vantage point of one adjacent to it, the distinction residing in temporal orientation (Figure 17.2). The default vantage point is that of the interlocutors, whose
S
S = subject of conception O = object of conception IS = immediate scope (onstage region) MS = maximal scope
O IS
MS
Figure 17.1 Viewing arrangement
(a) yesterday day
(b) tomorrow day
t
day G/VP
day G/VP t
IS
day
day
IS
(c) the next day day
day
day
RP
t IS G/VP Figure 17.2 Vantage point
day
t = time VP = vantage point RP = reference point G = ground
Cognitive Grammar
(a) week d d d d d d d d d
(b) roof
(c) bride m
(d) groom f
m
f
t Figure 17.3 Profiling of things
interaction defines the deictic center and constitutes the ground, which is commonly invoked as an offstage point of reference. In this respect the next day differs from tomorrow by invoking an onstage temporal reference point made explicit in the prior discourse, as in (2a). Also matters of perspective are departures from the canonical viewing arrangement, where the interlocutors are together in a fixed location. Though seemingly inconsistent, (2b) is perfectly coherent as the message on an answering machine. And despite their stability, trees have temporal properties when described dynamically from the moving perspective of a traveler, as in (2c). (2)
a. She arrived on Tuesday and left the next day. b. I’m not here now. c. Trees are getting less frequent.
Numerous kinds of prominence need to be distinguished for linguistic purposes. One pivotal to lexicon and grammar is conceptual reference, known in CG as profiling (Figure 17.3). An expression’s profile is the focus of attention within its immediate scope – the main object of conception in the sense of being the entity it designates (refers to). For example, week designates a span of time comprising seven consecutive days. Roof evokes the conception of a building, within which it profiles the structure on top that covers it. Bride and groom have the same conceptual content, that of a marriage ceremony, but are semantically distinct due to their different profiles. The expressions in Figure 17.3 profile things (abstractly defined below). Other expressions profile relationships, exemplified in Figure 17.4. Tall specifies that a thing lies beyond the norm on a scale of height. Above and below designate spatial relationships in which two things have different positions along the vertical axis. Fall profiles a relationship that evolves through time, as a mover occupies successively lower locations. A second kind of prominence is conferred on relational participants. Within a profiled relation, the trajector (tr) stands out as the primary figure, that is, the entity being located, characterized, or evaluated (Talmy 1978a). Another participant, called the landmark (lm), is often invoked as a secondary figure for this purpose. Like profiling, these notions are essential for semantic description. Note, for example, that trajector/landmark alignment is the only factor distinguishing the meanings of above and below, which have the same content and designate the same relationship. Because so much of our conception involves imagination, many aspects of construal fall under this rubric. One is motor and sensory imagery (Kosslyn 1980); for example, part of the meaning of bacon is an image of
265
266
LANGACKER
(a) tall (ADJ)
(b) above (P) VERTICAL AXIS
HEIGHT tr
(c) below (P)
(d) fall (V) VERTICAL AXIS
VERTICAL AXIS
tr
Im
Im
tr
tr
norm
t Figure 17.4 Profiling of relationships
its smell. More generally, mental simulation is now being recognized as a basic component of conceptual semantics (Barsalou 1999, Bergen 2012). Farther removed from direct experience, and pervasive in thought and language, are the products of metaphor (Lakoff and Johnson 1980a, Ko¨vecses 2005) and conceptual blending (Fauconnier and Turner 2002). Moreover, even in describing actual situations the overt object of description is often virtual (or fictive) in nature (Talmy 1996, Langacker 1999b, Matlock 2001, Langacker 2005b, Pascual 2014). For example, (3a) does not refer to any actual boy, fish, or event of catching. The linguistic referents are virtual instances of their types, imagined entities construed as being representative of certain actual instances. Metonymy is similar in that an overt referent is not per se the real object of concern (Ko¨vecses and Radden 1998, Panther and Radden 2004, Panther 2005, Handl and Schmid 2011). Thus (3b) describes the book (not its famous author), and (3c) indicates that they drank the bottle’s contents. Metonymic construal is an inferential process of mentally accessing one entity via another in the relevant conceptual substrate. (3)
a. Every boy caught a fish. b. Lakoff is out of print. c. They emptied the bottle.
With a conceptualist semantics that accommodates construal, the inherent meaningfulness of grammar (Wierzbicka 1988, Talmy 2000a, 2000b) becomes apparent. A symbolic account of grammar can thus be envisaged. It is claimed in CG that the elements of grammatical description, though often quite schematic, all have conceptual import. Like lexicon, grammar resides in form-meaning pairings, representing a continuum of symbolic structures. In this respect the account is unified. It is natural in view of the general function of language as a means of symbolic expression. It is also restrictive, in that CG posits the minimum required to fulfill that function: semantic structures, phonological structures, and symbolic links between
Cognitive Grammar
the two. Importantly, CG does not claim that grammar is predictable from meaning; it cannot be, since the semantic import of grammar is itself an essential aspect of linguistic meaning. Rather than being autonomous, grammar consists in ways of construing conceptual content and symbolizing that construal – schematized patterns of symbolization instantiated by specific symbolic expressions. Though seemingly problematic in this regard, grammatical classes find their place in a general view of linguistic categorization. Elements are categorized on the basis of either intrinsic or extrinsic properties (i.e. structure or function). These can be phonological, semantic, or symbolic, defining classes with any degree of generality. Vowels, for example, form a category based on either their intrinsic phonetic nature or their function as syllabic nuclei, and within the general class more specific classes can be recognized (e.g. high vowels, back rounded vowels, vowels that trigger palatalization). For symbolic structures (e.g. lexemes), occurrence in particular grammatical constructions provides an extrinsic basis for categorization. In terms of inherent properties, these distributional classes have varying degrees of semantic and/or phonological coherence. While shared features are a motivation for class membership, they are generally not fully predictive of it; at the extreme (e.g. in a minor pattern of past tense formation), members may have nothing in common beyond the very fact of occurring in the construction. However, the issue of predictability is distinct from the issue of what it takes to describe a category, and even should membership be arbitrary, description requires nothing beyond an assembly of form-meaning pairings. In such cases, the specification of category membership comprises an array of lower-level constructions, each of which incorporates a particular member. Grammatical classes are sometimes equated with distributional classes as a matter of definition (Croft 2001: ch. 1). So defined, there are no universal categories – not even noun and verb – because constructions vary greatly from language to language. But the CG account, where grammar resides in the implementation of semantic functions, entails a broader view of grammatical categories for which distribution is merely symptomatic (not diagnostic or criterial). CG thus posits certain categories whose degree of universality correlates with that of their functions. In particular, nouns and verbs reflect the universal communicative functions of referring to objects and events (and more generally, to things and occurrences); showing the importance of these functions is the large proportion of grammar – including nominal and clausal structure – devoted to their implementation. Their pivotal role in grammar does not imply that nouns or verbs per se constitute distributional classes. They do contribute to the characterization of constructions: as partial descriptions, for instance, we want to say that demonstratives combine with nouns, and that tense is marked on verbs. But since constructions have their own conceptual import, which may be inconsistent with the meanings of
267
268
LANGACKER
certain category members, it need not be the case that all nouns or all verbs occur in any particular construction. Thus in English, demonstratives do not occur with pronouns (which count as nouns under the CG definition), and beware – being limited to imperative use – does not take tense. The central CG claim that certain basic categories have general conceptual characterizations runs directly counter to the well-entrenched doctrine that grammatical categories cannot be defined semantically. The arguments supporting this doctrine are invalid, as they presuppose an objectivist semantics that recognizes neither construal nor the obvious mental abilities invoked by the CG characterizations (Langacker 2008a: ch. 4, 2015c). One argument is that members of different categories can be used in reference to the same objective entity; for example, fall as a verb (he fell) or as a noun (his fall). This ignores the possibility of a meaning difference residing in alternate ways of construing the same conceptual content. One can plausibly argue that the verb focuses specifically on the spatial relationship as it changes through time (Figure 17.4(d)), whereas the noun views the change holistically and highlights its status as a single occurrence. Another argument is the obvious inadequacy of familiar proposals for semantic definitions; for example, that a noun names an object and a verb names an event – fall names an event as a member of either category. But this does not preclude the possibility of other definitions being viable. The standard candidates – notions like object, event, property, and location – can function as category prototypes but are too specific to qualify as schematic characterizations valid for all category members. General characterizations have to be independent of any specific content. Instead of conceptual archetypes like object and event, they are taken as residing in basic cognitive abilities inherent (and first manifested) in those archetypes. One is the capacity for grouping, in which multiple entities function as a single entity for higher-level purposes. Another is the apprehension of relationships, where entities are connected by mental operations (e.g. assessment of relative position). A third is temporal scanning (Langacker 2008c): the ability to track a relationship through time and detect any change. Together with attentional focusing, these capacities afford plausible schematic definitions of basic categories. A crucial point is that an expression’s category is determined by its profile (onstage focus of attention), not by its content overall. The CG proposal is that a noun profiles a thing, defined abstractly in terms of grouping; a verb profiles a process, defined as a relationship scanned through time (Figure 17.4(d)); and certain other classes, including adjectives and prepositions, profile non-processual relationships (Figure 17.4(a)–(c)). It is argued that these conceptual characterizations provide a rationale for the existence of basic categories, allow precise descriptions of how they relate to one another, and lead to revealing linguistic analyses (Langacker 2015b).
Cognitive Grammar
Consider nouns, for example. Clearly reflecting their semantic characterization are the many nouns that profile groups, where multiple entities function collectively as a unitary whole: team, herd, convoy, stack, orchard, collection, alphabet, constellation, etc. Physical objects are prototypical because their unitary nature predominates at the conscious level; moreover, an object comprises a continuous expanse of substance, whose apprehension as such (the registration of continuity) amounts to a grouping operation. Since the schematic definition makes no reference to bounding, it covers mass nouns (characterized by qualitative continuity) as well as plurals. It also applies to abstract entities; for example, the many components of a theory combine to serve an overall function. The key role of profiling is evident from a noun like bride (Figure 17.3(c)), which invokes a relationship as its essential content but is a noun because its referent is a person (a kind of thing). Deriving a noun from a verb consists in shifting the profile from the verbal process to one of its participants (e.g. complain > complainer), or alternatively, to the abstract thing obtained by viewing it holistically (e.g. [a] fall as a single instance of falling). Grammatical constructions allow the formation of complex symbolic structures out of simpler ones. Constructions run the gamut from highly schematic patterns to specific instantiating expressions. They reside in assemblies of symbolic structures linked by correspondences, with no inherent restriction on either their form or their complexity. In a typical construction, component symbolic structures are integrated to form a composite symbolic structure based on correspondences (overlap) between certain substructures. This is itself symbolic, as the phonological integration – for example, by combining elements in a certain order – symbolizes the semantic integration. Because composite structures can function in turn as higher-level component structures, grammatical organization is to some extent hierarchical. As viewed in CG, however, constituency per se is non-essential to grammar and often variable (Langacker 1997), semantic and symbolic relationships being more fundamental. By way of illustration, Figure 17.5 sketches the grammatical organization of the day before yesterday. This symbolic assembly instantiates a complex constructional schema representing two levels of composition: the formation of a prepositional phrase; and the use of that phrase to modify a noun. At the first level, the noun yesterday combines with the preposition before, which profiles a non-processual relationship of temporal precedence. Their integration is effected by a correspondence (dotted line) equating the nominal profile with the preposition’s schematic landmark. Shading indicates that the landmark functions as an elaboration site which (as shown by the arrow connecting them) the noun specifies in finer detail. A heavy-line box marks before as the profile determinant, that is, it imposes its profile – hence its grammatical category – on the composite structure before yesterday. This in turn functions as a component structure, combining with the day at the second level. The nominal
269
270
LANGACKER
the day before yesterday d
d
d G
t the day
before yesterday d
d tr
t
Im
d G
t
d tr
Im
t
d G
t before
yesterday
Figure 17.5 A symbolic assembly
elaborates the schematic trajector of the prepositional phrase, and since it functions as profile determinant, its referent – a particular day – is also profiled by the full expression. The implies the contextual uniqueness of this referent, which is thereby interpreted as the one (the only one) immediately prior to yesterday. Central to the CG vision is the demonstration that descriptive notions established independently for semantic purposes also prove efficacious for describing grammar. An expression’s profile (conceptual referent) determines its grammatical category. The semantic notions trajector and landmark – needed to distinguish opposing pairs like above/below, before/after, like/please, etc. – are reflected in the grammatical relations subject and object (Langacker 1991: section 7.3, 2008a: section 11.2). Fully general definitions cannot be based on semantic role or even discourse status. Instead, they are characterized in CG as expressions which specify the trajector (primary figure) and landmark (secondary figure) in a profiled relationship; yesterday is thus the object of before in Figure 17.5. The traditional grammatical notions head, complement, and modifier likewise have conceptual characterizations pertaining to symbolic assemblies. At a given level of organization, the head (in one sense of the term) is the profile determinant: the component structure that determines the profile (hence the category) of the composite expression. A complement is then definable as a component which elaborates the head (or a salient substructure thereof), and a modifier as one elaborated by the head. In Figure 17.5, yesterday is thus a complement of before, while before yesterday modifies the day. From the CG perspective, describing grammatical structures without describing their meanings is comparable to omitting lexical meanings from a dictionary. One indication that grammar and lexicon represent
Cognitive Grammar
a continuum of meaningful structures is that the same factors (e.g. profiling, trajector/landmark alignment, degrees of specificity) figure in their construal of conceptual content. They cannot be segregated, as a large proportion of lexical items (defined in CG as any fixed expression) exhibit internal grammatical complexity, whether it be morphological (taste-ful-ly), syntactic (take the bull by the horns), or in between ([pro- vs.] anti-abortion). As always, elements toward opposite ends of the spectrum seem very different. A typical lexeme has rich conceptual content; for example, the verb complain designates a specific type of process. By contrast, the derivational suffix -er evokes a highly schematic process and profiles a processual participant; in a form like complainer, it contributes semantically by imposing its profile on the specific content of the verb stem. Thus it is not unexpected that there should be neurological differences in the processing of structures rich in content and structures largely devoid of it (Pulvermu¨ller, Cappelle, and Shtyrov 2013). But construal is a basic aspect of lexical semantics, and a meaning consisting mainly in construal is still a meaning. From the outset CG has been a usage-based framework (Langacker 1987, Barlow and Kemmer 2000). Linguistic structures consist in learned patterns of processing activity. Well-established structures, referred to as units, are abstracted from usage events, that is, instances of language use in all their detail and complexity. Units emerge through entrenchment: the reinforcement of recurring commonalities (Langacker in prep.). They are part of a language – conventional linguistic units – to the extent of being thoroughly mastered by individuals and widespread in a speech community. Once acquired, units may be activated in subsequent usage events. The set of units activated in apprehending a given expression constitutes its structural description, that is, its interpretation vis-a`-vis the language. In either speaking or understanding, activated units effect the categorization of the target expression. For a given facet of its structure, potential categorizing units compete for activation and the privilege of categorizing it. The winner is determined by a number of factors – degree of entrenchment, extent of overlap with the target, prior level of activation, and contextual influence (Elman and McClelland 1984, MacWhinney 1987b) – so the categorizing unit and the target may not be wholly compatible. An expression is conventional (‘grammatical’) when it fully conforms to all the units evoked to categorize it. However, the demands of language use are such that discrepancies are normal and often not even noticed. Another aspect of the usage-based approach is that specific units have an inherent advantage over more schematic ones in the competition for selection as categorizing structure. A specific unit has more content, hence more potential points of overlap with a target, each of which tends to activate it. This is responsible for the well-known phenomenon that – other things being equal – a specific structure blocks (or preempts)
271
272
LANGACKER
a more general pattern that could otherwise apply. As a consequence, lowlevel patterns are no less important than general ones and may well predominate in language processing. Moreover, a general construction is seldom describable by means of a single constructional schema, as the range of potential instantiations is neither fully nor evenly exploited. The actual distribution is spelled out by an array of lower-level units, including both specific expressions and schemas at varying levels of abstraction.
17.3 Structure as Interactive Activity The second phase of investigation, which aims at a unified account of structure, processing, and discourse, places greater emphasis on the CG view of language as dynamic and interactive. Dynamicity is fundamental, as structure consists in patterns of activity at the neural, psychological, and social levels. It unfolds through processing time, and how it unfolds – its time course – is essential to its characterization. Just as speech time is a basic dimension of phonological structure, conception time is a basic dimension of semantic and grammatical structure. Equally fundamental is the role of interaction, reflected in CG in several ways. There is first the abstraction of linguistic units from usage events, which center on the interlocutors and their engagement in a physical, social, and discourse context. The interlocutors are also the subjects of conception who apprehend expressions, effect their categorization, and negotiate their contextual interpretation. Additionally, their interaction defines the ground, which has numerous manifestations in meaning and grammar. A speaker’s linguistic ability comprises a vast assembly of semantic, phonological, and symbolic units with varying degrees of entrenchment and conventionality. In a given usage event, certain units are activated in apprehending the target expression and are thereby part of its structure. The assembly is unavoidably affected in a number of ways: activated units are reinforced; they adapt to the context; new structures start to coalesce as units; and units that are not employed have a tendency to decay. These constant adjustments via usage represent another dimension of dynamicity. Units form an assembly by virtue of being connected. From the processing standpoint, where structures reside in patterns of neural activity, they are connected by overlap or association; in either case one structure tends to activate the other. From the linguistic standpoint, structures are connected in relations of symbolization, categorization, and composition. Symbolization is the association of a semantic and a phonological structure. Categorization and composition are based on some kind of overlap: partial overlap, inclusion, or the special case of inclusion where one
Cognitive Grammar
structure is schematic vis-a`-vis the other. In categorization, the relation is one of schematicity or partial overlap depending on whether the target fully conforms to the categorizing unit or is in some way incompatible with it. In composition, components are included in the composite structure unless additional factors (e.g. a metonymic construal) introduce some inconsistency. And because the profile determinant imposes its organization on the whole, it is usually schematic in relation to it. An assembly is thus a set of connected elements: the global assembly of latent units representing a speaker’s established linguistic ability at a given time; or the local assembly comprising units and structures activated in the apprehension of a given expression. Importantly, assemblies cannot be properly understood in terms of the metaphors most commonly employed in linguistic thought and theory – they are neither networks nor trees. They are better understood in terms of the fundamental phenomena responsible for networks and trees, namely connection and grouping. An assembly differs in several ways from a simple network (e.g. the map of a subway system). For one thing, the connected elements vary in degree of entrenchment. They are also non-discrete, as connection is usually a matter of overlap. Moreover, a network per se is static, its nodes and links being all in place at any one moment. By contrast, linguistic assemblies are inherently dynamic, consisting of patterns of activity taking place through time. Finally, whereas a basic network is a ‘flat’ structure such that the nodes are all on the same level, an assembly includes elements at different levels of organization: by virtue of being connected, elements constitute a higher-order element with the potential to participate in further connections. Grouping is a matter of that potential being exploited, so that connected elements do function as a whole for some higher-level purpose. And since the connections at that level also define a higher-order element, grouping provides the basis for hierarchical organization (Figure 17.6). However, an assembly is not a tree structure. One reason is that the same elements are grouped simultaneously in alternative ways based on different functions. For instance, the segments of albums are grouped as ((al)(bums)) on purely phonological grounds, and as ((album)(s)) for symbolic purposes. The assembly for the day before yesterday includes the crosscutting hierarchies in Figure 17.7. Semantic considerations motivate the groupings in 17.7(a): the entire structure day before yesterday describes
(a) Elements (c) Grouping / Higher-Level Connection (b) Connection
Figure 17.6 Connection and grouping
(d) Hierarchical Organization
273
274
LANGACKER
(a) Semantic Grouping the
day
before
(b) Prosodic Grouping the
yesterday
day
before
yesterday
Figure 17.7 Semantic vs. prosodic grouping
(a) Simple Connection a
(b) Emergent Properties e
b
a
b
(c) Seriality
(d) Constituency e2
a
b
c
e1 a
b
c
Figure 17.8 Seriality and constituency
the entity for which the specifies contextual uniqueness. The hierarchical organization in 17.7(b) (also in Figure 17.5) corresponds to prosodic groupings; their functional motivation pertains to the packaging of information for efficient presentation in discourse. Another thing distinguishing assemblies from trees is that the emergence of a distinct composite whole – the basis for constituency – is a matter of degree. To be sure, a connection necessarily results in something more than just the connected elements; the configurations in Figures 17.6(a) and 17.6(b) are not equivalent. At issue is whether this ‘something more’ goes beyond the mere fact of connection. A minimal connection is that of elements simply co-occurring in a processing window. A structure departs from this baseline to the extent that the whole has emergent properties (‘e’ in Figure 17.8) making it more than just the sum of the components. Several factors may be involved: particular kinds of connecting operations (e.g. comparison, scanning, or assessment of relative position); the modification of components; the incorporation of additional content; or construal affecting the composite whole (e.g. the imposition of an overall profile). Assemblies allow a unified treatment of two basic kinds of linguistic organization often regarded as polar opposites: seriality and constituency. Seriality combines the absence of emergent properties with the inherent temporality of processing. It arises to the extent that elements are apprehended individually apart from being connected by temporal sequencing (>). An extreme case is a recitation of the alphabet such that the letters are accessed in chain-like fashion, each serving to activate the next. Constituency arises when, at each hierarchical level, emergent properties result in a composite structure which is clearly distinct from its components and participates as such in higher-level connections. These are matters of degree, so pure cases of either are rare at best. In running through a chain of associations, there is usually some awareness of the larger whole they belong to. Standard syntactic tree structures (prime examples of constituency) incorporate serial access (‘linear order’) as one dimension.
Cognitive Grammar
Assemblies also represent a unified approach to the classic issue of structure vs. function. Structure consists in groupings, which are motivated by the functions they fulfill. The function of a grouping is just its place within a larger whole (Harder 2010), as an element participating in higherlevel connections and the groupings they establish. For a particular grouping, therefore, structure versus function is a matter of analytical perspective on the same assembly: focusing on the elements connected to constitute it, or on its role in further connections giving rise to a more inclusive grouping. Functional description is thus inherent in a full structural description, not something extraneous to it. By providing means of symbolic expression, assemblies subserve the global function of language: its role in thought and communication. Particular expressions are assemblies of semantic, phonological, and symbolic structures serving more specific functions. One dimension of functional organization is that of semantic, phonological, or symbolic structures functioning as parts of larger ones; for example, a vowel functions as syllabic nucleus, and syllables as components of a word. Another dimension pertains to symbolization per se. When semantic and phonological structures are connected to form a symbolic structure (its two poles), they function respectively as the symbolized and symbolizing elements. Their symbolic association is symmetrical from a processing standpoint, in that either pole serves to activate the other. But there is also a definite asymmetry, as the raison d’eˆtre of symbolic structures is to allow the expression of meaning – not to render vocalizations meaningful. From this standpoint symbolic structures are reasonably described as serving semantic functions. To say that a lexeme ‘has’ a certain meaning is equivalent to saying that it fulfills the semantic function of expressing that meaning. Lexicon and grammar reside in assemblies of symbolic structures that effect the implementation of semantic functions. Like the structures they inhere in, these functions range from highly specific to maximally schematic. Whereas a lexical noun (e.g. tangerine) evokes the conception of a very specific type of object, nouns as a class subserve the schematic function of evoking the conception of a thing (characterized abstractly in terms of grouping). Also like implementing structures, functions can be internally complex, decomposing into subfunctions at multiple levels of organization. For example, a nominal (‘noun phrase’) serves the referential function of directing attention to a particular thing (out of the infinite set of conceivable referents). A basic strategy for doing so combines the subfunctions of type specification (by a lexical noun) and grounding (by a determiner); for example, this tangerine. Type specification may itself comprise the subfunctions of specifying a basic type and refining it via modification, as in this overripe tangerine. Semantic functions can further be characterized with respect to either conceptual content or role in symbolic assemblies. Being an adjective, overripe describes a property, reflecting the global function of describing the world. But in overripe tangerine it also
275
276
LANGACKER
functions as a modifier, defined in constructional terms as a component structure elaborated by the head. Because symbolic structures implement and embody semantic functions, their organization tends to be parallel; in a nominal like this tangerine, symbolic components (this and tangerine) correspond to semantic subfunctions (grounding and type specification). They are not equivalent, however, nor are they always congruent. Observe that the same semantic function can be implemented by very different structures reflecting alternate strategies for fulfilling it. For example, the expressions in (4) are structurally quite diverse, but they all qualify as nominals because they subserve the function of nominal reference. Semantic functions may thus be assembly elements in their own right, with a certain primacy and a status independent of any particular structures. (4)
this woman; those with children; mine; yesterday; eating tangerines; that Bush is a genius
To the extent that structures fulfill the same function, they constitute a system of opposing alternatives (Halliday and Matthiessen 2004); for example, a vowel system for the function of syllabic nucleus, or a determiner system for nominal grounding. A system is a kind of category for which the shared function provides an extrinsic basis. Typically it also has an intrinsic basis, as elements with a common function are likely to be similar. Systems vary in their degree of integration, the main factors being the number of members and their organization into cross-cutting subclasses. Within a loosely integrated system there is often a tightly integrated core: simple vowels (rather than diphthongs or syllabic resonants) in the case of syllabic nuclei; and demonstratives (rather than quantifiers or possessives) for nominal grounding. A core system represents the default means of implementing the function in question – the most fundamental, straightforward way of doing so. It is thus a baseline for expansion into a more elaborate system providing a wider range of alternatives. This illustrates a pervasive aspect of conceptual and linguistic structure: organization into strata, each a baseline for arriving at the next, which comprises structures of greater complexity requiring additional cognitive resources. Linguistic manifestations are quite diverse: the expansion of a category from its prototype; morphological layering, as in (((help)less)ness); privative oppositions, such as [a] vs. [a˜]; ‘markedness’ more generally; and as special cases, even seriality and composition (Langacker 2015d). Strata can also be ascribed to usage events, where individuals engage in interaction by way of describing situations as part of a discourse. Each stratum is a baseline for the next, which presupposes it: ((((individual) interactive) descriptive) discursive). Symbolic elements are typically multifaceted, with semantic functions at more than one level. For example, piss and urinate have the same descriptive content but differ in their expressive/
Cognitive Grammar
emotive value (individual) as well as their social import (interactive). Canonical imperatives (e.g. Turn around!) combine the description of an event with the interactive force of ordering its execution. By describing a relation of causation or logical consequence, therefore serves the discourse function of connecting two clauses. The phenomena traditionally subsumed under grammar pertain primarily to description. But since structures usually conflate them, these dimensions of functional organization cannot be disentangled let alone segregated in distinct components. In CG, grammar is taken as reflecting the interplay of all these factors. A basic point is that description is inherently interactive: a cooperative endeavor of the interlocutors, who apprehend the described situation in relation to the ground (Figures 17.1 and 17.2). Their role as offstage subjects of conception is the basis for two essential descriptive notions. Profiling is characterized as the intersubjective focusing of attention effected through symbolization. And in a nominal or a finite clause, grounding indicates the epistemic status vis-a`vis the interlocutors of the profiled thing or process. A grammatical assembly represents an amalgam of descriptive and discursive groupings. The semantic functions implemented by the latter fall under several broad headings: speech management, the connection of utterances, information structure, order of presentation, and the packaging of content. Consider (5a), in which the situation described (that of the hearer’s sister being a lawyer) is grounded via the interactive function of questioning. Because it includes the expectation that the hearer will respond, questioning implements the discursive function of turn taking – a basic component of speech management – as well as effecting the connection of successive utterances. So also has a connecting function, conveying that the utterance it initiates somehow follows from the preceding one. As one aspect of information structure, English marks discursively significant content with unreduced stress (small caps); hence in (5)(a) the reduced stress on lawyer downplays this notion as one being already accessible in the discourse. Another aspect of information structure is the topic function of your sister. As the basis for interpreting a clause, a topic tends to be initial, in which case it has a framing function, but in (5b) it is appended as an afterthought. Even when seemingly free, the order of presentation always has semantic import, if only by inducing a particular order of conception. Also significant is the packaging of content into grammatical structures and prosodic groupings. Compared to (5a–b), the apprehension of (5c) represents a slightly different conceptual experience because it compresses the same content into a single clause and a single prosodic window. (5)
a. So your S I S T E R , is S H E a lawyer T O O ? b. So is S H E a lawyer T O O , your S I S T E R ? c. So is your S I S T E R a lawyer T O O ?
277
278
LANGACKER
Discursive structures tend to have little content of their own, being supervenient on the content provided by descriptive elements. In (5), for example, the informational status of lawyer supervenes on its descriptive content (its meaning) – status as ‘old information’ presupposes information. At the phonological pole, symbolization of its status by reduced stress supervenes on the lexeme’s segmental content. Thus discursive structure is mostly offstage, in the sense of inhering in the organization of descriptive content rather than being an object of conception in its own right; speakers attend to the form and meaning of lawyer, but have no explicit awareness of reduced stress indicating old information. Among the elements that do figure in description, those considered ‘grammatical’ likewise have little onstage content, hence are not “discursively addressable” (Boye and Harder 2012). So linguistic structures – be they lexical or grammatical, descriptive or discursive – are indissociable aspects of assemblies comprising all these dimensions of organization. Assemblies are dynamic, consisting in patterns of processing activity. Processing runs concurrently on different time scales; for example, those involved in the articulation of a syllable, the grouping of symbolic structures in a clause, and the connection of clauses in a discourse. On a given time scale, processing occurs in a series of windows within which connections are made and groupings emerge. This basic seriality gives rise to hierarchy when elements appearing in successive windows, on one time scale, are connected and grouped within a single window on a larger time scale (Christiansen and Chater 2015). Speakers are more aware of groupings on certain time scales than on others. Especially important in spoken discourse are processing windows whose duration is such that they typically coincide with clauses (Chafe 1987, 1994, Langacker 2001). Because the structures appearing in these ‘clause-sized’ windows are delimited phonologically by various prosodic clues, Chafe refers to them as intonation units. From the conceptual standpoint they represent the information currently active or being attended to. As shown in (6a), the construction in (5a) is normally expressed iconically by intonation units corresponding to the topic and the clause it frames; their presentation in successive windows mirrors and reinforces the mental progression constituting a topic relation. Since apprehending their connection requires a processing window of longer duration, seriality predominates owing to the lesser awareness of groupings on the larger time scale. (6)
a. b. c. d.
//So your S I S T E R // is S H E a lawyer T O O ?// //So is S H E a lawyer T O O // your S I S T E R ?// //So is S H E a lawyer T O O /your S I S T E R ?// //S O // is your S I S T E R a lawyer T O O ?//
The other examples show alternate ways of packaging and presenting the same descriptive content. In (6b) the order of presentation is no longer iconic for the topic function, but instead reflects the status of topic
Cognitive Grammar
specification as an afterthought; word order being a limited resource, semantic functions are often in competition for its exploitation. Shown in (6c) is the alternative of packaging the clause and the topic in a single intonation unit. They still appear in successive windows, but on a smaller time scale, where groupings are less evident. This compression has the phonological consequence that the windows are no longer delimited by an evident pause (//), but at most by a slight hesitation (/). It has the semantic consequence of downplaying the topic and afterthought functions. In (6d), as a result of so appearing by itself in a clause-sized window, it is fully manifested phonologically and its connecting function is highly salient. Prosodic groupings are an integral part of the symbolic assemblies constituting lexicon, grammar, and discourse. While prosodic patterns can be characterized in purely phonological terms, and may be independent of any particular grammatical construction, they are nonetheless incorporated in constructions and contribute to the implementation of descriptive and discursive functions. Usually, for example, a prepositional phrase forms a prosodic grouping, as does a nominal in which it functions as a modifier: //the man/at that table//. This packaging, which dovetails with descriptive semantic groupings and reinforces their symbolization, is inherent in the prototypes for these constructions. However, this coalignment is easily overridden by other functions or processing considerations. If the prepositional object is itself modified in this fashion, the content is broken up into three prosodic windows, as it will not all fit in two windows on that time scale: //the man/at that table/in the corner//. Though still part of the assembly, the descriptively motivated groupings must then compete for salience with those effected prosodically. This is not unlike the discontinuous manifestation of an idiom prototypically realized as a contiguous phonological sequence (e.g. the beans were spilled); the idiom can still be recognized because the encompassing grammatical assembly includes the semantic connections it is based on (cf. Osborne and Gross 2012). A sequence of processing windows in which the conceptual content appears can equally well be thought of as a single window moving through the conceptual landscape. For a unified analysis of grammar and discourse, this moving window metaphor is more perspicuous than the standard compositional metaphor of building a whole out of smaller parts. In Figure 17.9(a), DT is the descriptive target: that portion of the interlocutors’ mental universe which is under discussion in a certain discourse. W1 and W2 indicate the position of the moving window at successive moments (on a given time scale). From moment to moment there is usually some overlap in the content it delimits – the content currently being attended to. The window’s progression through DT represents the sequence of access imposed by linguistic expression. In Figure 17.9(b), successive windows function in turn as immediate scope for the nominal and clause in a topic construction: //the burglar //
279
280
LANGACKER
(a)
(b) DT
DT W1 b
W2 W1
b DT = descriptive target W = window IS = immediate scope MS = maximal scope
G = ground S = speaker H = hearer
W2
f
k
m tr
f
k Im IS2 he found the key
IS1 the burglar
H
S G
MS
Figure 17.9 Moving window metaphor
he found the key//. The organization is mainly serial; descriptively, the whole is just the sequence of components (with no basis for positing a distinct composite structure or a single overall profile). However, there is more to linguistic meaning than just onstage descriptive content. Also included are discursive connections: the coreference of topic and clausal subject; the topic’s framing function; and the sequence of access induced by order of presentation. Other contributing factors – providing the substrate for description – are unexpressed aspects of DT as well as the interlocutors’ interaction and their offstage role as conceptualizing subjects. These are all part of the expression’s maximal scope, that is, the relevant scope of awareness (a window on a larger time scale). The moving window metaphor reflects the CG view of language structure as being dynamic, interactive, and contextually grounded. It has different entailments from the compositional metaphor, in which expressions are constructed prior to semantic and pragmatic interpretation. In terms of composition, the sequence in Figure 17.9(b) gives rise to a coherent overall conception – where he is identified as the burglar – only after the nominal and clause have been processed. In the moving window alternative, an expression emerges from a substrate that includes the interactive context, the previous expression, and currently active portions of the descriptive target; a coherent conception is thus available from the outset. So instead of composition, the task is one of dissociating DT into overlapping ‘chunks’ for expressive purposes. The interpretation of he as referring to the burglar is an automatic consequence of the overlap. Discourse is a cooperative, inherently intersubjective process in which the interlocutors, through the selection, symbolization, and packaging of content, achieve a measure of alignment in their scope of awareness and focus of attention. As a discourse unfolds, the interlocutors negotiate a shared apprehension of DT and continually update it. An important basis for updating is
Cognitive Grammar
the content presented in successive clause-sized windows, which delimit the amount of information that “can be fully active in the mind at one time” (Chafe 1994: 69). However, the updating depends on connections established in a more inclusive scope of awareness (a processing window on a larger time scale); in Figure 17.9(b) the pronoun-antecedent relation spans two clausesized windows. A construction can vary as to the size of the encompassing window or the structures it subsumes; notably, the pronoun-antecedent connection can be established in either a single sentence, as in Figure 17.9(b), or across sentences, as in (7a). This is just one indication that grammar and discourse form a continuum. (7)
a. //A burglar/got into the house.//↓ //He found the key/in a pot.//↓ b. //He found the key.//↓ //In a pot.//↓
Another is (7b), where the content of a sentence is split between two clause-sized windows, each delimited by an intonational contour with a terminal fall (↓). This alternate discursive packaging illustrates a tradeoff between the hierarchy typical of clauses and the seriality typical of discourse. Figure 17.10(a) represents the canonical packaging of the content in a single, complete clause. It indicates the highest level of composition, involving the clausal core he found the key and the prepositional phrase in a pot. While this is not the only possible constituency, it does reflect the most likely prosodic grouping. All that matters here, though, is the formation of a distinct composite structure with emergent properties: taken as a whole, the expression profiles the event, not the locative relationship. It is a clause by virtue of having a processual profile and appearing in a single window of attention.
(a)
(b)
DT
DT
w
w2
f
b
k
p
b
f w1
k
p
IS tr
f
m
Im k
p f
m
k
tr f
m tr
k Im
tr Im
p
p
tr
Im
IS1
Im
//He found the key.//↓
//In a pot.//↓
IS2
//He found the key / in a pot.//↓ S MS H
S MS
G
Figure 17.10 Packaging and ellipsis
H G
S
H G
281
282
LANGACKER
Sketched in Figure 17.10(b) is the serial alternative of presenting the content in successive clause-sized windows with separate intonational contours. With serial access there is no basis for positing a specifically linguistic structure equivalent to the composite structure in 17.10(a) – instead of all the content appearing in one window with a single overall profile, it is described in separate windows with separate profiles. The essential connection (the key as trajector of in the pot) is established in a larger scope of awareness. And despite the individual presentation, the content represents a coherent conception because the descriptions pertain to overlapping portions of DT. The second expression in (7b) is a case of ellipsis, such that In a pot is understood as a shortened version of He found the key in a pot. In the CG account this does not involve deletion, but is just a matter of selective description given the discourse context (Langacker 2012b). As shown in Figure 17.10(b), the key factor is that the content invoked by the first expression (W1) remains active as the second is being processed. It is thus included in the second window (W2), only a portion of which is put onstage and explicitly described. That portion is selected based on information structure, as it is mainly limited to the content in W2 that does not appear in W1. Being interpreted in the shadow of the preceding clause, which remains active to some degree, In a pot is apprehended as though it were part of the sentence in 17.10(a). Sequences like (7b) instantiate a productive discourse construction. It conflates a number of other constructions pertaining to clause-internal grammar, information structure, selective description, prosodic packaging, and turn taking. This latter is associated with the intonation contours and their terminal fall – an indication of completion which the interlocutor can interpret as an opportunity to take the floor. Accordingly, Figure 17.10(b) shows the option of the speaker and hearer exchanging their roles after the first expression, that is, the discourse sequence is coconstructed. This is unproblematic in CG, given that the interlocutors and their interaction are an integral part of linguistic meaning and the dynamic assemblies representing grammar and discourse. More generally, this framework seems capable of accommodating the parallelisms and higher-level structures investigated in dialogic syntax (Du Bois 2014).
17.4 Conclusion The initial phase of CG research envisaged a unified account of lexicon, morphology, and syntax. It was shown that a symbolic account of grammar is not only possible but also leads to more cogent analyses and the resolution of classic problems, such as “raising” (Langacker 1995), pronominal anaphora (van Hoek 1995, 1997), and double-subject constructions (Kumashiro and Langacker 2003).
Cognitive Grammar
The current phase envisages a unified account of structure, processing, and discourse. The agenda being more ambitious, the early results are at best preliminary and the coverage only fragmentary (e.g. Langacker 2012a, 2012b, 2014, 2015c). The outlines of a synthesis are nonetheless emerging. It represents an elaboration of CG (not a fundamental revision) because essential features were present from the beginning: the emergence of structure from usage; the central role of the interlocutors and their interaction; and discourse as higher-level grammar (which does not stop at the sentence level). CG does not purport or aspire to be a self-contained formal model. The objective is rather more limited: a coherent conceptual framework capable of supporting a comprehensive and integrated account of language structure. Ideally the framework should provide an array of descriptive notions allowing a principled and reasonably explicit characterization of the full range of structures encountered in natural language. At the same time, it should relate these structures to the myriad factors giving rise to them, such as processing, acquisition, language change, and social interaction. The requisite synthesis is certainly a work in progress, but at least a foundation has been laid.
283
18 From Constructions to Construction Grammars Thomas Hoffmann
18.1 Introduction While many issues remain controversial in modern linguistics, virtually all linguists agree that language is fundamentally a symbolic communication system. The central units of this system are linguistic signs, arbitrary and conventional pairings of form (or sound pattern/signifiant) and meaning (or mental concept/signife´; cf. e.g. de Saussure 2006 [1916]: 65–70). Most of the words of a language (excluding deictic elements and onomatopoeia, which additionally also exhibit indexical as well as iconic properties) are prototypical linguistic signs: the German words Herz and its Hungarian equivalent szı´v, for example, have the same underlying meaning ‘heart,’ but different associated conventional forms ([hɛʁts] and [siːv]). In addition to linguistic signs, many linguistic approaches postulate independent and meaningless syntactic rules that combine words into sentences. In contrast to such ‘items and rules’ grammars, several approaches have emerged over the past thirty years that claim that arbitrary form-meaning pairings are not only a useful concept for the description of words but that all levels of grammatical description involve such conventionalized form-meaning pairings. This extended notion of the Saussurean sign has become known as the ‘construction’ (which includes morphemes, words, idioms, as well as abstract phrasal patterns) and the various linguistic approaches exploring this idea were labeled ‘Construction Grammar.’ Instead of assuming a clear-cut division of lexicon and syntax, Construction Grammarians consider all constructions to be part of a lexicon–syntax continuum (a ‘constructicon,’ Fillmore 1988; see also Jurafsky 1992 and Goldberg 2003: 223). Examples from this continuum are given in (1)–(5) (employing a fairly informal description of the form and meaning parts; for various different approaches to the representation of constructions, cf. Hoffmann and Trousdale 2013):
From Constructions to Construction Grammars (1)
(2)
(3)
(4)
(5)
morpheme construction Un-construction: FORM: [[ʌn]-X] ↔ MEANING: ‘not X’ (e.g. unfair, untrue, unfriendly) word construction Apple-construction: FORM: apple [æpl] ↔ MEANING: ‘apple’ idiom construction Take-for-granted construction FORM: [X TAKE Y fə ɡɹɑːntɪd] ↔ MEANING: ‘X doesn’t value Y enough’ (e.g. She took him for granted., Her father takes her mother for granted.) phrasal construction I Comparative construction FORM: [X BE Adjcomparative ðən Y] ↔ MEANING: ‘X is more Adj than Y’ (e.g. John is taller than you., A mouse is smaller than an elephant.) phrasal construction II Resultative construction FORM: [X V Y Z] ↔ MEANING: ‘X causes Y to become Z by V-ing’ (e.g. She rocks the baby to sleep., The firefighters cut the man free.)
All the constructions in (1)–(5) are pairings of FORM and MEANING, with the arbitrary pairing of the two poles being expressed by a bidirectional arrow ‘↔’: The word apple in (2) is a classic Saussurean sign and thus by definition qualifies as a construction. Since we can identify a FORM pole ([X TAKE Y fə ɡɹɑːntɪd]) as well as a MEANING pole (‘X doesn’t value Y enough’) for the idiom in (3), it too can be described as a construction. Note that the meaning of (3) is not completely compositional (it does not just mean ‘take something as given,’ but crucially implies that something is not valued enough). Due to this non-compositional property, idioms such as (3) would also be stored as lexical items in ‘items and rules’ approaches. In contrast to the word construction in (2), the idiom in (3) is a construction that is partly substantive (i.e. whose phonological form is fixed in several places, that is, [fə] and [ɡɹɑːntɪd]), but also contains ‘slots’ in its subject and object position which can be filled by various elements (cf. Many people take their parents for granted./Her father takes her mother for granted./ John and Zoe take each other for granted.). Constructions that contain such slots are said to be schematic and enable creative language use by allowing speakers to fill these templates with appropriate linguistic material. As can be seen, the comparative construction in (4) can be said to be slightly more schematic than the idiom in (3) since the former only has one substantive element ([ðən]) and several schematic slots (for the subject X, the form of BE, the comparative adjective and Y). Finally, the Resultative construction in (5) is a completely schematic construction since it only contains slots for the cause X, the verb V, the affected complement Y, and the resulting state Z (and thus licenses such diverse structures as She made it worse./They wiped the table clean./He painted the wall red.).
285
286
HOFFMANN
Construction Grammar (together with the related framework of Cognitive Grammar, Langacker this volume Ch. 17) can now be seen as the standard syntactic approach of cognitive linguistics. Yet, thirty years ago, when the first constructionist studies appeared, their claim that form-meaning pairings, that is, constructions, are the fundamental units of the human language capacity, were nothing short of revolutionary. In this chapter, I will explain the then marginal role of constructions and outline the reasons for their historical renaissance that lead to the rise of Construction Grammar. The next chapter will then focus on the current major Construction Grammar approaches, and outline their shared assumptions as well as differences.
18.2 The Renaissance of Constructions in Linguistics The analysis of syntactic structures as constructions, that is, as formmeaning pairings, can already be found in several traditional grammars (and the use of constructions in this sense can be traced back at least as far as the twelfth-century medieval linguists; cf. Goldberg and Casenhiser 2006: 343). Interestingly, even early Chomskyan transformational models (1957, 1965) employed the notion of ‘constructions’ and included construction-specific rules. In his seminal Aspects of the Theory of Syntax, Chomsky, for example, speaks of the ‘Verb-with-Complement’ (1965: 190) construction take for granted, the comparative construction underlying John is more clever than Bill (1965: 178) or a causative construction similar to (5) (1965: 189; cf. Hoffmann and Trousdale 2013: 2). Yet, once Mainstream Generative Grammar had shifted its focus from language-specific computational models of native-speaker competence to an investigation of the mental principles and parameters underlying all human languages in the various Principles-and-Parameters models (Chomsky 1981, 1995, inter alia) constructions had ceased to have any theoretical importance. From then onwards, constructions were seen as mere epiphenomena in Mainstream Generative Grammar, that is, only a collection of structures that are the result of the interaction of universal principles and parameter settings (cf. Chomsky 1995: 129). Yet, while proponents of the Principles-and-Parameters theory claim to have achieved a high level of descriptive and explanatory adequacy (in the sense of Chomsky 1965: 24–26), Culicover and Jackendoff (cf. 2005: 25–37) draw attention to the fact that this alleged success is only achieved by ignoring a great number of linguistic phenomena. The standard approach of the Principles-and-Parameters paradigm is “to focus on the core system, putting aside phenomena that result from historical accident, dialect mixture, personal idiosyncrasies, and the like” (Chomsky 1995: 20). As it turns out, however, this results in a great many phenomena – all of which must be part of a speaker’s mental representation of language – which
From Constructions to Construction Grammars
cannot be accounted for by the Principles-and-Parameters theory. Besides, as Borsley (2006, 2007) points out, in order to achieve analyses that can capture the differences between clause types (as well as other abstract constructions), Principles-and-Parameters approaches heavily rely on invisible elements that head functional projections (such as vP, T, or C; cf. Chomsky 1995, 2000a, 2000b) and which have specific properties and interpretations – an analysis “that is little different from a constructionbased account that associates the same set of properties and interpretations directly” (Sag 2010: 488, Hoffmann and Trousdale 2013: 3). While many cognitive-functional studies from the 1970s later received a constructionist interpretation (Wulff 2013: 277), it seems safe to say that the first set of major publications that advocated a Construction Grammar approach appeared in the 1980s (the most important ones, arguably, being Fillmore 1985a, 1988, Fillmore, Kay and O’Connor 1988, Lakoff 1987 or Wierzbicka 1987, 1988). These publications questioned the then dominant view of the strict separation of the lexicon as the repository of conventional meaningful words and syntax as the domain of regular, purely structural operations. On the one hand, these authors showed that there is a class of linguistic expressions that clearly exhibits characteristics of both lexical idiosyncrasy as well as syntactic creativity that cannot be adequately accounted for by classical ‘items and rules’ analyses, namely idioms. On the other hand, they argued that many syntactic configurations possessed construction-specific properties that could not be explained by independent syntactic rules, but required speakers to store these phrasal structures as form-meaning pairs. In other words, the lexicon had to include much more syntactic information, and syntax exhibited much more idiosyncratic lexical behavior than previously assumed.
18.2.1 From ‘Periphery’ to Centre: Idioms as Constructions Concerning idioms, “[i]t is not an exaggeration to say that construction grammar grew out of a concern to find a place for idiomatic expressions in the speaker’s knowledge of a grammar of a language” (Croft and Cruse 2004: 225). The central property of all idioms is that they have a nonconventional meaning (Nunberg, Sag, and Wasow 1994: 492–93) that requires speakers to store them in their mental lexicon. The meaning of idioms like kick the bucket ‘to die’ or spill the beans ‘to divulge the information,’ for example, cannot be computed using the conventional, literal meanings of kick and the bucket or spill and the beans and must therefore be stored in the lexicon (cf. Croft and Cruse 2004: 248–53; Jackendoff 2002: 167–68). In addition to this, once idioms are inserted in a sentence any compositional, conventional reading (that would result in something like ‘kick (with a foot)’ ‘a physical container’ and ‘spill’ ‘a certain type of food’) must be precluded. It might seem at first that the best way to achieve this is to treat idioms as frozen, inflexible chunks that cannot be modified
287
288
HOFFMANN
internally once they are passed on into the syntax. Such analyses have been advocated within Mainstream Generative Grammar (cf., e.g., Chomsky 1981: 146, n. 94; cit. in Jackendoff 2002: 168), and they do indeed work for so-called inflexible ‘idiomatic phrases’ such as kick the bucket, which are syntactically treated as single chunks (Nunberg, Sag, and Wasow 1994: 497). As Nunberg, Sag, and Wasow (1994: 496–97) point out, however, the majority of idioms are “idiomatically combining expressions,”1 which are not frozen units but which show some degree of syntactic flexibility (examples from the British National Corpus (BNC) and the Corpus of Contemporary American English (COCA)): (6)
a. There . . . the beans are spilled! (BNC EVN W_biography) b. For three years, the government wrestled with the beans Boesky had spilled, trying to come up with airtight felony counts. (COCA 1991 MAG NatlReview)
In (6a) the idiom spill the beans appears in a passive sentence with the beans in subject position and in (6b) the beans is a clause-external antecedent to a zero relative clause (cf. the beans that/which Boesky had spilled). These structures cannot be explained by assuming that spill the beans is merely a single frozen idiom chunk. Instead, Nunberg, Sag, and Wasow (1994: 496) point out that the parts of idiomatically combining expressions have nonconventional meanings that allow for the compositional computation of the (non-conventional) semantics of these idioms. In spill the beans, for example, spill means ‘divulge’ and the beans essentially denotes ‘the information.’ Since the beans is thus an NP with an independent meaning in this idiomatically combining expressions, it can syntactically be separated from spill in passives (6a) or relative clause constructions (6b) without the idiom losing its non-conventional reading – so (6a) means ‘The information was divulged!’ and (6b) can be rendered as ‘the information Boesky had spilled.’ Due to their non-conventional meaning, idioms such as spill the beans definitely have to be learned by speakers. Yet, they cannot be analyzed as frozen chunks, since they have “have grammatical structure, structure of the kind that we ordinarily interpret by appealing to the operation of the
1
As Croft and Cruse (2004: 232) point out, Nunberg, Sag, and Wasow’s (1994: 496–97) distinction between idiomatically combining expressions and idiomatic phrases is closely related to Fillmore, Kay, and O’Connor’s (1988: 504–05) “encoding” versus “decoding” idiom dichotomy. In contrast to Nunberg, Sag, and Wasow, who focus on the internal compositionality of idioms, however, Fillmore, Kay, and O’Connor address the issue of whether a hearer can interpret an idiom even if he/she has never heard or learnt it before. Encoding idioms are then said to be idioms that a hearer would not have known to be the conventional expression of a certain meaning, but might be able to guess their meaning. Upon hearing answer the door for the first time, for example, hearers can infer that this means ‘open the door.’ With decoding idioms, such as kick the bucket, this is not possible since hearers cannot figure out the meaning ‘die’ from any of the parts of the expression. Croft and Cruse (2004: 232) rightly point out that the encoding/decoding distinction is not a completely objective one, since the decoding of an idiom depends on the skills (or luck) of an individual hearer, which is impossible to predict. They therefore argue that the idiomatically combining expression/ idiomatic phrases dichotomy is to be preferred over the encoding/decoding one.
From Constructions to Construction Grammars
general grammatical rules” (Fillmore, Kay, and O’Connor’s 1988: 504). The beans, for example, in (6) clearly functions like a normal NP, which means that phrasal information, something that would be reserved for the syntactic domain in items-and-rules approaches, has to be part of the lexical entry of spill the beans. Instead of maintaining a strict lexicon–syntax division, a logical conclusion is to treat idioms as complex constructions, that is, form-meaning pairings that can also contain phrasal and syntactic information: (7) (8)
FORM: [X1 [KICKTNS ðə ˈbʌkɪt]2] ↔ MEANING: ‘X1 die2’ FORM: [X1 SPILLTNS_2 [ðə biːnz]NP_3] ↔ MEANING: ‘X1 divulge2 the-information3’ (modeled on Croft and Cruse 2004: 252)
Examples (7) and (8) provide constructional accounts of the idiomatic phrase kick the bucket and the idiomatically combining expression spill the beans, respectively. As the informal notation shows, the full string kick the bucket [KICKTNS ðə ˈbʌkɪt]2 receives a single interpretation as the predicate ‘die2.’2 In contrast to this, spill SPILLTNS_2 and the beans [ðə biːnz]NP_3 each are linked to an independent meaning (‘divulge2’ and ‘the-information3,’ respectively). It might be objected that idioms are only a peripheral area of linguistic knowledge. Yet, as Fillmore, Kay, and O’Connor note, idioms do not only make up a large part of a speaker’s knowledge, they also “interact in important ways with the rest of the grammar” (1988: 504). Moreover, idioms themselves are not, as we have just seen, a homogeneous group of linguistic expressions. In addition to the question of internal compositionality, idioms can also be distinguished based on whether they are lexically filled (substantive) or lexically open (formal/schematic; Fillmore, Kay, and O’Connor 1988: 505–06): (9)
(10)
The jig is up. ‘the game is up; it is all over’ (adapted from Jackendoff 2002: 169; semantic definition from the Oxford English Dictionary Online s.v. jig 5., www.oed.com (last accessed 30 January 2016) b. It was almost like the last spiteful act of a man who knew the jig was up. (OED: 1923, E. Wallace Missing Million xii. 100) a. It takes one to know one. ‘only a person with a given personality, characteristic, etc., is able to identify that quality in someone else’ b. *It took one to know one. a.
(from Croft and Cruse 2004: 233; semantic definition from the Oxford English Dictionary Online s.v. take P7 c., www.oed.com (last accessed 30 January 2016)
2
Following, for example, Jackendoff’s Parallel Architecture model (2002) as well as constraint-based approaches such as, for example, Pollard and Sag (1994) or Boas and Sag (2012), individual components across the form-meaning levels are indexed by subscript numbers in this and all following examples.
289
290
HOFFMANN
The idioms the jig is up (9a) and it takes one to know one (10) are both completely lexically filled, that is, substantive. Yet, while the former at least licenses different tense forms (10b), the latter is even more tightly constrained in that the verb take always has to be realized in its third person singular indicative present tense form. Again, a constructionist approach is able to provide a straightforward analysis of this subtle difference: (11) (12)
FORM: [ðə dʒɪg1 BETNS_2 ʌp3] ↔ MEANING: ‘the-game1 be2 up3’ FORM: [[ɪt teɪks]1 wʌn2 [tə nəʊ]3 wʌn4] ↔ MEANING: ‘only1 a-person-witha-given-personality2 is-able1 to-identify3 that-quality-in-someone-else4’
The FORM pole of (11) indicates that the verb position has to be filled with a tensed word form of the lexeme BE (thus, e.g., also licensing strings such as the jig will be up). In (12), however, the verb position is filled by a specific word form (takes) that allows for no other tensed realization of TAKE. Arguably of greater theoretical importance are formal/schematic idioms, which exhibit syntax-like creativity. English, for example, has a number of idioms that not only have an open subject slot, but also contain VP-internal slots (Jackendoff 2002: 168): (13)
(14)
(15)
(16)
X TAKE NP to task a. I am certainly going to take you to task for leaving my party so early. (BNC HGV W_fict_prose) b. [. . .] the Dutch, after nearly fifty years, still take the Germans to task, albeit jokingly, for having confiscated thousands of bicycles during the occupation. (BNC EBW W_pop_lore) X SHOW NP the door a. I had both the girls taught, but they were idle, and I showed the teacher the door in the end. (BNC CD2 W_fict_prose) b. West Ham never showed me the door. (BNC CH7 W_newsp_tabloid) X GIVE NP the boot a. I thought I told you I gave her the boot. (BNC G07 W_fict_prose) b. And next day Mayor John Hughes gave Mr Rees the boot. (BNC CH2 W_newsp_tabloid) X PUT NPi in proi’s place a. Some of the boys who had just left school used to be mischieful when they brought the farm-horses in, but the smith had a few tricks to put them in their place. (BNC G09 W_non_ac_soc_science) b. It was six o’clock in the evening, four hours since the Deputy Public Prosecutor had summoned Zen to his office in the law courts. When he arrived he had been told to wait, and he had been waiting ever since. He was being put in his place, softened up for what was to come. (BNC HTT W_fict_prose)
From Constructions to Construction Grammars (17)
Xe GIVE NP a piece of proe’s mind a. Boy, am I going to give him a piece of my mind when I see him. (BNC JY6 W_fict_prose) b. The Full Moon in your own sign on the 10th urges you to give others a piece of your mind. (BNC CB8 W_pop_lore)
The examples in (13–15) illustrate that the phrasal slot in these idioms can be filled by various NPs (you (14a), the Germans (13b), the teacher (14a), me (14b), her (15a), Mrs Rees (15b)). The idioms in (16) and (17) interact with syntactic principles in an even more complicated way: The idiom X PUT NPi in proi’s place (16) not only has a VP-internal NP-slot, but the additional requirement is that this NP must be co-referential with a following pronoun that functions as the determiner in a noun phrase headed by place, which itself is embedded in a prepositional phrase headed by in; coreferentiality in (16) is signaled by the index i; cf. them and their in (16a) and he and his in (16b). As (16b) shows, these idioms again exhibit syntactic flexibility and the schematic NP can, for example, also be realized as the subject in a passive sentence. In contrast to this, the expression Xe GIVE NP a piece of proe’s mind (17) also has an NP and a determinative pronoun slot proe, but in this case the pronoun must be co-referential with the subject of an active clause (as signaled by the index e; cf. I and my in (17a) and you and your in (17b). In addition to the above parameters, Fillmore, Kay, and O’Connor (1988: 505) mention another distinction relevant for idioms: while all of the above idioms follow the general syntactic rules of English (and can therefore be labeled ‘grammatical idioms’), there are also a considerable number of idioms that deviate from the standard syntactic structure in significant ways (Jackendoff 2002: 169): (18)
(19)
(20)
a.
all of a sudden Mm. I’m not worn no make-up for ages, ’cos me eyes seem to be like, itching all of a sudden. (BNC KB1 S_conv) b. by and large By and large, there is general agreement as to what counts as doing economics or pharmacy. (BNC G0 R W_ac_polit_law_edu) Far be it from NP to VP. (usually NP is me) a. Far be it from us to condone tax evasion (BNC CFT W_advert) b. Far be it from me to speak ill of the dead (BNC FR9 W_fict_prose) How dare NP VP! a. Now, how dare you say that! (BNC: KB1 S_conv) b. and one comment was how dare these men gang up against us. (BNC HUXS_interview_oral_history)
The substantive idioms all of a sudden (18a) and by and large (18b) function as adverbials despite the fact that nothing in their internal syntactic structure is adverbial (Jackendoff 2002: 169). Etymologically it is, of
291
292
HOFFMANN
course, possible to explain these structures, but synchronically they constitute syntactic anomalies. As the substantive idioms (19) and (20) prove, however, it is not even possible to claim that all extragrammatical idioms enter the syntax as frozen chunks. Both Far be it from NP to VP and How dare NP VP! have phrasal NP and VP slots that can be filled creatively be various lexical items (us and condone tax evasion (19a), me and speak ill of the dead (19b) and you and say that (20a) and these men and gang up against us (20b), respectively). Finally, Fillmore, Kay, and O’Connor (1988: 506) note that some expressions (like all of a sudden) are idioms without special pragmatic constraints (idioms without pragmatic point), while others require specific pragmatic contexts to be used felicitously (e.g. greetings such as Good morning or the opening formula for fairy tales Once upon a time. . .; these are called idioms with pragmatic point). Idioms – non-conventional form-meaning pairings that undoubtedly have to be learnt and stored by speakers – combine lexical as well as syntactic properties and vary along several important parameters (idiomatic phrases versus idiomatically combining expressions, substantive versus formal, grammatical versus extragrammatical, with and without pragmatic point). Many researchers believe that the best way to account for the full range of idiomatic expressions is a constructional analysis. Since a constructional approach can capture the numerous and varied types of idioms, the next logical step was to investigate whether the (arguably much fewer) regular syntactic rules of a language could also be accounted for by constructions (Fillmore 1985a, Fillmore and Kay 1993, 1995, 1999): It appears to us that the machinery needed for describing the so-called minor or peripheral constructions of the sort which has occupied us here will have to be powerful enough to be generalized to more familiar structures, in particular those represented by individual phrase structure rules. (Fillmore, Kay, and O’Connor 1988: 534)
18.2.2
Beyond Idioms: Partly Substantive, Partly Schematic Clausal Configurations as Constructions Once researchers extended their constructional view to so-called regular syntactic configurations, it turned out that far from being maximally general, these also exhibited idiosyncratic properties. In one of the earliest constructionist publications, Fillmore (1985a) pointed out a complex type of WH-interrogative construction that can be ‘peppered up’ by introducing interjections such as the heck, the devil, in heaven’s name, etc.:
From Constructions to Construction Grammars (21)
(22)
(23)
(24)
a. What the heck did you see? b. *You saw what the heck? (Fillmore 1985a: 81) a. What the devil did you fix it with? b. *With what the devil did you fix it? (Fillmore 1985a: 81) a. I can’t imagine what in heaven’s name she cooked? b. *I couldn’t eat what in heaven’s name she cooked. (Fillmore 1985a: 81) a. What the heck did you choose? b. Who the heck’s fault do you think it is? c. Why the heck did you choose it? d. *Which the heck did you choose? (Fillmore 1985a: 82)
Examples (21)–(24) illustrate the major idiosyncrasies of the construction: while wh-initial interrogatives (What did you see?) normally have a corresponding in-situ echo question (You saw what?), the complex Expletive-Phrase WH-interrogative construction3 (21a) does not have an equivalent echo question (21b). Moreover, the interrogative construction in question only licenses a single wh-item in initial position (22a), all other phrasal material associated with the wh-item cannot be preposed (22b). Next, the expletive phrases are grammatical in (embedded) interrogative clauses (23a), but not in superficially similar embedded free relative clauses (23b). Finally, the construction is acceptable with a range of wh-items (24a, b, c), but explicitly bars which (24d). All of these constraints cannot be captured by a general (phrase structure or transformational movement) rule. Instead, the phenomenon requires a constructional account: (25)
Expletive-Phrase WH-interrogative construction FORM: [[X[+wh] A1. . .]XP[+wh] [. . .]S]2 ↔ MEANING: ‘emphatic1 interrogative2’ A = Expletive Phrase (the heck, the hell,. . .) X≠ which (modeled on Fillmore 1985a: 83)
Example (25) is a highly schematic construction, which nevertheless has very specific constraints on its initial slots (the first element must be a whword (X[+wh]), but not which (X≠ which) and the phrases that can fill the A slot must come from a limited number of expletive phrase constructions (A = Expletive Phrase [the heck, the hell,. . .]). Due to its high degree of schematicity, its meaning is consequently fairly vague (specifying only that it is an interrogative that has an emphatic quality to it). If speakers must (and obviously do) learn such a schematic construction, then it is 3
The label Expletive-Phrase WH-interrogative construction is mine; Fillmore (1985a) provides no special name for this construction.
293
294
HOFFMANN
only a small step to claim that regular wh-interrogatives can also be stored in a constructional template (such as FORM: [[. . . X[+wh]. . .]XP[+wh] [. . .]S]1↔ MEANING: ‘interrogative1’). Another syntactic configuration that appears to require a constructional account is the Comparative Correlative construction (also known as the ‘Covariational-Conditional construction’; McCawley 1988, Michaelis 1994, Culicover and Jackendoff 1999, Borsley 2004, Den Dikken 2005, Sag 2010, Cappelle 2011, Kim 2011, Hoffmann 2013a, 2014): (26)
[the [more tired]comparative phrase1 Ben was,]C1 [the [more mistakes]comparative phrase2 he made]C2
The construction in (26) has phrases (more tired, more mistakes) in clause-initial position that in a corresponding declarative clause would be realized in post-verbal position (27). In this respect, comparative correlatives pattern with wh-questions (28) and wh-relative clauses (29): (27)
(28)
(29)
Declarative clause: Ben was [tired] Ben made [more mistakes] Wh-question: [What] was Ben? [What] did he make? Wh-relative clause: A pilot shouldn’t be tired, [which] Ben was. The mistakes [which] he made. . .
In Mainstream Generative Grammar (cf. e.g. Chomsky 1981, 1995, 2000a, 2000b), the structural similarities of (26) and (28–29) are explained by a single transformational operation (which has, e.g., been called A-bar movement or WH-movement). Consequently, in this approach the mental representation underlying comparative correlatives (as well as whquestions and wh-relative clause) is maximally abstract and general. Yet, as Sag (2010) pointed out, the various structures accounted for by A-bar/ WH-movement (which he labels ‘Filler-Gap constructions’) are characterized by great variation across a number of other parameters (presence of a WH-element, syntactic category of the filler phrase, grammaticality of subject-verb inversion, etc.; cf. Sag 2010: 490). In contrast to wh-questions and wh-relative clauses, comparative correlatives, for example, consists of two clauses (C1: the more tired Ben was/C2: the more mistakes he made), each of which computes a semantic differential (the difference between how much more tired Ben was at time t2 than at an earlier point in time t1/the difference between how many more mistakes Ben has made at time t2 than at an earlier point in time t1; Sag 2010: 525–26; cf. also Beck 1997, Brasoveanu 2008, Cappelle 2011). In addition to this, the second clause C2 is interpreted as the apodosis/dependent variable for the protasis/independent variable specified by C1 (cf. Goldberg 2003: 220; e.g. the more tired
From Constructions to Construction Grammars
Ben was → the more mistakes he made; cf. Beck 1997, Cappelle 2011). Moreover, the construction consists of specific fixed, phonologically specified material ([ðə. . .]C1 [ðə. . .]C2) as well as schematic, open slots which can be filled freely by the speaker to create novel utterances (cf. the more she slept, the happier she felt; the richer the man, the bigger the car). All these idiosyncrasies (for further properties of the construction, cf. Hoffmann 2014) can again best be captured by a constructional analysis: (30)
Comparative Correlative construction FORM: [ðə [] comparative phrase1 (clause1)]C1 [ðə [] comparative phrase2 (clause2)]C2 ↔ MEANING: ‘[As the degree of comparative phrase1 increases/decreases with respect to clause1]independent variable [so the degree of comparative phrase2 increases/decreases with respect to clause2]dependent variable in a monotonic way’
(adapted from Culicover and Jackendoff [1999: 567])
Interestingly, as Fillmore, Kay, and O’Connor (1988: 506) point out, the existence of an abstract construction such as (30) does not preclude the possibility that speakers also store more substantive realizations in addition to the schematic construction: (31)
The bigger they are/come, the harder they fall.
The comparative correlative in (31) has a non-conventional meaning that goes beyond the literal interpretation afforded by (30) (cf. the OED definition, which states that (31) is “used proverbially to suggest that the effects of a downfall or defeat are more severe or humiliating for those of great power or prominence,” www.oed.com s.v. big, adj. and adv. P2). Consequently, both the substantive idiom (31) as well as the schematic template (30) must be stored by speakers, which raises the question of how these constructions are connected in the constructicon – an issue I will return to in the next section.
18.2.3
Fully Schematic Constructions: Argument Structure Constructions So far, we have focused on constructions with at least one substantive element. There are, however, also structures that require the postulation of completely schematic phrasal templates: (32)
Could he shriek himself unconscious. . .? (BNC W_fict_prose CJJ) b. Firefighters cut the man free. . . (BNC W_newsp_other_report K55) c. he had often drunk himself silly (BNC W_fict_prose CDN)
a.
295
296
HOFFMANN
Shriek is normally an intransitive verb (one that does not require an object; cf. Leila laughed and shrieked BNC W_fict_prose AD9), yet in (32a) it has two obligatory post-verbal complements (himself and unconscious) that seem to depend on each other: while he shrieked himself unconscious is fine, neither *he shrieked himself nor *he shrieked unconscious would be grammatical. Cut and drink, on the other hand, can be used transitively, that is, with an object (cf. I cut my fingernails all the time BNC W_fict_drama FU6 or he drank a large whisky BNC W_fict_poetry FAS). Yet, the use of cut and drink in (32b, c) is clearly different from these transitive uses: while in I cut my fingernails, the fingernails are actually cut, the fire fighters do not cut the man in (32b). Similarly, you can drink a whisky, but not yourself, as in (32b). On top of that, all the sentences in (32) also give the result that the shrieking, cutting, and drinking action has on the object (it falls asleep, is set free or loose). Many theoretical approaches (including Mainstream Generative Grammar as well as various types of dependency grammar-inspired theories) assume that it is the main verb that has the main function of selecting the number and types of obligatory and optional syntactic arguments in a clause. Yet, for the examples in (32) it seems implausible to assume that the verbs shriek, cut, and drunk license the post-verbal arguments. Structures such as these led Goldberg (1995: 224) to argue that it is not the main verb alone that projects the syntax and semantics of a clause (Boas 2013a: 236). Instead, she postulated a family of so-called Argument Structure constructions – abstract, completely schematic constructions that encode various basic event construals together with their obligatory arguments. The Argument Structure construction licensing the examples in (32) is, for example, known as the Resultative construction (Goldberg 1995, 2006, Goldberg and Jackendoff 2004, Boas 2003a, 2005a): (33)
Resultative construction FORM: [SBJ1 V4 OBJ2 OBL3]Resultative Construction ↔ MEANING: ‘Agent1 causes Patient2 to become Result-Goal3 by V4-ing’
(adapted from Goldberg 1995: 189, 2006: 73)
The construction in (33) is completely schematic; it only consists of syntactic slots for the subject (SBJ), verb (V), object (OBJ), and oblique, that is, OBL, which includes adjective phrases like the ones in (32) elements. Note how the MEANING pole in (33) is more explicit about the semantic contribution of the slot elements than in the other constructions we discussed so far: in (33), the SBJ1, OBJ2, and OBL3 slot in the FORM pole of the construction are identified, respectively, as the Agent1, Patient2, and Result-Goal3 participants of the event in the MEANING pole. This type of semantic role information is, obviously, also relevant for the other types of constructions discussed so far (and was merely omitted above for
From Constructions to Construction Grammars
the simplicity of exposition), but is of crucial importance for the following discussion. Besides, the MEANING pole also includes parts (‘cause . . . to become’) that are not associated with any element on the formal level, and which, consequently, constitute an arbitrary property of the construction. Argument Structure constructions are thus completely schematic templates that need to be filled (or ‘fused’; Goldberg 1995, 2006) with more substantive constructions. One central question of this process is how exactly a specific verb (such as shriek, cut, and drink) with its specific semantic and syntactic argument selection requirements fuses with the slots of an Argument Structure construction. Goldberg (1995, 2006) addresses this issue by first of all distinguishing the semantic roles of Argument Structure constructions and verb constructions: while the semantic roles of Argument Structure constructions are referred to as ‘argument roles,’ the ones of a specific verb are called ‘participant roles’ (Goldberg 1995: 43–66; 2006: 38–43). One major difference of these two types of semantic roles is that Argument Structure constructions encode cognitive generalizations of basic event types (e.g. “something causing something to change location . . . an instigator causing something to change state . . . or an instigator moving despite difficulty” Goldberg 1995: 39). As a result of this generality, the associated argument roles are also fairly general (cf., e.g., agent, cause, goal, instrument, path, patient, recipient, theme, etc.; Goldberg 2006: 39). In contrast to this, participant roles are rich frame semantic roles entailed by the meaning of a specific verb (Goldberg 2006: 39). The verbs in (32), for example, could be claimed to have the following participant roles: (34)
a. shriek b. cut c. drink
Frame Semantics (cf. Fillmore 1977a, 1982, 1985b, Petruck 1996, Croft and Cruse 2004: ch. 2, Croft 2012: 11–13, Boas this volume Ch. 34) eschews truth-conditional definitions and instead considers the meaning of a word to comprise world and cultural knowledge, experiences and beliefs (Boas 2013a: 237). Since verbal meaning is thus supposed to encode rich, encyclopedic information, the frame semantic participant roles associated with verbs are also highly verb-specific: shriek entails someone who shrieks, a ‘shrieker’ (34a), cut implies a ‘cutter’ and a ‘cut.object’ (34b) and drink requires a ‘drinker’ and a ‘liquid’ that is drunk (34c). Concerning the roles of a verb, Goldberg distinguishes between profiled and nonprofiled participant roles. In contrast to the latter, profiled roles (indicated by bold type) are prominent, focal points of a scene that syntactically must be obligatorily expressed in all uses of the verb or, which, when not overtly expressed, are always contextually identifiable (a phenomenon known as
297
298
HOFFMANN
definite null instantiation DNI; Goldberg 2006: 39, Croft 2012: 364, Boas 2013a: 237). To illustrate the difference between profiled and non-profiled participant roles, compare the following examples: (35)
For three hours we worked and drank, worked and ate, worked and packed. (BNC A6 T W_misc) b. He placed the dead rabbit carefully on the block then turned and looked at her . . . Now she waited as he probed and cut and then compared what had been exposed against the diagram spread across the double page. (BNC FRF W_fict_prose)
a.
In both (35a) and (35b), the objects of drink and cut (the liquid and the cut. object) are not syntactically expressed. Yet, while in (35a) it is completely immaterial what was drunk, in (35b) it is obligatorily entailed that something was cut that is contextually retrievable (the dead rabbit in (35b)). This conclusion is supported by the fact that a rejoinder such as I wonder what they drank is fine for (35a), while #I wonder what they cut is strange for (35b) (which shows that the liquid role in (35a) is unspecified, while the cut. object role is already contextually specified; cf. Fillmore 1986, Goldberg 2006: 41). The cut.object role is thus a profiled role that has a DNI in (35b), while liquid is a non-profiled, optional role in (35a). The participant roles of verb constructions can thus either be obligatory (profiled) or optional (non-profiled). In contrast to this, constructional argument roles are always syntactically obligatory. Still, Goldberg also distinguishes profiled from non-profiled constructional roles, but here the difference merely concerns the syntactic function associated with the argument role (Goldberg 2006: 40, Croft 2012: 366): those that are realized as core syntactic arguments (subject or object) are profiled argument roles; the ones that are realized as oblique arguments are nonprofiled argument roles. Thus, in the Resultative construction (33), all three argument roles are obligatory, but only the agent and patient are profiled (since they are realized as subject and object, respectively), while the result-goal is a non-profiled role (realized as an oblique syntactic argument). The most important issue then is, obviously, how the fusion of verbs and argument structure constructions actually works. For this, Goldberg (1995: 50, 2006: 39–40; cf. also Boas 2013a: 237–38) postulates two principles that regulate this process: 1) the Semantic Coherence Principle specifies that the participant roles of verbs can only fuse with constructional argument roles that are semantically compatible; in addition to this, 2) the Correspondence Principle requires that lexically profiled and expressed participant roles must be fused with a profiled constructional argument role (i.e. a subject or object slot – the only exception being verb constructions with three profiled participant roles, in this case one of these can be
From Constructions to Construction Grammars
fused with a non-profiled argument role of a construction). In order to illustrate these principles, the examples from (32) are repeated in (36) together with participant and argument roles: (36)
a.
Could heshrieker=agent shriek himselfpatient unconsciousresult-goal. . . ?
b.
(BNC W_fict_prose CJJ) Firefighterscutter=agent cut the mancut.object=patient freeresult-goal. . .
c.
(BNC W_newsp_other_report K55) hedrinker=agent had often drunk
himselfpatient
sillyresult-goal
(BNC W_fict_prose CDN)
In (36) all argument role slots of the Resultative construction (SUBJ: Agent, OBJ:Patient and OBL:Result-Goal) are filled. In (36a) the shrieker participant role of shriek can be considered to be semantically compatible with the agent argument role, and thus complies with the Semantic Coherence Principle. Since this is also the only profiled role of shriek, which, moreover, merges with a profiled constructional role (the subject slot associated with the agent role), the Correspondence Principle is also fulfilled. The fusion of the verb and argument structure construction is thus grammatical, with the more abstract construction supplying the slots for the patient and result-goal, which can be filled creatively independently of the verbal semantics. Similarly, (36b) and (36c) can be argued to obey the Semantic Coherence Principle (identifying the cutter and drinker participant roles as compatible with the agent and the cut.object4 and liquid as compatible with the patient slot) as well as the Correspondence Principle (with the profiled roles cutter and drinker as well as the cut.object realized as core syntactic arguments, that is, subjects and object, respectively). In addition to the Resultative construction, a broad range of other abstract argument structure constructions have been identified in the constructional literature, the most prominent being the following (constructional templates in (37–42), adapted from Goldberg 1995: 117, 142, 160, 208, 2006: 41, 73): (37)
(38)
4
Transitive construction: FORM: SBJ1 V4 OBJ2 ↔ MEANING: ‘Agent1 affects Patient2 by V4-ing’ Examples: She1 kissed him2. He1 sang [a song]2. They1 opened [the box]2. Intransitive Motion construction: FORM: SBJ1 V3 OBL2_path/loc
As mentioned above, however, this identification is less straightforward than in the other cases since the man in (36b) is clearly not the cut.object in this case but merely a patient affected by the action.
299
300
HOFFMANN
(39)
(40)
(41)
↔ MEANING: ‘Theme1 movesby Manner=V_3 GOAL2_path/loc’ Examples: He1 ran [out of the house]2. [The fly]1 buzzed [into the room]2. [People]1 strolled [along the river]2. Caused Motion construction: FORM: SBJ1 V4 OBJ2 OBL3_path/loc ↔ MEANING: ‘Cause1 causes Theme2 to move Goal3_path/loc by V4-ing’ Examples: She1 kissed [her shoes]2 [under the sofa]3. He1 sang [them]2 [out of the room]3. They1 sliced [the box]2 [open]3. Causative with construction: FORM: SBJ1 V4 OBJ2 + with-PPOBL3 ↔ MEANING: ‘Cause1 causes Patient2 to be in state by means of Instrument3_path/loc by V4-ing’ Examples: She1 loaded [the truck]2 [with books]3. He1 sprayed [the walls]2 [with paint]3. They1 heaped [the plate]2 [with mashed potatoes]3. Way construction: FORM: SBJ1 V4 [pro1 way]OBJ2 Obl3 ↔ MEANING: ‘Creator-Theme1 create-moves Createe-Way2 along Path3 by V4-ing’ Examples: She1 belched [her1 way]2 [out of the restaurant]3. He1 bought [his1 way]2 [into the club]3. They1 made [their1 way]2 [into the room]3.
(42)
Ditransitive construction: FORM: SBJ1 V4 OBJ2 OBJ3 ↔ MEANING: ‘Agent1 causes Recipient2 to receive Patient3 by V4-ing’ Examples: She1 sent him2 [an email]3. Jack1 passed her2 [the salt]3. [The waiter]1 served them2 [their dinner]3.
As Boas (2013a: 238) points out, Goldberg’s approach has not only inspired research into English Argument Structure constructions (for a most recent publication, see Perek 2015). There are also a considerable number of constructionist studies that have analyzed Argument Structure constructions in, inter alia, Dutch (Colleman 2009a, 2009b), Finnish (Leino and Östman 2008, Leino 2010), French (Bergen and Plauche´ 2001, Lambrecht and Lemoine 2005, Chenu and Jisa 2006), German (Michaelis and Ruppenhofer 2001, Boas 2003a,
From Constructions to Construction Grammars
Hilpert 2009), Hindi (Budwig, Narasimhan, and Srivastava 2006), Icelandic (Barðdal 2008), Japanese (Fujii 2004), Spanish (Gonza´lvez-Garcia 2010), Swedish (Olofsson 2014), and Thai (Timyam and Bergen 2010). Nevertheless, several assumptions of the above approach to Argument Structure constructions have also been questioned. Croft (2012: 364–74), drawing on work by Nemoto (1998), Boas (2003a), and Iwata (2005), shows that Goldberg’s definition of profiling is not without problems. In Frame Semantic and Cognitive Grammar approaches, where the term profiling originated, “[a] profiled concept is the part of a semantic frame that is denoted by a linguistic expression” (Croft 2012: 364). As pointed out above, however, unlike these purely semantic approaches, Goldberg has additional syntactic criteria for the profiling status of her roles – the syntactic obligatoriness of participant roles and the core argument status of argument roles (Croft 2012: 365). One problem of the criterion of syntactic obligatoriness is that it cannot be decided out of syntactic context, that is, outside of the context of Argument Structure constructions. As Croft (2012: 364–74) shows, this approach runs into significant empirical problems. Take the verb spray, which can, for example, appear in the Caused Motion construction (39) as well as Causative with construction (40): (43) (44)
Caused Motion construction Bob sprayed paint onto the wall. (from Iwata 2005: 387) Causative with construction Bob sprayed the wall with paint. (from Iwata 2005: 387)
In (43) and (44) spray seems to imply the frame semantic participant roles sprayer (Bob), liquid (paint), and target (wall). Yet, which of these participant roles are profiled? According to Goldberg (1995: 178; see also Croft 2012: 367), liquid and target, but not sprayer are profiled participant roles of spray: (45)
a. Waterliquid sprayed onto the lawntarget b. The mensprayer sprayed the lawntarget
(adapted from Croft 2012: 368)
In (45b), only sprayer and target are realized, but Goldberg (1995: 178) argues that the liquid role is still profiled since it is missing but contextually identifiable (i.e. a case of DNI; an analysis Croft 2012: 368 contests). In contrast to this, (45a) without the sprayer role is grammatical and has no DNI reading. Consequently, Goldberg offers the following participant role frame for spray (with angled brackets for profiled roles that can be suppressed via DNI): (46)
(Goldberg 1995: 178)
Yet, Croft (1998: 43) already showed that even the target role can be omitted without giving rise to any DNI reading:
301
302
HOFFMANN
(47)
The broken fire hydrantsprayer sprayed waterliquid all afternoon.
(adapted from Croft 1998: 43)
This kind of reasoning thus leads to a lexical construction spray in which no participant role is profiled and obligatory; if (45b) does not necessarily give raise to a DNI reading, the participant frame for spray would end up as (Croft 2012: 367–68). Even more problematic is the fact that Goldberg’s approach cannot explain why some verb constructions are occasionally incompatible with certain Argument Structure constructions: (48) (49)
The lawnSBJ:LOCATION gleamed with dewwith-PP:LIQUID *The lawnSBJ:LOCATION sprayed with waterwith-PP:LIQUID
The Argument Structure construction underlying (48) and (49) has a subject slot that specifies a location and a with-PP slot for a liquid. According to the Semantic Coherence Principle (cf. Croft 2012: 368), the target participant role of spray should be able to fuse with the location slot, as it does in the Caused Motion construction, cf. (43), and the liquid participant role should perfectly match the constructional liquid argument role. (Additionally, if all participant roles are nonprofiled, as argued above, the Correspondence Principle does not apply.) To explain the difference in grammaticality between (48) and (49) it is obviously not enough to just take into account the number and types of participant roles of a verb. In addition to this, the (dis)ability of verbs to fuse with certain Argument Structure constructions also depends on the verbal semantics, that is, the event type a verb describes (Croft 1998, 2012: 368). Iwata (2005: 387–90) uses the different licit grammatical contexts for smear and spray to illustrate this: (50)
a. They smeared paint on the wall. (Caused Motion construction) b. They smeared the wall with paint. (Causative with construction)
As (51) shows, smear, like spray, occurs in both the Caused Motion and Causative with construction. Furthermore, smear can also be argued to possess a similar set of participant roles (smearer, target, liquid). Yet, while spray can also be used in the Intransitive Motion construction (51a), smear cannot fuse with this Argument Structure construction: (51)
a. Paintliquid sprayed on the walltarget. (Intransitive Motion construction) b. *Paintliquid smeared on the walltarget. (Intransitive Motion construction)
Iwata (2005: 388) points out that the reason for this does not lie in different participant role profiles, but instead is tied to the frame semantic scene encoded by a verb: verbs like spray (or splash) encode a scene in which the manner of movement of the liquid (going in a mist) is construed as a subevent that can be focused on independently of the initial action of the external causer (the sprayer). Consequently, spray-verbs can fuse with
From Constructions to Construction Grammars
the Intransitive Motion construction, which encodes manner of movement of a theme to a location. In contrast to this, the manner of smearing is continuously construed as dependent on a co-occurring action of the external causer (e.g. manual movement). As a result, there is no independent subevent of smear (or daub, etc.) that could fuse with the Intransitive Caused Motion construction. Phenomena such as (51) have lead several Construction Grammarians (e.g. Nemoto 2005, Boas 2005a, 2008b, Iwata 2008, Croft 2012, Boas 2013a: 238) to argue for a greater prominence of the frame semantic meaning of verbs. On top of this, these researchers claim that abstract Argument Structure constructions such as (37–42) often overgenerate and that speakers cognitively seem to have stored more verb-specific Argument Structure templates. Take, for example, the Ditransitive construction: while the template in (42) might seem to license all types of ditransitive structures, it turns out that, depending on certain verb classes, different types of meanings can be expressed (Goldberg 1995: 38, Croft 2012: 375–80; see there for various other subclasses of the construction). (52)
a.
Central Sense: agent successfully causes recipient to receive patient Verbs that inherently signify acts of giving: give, pass, hand, serve, feed, . . . [Modulation: actual] e.g. She passed him the salt. b. Central Sense: agent successfully causes recipient to receive patient Verbs of instantaneous causation of ballistic motion: throw, toss, slap, kick, poke, fling, shoot, . . . [Modulation: actual] e.g. She threw him the ball. c. Agent causes recipient not to receive patient Verbs of refusal: refuse, deny [Modulation: negative] e.g. She denied him the answer.
As (52) shows, the three verb classes lead to crucially different meanings: while the verbs in (52a) and (52b) entail that the possessive relation of the Ditransitive construction is actually taking place, the verbs in (52c) entail a negative modulation (the recipient does not receive the theme). On top of that, verbs such as throw, toss, slap, etc. (52b) lead to an instantaneous ballistic motion reading of the Ditransitive construction that is absent from similar constructions with verbs such as give, pass, hand, etc. (52a). In order to address this issue, Goldberg (1995: 38) advocates a polysemy account in which the construction in (42) is enriched by various submeanings (such as whether the transfer of possession is actual (52a, b) or negative (52c); see also Croft 2012: 376). One problem with this approach is that it wrongly predicts that verbs of one class should also be used with any of the other polysemous meanings. Yet this is not the case: pass in the Ditransitive construction cannot mean ‘X passes Y to receive Z by instantaneous ballistic motion’ nor can refuse be made to mean ‘X refuses Y to receive Z (and Y successfully receives Z)’ (Croft 2012: 377). Due to this,
303
304
HOFFMANN
instead of postulating a single, maximally abstract, polysemous Ditransitive construction, Croft (2012: 378) argues that a family of related Verb-Class-Specific constructions have to be postulated. The following three Verb-Class-Specific constructions could then be said to license the distribution in (52): (53)
Verb-Class-Specific Ditransitive constructions (adapted from Croft 2012: 378) a. FORM: SBJ1 GIVING.V4 OBJ2 OBJ3 ↔ MEANING: ‘actual transfer of possession: Agent1 causes Recipient2 to receive Patient3 by V4-ing’ b. FORM: SBJ1 BALL.MOT.V4 OBJ2 OBJ3 ↔ MEANING: ‘actual transfer of possession via ballistic motion: Agent1 causes Recipient2 to receive Patient3 by V4-ing’ c. FORM: SBJ1 REFUSE.V4 OBJ2 OBJ3 ↔ MEANING: ‘negative transfer of possession: Agent1 causes Recipient2 not to receive Patient3 by V4-ing’
Croft (2012: 380) highlights that the three types of Verb-Class-Specific Ditransitive constructions interact differently with the intuitive semantics of the various verb classes: while the Ditransitive construction in (53a) adds no meaning to verbs such as give, hand, serve, etc., the one in (53b) adds a modulation of actual transfer of possession to verbs that themselves do not encode this meaning (cf. He threw a stone., e.g., implies no transfer of possession at all). Finally, verbs such as refuse or deny already imply a negative modulation, but no transfer of possession (cf. He refused to do it. → He did not do it.), so that (53c) only adds the latter meaning component to the basic verbal semantics. There is, however, evidence that suggests that even Verb-Class-Specific Ditransitive constructions overgenerate in some cases. Goldberg (1995: 130; see also Croft 2012: 378), for example, notes that not all refusal verbs can fuse with the (REFUSE.V-) Ditransitive construction (53c): (54)
She refused/denied/*prevented/*disallowed/*forbade him a kiss.
In addition to Verb-Class-Specific Ditransitive constructions, this appears to call for even more specific Verb-Specific constructions such as the following (adapted from Croft 2012: 379): (55)
a. FORM: SBJ1 REFUSEV4 OBJ2 OBJ3 ↔ MEANING: ‘negative transfer of possession: Agent1 causes Recipient2 not to receive Patient3 by refuse4-ing’
From Constructions to Construction Grammars
b. FORM: SBJ1 DENYV4 OBJ2 OBJ3 ↔ MEANING: ‘negative transfer of possession: Agent1 causes Recipient2 not to receive Patient3 by deny4-ing’
The right level of granularity of constructional schematicity remains a central issue in all Construction Grammars (and is a point that we will return to in the next chapter). As previous research has show, however, not all Argument Structure Constructions exhibit the same degree of productivity: while the Way construction (41), e.g., can fuse with a considerable number of different verbs (Goldberg 1995: 199–218), the Resultative construction (33) is lexically more constrained and thus much less productive (Goldberg and Jackendoff 2004, Boas 2005a, 2011a, Boas 2013a: 237–38). An important insight of Construction Grammar research thus is that constructional generalizations can occur at varying degrees of schematicity (Hilpert 2013a: 201–03) – even for a set of similar constructions such as the Argument Structure constructions in (33, 37–42). Mu¨ller (2006) raises another issue concerning the status of phrasal constructions such as (33, 37–42). All of these constructions seem to specify a fixed word order, the Resultative construction (33), for example, has the SBJ1↔Agent1 slot preceding the verbal slot, while OBL3↔Result-Goal3 follows OBJ2↔Patient2 in post-verbal position. Since English has a fairly fixed word order, this mapping of hierarchical structure into linear order is unproblematic for most declarative clauses. In German, however, a language with a freer word order, many different orderings of the constituents of the Resultative construction can be found: (56)
(57)
Er fischt den Teich schnell leer. he fishes the pond quickly empty. b. Den Teich fischt er schnell leer. the pond fishes he quickly empty. c. Schnell fischt er den Teich leer. quickly fishes he the pond empty fischt er den Teich nicht. d. Leer empty fishes he the pond not. (examples a–d from Mu¨ller 2006: 853) . . . (weil) er den Teich leer fischt. . . . (because) he the pond empty fishes.
a.
The meaning of the main clause examples (56a–d) as well as the subordinate clause (57) all roughly translate into ‘he quickly fishes the pond empty’ in English. Yet, while English has subject-verb order in main and subordinate clauses, German has verb second order in main clauses (56) and verb-final order in subordinate clauses (57). Now one way to account for this is to assume two sets of Resultative constructions: a main clause
305
306
HOFFMANN
variant with the verb slot in second position and a subordinate variant with verb in the final slot. Such an approach, however, would still have to explain that the verb order variation in (56) and (57) is not specific to Resultative constructions, but applies to all German Argument Structures constructions (and thus constitutes a structural generalization that speakers entrench fairly quickly). Alternatively, one could argue that constructional templates such as (33) only specify syntactic dominance but do not determine the linear order of constituents (see Mu¨ller 2006: 862). Then, the various structures in (56), for example, could all be explained by the interaction of the Resultative construction, which would not specify the linear precedence information, and, for example, an information structure construction that would specify that the initial constituent preceding the verb in (56) receives a topic interpretation. The latter explanation is feasible for the examples in (56), which do indeed differ with respect to their topic-comment structure (and which can therefore not be considered structural alternatives that are fully synonymous). More problematic, even for languages with a fixed word order such as English, however, are valency-changing operations such as passives: (58)
The man was cut free (by the firefighters).
In the passive sentence (58), the argument linking of argument roles and syntactic function is clearly different from the Resultative construction template in (33): in (58) the patient is realized as subject, while the agent, as expected in a passive sentence, is demoted to an optional oblique position (the complement of the optional by-PP). All Construction Grammar approaches will, of course, assume a Passive construction. Examples such as (58), however, beg the question of whether it makes sense to first fuse verbs with an Argument Structure construction such as (33) and then to assume the Passive construction to change the valency as well as word order of this template (since this has a transformational flavour which is strongly eschewed in Construction Grammar approaches; see the next chapter). In essence, three alternative types of analysis for this problem have been advocated in the constructionist literature: first of all, early constructionist approaches (Fillmore and Kay 1995, Michaelis and Ruppenhofer 2001: 55–57) treated Argument Structure constructions as well as the Passive construction as linking constructions that simultaneously were to unify with lexical constructions (such as the cut Verb construction). These analyses, however, were technically flawed in that the valence lists of the various constructions were represented as sets. Since set unification could not be defined in a way that allowed for the automatic unification of the various input constructions, these analyses were soon abandoned (see Mu¨ller 2006: 863–66, 2015: 309–14 for a more detailed discussion).
From Constructions to Construction Grammars
The second approach emerged from formal syntactic theories such as Head-Driven Phrase Structure (HPSG) and Sign-Based Construction Grammar (SBCG). Researchers within these approaches (inter alia, Mu¨ller 2006, Sag 2012) argue for a lexical construction account. In their analysis, the Passive construction and various Argument Structure constructions such as the Resultative construction (as well as various other constructions) are unary lexical constructions that take a lexical verb construction as their daughter node and link it to a mother node with the appropriate argument structure. A simplified version of a Lexical Resultative construction is provided in (59) (based on Mu¨ller 2006: 873 and reinterpreted as a feature description following Meurers 2001, Mu¨ller 2015: 281): (59)
Lexical Resultative construction (simplified and adopted version of Mu¨ller 2006: 873) (resultative) MTR [FORM: V SUBCAT ↔ MEANING X1 CAUSE Y2 to become Z4 by V2-ing] | (verbal) DTR [FORM: V SUBCAT ↔ MEANING: V2]
Such analyses are the constructionist equivalent of lexical rules: an input lexical item (the verbal daughter node) is related to an output lexical item (the Resultative mother node). Once the lexical construction in (59) has licensed a resultative verb (e.g. cut SUBCAT , this output construction can become the input of a lexical Passive construction that yields the required passivized verb form (e.g. cut SUBCAT . Only once all lexical constructions have applied, syntactic constructions can then create phrasal patterns. Lexical constructionist analyses have the advantage of significantly reducing the number of postulated constructions and of directly capturing generalizations such as the passive formation across all types of Argument Structure constructions. At the same time, they re-introduce a lexicon–syntax dichotomy, a distinction that has been refuted by much Construction Grammar and cognitive linguistic research. As a case in point, take the Ditransitive construction above: the data in (52)–(55) showed that a single abstract Ditransitive construction can not capture the distribution of the various semantic subtypes. Similarly, one would need not only one but (at least) two separate unary lexical constructions for verbs of instantaneous causation of ballistic motion and negative transfer of possession, respectively. This, of course, can easily be done and is not a substantial argument against lexical accounts. More
307
308
HOFFMANN
problematic, however, are idiosyncratic constraints on the interaction of Argument Structure constructions and the Passive construction: (60) (61)
He was given/passed/handed/served/fed the soup. a. He was thrown the ball. b. ?He was kicked the ball.
While a passive version of all verbs that inherently signify acts of giving in (60) are perfectly acceptable, not all verbs of instantaneous causation of ballistic motion seem to be able to appear in passive ditransitive structures; cf. (61a versus 61b).5 If the passive is a lexical construction independent of Argument Structure constructions, this type of restriction cannot be accounted for in a straightforward way. More generally speaking, a problem of lexical construction accounts is that they are too powerful and tend to overgenerate (Carpenter 1992; cf. also van Trijp 2014: 618–29). Besides, unary lexical constructions of the type given in (59) require that a mother node (output) always has a lexical daughter node (input). Yet, there are some verbs that only appear in passive sentences that lack an active equivalent/input (Koenig 1999: 37, Huddleston and Pullum 2002: 1435, Hilpert 2014: 42): (62)
a. Pat is reputed to be very rich. b. Kim is said to be a manic depressive. c. It is rumoured that there will be an election before the end of the year.
(examples from Hilpert 2014: 42)
Data such as (61) and (62) have lead to a third approach (Croft 2001: 215, 2013: 215–16, 220, Goldberg 2006: 22, fn. 3, van Trijp 2014: 624) which postulates active as well as passive Argument Structure constructions as separate, albeit related constructions. In addition to the Active Ditransitive construction (42), speakers should therefore also possess a Passive Ditransitive construction: (63)
Passive Ditransitive construction: FORM: SBJ2 BE V-en4 OBJ3 (by NP1) ↔ MEANING: ‘Agent1 causes Recipient2 to receive Patient3 by V4-ing’
As before, the data discussed above probably requires the postulation of additional, more specific passive constructions for the various verbs classes and Argument Structure constructions. Whether this is cognitively plausible remains an area of active research. One advantage of the phrasal analysis is that, in addition to the many systematic correspondences between the two types of constructions, such an approach also allows active and passive versions of Argument Structure constructions to exhibit
5
While a Google search for ‘was thrown the ball’ site:.uk yields 8,210 hits, the equivalent query ‘was kicked the ball’ site:. uk gives no single hit, www.google.de [last accessed February 14, 2016].
From Constructions to Construction Grammars
individual idiosyncratic properties and does not require one structure to be the input for the other (Hilpert 2014: 42).
18.3 Conclusion The present chapter backtracked the renaissance of constructions in (cognitively oriented) linguistics. From the mid-1980s onwards, more and more evidence accumulated that constructions are not only needed to model the behavior of ‘peripheral’ linguistic items (such as idioms), but also that in fact all levels of grammatical description involve such conventionalized form-meaning pairings (from morphemes to abstract syntactic clausal patterns). After this historical overview of the renaissance of constructions in linguistics, the next chapter will take a closer look at the most important Construction Grammar approaches that have emerged in recent years.
309
19 Construction Grammars Thomas Hoffmann
19.1 Introduction The previous chapter outlines the renaissance of constructions in linguistic research since the mid-1980s. This line of research culminated in a school of linguistic approaches that are now known as Construction Grammar. The main tenet of all Construction Grammar approaches is that form-meaning pairings, that is, constructions, are the fundamental units of the human language capacity and that all levels of syntactic description, from morphemes to abstract sentence patterns, can best be explained as constructional templates. In the present chapter, I will first give an overview of the other main assumptions that are shared by all major constructionist approaches to language. After that, I will outline the major differences between non-usage-based (such as Berkley Construction Grammar and Sign-Based Construction Grammar) and usage-based approaches (Cognitive Construction Grammar, Embodied Construction Grammar, Fluid Construction Grammar, and Radical Construction Grammar). Moreover, I will discuss the controversial issue of what counts as a construction (from Kay’s conservative competence-based notion to the usage-based interpretation of constructions as exemplar-based clouds), as well as the ontological status of meaningless constructions. In addition to that, I will also touch upon the nature of the structured inventory of constructions, the constructicon, and explore the advantage and limits of constructional inheritance in taxonomic networks. Finally, the chapter will also address the question as to how the meaning pole of constructions is analysed in the various approaches, which range from semantic paraphrases (Cognitive Construction Grammar) over first-order predicate logic (Fluid Construction Grammar) to Frame-based approaches (Sign-Based Construction Grammar).
Construction Grammars
19.2 Basic Assumptions of Construction Grammar Approaches Most of the research surveyed in the previous chapter has converged on a view of language that no longer maintains an artificial distinction between the lexicon and syntax. Instead, virtually all Constructionist Grammar approaches provide a uniform analysis of more idiosyncratic ‘peripheral’ as well as ‘core’ linguistic features as form-meaning pairings. They achieve this without recourse to transformations/derivations or the employment of empty elements. Instead, the mental grammar of speakers is claimed to consist of a network of schematic and substantive constructions (‘constructicon’) and it is the parallel activation of constructions that underlies a set of particular utterances (‘constructs’). In the last thirty years, this view of grammar as a mental network of constructions has received great empirical support by independent research on: • first language acquisition (inter alia Clark 1987, Tomasello and Brooks 1998, Brooks et al. 1999, Brooks and Tomasello 1999a, 1999b, Tomasello 1999, Da˛browska 2000, Rowland and Pine 2000, Tomasello 2003, Diessel 2004, Da˛browska and Lieven 2005, Rowland 2007, Diessel 2008, 2009, Da˛browska, Rowland, and Theakston 2009, Diessel 2013; see also De Ruiter and Theakston this volume Ch. 4), • second language acquisition (inter alia Ellis 2002, 2003, Gries and Wulff 2005, Ellis 2006b, McDonough 2006, McDonough and Mackey 2006, McDonough and Trofimovich 2008, Gries and Wulff 2009, Wulff et al. 2009, Ellis 2013; see also Tyler this volume Ch. 5), • psycholinguistics (inter alia Bencini and Goldberg 2000, Chang, F. et al. 2000, Chang F. 2002, Chang, F., Bock and Goldberg 2003, Dominey and Hoen 2006, Bencini and Valian 2008, Konopka and Bock 2008, Wardlow Lane and Ferreira 2010, Bencini 2013), • neurolinguistics (Pulvermu¨ller 1993, 2003, Pulvermu¨ller and Knoblauch 2009, Cappelle, Shtyrov, and Pulvermu¨ller 2010, Pulvermu¨ller 2010, Pulvermu¨ller, Shtyrov, and Cappelle 2013; see also Coulson this volume Ch. 32). Moreover, constructional approaches have also provided important new insight into the following fields: • the diachronic evolution of languages (see Bybee 2003, Himmelmann 2004, Bybee 2006, Noe¨l 2007a, 2007b, Bergs and Diewald 2008, Gries and Hilpert 2008, Hilpert 2008, Noe¨l 2008, Petre¨ and Cuyckens 2008, Traugott 2008a, 2008b, 2008c, Trousdale 2008a, 2008b, Trousdale and Gisborne 2008, Bergs and Diewald 2009, Noe¨l and Colleman 2009, Gries and Hilpert 2010, Patten 2010, Trousdale 2010, Barðdal and Eythðrsson 2012, Trousdale 2012, Barðdal 2013, Fried 2013, Hilpert 2013a, 2013b,
311
312
HOFFMANN
Trousdale and Traugott 2013, Petre¨ 2014; see also Bergs this volume Ch. 22), • cognitive sociolinguistics (Grondelaers 2000, Hollmann and Siewierska 2006, Grondelaers, Speelman, and Geeraerts 2007, Hollmann and Siewierska 2007, Grondelaers, Speelman, and Geeraerts 2008, Croft 2009b, 2010, Hollmann 2013; see also Hollmann this volume Ch. 33), and ¨ stman 2002, Leino and O ¨ stman • dialectal and discourse variation (see O 2005, Gu¨nthner and Imo 2007, Linell 2009a, Mukherjee and Gries 2009, Wide 2009, Hollmann and Siewierska 2011, Hoffmann 2011, Hoffmann ¨ stman and and Trousdale 2011, De Clerck and Colleman 2013, O Trousdale 2013; see also Tenbrink this volume Ch. 41). While the various constructionist approaches differ in non-trivial ways (see section 19.3 below), Goldberg (2013) identifies four basic assumptions that are shared by all Construction Grammar approaches: the lexiconsyntax continuum (section 19.2.1), taxonomic network organization and inheritance (section 19.2.2), surface structure orientation (section 19.2.3), cross-linguistic variability and generalization (section 19.2.4).
19.2.1 The Lexicon–syntax Continuum Most other syntactic theories assume a strict division between the lexicon as a repository of meaningful words and morphemes on the one hand and meaningless syntactic rules which combine these words into sentences (cf. e.g. Mainstream Generative Grammar) on the other. As pointed out above, however, Construction Grammarians do not uphold this strict lexicon–syntax distinction. Instead, all levels of grammatical knowledge involve form-meaning pairs, that is, constructions. The only difference between lexical constructions (such as the Apple-construction: FORM: apple [æpl] ↔ MEANING: ‘apple’) and phrasal/grammatical constructions (such as the Resultative construction: FORM: [SBJ1 V4 OBJ2 OBL3] ↔ MEANING: ‘Agent1 causes Patient2 to become Result-Goal3 by V4-ing’) is the degree of schematicity: while the former are fully substantive (have their phonological form filled), the latter are schematic (and thus contain slots that can be filled by various lexical constructions). Moreover, Construction Grammarians point out that the grammatical knowledge of a speaker does not only consist of these two extreme types of constructions. Instead, fully substantive and fully schematic constructions only lie at the opposite ends of a cline (cf. the previous chapter). In between, we find constructions that have both substantive and schematic parts, such as the Un-construction: FORM: [[ʌn]-X] ↔ MEANING: ‘not X’ or the X TAKE Y for granted-construction (FORM: [X TAKE Y fə ɡɹɑːntɪd] ↔ MEANING: ‘X doesn’t value Y enough’). A central claim of the constructionist approach is therefore that “all grammatical knowledge in the speaker’s mind [is stored], in
Construction Grammars
the form of . . . constructions” (Croft and Cruise 2004: 255). The full list of constructions that make up a speaker’s mental grammatical knowledge is then referred to as the ‘constructicon’ (in analogy to the lexicon, which in other theories only comprises words and morphemes: Fillmore 1988, Jurafsky 1992) (see Boas this volume Ch. 34).
19.2.2 Taxonomic Network Organization and Inheritance The constructicon is not seen as an unstructured list of constructions. Instead, all versions of construction grammars agree that the constructions of a language form a structured inventory, which can be represented by (taxonomic) networks (cf. Croft and Cruse 2004: 262–65). Figure 19.1 gives a partial representation of the various Ditransitive constructions (see Hoffmann this volume Ch. 18): At the bottom of Figure 19.1, we find specific utterances (so-called constructs) such as She refused him a kiss. While Construction Grammar approaches differ as to whether they consider constructs to be stored by speakers (see below), the other taxonomic levels are motivated by the empirical facts outlined in section 19.2 and thus need to be captured by all Construction Grammar approaches. Constructional networks are similar to prototype taxonomies (Rosch 1973, Ungerer and Schmid 2006: 7–63) in that there are basic-level items such as chair (the Verb-Class Specific Ditransitive constructions in Figure 19.1) that are situated between more general, superordinate terms such as furniture (the Ditransitive construction in Figure 19.1) and the more specific subordinate kitchen chair (the REFUSE.Verb-Specific constructions in Figure 19.1). Just like prototype taxonomies, constructional networks exhibit ‘default inheritance’ effects (cf., e.g., Croft and Cruse 2004: 262–65, Ginzburg and Sag 2000: 5–8, Goldberg 2003: 222–23): normally (by default) a subordinate construction will inherit all properties from its superordinate construction, but a more specific construction can also override inherited properties. All subordinate constructions of the most schematic Ditransitive construction, for example, inherit its transfer of possession meaning, but the REFUSE.VerbClass override the default modulation of actual transfer with their negative modulation. Cognitively, constructional taxonomies are also motivated by priming effects, of constructions with a similar form and/or meaning (cf., e.g., Bock 1986a, 1986b, Bock and Griffin 2000, Hudson 2010: 75–76; though as Croft 2013: 218–23 points out alternative visualizations such as geometric or network representations can also be used to capture these relationships between constructions). While taxonomic construction networks such as Figure 19.1 play a crucial role in all Construction Grammar approaches,1 an important 1
In Fluid Construction Grammar inheritance is not a default way of capturing constructional relationships, but implementations for it have been discussed in the literature (Bleys, Stadler, and de Beule 2011: 158–59).
313
Ditransitive cxn FORM: SBJ1 V4 OBJ2 OBJ3 ↔ MEANING: ‘Agent1 causes Recipient2 to receive Patient3 by V4-ing’
GIVING.Verb-Class-Specific Ditransitive cxn FORM: SBJ1 GIVING.V4 OBJ2 OBJ3 ↔ MEANING: ‘actual transfer of possession: Agent1 causes Recipient2 to receive Patient3 by V4-ing’
REFUSE. Verb-Specific constructions FORM: SBJ1 REFUSEV4 OBJ2 OBJ3 ↔ MEANING: ‘negative transfer of possession: Agent1 causes Recipient2 not to receive Patient3 by refuse4-ing’
She refused him a kiss.
REFUSE.Verb-Class-Specific Ditransitive cxn FORM: SBJ1 REFUSE.V4 OBJ2 OBJ3 ↔ MEANING: ‘negative transfer of possession: Agent1 causes Recipient2 not to receive Patient3 by V4-ing’
BALL.MOT.Verb-Class-Specific Ditransitive cxn FORM: SBJ1 BALL.MOT.V4 OBJ2 OBJ3 ↔ MEANING: ‘actual transfer of possession via ballistic motion: Agent1 causes Recipient2 to receive Patient3 by V4-ing’
They refused her the answer
Figure 19.1 Partial taxonomic construction network for the Ditransitve construction family
DENY.Verb-Specific constructions FORM: SBJ1 DENYV4 OBJ2 OBJ3 ↔ MEANING: ‘negative transfer of possession: Agent1 causes Recipient2 not to receive Patient3 by deny4-ing’
…
Construction Grammars
point of controversy concerns the role of frequency in the emergence of such taxonomies (cf. Croft and Cruse 2004: 276–78): on the one hand, “complete inheritance (constructional) models”2 (such as Sign-Based Construction Grammar or Berkeley Construction Grammar) assume that only “idiosyncratic morphological, syntactic, lexical, semantic, pragmatic or discourse-functional properties must be represented as an independent node in the constructional network in order to capture a speaker’s knowledge of their language” (Croft and Cruse 2004: 263). In contrast to this, usage-based (Kemmer and Barlow 2000, Bybee 2013) constructional approaches – including Cognitive Grammar (Langacker 1987, 2005a, Broccias 2013), Radical Construction Grammar (Croft 2001, 2013), and Cognitive Construction Grammar (e.g. Lakoff 1987, Goldberg 2003, 2006, Boas 2013a) – advocate that the frequent use of a construction can lead to it being cognitively entrenched, even if its properties can be completely derived compositionally by more abstract superordinate constructions (Goldberg 2006). In usage-based approaches the storage, that is, the mental entrenchment, of a construction depends on two types of input frequency effects (Hoffmann 2013a: 314–17): high token frequency leads to the entrenchment of phonologically filled constructions (Langacker 1987: 59–60, Croft and Cruse 2004: 292–93). Thus, even a construct such as She refused him the kiss can be become entrenched if it is encountered often enough. High type frequency of a pattern, that is, different lexicalizations such as She refused him the kiss, They refused her the answer, or Leeds United have refused Sky TV entry into Elland Road3 instead, due to the domain general capacity of pattern detection and schematization (Bybee 2010), lead to the storage of a more abstract construction (such as the REFUSE.Verb-Specific Ditransitive construction). Under this view, constructional taxonomies thus emerge in a bottom-up fashion. The resulting levels of schematicity are sometimes referred to using a classification first introduced by Traugott (2008a, 2008b; see also Bergs 2008, Trousdale 2008c, Traugott and Trousdale 2010, Fried 2008): entrenched, substantive constructions, which appear at the bottom of the mental constructicon (such as She refused him the kiss), are called ‘micro-constructions.’ A high type frequency of micro-constructions can then lead to the generalization of more abstract ‘meso-constructions’ such as the REFUSE.Verb-Specific Ditransitive construction (or at an even more abstract level, the REFUSE.Verb-Class Ditransitive construction, since there can be more than one level of meso-constructions). Finally, the most schematic constructions such as the Ditransitive construction at the top of Figure 19.1 are labeled ‘macro-constructions.’ Since most constructs could also be generated by drawing on existing micro- and meso-constructions, 2
Earlier complete inheritance models advocated monotonic inheritance relationships among constructions. Currently, however, all constructionist approaches assume default inheritance (Goldberg 2006: 215).
3
Source http://www.leeds.vitalfootball.co.uk/article.asp?a=427875 (last accessed February 15, 2016).
315
316
HOFFMANN
Ditransitive cxn Active Ditra
I cxn
Past cxn
whatsapp cxn her cxn the cxn pic cxn … I whatsapped her the pic Figure 19.2 Selected multiple inheritance relationships in (1)
for many phenomena the question of whether a macro-construction has to be postulated at all is an empirical one from a usage-based perspective. Barðdal (2008, 2011), for example, has shown that the productivity of abstract constructions can be seen as an inverse correlation of type frequency and semantic coherence, with highly abstract macro-constructions only arising if the underlying meso-constructions have a high type frequency and a high degree of variance in semantic distribution. The majority of Construction Grammar approaches assume that constructional taxonomies are multiple inheritance networks (Lakoff 1987, Langacker 1987, Fillmore Kay, and O’Connor 1988, Wierzbicka 1988, Hudson 1990, 2007, Goldberg 1995, Booij 2010), which allow for a specific construction to inherit from more than one superordinate construction. As Figure 19.2 shows, a construct such as (1) is, for example, supposed to inherit from the lexical constructions involved as well as the involved Argument Structure constructions and Tense-Aspect constructions: (1)
I whatsapped her the pic whilst talking to her. http://community.babycentre.co.uk/post/a28710529/long_distance_ announcement (last accessed February 15, 2016)
Figure 19.2 is, obviously, a highly simplified representation of the constructional interaction underlying the example in (1). For (1), for example, the exact formalization of the Tense and Aspect constructions as well as the precise nature of how (1) inherits from these constructions is much more complicated than implied by Figure 19.2. The details of this are beyond the scope of the present chapter and readers are referred to Croft (2012) for the most in-depth constructionist analysis of tense and aspect as well as their interaction with the semantics of verb constructions. What Figure 19.2 illustrates, however, is the idea that inheritance networks allow actual expressions (so-called ‘constructs’4) to be freely formed as long as the constructions they consist of do not conflict (cf. Goldberg 2006: 22). Thus, Construction Grammars can also account for the creative use of language (Chomsky 1957, 1965, Goldberg 2006: 22). 4
NB: in Sign-Based Construction Grammar, SBCG (Michaelis 2010, 2013; Boas and Sag 2012), the term ‘constructs’ is used in a slightly different way as the name of a special type of constraint (namely, a type constraint on local trees). In almost all other constructionist publications, however, people use the term in the sense of an actual token of linguistic performance.
Construction Grammars
There are, however also cases that cannot easily be explained by inheritance and which therefore require an alternative constructional analysis (Mu¨ller 2006, 2015, Sag, Boas, and Kay 2012: 9–14). Take, for example, the great-Nkinship construction (Kay 1973): (2)
(3)
(4)
a. great grandfather b. great great grandfather c. great great great grandfather d. great great great great grandfather . . .. a. great grandmother-in-law b. great great grandmother-in-law c. great great great grandmother-in-law d. great great great great grandmother-in law . . . great-Nkinship FORM: [[ɡɹeɪt]1-Nkinship2]N3 ↔ MEANING: ‘Akinship2 at one further generation of ancestry1’N
As (2) and (3) show, by adding great to kinship terms like grandfather or mother-in-law it is possible to refer to one further generation of ancestry (so one’s great grandmother is one of your parents’ grandmothers; one’s great great grandmother is one of your parents’ great grandmothers, etc.). Intuitively, these data might lead one to postulate an abstract great-Nkinship construction such as (4): on the FORM level the construction has a substantive element [ɡɹeɪt]1 which has a MEANING of ‘at one further generation of ancestry1.’ On top of that, there is a schematic Nkinship2 slot that provides the generation and kinship information that the full construction shifts back by a generation. Using constructional inheritance one can then account for words such as great grandfather (2a), which can be said to inherit from the lexical construction grandfather as well as the great-Nkinship construction (4). Inheritance accounts, however, fail for recursive patterns such as great great grandfather (2b), great great great grandfather (2c, etc.). Inheritance is an ‘all-at-once’ phenomenon (see Figure 19.2), that is, all input constructions from which a specific construct inherits apply simultaneously. If only two input constructions are available, that is, the great-Nkinship construction (4) and an Nkinship construction (such as grandfather), inheritance is only going to license the construct great grandfather (and multiple inheritance from the same input construction does not give a different result; Mu¨ller 2015: 312–14). What is needed is a constructional analysis that allows for the construct great grandfather to become the input of the great-Nkinship construction (4) (to yield great great grandfather) and for this procedure to apply recursively, to produce, for example, (2c, d). One way to tackle this is to postulate a lexical construction that allows the output kinship term great grandfather to become itself the input to (5) that then yields great great grandfather (see Sag, Boas, and Kay 2012: 9–14):
317
HOFFMANN
(5)
Lexical great-Nkinship construction (simplified version) MTR r
[FORM: [[g eΙt]1-Nkinship2]N3 ↔ MEANING: ‘Akinship2 at one further generation of ancestry1’N] | DTR [FORM: Nkinship2 ↔ MEANING: ‘Akinship2’N]
As mentioned in Chapter 18, however, one disadvantage of lexical constructions is that they reintroduce a type of lexicon–syntax distinction into Construction Grammar and that they also clearly distinguish an input and an output construction. While recursive phenomena such as (2) and (3) can clearly not be captured by vertical taxonomic inheritance relationships, it is also not the case that unary lexical constructions are the only solution for such data. An alternative that is, in accordance with usage-based approaches, ‘output-based’ or ‘product-oriented’ (cf. Croft and Cruse 2004: 300–02) draws on bidirectional horizontal construction relationships, similar to morphological correspondences in word-based approaches to morphology (Haspelmath and Sims 2010: 47; see also Booij 2013): (6)
Nkinship construction
Great-Nkinship construction r
318
FORM: Nkinship2
FORM: [[g eΙt]1-Nkinship2]N3
↔
↔
MEANING:
MEANING: ‘Akinship2 at one further generation of
‘Akinship2’N
ancestry1’N
In (6), the Nkinship construction and Great-Nkinship construction are two independent constructions that are linked by a bidirectional correspondence link. This horizontal link enables an ‘output’ construction (e.g. great grandfather) to become associated as an instance of the Nkinship construction (and thus, in turn, to turn into the ‘input’ for a further recursive application of the Great-Nkinship construction). More importantly, the bidirectional link also allows speakers to derive an ‘input’ construction from an observed ‘output’ construction: if one encounters the construct great great great great grandfather, the horizontal link in (6) allows one to entrench the great great great grandfather construction, even if one did not have a stored template for this ‘input’ to start with.
Construction Grammars
Constructions such as Great-Nkinship construction might seem marginal. However, horizontal constructional links, such as the one given in (6), are also needed to model derivational morphology as well as, for example, the active–passive correspondence. Besides, whereas unary lexical constructions are input-dependent, bidirectional constructional correspondences can also capture data for which no input construction can be found, as in examples (7) and (8) from Haspelmath and Sims 2010: 49): (7)
(8)
V N ADJ attract attraction attractive suggest suggestion suggestive prohibit prohibition prohibitive V N ADJ – illusion illusive – aggression aggressive
The data in (7) seem to call for a constructional analysis in which the verb acts as the input to two unary lexical constructions, one that creates nouns (attract+ion, suggest+ion and prohibit+ion) and one that licenses adjective (attract+ive, suggest+ive and prohibit+ive). Yet, such an analysis fails for the data in (8) for which only the ‘output’ noun and adjective forms exist, not ‘input’ verb forms. In a bidirectional constructional correspondence approach, these data receive a straightforward analysis: there are three schemas (a verb construction, a noun construction, and an adjective construction), which are all linked via horizontal correspondence rules. Most micro-constructions of one of these schemas will have corresponding micro-constructions in the other schemas (7). Yet, since the schemas are independent constructions, they can also have micro-constructions that do not correspond to a micro-construction in one of the schemas (8). Returning to the vertical network, Goldberg (1995: 75–81; cf. also Boas 2013a: 245–45) distinguishes the following four types of taxonomic links: 1. polysemy links: as mentioned in section 18.2, Goldberg analyses ditransitives as a polysemous construction. Under this view, the GIVE.VerbClass-Specific Ditransitive construction in, example (53a) in the previous chapter, represents the central meaning and others such as the BALL.MOT.Verb-Class-Specific construction, example (53b) in the previous chapter, or REFUSE.Verb-Class-Specific construction, example (53c) in the previous chapter, are extensions of the central meaning and linked to the GIVE.Verb-Class-Specific Ditransitive construction via so-called polysemy links. Now it is uncontroversial that all of the various Ditransitve constructions share a great number of form as well as meaning characteristics in a family resemblance fashion. On top of that, during language acquisition GIVE.Verb-Specific constructions have the highest type frequency and thus furthermore lead to the entrenchment of the GIVE.Verb-Class-Specific Ditransitive
319
320
HOFFMANN
construction as the most prototypical construction in the network (Goldberg 2006: 75–79). In light of the arguments presented in the previous chapter (cf. also Croft 2012: 375–83), however, it remains questionable whether polysemy links are needed for the many VerbClass-Specific and Verb-Specific constructions required to adequately model the English ditransitive network. 2. subpart links: these links are postulated for constructions that exist independently as well as a subpart of a more complex construction. As an example, Goldberg (1995: 78–79) mentions the Intransitive Motion construction, example (38) in the previous chapter, whose form and meaning are also part of the Caused Motion construction, example (39) in the previous chapter. 3. instance links: these types of links are supposed to exist between more schematic constructions and more substantive constructions that are supposed to instantiate them (Goldberg 1995: 79–81). The taxonomic link between the REFUSE.Verb-Class-Specific Ditransitive and REFUSE. Verb-Specific Ditransitive construction in Figure 19.2, for example, is an instance link (in fact, all subordinate constructions are instance links of their superordinate constructions in Figure 19.2). 4. metaphor links: finally, Goldberg (1995: 81–89) argues that taxonomic links also exist between a construction and another one that is said to be a metaphorical extension of the former. A metaphor link is established between the two constructions since the Resultative construction, example (33) in the previous chapter, for example, They kicked him black and blue, can be argued to be a metaphorical extension of the Caused Motion construction, example (39) in the previous chapter, for example, They kicked the ball into the goal, in which the source domain meanings Motion and Location are mapped onto the target domain meanings Change and State. It is still an open question whether all of the above vertical taxonomic links are actually established in a speaker’s mental constructicon. However, the general need for vertical as well as horizontal network links in the mental constructional network can be considered a central tenet of mainstream Construction Grammar.
19.2.3 Surface Structure-orientation Mainstream Generative Grammar (Chomsky 1995, 2000a) advocates a view of language in which phonology, syntax, and semantics are three independent modules. In this approach, abstract syntactic representations are manipulated independently by transformations or derivations and only once these syntactic operations are completed is the semantics of the output computed. This is in stark contrast to Construction Grammar approaches, which – as is seen above – treat linguistic elements and
Construction Grammars
structures as holistic pairings of phonological/syntactic form and semantic/pragmatic meaning. A result of this holistic view is that Construction Grammar places much more emphasis on surface structure, that is, the concrete utterances that a hearer is exposed to, than Mainstream Generative Grammar does (which is much more interested in invisible syntactic operations that cannot be directly observed in the output; see Goldberg 2002, Culicover and Jackendoff 2005, Goldberg 2013: 20).
19.2.4 Cross-linguistic Variability and Generalization Since constructions are arbitrary pairings of form and meaning, it should not be surprising to find that phrasal and clausal constructions – just like word constructions – vary greatly across languages. At the same time, this does not mean that Construction Grammarians deny that a great number of cross-linguistic generalizations exist. Yet, unlike other approaches, Construction Grammar claims that these generalizations derive from domain-general cognitive processes that the functions constructions have to fulfill (Croft 2001, Haspelmath 2008, Evans and Levinson 2009a, Boas 2010c, Goldberg 2013: 23–26).
19.3 Approaches Despite the common core of assumptions just outlined, the various major constructionist approaches that have been developed over the past thirty years differ significantly with respect to their view on several non-trivial issues. It is these differences that explain why currently there is still a need to speak of Construction Grammars and not a single Construction Grammar theory. Currently, the major Construction Grammars are the following: • Berkeley Construction Grammar (Fillmore 1985b, 1988, Fillmore, Kay, and O’Connor 1988, Michaelis 1994, Michaelis and Lambrecht 1996) • Sign-Based Construction Grammar (Boas and Sag 2012, Michaelis 2010, 2013) • the Parallel Architecture model (Jackendoff 2002, 2013, Booij 2013) • Fluid Construction Grammar (Steels 2011c, 2013, van Trijp 2014) • Embodied Construction Grammar (Bergen and Chang 2005, 2013) • Cognitive Construction Grammar (Lakoff 1987, Goldberg 1995, 2006, Boas 2013a) • Radical Construction Grammar (Croft 2001, 2012, 2013) These various approaches differ and agree with each other in a way that can best described by seeing Construction Grammar as a prototypical category with all of these approaches as exemplars that exhibit family
321
322
HOFFMANN
resemblance effects. The main issues that distinguish these approaches from each other (or in some cases unite them) are (see also Croft and Cruse 2004: 265–90, Goldberg 2006: 213–26): 1. Constructional storage: does a theory adopt a complete inheritance or usage-based position? 2. Formalization: is the theory fully formalized or not? 3. Processing: does the theory explicitly try to account for language production and processing? 4. Constructional types: are all constructions necessarily seen as formmeaning pairings or are defective constructions (with form but without meaning and vice versa) also postulated? 5. Theory of meaning: is the meaning pole of constructions represented in terms of truth-conditional, model theoretic terms or in a way that is compatible with cognitive semantic approaches such as, for example, embodiment simulation or frame semantics? Arguably the most important issue is the first one: some researchers still defend the competence-based complete inheritance approach, which only postulates something as a construction if it has some idiosyncratic pairing of morphological, syntactic, lexical, semantic, pragmatic, or discoursefunctional properties. Kay, for example, argues that one should consider as constructions only those linguistic phenomena that a speaker needs to know to “produce and understand all possible utterances of a language and no more” (2013: 32). In stark contrast to this, and in line with virtually all of the findings from the empirical studies on first and second language acquisition as well as psycho- and neurolinguistic evidence mentioned at the beginning of section 19.2, Bybee (2006, 2010, 2013) advocates a usagebased Construction Grammar approach. As she points out, the mental grammar of speakers is shaped by the repeated exposure to specific utterances and in which domain-general cognitive processes such as categorization and cross-modal association play a crucial role in the entrenchment of constructions. Consequently, specific phonological linguistic usage-events (exemplars) are stored together with rich semantic and pragmatic information first and any generalization (including the taxonomic network in Figure 19.1) is the result of an emergent, bottomup categorization and generalization process. The second issue concerns the question of formalization: some researchers aim to make all of their assumptions and postulated units formally explicit in order to allow for “more precise empirical prediction, enhanced comparability of analyses across languages, and theoretical clarity” (Sag, Boas, and Kay 2012: 3). In contrast to this, other Construction Grammarians argue that “grammatical categories and roles are not general across constructions but are only defined with respect to particular constructions” (Goldberg 2006: 216), which precludes the use of construction-independent, general formalizations.
Construction Grammars
The remaining issues are: third, whether a Construction Grammar theory is explicitly designed to model language production as well as processing or whether it is developed as a process-neutral model of linguistic knowledge. Fourth, whether cognitive generalization can also lead to defective constructions with only a form or meaning pole. Finally, the approaches differ greatly as to the semantic model adopted for the meaning pole. In the following, I will present and contrast the various Construction Grammars with a special focus on the five central issues of controversy.
19.3.1
Complete Inheritance Approaches: Berkeley Construction Grammar and Sign-Based Construction Grammar Berkeley Construction Grammar (BCG) and Sign-Based Construction Grammar (SBCG) are the only constructionist approaches that currently still advocate a complete inheritance approach. Historically, BCG5 is the oldest Construction Grammar with first publications of this framework surfacing in the late 1980s to early to mid-1990s (Fillmore 1985b, 1988, Fillmore, Kay, and O’Connor 1988, Michaelis 1994, Michaelis and Lambrecht 1996). The main initial motivation behind BCG was the goal to develop a grammatical framework that would analyse core as well as peripheral grammatical patterns by the same mechanisms (see section 19.2 and Fillmore 2013). BCG is a formalized theory that uses ‘boxes within boxes’ representation and the attribute-value-matrixes (AVM), yet relied on a problematic definition of construction unification (Mu¨ller 2006: 863–66, 2015: 309–14, Sag, Boas, and Kay 2012: 6). This was one of the reasons why Paul Kay later started to collaborate with Ivan Sag, Hans Boas, and Laura A. Michaelis on an approach called Sign-Based Construction Grammar. SBCG is also a formalized Construction Grammar theory that evolved out of ideas from BCG and construction-based Head-Driven PhraseStructure Grammar (HPSG) (Sag 1997, Ginzburg and Sag 2000; cf. Michaelis 2013). Like BCG and HPSG, SBCG employs AVMs or feature structures to model linguistic phenomena. However, unlike BCG but similar to HPSG, the feature structures of SBCG are typed, that is, arranged in a hierarchical inheritance classification (Carpenter 1992). BCG and SBCG are processneutral approaches that make no predictions about the actual online parsing or production of constructions. Moreover, both approaches draw on meaningless phrasal constructions to license sentences and thus postulate the existence of defective constructions. A case in point is subjectverb agreement in English as exhibited in He3PS.SG loves3PS.SG you versus They3PS.PL love3PS.PL you. As (9) shows, SBCG models this phenomenon by the postulation of a construction (more precisely a constraint on feature 5
Originally, publications with the framework only referred to it as Construction Grammar. The attribute ‘Berkeley’ was added only recently to distinguish it from other constructionist approaches (Sag, Boas, and Kay 2012: 2, Fillmore 2013).
323
HOFFMANN
structures) that specifies that a verb has to be marked as third-person singular (3sing) if its subject (XARG) argument also has this feature: (9)
Subject-verb agreement in SBCG ⎡FORM ⎢ ⎡ ⎡ AGR 3sing ⎢ ⎢SYN ⎢ CAT ⎢ ⎣ XARG Y: [AGR 3sing] ⎣ ⎣
⎡⎡ ⎡ ⎢⎢ ⎢ ⎢⎣ ⎣ ⎢ ⎣
324
(SBCG analysis of subject-verb agreement modeled on Sag, Wasow, and Bender 2003: 107, Sag, personal communication; from Hoffmann 2013a: 310)
Turning to the meaning pole of constructions, early BCG work was already influenced by insights from Frame Semantics (Petruck 1996) and later by its technological implementation in the FrameNet database (http://framenet.icsi.berkeley.edu/) (cf., e.g., Fillmore, Johnson, and Petruck 2003, Boas 2006, 2009). Early HPSG Construction Grammar (e.g. Ginzburg and Sag 2000) and SBCG (Sag 2010) represented the meaning of constructions by model theoretic, truth-conditional semantic theories such as Situation Semantics or Montague Possible Worlds Semantics. Recent SBCG approaches, however, also implemented Frame Semantics as their semantic model (Sag 2012: 87–88, Fillmore, Lee-Goldman, and Rhomieux 2012).
19.3.2
Models of Human Language Processing: Parallel Architecture, Fluid Construction Grammar, and Embodied Construction Grammar As pointed out above, all other Construction Grammar approaches subscribe to a usage-based view. Out of these, Parallel Architecture (PA) (Jackendoff 2002, 2013), Fluid Construction Grammar (FCG) (Steels 2011c, 2013), and Embodied Construction Grammar (ECG) (Bergen and Chang 2005, 2013) are all explicitly designed to model language processing and production. On top of that, all three are formalized syntactic approaches (though only FCG and ECG use feature structures, while PA uses subscripts to link atomic constructional elements across the various constructional poles, similar to the notation system used in the present chapter). Besides, in PA, phonology, syntax, and semantics are described as independent generative components within a speaker’s mental grammar. Constructional units (i.e. pairings of phonological, (morpho)syntactic, and semantic structures), called ‘lexical items’ in PA, then function as smallscale interface rules between the three generative components. In addition to normal form-meaning pairings, PA also postulates defective constructions, that is, abstract syntactic principles without meaning (such as the VP-construction [VPV NP]) or semantic principles without syntactic form, such as reference transfer in I have read Shakespeare (Shakespeare = ‘books/
Construction Grammars
plays by the author Shakespeare’). As an example of such a defective construction, compare the following PA Subject-Verb Agreement construction, which has no meaning pole but simply links the first grammatical function GF (i.e. the subject) with the tense-carrying verbal element (T) and, via coindexation of subject and an agr(eement) suffix, ensures feature concord: (10)
Subject-verb agreement in PA [GFi(>. . .)]k , [S. . . T + agri. . .]k (Culicover and Jackendoff 2005: 192)
Concerning the representation of the meaning pole of constructions, PA employs Jackendoff’s Conceptual Semantics (2002: 267–421), a mentalistic, decompositional conceptual theory. FCG is a computational formalism that developed out of research into artificial intelligence and language evolution, which makes no claims about psychological validity (Steels 2011a: 3). In contrast to this, ECG is explicitly designed as a computational model of the cognitive and neural mechanisms that underlie human linguistic behavior, and it draws on insights from cognitive linguistic research (Bergen and Chang 2005, 2013). These differences in objectives are also reflected in the semantic theory adopted by the two approaches: while FCG standardly employs truth-conditional first-order predicate calculus,6 ECG relies on models of mental simulation and embodied schemas. Moreover, while FCG allows for defective constructions, constructions in ECG are always formmeaning pairings (though purely form or meaning schemas are nevertheless considered to exist). Again, consider how the two approaches encode subject-verb agreement: (11)
Subject-verb agreement in FCG (?subject-unit (syn-cat (= =1 NP (number?n) (person?p)))) (?predicate-unit (syn-cat (= =1 verb (number?n) (person?p))))
(12)
Subject-verb agreement in ECG construction Intransitive-Clause ... constructional constituents theme : R E F - E X P R verb : V E R B constraints verb.person ↔ theme.person verb.number ↔ theme.number ...
(adapted from Steels and de Beule 2006: 220)
6
Yet, since FCG is more of an approach than a dogmatic theory, it is not surprising that some FCG implementations also use, for example, Frame Semantics or other semantic theories (cf. Steels 2011b: 79).
325
326
HOFFMANN
form constraints themef before verb meaning ... (adapted from Bergen 2003: 8)
Similarly to (9), the FCG analysis in (11) ensures that the subject (the subject-unit) and the verbal element (predicate-unit) have identical person and number features by token identical variables (cf. the variables ?n and ?p). Due to the flexibility of FCG, however, individual researchers can also write FCG grammars (www.fcg-net.org/) that would avoid defective constructions and only include form-meaning pairings. How this could be implemented is illustrated by the ECG construction in (12). Here, subject-verb agreement is not modeled as an independent constraint, but as a subpart of a construction: in (12), the Intransitive-Clause construction. The agreement constraints (verb.person ↔ theme.person and verb.number ↔ theme.number) would then have to be repeated for all constructions in which subjectverb agreement is obligatory.
19.3.3
Non-formalized Usage-based Approaches: Cognitive Construction Grammar and Radical Construction Grammar The remaining two approaches, Cognitive Construction Grammar (CCG) (Lakoff 1987, Goldberg 1995, 2006, Boas 2013a) and Radical Construction Grammar (RCG) (Croft 2001, 2012, 2013) are both usage-based approaches that are not intended to directly model language production and processing and that eschew formalization (since grammatical categories and roles are considered construction-specific and elude simple crossconstructional generalizations). Both CCG and RCG are influenced by previous research into Cognitive Grammar (CG) (inter alia, Langacker 1987, 2000, 2008a, 2009, Broccias 2013). Following CG, CCG, and RCG take language to be grounded in embodied human experience and languageindependent cognitive processes such as association, automatization, schematization, and categorization. One major difference, however, is that CG sees constructions as pairings of a semantic pole with a phonological one, whereas CCG and RCG also postulate (morpho-)syntactic information in their form pole. Neither CCG nor RCG allow for defective constructions (nor does CG). Consequently, subject-verb agreement in these approaches is modeled similarly to ECG, that is, as constraint/construction that is a subpart in the various relevant constructions, such as the Intransitive construction (I N T R S B J I N T R V ) or the Transitive construction (T R S B J T R V T R O B J ) (cf. Croft and Cruse 2004: 288, Croft 2013: 220). With respect to the semantic model underlying their meaning pole, both CCG and RCG draw on cognitive theories (such as Frame
Construction Grammars
Semantics) and take meaning to be the result of cognitive construal and not truth-conditional constraints (though see section 18.3 for differences concerning the prominence of verbal semantics). CCG and RCG are therefore closely related to each other and are also the most widely used Construction Grammar frameworks. This is particularly true for usage-based studies that adopt a corpus-based approach. In theory, results from corpus studies, that is, the analysis of the documented authentic use, type, and token frequency of constructions, are highly relevant for any usage-based approach. Take, for example, collostructional analyses (for an overview, cf. Stefanowitsch 2013), the family of quantitative corpus-linguistics methods that aim to statistically identify significant associations of word constructions and phrasal/clausal constructions (simple and distinctive collexeme analysis, e.g., the types of verbs that are significantly associated with a specific Argument Structure construction) or significant co-occurrence patterns of two slots in a construction (co-varying collexeme analysis, e.g., between the verb and the theme slot of the Ditransitive construction). Virtually all major publications using these methods are framed in a CCG/ RCG framework (inter alia, Stefanowitsch and Gries 2003, Gries and Stefanowitsch 2004a, 2004b, Stefanowitsch and Gries 2005, Gilquin 2006, Scho¨nefeld 2006, Stefanowitsch 2006c, Hilpert 2008, Stefanowitsch and Gries 2008, Colleman 2009a, 2009b, Mukherjee and Gries 2009, Wulff et al. 2009, Gries and Stefanowitsch 2010). Still, important differences between CCG and RCG can nevertheless be found: while CCG has a stronger focus on trying to capture and motivate language-specific generalizations (Goldberg 2006: 226), RCG, due to its strong typological background, is non-reductionist: it emphasizes that all grammatical categories are language-specific and constructionspecific and consequently assumes no formal syntactic structure other than the part–whole structure of constructions and the grammatical roles that occur in constructions. Moreover, RCG advocates an exemplar semantics model of the syntax–semantics mapping in which specific situation types are organized in a multidimensional conceptual space. Formal construction types are then said to have a frequency distribution over that conceptual space.
19.3.4 Summary of Construction Grammar Approaches Table 19.1 broadly summarizes the properties of the seven Construction Grammar approaches presented above. As Table 19.1 shows, Construction Grammar approaches still disagree on important theoretical and empirical points. It will therefore be interesting to see whether the various approaches will converge on these issues in the next couple of years or not. Still, from a cognitive linguistic point of view, a clear positive trend toward more cognitively oriented approaches
327
Table 19.1 Summary of properties of Construction Grammars Property
BCG
SBCG
PA
FCG
ECG
CCG
RCG
constructional storage
complete inheritance yes no yes Frame Semantics
complete inheritance yes no yes Frame Semantics
usage-based
usage-based
usage-based
usage-based
usage-based
yes yes yes Conceptual Semantics
yes yes yes Model Theoretic
yes yes no Embodied Simulation
no no no Frame Semantics
no no no Frame Semantics
formalization processing-oriented defective constructions theory of meaning
Construction Grammars
to the meaning pole of constructions can already be observed – something that should be highlighted more explicitly by future constructionist research.
19.4 Conclusion Construction Grammars are a family of cognitive linguistic approaches that provide a psychologically plausible theory of human language. Over the past thirty years, the notion of constructions, cognitive pairings of form-meaning, has spawned a large body of research. Construction Grammars offer a cognitive syntactic framework that has successfully provided an explanatory account of first as well as second language acquisition and has received support from psycho- as well as neuro-linguistic studies. Considering their empirical success, it is somewhat surprising that so far only a handful of authoritative summaries of Construction Grammars exist. There are numerous research articles, monographs, and edited volumes and even two journals specifically dedicated to constructionist research: Constructions and Frames and the open access online journal Constructions. Yet, so far only one handbook (Hoffmann and Trousdale 2013) and two textbooks have been published (Ziem and Lasch 2013, Hilpert 2014, both mostly focusing on CCG/RCG approaches). The major reason for this is that so far one cannot identify one single Construction Grammar theory. As the present chapter has shown, despite important shared theoretical commitments, the various constructionist approaches still differ with respect to important empirical and theoretical points. It will therefore be interesting to see whether the various approaches converge on certain assumptions within the next couple of years (concerning, e.g., the lexical versus phrasal construction controversy or the question of defective constructions). Despite these controversies, however, it is important to point out again the empirical descriptive and explanatory success that Construction Grammars have already had as cognitive theories of language and that future constructionist research will undoubtedly contribute to a deeper understanding of human language and the mind.
329
20 Cognitive Linguistics and Pragmatics Kerstin Fischer
Over the past decades, research in the areas of cognitive linguistics and of pragmatics and discourse have increasingly converged since, in many respects, the concerns of cognitive linguistics and (many areas within) pragmatics are similar, namely to investigate language in its dependency on those aspects that characterize human speakers. While the definition of pragmatics is notoriously difficult (cf. Levinson 1983: ch. 1), the areas it deals with include, crucially, the relationship between speech events and their contexts, language usage, verbal (and multimodal) interaction, larger units such as text, and spoken discourse. These areas have also increasingly been addressed in cognitive linguistics over the past years (e.g. Langacker 2008a, Nikiforidou and Fischer 2015). In earlier years, cognitive linguistics focused on how cognitive processes such as categorization and analogical reasoning contributed to motivating linguistic structure (e.g. Lakoff 1987). Because the claim that human, embodied cognition influences grammatical structure was the most novel and controversial, these early developments focused on the description of the language system and not so much on language use, leaving language use aside to be dealt with in the framework of pragmatics and discourse analysis (see Fischer 2002). Several developments caused these barriers to drop, especially the usage-based model (Langacker 1988, Bybee 2010), but also the rise of grammatical theories (in particular, cognitive grammar and construction grammar). These theories invite the inclusion of semantic and pragmatic information as well as make possible the gathering of increasing evidence that grammatical units emerge from use (in discourse and interaction) (cf. Langacker 2008a: 458ff.). These developments render a strict division between language structure and use impossible, opening up cognitive linguistics to domains previously left to discourse analysis or pragmatics. For instance, Langacker (2008a) suggests that large schematic structures, such as ‘story,’ should be included in cognitive grammar, and construction grammar has become
Cognitive Linguistics and Pragmatics
a tool to represent in detail linguistically encoded pragmatic and even interactional knowledge (e.g. Fillmore, Kay, and O’Connor 1988, Kay 2003, Auer and Pfa¨nder 2011, Fischer 2015a). However, in spite of their increasingly shared focus on language use, the two disciplines still differ generally in the perspectives they are taking, and there are differences in objectives, ontologies, and methodologies: cognitive linguistics investigates language phenomena primarily from a cognitive perspective, whereas work in pragmatics and the study of discourse approaches the functional use of language in context from a socially situated perspective. These two perspectives converge in many respects, yet there are also areas in which the two approaches lead to different results and in which they could profitably complement each other. In the following, I will discuss three areas in which the different contributions of the two disciplines become visible; I will use these examples to illustrate both the differences between the two disciplines and how they can complement each other. The first example considered concerns the extent to which aspects of situated use are part of grammatical knowledge; for instance, knowledge about situation-specific language use could be part of people’s grammatical knowledge, or it could be handled by other types of knowledge, for instance, pragmatic knowledge about situated social practices or frame semantic knowledge. Related to this issue is the question of the size of linguistic units to be assumed. While cognitive linguistic approaches to grammar also include larger, semi-fixed units such as idioms and proverbs (e.g. Fillmore, Kay, and O’Connor 1988), it is an open question as to whether it makes sense to assume units that encode situational practices (cf. Nikiforidou and Fischer 2015). The example considered will be recipes, a particular text-type with relatively predictable components and structure. The case study will illustrate one area in which pragmatic and cognitive linguistic analyses yield very similar, in fact indistinguishable results. The second example considered, discourse particles in similar structural contexts but in different speech situations, will demonstrate how a cognitive linguistic approach can add to a pragmatic analysis in ways that enriches a pragmatic/discourse analytic description considerably. In particular, a pragmatic analysis of the functions of individual lexical items is limited in terms of generalization, whereas a cognitive linguistic analysis can go beyond description and explain how speakers can actually learn the language and interpret utterances by introducing general mechanisms, such as constructions and frames, that help generalize beyond individual patterns. The third example considered concerns the interactional situatedness of frame-semantic knowledge and thus serves to show how the pragmatic analysis of interactions can contribute to our understanding of a frame and how it relates to language use in interaction.
331
332
FISCHER
20.1 Case Study I: Recipes Recipes are generally taken to be prototypical examples for registers (e.g. Ferguson 1982), where registers are seen as conventionalized linguistic varieties with identifiable linguistic properties related to particular situations of use (cf. Biber 1994, Biber and Finegan 1994, Crystal 2001, Halliday and Matthiessen 2004). Recipes are generally structured according to different phases in the food preparation process. The first phase concerns decision making, that is, the question of what to cook; this part of the recipe comprises a projection of the results of the process and is intended to help the reader to make a choice for or against the recipe. The second phase concerns the selection of ingredients and is often presented as a list. The third phase concerns the food preparation process itself. This structure corresponds to what a reader might “demand to be told in a recipe” (Fisher 1983: 17). Fisher (1983: 23) thus argues that a prospective reader of a cookbook will need three types of information: name, ingredients, and method. These requirements have been found to have developed over time; Arendholz et al. (2013) find, for instance, that the recipes of the fifteenth century do not present lists of ingredients. Cotter (1997) breaks recipes down into what she calls “narrative components”; she suggests a general bipartite structure consisting of ingredients and directions (1997: 59) with additional optional components that serve orientation and evaluation functions (1997: 60). For instance, an optional component is a fixed expression like ‘serves 6’ which can occur at the beginning or end of the recipe. The different ‘narrative components’ of recipes are conventionalized to different degrees. While decision making is not conventionalized in any way, descriptions of ingredients and the food preparation instructions may contain highly conventionalized elements. The ingredients are usually presented as an ordered list, where the order of ingredients reflects either the timing in the food preparation process or the type of ingredients used (for instance, spices or ingredients for the garnish may be presented separately). The food preparation process is often presented with characteristic linguistic features, such as the imperative mood (Ferguson 1982). Accordingly, Cotter argues that the imperative is “the recipe’s most distinguishing syntactic feature” (Cotter 1997: 56). Fisher (1983: 21) even argues that most recipes start out with the same word: “Take. . .”. Similarly, in Fischer (2013) I have analyzed four different cookbooks written for very different audiences, and I found imperative mood to be present in all recipes, too. The imperative has also been found in conversational recipe telling, where conversationalists switch back and forth between instructions, which are then presented in the imperative mood like their written counterparts, and narrations, which are then presented in the first-person past indicative (Norrick 2011: 2745).
Cognitive Linguistics and Pragmatics
Besides the characteristic use of the imperative, both Fisher (1983: 21–22) and Cotter (1997: 57) identify the absence of temporal adverbs as a formal characteristic of recipes. They also argue that recipes rely heavily on their readers’ background knowledge, for instance with respect to the interpretation of scalar or vague descriptions like ‘extra thick,’ which may require sophisticated background knowledge on food preparation processes. This is also reported by Culy (1996), who describes null objects as a distinguishing characteristic of recipes. Again, I have found instances of all these features in the four cookbooks written specifically for various audiences I analyzed, even though to different degrees (Fischer 2013). The extent to which cookbooks presuppose contextual and background knowledge generally depends on their target audience: Cotter finds that cookbooks for general audiences are explicit and detailed whereas those for a special readership allow for more brevity (1997: 69). Thus, in her view, audience design shows especially in the amount of background knowledge presupposed (see also Diemer 2013), which was generally supported by my own study; for instance, in a cookbook for girls, in general little background knowledge on cooking was assumed; nevertheless, some concepts certainly exceeded the general background knowledge of kids and even teenagers. However, in Child et al.’s My Life in France (2006), an introduction to French cooking for a general adult American audience, all steps were made explicit. Waxman (2004: 249) predicts that recipes will develop to become increasingly simpler, more explicit, and more intuitive, providing more background information in the future. To summarize, with recipes we have a text type that is structured in certain recognizable ways, yet whose structure is entirely driven by the function of recipes to support the reader in a decision concerning what to cook, in selecting the ingredients and in preparing the food. If any of these functions is not relevant, or other functions are required as well, recipes may well deviate from this structure. For instance, while the cookbook for girls analyzed comprises description (name), ingredients, and food preparation as components, these are embedded in a much richer and highly structured grid (Schriver 1997), which also covers tools required, illustrations, tips, and other elements. Moreover, much variation can be found in those formal features that are taken to be characteristic of recipes, which mostly concern the absence of linguistic properties that would be expected in other text types, such as narratives, for instance, determiners, temporal adverbs, or overt subjects and objects. The lack of these features is, however, unproblematic in recipes because the context-dependency of recipe use makes up for the missing information since it either concerns background knowledge that a person with some cooking experience can be expected to have, or information that will be situationally available once one has started the meal preparation process. The linguistic features identified may therefore not necessarily be conventional characteristics of recipes; instead, they
333
334
FISCHER
may result from the fact that recipes are expected to be read in very predictable environments involving objects and actions that may previously have been introduced (like the ingredients) or that can be expected to belong to readers’ knowledge of kitchens and cooking (cf. Clark 1999). If for some reason contextual or domain knowledge cannot be presupposed, like in recipes for children, these features are rare (Fischer 2013). Furthermore, the marked absence of certain linguistic features, even though characteristic of recipes, is not constitutive of recipes since it can ¨ stman 2005), brief instructions, also be found in newspaper headlines (O and other text types (e.g. Matsumoto 2015). Finally, the functionally driven structure of recipes becomes apparent when very different media are used, as in, for instance, food blogs. Diemer and Frobenius (2013) find very similar content in recipes in food blogs and in cookbooks, though food blogs showed additional components, such as date and a link to an archive. Jensen (2016) compares recipes from cookbooks and food blogs directly and finds that regarding rhetorical structure, both text types are very similar, but that cookbooks display a much more varied layout structure. From a pragmatic perspective, the patterns observable can be clearly described: there are typical ways in which recipes are written. These typical ways are functionally motivated but can be taken to be conventionalized: socially shared practices that shape future behavior and account for the linguistic peculiarities observable. While some approaches within pragmatics can content themselves with describing these practices (e.g. conversation analysis; see Hutchby and Wooffitt 1998), most would still want to motivate the patterns observable by relating them to communicative function (e.g. Johnstone 2007). From a cognitive linguistic perspective, in general only those aspects may enter the description that can be demonstrated to be cognitively valid for the speakers; from this perspective, we need to ask whether recipe writers really orient to these characteristics as part of their conventional knowledge, or whether they simply attend to the functions they want their recipe to fulfill: to inform, for instance, the reader about the dish, to provide a list of ingredients, and to instruct the reader how to prepare the food. In the second case, knowledge about the structure of recipes would not be linguistic knowledge, but instead it would be knowledge about what is involved in teaching someone to cook a certain dish, of which the structural components would just be an epiphenomenon. In cognitive linguistics, we have several options: the expected components of recipes could be represented as part of one ¨ stman 2005) or as a schematic sign in large recipe construction (e.g. O cognitive grammar (cf. Langacker 2008a); this would state adequately that recipes generally comprise the parts discussed above. It is also possible to indicate for the food preparation instructions that they are usually delivered in imperative mood, and that subjects, objects, and adverbs may be spared. Alternatively, we can assume a frame semantic representation
Cognitive Linguistics and Pragmatics
of the cooking process, which does not include any representation of linguistic properties; instead, from this perspective, these are the result of the attended-to function of brevity and of the availability of situational contextual information, available in the restricted (and allegedly shared; cf. Clark 1999) universe of the preparation of the particular dish. Eventually, however, the results of pragmatic and cognitive linguistic analyses of recipes are very similar and very difficult to disentangle – which is as to be expected from the perspective of the usage-based model (Langacker 1988); activities that are often repeated and constitute a part of our encyclopedic knowledge about how things are done can be assumed to be stored schematically and created again on the basis of available (functional) resources (Langacker 1991); recipes are thus an example of a linguistic phenomenon for which a pragmatic and a cognitive linguistic perspective yield very similar results.
20.2 Case Study II: Discourse Particles The second example analysis concerns the meanings of discourse particles in different situations. This case study will illustrate the limits of a purely pragmatic account and indicate at the same time how a cognitive linguistic approach may go beyond the results achieved in a pragmatic analysis. When we consider the functions of discourse particles such as ah, oh, well, yeah, and yes in different situations, we find not only that the frequencies with which they are used differ significantly according to the situation of use, but also that the same discourse particle in very similar contexts and even sequential contexts is used with different functions (Fischer 2000a). The German discourse particle ne, for instance, elicits (preferably positive) feedback from the hearer. Usually it also serves as a turn-yielding signal, such that the speaker-role changes without overlap after the occurrence of ne:1 (1)
09I073: 09K072:
es ist jetzt hinten bei dir ziemlich fertig, ne? [your model is now pretty complete in the back, PTC?] ja wu¨rd ich mal sagen [yes < – - > I would say so]
In this example, the turn-final discourse particle ne serves to elicit feedback from the communication partner. It occurs at the end of an assertion that presents a prediction about the communication partner’s current state, to which the communication partner responds by producing the expected feedback and by confirming the communication partner’s prediction. 1
The data stem from a corpus of toy-airplane construction dialogs (Sagerer, Eikmeyer, and Rickheit 1994) in which an instructor explains to a constructor how to build a toy airplane from wooden parts. represents a pause between 300 and 500 milliseconds, represents a pause between 500 milliseconds and one second, word and {word} represent parallel speech, word represents a word uttered quietly, encodes a human sound event, here breathing, describes a word that could not be heard with sufficient clarity. The transcription conventions are detailed in Fink, Johanntokrax, and Schaffranietz (1995).
335
336
FISCHER
In the next two examples, ne occurs in utterances with the same formal properties, that is, at the end of a statement and with rising intonation contour: (2) (3)
19K002 19I002 13K011 13I013
schraube ich rein, ne? [I screw it in, PTC?] ja, schraubst du rein [yes, you screw it in] a¨hm soll der, also die eine Seite ist dann praktisch frei ne? [um should it, so the one side is free then PTC?] ja. [yes.]
In all instances, the respective communication partner responds by affirming the proposition made by the speaker. We can conclude that the discourse particle ne functions to secure a shared understanding with the communication partner in situations in which a good candidate interpretation is available (hence the position at the end of an assertive clause). In asking confirmation from the communication partner, which is usually promptly provided, it also functions as a turn-yielding signal. Furthermore, by taking the partner into account, it also serves politeness functions. Pragmatic analysis can now still go further by revealing more detail about contexts of application, subtle communicative effects, and different prosodic deliveries, for instance (e.g. Ehlich 1986, Willkop 1988), yet the current analysis serves to illustrate that pragmatic methods tease out the interactional functions discourse particles may have in different situations. Now, what we can observe is that there are other discourse particles that serve very similar functions, for instance ja [yes]: (4)
21I020
(5)
21K020 11I062
(6)
11K063 13K040 13I042
jetzt u¨berschneiden sich quasi so die a¨hm eben haben sich ja nur die beiden u¨berschnitten, ja? [now they overlap so they um before only two of them overlapped, PTC?] ja. [yes.] du hast erstmal den roten Wu¨rfel, ja? [first of all you have the red cube, PTC?] ja. [yes.] und der Fu¨nfertra¨ger ist jetzt dadrunter, ja? [and the five-hole carrier is below now, PTC?] nein, oberhalb. [no, on top.]
These examples illustrate that also instances of ja [yes] in turn-final position at the end of an assertive clause elicit immediate feedback from the communication partner; example (6) shows that the feedback does not always have to be positive. Besides ja and ne, also oder [or], gell (checking particle like ne) and nicht wahr [literally: not true] occur in the same position with the same function, to elicit feedback from the communication partner, which is responded to by the partner by means of a confirmation:
Cognitive Linguistics and Pragmatics (7)
13K092 13I094
(8)
12K015
(9)
12I021 18I001
18K002
und das wird dann da seitlich draufgeschraubt oder? [and thus is then screwed onto it from the side or?] ja, genau, aber mach(e) das erstmal so. [yes, exactly, but first do it so.] ja, Moment, ich habe das noch nicht ganz. ich darf auch reden, gell? [yes, just a moment, I haven’t got it yet. I may also speak, PTC?] ja ja. [yes yes] also du siehst ja was das ist, {nicht wahr?} erkennst du es? [so you see what this is, {PTC?} do you recognize it?] ein Flugzeug. [an airplane.]
Less common and less conventionalized are also the following instances. In the first two examples, the instructor responds immediately with a confirmation, in the third example, the instructor responds with a delay: (10)
12K081 12I088
(11)
12K123
12I130 (12)
12K042
12I049
dann sind das die Tragfla¨chen oder was? [so these are the wings or what?] ja, genau, das sind die die hinterste Tragfla¨che, genau, [yes, exactly, these are the the rear wing, exactly,] ja, jetzt sind es ja wieder, jetzt sind es ja wieder drei, was? {weil} jetzt steht ja das eine ab vorne. [yes, now there are again now there are again three, what? {because} now one the one sticks out in the front.] ja, richtig, jetzt sind es wieder, sind es wieder drei, genau. [yes, correct, now there are again, there are again three, exactly.] und die Muttern ko¨nnen u¨berstehen oder wie? nee, das [and the bolts may stick out or how? no, that] das macht nichts [that doesn’t matter]
What all of these instances indicate is that the function, to elicit a confirmation for a candidate understanding from the communication partner, is encoded in the structural position in which the discourse particles (or speech formulas) occur, not in the discourse particles themselves. At least ja [yes] and oder [or] can fulfill many other functions besides seeking confirmation; that is, there is nothing in the meanings of ja [yes] or oder [or] that would encode this function. Instead, it is the structural position (at the end of an assertive utterance produced with rising intonation) that carries this meaning, which is evident from the fact that even nonconventional particles such as was [what] fulfill the same function if used with rising intonation in the turn-final position of an assertion. We can therefore interpret the structural position as a schematic form-meaning pair, as a grammatical construction in the construction grammar sense (Goldberg 1995, 2006), and thus generalize over different discourse particle lexemes. This perspective goes beyond the pragmatic approach by introducing a grammatical perspective that states generalizations over
337
338
FISCHER
lexical items (Fischer 2010). In this approach, the construction consists of the utterance- and turn-final position of an assertive clause with rising intonation on the form-side, and the function to elicit a confirmation from the communication partner, yielding the turn to her, on the meaning-side of the construction (see also Fischer 2000a, 2000b, 2010, 2015a). The cognitive linguistic perspective thus extends the pragmatic analysis by allowing to generalize over different lexical items. However, we can take the cognitive linguistic perspective on the data even further. What we can find is that the same discourse particle item in the same construction may still not fulfill the same functions if used in different situations; using pragmatic methods, this phenomenon has been widely acknowledged and is usually accounted for by explicating the context conditions in which the data investigated occur. An example is Heritage’s accumulation of functions of the English interjection oh in various different contexts (e.g. Heritage 1984, 1998, 2002). Whereas pragmatics has few means for connecting these different functions, a cognitive linguistic perspective can account for the range of different interpretations, which I am going to illustrate in situations in which the communication partner does not play the same role, as in instruction dialogs between a human and an artificial communication partner. The second data set investigated thus also comprises task-oriented interactions, namely instructions to build a toy airplane; however, in this corpus (Brindo¨pke et al. 1995), participants believed that they were giving these instructions to an artificial communication partner, an automatic speech processing system (which was actually simulated believably by two human wizards; see Fischer 2000a for details), and the construction results were displayed as images on participants’ screens. In these data, discourse particles such as ne, ja, and also are used, too, yet in much lower numbers (see Fischer 2000a). Furthermore, the functions of these discourse particles may be different from those identified for the human–human instruction dialogs. For instance, turn-final uses of ne after assertions that make predictions about a current or future state occur also in this corpus; however, in the simulated human–computer interaction data, all occurrences of ne are found in reduced volume. The speaker thus does not present the utterance containing ne for the simulated speech processing system to consider: (13)
13MS157
K 13MS158
muss es jetzt noch richtig stehen? a¨hm ja, ne die beiden Holz/ a¨hm [ does it have to be placed correctly? um yes, PTC the two wood/ um] ich konnte Ihre Äußerung leider nicht verstehen [unfortunately I could not understand your utterance] ich habe u¨berlegt. a¨hm die beiden Dreierholzsta¨bchen . . . [I was thinking um the two wooden threeholebars . . . ]
Cognitive Linguistics and Pragmatics (14)
30KR123:
K: 30KR124:
fertig ne ja jetzt [ ready PTC well now ] bitte sprechen Sie lauter [ please speak up] a¨hm das Zusammenbauen ist jetzt beendet. [um the construction is now completed]
Thus, in these dialogues the discourse particle ne occurs in structurally very similar contexts as in the human-to-human interaction data, that is, at the end of turn-final assertions. However, instances of ne in the humancomputer dialogues are not presented to the communication partner, and a response is not expected as it would be in the interactions between human communication partners. We can conclude our comparison of the functions of the discourse particle ne in instructions directed at a human and at an artificial communication partner by stating that even though this particle occurs in formally very similar grammatical constructions, its functions in the two data sets are different. While in human–human communication ne mostly functions to involve the communication partner and to relinquish the turn to her, in human–computer interaction, it mostly occurs in self-talk and serves to confirm one’s own line of reasoning. Since the structural contexts in which the two discourse particles occur in the two corpora are very similar, constructions cannot account for the different uses across situations. Instead, it is our understanding of the kinds of actions carried out and the fact that certain functions, such as relationship management and the interactional establishing of meaning, can be assumed to be much less relevant to speakers in the communication with artificial communication partners. Thus, our knowledge of relevant functions and activities serves as a background for our interpretation of the functions of the discourse particle occurrences. For instance, regarding example (13) above, we could interpret the occurrence of ne as an attempt to elicit a response from the partner – however, once it is known that the communication partner is a computer, these functions are quite unlikely. Thus, here our knowledge of attended-to functions and activities serves as a resource for the interpretation of the functions of discourse particle occurrences. Cognitive linguistics provides us with a means to represent such background knowledge; in particular, Fillmore’s (1982) interactional frames provide exactly the kind of framework to capture the background knowledge people carry into interactions (see also Fischer 2000a, Matsumotu 2010, 2015, Fischer 2015b). Such a frame encodes relevant communicative functions; for instance, for the description of the confirmation-seeking construction, we can assume that securing understanding and turn-
339
340
FISCHER
management are functions that are attended to in natural conversations between people. Independent evidence that these are indeed functions that characterize the communicative tasks conversational participants generally orient to comes from the frequent backchannels speakers produce (e.g. Gardner 2001, Bavelas, Coates, and Johnson 2002) as well as from explicit requests for signs of understanding; for example: (15)
21I021:
21K021:
und jetzt u¨berschneiden die sich dreimal. verstehst du? [and now they overlap in three places. do you understand? ] nee. [no]
Furthermore, much interactional work is devoted to turn-taking (Sacks, Schegloff, and Jefferson 1974), with respect to which discourse particles work as turn-allocation devices (e.g. Gu¨hlich 1970). It cannot be resolved whether the turn-yielding function of the discourse particles in the construction under consideration is an attended-to function or whether it is just an epiphenomenon of its function to elicit feedback from the partner (see Fischer 2000b). However, there is clear evidence that speakers in general employ means to signal that their turn is coming to a close (e.g. Local 1996, Selting 1996, Ford, Fox, and Thompson 1996). Moreover, I have argued above that speakers are fulfilling interpersonal functions as well when they involve the communication partner. Besides explicit conventional phrases dedicated to what is considered polite in a given speech community, such as please, thank you, or you are welcome (e.g. Aijmer 2005), many other interactional mechanisms have been identified that serve to save the faces of speakers and their partners (cf. Brown and Levinson 1987). Thus, there is independent evidence to state that the three functions potentially involved in the use of confirmation-seeking discourse particles like those under consideration are part of a general set of attended-to tasks in human conversation that constitute the background against which discourse particle occurrences are interpreted. Thus, my suggestion is that we can assume the existence of a cognitive frame that encodes general tasks in communication, such as that speakers need to secure successful perception and understanding of their utterances, that they need to attend to conversational flow and thus to turntaking, and that they need to do face work. These tasks constitute general tasks speakers can be assumed to attend to in conversation with other humans. Now, when people talk to computers, their understanding of what their tasks consist in is very different (cf. Fischer 2006); they are much more concerned with securing perception and establishing understanding of propositional information, and they are not very concerned about politeness and face-saving, even though politeness routines do occur. Thus, we can supplement the pragmatic analysis of the different functions of
Cognitive Linguistics and Pragmatics
discourse particles in the different situations with a communicative background frame that is activated for typical human-to-human interaction situations, yet which is suspended and replaced by different, adjusted models of communicative tasks that guide speakers’ interpretations in special situations, such as when having to talk to a computer. Different frames representing speakers’ background knowledge and communicative tasks associated with a given situation thus account for the different sets of functions that discourse particles may fulfill in different situations – and this time across constructions, since the interpersonal regulation function would also be missing, in, for instance, initial instances of discourse particles when uttered to a computer (cf. Fischer 2000a, 2006). Cognitive linguistics thus contributes the notion of grammatical constructions as form-meaning pairs and the notion of interactional frames as resources that help to make sense of the findings of pragmatic analyses. To sum up, this case study provides an example of how a pragmatic analysis can be meaningfully extended by means of concepts and methods from cognitive linguistics.
20.3 Case Study III: Frames The third case study considered in this chapter illustrates how a pragmatic approach can enrich a frame semantic representation and put it into perspective regarding its usage conditions (cf. also Fischer 2017). Here the pragmatic analysis shows the contingency of the frame instantiation on various interactional factors and thus provides a useful extension of frame analysis. Fillmore and Atkins (1992) demonstrate how a word’s frame can be developed from corpus instances, and this approach has been taken up and developed further in the FrameNet project (e.g. Baker 2014). In the following, I carry out such an analysis for a noun with the aim to analyze speakers’ underlying knowledge that they bring to bear in an interaction. The frame investigated is the ‘lamp’ frame: what it means to operate a common tabletop lamp. The corpora I use for this purpose are tutoring data in which one adult ‘expert’ explains the object under consideration to a language learner. Thus, in contrast to Fillmore and Atkins (1992) and other FrameNet work, I use very specific corpora to address the meanings of the word directly. Now, the data I analyze differ in controlled ways: corpus I is a collection of tutoring sessions between parents and their children, recorded by Katharina Rohlfing at Bielefeld University (Rohlfing et al. 2006). Corpora II and III (Lohan 2011, Lohan et al. 2012) are collections of tutoring sessions in which adults explain the same household objects to a language learner, yet this time not to a child, but to a robot with a childlike appearance (the iCub robot built by the IIT in Genoa, Metta et al. 2010). Corpora II and III differ with respect to the robot’s behavior. In both corpora, the robot looks at the objects demonstrated, at the tutor or around
341
342
FISCHER
Figure 20.1 A tutor introduces a lamp to the iCub robot
and points at the objects presented. However, in corpus II, the robot does so in direct response to the tutor’s behavior. In particular, if the robot perceives that the tutor is looking toward the robot, the robot looks back; if it perceives that the tutor is looking at the object, the robot also looks at the object; and if the tutor looks elsewhere, the robot tries to follow the tutor’s eye gaze. In corpus III, the robot uses the same behaviors, yet produces them at random. The impression created is one of liveliness, and the robot’s gaze behavior is in fact similar to the behavior of young children (Gogate, Bolzani, and Betancourt 2006). Figure 20.1 shows the experimental set-up with the iCub robot. The coding scheme for the data was iteratively developed (cf. also Ziem 2008): each utterance was assigned a ‘topic’ corresponding to one suspected frame element, that is, information that speakers regard relevant for knowing a lamp. Once an inventory of such frame elements was established, the corpus was annotated again to ensure maximal distinctiveness between and maximal consistency within each category. Consider an example interaction from the human–robot interaction data:2 (16)
2
dies ist eine Lampe? - - und die Lampe hat die Eigenschaft, dass a¨h - sie leuchtet. - - indem man – an diesem Kn- Knauf hier (at=lengthening)zieht(/a), also das heißt den Knauf nach unten bewegt, geht die Lampe an. – zieht man den (at=prominent)nochmal(/a), - - geht sie wieder aus. (2) und wie immer nach unten. geht an, - - und dann wieder aus. [this is a lamp? – and the lamp has the property, that it uh – it gives light. – by pulling – this handle here, so this means by
Transcription conventions: – signals a pause of 300 to 500 milliseconds; – –signals a pause of 500–999 milliseconds; (2) signals a pause of two seconds; (at=slow) word (/a) signals that a word is spoken in a particular way, here slowly; punctuation marks (.,?) signal falling, slightly falling, and rising intonation contours.
Cognitive Linguistics and Pragmatics
Table 20.1 Frame elements of the lamp frame frame element
example
intro
iCub? das hier ist eine Lampe iCub? this here is a lamp
function
sie macht Licht wenn man sie einschaltet it gives light when it is switched on
parts
die Lampe hat einen Schalter the lamp has a switch
sub-act
ich nehme meine rechte Hand und ziehe an der Kordel I take my right hand and pull on the string
manner
und ziehe wieder – von oben nach unten ganz stramm senkrecht and pull again from up to down completely straight vertically
result
dann ist die Lampe an then the lamp is on
resource
also als erstes muss diese Lampe Strom haben so first of all the lamp needs electricity
background
und wenn es dunkel ist, machen viele Leute gerne Licht an and when it is dark, many people like to turn on the light
moving the handle downwards, the lamp turns on. – by pulling again, - - it is turned off. and downwards as usual. turns on, - - and then off again.]
In the example, the participant first introduces the lamp (this is a lamp), then explains its function (has the property to give light), then describes the manner of its operation (pulling) and its result (light is on), after which he repeats the action to illustrate how it is turned off again. The iteratively established inventory of information considered relevant by the tutors in the three corpora thus comprises the categories listed in Table 20.1. Now, when analyzing the different interactions, significant differences between the corpora emerge; for instance, parents in the child-directed interactions restrict themselves to only a subset of the possible features of the lamp frame – for example, they introduce the lamp and illustrate the result of switching it on and off (exx. 17 and 18); very few also mention the string by means of which it is operated (ex. 19), for instance: (17)
VP005_3-s503: Alicia, guck mal. [Alicia, look.] VP005_3-s504: Guck mal. [look] VP005_3-s505: Hier macht man die Lampe an und hier macht man sie wieder aus. [here you turn the lamp on and here you turn the lamp off again.] VP005_3-s506: Guck einfach nur ziehen und klack nochmal und aus. [look simply pull and click again and off.] VP005_3-s507: Hier ziehen und nochmal. [pull here and again.] VP005_3-s508: Jetzt ist sie wieder aus. [now it is off again.]
(18)
VP007_1-s1140 Jasmin, Jasmin, guck mal. [Jasmin, Jasmin, look.] VP007_1-s1141 Hier, das ist die Lampe, ne? [here, that’s the lamp, isn’t it?] VP007_1-s1142 Und wenn man hier dran zieht, siehst du. [and when you pull on it, you see.]
343
344
FISCHER
VP007_1-s1143 Dann geht die Lampe an. [then the lamp turns on.] VP007_1-s1144 Und wenn man nochmal dran zieht, so, dann geht die Lampe wieder aus. [and when you pull again, like this, the lamp goes off.] VP007_1-s1145 Nochmal. [again.] VP007_1-s1146 Guck mal so. [look like this.] VP007_1-s1147 Hm. [hm.] VP007_1-s1148 Und wieder aus. [and off again.] VP007_1-s1149 Ja. [yes.] VP007_1-s1150 Kapiert. [got it.] (19)
VP006_3-s744 Gerry, kleine Maus, guck mal hier. [Gerry, little mouse, look here.] VP006_3-s745 Schalter und ah. [switch and ah.] VP006_3-s746 Super ja. [super yes.] VP006_3-s747 Da guckst du was. [that’s something, isn’t it.] VP006_3-s748 Ziehen, an. [pull, on.] VP006_3-s749 Und nochmal ziehen, aus. [and pull again, off.]
In contrast, in the interactions with the iCub robot, tutors introduce the lamp and mention its function, the result of pulling the string, but in addition, they explain further aspects. For instance, they may mention some parts, define subactions, provide manner information, explain the necessary resources for the lamp to work and provide background information; some examples of these kinds of explanations can be seen below: (20)
ja. – diese Lampe erzeugt ja Licht, wenn ich Licht haben mo¨chte ziehe ich an dieser kleinen Kette hier, – dann leuchtet die Glu¨hbirne in Inneren und ich habe Licht. [yes, this lamp creates light, when I want light I pull on this little chain here, then the lightbulb inside glows and I have light.]
(21)
das ist eine Lampe. (2) die Lampe hat einen Schalter, – das ist der Schalter. (1) ziehe ich den Schalter geht die Lampe an. (1) die Lampe ist an. (2) ziehe ich den Schalter wieder ist die Lampe aus. die Lampe ist jetzt aus. [this is a lamp. (2) the lamp has a switch, – this is the switch. (1) when I pull the switch the lamp turns on. (1) the lamp is on. (2) when I pull the switch again the lamp is off. the lamp is now off.]
(22)
so, das ist eine Lampe, mit der man Lampe a¨h ist ein elektrisches Gera¨t, wird mit Strom versorgt und wenn ich die anschalte wird’s hell in nem dunklen Raum, und um die anzuschalten muss ich unten an dieser Kordel ziehen, – dann geht das an, – und die rastet so’n bisschen wieder ein in ihrem urspru¨nglichen (unclear), a¨h (1) Ursprungszustand, und wenn ich nochmal ziehe, also die gleiche Ta¨tigkeit nochmal ausfu¨hre, dann ist sie wieder aus. [so, this is a lamp, with the lamp can is an electrical device, is operated with electricity and when I turn it on it becomes light in a dark room, and to turn it on I need to pull down the string here, – then it turns on, – and it positions itself a bit in its original position, and when I pull again, that is carry out the same action again, then it is off again.]
These differences between parent–child interactions and human–robot interactions are significantly different for function, parts, manner, and
Cognitive Linguistics and Pragmatics
Table 20.2 Average number of frame elements of the lamp frame mentioned and results of a univariate ANOVA for human–robot interaction with contingent response (HRI-cont), random response (HRI-rand) and childdirected speech (CDS)
intro
Corpus I CDS
Corpus II HRI-cont
Corpus III HRI-rand
F(2,61)
p
0.320
0.619
0.444
1.852
0.166
function
1.200
1.809
1.944
4.122
0.021
*
parts
0.080
1.000
0.722
9.987
0.000
***
sub-act
0.000
0.048
0.056
0.654
0.523
manner
0.000
0.476
0.444
3.866
0.026
result
1.680
1.286
1.777
0.839
0.437
resource
0.000
0.000
0.222
3.813
0.027
background
0.000
0.143
0.055
1.295
0.281
* *
resource (see Table 20.2). Furthermore, only participants tutoring the robot mention subactions and background information, but the difference is not statistically significant, as many robot tutors do not mention these. So whether tutors explain a lamp to a child or a robot causes considerable differences. Crucially, however, tutors also produce significantly different descriptions depending on whether they interact with the robot that is responding contingently or to the robot with the same behaviors displayed randomly. The statistical analysis reveals that tutors who interact with the contingently responding robot explicitly introduce parts of the lamp more frequently (p=.020), and they tend to provide more information about its manner of function (p=.094). Conversely, only the tutors who interact with the randomly responding robot mention outside resources like electricity (p=.053), and they introduce the lamp’s function significantly more often (p=.040). So we can identify a common core for the three corpora: All tutors across conditions introduced the object and presented the result of operating it, namely to get light. Beyond that, frame elements are considered relevant to different degrees depending on the situatedness of the communication partner, and we can see that the less responsive the communication partner is, the more situationally removed the explanations become. We thus need to distinguish between frame semantics as potentially available as a knowledge resource for the community and the possibly much more restricted resource that guides speakers’ productions in interaction with a particular communication partner. That speakers’ concepts of the communicative situation determine their speech events is a consequence of the cognitive approach and thus generally predictable.
345
346
FISCHER
While early cognitive linguistic work focused on how general conceptualizations influence language structure, results like those presented here indicate that also concepts of the communicative situations themselves influence language use – which is generally in accordance with cognitive linguistics and a natural extension. However, that subtle interactional differences in the timing of non-verbal behaviors of the communication partner may cause such different understandings of the affordances of a communication situation suggests that discourse, and hence an understanding of what the communication situation is all about, is really an interactional achievement. This points to social aspects of communication that are easily overlooked (see Bavelas 2005). Thus while cognitive models – for instance, of the communication partner – determine linguistic choice, these models are updated continuously and subtly in the course of the interaction (Fischer 2016). Speakers’ conceptualizations of the communication situation and the communicative tasks they see themselves faced with are therefore not their own, solitary products but interactionally established. As shown here, the effects of such interactionally achieved conceptualizations may not only concern linguistic choices (Fischer et al. 2013, 2014), but also the knowledge structures evoked, such as frames for simple household objects. To sum up, the analyses of tutors’ explanations for different communication partners with different levels of responsiveness show that the semantic frame that is relevant for the respective speech situation differs considerably in the way it is instantiated or even evoked as a cognitive representation. While a core of features could be identified, most aspects that tutors thought noteworthy for their particular communication partners differed across situations – situations that differed according to the nature of the communication partner, child versus robot, but also in response to subtle interactional timing differences.
20.4 Conclusion In this chapter, I have illustrated three ways in which pragmatic analysis and cognitive linguistic work may interact and benefit each other. While the concepts and methods available in each discipline are clearly different, the results of both may converge, as in the case of genre analysis, or they may complement each other, as in the case of discourse particles and frame analysis. Thus, together they contribute to a much more comprehensive basis for understanding language and its use.
21 Fictive Interaction Esther Pascual and Todd Oakley
21.1 Introduction Cognitive Linguistics rests on the assumption that language is the result of overall cognitive processes, and reflects both human cognition and our most basic experiences. This view originally led to the almost exclusive focus on our sensorimotor life, within an understanding of our body as a solitary biological organism moving in space (e.g. Lakoff and Johnson 1980a, 1999, Talmy 1996 [2000]). Critical as embodiment certainly is to our conceptualization of experience and language, such an approach should not undermine the role of social interaction as an essential and unavoidable aspect of human existence: we are born and raised into a world of conversational participants, who are – throughout our lives – an intricate part of our direct everyday environment. Indeed, interaction is the first communicative form mastered by human babies (Trevarthen 1979), children acquire language in conversation (Bruner 1983, E. Clark 2003), and language as a system emerged as an oral means of face-to-face communication and has been used exclusively in sequential interaction for most of its history. Critically, conversation is also the canonical form of verbal communication (H. H. Clark 1996), probably present in all languages and cultures (Sacks, Schegloff, and Jefferson 1974), and is thus likely a fundamental part of what makes us human (Levinson 2006). Conversation is furthermore the context in which language becomes meaningful, both for children (E. V. Clark 2003) and for adults (Bakhtin 1979 [1986], Goodwin 1995). If we assume that language mirrors our most basic human experiences, it is not far-fetched to hypothesize that the structure of language reflects its interactional dimension, an idea that is in fact far from new (Voloshinov 1929 [1986], Vygotsky 1934 [1962], Bakhtin 1975 [1981]). A substantial amount of work within Cognitive Linguistics has recently extended the original idea of language as modeled by both cognition and physical and physiological experience to cover the social and communicative dimensions as factors
348
PA SCUA L A N D OA K L EY
that also frame grammar (Pascual 2002, Verhagen 2005, Pascual 2006b, 2014) as well as semiotics (P. Aa. Brandt 2004, Zlatev 2007a, Zlatev et al. 2008, Sinha 2009, Oakley 2009, 2014). In this chapter we will try to show how Talmy’s (1996 [2000]) notion of fictivity, an ontological realm between reality and fiction, but which appears fundamental in the conceptualization of both the real and the unreal, can also be communicative in nature (Pascual 2006b). Specifically, we present the premise that there is a conversational basis for language, which serves to partly structure cognition, discourse, and grammar. Stemming from this tenet, we discuss the notion of fictive interaction or ‘FI’ (Pascual 2002, 2006b, 2014), namely the use of the template of face-toface interaction as a cognitive domain that partially models: (i) thought (e.g. talking to oneself); (ii) the conceptualization of experience (e.g. ‘A long walk is the answer to headache’); (iii) discourse organization (e.g. monologues structured as dialogues, Pascual 2006a, Xiang and Pascual 2016); and (iv) the language system and its use (e.g. rhetorical questions). In language, fictive interaction may occur at different levels: (i) the intersentence (e.g. ‘Any questions? Call us’); (ii) the sentence (Langacker’s 1999b ‘virtual speech acts’ e.g. ‘Why bother?’); (iii) the clause (e.g. ‘They felt, Oh no!’); (iv) the phrase (e.g. ‘the attitude of yes, I can do it’); (v) the word (e.g. ‘forget-me-nots’); and in some languages even (vi) the morpheme (e.g. a direct speech infix marking future tense in Aikanã, van der Voort 2013, 2016; see section 21.4.1).
21.2 The Conversational Conceptualization of Reality Various scholars have postulated that the structure of thought mirrors the turn-taking pattern of ordinary intersubjective conversation (cf. Voloshinov [1929] 1986, Vygotsky [1934] 1962, Bakhtin [1975] 1981; see overview in Melser 2004). We talk to ourselves as well as to imaginary friends, animals, or inanimate entities, and often speak about our thoughts and opinions in terms of an inner dialogue or argument between different versions of ourselves (Pang 2005). Similarly, the frame of the conversation may model the conceptualization of experience as a nonactual kind of interaction (Cooren 2010, 2012, Cooren and Sandler 2014). We commonly construe voting as speaking (Coulson and Oakley 2006); an artwork as ‘interacting’ with its artist or viewers (Sullivan 2009, 2016) and evidence as ‘telling’ something to those who draw the inference (Coulson and Pascual 2006, Pascual 2008a, 2008b). Consider this example about a piece of scientific evidence: (1)
I, strongly, strongly, believe that there was a period before the Big Bang, that the singularity was eliminated . . . To me the singularity is not an indication that there was a first moment of time, but that general
Fictive Interaction relativity is an incomplete theory. It’s general relativity shouting at us, screaming at us, I am not the end, there is more to understand. (Prof. Smolen, BBC Horizon, documentary ‘What Happened Before The Big Bang’ 2015, 22:36 min.)1
Theories may thus be conceptualized as conversational participants, fictively telling scholars where to look next. Moreover, the human body may not just be a source domain for a great number of conceptual metaphors (see e.g. Gibbs this volume Ch. 27), but it may also be a target domain, construed as a conversational partner, as in these common expressions in Catalan: (2)
a.
El cos diu prou. Lit. ‘The/My/Your body says stop.’ ‘X [the owner of the ‘speaking’ body] is exhausted.’ b. Córrer/marxar cames ajudeu-me. Lit. ‘to run/leave legs, help me’ ‘to run away very quickly and desperately’ c. Menjar a cor que` vols, cor que` desitges. Lit. ‘to eat at heart, what do you want[?], heart, what do you wish[?]’ ‘to eat as much as one wants/voraciously/indulgently’
Here, a given physical state (fatigue) or action (running and eating fast) is presented by either one’s body demonstrating that state through an order to bring it to a halt in (2a) or by addressing a body part associated with the action referred to with an imperative – to one’s legs for running in (2b) or a question – to one’s heart for willingness to eat in (2c). Even the unequivocally physical experience of dance is typically construed and spoken about in communicative terms, as “a conversation between body and soul,” a “dialogue between dancers,” or a means to “tell a story” to an audience. Amateur as well as professional dancers and choreographers often define dance as “physical dialogue” or “dialogue of movement,” which uses “corporeal body languages and vocabularies,” “messages” consisting of “words,” used more or less “eloquently” to produce “questions,” “responses”, or “contradictions” (Pascual and Brandt 2015). In this case, the sociocommunicative domain conceptually structures the physical domain (dance as conversation versus conversation as dance). The very idea of conversation being so basic to human existence as to model physical experiences challenges the prevailing paradigm in Cognitive Linguistics, according to which the physical domain is the constitutional ground for the social, mental, and communicative domains (Lakoff and Johnson 1980a, Sweetser 1990). The speech-act domain is at least as central to cognition (and thus the conceptualization of experience, language, and language use) as the physical and epistemic domains (see also P. Aa. Brandt 2004, Oakley 2009, 2014, Kövecses 2015).
1
Thanks to Sergeiy Sandler for drawing our attention to this example.
349
350
PA SCUA L A N D OA K L EY
21.3 Fictive Interaction as Discourse Structuring Device Let us begin by considering this example from Al Gore’s documentary film An Inconvenient Truth (Turner 2010: 110): (3)
Future generations may well have occasion to ask themselves, “What were our parents thinking? Why didn’t they wake up when they had the chance?” We have to hear that question from them now.
In this example, Gore’s argument is presented through the image of yet-tobe-born generations in collective excoriation in the form of a rhetorical question. The audience is construed as coerced bystanders of this selfflagellation. Gore’s imagined dialogue between future and present generations is but one of innumerable examples of fictive interaction as a discursive strategy. It is a prevailing technique across cultures and historical time, from the ancient Chinese text Zhuangzi (4th century B C ), in which fictional characters, animals, nature, and even abstract concepts engage in philosophical debates (Xiang 2016), to Plato’s dialogues (4th century B C ), such as Xenophon’s conversations with Socrates (Bakhtin 1975 [1981], 1979 [1986]), not to mention Aesop’s fables, and many passages from the Hebrew Bible. Intersubjective structures such as fictive questions followed by their corresponding answers may also be used to organize a string of discourse. Take this example from Zhuangzi (see also Xiang and Pascual 2016). (4)
道行之而成,物谓之而然。恶乎然?然于然。恶乎不然?不然于不然。物固 有所然,物固有所可。无物不然,无物不可。 ‘Dao operates, and the given results follow; things receive names and are what they are. Why are they so? They are said to be so! Why are they not so? They are said to be not so! Things are so by themselves and have possibilities by themselves’ (Lin translation 1942).2
In (4) the author takes the role of both questioner and answerer, thereby voicing and responding to doubts or objections that may arise in the mind of potential readers, who may be vastly distant in time and space. Such dialogic structures, for which (4) epitomizes their use as a general argumentative strategy, are found in a large number of other influential ancient texts in different languages, such as the Bible (Miller 1996 [2003]), the Quran (Badarneh 2003), the Soliloquies by Augustine of Hippo (cf. Bakhtin 1963 [1984]), and Medieval witchcraft pamphlets (Chaemsaithong 2013, 2016). Non-genuine question–answer pairs and similar interactional structures further appear to be often used to structure more recent written texts and oral monologues, such as contemporary philosophical discourse (L. Brandt 2008, 2013, Oakley 2009: ch. 3, Turner 2010); modern public discourse (Fairclough 1994, Vis, Sanders, 2
Thanks to Mingjan Wesley Xiang for this example.
Fictive Interaction
and Spooren 2012); classroom discourse (L. Brandt 2008, Oakley 2009: ch. 3, Oakley and Brandt 2009); and legal argumentation (Pascual 2002, 2006a, 2014, Chaemsaithong 2014, 2015, Oakley and Tobin 2014). In sum, the pervasive presence of dialogism in written as well as oral communication across time, culture, and languages suggests that some normative model of turn-taking is itself a generalized mental operation on a par with bodily experience.
21.3.1 Unmarked or Obligatory Fictive Question–Answer Pairs English fictive question–answer pairs may express topicalization (Li and Thompson 1976), as well as focus or conditionality (see overview in Pascual 2014: ch. 2). Examples are: (5)
a.
Ages 6 to 12? I’m a killer. (President Obama, Rolling Stone interview, 25 October 2012) b. . . . You want me to go to the Waverly ballroom? . . . Do you know what I’ll get for it? I’ll get heartache, that’s what I’ll get! I’ll get a big night of heartache! (Marty, teleplay by Paddy Chayefsky, 1953) c. You have enemies? Good. That means you’ve stood up for something in your life. (quote from Winston Churchill)
We find similar correlations of interrogative and topic, as in (5a), focus, as in (5b), and conditionals, as in (5c), in many other Indo-European languages, as well as in Mandarin Chinese and Korean. More interestingly, fictive question–answer pairs seem to be even the default or the only grammatical means to express these meanings in some languages without a written code (Pascual 2014: ch. 2), like the Papuan language Hua (Haiman 1978). It seems that in a large number of unrelated signed languages (which lack a written code but are generally used by literate signers), a question–answer scenario is also sufficiently entrenched as a representational resource for managing a wide range of other speech acts, thereby modeling information structure (Jarque 2016). Indeed, signed languages tend to use a non-actual interrogative mood as the most unmarked structure for topicalization, focalization, conditionality, connectivity, and relativization (Jarque 2016). A similar pattern as the one used by signed languages for relativization can also be observed in Latin (Lehmann 2008: 9): (6)
Cave tu idem faxis alii quod servi solent! beware you.NOM.SG ACC.SG.N:same do:PRF.SBJ:2SG other:NOM.PL REL: ACC.SG.N slave:NOM.PL use:3PL Lit. ‘Don’t you do the same, what do others do?’ ‘Don’t you do the same that the other slaves tend to do!’
In (6), the wh-interrogative ‘idem faxis alii’ (‘what do others do?’) constitutes a restrictive relative clause, indicating that the canonical declarative relative clause is modeled by an underlying question–answer structure. At the
351
352
PA SCUA L A N D OA K L EY
very least, this suggests that questions are not simple derivatives of declaratives (which was the basis of early transformational grammars). Rather, the unmarked question–answer structure of the above example suggests that the workings of a default interactional template between two persons with unequal access to shareable information comprises a conceptual primitive of language structure, use, and acquisition. It should be noted that there are similarities between wh-questions and relative clauses in more unrelated languages (see overview in Pascual 2014). In fact, interrogative markers are one of the most common sources for relative markers in the languages of the world (Heine and Kuteva 2002). Qualitative research on language pathology offers further evidence for the claim that the frame of the conversation may be projected onto the ongoing discourse, thereby serving to structure it. It is for example not unusual for individuals suffering from Broca’s aphasia to use the question–answer template for topicalization, focalization, and conditionality (e.g. Beeke, Maxim, and Wilkinson 2007). Take this exchange between patient and researcher (Versluis and Kleppa 2008 [2016]): (7)
PATIENT: INTERVIEWER: PATIENT: INTERVIEWER:
PATIENT:
E::::h muito calo::r? ‘Hmmm very hot?’ Mh! A` noite! ‘At night!’ Ah, tá. A` noite faz muito calor, lá? Ah! Quando faz muito calor de dia, o senhor anda à noite? ‘All right. At night is it very hot there? Ah! When it is hot during the day, you walk at night?’ A` noite. ‘At night.’
In this short piece of dialogue, the patient presents a conditional through a question–answer structure, as is clear in the interviewer’s paraphrase. Thus, when grammar and vocabulary fail, speakers with Broca’s aphasia tend to rely heavily on the scaffolding of face-of-face conversation, namely paralinguistic information as well as the interlocutor’s contributions, as compensatory strategy. Since speakers with Broca’s aphasia have their communicative competence intact (intonation, gesture, turntaking), despite often severe grammar and vocabulary difficulties, they organize their discourse through the conversational frame, partly using the communicative performance of their interlocutors as part of their discourse.
21.3.2 Sentential Fictive Interaction Fictive interaction is evident at the sentential level of speech acts (Langacker 1999b), such as greetings, assertions and negations, apologies,
Fictive Interaction
commands, as well as in a range of instances typically analyzed under the aegis of politeness theory (Brown and Levinson 1987).
21.3.2.1 Fictive Greetings Conventional means of opening and closing encounters (the prototypical conversational feature) is a natural way of framing a situation that is not, in itself, an actual conversation. Consider this popular 1957 song originally written by the Everly Brothers, which ranked 210th on Rolling Stone magazine’s list of ‘The 500 Greatest Songs of All Time’: (8)
Bye bye love/Bye bye happiness/Hello loneliness/I think I’m-a gonna cry-y Bye bye love/Bye bye sweet caress/Hello emptiness/I feel like I could die Bye bye my love goodbye [. . .] Goodbye to romance that might have been. . . Bye bye love/Bye bye happiness/Hello loneliness/I think I’m-a gonna cry-y/Bye bye love/Bye bye sweet caress/Hello emptiness/I feel like I could die/Bye bye my love goodbye
Here, a romantic break-up and the heartache it entails is not presented in spatial terms as ego inhabiting a given container (being in versus out of love or a relationship) or as a journey from one state/space to another, but in interactional terms. Naturally, as opposed to the narrator’s lover in ‘Bye bye my love, goodbye,’ caresses, romance, or inner feelings cannot be actual addressees of an interaction. That non-verbal entities can fictively greet as well as receive greetings, is illustrated by an ad from the Dutch travel agency D-reizen: ‘July says hello! Say hello July and hello vacation’ (‘Juli zegt hallo! Zeg hallo juli en hallo vakantie’ in the original). It should be pointed out that in different languages, greetings – as well as politeness markers (‘please,’ ‘thank you’), for that matter – are used fictively to express strong irritation, disagreement, or anger (Pascual 2014: 44), as in ‘Hello! You play to win the game!’
21.3.2.2 Fictive Apologies It is not uncommon for non-actual apologies to be addressed to inanimate entities, as in this newspaper headline on the astronomical demotion of Pluto from planet to planetoid: (9)
So sorry, Pluto, you are not a real planet. (www.nwitimes.com, August 25, 2006)
The journalist ‘apologizing’ to Pluto in (9) is naturally not deluded enough to think that a celestial body can have knowledge of and care about how inhabitants of another celestial body categorize it, let alone that it can factually engage in conversation. The fictive apology is merely a discourse strategy, which was apparently considered an effective and appealing means of presenting the news, since it appeared in slightly different versions in different media and blogs. As Demeter (2011, 2016) shows in two large corpus studies, fictive apologies are not only the work of
353
354
PA SCUA L A N D OA K L EY
humorous journalism (see also Pascual 2014: 43), but also occur in serious discourses, such as quality periodicals and parliamentary debates, and in as diverse languages as English, Romanian, and Hungarian.
21.3.2.3 Fictive Affirmations/Negations Similar to the examples above, many non-conversational situations are often framed as ‘disagreeable’ with one party negating the intentions of another, as in: (10)
I’m having a bad hair day; I tried to curl it this morning, and it just said ‘No.’
Consider also the text in the commemorative card from old Belgium Queen Fabiola’s Funeral that reads (in four languages): (11)
Always say Yes to God
This is an example of the subscript that often appears embedded in an encouraging command and is frequently employed in argumentation and marketing in which deciding, taking a stand, or committing oneself to or against something is framed in terms of an implicit proposition to which one responds with a yes/no answer (Brandt and Pascual 2016).
21.3.2.4 Fictive Commands Yet another common occurrence of sentential FI is an apparent command to non-specific agents capable of inducing some affective state. Consider this headline from an advertorial encouraging travel lovers to continue flying despite recent terrorist attacks: (12)
Take That, Terrorism! 7 Ways to Travel Without Fear (YahooTravel, 20 November 2015, www.yahoo.com/travel/by-wendy-perrin-the-terroristattacks-in-paris-063849605.html)
Here, the headline exclamation, a fictive enunciation addressed at terrorism in general (as opposed to some specific terrorist group) puts the reader in the position of being able to ‘tell terrorists off’. While it is commonplace to expect certain speech acts to change social reality through performatives with the proper illocutionary force, we can also fictively construe root actions and events as preventative or at least empowering. In this case, the proposed seven ‘ways to travel without fear’ result in a speech act with the illocutionary force of defying terrorism. Other fictive commands have become conventionalized, like the Call me crazy/biased-construction (Pascual 2002: 180–82, 2014: 42). Fictive commands, just as fictive apologies, greetings, assertions, or negations, may also be used as head nouns or be embedded in a grammatical constituent, as in the ‘Sorry I’m Not Sorry’ sweaters or t-shirts, brands like ‘Tousle Me Softly’ products, Herbal Essences’ ‘Hello Hydration’ collection, or its equivalent counterpart, the ‘Bye Bye Dry’ skin
Fictive Interaction
protectant cream (cf. Brandt and Pascual 2016). This is the topic of the next section.
21.4 Fictive Interaction as Intra-sentential Grammatical Constituent Direct speech may be used fictively as a means to express a large number of meanings, typically mental or emotional states (Pascual 2002, 2006b, 2014). Consider the attested examples in (13), from Mandarin, Polish, Hebrew, and Dutch: (13)
看 完 保证 让 你 大呼 过瘾 的 影史 最 强 科幻片 (http://mt.sohu.com /20150723/n417393929.shtml) Lit. ‘The best Sci-fi movie in film history that surely will make you shout/say “(this is) so satisfying’’.’3 b. Wszystko jest nie tak, człowiek boi sie˛ jej o cokolwiek zapytac´, bo zaraz lodowe spojrzenie typu “jestem tobą rozczarowana”. Lit. ‘Everything is wrong, you are afraid to ask about anything because you may immediately get the ice-cold look of the type “you disappoint me”.’ (Królak 2008: 93, 2016: 242) c. [ ַּכמּותkamut] Lit. ‘how-much/how-many-ness’ [from ‘[ ’כמהkama]: ‘how many’/‘how much’] ‘quantity’ d. Tien regels voor een ‘ja-maar’-vrij-leven Lit. ‘Ten rules for a ‘yes-but’-free-life’ (subtitle of the book Ja-maar, wat als alles lukt?], ‘Yes-but, what if everything works out?’, by Berthold Gunster, Amsterdam: Bruna 2009) a.
These examples illustrate the use of a fictive enunciation (a declarative in 13a, 13b, and 13d, it being an answer to a previous fictive turn in (13d), and a question, rhetorical, or information-seeking in (13c), filling the grammatical slots of a clause, phrase, lexical item, and a morpheme, respectively. A cross-linguistic study based on bibliographic research and the examination of a self-gathered database involving a total of 73 languages, both spoken and signed, found no single language without fictive speech as grammatical construction (Pascual 2014: ch. 4). Across languages, fictive direct speech typically serves to express: mental and emotional states, states of affairs, purpose, causation (Pascual 2002, 2014), and in some cases even evidentiality or future tense (e.g. ‘The sky says “I rain”’ for ‘It will rain’ in Wari’, Everett 2011). Little as it has been studied, this constitutes a productive, ubiquitous construction in a vast number of languages of the world (Pascual 2006b, 2014), and one that is frequently
3
Thanks to Mingjan Wesley Xiang for this example.
355
356
PA SCUA L A N D OA K L EY
used as communicative strategy in language pathology (Versluis and Kleppa 2008 [2016]). There are naturally important differences between the manifestation of fictive interaction constructions in different languages. For instance, direct speech compounds and modifier-noun combinations such as ‘Not happy? Money back! guarantee’ or ‘Hi honey, I’m home happiness’ are possible and frequently used in Germanic languages, but are relatively rare in Mandarin Chinese (Xiang p.c.) and can only appear as phrases or clauses in Romance or Slavic languages. Intra-sentential fictive interaction is widespread across different sociolinguistic groups, being used professionally by the greatest writers (e.g. Shakespeare, Tolstoy, Dessing or Szymborska, see Pascual and Sandler 2016), to the most prominent linguists (e.g. Chomsky, Langacker, see Pascual 2006b), or orators (e.g. Clinton, Obama, see Pascual 2014), as well as by the modern youth (Streeck 2002). Furthermore, fictive speech is used as effective communicatively strategy in very different genres and settings, including but not restricted to: marketing (Brandt and Pascual 2016); legal argumentation (Pascual 2002, 2006a, 2014); journalistic discourse (Królak 2008, 2016, Demeter 2011, 2016, Chaemsaithong 2014); or political argumentation (Coulson and Oakley 2006, Cienki and Giansante 2014). Fictive direct speech also seems to be preferred to indirect speech or descriptive alternatives by adults with Broca’s aphasia (Bánréti 2010, Versluis and Kleppa 2008 [2016]) and by children suffering from autism (Dornelas and Pascual 2016). In language pathology, non-quotational speech is often used in conversation as an adaptation strategy to compensate for poor grammar and word-finding difficulties, as a means to express needs, describe mental and emotional states, and refer to individuals and things.
21.4.1 Grammaticalized Fictive Speech Similar to what we discussed regarding fictive question–answer pairs in section 21.3.1, there also seem to be differences in degree of grammaticalization of fictive speech related to orality. Intra-sentential fictive speech appears fully grammaticalized in many unrelated languages with no or a poorly used writing system, often being the only or the most unmarked construction for expressing certain meanings (Pascual 2014: ch. 4, Da˛browska 2015a). In particular, fictive speech is frequently unmarked in signed languages (Jarque and Pascual 2016) and obligatory for certain meanings in a large number of unrelated languages without (widespread) writing (e.g. Spronck 2016, van der Voort 2016). Take the following examples, illustrating the use of fictive direct speech to express mental and emotional states, and causation in Kombai (Papuan) and Manambu (Australian), respectively:
Fictive Interaction (14)
a.
Kharabumano khe fenemora ma-khe-y-e-ne be.astonished.3PL.NF he how do.3SG.NF-Q-TR-CONN-say.SG Lit. ‘They were astonished: how does he do this?’ ‘They were astonished because of the things he did.’ (de Vries 2010: 35) b. asayik wa-ku gra-na father+ D A T say- C O M P L E : S S cry-A C T . F O C +3F E M . S G S U B J . N Lit. ‘Having said [the baby] ‘because of father’ she is crying.’ ‘She is crying because of her father.’ (Aikhenvald 2008: 391)
With respect to (14a), Kombai uses the verb ‘say’ to cover a wide range of situations about reasons. De Vries concludes that in languages such as Kombai and Kwaza (Brazil), “fictive interaction constructions with direct speech have become unmarked and are obligatory in certain domains, fully integrated in the grammar, used by all speakers in all contexts, and highly frequent” (2010: 35). The presentation of emotional and mental states (including intentions, attempts, as well as reasons or purposes) through a fictive enunciation, is very common cross-linguistically. In some languages of the world, states of affairs are also most frequently or invariably expressed through fictive speech, either addressed at the discourse topic, as in the following example from Wari’ (ChapacuraWanham) or ascribed to the interlocutor as a commentary about the discourse topic, as in the example below from Kwaza (isolate): (15)
a.
Om ca pi ra na ne Mapco not:exist agr:n:real.pst/prs finish 2sg:real.fut consent 3 n corn Lit. ‘The corn does not consent (when it is told) “Be finished”’ ‘The corn will never finish’/‘There is a lot of corn’ (Everett 2003 [2008]: 395) b. eromutsa-xa-´te-xa-ta ´nãi-xa-re wrist-2-purp-2-csolike-2-int Lit. ‘Is it for you to (say) ‘you wear it on the wrist’?’ ‘Is it a wristwatch?’ (van der Voort 2002: 318)
Examples (14b), (15a) and (15b) show fictive interaction constructions can profile the speaker (fictive speech ascribed to a pre-linguistic baby to present the baby’s reasons for crying), the addressee (the inanimate corn refusing to obey an order to stop growing as a way to present its abundance), or the topic (what one would say about a watch as a means of categorizing it). That is to say, the conversational frame may be universal among languages and may be used to express different meanings and functions in different grammatical positions, but the particular facets of the frame vary greatly from one language or context to the other. Fictive speech may also structure evidentiality marking in signed languages (Jarque and Pascual 2015) as well as spoken languages (Spronck 2016). Perhaps more spectacularly, fictive speech may serve to express future tense, as in the unrelated languages of the Amazon Wari’ (Everett
357
358
PA SCUA L A N D OA K L EY
2003 [2008], 2011) and Aikanã, an isolate language of the Rondonia area (van der Voort 2016). Consider this example from the latter: (16)
Hari-t∑a-re-mia-e˜ [bathe-1p-FUT-2p-DEC] Lit. ‘You people (say): “we will take a bath”.’ ‘You people will take a bath.’ (van der Voort 2009, 2016)
In Aikanã the future tense is marked by pronominal reduplication, just as actual quotation and the expression of intentionality. Apparently, this structure is so ingrained in the language that it also appears in the syntax of the Brazilian Portuguese spoken by the Aikanã (van der Voort 2016). At an even further stage of grammaticalization are common instances of wh-question markers for relative or conditional pronouns (e.g. ‘whether’), questions for indefinite pronouns, certain discourse markers or phrases (e.g. Korean ‘n-ka when-(be)-prs-q’, lit. ‘when is it?’ for ‘some time’, Rhee 2004: 411–20); or verbs of communication grammaticalized into a complementizer or auxiliary verb (Güldeman and von Roncador 2002). Significantly, all these are fairly to very common cross-linguistically. To sum up, the conversation frame not only serves to model information structure at the level of discourse, but can also play a role in the formation of core grammatical and lexical categories.
21.4.2 Is Fictive Speech Always Constructed Speech? Tannen (1986) focuses considerable attention on instances of what she calls “constructed speech”, that is instances of “demonstrations” (à la Clark and Gerrig 1990) used for hypothetical or counterfactual quotes, pieces of speech that never occurred or would never occur, as well as for the presentation of mental and emotional states through a direct speech not representing a previously produced utterance. Much less attention has been paid to verbatim speech used fictively, namely the use of literal quotations from previous discourse for more than making mental contact with prior discourse. Take the following extract from an autobiographical novel: (17)
A woman in a Disneyland sweatshirt stood in the doorway taking pictures of my sink, and I intentionally bumped her arm so that the prints would come out blurry and undesirable. “Hey!’ she said. “Oh, ‘Hey’ yourself.” I was in a fever, and the only thing that mattered was this apartment. (Dress Your Family in Corduroy and Denim, by David Sedaris, New York and Boston: Back Bay Books, 2004: 173)
In (17), the literal repetition of an utterance in the preceding turn appears embedded in a construction with a different communicative function,
Fictive Interaction
now functioning as a comment on it (cf. Pascual 2002: 203, 219). Consider now: (18)
As for the Brazil pictures, to quote Dorothy in “The Wizard of Oz”: Toto, I’ve a feeling we’re not in Kansas any more!:-) (personal email, March 9, 2015) b. The ‘I do’ fear: Dealing with commitment phobia (http://www .shaaditimes.com/love/gender-relations/commitment-phobia-050920/1)
a.
The expression “Toto, I’ve a feeling we’re not in Kansas any more” in (18a) comes from a particular utterance in a work of fiction and has now become part of popular American culture to express vertiginous experiences. This fictional quotation even appears conventionalized as entry in an urban dictionary to define an unusual situation or state of affairs. The “‘I do’ fear” nominal in (18b) is a creative one, emerging from a formulaic expression associated with the wedding ritual, which serves to set up the whole marital status (compare with “I do ring,” “I do couple,” “I do dress,” “I do kiss,” etc., Pascual, Królak, and Janssen 2013: 354, Pascual 2014: 78, 79). Interestingly, children with Autistic Spectrum Disorder often use verbatim or paraphrased repetitions of previous interactions (or sociocommunicative formulae) as communicative strategy. Consider this piece of dialogue between a Brazilian twelve-year-old boy with severe autism and his therapist, playing a game in which the child needs to choose pictures of people and objects in a map of a city (Dornelas and Pascual 2016: 351): (19)
CHILD: THERAPIST: CHILD:
THERAPIST: CHILD:
Sino! ‘Bell!’ O sino fica aonde? ‘Where is the bell supposed to be?’ Respeitáve::l público (.) Com vocês o [melhor] tocador de sino de todos os tempo::s ‘Ladies and gentlemen, boys and girls! With you is the best bell player of all times!’ Ele trabalha no circo? ‘Does he work in the circus?’ Sim! ‘Yes!’
In order to explain that he wants to place the picture of a bell in a circus, the child enacts a fictive circus director introducing such a bell as in a circus act. The vocative “Respeitável público!” is only used in the context of the introductory speech of Portuguese-speaking circus directors, and is thus far from being a constructed instance, just as the “the best X of all times!” is part of the schema. Its close sociocultural association with the speech of circus directors is what renders it suitable as a metonymic means of setting up the entire circus frame. In all these cases an enunciation serves to express or refer to a non-conversational message or reality, while not being an instance of creative or ‘constructed’ speech.
359
360
PA SCUA L A N D OA K L EY
21.5 Conclusions In this chapter, we provide evidence to suggest that fictivity, originally solely applied to conceptual phenomena modeled by the physical world, as in fictive motion and change (see overview in Matlock this volume Ch. 29), may also reflect intersubjectivity, going above and beyond Langacker’s (1999b) ‘virtual speech acts’ to cover all levels of discourse and grammar, and being widespread across languages (spoken and signed), genres, and sociolinguistic groups. We hope to have shown that the conversation conceptualization of reality forms a pervasive means of structuring language at several layers of analysis. Not only is it a widespread rhetorical strategy for structuring whole chunks of discourse, but it also serves as a pervasive technique for expression at the sentential, intra-sentential, and morphological layers, thereby implicating that fictive interaction is a motivating force behind basic linguistic issues of topicalization, focalization, and even grammatical tense. More specifically, we argue that fictive interaction: (i) is a productive construction, highly widespread across different language families and modalities; (ii) is frequently used for a great number of meanings or functions in a wide range of genres and by speakers from different sociolinguistic backgrounds; and (iii) may be used as a communicative strategy by professionals in written and oral communication, as well as by speakers with language impairments in everyday conversation. In sum, fictive interaction may be a universal and effective construction reflecting the interactional nature of language. More generally, fictive interaction captures the sociocognitive and communicative facets of grammar, and should be regarded as a basic schema that influences the lexicon and grammar of any given language in multiple and predictable ways. We conclude that language does not just reflect cognition and embodiment, as assumed in cognitive linguistics, but also emerges from and reflects discourse (Givón 1979, Tomasello 2003, Rhee 2004), or more specifically, talk-in-interaction (Janssen 2007, Pascual 2014, Pascual and Sandler 2016). Language – from semiotics to grammar to discourse structure – mirrors the fact that human beings are ultra-social, gregarious primates that regularly interact and communicate with family, friends, enemies, and strangers, an unprecedented set of behaviors among biological agents (Seabright 2010). Language is thus a communication system that enables us to maintain interaction and constantly align with our addressee(s), since a conversational turn typically responds to some prior turn and anticipates a response based on the assumed knowledge and communicative intentions of the addressee(s). This primordial facet of language, and the conversational form it takes, becomes a model that partly serves to structure cognition, grammar, and discourse in fundamental ways.
22 Diachronic Approaches Alexander Bergs
This chapter discusses the role of cognitive factors in diachronic approaches to language, that is, in language change. It begins with a section which presents some more general principles from cognitive linguistics (such as analogy or iconicity) and how these relate to more general problems and phenomena in linguistic change (such as directionality). The following three sections concentrate on three major theoretical lines of research in the study of linguistic change, all of which have been substantially influenced by cognitive linguistics: grammaticalization, lexicalization, and constructionalization.
22.1 Cognitive Factors in Linguistic Change Cognitive linguistics is an inherently synchronic, present-day discipline, focusing on the cognitive underpinnings of language processing in production and perception. In most cases, it studies the actual linguistic behavior of speakers as it can be observed in interviews, experiments, tests, and the like. These, naturally, are not available in language history, so diachronic approaches in cognitive linguistics, or even “historical cognitive linguistics” (Winters, Tissari, and Allan 2010, Pentrel 2015) at first sight appear to be a strange or even contradictory ideas. And yet, language and linguistic structures have always been subject to variation and change. It therefore follows that not only synchronic phenomena should be related to cognitive principles and factors, but also their diachronic development (see, e.g., Labov 2010). In other words, if we accept that synchronic speaker behavior is at least in parts related to cognitive factors, then we must also assume that synchronic variation is somehow connected to these factors. And even though not all variation leads to change, all change is based on variation – which means that change, or linguistic diachrony, can and should also be linked to cognition.
362
BERGS
In the following, we will look at some more general factors and the role they play in linguistic change: analogy, iconicity, and frequency.
22.2 Analogy Analogy and analogical thinking have sometimes been described as “the warp and woof of cognition” (Antilla 2008) – that is, as one of the essentials, or even the core of cognition (Hoftstadter 2013). Consequently, cognitive science has identified analogy and analogical thinking (or reasoning) as factors in learning, problem solving, categorization, decision making, and creativity, among others. And analogy has also already been discussed as a major force in linguistic change at least since the neogrammarians, that is, long before the advent of cognitive linguistics. Osthoff and Brugmann (1886: XIII–IX), for example, explain that: Second: since it is clear that the association of forms, that is, the new creation of linguistic forms, through analogy plays a very important role in the life of modern languages, one can safely accept this kind of linguistic innovation for the older and the oldest periods, moreover, this explanatory principle is also to be used in the way as it is used to explain the linguistic phenomena in later periods, and it is not in the least remarkable, when analogical formations show up in the same degree or even more in the older and oldest periods than in the the younger and youngest. [my translation]
What does that mean in concrete terms? A very simple, but well-known example of analogical thinking is four-way, proportional analogy: hand ðXÞ=finger ðX1 Þ ¼ foot ðYÞ=toeðY1 Þ ðXÞ=ðX1 Þ ¼ ðYÞ=ðY1 Þ: This kind of analogy allows us to easily fill any gaps in the equation, provided we see the similarity between the elements compared: glove ðXÞ=hand ðX1 Þ ¼ sock ðYÞ=??ðY1 Þ where ðY1 Þ ¼ foot: Many documented morphological changes can be explained that way. For example, I-Umlaut in Old English would predict that the plural of book should be beek, like foot and feet. I-Umlaut existed because, in the pre-Old English period, we find regular plural inflection with -iz. This forms the context for I-Umlaut, that is, the fronting and raising of the previous vowel. Note that this change did not take place in the singular, since here we do not find the relevant phonetic context. Following a fronting and/or raising, the affected vowels were often unrounded. Later on, the inflectional suffix was reduced or lost, so that the phonetic conditioning environment was also lost. Yet, the shifted vowels now remained in their new place. Following independent changes such as the Great Vowel Shift
Diachronic Approaches
Table 22.1 I-Umlaut and analogy (Pre-)OE
I-Umlaut
unrounding
Loss of I-Umlaut
GVS
PdE
environment *fo:t *fo:t-iz (plural) *bo:k *bo:k-iz (plural)
fo:t fø:t-iz bo:k bø:k-iz
fo:t fe:t-iz bo:k be:k-iz
fot fe:t bo:k be:k
fu:t fi:t bu:k *bi:k
fʊt
bʊk
but correct: bʊks
(GVS) and other minor changes, this led to the present-day forms. This is summarized in Table 22.1. So the prediction of I-Umlaut would be the modern plural beek instead of books. This is obviously not the case. But in the late Old English and early Middle English period, we find evidence for be(e)k as the plural of book (1): (1)
He . . .þe [fif] bec wende Genesis, Exodus, Vtronomius, ‘He the [five] books translated: Genesis, Exodus, Deuteronomy, Numerus, Leuiticus Numbers, Leviticus.’ (c. 1225 Homily: Sicut Oves absque Pastore; Worcester F.174, OED s.v. book)
This suggests that there must have some variation and competition between the seemingly irregular (but highly predictable) beek and the regular ‘new’ form books, which was formed in analogy to other ‘regular’ plural forms. This regular form is also first documented in the Middle English period, but quickly gained ground and became the dominant form in the Middle English period. Interestingly, this analogical change affected book but not other words which could also have been changed, such as foot – even though we find the occasional regular plural for foot (i.e. *foots) in Old English (2): (2)
þa scancan scyllum biweaxen, The legs scales (are with) covered, ‘Copper scales tinge the legs and fallow feet’ (OE Phoenix 311, OED s.v. foot)
fealwe fotas yellow feet
With foot, eventually, the seemingly irregular form feet won and became the dominant form for present-day English. What this shows is that analogical thinking and analogical formations of this kind are not random as such (since each of these formations is entirely regular and predictable once it kicks in), but that it is impossible to predict which analogy will actually survive the test of time. Why did book become regular, while foot stayed irregular? This problem lies at the heart of what has come to be known as Sturtevant’s Paradox: “Phonetic laws are regular but produce irregularities [foot-feet; AB]. Analogic creation is irregular but produces
363
364
BERGS
regularity [book-books; AB]” (Sturtevant 1947: 109). Note that although analogy can be characterized as an irregular process, it still seems to follow certain pathways. Kuryłowicz, for example, coined the idea of rain pipes and gutters. We can never be sure when exactly it will rain (i.e. when analogy strikes), but when it does we have pipes and gutters that influence the flow of the water in certain ways (i.e. when analogy strikes, it will do so in certain more or less predictable ways). Kuryłowicz formulated six laws of analogy that act like pipes and gutters in morphological change. Similarly, Man´czak also formulated nine tendencies for analogical changes in morphology (see Hock 1991 for a comprehensive outline and fair report on the widely debated status of Kuryłowicz’s laws and Man´czak’s tendencies). Analogy and analogical thinking go beyond the simple (but powerful) morphological four-way analogy described above. Analogy has also been implicated in metaphorical language use, in syntactic change, and as a major factor grammaticalization. Both metaphor and analogy are based on the recognition (or presentation) of (structural) similarities between two entities (cf. Sullivan this volume Ch. 24). A simple metaphor like The mind is a computer maps (or in some frameworks ‘blends’; see Oakely and Pascual this volume Ch. 26) certain elements and their relations from one source domain (computer) onto a target domain (mind). In doing so, the speaker creates an analogy between the two, as is easily visible when the word like is inserted: The mind is like a computer. We can also express these relations as individual four-way analogies: the RAM (Rapid Access Memory) relates to the computer like the working memory relates to the mind, the electrons relate to the computer like the neurons relate to the mind and so on (cf. Gentner et al. 2001). Some approaches have even suggested that analogy as the more general process subsumes metaphor as a more specific one. In any case, there seems to be a very close relationship between the concepts, and a cognitive principle such as analogy can help us to shed further light on the creation and development of a pervasive linguistic phenomenon like metaphor. Analogical thinking also plays a role in more complex syntactic change. For example, in present-day spoken German we witness some major changes in subordinated clauses introduced by weil ‘because.’ These clauses usually require verb-final order (the canonical word order for subordinate clauses in German), yet, when introduced by weil we find an increasing number of verb second-order clauses (the canonical word order for main clauses). This is briefly illustrated in (3) and (4). (3)
Ich konnte nicht kommen, weil ich krank war. I could not come because I sick was [canonical, verb final] ‘I could not come, because I was sick.’
Diachronic Approaches (4)
Ich konnte nicht kommen, weil ich war krank. I could not come because I was sick [V2] ‘I could not come, because I was sick.’
The order in (4) is mostly restricted to informal spoken German and is said to reflect a greater deal of subjectivity (Keller 1994). Its structure, however, could be due to analogical thinking, in that subordinate clauses are seen in analogy to main clauses, and thus receive V2 order. A second example comes from the history of English. The modal auxiliaries of present-day English (can, could, may, might, must, should, will) originated in the Middle English period. Before this time, most of them behaved more or less like regular lexical verbs, as in (5). (5)
Ge deliað and ne cunnon halige gewritu. You go astray and no can holy writings ‘You go astray and do not know/understand the holy writings.’ (Anglo Saxon, Gospel of Mathew xxii, MS. Corpus Christi College, Cambridge, No. CXL.)
Clearly, the verb cunnon ‘can’ in example (5) is not an auxiliary like modern English can. It is a full, lexical verb meaning ‘to know (something)’ or ‘to understand (something).’ It is only during the late Old English and early Middle English period that the group of so-called pre-modal verbs, including can, begins to develop. The interesting point is that these verbs did not all change at the same time (e.g. may seems to have been among the first to change, can and will seem to have been slower; cf. Traugott 1972: 198–99, Plank 1984). They gradually began to replace the subjunctive inflection, which disappeared from the language round about the same time. This change took place over a very long time, and throughout the Middle English period we find morphosyntactic variation in the sense that full lexical verbs appear side by side with modal-like verbs of the same root (cf. Traugott 1972: 198, Plank 1984, Fischer 2003 for a comprehensive overview). It is a plausible and interesting idea to think about this development in terms of analogy. On the one hand, it is clear that this is no simple fourway analogy like the one described above. On the other hand, however, the gradual developments noticeable here suggest that the basic mechanism for this could have been analogical thinking. Speakers perceived enough structural and functional similarities between two or more forms in order to treat them similarly, at least in some respects. So we witness the gradual emergence of the present-day NICE properties of modals in English: they take Negation, Invert in questions, are used in Code (e.g. question tags) and for Emphasis. Over time, and through frequency effects, we see more and more developments like these and finally the emergence of a new word class of modal auxiliaries, which is united through structural and functional resemblances between its members. While this is only one textbook example from the history of English, it is very interesting to think about the development of word classes in languages generally in terms of
365
366
BERGS
analogy – even more so since this is perfectly in line with recent findings in the study of frequency in language use and construction grammar (see below).
22.3 Iconicity Iconicity is the second major general factor to be discussed here. Simplifying a bit, iconicity is taken to mean that the signifier in some way actually resembles the signified (for a more detailed picture, see, e.g., de Cuypere 2008). This of course flies in the face of one of the most cherished linguistic axioms, the symbolic nature of language. Symbols, in contrast to icons, are related to what they signify through convention, rather than resemblance. Saussure himself stressed that the mental image of a tree, for example, is only arbitrarily and conventionally linked to the sound image [tri:] (or German [baum], or French [arbrə], or Mohawk [o´:kwire]). This fact, in some sense, forms part of the motivation for developing the idea of language as an autonomous cognitive faculty. Iconicity, in contrast, claims that language can be linked in interesting ways to the language-external world, and its perception and processing through more general cognitive principles (such as gestalt psychology). To use a very simple example: when speakers perceive a ‘more’ in the outside world, such as several entities rather than one, they may want to reflect that in their language – for example, through plural marking. Plural morphology, in turn, tends to be semiotically (phonetically or graphemically) ‘more’ than singular morphology: English tree – trees, German Baum – Ba¨ume, French arbre – arbres, Mohawk o´:kwire – o´:kwire’shon:’a. Needless to say, this is only a tendency, that is, there are a couple of exceptions to this principle (sheep – sheep etc.). Nevertheless, it is amazing to see that the majority of the world’s languages have plural forms that are substantially ‘longer’ than the corresponding singular forms. Iconicity would claim that this is a reflection of the perception and processing of the outside world in language (other accounts would probably point out that from a systemtheoretic perspective having longer plurals also makes sense since plurals tend to be less frequently used than singulars and thus need less sign efficiency; for reasons of space, this idea cannot be pursued here; see Haspelmath 2005 for an interesting summary of the pros and cons). Current considerations of iconicity go beyond morphological naturalness. Several studies (e.g. Haiman 1985b) also point out that syntax might be subject to iconicity. Consider Caesar’s famous ‘Veni, vidi, vici’ (‘I came, I saw, I conquered’). This paratactic structure obviously reflects real line events in its order, that is, Caesar first came, then saw, then conquered – even though this is not necessarily what this utterance says. Similarly, an utterance like ‘they had a baby and got married’ suggests that they first had the baby and then got married, even though this is not explicitly stated
Diachronic Approaches
in the sentence, since ‘and’ as a conjunction does not semantically specify any particular order. The iconic ordering of events in this utterance goes back to pragmatic (relevance theoretic) principles, that is, the iconic interpretation is encoded and decoded on the basis of pragmatic principles in reflection of the outside world. Similarly, there seems to be a tendency to group together linguistic elements that also belong together conceptually or in the external world. In this sense, syntax is perhaps not as autonomous as some theoretical accounts would claim. Rather, like morphology, it can reflect the outside world and how we perceive it. This has consequences, for instance, for the development of word order in English. Old English, as a more synthetic language, allowed for more or less free word order, like Latin. With the dramatic reduction of inflectional morphology in late Old English and early Middle English we witness a fixation of word order in order to be able to signal syntactic functions. This fixation of present-day SVO order in English could have been introduced through a transitional phase in which a word order such as SXV (verb final) turned into TVX (topic-verb-complement) and ultimately SVX (Vennemann 1974; cf. Meurmann-Solin, Los, and Lo´pez-Couso 2012). The TVX phase is motivated by the principle of iconicity: while ‘subject’ is a strictly grammatical term, ‘topic” reflects the actual information structure of the message, beginning with what was most salient in the situation that was being reported. Iconicity requires that the most salient element in a message should come in the most salient position, and this is initially. So, in a nutshell, iconicity as a cognitive principle can not only play an important role for synchronic morphology and syntax, but also can also give us some clues about diachronic aspects, including the directionality of change (toward greater iconicity as a rule of thumb).
22.4 Frequency The third cognitive factor that is intricately tied up with linguistic change is the role of frequency. Ever since Weinreich, Labov, and Herzog’s seminal 1968 article, language variation and linguistic change have come to be viewed as inseparable. Language variation, in turn, is ultimately also a matter of frequency. If there are two alternative forms, A and B, one of the two is usually the more frequent one. Cognitive psychology has spent a considerable amount of energy and time studying the effect of exposure and frequency of exposure. It has identified two potentially conflicting factors: saliency on the one hand, and high frequency on the other. Apparently, the (human) mind is susceptible to both forms of input, that is, there is a tendency to notice, remember, and perhaps even re-use both salient (albeit rare) forms and less salient, but common forms, which are encountered at a high frequency. And sometimes both factors coincide.
367
BERGS
Percentage
368
Time Figure 22.1 The S-curve of diffusion
This is part of the explanation for why very frequent verbs like to be can maintain their often irregular forms. Their occurrence is so frequent that the mind can easily store even irregular forms. Vice versa, this also explains why low frequency items are often subject to analogy and get regularized. German used to have irregular plurals for words like Lexikon ‘dictionary’ (pl. Lexika) and Atlas ‘atlas’ (pl. Atlanten) – which now also have acceptable regular alternatives Lexikons and Atlasse. Only less than half of the approximately 330 strong verbs in Old English still survive in presentday English, mainly because many were turned into weak verbs due to analogy. Research in frequency effects has demonstrated that frequency of exposure is one of the major factors in linguistic change (Bybee 2007). In other words, the more often speaker/hearers are exposed to some particular form, the more likely they are to replicate this form, which in turn increases its overall frequency. This is the basis for the ever-present S-curve of diffusion (Figure 22.1). This S-curve shows one of the most pervasive patterns in cultural evolution, and it applies not only to linguistics, but also to economy, epidemiology, culture, and many other fields (cf. Denison 2010). Changes begin with one or more innovators. If a particular change catches on (and it is important to note that not all changes do catch on), it gets copied by so-called early adopters who help to get it ‘off the ground.’ This leads up to a statistical tipping point (the early steep part of the S-curve) where the change begins to gain ground very quickly and a number of people adopt the new form. This characterizes the very steep middle part with lots of new speakers using the new form (via the mechanism outlined above). After a certain time, the development slows down again (qua statistical saturation phenomena), until a late majority adopts the new form. The development ends with a few laggards and a sort of stop in the development, that is, the development approaching, but often not reaching 100 percent of the population. Note that the S-curve pattern not only applies to diffusion among speakers, but can also be found languageinternally; for example, in the lexical diffusion of sound change or of morphological forms (see Ogura 2012 for details).
Diachronic Approaches
All three aspects just discussed – analogy, iconicity, frequency – have been well documented in historical linguistics, and they have sparked some very productive analyses and insights into linguistic change. It is worthwhile to re-consider these in the light of cognitive psychology and cognitive linguistics. Doing so not only leads to a deeper and more critical understanding of these concepts and their pervasiveness, but also helps in exploring the phenomenon of directionality in linguistic change. Directionality can be investigated from a number of different perspectives. On the one hand, one can look at language-internal directionality, i.e. the way in which changes proceed through the linguistic system. From this point of view, all three factors can help to explain why changes have a strong tendency to happen in certain ways, but not in others. For example, why is it more common for languages to eradicate irregularity in favor of regular forms? Why do languages retain some irregular forms? Why is there a discernible tendency toward SO word order in English? All these are questions that can be tackled on the basis of analogy, iconicity, and frequency. On the other hand, a second aspect of directionality relates to speakers. The simple question is why individual speakers, sometimes even across multiple generations, all move in the same direction in linguistic change; for example, toward do-periphrasis and modal auxiliaries, or one more step along Jespersen’s cycle, or away from irregular verbs. Sapir (1921) described this as drift. The answer seems to lie both in language-internal (systemic) factors and in principles of cognitive psychology, as outlined above. On the basis of the uniformitarian principle (cf. Lass 1997, Bergs 2012b) one can assume that all speakers, even across generations, are subject to the same kind of cognitive constraints and principles (for a critical discussion, see Bergs and Pentrel 2015, Pentrel 2015). If this is indeed the case, it follows that speaker/hearers will construct similar analogies, will perceive the same sort of iconicity, and are subject to the same frequency effects as other speakers in their speech community, and perhaps even beyond.
22.5 Grammaticalization In this section, I will discuss grammaticalization, one of the most prominent approaches to linguistic change, from a cognitive linguistic point of view (see Brems and Hoffmann 2012 for a recent summary). Grammaticalization can be broadly defined as the development of more grammatical material out of less grammatical material. A textbook example would be the development of the English future construction going to/ gonna V out of a lexical verb go meaning physical movement. The typical cline that can be identified for this and other grammaticalization processes is shown in (6) below.
369
370
BERGS
(6)
content word > grammatical word > clitic > inflectional affix > zero (Hopper and Traugott 2003: 7)
These processes can be studied from at least two points of view. On the one hand, there are more formal or structural studies in grammaticalization (e.g. Lehmann 2015) that mostly focus on the structural changes involved (such as semantic and phonetic bleaching, increased obligatoriness, reduced scope). On the other hand, studies such as Traugott (1982) also focus on the context of grammaticalization and its functional, pragmatic aspects. She points out, for example, that grammaticalization can lead not only to more ‘grammatical’ material (in the traditional sense), but also to pragmatically enriched, intersubjective constructions via a process of subjectification. So, gonna today is not just a future marker in present-day English, but also marks subjective viewpoints as in (7). (7)
JOSH-ELLIOTT# (Off-camera) This is true. And I just wanna say I’ve never been less happy to be a man cause it means I have to stay on this side of the table today. LARA-SPENCER# (Off-camera) I’ll switch with you. JOSHELLIOTT# (Off-camera) Because right there, that’s gonna be Tyra Banks, everybody. And right here is gonna be Sherrod Small. (Inaudible). LARASPENCER# (Off-camera) Do you wanna, do you wanna switch? TYRA BANKS (CREATOR/PRODUCER/HOST OF ‘AMERICA’S TOP MODEL’) How you doing? JOSH-ELLIOTT# (Off-camera) No, I’m fine. No. All right.) (COCA, ABC, Toe to Toe; is the mirror your friend? 2012)
In (7) the utterance that’s gonna be Tyra Banks does not refer to any future event, as going to V usually does. Rather, it has some pragmatic, epistemic function and signals the speaker’s perception and surprise while at the same time presenting this to the audience (see Traugott 1995: 35 for a detailed account). This leads to two interesting questions from a cognitive point of view. On the one hand, we need to look at the whys and hows of the structural aspects of grammaticalization. On the other hand, we can also ask about the cognitive underpinnings of the functional or pragmatic perspective. We will first focus on the structural aspects, and exemplify these with the development of future gonna (as discussed in Hopper and Traugott 2003: 61). The key mechanisms here are reanalysis and analogy. Reanalysis refers to the analysis of syntagmatic structures by speaker/ hearers which differs in some way from the previous analysis. In a first step, the element to in a sentence like John is going to visit Bill is no longer seen as a preposition belonging to the purposive clause to visit Bill, but it is rather reanalyzed as belonging to the verb going: John is going [to visit Bill] > John [is going to] visit Bill. Note that this process can be motivated by metonymical aspects: in order to visit Bill, for example, one necessarily has to go there. In other words, going someplace is part and parcel of the act of visiting. Furthermore, it also implies that the visiting itself will be a future
Diachronic Approaches
event, since first the act of going must take place. This leads to a new underlying structure: going [to visit Bill] becomes [going to] visit Bill, with a complex verb-particle combination followed by a bare infinitive complement clause. Note that the status of the verbs involved also changes. After the reanalysis, going is no longer treated as a full verb, but becomes an auxiliary. At a first stage, this is usually followed by verbs of action. On the paradigmatic axis, we see analogy at work in the next step. Once reanalysis has taken place and the construction no longer necessarily has movement or purposive readings, other verbs can fill the verb slot following going to. So we also begin to find non-action verbs such as like, which do not allow for a movement or purposive reading, and that can now fill this slot by being seen as analogous to action verbs such as visit. Once we see verbs such as like, which are semantically incompatible with the original structure of go as a directional verb of movement, in this structure we have positive evidence (by actualization) that reanalysis must have taken place. This step then usually involves a massive increase in frequency for the construction, which in turn accelerates the process itself (see above). In a final step then, going to as an auxiliary (not as a full verb!) undergoes univerbation into gonna (again, a consequence of frequency). Hopper and Traugott (2003: 77–87) point out that the first reanalysis is motivated by metonymy and that the extension by analogy in the second step involves metaphor. Ferrari and Sweetser (2012) analyze the development of future going to/ gonna from a pragmatic, or cognitive-functional point of view. In their account, this development involves the addition of extra mental spaces to the original go plus infinitive construction. On the one hand, there is a new Future Focus Space, which is added to the Present Content Space, and a speaker’s Epistemic Assessment Space, above the Ground Base. So through the added possible inference that the “mover will reach GOAL in future and achieve purpose” (2012: 60), the purpose space which contains the content base: present turns into a future space with the content base; present which specifies that the event has not taken place yet, but will take place when the goal is achieved. Furthermore, the ground base is extended now to include information on the evidential status of the claim. Note that the latter is particularly important to specify the difference between the going to future and the will future. With the latter, the speaker commits himself to a statement about the future, but there is no assessment of the available data in the present (Ferrari and Sweetser 2012: 59). In sum, cognitive linguistics can enrich our understanding of grammaticalization both from a structural point of view (by highlighting important notions such as analogy, metaphor, metonymy, and the role of frequency) and from a cognitive-functional point of view, as shown by Traugott (1995) and Ferrari and Sweetser (2012), for example.
371
372
BERGS
22.6 Lexicalization Lexicalization is the development of new lexical material in a language (see Brinton 2012, Brinton and Traugott 2005 for an overview). This does not necessarily mean that this is the opposite process of grammaticalization (which is usually referred to as degrammaticalization). Lexicalization, rather, deals with a number of different sources for new lexical material, one of which can be degrammaticalization. The others include processes such as word-formation, univerbation, fusion, idiomatization, and separation: Word-formation Compounding Derivation Blending Back formation Loan translation Univerbation Fusion Idiomatization Separation Degrammaticalization
redneck, nationwide regifter, sidler neuticles < neuter+testicles, cyborg < cybernetic organism co-vary < co-variation, flake < flaky flea market from French marché aux puces, paper tiger from Chinese zhilaohu (紙老虎) bread-and-butter (approach), run-of-the-mill (shirt) God be with you > goodbye, scıˉrg˙ereˉfa> shirereve > sheriff spill the beans, shoot the breeze ism, ex up the ante, ifs and buts
Clearly, all of these examples show how new words can enter the lexicon of a language, or at least the mental lexicon of individual speakers. In other words, these processes may constitute the source for new words, which may then undergo lexicalization. Cognitive linguistics is concerned with lexicalization in so far as it can contribute a number of important aspects to a theory of lexicalization. A theory of schemas and prototypes (e.g. Langacker 1987), for example, can explain derivation and compounding (e.g. adjective-noun compounds). These, Ungerer argues (2007: 653ff.) following Langacker (1987), show striking parallels to syntactic phrases: the noun climber, for example, derived from climb plus suffix –er, shows a similar trajector (the entity that climbs) and landmark schema (movement upwards in and/or across a defined space) as the prepositional phrase across the table; the compound Christmas tree shows a similar composite structure (albeit with encyclopedic enrichment) as the adjective-noun combination tall tree. Similarly, conceptual integration (see Fauconnier and Turner 2002, Oakley and Pascual this volume Ch. 26) can also account for many morphological blends and compounds. From the two input spaces of computer and virus and a generic space we get the blended space of computer virus, which integrates many of the aspects of both input spaces, but may also add a few extra features, which were not present in either input space (such as the fact that computer viruses cause damage until they are removed; biological viruses do not necessarily cause permanent
Diachronic Approaches
problems – influenza can pass after a few weeks even if it is untreated). Similarly, blends such as cyborg can also be treated as the product of conceptual integration. For non-compositional (exocentric) compounds we can assume that they either gain their extra, non-compositional meaning in the act of creation (e.g. egghead) or that they start out more or less compositional at the beginning, but then become exocentric through processes such as metaphor and metonymy (e.g. redneck). Moreover, the study of the role of frequency (as a major factor in cognitive, usage-based linguistics) can also highlight important aspects of how new words are added to the lexicon. It seems highly plausible that this is a matter of either frequency of exposition or of saliency, as discussed above. New words which are encountered with a certain high frequency (such as computer virus, unfriend, Facebook) are obviously more likely to enter the inventory of a language (be lexicalized) than low frequency items (e.g. uncandidate, de-boyfriend). And yet, even low-frequency items can, in some cases, successfully be lexicalized, provided there is a certain ‘saliency’ associated with them. In German, the word unkapputbar ‘un-broken-able’ was coined by the Coca-Cola company in the late 1980s and was used between 1990 and 1992 to describe the new plastic bottles. Not only does this new word violate certain word formation rules (adjectives like ‘kaputt’ do not allow –bar ‘able’ as suffix), it should also be subject to semantic blocking, since there are already two well-established words to describe the fact: unzerstörbar ‘undestroyable’ and unzerbrechlich ‘unbreakable.’ And yet, this new word became lexicalized and is now even included in the Duden (the standard German dictionary). This may have to do with high amount of saliency that this word used to have, despite, or perhaps rather because, of its obvious rule-breaking form. Another new word, formed in analogy, is unplattbar ‘unflatable,’ which was used to advertise new bike tires. While the jury is still out on this one, I am fairly skeptical that it will be just as successful as its model from the 1990s.
22.7 Constructional Changes and Constructionalization The last ten years or so have seen the development of a new theory of syntactic change: constructional change or constructionalization (Bergs 2012a, Traugott and Trousdale 2013). These terms are firmly rooted in construction grammar (see Hoffmann this volume Chs. 18 and 19). Some of the key questions here are how constructions can change on the form or meaning side, or both. Traugott and Trousdale define the former (change only on one level) as constructional change, and the latter (concomitant change on both form and meaning side) as constructionalization. One example for constructional change is the fusion of future going to into gonna (also discussed earlier as grammaticalization in section 22.2). Going to can be analyzed as a construction in the technical sense, in that it
373
374
BERGS
conventionally associates a particular form with a particular meaning .1 The univerbation of going to into gonna in this case is only a change on the form side, which does not affect the meaning side in any way. Similarly, when the meaning of a word shifts (e.g. Old English quene ‘any woman’ to Modern English queen ‘royal female leader’), we also see constructional change, as the form side remains constant, whereas the meaning side shifts. In constructionalization, both form and meaning shift in tandem. Traugott (2015) shows that the overall development of going to can be seen as such a case of constructionalization. She argues (2015: 70ff.) that future BE going to did not simply originate from a purposive construction of some sort (as previously suggested by most accounts in grammaticalization) but rather through the unification of the micro-construction go with the Purpose Construction (with coreferential subject), the pre-progressive -ing construction, a purpose clause construction with to V, and optionally even a passive clause. Traugott suggests that we see an expansion of this in the seventeenth century due to the increased use of go in this construction (2015: 71). We now find intention to act at a later time as part of the pragmatic meaning of (rather than a purposive reading), whenever motion was not an available or likely reading. Note that this is also what Ferrari and Sweetser (2012) show in terms of mental spaces (see above). This, in some sense, led to a mismatch between Clause 1 (go) and Clause 2 (to V), which ultimately may have led to the reanalysis of the bi-clausal structure of this overall construction and the emergence of BE going to as an auxiliary in the eighteenth century. Note that constructionalizations such as this one may consist of several constructional changes (such as meaning shifts, changes in form, or unification with hitherto unavailable constructions). Eventually, however, we see a shift of both form and meaning in tandem. Another important aspect of construction grammar that should be mentioned here is its rootedness in language use. Constructions are defined as learned form-meaning pairs. This learning can take place at any point during a speaker’s lifetime, not just in childhood during first language acquisition. This means that change does not necessarily only take place during first language acquisition. The development of new constructions simply depends on the frequency of exposition. New words (which are, by definition, constructions) can be learned through a high token frequency. The more often speakers hear a word like truther (someone who tells the truth), the more likely they are to repeat it and integrate it into their constructicon. A high degree of type frequency, in turn, helps to build more abstract constructional schemata (e.g. the more often speakers hear because NP (‘I want to go there because food’) the more likely they are to develop a new semi-schematic construction. 1
Obviously, this is a somewhat simplified description; see Bergs 2010 for a fuller account and further references.
Diachronic Approaches
22.8 Summary and Conclusion The present chapter tried to show that, despite initial skepticism, cognitive linguistics and diachronic problems are actually compatible and that principles and factors from cognitive linguistics (such as analogy, iconicity, and frequency effects) can be very illuminating for the study of linguistic change. This was further illustrated by tracing the cognitive underpinnings of three major theories of language change: grammaticalization, lexicalization, and constructionalization. All of these implicitly or explicitly work with ideas from the core of cognitive science and cognitive linguistics. Something that the present chapter could not do, for reasons of space, is to reverse this perspective. The focus here was on the contribution of cognitive linguistics to the study of language change. But what can historical linguistics contribute to the study of cognition and language? It was pointed out at the beginning that one of the key assumptions of historical cognitive linguistics seems to be that speakers in the past essentially processed language in the same way as speakers in the present. But we need to say very clearly that this is only an assumption. Pentrel (2015) is one of the first studies to critically look into this problem. She develops a so-called ‘feedback loop’ between historical linguistics and cognitive linguistics, where both disciplines feed into each other. This is particularly interesting, according to Pentrel, when predictions derived from presentday studies in cognitive linguistics are not met in our study of historical data. It is to be hoped that future research will look further into this question, so that eventually we can arrive at a much fuller understanding of how speakers and hearers in the past processed linguistic material.
375
Part IV
Conceptual Mappings
23 Opening Commentary: Conceptual Mappings Eve Sweetser
Humans seem, universally and productively, to use labels from one frame or mental space to refer to another. English is by no means unusual in referring (1) metaphorically to the past as behind us (metaphoric use of a spatial label for a time concept), (2) metonymically to an organization by naming the location of its primary organizers (Hollywood, metonymic for the American ‘big’ film industry centered there), or (3) to a dramatic role by the actor’s name or vice versa (Daniel Radcliffe is still unaware that Hermione is attractive, in the second Harry Potter movie; or alternatively, Juliet’s sick today so the understudy will be performing). English is even more ordinary in uses like This little girl in the photo is my grandmother – which involves, as Fauconnier (1994, 1997) pointed out, an Identity function which treats a past little girl as ‘the same person’ as a later old person. These examples are unremarkable, and would be considered literal by most literary and linguistic analysts, but they rely on a meaning relation, called a mapping, which represents what has been described, in stylistics or rhetoric, as figurative language. Such usage has received much attention in cognitive linguistics work, and thus the very concept of what counts as literal and figurative has come to be re-defined. Broadly speaking, cognitive linguistics treats the boundary between literalness and figuration as a matter of the patterns of use of complex knowledge structures, called frames, or domains. Lakoff and Johnson’s work on conceptual metaphor (1980a, 1999) brought metaphoric mappings onto the theoretical center stage, where they have certainly remained. Reference problems started Fauconnier off on his investigation of other kinds of mappings, those between mental spaces. In his analysis, the sentence In the picture, the girl with green eyes has blue eyes is not contradictory because it maps information between two mental spaces – that of reality and that of the picture; one reading is that the real-world green-eyed girl has a blue-eyed depiction. Since then, the topic of mappings between domains, frames, or mental spaces has become
380
SWEETSER
an important subject of investigation. Fauconnier’s theory of Mental Spaces (1994, 1997) and Fauconnier and Turner’s Blending Theory (2002) have made it clear that metaphor is only one part of a basic and varied human ability to map from one space to another. How the mind lets one thing map to another, and in what varied ways that affects language, has become one of the central areas of cognitive linguistics work. Some of the differences between types of mappings identified depend on the nature of the proposed ‘projection’ – transfer of conceptual structure between domains. Metaphor is specifically a unidirectional mapping between domains, or more specifically between a Source frame and a Target frame (see Sullivan 2013a, Dancygier and Sweetser 2014). As Lakoff and Johnson have argued, metaphor is not similarity-based: one cannot say that there is something literally similar between objects’ location behind me, and the past-ness of an event. They have also made us aware that metaphoric mappings shape reasoning, not just reference: inferential patterns are mapped from the Source to the Target, so that if we are thinking about crime as a disease in one case and as a dangerous enemy in another, we reason about it differently (cf. Thibodeau and Boroditsky 2011 as one high-profile study on this). A disease needs a cure, a dangerous enemy needs to be fought or killed; this may suggest that crime needs treatment as a social problem, or punishment of criminals. Contrary to what much of the earlier work on metaphor proposes, in Conceptual Metaphor Theory, metaphoric mappings are seen as systematic and coherent, including not only inferential mappings, but what is sometimes called structure preservation; for example, in mapping Life onto a Day or a Year, the beginning of the Life will be necessarily mapped onto the beginning of the Day, and the end of the Life onto the end of the Day. It appears that aspectual and sequential structure is particularly well preserved in cross-frame mappings. Conceptual metaphor has also been found to be pervasive, and not always conscious or intentional. Whole areas of human cognition seem to be mostly metaphorically structured; it is hard for an English speaker to talk for long about Social Status or Group Membership without any metaphoric use of words referring to physical Height (high, low, upper) or Centrality and Containment. And cross-linguistically, spatial vocabulary seems omnipresent in references to Time. Some of this appears to be due to the cognitive pervasiveness of primary metaphors, motivated by experiential correlation of two frames in Primary Scenes (Grady 1997b, C. Johnson 1997) experienced regularly by young children during cognitive development. For example, small children co-experience Warmth (from a caregiver’s body) and Affection, resulting eventually in an adult metaphoric mapping where an affectionate person can be called warm, even when no correlation exists with physical warmth (the person could be hypothermic at the time).
Conceptual Mappings
This experiential grounding of metaphor has become an important argument in the discussions central to cognitive linguistic enterprise – those clarifying the connection between language and human cognition. Moreover, the various levels of embodied and social experience also motivate an understanding of conceptual metaphors as working at various interconnected levels of conceptual and linguistic complexity. A hypothesis of cognitive metaphor theory is that perhaps, ultimately, all complex metaphor systems can be motivated as combinatory products of multiple simpler metaphors – in fact, possibly all metaphors ultimately go back to primary metaphors. It is now often understood that metaphoric mappings are one particular subclass of inter-frame or inter-space mappings. The theory of Blending, or Conceptual Integration (see Fauconnier and Turner 2002), discusses unidirectional mappings between source and target as single-scope blends, but also makes clear that not all blends are single-scope. Indeed, in some cases, the productive result of the mapping between two spaces is the creation of a new space involving new frames – frames which may not have been present in either input space. One salient example is the Regatta blend (Fauconnier and Turner 2002): the crew of the modern catamaran Great America II understood themselves to be ‘racing’ the nineteenthcentury clipper ship Northern Light (which had set the then still-standing record time for passage around the Horn) during their attempt to set a new time record for the passage. The press coverage referred to them keeping “a lead.” Since neither ship’s passage was an actual race against other ships traveling the same course at the same time, only in the blend can there be a Race frame, since only in the blend are Northern Light and Great America II imagined to be co-present and co-temporal. The existence of the level of meaning that emerges on the basis of simpler conceptual structures being blended has opened up a new approach to the study of meaning, less focused on standard linguistic forms, but more interested in the nature of the creativity involved. The interest in emergence of new meanings is specific to the cognitive linguistics enterprise and has opened up broad areas of investigation, especially into realms of fiction, poetry, and art. As I noted above, structure-preservation has sometimes been argued to be a characteristic of metaphoric mappings. But the observation is equally true for metaphoric blends and multiscope non-metaphoric blends. We cannot map the start of Great America II’s trip onto the end of Northern Light’s trip to create a race, any more than we can map plant sprouting onto human death to build a metaphor. By this analysis, metaphor is one kind of – either conventional or creative – blending, a mapping between frame-structured mental spaces. As this suggests, the types of mappings identified are one direction of investigation, but it is equally important to search for general mechanisms, characteristic of various kinds of mappings.
381
382
SWEETSER
For comparison, the figure known as metonymy is a mapping within the same frame rather than between frames: mention of one frame-element evokes another, or the larger frame (see Panther and Radden 1999, Panther, Thornburg, and Barcelona 2009, Dancygier and Sweetser 2014). It is based in experiential correlation. In the case of categorial metonymy, the relationship is between a prototypical subcategory and a larger category. In many European languages, for example, the same word is used for the oak genus Quercus and the species Quercus quercus (the ‘European oak’); European oaks were common and correlated highly with experience of oaks. Frame metonymy, on the other hand, allows one element of a frame to stand for others or for the frame. Unlike metaphor, metonymy can be multidirectional; it seems that the crucial factor is reliably evoking the metonymically labeled (and experientially correlated) entity via the label for the frame-associated entity. The New York Times can thus refer to a physical copy of the newspaper in question, to a news website, to the company which produces the physical paper and the website, to the editorial board of that company (The NYT opposes the Death Penalty), and even to the building in Manhattan where the company’s headquarters is lodged. As many researchers have pointed out, primary metaphors may have their roots in experiential correlation – but they transcend it, so that we say prices rose when there was in fact no literal upwards motion. Metonymic mappings have no such roots in metaphor, but since frame structure is based in experiential correlation, they can be thought about in terms of experiential grounding as well. So one way of thinking about mappings is that they all start in experiential correlation. Metaphor, metonymy, and now also multiple-scope blends do not exhaust the possibilities of mappings. For example, one very crucial aspect of mental space structure, brought to prominence in Fauconnier (1985, 1997) as well as by the literature on Frame Semantics (Fillmore 1982, 1985a; see Bergen and Chang 2005 for an implemented integration of frame semantics into grammar), is the mapping between roles and values, or fillers. We can see this as a cross-space mapping, in the case (for example) of theatrical roles: there is a base space containing people (actors), and a play space containing characters. As Fauconnier remarked, this mapping allows label transference in either direction. In the movies of The Hobbit, Benedict Cumberbatch played the role of Smaug the Dragon, and Martin Freeman the role of Bilbo Baggins; we can thus say, Wow, I met Smaug! if we shake Cumberbatch’s hand, or Benedict Cumberbatch terrorizes Martin Freeman (or Bilbo). In other cases, we might see it as a within-frame mapping: a frame like Mother necessarily involves a filler or value for the role of mother, and one for the role of child. Thus, Oedipus’ mother is a role; Jocasta and Merope are possible individual fillers of the role, in different mental spaces.
Conceptual Mappings
Perhaps the most fascinating cross-space mapping is that of Identity. Via very complex (and very backgrounded) models of personhood, temporal sequence, and causal structure, we quite automatically construe a person’s baby self and her 90-year-old self as ‘the same’ person. Here’s my mom at age two makes perfect sense as a description of a photo, using a current roledescriptor (not appropriate to the baby) to refer to the past entity pictured. These are remarkable sentences in their own right, and manifest some of the deepest underpinnings of our cognitive structure. How do humans do this? We do not fully know, though it must be interrelated with our general models of causation, time, and long-term object recognition or identification. But language certainly shows it saliently. Finally, a rather different category of mapping, much less discussed in linguistics, is that between shapes and more or less schematic images. We do such ‘matching’ constantly in visual processing: and it is the basis for crucial mappings both in metaphor and in counterpart relations. Calling a particular model of Volkswagen a Beetle or a Bug (cf. French Coccinelle ‘ladybug’) says nothing about either identity or metaphoric mappings – it refers solely to the overall mapping of the car shape onto the bug shape; Lakoff and Johnson (1980a) labeled this an image metaphor. Iconicity is a basic cognitive factor early recognized by Peirce (1932); iconic mappings between spaces were first systematically analyzed by Taub (2001). It is prominent in both language and other representational systems: they are what allows us to recognize the word meow as representing a cat noise, or a stick-figure drawing (or a Cubist portrait) as a representation of a person. However, visual processing and iconicity are also central to our understanding of how mappings may be represented, or constituted, through embodied aspects of communication such as gesture, and in linguistic systems such as signed languages. Again, like Identity, image mappings draw on the very base of human cognitive structure – our ability to see shapes, hear patterns of tones, and so on. Iconic mappings are extremely transparent to us – the stick figure seems ‘like’ the person. But in fact this is no more a simple resemblance than are metaphoric mappings. The study of signed languages, and more recently of co-speech gesture, has brought iconic structures into analytic prominence, and we can hope for much more examination of these kinds of mappings. As far as we know, all of these kinds of mappings are universal in humans and their languages. Categories and basic frame structures seem to be cognitively present in many species, although a non-linguistic animal cannot use a linguistic sign metonymically. But humans seem uniquely capable of certain levels of complex multispace mappings and blends, as well as of exploiting them for linguistic reference. The insights gleaned from the studies reviewed above paint a picture of both unity and diversity in mappings. Cognition is a network of interconnected mappings with varying cognitive underpinnings – between roles
383
384
SWEETSER
and values, between frame elements and frames, between input spaces and blended spaces, between time frames, and between images. Language reflects those mappings systematically in its varied use of labels. The directionality, or the systematicity, varies depending on the kind of mappings, because the cognitive bases of the mappings are different. Finally, the study of mappings connects linguistic usage to other forms of human activity, where analogy, iconicity, projection, and creativity are at the center of communicative behavior. This gives a truly new meaning to the idea of language as ‘generative’ – it is generative because cognition is generative: able to combine concepts, sometimes in new and varied ways, but always systematically. And, unlike some formal generative grammar models, this model accounts equally for aspects of language traditionally considered ‘figurative’ and ones which have been thought of as ‘literal.’
24 Conceptual Metaphor Karen Sullivan
Metaphor is widely viewed as a poetic way of using language. Writing guides define metaphor as “a kind of poetry” (Chakravarty and Boehme 2004) or a “figure of speech” (Johnson 1991), for example. Though this traditional view of metaphor remains prevalent, it is challenged by increasing evidence that metaphor is a conceptual phenomenon rather than a purely linguistic one. Three and a half decades ago linguists began questioning the traditional view of metaphor. Linguistic researchers, most famously Lakoff and Johnson (1980a), suggested that the metaphoric language in poetry and speech was in fact a by-product of metaphoric thought. On this view, English speakers call a difficult project a marathon not in order to be poetic, but because they actually think of the project as a long, difficult race, for example. If metaphor exists at a conceptual level, this explains why semantically related words often have similar metaphoric uses. For instance, many words related to self-propelled motion metaphorically describe projects. A project can be a marathon if it is difficult but a cakewalk if it is easy. A person can get moving on a project, then pick up speed, possibly get off course but be back on track in time to cross the finish line. If metaphor were just a matter of language, there would be no reason for more than one word related to motion, such as marathon, to have a metaphoric sense related to projects. Instead, the general concept of self-propelled motion seems to be connected to the concept of projects. In other words, there is a conceptual metaphor that allows human beings to think about, and describe, projects in terms of self-propelled motion. There is nothing unusual about the conceptual metaphor linking selfpropelled motion to projects. Human cognition seems to be full of metaphors, many of which are apparent in sets of words related to one concept, such as motion, that can all be applied to another concept, such as projects. To take another example, morality is metaphorically understood as cleanliness. People with dirty minds have immoral thoughts and those with filthy minds are even worse. Pure, clean-minded individuals with a conscience as
386
SULLIVAN
white as snow sometimes admonish the dirty thinkers to clean up their acts or cleanse the stain of their sins. There is no purely linguistic reason that clean, white, and pure should all metaphorically indicate morality, or that dirt and filth should come to mean ‘immoral.’ However, if human beings actually reason about morality as cleanliness, then it is natural that many words related to cleanliness will also metaphorically refer to morality. A conceptual view of metaphor allows us to make (at least) four predictions. First, even though metaphor has traditionally been considered a poetic device, we would expect metaphor to be apparent in everyday language as much as in poetic language. Second, metaphors should work in basically the same way in every human language, with some differences due to cultural influences and historical coincidences. Third, if metaphor is conceptual, it would be predicted to appear in visual communication as well as linguistic communication. Fourth, it could be expected to affect not only visual communication, but also all kinds of cognitive tasks that do not require language. In recent years, metaphor has been examined in all four areas mentioned above: in everyday language, cross-linguistically, in visual communication, and in other non-linguistic cognition. Results support the hypothesis that at least some metaphors are conceptual. The conceptual status of other metaphors remains unclear.
24.1 Evidence for Conceptual Metaphor Once metaphor was hypothesized to be conceptual, researchers began testing this hypothesis in various ways. The first kind of evidence in support of conceptual metaphor was the observation that the same metaphors appear in ordinary and literary language (discussed in 24.1.1). Soon afterwards, researchers began to look at metaphors across multiple languages (24.1.2). Finally, researchers looked at metaphors outside of language, first in visual communication (24.1.3) and then in psycholinguistic and psychological studies (24.1.4). Visual and psychological evidence of metaphor offer a strong argument that metaphor is conceptual rather than purely linguistic.
24.1.1 Metaphor in Everyday Language Metaphor, as a cognitive tool, could be expected to occur in ordinary language as well as literary language. Human beings have the same basic cognitive resources whether they are writing a poem, planning a meal, working on a relationship, or designing a nuclear reactor – even though the language they use in these situations is very different. Lakoff and Johnson’s influential 1980 text Metaphors We Live By showed that metaphor is by no means limited to literature or poetry. Everyday language is packed full of metaphors, which are often the same conceptual
Conceptual Metaphor
metaphors that had previously been observed in literary language. For example, Robert Frost’s popular poem ‘The Road Not Taken,’ which begins “Two roads diverged in a yellow wood,” describes a choice between two paths leading to two different destinations, which can be metaphorically interpreted as two life choices leading to two different life goals. Lakoff and Johnson (1980a) pointed out that ordinary language uses the same metaphors. For example, there is nothing particularly poetic about wondering where you are going in life, or contemplating whether you should follow an academic path or enter the political arena. Similarly, many famous poems describe the human lifetime as a natural cycle of light and heat, such as a year or a day. In this metaphor, youth is conceptualized as spring or morning, maturity is summer or noon, old age is autumn or evening, and death is winter or night. In Shakespeare’s Sonnet 73, for example, old age is represented by signs of autumn, such as falling leaves, bare branches, and cold; and also by the twilight that comes in the evening before night. Though Sonnet 73 demonstrates metaphoric complexity not typical of ordinary language (Lakoff and Turner 1989, Sweetser and Sullivan 2012), metaphors representing the human lifetime as a cycle of light and heat, such as a year or a day, are commonplace enough. For instance, everyday speakers talk about a December–May romance (between an older person and a much younger person); spring chickens (‘young people’); or the sunset years (‘old age’). The similarities between literary and everyday metaphor led Lakoff and Johnson (1980a) to conclude that the same conceptual metaphors underlie all human language use.
24.1.2 Metaphors Around the World Metaphor, to the extent that it is cognitive, should be found in every human culture and evidenced in every human language. Conceptual metaphor studies initially focused on English, but quickly spread to encompass other world languages and eventually a range of less-spoken languages. Though the cross-linguistic study of metaphor is almost as old as conceptual metaphor theory itself, it remains a vibrant research area. Even now, the metaphors in most of the world’s languages have not yet been documented, or have just recently begun to be examined (see Gaby and Sweetser this volume Ch. 39 ). One of the most obvious similarities across languages is that cognates in related languages often share metaphoric uses. In English, an intelligent person metaphorically emits light, as in bright student, a dim individual and brilliant person, all of which include adjectives of light-emission that modify nouns denoting humans. Metaphoric phrases such as a brilliant person can be translated directly into related languages while retaining the metaphor. For example, Spanish non-metaphoric brilliante ‘brilliant,’ like non-metaphoric English brilliant, refers to literal light emission, as in una luz brilliante ‘a brilliant light.’ When brilliante modifies a noun denoting humans, it
387
388
SULLIVAN
metaphorically refers to intellect, as in English. For example, una persona brilliante ‘a brilliant person’ means much the same as in English. However, even related languages show differences in their expression of metaphor. Spanish lacks a word for ‘low-level light emission’ with a metaphoric sense analogous to dim ‘stupid’ in English. Instead, Spanish has the expression tener pocas luces ‘to have few lights,’ which metaphorically means ‘to be stupid,’ much like none too bright in English. Spanish tener pocas luces evokes the same metaphoric concept as English dim, but the expressions involve no cognates or other linguistic similarities. The expressions are related cognitively but not linguistically. A purely linguistic account would consider these metaphors to be unconnected, but a conceptual theory of metaphor captures their underlying similarity. In fact, entirely unrelated languages may have the same underlying conceptual metaphors. For example, ‘anger’ is described in terms of a H O T F L U I D in languages as diverse as English, Japanese, Chinese, and Hungarian, though the metaphor demonstrates superficial differences across these languages (King 1989, Matsuki 1995, Ko¨vecses 1995, 2002). In English, you might use the H O T F L U I D metaphor if someone cuts you off in traffic during your morning commute, leading you to simmer about the incident. If your day does not improve, you might let off steam by complaining to your co-workers. If this fails, any little event might make your blood boil, and an angry email could have you steaming at the ears and might just make you explode. All of the italicized expressions describe a hot liquid in a container, such as a kettle. The contained liquid can be heated to a simmer, then to a boil, and then might release steam, boil over, or even explode. These stages of boiling are metaphorically applied to the stages of angry behavior. Once initiated, anger can remain mild (simmer), become more serious (boil), and may then result in behavior such as complaints, rage, or angry outbursts (exploding). These correspondences between anger and a hot fluid are relatively specific. Nevertheless, similar, though non-identical, correspondences are found in unrelated languages around the world. In English, the ‘hot fluid’ is contained in the head. Angry people steam at the ears, or blow their tops, and are visually represented with fire coming out of their heads. In Japanese, the H O T F L U I D metaphor is similar, except that the fluid boils inside of the belly rather than the head. In Hungarian, either the head or the entire body may ‘contain’ the fluid. In Chinese, the fluid can boil in various places, such as the belly, heart or spleen. Moreover, in Chinese the fluid is not actually hot, but is understood to be the “life energy” called qi (King 1989). Since Chinese culture has the concept of qi, it is used in the place of the more generic F L U I D conceptualized in other cultures. The metaphors in these languages are basically similar, but with variations, some of which are due to cultural influences. Many other metaphors are found across unrelated languages. For example, understanding a concept is often metaphorically conceptualized as
Conceptual Metaphor
seeing an object. In English, if someone is reading a difficult book, at first the topic might seem obscured by the opaque writing, and the reader might be in the dark. Eventually, the reader might glimpse the author’s intent and perceive the general shape of the argument. Finally the main point becomes crystal clear, and the examples shed light on some difficult questions. Here, ignorance is described as darkness or opacity, both of which impede visibility. Understanding is described as the ability to see. Factors that enable sight, such as clarity and brightness, metaphorically assist understanding. Again, a whole range of vocabulary related to one concept, vision, is here applied to another concept, understanding. This is the same conceptual metaphor underlying the earlier examples bright, dim, brilliant, and the rest, at the start of this section. Across the world’s languages, understanding is frequently described in terms of seeing. This metaphor has been documented throughout IndoEuropean languages (Sweetser 1990) and in over a hundred non-IndoEuropean languages (Haser 2003). However, the metaphor is not universal. For example, many Australian languages describe understanding as hearing, rather than seeing (Evans and Wilkins 2000, Gaby and Sweetser this volume Ch. 39). Evans and Wilkins (2000) attribute this to a cultural emphasis on auditory communication. It seems that every human language describes understanding as perceiving, though the perception may occur through seeing or hearing. The examples presented here are among dozens of metaphors that have been documented across unrelated languages around the world. All human languages have metaphors, and most seem to share some metaphors, including complex structures such as the H O T F L U I D anger metaphor described above. However, metaphors are shaped by cultural considerations, such as the concept of qi in Chinese and the importance of auditory information in Australian languages. Metaphors also evolve in language-specific ways that cannot be easily attributed to cultural factors, such as whether anger is ‘located’ in the head, the body or the stomach. Within-language differences, such as the various possible ‘locations’ of anger in Chinese, show that culture does not directly determine all cross-linguistic differences in metaphor systems. Some of these variations are arbitrary selections from a limited range of possibilities, such as container-like body parts (the head, the stomach, etc.). Despite these differences, the similarities across unrelated languages are too great to have arisen through chance. These similarities suggest that human experience and cognitive abilities form the foundation of metaphor systems in all human languages.
24.1.3 Visual Metaphors Metaphor seems to be found in all language use. Nevertheless, linguistic evidence alone cannot demonstrate that metaphor is conceptual rather
389
390
SULLIVAN
Figure 24.1 Angry man with steaming ears and exploding head
than linguistic. The most easily observed non-linguistic metaphors are visual, such as in artwork and other visual communication. Most of the metaphors in visual media can be ascribed to the same conceptual metaphors found in language. For example, the H O T F L U I D anger metaphor is frequent in cartoons, comics, and drawings in Western culture. In children’s cartoons, for example, characters frequently turn red when they become angry. Next, steam shoots out their ears, they may jump around like a boiling kettle, and finally the top of their head pops off in a fiery explosion (see Figure 24.1). These stages of angry behavior are almost identical to those in the metaphoric narrative about a bad day at work, described in the previous subsection. Another example familiar from Western cartoons is the lightbulb that appears over a character’s head, indicating an idea. This lightbulb sheds light on a topic, replacing ignorance with understanding, much as in the example of reading an obscure book described above. In general, the same conceptual metaphors surface in language, in comics (Narayan 2000), in fine art (Deignan 2005, Sullivan 2006b, 2009, 2016), film (Forceville 2002), and in the gestures that accompany speech (Cienki 1998, Cienki and Mu¨ller 2008a, Cooperrider and Goldin-Meadow this volume Ch. 8, Sweetser this volume Ch. 23).
Conceptual Metaphor
Some visual mediums affect the choice of particular metaphors. For example, comics excel at representing shadow and light, and so tend to exploit metaphors that relate to these themes (Narayan 2000). Visual metaphors range from the lightbulb example described above, to more subtle examples such as a darkened environment metaphorically representing a character’s lack of awareness. Both of these visual examples involve the same conceptual metaphor, by which ignorance is represented as darkness and understanding is conceptualized as seeing. Some metaphors that surface in visual communication are not found in language. For example, English speakers gesture to the left to metaphorically indicate the past, and to the right to refer to the future, following the left-to-right writing order in English (Fuhrman and Boroditsky 2010). This left-to-right depiction of time does not appear in spoken English. Speakers say the week before but not the week to the left, and speakers can leave the past behind but not leave it to the left, for example. Visual communication, like linguistic communication, is rife with metaphor. Moreover, the metaphors found in visual communication seem to be largely the same as those that appear in language. This supports the argument that metaphors are conceptual, not merely linguistic.
24.1.4 Psycholinguistic Evidence Metaphor, then, is a fundamental trait of visual and linguistic communication. At least some metaphors have also been shown to affect other aspects of cognition, such as reasoning, judgments, and memory. For example, feelings of affection are metaphorically understood as warm sensations, as in linguistic expressions such as icy stare or warm reception. This metaphoric correspondence appears to affect non-linguistic human behavior. For example, when people are asked to remember either an experience of social exclusion or an experience of social inclusion, the people who remember the negative exclusionary experience judge the temperature in a room to be colder than those who remember the positive experience (Zhong and Leonardelli 2008). Affectionate feelings are not just described as warmth, but also apparently make people feel warmer. When people play a computer game of ‘catch’ with an onscreen ball – supposedly with other players online, but actually pre-programmed to either include them in the game or to rarely throw them the ball – the players who are excluded are more likely to afterwards choose a hot drink over a cold one, using physical warmth to compensate for the apparent unfriendliness of the online players (Zhong and Leonardelli 2008). Numerous other studies provide evidence of a range of metaphoric correspondences. As mentioned in section 24.1, morality is metaphorically understood as cleanliness, and immorality as dirtiness. When people are asked to remember a time they behaved morally or a time they were immoral, and afterwards are offered a choice between a free pencil or
391
392
SULLIVAN
a hand wipe, the people who recalled their own immoral behavior are more likely to choose the hand wipe (Zhong and nquist 2006). Apparently, they feel that they can wash away their immorality, even though this moral stain was metaphoric. Other metaphoric correspondences that have been shown to have a psychological reality include the conceptualization of social status as being located in a high, upward position, as in high status or he’s under my supervision (Soriano and Valenzuela 2009); the conceptualization of importance as physical size, as in a big event or huge significance (Valenzuela and Soriano 2008); evil as dark and good as light (Frank and Gilovich 1988, Meier, Robinson, and Clore 2004, Pen˜a, Hancock, and Merola 2009; cf. Nickels 2007); intelligence as brightness (Sullivan 2013b, 2015); and various metaphoric conceptualizations of Time discussed in Part VI (Gaby and Sweetser this volume Ch. 39). No one knows how many other metaphors could be shown to have psychological validity, and much work remains to be done in this area. Nevertheless, it is apparent that some metaphors can affect human reasoning.
24.2 The Conceptual Metaphor Theory (CMT) Formalism Traditional Conceptual Metaphor Theory (CMT) as initiated by Lakoff and Johnson (1980a) provides a format for labeling and modeling metaphors. The CMT formalisms have the advantage of simplicity and a long history of usage. They remain popular today, alongside newer, more flexible formalisms such as Blending Theory (Grady, Oakley, and Coulson 1999, Fauconnier and Turner 2002, Oakley and Pascual this volume Ch. 26). Section 24.1.2 above gave various examples that described the comprehension of a concept as seeing an object, such as when a person is in the dark about a topic but then a brilliant idea sheds light on the issue. In these examples, the concept of understanding is represented as another concept, seeing. In CMT terms, the concept of seeing is called the source domain and understanding is the target domain of this metaphor. Cognitive structure is said to “map” from the source domain to the target domain, so that the target domain can be understood in terms of the source. Metaphors are named in the format ‘T A R G E T D O M A I N is S O U R C E D O M A I N ,’ usually in small caps, so that the metaphor mapping from seeing to understanding would be labeled U N D E R S T A N D I N G I S S E E I N G . Source and target domain names are typically represented in small caps even when not in a metaphor name, as in U N D E R S T A N D I N G or S E E I N G . The other metaphors mentioned in this chapter have similar names and notation. For instance, M O R A L I T Y is understood as C L E A N L I N E S S via the metaphor called M O R A L I T Y I S C L E A N L I N E S S ; a human L I F E T I M E is described as a L I G H T / H E A T C Y C L E via A L I F E T I M E I S A L I G H T / H E A T C Y C L E ; I M P O R T A N C E is conceptualized as S I Z E via
Conceptual Metaphor I M P O R T A N C E I S S I Z E ; and S T A T U S
is represented as physical height via S T A T U S
IS UP.
There is some variation in metaphor names in the literature. For example, I M P O R T A N C E I S S I Z E is sometimes called I M P O R T A N T I S B I G , and U N D E R S T A N D I N G I S S E E I N G may be labeled K N O W I N G I S S E E I N G . It is usually clear from context that the names represent the same conceptual metaphor. Variation in names is unfortunately sometimes taken as a license to invent new names for metaphors as a shortcut to analysis. For example, a student new to CMT might analyze the Prime Minister is a dimwit as involving P R I M E M I N I S T E R I S D I M . Of course, this new ‘metaphor’ is unnecessary – the linguistic expression involves the well-documented metaphor U N D E R S T A N D I N G I S S E E I N G , in which the absence of emitted light maps to stupidity. The unnecessary introduction of new cognitive metaphors runs contrary to the spirit of CMT, the goal of which is to generalize across instances of metaphor. CMT also provides methods for representing the structure that is mapped from the source domain to the target. One method is to label the correspondences within a metaphor, or ‘mappings,’ using the same format as a metaphor name. For example, the mapping T H I N K E R S A R E V I E W E R S in the metaphor K N O W I N G I S S E E I N G can be given a label that looks just like the name of the metaphor. Names of mappings may be listed under the metaphor title, as in the following: KNOWING IS SEEING THINKERS ARE VIEWERS IDEAS ARE OBJECTS INTELLIGIBILITY IS VISIBILITY SOURCES OF KNOWLEDGE ARE LIGHT SOURCES INTELLIGENCE IS LIGHT EMISSION
More often, though, each domain is represented as a circle, and mappings are connected by arrows between the domains, as in Figure 24.2. Note that the names of metaphors and mappings have the format T A R G E T I S S O U R C E , in which the target domain concept is first and source domain concept is second. In diagrams such as Figure 24.2, the source domain is instead given to the left (first) and the target domain to the right (second). In both the list and the diagram formats, source and target domain information must be kept separate. That is, in the list format, all mappings should give the target concept first and the source concept second, following the pattern T A R G E T I S S O U R C E . It would be incorrect to write O B J E C T S A R E I D E A S instead of I D E A S A R E O B J E C T S . In the diagram format as in Figure 24.2, each circle should contain only concepts from that domain. For example, ‘I D E A ’ should never be listed in the domain of ‘S E E I N G ,’ because I D E A S are related to U N D E R S T A N D I N G , not S E E I N G (see section 24.3 for a discussion of what material is included in a domain).
393
394
SULLIVAN
SEEING
UNDERSTANDING
VIEWER
THINKER
OBJECT
IDEA
VISIBILITY
INTELLIGIBILITY
LIGHT SOURCE
SOURCE OF KNOWLEDGE
LIGHT-EMISSION
Figure 24.2 A CMT diagram of U N D E R S T A N D I N G
INTELLIGENCE
IS SEEING
The CMT formalism has built-in limitations that can be either convenient or counterproductive, depending on the complexity of the structures that need to be represented. Most strikingly, CMT involves only two domains and unidirectional mappings from S O U R C E to T A R G E T . Many metaphors fit this template. However, metaphoric structures that are more complex cannot be fully captured in CMT. Other formalisms without CMT’s restrictions have been developed to represent a wider range of conceptual structures. Chief among these is Blending Theory, also called Conceptual Integration Theory (Grady, Oakley, and Coulson 1999, Fauconnier and Turner 2002). (This model is discussed in Oakley and Pascual this volume Ch. 26.)
24.3 From Embodied Experience to Complex Correspondence Metaphors are mostly unidirectional, meaning that they map in only one direction, from the source domain to the target domain. The domain of U N D E R S T A N D I N G is conceptualized as S E E I N G , but S E E I N G is not conceptualized as U N D E R S T A N D I N G . For instance, a thinker may be called bright to metaphorically convey the meaning ‘intelligent,’ but a lamp is never called intelligent with the intended meaning ‘bright.’ Similarly, an argument may be opaque (‘incomprehensible’), but a non-transparent container cannot be called incomprehensible (‘opaque’). Words with non-metaphoric meanings related to S E E I N G often metaphorically refer to U N D E R S T A N D I N G , whereas words with non-metaphoric meanings about U N D E R S T A N D I N G do not metaphorically refer to S E E I N G . In general, metaphors map from concrete bodily experiences, such as S E E I N G , to more abstract concepts, such as U N D E R S T A N D I N G . Notice how the other metaphors named in the previous section also have concrete source
Conceptual Metaphor
domains and abstract target domains: in M O R A L I T Y I S C L E A N L I N E S S , physical C L E A N L I N E S S is a more concrete concept than M O R A L I T Y ; in I M P O R T A N C E I S S I Z E , physical S I Z E is more concrete than I M P O R T A N C E ; and in S T A T U S I S U P , physical height (U P ) is more concrete than hierarchical S T A T U S . This pattern is typical of metaphors. The tendency to map from concrete to abstract may be what makes metaphors useful. Human beings use their bodies and senses every hour of their waking lives. These bodily and sensory experiences naturally build up complex cognitive structures related to space, movement, and sensory inputs. Abstract concepts such as M O R A L I T Y and I M P O R T A N C E arguably require more effort to understand than concepts that can be perceived with the senses, such as C L E A N L I N E S S and S I Z E . Metaphors allow the welldeveloped cognitive structures from the ‘easily’ understood concrete domains to be applied to the more ‘difficult’ abstract domains. For example, a small child already knows that people sometimes can only S E E objects with the help of a light source (such as the sun or a lamp). The child might not yet have learned that people can sometimes only U N D E R S T A N D concepts with the aid of a knowledge repository (such as a book or a teacher). The former concept in the domain of S E E I N G does not need to be taught, because children quickly learn from their own experience that they can see only when there is light. The second concept, related to U N D E R S T A N D I N G , is less intuitive. A metaphor such as U N D E R S T A N D I N G I S S E E I N G allows the rich first-hand information from a concrete domain, such as S E E I N G , to be applied to a more abstract domain, such as U N D E R S T A N D I N G . When reasoning from a source domain is applied to a target domain, this is called an entailment. In S E E I N G , light sources sometimes make objects visible. Transferred to U N D E R S T A N D I N G , this relation gives us the entailment that sources of knowledge (such as experts or books) make ideas comprehensible. Entailments make metaphor cognitively useful, because they provide new ways of thinking about the target domain. Metaphor is embodied in that it comes from the body. Metaphor takes cognitive structures built by physical, sensorimotor experiences, and maps these structures to more abstract domains. At some point in their development, children apparently acquire the ability to get ‘double duty’ out of their physical experiences by metaphorically mapping them to more abstract concepts. One theory that explains how this happens is the Conflation Hypothesis.
24.3.1 The Conflation Hypothesis Johnson (1997) argues that infants and young children do not yet consciously understand the difference between the concepts that later become the source and target domains of a metaphor. Small children conflate domains such as U N D E R S T A N D I N G and S E E I N G , and fail to recognize that they can understand things without seeing them (Johnson 1997).
395
396
SULLIVAN
Figure 24.3 When two containers hold the same volume of fluid, children will judge the taller container (here, the one on the left) to contain more fluid
Children conflate other source and target domains, such as U P and M O R E , the domains that form the basis of the metaphor M O R E I S U P in expressions such as stocks rose. Children’s failure to consciously differentiate M O R E and U P is illustrated by Piaget’s experiments (1972, 1990). As a child, I was a subject in an experiment along these lines. An experimenter came to our kindergarten class and showed us how he poured all the liquid from a tall, narrow beaker into a short, wide beaker (see Figure 24.3). The short beaker filled, but did not overflow. To us, this was a magic trick – we thought that some of the liquid had somehow disappeared. The second beaker was shorter than the first, so we assumed it held less liquid. The experimenter poured the liquid back into the tall beaker, and mysteriously the missing liquid was back again, filling the tall beaker. We thought that M O R E and U P were the same, so a taller beaker would always hold more fluid than a shorter beaker, regardless of width. My classmates and I were typical of kindergarten-aged children. Repeated experiments have shown that children consistently judge a container with a higher level of liquid to contain a greater quantity of the liquid, regardless of the width of the container (Piaget 1972, 1990). Later in life, children learn to differentiate domains such as M O R E and U P . The link between the domains becomes a metaphoric connection. When the concepts are conflated, any reasoning that applies to one concept necessarily applies to the other. Once the concepts are separate, then reasoning related to the more concrete concept can be selectively applied to the more abstract concept. Some degree of conflation seems to persist into adulthood. Even though adults would not be fooled by Piaget’s trick with the two beakers, adults
Conceptual Metaphor
still are more affected by height than width in judging volume (Pearson 1961, 1964, Frayman and Dawson 1981). Marketers can take advantage of this phenomenon by packaging products in taller boxes or jars, convincing consumers that the package contains a larger quantity of the product (Garber, Hyatt, and Boya 2009).
24.3.2 Primary and Complex Metaphor Conflation leads directly to the type of metaphor called primary metaphor. Primary metaphors are those that are based on real-world co-occurrences, such as M O R E and U P , or U N D E R S T A N D I N G and S E E I N G . These domains actually co-occur in the world and are conflated by children. The co-occurrence of two domains such as M O R E and U P is sometimes called a ‘primary scene’ (Grady 1997b, Grady and Johnson 2002). The domains M O R E and U P combine in a primary scene whenever we stack up books, add sand to a sandcastle, or throw clothes in a heap. As we add books, sand, or clothes, the stack becomes higher. Height (U P ) increases along with quantity (M O R E ). As adults, we know that this correlation is not universal, but we still find uses for the connection between M O R E and U P . This connection becomes the basis of the metaphor M O R E I S U P , as in stocks rose or funds are getting low. Because U P and M O R E co-occur in real-world primary scenes, when these concepts are distinguished and become source and target domains, the resultant metaphor M O R E I S U P is a primary metaphor. Many other metaphors evolve through primary scenes. For instance, I M P O R T A N C E and S I Z E often are correlated in the world. For children, their parents are both big and important. Big animals can hurt you more than small ones. A big piece of cake is more satisfying than a small one. Furthermore, visual perspective makes closer objects seem bigger, and closer objects are more likely to affect you. These real-world correspondences make I M P O R T A N C E I S S I Z E a primary metaphor. Similarly, children learn to associate A F F E C T I O N with W A R M T H through physical contact that indicates affection, such as when they are held by their parents or caretakers. The metaphor A F F E C T I O N I S W A R M T H , as in expressions such as warm welcome, is based on this co-occurrence and therefore is a primary metaphor. A metaphor may be extended from the initial primary correspondence and acquire complex mappings. These mappings connect elements in the source and target domains that do not themselves co-occur in the world, but which are related to elements that do. For example, human beings are not literally capable of emitting cold when they dislike someone, only warmth when they embrace or touch. Therefore the mapping L A C K O F A F F E C T I O N I S C O L D , as in chilly reception, is a complex mapping based on the primary metaphor A F F E C T I O N I S W A R M T H . Affection and warmth cooccur. Unfriendliness and chilliness do not, but are connected by
397
398
SULLIVAN
a mapping on the basis of analogy with the relation between affection and warmth. Complex mappings also arise when primary metaphors are combined into complex metaphors. Many well-studied metaphors are complex metaphors, such as the mapping from B U I L D I N G S to T H E O R I E S (as in the foundation of an argument), which is not a primary metaphor because theories and buildings do not tend to co-occur (Grady 1997a, Grady and Johnson 2002). However, certain aspects of buildings do co-occur with particular attributes of theories. For example, buildings stand erect for as long as they exist, much as trees, people, and piles of stones remain erect while they are alive or otherwise in existence. This primary correspondence is P E R S I S T I N G I S R E M A I N I N G E R E C T . This and other primary metaphors combine to form T H E O R I E S A R E B U I L D I N G S , as Grady and Johnson (2002) describe. Grady and Johnson also observe that complex metaphors have ‘gaps’ where no primary metaphors supply material. For example, they note that we do not say that theory has no windows via B U I L D I N G S A R E T H E O R I E S , because W I N D O W S are not a part of B U I L D I N G S that relate to P E R S I S T I N G I S R E M A I N I N G E R E C T or any other primary metaphor.
24.3.3 Image Schemas in Metaphor From birth, humans repeatedly experience certain spatial configurations using more than one of their senses. For example, babies see containers, feel them with their hands, and may experience themselves as containers for food. Babies see objects move from place to place, they hear noisy moving objects, they feel movement, and once they can crawl they move their bodies from one location to another. Experiences such as ‘Containment’ and ‘Movement’ are collected by multiple senses and then abstracted away from these senses, so that children can recognize Containment or Movement regardless of whether they see, feel or hear it. Abstractions such as Containment and Movement are called ‘image schemas.’ Image schemas are simple cognitive structures that represent spatial configurations independently of a single sensory modality. For example, the Containment schema represents a bounded region with an inside and an outside, which objects can enter and exit. The Path schema involves a trajectory that objects can follow when they move from one location to another. The Link schema includes two connected objects. Researchers disagree on how many image schemas should be recognized. Some definitions of image schemas include force-dynamic information (Talmy 1988a, 2000a, 2000b), whereas others are strictly limited to spatial configurations (Mandler and Paga´n Ca´novas 2014). However, there is general agreement that image schemas omit all information specific to one sensory modality, such as color, which can only be
Conceptual Metaphor
perceived visually. Image schemas have no specific size or shape, and belong to no particular time or place. This schematicity allows useful generalizations. For example, a tiny brown bottle held in the hand, and a huge red house seen in the distance, are equally recognized as containers. The Containment schema supplies the information that objects can enter and exit both the bottle and the house, that objects inside the bottle and house will move if the containers are moved, contained objects may be hidden, and other useful inferences.
24.3.4 The Invariance Principle Image schemas play a special role in CMT because they are preserved when metaphors map from a source domain to a target domain, insofar as allowed by the target-domain structure. The preservation of image schemas in metaphor is called the Invariance Principle (Lakoff 1993). As Lakoff describes, “What the Invariance Principle does is guarantee that, for container schemas, interiors will be mapped onto interiors, exteriors onto exteriors, and boundaries onto boundaries . . . and so on” (Lakoff 1993: 215). For example, when A N G E R is conceptualized as A H E A T E D F L U I D I N A C O N T A I N E R , the interior of the container maps to the inside of the head (in the English version of the metaphor). The hot fluid maps to the anger ‘inside’ the head. A F L U I D is carried everywhere its container moves. This generates the entailment that in the target domain, A N G E R is carried along wherever an angry person goes. A contained F L U I D may be hidden inside the container, but it can also exit the container. Similarly, A N G E R may not always be apparent, but it can ‘exit’ the container as angry behavior. The elements and relations in the Container schema are all mapped from A H E A T E D F L U I D I N A C O N T A I N E R to A N G E R . The invariance of image schemas in metaphor might be a relic of conflation. The factors that are ignored during conflation, such as the relative size of containers and contents, are much the same as the factors that are omitted from image schemas. When children conflate M O R E and U P in Piaget’s experiment described in section 24.3.1, they are ignoring the role of container shape, and also failing to consider that containers’ volumes must be greater than that of their contents. Shape and size are not part of image schemas, nor does the Container schema specify that a container must be larger than its contents (Mandler and Paga´n Ca´novas 2014). When children conflate U N D E R S T A N D I N G and S E E I N G , they ignore the distinction between vision and the other senses that are used to acquire knowledge. Image schemas also erase distinctions between the senses. In sum, image schemas and primary scenes attend to the same schematic information and omit the same types of information. This suggests that image schemas are active during conflation.
399
400
SULLIVAN
Image schemas appear to be acquired in the first months of life (Mandler and Paga´n Ca´novas 2014), long before conflation gives way to metaphor. It is possible that image schemas continue to connect two domains even once these have been differentiated into a source and a target domain. Whatever the origins of image-schema invariance, the schemas tend to be preserved in mappings from a source domain to a target domain.
24.4 Frames in Metaphor Semantic frames, like image schemas, are a type of structure based on real-world experiences. Like image schemas, frames seem to be preserved in metaphoric mappings. A frame is “a script-like conceptual structure that describes a particular type of situation, object, or event and the participants and props involved in it” (Ruppenhofer et al. 2010: 5). When frames describe events, for example, they specify who did what, with what instrument, at what time, and so on. Frames are often evoked by verbs, and some of their frame roles are roughly comparable to traditional thematic roles like Agent and Patient. In a simple sentence like Marc exercised, for example, Marc would be traditionally considered the Agent of the verb exercise. In the transitive sentence Marc exercised his biceps, Marc is again the Agent, and his biceps would traditionally be considered a Patient. However, semantic frames solve some of the problems with thematic roles. For example, frames explain why certain roles are sometimes present in language, and sometimes not. The verb exercise can be either intransitive (as in Marc exercised) or transitive (as in Marc exercised his biceps). Why can exercise optionally take an object? Frames answer this question by recognizing that the addition of his biceps does not change the meaning of exercise, because exercise necessarily requires moving the body or one of its parts, whether or not this information is linguistically expressed. That is, the body (or one of its parts) is an element of the conceptual structure of the Exercising frame. Semantic frames, then, capture the conceptual structure required to understand a verb, whether or not this structure is actually apparent in a given sentence. Frames often contain more conceptual elements than are apparent in language. These elements must exist cognitively even if they are not expressed linguistically. For example, in Liz took revenge, the word revenge evokes the Revenge frame.1 An act of revenge has to minimally involve certain elements, whether or not these are expressed in language. 1
The full structure of frames cited in this chapter may be viewed in the publically available data of the FrameNet project (FrameNet [2015], described in Ruppenhofer et al. [2010] and discussed in Boas [this volume Ch. 34] and David [this volume Ch. 35]). The discussion of the Revenge and Exercising frames in this section is adapted from Sullivan (2013a).
Conceptual Metaphor
Someone must, for example, commit the act of revenge (an Avenger; here, Liz). The revenge must be undertaken in response to some past offense (an Injury; not expressed in the sentence here); and someone has to have committed this injury (the Offender; also not expressed here). Without these elements, revenge cannot happen. For example, if Liz never suffered an Injury, the sentence Liz took revenge makes no sense. Similarly, if the Injury was an accident and no Offender can be blamed, Liz cannot take revenge. Further evidence for conceptual frame roles is that these roles always are available to be expressed in language, even when this is not required. For example, Liz took revenge on her husband for years of snoring includes the Injury (years of snoring) and the Offender (her husband). The Revenge frame is the same as in Liz took revenge, and the meaning of Revenge has not changed. Rather, some of the roles in Revenge are now directly mentioned in the sentence. Additionally, Liz took revenge on her husband for years of snoring tells us the fillers of some of the frame roles in Revenge. In a schematic structure such as the Revenge frame, roles like Avenger and Injury generalize over many potential situations and individuals. The Revenge frame itself does not tell us who the Avenger is, only that someone has this role. In specific instances of Revenge, frame roles may receive fillers specifying the identity of the Avenger, the type of Injury, and so forth. In the above sentence, Liz fills the Avenger role in the Revenge frame, her husband fills the Offender role, and years of snoring is the Injury. These roles may or may not be filled, but they are necessary to any instance of Revenge. Many words are interpretable only if we have some kind of access to frames and their elements. Returning to our previous example, the verb exercise only makes sense in terms of the frame of Exercising. This frame includes elements such as a person with a body (an Agent), effortful movement of the body (Means), and strengthening or otherwise improving the body (the Purpose of the effortful movement). Without any of these elements, the concept of Exercising is meaningless. On the other hand, when we hear the verb exercise, all of this structure is automatically available. If you tell me that “Marc exercised today,” I know already that there is a Purpose role in the exercise frame. I could ask, “Why did he do that?” You could respond, “He’s working on muscle tone,” thereby giving me the filler for the Purpose role in the Exercising frame. Since the Purpose role was already part of the frame evoked by Marc exercised today, the role is available to be filled later on in the conversation. One way to show how individual words evoke frames is to draw an arrow from the word to a representation of the conceptual frame. Frames are usually depicted as a list, often bordered by a rectangular box, as in Figure 24.4. In this diagram, the verb exercise evokes the Exercising
401
402
SULLIVAN
EXERCISING frame
AGENT exercise
BODY or BODY_PART MEANS PURPOSE ... etc.
Figure 24.4 The verb exercise evokes the Exercising frame
frame. The frame roles of Exercising are listed in a box representing the Exercising frame.
24.4.1 Frames and Domains The relation between frames and metaphors has been described in various ways. It has been proposed that the source and target domains of metaphors (collectively called ‘metaphor input domains’) include frames (Sullivan 2006a, 2013a, David this volume Ch. 35), that metaphor input domains designate the same structures as frames (Croft and Cruse 2004), that domains can themselves be considered frames (Moore 2011b), or that domains are ‘super frames’ that combine multiple frames (Materna 2010). The debate over frames and metaphor input domains is more a matter of terminology than of content. Metaphor input domains are clearly structures that can combine multiple small-scale frames. If the combinations of frames in metaphors can be demonstrated also to exist outside of metaphor, this would support labeling the combinations ‘frames,’ ‘big frames,’ or ‘super-frames’ or another more general label that does not contain the word ‘metaphor.’ If, however, these combinations are specific to metaphor, it would be more accurate to retain a label such as ‘metaphor input domain.’ Some frames that occur in metaphor clearly occur elsewhere. For example, several frames related to the physical body are interconnected both in metaphor and in general cognition. For example, the Exercising frame interacts with other frames analyzed by the FrameNet team which are accessible on their website, such as Ingestion, Observable_body_parts, and others. Although each of these frames contains its own roles and relations, the frames themselves are related, too. We know, for instance, that the Ingestion of beneficial foods, combined with Exercising, will strengthen and tone various Observable_body_parts, which in turn will permit more powerful force exertion (Physical_strength) and help maintain good health (avoiding Medical_conditions and Experience_bodily_harm). These frames are related even when they are used non-metaphorically.
Conceptual Metaphor
The relation between the frames is also apparent in metaphoric language. For instance, example (1) combines several of the above frames in the B O D Y domain, and maps these to the M I N D target domain. (1)
Brain training – finding ways to stretch and exercise your brain – . . . will boost your mental power and leave your brain sweating. (missiontolearn 2009. ‘Brain Training: 25+ Sites and Tools to Exercise Your Brain.’ Mission to Learn. Retrieved from www.missiontolearn.com/2009/ 07/brain-training-exercises/ [accessed August 2, 2015]; boldface mine)
Here, Exercising (evoked Body_Movement (stretch) is conceptualized as leading to improved Physical_Strength (power), with such intensity that it results in Excreting sweat (sweating). Although the brain is not literally capable of any these processes, the frames are carried over from the B O D Y domain. The relations between the frames are also mapped, so we know that the Exercising causes the increase in Physical_Strength and the Excreting of sweat. Example (2) maps from B O D Y to a different target domain, E C O N O M Y . Again, a range of frames are integrated in the B O D Y domain and map together to the target domain. For instance, (2) demonstrates the interrelation between frames such as Intoxication (the effects of drugs), Experience_bodily_harm (harmful), and Experiencer_obj, in which a Stimulus provokes a response in an Experiencer (stimulating). These frames are mapped by A N E C O N O M Y I S A B O D Y . In the E C O N O M Y target domain, the frames generate entailments about the positive and negative results of a currency devaluation. (2)
This devaluation had the effects of drugs: it could prove stimulating to the economic body in the short term – but it might be harmful to it in the longer term. (Vaknin, S. (2007). An Evaluation of the Devaluation [eBook]. Retrieved from http://samvak.tripod.com/nm024.html [accessed August 2, 2015]; boldface mine)
In examples such as (1) and (2), it matters which words are chosen to evoke the B O D Y domain, and which frames these particular words evoke. Frame structure evoked by particular items is ‘profiled,’ in Langacker’s sense (1990a), relative to the other structure in the source and target domains of the relevant metaphor. For instance, the item exercise in the phrase mental exercise evokes the frame of Exercising, so this frame is more crucial to understanding the phrase mental exercise than the other frames in the B O D Y domain. Profiled frame structure is more active than non-profiled (backgrounded) frame structure in generating inferences in a particular instance of a conceptual metaphor. Figure 24.5 represents the B O D Y domain as evoked by the metaphoric phrase mental exercise. This phrase profiles the Exercising frame in the B O D Y source domain, because the relevant aspect of the B O D Y is related to Exercising. The arrows in Figure 24.5 represent processes of evocation.
403
404
SULLIVAN
EXERCISING frame
exercise
AGENT BODY or BODY_PART MEANS PURPOSE ... etc.
BODY domain
EXERCISING frame (evoked by exercise): AGENT BODY or BODY_PART MEANS PURPOSE ... etc. ... etc.
Figure 24.5 The item exercise evokes the Exercising frame and the B O D Y domain
The item exercise evokes the Exercising frame, and once the phrase is recognized as metaphoric, the B O D Y domain is evoked via the Exercising frame, which forms part of the structure of the B O D Y domain. Other frames in B O D Y , such as Physical_Strength, are backgrounded relative to the profiled frame Exercising.
24.4.2 The Extended Invariance Principle The Invariance Principle, introduced in section 24.3.4, poses a restriction on metaphoric structure, in that “Metaphorical mappings preserve the cognitive topology (that is, the image-schema structure) of the source domain, in a way consistent with the inherent structure of the target domain” (Lakoff 1993: 215). The principle was formulated to account for the mapping of image schemas. Fortuitously, however, the principle is phrased in such a way that it can be easily extended to apply to frames as well as image schemas. If the definition of ‘cognitive topology’ includes frame structure as well as image-schema structure, then the preservation of frame elements and relations in metaphors such as T H E M I N D I S A B O D Y is predicted. This interpretation is called the ‘Extended Invariance Principle’ (Ruiz de Mendoza 1998, Sullivan 2013a). Indeed, metaphors preserve all aspects of frame structure, down to the specific relations between frame roles. In the Exercising frame, for example, the Body role must refer specifically to the Agent’s Body. This relation carries over when Exercising is mapped from the B O D Y domain into the M I N D domain via T H E M I N D I S A B O D Y . In the M I N D domain, the Mind that is ‘exercised’ must be specifically the Mind of the individual who is ‘exercising.’ The Purpose of Exercising is usually to improve the Body. In the M I N D domain, this relation persists, in that the Purpose of the relevant mental ‘exercise’ is to improve the Mind in some way. Relations between frames roles map unless they are inconsistent with material already in the target domain, as predicted by the Extended Invariance Principle. In sum, frames
Conceptual Metaphor
appear to play a role in metaphor that is as privileged and significant as that of image schemas.
24.5 Summary: How A N G E R Becomes A
HEATED FLUID
Conceptual metaphors arise from embodied experiences. Metaphors begin with two conflated domains, which are eventually disentangled into a source and a target domain, as outlined in section 24.3.1. These domains remain connected by a system of metaphoric mappings. The mappings allow the target domain, which is usually more abstract, to be understood in terms of the source domain, which is usually more concrete. For example, A N G E R I S H E A T begins with the simple sensation of increased body temperature induced by angry feelings. The experiences of heat and anger are first conflated, then distinguished as the domains A N G E R and H E A T , which are related by the metaphor A N G E R I S H E A T . A N G E R I S H E A T is a primary metaphor, because it is based directly on the co-occurrence of A N G E R and H E A T in real-life experience. Primary metaphors based directly on embodiment, such as A N G E R I S H E A T , may be extended through the addition of complex mappings or by the combination of multiple primary metaphors into a complex metaphor (as illustrated in section 24.3.2). For example, A N G E R I S H E A T combines with other primary metaphors, such as C A U S E S A R E P H Y S I C A L F O R C E S (by which the cause of anger is conceptualized as a heat source) and I D E A S A R E O B J E C T S (by which the angry condition is conceptualized as a physical substance), resulting in the complex metaphor A N G E R I S A H E A T E D F L U I D IN A CONTAINER. The structure of metaphors is to some extent predictable, in that all metaphors preserve image schemas and frames. For example, A N G E R I S A H E A T E D F L U I D I N A C O N T A I N E R retains the image schemas and frames from A H E A T E D F L U I D I N A C O N T A I N E R and maps these to A N G E R , as predicted by the Invariance Principle. For example, the inferences of the Containment schema are preserved in the target domain A N G E R . The metaphor also preserves frames, such as the Cause_change_of_phase frame, in which a Source of temperature change heats a Patient from an Initial_state with a Result change in phase (here, from liquid to gas). In A N G E R , the Source is conceptualized as the cause of the anger, the Patient is the anger itself, which passes from a fluid Initial_state to the boiling Result (angry behavior). The progression from Initial_state to Result, and other relations between elements in the Cause_change_of_phase frame, are mapped coherently to the domain of A N G E R . In A N G E R I S A H E A T E D F L U I D I N A C O N T A I N E R , as in other complex metaphors, the combined structures of primary metaphors, preserved image schemas, and frames result in a powerful cognitive tool for understanding an abstract concept – in this case, the emotional, behavioral, and social consequences of A N G E R .
405
406
SULLIVAN
Metaphors are apparent in all human languages, in visual communication, and in other non-linguistic cognition (sections 24.1.2, 24.1.3 and 24.1.4, respectively). The primary metaphor A N G E R I S H E A T may be universal, since all humans experience increased body temperature when angry. The complex metaphor A N G E R I S A H E A T E D F L U I D I N A C O N T A I N E R , while not universal, is found across a range of unrelated languages (section 24.1.2). The metaphor appears visually, such as when cartoon characters steam at the ears or their heads explode with rage (Figure 24.1). There is evidence that A N G E R I S A H E A T E D F L U I D I N A C O N T A I N E R affects cognition (Nayak and Gibbs 1990, Gibbs 1994), though more work is needed to demonstrate that the metaphor can be active even when no language is involved. The embodied origins of complex metaphors such as A N G E R I S A H E A T E D F L U I D I N A C O N T A I N E R are not obvious when we read descriptions of physically impossible scenes, such as he was steaming at the ears, nor when we see physically impossible imagery, such as a cartoon character with an exploding head. Nevertheless, when these metaphors are divided into their component primary metaphors, image schemas and frames, their embodied basis becomes evident. As children grow and accumulate sensorimotor experiences, they form cognitive structures that allow them to make the most of their sensorimotor skills. Cognitive structures such as image schemas and frames let human beings apply inferences derived from previous examples of Containment or Exercising, for instance, to new examples of the schemas. This is useful because a new type of container will automatically be recognized as having an inside and an outside, as causing its contents to move when it moves, and so on. Metaphors go a step further. Conceptual metaphors allow humans to apply inferences based on physical experience to new domains, even abstract domains such as A N G E R or the E C O N O M Y . This multiplies the mileage provided by frames, image schemas, and the other cognitive structures extracted from embodied experiences. Conceptual metaphors are an efficient way to re-use these cognitive structures in other domains. On this view, metaphors are much more than ‘a form of poetry’ or a ‘figure of speech’; they are, in fact, the basis of much abstract reasoning and thought.
25 Metonymy Jeannette Littlemore
25.1 Introduction Metonymy is a process whereby one entity or event is used to refer to another, related, entity or event. For example, in the sentence the Number 10 knives were out for the Chancellor, which appears in the British National Corpus (BNC), Number 10 refers metonymically to the UK government. In the USA, the corresponding term is The White House; and in South Korea, it is The Blue House. All of these examples involve a P L A C E F O R 1 I N S T I T U T I O N metonymic relationship. However, this is not the only kind of metonymic relationship. There are many others. The word Hoover can be used metonymically to mean vacuum cleaner, via a P R O D U C E R F O R P R O D U C T relationship, or we might say that we need a drink, to refer specifically to alcoholic drink, which would evoke a W H O L E F O R P A R T metonymic relationship. We might say that we need some muscle, when what we need is a strong person to help us move some furniture, thus evoking a D E F I N I N G P R O P E R T Y F O R C A T E G O R Y metonymic relationship, and so on. Because metonymy is a property of our conceptual system, evidence of it can be found in a wide range of forms of expression, including language, art, music, dance, and film, where it operates as a kind of communicative shorthand, allowing people to use their shared knowledge of the world to communicate with fewer words than they would otherwise need. Metonymic relationships also underlie many of the grammatical features found across languages, though a full discussion of this is beyond the scope of this chapter (for good accounts of this, see Panther and Thornburg 2009, Ruiz de Mendoza Iba´n˜ez and Otal Campo 2002, Ruiz de Mendoza Iba´n˜ez and Mairal Uson 2007). It serves a wide variety of communicative functions, such as relationship-building, evaluation, humor, irony, and 1
In the cognitive linguistic literature on metonymy it is conventional for overarching metonymy-producing relationships, or conceptual metonymies, such as these, to be presented in small capitals, in order to emphasise their parallels with conceptual metaphor.
408
LITTLEMORE
euphemism. We are often unaware of the fact that we are using it, and at times it can be very difficult to spot. When used well, metonymy can have significant persuasive powers, but its use can lead to severe misunderstandings; metonymic meanings can be very subtle and easily missed, especially in communication between people with different linguistic or cultural backgrounds. I begin this chapter with a discussion of how metonymy is defined and how it relates to metaphor and literal language. I then discuss the various models that have been proposed within cognitive linguistics to explain how metonymy operates, and present the main findings from neurolinguistic and developmental research into metonymy. I then outline some of the key functions that metonymy performs in language and other forms of communication Finally, I look at variations and similarities in the ways in which metonymy manifests in different languages, and discuss the implications this has for language learning. Throughout the chapter, the discussion is illustrated with real-world examples of metonymy.
25.2 What Is Metonymy and How Is It Different From Literal Language or Metaphor? Unlike metaphor, which usually involves a comparison between two unrelated processes or entities, metonymy is a process whereby one process or entity is used to refer to another, to which it is closely related or even forms part of. The best way to illustrate this is with an authentic example such as the following from the British National Corpus (BNC): (1)
This created something of a diplomatic flurry between London and Washington, as the Americans sought assurances that the British were not trying to sabotage their plans. (BNC)
In this example, London and Washington are metonyms for the British and US governments, or rather the people who work for them. This is a common metonymic relationship, in which a place stands for the people or organizations that are based there. The simplicity of this example belies the complexity of metonymy and the considerable controversy that surrounds its nature and scope. This complexity means that it has been defined and modeled in a number of different ways (see, for example, Barcelona 2000, Radden and Ko¨vecses 1999, Ruiz de Mendoza Iba´n˜ez 2000, Barcelona 2003, Panther and Thornburg 2007, Benczes, Barcelona, and Ruiz de Mendoza 2011, and Littlemore 2015). However, in cognitive linguistics there is general agreement on three things: that metonymy is fundamentally conceptual in nature, that it is experientially grounded, and that it involves contiguous
Metonymy
elements of some kind (Gibbs 1994, Barcelona 2011, Denroche 2015, Littlemore 2015).2 It has been argued that metonymy sits on a cline somewhere between literal language and metaphor (Dirven 1993 [2003]). We can see this by looking at the following three citations from the British National Corpus, all of which contain the word nose: (2) (3) (4)
Her mother sat by the fire, her knitting lying in her lap, eyes closed and spectacles slipped down her nose. (BNC) To the modern nose, it stinks. (BNC) After the initial reaction of lowering the nose and checking the speed, a quick assessment of the situation is necessary. (BNC)
In the first of these examples (2), the word nose is being used to refer to a part of the body and is clearly literal. In (3) it refers to the act of smelling, via a process of metonymy. In (4), it refers to the front of an aeroplane and is clearly a metaphor. Dirven’s cline has the potential to be used in the context of language teaching. Indirect evidence for this cline has been found by MacArthur and Littlemore (2008) and Littlemore and MacArthur (2012), who discovered that when learners of a foreign language were exposed to corpus lines, they were able to move fairly freely between purely literal senses, through metonymic senses, to metaphorical ones. In other words, they appeared to be using the continuum of senses as they worked with the data, using one sense that they understood to help them work out the meaning of another. So, for example, when looking at uses of the denominal verb shoulder in the BNC, the Spanish learners of English who participated in the study encountered the verb used metonymically to mean ‘carry,’ as in sometimes Andrew had been so drunk that Iain had shouldered him all the way home (BNC) as well as metaphorical uses like she shouldered the blame for what the media gleefully called: ‘Malice at the Palace’ (BNC). Although this second use of shoulder might prove somewhat opaque to the non-native speaker in another context, when encountered alongside other related senses of the verb in the corpus lines, learners had little difficulty in understanding it, and also appreciated its entailments. Another way of distinguishing between metaphor and metonymy is to focus on the fact that whereas metaphor involves the juxtaposition of two different Idealised Cognitive Models (ICMs), or frames, metonymy draws on relationship within a single ICM (frame). ICMs are abstract mental representations that people have of ‘typical’ situations in life; they are encyclopedic, flexible, and somewhat idiosyncratic (Littlemore 2015).3 They encompass the cultural knowledge that people have and not 2
One important exception here is Ruiz de Mendoza Ibáñez and Diez Velasco’s (2003) model, which questions the role of contiguity, and focuses instead on group membership.
3
Lakoff (1987) lists five types of ICM: propositional ICMs, image schema ICMs, metaphoric ICMs, metonymic ICMs, and symbolic ICMs. What we are referring to here are propositional ICMs. The fact that Lakoff includes metaphor and metonymy in his list is somewhat infelicitous as these are best seen as operational or ‘dynamic’ cognitive processes rather than non-operational cognitive models. This view is also expounded by Ruiz de Mendoza Ibáñez (1998).
409
410
LITTLEMORE
Crockery and cutlery (plate dish, knife, fork...)
Meal (Breakfast, lunch, dinner, etc.) FOOD AND DRINK
People eating or serving the food (waiter, diner, etc.)
Furniture and buildings associated with food (Table, trolley, drinks cabinet etc.)
Types of food (steak, rice, potatoes, wine, etc.)
Figure 25.1 The ‘food and drink’ ICM
necessarily a reflection of ‘the real world.’ One example of an ICM is ‘food and drink’ ICM, which is illustrated in Figure 25.1. This ICM sanctions a number of metonyms that are conventional in English and in other languages. For example, in Japanese, breakfast, lunch, and dinner are referred to as asagohan (morning rice), hirugohan (midday rice), and bangohan (evening rice), which relies on a metonymic link between meals and types of food; in French one can say elle fait une bonne table (she makes/does a good table), meaning she cooks well, thus linking food to the furniture associated with it; or in English we say that a restaurant has a good cellar, meaning that its wine is good. A waiter might ask which of the diners is the steak thus linking the people eating the food to refer to the food itself, and a good restaurant in a restaurant guide might have three knives and forks, which implies that serves good food. People use the term dish to refer to the food that is in the dish and they use the term glass to refer to the content of that glass. Dirven’s continuum and the notion of ICMs provide a good starting point for the identification and analysis of metonymy, but when we start to look at more examples of metonymy, we see that metonymy itself can be broken down into different types. Various taxonomies of metonymy
Metonymy
Over-arching metonymy categories Whole and part metonyms
ICMs
Examples of metonymies
Thing and part
E.g. Part for whole The perfect set of wheels (BofE) E.g. Ends for whole scale Young and old alike (BofE) E.g. Material for object Use only a 3-wood off the tee (BofE) E.g. Subevent for whole event Jay and Denise are to walk up the aisle (BofE) E.g. Category for member of category Fancy coming round for some drinks (CorTxt) E.g. Salient property for category The brothers needed some muscle (BofE) E.g. Time for action They summered at Ville d’Avray (BofE) E.g. Thing perceived for perception Head not so great (CorTxt) E.g. Effect for cause Because you live on a fast road… (BofE) E.g. Producer for product She took out the hoover (BofE) E.g. Controller for controlled Rommel was in retreat (BofE) E.g. Possessed for Possessor he married money and became and M.P. (BofE) E.g. Container for contents Sits you down and have a glass (BNC) E.g. Place for inhabitants The whole town is on the verge of starvation (BofE) Modified form for original form E.g. LOL (for ‘laugh out loud’)(CorTxt)
Scale Constitution Event
Category and member
Part and part metonyms
Category and property Action Perception Causation Production Control Possession
Containment Location
Modification
Figure 25.2 Key metonymy types in Radden and Kövecses (1999) taxonomy (adapted from Littlemore 2015)
types have been proposed in the literature. One of the most influential semantically inspired taxonomies of metonymy types is that proposed by Radden and Ko¨vecses (1999). The model is not exhaustive, but it does contain a very wide range of the types of metonymy that appear in everyday language, as we can see in Figure 25.2. Another model of metonymy, which has made a key contribution to the metonymy literature, is that proposed by Ruiz de Mendoza Iba´n˜ez and Diez Velasco (2003), who argue that all instances of metonymy can be described as one of two types: either as a T A R G E T I N S O U R C E metonymy, where the metonymic term is part of its referent, or as a S O U R C E I N T A R G E T
411
412
LITTLEMORE
metonymy, where the referent is part of the metonymic term. The example that they give of a T A R G E T I N S O U R C E metonymy is the use of the term the Pill to refer specifically to the contraceptive pill. Here, the Pill is used to refer to a specific type of pill, which means that the target of the metonymy (the contraceptive pill) is a subset of the domain covered by the general word pill. The example that they give of a S O U R C E I N T A R G E T metonymy is the use of hand in expressions such as all hands on deck. Here, the hands refer to the sailors who are doing hard physical work so the hands are simply part of the domain. They adopt a very broad definition of what is meant by the source and target, which allows them to classify a wide range of metonymies as target-in-source metonymies. Other researchers, such as Barcelona (2011) and Handl (2011), have shown that some examples of metonymy appear to be more prototypically metonymic than others, and therefore see metonymy as operating within a radial category. Consider, for example, the following: (5) (6) (7) (8) (9)
the odds had lengthened against effective leadership from the White House (BNC) This does not mean that speed must be avoided, nor that the unexpected need offend the eye. (BNC) The doctor, short, plump with thinning white hair, raised his eyebrows. (BNC) the Tsar got out of the Imperial train for a stroll at Binul Station and the train left without him (BNC) toothbrush versus paintbrush
Examples (5) and (6) are arguably the most prototypical. It is not literally possible for the White House (i.e. the building) to provide leadership, and it is not literally possible for an eye to be offended. However, it is quite possible for people to raise their eyebrows and for this to be metonymically interpreted as meaning that they were surprised, which suggests that example (7) constitutes a less prototypical form of metonymy. This difference in metonymy types has been observed by Warren (2006), who labels the first type referential metonymy, and the second type propositional metonymy. In referential metonymy, one concept (in this case, the eye) is used to refer to a related concept (in this case, ‘taste,’ which is metonymically related to the eye in that one needs to see something in order to appreciate it). In propositional metonymy, one proposition (in this case raising one’s eyebrows) is used to refer to another (in this case, the experience of being surprised). As well as one being ‘literally’ true and the other not, another difference between these two types of metonymy is that for referential metonymy, there is no other word for the phenomenon, other than ‘metonymy.’ With propositional metonymy on the other hand, there are other terms for the same phenomenon, such as ‘inductive generalisations,’ ‘invited inferences,’ and ‘implications’ (Warren 2006: 10). In example (8), we have an even less ‘prototypical’ form of metonymy.
Metonymy
Although the train left without the Tsar, it was the driver of the train who was responsible for this. It is not clear in this example whether the writer is referring to the train or the driver or both. In other words, this metonymy is underspecified in that both the vehicle and the target contribute to the overall meaning. ‘Underspecified metonymies’ thus have distinct subdomains, both of which determine the meaning. The examples in (9) are the least prototypically metonymic of all. These are ‘domain highlighting’ metonymies, which have no distinct subdomains whatsoever, just differences in construal or viewpoint that result from the juxtaposition of the different words (Langacker 1993). Whereas a toothbrush is a brush that is used to brush one’s teeth, a paintbrush is a brush that is used to apply paint to a wall; the relationships between the constituents in these two compounds are therefore different, and highlight different domains. Handl (2011) provides experimental evidence indicating that many metonymies are in fact often understood in an underspecified manner. She found that the informants in her study usually thought that the metonymy was referring to both the term and its referent. She assigned different weights to the two parts of the metonymy based on frequency of response. In prototypical metonymies, the weight of the vehicle was found to be relatively high. That is to say, the actual words of the metonymy play an important role in its meaning, whereas when we move out toward the periphery of the radial category, the vehicle recedes into the background. In the more intermediate types of metonymy, both the vehicle and its referent play an equally important role. Another non-prototypical form of metonymy is ‘illocutionary’ metonymy, which has been proposed by Panther and Thornburg (1998) in their work on the role of metonymy in pragmatic inferencing. For example, one might ask a friend have you got any scissors? which in fact means ‘can I borrow the scissors?’ It is not always easy to distinguish between metonymy and metaphor, partly because metonymy actually lies at the heart of much metaphor (see, e.g., Barnden 2010, Dirven and Po¨rings 2003). In his work on ‘metaphtonomy,’ Goossens (2003) identifies four main ways in which metaphor and metonymy interact. The first involves ‘metaphor from metonymy’ in which the experiential basis of a metaphor consists of a metonymy. It has been suggested that most, if not all, primary metaphors (such as E M O T I O N A L C L O S E N E S S I S W A R M T H ) have a metonymic basis, in that they begin life as metonymic extensions from bodily experiences, which are then projected onto abstract domains (Ko¨vecses 2013). The second way in which metaphor and metonymy interact involves ‘metonymy within metaphor.’ This occurs when a metonymy functioning in the target domain is embedded within a metaphor. For example, in the expression I’ve got you, meaning ‘I understand what you are saying to me,’ you stands metonymically for the information that you are trying to convey to me, but the whole exchange takes place within the conduit metaphor (see Lakoff and Johnson 1980a), according to which conversation and/or information exchange is construed metaphorically as the transfer of a physical entity from one person to
413
414
LITTLEMORE
another. The third way in which they interact is referred to as ‘demetonymisation within metaphor.’ For example, the expression, scatter brained, which means to be flighty and disorganized, appears, at first sight, to involve a metonymy in which the brain stands for thinking, but the expression is, in fact, only ever used in an abstract sense and is therefore always a metaphor. The expression therefore loses its apparent metonymic element. The fourth way in which metaphor and metonymy interact involves ‘metaphor within metonymy.’ For example, if we talk about a department being the eyes and ears of an organization, the eyes and ears metonymically represent seeing and hearing, but they operate within an overarching personification metaphor in which the organization is seen as a person.
25.3 Neurolingustic and Developmental Studies of Metonymy One way to discover more about the nature of metonymy and about how it relates to metaphor and literal language is to investigate findings from research into how it is processed in the brain. By doing so, we can find out more about similarities and differences in the neurological processing of metaphor, metonymy and literal language, the role of context in metonymy comprehension, and the effect of syntax on metonymy recognition and production. All of these areas have been investigated using eye tracking, brain scanning, reaction time studies, and metonymy comprehension tasks that have been administered to individuals across different age groups. Space does not permit an in-depth discussion of all this work, so the aim here is to focus on specific findings from neurolingustic and developmental studies of metonymy that tell us more about the nature of metonymy and its relationship to metaphor. Research using both eye-tracking software and brain-imaging techniques suggests that people process conventional metonymy in much the same way as they do literal uses of language. Eye-tracking studies (e.g. Frisson and Pickering 1999) have shown that people do not pause any longer when asked to read conventional metonymic sentences than they do when reading literal sentences, but that when they are presented with a novel metonymy it takes them much longer to do so, and that metonymies from more complex ICMs take longer to process than ones from simple ICMs. Frisson and Pickering conclude from their findings that individuals develop an underspecified, schematic idea of a word’s meaning and then home in on the appropriate (literal or metonymic) sense by activating relevant parts of the meaning as soon as sufficient contextual information has been provided. Findings from brain imaging studies (e.g. Rapp et al. 2011) concur with those from eye-tracking studies in that they have shown conventional metonymies to be processed in the same part of the brain as literal language, whereas novel metonymies are processed in a different part of the brain. These
Metonymy
findings thus show us that novel and conventional metonymy are clearly very different things, a fact which needs to be taken into account within theoretical models of metonymy. Researchers investigating the rate at which metonymy comprehension develops in children have found that children start to understand metaphor and metonymy at roughly the same time, but that metonymy comprehension develops at a faster rate than metaphor comprehension, and that metonymy comprehension is more strongly related than metaphor to vocabulary size (Rundblad and Annaz 2010). Metonymy comprehension thus appears to develop alongside vocabulary size; it develops more rapidly than metaphor; and it follows a similar trajectory to the development of literal language comprehension, although the trajectory is somewhat slower. These findings suggest that, from a neurological perspective, metonymy is different from both metaphor and literal language but that it shares important properties with both, and that novel and conventional metonymy are processed differently.
25.4 The Functions of Metonymy In the chapter so far, we have focused mainly on what metonymy is, and on how it is processed. In this section we turn to the question of what metonymy does, and explore the functions that it performs in communicative settings. As well as serving a relatively straightforward referential function, it can be used for a wide variety of communicative functions, such as evaluation, relationship-building, humor, irony, and euphemism. In this section, we explore some of these functions by looking at a number of contextualized examples of metonymy.
25.4.1 The Evaluative Function of Metonymy Metonymy is often used to convey evaluation. One way in which it does this is by drawing attention, via the T R A I T F O R P E R S O N relationship, to a particular feature of an object or person and using this feature to refer to the whole person. We can see this in the following example where it is used to convey the protagonist’s attitude toward one of the other characters in the novel: One of the girls, with bad acne, said ‘Give me ten and I’ll flog you off.’ ... They all laughed at this – giggling, silly, pushing at each other. Jonjo remained impassive. ... The girls whispered to each other, then Acne said, ‘We don’t know.’ –William Boyd (2009) Ordinary Thunderstorms, London: Bloomsbury, p. 107.
415
416
LITTLEMORE
By referring to one of the girl as ‘Acne,’ the author shows how she is both unattractive and dehumanized in the eyes of the protagonist, Jonjo. Another way in which metonymy conveys evaluation is through the P A R T F O R W H O L E relationship, where the part being referred to has strong positive or negative connotations, as in the following examples, both of which are taken from Littlemore and Tagg’s (2016) study of the role of metonymy in text messaging: (10) (11)
Ok.Get ur fucking lazy, no good for nothing arse in here right now (CorTxt) Cool cheers brain u’re too kind! Haha. See u later alligator! (CorTxt)
In some contexts, even the mere presence of metonymy can indicate evaluation. It has been shown that emotionally charged capital city names are significantly more likely to be referred to metonymically than those that are less emotionally charged (Brdar and Brdar-Szabo´ 2009). Moreover, in their study of the metonymic use of the terms Beijing and Taipei in Chinese and Taiwanese newspapers to refer to the governments of those countries, Zhang, Speelman, and Geraarts (2011) found that capital names were more likely to be used metonymically if the country being referred to was being negatively evaluated. The evaluative function of metonymy is also highlighted by Biernacka (2013) who, in her study of the use of metonymy in focus-group discussions of terrorism, found that metonymic clusters, where several metonyms were used in quick succession, tended to coincide with intense discussions of particularly controversial and emotional topics. The metonyms in these ‘emotionally intense’ clusters were used by the participants to describe how they saw themselves in relation to the situation under discussion and how they felt about their relations with others.
25.4.2 Metonymy as a Carrier of Social Attitudes Because of its ability to convey subtle evaluation, metonymy can reveal implicit social attitudes. We can see this process at work in the following metonyms, all of which can be used to refer to one’s wife in British English: (12) (13) (14)
He actually bumped into Her Indoors, who peered, looked bewildered and retired. (BNC) I promised the missus I’d be home by eleven. (BNC) Simon Draper took to calling her ‘She who must be obeyed’ – if not to her face. (BNC)
All of these examples rely on the ‘humorous’ use of a T R A I T F O R P E R S O N metonymy, and reduce women to either figures of fun or to their roles as wives, especially housewives, who spend most of their time indoors. Metonymy can also be used to inadvertently convey underlying prejudice toward disabled people. Consider, for example, the following sentence, which appeared in a news article reporting on the fact that British
Metonymy
Paralympian athlete Claire Harvey reported that she had to drag herself from a Qatar Airways flight after being left on board. Before the flight, Ms Harvey said she heard staff talking about getting “the wheelchair” on board – a term she eventually realised referred to her. –BBC News Website, November 9, 2015
By referring to Claire Harvey as the wheelchair, the member of staff appears to be seeing her purely in terms of her disability. This may have made sense from a practical point of view but the fact that it was uttered in her presence makes it somewhat dehumanizing. A further way in which metonymy acts as a carrier of social attitudes is demonstrated in its role as a vehicle for euphemism (Gradecˇak-Erdeljic´ 2004). Going to the bathroom, the little boys room, the cloakroom, and spending a penny are all metonymic euphemisms in British English for the act of going to the toilet (itself a euphemism) which, in Western society, is something that people tend to avoid referring to directly.
25.4.3 Metonymy as a Marker of Group Membership and Cohesion In their study of figurative language, genre, and register, Deignan, Littlemore, and Semino (2013) show how metonymy is often used to create and sustain group membership, highlighting shared knowledge between members of the group and at times keeping outsiders out of the group. They focus in particular on two communities of practice: staff working in a children’s day nursery and parents associated with a children’s football club. Examples included the term loose nappy by the nursery workers to mean that a baby had problems with its bowels, as well as numerous terms such as stand on it, hit it, and curve it in, to refer to various things that one might do with a football, none of which correspond to the literal senses of the words used. The language used by members of sports clubs appears to be particularly rich in metonymy. For example, in climbing clubs, it is common to ask have I got a foot? or have I got a hand? meaning ‘is there a hand hold (or foot hold) nearby that I am likely to be able to reach?’ Discourse-community specific metonymy is also prevalent in sportsrelated journalism. In the following example, an expert on sports finance, Jeremy Claine, is commenting on the way in which Chelsea football is being run by its owner, Roman Abramovic: (15)
There has clearly been discord in the dressing room and in the boardroom. (Jeremy Claine, Sports Finance Expert, Grant Thornton, speaking on the Today Programme, Radio 4, December 2015)
Here, the speaker sums up, via two different P L A C E F O R P E R S O N metonymies (which draw on different yet overlapping ICMs), the fact that neither the players nor the management are happy with the way in which the club
417
418
LITTLEMORE
is being run. These uses of metonymy would be familiar to people used to listening to the genre of sports news reporting, but may be unfamiliar to those who are less well-versed in the genre. Novelists are aware of the ways in which the use of metonymy can help create and develop bonds within families and groups of friends, and are able to exploit this to make their characters and settings feel more authentic. We can see an example of this in Maggie O’Farrell’s novel, Instructions for a Heatwave: Imagine then, the uproar when Aoife announced one night over dinner – Monica and Joe were there but Michael Francis was not – that she wasn’t going to help at Sunday school classes, that she had been to see the priest that very day to say she wouldn’t be doing it. She didn’t want to be a teacher, she wasn’t good with children, she could think of nothing worse. It was one of the Riordan family’s louder uproars. Gretta hurled a plate of spinach to the floor. She would later deny this and say it had slipped from her hand. Either way, spinach ended up on the carpet, and there would be a green stain there for years, always referred to within the family as ‘the Sunday-school stain.’ Gretta said she would die of shame, that Aoife would be the death of her, that she didn’t know what to do with her. –Maggie O’Farrell (2013) Instructions for a Heatwave. London: Tinder Press, p. 95.
In this example, the metonymic term the Sunday-school stain has a particular meaning for this family, which is associated with a particular event. The term would be meaningless to people outside the family group, or who are not familiar with the event. O’Farrell uses this device to make the reader feel that they understand this family.
25.4.4 Playful and Creative Uses of Metonymy A powerful way in which people use metonymy to build and maintain relationships is to use it creatively as a form of language play.4 In their study of creative metonymy use in text messaging, Littlemore and Tagg (2016) identify two non-mutually exclusive forms of creativity involving metonymy: one based on meaning and one based on form. An example of meaning-based metonymic creativity can be found in the following extract, in which two different metonyms from different ICMs are juxtaposed within the same exchange: A: B:
4
Just in case you need rescuing from work or dissertation or both, we are meeting again tomo at 6 in staff house. Hope week 3 ok. It’s a toss up between that and going for a balti with cherry blossoms. Not sure what would be more fun.
For an in-depth discussion of the social functions of language play and creativity see Carter (2004), Cook (2000) and Swann (2006).
Metonymy
A: B: A:
Ooh i wouldn’t like to have to make that decision. . . [time passes] Sorry cherry blossoms and balti win out. Maybe we could meet up tomorrow if you fancy. NAME119’s busy tonight and haven’t asked about tomorrow yet. Damn, passed over for a cherry blossom. Yeah give me a shout if you’re doing anything tomo. Happy balti – and happy end of course.
This example contains repeated metonymic references to cherry blossoms and baltis. The two texters in this exchange were language teachers at an English Language school and they used the term cherry blossoms to refer to a group of Japanese students from the Cherry Blossom College in Japan. A balti is a curry dish popular in Birmingham (UK) and is itself a metonym in that it is derived from the vessel in which these types of curries are normally cooked. In this example it has a further metonymic use as it means to go out to a restaurant and eat a balti. Thus we have two different metonymy types (P L A C E F O R P E R S O N and O B J E C T F O R E V E N T ) being used in close proximity. Although both metonyms are relatively conventionalized, at least for these texters, the way in which they are repeatedly juxtaposed here suggests a degree of deliberate creativity. The creativity rests on an awareness of the two very different ICMs being evoked: one surrounding Japanese culture and one surrounding spiced Indian food. Coherence is provided by the fact that both refer to ‘foreign’ or even ‘exotic’ cultures (Japanese and Anglo-Indian), but humor is implied by the contrasting features of Japanese and Anglo-Indian culture. As well as drawing on and reinforcing the texters’ shared knowledge and thus strengthening their relationship, the humor here appears to be used to mitigate a potentially face-threatening situation. Texter B is turning down texter A’s offer of a night out, and appears to be using the references to baltis and cherry blossoms to cover any potential social awkwardness. The second type of creativity involving metonymy, form-based creativity, involves repetition, structural parallelism, and deviations from expected forms. The following example contains structural parallelism and it is this parallelism which provides metonymic meaning to the word allotmenting: (16)
Not been much fun cycling let alone allotmenting. CorTxt
Deviation from expected forms involves the manipulation of fixed expressions and idioms, as in the following examples: (17) (18) (19)
Make yourself a cup of tea on me Never in blue jeans Smell ya later
These examples involve the creative manipulation of the metonymic expressions have one on me (which usually refers to an alcoholic drink at the bar), forever in blue jeans (meaning overly casual) and see ya later. Many
419
420
LITTLEMORE
creative uses of metonymy found in the text messaging corpus involved both meaning and form-based creativity.
25.4.5 The Use of Metonymy to Create Textual Coherence A final function of metonymy is to provide textual coherence. This feature of metonymy has been explored in depth both in literature (Al-Sharafi 2004) and in everyday language use (Brdar-Szabo´ and Brdar 2011). Metonymy lends itself particularly well to this function as both the source domain and the target domain can serve as the point of reference, and both remain ‘available’ for access by the reader long after their first mention, a principle which is referred to as the ‘Domain Availability Principle’ (Ruiz de Mendoza Iba´n˜ez and Pérez Herna´ndez 2003). Because the source domain is so large and flexible, it allows for forms of anaphoric reference which, at first sight may not seem to be syntactically ‘logical.’ For example, in a single text, one might see a reference to 9/11, the attacks, the Twin Towers, all of which provides access to the same domain. Metonymy can also contribute to textual coherence in a much more straightforward way, by means of repetition, as we saw above in the balti and cherry blossom example, where the repeated use of these metonyms provided continuity in terms of both the subject matter and the lighthearted attitude that both speakers sought to convey with respect to that subject matter. A third way in which metonymy can provide textual coherence is via the re-literalization of metonymic references. We can see an example of this in the following extract: Estha watched as they walked along the railing, pushing through the crowds that moved aside, intimidated by Chacko’s suit and sideways tie and his generally bursty demeanour. ... The Sitting Man with the cap and epaulettes, also intimated by Chacko’s suit and sideways tie, allowed him in to the baggage claim section. When there was no railing left between them, Chacko kissed Margaret Kochamma, and then picked Sophie Mol up. ‘The last time I did this I got a wet shirt for my pains,’ Chacko said and laughed. He hugged her and hugged her and hugged her. He kissed her bluegreyblue eyes, her entomologist’s nose, her hated redbrown hair. Then Sophie Mol said to Chacko, ‘Ummm. . . excuse me? D’you think you could put me down now? I’m ummm. . . not really used to being carried.’ So Chacko put her down. Ambassador Estha saw (with stubborn eyes) that Chacko’s suit was suddenly looser, less bursty. –Arundhati Roy (1997/2009) The God of Small Things, London: Harper Collins, p. 147.
Metonymy
In this extract, the fact that Chacko appears to be bursting out of his suit refers metonymically to his high level of confidence, which is reflected in the way he walks and carries himself. In the final sentence, when he has been rebuffed by his daughter (Sophie Mol), he seems physically deflated to the extent that his suit actually appears to have literally become looser and less bursty.
25.5 Cross-linguistic Variation in the Use of Metonymy and Its Implications for Language Teaching The way in which metonymy is used varies significantly across languages. Studies have revealed, among other things: variation in the metonymic use of place names to refer to events that took place there (Brdar and BrdarSzabo´ 2009), variation in the ways in which metonymy is used in pragmatic inferencing and indirect speech acts (Panther and Thornburg 1999), variation in the use of syntax to mark metonymic meaning (Brdar-Szabo´ and Brdar 2011, Ruiz de Mendoza Iba´n˜ez 2008) and variation in the degree of tolerance that speakers have for loosely contextually associated, novel metonymic shifts (Slabakova, Amaro, and Kang 2013). Moreover, the same underlying metonymy-producing relationships have been found to be exploited in very different ways in different languages (Ko¨vecses 2005). Hilpert (2007) found substantial differences in the ways in which body parts are used metonymically in different languages. For example, in some languages, the word ear is used metonymically to refer to ‘paying attention,’ whilst in other languages, ears are used to refer to ‘obedience’ via a C A U S E F O R E F F E C T metonymy. In some languages, the word belly is extended via a C O N T A I N E R F O R C O N T A I N E D metonymy to refer to ‘pregnancy’ and via a C A U S E F O R E F F E C T metonymy to refer to ‘offspring.’ In some languages, the word forehead is grammaticalized to mean ‘in front’ and the word buttocks is grammaticalized to mean ‘behind.’ Bodypart metonymies are thus subject to substantial cross-linguistic variation. The resulting expressions may be less than transparent to people who do not speak those languages fluently. This variation has implications for language learning. Littlemore et al. (in press) have shown that metonymy presents difficulties for language learners, particularly when the meanings carry strong cultural connotations, or are related to less central features of the source term. They found that the learners in the study displayed a tendency to process metonymy as if it were metaphor. For example, the expression married money was interpreted metaphorically to mean ‘earn a lot of money,’ rather than metonymically, meaning to marry someone with money, which is its actual meaning. The fact that learners experience difficulties with metonymic meaning and that they appear to treat metonymy as if it were metaphor suggests that they may be unfamiliar with the device. It may therefore be
421
422
LITTLEMORE
beneficial to raise their awareness of metonymy as a linguistic device, and to discuss the ways in which it varies across languages. An explicit focus on metonymy may therefore be beneficial to foreign language learners. It may be useful to attempt to raise learners’ awareness of ubiquity of metonymy through the use of examples, discuss with them how contextual cues can be used to understand metonymy, outline language- or culture-specific barriers to metonymy comprehension, and stimulate metonymy-guided reasoning (see Barcelona 2010). Future research could usefully investigate the advantages of employing such approaches.
25.6 Conclusion Metonymy is subtle, useful, and widespread. This chapter has provided a very brief introduction to metonymy and to the growing amount of research that is currently being conducted in the area. Findings from this research indicate that metonymy is at least as interesting, and as complex, as metaphor, in terms of its underlying conceptual structure, the functions that it performs in everyday communication, and the challenges that it presents in cross-linguistic settings. Over the coming years we can expect to see more research being conducted into this elusive yet fascinating trope.
26 Conceptual Blending Theory Todd Oakley and Esther Pascual
26.1 Introduction: Meaning Construction in Context Standing in line at a local Starbucks Coffee, one of the authors overheard the following remark by a fellow customer to his companion: ‘Oh . . . he was your next-door neighbor before you lived there.’ The referent in question was never the addressee’s next-door neighbor in the conventional sense of synchronic occupation of adjacent dwellings, and if we were philosophers of language in a model-theoretic mold, such an utterance would be demonstrably false, if not altogether meaningless. But such utterances are the everyday results of interactions between cognitively modern human beings, and we doubt the addressee had any trouble understanding it. These are precisely the types of utterances on which Gilles Fauconnier (1985, 1994, 1997) built his theory of mental spaces, according to which natural language is a process of building conceptual scenes and scenarios as we think and talk, and otherwise interact. Meaning emerges from elements, roles, values, and relations that inhabit or form a conceptual domain referred to as a ‘mental space.’ Mental spaces are either ontological domains within which the same element can be construed (e.g. Reality, Fiction, Dream, Counterfactuality) or can represent the ‘world’ of an individual, entity, or concept as opposed to another one (e.g. Barack Obama, Disneyland, Chinese medicine, Nobel Prize winner). Human minds fluidly and (seemingly) effortlessly build up meaning through the construction, connection, and integration of these mental spaces, where the truth or falsity of a statement is really a rarified outcome of a much more general process linking thought, language, and context. Back to the Starbucks queue: (1)
Oh. . . he was your next-door neighbor before you lived there.
Comprehension of (1) entails the creation of a mental space configuration such that it is possible to refer to a past inhabitant of a dwelling as
424
OA K L EY A N D PA SC UA L
n
R (present)
n’’
P (past)
Figure 26.1 Example 1 (Classic mental spaces analysis)
erstwhile neighbor to a present occupant at an adjacent dwelling. So, then, in some non-trivial sense, the addressee of (1) is predicated to share a social relationship with the referent of (1). A fruitful way to model this relation is to posit a mental space ‘R’ for speaker’s reality at the time, within which the referent ‘he’ is construed. In this space, the element n refers to ‘he’ as a rolefilling topic of conversation. It signifies that n is known to both discourse participants: perhaps ‘he’ is a colleague at work or a common client. To understand (1) requires a new mental space, P, referencing a past situation. In P, the addressee and referent occupy the same social-geospatial role of ‘next-door neighbor,’ but at different times. As Figure 26.1 shows, a meaningful and salient spatial alignment of roles emerges, but it does not show the precise temporal alignment of their values, where the role n is preserved (represented by the pragmatic connector running from R (present) to n’ in P (past). The two mental spaces specify counterparts of the same individual, inhabiting either present R or past P. This curious example illustrates the fact that natural language rarely behaves well according to the postulates of model theoretic accounts of language, as there is no unequivocal one-to-one relation between linguistic entities and the objective world, whatever that may be. This naturally has important implications for semantic theory, since it entails that meaning is not found in words or in other tangible linguistic units. Rather, meaning is constructed within a given conceptual configuration set up by language users at a given point in discourse and interaction. Thus, an understanding of the functional-cognitive foundations of meaning construction (wherein reference is a critical and defining practice) requires that any cognitive linguistic model of meaning construction regard such counterfactual, even illogical utterances as central rather than peripheral cases.
26.2 Ersatz Neighbors Now, imagine that you are the addressee. A week later, you run into the very person who is the referent of (1). You happen to know him
Conceptual Blending Theory
xα
xβ
nα
nβ
OR (present) NEIGHBOR
NEIGHB NEIGHBOR (past)
xα’
nβ’
Ersatz Neighbor Figure 26.2 Example 1 (ersatz neighbors blend)
professionally, and now decide to greet him with the hardy salutation: ‘Howdy, neighbor!’ Initially nonplussed, he is unsure what to make of this new title, and you then explain the situation. When recognizing his new role, he proceeds to swap stories with you about the neighborhood. Who lived where, when? Who left because of divorces and neighborhood scandal, etc. At this moment you two are honorary neighbors reminiscing about the neighborhood you never shared. What are we to make of such a fanciful, but nevertheless commonplace, discourse situation? Fauconnier and his collaborator, Mark Turner, extended the theory of mental space to focus on such acts of conceptual gymnastics (1996, 1998, 2000, 2002). Utterance (1) can be regarded as a linguistic tipping point for developing a full-bodied scenario of non-genuine next-door neighbors gossiping about their neighborhood. These as if activities require an extended model known variously as ‘conceptual integration’ or ‘conceptual blending.’ The imagined interaction between ersatz neighbors – and in fact just the conceptualization of these ‘neighbors’ who never dwelled in adjacent homes at the same time – is represented in Figure 26.2, where the two mental spaces function as so-called ‘input’ spaces to an emergent blended space. The blended elements of xα’ and nβ’ signify a distinct conceptual achievement, characteristic of human beings, that reveals much about the nature of cognition, but perhaps even more about the nature of language and discourse. As will become apparent in the following
425
426
OA K L EY A N D PA SC UA L
pages, such flights of fancy are really just the beginning; much of our workaday cognitive operations require some form of conceptual integration or blending, which, as (1) implies, often requires a conceptual ‘compression’ of roles, values, identity, and other relations over extended temporal and spatial associations. It is in this blended, compressed relationship that your colleague functions as both your and not your next-door neighbor. Now that we have achieved a basic understanding of what mental spaces are and how they can be integrated or ‘blended’ to achieve workable, if sometimes fanciful, conceptual scenarios, we proceed to outline the basic elements, governing principles, and possible vital relations between input spaces. After this, we discuss work on blending in grammar, and then proceed to pay particular and sustained attention to the explanatory power of Blending Theory to account for meaning construction and discourse phenomena in written, spoken, and multimodal communication. We will examine instances of conceptual blends in language for specific purposes, ranging from marketing to humor and journalism. Three different kinds of blending networks used to achieve different communicative goals will be discussed: (i) mirror blends, especially those involving a split self; (ii) double-scope blends, some with double grounding (Broˆne and Feyaerts 2005), namely blends motivated by metaphor and metonymy; (iii) double-scope blends with material anchors, modeled by physical structure in the here and now (Liddell 1995, 1998, Hutchins 2003 [2005], Williams 2004); and (iv) simplex blends, including fictive interaction blends, in which an input space is fused with the frame of the conversation (Pascual 2008a, Turner 2010). We conclude with an elaborate double-scope blending analysis of an American pharmaceutical advertisement with quadruple grounding spaces.
26.3 Basic Principles The primary cognitive function of conceptual blending is to create scenes that fit human scale (e.g. Fauconnier and Turner 2000, 2002). Much of what human beings think about, talk about, and act upon operate over scales of time and space that are either too small or too large to fit comfortably with our range of common experience. Thus, the basic principle of conceptual blending is to ‘compress’ that which is inherently diffuse and ‘decompress’ that which is inherently condensed. Consider a situation in which a father tries to explain to his young daughter the relationship between the Sun and the Earth using produce at the dinner table: (2)
This hazelnut is the Earth, which rotates like this around the Sun, this orange.
Conceptual Blending Theory
The ready-to-hand items can be manipulated to represent an unready-tohand relationship between celestial bodies, such that the father can demonstrate a complex, diffuse, and unfamiliar relationship to the daughter. In (2), we have a network of mental spaces – one for fruit, the other for planets – that can be productively aligned and momentarily blended to make concrete a relationship not easily distilled from basic human experience. The same basic operation would be at work for purposes of decompression: (3)
The hazelnut is an electron, which rotates around the nucleus, this orange.
Here the ready-to-hand foodstuff comprises the familiar means of ‘exploding’ out a mode of observation and experience that is, in principle, unavailable to denizens of the macroscopic world. Since the elements in input space 1, the fruit to be mapped and fused with the concepts at issue, are objects in the real world, present in the situation of communication, they are called ‘material anchors’ (see Hutchins 2003 [2005], Williams 2004). Interestingly, such material structures compressing or decompressing entities to a human scale may also constitute the addresser’s body, as when body parts or gestures accompanying language stand for space elements or relationships, or represent conventionalized lexical items, as in sign languages (Liddell 1995, 1998, Dudis 2004, Parrill and Sweetser 2004, Parrill, Tobin, and Turner 2010).
26.4 Compression of Vital Relations Such compressions and decompressions are part and parcel of blending processes – while fundamentally unpredictable and indeterminate in nature – nevertheless, they operate according to regular and routine conceptual connections. These ‘vital relations’ (Fauconnier and Turner 2002) cover the gamut of common concepts, including Time, Space, Cause-Effect, Change, Identity, Part-Whole, Representation, Category, Similarity/Difference (Analogy/Disanalogy), Intentionality, and Uniqueness. For instance, the blended space in (1) preserves a relation of identity between fillers of these social roles even as it relaxes any strict relation of time. By contrast, the material anchor blend in (3) creates a new part–whole relation between hazelnut and orange that is not a relation in the Produce input space, but which builds on the analogy and representation possible once the speaker constructs the network. These relations then provide the basis for mental space connections between elements and relations in any given network of spaces. In (1), both ‘he’ and ‘you’ fill identical roles in the blended space, but not in the input spaces, and are compressed into Uniqueness. In (2) and (3), the choices of hazelnut for Earth/electron and orange for Sun/nucleus are motivated by the property ‘spheroid’ and relative size, making the smaller hazelnut a good candidate for mappings to its analogical
427
428
OA K L EY A N D PA SC UA L
counterparts, Sun and nucleus, both of which are empirically ‘bigger.’ The fruit and planets are mapped by Similarity and compressed into Uniqueness in the blend, where the one entity in the here and now stands for spheres in outer space or the micro-world. In this respect, the basis of the conceptual blend, whereby the produce represents the planets/atomic parts, exploits the conceived similarity of “shape.”
26.5 Composition, Completion, and Elaboration How is blending achieved? Assuming the activation of a familiar mental space (e.g. Present Reality) and its counterpart (e.g. Past Reality), the speaker projects selective elements and relations from the one to the other, for there are many aspects of those spaces that are not immediately relevant to the configuration at hand. The composition of a conceptual blend begins with some kind of exploitable vital relation available for recruitment (i.e. projection) across spaces and into the blend. This fact is perhaps easier to grasp with example (2). The ready-to-hand, spherical foodstuff offers an occasion to project shape and relative size from the mental space, Produce, to the mental space, Celestial Body. Once the father has conscripted these items to serve a different purpose, he puts them into a proximal spatial relationship resembling an orbit. Assuming this process coincides with the utterance of (2), we can infer that the composition of the blend overlaps with this discourse event. At this moment, we have composed a Produce as Celestial Bodies blend (see Figure 26.3) with a ‘roundish’ hazelnut occupying the salient categorical entity, orbiting third planet, with the orange occupying the salient categorical entity, stationary gravitational center. Suppose now that the father replaces the first orange with a second one, with a conspicuous dark spot thereon. He says, ‘No, let’s use this one. It will be better for talking about other things.’ The composed blend has now more clearly undergone completion, for additional information about suns has been projected onto a unique feature of the dinner table orange. After the father has explained the orbital relation of these bodies, he then directs his daughter’s attention to the dark blemishes on the orange. ‘See these, these are sun spots, and they can greatly affect what happens here on Earth.’ At this point, the father exploits a property from the Produce space to illustrate a phenomenon specific to stars, namely spots of concentrated electromagnetic radiation, to say something more about the solar system. This saying something more is a common discursive effect of blending. We project additional elements and relations in order to structure the blend for the purpose of elaborating knowledge of one or more input spaces. In this case, the blemished orange is a serendipitous property, enabling the parent to open a conversation about a separate but related matter. Or, harkening back to (1), if so-and-so were to reminisce
Conceptual Blending Theory
Earth
Hazelnut Orange
Sun
PRODUCE
CELESTIAL BODY
Hazelnut
Orange
PRODUCE as CELESTIAL BODIES Figure 26.3 Examples 2 and 3 (material anchors)
with his ersatz neighbor about the ‘good old days on the block,’ he would be elaborating the blend.
26.6 Simplex, Mirror, Single-scope, and Double-scope Networks Fauconnier and Turner (2002) identify four distinct types of mental space networks to account for a range of conceptual operations. We prefer to use the term ‘integration’ when discussing simplex networks, for the network is the outcome of the basic process of fusing specific values with their respective roles from a ‘frame,’ in the sense of Fillmore (1976, 1982, and see Sweetser this volume Ch. 24 and Boas this volume Ch. 34). The most common illustration of a simplex network is the integration of Kinship relations, as in: (4)
Eric is my father.
The speaker creates a mental space that identifies the value ‘Eric’ with the role ‘father’ from the Family frame, while the possessive pronoun ‘my’ identifies the speaker as ‘ego.’ Such simplex integration of a role in a given frame and a value in a specific mental space may also be metaphorical, as in other examples of the ‘X is the Y of Z’ construction, such as
429
430
OA K L EY A N D PA SC UA L
Wordsworth’s “The Child is Father of the man” (Fauconnier and Turner 2002: 142) or this ad reading “Virgin is the parent of Virgin Music” (Joy, Sherry, and Deschenes 2009: 43).1 Similarly, in ‘he was your next-door neighbor,’ without the additional complement, we have something akin to a simplex integration of role and value. If one identifies a specific writing implement laying on a desk, picks it up, and puts it back in a box of red pencils that happens to be laying adjacent to a box of blue pencils, then one is creating a simplex mental space based on a type-token relationship. Another type of simplex blend is the fictive interaction blend (see Oakley and Pascual this volume Ch. 26), which emerges from the fusion of a mental space and the frame of the Conversation (Pascual 2008a, Turner 2010). If we say that ‘Thunder announces a coming storm,’ that ‘The bean burrito is California’s answer to France’s Croque Monsieur’ (Fauconnier and Turner 2002) or that ‘Drug makers listen in while bacteria talk’ (The New York Times, February 27, 2001), we are construing and presenting a non-conversational event or relationship in terms of a conversation. It should be pointed out that simplex networks rarely exist ‘in the wild,’2 for they are quickly integrated into complex networks for thinking, talking, and acting – we regard as conceptual blends, of which there are three. Consider the Dutch advertisement for Douwe Egberts ‘Return-Ticket Coffee’ (Figure 26.4), which reads in English as: (5)
New Return-ticket Coffee. Coffee for your outgoing and your return trip. A good trip. It starts with Douwe Egberts.
The two input mental spaces set up share the same organizing frame of Travel with identical railway travelers who drink coffee during both legs of the journey. Each mental space specifies a different temporal aspect of an event; otherwise, they mirror one another in all salient respects. In this case, the Food frame is subordinate and thereby does not recruit any elements, roles, or relations that would clash with the Travel frame. The blend itself extends these two spaces to produce the logically impossible scenario of meeting yourself coming-and-going somewhere along the line. In the blended space, but not in either input, you can make eye contact and toast yourself with coffee (see Figure 26.5).3 Classical examples of mirror networks are the boat race between Great America II and Northern Light, traveling the same route in different centuries (Fauconnier and Turner 1998: 154–56) and the riddle of the Buddhist monk,
1
For a discussion on metaphor as an instance of conceptual blending and the distinction between conceptual metaphors in a given language and culture, and specific, often one-time blends, please see Coulson (1995), Grady, Oakley, and Coulson (1999), Grady (2005), Fauconnier and Turner (2008), and Fauconnier and Lakoff (2013).
2
Instances of lexicalized fictive interaction are the exception, as discussed in Pascual and Oakley (this volume Ch. 21).
3
The ersatz neighbor blend is likewise a mirror network.
Conceptual Blending Theory
Figure 26.4 Image: Douwe Egberts ‘Return-Ticket Coffee’
in which one needs to guess when a monk walking up and down a mountain meets ‘himself’ (Fauconnier 1997, Fauconnier and Turner 1998, 2002). In contrast to mirror networks, single and double-scope blends often exploit analogical mappings of role/value relations between different organizing frames, where one frame dominates the conceptual relationship. Such activities arise from single-scope mental space networks, as in this example from Groucho Marx’s (1959) autobiography: (6)
Today, with actors, musicians and all the affiliated crafts unionized, it is hard to conceive the relationship that existed in those days between the actor and the theater manager. What Henry the Eight was to English history and Torquemada was to the Spanish Inquisition, the theater manager was to vaudeville. His powers were absolute. (66)
431
432
OA K L EY A N D PA SC UA L
Train Travel
Train Travel
coffee
OUTGOING TRIP
coffee
RETURN TRIP coffee
Train Travel
SPLIT-SELF TRAVELER Figure 26.5 Mirror network blend (Douwe Egberts)
In example (6), an analogical relationship is set up between theater manager and vaudeville actor on the one hand and absolutist ruler and subject on the other. The input space for Dictators represents a broad scenario of absolute rule, with the more scenic ‘daughter’ spaces supplying the exemplar personages from Tudor England and the Spanish Inquisition (see Figure 26.6).4 In the blend, any particular theater vaudeville actor is under the same absolute rule as a subject in the Court of Henry VIII or, worse still, a Sephardic Jew in fifteenth-century Spain or Portugal. Hyperbole aside, the emergent meaning of blending a theater-manager with one of these historical figures is meant to highlight the lopsided power-relationship between these two roles, a relationship characterized by a dominant conception of the American employer/employee relationship before the Progressive movement of the early twentieth century. While there are many potential clashes of conceptual structure between the semantic frames for Absolute Rule and Theatricality, no such clashes need to be negotiated in the blend or network at this stage. The frame for Absolute Rule comprises the organizing frame for the network; thus, it is single-scope.
4
Oftentimes, as in the case above, the input spaces comprise scenarios, defined here as situations with types of agents, actions, instruments, and constraints, while scenes are particularized instantiations of one or more of these elements and relations.
Conceptual Blending Theory
Absolute Rule
Theatricality
Rulers
Managers
Subjects
Actors
DICTATOR R
Absolute Rule
VAUDEVILLE V Managers gers e are Dictators ors o Actors are Subjects
ABSOLUTE THEATER Figure 26.6 Example 6 (single-scope network)
Such single-scope networks are open to further elaboration by bringing additional elements of theatricality into the blend, changing the heretofore single-scope into a double-scope network. A piquant example of this comes from Mel Brooks’s 1981 film, History of the World, Part I, where the Spanish Inquisition is dramatized as a musical. In this instance, Torquemada (played by Brooks) is the lead in a Broadway-style musical. One of his minions introduces him with the following word-play salutation: (7)
Torquemada. Do not implore him for compassion. Torquemada. Do not beg him for forgiveness. Torquemada. Do not ask him for mercy. Let’s face it. You can’t torq-im-odda-anything!
Here, the phonetic form of the character’s Spanish name, Torquemada, is integrated with the English utterance ‘talk him out of anything,’ motivated by a vital relation of similarity. And so ensues an elaborate vaudeville-inspired musical. It is the double-scope blending of absolute rule with musical theater for comic effect. One could say that the mental space for Musical Theater “drowns out” the content of the Spanish Inquisition space. Such instances of human creativity and satire are carnival examples of double-scope mental space networks, where the focus of attention is solely on the blended space, where the humor emerges (see Figure 26.7). In contrast to Figure 26.6, where the source input space remains the dominant focus of attention, this double-scope network focuses extended attention on the blend itself, but with recruitment of additional conceptual structure from scenic spaces of the Musical Theater scenario. Thus,
433
434
OA K L EY A N D PA SC UA L
Theatricality
Absolute Rule
Torquemada
Mel Brooks
Jews
Actors/Singers Swimmers
Nuns
SPANISH H ON N INQUISITION
MUSICAL MU M THEATRE Torquemada as lead in a Broadway musical
Absolute Rule Actors as singing inging Jewish prisoners J p soners
Nuns as synchronized swimmers
ABSOLUTE THEATER Figure 26.7 Example 7 (double-scope network)
the scene of the Spanish Inquisition in the previous example is now the input space for the blend, as opposed to merely one of two referential ‘daughter’ spaces. The fact that the point of the blend is to spoof history rather than use history as a means of teasing out an analogical relation, shifts the entire networks toward Musical Theater. Such phonologic and semantic blends are common, but not restricted to, humoristic discourse (e.g. Coulson 2001 [2005], 2005, Feyaerts and Broˆne 2005, Fujii 2008, Krikmann 2009). They are also frequently used in advertising, where they may also be integrated with visual input spaces (e.g. Lundmark 2005, Joy, Sherry, and Deschenes 2009), and some may occur in everyday language use and even become conventionalized (e.g. “brunch,” “motel,” “nectarine”; see Fauconnier and Turner 2002). Perhaps more surprisingly, syntactic and morphological blending may be fully grammaticalized in language (Fauconnier and Turner 1996, Mandelblit 1997).
26.7 Optimality Principles Conceptual blends achieve their cognitive and communicative effects by variably satisfying different constraints. Fauconnier and Turner (2000, 2002) identify six such optimality principles. Importantly, no blend satisfies each constraint equally well; in fact, the completion and elaboration of blends often requires the relaxation of one constraint in favor of another, as shown below.
Conceptual Blending Theory
The integration constraint posits that a blend forms a single scene or unit and be ‘manipulated’ as such. This constraint reflects the principle of compression. When the father at the dinner table takes the orange and hazelnut and spatially manipulates them into an orbital array, the blended mental conceit of an orange/Sun–hazelnut/Earth are thereby animated as a single scenario, such that the father can now move the hazelnut in an arc and have it signify the movement of the Earth. Material anchor blends must satisfy the integration constraint to a degree more significant than with other blending types. The topology constraint stipulates that it is optimal to preserve the relationships between elements in the input spaces and in the blended space. While it is possible to anchor a dinner-table solar system with salt and pepper dispensers, the fact that an orange and hazelnut are spheroids of differing size has the effect of preserving gross geometric shape in the blended space. (Other instances of material anchor blending relax such iconic properties of their proxies, but it may be easier – especially for children wielding a limited range of categorical strategies – to attend to the blended concept when general topological structure remains active in attention, perception, and working memory.) The web constraint specifies that it is optimal to maintain mappings between the input spaces in the blend. Thus it is critical for the ensuing interaction between ersatz neighbors that they maintain the role-value connections between the input spaces if elaboration of the blend is to enable the non-actual reminiscences of the ‘good old days in the neighborhood.’ If, for instance, one of the discourse participant’s attention drifts, the identity mappings between these two input spaces may be momentarily interrupted, making it difficult for one of them to track the discourse and respond appropriately, requiring a discursive ‘repair’ of the blend. The unpacking constraint stipulates that the blend should supply the basis for inferring the structure of the input spaces. Blends that optimize the unpacking constraint are best exemplified by instances of so-called ‘double grounding blends,’ involving ambiguity, as in the news headlines in (8) and (9) (Broˆne and Feyaerts 2005): (8) (9)
U.S. slowdown punctures Michelin’s profits. The Agnelli family is once again in the driver’s seat at Fiat.
Both examples are puns, exploiting the literal and metaphorical meanings of the verb ‘puncture’ in (8) and of the preposition ‘in the driver’s seat’ in (9). In each case, the blend forces the reader to unpack the contents of each input space (tires/company profits; car’s interior/dominant member of an organization) and dwell for some time on the relationship. That is, the metonymic relationship between tires and the tire company Michelin and between a driver’s seat and the car company Fiat, and the conceptual metaphors M O R E I S U P , L E S S I S D O W N and L E A D I N G A N O R G A N I Z A T I O N I S D R I V I N G A V E H I C L E , respectively (Figure 26.8).
435
436
OA K L EY A N D PA SC UA L
X
Y
UP
MORE
VERTICALITY
QUANTITY Profits
Puncture
US Economic Slowdown MICHELIN
PUNCTURE Figure 26.8 Example 8 (grounded blend: Michelin)
The point of the blend is to serve as a springboard for appreciating the tensions between the input spaces. Psycholinguistic reaction time experiments suggest that double grounding blending is a psychologically real phenomenon, which readers find more cognitively demanding but also more aesthetically satisfying (Broˆne and Coulson 2010). The metonymy constraint stipulates that items related via contiguity or some other part–whole relation compress to identity in the blend. The cartoon in Figure 26.9 exemplifies metonymic ‘tightening’ (Fauconnier and Turner 2000). (10)
Suddenly, Father Schober was not so sure whether he really should have bought the new crucifix at Ikea (accessed from Catholic Humor, May 22, 2015, by Martin Perscheid, www.catholichumor.org/2014/06/suddenlyfather-realized-he-should-not.html)
Here, the clergy cannot help but feel as though he were about to nail reallife Jesus upon the Cross, this despite knowing that he is merely engaging in a bit of do-it-yourself iconographic assembly. The humor comes from the discrepancy between these two scenarios – the quotidian act of assembling furniture and an act of torture. The good reason constraint postulates that any element that happens to emerge in the blend, however incidental to the framing structure of the input spaces, is given significance in the blend. Consider the Dutch advertisement for eggs in Figure 26.10. (11)
Ei love you Hou jij van eieren? Dan heb je geluk. Want eieren houden ook van jou! Lit. Egg/I love you. Do you love eggs? Then you’re in luck. ‘Cause eggs love you too!’ (www.eiloveyou.nl/)
Conceptual Blending Theory
Figure 26.9 Example 10 (Ikea cartoon)
Figure 26.10 Example 11 (‘I love you’)
437
438
OA K L EY A N D PA SC UA L
This advertisement exploits the phonological overlap between the Dutch common noun (‘ei’, ‘egg’, pronounced [ai]) and the English first person singular pronoun (‘I’) to create a fictive interaction blend (see Pascual and Oakley this volume Ch. 21) in which the expression of affection ‘I love you’ is now ascribed to an egg addressing the reader. Of immediate relevance is the image of an egg wearing a knit hat. Why this image? The obvious answer is to personify the egg as an addressor, a sentient being who needs to be protected from the elements. The presence of the hat, otherwise irrelevant for eggdom, materializes as an important indicator of personhood – putting the ‘I’ in ‘ei,’ so to speak. (Notice too that the photographed egg now entertains an intrinsic front, such that it is ‘facing’ the reader, something more difficult to achieve without the hat.) Consider again the Father Schober cartoon (Figure 26.9), in which the initial reaction to it by one of our friends is instructive. After laughing, she uttered ‘i-keel-ya’ [ajkilja] or ‘I kill you,’ inspired by the name of the company, Ikea, as if Father Schober were announcing his intentions to his victim. It is at this point that the cognizer finds additional good reason to utter the company’s name.
26.8 Excursus: Whither the Generic Space? Readers familiar with the literature on conceptual blending will notice something missing from each of these blending analyses, namely the generic space. If one were to read, for example, Fauconnier and Turner (1996, 1998, 2000, 2002), but also Oakley (1998) or Grady, Oakley, and Coulson (1999), one would see both diagrams and text given over to a mental space housing generic structure common to all the spaces in the network. Such stipulations, once commonplace, are becoming less frequent, in part because perhaps no feature of the blending framework has engendered as much controversy among blending theorists themselves (cf. Brandt and Brandt 2005, Coulson and Pascual 2006, Oakley 2012). If one regards mental spaces as essentially scenes and scenarios activated as we think and talk (and not simply ‘conceptual packets’), the generic structure of that space would be an emergent property of direct mental engagement with the scenes themselves; thus a generic space would be an output of the network rather than an input to the network. Given that these mental space networks come and go quickly, are used to illuminate parts of one space at the expense of another, or otherwise are devised for local, online conceptualization, the generic space may just be more an artifact of analysis – a form of mental scaffolding for the theorist – than a reflection of the processes unfolding during discourse production and processing. There is to date, in sum, no consensus among conceptual blending theorists as to the need for a generic space.
Conceptual Blending Theory
26.9 Blending and Discourse Most of the aforementioned examples are attested utterances but with many of the facets of their rhetorical situations redacted for analytic and expository convenience. In this final section we present a full case study of a pharmaceutical commercial from the United States. This analysis exhibits how the conceptual blending framework can systematically capture features of rhetorical situations and their relation to semiotic material that rhetoricians have only investigated impressionistically. The following case study takes its place among a range of similar studies presented in several venues, with Oakley and Hougaard’s (2008) edited volume and two special issues on blending (Coulson and Oakley 2000, 2005, Dancygier 2006) as signal exemplars. But first we discuss the role of context in blending configurations.
26.10 Grounding: Basic and Extended Communication Space Networks All acts of communication take place among situated participants endowed with skills at intersubjective interaction. As Sanders, Sanders, and Sweetser (2009: 25) note: any communicative use of language necessarily involves the presumption that the speaker has mental states, and that she is expressing some content of her mental states, in some speech setting, using some set of linguistic forms.
The authors then introduce the notion of a basic communicative spaces network that participants access ‘for free’ – meaning that such a space of communication is immediately accessible and does not have to be built up or elaborated as other content mental spaces, but which constitutes the deictic center of those mental space networks. Similar but not identical to Langacker’s (1987, 1991, 2008a) notion of grounding and Clark’s (1996) common ground, a basic space of communication operationalizes the basic rhetorical nature of language that has posed a challenge for speakers and analysts of discourse: The space of communication is taken to be nonsalient or lacking in prominence during the speech event (for the ostensible object of the speech events is not the event itself), yet they remain highly prominent, insofar as every symbolic act issues from it, returns to it, thereby altering it. That is, all content mental space networks are anchored to some rhetorical situation, replete with persons, exigencies, and constraints (cf. Bitzer 1969). A brief review of the previous examples should suffice to illustrate the implicit presence of communicative spaces networks that we term ‘grounding spaces’ (cf. Coulson and Oakley 2005, Oakley and Coulson
439
440
OA K L EY A N D PA SC UA L
2008, Oakley and Kaufer 2008).5 In order to understand utterance (1), we need to appeal to a minimal context that includes a bystander overhearing two men talking in English while waiting in line. From this vantage point, the bystander takes this comment to be ‘off the record,’ ‘small talk,’ while waiting in line. But we can push this example a little further, and likewise change the space of communication from a Starbucks queue to the authors’ laptops; the bystander opportunistically retrieves ‘field data’ for a suitable example of mental spaces. What is more, the act of imagining this utterance as the basis for the ersatz neighbor blend is itself an elaboration on the basic communication space from which the utterance first appeared. In addition, example (2) makes little sense absent the intentional communicative situation between father and daughter at the dinner table. This is how the overall context of production and interpretation both motivates and constraints blending operations (Coulson and Pascual 2006). The basic communicative space offers blending theorists interested in discourse the means of modeling or otherwise accounting for the fundamental presumption of communicative intentions in specific situations. While obvious, it becomes clear to anyone working with complex textual artifacts that such communicative spaces can be extended into fictive and fictional variants once the fundamental process of one-person-addressinganother becomes a representational resource for rhetorical elaboration in the satisfaction of given communicative goals.
26.11 Case Study of Symbicort and ‘Big Bad Pharma’: Conceptual Blending in Discourse, Interaction, and Rhetoric Our final example of blending is particularly complex: a televised pharmaceutical advertisement for Symbicort, a medication for alleviating symptoms associated with Chronic Obstructive Pulmonary Disease (COPD). It entails multiple embedded instances of double-scope blending that are rhetorically presented through active and fictive forms of direct speech. The central enabling conceit involves a grandfather reading a picture book of the Three Little Pigs to his young grandson (Figure 26.11). As he exclaims, “And the wolf was huffing and puffing,” the child interjects with “kinda like you sometimes, Grandpa.” He then responds, “Well . . . when you have COPD it can be hard to breathe . . . it can be hard to get air out, which can make it hard to get air in,” at which time the camera shot shifts to inside the story world, with a wheezing Big Bad Wolf wearing the same green sweater as grandpa, unable to blow down the straw house (Figure 26.12).
5
Brandt and Brandt (2005) offer a similar construct known as ‘semiotic space.’
Conceptual Blending Theory
Figure 26.11 Symbicort ad: grandfather and grandson
Figure 26.12 Symbicort ad: wolf with COPD
As the ad proceeds, the grandpa Big Bad Wolf consults a pulmonologist, represented as a she-wolf (Figure 26.13), who then prescribes Symbicort, thus enabling the Big Bad Wolf to help his grandson blow out the candles on his birthday cake (Figure 26.14). In the case of the Three Little Pigs story, an anthropomorphized wild animal (the wolf) antagonizes a pen of domesticated animals (the pigs), a common occurrence, for wolf packs find livestock an easy catch. The critical point from the folktale is that the Big Bad Wolf’s salient weapon of choice is not his fangs but his lungs. He gains access to prey by blowing down the fortress. The blowing is some opportunistic conceit that combines the fact that wolves howl with an implausible possibility of
441
442
OA K L EY A N D PA SC UA L
Figure 26.13 Symbicort ad: wolf and she-wolf
Figure 26.14 Symbicort ad: party
them using that breath to collapse the barrier between predator and prey. This scenario allows for completion and elaboration such that, in the blended scenario of a ‘lupus-genic’ cyclonic wind, there are some materials capable of withstanding such forces better than others. With each disaster and escape the three little pigs follow good engineering principles and build consecutively stronger abodes. While the wolf makes short work of straw and stick houses, he runs into trouble with brick. This fable blend is important for our purposes because the blend itself functions as an ‘input space’ for an elaborated blend.
Conceptual Blending Theory
• Protection from weather & intruders
• Wolves prey on livestock
• Intentionality • Pigs ass livestock l • ToM
ile Domicile
Predation Pred d
• 3 Pigs dwelling in houses made of different materials • Wolf blow blows down houses to access livestock.
Each house gets harder to blow down
Wolf “Blowhard” Figure 26.15 Three Little Pigs fable
Our analysis focuses on the composition of a blend based on selective projection from Domicile and Predation input spaces. In the Domicile space, intentional agents seek refuge in a dwelling from the elements and intruders. It stands to reason that some materials are stronger and more resistant than others: a house made of straw is weaker than a house made of sticks, and a house made of bricks is far stronger than both straw and stick. Builders can imagine intentional mental states of potential intruders and plan accordingly; hence, the elements Theory of Mind (‘ToM’ in Figure 26.15) and Intentionality are salient characteristics of the fabled pigs in the blend. The logic of this space contributes the pragmatic scale of increasing imperviousness, such that straw, stick, and brick ‘line up’ in ascending order of strength. The Predation space contributes the role of Wolf as an apex predator of wild and domesticated animals, but primarily of livestock, such as pigs. In this space, the apex predator seeks to capture prey by any means possible. The blend, which we like to call the ‘Wolf Blowhard’ space, recruits the naturalistic predator–prey relationship between wolves and pigs, while concurrently recruiting knowledge about human dwellings, such that they are intentionally built to offer protection, and that their builders, in anticipating the nefarious intentions of potential intruders (Theory of Mind), select building materials of increasing strength. In the blend, the weapon of choice wielded by the wolf is his breath, and the defensive tactic chosen by the pigs are the building materials. The emergent logic of the blend is that each escape and bivouac of the pigs brings with it a corresponding increase in
443
444
OA K L EY A N D PA SC UA L
Elderly Patient (g)
Big Bad Wolf (bbw)
Pulmonologist (p)
She-wolf (sw)
Three Little Pigs Tale
COPD
bbw’ g’ sw’ p’
COPD Big Bad Wolf
BBW/G cannot even blow down the house of straw, or candles on a birthday cake
Figure 26.16 Big Bad Wolf with COPD (composition and completion)
the imperviousness of the building materials, which, in turn, increases the amount of energy necessary for the wolf to blow the house down. According to legend, this wolf succeeds in blowing down the first two houses, but then fails in his attempts with the brick house. Such is the blended scenario of a predator whose weapon of choice is his lungs. What if his weapon of choice is defective? This entrenched narrative highlights and selectively projects the ‘huffing and puffing’ actions of said wolf, an unusual tactic for apex predation, but one that nicely dovetails with the condition of Chronic Obstructive Pulmonary Disease (COPD), a combination of emphysema and chronic bronchitis. While the wolf’s huffing and puffing is healthy, grandpa’s huffing and puffing is symptomatic of disease. In this blended scenario, the Big Bad Wolf suffers from COPD and cannot even blow down a house made of straw, the easiest of the three houses (Figure 26.16). Here we have a grandfather (senior citizen) who gets transported into the story world of the Three Little Pigs as the Big Bad Wolf. Unfortunately, this elderly Big Bad Wolf suffers from COPD, which does not allow him to breathe normally. He cannot get enough air in or out to help his grandson wolf pup blow out candles on a birthday cake, let alone blow down an entire house made of straw. It is the elaborated concept of a COPD Big Bad Wolf that motivates an additional narrative of a medical consultation with a pulmonologist (lung specialist). In the fable-world of the Big Bad Wolf, the pulmonologist is a she-wolf who then prescribes Symbicort to her patient, since, of course, the story-world of this Anglo-European fable
Conceptual Blending Theory
• (Senior) wolf suffering from shortness of breath • Huffing & puffing are symptoms of COPD • Help from other pack members
• Patient visits pulmonologist • Pulmonologist can prescribe medicine • She/he considers contraindications & side effects
COPD OPD Big Bad Wolf
Pulmonologist • Wolf consults pulmonologist
• Pulmonologist/she-wolf assesses symptoms & history • Prescribes Symbicort Canis lupus Consultation
Big g Bad Wolf breathes normally no ormally & can blow houses down ho
Figure 26.17 Big Bad Wolf with COPD (elaboration)
nevertheless takes place in an Americanized fee-for-service system where access to many drugs is through a physician (Figure 26.17). The network now reaches a state of elaboration, with the Pulmonologist space playing an even greater role in the functioning of the blended conceit. Here the specialist considers different treatment options, as well as the contraindications and possible side effects of each medication. In this enactment, the she-wolf pulmonologist prescribes Symbicort to the elderly Big Bad Wolf, and we begin to see its positive effects. The Big Bad Wolf can now engage in pleasant interactions with his grandson/pup; he now passes by the pigs’ straw house. They notice a healthier and happier Big Bad Wolf and begin to get nervous. The narrative then breaks back to the two-shot between grandson and grandfather, in which the grandfather enacts the wolf at the ad’s end. The complexity of this advertisement inheres not only in the blended conceit of a COPD Big Bad Wolf, but in its narratological structure, for which conceptual blending theory has also been helpful in elucidating. Recent research in blending (e.g. Hougaard 2005, Coulson 2006a, 2006b, Pascual 2008a, Dancygier 2012a, Xiang and Pascual 2016) has come to focus on networks of mental spaces embedded within mental space networks of discourse and interaction. Though the terminology is not settled, there is a consensus that conceptual blending should model the dynamics of discourse and interaction. This advertisement brings out the human complexities inherent in discourse and interaction.
445
446
OA K L EY A N D PA SC UA L
26.12 Communicative Spaces of Symbicort Our analysis of the Symbicort ad leads us to posit four distinct types of Communicative Spaces and their interconnections, captured in Figure 26.18. There are four discourse formations for evoked scenes of speech, constituting the rhetorical situation of this advertisement. Our exposition thereof follows an internal to external trajectory. Grounding space alpha (α) represents the establishing scene we call ‘story time’ (similar to the Story Viewpoint Space in Dancygier 2008b, 2012a). Here, a grandfather and his grandson sit on a couch reading the story of the Three Little Pigs. Everything in this space is happening before the viewer’s eyes in real time. Grounding space beta (ß), on the other hand, provides default schemas of doctor/patient interactions. In this particular instance, we have the grandfather (a senior citizen, with all the connotations that evokes) and a pulmonologist. In the narrative, the events of this encounter are past relative to alpha, as this interaction forms the basis of the grandfather’s ostensible reported speech to his grandson. But there is a third grounding space enabled by the blended conceit, wherein the grandfather is the COPD Big Bad Wolf and his pulmonologist is a she-wolf. There are also other potential and incidental discourse participants, such as the wolf-pup, as well as the grandmother she-wolf and other members of the family/wolf pack, depicted as
Story time • Grandpa • Grandson
Medical Consultation • Patient • Pulmonologist
• Present → γ • Future → β • Past → δ
• Past → α
Grounding β
Grounding α Characters • Big g Bad Wolf
• Wo Wolf Pup • She- Wolf Pulmonologist • • • •
Pack members Co-present w/ α Future relative → β Past → δ
Grounding γ Figure 26.18 Symbicort ad: grounding spaces
•
Mass-media Broadcast Astra Zenaca
• Viewer • Physician (your doctor)
• Future → α, β, γ
Grounding δ
Conceptual Blending Theory
attending a birthday party, and let us not forget the three little pigs hiding in the straw house within earshot of the main character. This grounding space is co-present alpha but future relative to beta. This grounding space is represented through visual information, with the she-wolf being the only wolf character speaking. The dominant communicative mode in the gamma (γ) space is through pantomime, facial expression, and gesture. The final grounding space is delta (δ), which corresponds ontologically to the advertisement itself. This space oscillates between being co-present with alpha, but its illocutionary force makes it logically future relative to all the grounding spaces. That is, when the advertisement asks the viewer to “ask your doctor if Symbicort is right for you,” it is, in effect, scripting a future encounter with the viewer’s own internist, presumably one who is not a she-wolf. It anticipates the content of your next medical consultation. Such scripted interactions are commonplace proleptic blends in pharmaceutical advertising. The integration of multiple discourse formations is not altogether seamless. Recall that virtually everything related to the audience is ostensibly emanating from the story time space. The grandfather begins by relating his medical condition to his grandson, which retains prosodic features of an adult talking to a young child (e.g. higher vocal register, elongated vowels, and other forms of exaggerated vocalization), but with adult content. The speech of the she-wolf pulmonologist is then represented as reported speech of the conversation with the grandfather. However, the physician recommends that “you should tell your doctor if you have a heart condition or high blood pressure before taking it [Symbicort],” which breaks the frame of doctor/patient consultation, since she is supposed to be speaking directly to her patient, but is, in fact, speaking to the viewer – the viewer’s role shifts from bystander to addressee. Importantly, the doctor’s speech is in voice-over and not directly animated, as it was when introducing the product to the COPD wolf. The network is as complex as can be, and still effective in attaining its persuasive communicative goal.
26.13 Conclusion Conceptual blending is a developing and thriving area of research in cognitive linguistics, such that nearly all who specialize in the fields of discourse, literary analysis, and rhetorical analysis must be familiar with its basic operations and principles. We have outlined just these basic operations and principles in the opening sections of this chapter with only intermittent reference to examples appearing elsewhere in the literature. Our second goal was to focus attention on the processes involved in
447
448
OA K L EY A N D PA SC UA L
understanding the dynamics of blending as a discursive and rhetorical phenomenon, with a special focus on pharmaceutical advertisements appearing on United States television and streaming systems. Our analysis emphasizes that: 1) mental spaces are best regarded as scenarios and scenes within more elaborate networks, and 2) the same dynamics of construing content also apply to the construal of acts of communication itself, necessitating, as they do in the above case study, to unpack the complex scenes and scenarios of interaction we experience in these complex but quotidian cultural products. Conceptual blending constitutes a general framework and model for grasping the details of these activities.
27 Embodiment Raymond W. Gibbs, Jr.
27.1 Introduction Embodiment is a hot topic these days within cognitive linguistics, and cognitive science more generally. Many aspects of language and thought refer to bodily experience in one form or another. For example, political discourse is a great place to find instances of bodily language. Consider part of a newspaper editorial regarding California tax law and Gov. Jerry Brown:1 As a string of bill vetoes last week shows, if Gov. Jerry Brown doesn’t like something, it won’t happen. That’s why it’s disappointing to hear his caustic, flip statements ridiculing any attempt to change Proposition 13, the state’s ironclad lid on property taxes . . . Brown has the leadership clout, skill and political capital to fix the imbalance that Prop. 13 causes. But he wants to play it safe, dodging an issue that undermines California’s long-term fiscal health.
This narrative refers to several bodily actions, such as flip statement, clout, imbalance, play it safe, dodging an issue, along with other figurative phrases, such as ironclad lid on property taxes, which are all intended to convey metaphorical meanings. Embodied metaphors have historically been treated as linguistic anomalies that can only be interpreted via special, effortful comprehension processes (Gibbs 1994). After all, linguistic processing was assumed to be a specialized module of mind that is divorced from bodily experience, which presumably makes it more difficult to understand bodily based speech and writing. These beliefs arose from a traditional view of mind within cognitive science that sees it as being quite distinct from the body, even if many scholars readily agree that human minds have important connections to human brains. Still, early definitions of ‘cognition’ refer to intelligent 1
www.sfgate.com/opinion/editorials/article/Gov-Brown-dodges-Prop-13-reform-6562906.php.
450
GIBBS, JR.
activity that can be physically manifested within many different devices, ranging from thermostats, computers sitting on your desk, and robots, to human bodies. The study of cognition was analogous to the study of flying. Cognitive scientists were interested in cognition regardless of its physical instantiations just in the way that one could try to characterize flying regardless of whether it was in the context of jets, birds, butterflies, or paper airplanes. Over time in the history of cognitive science, scholars became increasingly aware of how different kinds of intelligent behavior were tied to the physical presence of the organism or artifact under study. For example, the fact that human beings have the bodies they do and move about in the world in particular ways fundamentally constrains how they think and adapt to real-world challenges. There may be some commonalities between organisms that live and move on their own compared to computational devices that do not (e.g. thermostats and non-movable computers). Still, human evolutionary history, including that related to bodily forms and capabilities, gives human beings a specific viewpoint of understanding and acting in the world which specifically shapes human cognition and language. My claim in this chapter is that cognitive linguistic research over the past thirty years has played a major role in alerting scholars to the constraining presence of embodiment in human meaning-making and action. Many cognitive linguistic studies have also been extended to demonstrate the importance of embodiment in meaningful expressions within other non-linguistic and multimodal domains (e.g. gesture, film, music, and dance). However, cognitive linguistic ideas on embodiment may be limited by the traditional focus on language on the page, which misses critical facets of embodied experience that guides real-life communicative interactions. Addressing these concerns will be critical as cognitive linguistics moves forward better to characterize the constraining role that bodies in action, often in collaboration, have in the creation and performance of linguistic and non-linguistic meaning.
27.2 Defining Embodiment Embodiment refers to the ways persons’ bodies and bodily interactions with the world shape their minds, actions, and personal, cultural identities. Embodied accounts of mind and language embrace the idea that human symbols are grounded in recurring patterns of bodily experience, and therefore reject traditional dualistic, disembodied views of human cognition and linguistic meaning. The study of embodiment demands recognition that thought and language arise from the continuous dynamic interactions between brains, bodies, and the world. Cognitive linguistic research has long embraced the idea that “Our construal of reality is likely
Embodiment
to be mediated in large measure by the nature of our bodies” (Evans and Green 2006: 2). The journal Cognitive Linguistics, for instance, explicitly ties its themes to embodiment in emphasizing “the experiential background of language-in-use.” Various bodily aspects, including anatomical, physiological, perceptual, neurological, and genetic functions of the body, constrain the range of our linguistic abilities and the specific character of our meaningful experience. Lakoff and Johnson (1999) suggested that there are three levels of embodiment which together shape the embodied mind. Neural embodiment concerns the structures that characterize concepts and cognitive operations at the neurophysiological level. The cognitive unconscious consists of the rapid, evolutionarily given mental operations that structure and make possible conscious experience, including the understanding and use of language. The phenomenological level is conscious and accessible to consciousness and consists of our awareness of our own mental states, our bodies, our environment, and our physical and social interactions. Many cognitive scientists agree that an embodied understanding of mind and language require attention to all three levels of embodiment, and most importantly, the interactions between them (Gibbs 2005a). Cognitive linguistic analyses of language have explored all three levels of embodiment as possibly motivating the creation and maintenance of linguistic structure and behavior. It is important to note, though, that all these general views of embodiment emphasize human bodies as individual entities. Each person presumably understands the world and his or her experiences in it from a solitary point of view given his or her own bodily sensations and actions. However, the individual orientation misses critical dynamics of bodies in interaction, something which is critical to our ability to participate in shared meaning-making activities (Gibbs 2005b, Blomberg and Zlatev 2014). This is an important issue for cognitive linguistics to more completely explore in future research studies (see below).
27.3 Embodied Metaphor Cognitive linguistic research primarily explores systematic relations in language to infer underlying linguistic, cognitive, and neural structures that motivate the creation and use of different words and linguistic expressions. Several early studies, for example, demonstrate the importance of people’s spatial thinking in specific linguistic structures (Langacker 1987, Talmy 1988a, Langacker 1990a). Still, studies of embodiment in cognitive linguistics have mostly arisen in the context of research on metaphorical thought and language. For example, consider the following shortlist of conventional expressions:
451
452
GIBBS, JR.
(1)
Greta is making good progress toward her PhD degree. John has already reached several career goals. David ran into a rough patch trying to solve the difficult math problem. Sandra was completely stuck figuring out what to do after her divorce.
Lakoff and Johnson (1980a, 1999) argued that these expressions are conventional manifestations of an underlying metaphor in thought, namely the L I F E I S A J O U R N E Y conceptual metaphor. The schematic phrase L I F E I S A J O U R N E Y represents only a convenient summary description of the rich set of mental mappings that characterize the complex relationship between target (L I F E ) and source (J O U R N E Y ) domains. The L I F E I S A J O U R N E Y conceptual metaphor gives rise to a diverse set of mappings, which include, at the very least, the following source to target domain correspondences (Ko¨vecses 2010). travelers → people leading a life motion along the way → leading a life destination(s) → purpose(s) of life different paths to one’s destination(s) → different means of achieving one’s purpose(s) distance covered along the way → progress made in life locations along the way → stages in life guides along the way → helpers or counselors in life These correspondences are not simply a list of the features in common between L I F E and J O U R N E Y , precisely because conceptual metaphors help establish a more structured conceptualization of the target domain, which people then speak of using mostly conventional words and phrases. People understand parts of their lives in terms of their varied bodily experiences of taking journeys, which provide the embodied grounding of source domains in metaphorical concepts.
27.4 Image Schemas The source domains in conceptual metaphors are primarily imageschematic because they emerge from recurring patterns of embodied experience. Image schemas can generally be defined as dynamic analog representations of spatial relations and bodily movements in space (Johnson 1987). For example, the image schema S O U R C E –P A T H –G O A L , which underlies the J O U R N E Y source domain, emerges from a variety of bodily experiences, such as when a person starts moving from one point toward another along some path with the intention of reaching a specific destination or whenever we reach out to grab hold of an object (i.e. reaching from a starting point, moving along a path, reaching and grabbing the object), or when we move our eyes from focusing on one object in the world across to another (i.e. moving from a source or starting point along some path to reach a goal).
Embodiment
Common linguistic expressions such as John walked from home to the store are understood as conveying a S O U R C E –P A T H –G O A L meaning. This image schema, like all others, can be metaphorically extended to create, in this case, journey metaphors related to expressions such as His career was off-track and Their relationship hit a dead-end street. Cognitive linguistic studies have suggested that there are likely many dozens of image schemas (e.g. Johnson 1987, Ko¨vecses 2010), and these likely play a role in motivating a wide variety of linguistic structures and meanings (e.g. from prepositions to narratives) (see Clausner and Croft 1999, Hampe 2005, Oakley 2007). Different behavioral evidence from psycholinguistics, cognitive psychology, and developmental psychology supports the possibility that image schemas have important psychological validity in several linguistic and non-linguistic domains (Gibbs 2005a, Gibbs and Colston 1995, 2012). For example, people have intuitive understandings of how different image schemas arise from bodily experience and help motivate the various related meanings of polysemous words (Gibbs 1994). Thus, people recognize that certain metaphorical uses of the word stand (e.g. I can’t stand my job) are related to non-metaphorical uses (e.g. The chair can’t stand on three legs) because of common image-schematic structures. Experimental studies also show that image schemas play a role in the immediate processing of verbs in sentence comprehension. For instance, people exhibit a strong tendency to associate an upward vertical image with the word respect, and a horizontal right-arrow image with point at (Richardson et al. 2003). This pattern was found both when people had to choose the most relevant image from a set of four, as well as when participants could create their own images on a computer. A later study presented people with auditory sentences such as The girl hopes for a pony while they visually saw a picture of either the girl and pony together in a vertical or horizontal manner. Not surprisingly, people responded faster when the girl and pony were arranged in a vertical manner (i.e. pony above girl) than when the two were seen in a horizontal (i.e. side-by-side) manner. These findings suggest that the spatial, image-schematic character of verb meanings are inferred even when these refer to abstract events. Other recent experiments provide evidence that image schemas also shape the processing of verb-particle constructions (VPC), including both spatial senses (e.g. The plane touched down) and non-spatial senses (e.g. The car has broken down) of the last word, or particle (Yang 2016). These data suggest that people’s understandings of VPC were influenced by the image schema representations for non-spatial or metaphorical statements. One possibility is that VPCs are understood as entrenched constructions, in the sense of Construction Grammar (Goldberg 2006), yet are still tied to their embodied motivations via image schemas. Many debates exist over the exact status of image schemas as representational entities in human minds (Hampe 2005, Mandler and Paga´n Ca´novas
453
454
GIBBS, JR.
2014). Most cognitive linguists view image schemas as basic cognitive structures that can be accessed during language use, even if these entities no longer retain their roots in embodied experience. Others see image schemas as emergent products of meaning construction processes in context, rather than as stored entities in long-term memory (Gibbs 2005b). Image schemas may also reflect cultural ideas and may not have a purely bodily grounding (Kimmel 2005, Zlatev 2005). Finally, recent proposals even suggest that images schemas are topological neural mappings which assist in the construction and understanding of both concrete and abstract ideas (Lakoff and Narayanan, in prep.).
27.5 Primary Metaphors One of the important developments in understanding body and language relations comes from the study of primary metaphors (Grady 1997a, Grady, Oakley, and Coulson 1999). Certain conceptual metaphors are more primary than others in human thought and experience. Strong correlations in everyday embodied experiences lead to the creation of ‘primary’ metaphors, which include some of the following examples: INTIMACY IS CLOSENESS
We have a close relationship. DIFFICULTIES ARE BURDENS
She’s weighed down by responsibilities. AFFECTION IS WARMTH
They greeted me warmly. IMPORTANT IS BIG
Tomorrow is a big day. MORE IS UP
Prices are high. SIMILARITY IS CLOSENESS
Those colors aren’t the same, but they’re close. STATES ARE LOCATIONS
I’m close to being in a depression. CHANGE IS MOTION
My health has gone from bad to worse. PURPOSES ARE DESTINATIONS
He’ll be successful, but isn’t there yet. CAUSES ARE PHYSICAL FORCES
They pushed the bill through Congress. KNOWING IS SEEING
I see what you mean. UNDERSTANDING IS GRASPING
I’ve never been able to grasp complex math.
Embodiment
These metaphorical correlations arise out of our embodied functioning in the world. In each case, the source domain of the metaphor comes from the body’s sensorimotor system. A primary metaphor is a metaphorical mapping for which there is an independent and direct experiential basis that can be expressed within language, as seen in the above list. A complex metaphor, on the other hand, is a self-consistent metaphorical complex composed of more than one primary metaphor. For instance, the three primary metaphors P E R S I S T I N G I S R E M A I N I N G E R E C T , S T R U C T U R E I S P H Y S I C A L S T R U C T U R E , and I N T E R R E L A T E D I S I N T E R W O V E N can be combined to create a conceptual metaphor T H E O R I E S A R E B U I L D I N G S that nicely motivates the metaphorical inferences that theories need support and can collapse, etc., without any mappings such as that theories need windows. The study of primary metaphors in cognitive linguistics has investigated many dimensions of thought and language (Ko¨vecses 2010). Primary metaphors are often claimed as providing cross-cultural and cross-linguistic linkages given their fundamental bodily basis, something that human beings all presumably share. Nonetheless, primary metaphors may be combined in varying ways, as shown above, to give rise to varying complex metaphors across different cultural contexts (Lakoff 2008). Primary metaphors may serve as critical inputs to conceptual integration processes, as seen in conceptual blending theory (Grady 2005). However, there are also debates over whether primary metaphors are really metonymic, and not metaphorical, because of the way source domains come to represent, or stand-for, target domains (Ko¨vecses 2013). Much recent research from cognitive and social psychology is also consistent with the presence of primary metaphors in non-linguistic experience. For instance, linguistic examples show that people often express their thoughts about similarity in terms of spatial distance as in Their opinions in this case couldn’t be further apart, which many interpret as being motivated by the primary metaphor S I M I L A R I T Y I S C L O S E N E S S . One study showed that people judged two words (e.g. grief and justice) as being more similar when they were closer to one another than more distant (Casasanto 2008a). Similarly, people in the same series of studies decided that two colored squares were similar in color faster when the squares were closer together, while dissimilar colors were judged faster at longer distances apart. When people physically engage in certain actions, this can also lead them to experience primary metaphors that influence their social judgments. Having people hold warm, as opposed to cold, cups of coffee for a few minutes led them subsequently to judge another person’s interpersonal traits as being warmer (Williams and Bargh 2008), a finding that is consistent with the primary metaphor A F F E C T I O N I S W A R M T H . Asking people to recall an immoral deed, as opposed to an ethical one, made them more likely to choose an antiseptic wipe, rather than a set of pencils, as a free gift after the experiment (Zhong and Liljenquist 2006). These
455
456
GIBBS, JR.
empirical results are entirely consistent with people’s understandings of the primary metaphors G O O D I S C L E A N and B A D I S D I R T Y . When asked to determine whether a fictitious person is suitable for a job, people judged job applicants to be better if they were also holding a heavier, rather than a lighter, clipboard (Ackerman, Nocera, and Bargh 2010), which surely reflects the common idea that I M P O R T A N C E I S W E I G H T . Finally, one study had participants recount positive or negative autobiographical experiences as they moved marbles upward or downward between two cardboard boxes (Casasanto and Dijkstra 2010). People spoke more quickly about positive memories when moving the marbles up, and negative memories when they moved the marbles down, a result that is consistent with two primary metaphors for positive (e.g. G O O D I S U P ) and negative (e.g. B A D I S D O W N ) experiences. These experimental results are representative of a large literature demonstrating the role that primary metaphors have not only in language, but also in different nonlinguistic experiences (Landau, Robinson, and Meier 2015). Simply experiencing an embodied source domain (e.g. feeling warm) can automatically prime people to think about a specific target domain (e.g. affection).
27.6 Neural Theory of Language The neural theory of language is a project aimed at determining the neural substrates of thought and language (Feldman 2006, Gallese and Lakoff 2005, Bergen and Chang 2013, Lakoff and Narayanan, in prep). Several computational models have been created that demonstrate how sensorimotor information may be activated in precise ways during reasoning and language understanding. In this way, the neural theory attempts to establish links between the neural and cognitive unconscious levels of embodiment. These studies formally explore the implications of cognitive neuroscience research showing the existence of ‘mirror neurons’ in the pre-motor cortex that are activated when people merely see specific actions, imagine doing those actions, and even hear language referring to those actions (Gallese and Lakoff 2005). Mirror neurons associated with grasping become active when people see others grasping objects, when they imagine grasping objects, or when they hear the verb grasp, to take one instance. Thus, people understand abstract verb meaning (e.g. grasp the concept) by simulating sensorimotor actions. One neural model, called KARMA (knowledge-based action representations for metaphor and aspect), characterized how we may reason about economic events using embodied metaphors, including those related to moving steadily, falling, taking small steps, plunging deeper, struggling, and starting to emerge (Narayanan 1997). KARMA is capable of drawing various appropriate inferences about economic events (e.g. related to
Embodiment
different goals and aspects) via mental simulation of the physical actions referred to in economics texts. This work provided an existence-proof of how a structured neural network used to control high-level motor schemas also operates during abstract reasoning and linguistic interpretation. A more recent development in the neural theory claims that certain neural pathways, called ‘cascades’ are critical to the creation and understanding of specific verbal metaphors (Lakoff and Narayanan, in prep). Cascades control activations across various regions of the brain which can be combined to create more complex neural networks, such as seen in motor control and decision-making. These neural pathways are the basis on which embodied metaphors may be combined with specific frame-based knowledge of specific situations to create very specific verbal metaphors. For example, a cascade containing primary metaphors such as K N O W I N G I S S E E I N G , I D E A S A R E E N T I T I E S , C O M M U N I C A T I O N I S E N T I T Y T R A N S F E R , M U L T I P L E X I S M A S S , P U R P O S E S A R E D E S T I N A T I O N S , and D I F F I C U L T I E S A R E O B S T R U C T I O N S T O M O T I O N can connect with situational knowledge to produce metaphorical expressions such as spill the beans, and let the cat out of the bag. The neural theory of language has received empirical support from experimental research in cognitive neuroscience. To take just a few examples, studies have explored the neural processing of metaphorical action phrases, such as grasp the concept, and revealed nearly identical activation in motor areas of the brain as when people interpret literal action statements (e.g. John grasped the straw) (Desai et al. 2011). Listening to taste metaphors (e.g. She looked at him sweetly) showed an increased activation in brain areas most associated with gustatory perception compared to when people heard non-gustatory paraphrases (e.g. She looked at him kindly) (Citron and Goldberg 2014). Finally, listening to metaphors based on tactile experiences (e.g. Sam had a rough day) demonstrated increased activation for tactile regions of the brain’s sensorimotor region compared to when people heard nonmetaphorical paraphrases (e.g. Sam had a bad day) (Lacey, Stilla, and Sathian 2012). All these findings are consistent with the idea that embodied metaphor understanding involves modality-specific neural processing. Several cognitive scientists take issue with an embodied perspective on language understanding (Mahon and Caramazza 2008, Pinker 2007, Hickok 2009). They question whether thought and language use are necessarily modality-specific as opposed to operating within amodal areas of the brain. For example, Casasanto and Gijssels (2015) argued “that source-domain representations are implemented in multimodal or amodal brain areas, not in modality-specific simulations.” This possibility raises doubts about the embodied, image-schematic nature of linguistic understanding and nonlinguistic reasoning. However, specific image-schematic concepts are manifested across several bodily modalities and are not restricted to a single dimension of neural experience. Consider the metaphorical expression Sam had
457
458
GIBBS, JR.
a rough day. Our experiences of roughness may be more closely related to the modality of touch, yet we also have sensations of roughness for taste, vision, audition, and even smell, similar to the way that smooth can be used metaphorically to talk about a smooth taste, a smoothing tone of voice, and so on. These observations pose a challenge for neuroscience studies that insist on modality-specific effects if verbal metaphor processing, for example, is believed to be embodied. Future studies are needed to uncover more distributed effects across several regions of sensorimotor cortex as we move forward in characterizing the neural basis for language use. Moreover, even if sensorimotor brain areas have some role in understanding of abstract language, it is not always clear exactly when and how this knowledge is accessed and integrated during online language processing.
27.7 Embodied Simulation in Language Understanding Psycholinguistic studies have also explored the role that embodied action plays in language production and understanding. Most of these studies present findings that are consistent with the ideas from the neural theory of language and, more specifically, claims from simulation semantics. One general theory, called the embodied simulation hypothesis, claims that embodied simulations enable people to project themselves into the minds and actions of others, including the objects and events referred to in texts. People create embodied simulations of speakers’ messages that involve moment-by-moment ‘what this must be like’ processes, which make use of ongoing tactile–kinesthetic experiences. For example, studies on the ‘action-compatibility effect’ indicate that people are faster to make comprehension responses for sentences like John opened the drawer when they have to move their hands toward their bodies to push a comprehension button than when they have to move their hands away from their bodies (Glenberg and Kaschak 2002). The reverse pattern of results was observed when people heard sentences implying movement away from the body such as John closed the drawer. People interpret sentences by imagining themselves engaging in the very actions specified in the language, which in turn directly influences their embodied comprehension task (e.g. moving their hand to push a comprehension button). Similarly, another study presented participants with sentences such as The carpenter hammered the nail into the wall. After reading the sentence, participants saw a picture of a nail in either the vertical or horizontal position and had quickly to judge whether that object was mentioned in the sentence. People were faster to make their ‘yes’ decisions when the picture was in the same spatial orientation implied by the sentence just read (Zwaan, Stanfield, and Yaxley 2002). Thus, people were faster to say ‘yes’ when the picture showed the nail in the horizontal orientation than
Embodiment
when it was shown upright, or in the vertical position. However, those who have first read the sentence The carpenter hammered the nail into the floor were faster to say ‘yes’ to the nail picture that presents it in a vertical position than to one in the horizontal orientation. Readers drew these inferences not simply because of their abstract knowledge of the world, but because they imagined themselves hammering a nail in a particular spatial manner. Experimental studies also indicate that embodied simulations are constructed incrementally during speeded sentence comprehension. Consider the following statement: Before/the/big race/the driver/took out/his key/and/ started/the/car. The implied direction of the key turn in this case is clockwise. In one experiment, participants read through each sentence by rotating a knob after each chunk of words, indicated by the slashes in the above example (Zwaan and Taylor 2006). Participants turned the knob in either a clockwise or a counter-clockwise manner. The notable result was that people were faster to read and comprehend the verb started when they made their knob turns in a clockwise direction than when making counter-clockwise rotations. People essentially understand the key verb started by constructing an embodied simulation of the implied movement the car driver had to perform in order to turn this key and start the engine. This illustrates that people do not wait till the end of the sentence to initiate their simulation processes. Embodied simulation processes, quite notably, help people understand metaphorical language referring to actions that are impossible physically to perform in the real world. Consider the simple abstract phrase John couldn’t grasp the concept of infinity. A concept is an abstract entity and it seems odd to think that one can physically grasp something that does not physically exist. But scientific research has convincingly revealed that people perform embodied simulations when interpreting verbal metaphors (Gibbs and Colston 2012). Understanding abstract events, such as grasping the concept, is constrained by aspects of people’s embodied experience as if they are immersed in the discourse situation, even when these events can only be metaphorically realized. For example, studies show that having people first make a hand movement, such as reaching out to grasp something, subsequently facilitates the speed with which they comprehend a metaphorical phrase such grasp the concept (Wilson and Gibbs 2007). Even if people are unable physically to grasp a concept, engaging in relevant body actions primes the construction of an embodied simulation to infer the metaphorical meaning of grasp the concept. Thus, people interpret the word concept as a metaphorical object that can be grasped, studied, and understood. Having people perform, or even imagine performing, a relevant bodily motion, such as tearing something apart, enhances the embodied simulations constructed when understanding metaphorical phrases. Creating embodied simulations when hearing metaphorical statements, such as Your relationship was moving
459
460
GIBBS, JR.
along in a good direction, can also affect people’s subsequent bodily actions, such as walking toward a target object (Gibbs 2013). These different experimental results highlight the degree to which people use their perceptual and motor systems for simulation purposes (see Matlock and Winter 2014). In some cases, the meaning products that one infers when understanding a speaker’s utterance will be relatively crude, primarily because this set of products will be ‘good enough’ for the purposes at hand. At other times, people may engage in more elaborate, even highly strategic, embodied simulation processes as they tease out numerous meanings and impressions from an utterance in context. Reading literary texts, such as fiction and poems, prompts people to create fuller, even more embodied, simulations which underlie our aesthetic interpretations and appreciation of poetic discourse (Gibbs and Blackwell 2012, Willems and Jacobs 2016).
27.8 Embodiment in Nonlinguistic Domains Research on embodied language and thought within cognitive linguistics has had a profound influence on the study of embodiment in many nonlinguistic domains. A vast number of studies have explored the various roles that image schemas, primary metaphors, embodied conceptual metaphors, and embodied simulations play in the ways in which people produce gesture, mathematics, cartoons, photographs, ads, films, music, dance, and other expressive acts (Lakoff and Nu´n˜ez 2002, Mu¨ller 2007, Cienki and Mu¨ller 2008a, Forceville and Urios-Aparisi 2009, Zbikowski 2009, Kappelhoff and Mu¨ller 2011, Stec and Sweetser 2013, Coegnarts and Kravanja 2015, Gibbs 2016b). These empirical analyses show in detailed ways how people’s tacit understandings of embodied experience, and many of its metaphorical extensions, are critical to the creation and performance of many nonlinguistic behaviors. In some cases, these nonlinguistic images or events are expressed in coordination with speech or writing that is also motivated by enduring patterns of bodily activity. More broadly, however, this work, again emerging from cognitive linguistic theories, offers compelling evidence for the claim that embodied concepts are not purely linguistic in their origin, but are psychologically real characteristics of how people think and engage in many kinds of nonlinguistic expressive action, contrary to the arguments of Murphy (1996) and Pinker (2007).
27.9 Some Limitations and Future Directions One limitation of much cognitive linguistic work on embodiment is that it derives from the analysis of ‘language on the page.’ An analyst looks at
Embodiment
a linguistic expression (e.g. John overcame many obstacles on the way to getting his PhD), determines its communicative meaning, and from this posits an embodied, conceptual structure that presumably motivates the existence of the statement (e.g. L I F E I S A J O U R N E Y ). Underlying this reverse engineering practice are two guiding assumptions. The ‘full specificity’ of meaning assumption maintains that cognitive linguistic analyses should characterize the entire conceptual, embodied motivation for some linguistic expression. The ‘discreetness’ of meaning assumption holds that the analysis of each word or utterance is separate from any other expression encountered. Thus, a standard cognitive linguistic analysis of embodied, meaning must specify all the relevant embodied motivations for an utterance and characterize these motivations as they apply to the full word or expression on the page. Both of these assumptions are questionable, however, in the context of real-time language use. First, people do not necessarily compute or infer the full meaning of any word or expression regardless of the overall context. Instead, they create representations that are just ‘good enough’ for the present purposes. Thus, people need not construct a complete analysis of the embodied image schemas and metaphors, for example, which presumably motivate the creation of some linguistic expression, each time that statement is encountered. Similarly, the kinds of embodied simulations people construct when understanding language may vary given changing circumstances. Listeners may, for example, adopt different viewpoints in construing the meaning of a spoken utterance or written narrative, some of which is surely guided by linguistic style (Dancygier and Sweetser 2014). Exactly what sort of embodied understanding people construe for any discourse will always depend on the situation, the goals and motivations of the participants, and the specific understanding task (Gibbs and Colston 2012). The idealistic, canonical analyses of embodied motivations for language do not necessarily reflect what real people in real situations understand about linguistic meaning. Second, people do not approach language from the analyst’s perspective, but interpret meaning in an incremental manner (i.e. word-by-word). The continuous processing of language also includes meaning products that are interpreted from previous discourse or from currently present gestural or prosodic information. People’s production and understanding of language never begins at some context-free point. After a person understands some utterance in context, the brain/mind/body does not go blank to process the next utterance from a neutral perspective. Human cognition and language are always situated and influenced by the history of people’s experiences up to that particular moment in time, including, once again, a variety of other bodily information given the current conversational interaction. Third, the interactive nature of ordinary language use suggests that embodiment is not located within individual brains and bodies, but is
461
462
GIBBS, JR.
shared and coordinated across conversational participants. Much cognitive science research demonstrates, for instance, the emergent, implicit coordination between people in terms of their speech, prosody, gaze, posture, gestures, and body positioning during conversation (Shockley, Richardson, and Dale 2009). It is precisely for this reason that the very concept of ‘embodiment’ must be understood as distributed and in constant flux given the dynamics of any interpersonal interaction. One implication of this idea is that embodied simulations are constructed from a range of bodily actions, especially those that are emergent from how two or more speakers attempt to coordinate in any moment in time. Some scholars are calling for “an interactive and multimodal reappraisal of embodied language comprehension” in order better to capture simulation semantics, for instance, within real language use (Kok and Cienki 2014: 1). What do these limitations imply for the future practice of cognitive linguistics? First, it will be important to acknowledge the probabilistic, partial nature of embodied influences on thought and language in the context of real language use. Recognizing that embodied meaning is rarely fully specified requires attention to bodies in interaction over time within any discourse situation. Moreover, the analysis of the embodied motivations for linguistic expressions must not start from scratch, but should seek out how the context of previously understood meaning, many of which are also embodied (e.g. previously inferred image schemas or embodied metaphors in the same conversation), shapes meaning construal. Online linguistic processing, because of its continuous nature, does not start once an utterance has been completely read or heard. Cognitive linguists would do well, then, to characterize the unfolding of meaning construction in a more incremental manner (e.g. word-by-word given the context). Finally, the coordination people achieve in discourse is not simply a matter of matching up fully specified embodied representations within each person’s head or mind. The collaborative nature of embodied meaning dynamically arises from the interacting influence of many different bodily segments (e.g. gesture, tone of voice, posture) that are, once more, distributed across individuals. Acknowledging these embodied complexities, ranging from incremental processing to bodily coordination across individuals, will not require that all cognitive linguists become psychologists. Still, cognitive linguistic analyses can become more sophisticated better to recognize the diverse ways in which embodied meanings both arise historically and are assembled in very contextually specific forms.
28 Corpus Linguistics and Metaphor Elena Semino
28.1 Introduction: The Relevance of Corpus Linguistics to the Study of Metaphor Research on metaphor increasingly exploits, in a variety of ways and for a variety of purposes, the computer-aided methods developed within corpus linguistics. In this chapter, I begin by briefly introducing corpus linguistics and its relevance to the study of metaphor. In section 28.2, I introduce the different types of corpora and corpus-linguistic methods that can be applied to the study of metaphor. In section 28.3, I discuss the different types of contributions that corpus methods have made to metaphor theory and analysis. In section 28.4, I provide a concrete example of corpus-based research on metaphor, focusing on the use of metaphor to express the experience of pain. In section 28.5, I provide some concluding remarks. Corpus linguistics involves the construction of large digital collections of authentic texts (corpora) and their investigation through dedicated software tools (e.g. McEnery and Hardie 2012). The methods of corpus linguistics were initially primarily applied to the study of lexis and grammar, but have recently been extended to a wider range of areas, including discourse analysis, translation studies and (first and second) language acquisition, as well as other branches of the humanities and social sciences. Why and in what ways, however, are corpus methods relevant to the study of metaphor? In principle, the understanding of any linguistic phenomenon can benefit from being systematically analyzed in large quantities of naturally occurring data, that is, from the kind of analysis that corpus methods make possible. This applies particularly to any phenomenon that is claimed to be frequent in language, and that is given center stage in theorymaking at least in part because of its frequency. Metaphor is such a phenomenon, particularly as it is viewed within cognitive linguistics.
464
SEMINO
Lakoff and Johnson’s (1980a) Conceptual Metaphor Theory (CMT) based its claim that metaphor is central to thought on the pervasiveness and conventionality of metaphor in language. In Lakoff and Johnson’s Metaphors We Live By, and in many subsequent CMT studies, linguistic examples are the main or sole type of evidence that is provided for the existence of particular conceptual metaphors as mappings between ‘source’ and ‘target’ domains in conceptual structure. Subsequent developments of CMT, as well as some alternative theoretical accounts of metaphor in cognition, are also founded, at least in part, on linguistic evidence, such as Grady’s (1997a) theory of primary metaphor and Fauconnier and Turner’s (2002) account of metaphor in Blending Theory. The same applies to much work on metaphor across languages and cultures, and to claims about the ‘universality’ or otherwise of particular metaphors (e.g. Ko¨vecses 2005). The linguistic evidence provided in early work within CMT, however, consisted of decontextualized examples invented or remembered by the authors themselves. In contrast, over the last few decades, research on conceptual metaphors has increasingly made use of authentic data, including by studying patterns of metaphor use in a variety of electronic corpora. Later in this chapter I describe in detail the theoretical contributions that this change of methodology has made possible. The CMT account of metaphor has also inspired, more or less directly, a large amount of research on the frequencies, forms, and functions of metaphor in particular texts, text types, or discourses. This kind of work varies in the extent to which it explicitly contributes to metaphor theory, but is firmly based in the analysis of language use in context. Here corpus methods tend to play an important role, often in combination with qualitative analysis, as I show below.
28.2 Different Corpora and Tools for the Corpus-based Study of Metaphor Before discussing the contribution made by corpus-based studies to metaphor theory and analysis, in this section I discuss the different kinds of corpora and tools that have been exploited in the study of metaphor.
28.2.1 Types of Corpora In corpus linguistics, a corpus is generally defined as: [a] set of machine-readable texts which is deemed an appropriate basis on which to study a specific set of research questions. The set of texts . . . is usually of a size which defies analysis by hand and eye alone within any reasonable timeframe. (McEnery and Hardie 2012: 1–2)
Corpus Linguistics and Metaphor
A wide range of corpora are used by metaphor scholars, depending on the goals of the research. Some studies make use of large pre-existing generalpurpose corpora, such as the 15 million-word Italian Reference Corpus (Deignan and Potter 2004), the 100 million-word British National Corpus (e.g. Stefanowitsch 2006b) and the much larger Bank of English Corpus (e.g. Deignan 2005). These studies tend to make or test generalizations about metaphor use in a whole (national) language, such as British English, or across two languages, such as English and Italian. In contrast, much research on metaphor is concerned with the forms, functions, and implications of metaphor use in particular texts or text types. In the former case, corpus methods are applied to digital versions of complete existing texts, such as the Bible and the Koran in Charteris-Black (2004) or Sylvia Plath’s Smith Journal in Demje´n (2015). When the focus is metaphor use in a particular genre or register, the analysts may either study a subsection of a larger existing corpus, or, more frequently, construct their own corpus or corpora. Skorczynska and Deignan (2006) for example, compare the use of metaphor in two specially constructed corpora of business research articles (e.g. Management Science) and articles from business periodicals (e.g. The Economist). Demmen et al. (2015) analyze the use of V I O L E N C E metaphors in a dedicated corpus of interviews with and online forum posts by patients with cancer, family carers, and healthcare professionals. Similarly, L’Hoˆte’s (2014) study of the metaphors and narratives of New Labour in the UK involved the construction of three comparable corpora of texts produced by different UK political parties, including manifestos and leaders’ speeches. These specially constructed corpora are seldom larger than a million words, but are nonetheless big enough to allow generalisations for the relevant text types. Finally, the extent of variation in the size of corpora exploited in metaphor research can best be appreciated by considering studies that lie at opposite ends of the continuum. Steen et al.’s (2010) work, for example, is based on the manual analysis of four samples from different sections of the British National Corpus that are small enough to be manually annotated for metaphor (approximately 50,000 words each). In this case, corpus methods were only used to search the annotations themselves. At the other extreme, Veale (2012) creatively used the whole World Wide Web as a corpus in order to study, among other things, the forms and uses of similes in English.
28.2.2 Types of Corpus Tools Variation can also be observed in the specific corpus tools that are employed to find and analyze occurrences of metaphor in corpora of different sizes (see also Stefanowitsch 2006a: 2–6). In spite of continued progress in automatic metaphor identification (e.g. Mason 2004, BerberSardinha 2010, Neuman et al. 2013), most corpus-based studies of
465
466
SEMINO
metaphor involve searching the data for words or phrases that are likely to be used metaphorically, or to occur in close proximity to relevant uses of metaphor. ‘Concordancing’ tools provide each instance of the search term in the corpus on a separate line, accompanied by the immediately preceding and following co-text. Some studies involve the concordancing of expressions that are likely to be used metaphorically at least some of the time in the data, that is, vocabulary associated with a particular source domain or vehicle grouping. Deignan (2005), for example, concordanced animal terms such as rabbit and squirrel in the Bank of English to study the realization of the Animal source domain in English. Semino et al. (2015) concordanced words such as journey and path in a corpus of online posts by people with cancer, in order to study the use of J O U R N E Y metaphors by this particular group of patients. Where the focus of the research is the metaphorical construction of a particular topic or experience, on the other hand, the selected search terms realize the target domain of the potential metaphors. Stefanowitsch (2006b), for example, concordanced in the British National Corpus a set of emotion words such as happiness and sadness in order to test and refine the claims made about metaphors for emotions in CMT. In both types of approaches, the co-text of each occurrence of the search term has to be carefully scrutinized, first, in order to see whether the search term or some surrounding words are in fact used metaphorically, and second, to identify any patterns that are relevant to the goals of the study. When specialized corpora are used, it is not uncommon for a small representative sample of the data to be analyzed manually for relevant metaphorical expressions first, and then for those expressions to be concordanced in the whole corpus (e.g. Charteris-Black 2004). Different approaches can be used to identify more open-ended sets of metaphorical expressions. Some studies involve the concordancing of ‘signalling expressions’ (Goatly 1997) or ‘tuning devices’ (Cameron and Deignan 2003), that is, words or phrases that are often found in close proximity to metaphorical, or generally figurative, uses of language. Cameron and Deignan (2003), for example, concordance expressions such as sort of and like in order to investigate metaphor use in corpora of difference sizes. Veale (2012) employed standard search tools to look for structures associated with similes in the World Wide Web, including, for example, ‘as. . . as.’ A different technique involves the exploitation of semantic annotation tools such as the USAS tagger, which is part of the online corpus comparison software Wmatrix (Rayson 2008, ucrel.lancs.ac.uk/wmatrix/). The USAS tool annotates each word in a corpus with a tag for the main semantic domains it belongs to. For example, the word war is allocated to the semantic domain ‘Warfare, defence and the army; weapons.’ Where a semantic domain in USAS is likely to correspond to an important metaphorical source domain in a corpus, this tool can be used to search for
Corpus Linguistics and Metaphor
open-ended sets of metaphor candidates. This enables the analyst to go beyond the limitations of traditional lexical concordances (involving preselected words or phrases; see Koller et al. 2008). For example, concordancing the word war in a 500,000-word corpus of online writing by cancer patients (Semino et al. 2015) provides all nine occurrences of that word in the corpus, which can then be checked for metaphoricity. In contrast, concordancing the ‘Warfare’ USAS tag in the same corpus results in 181 occurrences of words that share that tag. This leads to the identification of a much wider set of war-related metaphors for cancer, including expressions such as chemo veteran and invaders for cancer cells (see Demmen et al. 2015, Semino et al. 2015). In a study of metaphors for mental states in Sylvia Plath’s Smith journal, Demje´n (2015) uses the same technique to search both for vocabulary that relates to potentially relevant source domains (e.g. the USAS semantic domain ‘Darkness’) and for targetdomain vocabulary (e.g. the USAS semantic domain ‘Thought, belief’). These kinds of studies also often tend to involve a first stage where a set of relevant semantic domains is identified via a manual analysis of a subsection of the corpus, and a second stage where these domains are systematically investigated in the whole corpus by means of the semantic tagger. Some recent studies additionally incorporate ‘keyness’ analyses as part of their methodology. In corpus linguistics, ‘keywords’ are words that are used much more frequently (to a statistically significant extent) in one’s corpus of data as compared with a (usually larger) reference corpus. Similarly, for semantically annotated corpora, a key semantic domain is a domain that is statistically significantly overused in one’s corpus of data in comparison with a reference corpus (Rayson 2008). Particular expressions or semantic domains do not have to be ‘key’ in a corpus to be of interest for the purposes of metaphor analysis, as not all instances of that expression/domain may be metaphorically used in the corpus under analysis, the reference corpus, or both (e.g. news data is always likely to include frequent literal references to wars). However, when a metaphorically used expression or domain is key in a corpus in statistical terms, strong claims can be made about the dominance of that particular kind of metaphor in that data. For example, in L’Hoˆte’s (2014) New Labour corpus, both the words tough and strong and the USAS semantic tag ‘Tough/strong’ are key as compared with a corpus of Labour texts from an earlier period. This suggests that metaphors to do with physical strength are systematically used in the New Labour corpus to counter the party’s earlier ‘soft’ image in British politics. Similarly, some signaling devices for metaphoricity are keywords in Partington, Duguid, and Taylor’s (2013) corpus of opinion pieces in British newspapers, as compared with a corpus of reviews in the Times Literary Supplement. This finding is used to support a more general claim that humorous metaphors and similes are much more frequent in the former corpus than the latter.
467
468
SEMINO
A few studies also make use of collocation tools for metaphor analysis. The ‘collocates’ of a word are words that are used unusually frequently in close proximity to that word. Collocations can be calculated on the basis of different measures of statistical significance (see Brezina, McEnery, and Wattam 2015). L’Hoˆte (2014), for example, considers the collocates of the metaphorically used word tough in her New Labour corpus, in order to see what policy areas and initiatives are described in these terms in her data (e.g. the New Labour slogans tough on crime). Semino (2008) considers the collocates of the adjective rich in the British National Corpus to test Lakoff’s (1993) claim that the expression a rich life is evidence of the existence of a conventional conceptual metaphor A P U R P O S E F U L L I F E I S A BUSINESS. Having discussed the types of corpora and tools that can be employed for metaphor analysis, I now turn to the main contributions that this kind of approach has made to the study of metaphor.
28.3 Contributions of Corpus-based Studies of Metaphor Corpus linguistics is an approach to the study of language that is not associated with any general or specific theory. However, the use of corpus methods rests on the assumption that actual linguistic behavior is not only worth studying systematically, but also needs to be accounted for by any theoretical model of language. As such, corpus linguistics is particularly consistent with the usage-based models of language that have been proposed within cognitive linguistics. When it comes to metaphor in particular, corpus methods have been used to contribute to metaphor theory and analysis in a number of ways. In this section, I start from some general contributions, especially concerning CMT. I then consider the findings that corpus methods have led to in studies that approach metaphor from historical, cross-cultural, discoursal, and practical perspectives.
28.3.1 General Implications for Metaphor Theory In the first book-length study of metaphor from a corpus linguistic perspective, Deignan (2005) provided evidence from the Bank of English corpus to support the most general claims made in CMT, especially concerning the frequency of conventional metaphorical expressions, and the ways in which they form patterns that can be taken as evidence for particular conventional conceptual metaphors. However, Deignan also showed systematically how the de-contextualized linguistic examples used in CMT are often not sufficiently representative of actual language use to be cited as evidence for claims about conceptual metaphors. In addition, Deignan used corpus methods to reveal patterns of metaphor use that were not considered or accounted for in (conceptual) metaphor
Corpus Linguistics and Metaphor
theory, at least up to that point. These include, for example, some systematic associations between particular source domains and linguistic metaphors belonging to different word classes, as well as the fact that some words have different conventional metaphorical senses for different morphological inflections (e.g. rock as a singular noun vs. rocks as a plural noun). Deignan also found evidence for a greater influence of target domains in metaphorical mappings than suggested by Lakoff’s (1993) Invariance Hypothesis. Some subsequent corpus-based studies have questioned or refined specific claims about conceptual metaphors. Semino (2008) finds no evidence in the British National Corpus for Lakoff’s (1993) claim that the expression rich life is a realization of the conceptual metaphor A P U R P O S E F U L L I F E I S A B U S I N E S S . Rather, corpus evidence is provided for a different conventional pattern that accounts for the expression rich life alongside other metaphorical uses of rich, such as rich soil and rich culture. Stefanowitsch (2006b) conducts a systematic analysis of the metaphorical expressions that co-occur with emotion words in the British National Corpus. He provides broad support for earlier claims about emotion metaphors in CMT (Ko¨vecses 2000). However, he also reveals patterns that were not discussed in previous studies, and provides information about the frequencies of different metaphors in British English, as represented in the corpus. Stefanowitsch’s (2006b) study also empirically addresses the question of whether there are emotion-specific metaphors, and what metaphors can be described as such. Studies such as Musolff (2006), in contrast, use corpus evidence to make suggestions about the level of generality at which claims about conceptual metaphors are best made. Musolff analyzes the use of metaphor in a bilingual English–German corpus of news reports about the EU. He finds that the use of metaphor in the corpus is best accounted for not in terms of mappings involving broad conceptual domains such as Marriage, but rather in terms of more specific ‘scenarios,’ such as End-of-honeymoon and Adultery. More generally, corpus-based studies of metaphor tend to reveal linguistic patterns that are difficult to account for in terms of CMT, and of metaphor theory more generally. For example, as mentioned above, the singular noun rock tends to have a positive meaning when used metaphorically (the rock on which society is built), while the plural rocks tends to have a negative meaning (The marriage has been on the rocks for a while) (Deignan 2005: 158–59). Furthermore, some A N I M A L metaphors seem to occur only as nouns (e.g. cow as a derogatory term); others only occur as verbs (e.g. horsing around); yet others occur both as nouns and verbs, but not necessarily with the same meaning (e.g. racist pigs and pigging out on food) (Deignan 2005: 153). In German, Boot (‘boat’) and Schiff (‘ship’) have similar literal meanings but different conventional metaphorical uses: Boot is used in expressions such as being ‘in the same boat’ or to describe a place as
469
470
SEMINO
having no space for newcomers; in contrast, Schiff tends to be used as part of metaphorical descriptions of difficult enterprises (Zinken 2007). Such findings from corpus-based studies have considerable theoretical implications. In her 2005 book, Deignan comments that “[t]he data discussed so far suggest that each linguistic metaphor has a life of its own” (Deignan 2005: 166). On the basis of patterns such as Boot vs. Schiff in German, Zinken (2007) proposes the notion of ‘discourse metaphors’ – form-meaning pairings that develop in communication as part of a ‘conceptual pacts’ among interlocutors, and that constitute a mid-point between novel metaphors and fully conventionalized ones. Other metaphor scholars explain these findings in terms of Dynamics Systems Theory (e.g. Cameron 2011). From a dynamic systems perspective, the meanings and functions of metaphorical expressions arise from the dynamic interaction of lexico-grammatical, semantic, cognitive, pragmatic, and affective factors in actual contexts of communication (e.g. Cameron and Deignan 2006, Gibbs 2016a). Within this account, conceptual metaphors are only one of the factors involved in the development of metaphorical meanings, including conventional ones. Johansson Falck and Gibbs (2012) specifically show the relevance of embodied simulations associated with particular words as one of these interacting factors. Like Zinken (2007), they consider a pair of words (road and path) that have similar literal meanings, but different metaphorical meanings. Corpus evidence from the British National Corpus shows that path is typically used metaphorically to refer to ways of living, and often to suggest (potential) difficulties. In contrast, road tends to be used metaphorically to describe purposeful activities. Johansson Falck and Gibbs compare the corpus findings with the mental imagery that people reported in a questionnaire about their experiences with paths and roads. The triangulation of participant responses and corpus findings suggest that people’s embodied experiences with the entities referred to by path and road can explain the (different) metaphorical meanings that the two words have in the corpus.
28.3.2
Corpus-based Findings on Historical, Cultural, and Register Variation in Metaphor Use Corpus methods are also contributing to the study of variation in metaphor use in the history of a language, across languages, and in different types of texts. The availability of historical corpora of English in particular has led to several diachronic studies of metaphor. As Tissari (2001: 238) puts it, these studies show both “stability and change” in metaphor use over time. Tissari (2001) compares some of the central metaphors for love as an emotion in corpora representing, respectively, Early Modern English and Present-Day English. An analysis of concordances for the string ‘love’ in both corpora showed that the main broad source domains that are applied
Corpus Linguistics and Metaphor
to love appear to be the same (C O N T A I N M E N T , A M O U N T , and E X C H A N G E ). However, some differences can be observed in how these source domains are realized. For example, E X C H A N G E metaphors seem to be realized more often by words associated with agriculture in the Early Modern English corpus than in the Present-Day English corpus. This reflects a decrease in the cultural centrality of agriculture over time. Similar findings arise from studies that exploit the Historical Thesaurus of English to study conceptual mappings across the history of the language. Alexander and Bramwell (2014) show, for example, how the metaphorical description of wealth in terms of large body mass (e.g. fat and the now archaic cob) has decreased over time, as the positive association of looking well-fed have waned. These studies often have broader theoretical implications. In a diachronic analysis of metaphors for intelligence in the Historical Thesaurus, for example, Allan (2008) finds evidence to support the importance of the metonymic basis of many conventional metaphors. Some of her findings also question the claim that metaphors tend to map concrete experiences onto abstract ones. For example, folk etymology may suggest that dull as a description of lack of intelligence involves a concrete-toabstract mapping from the source domain of (lack of) physical sharpness. However, historical evidence reveals that the word was first used for lack of intelligence, and thus suggests that it is the more concrete physical use that was derived from the one that relates to intellectual abilities (Allan 2008: 186). Many corpus-based cross-linguistic studies of metaphor have built on and extended the work of Ko¨vecses and others on the relationship between metaphor and culture, broadly conceived (e.g. Charteris-Black 2003, Deignan and Potter 2004, Chung 2008). These studies also tend to reveal both similarities and differences in the metaphors used in languages associated with different cultures. The similarities often confirm the embodied basis of many metaphors. The differences emphasize cultural relativity, not least in perceptions of the body itself. Simo´ (2011), for example, builds on Charteris-Black’s (2001) research on English to compare the metaphorical use of blood in corpora of Hungarian and American English. She finds that the main target themes for blood metaphors are the same: E M O T I O N and E S S E N C E . However, there are differences in the connotations and frequencies of different metaphorical expressions involving blood. For example, the expression in cold blood/hidegve´rrel is always negative in American English but equally split between positive and negative uses in Hungarian. In Hungarian, hidegve´rrel is also particularly frequent in sports reports. More generally, Simo´ finds evidence of within-culture variation by comparing blood metaphors in each section of each corpus. Indeed, a large number of studies have used corpus methods to study the forms and functions of metaphor in different text-types, including news reports (e.g. Koller 2004, Musolff 2004), political speeches (Charteris-Black 2005), religious texts (Charteris-Black 2004), and so on. These studies often
471
472
SEMINO
tend to aim to answer questions that pertain to those text-types and their contexts of use, rather than or in addition to furthering metaphor theory. For example, Koller (2004) shows how the dominant S P O R T S and W A R metaphors in reports about the economy are both closely linked to each other and a reflection of a sexist bias. Charteris-Black (2012) uses corpus methods to analyze metaphorical expressions in a large corpus of interviews with people with depression. Among other things, he shows that men and women tend to use similar metaphors, but suggests that the few differences that do emerge may have relevance in the therapeutic process. More generally, these studies reveal the potential pitfalls involved in making generalizations about metaphor in language and cognition without taking into account variation at the level of speakers/writers, registers, genres, and discourses (see also O’Halloran 2007, Deignan, Littlemore, and Semino 2013). A more specific contribution of corpus-based studies of metaphor in particular text-types is the systematic investigation of the functions of metaphors in discourse. L’Hoˆte (2014) shows the way in which metaphor is used to project a political party’s new identity in a corpus of New Labour texts in the UK from 1997 to 2007. Partington, Duguid, and Taylor (2013: 131ff) concordance tuning devices in a corpus of reviews in British broadsheets and show the ways in which creative, humorous metaphors and similes are used to entertain the reader. These expressions tend to involve incongruous comparisons and proper names of various kinds, and to rely heavily on the readers’ cultural knowledge (e.g. a sort of Goldilocks on crack and an unexpected, juicy cross between meaty liquorice and Noggin the Nog’s bum). Veale’s (2012) study of similes on the World Wide Web reveals, among other things, how certain simile structures tend to be used creatively and humorously. For example, the structure ‘about as. . . as. . .’ is shared by many ironic similes, especially when the adjective following the first as is positive (e.g. about as useful as a chocolate teapot). Corpus-based approaches to particular figurative phenomena within specific genres can also have broad relevance for metaphor theory. This is the case with Dorst’s (2015) work on similes in fiction and two recent studies on metaphor clusters and mixed metaphors (Kimmel 2010, Semino 2016). In particular, Semino (2016) shows how people often use the label ‘mixed metaphor’ to describe the use, in close proximity, of metaphorical expressions that actually involve the same broad source domain, such as the source domain of Music in: to keep things on a different note, instead of the same record getting played over and over again. This suggests that people are particularly sensitive to contrasts between nearby metaphors when those metaphors are similar enough to each other for their differences to be consciously noticed.
28.3.3 Corpus-based Approaches to Metaphor and Practice Finally, some corpus-based approaches to metaphor have implications for practice in different areas. Skorczynska and Deignan’s (2006) comparison
Corpus Linguistics and Metaphor
of a corpus of business-research articles and a corpus of businessperiodicals articles revealed differences that are relevant to the choice of teaching materials for non-native speakers who are preparing for higher education courses in English. Skorczynska and Deignan suggest that the use of magazines such as The Economist is not sufficient to prepare students for reading academic papers on business and economics. Skorczynska (2010) similarly found that the metaphors in a Business English textbook differ considerably, both in kind and in frequency, from those found in a corpus of specialist business articles. These studies have implications for the choice of teaching materials in English for specific purposes, as well as for a further appreciation of variation in metaphor use depending on audience and text-type. A corpus-based study of metaphors for cancer and the end of life at Lancaster University, in contrast, has implications for communication in healthcare. The study involved a combination of manual and computeraided analysis of a 1.5 million-word corpus of interviews with and online posts by patients with advanced cancer, family carers, and healthcare professionals. A variety of patterns of metaphor use were identified by means of a combination of lexical and semantic concordances. Among other things, the analysis showed differences among patients and healthcare professionals in the use of V I O L E N C E metaphors such as fighting cancer, with patients using them significantly more frequently than healthcare professionals. The study also showed that, as many have argued (e.g. Sontag 1979), violence-related metaphors can be detrimental to patients’ morale and self-esteem, as when a patient with a terminal diagnosis says I feel such a failure that I am not winning this battle. However, for a substantial minority of patients, these metaphors can be empowering, as when a patient says Cancer and the fighting of it is something to be proud of (see Demmen et al. 2015, Semino et al. 2015). These findings suggest a degree of individual variation in the use of metaphor for illness that is relevant to communication training and practice in healthcare (see Demje´n and Semino 2017). Overall, this section has shown the variety of contributions that corpusbased approaches can make to the understanding of metaphor in language and cognition. In the next section I provide a more extensive example of corpus-based analyses of metaphor in relation to physical illness.
28.4 Corpus Approaches to Metaphor and Pain The experience of physical pain is well known to be difficult to articulate in language (e.g. Scarry 1987). English, for example, has relatively few words that specialize in the description of pain sensations, and all of them are quite general (e.g. pain, painful, hurt, ache, aching, sore). This poses a challenge in healthcare, particularly when pain becomes chronic: in that
473
474
SEMINO
context, diagnosis, and treatment rely to a significant extent on the sufferer’s ability to express their pain, and on healthcare professionals’ sensitivity to patients’ descriptions. It is also well known, including among clinicians, that figurative language is an important resource in articulating pain sensations. In the words of a consultant at the National Hospital for Neurology and Neurosurgery in London: Attempts to truly describe pain indeed appear as difficult as they are frustrating, yet the need to communicate is overwhelming, and I suggest that the only option available is the resort to analogy . . . (w)hether by means of metaphor or simile. (Schott 2004: 210)
Ko¨vecses (2008) discusses the main metaphors used in English to convey pain experiences, from the perspective of CMT. He observes that “pain is conceptualized metaphorically in terms of its potential causes” (Ko¨vecses 2008: 28). More precisely, pain sensations tend to be expressed metonymically or metaphorically in terms of causes of damage to the body. For example, the expression a burning pain is used metonymically when the pain is caused by skin contact with a flame, and metaphorically when the pain does not result from contact with sources of heat, but can be compared with the sensation caused by such contact (e.g. a burning pain caused by acidity in the stomach). Ko¨vecses (2008) proposes a number of specific conceptual metaphors that are all part of this broad pattern: A sharp stab of pain made her sit back down. A massive killing pain came over my right eye . . . I clawed at my head trying to uproot the fiendish talons from their iron grip. P A I N I S F I R E Pain is fire that can devour the whole body. (Ko¨vecses 2008: 28; emphasis in original) PAIN IS A SHARP OBJECT
PAIN IS A TORMENTING ANIMAL
The connection with metonymy emphasizes the main motivation for such metaphors. The most prototypical and intersubjectively accessible kind of pain results from damage to bodily tissues, as in the case of cuts, burns, etc. Other types of more subjective and invisible pain (e.g. migraine) are described metaphorically in terms of properties or processes that cause damage to the body (e.g. a splitting headache). In Semino (2010) I discuss these metaphors from the perspective of theoretical and empirical work on embodied simulation. I also point out that language-based diagnostic tools for the diagnosis of pain also heavily rely on linguistic metaphors that draw from a wide range of causes of bodily damage. The McGill Pain Questionnaire (Melzack 1975), for example, requires patients to choose appropriate descriptors from a list that includes words such as stabbing, burning, drilling. In the same study, I also provide corpus evidence to support and refine Ko¨vecses’s (2008) claims about conventional conceptual metaphors in English. The top sixty-five collocates of the word pain in the British National Corpus include eight expressions that have literal, basic meanings to do with causes of bodily
Corpus Linguistics and Metaphor
damage: searing, sharp, stabbing, lanced, seared, stabbed, stinging, burning.1 The corpus data provides evidence for a conventional conceptual metaphor that can be expressed as P A I N I S C A U S E O F P H Y S I C A L D A M A G E , which can in fact be seen as a ‘primary’ metaphor, in the terms used by Grady (1997a). The metaphors proposed by Ko¨vecses (2008) for different kinds of bodily damage can be seen as specific-level variants of this more general and basic metaphor (see also Lascaratou 2007 for metaphors and metonymies for pain in Greek). In addition, an analysis of concordance lines for pain shows that some expressions that are included in the McGill Pain Questionnaire are never or seldom used in the corpus to describe pain sensations (e.g. taut). This may cast some doubts on the decision to include such expressions in the questionnaire. Corpus methods can be further exploited to investigate how chronic pain sufferers use figurative language creatively to express particularly intense and distressing experiences. As an example, I searched for the structure ‘feels like’ in a 2.9 million-word corpus of contributions to an English language online forum for sufferers of Trigeminal Neuralgia.2 Trigeminal Neuralgia (TN) is a rare but distressing condition that causes unpredictable pain in the face of such intensity that all aspects of the sufferer’s life can be affected. The search for ‘feels like’ in the corpus returned 254 occurrences of the expression. As I expected, most instances function as a tuning device, that is, they introduce a simile that describes the writer’s experience of a TN attack, such as the following (NB: original spellings have been retained but proper names have been changed): (1) (2) (3)
feels like needles poking my eyes and ears sometimes it feels like I am covered in insects and this can be more distressing than the pain At the moment, it feels like someone of considerable weight (not you jenny lol) is standing on the side of my face, wearing stilletto’s whilst pouring red hot liquid into my ear. When I get the bolts it feels like someone has decided that I deserve to be cattle prodded!
Example (1) is a relatively creative instantiation of the conventional tendency to describe pain in terms of the insertion of sharp objects into the body (cf. Ko¨vecses’s 2008 P A I N I S A S H A R P O B J E C T ). Example (2) does not involve pain as such, but an uncomfortable symptom of TN that is described as more distressing than the pain itself. The expression covered in insects primarily conveys a physical sensation, but also has associations of repulsion and disgust, which can account for the distressing nature of the experience. Example (3) is one of many examples in the corpus where different causes of physical damage
1
Collocates of pain were computed on the basis of log-likelihood (Dunning 1993) and within a window span of one
2
I am grateful to the UK Trigeminal Neuralgia Association for allowing me access to their online forum.
word to the left and one word to the right of the search string.
475
476
SEMINO
(pressure, penetration with a pointed object, heat, electric shock) are combined into a rich, dynamic and creative metaphorical scenario (see also Deignan, Littlemore, and Semino 2013: 279ff.). This kind of description goes well beyond the single-word options included in language-based diagnostic tools such as the McGill Pain Questionnaire. A systematic analysis of conventional and creative figurative descriptions of pain, in both general and specialized corpora, can therefore be used both to assess the validity of existing pain questionnaires and to inform the creation of future language-based diagnostic tools.
28.5 Conclusions Corpus approaches have made and are continuing to make a variety of important contributions to the study of metaphor as a linguistic and cognitive phenomenon. These contributions involve not just metaphor theory, but also the understanding of communication in a variety of contexts, as well as practice in areas such as education and healthcare. The chapter has also shown the ingenuity and methodological eclecticism involved in corpus-based studies of metaphors. Researchers do not just use a variety of corpus tools, but do so creatively, in order to identify the widest possible variety of potential metaphorical expressions in their data. An initial manual analysis often provides the springboard for computer-aided analysis. Even more importantly, decisions about metaphoricity and about the meanings and functions of metaphors in the data involve detailed qualitative analyses of the output of corpus tools. Two final points are in order, which I have not been able to do justice to in the course of the chapter. First, experimental approaches to the study of metaphor (e.g. Bowdle and Gentner 2005, Casasanto 2008b) can benefit from using as stimulus materials authentic and, ideally, frequent linguistic expressions: corpora are a potentially useful source of such materials. Second, any large-scale linguistic investigation will come across metaphorical uses of language. Therefore, corpus-based studies of language generally can benefit from, or possibly even require, the findings and insights that are developed by metaphor scholars. It is no coincidence, for example, that metaphor is the topic of one of the guides that followed the creation of the first corpus-based dictionary of English, the Collins Cobuild dictionary (Deignan 1995). No systematic corpus-based investigation of word meanings and discourse patterns in any language can avoid dealing with the role of metaphor in the lexicon and in language use. Future corpus linguistic research on metaphor is likely to involve the use of ever more sophisticated corpus and computational tools, the analysis of larger and more varied corpora and further theoretical, analytical, and practical advances.
29 Metaphor, Simulation, and Fictive Motion Teenie Matlock
29.1 Introduction People constantly imagine physical motion, all sorts of motion. They imagine volitional agents, including themselves, walking, riding bicycles, and reaching out for objects. They imagine non-volitional agents in motion, too, such as clouds floating across the sky and water gushing out of faucets. People imagine motions they have done in the past, such as running across a finish line or hiking up a hill. They also imagine things they have never seen before, such as a trapezoid spinning in mid air. People are also able to imagine various types of movement along a path, such as fast or slow, easy or labored, or straight or meandering. They can mentally experience motion tacitly, with no clear mental image. This chapter discusses how people’s experience with physical motion figures into their understanding of fictive motion and other kinds of metaphorical motion. As will be discussed, their ability to mentally enact movement underlies their understanding of figurative spatial descriptions that use motion language but express no motion, such as A scar runs down his leg and Interstate 5 goes from Sacramento to Los Angeles. Such language is pervasive in English and other languages, and has received much attention in cognitive linguistics. This chapter begins with an overview of metaphors that feature motion as a source domain, and moves on to discuss fictive motion, including past theoretical and recent cognitive science work on how it is used and understood. It goes on to discuss mental simulation and its role in processing such language, and then ends with ideas for future research.
29.2 Metaphorical Motion People use metaphor because it allows them to understand things that are relatively abstract in terms of things that are relatively concrete. Metaphors
478
M AT L O C K
do this by drawing on people’s knowledge of the physical world, including their implicit knowledge of the body and space (see Lakoff and Johnson 1980a, Johnson 1987, Lakoff 1987, Gibbs 1994, Lakoff and Johnson 1999). Given the primacy of movement in everyday human experience, motion serves as a good source domain for other, more abstract domains (Lakoff and Johnson 1980a, Lakoff 1993). In metaphors that have a source domain of motion, the source domain often implies some kind of directed movement along a path, often toward a goal. Certain aspects of movement can be emphasized and mapped on to a target domain, such as direction, as in The proposal is going up to the top and The proposal is coming down from the top, and force, as in Let’s charge ahead with this plan and Let’s ease ahead with this plan. Also, the target domain is often non-physical, as in I’ll come back to that part of the argument later and Go slowly through the early part of your presentation. Motion metaphors share many features like these, and exist as part of a large system of metaphors that involves the mapping of Location/Motion to Event Structure (for basic overview, see Dancygier and Sweetser 2014). One popular motion metaphor, l o v e i s a j o u r n e y , is seen in statements such as Their marriage is moving in the right direction and Their relationship is a smooth ride toward marital bliss. With this conceptual metaphor, a romantic relationship is tacitly construed as a vehicle that contains the lovers and ‘moves’ along a path toward their common goals. When goals change or there is trouble, the path can become bumpy and difficult, or both parties in the relationship can take different paths, reflected in statements such as Their relationship hit a rough patch or The young couple decided to take different paths (for comprehensive discussion of l o v e i s a j o u r n e y , see Lakoff 1993, Gibbs and Nascimento 1996). t i m e i s m o t i o n is another ubiquitous metaphor with a source domain of motion. This metaphor is reflected in statements such as We are moving closer to Election Day and Election Day is coming soon. In both statements there is ‘movement’ along a path; in the first case, it is ‘movement’ through time (ego-moving construal), and in the second, ‘movement’ of time (timemoving construal) (for discussion of t i me i s m o t i o n , or the more general t i m e i s s p a c e metaphor, see Clark 1973, Radden 1997, Evans 2004, Moore 2014; see also Sweetser and Gaby this volume Chapter 38). Another popular motion metaphor is p o l i t i c a l c a m p a i g n s a r e r a c e s , reflected in statements such as The politicians are sprinting toward the finish line and Obama moved ahead of Romney. With this metaphor, political candidates are construed as ‘moving’ toward a finish line, a landmark that represents victory. The competitors can ‘move’ quickly or slowly, and they can move ahead or fall behind each other in the race (see Matlock 2013). In all three of these metaphors and many other motion metaphors, there is, of course, no physical motion along a path. Motion is used to structure our understanding of a relatively abstract domain (romantic love, time, political campaign).
Metaphor, Simulation, and Fictive Motion
In some cases of figurative language and thought, motion along a path is used to structure our understanding of physical space. Initially this seems counterintuitive because physical space is inherently concrete and often serves as a source domain. Why is motion recruited to describe what is concrete? The next section will discuss fictive motion, a type of figurative language that uses motion along a path to describe static physical form. Such language is worthy of attention because the target domain is seemingly concrete, simply the configuration of a physical object in physical space. In addition, it is pervasive across languages, and many interesting claims have been made about how it is processed.
29.3 What is Fictive Motion? The figurative statements The road goes from Fresno to Modesto as well as A tunnel runs through the mountain convey information about the shape and configuration of a physical object. They express that an object is linearly extended through physical space (e.g. road extends from Fresno to Modesto). Though motion language is used, no overt motion occurs. Such statements are often found in descriptions of physical objects, such as geographical or geological structures (see Davis 2013), architectural structures (Caballero 2006), and archaeological artifacts (see Di Giuseppantonio Di Franco, Matthews, and Matlock 2015). They are often found in descriptions of routes in guidebooks (see Tsai and Hsieh 2013) and in mathematics education (see Marghetis and Nu´n˜ez 2013), and they occur in many languages, including English (Talmy 1996), Spanish (Rojo and Valenzuela 2003), Japanese (Matsumoto 1996), Thai (Takahashi 2001), Hindi (Mishra and Singh 2010), Finnish (Huumo 1999), Estonian (Taremaa 2013), Yukatek Maya (Bohnemeyer 2010), and Serbian (Stosic and Sarda 2009). Leonard Talmy was the first cognitive linguist to analyze the conceptual structure of such statements. He argued that its interpretation involves an implicit, fleeting sense of motion despite an absence of explicit movement (see Talmy 1996, 2000a). For instance, with The road goes from Fresno to Modesto, the reader or listener experiences a sense of movement along some portion of the path to Modesto. Talmy provided a taxonomy of various sources of motion for fictive motion, including coextension path fictive motion (also known as linear path), as in The road runs along the coast and The mountain range goes from Canada to Mexico, shadow emanation path fictive motion, as in The statue threw its shadow across the yard, pattern path fictive motion, as in My wet hair left a trail of water spots on the floor, and access path fictive motion, as in The bakery is across the street from the bank (see Talmy 2000a for discussion of these and other types of fictive motion). This chapter focuses on coextension path fictive motion because this type shares many properties with actual
479
480
M AT L O C K
motion and other metaphorical kinds of motion, especially directed motion along a path, and because this form of fictive motion has received more scholarly attention than other forms (see also Zlatev this volume Chapter 11).1 As noted by Talmy (1983, 1996, 2000a), fictive motion descriptions share conceptual structure with actual motion descriptions. For example, the actual motion statement The horse runs along the cliffs includes a mover (horse), movement (running), and an implied path in a physical environment (path along cliffs). The fictive motion statement The highway runs along the cliffs also includes a path in a physical environment (highway along cliffs). However, it lacks a mover. In the absence of a mover, the language user subjectively ‘moves’ by simulating movement or visually scanning along some portion of the subject-noun phrase referent, for instance, momentarily imagining motion along the highway (see also Langacker 1986, 1987, Matsumoto 1996). Fictive motion sentences have many interesting properties related to how the subject-noun phrase referent (hereafter, subject-NP referent) is conceptualized. It is always stationary. An animate subject-NP referent sounds unacceptable unless it is construed as being still. A statement such as A king snake runs along the fence makes sense only if the king snake is dead, lying completely still, or in a visual depiction, such as in a painting or photograph. The subject-NP referent in fictive motion sentences can also be linearly extended in various directions, horizontally, as in A bike path runs along the lake, and vertically, as in A glacier runs down the north side of the hill. The subject-NP referent can also be conceptualized as large or small, as in The Milky Way runs through the easternmost regions of the constellation and The spermine molecule runs along the minor groove of a third duplex. There are also two ways to conceptualize a fictive motion subject-NP referent, depending on whether it is generally associated with movement in the world. With Type 1 fictive motion, the subject-NP referent is a path that can be traversed in some way, for instance, by driving, walking, biking, as in The highway runs along the cliff, A trail climbs up the back of the hill, and A bike path goes along the lake. With Type 2 fictive motion, the subject-NP referent is an object that is typically not traversed, for instance, A fence follows the coastline, A table runs along the back wall, and The parking lot goes from the soccer field to the gymnasium. With some Type 2 fictive motion, the subject-NP referent is an inherently long object (e.g. fence), but in others, it is not (e.g. table) (for discussion of Type 1 and Type 2, see Matlock 2004a). In addition, fictive motion sentences are inherently imperfective. With The mountain range goes from Canada to Mexico, the mountain range is 1
Lakoff and Turner (1989) have attributed figurative uses like these to f o r m i s m o t i o n , which maps motion on to physical form. This is not inconsistent with the idea of fictive motion; it is simply a more general way to characterize the phenomenon.
Metaphor, Simulation, and Fictive Motion
expected to not change (i.e. it should retain its structural integrity and position indefinitely). For this reason, imperfective aspect is often not used with fictive motion descriptions. This is not to say that imperfective aspect is never used with fictive motion. It can be used to describe a subject-NP referent that is linearly extended in real time, with the expectation of even further extension, as in The new cable is now running from the library to the gym, and will reach the street by Wednesday, said by a foreman on a construction site. Imperfective aspect can also be used with fictive motion when the person speaking is moving through physical space, for instance, while talking on a cell phone, as in The path I’m on is going through a densely wooded area, but it should come to a clearing soon. Imperfective aspect can also be used with fictive motion when the temporary nature of the subject-NP referent is emphasized, as in Before the cleanup crew came, the phrase ‘Food not bombs’ was running across the side of the building (see also Langacker 2005b, Matlock and Bergmann 2014). Early cognitive linguistics research on the semantics of fictive motion provided many new and interesting insights on a pervasive type of nonliteral language. It also set the stage for subsequent research on how it is understood and used in everyday communication. The next section summarizes some empirical research that has tested Talmy’s and other cognitive linguists’ original observations and claims about coextension path fictive motion, including the idea that it involves construal of motion along a path.
29.4 How is Fictive Motion Understood? In recent years, more and more cognitive linguists are using behavioral methods to test theories about how metaphorical meaning is understood in everyday language (for discussion, see Gibbs 1996, Matlock and Winter 2015). Some of that work has used psychological experiments to test how fictive motion is comprehended in real time, and what underlies the purported fleeting sense of motion. Matlock (2004b) used narrative comprehension tasks to test whether fictive motion interpretation includes any sort of implied motion. The logic was that if people mentally experience a fleeting sense of motion while processing fictive motion, then thinking about actual motion right beforehand should influence how that fleeting sense of motion unfolds in real time. Time to mentally process fictive motion sentences should vary if there are differences. In each trial in Matlock (2004b), participants first read a story about a protagonist traveling through a spatial scene. Next they had to decide whether a fictive motion target sentence was related to the story they had just read. For example, they read about a man who was driving either slowly or quickly across a desert and then decided (i.e. pressed a ‘yes’ or ‘no’ button) to indicate whether the sentence Road 49
481
482
M AT L O C K
crosses the desert was related to the man’s drive across the desert. The decision required people to ‘re-play’ how the actual motion unfolded in the story. All decision times were averaged across participants in the two conditions (in the slow condition and in the fast condition) and compared. Analysis of the data showed that people were generally slower to make decisions about fictive motion sentences after reading about slow travel than they were after reading about fast travel, suggesting that they mentally re-enacted movement they had read about while processing fictive motion. It is important to note that in a control study with the same stories, people were no slower or faster when reading target sentences that were not fictive motion (e.g. Road 49 crosses the desert). Other experiments in Matlock (2004b) revealed consistent results for fictive motion: slower fictive motion decision times with long distances (versus short) and with cluttered terrains (versus uncluttered). The results of these experiments provided evidence to support the idea that understanding fictive motion involves a fleeting sense of motion along a path. Other behavioral work has explored whether fictive motion is conceptualized like actual motion by looking at whether and how it would permit space–time construal. Over the years, many cognitive scientists have tested whether conceptions of time are influenced by conceptions of space. In one experiment by Boroditsky and Ramscar (2002), participants imagined themselves moving toward a physical object or a physical object moving toward them before answering this ambiguous question: Next Wednesday’s meeting has been moved forward two days. What day is the meeting now that it has been rescheduled? A Monday response to this query suggests “moving” the meeting two days further into the past (relative to Wednesday), and a Friday response suggests “moving” two days further into the future (relative to Wednesday). The results of the experiment indicated that the way participants conceptualized motion influenced how they reasoned about time. People were more likely to provide a Friday response after imagining themselves moving toward an object, but more likely to provide a Monday response after imagining the object moving toward them.2 Matlock, Ramscar, and Boroditsky (2005) used a similar approach to investigate whether fictive motion would have a similar affect on reasoning about time. Participants in one study read a fictive motion sentence (e.g. The bike path runs alongside the creek) or a non-fictive motion sentence (e.g. The bike path is next to the creek) and then drew a picture to show their understanding of the sentence. Next they answered the ‘move forward’ question, namely, Next Wednesday’s meeting has been moved forward two days. What day is the meeting now that it has been rescheduled? In brief, thinking about fictive motion led to more Friday responses (than Monday 2
For similar work and alternative explanations, see McGlone and Harding (1998), Moore (2006), Núñez, Motz, and Teuscher (2006), Teuscher, McQuire, Collins, and Coulson (2008).
Metaphor, Simulation, and Fictive Motion
responses), but thinking about non-fictive motion did not (i.e. equal likelihood of a Monday or Friday response). In addition, their drawings of fictive motion descriptions included more motion elements than drawings of non-fictive motion descriptions (e.g. people jogging) (see Matlock, Ramscar, and Boroditsky 2003). Another experiment showed magnitude effects for fictive motion, including more Friday responses with a larger number of scan points (e.g. Twenty pine trees run along the edge of the driveway versus Four pine trees run along the edge of the driveway). Yet another experiment tested effects related to viewpoint, including more Friday responses with an ego-moving perspective away from the assumed position of the reader (e.g. The road goes all the way to New York versus The road comes all the way from New York).3 This set of experiments suggests that fictive motion construal behaves like actual motion construal: it affects reasoning about time in a way that mirrors the way actual motion does. People engage in thought about fictive motion and then that influences reasoning about ‘where’ temporal landmarks are and how they move. It can also encourage thought about actual motion (i.e. prompt people to draw motion in static form in fictive motion depictions). Other behavioral work has explored how linear fictive motion is construed by examining how it is represented in static form. In one experiment in Matlock (2006), participants were asked to generate line drawings to represent their understanding of Type 1 fictive motion sentences, such as The highway runs along the coast, or similar non-fictive motion sentences, such as The highway is next to the coast. In general, people tended to draw longer subject-NP referents (highways) in their depictions of fictive motion than they did in their depictions of non-fictive motion. In a second experiment, people drew depictions of Type 2 fictive motion sentences or comparable non-fictive motion sentences, such as The tattoo runs along his spine or The tattoo is next to his spine. For this experiment, all subject-NP referents were objects that can be construed as short or long (independent of fictive motion construal); a tattoo can be a long object, such as a snake or a sword or a short object, such as an apple or a heart. The results revealed that, on average, people drew longer subject-NP referents in their fictive motion depictions than they did in their non-fictive motion depictions, suggesting that fictive motion brings on a greater degree of linear extension than does non-fictive motion. Fictive motion understanding has also been explored in eye-tracking experiments. In a study by Matlock and Richardson (2004), people were presented with pictures of spatial scenes on a computer screen while listening to descriptions of the scenes. Each scene showed a vertical and a horizontal subject-NP referent (e.g. back of building with a ladder and 3
The results of Matlock, Ramscar, and Boroditsky (2005) have been replicated with a wider range of stimuli, for instance, in Ramscar, Matlock and Dye (2010). Consistent results have also been discovered in experiments with ordered sequences of numerals and ordered sequences of letters in Matlock et al. (2011) and with visual paths in Matlock (2010).
483
484
M AT L O C K
clothesline). Some descriptions had fictive motion (e.g. The cord runs along the wall) and others did not (e.g. The cord is on the wall). While people viewed scenes and listened to sentences, their eye movements were recorded and measured,4 making it possible to pinpoint where they directed visual attention. The results showed that people spent more time visually attending to the region of the scene that showed the subject-noun phrase referent while listening to fictive motion sentences (i.e. eyes often scanned the cord) than they did while listening to non-fictive motion sentences (i.e. eye movements were distributed across the scene) (for similar results in Hindi, see Mishra and Singh 2010). In a later priming study by Richardson and Matlock (2007), people first listened to a terrain description that would lead them to infer either easy or difficult movement (e.g. The valley is covered with dust, The valley is covered with ruts), and then heard a sentence with or without fictive motion (e.g. The road runs through the valley, The road is in the valley). Next they viewed a scene on the screen (e.g. valley). One of the main results was that people directed more visual attention to the region of the scene associated with subject-NP referents (e.g. road) when they had been primed with difficult terrain descriptions (e.g. ruts in a valley) than they did when they had been primed with easy terrain descriptions (e.g. dust in a valley). The results suggest that the way people construed fictive motion was influenced by inferences they made about how easy or difficult it would be to ‘move’ through the environment. Difficult terrain led to more difficult fictive motion, and hence, more visual attention to the region associated with the ‘motion.’ Last, some cognitive neuroscience research has investigated patterns of brain activation while people are interpreting fictive motion sentences. In a study by Saygin et al. (2010), participants read (and heard) three kinds of actual motion sentences (The wild horse crossed the barren field), no motion sentences (The black horse stood in the barren field), and fictive motion sentences (The hiking trail crossed the barren field). The goal of the study was to determine whether activation associated with MT+, which is sensitive to visual processing of motion, would arise with both literal and non-literal language. One of the main findings was that fictive motion showed less MT+ activation than actual motion, but more than no motion, suggesting that areas of the brain areas associated with visual processing motion may underlie fictive motion processing (see Romero et al. 2013 for consistent results and discussion of lower-limb somatotopic effects, and see Wallentin et al. 2005 for comparable results in Danish). Together, these studies support cognitive linguistic claims that fictive motion involves some kind of dynamic construal, such as scanning or linearly extending a subject-NP referent, and that it behaves in some 4
Eye tracking enables researchers to track eye movements in the time course of processing visual and linguistic information. For discussion of this technique, see Tanenhaus and Spivey-Knowlton (1996) and Richardson, Dale, and Spivey (2007).
Metaphor, Simulation, and Fictive Motion
ways like actual motion along a path. Still, what drives or enables this ability? How might it be like actual motion? The next section turns to research on mental simulation to gain a better understanding of what may be driving the fleeting sense of motion.
29.5 Mental Simulation There are many exciting cognitive science theories about how people think, reason, and communicate. One proposal that has gained tremendous support in the past two decades is mental simulation. Lawrence Barsalou characterizes a simulation as a re-enactment of perceptual, motor, and introspective states that are built up in the body and mind through interactions with the physical world (Barsalou 2008). Barsalou argues that over time people mentally accumulate collections of their experiences, especially recurrent perceptions and physical actions. For example, over the course of many years a given person sees many different types of bicycles, watches other people riding bicycles, and occasionally rides bicycles. Over time, those experiences form a simulator, or a distributed schematic representation that includes implicit information about how bicycles look, how they move, and how humans use them. That knowledge is then re-enacted, or simulated, in specific situations, such as reading about a particular bicycle or getting on a new bicycle. Importantly, more often than not simulations are underspecified and unconscious (see especially Barsalou 2009). If a friend reports that she sold her old red bicycle, you are unlikely to mentally simulate small or insignificant details about the bicycle, such as links on its chain or tread on its tire. You may not even form a visual image of a bicycle (see also Glenberg 1997 and Zwaan 2004, who offer compatible accounts of simulation). In recent times, cognitive scientists have shown that simulation is central to language understanding (see Zwaan 2009 for discussion), a development that has posed serious challenges for the long-standing view of language processing, namely, that it is the product of routinized concatenation of words and other symbols that are housed in the mind (for discussion, see Barsalou 1999, Gallese and Sinigaglia 2011). Glenberg and Kaschak (2002) investigated how comprehension of action statements that implied motion toward or away from the body would be influenced by direction of physical movement. Participants in one experiment were asked to judge whether English sentences such as Open the drawer or Close the drawer were sensible by moving the hand toward or away from the body to press a ‘yes’ or ‘no’ button. The results showed that people were generally faster to make their judgments when direction of hand movements was consistent with direction implied in the sentence (e.g. open versus closed), suggesting that simulation of motion direction is grounded in
485
486
M AT L O C K
actual movement. In another study on mental simulation in language comprehension, Zwaan, Stanfield, and Yaxley (2002) asked participants to read sentences about an object in a particular location and then view a line drawing of the object. Next they were asked to decide quickly whether the object had been mentioned in the sentence. The results indicated that people were generally faster to make the decision when the object’s location in the drawing was consistent with information that was implied by the sentence. For example, they were faster to decide whether an eagle had been mentioned in a sentence about a ranger seeing an eagle in the sky (as if it were in flight) than when they had seen a picture of an eagle with its wings spread (as if it were flying rather than an eagle perched in a tree). This suggested that people were simulating physical locations and actions associated with objects while processing sentences about those objects. In yet another experiment, Spivey and Geng (2001) asked participants to wear an eye-tracker device and listen to stories while sitting in front of a blank screen. Each story described events occurring in different places in a scene, for instance, on different floors in a tall building. As each story ‘moved’ from place to place in the scene, participants unknowingly moved their eyes in a consistent way, for instance, along the same access and progressively more upward while hearing about events on higher and higher floors in a building. The results suggest that people simulate the experience of visually scanning scenes when they are imagining scenes (for more general discussion of mental simulation, especially embodied simulation, see Gallese 2003, Gibbs 2005a, Zwaan 2009, Bergen 2012, Glenberg and Gallese 2012). What about simulation in the processing of non-literal language, especially motion metaphors? The experiments on mental simulation mentioned in this section are consistent with those on fictive motion discussed in section 29.4. For instance, the processing time differences observed for fictive motion target sentences in Matlock (2004b), such as Road 49 crosses the desert, are easily explained by simulation. The way people mentally re-enacted motion while reading and making a decision about Road 49 crosses the desert relied on their prior interpretation of motion in the travel stories, and how it took place – for instance, slowly or quickly. In addition, the results reported in Matlock, Ramscar, and Boroditsky (2005) can also be explained in terms of simulated motion. It is reasonable to assume that participants mentally simulated motion along the subject-NP referent along the bike path with The bike path runs alongside the creek, and that encouraged many to provide a Friday response to the question Next Wednesday’s meeting has been moved forward two days. What day is the meeting now that it has been rescheduled? Empirical research also supports the idea that mental simulation is involved in processing other kinds of metaphorical motion, including metaphors that portray romantic relationships as journeys. In a study by Gibbs (2013), participants heard a story about a relationship that was
Metaphor, Simulation, and Fictive Motion
described as a successful journey or a relationship that was described as an unsuccessful journey. After listening to the story, the blindfolded participants were asked to ambulate to a marker forty feet away while reflecting on the story. The main result was that the story about the successful journey led to further walking along the path than did the story about the unsuccessful journey. Another example of research along these lines is Wilson and Gibbs (2007). In one experiment they had participants first learn to perform specific actions with specific symbols, for instance, swallow whenever they saw this symbol: #. In the experiment that followed, these participants were shown the symbol and had to do the action. Immediately after they were given a metaphorical phrase, such as Swallow your pride or Grasp a concept, and had to press a space bar once to show they had understood the phrase. When the action symbol and the phrase were consistent (# with Swallow your pride) people were generally faster to press the space bar. These results, which suggest that physical actions facilitated the understanding of metaphorical phrases, support the idea that simulation is part of understanding metaphor, a position held by Bergen (2007), Gibbs and Matlock (2008), Lakoff (2014), and many others.
29.6 The Path Forward The constellation of work on fictive motion and other motion metaphors in cognitive linguistics dovetails nicely with a growing interest in mental simulation, and its role in understanding figurative language. The idea that we re-enact motion or visually scanning along a path while understanding metaphorical statements such as The road goes along the coast and Their marriage is moving in the right direction is supported directly and indirectly by a large body of cognitive science research, including behavioral and neurological experiments in the lab (see section 29.5). Still, there are some gaps to fill before we can reach a clear understanding of how simulation works, especially in real time and how it might vary across situations. Some of the work on fictive motion touches on the fleeting aspect of motion. There are many questions here, including how the movement transpires when somebody is interpreting a sentence such as The road goes along the coast. It is reasonable to assume that there is probably no explicit or conscious simulated movement or scanning along the road (or any subject-NP phrase referent). This would be in line with Barsalou (2009), who, as mentioned, claims that simulations are partial. For example, even if you frequently drove from Sacramento to Los Angeles, it would be impossible to simulate all details upon hearing the statement Interstate 5 goes all the way from Sacramento to Los Angeles. You would be unable fictively to scan visually or move through every town along the way or all the landscape going north to south. Also, you might process some ‘crude’ aspects of movement in interpreting this statement (e.g. going south),
487
488
M AT L O C K
but more than likely you would not think about whether the movement involved a car, truck, or motorcycle. Again, this would be in line with Barsalou (2009), who argues that simulations are often not conscious. Also, it is unclear who you would ‘see’ as being the mover. This gets at subjectivity. Clearly you experience some type of movement in interpreting Interstate 5 goes all the way from Sacramento to Los Angeles, but is it your own movement or visual scanning? Does the simulation of movement proceed in a direction that goes away from you, or does it transpire left to right in front of you? It is likely that these and other construals are possible, but only further experiments will provide a clearer picture of how it works. Expanding on some of the work that Jordan Zlatev and colleagues have done on subjectivity would be informative (e.g. Blomberg and Zlatev 2015). In this same vein, doing more experiments that tap into viewpoint would also be helpful (see Dancygier and Sweetser 2012 for an array of research on viewpoint and its importance in cognitive linguistics). Such follow-up work would give us a better understanding of how simulation plays out in real time. It would also be useful to conduct more cross-linguistics research on fictive motion, especially how it varies across languages. One underexplored area is manner of motion. What effect does manner of motion have on simulation in figurative language? Satellite framed languages have many degrees of freedom in terms of how manner of motion is expressed in a verb. English speakers, for instance, can use many different manners of motion verbs to describe how a road ‘moves’ through physical space: it can go, run, race, meander, climb, etc. How are various manners of ‘movement’ expressed in fictive motion descriptions in languages that do not code manner (as richly)? Closely related to this, what part of the ‘movement’ of fictive motion is most salient across languages? Fictive motion in Spanish places much focus on the beginning and end states of motion events, but places less focus the ongoing nature of motion events (see Rojo and Valenzuela 2003). All the lab research on metaphorical motion mentioned in this chapter was devoted to the study of language comprehension. It would be useful and informative to carry out research on how fictive motion and other metaphorical motion is produced in natural discourse. Some researchers are beginning to explore this, including analysis of fictive motion gestures in discussions of mathematics and in discussions of spatial layouts (see Bergmann and Matlock in prep., Marghetis and Nu´n˜ez 2013). Such research would give us a better sense of when and how figurative motion is used, and how useful it is to achieving mutual understanding in a conversation. Does it help establish common ground (Clark 1996), and if so, when and how? Far more work on the pragmatic utility of fictive motion is needed. It would also be informative to examine how fictive motion and other motion metaphors evolve and change in everyday use in a relatively short time period. A good example of this is seen in the way
Metaphor, Simulation, and Fictive Motion
people talk about the web. When it became popular in the mid- to late 1990s, people needed a way to talk about information search and access. Within no time, they started talking about it in terms of motion. Over the years, their language has changed. In particular, the range of motion verbs used is far narrower than it once was. These days, people tend to use go nearly exclusively, as in I went to the American Airlines website (for discussion of metaphors of web use, see Matlock et al. 2014). Last, despite good overall consensus on what mental simulation is and does, it means different things to different researchers. For behavioral scientists interested in motor actions, simulation is viewed as the mechanism that anticipates future actions (e.g. looking at an apple with the intention of picking it up and anticipating how the hand should be held in order to grasp the object), but for language theorists interested in how people interpret meaning, mental simulation is viewed as a ‘higher-level’ process that involves concepts and categories (for discussion of various views of mental simulation see Gallese and Sinigaglia 2011). Additional work on how simulation works in the understanding of metaphor, especially given that it involves the understanding of abstract information, could shed more light on how simulation generally works across thought and language. So, we still have a way to go before we arrive at a full understanding of how it is used, what drives it, and why it is used.
489
Part V
Methodological Approaches
30 Opening Commentary: Getting the Measure of Meaning Chris Sinha Cognitive linguistics has, as many authors have noted, been a significant contributor to, as well as an indispensable part of, the fundamental paradigm shift in the cognitive sciences away from abstract formalism and toward the embodied, enactive, ecological framework that has been called a “Second Cognitive Revolution” (Sinha 2007, among others). More recently, it has been claimed that cognitive linguistics itself is undergoing profound orientational shifts or ‘turns.’ Two such turns have been identified: the ‘quantitative turn’ in methodology, as described by Laura Janda, one of its leading proponents, in the following chapter in this part of the Handbook; and the ‘social turn’ in theory, claimed and argued for by Peter Harder (2010), which is principally represented in this volume by Esther Pascual and Todd Oakley (Chapter 21) and Jordan Zlatev (Chapter 11) (see also Verhagen 2005, Zlatev et al. 2008, Danziger and Rumsey 2013, Pascual 2014). The quantitative turn is the fruit of the convergence of a number of important developments in data type and quantity as well as technologies for data production, storage, and analysis which have become available to researchers in the last few years. These include the construction of naturallanguage corpora in an ever-increasing number of languages, the rapid development of brain-imaging technologies and the development of software tools for the management and statistical analysis of large datasets. These tools, their ranges of application and key examples of their use in empirical research are described in detail in the chapters comprising this part of the present volume. What I wish to focus on in this overview is the relationship between theory and methodology, including, in particular, what I see as a striking disconnect between the ‘social turn’ in theory and the ‘quantitative turn’ in method. Let us begin by asking a simple but fundamental question: What is it that really distinguishes cognitive linguistics from what is still its main theoretical adversary, formal linguistics? The answer, surely, is that, for cognitive
494
SINHA
linguists, meaning is central and is what motivates form and expression. The ‘cognitive’ in cognitive linguistics is the cognitive system in its embodied, ecological entirety, interfacing with the physical and social world. It is the coupling of this cognitive system with the world that is the source of meaning, and meaning is the source of linguistic form. The ‘social turn’ in cognitive linguistics is, at base, a shift of emphasis from the single (but never isolated) individual cognitive system that motivates meaning to the socially distributed and socially reproduced meanings acquired and used by the individual speaker–hearers. The social turn has been implicit in the notion of a usage-based view of language from the very start. It is not a negation of the cognitive paradigm but an enrichment of it, which brings renewed attention to those aspects of human cognitive capacity and cognitive process that underpin linguistic and other communicative interaction. The social turn also, perhaps, re-emphasizes a degree of continuity of the cognitive linguistic enterprise with Saussurean structuralism and Prague School functionalism, while retaining the theoretical innovations that mark it out from those predecessors. At any rate, we can reasonably maintain that the social turn sharpens our awareness of the deep divide that separates formal linguistics (for which social norms are nothing other than manifestations of universal mental forms; Itkonen 2003) from cognitive linguistics. If we accept the irreducibly social nature of linguistic meaning (which is not the same as saying that it is either arbitrary or unconstrained), what then follows, in terms of the appropriate ways in which meaning can be analyzed? And how does this relate to the quantitative turn in methodology? To put the question in a provocatively simplistic way: Can meaning be measured? Much ink has been spilled by philosophers and social scientists, as well as linguists, in trying to answer more sophisticated variants of this question in a debate that can be traced back to nineteenth-century discussions of the difference between the natural sciences (Naturwissenschaften) and the human sciences (Geisteswissenschaften). This debate has always revolved around three profoundly interlinked propositions. The first is that the objects of these two scientific enterprises are different, and that the ‘object feature’ that distinguishes the human sciences from the natural sciences is meaning (understood as both a social and a psychological phenomenon). The second is that this difference has consequences for the aims of the natural and the human sciences, such that (roughly speaking) the former aims at the establishment of invariant, law-like relationships, and the latter aims at the understanding and elucidation of particular events in human life, in all their psychosocial richness and complexity. In the classical philosophy of method, these aims are designated by the German terms Erklarung (explanation) and Verstehen (understanding). The third proposition is that these differences in object and aim are reflected in a difference in the respective methods appropriate for the natural and human sciences. The former is experimental and nomothetic, based in
Getting the Measure of Meaning
measurement. The latter is observational and idiographic, based in interpretation. The former employs what we know as quantitative methods; the latter employs what we know as qualitative methods. The founder of experimental psychology, Wilhelm Wundt (1832–1920), believed that experiment and measurement were appropriate only for obtaining psychophysical data, while higher psychological processes, as well as cultural and linguistic analyses, demanded interpretive and descriptive qualitative methods. From the early twentieth century onwards, however, it was recognized that social and psychological sciences, in attempting to account for social and psychological regularities, could not rely on qualitative methods alone. Striving for scientificity, spurred by the positivist view that explanation depends on prediction and control and greatly aided by advances in statistics, social and psychological sciences enthusiastically adopted quantitative methods, both experimental and correlational. This did not mean the total demise of qualitative methods, but it did mean that the qualitative method par excellence, ethnography, came to be associated with (and to some extent quarantined in) the discipline of anthropology. In psychology, especially, qualitative and interpretive methodologies were long banished to the same wasteland of ‘unscientific’ or ‘soft’ methods in which ‘introspection’ languished, anathematized by behaviorists because the practitioners of qualitative methods persisted in speculating and hypothesizing beyond objectively quantifiable data. Behaviorism, as we know, was overturned in the first cognitive revolution, but ‘opening the black box’ did not bring a concomitant overthrow of the predominance in psychology of experimental and quantitative methods. If anything, the prestige of cognitive psychology (and, more recently, cognitive neuroscience) has lent even more weight to experimental paradigms. At the same time, the dawning of the era of Big Data in social sciences has meant that new and powerful techniques and tools (often borrowed from the natural sciences) can be applied in ways that were unimaginable in the pre-digital age. It is these same tools that are increasingly being used by linguists, whether they are investigating corpora for specific co-occurrences and correlations or using cladistic algorithms to establish family relations between languages in historical (and prehistorical) time. Of course, qualitative methods never went away – indeed, in anthropology, “Thick Description” as a method (Geertz 1973) regained its original pre-eminence as a result of the decline of structuralism. More generally, the last twenty years or so of the twentieth century witnessed something of a revival in qualitative methods in microsociology (Goffman 1974) and interactional linguistics (as exemplified by Conversation Analysis; see Sacks, Schegloff, and Jefferson 1974), as well as in psychology. The nomothetic/ideographic distinction was recast by the distinguished and influential psychologist Jerry Bruner as a distinction between
495
496
SINHA
paradigmatic sciences (concerned with form, structure, and rule) and narrative sciences (concerned with meaning and context) (Bruner 1990). It is fair to say that, despite the continuing dominance of the landscape of social and psychological sciences by quantitative methodology, the advocates of qualitative approaches have gained substantial ground. Qualitative methods are no longer confined to the monodisciplinary niche of anthropology, and (even if they sometimes continue to be accorded a secondary status) their methodological validity is no longer contested as a matter of principle in psychology. Why, then, have discussions of methods in cognitive linguistics largely ignored the question of qualitative methods, instead being frequently confined to arguing the relative merits of quantitative methods versus ‘good old-fashioned intuitions’? I suggest that the main reason is that an increasing preoccupation with analyses of large databases has reinforced the adherence of many cognitive linguists to the methodological stance that treats language as an autonomous object of scientific inquiry. To avoid misunderstanding, I do not claim that this ‘language-as-object’ view (we could call it ‘methodological reification’) is wrong, tout court. Indeed, the mainstream traditions of linguistic science have long viewed language as a synchronic system, taking a third-person ‘externalist’ perspective; and the working assumptions of many subfields of linguistics (including linguistic typology) reflect this stance. However, the language-as-object view is not the only valid one, and if it is taken to be so it is constricting and incomplete. Although it may represent the historically dominant paradigms in linguistics, it does not represent the whole of language science; in particular, it does not represent those research enterprises that are primarily interested in languages as communicative tools and in speech/ language as social practice. The problem, at root, is this. If the ‘object’ of (cognitive) linguistics is taken to be nothing other than the regularities discernible from the analysis of the facts of language use, meaning as the fundamental feature of language and linguistic communication will be displaced from center stage. The reason, as has been pointed out by numerous philosophers and social scientists, is that meaning implicates the intentions of social actors, the contexts within which these intentions are formed and the understanding by the actors of the contexts (including of the intentions of other actors). No quantitative method, however sophisticated, can analyze this interplay of meaning, situation, and intentionality, not least because any adequate analysis inevitably implicates the intentionality of the analyst. This is just one aspect of the importance accorded by qualitative theories of method to the notion of reflexivity, the recognition on the part of the researcher of her or his engagement with the research participants, and with their world of meaning, or Lifeworld (Schutz 1966). The language-as-object view neither requires nor is conducive to reflexivity; and unlike many social sciences (including anthropology and
Getting the Measure of Meaning
cultural psychology), linguistics does not traditionally value reflexivity as an inherent part of the research process. Indeed, many linguists would deny that linguistics is a social science, preferring to emulate the supposedly objective, disengaged stance of the natural sciences. It is difficult, however, to reconcile this stance with the meaning-oriented theoretical perspective of cognitive linguistics without attenuating the cognitive perspective to the point where it becomes an optional add-on to an essentially data-driven enterprise. Cognitive linguistics, I suggest, stands at a crossroads. Quantitative methods will, doubtless, and rightly, continue to play an important role; but they are not a Holy Grail of language analysis. Quantitative methods are part of the methodological toolkit available to cognitive linguists, not the whole of it. Other methods include computational simulation (which I do not have space to discuss in this overview) and crucially, as I have argued, qualitative methods. If the social and cultural turn in cognitive linguistics is to really gain ground, it will have to be accompanied by a qualitative turn in methodology with concomitant attention to reflexivity in the research process. Such a methodological completion of the theoretical social turn will mean that cognitive linguistics will continue on the interdisciplinary trajectory that has been so important for its development from its inception, requiring the full embrace of its social-scientific dimension. The alternative, to fixate exclusively on quantitative analysis, would not only neglect this dimension but ultimately would also diminish the interdisciplinary impulse of cognitive linguistics, reducing it gradually to a subdiscipline within linguistics.
497
31 The Quantitative Turn Laura A. Janda
31.1 Introduction The quantitative turn in cognitive linguistics is a force to reckon with. In this chapter, I track the history of our quantitative turn, which has been facilitated by a confluence of three factors: the usage-based nature of the cognitive linguistics framework, the advent of electronic archives of language data, and the development of statistical software. I give an overview of the types of statistical models cognitive linguists are turning to, illustrated by the kinds of research questions that are being asked and answered using quantitative tools. I also discuss the opportunities and dangers that we face now that we have taken our quantitative turn.
31.2 What Brought about the Quantitative Turn? A survey of articles published in the journal Cognitive Linguistics (Janda 2013a) gives us a perspective on the quantitative turn in cognitive linguistics (see also Janda 2013b). Figure 31.1 presents the distribution of articles in the journal from its inaugural volume in 1990 through the most recent complete volume in 2015, according to whether or not they presented quantitative studies.1 Figure 31.1 reports percentages of quantitative articles for each year. A thick line marks 50 percent to make this visualization clearer. On the basis of this distribution we can divide the history of Cognitive Linguistics into two eras: 1990–2007 – when most articles were not quantitative; and 2008–2015 – when most articles were quantitative. In 1990–2007, twelve out of eighteen volumes had 20–40 percent quantitative articles. 1
This survey includes only articles proper, excluding review articles, book reviews, overviews, commentaries, replies, and squibs. For the purpose of this survey we define a ‘quantitative article’ as an article in which a researcher reports numbers for some kind of authentic language data.
The Quantitative Turn
percent quantitative articles 90% 80% 70% 60% 50% 40% 30% 20% 10% 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015
0%
Figure 31.1 Percentage of articles presenting quantitative studies published in Cognitive Linguistics 1990–2015
The lowest points were 1994, with one out of twelve articles, and 2002, with one out of eleven articles. In 2005 the move was in the other direction, with ten out of nineteen articles. At present the publication of quantitative articles seems to be leveling off at the rate of about 75 percent. Quantitative articles have always been with us; no year has ever been without quantitative studies. Three quantitative articles appeared already in the very first volume of Cognitive Lingustics: Goossens 1990 (with a database of metaphorical and metonymic expressions), Delbecque 1990 (citing numbers of attestations in French and Spanish corpora), and Gibbs 1990 (presenting experimental results). However, 2008 is the year in which we definitively crossed the 50 percent line, and it is unlikely that we will drop below that line again in the foreseeable future. This survey indicates approximately when quantitative studies came to dominate our scholarly output. It also shows us that cognitive linguistics has always engaged in quantitative studies, yet there is no reason to expect quantitative studies to entirely eclipse non-quantitative studies either. I do not mean to imply that there is a dichotomy between quantitative versus non-quantitative studies. A variety of valuable types of studies require no quantitative analysis, such as descriptive linguistics, theoretical works, and overviews of the state of the art. Conversely, an ideal quantitative study relies on linguistic description, expands our theoretical framework, and thus contributes to the state of the art. Thus, in a sense, quantitative studies depend on and ideally integrate non-quantitative components, though the reverse is not necessarily true. Although this survey is based on a single journal, Cognitive Linguistics is the signature journal of our field and it reflects the recent history of cognitive linguistics as a whole. Evidence from conferences and textbooks
499
500
JANDA
devoted to quantitative studies points in the same direction. Since 2002 there have been six bi-annual meetings of Quantitative Investigations in Theoretical Linguistics, a conference series devoted to statistical analysis of language data predominantly from the point of view of cognitive linguistics. QITL has grown over the years from a workshop with only a dozen speakers to a three-day event. Three of the authors of the five textbooks on the use of statistical methods in linguistics that I cite in section 31.2.3 have close ties to cognitive linguistics: Harald Baayen, Stefan Gries, and Natalia Levshina. How did we reach the quantitative turn? As is usually the case with historical developments, there was no single cause, but rather a combination of factors that pushed and pulled cognitive linguistics in this direction. Pushes have come from the theoretical framework of cognitive linguistics, which has proved to be fertile ground for developing research questions that rely on analysis of observed data. Pulls have come from the attraction of vast data resources and the access to sophisticated tools for their analysis.
31.2.1 A Usage-based Model of Language is Data-friendly Cognitive linguistics is a usage-based model of language structure (Langacker 1987: 46, 2013: 220). In other words, we posit no fundamental distinction between ‘performance’ and ‘competence,’ and recognize all language units as arising from usage events. Usage events are observable, and therefore can be collected, measured, and analyzed scientifically (Glynn 2010a: 5–6). In this sense, cognitive linguistics has always been a ‘data-friendly’ theory, with a focus on the relationship between observed form and meaning. Linguistic theories that aim instead to uncover an idealized linguistic competence have less of a relationship to the observation of usage, though there are of course notable exceptions.2 Even the question of what constitutes data in linguistics is controversial, and largely dependent upon the theory that one uses. Some researchers refer to constructed examples and individual intuitions as data, while others prefer to use corpus attestations or observations from acquisition or experiments. Introspection certainly plays an important role in linguistic analysis and indeed in the scientific method in general (cf. section 31.3.2), but reliance on introspection to the exclusion of observation undermines linguistics as a science, yielding claims that can be neither operationalized nor falsified (cf. section 31.4.2). It may seem attractive to assume that language is a tightly ordered logical system in which crisp distinctions yield absolute predictions, but there is no a priori reason to make this assumption, and usage data typically do not support it. Instead, we find complex relationships among factors that motivate various trends 2
For overviews of the use of corpus linguistics across various theoretical frameworks, see Joseph 2004 and Gries 2009b.
The Quantitative Turn
in the behavior of linguistic forms. A usage-based theorist views language use as the data relevant for linguistic analysis, and this gives cognitive linguistics a natural advantage in applying quantitative methods, an advantage that we have been steadily realizing and improving upon over the past quarter century. It is crucial to distinguish between the linguist’s own introspection about data (perhaps augmented by introspection solicited from a few colleagues) and the systematic elicitation of the intuitions of naı¨ve informants under experimental conditions, which is a legitimate scientific method that normally involves quantitative analysis. The difference is that whereas the linguist’s introspection does not necessarily yield reliable, replicable results, the elicitation of native speakers’ intuitions can yield such results. Introspection on the part of linguists can present numerous problems in that there are disagreements between linguists (cf. Carden and Dieterich 1980, Cowart 1997, Anketa 1997); their intuitions about mental phenomena are often inaccurate (Gibbs 2006); and last but not least, linguists’ intuitions may be biased by their theoretical commitments (Da˛browska 2010). Even if we put aside the issue of whether a linguist can report viable intuitions about language data, it is a fact that a linguist is an individual speaker, and there is abundant evidence that different speakers of the same language have different intuitions about linguistic forms. Given the fact of inter-speaker variation, it is more reasonable to assume that there is not just one model, but instead many models of the grammar of a given language (Da˛browska 2012, Barth and Kapatsinski 2014, Gu¨nter 2014). Every speaker, linguist or not, has to some extent a unique experience with the use of his or her native language, and a usage-based theoretical framework is well equipped to accommodate this fact.
31.2.2 Advent of Electronic Language Resources Recent history has impacted the practice of linguistics through the development of language corpora and statistical software. Today we have access to balanced multipurpose corpora for many languages, often containing hundreds of millions of words, some even with linguistic annotation. Modern corpora of this kind became widespread only a little over a decade ago, but have already become the first resource many linguists turn to when investigating a phenomenon. Many languages have national corpora, and open corpora are being built, providing free access not only to the linguistic forms and annotation in the interface, but also to the code itself, facilitating further exploration of data. A free resource that has attracted linguists is the Google Books Ngrams Corpus, which has a function that charts the frequency of words and phrases in a few of the world’s largest languages. In addition to corpora of written language, spoken corpora are becoming available, and some resources are even
501
502
JANDA
multimodal. For example, the UCLA NewsScape Library is an archive of billions of words in several languages, along with associated sound and images captured from television newscasts. The attraction of all this data is predictably compelling, particularly for linguists who view usage events as linguistic data. It is no surprise that a large portion of the quantitative studies undertaken by cognitive linguists have involved the analysis of corpus data, either alone or in comparison with experimental results (see Gries this volume Ch. 36 for more details concerning corpus linguistics).
31.2.3 Advent of Analytical Tools At approximately the same time that electronic corpora emerged, statistical software likewise became widely available. Thus linguists have at their disposal the means to explore the structure of complex data. The tool of choice for cognitive linguists is primarily ‘R’ (R Development Core Team 2010), which is open-source, supports UTF-8 encoding for various languages, and has a programming package, ‘languageR,’ specially developed by Harald Baayen for linguistic applications. A natural place to turn to for inspiration in the use of analytical tools is computational linguistics.3 Computational linguistics has of course been around since the 1950s, and computational linguists have considerable expertise in digital exploration of language data. However, the goals of cognitive linguistics and computational linguists have traditionally differed significantly due to the theoretical focus of cognitive linguistics (though there is good potential for collaboration, cf. section 31.4.1). Therefore, in addition to drawing on the capacities of computational linguistics, we have looked for leadership to other disciplines that also deal with human behavior but took the quantitative turn earlier, in particular psychology (in addition to sociology and economics). We linguists are still in a formative period where we have not yet settled on a set of best practices for use of statistical methods. A pioneering work in bringing statistical methods to linguists was Butler’s 1985 textbook. But ten years ago this textbook was out of print and there were very few alternatives. Since cognitive linguistics took its quantitative turn in 2008, several texts have been published such as Baayen (2008), Johnson (2008), Larson-Hall (2010), Gries (2013c), Levshina (2015). These books, together with scholarly works, are helping to establish norms for the application of statistical models to linguistic data and analysis. However, the field of statistics is itself in a state of considerable flux, particularly in the area of non-parametric models (especially relevant for us, since linguistic data is usually non-parametric; see section 31.3.1.2), adding an
3
See, for example, the journal Computational Cognitive Science at www.computationalcognitivescience.com/.
The Quantitative Turn
extra challenge for cognitive linguists as relative late-comers to quantitative analysis.
31.3 What Does the Quantitative Turn Bring Us? An introduction to statistical methods goes beyond the scope of this chapter and is better addressed by the textbooks cited above, so I will give only a bird’s-eye view, sprinkled with illustrative examples of how cognitive linguists are applying such methods. The scope of this overview is restricted to tracking some trends and discussing the relationship between quantitative methods and introspection.
31.3.1 Quantitative Methods in Cognitive Linguistics The goal of this section is to illustrate how quantitative methods are being used in cognitive linguistics and to identify some methods that are likely to stand the test of time. All statistical models are subject to assumptions and limitations concerning the nature of the data that need to be carefully observed and many models also facilitate the measurement of effect sizes which should be applied wherever possible, but since these issues are covered in textbooks, neither of them will be addressed in detail here. 31.3.1.1
Is A Different from B? Chi-square Test, Fisher Test, Binomial Test, T-test, ANOVA The main idea of this set of tests is to find out whether there are significant differences between two (or more) measured phenomena. Just because two numbers are different does not mean that there is a statistically significant difference between them. This set of tests aims to discover whether there is sufficient reason to reject the ‘null hypothesis.’ The null hypothesis is the default position according to which there is no difference between the measured phenomena. If the null hypothesis is true, the observed difference can be accounted for by random fluctuations in samples taken from a larger population of observations in which there is no difference. If the null hypothesis is rejected, the observed difference is unlikely to be accounted for by such fluctuations. Languages often give speakers choices, for example the choice between: A) the ditransitive (read the children a story), and B) the prepositional dative (read a story to the children) constructions in English. Corpus or experimental data might reveal a pattern such that there is more use of choice A in one environment (X) than in another environment (Y). But is the difference between the measurements of A and B a significant difference? In other words, is there reason to believe that there is a real difference between the frequency of A and B, or might the difference we observe be just a matter of chance (the null hypothesis)? A chi-square test can tell us the probability
503
504
JANDA
that the observed difference is significant. Chi-square tests have been used, for example, to test differences between the two English constructions listed above (Stefanowitsch 2011, Goldberg 2011), the difference between physical and metaphorical understanding of English path versus road (Falck and Gibbs 2012), and the difference in the use of SVO constructions between a child and his mother (Theakston et al. 2012). While a chi-square test can give an overall evaluation of whether there is something significant in a matrix of numbers, the Fisher test is useful when trying to find exactly which of those numbers deviates significantly from the overall distribution of the matrix. The Fisher test was brought to the attention of cognitive linguists by Stefanowitsch and Gries (2003, 2005) in collostructional analysis, where the point was to find out which words (such as disaster, accident) were more or less attracted to constructions (such as an N waiting to happen). This application of the Fisher test has since come under criticism (Bybee 2010: 97–101, Baayen 2011: 315, Schmid and Ku¨chenhoff 2013, Ku¨chenhoff and Schmid 2015),4 primarily for the use of numbers on very different scales (especially when some of these numbers are estimated rather than actual numbers), and for the use of the p-value as a measure of collostruction strength. However, when used on actual (not estimated) numbers of low values (tens or hundreds rather than tens of millions), the Fisher test is a useful way to probe the relationships among values in a matrix.5 If you know the overall distribution of a phenomenon, a binomial test can tell you whether the frequency of that phenomenon in your sample is significantly different from that in the overall distribution. Gries (2011) compared the frequency of alliterations in the British component of the International Corpus of English (the ICE-GB, here taken to reflect the overall distribution of alliteration in English) with the frequency of alliteration in lexically specified idioms such as bite the bullet (as opposed to spill the beans with no alliteration). The binomial test showed that the frequency of alliteration in English idioms is indeed significantly higher than in English overall. If two groups of items (e.g. two different semantic groups of lexemes – let’s call them A and B) each get a set of scores (e.g. acceptability scores), those two sets of scores will probably overlap. If the means of scores of the two groups are different, how do we know whether there is a significant difference between group A and group B? In other words, how do we know whether the difference in means is likely to reflect a real difference, or just chance variation in a situation where A and B actually behave the same in a larger sample? A t-test can handle a simple comparison of two groups. ANOVA (‘analysis of variance’), which is an extension 4
See also Gries’s responses to this criticism in Gries 2014b and Gries 2015a.
5
A relevant example of the application of the Fisher test is presented here: http://emptyprefixes.uit.no/semantic_eng .htm.
The Quantitative Turn
of the t-test, compares the between-group variation in scores with the within-group variation in scores, making it possible to compare more than two groups or more than one variable across the groups. Da˛browska, Rowland, and Theakston (2009) wanted to investigate the nature of longdistance dependencies such as Who1 did Mary hope that Tom would tell Bill that he should visit ______1 ? Da˛browska, Rowland, and Theakston’s hypothesis was that spontaneously produced long-distance dependencies follow the lexically specific templates WH do you think S-GAP? or WH did you say S-GAP?, where S-GAP is a subordinate clause with a missing constituent, and the majority of the remaining attestations are minimal variations on these patterns. They conducted an experiment in which children and adults were asked to repeat long-distance dependencies that did versus did not follow the lexically specific templates. An ANOVA analysis showed that children rely on lexically specific templates as late as age six, and that even adults are more proficient with long-distance dependencies that match the templates. These results support the usagebased approach, according to which children acquire lexically specific templates and make more abstract generalizations about constructions only later, and in some cases may continue to rely on templates even as adults.
31.3.1.2
What Factors are Associated with A? Correlation, Regression, Mixed Effects Regression, Classification and Regression Trees, Naive Discriminative Learning Suppose you want to find out what factors contribute to a given phenomenon, such as reaction time in a word-recognition task. The reaction time (A), termed the dependent variable in this example, may be related to various other phenomena such as frequency, length, and morphological complexity (B, C, D, etc.), known as independent variables. Correlation and regression are a family of models that can be used to explore such relationships. Correlation refers to the degree of relationship between two variables, such that the stronger the correlation, the better we are able to predict the value of one variable given the value of the other. Let’s say, for example, that we want to explore the relationship between the corpus frequency of a word and reaction time in a word-recognition experiment. A likely outcome would be that there is a correlation, such that the higher the frequency of a word, the shorter the reaction time, and thus it is possible to fit a line to a plot of data where one variable (frequency) is on the x-axis and the other variable (reaction time) is on the y-axis. If there is a correlation, given the frequency of a word it is possible to use the slope and intercept of the line to predict the reaction time, and conversely, given the reaction time associated with a word it is possible to predict its frequency. Notice that the prediction goes both ways. A big caveat with correlation is that prediction is not the same as causation: an association between
505
506
JANDA
frequency and reaction time does not necessarily mean that higher frequency causes shorter reaction times (or the converse). Even if you can use the value of B to predict the value of A with 100 percent accuracy, correlation tells you only that there is a relationship, not that B causes A. However, linguists are not immune to the temptation to assume causation when correlation is found (for a survey of correlation in relation to this problem, see Ladd, Roberts, and Dediu 2015). Another problem with interpreting correlation is that an apparent association between variables A and B might well be caused by other variables that have not been taken into account. The larger the dataset, the easier it is to find spurious relationships such as a positive correlation between linguistic diversity and traffic accidents (overlooking more telling factors such as population size and GDP; see Roberts and Winters 2013). Correlation has been used in a wide variety of studies. For example, in a study of long-distance dependencies, Ambridge and Goldberg (2008) found a correlation between the backgrounding of a clause (measured by a negation test) and the difficulty of extracting a clause (measured by the difference between acceptability in questions versus declaratives), such that verbs like know and realize behaved very differently from verbs like think and believe. In a study of Polish prefixed verbs, Kraska-Szlenk and Z˙ygis (2012) discovered a correlation between the reported morphological transparency of a prefixed verb and its acceptability rating by experiment participants. A regression analysis allows you to consider the relationship between a dependent variable (A) and a set of independent variables (factors associated with A). Linear regression is based upon the same calculations as correlation, since the line of best fit in a correlation is the regression line, defined by the regression equation. Because the correlation is generally not perfect, there is a difference between the predicted values and the actual values, and this difference is referred to as the ‘residual error.’ The standard error of estimate (which is an estimate of the standard deviation of the actual scores from the predicted scores) gives us a measure of how well the regression equation fits the data. Because regression is based upon the same calculations as correlation, it also inherits the same drawbacks, namely that by default it assumes a linear relationship (though this can be modified), it cannot tell us anything about causation, and any association that we find might actually be the result of other variables that we have not taken into account. Regression models come in a variety of types and all involve the prediction of a dependent variable based upon one or more independent variables (also called predictors). Ideally the independent variables should be independent not just of the dependent variable, but also of each other (thus avoiding what is called ‘collinearity’). In logistic regression (named after the logistic function used to divide all values into a categorical choice between two levels), the dependent
The Quantitative Turn
variable has only two values, and this is particularly useful for linguistic phenomena that involve a choice between two forms. The goal of a logistic regression model is to predict the probability that a given value (e.g. initial versus final position) for the dependent variable will be chosen. If the dependent variable has an ordered set of more than two values (such as the values low, medium, and high acceptability), it is possible to use an ordinal regression model. The use of regression, and in particular logistic regression, has become fairly common in cognitive linguistics. For example, Diessel (2008) tested the hypothesis that there is an iconic relationship between the position of a temporal adverbial clause (which can come before or after the main clause) and the order of the event reported in the adverbial clause as prior, simultaneous, or posterior to the event in the main clause. In other words, the prediction is that a speaker is more likely to produce After I fed the cat, I washed the dishes than I washed the dishes after I fed the cat. Diessel constructed a logistic regression model to explore the relationship between the position of the adverbial clause (initial versus final) as the dependent variable (the factor that is being predicted), and as independent variables conceptual order (iconicity), meaning, length, and syntactic complexity. Mixed effects models are regression models that can take into account ‘random effects,’ which are the effects introduced by individual preferences. Mixed effects models are commonly used in experimental studies where random effects account for the behavior of individual stimuli and/ or participants, and such models make it possible to arrive at generalizations that go beyond a specific sample of speakers or data. Random effects are relevant when we need to cope with what are called ‘repeated measures,’ such as in an experiment where multiple measurements are taken from each participant. In a word-recognition task where each participant responds to a set of words, some participants will be faster in general than others, so the baseline speed of each participant needs to be taken into account as a random effect. Random effects are opposed to fixed effects, which have a fixed set of values such as those for sex and age for experimental participants or tense, number, and person for verbs. For example, lexemes might act as random effects in a model, since they can have individual patterns of behavior. Janda, Nesset, and Baayen (2010) and Nesset and Janda (2010) applied a mixed effects model to a historical change underway in Russian verbs. In this model the individual verbs are a random effect since each verb has its own tendencies in relation to the ongoing change: some verbs use more of the innovative forms while others tend to resist innovative forms. In a study of the relative success of anglicisms in Dutch, Zenner, Speelman, and Geeraerts (2012) treated the concept expressed as a random effect, along with a number of fixed effects: relative length of anglicisms versus Dutch equivalents, lexical field, era of borrowing, ‘luxury borrowing’ (when a Dutch equivalent exists) versus necessary borrowing (when there is no Dutch equivalent),
507
508
JANDA
era of borrowing, concept frequency, date of measurement, register, and region. Regression models rest on assumptions that are often violated by linguistic data. Linear regression is a parametric model, which means that it tests hypotheses about population parameters. In other words, this type of model assumes that data should follow the bell curve of what statisticians call a normal distribution. Corpus data is, however, usually highly skewed, thus rendering linear regression less appropriate. Logistic regression assumes that all of the combinations of the various levels of all variables should be represented in the dataset. However, linguistic data often involve systematic gaps where certain combinations of the relevant variables are necessarily absent. There are at present at least two alternatives to regression models that offer the advantage of being nonparametric tests that also do not require all levels of variables to be observed in the dataset: classification and regression trees and naı¨ve discriminative learning. The classification and regression tree model (‘CART’; Strobl, Tutz, and Malley 2009) uses recursive partitioning to yield a tree showing the best sorting of observations separating the values for the dependent variable. Figure 31.2 shows an example of a CART tree from Baayen et al. 2013, showing the behavior of the Russian verb gruzit’ ‘load’ with respect to two grammatical constructions: the ‘goal’ construction, as in load the truck with hay, versus the ‘theme’ construction, as in load the hay onto the truck. The terminal nodes at the bottom of the tree show the number of examples in each node (‘n=’) and plot the distribution of theme versus goal uses for those examples. The top node of the tree (node 1) takes the entire dataset and makes the cleanest first division by finding the independent variable that is most effective at separating the goal uses from the theme uses, namely VERB: the ‘load’ verb prefixed in na-, za- or without prefix (the left branch) prefers goal use (represented by the light grey bars in the terminal nodes) more than when prefixed in po- (the right branch), where theme use (dark grey bars in terminal nodes) is strongly preferred. On the right side at node 13, the po-prefixed verb forms are further sorted into reduced constructions (yes), where a few goal uses are attested (light grey in node 15) versus full constructions (no), where only theme uses are attested (node 14). Most of the goal uses appear to the left, where we see that at node 2 the most important factor is whether the verb form is a participle (yes) or not (no): nearly all these examples are goal uses, though a few theme uses are found for the za-prefixed verb (dark grey in node 5). A CART tree can literally be understood as an optimal algorithm for predicting an outcome given the predictor values, and Kapatsinski (2013: 127) suggests that from the perspective of a usage-based model, each path of partitions along a classification tree expresses a schema, in the Langackerian sense (Langacker 2013: 23), since it is a generalization
1 VERB p < 0.001
{na, za, zero} 2
13
PARTICIPLE p < 0.001
REDUCED p < 0.001
yes
no 6 VERB p < 0.001
{na, za}
1
yes
no 1
Node 11 (n = 169) 1
Node 12 (n = 117) 1
Node 14 (n = 634) 1
Node 15 (n = 69) goal
za Node 9 (n = 208) goal
goal
goal
1
Node 8 (n = 147)
1
0.8
0.8
0.6
0.6
0.6
0.6
0.6
0.6
0.6
0.6
0.4
0.4
0.4
0.4
0.4
0.4
0.4
0.4
0.2
0.2
0.2
0.2
0.2
0.2
0.2
0.2
0
0
0
Figure 31.2 CART tree for Russian gruzit’ ‘load’ from Baayen et al. 2013
0
0
0
0
theme
0.8
theme
0.8
theme
0.8
theme
0.8
theme
0.8
theme
0.8
theme
goal theme
1
Node 5 (n = 248)
goal
na Node 4 (n = 328)
10 REDUCED p < 0.001
goal
za
yes
no
zero
7 VERB p < 0.001
goal
3 VERB p = 0.027
{na, zero}
po
0
510
JANDA
over a set of instances. For example, in Figure 31.2, node 11 is a generalization over 169 examples in which finite (non-participial) unprefixed (zero) forms of Russian ‘load’ in full (not reduced) constructions show a strong tendency (over 80 percent) for theme use. Naı¨ve discriminative learning (Baayen 2011, Baayen et al. 2011) is a quantitative model for how choices can be made between rival linguistic forms, making use of a system of weights that are estimated using equilibrium equations, modeling the usage-based experience of a speaker. Both CART and naı¨ve discriminative learning offer means for measurement of the importance of variables and validation of results. A CART random forest analysis uses repeated bootstrap samples drawn with replacement from the dataset such that in each repetition some observations are sampled and serve as a training set and other observations are not sampled, so they can serve for validation of the model and for measurement of variable importance. Naı¨ve discriminative learning partitions the data into ten subsamples, nine of which serve as the training set, reserving the tenth one to serve for validation. This process is repeated ten times so that each subsample is used for validation. Baayen et al. (2013) test the performance of regression against classification tree and naive discriminative learning models across four datasets and find that the three models perform very similarly in terms of accuracy and measurement of the relative importance of variables.
31.3.1.3
What is the Structure of Relationships among a Group of Items? Cluster Analysis, Multidimensional Scaling, Correspondence Analysis A given linguistic item, for example, a lexeme, might be measured in many different ways, yielding an array of data; and a group of lexemes could then each have an array. The linguist might want to ask: which of these items are more similar to others, how can these items be grouped? Cluster analysis, multidimensional scaling, and correspondence analysis take as input arrays of data associated with a set of items and use various mathematical techniques to arrange the items into a ‘space’ of two or more dimensions. Janda and Solovyev (2009) approached the relationships within two sets of Russian synonyms, six words meaning ‘sadness,’ and five words meaning ‘happiness,’ by measuring the relative frequency distribution of the grammatical constructions for each word in a corpus. The output of a hierarchical cluster analysis shows us which nouns behave very similarly as opposed to which are outliers in the sets. These results largely confirm the introspective analyses found in synonym dictionaries, and point to asymmetries between metaphorical uses of grammatical constructions and concrete ones. Multidimensional scaling has been used in various ways in cognitive linguistics; for example, to map out the functions of grammatical case in
The Quantitative Turn
Slavic languages (Clancy 2006) and to map the relations of aspect and expressions for spatial location (Croft and Poole 2008; see also Janda 2009). Eckhoff and Janda (2014) used correspondence analysis to measure distances between verbs according to the frequency distributions of their grammatical forms, yielding a sorting that suggests that there was indeed a difference in behavior between perfective and imperfective verbs in Old Church Slavonic.
31.3.2 Role of Introspection There should be a healthy balance between introspection and observation in any scientific inquiry. Introspection is the source of inspiration for hypotheses, which are then tested via observation. When it comes to analysis, introspection is indispensable in order to interpret the results and understand what they mean for both theory and facts of language. The data do not speak for themselves; we need introspection in order to understand what they mean. The critical eye of introspection is necessary to ferret out suspicious results and alert us to problems in design and analysis. Whereas theory should of course be informed by data, theoretical advances are typically born through introspection. Introspection is irreplaceable in the descriptive documentation of language. In fieldwork, a linguist interacts with speakers and posits the structure of a grammar based on a combination of observations and insights. The foundational role of descriptive work and reference grammars is not to be underestimated, for without this background we would have no basis for stating any hypotheses about language at all.
31.4 Where Does the Quantitative Turn Lead Us? Like any journey, taking the quantitative turn both opens up new opportunities and exposes us to new perils. It is worth taking stock of the pros and cons of this situation.
31.4.1 Opportunities The most obvious advantage to taking the quantitative turn is of course the opportunities we gain to discover structures in linguistic data that would otherwise escape our notice. In addition, we can bolster the scientific prestige of our field and foster greater accountability and collaboration. It is essential for the legitimacy of our field to secure and maintain the status of linguistics as a science. In applying quantitative measures we are developing linguistics as a discipline, following psychology and sociology in bringing the scientific method best known from the natural sciences to the fore. Cognitive linguists are on the leading edge in terms
511
512
JANDA
of implementing data analysis in the context of a theoretical framework and we may well have a historic opportunity now to show leadership not only within cognitive linguistics, but in the entire field of linguistics. We can establish best practices in quantitative approaches to theoretical questions. One important step we can take as a community is to make a commitment to publicly archive both our data and the statistical code used to analyze it. This will help to move the field forward by providing standards and examples that can be followed. In so doing, we can create an ethical standard for sharing data, stimuli, and code in a manner explicit enough so that other researchers can access the data and re-run our experiments and statistical models. Publicly archived linguistic data and statistical code have great pedagogical value for the community of linguists. As anyone who has attempted quantitative analysis of linguistic data knows, one of the biggest challenges is to match an appropriate statistical model to a given dataset. Access to examples of datasets and corresponding models will help us all over the hurdle of choosing the right models for our data. We can advance more efficiently if we pool our efforts in a collective learning experience. In many cases, funding agencies require researchers to share their data, adding further motivation for public archiving of data. Ultimately, the most important reason for making data publicly accessible stems from the basic principles of the scientific method, namely that scientific findings should be falsifiable and replicable. Researchers should be held accountable for their findings and only findings that can be replicated can be considered valid. One good option for linguists is the Tromsø Repository of Language and Linguistics (‘TROLLing’ at opendata.uit.no), a professionally managed, free, and open international archive of linguistic data and statistical code built on the Dataverse platform from Harvard University. As cognitive linguists become more familiar with quantitative methods, the opportunity for joining forces with computational linguists also increases. We can bring to the table valuable descriptive analyses and theoretical perspectives that can enrich collaboration in the building of better natural language processing and language technology applications.
31.4.2 Dangers There are at least two types of dangers lurking just beyond the quantitative turn. One involves over-reliance on quantitative methods, and the other involves various kinds of misuse or neglect of data. In the face of these dangers we can lose sight of the bigger picture of our theoretical principles and values. If taken too far, quantitative research runs the risk of triviality and fractionalization of the field. It is very easy for researchers to be seduced by fancy equipment and sophisticated software to the point that these
The Quantitative Turn
receive more attention than relevant linguistic principles. The most harmless negative outcome of this situation are shallow studies that do little or nothing to advance the field because they involve number-crunching without any real linguistic or theoretical goal. The potential outcome is a cognitive linguistic version of ‘cargo cult science’6 in which linguists perform empty rituals of calculations in hopes of conjuring up publishable results. More problematic is the substitution of ‘quantitative’ for ‘empirical’ and ‘scientific’ in the minds of researchers. The use of quantitative methods in a study does not make it better or necessarily any more empirical or scientific than language documentation or qualitative analysis. Confusion of these concepts could result in the marginalization of many of the traditional endeavors of linguists that could then be disadvantaged in the selection of works presented at conferences and in publications. We thus risk erosion of the core of our field, linguistic description and theoretical interpretation, which are also the source for research hypotheses. As Langacker stated in 2015, “linguistic investigation is a highly complex and multifaceted enterprise requiring many kinds of methods and expertise”7 and these various kinds of expertise should ideally be mutually supportive. In the age of big data, it becomes far too easy to find results simply because as the number of observations increases toward infinity (or just millions and billions), statistical tests are able to find effects that are infinitesimally small and therefore meaningless. To some extent this can be corrected for by the use of effect sizes as a check on results. However, Kilgarriff (2005) argues that since languages do not behave in a random fashion, the use of statistics to test null hypotheses is perhaps misguided to begin with. There will always be some patterns in linguistic data. The linguist’s job is to bring enough insight to the enterprise to know what is worth looking for and to distinguish between results that have a real impact on the advancement of our science and those that do not. Focus on big data analysis also threatens to marginalize languages themselves. Only a tiny fraction of the world’s languages have the resources to support large corpora, experimental studies, and comprehensive language technology coverage. The quantitative turn has the potential to exacerbate the existing imbalance between the few languages that many linguists study and the majority of languages that are largely ignored. We should not engage in an arms race to find out who can show off the most complex statistical models. It is usually the case that the simplest model that is appropriate to the data is the best one to use, since the results 6
This term is used by Feynman (1992) to compare inept scientists to ‘cargo cult’ south sea islanders, who, after experiencing airlifts during WWII, constructed mock runways manned by mock air traffic controllers, in hopes that this would cause more airplanes to land and bring them cargo.
7
Quoted from Langacker’s presentation at the ‘Theory and Method’ panel at the International Cognitive Linguistics Conference (2015a).
513
514
JANDA
will be most accessible to readers. Sometimes the structure of the data dictates a more complex model, but very complex models carry with them the disadvantage that they are well understood only by the statisticians who developed them. Overuse of ‘black box’ methods will not enhance the ability of linguists to understand and communicate their results. Wherever numbers are involved, there is a temptation to misrepresent them. Most academic fields in which researchers report statistical findings have experienced scandals involving fudged data or analyses, and current pressures to publish present an incentive to falsify results in hopes of impressing reviewers at a prestigious journal. Data sharing and best practices (cf. section 31.4.1) can help us to protect our field from this kind of dishonor. While transparency does not guarantee integrity, it does make some kinds of fraud easier to detect, and it always improves the quality and depth of scholarly communication. Major corporations such as Google, Amazon, Apple, and Facebook, along with hacking and spyware operations and state governments, have access to massive quantities of human language data. The lure of developing mining techniques via language analysis is part of what Kelly (2010) terms the ‘technium,’ the collective of archives and devices that constitute an organism-like system with a powerful momentum. This technology is advancing rapidly, and like it or not, we as linguists are contributing to it by improving our understanding of languages. This development is unstoppable; our only defense is to keep as much of it as possible in the public domain rather than behind clandestine corporate, state, and criminal firewalls.
31.5 Conclusion Since about 2008, cognitive linguistics has shifted its focus, and is now dominated by quantitative studies. On balance, the quantitative turn is a hugely positive step forward since it puts powerful new tools into the hands of cognitive linguists. Time always brings changes, and changes always bring challenges, but in this case the pros clearly outweigh the cons. Our field can gain in terms of scientific prestige and precision and collaboration. We can show leadership in best practices and the norming of application of statistical models to linguistic data. At the same time, I hope we can retain a humble attitude of respect for our venerable qualitative and theoretical traditions, which we should continue to nurture. If anything, we need qualitative and theoretical insights now more than ever in order to make sense of all the data at our command because those insights are the wellspring for hypotheses and the yardstick for interpretation of results.
32 Language and the Brain Seana Coulson
Could it really be true that language ability derives from a biological process in an organ contained in the skull? For me this idea was first driven home in the early days of functional MRI while processing data after a late night experiment in which we had asked a fellow post doc to read a series of sentences that either did or did not make sense. Huddled around the monitor, my collaborator and I watched rapt as the computer slowly rendered the activations from our first subject, high-fiving one another before leaning in to examine the splotches of orange and red atop a pixelated brain. More recently I felt that same fluttery feeling while watching a very articulate twenty-four-year-old suddenly stutter and sputter midway through a recitation of ‘Twinkle twinkle little star’ – especially because it was abundantly clear that the cause of his disfluency was the application of a magnetic pulse to his temple via transcranial magnetic stimulation (TMS). Beyond the recognition of the folly of dualism, such experiences raise the question of how the human brain supports language, and how the neural basis of language relates to that of our other cognitive abilities. Many traditional assumptions about these questions derive from linguistic theories that posit an innate universal grammar (e.g. Chomsky 1986). Such approaches lead naturally to the idea that brain regions have innately specified functional properties, that particular language processing operations can be ascribed to discrete regions of the brain, and that the neural substrate of language will be similar in different individuals and will be stable over time. However, as reviewed below, neuroscience research undermines many of these assumptions, and suggests a more complex relationship between brain and language ability. Rather than a fixed architecture for language, the data suggest that the brain regions supporting language ability change over the course of development, and involve a combination of domain-general and domain-specific components. I begin the chapter with a basic introduction to the organization of the human brain. Next is an overview of the study of semantic memory, and
516
COULSON
what cognitive neuroscientists have managed to glean about the neural representation of conceptual structure. From here I consider one of the oldest questions in neuropsychology, namely the location of the brain’s language center. This question, however, turns out to be deceptively simple as the neural networks that underlie language ability vary in ways that complicate straightforward mappings between linguistic functions and particular brain regions. I go on to review research on neural development and plasticity that highlight the diverse ways the brain supports language ability, and conclude with speculations regarding their implications for linguistics.
32.1 A Brief Tour of the Human Brain The human brain has the same basic structure as that of other mammals, but with a particularly large cerebral cortex, a thin (5 mm) sheet of brain cells, that is, neurons, folded to fit within the relatively small confines of the skull. The brain’s wrinkly appearance reflects this folding structure that includes both hills – known as gyri – and valleys – known as sulci. Below the cortex lie a number of subcortical structures, such as the basal ganglia, the amygdala, the thalamus, and many others, containing clusters of neurons known as nuclei and serving as neural relay centers that transmit signals both to the body and to different regions of the cortex. Together the cortex and the subcortical structures form the gray matter, so named because a neuron’s cell body is gray in appearance. By contrast, the long thin axon that protrudes from the neuron’s cell body is white, due to the presence of myelin, a fatty white sheath that speeds the conduction of electrical signals to cells in potentially distant regions of the brain. The bundles of axons that connect neurons into networks are thus known as white matter, or alternately fiber tracts. As with other mammals, the cortex in humans is composed of two hemispheres, each of which can be divided into four lobes: the frontal lobe, the parietal lobe, the temporal lobe, and the occipital lobe (see Figure 32.1). The central sulcus, or Rolandic sulcus, is an important landmark, dividing the frontal lobe from the parietal lobe. The pre-central gyrus is located in front of the central sulcus and is the site of primary motor cortex. Electrical stimulation of these areas can lead to complex movements of the limbs, and damage to motor cortex can lead to paralysis. The post-central gyrus is located behind the central sulcus and is the site of primary somatosensory cortex. Electrical stimulation of these areas leads to the sensation of touch, temperature, and pain on particular regions of the body. Interestingly, both motor and somatosensory cortices are similarly organized into maps of the body, with left hemisphere motor and somatosensory cortex representing the right side of the body, and right hemisphere motor and somatosensory cortex representing the left. Moreover, within each hemisphere, the general organization of the body map is similar in motor and somatosensory cortices, with neurons
Language and the Brain
Central Sulcus
Parietal Lobe
Frontal Lobe
Parietooccipital Sulcus
Occipital Lobe
Sylvian Sulcus
Basal parieto-temporal line
Temporal Lobe Figure 32.1 The left hemisphere of the human cerebral cortex with major landmarks indicating its division into four lobes: the parietal lobe, the occipital lobe, the temporal lobe. and the frontal lobe
controlling and sensing the feet and legs up near the top, neurons representing the hands and arms toward the middle, and neurons representing the face and the inside of the mouth toward the bottom. The occipital lobe lies behind the parietal lobe, delimited by the parietooccipital sulcus, at the back of the brain. Damage to the occipital lobe results in blindness, and experiments with primates suggest the main function of the occipital lobe is visual processing. The temporal lobe connects to the occipital lobe at the base, delimited only by an imaginary line known as the basal parietotemporal line. Among other things, this portion of the temporal lobe is involved in high-level visual processing operations such as object and face recognition. The temporal lobe is also important for the formation of long-term memories, especially structures in the medial (interior) section, such as the hippocampus and adjacent structures. A deep fissure known as the Sylvian sulcus delimits the boundary between the temporal lobe and the parietal lobe. This region of the temporal lobe, that is, the superior temporal gyrus, is the site of primary auditory cortex.
32.2 Semantic Memory and Embodiment One area of cognitive neuroscience with obvious relevance to cognitive linguistics is the study of semantic memory, or how conceptual structure
517
518
COULSON
is represented in the brain. In particular, the past decade has witnessed an upsurge in research investigating embodied or grounded cognition (see Glenberg 2015 for a review). Embodied cognition is the idea that concepts are grounded in brain systems for perception and action, and its advocates frequently cite work in cognitive semantics as helping to motivate the approach (see, e.g., Gallese and Lakoff 2005, Bergen 2012). Proponents of embodied cognition suggest that words and concepts are learned through sensorimotor experience so that the subsequent retrieval of concepts involves a simulation process that recruits sensorimotor processing regions of the brain. Neuroimaging research confirms some of the key predictions of embodied cognition, revealing the involvement of perceptual and motor systems in conceptual tasks. Action concepts are among the most investigated, and studies of action language indicate the activation of brain regions involved in movement and motor planning, including primary motor cortex in the frontal lobe, and primary somatosensory cortex in the parietal lobe (e.g. Pu¨lvermuller, Shtyrov, and Ilmoniemi 2005). Moreover, access to action concepts has been shown to be impaired in neurological conditions that affect the ability to move the body, such as Parkinson’s disease and apraxia (see Binder and Desai 2011 for a review). Such research suggests that we use brain areas involved in moving the body to understand language about human action. There is also evidence that linguistic tasks activate the various sensory systems in the brain to simulate the visual, auditory, and even olfactory features of linguistic referents. For example, sentences about moving objects activate lateral temporal lobe regions near the visual motionprocessing pathway (e.g. Saygin et al. 2010). So, too, do sentences about action, as one might expect if such concepts recruit visual simulations of the described actions (Kable et al. 2005). Further, the retrieval of color concepts activates temporal lobe regions adjacent to color-selective regions of the fusiform gyrus (Simmons et al. 2007). Sound concepts activate superior temporal and temporo-parietal regions of auditory cortex (Kiefer et al. 2008). Finally, the retrieval of information about how things smell activates olfactory cortex on the medial aspect of the temporal lobe (Gonzalez et al. 2006). The conceptual recruitment of modality-specific cortex, that is, regions of the brain that specialize in particular sorts of perceptual or motoric processing, is a good fit with process models of memory that emphasize memory as a consequence of neural processing (see Hasson, Chen, and Honey 2015 for a review). According to these models, memory is not stored in a particular ‘memory region’ of the brain. Rather, memories are stored in the neural circuits that process each domain of experience. Accordingly, visual memories are stored in visual cortex; auditory memories are stored in auditory cortex; and motor memories are stored in motor and somatosensory cortex. Even process models of memory, though, posit attentional
Language and the Brain
processes supported by fronto-parietal circuits to manipulate and control the information, and medial temporal lobe circuits to consolidate it. Similarly, there is reason to think that language relies heavily on these supra-modal processes.
32.3 Embodied Abstraction and Convergence Zones Although there is a fairly broad consensus that modality-specific regions of cortex play some role in the representation of concepts, there is disagreement regarding the importance of that role, especially for abstract concepts with no obvious sensorimotor associations (see Meteyard et al. 2012 for a review). Mahon and Caramazza (2008), for example, have questioned whether modality-specific activations are necessary aspects of conceptual structure, arguing instead that these activations are merely a side effect of amodal symbolic representations. More convincingly, however, Binder et al. (2009) argue that modality-specific activations are not sufficient to explain our language ability, because conceptual tasks recruit a network of brain areas whose function is neither sensory- nor motor-based. In keeping with the suggestion that supra-modal cortices play an important role in semantic memory, neuroimaging studies of semantic tasks consistently reveal activity in frontal and pre-frontal regions thought to control the top-down activation and selection of information, as well as temporal and parietal lobe regions that are not tied to a single sensory modality (Binder and Desai 2011). The temporal lobe regions in particular have been argued to host amodal semantic memories. These brain regions are larger in humans than in other primates, and respond to multiple modalities of stimuli as opposed to purely visual ones (Orban, Van Essen, and Vanduffel 2004). Moreover, their degeneration in semantic dementia coincides with the breakdown of knowledge about the world (Mummery et al. 2000). These supra-modal processing regions are likely to be convergence zones, that is, brain regions that integrate input from a range of unimodal input streams, and are hypothesized to be important for the formation of more abstract concepts (Damasio 2010). A given convergence zone receives input from one or more perceptual areas, and sends feedback to them via re-entrant projections. Moreover, a convergence zone also sends feedforward signals along to the next level in the hierarchy, and receives return projections from these higher-level convergence zones. For example, a low-level convergence zone might link neural codes for the color and shape of an apple, receiving input from parts of the visual system coding color and shape, and sending signals along to a higher-level convergence zone that might link codes for the apple’s color, shape, taste, and feel. In some sense we can see convergence zones as a general architectural feature of the brain, with hierarchically organized convergence zones
519
520
COULSON
linking representations with varying levels of abstraction. The lowest level of the hierarchy involves cell assemblies or neuronal ensembles, coordinated patterns of firing that arise naturally via synaptic strengthening between cells that fire in temporal proximity during the coding of an experience. More colloquially, cells that fire together then wire together to form neuronal ensembles so that neurons coding correlated aspects of experience can be activated together. The existence of feed-forward and feedback loops in a hierarchically organized set of neuronal ensembles is a key component of convergence zones, and makes them an efficient way to code information in memory. One function of convergence zones, then, is to reactivate neuronal ensembles in a constructive knowledge retrieval process involving a wave of activations in potentially disparate brain regions. For small local ensembles, this process might take only tens of milliseconds. The reactivation of larger ensembles, however, might last anywhere from hundreds of milliseconds to over a second. Apart from these stable modes of firing, convergence zones can activate subsets of their feedback connections in a way that produces novel combinations of activity (Damasio and Damasio 1994). Artificial neural networks provide a computational model of how experience-dependent weight changes between individual processing units can extract generalizable features of its input so that a network can both respond correctly to items it has previously encountered, and adaptively to novel ones based on their similarity to known items.
32.4 Where are the Language Areas of the Brain? We have seen that there are areas in the brain that specialize in the processing of different kinds of sensory information – that is, visual, auditory, somatosensory, and even olfactory information – as well as areas in the brain that specialize in the production of movement of the body. Moreover, in keeping with embodied theories of meaning, we have seen that sensorimotor processing regions play an important role in semantic memory, as brain regions that code our experience in the world are activated in our concepts of it. We have also seen that the neural representation of conceptual structure recruits supramodal convergence zones that comprise more schematic codes involved in the reactivation of sensorimotor regions, the detection of correlations across sensorimotor regions, and potentially the derivation of abstractions. In this section we turn to the issue of whether there are areas of the brain that specialize in the processing of linguistic information, beginning with the classical nineteenth-century model and ending with the network model prevalent in cognitive neuroscience today.
Language and the Brain
According to the classical model, the neural substrate of language involves two regions of the left hemisphere: Broca’s area in the inferior frontal gyrus, related to the production of language, and Wernicke’s area in the posterior superior temporal gyrus, related to its comprehension (see Figure 32.2). These classical models were inspired by the study of patients with brain damage, and the attempt to connect particular linguistic deficits with damage to specific parts of the brain. Broca, for example, described a series of patients who were able to understand language, but experienced profound difficulties producing it. This expressive aphasia was linked to damage to the left hemisphere frontal lobe, in the inferior frontal gyrus. Wernicke described patients with damage to the posterior region of the left hemisphere, toward the back of the superior temporal gyrus. Although these patients could speak fluently, their speech contained both phonological paraphasias, words or word-like utterances that contain a subset of phonemes from the target word, such as ‘feffort’ for ‘effort,’ and semantic paraphasias, words that are wrong, but somewhat related to the target word, such as ‘cat’ for ‘dog.’ Also known as receptive aphasia, the most salient symptom of Wernicke’s aphasia, however, is a profound deficit in language comprehension.
Primary Motor Cortex
Primary Somatosensory Cortex
Primary Visual Cortex
Broca’s Area
Wernicke’s Area Primary Auditory Cortex
Figure 32.2 Image of the left hemisphere with approximate locations of the two classical language areas: Broca’s area in the inferior frontal gyrus and Wernicke’s area in the posterior (back) region of the superior temporal gyrus. Also shown are primary motor cortex (on the pre-central gyrus), primary somatosensory cortex (on the post-central gyrus), primary auditory cortex (on the superior temporal gyrus), and primary visual cortex (in the calcarine sulcus).
521
522
COULSON
The classical model holds that sound images of words are stored in the posterior temporal lobe region now known as Wernicke’s area, while their motor images are stored in the inferior frontal gyrus region now known as Broca’s area (see Wernicke 1977). Consequently, Broca’s aphasics can comprehend language, but not produce it, while Wernicke’s aphasics can produce language but not understand it. Wernicke also suggested that the two language areas are connected by a fiber tract known as the arcuate fasciculus, important for word repetition tasks in which speech production is not initiated by conceptual processing, but relies critically on the transferal of sound images from Wernicke’s area to Broca’s area for the selection of the appropriate motor images. Neuroimaging techniques for studying brain activity in healthy adults first became available with the development of positron emission tomography (PET) in the late 1980s, a time when modular processing models were quite popular in cognitive science and psychology (e.g. Fodor 1983, Levelt 1989). It is perhaps unsurprising then that early work in cognitive neuroscience was aimed at mapping cognitive faculties onto the brain in a fairly straightforward fashion. The assumption behind this early work was that cognitive processes could be broken down into discrete modules, that is, specialized processors that are extremely fast acting and efficient, each of which is instantiated in a distinct part of the brain. On this sort of a modular approach it was natural to try to identify the portion of the brain that realized each aspect of linguistic knowledge, such as phonology, lexicon, syntax, and semantics. Consequently, whereas nineteenth-century neurologists interpreted the classical language areas in light of their knowledge about brain organization for sensorimotor function, twentieth-century researchers attempted to ascribe linguistic functions to particular regions of the brain. Accordingly, whereas Broca noted the proximity of the frontal lesions that give rise to expressive aphasia to primary motor cortex, twentiethcentury neuropsychologists were more taken with these patients’ agrammatic output, and preferred an account of Broca’s area as the seat of grammar (Goodglass 1976). Similarly, whereas Wernicke noted the proximity of the posterior lesions that give rise to receptive aphasia to primary auditory cortex, twentieth-century neuropsychologists were more taken with these patients’ semantic and phonological paraphasias, and suggested that this area was the seat of the mental lexicon (e.g. von Stockert and Bader 1976). Inspired in part by generative approaches to grammar, one assumption of this sort of localization approach is that each brain region has a single domain-specific function. However, neuroimaging studies have typically revealed that a given language function recruits multiple different brain areas, and, likewise, that some brain areas are involved in multiple different aspects of language function (see Price 2012 for a review). For example, left inferior frontal regions, argued by some to be the site of syntactic
Language and the Brain
transformations (e.g. Grodzinsky 2000), have also been shown to be activated by tasks that would seem to require knowledge of phonology (Zatorre et al. 1996) and word meaning (Murtha et al. 1999). These findings suggest distinct portions of the grammar do not map discretely onto particular regions of the brain. Another prediction of traditional approaches to language – that is, accounts that presume that language ability is supported by a languagespecific brain module – is that some brain regions are exclusively dedicated to language processing. However, neuroimaging research indicates that many of the brain areas activated by language are also activated by nonlinguistic tasks. For example, cells in the superior temporal sulcus (STS) that have been implicated in speech processing have also been shown to be important for face processing, the perception of biological motion, audiovisual integration, social perception, and theory of mind (see Hein and Knight 2008 for a review). Further, using fMRI, Mu¨ller and Basho (2004) demonstrated that the very same regions of Broca’s area were activated in a semantic decision task, a tone discrimination task, and a visuo-motor finger-tapping task. These data suggest that not only does Broca’s area contribute to linguistic tasks besides grammatical processing, it plays a role in a diverse array of tasks beyond language (see also Cabeza and Nyberg 2000). Just as each brain region does not have a single function, many functions are not localized to a single area. This seems to be especially true for grammar. Using a manipulation commonly employed to localize grammatical processing, Blank et al. (2016) scanned participants as the participants matched sentences to pictures depicting events described by objectrelative clauses (‘the circle that the star is greeting’) and subject-relative ones (‘the star that is greeting the circle’). Unlike many prior investigators, however, Blank and colleagues analyzed their data by identifying each participant’s language sensitive brain areas via the use of a localizer scan that compared activations to well-formed sentences versus a string of nonwords. This technique greatly improved the statistical power of their analysis over conventional approaches, and enabled them to detect changes in language-sensitive brain areas whose exact location varies from individual to individual. Blank et al. (2016) found that their manipulation of grammatical complexity did not activate a single area, but rather modulated brain activity in all of the language areas identified in their localizer task. This included brain regions in the inferior frontal gyrus that have traditionally been associated with grammatical processing, as well as other left hemisphere language areas such as the middle frontal gyrus, the anterior temporal lobe, the middle temporal gyrus, and the superior temporal gyrus (the latter being traditionally associated with phonological and semantic processing). These findings suggest that grammatical processing is not localized to a single brain area, but rather is a distributed property of the entire
523
524
COULSON
language system. While potentially compatible with a number of different approaches to syntax, this sort of distributed architecture seems particularly amenable to an account of grammatical knowledge as a structured inventory of linguistic units (e.g. Langacker 2008a).
32.5 Delocalized Views of the Neural Substrate of Language Another method for identifying the neural substrate of language is direct electrical stimulation performed in patients scheduled to undergo brain surgery. Whereas neuroimaging reveals brain regions whose activity is correlated with language comprehension and production, electrical stimulation can identify brain regions that are crucial for those processes. In this procedure, the neurosurgeon electrically stimulates a small patch of cortex as the patient performs naming or repetition tasks. In this way, she can identify regions that disrupt normal language production and avoid them during the subsequent resection. These ‘essential’ language sites have been identified in a range of left hemisphere frontal, temporal, and parietal regions, varying greatly from patient to patient. Within a given individual, though, these sites tend to be relatively discrete, less than a square centimeter in area, and their impact on language is highly replicable across testing sessions (Ojemann 1983). Direct electrical stimulation also reveals that stimulation of a variety of sites can elicit paraphasic errors, corroborating neuroimaging data discussed above that suggest a delocalized view of language function. Phonological paraphasias, for example, have been induced by stimulation to the inferior parietal lobe, the posterior (back) end of the middle and inferior temporal gyri, and in the inferior frontal gyrus (Duffau et al. 2005). Similarly, semantic paraphasias have been induced by stimulation of a variety of different brain areas in the temporal and frontal lobes, as well as stimulation of the fiber bundles that connect them (Duffau et al. 2005). These sorts of data support a delocalized view of language as arising not out of computations within a single area, but rather from parallel processing among distributed, interconnected groups of neurons. Such findings resonate with work in computational neuroscience with artificial neural networks that demonstrate how to leverage the processing power of idealized neural units. These simulations indicate a neuron’s computational power derives from its interactions with other neurons, and suggest that sophisticated cognitive functions require large networks of interconnected units (see Bechtel and Abrahamsen 2002 for a review). Note that support for a delocalized view of language processing is not a complete denial of functional specificity in the brain. For example, it is clear that different networks of brain regions mediate the processing of sound and the processing of meaning. According to dual stream models, the language system can be divided into two distinct, but interacting,
Language and the Brain
processing streams that parallel the organization of the visual system. Visual information is assumed to be partitioned into a ventral ‘what’ stream involving temporal lobe structures and dedicated to processing object identity, and a dorsal ‘how’ or ‘where’ stream involving parietal lobe structures that help mediate visually guided motion through space (Mishkin, Ungerleider, and Macko 1983). Similarly, the speech stimulus is initially processed in an auditory association area in the superior temporal lobe, and is partitioned into a ventral ‘what’ stream involving temporal lobe structures that maps sound onto meaning, and a dorsal ‘how’ stream involving frontal and parietal lobe structures, that maps sound onto articulatory characteristics (Hickok and Poeppel 2007). Examining the real-time brain response during language tasks reinforces the idea that the performance of a complex task requires a network of cells, often in multiple different brain regions, operating in parallel. Reading a word, for example, involves cognitive steps including visual encoding of the stimulus, potentially matching it to its phonological instantiation, and retrieving associated semantic information, all via a dynamic series of neural processes unfolding at the millisecond level. Event-related potentials (ERPs) derived from the EEG reveal a series of components sensitive to manipulations of the perceptual demands of word recognition (the P1, N1, and P2 components), the semantic demands of word recognition (the N400), and the demands on memory (the N400, whose amplitude is reduced by repetition) and the late positive complex (LPC), whose amplitude is typically enhanced by the recognition of a previously presented word. Evoked magnetic fields (EMFs) derived from the MEG reveal a similar series of components, and the greater spatial resolution of the MEG affords more accurate localization of the neural sources of these components. MEG studies of word comprehension suggest words are processed via repeated activation cycles in an extended network of brain areas. When the stimulus is a spoken word, the initial activation is in bilateral superior temporal cortices, that is, regions of auditory cortex; when it is a written word, the initial activation is in striate and extra-striate cortices, that is, regions of low-level visual cortex. Activity then spreads anteriorly – along the bottom of the temporal lobe for visual stimuli, and along the middle temporal gyrus for auditory stimuli – eventually involving the left superior temporal sulcus, left inferior prefrontal regions, and bilateral medial prefrontal areas for both written and spoken words. Importantly, following the initial feed-forward sweep of activity along the visual and auditory ‘ventral’ streams, MEG recordings suggest waves of activity spread back and forth between activated temporal and frontal lobe areas. The areas identified by MEG are the same supra-modal areas activated in fMRI studies of semantic processing, and the repeated waves of activity are consistent with activations in a hierarchical system of
525
526
COULSON
convergence zones. The computational role of cells in these areas likely changes as a function of time and the nature of the input (feed-forward versus feed-back), making the attribution of a single function to a particular cortical area difficult, and potentially misleading. Indeed, there is some evidence to suggest that cells in the same region of frontal cortex participate in very different aspects of language production at different points in time. Sahin et al. (2009) used intra-cranial electrodes implanted in the inferior frontal gyrus (Broca’s area) to record brain activity in three epileptic patients undergoing a procedure to localize the source of their seizures. In an attempt to localize brain regions important for lexical, grammatical, and phonological processing, patients first saw a prompt, such as ‘Read word,’ ‘Every day they,’ or ‘Yesterday they,’ followed by the presentation of the target word, ‘to walk.’ To this prompt, the patients responded (respectively) ‘walk’ for the repeat condition, ‘walk’ for the null inflectional condition, or ‘walked’ for the overt inflectional condition. These recordings revealed a peak at 200 ms related to lexical processing, a peak at 320 ms related to grammatical (inflectional) processing, and a peak at 450 ms related to phonological encoding. In contrast to simple localization approaches of the twentieth century, cognitive neuroscientists now view cognitive functions as arising out of the dynamic interactions of distributed brain areas organized in networks (Bressler and Menon 2010). A given brain region can participate in more than one network, and some regions participate in many. Frontal-parietal networks in particular have been observed to couple with a broad array of other more specialized networks as a function of task demands (Cole et al. 2013). Because a given region can participate in numerous different networks, its contribution to any given ongoing computation differs as a function of its inputs, and may evolve over time.
32.6 Variability in the Neural Substrate of Language Studies of the neural substrate of language thus reveal an impressive amount of variability from person to person, and from moment to moment as different networks of brain areas are engaged by languageprocessing tasks. Variability is also apparent if we examine differences on the time scale of an individual lifespan, and note how the neural substrate of language changes as a function of age and experience. For example, in toddlers with little language experience, speech activates a broad network of areas bilaterally in the frontal, parietal, and occipital lobes; by age three, however, the same stimuli also activate classical left hemisphere language areas in the superior temporal gyrus (Redcay, Haist, and Courchesne 2008). Neural circuits functionally specialized for language processing emerge progressively from a competitive process
Language and the Brain
of neural pruning, in which less efficient connections are eliminated, and a more dynamic process of experience-dependent tuning that strengthens useful connections and weakens the less useful ones. In this way, the bilateral and broad ranging neural substrate of language in early childhood becomes increasingly lateralized to the left hemisphere in adulthood, and focused in the classical left frontal and temporal lobe language regions. However, because the brain is a physiological organ and subject to the forces of senescence, the neural substrate of language changes again as the brain ages. Whereas younger adults tend to show more evidence for discrete, functionally specific networks of co-active brain regions, older adults show evidence of de-differentiation as the functional specificity of networks is reduced with age. Age-related changes include cortical thinning and changes in connectivity between brain regions (Meunier, Stamatakis, and Tyler 2014). Particularly relevant for language processing are age-related increases in the connectivity across hemispheres and concomitant physiological changes in bilateral activation (Cabeza 2002). Compared to younger adults, sentence-processing tasks in older adults lead to increased right hemisphere frontal activations (see Shafto and Tyler 2014). Changes in the brain regions that participate in language processing across the lifespan thus suggest a dynamic functional architecture (Blumstein and Amso 2013), without a strict mapping between specific brain structures and particular language processing operations. The extent to which language can be realized in multiple different ways in the human brain is also evident in the capacity of the brain to reorganize in the face of damage. For example, aphasic patients who suffer from communicative deficits frequently improve over the course of the first year following their stroke. Neuroimaging studies of recovering stroke patients suggest they undergo several distinct stages of neural reorganization (Saur et al. 2006). In the first few days following the traumatic event, activation is broadly reduced in the left hemisphere – even in areas not damaged by the stroke. A few weeks later, language stimuli elicit increased right hemisphere activation, especially in the homologues of classical left hemisphere language areas. Patients’ language ability has typically stabilized a year after their stroke, at which time the degree of left hemisphere activation is predictive of recovery. Language activations in these patients typically include left hemisphere areas around the damaged region, as well as brain areas not typically activated in language tasks (Thompson 2000, Turkeltaub et al. 2011). While these sorts of findings indicate that the adult brain can be organized in multiple different ways to support language, the developing brain is even more plastic. Because a great deal of brain development depends on experience, the possibility for neural reorganization is most evident for
527
528
COULSON
brain damage that occurs very early in the child’s life. Developmental plasticity is especially apparent in children who have suffered strokes during birth, known as perinatal lesion (PL) patients (see Stiles et al. 2005 for a review). For example, a longitudinal study that tracked language development in a small cohort of PL patients revealed that children with left hemisphere lesions were delayed in almost all language development milestones such as the age at which they uttered their first words, the onset of multiword utterances, the use of complete sentences, mastery over morphology, and the utilization of complex grammatical constructions (Bates et al. 1997). Given that the size of these children’s lesions was quite large, the incidence of language delay is unsurprising. By age five, however, the performance of these same PL children on tests of basic grammatical competence resembled that of their typically developing peers (Bates et al. 2001). Besides indicating the capacity of the developing brain to utilize alternative pathways for basic language ability, this study revealed that children with PL displayed a different pattern from adult stroke patients in the relationship between language deficits and lesion location (Bates et al. 2001). For example, whereas damage to the left superior temporal lobe (i.e. Wernicke’s area) is associated with language comprehension deficits in adults, it was associated with delays in language production in children. Further, there was a weak association between language comprehension deficits in PL children and damage to the right hemisphere (Bates et al. 2001), suggesting the right hemisphere activation observed in neuroimaging studies of language in toddlers is functionally involved in language processing, and not a mere epiphenomenal accompaniment to it. The qualitative difference in language deficits experienced by children and adults with left hemisphere lesions stands in contrast to analogous comparisons of spatial cognition. Although children with PL experience less severe visuo-spatial processing deficits than adult stroke patients, the structure-function mapping is qualitatively similar – both children and adults with right hemisphere lesions experience greater difficulty attending to global information about the overall configuration of visual diagrams, with preserved attention to local detail, and both children and adults with left hemisphere lesions experience greater difficulty attending to local detail, with preserved attention to global features (Stiles et al. 1996). These findings suggest that the neural substrate for spatial cognition is more highly specified than that for language ability and emerges earlier in development (Stiles et al. 2005). Presumably the different impact of damage to left hemisphere posterior temporal lobe regions in children and adults reflects the protracted nature of language development, and the way in which learning to produce language builds on neural representations used to understand it.
Language and the Brain
32.7 Neural Plasticity and Nativism Finally, we turn to another traditional assumption regarding brain organization for language, namely that different regions of the brain have innately specified functional properties. In view of the similarities in the overall organization of the human brain across individuals, this assumption is clearly true. Different regions of the brain are populated by different kinds of neurons, and vary in the sorts of input and output connections to other brain regions. While many of these factors are under genetic control, recent decades have provided vast evidence for the substantial role of experience in brain development in childhood and continuing throughout the lifespan (Stiles 2008). Rather than a somewhat deterministic set of processors coming ‘online,’ developmental neurobiology suggests that the emergence of functional specialization in the brain involves continual interplay between intrinsic processes of differentiation and extrinsic factors that shape and influence the direction of the emerging networks (Stiles and Jernigan 2010). The earliest stages of brain development are known to be under genetic control, as eliminating particular genes (e.g. EMX2 and PAX6) in mice has been shown to alter the relative proportion of motor and somatosensory cortex to visual cortex (O’Leary and Nakagawa 2002). By contrast, later stages of brain development in which brain regions acquire their particular functional characteristics are predominantly influenced by experiential factors. A great deal of functional specialization, for example, is attributable to receiving inputs from particular nuclei in the thalamus (a subcortical structure that serves as a relay for sensory inputs from the eyes, ears, skin, and so forth). In an experiment in marsupials, large portions of the cortex were removed prenatally, before the development of neuronal projections from the thalamus, so that the resultant specialization of cortex was largely attributable to the thalamic inputs. Huffman et al. (1999) found that even when three-fourths of the cortex was removed, the proportion and relative position of visual, auditory, and somatosensory cortex was remarkably similar to that in neurotypical animals. It turns out that the role of thalamic neurons in the specialization of sensory cortex lies not in their intrinsic chemical signaling properties, but, rather, in the sort of information they carry. Consequently, auditory cortex that receives inputs from the eyes takes on many of the properties we associate with visual cortex (Frost and Metin 1985, Roe et al. 1993). Although most of the work in developmental neurobiology has been done in animals, research with congenitally blind and deaf people has revealed a similar capacity for cross-modal plasticity in human brains. In congenitally deaf individuals, for example, temporal lobe regions more typically associated with auditory processing participate in the
529
530
COULSON
processing of peripherally presented visual stimuli (Neville 1995), as well as in tactile processing (Leva¨nen, Jousmaki, and Hari 1998). Similarly, in congenitally blind individuals, occipital brain regions that would normally develop into visual cortex are recruited for both auditory and somatosensory processing (Kujala, Alho, and Na¨a¨ta¨nen 2000). Indeed, the recruitment of occipital regions for non-visual processing has been demonstrated in people whose blindness did not begin until adulthood, suggesting that large-scale reorganization can sometimes occur in the mature human brain (Kujala et al. 1997). Plasticity associated with altered sensory experience also extends to language, as considerable research indicates portions of occipital cortex are recruited for language processing in the congenitally blind. For example, along with somatosensory regions of the parietal cortex, Braille reading activates a number of occipital regions, including primary visual cortex (Bu¨chel et al. 1998). Moreover, this activity is functionally significant as the application of TMS to occipital cortex impairs the ability to read words presented in Braille (Cohen et al. 1997). Although the latter finding could be explained by a role for occipital cortex in somatosensory processing, occipital cortex is also activated during auditory sentence processing in congenitally blind individuals, and activity in these occipital areas is modulated by the linguistic complexity of the sentences (Roeder et al. 2002). Finally, these regions of occipital cortex are activated when blind participants perform language production tasks, generating verbs in response to nouns presented either in Braille (Burton et al. 2002a) or as spoken words (2002b). Congenital deafness is similarly associated with an altered language substrate, though this comparison is somewhat complicated, as most deaf individuals use sign languages, such as American Sign Language (ASL). Despite having similar linguistic properties to spoken languages, sign languages might be expected to recruit different brain regions because they are expressed with the hands and arms rather than the mouth, and perceived primarily with the visual rather than the auditory system (see Hickok, Bellugi, and Klima 1998 for a review). Nonetheless, neuroimaging studies indicate the comprehension of ASL sentences activates traditionally defined ‘language regions’ in frontal, temporal, and parietal cortex in deaf signers, as well as in hearing adults who grew up signing with their deaf parents (Bavelier et al. 1998). Unique to deaf signers, however, may be the recruitment of auditory cortex for sign language comprehension. In a comparison of activity elicited by meaningful signs versus phonologically legal non-signs, meaningful signs activated bilateral regions of secondary auditory cortex in the superior temporal gyrus of deaf participants but not hearing participants who were viewing the same materials (Petito et al. 2000). Cardin et al.
Language and the Brain
(2013) compared brain activity elicited by British Sign Language (BSL) in hearing non-signers with that of two different groups of congenitally deaf participants – native signers of BSL and non-signers who communicated orally in English, relying on lip-reading for comprehension. Activation in bilateral auditory cortex was greater in both deaf groups than in the hearing non-signers, suggesting auditory deprivation is a key factor in the recruitment of auditory cortex for visually mediated language processing.
32.8 Conclusion We have seen that early ideas about how the brain supports language – that is, that discrete neural components support phonology, syntax, and semantics – have proven to be overly simplistic. Language ability cannot be mapped straightforwardly onto the brain because it emerges from the dynamic interaction of a highly interconnected series of brain areas. Moreover, the computational role of a particular brain area seems to change over time, depending on the sorts of inputs it receives from other parts of its network. In fact, the very idea of a brain region as having innately specified functional properties has been replaced with a far more nuanced account in which the functions of brain regions arise from the complex interplay of intrinsic, genetic factors and extrinsic, experiential ones. Cognitive linguistics, with its commitment to cognitively plausible theories, is well placed to accommodate recent discoveries about the neural substrate of language (Da˛browska 2004). For example, there is a natural affinity between the dual stream model that proposes speech sounds are processed in parallel streams for their articulatory and meaning components (Hickok and Poeppel 2004) with the treatment of linguistic expressions in cognitive grammar as symbolic links between a form, or phonological pole, and its meaning, or semantic pole (e.g. Langacker 2008a). Moreover, as is consistent with the approach in construction grammar (Goldberg 2003), recent neuroimaging data support an overlap in the brain regions involved in the processing of individual words and the processing of sentences (Blank et al. 2016). Finally, usage-based approaches in which grammatical categories emerge from instances of linguistic interaction (Bybee 2000, Croft 2001) are inherently compatible with process accounts of memory. On such accounts, a memory is not a representation in the brain, but an inherent property of neural circuits shaped by experience. Indeed, recent findings regarding the importance of experience for brain development and evidence for neural plasticity throughout the lifespan are far more compatible with usage-based approaches common in cognitive linguistics than the traditional vision of UG.
531
532
COULSON
In sum, language ability builds on fundamental aspects of brain function, such as the ability to perceive and move through the world. But at the same time, language enables us to go substantially beyond life in the sensori-moment, to derive abstractions, detect relational structures, and transmit cultural knowledge. The variability in the neural substrate of language suggests that there is no predetermined implementation of language in the brain. Further, while there are clearly more and less efficient ways for the brain to support language, the emergence of an optimal implementation requires a good deal of experience-dependent tuning. Although we began with the idea that language derives from brain activity, it is important to remember that, because of the key role language experience plays in neural development, the brain is as much a product of language ability as a prerequisite for it.
33 Cognitive Sociolinguistics Willem B. Hollmann
33.1 Introduction This chapter provides an overview of cognitive sociolinguistics – the approach to language study in which cognitive linguistics and sociolinguistics are combined, ideally for mutual benefit – presented in such a way as to make it clear how new research projects may be developed. To achieve this, I first discuss the scope of cognitive sociolinguistics, which sets out the playing field (section 33.2). I pinpoint the main pillars of both cognitive linguistics and sociolinguistics; that is, the sciences out of which they have each emerged. The suggestion here is that any and all theoretical constructs from these foundational areas represent valid bases for research questions in cognitive sociolinguistics and/or useful theoretical perspectives for interpreting instances of linguistic variation. The same applies to methodology: data collection and analytical methods familiar from cognitive linguistics and sociolinguistics, as well as the fields of inquiry they have arisen from, can and should be explored and exploited by cognitive sociolinguistics – though of course not necessarily all simultaneously. I then move on to highlighting the main methodological and especially theoretical achievements of cognitive sociolinguistics thus far (section 33.3.1). Whilst such a demonstration is important in any overview of the field, in the context of this chapter they take on particular prominence as clues to directions in which future research may be extended. Methods that have yielded significant insights are likely to continue to do so, when applied to novel variation data. Challenges to and modifications of usagebased cognitive or sociolinguistic theory invite more in-depth investigation in the context of additional data. Yet of course not all future cognitive sociolinguistic research should set out to test and further refine theoretical insights already gained. In a field as young as this the opportunities for exploring different methods and formulating entirely new research questions are manifold. I will point to
534
HOLLMANN
a number of fields from which inspiration may be drawn regarding novel methods and/or questions (section 33.4). I sketch (in section 33.3.2) the rather peculiar and unhelpful sociological situation that cognitive sociolinguistics still finds itself in, despite recent indications of improvement. In the conclusion to the chapter I return to this situation and offer a very practical suggestion for advancing the field more rapidly than is currently the case (section 33.5).
33.2 Defining the Scope: Linguistic Variation in Cognition, Culture, and Society A discussion of the overlap between cognitive and sociolinguistics depends on a definition of the territory, basic principles, and methods of these two fields. Both are in fact large and diverse, so I will not attempt to describe them comprehensively. Instead, my description is based on a conception of cognitive linguistics and sociolinguistics as radial categories, and a decision to focus on the central areas. There is no implication, however, that more peripheral areas are in any sense less worthy of scholarly pursuit. The foundations of cognitive linguistics are easier to pinpoint than those of sociolinguistics, which is why I elaborate more on the latter, below. As a generalization, I note that where in cognitive linguistics the main foundational science has had more impact on a theoretical than a methodological level, the relation between sociolinguistics and influential disciplines outside linguistics appears to have been, at least to some extent, the reverse.
33.2.1 The Foundations of Cognitive Linguistics At the heart of the cognitive linguistic enterprise lies the view of language as a manifestation of general cognitive abilities, and a commitment to theory-building on that basis (Langacker 1987, 1991, Croft and Cruse 2004). The main foundation of cognitive linguistics, other than linguistics itself, is cognitive psychology (see, e.g., Anderson 2014 for an overview). The influence of concepts from cognitive psychology is omnipresent in cognitive linguistics; for example, notions such as prototypes, schemas, and figure-ground organization. On the other hand, experimental methods are not as central to cognitive linguistics as they are to cognitive psychology (though see, e.g., Bergen 2007, Gonzalez-Marquez, Becker, and Cutting 2007 for useful guidance), with Gibbs (2007) arguing that this is for good reason.
33.2.2 The Foundations of Sociolinguistics Sociolinguistics views language through the lens of culture and society. Although the scope of the field may include areas as diverse as education,
Cognitive Sociolinguistics
policy, and politics, either in a descriptive or interventionist way, the core of the field is the study of linguistic variation in the context of linguistic, cultural, and social variables. This approach is known as variationist sociolinguistics. With respect to foundational disciplines, sociolinguists mostly point to anthropology (e.g. Gumperz 1962, 1964, Lavenda and Schultz 2014 offer an up-to-date overview of so-called four-field anthropology, which includes linguistics) and sociology (see Macionis 2013 for an overview).
33.2.2.1 Sociolinguistic Research Methods Regarding methodology, sociolinguists have drawn much inspiration from the qualitative and quantitative methods used in anthropology and sociology, respectively. The main qualitative method borrowed from anthropology is the ethnographic paradigm, which typically involves careful participant observation as a member of or trusted party within the community (see, e.g., Schreier 2013, or for a cognitive sociolinguistic perspective Clark and Trousdale 2013 and references cited therein). The ethnographic method is associated particularly with what Eckert calls second and third wave sociolinguistics. Eckert (2012) suggests that the first wave of variation studies was initiated by Labov’s (1966) study on New York English, which linked variation to a number of more or less stable social variables. The relevance of such macrosociological variables – mainly age, gender, ethnicity, neighborhood, and social class – has been confirmed in many studies since. Second wave research problematizes the consideration of variation in such broad-stroke terms, and thus “turned to ethnographic methods to get closer to the local dynamics of variation” (Eckert 2012: 90). The pioneering work of Milroy (1980) on social networks in Belfast exemplifies this (cf. Clark and Trousdale 2013 for guidance on undertaking social network analysis). In the third wave (e.g. Irvine 2001, Bucholtz and Hall 2005, Bucholtz 2010), finally, we see a move away from the perspective of social variables impacting on speakers’ language to a consideration also of the possibility that speakers may actively use language to construct social reality, in particular their identity (Eckert 2012: 94). As the remainder of the chapter will continue to make reference to Eckert’s wave model, it is worth pointing out a significant (potential and real) misunderstanding: Eckert’s model does not imply that second wave sociolinguistics has replaced the first wave approach, and that the third wave has consigned both earlier waves to history. Eckert herself (2012: 88) points to one reason why a strictly chronological interpretation is incorrect: Labov’s (1963) study of Martha’s Vineyard, which depended on ethnographic analysis of the very local dynamics of this small island community, predated his New York study. In addition, Blommaert (2014: 5) traces the emergence of the third wave insight that language and social structure are co-constructive to Gumperz and Hymes
535
536
HOLLMANN
(1972), which further upsets a strictly chronological view of Eckert’s waves. But perhaps most importantly, as is obvious from leading journals in the field (e.g. Journal of Sociolinguistics, Language in Society, Language Variation and Change) first and second wave studies continue to be carried out and to yield novel insights into the interplay between linguistic variation and society and culture. Moreover, Meyerhoff and Stanford (2015: 11) point out that without continued first and second wave research third wave studies will never be possible for less well described languages and varieties, as third wave studies depend on the solid descriptive work provided by first and second wave work. The reality of this misconception is apparent in the work of some sociolinguists themselves; see, for example, Blommaert’s insistence that the move away from “major assumptions of mainstream sociolinguistics” (2014: 8) toward “a more complex, dynamic and multifaceted view of sociolinguistic realities [is now] largely completed” (2014: 8–9), bringing into view a perspective “organized on an entirely different footing from that which characterized the Fishmanian, Labovian and Schegloffian sociolinguistic world” (2014: 14).1 I would suggest that Eckert’s wave metaphor is somewhat unfortunate, as it might encourage a view whereby the waves follow each other in neat succession, with each wave coming in as the previous one has broken and is receding. It seems more helpful to think of three layers instead: layers, as in brickwork, for example, do not replace one another, and higher layers can only be added once lower ones are firmly in place. Moving on to the influence of sociology on sociolinguistic methods, the main quantitative method starts from data collection from suitably stratified random samples. These are randomly selected sets of participants drawn from a population which the researcher divides into (what are expected to be potentially meaningful) subgroups (strata); for instance, of different ages, genders, and so on. The researcher then explores patterns in the data in relation to one or more potentially relevant social and linguistic factors. In the latter case, they typically use multifactorial analysis, which allows one to study the possibility that some factors may conspire to produce a certain effect, even if individually their effect is not necessarily significant. Tagliamonte (2006) offers a useful variationist introduction to sampling and multifactorial analysis. However, Gries has recently criticized the software she and many other sociolinguists use, VARBRUL/GOLDVARB. Gries (2013b: 8) suggests that logistic regression analyses are superior, and he points to some other sociolinguistic studies such as Mendoza-Denton, Hay, and Jannedy (2003) and Johnson (2008) 1
Blommaert and Rampton appear to suggest that there are in fact knock-on effects on linguistics at large, as they claim that “the denotational and propositional meanings of words and sentences lose their pre-eminence in linguistic study” (2011: 5). This seems misguided, however: one need look no further than several other chapters in the present volume to seriously question whether there is a decrease in interesting research on these aspects of linguistic meaning.
Cognitive Sociolinguistics
which demonstrate this. Other useful introductions to quantitative methods, including regression, are Gries (2009a) and Field (2013).
33.2.2.2 Sociolinguistic Theory Regarding theory, sociolinguists view language in relation to cultural and social variables such as gender and social class, yet they have not drawn on notions developed by anthropologists and sociolinguists in quite the same way as, for example, Fillmore, Lakoff, Langacker, and Talmy have made use of cognitive psychological constructs. This has led to criticism, even from within sociolinguistic ranks, with Fishman famously suggesting that people in the field “created their sociology as they went along,” leading to “self-imposed underexposure to serious sociological stimulation” (1991: 130–32).
33.3 Taking Stock: Overlap and Symbiosis to Date Considerable methodological and theoretical advances have accrued from combining cognitive and sociolinguistics. Cognitive linguists have benefitted from the additional dimensions added by considering what sociolinguists refer to as external factors (e.g. age, gender, social class), and by determining the relative contribution of such factors by means of multifactorial analysis. Conversely, insights from cognitive linguistics are allowing sociolinguists to show how the interrelation between language and society plays out on the level of individual speakers’ cognitive abilities and processes, which puts their accounts on a more solid foundation.
33.3.1 Major Methodological and Theoretical Advances to Date Gries’s (2003a) study is an early illustration of how considering an external factor may enrich a previously language-internal account. Medium (i.e. speech versus writing) added a new explanatory dimension to Gries’s own (1999) study of variation in particle placement. He also demonstrates the advantages of multifactorial analysis compared to monofactorial analysis. (The Leuven Quantitative Lexicology and Variational Linguistics group often explicitly state that their use of this statistical method has been inspired by variationist research; see below for discussion of Grondelaers, Speelman, and Geeraerts 2007.) Kristiansen (2003, 2006, 2008) shows how cognitive linguistic accounts of prototypes and construal – in particular metonymy – offer more insight into the way in which sociolinguistic notions of variants, styles, and social values are connected. She proposes that all these notions are prototypebased categories, with more versus less central examples and members. And these categories are linked to one another in a metonymic fashion. Thus, certain linguistic features may invoke styles (which are defined in
537
538
HOLLMANN
sociolinguistics as sets of not just linguistic features, but including also other aspects of behavior, such as physical comportment, dress and taste in music). Styles, in their turn, may metonymically activate cognitive cultural models of speakers associated with those styles, crucially including the social values associated with those speakers. The influence of Kristiansen’s work is clearly felt in Soukup’s (2013) third wave study. Her discussion of Schilling-Estes’s (2002) notion of Speaker Design is explicitly couched in Kristiansen suggestions concerning the conceptual nature of and links between linguistic features, styles, and representations of speakers I discussed above. Together, they place Soukup’s discussion of the phenomenon of Speaker Design on a firmer cognitive footing. Soukup (2013) also makes several useful suggestions regarding speech perception and speaker evaluation tests that will allow researchers to avoid the pitfall of ascribing social value to linguistic variants when speakers themselves do not necessarily perceive such links (see also Johnstone and Kiesling 2008). The research by Grondelaers, Speelman, and Geeraerts has pioneered the extension of Langacker’s (1993) notion of reference points to the arena of linguistic variation. Grondelaers, Speelman, and Geeraerts (2007, 2008) show that the distribution of presentative er in Belgian and Netherlandic Dutch is subject to dialectal differences but that reference points represent another relevant factor. Langacker defines reference points as constructions describing entities which facilitate cognitive access to other entities. In example (1), below, mentioning the surgeon makes it easier for the hearer to subsequently activate their representation of His wife, as they presumably know that the surgeon is (likely to be) married: (1)
Do you remember that surgeon we met at the party? His wife just filed for divorce. (Langacker 2008a: 83)
Presentative er is less frequent in sentences that provide good reference points regarding the following NP. In example (2), below, the sentenceinitial locative narrows down the range of the hearer’s expectations as to the possible entities denoted by the grammatical subject of the sentence, as they presumably know that this newsroom (or newsrooms in general) may be used for functions which include wine and snacks. In (3), by contrast, the temporal adjunct is not much help at all: the number of things that happened in 1977 is vast. (Both examples are from Grondelaers, Speelman, and Geeraerts 2008: 158, with slightly modified glosses.) (2)
(3)
In het redactielokaal staan enkele flessen wijn en wat borrelhapjes. in the newsroom stand some bottles wine and some appetizers ‘In the newsroom there are some bottles of wine and some appetizers.’ In 1977 was er een fusie tussen Materne en Confilux. in 1977 was there a merger between Materne and Confilux ‘In 1977 there followed a merger between Materne and Confilux.’
Cognitive Sociolinguistics
The suggestion, then, is that er functions as an instruction to the hearer along the lines of: “the entity I am about to mention is rather unpredictable.” Clark (2007) focuses on style shifting in her ethnographically collected data from a juvenile pipe band in West Fife, Scotland. She studies what is often called ‘performance speech,’ where a speaker puts on a style that they do not normally use; for example, because they are imitating someone. She reports on several conversations which make it clear that in words such as out the high rounded monophthongal variant (oot) is perceived as ‘Scottish’ or ‘slang,’ with the standard diphthongal realization perceived as ‘proper’ or ‘posh.’ Her data show on a very fine-grained and concrete level how speakers come to associate certain features with cognitive representations of certain speakers. For instance, one of her speakers relates a story about a teacher who tells one of his pupils to get out of the classroom when he mishears something the pupil says as a swearword. The speaker normally uses oot but when she adopts what Clark labels ‘teacher style,’ the teacher’s direct speech features the diphthongal pronunciation of out. Citing Langacker (1987) and Tomasello (2000), Clark uses the notion of bottom-up learning to suggest that these sorts of situations lead speakers to connect linguistic forms and social meaning and that once connected, the links may become entrenched in speakers’ minds. Hollmann and Siewierska (2007) show the relevance of two cognitive linguistic concepts to linguistic variation: frequency effects, particularly concerning reduction, and schemas. Their data set consists of transcribed oral history interviews obtained from the North West Sound Archive.2 Focusing on 1Sg possessive my, the authors report that in addition to the full form with the diphthong /aI / Lancashire speakers also employ a range of reduced forms: [mi], [ma], and [mǝ]. Whereas variationist sociolinguists would investigate the distribution from the viewpoint of social variables such as social class and linguistic variables such as the phonetic environment, Hollmann and Siewierska suggest that reduction is conditioned at least in part by token frequency. Broadly speaking, high token frequency of [my N] strings appears to promote reduction, as indeed one would expect on usagebased grounds (e.g. Bybee and Scheibman 1999); see also the grammaticalization literature (e.g. Hopper and Traugott 2003). (Hollmann and Siewierska also offer an explanation in terms of the alienability hierarchy familiar from linguistic typology; see section 33.4.2, below, where I discuss typology as a valuable source of inspiration and information for cognitive sociolinguistic research.) 2
Unfortunately, the North West Sound Archive has ceased to exist. The collection has been distributed across Manchester Central Library, Liverpool Central Library, and Lancashire County Council Archives in Preston (www .nwsoundarchive.co.uk/default.aspx, accessed January 12, 2016).
539
540
HOLLMANN
Hollmann and Siewierska find that not all of the variation in reduction is neatly accounted for with reference to token frequency. For example, my niece and my stepsister are infrequent, yet often feature reduction of the possessive in their data. The explanation they offer appeals to schemas, type frequency, and productivity (Bybee 1985): given that a fairly large number of other kinship terms (including brother, dad, father, husband, mother, mom, sister, son, wife) do co-occur often with my, those noun phrases were the likely starting point of the reduction of the definite article. Given their semantic similarity, this may have led to the emergence of a schematic representation of a reduced possessive form followed by a kinship term. Due to its relatively high type frequency this schema may subsequently have attracted new members, including my niece and my stepsister. Clark and Trousdale (2009) and Clark and Watson (2011) suggest further refinements to our understanding of frequency effects in language change. These studies concern TH-fronting in West Fife and t-to-r lenition in Liverpool English, respectively. The picture emerging from Clark and Trousdale (2009) in relation to the role of frequency is a complex one, as was to be expected since the change in question is a not a reductive one (both the standard dental and the non-standard labiodental variants are fricatives). Clark and Trousdale use multifactorial analysis to show that a multitude of factors are at play. Phonological priming is one of them, which fits naturally in a cognitive linguistic perspective. Yet the strongest factor of all is “community of practice/friendship group membership,” which shows not only how cognitive and social factors may be intertwined, but also that social factors may at least in some cases turn out to be dominant. Clark and Watson (2011) is a first wave study involving oral history interviews obtained from the sound archive also used by Hollmann and Siewierska (2007, 2011). The relevant phonological change is an instance of reduction, with intervocalic word-final /t/ being realized not as a plosive but as a tap [ɾ] or a rhotic approximant [ɹ]. The results are to some extent interpretable from the viewpoint of high token frequency causing reduction. For instance, Clark and Watson observe high levels of reduction in bit, which also has a high token frequency. However, they suggest that their data raise two main questions for the usage-based model. First, toward the lower end of the frequency scale there is an abrupt cut-off point, below which words simply fail to participate in t-to-r reduction. Clark and Watson suggest that this categorical effect remains as a problem for the cognitive linguistic model of reduction as a gradient phenomenon. The second issue they raise concerns the comparatively low level of t-to-r reduction in it. At first blush this represents a problem for the model, as the pronoun is of course highly frequent. But Clark and Watson suggest that
Cognitive Sociolinguistics
schema theory in cognitive linguistics has not taken on board yet the notion that strong schemas (those with a high type frequency, and relatively low-frequency instances) are resistant to change. They refer to a study by cognitive and developmental psychologists McNeil and Alibali (2002), which in its turn cites a wealth of other studies to support this suggestion. The importance of adding this insight to cognitive linguistic theory is that it helps explain the otherwise unexpectedly low frequency of t-reduction in it in Clark and Watson’s data. Looking at bigrams of t-final words, it is at the top of the list, with as many as twenty-seven collocates in their data, whereas bit has only two collocates. Clark and Watson argue that “[b]ecause the [i(t)#V] schema is instantiated by so many bigrams, it is strong, and so is able to resist the change from t-to-r for longer” (2011: 542). Hollmann and Siewierska’s (2011) consideration of non-standard data has added yet a different dimension to the cognitive linguistic understanding of the causal relation between token frequency and reduction. They show that the reduction of definite articles in Lancashire dialect is not always in proportion to their frequency. Specifically, the words farm and especially smallholdings (which describes a type of small-scale farming, popular in the region) collocate with reduced definite articles more often than one might expect based on their frequency. Hollmann and Siewierska hypothesize that definite article reduction indexes Lancashire identity, and argue in general terms that if a reduced variant comes to perform this sort of indexical function, then even infrequent forms may be reduced, if those forms describe aspects of local experience or culture that are perceived to be significant – which is probably true for farming and smallholdings in Lancashire. Nycz’s (2013) first wave study, relying on traditional sociolinguistic interviews (see Tagliamonte 2006: 37–49), provides yet further evidence for the need to consider that frequency effects may manifest themselves in a variable manner, depending on the social context in which features are embedded. Her study draws on exemplar phonology (e.g. Pierrehumbert 2001, 2003a, 2003b, 2006, Bybee 2001) to help explain certain observations in relation to the vowel system of Canadian immigrants in New York. She finds that the usage-based model allows for a satisfactory explanation of both gradience and frequency effects in her data. In addition, she argues that the comparative lack of a change of Canadian raising in her data relative to the unmerging of the vowels in cot-caught (which are identical in Canadian English) is unsurprising given a model that suggests that social information is stored along with linguistic information. Since Canadian raising is such a salient feature of the variety it is considerably more likely to be stored as Canadian than is the less salient cot-caught merger. Therefore, in a context in which Canadian identity is strongly invoked, as was the case in Nycz’s interviews, it is unsurprising that the raised diphthong /aʊ/ fails to undergo much lowering.
541
542
HOLLMANN
33.3.2
An Inward-Looking Perspective as an Obstacle to Further Progress Having described the main methodological and theoretical achievements of cognitive sociolinguistics, I would like to make a sociology-of-science observation concerning cognitive sociolinguistics. Hollmann and Siewierska (2011: 49) suggest that cognitive sociolinguistics is characterized by a relative lack of interest from sociolinguists. Similar views have been expressed by Gries (2013: 6), Hollmann (2013: 509) and Pu¨tz, Robinson, and Reif (2014: 18). In section 33.3.1, above, I have discussed recent studies by Clark, Watson, Nycz, and Soukup, all of whom selfidentify as sociolinguists. Evidently, then, the cognitive sociolinguistic enterprise is beginning to bring scholars from both fields around the same table, with methodological and theoretical benefits now clearly flowing bidirectionally. However, this does not mean that it is no longer worth drawing explicit attention to these benefits, in particularly the sociolinguistic camp. To illustrate that there is still an issue, Gries criticizes Labov’s model of the way in which speakers store social information associated with linguistic variants, the so-called ‘sociolinguistic monitor.’ Labov’s most recent description of this construct, in a volume on cognitive sociolinguistics, is as follows: Information on social variation is probably stored separately, in a sociolinguistic monitor (Labov, Ash, and Boberg, 2006) which outputs estimates of social distance, and social attributes of the speaker based on previously acquired data. (Labov 2014: 23)
Gries points out that in a usage-based perspective social information is seen as stored as an integral part of linguistic knowledge. Gries is right in saying that this is incompatible with the view of a separate monitor, but what is even more remarkable is the development of Labov’s thinking about the sociolinguistic monitor over time. The (2006) study referred to in the quote from Labov, above, does not in fact discuss the monitor, but Labov et al. (2006: 108, fn.4) do: A growing body of sociophonetic research shows that listeners store and remember information on speaker identity and speech rate even when this information is irrelevant to the main communicative message (Bradlow, Nygaard and Pisani 1999, Hay, Warren and Drager [2006]). This information is either stored and remembered as a discrete social judgment of the speaker or retrieved from remembered exemplars of lexical items (Pierrehumbert 2002). The SLM might be conceived of as a separate processing and storage module, or as the capacity to do a calculation ‘on the fly’ at any time by an inspection of remembered tokens. For the moment, our research is neutral on this issue.
The repetition of this quote in Labov et al. (2011: 434) indicates that the researchers remained uncommitted regarding the cognitive status of the
Cognitive Sociolinguistics
sociolinguistic monitor for some time, yet the subsequent (2014) claim that it is separate suggests that Labov has now actively distanced himself from one of the main tenets of cognitive linguistics. The case of Labov’s sociolinguistic monitor is particularly striking. However, as the special issue of Journal of Pragmatics on cognitive linguistics shows, other prominent sociolinguists (Guy 2013, Preston 2013) also make little attempt to engage explicitly with cognitive linguistic research.
33.4 Looking Ahead: Pushing the Cognitive Sociolinguistic Boundaries In terms of generating ideas for new research, all aspects of theoretical progress described in section 33.3.1, concerning styles, frequency effects, and so on, provide a wealth of opportunities for further research. Conclusions reached on the basis of a single variety are in need of further testing. Conversely, new theoretical insights may help to explain variable aspects of a variety that has not been investigated yet.
33.4.1
Maximizing the Potential Benefit of Cognitive Psychology, Anthropology, and Sociology The foundational sciences of cognitive and sociolinguistics also host many possibilities for new research questions to be generated and/or new perspectives to be taken on variationist data. One way to achieve this is to engage more seriously with the fields on which cognitive linguistics and sociolinguistics, and thus ultimately also cognitive sociolinguistics, are based (see section 33.2, above). The discussion of Clark and Watson (2011), above, illustrates this. They found that they could only explain some aspects of the variability by transferring an insight regarding schema strength from cognitive psychology to cognitive linguistics. It is conceivable that other constructs from cognitive psychology used by cognitive (socio)linguists would also benefit from closer examination of recent psychological literature, not least since many cognitive linguistic constructs were defined in the era when this approach to language first emerged; see, for example, the usual references in cognitive linguistic discussions of prototypes to Rosch and her collaborators (e.g. Rosch 1975, Rosch and Mervis 1975). Mallinson provides a concrete example of how sociolinguistics does not appear to maximize the possible benefits of recent insights from sociology. She argues that although some early sociolinguistic work defined social class in line with then current views in sociology, present-day sociolinguists have insufficiently considered recent work by sociologists that problematizes “male and Eurocentric biases of socioeconomic indices,
543
544
HOLLMANN
the validity of socioeconomic indices, and variability in locating people within class categories” (2009: 1039). To sum up, a critical investigation of sociolinguistic and cognitive linguistic concepts from the point of view of (current) research in cognitive psychology, sociology, and anthropology may lead to new questions and to a more accurate understanding of the cognitive and social-cultural dimensions of linguistic variation. Areas of inspiration need not be restricted to the sciences listed here, as other fields of inquiry are relevant to cognitive sociolinguists as well, prominently including a number of subdisciplines of linguistics itself.
33.4.2 Drawing on Relevant Linguistic Subdisciplines Croft (2009b) argues that cognitive linguistics has focused too much on cognition and too little on interaction. He shows how insights from linguistic and philosophical pragmatics may extend the theory so as to include social cognitive abilities; see Hollmann (2013: 493–97) for a summary of what Croft defines as “social cognitive linguistics” and for discussion of the connection to studies on intersubjectivity by, for example, Sinha (1999, 2004) and Verhagen (2005). However, the subdiscipline I want to foreground here as a possible source of inspiration and guidance for cognitive sociolinguistics is linguistic typology. Typological variation and universals have of course been central to several cognitive linguists (e.g. Talmy 1985, Croft 2001). Furthermore, Chambers’s (2001) study on vernacular universals has led to a productive research program; see, for example, the contributions in Filppula, Klemola, and Paulasto (2009). That volume includes several studies from typologists, who had more or less simultaneously started to apply their theory and methods to dialect variation (see, e.g., Kortmann 1999, 2003). But despite existing links between cognitive and sociolinguistics on the one hand and typology on the other, cognitive sociolinguistic research has not drawn very much upon this field. On the theoretical level, typology offers variationist study an opportunity to view findings in a broader perspective – in many cases a far broader perspective, in fact, since typologists may have studied a linguistic phenomenon in a particular dialect that has attracted the attention of a (cognitive) sociolinguist in hundreds of languages. This is especially likely for morphosyntactic variables, which have traditionally occupied typologists more than phonology. Hollmann and Siewierska (2007) provide an example of the explanatory potential of typological universals, as was hinted in section 33.3.1, above. In their study of reduction in 1Sg possessive forms in Lancashire, they link the phenomenon to typological research on (in)alienability. Inalienable possession is coded differently from alienable possession in many of the
Cognitive Sociolinguistics
world’s languages. The distinction between these two types of possession is gradient rather than categorical, with Nichols’s (1988: 572) hierarchy being as follows: body parts and/or kinship terms > part-whole > spatial relations > culturally basic possessed items > other
Oversimplifying a little, in some languages inalienable possession involves less linguistic distance between the possessor and the possessed, as in the Paamese examples (4–5), below (from Crowley 1996: 389–411). (4)
(5)
vati-n head-3 S G ‘his head’ ani emo-n coconut CLF-3SG ‘his drinking coconut’
Typologists usually explain this effect with reference to iconicity: linguistic distance reflects conceptual distance. The typological perspective leads Hollmann and Siewierska to hypothesize that the reduced forms of my might be found more with inalienably than alienably possessed nouns, since reduction could be analyzed as representing a higher degree of formal integration between the possessive marker and the possessed noun. The Lancashire data support the hypothesis, which poses an interesting challenge for traditional descriptions of (standard) English, as these do not suggest that there are any alienability effects in this language. As explained in section 33.3.1, above, Hollmann and Siewierska also offer an explanation based on frequency. Here, they follow Haspelmath’s (2006) suggestion that alienability effects are in fact caused by frequency effects, with NPs with inalienably possessed nouns (e.g. his head or his brother) typically also being highly frequent and alienably possessed ones (e.g. his drinking coconut) less so. In line with what Haspelmath would predict, Hollmann and Siewierska find that frequency allows for an explanation of more aspects of the variation than alienability. The interface between linguistic typology and dialectology has yielded several other studies which testify to the value of typological universals in the context of variation in and across varieties of a single language. Another example which invokes a well-known typological hierarchy is Siemund’s (2002, 2002) insightful research on the use of animate 3Sg pronouns for inanimate objects in the southwest of England, Newfoundland, and Tasmania. There are differences between the precise way in which animate pronouns are used for inanimates across these varieties, but Siemund observes that all the variation obeys Sasse’s (1993) hierarchy of individuation, which represents a scale from easily individuated entities (such as people) to entities for which individuation is hard or impossible (substances which are normally referred to by means of mass nouns).
545
546
HOLLMANN
The generalization to draw from these studies is that typological universals may both give rise to new cognitive sociolinguistic research questions and help in interpreting variationist data. Typology has yet more to offer to cognitive sociolinguists. Elicitation questionnaires involving translation tasks of sentences carefully designed to test the researcher’s hypotheses are characterized by Anderwald and Kortmann as “the typological ‘u¨ber’-method” (2013: 315). Typologists tend to distribute such questionnaires among colleagues or students who are native speakers of relevant languages. They usually start with a small-scale pilot before finalizing the questionnaire and expanding the number of participants/languages. This method is valuable also when transposed to variationist context, although there are some caveats. One caveat is the difficulty of representing, in writing, many nonstandard variants. This problem does not arise for variables such as word order (e.g. in ditransitives, see Siewierska and Hollmann’s 2007 corpusbased analysis),3 but phonological variation in particular is difficult to capture. Anderwald and Kortmann (2013: 315–16 and references cited therein) discuss the use of “eye-dialect,” such as oot as a representation of the Scottish pronunciation of out in Clark (2007; see section 33.3.1, above) but that does not always work. The reduced definite article in Lancashire and Yorkshire, for instance, is often realized as t’ in literary and laypersons’ representations of local dialect, but the phonetic reality is rather more complicated, with a number of distinct variants; in Lancashire [t], [θ], and [ʔ] (Hollmann and Siewierska 2011). Such different variants might display different distributions and therefore a single eye-dialect representation such as t’ is unsatisfactory. Another issue is the representativeness of the data obtained. Linguists are hardly ordinary language users and so their judgments and translations may not represent genuine vernacular. Linguists are also a rather narrowly defined group in terms of social variables that variationist research tends to be interested in, such as socioeconomic class. It may seem unsurprising, for these and other reasons, that sociolinguists usually construct their sample very differently from typologists (see, e.g., Feagin 2013 for guidance on matters such as informant selection from a sociolinguistic perspective). However, typologists’ questionnaires have several advantages as well, which mean that they still deserve a place in the (cognitive) sociolinguist’s toolkit. 3
Corpus methods are covered by Gries (this volume Ch. 36), but it is worth noting here that there are a number of dialect corpora available, especially for English. For British English varieties the choice includes the Freiburg English Dialect Corpus (FRED) and the Helsinki Corpus of British English Dialects (HD); for American English two options are the Nationwide Speech Project Corpus (NSP) and the Santa Barbara Corpus of Spoken American English (SBCSAE). For comparison of world Englishes, the ICE Corpus is the best place to start; this may be very interesting in conjunction with eWAVE (discussed in the main text, below). For other languages the selection is more limited. The main varieties of Chinese can be investigated with the Linguistic Variation in Chinese Speech Communities Corpus (LIVAC), whilst for Spanish and Portuguese the Corpus del Español and Corpus do Português are currently being expanded so as to enable regional dialect analysis.
Cognitive Sociolinguistics
First, it is reasonable to assume that linguists’ judgments are less polluted by the kinds of normative pressures which ordinary language users are subject to (and which many sociolinguists seek to avoid; see, e.g., Cornips and Poletto 2005). Second, linguists tend to be prepared to complete far longer questionnaires than laypersons. This is particularly helpful where the feature under investigation is very rare in natural discourse and/or where differences in acceptability are subtle.4 On a related note, a linguist filling out a typological questionnaire will often volunteer helpful additional information. This may concern, for instance, another minimal variation of a test sentence which would sound more or less acceptable in the language in question, or an informed suggestion as to why a certain difference in the acceptability of two sentences may obtain. Kortmann and Lunkenheimer (2013) have employed this questionnaire method to build the Electronic World Atlas of Varieties of English (eWAVE). Modeled on the World Atlas of Language Structures (WALS; Dryer and Haspelmath 2013), it contains information from seventy-six varieties of English on 235 morphosyntactic variables. It may be a useful resource for cognitive sociolinguists in a multitude of ways. One could for instance use it to discover some features of English dialects one was not previously aware of. These may lead to new research questions; for example, because the frequency information in eWAVE is fairly crude, classifying a feature in terms of whether it is ‘pervasive or obligatory,’ ‘neither pervasive nor extremely rare,’ ‘extremely rare,’ or absent. A variety where a feature is reported to have medium-high frequency may then invite an investigation of possible constraints on its variability. As with any database, some care must be taken when using eWAVE. One issue emerges from the limited number of informants, in many cases just one per variety. Moreover, the classification of varieties is relatively broadstroke. For example, northern English dialects are conflated into just a single category: ‘English dialects in the North of England.’ This makes the frequency information difficult to interpret. Consider definite article reduction, which is characterized in this area as ‘neither pervasive nor extremely rare.’ Given that definite article reduction does not appear to occur in Northumberland (Jones 2002: 325–26) and assuming that the (single) informant for these northern dialects is aware of that, does that perhaps mean that this informant would characterize the frequency in Lancashire and Yorkshire as pervasive, but considering Northumbria as well has arrived at a kind of average for the whole region? Or has the description of this feature in eWAVE, as ‘use of zero article’ rather than reduction, led the informant to suggest that the feature is not pervasive, since full omission is only the endpoint of a cline in these 4
See Hoffmann (2013b) for guidance on how such issues may also be addressed by means of magnitude estimation acceptability judgments. Borrowed from psychophysics and inspired also by methods in psychology, magnitude estimation was pioneered in linguistics by generative grammarians (see especially Cowart 1994, 1997), but its value is by no means restricted to that framework.
547
548
HOLLMANN
varieties, which also includes reduced but non-zero forms such as [t], [θ], and [ʔ] (see above)? Finally, information about what other varieties also have zero definite articles may certainly open up avenues for cognitive sociolinguistic research, but care is needed in comparing varieties. To illustrate the issue, I note that eWAVE suggests that zero definite articles are similarly frequent in Sri Lankan English as in northern English dialects. However, the phenomena are almost certainly completely different. Definite article omission in Sri Lankan English is unlikely to be a gradient, repetition-driven process; instead, it is probably caused by an absence of definite articles in Sinhala (Lyons 1999: 323) and also in Tamil (which has demonstrative forms but no grammaticalized definite article; cf. Rhenius 1836: 219, Lehmann 1989: 148–49).
33.5 Conclusion This chapter has aimed to provide guidance to future research in cognitive sociolinguistics. In order to make the most of both cognitive linguistics and sociolinguistics, it is important to be aware of the historical and intellectual foundations of these fields. Countless opportunities for further research may emerge upon careful (re)consideration of the sciences in question – cognitive psychology, sociology, and anthropology – as well as of other subdisciplines of linguistics itself, such as pragmatics and typology. Authors and reviewers (e.g. Hollmann 2013, Koller 2014, Pu¨tz, Robinson, and Reif 2014, Jensen 2015) still tend to refer to cognitive sociolinguistics as an ‘emerging’ paradigm. Yet as a named field of inquiry it now goes back almost a decade and a half (Geeraerts 2003), and prior to that several other usage-based cognitive linguists had already suggested, at least in a programmatic manner, that social information should be seen as part of linguistic knowledge (Hudson 1996, Langacker 1999a). Two decades after the publication of Hudson (1996), then, it seems appropriate in a handbook chapter such as this to take stock of the degree of interplay so far between cognitive linguists and sociolinguists. In section 33.3.1 I summarized a considerable number of significant theoretical insights gained so far. I also observed (section 33.3.2) that although cognitive linguists have been rather more keen to take on board insights from sociolinguistics than vice versa, there is now a generation of scholars who self-identify as variationist sociolinguists but who draw increasingly heavily on cognitive linguistics, and present and publish their work in peer-reviewed cognitive linguistic venues. In comparison, a search of major journals in sociolinguistics yields very few hits for contributions from cognitive linguists. In order to further enhance collaboration and progress, the time may have come for usage-based cognitive linguists to take their work to such venues and establish cognitive sociolinguistics more firmly – first by exposure, then by entrenchment.
34 Computational Resources: FrameNet and Constructicon Hans C. Boas 34.1 Introduction This chapter discusses how ideas, concepts, and methodologies from cognitive linguistics have been applied to the creation of two specific computational resources for linguistics research.1 After this introductory section, section 34.2 discusses one of the most prominent computational resources inspired by cognitive linguistics, the FrameNet lexical database (http://framenet.icsi.berkeley.edu), which is structured according to the theory of Frame Semantics (Fillmore 1982). Section 34.3 shows how the FrameNet methodology and representational format for creating lexical entries has been expanded for the creation of a so-called ‘constructicon,’ aimed at creating entries for grammatical constructions. Section 34.4 briefly outlines how FrameNet and the Constructicon can be used for research in cognitive linguistics.
34.2 From Frame Semantics to FrameNet 34.2.1 Historical Background Fillmore’s seminal 1968 paper ‘The case for case’ proposed a list of semantic roles aimed at identifying grammatically relevant facets of a verb’s meaning in terms of a set of labels such as Agentive, Instrumental, Dative, Factitive, and Locative. These labels served to identify the role played by each of the verb’s arguments in the event it denotes. The collection of sets of different semantic roles came to be known as 1
Thanks to Oliver Čulo, Marc Pierce, and Josef Ruppenhofer for valuable feedback on earlier drafts of this chapter. The usual disclaimers apply. Work on this chapter has been supported by a Fellowship for Advanced Researchers from the Alexander von Humboldt Foundation, Germany, as well as the grant ‘Cognitive modeling and meaning construction across languages: theory, implications, and applications’ from the Spanish Ministry of Economy and Competitiveness (grant # FFI2013-43593-P).
550
BOAS
case frames, which specify a verb’s semantic valency. One of the major ideas behind these semantic roles was that they are ordered in a hierarchy depending on their realization in terms of grammatical function. This hierarchy put the semantic role Agentive at the top, followed by Instrumental, Objective, and other semantic roles. Depending on a verb’s valency, different semantic roles are mapped as subject. For example, the case frame of the verb to open consists of three semantic roles: Agentive, Instrumental, and Objective. When all three semantic roles are present, the Agentive is realized as the subject (e.g. Kim opened the door), but when the Agentive is not present, the next semantic role on the list is realized as the subject (e.g. The key opened the door). Following the initial success of Fillmore’s proposals, several problems emerged regarding the status of semantic roles, including issues with the definition and granularity of semantic roles, problems reflecting cross-role generalizations, and problems of one-to-one correspondence (for an overview, see Levin and Rappaport Hovav 2005). As a result, during the 1970s Fillmore moved away from his initial proposal that semantic roles are primary and that there should only be a rather limited set of semantic roles (see Fillmore 1975, 1976, 1977b, 1977c, 1978, 1979). The goal of this move was to develop an elaborated model of the semantics of understanding, in contrast to a truth-conditional semantics, the prevalent approach to linguistic semantics at the time. More specifically, Fillmore suggested that so-called semantic frames should be regarded as primary for the description and analysis of meaning (and its syntactic relevance), and that semantic roles should be defined in terms of their semantic frames. The early 1980s saw the first fully developed version of Frame Semantics, which was articulated as a model based on the full and rich understanding required for the production of a text as well as the understanding of a text (Fillmore 1982). The central idea regarding the status of semantic frames for the understanding of words and texts is summarized by Fillmore and Atkins (1992: 76–77) as follows: A word’s meaning can be understood only with reference to a structured background of experience, beliefs, or practices, constituting a kind of conceptual prerequisite for understanding the meaning. Speakers can be said to know the meaning of the word only by first understanding the background frames that motivate the concept that the word encodes. Within such an approach, words or word senses are not related to each other directly, word to word, but only by way of their links to common background frames and indications of the manner in which their meanings highlight particular elements of such frames.
One of Fillmore’s examples illustrating the central role played by semantic frames concerns the meaning of the term bachelor. Fillmore points out that while this term typically refers to an unmarried man, not all unmarried men such as the pope or Tarzan would typically be characterized as
Computational Resources: FrameNet and Constructicon
bachelors. The reason, according to Fillmore, is that the meaning of bachelor makes reference to other extrinsic entities, values, and concepts, such as religious customs (e.g. certain societies and religious institutions do not allow certain people to marry) and more general cultural knowledge. For example, the figure of Tarzan, though living with a female (Jane), would not fit the description of bachelor, since the cultural norms of Western society as captured by the Personal_Relationship frame do not apply to his living environment (the jungle).2 These observations led Fillmore to the insight that it is necessary to characterize the meanings of words in terms of experience-based schematizations of events and objects in the speaker’s world (see Petruck 1996, Croft and Cruse 2004, Boas 2013c). Thus, the semantic frame evoked by the word bachelor is much more complex than that suggested by simple dictionary definitions or checklists of semantic features. The reference to extrinsic knowledge structured in terms of semantic frames can be of various sorts and levels of complexities. For example, they may refer to events, such as Giving_birth (to birth, to bear) or death (to croak, to die, death); relations, such as Personal_relationship (friend, bachelor); states, such as Being_located (to find, situated); entities, such as Gizmo (appliance, device, machine); scales, such as Temperature (hot, freezing); and person and spatial deixis.3 The verb to avenge is an example of an event. In order for a speaker of English to be able to interpret a sentence such as Rick avenged the death of his pet armadillo by killing the coyote, one must have knowledge about the various events leading up to the point in time when the sentence is uttered. This knowledge is captured by the Revenge frame, which is evoked by the Lexical Unit (LU; a ‘word’ in one of its senses4) to avenge, as well as other semantically related LUs such as to revenge, revengeful, and revenger. The frame definition of a frame is a prose description of a situation involving various participants and other conceptual roles, each of which constitutes a Frame Element (FE), marked in small caps. The Revenge frame One person (we call him the O F F E N D E R ) did something to harm another person (what he did we call the I N J U R Y and his victim we call the I N J U R E D _ P A R T Y ); reacting to that act, someone (the A V E N G E R , possibly the same individual as the I N J U R E D _ P A R T Y ) acts so as to do harm to the O F F E N D E R , and what he does we call the P U N I S H M E N T .
Frames such as the Revenge frame capture the rich knowledge that speakers associate with words. The items in small caps are the so-called Frame Elements (FEs), which are situation-specific types of more general 2
Following standard practice in Frame Semantics, names of frames are given in Courier New font.
3
Fillmore’s use of the term ‘frame’ is somewhat related to work in artificial intelligence as found in Minsky (1977) and psychology (Schank and Abelson 1977). For an extensive discussion of the use and various meanings of the term ‘frame,’ see Ziem (2008) and Busse (2012).
4
A LU can also consist of a multiword expression, such as give the slip or put into words.
551
552
BOAS
semantic roles such as A G E N T , P A T I E N T , or I N S T R U M E N T (see section 34.2.2.2 for details on the frame hierarchy allowing for generalizations).5 The FEs, which are participants in the frame, are also explicitly defined. For example, the A V E N G E R is the person who enacts revenge (Rick in the example above), the O F F E N D E R is the original offender (not mentioned in the example above), the I N J U R E D _ P A R T Y is the offender’s victim (also not mentioned in the example), the I N J U R Y is the result of the offender’s act (the death of his pet armadillo), and the P U N I S H M E N T is the avenger’s act (by killing the coyote). Applying the FE labels to the individual constituents in our example sentences yields the following representation: (1)
[Rick] avengedtgt [the death of his pet armadillo] [by killing the coyote]. [DNI] [ DNI]
The LU evoking the semantic frame is called the target LU, indicated by the superscripted ‘tgt’ following avenged in (1). Each constituent representing an FE of the frame evoked by the target LU is in square brackets, with the name of the FE in subscript following the opening bracket. One important feature of Frame Semantics is that it is also concerned with documenting the types of FEs that can be omitted under certain circumstances (Fillmore 1986). The phenomenon known as Null Instantiation (NI) covers three different subtypes. First, it covers Definite Null Instantiation (DNI), which is lexically specific, understood from discourse, and for which knowledge of the missing material is required for determining the referent. In (1) above, both O F F E N D E R and I N J U R E D _ P A R T Y are characterized as DNI, since their identity can be construed from discourse. Second, Indefinite Null Instantiation (INI) involves lexically specific, intransitive uses of transitive verbs such as eat, drink, sew, and bake, as well as knowledge about the category of missing material, even if it is not mentioned in the previous discourse or if it is not recoverable from context, as in John baked again. Third, Constructional Null Instantiation (CNI) describes situations in which particular grammatical constructions such as the imperative or agentless passive may license the lexical omission. Frame Semantics and FrameNet, as shown below, are interested in documenting null instantiation properties because they are crucially important for our understanding of how the semantics of frames may be realized syntactically.6 We now turn to a discussion of how the principles of Frame Semantics have been applied in the creation of FrameNet.
5
Earlier work by Fillmore distinguishes between frames, scenes, and scenarios. However, since the mid-1990s, the term ‘frame’ has become the main concept around which Frame Semantics and FrameNet have evolved. For a discussion of the similarities and differences between these terms, see Schmidt (2009).
6
For further references, see Lambrecht and Lemoine (2005), Ruppenhofer and Michaelis (2010) and Lyngfelt (2012).
Computational Resources: FrameNet and Constructicon
34.2.2 FrameNet The FrameNet project, founded in 1997 at the International Computer Science Institute in Berkeley, is primarily concerned with compiling FrameNet (framenet.icsi.berkeley.edu/fndrupal/), a lexicographic database according to the principles of Frame Semantics. Its aim is to provide rich frame semantic knowledge about the core English vocabulary based on manually annotated corpus data, including valence descriptions for each item analyzed (Fillmore, Johnson, and Petruck 2003).7 FrameNet, which is freely available for academic research, currently contains about 1,200 frame definitions (including definitions of their respective Frame Elements), together with lexical entries for more than 13,100 Lexical Units evoking frames, more than 200,000 annotated corpus examples, and nearly 1,800 frame-to-frame relations illustrating how semantic frames are connected to each other via a hierarchy (as of September 2015). The following sections provide more details about (1) the workflow of the FrameNet project, (2) the structure of frames and lexical entries in the FN database, and (3) the frame hierarchy. 34.2.2.1 FrameNet Workflow The workflow of the FrameNet project consists of three main stages. At the first stage, the FN staff proposes new semantic frames and LUs that evoke them. This process typically involves the identification of a particular sense of a specific word that represents the prototypical meaning of that word and the frame more generally. An example is the LU avenge, discussed above, whose meaning expresses the prototypical meaning of the Revenge frame. FN lexicographers arrive at preliminary frame descriptions by first comparing definitions of prototypical LUs in various dictionaries, and then discussing these definitions and comparing them with their own intuitions. At that point, FN lexicographers also carefully study the contexts in which the relevant LUs appear in electronic corpora such as the British National Corpus and the American National Corpus. This empirical verification helps FN lexicographers with arriving at clearer initial definitions of semantic frames and their FEs. Next, FN lexicographers search for other LUs evoking the same frame including verbs, nouns, and adjectives. The basic criterion for defining the boundaries of a frame is that all LUs should evoke the same type of event and share the same inventory and configurations of FEs (see Atkins, Rundell, and Sato 2003, Petruck et al. 2004, Ruppenhofer et al. 2010, Ruppenhofer, Boas, and Baker 2013). A combination of objective criteria, such as dictionary definitions, entries in thesauruses or in other databases such as WordNet and PropBank, and corpus examples, together with the lexicographer’s linguistic intuitions, 7
FrameNet has also produced frame-semantic annotations of continuous texts to show how semantic frames can contribute to the understanding of full texts. For details, see https://framenet.icsi.berkeley.edu/fndrupal/index.php? q=fulltextIndex.
553
554
BOAS
form the basis for careful discussions about defining the boundaries of frames and identifying the LUs that evoke it. Based on the frame description and the list of LUs evoking the frame, FN staff extracts examples sentences from electronic corpora for each target LU (see Atkins, Rundell, and Sato 2003, Fillmore, Johnson, and Petruck 2003, Boas 2005b). The second stage of the FN workflow consists of annotating example sentences that FN lexicographers extracted from electronic corpora. To achieve this goal, trained annotators use a software called the FN Desktop to manually annotate about ten to twenty sentences per LU (Ruppenhofer et al. 2010, Fillmore, Johnson, and Petruck 2003). The number of sentences suitable for annotation depends on the frequency of a particular LU in the corpus, as well as on the syntactic complexity of sentences in which a given LU occurs. For example, since FN’s main emphasis is on compiling a lexicographic database, annotators typically do not chose to annotate complex sentences involving multiple complex grammatical constructions. As shown in Figure 34.1, the FN desktop software allows annotators to choose between different frames, which are displayed in an alphabetically ordered list on the left. By clicking on the name of a frame, annotators see the list of FEs as well as the list of LUs. Clicking on a specific LU such as to avenge pulls up the relevant subcorpora (based on syntactic contexts) available for annotation. Clicking on one of the subcorpora displays its contents in the upper part of the screen. When none of the sentences in a subcorpus are annotated, only the target LU evoking the frame (here: to avenge) is marked in black (as
Figure 34.1 Subcorpora of avenge for annotation in the FN Desktop
Computational Resources: FrameNet and Constructicon
in the top line in Figure 34.1). Clicking on one of the sentences in the top part of the screen leads to it being displayed in the middle part of the screen, with the sentence ready for annotation (here: the sentence With this, El Cid at once avenged the death of his son. . .). Annotators then read the sentence and identify, for each relevant constituent, its FE status; they then scroll with the mouse over the constituent to mark it, and then they choose from the list of FE names in the lower part of the screen the proper labels for each FE. Clicking on it marks the constituent as a particular FE. Only the dependents of one frame-evoking target LU are annotated per example sentence. Automated scripts then fill in additional information for each FE, specifically labels for Grammatical Function (GF) and Phrase Type (PT), which can be corrected by the annotator if necessary. As a result, each relevant constituent is annotated with three layers of information. For example, the phrase With this at the beginning of the sentence in Figure 34.1 is annotated with the FE label P U N I S H M E N T , the GF label Complement (Comp), and the PT label PP. The constituent El Cid is annotated with the FE label A V E N G E R , the GF label External Argument (Ext), and the PT label NP. Finally, the constituent the death of his son is labeled with the FE label I N J U R Y , the GF label Object (Obj), and the PT label NP. The resulting annotation thus consists of three layers combining, for each relevant constituent, information about FE, GF, and PT status. Once annotators have labeled for each LU in a frame ten to twenty example sentences, the annotation phase of the workflow is completed.8 Based on this information, a set of scripts automatically summarizes the valence of each LU by abstracting over the annotated corpus examples (Ruppenhofer et al. 2010). The third stage of the FN workflow consists of producing lexical entries for each LU, based on the information contained in the FN database (see Baker, Fillmore, and Cronin 2003). To achieve this goal, automated scripts generate lexical entries that include a brief definition of the LU, the name and definition of the frame it evokes, valence tables summarizing the exhaustive information about every attested combinatorial possibility of FEs and their syntactic realizations (i.e. GF and PT), and annotated example sentences. The following section details the structure of lexical entries from the perspective of a user accessing the FrameNet database online.
34.2.2.2 FrameNet Data There are different methods of accessing the data in FrameNet. The first method is the download of FN data as a stable data release (see Baker 2012). This method requires registration with FrameNet. The second and perhaps 8
The description of the FN workflow may sound linear, but it is not. For example, when compiling lists of LUs evoking a semantic frame, the group of FN lexicographers has to take into consideration what types of subcorpora (based on syntactic contexts) will be used for annotation later on. Similarly, when annotators apply the frame descriptions and FE definitions, compiled by the FN lexicographers, to corpus examples, they may discover inconsistencies, which in turn need to be discussed with the FN lexicographers. The result of this exchange may lead to updated frame definitions, the definition of new frames that are closely related to the original frame, and the re-assignment of LUs to other frames.
555
556
BOAS
Figure 34.2 Description of Revenge frame
more popular method is the direct online access via FN’s website at http:// framenet.icsi.berkeley.edu/fndrupal/. This method does not require any registration and offers the user the most updated version of the FN database as well as multiple ways of viewing FN data. The search option on FN’s home page is the most prominent way of finding a lexical entry. For example, a search for to avenge returns a table including the closest match, in this case the frame Revenge, which is evoked by avenge. In the case of polysemous words, this table includes multiple search results, one per LU/semantic frame. Clicking on the name of the frame leads to the display of the so-called frame report, which contains the frame’s description, the definition of FEs, and the list of LUs evoking the frame. As can be seen in Figure 34.2, illustrating the Revenge frame, the frame description consists of a few lines of prose text explaining how the individual FEs, which are color coded, relate to each other in different ways. The frame description is augmented by a few clear examples from the corpus, which are color coded according to the FEs they contain. The frame description is followed by a list of FE definitions, typically differentiating between core and non-core FEs. Each FE definition is exemplified by a corpus example illustrating the use of the FE in a specific context. Core FEs are those types of FEs that uniquely define a frame, such as the FEs A V E N G E R , O F F E N D E R , and I N J U R E D _ P A R T Y in the Revenge frame. Non-core FEs are peripheral FEs used to describe aspects of events more generally, such as T I M E , P L A C E , and M A N N E R . In contrast, so-called extra-thematic FEs do not conceptually belong to the frame in which they occur, that is, they situate an event against the backdrop of another state of affairs (Boas 2013b). For example, in Sue baked the cookies for me, the PP for me is an extra-thematic R E C I P I E N T FE, which is not an important part of
Computational Resources: FrameNet and Constructicon
a situation in which some edible entity is created (for details, see Fillmore and Baker 2010 and Ruppenhofer et al. 2010). Following the definitions of FEs, we find specifications about so-called frame-to-frame relations, which specify how a particular frame is related to other frames in the FN hierarchy. In the case of the Revenge frame, we learn that it inherits from a more general frame Rewards_and_Punishments. The last part of the frame report contains a list of LUs evoking the Revenge frame, together with links to their lexical entry reports and their annotation reports.9 Clicking on the annotation report link for avenge displays all sentences that have been annotated with frame-semantic information (color-coded FEs), and which form the basis for the lexical entry report. Clicking on the lexical entry report link for avenge opens up a new web page with more details. Figure 34.3 presents the first part of the lexical entry report of avenge. The first lines at the top of the lexical entry report identify the part of speech as verb and the frame it evokes as the Revenge frame. The relatively short prose definition is followed by a table illustrating how the FEs of the Revenge frame are syntactically realized. The column on the left lists the name of the respective FEs, the middle column displays the total number of annotated example sentences containing that FE found in the annotation report (see below), and the column on the right presents the full range of syntactic realizations for each FE.10 For example, the FE A V E N G E R occurs in thirty-three annotated example sentences, and in these sentences it is realized syntactically either as an external noun phrase (NP.Ext) (e.g. Swegen is also said to have invaded England later to avenge his brother), as an external possessor phrase (Poss.Ext) (e.g. Though Satan’s motive for avenging against God is not made quite clear. . .), or it is not overtly realized because it is constructionally null instantiated (CNI) (e.g. I want Leila avenged). Clicking on a hyper-linked number will display the annotated example sentence(s) containing that FE in the lower part of the screen. Scrolling down to the second part of the lexical entry report presents the user with a table listing the valence patterns of that LU, as shown in Figure 34.4. The valence table records all the attested combinations of FEs and their various syntactic realizations as they occur in the annotated corpus sentences. The column on the left lists the total number of annotated example sentences illustrating a particular Frame Element Configuration (FEC). Clicking on a hyperlink will display the corresponding annotated example sentence(s) in the lower part of the screen. The valence table for avenge lists a total of six FECs, together with their various syntactic realizations. For 9
FrameNet also documents multiword expressions such as phrasal verbs (e.g. pick up, take off, take up), words with selected prepositional complements (e.g. depend on, object to, proud of, fondness for), support constructions (e.g. take comfort in, put emphasis on, at risk, under arrest), combinations in which selected prepositional complements are combined with a particle or a noun (e.g. put up with, i.e., tolerate, take comfort in, take under consideration), and transparent nouns (e.g. my gem of a wife, in a part of the room, on this part of the shelf) (see Fillmore 2008).
10
As a lexicographic database, FrameNet does not provide information about the LU’s frequencies.
557
558
BOAS
Figure 34.3 First part of lexical entry report of avenge
example, the first FEC consists of [A V E N G E R , I N J U R E D _ P A R T Y , I N J U R Y , O F F E N D E R , P U N I S H M E N T ] while the second FEC consists of [A V E N G E R , I N J U R E D _ P A R T Y , I N S T R U M E N T , O F F E N D E R , P U N I S H M E N T ]. Each FEC lists the range of syntactic realizations: the first FEC has two different syntactic realizations, the second FEC has only one particular way in which the FEs are realized syntactically, and the third FEC [A V E N G E R , I N J U R E D _ P A R T Y , O F F E N D E R , P U N I S H M E N T ] has a total of four syntactic realizations. The valence table for avenge contains a total of seventeen different options for the syntactic realization of the Revenge frame’s FEs. Providing such detailed information makes it possible to investigate similarities and differences between LUs evoking the same frame. For example, if one wants to find out how verbal LUs such as avenge, get back at, get even, retaliate, and revenge realize the Revenge frame’s FEs, one can compare the variation of the FE’s syntactic realizations. Table 34.1 shows
Computational Resources: FrameNet and Constructicon
Figure 34.4 Second part of lexical entry report of avenge
the comparison of the variation in syntactic realization of the five verbal LUs evoking the Revenge frame. Table 34.1 shows how the verbal LUs evoking the Revenge frame realize its FEs in different ways. The comparison suggests that there are some general syntactic realization patterns that are shared by all LUs; fo example, the A V E N G E R is realized as the external NP (and it can also be
559
Table 34.1 Syntactic variability of FEs with verbal LUs evoking the Revenge frame11
AVENGER NP.Ext
Poss.Ext CNI
DNI
INI
11
avenge retaliate get back get even revenge avenge avenge retaliate get back get even revenge
DEPICTIVE
INJURED_ PARTY
INJURY
avenge
avenge
avenge retaliate get back get even get even
MANNER
INSTRUMENT OFFENDER
PUNISHMENT PURPOSE
avenge retaliate get even
retaliate
avenge get even revenge
avenge retaliate get back get even revenge
TIME
Table 34.1 lists only the syntactic variability of FEs with verbal LUs evoking the Revenge frame. Similar tables comparing the syntactic variability in the realization of FEs in the Revenge frame can be compiled for other parts of speech; for example, the adjectives retributive, retributory, revengeful, vengeful, and vindictive, as well as for the nouns avenger, payback, retaliation, retribution, revenge, revenger, sanction, and vengeance.
NP.Obj
avenge revenge
PP[for].Dep
avenge retaliate get back get even avenge
PP[of].Dep PP[in].Dep PP[on].Dep
get back retaliate
PP[against].Dep PP[by].Dep PP[with].Dep
avenge
retaliate avenge retaliate revenge avenge retaliate
retaliate revenge
get even
PPing[by].Dep PPing[for].Dep VPto.Dep AVP.Dep AJP.Dep
avenge
avenge avenge retaliate avenge retaliate
revenge avenge retaliate revenge
retaliate
avenge retaliate
562
BOAS
constructionally null instantiated), and the P U N I S H M E N T can be null instantiated (INI). However, this is where the commonalities between the LUs end. As the data in Table 34.1 show, most syntactic realizations of a particular FE occur with only one, two, or sometimes three verbal LUs. In other words, there is only little overlap between a frame’s FEs and how they are realized syntactically with different LUs evoking that frame.12 These observations may appear rather tedious at first glance.13 However, they have important ramifications for investigating linguistic phenomena such as linking between semantics and syntax (cf. Pinker 1989, Rappaport Hovav and Levin 1998, Boas 2008a), profiling of verbal participant roles to account for their syntactic distribution (Goldberg 1995, Nemoto 1998, Boas 2003b/2011a, Croft 2012), and verb classification (Levin 1993, Baker and Ruppenhofer 2002, Croft 2003, Boas 2008b, 2011b, Engelberg et al. 2011, Faulhaber 2011). In other words, the detailed information in the lexical entries of FN offers valuable insights into determining whether semantically related verbs (or adjectives/nouns) exhibit similar syntactic distributions or whether they differ and how. These insights, in turn, have important implications for our understanding of how lexical information is organized and how it interacts with other types of linguistic processes. The information in valence tables such as those in Figure 34.4 above are also helpful for research in Construction Grammar, specifically in the area of argument structure constructions. Goldberg’s (1995) seminal work emphasizes the importance of argument structure constructions for the licensing of particular syntactic phenomena such as ditransitive, causedmotion, resultative, and other related constructions. On this view, lexical entries of verbs may fuse with independently existing meaningful argument structure constructions, which then contribute additional meaning to a verb’s lexical entry, thereby licensing the different argument structures commonly found with verbs. However, the fusion of lexical entries with Goldberg’s argument structure constructions is sometimes problematic. Various authors such as Boas (2003b, 2005b), Nemoto (2005), Iwata (2008), and Herbst (2011) have pointed out that it is often not possible to easily restrict the fusion of verbs and constructions because of the sparse amount of information contained in Goldberg’s lexical entries. The alternative view, known as the lexical-constructional approach, holds that information in lexical entries should be more detailed in order to allow for a more regulated process of fusion between constructions and lexical entries. More recently, various researchers have suggested combining the two approaches, thereby arriving at a network of interrelated constructions spanning from low-level so-called mini-constructions representing 12
Recall that even though the BNC is a relatively large corpus, it is not exhaustive. This means that FrameNet does not make any claims about complete coverage, but only about the distribution of LUs as they occur in the BNC.
13
The valence tables of LUs evoking frames with many more LUs, such as the Self_motion frame, exhibit an even greater degree of variability regarding how FECs may be realized syntactically. See Boas (2001, 2008c).
Computational Resources: FrameNet and Constructicon
Intentionally_affect
Response
Rewards_and_punishments
Revenge
Fining
Corporal_punishment
Execution
Figure 34.5 Partial FrameGrapher representation of Revenge frame
individual senses of words (comparable with FN-like entries, together with additional information) to higher-level schematic argument structure constructions (see Welke 2009, Boas 2010a, 2011b, Herbst 2014). Besides the lexical entry reports and annotation reports, there are other ways of accessing FN data. One alternative way of accessing FN data is via the FrameGrapher visualization tool (see https://framenet.icsi .berkeley.edu/fndrupal/FrameGrapher). While some of this information is also available as a part of the lexical entry report, the FrameGrapher allows for a much more intuitive method for exploring how frames and their frame elements are connected to each other in the frame hierarchy. Recall that the status and granularity of FEs in FN are quite different from those of the original semantic roles proposed by Fillmore (1968). Instead of positing a small set of semantic roles ordered in a strict hierarchy, FN currently contains more than 1,200 frame descriptions including definitions of the frame-specific FEs. The number of FEs is growing exponentially as new frames are added to the FN database, and they are systematically related to each other via so-called frame-to-frame relations, such as those representing complex events (SUBFRAME, PRECEDES), ‘systematic’ relations (CAUSATIVE OF, INCHOATIVE OF), and generalizations (PERSPECTIVE ON, INHERITANCE, USING). For details, see Fillmore and Baker (2010) and Ruppenhofer et al. (2010). The FrameGrapher tool allows users to explore the hierarchy of frames in which some frames are instances of others; some are components of others, etc., thereby allowing for systematic accounts of how the semantics of frames that are instances of each other are realized similarly at the semantic level; see, e.g. Boas (2010b). Another advantage of the FrameGrapher visualization tool is that it allows researchers to identify situation-specific frames and their FEs as instantiations of high-level frames such as Event, Action, and Intentionally_act, together with their more general semantic roles such as A G E N T , P A T I E N T , I N S T R U M E N T , etc. These insights can be used for research on linking (for details, see Van Valin and Wilkins 1996, Ziem 2008, and Fillmore and Baker 2010); for example, using the FrameGrapher to display the
563
564
BOAS
relationships holding between the Revenge frame and other related frames yields the graphic representation in Figure 34.5. Figure 34.5 shows the Revenge frame inheriting from the Rewards_and_Punishment frame, which in turn inherits from the Intentionally_affect and Response frames. Other frames inheriting information from the Rewards_and_Punishment frame include the Fining, Corporal_punishment, and Execution frames. Note that the display of information in Figure 34.5 is optimized for size and layout, that is, the FrameGrapher only displays two parent generations of the Revenge frame. If interested in learning more about higher-level generations, a user may also display that information. In the case of the Intentionally_affect frame (one of the parent frames of the Revenge frame), this includes the Intentionally_act and Event frames, which can be regarded as higher-order parent frames of the Revenge frame.14 Users may also view FN data with the FrameSQL search tool, which allows systematic searches of FN frames for specific combinations of parts of speech, phrase types, and grammatical functions (http://sato.fm .senshu-u.ac.jp/frameSQL/fn2_15/notes/) (Sato 2008). One important feature of the FrameSQL search tool is its ability to also access the databases of FrameNets for other languages, such as Japanese, German, and Spanish. This enables users to compare semantic frames and their syntactic realizations with different LUs across languages, thereby serving as a valuable comparative database for contrastive and cross-linguistic research, as section 34.2.3 shows.
34.2.3 FrameNets for Other Languages Initial research in the 1990s explored the feasibility of applying framesemantic insights to the systematic analysis of the lexicons of languages other than English (Heid 1996, Fontenelle 1997). Subsequent research demonstrated in greater detail how semantic frames derived on the basis of English data could be employed for the creation of FrameNets for other languages (Fillmore and Atkins 2000, Boas 2001, 2002, 2005b, Petruck and Boas 2003). The results of this research inspired the creation of FrameNets for other languages, most notably Spanish (Subirats and Petruck 2003, Subirats 2009), Japanese (Ohara et al. 2004, Ohara 2009), German (Burchardt et al. 2009), and Swedish (Borin et al. 2010). While the technical resources and workflows of the non-English FrameNets differ slightly from each other, they all produce FN-style entries similar to the English original discussed above, including detailed information about how the semantics of a given frame are realized syntactically by different LUs evoking that 14
FN has also conducted full text annotations, in which all frame-evoking LUs in a given sentence were analyzed according to the semantic frames they evoke. See the FrameNet homepage for full text annotations.
Computational Resources: FrameNet and Constructicon
frame in different languages.15 This information allows researchers to systematically conduct contrastive and comparative research on topics in lexical semantics such as polysemy (Fillmore and Atkins 2000, Boas 2005c, Willems 2012), typological differences in profiling properties (Ohara 2009, Petruck 2009), and the interface between the lexicon and syntax (Hasegawa et al. 2010, Bouveret 2012). The multilingual data contained in the FrameNets for various languages is also relevant for cross-linguistic research on argument structure constructions, as shown by Boas (2003a), Leino (2010), and Timyam and Bergen (2010), among others.16
34.3 The Constructicon While the primary focus of the lexicographic work by FrameNet researchers is on the semantic and syntactic properties of lexical items, there has also been considerable interest in exploring ways of creating a so-called constructicon, a database of English grammatical constructions, annotating sentences by noting which parts of them are licensed by which specific constructions (Fillmore 2008). This idea is a natural extension of FN’s work on the lexicon, because in Construction Grammar (CxG), the sister theory of Frame Semantics, there is no strict separation between the lexicon and syntax. In this view, language consists of a large structured inventory of constructions, pairings of forms and meanings (where frames constitute the semantic side of constructions), which vary in size and complexity. These include, among others, non-lexical constructions such as highly schematic constructions and meaningful argument structure constructions, as well as (partially) idiomatic constructions, complex words, words, and morphemes (Goldberg 2006). Over the past thirty years, construction grammarians have identified and analyzed numerous constructions in an array of languages.17 A one-year long pilot study from 2007 to 2008, known as ‘Beyond the Core,’ permitted investigation of the feasibility of extending FN’s analytical and technical apparatus for lexical analysis to investigate and document grammatical constructions. To allow for a systematic analysis of grammatical constructions in corpus examples, FN programmers modified the structure of the FN database and revised the annotation software to allow for the localization and annotation of sentences in the FN corpora in terms of the grammatical rules that licensed the structures 15
For an overview, see the contributions in Boas (2009).
16
For an overview, see the contributions in Boas (2010a). There are also a number of domain-specific FrameNets, such as one for soccer language (Schmidt 2009), biomedical language (Dolbey 2009), and legal language (Bertoldi and de Oliveira Chishman 2011). More recently, an effort has been under way to create a FrameNet-like resource for teaching German as a foreign language at colleges in the United States; see http://coerll.utexas.edu/frames/, Boas and Dux (2013), and Boas, Dux, and Ziem (2016).
17
For an overview, see the contributions in Hoffmann and Trousdale (2013).
565
566
BOAS
found in them. The modified database and annotation software enabled FN researchers to identify, analyze, and annotate constructions in a very similar way as LUs. This is because LUs, too, are (lexical) constructions whose form pole is one or more word-forms, and whose meaning pole is usually represented as a specific semantic frame. Similarly, non-lexical constructions such as the subject-predicate construction or the genitive construction are also form-meaning pairings in which there is a clear form side of the construction. They differ, however, from lexical or partially idiomatic constructions in that the meaning evoked is either extremely vague or underspecified (cf. Baker 2012, Boas 2011c). Consider an example such as the construct Kim doesn’t like citrus fruit, let alone grapefruit. A construct is a linguistic form that instantiates one or more constructions. In this example, the construct instantiates the Let-alone construction,18 in which the phrase let alone functions as a conjunction with very specific semantic-pragmatic constraints on the pieces that it joins (Fillmore, Kay, and O’Connor 1988). The construct also instantiates other constructions, such as the non-lexical Subject-predicate and Negation constructions and the individual words (except let alone), which are lexical constructions (i.e. LUs evoking a particular semantic frame). Finding and annotating constructions follows roughly the same steps as the lexicographic workflow in FN, that is, just like lexicographers first choose a frame to analyze, construction grammarians first choose a construction and then search for corpus data that enables them to arrive at an adequate description. The first step in this process often builds on existing research in CxG, which over the past thirty years has already compiled a rich description and analysis of grammatical constructions. This step involves the formulation of a prose description of the construction, together with a definition of construct elements (CEs), parallel to that of frames and their corresponding FEs. Then, construction grammarians conduct corpus searches to extract relevant example sentences for annotation, which are then annotated with the revised FN Desktop software also used for lexical annotation. The annotation of constructions in corpus examples involves a procedure very similar to that of LUs discussed in section 34.2.2.1 above. Recall that in lexical annotation, the frame-evoking LU is already identified in a sentence. In construction annotation, annotators may find a counterpart in the socalled construction-evoking element (CEE), which is specific lexical material central for evoking the construction, such as the phrase let alone. Then, annotators identify and use the FN desktop software to mark CEs such as, in the case of the Let-alone construction, F I R S T _ C O N J U N C T (citrus fruit) and S E C O N D _ C O N J U N C T (grapefruit), which are constituent parts of a construction, similar to FEs in lexical annotation. In some cases, however, there may not be any CEE, as in abstract constructions such as 18
Following Fillmore, Goldman, and Rhomieux (2012), names of constructions are represented in italicized Courier New font.
Computational Resources: FrameNet and Constructicon
Table 34.2 Lexical and constructional description and annotation compared (Fillmore 2008: 9) Lexical FrameNet
Constructicon
Frame descriptions describe the frames and Constructicon entries describe the constructions and their components, set their components, set up FE names for up construction elements (CEs, the annotation, and specify frame-to-frame syntactic elements that make up relations; lexical entries are linked to a construct), explain the semantic frames, valence descriptions show contribution of the construction, specify combinatory possibilities, entries link construction-to-construction relations, valence patterns to sets of annotated and link construction descriptions with sentences. annotated sentences that exhibit their type. The FEs are given names according to their The CEs are named according to their role in the frame, and provide labels for the function in the constructs, they provide the phrases in the annotations that give labels on words and phrases in annotated information about the FE. sentences. The syntactic properties – grammatical Phrase types are identified for constituents functions and phrase types – are identified that serve as CEs in a construct; for for all constituents that realize frame constructions that are headed by lexical elements. units, grammatical function labels will also be relevant. Example sentences are selected that Example sentences are selected and illustrate the use of the lexical units annotated for the ways they illustrate the described. use of the construction. Annotations identify the LU, the FEs, and the Annotations contain labels for the CEs and GFs and PTs of the segments marked off. identify, for lexically marked constructions, the relevant lexical material. Valence patterns are identified, and linked to Varieties of construct patterns are identified the annotations. and linked to the annotations. Frame-to-frame relationships are Construction-to-construction relationships documented and displayed in a separate are identified and (will eventually be) resource. displayed
Subject_Predicate, Gapping, and Right_Node_Raising, which have no overt lexical material signaling the presence of a construction. In such cases, annotators only employ the CE labels to identify the different parts of the construction. Besides the identification of CEs, annotations on different layers may also include information about grammatical functions and phrase types, parallel to FN’s lexical annotation. These added annotation layers are intended to capture possible variations in the realization of a construction.19 Table 34.2 compares the similarities between lexical and constructional description in the FN lexicon and the FN constructicon. After the annotation process is complete, the construction descriptions, together with their annotated example sentences, are stored in the Constructicon, an extension of the original FN database. It currently consists of roughly seventy-five construction entries documenting different 19
For more details on the formulation and annotation of constructions, see Fillmore (2008) and Fillmore, Lee-Goldman, and Rhomieux (2012).
567
568
BOAS
Figure 34.6 First part of Way_manner construction entry
types of constructions according to the kinds of constructs they create. These include frame-bearing constructions, valence-augmenting constructions, constructions without meanings, contextually bound constructs, pumping constructions, exocentric and headless constructions, and clause-defining constructions.20 Using FN’s FrameSQL search interface, users can search for specific construction entries. Each construction entry consists of a construction description, together with definitions of the CEs, and a list of annotated example sentences together with summary tables highlighting the different ways that a construction’s CEs are realized. To illustrate, consider a sentence such as She elbowed her way into the meeting, in which the verb to elbow appears with a possessive pronoun and the noun way. Goldberg (1995) offers a detailed treatment of the so-called way-construction in terms of an argument structure construction with its own meaning that is capable of fusing with verbs that can be interpreted as denoting motion. The first part of the construction entry for the Way_manner construction is illustrated in Figure 34.6. Above the construction description we see that this construction evokes the Motion frame, and it inherits from the Way_neutral construction. This information is followed by a general prose description, including the semantics of the construction. Beneath the description we find references to publications on the Way_manner construction. Under the construction description we find the definitions of CEE(s) and CEs, as in Figure 34.7. Non-lexical constructions without meaning (or without very little clearly identifiable meaning) such as Subject_Predicate, Gapping, and Right_Node_Raising are not evoked by a CEE. Lexical constructions, (semi-)idiomatic constructions, 20
Like the frame hierarchy, constructions are also hierarchically organized, according to constructional inheritance and CE inheritance. For details, see Lee-Goldman and Petruck (in prep.).
Computational Resources: FrameNet and Constructicon
Figure 34.7 Second part of Way_manner construction entry (partial)
Figure 34.8 Third part of Way_manner construction entry: partial summary
argument structure constructions, and other meaningful constructions will list a specific CEE. In the case of the Way_manner construction, this is the noun phrase one’s way, where one’s is coindexed to the T H E M E . One special feature of the Way_manner construction is the fact that its CEs are directly linked to the FEs of the Motion frame.
569
570
BOAS
The third part of a construction entry provides a summary of how the construction’s CEs are realized syntactically. This summary is based on the annotated example sentences that accompany each construction entry.21 While the types and granularity of information displayed differs from construction to construction, they are still parallel to the valence tables found in the FN lexical entries (see Figure 34.3 above). In the case of the Way_manner construction, Figure 34.8 shows that the T H E M E is always realized as an external NP, and that the I N T R A N S I T I V E _ M A N N E R _ V E R B appears in different forms such as finite (VPfin) and the progressive form (VPing). The CEE is always a NP, while the D I R E C T I O N is realized as a dependent ADVP or PP.
34.4 Using FrameNet and the Constructicon for Research in Cognitive Linguistics The wealth of lexical data contained in FrameNet has served as the basis for a broad spectrum of research in cognitive linguistics. One major area is metaphors, which have been at the center of Conceptual Metaphor Theory since Lakoff and Johnson (1980a). Understanding the internal meaning of metaphors is quite difficult since Conceptual Metaphor Theory does not provide tools for systematically defining metaphoric concepts and their components. More recently, Croft (2009a), Sullivan (2013a), and Gemmell (2015) developed alternative proposals by employing frame-semantic data and criteria to arrive at more systematic analyses of metaphors (see also Sullivan this volume Ch. 24). FrameNet data have also been instrumental in the design and operations of a large-scale research project called MetaNet, which is concerned with creating a computational system capable of understanding metaphors in a variety of languages.22 By using FrameNet data, the MetaNet project is capable of extracting linguistic manifestations of metaphor from texts and understanding them automatically (see David this volume Ch. 35).23 As discussed in section 34.2.2.2 above, FrameNet data have also been used in other research relevant to cognitive linguistics, such as verb classification, profiling of verbal participant roles, and the licensing of argument structure constructions. The recurring theme in most research on these topics is the significance of access to detailed lexical information about a word’s semantic and syntactic properties. Time and again, researchers relying on frame-semantic information in FrameNet have found that previous accounts of these linguistic phenomena were
21
Because of space limitations, the annotated example sentences contained in the construction entry have been omitted here.
22
See www.icsi-berkeley.com/icsi/projects/ai/metanet.
23
For a novel way of applying semantic frames to the study of oral poetics, see Boas (2016).
Computational Resources: FrameNet and Constructicon
inadequate because they were too coarse-grained, missing important frame-semantic details relevant for a systematic account. On a more general level, the data contained in FrameNet (as well as FrameNets in other languages) can be regarded as a multipurpose repository of detailed frame-semantic information for research in Construction Grammar. Recall that constructions are combinations of form and meaning, and that the meaning pole of constructions can be represented by semantic frames and the LUs evoking them. The systematic integration of semantic frames into the analysis of grammatical constructions is first demonstrated by Fillmore (1988: 43), who shows that “The semantic interpretation of a sentence will be accomplished by unifying . . . semantic information from the semantic frames activated by the predicator with those introduced by the obligatory and optional companions (the complements and adjuncts) of the predicators.” In subsequent constructional research such as Fillmore and Kay (1993), Kay and Fillmore (1999), Michaelis and Ruppenhofer (2001), Boas (2003a), Iwata (2008), Croft (2009a), and Hasegawa et al. (2010), frame-semantic information of the type contained in FrameNet plays a crucial role in the representation and licensing of a variety of constructions. Frame Semantics has also played an important role for Translation Studies, as a means of analysis as well as a method for finding correct translations (Vannerem and Snell-Hornby 1986, Kussmaul 2010). In contrast to other theories of meaning, which are more logic- and truthoriented, Frame Semantics explicitly includes cultural background knowledge into its considerations. Furthermore, the notion of perspectivization is significant in Frame Semantics, a notion also crucial to translation which often involves shifts in perspective. As Ellsworth et al. (2006) note, many of the semantic translation shifts they find in the analysis of translations of the Hound of the Baskervilles (first published in 1902), include shifts in perspective. Interestingly, at least some of the mismatching frames instantiated in the original and the translation can be linked to each other by means of frame-to-frame relations, indicating at least semantic similarity, if not an exact match between original and translation. Also, the authors add that despite these shifts in perspective, readers will be able to reconstruct a similar overall scene in different languages. Frame Semantics has also been used to analyze originals and translations with respect to cultural differences; for example, with regard to differences in institutional or interpersonal relations (Rojo 2002). Construction Grammar has recently also been adopted in Translation Studies. Rojo and Valenzuela’s (2003) study of the translation of constructions goes so far as to analyze the effect of construction shifts on the translation process. Measuring reading and production times using keystroke logs and eyetracking, they find that translating a resultative construction from English into a predicative construction in Spanish takes
571
572
BOAS
significantly longer than translating an English predicative construction into a Spanish one. Another approach aims at combining frame semantic and constructional analysis in order to assess how far different constructional constraints between two languages affect not only form, but also meaning (Cˇulo 2003). Central to this approach is the function of a construction. The notion of function, though usually on a more abstract text level, is a pervasive one in Translation Studies. A construction such as the direct object in sentence initial position has the (micro-linguistic) function of highlighting an element, or in other words guiding the readers’ attention. This function can be reproduced to various degrees in a translation, or, if of minor relevance in the context, can be ignored. As stated above, there is a variety of constructions which can reproduce this focus in English, such as clefting or topicalization. In some cases, though, this leads to syntactic and lexical adaptions which may change the main verb (and thus the central frame) of a sentence. The frame shifts which are induced can, however, be analyzed and described in a similar manner as in Ellsworth et al.’s (2006) above-mentioned analysis. Often, the frames used in original and translation can be related to each other by means of frame-to-frame relations as defined in FrameNet (see Boas 2013c). More recently, two major efforts have begun to integrate framesemantic information into more formalized versions of Construction Grammar. While these efforts do not explicitly refer to FrameNet, it is obvious that the types of frame-semantic representations proposed by the various initiatives have in mind the types of information contained in FrameNet. The first is Sign-Based Construction Grammar (SBCG) (Sag 2012, Michaelis 2013), which is inspired by Berkeley Construction Grammar (Fillmore 2013) and Head-Driven Phrase Structure Grammar (Pollard and Sag 1994). Regarding grammar as an inventory of signs, SBCG emphasizes the importance of frame-semantic information by coindexing names of FEs from the meaning part of a construction with information about the construction’s argument structure.24 The second constructional approach integrating frame-semantic information is Embodied Construction Grammar (ECG) (Bergen and Chang 2013), which has a much broader scope than SBCG. ECG is interested in determining the cognitive and neural mechanisms involved in the usage of human language. By incorporating constructions into models of language use (as opposed to languages as mere descriptive objects), ECG aims to validate the status of constructions through observations of behavior in natural and experimental settings, including the computational implementation of the ECG formalism that allows it to be predictive of language use. Part of the formalism involves the architecture of constructions, in 24
For specific accounts integrating frame-semantic information into SBCG, see Hasegawa et al. (2010), Fillmore, LeeGoldman, and Rhomieux (2012), Kay and Sag (2012), and Sag (2012).
Computational Resources: FrameNet and Constructicon
which the meaning block (using frame-type semantic roles) is bound together with the form block, thereby allowing for the possibility of binding different schemas to each other (see, e.g., Bergen and Chang 2005, Chang 2008).25 The FrameNet Constructicon also represents a significant resource for research in cognitive linguistics. It presents, for the first time, a corpusbased inventory of different types of constructions whose descriptions are based on one of the basic principles of CxG, namely the idea that a difference in form signals a difference in meaning. This principle has been crucial for the identification and description of individual constructions currently contained in the Constructicon. While the current inventory of roughly seventy-five constructions in the Constructicon is only a small number, it nevertheless confirms that the methodology underlying the creation of the Constructicon is systematic and yields clearly identifiable results that can be systematically organized in a hierarchy of constructions (similar to the hierarchy of frames). Using the same methodology from the one-year-long pilot project on the Constructicon, it should in principle be possible to identify and describe most, if not all, remaining constructions of English.26 What is more, the Constructicon provides researchers, for the first time, with a broad array of corpus data illustrating the distribution and variability of constructions. This is a major departure from traditional research in cognitive linguistics during the twentieth century, which typically relied on only very few examples to arrive at hypotheses.
25
For a different approach that heavily relies on formalisms from traditional computational linguistics, see Steels (2013) on Fluid Construction Grammar (see also www.fcg-net.org). For differences between Fluid Construction Grammar and Sign-Based Construction Grammar, see Van Trijp (2013).
26
For a similar project focusing on the compilation of constructions in Swedish, see Forsberg et al. (2014). For proposals regarding the systematic creation of a Constructicon for German, see Boas (2014).
573
35 Computational Approaches to Metaphor: The Case of MetaNet Oana A. David
35.1 Introduction Contrary to many speaker’s intuitions about their own language, metaphor is the norm rather the exception in most spoken and written discourse, from the everyday to the poetic. Signed language and gestural communication are also not immune to figurative expression, as has been found repeatedly in the flourishing field of gesture research (Cienki and Mu¨ller 2008b). And finally, due to its basis in all forms of cognition, conceptual metaphor is pervasive not only in spoken and written media, but also in visual forms of expression, such as advertisements, works of art, and in recent years, internet memes. Within cognitive linguistics, cognitive science, and communication studies, there is now a well-established acceptance that metaphor pervades all human thought and communication, and this acceptance is backed by ample empirical data gathered over several decades of interdisciplinary research (see Lakoff 2012 for an overview). Yet, precise formalization of metaphor, known to be a conceptual rather than solely linguistic phenomenon, is not yet widely shared across these fields, nor is an understanding of linguistic and extra-linguistic manifestations of metaphoric thought. Further, in light of recent research on the inseparable connection between grammar and metaphor (Sullivan 2007, Steen 2007, Sullivan 2013a), it has become clear that important generalizations can be stated about the metaphor interface with language that can prove useful not only in theoretical developments in the field, but also in designing Natural Language Processing implementations that are more responsive to this rich domain of human expression. The current chapter makes a case at the same time for a structured metaphor taxonomy that is integrated with findings from other areas of studies in language and cognition, and for a computational implementation
Computational Approaches to Metaphor: The Case of MetaNet
of such a taxonomy toward the goal of making the identification and analysis of metaphoric language more efficient and accurate. Since the sound theoretical backbone of Conceptual Metaphor Theory was established by Lakoff and Johnson (1980a, 1999), Reddy (1979), Ko¨vecses (1986), Dancygier and Sweetser (2014), and many others over several decades, we see in recent years an empirical bent in metaphor research that has brought several waves of much-needed innovation to the field. First, there is a move away from top-down analyses that rely on examples derived from introspection, whereby metaphor researchers intuit that a given metaphor should exist based on certain salient examples. Instead, we are now seeing bottom-up corpus-driven approaches to metaphor analysis (e.g. Deignan 2005, Stefanowitsch 2006b, Martin 2006), guided by the principle that metaphor discovery can only happen over large and diverse linguistic data sets, and that the theory should in turn be enhanced with newly found instances of usage. This corpus-driven line of research has led to advances in computational methods for automated and semi-automated metaphor detection. For instance, Mason (2004) uses the selectional preference of verbs as inferred from WordNet (Fellbaum 1998) to predict metaphor via a learning process that rules out normally selected-for arguments (for instance pour water would be ruled out as metaphoric, so, by a process of elimination, pour funds would be metaphoric). There is also much important work in the domain of metaphor in discourse. Toward that end, the Pragglejaz Group put forth a metaphor identification procedure that analyzes metaphors in context, using localized meanings of lexical items in running text to identify potential metaphors arising from particular collocations (Pragglejaz Group 2007). Finally, seed-based cluster models and statistical metaphor identification methods have also produced results over unrestricted texts (Shutova, Sun, and Korhonen 2010, Shutova, Teufel, and Korhonen 2012) using algorithms and statistical methods that operate over large clusters of words that are marked in accordance with their likelihood of evoking metaphoric source and target domains (for a review of this literature, see Shutova 2010). To complement these computational mechanisms for metaphor detection and analysis, in this chapter I introduce the MetaNet Metaphor Repository and Identification System developed by a team of linguists, computer scientists, and cognitive scientists as part of a multicampus collaboration starting in 2011, with much of the computational development occurring at the International Computer Science Institute in Berkeley, California. Relative to existing methods in metaphor computation, the system developed for MetaNet is unique in several respects. First, it combines a top-down theory-faithful CMT approach with a bottom-up datadriven approach, implementing these simultaneously as a central, defining feature of the system. This merger of methods rests on the project’s adoption of the classic definition of conceptual metaphor; namely, metaphor is
575
576
DAVID
a deeply engrained conceptual phenomenon, whereby target domains of many kinds are systematically reasoned about in terms of source domains that have origins in concepts formed via embodied experience (see Sullivan this volume Ch. 24). Due to the embodied nature of the formation of source domains during early development, primary metaphors are believed to be universal and language-independent (Grady 1997a). This definition of metaphor, with the strong emphasis on the embodied nature of source domains and the unidirectional mapping from source to target domains, was put forth in Lakoff and Johnson (1980a, 1999) and was expanded on in subsequent Conceptual Metaphor Theory research. The MetaNet automated metaphor identification system is meant to be reflective, to a certain extent, of how human minds process metaphoric language. That is, when reasoning about relatively more abstract notions, human beings use linguistic and extra-linguistic cues to evoke deep semantic structures that are grounded in more accessible embodied experience with the world, such as force-dynamic interactions with objects and other entities, with motion through space, with one’s own bodily experiences and sensations, and with vertical and horizontal orientation. Human beings are not believed to learn metaphors arbitrarily, nor piecemeal, but instead to acquire them by domain co-associations in primary experiences, which form the foundational primary metaphors upon which later, more complex abstract mappings are layered. This includes understandings of ideologies, economics, politics, complex emotions, interpersonal relationships, and sociocultural norms. An example is the expression tax relief, which construes the relatively more abstract domain of taxation, a financial and social activity, in terms of a concrete notion of physical relief from burdens or suffering. The above definition of metaphor restricts the empirical domain of metaphor research. Given the design of the metaphor identification system, this also restricts the metaphoric phenomena identifiable to only those that appear in an overt linguistic format that provide at least one source-domain-evoking and at least one target-domain-evoking element. An expression such as glass ceiling, for instance, although highly metaphoric, would not be detected by such a system for lack of a target domain linguistic element (the target domain here remains implicit and contextually recovered). This definition of metaphor excludes other types of figurative language, such as metonymy, analogy, and conceptual blends, as well as other metaphor phenomena not grounded in primary metaphors, such as image metaphors (e.g. hourglass figure), which do not provide this type of asymmetric, systematic conceptual mapping grounded in primary embodied experiences (Lakoff and Turner 1989). This theoretically driven distinction sets the MetaNet system apart from other metaphor computing systems, which tend to take a more allencompassing view of metaphor as any type of figurative language. Although more narrow in empirical scope, the value of the MetaNet
Computational Approaches to Metaphor: The Case of MetaNet
system lies precisely in its ability to link linguistic metaphors (LMs)1 with deeper semantic domains and mappings; consequently, an LM such as tax relief would be found to have something in common with an LM such as tax burden, via their higher-level association with the primary metaphor d i f f i c u l t i e s a r e b u r d e n s (in this case, taxation is construed as a type of financial difficulty). The MetaNet system is designed with a set of practical applications in mind in order to serve as a useful tool for metaphor analysts. Namely, a linguist interested in the metaphoric content of a particular text can use this tool in a number of tasks – to reinforce or challenge his own intuitions about what the metaphors are in a particular domain, to discover metaphors over prohibitively large corpora, where hand-annotation would be time-consuming, and to reveal potential patterns and connections among conceptual metaphors that would otherwise take a human analyst much longer to uncover. Further, the tool helps analysts perform these tasks in a way that maintains consistency in their own work, as well as encourages consistency across analyses done by multiple analysts throughout the discipline. Finally, with each application of the system, the annotated LM database is augmented, storing data representing both the breadth and the frequency of metaphors, while furnishing metadata about the corpus, genre, register, and other potentially useful information that can help shed light on patterns in actual metaphor usage. This collected data can then be used to give feedback into the metaphor repository, in turn enhancing and diversifying existing metaphoric networks, filling in any gaps that may exist in inter-metaphor relations. Two crucial components of the system are a well-designed semantic frame and metaphor ontology, and a connection of metaphoric language with grammatical constructions. The former feature of the design is in large part modeled on the configuration inherent in lexicographic semantic frame databases such as FrameNet (Ruppenhofer et al. 2016, Petruck 2013), which have enabled semantic frame role labeling and frame-based sentence annotation techniques to be applied automatically to large data sets. These tools are increasingly prevalent in the fields of both cognitive linguistics and NLP, being extensively applied to English as well as other languages. In addition to the popular English FrameNet resource, other examples include the FrameNet databases designed for Swedish (Borin et al. 2009), German (Boas 2009), Brazilian Portuguese (Saloma˜o et al. 2013), and Japanese (Ohara et al. 2004). MetaNet uses much of the data structure in FrameNet, such as having numerous distinct yet interconnected frames that interact with grammatical patterns in a constructional database. 1
Here, linguistic metaphor is defined simplistically as any linguistic manifestation of a conceptual metaphor, and not as a separate type of metaphor parallel to conceptual metaphor. For instance, tax relief, tax burden, and heavy taxes are three LMs that evoke the same conceptual metaphor, t a x a t i o n i s a b u r d e n with one entailment, r e l i e v i n g taxation is relieving a burden.
577
578
DAVID
35.2 Human Cognition Meets Metaphor Computation To observe the utility of a tool like MetaNet in action, and to demonstrate the dual implementation of a semantic and a grammatical component to the metaphor identification system, let us imagine we are faced with a common real-world text. For instance, consider the following excerpt from President Barak Obama’s final State of the Union Address (January 12, 2016):2 All these trends have squeezed workers, even when they have jobs; even when the economy is growing. It’s made it harder for a hardworking family to pull itself out of poverty, harder for young people to start on their careers, and tougher for workers to retire when they want to.
A human reader of this paragraph effortlessly parses the linguistic metaphors, interpreting them against the background of the existing cultural models she is accustomed to and in large part takes for granted. For instance, the reader can parse squeezed workers metaphorically, realizing that no humans are actually, physically squeezed. Instead, there is an inference, based on the metaphor f r e e d o m o f a c t i o n i s f r e e d o m o f m o t i o n , that worker’s ability to perform a job is institutionally and systematically encumbered by limiting policies, lack of job opportunities, etc. There is a further inference that the ‘squeezer’ (i.e. via the metaphor i n s t i t u t i o n s a r e p e o p l e ) is an intentionally acting agent, purposely encumbering workers’ actions, rather than this being an involuntary encumbrance. The fact that the verb is ‘squeeze’ also indicates that these limitations are also harmful to health and to life via s o c i a l h a r m i s p h y s i c a l h a r m . The human reader proceeds equally unconsciously to parse the many other metaphoric expressions in the text, such as pull itself out of poverty, which evokes the metaphors n e g a t i v e s t a t e s a r e l o w l o c a t i o n s and i m p r o v i n g a s t a t e i s mo v i n g u p w a r d . From this we infer that the agency of making this situation-improving state-change lies with the working families themselves, rather than with the governing institutions. Tasked with the dissection of a short paragraph such as this, a trained metaphor analyst can thus produce a synthesis of the main primary metaphors, metaphoric entailments, and necessary cultural models needed to understand how metaphor is structuring the political message expressed therein. However, what if the paragraph is only one small piece of a much larger text the analyst is interested in? What if the analyst would like to compare this State of the Union Address with previous ones by the same president, or better yet, has a corpus of all recorded State of the Union addresses and would like to do a comparative study of metaphoric discourse across presidents with respect to how not only poverty but also war, health care, and gun control are discussed over many years? 2
www.whitehouse.gov/the-press-office/2016/01/12/remarks-president-barack-obama-%E2%80%93-prepareddelivery-state-union-address.
Computational Approaches to Metaphor: The Case of MetaNet
For these types of data-intensive queries, an automated metaphor identification system can prove a boost to efficacy. At this juncture, several statements must be made about how all of this metaphoric processing happens in the human mind. First, the metaphors are not in the words themselves; rather, the words evoke a more complex network of conceptual structures, which can manifest linguistically in a variety of ways. Above, squeeze workers could very well have been expressed in a syntactic variant, such as led to the squeezing of workers or as a lexicosemantic variant like wrung workers or strangled workers. Regardless of these syntactic and semantic possibilities, the same metaphoric mappings are operating unconsciously in all of these variants on the expression of the metaphor. The mind knows to parse these expressions not on an individual one-by-one basis, but against the backdrop of shared inferences. Second, there are elements of this metaphoric network that are either conceptually universal, or high-frequency, or both, and can in fact be predicted to appear cross-linguistically. For instance, we are not surprised that poverty (as a type of state) is construed as an obstacle to motion (as pull out of poverty suggests3) and also as a low location (such as an often-found variants raise out of poverty and climb out of poverty), and we would never expect to find an expression in any language in which poverty is construed as a high location. Cross-linguistically, it has been observed that negative states, statuses, conditions, etc. are construed low on a vertical scale, and positive ones as high (Ko¨vecses 2005). Similar cross-linguistic similarities have been observed for any number of so-called primary metaphors, of which n e g a t i v e s t a t e s a r e l o w l o c a t i o n s is just one. These predictions about high-level primary metaphors can be leveraged to train an automated metaphor detection system to assign potential metaphoric expressions to correct candidate conceptual metaphors underlying those expressions. The fact that metaphor does not reside in the individual words or word collocations is good news for computational modeling and corpus identification of metaphoric expressions, because it saves us from specifying lengthy, dictionary-like lexical ontologies in which each word or collocation corresponds to one metaphor, or vice versa. A system that requires an input of every possible target domain-source domain word combination in order to identify these collocations in texts would not be economical. Instead, we only want to design a relatively small conceptual network in which the semantic frames that feed the source and target domains of metaphors are enriched with specific lexical information, and only as the latter emerges from the evidence in the data. Subsequently, when those 3
Although pull out of poverty could be argued to be evoking a scene involving someone getting help in overcoming an obstacle on a horizontal, and not necessarily a vertical path, the corpus data shows that poverty is most often construed as a low location – e.g. pull out of poverty, raise out of poverty, leap/climb out of poverty. On the other hand, few examples arise in which poverty is unambiguously an obstacle on a horizontal path (e.g. the wall of poverty, get over the hurdle of poverty, etc.). This suggests that pull is evoking an aid to vertical motion from a lower to a higher location.
579
580
DAVID
metaphor
metaphor
metaphor
metaphor
metaphor
is target frame in
is source frame in
semantic frame
lexical item 1
lexical item 2
semantic frame
...
lexical item 1
lexical item 2
...
collocation in usage, e.g., “tax burden”
Figure 35.1 Schematic conceptual network
lexical items are found in running text, they are used as pointers to some neighborhood in an existing, fully mapped semantic space. Figure 35.1 schematically illustrates the relationship among lexical items, such as tax and burden, semantic frames, such as Taxation and Physical burden, and metaphors, such as d i f f i c u l t i e s a r e b u r d e n s , and the more specific subcase, t a x a t i o n i s a b u r d e n . This is a simplified depiction of how a computational metaphor processing systems can approximate human metaphor inferencing patterns. Upon hearing a metaphoric collocation, in the mind of a human interpreter an expression such as tax burden would set off a cascade of inferences, leading the interpreter to not only infer the immediate metaphor that t a x a t i o n i s a b u r d e n , but also, by virtue of the metaphor network that metaphor is embedded in, realize connections to other metaphors. Namely, t a x a t i o n i s a b u r d e n , as a subcase of d i f f i c u l t i e s a r e b u r d e n s , is not unrelated to higher-level metaphors such as a c t i o n i s m o t i o n , and i m p e d i m e n t s t o a c t i o n a r e i m p e d i m e n t s t o m o t i o n a l o n g a p a t h , and the inferences that e a s e o f a c t i o n i s r e l i e f f r o m b u r d e n s . A network-based modeling of metaphor relations, such as that implemented in the MetaNet system, helps computational automated metaphor identification to better reflect how humans organize these concepts at a higher level, and how they deduce both similarity and difference among seemingly idiosyncratic forms of linguistic expressions. In the following sections of this chapter, I detail the basic conceptually grounded components needed for designing a computational
Computational Approaches to Metaphor: The Case of MetaNet
system capable of automatically identifying potential metaphoric language in any type of text, with examples of how such components follow from hypothesized human cognition patterns. Further, the system not only identifies the metaphoric language, but also provides candidate conceptual metaphors evoked by the linguistic expressions, and, by virtue of this suggestion, classifies the linguistic expression in a class of similar, previously detected linguistic expressions. This type of system has the potential for discovering patterns within and across languages.
35.3 The MetaNet Architecture The computational system proposed is designed to perform two functions, and is thus divided into two discrete yet interdependent parts. First, the metaphor repository is organized as a structured set of ontologies, including the source and target domain frames, their frame elements, frame-toframe and metaphor-to-metaphor relations, and the lexical units evoking particular frames and metaphors. Frames, lexical items, and metaphors represent primitives in the repository, while frame and metaphor families are built up from these primitives. An important notion is that of cascade – a package of pre-defined hierarchical and ‘makes use of’ relations among frames or metaphors (Stickles et al. 2016), specified only once at a very high, schematic level (David, Lakoff, and Stickles 2016). Subsequently, any novel or even creative linguistic occurrence of a conceptual metaphor does not require analysis from scratch, but instead makes use of frequently evoked, entrenched structures. The implementation of cascades relies crucially on well-defined and consistent frame-to-frame and metaphor-tometaphor relations. Much of this hierarchical organization is conceptually similar to semantic hierarchies and inheritance structures proposed for other network theory-based cognitive linguistic approaches, such as Word Grammar (Hudson 2007, 2010), and is also implemented to an extent in frame-to-frame relations in FrameNet (Ruppenhofer et al. 2016, Petruck 2013). All of these entities and their interrelations are stored in the repository, alongside additional data such as illustrative example sentences and usage information. The MetaNet repository is configured to house most of the metaphor networks encapsulating at least two hundred primary metaphors and metaphoric entailments, but also additional general and specific metaphors, resulting in a total of over eight hundred. This includes metaphors for time, events, actions, the mind, the self, morality, social interaction, emotion, and perceptual and sensorimotor representations (e.g. s i m i l a r i t y i s c l o s e n e s s ). It also currently includes several specialized domains of metaphoric analysis pertaining to non-primary but highly prevalent general and complex metaphors. Most notably, the repository
581
582
DAVID
Data visualization, Statistical analysis
Annotated database of LMs
Metaphor and frame repository
Ontology visualization, Custom ontology, External resources (WordNet, FrameNet)
Metaphor Identification
Text resources (e.g. English GigaWord, BNC)
Figure 35.2 MetaNet integration of three systems – repository, identification, and annotated data
includes those general metaphors pertaining to common social, political, and economic abstract concepts such as democracy, poverty, governance, taxation, and myriad culturally defined social issues. The repository is embedded in a larger computing pipeline, which includes several components external to but interacting with the custom-made metaphor identification system, as shown in Figure 35.2. Raw texts are first processed using standard parsers, and the repository is used in the detection of metaphoric collocations. Then, any metaphoric collocations that are found are submitted to a growing database of LMs, where the data supporting existing metaphors are stored. If, for instance, the system is constrained by source domain to identify not only taxation as a burden but any other type of economic practice construed as a burden, the system will also identify sentences containing phrases such as heavy fees, rent burden, and payment relief. These would be identified by virtue of the network structure that the frames in which the LUs fees, rent, and payment share with the frame in which tax is located. This system has already been implemented and tested in discovering metaphors for poverty (Dodge, Hong, and Stickles 2015, Dodge 2016), gun control discourse (David, Lakoff, and Stickles 2016), and drug abuse (Stickles, David, and Sweetser 2014).
35.3.1 Component 1: The Repository The repository is a body of interconnected ontologies defining the taxonomic entities needed in conceptual metaphor analysis. These ontologies make it possible for the metaphor identification mechanism to use grammatical construction-matching patterns to identify potential LMs in unstructured texts. Figure 35.3 schematically illustrates the primitives used toward this end in the repository, showing the representation of frames and metaphors. Frames contain frame elements, and metaphors
Computational Approaches to Metaphor: The Case of MetaNet
Frame 1 frame role 1 frame role 2 frame role 3 … LU1
LU2
LU3
…
Metaphor Frame 1
Frame 2
frame role 1
frame role 1
frame role 2
frame role 2
frame role 3
frame role 3
…
…
Figure 35.3 Schematic representation of frames and metaphors
contain mappings from source domain frame elements to target domain frame elements. The representation in Figure 35.3 illustrates the important association between lexical units (LUs) and frames, and the subsequent association between frames and metaphors. LUs are assigned to frames in much the same way as they are in FrameNet (Ruppenhofer et al. 2010), namely by likelihood that the LU will evoke the frame in any given sentential context. LUs are associated with frames in a one-to-many association, regardless of any broader polysemy structure that word may participate in. For instance, the word break may evoke any number of frames, such as Separation (break off a piece of chocolate) and Nonfunctionality (the laptop is broken), and is thus listed under both frames. Information about the LU-to-frame associations used in MetaNet is gathered both from existing FrameNet frames, and from custom-made frames that are explicitly designed within the context of this system for the purposes of metaphor analysis. Frames are included in the repository often with a metaphor in mind. For instance, faced with an expression such as our relationship is broken, the Nonfunctionality source domain frame (stemming from r e l a t i o n s h i p s a r e f u n c t i o n a l o b j e c t s ) is more pertinent than would be a Separation frame, whereas the latter frame might be better suited as the source domain frame for an expression such as we must break ties with tradition. Metaphors do not receive
583
584
DAVID
is a subcase of
is a subcase of
Bodily harm
Harm
Causation
harming entity
harming entity
causal agent
harmed entity
harmed entity
affected entity
harming action
harming action
causal action
harming effect
harming effect
effect
harm
cripple strangle Figure 35.4 Relations among frames
Metaphor: SOCIAL HARM IS BODILY HARM Metaphor: POVERTY IS BODILY HARM Social harm
Poverty
Bodily harm
soc.harming entity
poverty-causing entity
harming entity
soc.harmed entity
poverty-sufferer
harmed entity
soc.harming action
social conditions leading of poverty
harming action
soc.harming effect
effects of poverty
harming effect
Figure 35.5 Relations among metaphors, and relationship of frames to metaphors
direct LU assignment, but are supplied with information about the frame elements that are mapped from the source domain frame to the target domain frame. As shown in Figure 35.4, frames are entities that have relations to one another, and this relationship is hierarchical not only in terms of how frames relate to each other, but also in terms of how the frame elements relate to each other. For instance, we can see Bodily harm as a type of Harm, which in turn is a type of Causation, where the more schematic causer, affectee, and causal process have local, specific instantiations in the more specific frames. Another important effect of this type of semantic layering is that lexical items are categorized by degrees of semantic specificity. In the example in Figure 35.4, words like cripple and strangle specifically evoke bodily harm, while verbs like harm do not refer exclusively to bodily harm. More general LUs can evoke more specific frames, but the reverse is not true. At the same time, metaphors (Figure 35.5) are also entities structured relationally to each other, as well as to the frames that populate their
Computational Approaches to Metaphor: The Case of MetaNet
source and target domains. Subcase relations among frames that feed the source and target domains of the metaphor are reflected as subcase relations among metaphors. As an example, Figure 35.5 can be posited as the metaphoric cascade for an expression such as crippling poverty. The latter captures two generalizations about the ontology. First, frames that are the source and target domains of metaphors are themselves simultaneously embedded in their own relational networks with other frames. The frame Poverty, for instance, is simultaneously the target domain of the metaphor p o v e r t y i s b o d i l y h a r m as it is a subcase of the Social harm frame. In parallel, the more specific metaphor p o v e r t y i s b o d i l y h a r m is a subcase of a more general s o c i a l h a r m i s b o d i l y h a r m (as represented by the nested boxes), which itself is potentially embedded in even more general metaphors. The subcase relationships among metaphors are in place when there are equivalent subcase relationships among frames, but not all subcase relationships among frames necessarily receive an equivalent metaphor entry.
35.3.2 Component 2: The Metaphor Extractor The metaphor identification mechanism is separate from the repository, yet crucially incorporates it as part of its functioning. While the repository is a place to store and organize information about conceptual, grammatical, and lexical structures involved in metaphor decoding, the metaphor identification mechanism uses custom scripts to implement information from the repository in seeking new instances of metaphors in the wild. Its purpose is to detect automatically potential candidates for linguistic metaphors (LMs) from large bodies of texts, such as the English Gigaword Corpus (Graff and Cieri 2003). It comprises several interdependent components, whose sequence is diagrammed in Figure 35.6. First, corpus texts are pre-processed for uniformity and compatibility with the part-of-speech and dependency tagging conventions adopted (Hong 2016). Then, texts are subjected to part-of-speech and dependency parsing, and that information is subsequently interpreted by a patented constructional pattern-matching system. The system consists of several predetermined schematic grammatical constructions covering many, but not all, possible metaphoric collocations in which at least one constituent evokes the source domain and one evokes the target domain of a metaphor.4 These constructional patterns are rough generalizations over actual grammatical constructions, and they are designed so that the metaphor identification system is able to assign source and target domain frames appropriately, depending on where the lexical unit appears in the sentence. Frames in the 4
Some examples of the grammatical patterns include: verb-object (cure poverty), noun-noun (poverty trap), and nounof-noun (jaws of poverty), among others. See Dodge, Hong and Stickles (2015) and David, Lakoff, and Stickles (2016) for a detailed discussion and exhaustive list of construction-matching patterns currently implemented in the system, as well as the details of the semantic libraries and ontologies used.
585
586
DAVID
real-world text
POS tagger/ Dep Parser
Cxn matching metaphoricity scoring
repository data
LU to frame matching conceptual metaphor
Figure 35.6 The metaphor identification pipeline
“pillars of democracy” POS/dependency tagging
pillar.n
democracy.n cxn matching pattern: NounS-of-NounT pillar.n (T)-of-democracy.n (S)
ABSTRACT CONCEPTS ARE PHYSICAL STRUCTURES
THEORIES ARE PHYSICAL STRUCTURES (e.g., build a theory)
IDEOLOGIES ARE PHYSICAL STRUCTURES target
Ideology
source
Physical Structure
is a subcase of Democracy
LUs: democracy, democratic, democrat
LUs: pillar, building, structure, framework
Figure 35.7 Sample metaphor identification run identifies pillars of democracy as metaphoric
repository are not themselves tagged as being candidates for source or target domain slots in any given metaphor; it is thus the job of the constructionmatching pattern to make this assignment by virtue of having slots designated for target and for source domain frames. The frames are subsequently recognized via the lexical items populating the constructional slots, which in turn are associated with frames in the repository frame entries. Figure 35.7 shows an illustration of the pipeline for one particular linguistic metaphor that could easily be found in a large corpus, particularly one focusing on political and news texts (as is the case with the English Gigaword Corpus).
Computational Approaches to Metaphor: The Case of MetaNet
In the automated metaphor identification run illustrated in Figure 35.7, the system is tuned to search the corpus based on a target domain of interest, in this case Democracy, which already contains several LUs that can potentially evoke this frame. The system can be constrained to either search for a particular target frame, a hand-picked subset of target frames, a known target frame family (e.g. all frames pertaining to Ideas/ Thinking) or by target frames in general. For instance, an analyst can have a set of blog entries that they suspect ahead of time to contain many metaphors on issues of democracy and/or freedom and/or rights. Alternatively, the analyst might not know anything about the metaphors in that set of blog entries and may want the system to crawl for all possible metaphors that the system is pre-equipped with. It can also be tuned to search by source domains, for instance, to show all metaphors for which the target domain is construed as object manipulation versus motion along a path. Finally, all constraints can be removed, and the system can be told to find all possible metaphors, fitting all available grammatical construction matching patterns. The ‘wild card’ by target or source domain search method is possible precisely due to the specification of grammatical construction matching patterns. Thanks to insights from recent writings on the interdependence between metaphoric structure and grammatical structure (Croft 1993, Sullivan 2013a, Stickles, David, and Sweetser 2014), we know to expect certain constructional slots to always be target-domain or source-domain evoking. The efficacy of the system comes in the form of its expandability from the compendium of primary metaphors (a relatively fixed set) and of more general metaphors; the latter can be a language-specific or a commonly occurring set of non-primary metaphors. However, we need not encode every possible specific metaphor into the repository in order to ensure a metaphor identification result for a specific metaphoric collocation. In the example in Figure 35.7, the metaphor d e m o c r a c y a r e p h y s i c a l s t r u c t u r e s happens not to be a metaphor in the repository, and yet the expression pillars of democracy is nevertheless recovered as a potential LM with a high metaphoricity score. This is by virtue of the subcase relation between the frame Democracy and the frame Ideology, where it is the latter frame, and not the former, that acts as the source domain of a general metaphor i d e o l o g i e s a r e p h y s i c a l s t r u c t u r e s . This more general metaphor has many more LMs than just those pertaining to democracy, such as tear down fascism, dismantle totalitarianism brick by brick, and communism crumbled at the end of the Cold War. We can then capitalize on a network that already provides us with ‘heavy traffic nodes’ in the conceptual structure, where most of the inferences are packed for a particular metaphor, while minimally adding frame subcase relations when interested in a new abstract domain for which existing general metaphors can potentially be informative.
587
588
DAVID
Upon performing a crawl, the metaphor identification system then provides all candidate LMs found in the text, each accompanied by their metaphoricity scores. Metaphoricity is scored in a gradient manner, such that an LM can be evaluated as more-or-less likely to be metaphoric, rather than metaphoric or non-metaphoric. The metaphoricity score is an index of the certainty with which the system can make the suggestion that any given LM be considered metaphoric, based on pre-specified parameters. The metaphoricity score is in part calculated based on the distance between the immediate frame (e.g. Democracy) and the higher-level frame for a candidate metaphor, measured in numbers of subcase relations between the two (Hong 2016). It is also measured in terms of the relationship between the source and target domain frames. A collocation will pick out two LUs (for illustrative purposes, A and B), each of which connects to a frame (e.g. A evokes frame C, and B evokes frame D). The collocation will receive a low metaphoricity score under two circumstances: 1) if A evokes a frame that is itself a subcase (or a subcase of a subcase) of D, and 2) if A and B both evoke D. For instance, cure cancer is ruled out as metaphoric because cancer and cure both evoke the Disease frame; however, cure poverty would receive a high metaphoricity score since the Disease source domain frame and the Poverty target domain frame do not find themselves in the same frame network (Dodge 2016). Finally, to expand on the ability of the extractor to identify more and more LMs of a similar kind as pillars of democracy, additional LUs can be added to both the source and target frames in the lowest-level metaphor in the cascade. In the case illustrated in Figure 35.7, adding brick and mortar in the Physical structure frame can ensure that the expression the brick and mortar of democracy is extracted with a high metaphoricity score on a subsequent run.
35.4 Conclusions Adapting standard NLP tools for corpus text processing to accommodate metaphor identification is an important component in CMT theory development, in empirical enrichment of the study of metaphor, and in improving computational linguistic techniques better to reflect diverse figurative forms of expression. Here, the MetaNet metaphor repository and identification system was presented as one possible solution to accurately and automatically detect metaphoric collocations. Researchers in any field working with linguistic data evaluate texts not only in terms of the percentages and distributions of linguistic collocations within a domain of interest, but also in terms of how those linguistic collocations relate to bigger conceptual structures that are informative in the understanding of human decision-making and reasoning processes. For this reason, the proposed textual analysis
Computational Approaches to Metaphor: The Case of MetaNet
architecture – consisting of both a hand-built frame and metaphor ontology and an automated metaphor identification component – can prove fruitful. Most importantly, by implementing a structured conceptual network model to both conceptual and computational analyses of metaphor, diverse linguistic metaphors can be organized and understood against the backdrop of complex cultural models and shared schemas grounded in embodied cognition.
589
36 Corpus Approaches Stefan Th. Gries
36.1 Introduction 36.1.1 General Introduction The core question at the heart of nearly all work in cognitive/usage-based linguistics is, how do characteristics of the cognitive system affect, or at least correlate with, the acquisition, representation, processing, use, and change of language? Thus, ever since Lakoff’s (1990: 40) cognitive commitment “to providing a characterization of general principles for language that accords with what is known about the mind and brain from other disciplines,” cognitive/usage-based approaches have revolved around notions such as exemplars and entrenchment; chunking and learning; association and contingency; categorization, prototypicality and schematicity, as well as cue and category validity; productivity and creativity; and analogy and similarity. Even though these notions all involve human cognition and have been addressed with quite some empirical rigor in psychology or psycholinguistics, much early cognitive linguistic research was based on introspection just as much as the generative approach against which much of it was arguing. However, in the last twenty to twenty-five years or so, there has been a greater recognition of the problems that arise when linguists do not (try to) back up their often far-reaching claims and theories with robust data. With regard to polysemy networks, for instance, Sandra and Rice (1995) has been a wake-up call in terms of how they discuss both corpuslinguistic and experimental ways (combined with statistical analyses) to put the study of polysemy networks etc. on firmer empirical grounds. Nowadays, cognitive/usage-based linguistics is characterized by a more widespread adoption of corpus data as a source of relevant linguistic data and quantitative/statistical tools as one of the central methodologies, and the field is now brimming with new corpus-based methods and statistical tools (cf. Ellis 2012 or Gries and Ellis 2015 for recent overviews). This
Corpus Approaches
chapter provides an overview of how corpus data and statistical methods are used in increasingly sophisticated ways in cognitive linguistics. In section 36.1.2, I briefly introduce some of the most central corpuslinguistic notions as well as how methods and goals of corpus-linguistic research are related to cognitive linguistics before turning to examples and applications in sections 36.2 to 36.4. Specifically and for expository reasons alone, I discuss (more) lexical examples in section 36.2 and (more) syntactic examples in section 36.3; I will then turn to selected applications of quantitative corpus linguistics in other fields in section 36.4. Section 36.5 mentions some necessary future developments.
36.1.2 Corpus linguistics: Methods and Goals Corpus linguistics is the study of data in corpora. The notion of a corpus can be considered a prototype category with the prototype being a collection of machine-readable files that contain text and/or transcribed speech that were produced in a natural communicative setting and that are supposed to be representative of a certain language, dialect, variety, etc.; often, corpus files are stored in Unicode encodings (so that all sorts of different orthographies can be appropriately represented) and come with some form of markup (e.g. information about the source of the text) as well as annotation (e.g. linguistic information such as part-of-speech tagging, lemmatization, etc. added to the text, often in the form of XML annotation). However, there are many corpora that differ from the above prototype along one or more dimensions. The most important of these dimensions are probably size (from a few narratives narrated in an underdocumented or even already extinct language to corpora of many billions of words) and degree of naturalness of the data they contain (from completely spontaneous dialog between two speakers to experimental, highly constrained situations). From one point of view, the technical aspects of corpus-linguistic analyses are quite simple and limited: nearly all kinds of corpus analysis are based on using specialized corpus software or, more usefully, programming languages, to retrieve from corpora examples of words or, more generally, constructions (in the Construction Grammar sense of the term), which are then analyzed in one of two different ways: 1. Analyses in which the construction(s) of interest is/are analyzed in a fairly or completely decontextualized manner. That is, one might consider frequency lists: how often do words x, y, z occur in (parts of) a corpus? Dispersion: how evenly are syntactic constructions a, b, c distributed in a corpus (e.g. in the paternal input to a child)? Collocation/ colligation: which words occur how often in, or with, another construction? And so on. In many such cases, there is no more detailed analysis of the (contexts of the) uses of the construction of interest – rather, the
591
592
GRIES
analysis of frequencies is based on observed frequencies or (conditional) probabilities of occurrence, or specialized measures of dispersion or co-occurrence/association. 2. Analyses in which the construction(s) of interest is/are studied by means of context, usually on the basis of concordances, that is, displays of instances of (a) construction(s) and the detailed annotation of contextual and/or (b) linguistic (phonological, morphological, syntactic, semantic) features. On the one hand, this might not only seem like quite a limited number of methods but also like a limited number of methods providing nothing but frequencies of occurrence or co-occurrence of character strings. In a sense, that is correct. On the other hand, however, the above corpus methods actually provide an immensely wide range of useful applications for cognitive and/or usage-based linguistics because of what may be the most important assumption uniting usage-based and corpus linguistics, the so-called distributional hypothesis. This assumption/working hypothesis assumes that the similarity of linguistic elements with regard to their functional characteristics – semantic, pragmatic, etc. – is reflected in their distributional similarity/characteristics. Harris (1970: 785f.) put it best: If we consider words or morphemes A and B to be more different in meaning than A and C, then we will often find that the distributions of A and B are more different than the distributions of A and C. In other words, difference of meaning correlates with difference of distribution.
This means that most of the central notions of cognitive/usage-based linguistics have close analogs in corpus linguistics or can be operationalized (more or less directly) on the basis of corpus data. Consider, for instance the following notional parallels between cognitive and corpus linguistics: just as cognitive linguists began exploring the notion that there may not be a clear-cut divide between lexis and syntax (e.g. Langacker 1987), corpus linguists independently began to do the same (e.g. Sinclair 1991). In the same vein, just as Langacker (1987: 42, 57) rejected the rule-list fallacy and proposed the notion of automatically deployable units, corpus linguists were discussing Sinclair’s Idiom Principle, according to which speakers make use of semi-preconstructed expressions. Just as cognitive linguists were adopting the notion of (argument structure) constructions (Goldberg 1995, 2006: 5), corpus linguists were considering Hunston and Francis’s (2000: 37) patterns, etc. The mutually beneficial parallels do not end with the above more theoretical correspondences. As usage-based linguistics is becoming more empirically rigorous and psycholinguistic, it finally takes more notice of psychological and psycholinguistic findings and relies more on exactly the kinds of data that corpora provide: token and type frequencies of (co-)occurrence. For instance, learning and cognition – linguistic and
Corpus Approaches
otherwise – are massively affected by frequency, recency, and context of usage (e.g. Ebbinghaus 1885, Bartlett 1932 [1967], Anderson 2000), which means that “Learners FIGURE language out: their task is, in essence, to learn the probability distribution P (interpretation|cue, context), the probability of an interpretation given a formal cue in a particular context, a mapping from form to meaning conditioned by context” (Ellis 2006b: 8). Paraphrasing Ellis immediately leads to two recognitions. First, a psychologically/cognitively informed approach to language presupposes and requires at the same time the precept that meaning is ultimately distributional in nature, and that is exactly what corpus frequencies provide proof of, including the insights that: 1. High(er) observed token frequency of occurrence (e.g. how often does a construction occur?) or conditional probabilities (e.g. how often does a construction occur given the presence of another construction?) are correlated with entrenchment (via the power law of learning), predictability, phonetic reduction, early acquisition, reaction times (in, say, lexical decision tasks, word and picture naming tasks) etc. 2. High(er) type frequencies or frequencies of co-occurrence (e.g. how many different word types occur in particular slots of constructions?) are essentially an operationalization of productivity, and thereby are correlated with generalization in acquisition, grammaticalization effects, higher predictability or lower degrees of surprisal, etc. 3. High frequencies of co-occurrence or high degrees of association are often associated with high degrees of contingency (often of form and function as in the Competition Model; see MacWhinney 1987a) and again higher predictability/lower surprisal. Second, the reverse perspective works just as well: rather than going from explaining how different kinds of corpus data are related to central notions from usage-based linguistics (as above), one can show how many (other) cognitive linguistic notions are readily operationalized in corpus terms, including the insights that: 1. Prototypicality and basic-level category status are often correlated with token frequency of occurrence, diversity of contexts, and high diachronic and or acquisitional frequency as measured in corpora. 2. Recency of usage events can be operationalized on the basis of dispersion of items in corpus data: how regular, or even-spaced, are occurrences of a particular expression, or how rare but then clumpy are they in a corpus? 3. Salience of (expressions in) usage events can be operationalized on the basis of how predictable (an expression in) a usage event is, for instance on the basis of conditional probabilities or of something derived from them.
593
594
GRIES
4. Context and form-function contingency are available from concordances (generated from, ideally, highly annotated corpora) and subsequent association measures that quantify the strength and, ideally, the direction of association of elements. 5. More generally, learning is not only based on recognizing contingencies from the masses of exemplars of usage events stored, but also driven by prediction errors: we notice, and learn from, more when our predictions/expectations (about what linguistic element comes next, what context a linguistic element is used in etc.) are incorrect than when they are confirmed, which is something that can be operationalized as (negative binary logs of) conditional probabilities of elements in corpus data, that is, surprisal, etc. In sum, while corpus linguistics ‘only’ provides distributional data, such distributional data are exactly what is relevant to the cognitive-/usagebased linguist via (i) the distributional hypothesis in general and via (ii) the close correspondences between, and often very direct operationalizations of, cognitive notions into corpus-based distributional patterns. In the following sections, I will discuss a variety of cognitive linguistic applications that involve (various combinations of) the above-mentioned kinds corpus linguistic data and their statistical exploration.
36.2 Syntax-lexis, with an Emphasis on Lexis Given its historical association with dictionary-making, corpus linguistics has always had a strong emphasis on the analysis of lexical items. Concordances and collocations have long helped lexicographers to tease apart multiple senses of polysemous words or differences in how near synonymous words are used. Especially for collocations, corpus linguists often rely on association measures to separate the wheat (frequent co-occurrence that reflects interesting semantic and/or functional characteristics) from the chaff (frequent co-occurrence that reflects little of semantic interest, such as the fact that most nouns co-occur with the a lot). Early examples of the study of co-occurrence in cognitive linguistics are Schmid (1993), Kishner and Gibbs (1996) on just, and Gibbs and Matlock (2001) on make. While these studies of collocation and colligation were groundbreaking at the time, they were still largely monofactorial in nature: Uses of (senses of) words were annotated for, and cross-tabulated with, co-occurrence patterns, but often no deeper quantitative analysis was conducted; a few selected examples will be discussed in section 36.2.1. The current state of the art, however, is that multidimensional cooccurrence data are gathered and statistically analyzed accordingly. Gries (2010b) distinguishes two different ways in which analyses can be multidimensional, which will be exemplified in sections 36.2.2 and 36.2.3.
Corpus Approaches
36.2.1 Early Univariate Studies Much early corpus work in cognitive linguistics was concerned with conceptual metaphor and metonymy and was based on concordance data resulting from (i) completely manual searches, (ii) searches for source and/or target domain expressions, and/or (iii) searches in corpora annotated for conceptual mappings; such searches may be based on individual expressions or sets of expressions (either defined a priori, on the basis of lists, or using some exploratory procedure). Stefanowitsch’s (2006b) metaphorical pattern analysis is an example: for several basic emotions, he chose a lexeme and retrieved up to 1000 matches from the British National Corpus, uncovering altogether 1443 metaphorical patterns, which are then classified in terms of the (kinds of) metaphors they instantiate and compared to a representative non-corpus-based study. The results indicate that this kind of analysis is superior to the mere introspective/opportunistic listing of metaphors, in particular in how this kind of analysis can be used to discover many metaphors that more traditional study did not find (see Stefanowitsch 2004 for a similar analysis contrasting H A P P I N E S S metaphors in English and German and Hilpert 2006a for a similar study of metonymies with eye). Finally, Gries (2006) is a corpus-based study of the many metaphorically and metonymically related senses of 815 instances of the highly polysemous verb to run in British and American English. Studies of the above kind helped pave the way to quantitatively more complex analyses exploring the distribution of linguistic items in many different dimensions.
36.2.2
Multidimensional1 Approaches: Behavioral Profiles and Cluster Analyses The first sense of multidimensional, multidimensional1, refers to the fact that concordance lines of (senses of) a word are annotated for many different characteristics – morphological, syntactic, semantic, contextual ones – and all of these dimensions are used in a statistical analysis at the same time, but separately. One example for this approach that has become more widely used is the behavioral profile (BP) approach (see Gries 2010b for an overview). Concordance lines are annotated for many features, and then the senses of polysemous words, or the near synonyms in point, are compared with regard to the percentages with which different features are attested with them. Gries (2006) applied this method to cluster senses of to run, and Divjak (2006) studied Russian verbs meaning ‘to intend,’ and both find that the percentages of co-occurrence phenomena reliably distinguish senses and near synonyms respectively. In addition, Gries (2006) also showed how co-occurrence percentages can be used to study the similarity of senses, their positions in networks, whether to lump or split them, and how more generally different types and aspects of corpus data help
595
596
GRIES
identify the prototypical senses of words (i.e. type and token frequencies, earliest historical attestations, earliest language acquisition attestations, etc.). A variety of more complex follow-up approaches to BP analyses have been pursued, too. For example, the behavioral profiles of, say, near synonyms with linguistic patterns in their contexts can be submitted to exploratory statistical tools such as hierarchical cluster analyses. Divjak and Gries (2006) is a case in point. They studied nine Russian verbs meaning ‘to try’ and analyzed the similarity of BP co-occurrence percentages with cluster analyses and some follow-up exploration and found that this lexical field falls into three different groups (of three verbs each), which reflect different idealized cognitive models of trying. Even more interesting, though, is that Divjak and Gries (2008) showed that the clusters obtained on the basis of the corpus analysis are very strongly replicated in sets of sorting and gap-filling experiments with native speakers of Russian, lending further support to the method in particular, but also to the idea of converging evidence in general. Since then BPs have been used in more specialized settings, as when Janda and Solovyev (2009) use a downsized version of BP data – the constructional profile, the relative frequency distribution of the grammatical constructions a word occurs in – to explore synonyms, but also in more general settings, as when Divjak and Gries (2009) compare the use of phasal verbs in English (begin versus start) and Russian (nacˇinat’/nacˇat’, nacˇinat’sja/nacˇat’sja, and stat’). Among other things, they find that English speakers’ choices are driven by semantic characteristics of the beginners and beginnees whereas Russian speakers’ choices are more driven by aspectual and argumentstructural characteristics.
36.2.3
Multidimensional2 Approaches: Regression and Correspondence Analysis The second sense of multidimensional, multidimensional2, refers to the fact that concordance lines of (senses of) a word are annotated for many different characteristics – morphological, syntactic, semantic, discoursepragmatic characteristics – and all of these dimensions are used in a statistical analysis together. That is, multidimensional1 uses the information of how a linguistic item – a morpheme, a word, a sense, etc. – behaves in each of many dimensions such as what are the percentages with which sense x has different kinds of subjects or what are the percentages with which sense x has different kinds of objects? For example, if one annotates n = 2 dimensions of variation – for example, the percentages of different subjects of senses a to f and the percentages of different objects of senses a to f – then multidimensional1 analysis uses that information in the shape of combining results from n = 2 two-dimensional frequency/percentage tables. However, that does not include the co-occurrence percentages of
Corpus Approaches
sense x’s different subjects with its different objects – this is what multidimensional2 does in the shape of one three-dimensional table: sense (a to f) × subject (all subject types) × object (all object types). The advantage over the BP analysis is, therefore, that higher-level co-occurrence information is included, which is more precise and cognitively more realistic (although recall the strong experimental validation of the BP approach). The disadvantage is that this can easily lead to very sparse data sets, as when many features are annotated so that many combinations of features are rare or even unattested. Two types of multidimensional2 applications are particularly interesting. First, there are exploratory approaches such as those using (multiple) correspondence analysis (MCA), a method applied to multidimensional frequency data that is similar to principal component analysis. One such application to a polysemous word is Glynn’s (2010b) study of bother. Glynn followed the work discussed in section 36.2.2 and annotated uses of bother for a large number of features and applied MCAs to different parts of the multidimensional frequency table to find different clusters and “semantically motivated distinction[s] between two sets of syntactic patterns” (Glynn 2010b: 256) – an agentive and a predicative construction. In order to test the patterns suggested by the exploratory tool, Glynn then added a second type of multidimensional2 application, confirmatory approaches based on regression analyses, showing that, just like BPs, a careful multidimensional analysis of corpus data with powerful statistical tools can reveal cognitively and constructionally interesting regularities impossible to discover by intuition or eyeballing. Similar applications in the domain of semantics include Glynn (2014a) and Desagulier (2014). Additional examples for similar multidimensional2 applications involve binary as well as multinomial or polytomous logistic regressions. As for the former, Gries and Deshors (2012) compared the uses of may and can by native speakers of English and French to see how well syntactic and semantic features allow to predict speakers’ choices, but also to determine which variables distinguish the native speaker’s from the learners’ use of may and can; the results were then interpreted against the background of processing principles. As for the latter, Arppe (2008) studied four common Finnish verbs meaning ‘to think’ by, as usual, annotating them for a variety of linguistic characteristics and then identifying the linguistic characteristics that allow best to predict speakers’ choices; later work by Divjak and Arppe (e.g. 2010) extended such regression approaches to the identification of prototypes in a way inspired by Gries (2003b), who uses linear discriminant analysis to the same end. Regardless of which multidimensional approach is chosen, the combination of comprehensive annotation and multifactorial/-variate analysis has proven to yield insightful results regarding a variety of the abovementioned central notions of cognitive linguistics on the level of lexical items, including the degree to which words/senses are entrenched, the
597
598
GRIES
association/contingency of formal and functional elements, matters of categorization (graded similarity versus discreteness of senses, prototypes of senses), and many more. For more examples regarding the corpus-based exploration of metaphor and metonymy, the reader is referred to the collection of papers in Stefanowitsch and Gries (2006) (for more examples highlighting in particular statistical applications, cf. Glynn and Fischer (2010) and Glynn and Robinson (2014).
36.3 Syntax-lexis, with an Emphasis on Syntax Somewhat unsurprisingly, the corpus linguistic tools used on the more syntactic side of the continuum are quite similar to those on the more lexical side of things. Again, concordances are used to explore the use of syntactic patterns, or constructions, in their context, and colligations/ collexemes – tables of words occurring in syntactically defined slots of constructions – are used to explore the ways in which constructional slots are filled. One major difference of course is concerned with the searchability of constructions, since corpora that are annotated for constructions in the general sense of the term do not exist. Thus, corpus searches for constructions typically rely on the use of regular expressions to retrieve (parts of) words, part of speech tags, parsed corpora, or combinations of all these things with lots of subsequent manual disambiguation. In the following sections, I will first discuss a recent development in the study of colligations/collexemes, which is a simple monofactorial topic, before I turn to corpus linguistically and quantitatively more involved topics.
36.3.1
Monofactorial Approaches: Frequencies, Percentages, and Collostructions One recent prominent approach in the study of constructions – the way they fill their slots and what that reveals about their semantics/function – is collostructional analysis (Stefanowitsch and Gries 2003, Gries and Stefanowitsch 2004a, 2004b, Stefanowitsch and Gries 2005). By analogy to collocations, Gries and Stefanowitsch proposed to study the functions of constructions by not just looking at how frequently words occur in their slots (e.g. which verbs occur in the verb slot of the way construction and how often) but by computing measures of association (most often pFisher-Yates exact test) that quantify how strongly (or weakly) a word and a construction are attracted to, or repelled by, each other. This family of methods has some psycholinguistic foundation and has been widely adopted in studies on near-synonymous constructions (alternations), priming effects (Szmrecsanyi 2006), first- and second-language acquisition and learning of constructions (see section 36.4.2), constructional change over time (Hilpert 2006b, 2008), etc. For alternations, for instance, the method
Corpus Approaches
was precise enough to discover the iconicity difference between the ditransitive (small distances between recipient and patient) and the prepositional dative (larger distances between recipient and patient; cf. Thompson and Koide 1987). In the last few years, a variety of studies have been published which also document the validity of the method experimentally. Gries et al. (2005) demonstrated how collexeme analysis outperforms frequency and conditional probabilities as predictors of subjects’ behavior in a sentence completion task, and the follow-up of Gries, Hampe, and Scho¨nefeld (2005) provided additional support from self-paced reading times (cf. also Gries 2012, 2015a, 2015b) for comprehensive discussion and rebuttals of recent critiques of the method. Lastly, collostructions have been coupled with more advanced statistical tools – such as cluster analysis or correspondence analysis – to discover subsenses of constructions (cf. Gries and Stefanowitsch 2010), or they have been refined better to incorporate senses of constructions (e.g. Perek 2014) or senses of verbs (e.g. Bernolet and Colleman, in prep.).
36.3.2
Multidimensional2 Approaches: Regression and Correspondence Analysis The previous section already mentioned the use of advanced statistical tools in the analysis of constructions; in the terminology of section 36.2, these tools are multidimensional2 and I will again discuss examples using exploratory and confirmatory approaches; for expository (and historical) reasons, I will begin with the latter. Among the very first multifactorial approaches in cognitive corpus linguistics were Gries’s (2000, 2003a) studies of the constructional alternation of particle placement, that is, the two constructions instantiated by Picard picked up the tricorder and Picard picked the tricorder up. He annotated examples of both constructions from the British National Corpus (BNC) for a large number of phonological, morphological, syntactic, semantic, and discourse-functional parameters and used a classifier – linear discriminant analysis – to identify the factors that make speakers choose one construction over another in a particular discourse context, to discuss their implications for language production, and to identify prototypical instances of both constructions. Since then, this type of approach – multifactorial modeling syntactic, but now also lexical, alternatives with regression-like methods – has become very prominent both within and outside of cognitive linguistics proper, and within cognitive linguistics, there are at least some studies that show how well this approach helps explore such alternations; Szmrecsanyi (2006) and Shank et al. (2016) are two cases in point using logistic regressions (the latter being a diachronic study) (see Levshina, Geeraerts, and Speelman 2014 for the additional tool of classification and regression trees). Gries
599
600
GRIES
(2003b) showed that the predictions of such methods correlate very strongly with results from acceptability ratings. Exploratory approaches in this domain involve the method of multiple correspondence analysis. One particularly interesting example involves the cross-linguistic corpus-based study of analytic causatives in English and Dutch. On the basis of data from the newspaper component of the BNC (approximately 10 m words) for English and an equally large sample from the Twente and the Leuven News corpora, Levshina, Geeraerts, and Speelman (2013) retrieved approximately 4000 examples of causatives in both languages, which were annotated for the semantic classes of the causer and the causee as well as for one of many different semantic verb classes. An MCA was then used to determine the conceptual space of the causatives in the two languages. Among other things, this bottom-up procedure provided a two-dimensional representation (of an ultimately three-dimensional) conceptual causative space with clear support for a previous merely theoretical typology of causative events. In addition, a follow-up analysis of the results of separate analyses of the English and the Dutch data showed that the two languages’ conceptual causative space is, overall, similar but not identical, and in the follow-up there is discussion about how both languages’ data points are located differently in causative space.
36.3.3
Straddling the Boundaries of Lexis and Syntax: Idioms and Multiword Units As mentioned above and for purely expository reasons, sections 36.2 and 36.3 in this chapter upheld a distinction that many cognitive and corpus linguists do not make anymore: the one between syntax and lexis. In fact, many of the earliest studies in Construction Grammar focused on items straddling the ‘syntax-lexis boundary,’ namely constructions that were traditionally called idioms (cf. Wulff 2008 for the probably most rigorous cognitive and corpus linguistic study of idiomaticity). At that time, and in fact until recently, it was part of the definition of the concept of construction that an expression being considered a candidate for constructionhood exhibited something that was not predictable from its constituent parts and other constructions already postulated. While, in Goldbergian Construction Grammar, this notion of unpredictability is no longer a necessary condition, there is now also a growing body of research on the psycholinguistic status of multiword units (MWUs, also often called lexical bundles), that is, expression consisting of several contiguous words. On the one hand, MWUs do not seem good candidates for constructionhood since they are often not even ‘proper’ phrasal elements, do not have a particularly unified semantic/functional pole, and have little that is unpredictable about them; but, on the other hand, many of them, at some point, become retained in speakers’ minds
Corpus Approaches
and, thus, most likely also give rise to processes of chunking (cf. Bybee 2010: Ch. 3, 8). Many such studies are experimental in nature but usually take their starting point from corpus frequencies of MWUs. For instance, Bod (2000) showed that high-frequency 3-grams (e.g. I like it) are reacted to faster than lower-frequency 3-grams (e.g. I keep it), and Lemke, Tremblay, and Tucker (2009) provided evidence from lab-induced speech that the last word of a 4-gram is more predictable than expected by chance, which they interpreted as showing that MWUs are stored as lexical units; similar findings are reported by Huang et al. (2012) based on the comparison of transitional probabilities in corpus data and eye-tracking data (cf., for more discussion, Snider and Arnon 2012 or Caldwell-Harris, Berant, and Edelman 2012). Again, the analysis of many of the central notions of the cognitive/usagebased approach to language benefits in multiple ways from the combination of fine-grained annotation of corpus data and powerful statistical tools, which elucidate complex patterns and interactions in the data that defy introspective or simple monofactorial analysis: notions such as chunking and entrenchment of words into MWUs, association and contingency of words in constructional slots (which are based on the validity of cues and constructional categories), the implications of this for learnability and processing, and so on, all these are areas where state-of-the-art quantitative corpus linguistics can be very useful (for more examples, cf. Stefanowitsch 2010 and the papers in Gries and Stefanowitsch 2006, Rice and Newman 2010, Scho¨nefeld 2011, Divjak and Gries 2012, and Gries and Divjak 2012).
36.4 Other Linguistic Subdisciplines 36.4.1 Structural Subdisciplines For purely technological reasons, corpus linguistics has been particularly involved in studies on lexis and syntax. However, given increasingly more and larger resources as well as the ongoing development of new techniques and tools, there is now also a considerabe body of corpus-based cognitive linguistic research in domains other than ‘pure’ syntax or lexis. While space does not permit an exhaustive discussion, the following highlights some examples. Some of the more influential recent studies on phonological reduction were not cognitive linguistic in a narrower sense, but certainly compatible with current cognitive linguistic work on processing. As one example, Bell et al. (2003) is a comprehensive study using regression analyses of how the pronunciation of monosyllabic function words (in the Switchboard corpus) is affected by disfluencies, contextual predictability (measured in terms of transitional probabilities; earlier studies used the association measure MI), and utterance position.
601
602
GRIES
To mention one more recent example, Raymond and Brown (2012) used binary logistic regression to study initial-fricative reduction in Spanish. Their study is remarkable for the range of variables they take into consideration to shed light on why many studies of frequency effects come to contradictory results. Maybe the most important conclusion is that, once contextual probabilities are taken into account, non-contextual frequencies did not yield any robust results, a finding strongly supporting the view that simple frequencies of occurrence are often not enough. Another area in which corpus-based studies have had a lot to offer to cognitive linguistics is morphology. There is a large number of studies by Bybee and colleagues (nicely summarized in Bybee 2010) that revealed how frequency of (co-)occurrence affects chunking or resistance to morphosyntactic change, to name but some examples, and that have been integrated into a usage-based network model of morphology. A different though ultimately related strand of research is work on morphological productivity, specifically on how to measure it best and how relative frequency – the difference in frequency of derived words (e.g. inaccurate) and their bases (e.g. accurate) – affects productivity as well as morphological processing, which in turn informs theoretical discussions of decompositional versus non-decompositional approaches (cf. Hay and Baayen 2003 or Antic´ 2012 for a more recent contribution). The following summarizes a few other studies that involve, or are defined by, morphological elements. Berez and Gries (2010) explored the factors that trigger the absence/presence of the middle marker d in iterative verbs on the basis of a small corpus of Dena’ina narratives. Traditionally, d was considered a reflex of syntactic transitivity, with semantics playing a less important role. However, a binary logistic regression and a hierarchical configural frequency analysis of their data showed that, while transitivity is a relevant predictor, the semantic type of iterativity (and its position on a scale from concrete to abstract) resulted in an even higher degree of predictive power. Teddiman (2012) showed how subjects’ decision which part of speech to assign to ambiguous words in an experiment are very strongly correlated (rS = 0.87) with the words’ preferences in the CELEX database. For instance and on the whole, words such as pipe and drive (mostly used nominally and verbally respectively) were typically assigned to be nouns and verbs respectively. Just as there are phenomena somewhere between, or in both, lexis and syntax, so there are phenomena somewhere between, or in both, phonology and morphology. An example of the former is Bergen (2004) on phonaesthemes. While the main point of his study involved a priming experiment, one section of it showed how some phonaesthemes such as gl-, sn-, and sm- are significantly more often attested with their phonaesthemic meanings of ‘light’ and ‘nose/mouth’ than expected by chance, which
Corpus Approaches
raises interesting issues for classical morphological theory, into which phonaesthemes do not fit very well, and statistical learning of speakers. An example of the latter, a phenomenon ‘in’ both phonology and morphology, is blends, formations such as motel (motor × hotel) or brunch (breakfast × lunch). In a series of studies, Gries showed how coiners of such blends have to strike a balance between different and often conflicting facets of phonological similarity and semantics while at the same time preserving the recognizability of the two source words entering into the blend (where recognizability was operationalized in a corpus-based way). Again, this corpus-informed work sheds light on a phenomenon that traditional morphology finds difficult to cope with. Finally, Sokolova et al. (2012) as well as Backus and Mos (2011) connect morphology and syntax. Using their variant of BPs, constructional profiles, the former explore nearly 2000 examples of the Russian locative alternation with грузить and three of its prefixed forms from the Modern subcorpus of the Russian National Corpus. They model the constructional choice using three predictors – prefix, number of participants, finite/ participle form of the verb – in a logistic regression and find, among other things, significant differences between the four different verbs, which is particularly interesting since the “three perfectives are traditionally considered to bear semantically ‘empty’ prefixes” (2011: 67); thus, the corpus-based approach goes against received wisdom and shows that the meanings of the verbs and the constructions interact. The latter study is concerned with the productivity and similarity of two Dutch potentiality constructions – a derivational morpheme (-baar) and a copula construction (SUBJ COPfinite te INF) and is a nice example of how corpus data are used complementarily with other kinds of data, here acceptability judgments. They report the results of a distinctive collexeme analysis to determine which verbs prefer which of the two constructions in the Corpus of Spoken Dutch and follow this result up with a judgment experiment to probe more deeply into seemingly productive uses of the constructions. They found converging evidence such that acceptability is often correlated with corpus frequencies and preferences (see the chapters in Scho¨nefeld 2011 for more examples of converging evidence).
36.4.2 Other Subdisciplines Apart from the structurally motivated subdisciplines of phonology, morphology, lexis, and syntax, corpus-based work in cognitive/usage-based linguistics has also had a particular impact on the following three areas, each of which is influenced particularly by one central figure/research group and which will be discussed very briefly in what follows: first language acquisition, second/foreign language acquisition, and cognitive sociolinguistics.
603
604
GRIES
In L1 acquisition research, the work done by Tomasello, Lieven, and colleagues has been among the most influential corpus-based work in cognitive linguistics (see Tomasello 2010 for a fairly recent overview). One currently ‘hot’ area is concerned with how children learn what not to say, that is, how they learn to avoid overgeneralizations – by negative entrenchment or statistical preemption (see Stefanowitsch 2011, Goldberg 2011, Ambridge et al. 2012, Robenalt and Goldberg 2015, among others). In addition, this field is also slowly embracing more computational methods, such as Da˛browska and Lieven (2005) on the development of early syntax using the traceback method, or larger data bases, such as Behrens (2006), who explores parts-of-speech information as well as 300 K NPs and 200 K VPs with regard to how distributions in children’s data over time come to approximate the (stable) distributions in adult data. In L2-acquisition/foreign-language learning research, the most influential work is by Nick Ellis and colleagues. One early influential study is Ellis and Ferreira-Junior (2009), who retrieve all instances of three argument structure constructions from the ESL corpus of the European Science Foundation project to compute type frequencies, type-token distributions, and collexeme strengths (of verbs and constructions) to test for Zipfian distributions and identify first-learned and path-breaking verbs. Gries and Wulff (2005, 2009) are studies of alternations (the dative alternation and to/-ing complementation) that combine corpus data with experimental results; Ellis, O’Donnell, and Ro¨mer (2014b) correlate results from a gapfilling experiment with German, Czech, and Spanish learners of English to the frequencies of verbs in the same constructions (entrenchment), the associations of these verbs to the constructions (contingency), and their semantic prototypicality using multiple regression. They find that each factor makes its own significant contribution to the frequency with which learners provide verbs for constructions. Last but not least, there is now some interest in cognitive sociolinguistics, mostly stimulated by work done by Geeraerts and colleagues. Studies such as Glynn (2014b) or Levshina, Geeraerts, and Speelman (2014) argue in favor of adding predictors covering dialectal, geographic, thematic variability to their statistical analyses. While the results reported in such studies indicate that including these dimensions in statistical modeling increases the overall amount of explained variability in the data, it seems to me as if it still needs to be shown to what degree such findings also inform the cognitive aspects of the phenomena thus studied; Pu¨tz, Robinson, and Reif (2014) is a recent interesting collection of work representative of this approach.
36.5 Concluding Remarks and Future Developments As the previous sections have demonstrated, corpus linguistic methods and subsequent statistical analysis have become very important for
Corpus Approaches
cognitive and exemplar-/usage-based linguistics. The type of exemplarbased approaches that many cognitive linguists now embrace are particularly compatible with the distributional data that corpora provide, and cognitive and corpus linguists have independently arrived at many shared notions and perspectives. In this final section, I would like very briefly to provide some comments on where I think cognitive linguistics can and should evolve and mature further by incorporating insights from quantitative corpus linguistics.
36.5.1 More and Better Corpus Linguistic Methods One important area for future research is concerned with refining the arsenal of corpus linguistic tools. First, there is a growing recognition of the relevance of association measures in cognitive/usage-based linguistics. However, with very few exceptions, such association measures are bidirectional or symmetric: they quantify the attraction of x and y to each other as opposed to the attraction of x to y, or of y to x, which would often be psychologically/psycholinguistically more realistic. Gries (2013b), following Ellis (2006a) and Ellis and Ferreira-Junior (2009), discussed and validated a directional association measure from the associative learning literature on the basis of corpus data, which should be interesting for anybody dealing with association and contingency, say in language learning/acquisition. Similarly, the entropies of the frequencies of linguistic elements are an important element qualifying the effect of type frequencies in corpus data (cf. Gries 2013b, 2015a), which in turn affects productivity and flexibility/creativity of expressions (cf. Zeschel 2012 and Zeldes 2012) as well as their learnability. Second, there is now also a growing recognition that corpus frequencies of x and y can be highly misleading if the dispersion of x and y in the corpus in question is not also considered: if x and y are equally frequent in a corpus but x occurs in every corpus file whereas y occurs only in a very small section of the corpus, then y’s frequency should perhaps be downgraded, and Gries (2008, 2010a) discussed ways to measure this as well as first results that indicate that, sometimes, dispersion is a better predictor of experimental results than frequency. The field should also consider further individual differences, which will require that researchers take seriously how corpora represent speakers’ individual contributions to the data. Studies such as Street and Da˛browska (2010) or Caldwell-Harris, Berant, and Edelman (2012) and others show clearly that the ‘native speaker,’ about which all linguistic theories like to generalize, is merely a convenient fiction, given the huge individual diversity that both corpus and experimental data reveal very clearly. Finally, there will be, and should be, an increase of corpus-based studies that involve at least some validation of experimental data, as in many of the studies from above.
605
606
GRIES
36.5.2 More and Better Statistical Tools Another area that is much in flux involves the development of statistical tools. One approach that is gaining ground rapidly is the technique of new regression-like methods. On the one hand, the technique of mixed-effects (or multilevel) modeling is becoming more frequent, since it allows the analyst to handle subject/speaker-specific (see above) and, for example, word-specific variation, as well as unbalanced data, much better than traditional regression tools; in addition, other methods such as multimodel inferencing and random forests promise to address potential shortcoming of often very noisy and collinear datasets (see Kuperman and Bresnan 2012 or Gries 2015b for the former, as well as Matsuki, Kuperman, and Van Dyke 2016 and Deshors and Gries 2016 for the latter). On the other hand, new classification tools such as Bayesian network and memory-based learning (cf. Theijssen et al. 2013), with its capacity for the modeling of causal effects, in a way reminiscent of structural equation modeling, and naı¨ve discriminative learning (cf. Baayen 2010), with its higher degree of cognitive realism, are becoming important promising new alternatives; in addition, the modeling of nonlinear relations between predictors and responses by means of, say, generalized additive models, is slowly becoming more frequent in linguistics (see, e.g., Wieling, Nerbonne, and Baayen 2011). Hopefully it will also catch on in cognitive linguistics. Finally, I hope that exploratory/ bottom-up techniques will become more frequently used. While cluster and correspondence analyses are already in more widespread use, methods such as network analysis (see Ellis et al. 2014a for a very interesting application of graph-based algorithms to verbs in slots of constructions) or longitudinal connectionist or exemplar-based simulations hold much promise for cognitively more realistic statistical evaluations; for an interesting example of a longitudinal corpus study, on German was. . . fu¨r (‘what kind of. . .’ questions, see Steinkrauss 2011). While this chapter could only provide the briefest of overviews of the impact that corpora and quantitative methods have had on cognitive linguistics, it is probably fair to say that such methods are taking the field by storm. It is to be hoped that this development and maturation of the field continues as individual scholars increase their repertoire of corpus and quantitative skills and their engagement with experimental data, and as more and more fruitful connections with neighboring disciplines – e.g. corpus linguistics or psycholinguistics – provide opportunities for interdisciplinary research.
37 Cognitive Linguistics and the Study of Textual Meaning Barbara Dancygier
Texts are linguistic artifacts. They use linguistic resources in rich and complex ways, creating a range of meanings: stories, accounts of experience, reflections on events and ideas, descriptions of travel, etc. While the specific content, the way the text is situated in time, or the writer’s life are interesting objects of study, they are not naturally defined as linguistic questions. At the same time, it is inherently important to ask how language participates in the emergence of textual meaning. In this chapter, I discuss some examples and strategies, to show how cognitive linguists can approach texts, and how that kind of work supports the cognitive linguistic enterprise. The linguistic study of the form and meaning of texts has a long tradition, and occupies a space overlapping with many disciplines: rhetoric, stylistics, text linguistics, discourse analysis, narratology, etc. The disciplines which share some goals and methods with cognitive linguistics are cognitive stylistics, cognitive poetics, and cognitive narratology (for some examples of representative work, see Bex, Burke, and Stockwell 2000, Semino and Culpeper 2002, Stockwell 2002, Gavins and Steen 2003, Herman 2003, Broˆne and Vandaele 2009, Sanford and Emmott 2012, Herman 2013). With their different goals and tools, these approaches to the study of texts offer a complex picture of potential research questions. They also rely on cognitive linguistics tools in various ways and with varying depth, but some works, especially the studies in Broˆne and Vandaele (2009), openly argue for cognitive linguistic engagement with textual study, especially the study of literature. There are thus various approaches which show that cognitive linguistics work is applicable to textual study, but cognitive linguistics also needs to be clear on the role of textual analysis as an inherent part of the discipline’s daily practice. Cognitive linguistics, with its overt interest in how linguistic choices are correlated with meaning, can contribute to textual analysis in important
608
DANCYGIER
ways; by the same token, cognitive linguists can learn from looking at texts. In what follows, I analyze three groups of textual examples, literary and non-literary ones, showing in each case how cognitive linguistics tools support textual interpretation, and also how textual analysis contributes to a better understanding of the mechanisms of meaning construction. Cognitive text analysis, a term I propose here, is a two-way street, with clear benefits for both the interpretive and the theoretical goals. Early work on Conceptual Metaphor Theory (see Sullivan this volume Ch. 24) was soon extended to approach texts – especially literary texts. Following in the footsteps of Lakoff and Johnson’s 1980 book (1980a), Lakoff and Turner proposed a range of applications of cognitive linguistic concepts to the study of literature (see Lakoff and Turner 1989, Turner 1989, 1991, 1996). The work further prompted the emergence of a discipline of cognitive poetics, which is now practiced by different scholars, with a range of different goals (consider Tsur 1992, Freeman 2002, Stockwell 2002, Broˆne and Vandaele 2009, Freeman 2009). Some prominent metaphor theorists added much to the discussion, also clarifying the link between cognitive poetics and linguistics (Sweetser 2004, Sullivan and Sweetser 2009, Sweetser and Sullivan 2012). In this chapter, I discuss various texts, including literary texts, showing how cognitive linguistics can engage with what I refer to as cognitive text analysis. In section 37.1, I consider the effects of applying mappings as analytical tools; in section 37.2, I look at the issues raised by story-construction needs in narrative discourse; and in section 37.3, I show how literary texts build on the basic aspects of interaction. Section 37.4 adds the perspective of embodiment and experiential grounding, while section 37.5 reviews the findings, and outlines some future research directions.
37.1 Mappings and Figuration In this section, I look at several journalistic texts. The first two rely in different ways on the frame (or domain) of Illness (problems affecting the functioning of the body) to metaphorically represent problems affecting areas of social life, such as the economy and the political system. The third text talks about an actual illness, using other metaphors to represent the impact on the sick person. The comparison shows interesting aspects of the role of figurative forms, such as metaphor and simile. In an NPR text published on August 27, 2011, entitled Policymakers Seek Prescription For Ailing Economy, Scott Horsley discusses the then economic situation in terms of Illness: (1)
Think of the U.S. economy as a patient who is bedridden after a long illness. At an annual policy conference this weekend in Jackson Hole, Wyo., what they’re discussing is not just about how to get the patient up and hobbling around, but how to get him running wind sprints . . . Fed Chairman Ben
Cognitive Linguistics and the Study of Textual Meaning Bernanke says there’s no reason to think the economy won’t get up and running again, even if it’s taking longer than anyone would like . . . Stephen Cecchetti, chief economist at the Bank for International Settlements, warns too much government debt is like a weight on that economic runner . . . Fever-driven politics may be one of the biggest factors keeping the economy from getting out of bed.
The excerpt builds a rather rich image based on the common conceptual metaphor (see Sullivan this volume Ch. 24) C O M P L E X S Y S T E M S A R E B O D I E S . It starts with the construction Think of X as Y, which sets up a figurative correlation between two frames. The economy is thus talked about as a person who is bedridden (the primary metaphor B A D / S I C K I S D O W N ) and needs to be up and running, out of bed (G O O D / H E A L T H Y I S U P ). Additionally, the Illness frame is combined here with the Sports domain – where running wind sprints represents not just being healthy, but also very fit. There is a natural embodied link between good health and the level of fitness, but the competitive nature of running sprints expresses more specific expectations about how well the economy should be doing in achieving goals, through the frame of Speed, and Ability worth a trophy. The formal and interpretive features of the excerpt are further complicated by the use of constructions. The first construction is a simile, describing debt being like a weight on that economic runner. In earlier work, simile has been seen as a more explicit and focused metaphor (see, e.g. Gentner and Bowdle 2001, 2008). Recent work (Dancygier and Sweetser 2014), however, shows that the differences between metaphor and simile are quite significant. First of all, simile is a constructional form, using a copula construction and an expression of comparison, with well-defined slots: X is like Y, or X is as Y as Z (for further discussion of constructional subtypes of simile, see Moder 2008, 2010). Metaphor only looks similar to simile in the standard formulation of mappings, such as I N T I M A C Y I S C L O S E N E S S or S P O R T S C O M P E T I T I O N I S W A R , while its typical use has no constructional features of any kind, as in She is my closest friend or the Canucks fought bravely in the playoffs, but lost – these examples represent the mappings, but are best described as metaphorical expressions, not mappings themselves. The expression like a weight in (1) is thus a simile, even though it also relies on the primary metaphor D I F F I C U L T I E S A R E B U R D E N S ; additionally, it builds on the metaphoric correlation set up in the preceding sentences to use a denominal adjective (economic) to further construe economy in sports terms (runner). Such modification constructions have been discussed as blends (see Oakley and Pascual this volume Ch. 26), which combine two inputs (Running and Economy) into one compact phrase. In effect, the short (and rather clear) excerpt relies heavily on mappings (complex metaphors, primary metaphors) and constructions (simile, Think of X as Y, denominal adjectival modification). One interesting feature of such use, then, is the possibility of extending and combining mappings, while also treating the correlation between three frames (Illness, Sports, and
609
610
DANCYGIER
Economy) as a flexible figurative resource which can be expressed through different forms. In this example, as in many others, what appears to be the primary source of meaning is a broad figurative pattern, realized through various specific forms of figurative language. The functioning of figuration as an underlying pattern is best observed in extended discourse – spoken or written – and should be accounted for (see Dancygier and Sweetser 2014, and Dancygier 2017a for a discussion of the issue). The text quoted in (1) is short (even without the ellipsis). The figuration in (1) feels compact, and has a clear effect on the message. In another text, How American Politics Went Insane, by Jonathan Rauch (The Atlantic, July/ August 2016), the use of the Illness frame is different. The text is very long, offering a complex historical and political argument, and is only occasionally punctuated with summary phrases describing the state of American politics as sick and neurotic. Overall, Rauch talks about a condition he calls the chaos syndrome. (2)
What we are seeing is not a temporary spasm of chaos but a chaos syndrome. . . . Chaos becomes the new normal – both in campaigns and in the government itself.
The author evokes the frame of Illness (syndrome) to then use another construction based on a blend of frames. Specifically, X is the new Y is a construction which profiles a previously not considered item X as now occupying the same position that Y used to occupy – so in the frame of behaviors or situations, the item previously considered exceptional (chaos) begins to count as the most common representative of the frame (normal) (see Dancygier 2011). Later in the text, the writer also uses a simile: (3)
. . . Americans have been busy demonizing and disempowering political professionals and parties, which is like spending decades abusing and attacking your own immune system. Eventually, you will get sick.
Finally, the text also uses the Think of X as Y construction, to set up a figurative correlation between two frames: (4)
Think of them (the political establishment) as analogous to antibodies and white blood cells, establishing and patrolling the barriers between the body politic and would-be hijackers on the outside.
Example (4) also uses the word barrier, evoking an image schema which is typically further elaborated to suggest unwanted influences from the outside and protection against them. Further, the writer talks about viruses attacking the political system to then sum up with the text quoted in (5), where another example of adjectival modification constructions suggesting a blend of Illness and Political System is used (political head cold): (5)
Being a disorder of the immune system, chaos syndrome magnifies other problems, turning political head colds into pneumonia.
Cognitive Linguistics and the Study of Textual Meaning
These two journalistic texts suggest several observations. In a text, a figurative pattern, such as the correlation between Illness and a Complex System, may be set up and elaborated through a range of forms: conceptual metaphor, primary metaphor, image schemas, simile, and a range of constructions specializing in blending frames in new ways (Think of X as Y, X is the new Y, denominal adjectival modification). In other words, specific types of mappings are textually less salient than the overall figurative pattern. A cognitive linguistic analysis has the power not only to identify the specific forms, such as metaphor or simile, but also to look at a broader understanding of a range of related forms. It can also include a close look at constructional resources at the service of figuration (something a more traditional analysis would not offer). At the same time, it may be the case that a very similar frame (Illness) can serve different meaning-construction purposes, either focusing on the expected consequences of a quick and full recovery, or on defining a situation through an illness to better understand the negative sides of it. Both examples thus show that textual analysis can help linguists see the meaning potential of frames, beyond the standard usage, because texts like the ones above use the frames to prompt an understanding of an abstract situation through a specific type of embodied experience. Saying that the economy should improve is a very different statement from the one expressing the expectation that it should be running sprints. Similarly, simply describing a political situation may not evoke much response, but construing it as a dangerous illness has emotional and affective consequences. I want to look at one more example of a text – this time actually describing a serious illness. What we saw above might suggest that Illness itself is a rich frame, not requiring further figurative elaboration. But this is not the case. In the text in Vanity Fair (September 2010) in which Christopher Hitchens first announced that he had been diagnosed with cancer, he used a range of figurative expressions. I will discuss only one here. (6)
It took strenuous effort for me to cross the room of my New York hotel and summon the emergency services . . . now that I view the scene in retrospect I see it as a very gentle and firm deportation, taking me from the country of the well across the stark frontier that marks off the land of malady.
Hitchens then goes on to describe the new land – noting that the locals are pleasant and smiling, though their sense of humor is not too good, and the cuisine is unattractive. While he does describe some symptoms, he is mainly talking about the sense of separation and forceful removal from his ordinary life routines. Importantly, his metaphor of gentle deportation relies on the schema of a barrier, but very differently from what we saw in example (4). There, the image was of unwanted agents crossing into the body politic from outside; here, it is the image of someone experiencing the abrupt and unwanted change of environment. The first text talks about
611
612
DANCYGIER
crossing barriers as a source of illness, Hitchens talks about illness itself as forcing one to cross barriers. Same image-schematic structure, but different viewpoints (being on one side of a barrier versus being forced to cross), and so the expressions yield different meanings. Overall, cognitive linguistics offers a very useful set of tools, which can help us understand the differences of construal and the resulting differences in reception, but the ways the tools are to be used are important too. Each of the texts mentioned could of course be discussed in ways similar to what metaphor analysis, image schematic analysis, and focus on constructions yield, but doing it without cross-text comparison would be less specific and effective. A look at various textual uses of the same frame, Illness, shows important facts about how such structures are used figuratively, beyond a rather obvious recognition of conceptual metaphor. A full and informative analysis in these cases requires that analysts recognize the multiple layers of figuration which jointly yield the meaning potential of a text based on the Illness frame. One needs a good understanding of constructions and their figurative potential, a clear view of the meaningful role of schemas, and an understanding of the viewpoint effects of such figurative construals – the experience of trying to emerge from an illness, of wanting to protect oneself from an illness, and of feeling the alienating consequences of an illness are all different experiential construals. Crucially, they are all based on frame structure, re-construal of frames through metaphor, image-schemas, and constructions (simile, X is the new Y, denominal adjectival modification, Think of X as Y, etc.). Cognitive linguistics plays a central role in making such analyses possible.
37.2 Story Construction In Dancygier’s cognitive linguistic study of narrative form and meaning (2012a), she argues that interpretation of longer texts, such as novels, can fruitfully be viewed from the perspective of low-level lexical and grammatical choices. In this section, I draw on that work to show how cognitive linguistic tools help explain some important aspects of narrative interpretation. I focus on the consequences of pronominal choices and textual representation of emotions, though it should be clear that the analytical potential is not restricted to these issues. Examples (7) and (8) come from Ali Smith’s novel, The Accidental (2005). In (7), a character, Michael Smart, an English professor, browses through books. He is depressed, because he has been suspended from his job as a result of repeated sexual relations with his students: (7)
The book had fallen open at the symptoms of hypothermia. He couldn’t believe how many of the symptoms he had. He definitely felt cold and tired, he felt this all the time . . . Yes, he had physical and mental lethargy and had been unable to answer questions or directions. That was true.
Cognitive Linguistics and the Study of Textual Meaning That was what he felt like, inside, all the time . . . Yes, he was very shaky, even right now . . . Definitely, he was pale. He couldn’t believe how pale he was when he got up in the mornings. He had extreme ashen pallor. Not just him; it was as if everything around him in the world had extreme ashen pallor too. Michael stood up. He was shaking. It was exposure. (pp. 265–66)
The narration here relies on one of the hotly discussed forms of speech and thought representation – Free Indirect Discourse. Michael’s thoughts, such as Definitely, I am pale, are represented through the distanced perspective of the narrator, and so in Past Tense and third person, but the flow of discourse is the flow of Michael’s reflections, and there are telltale evaluative adverbs, like definitely, maintaining the character’s perspective. This form has often been discussed, by narratologists and stylisticians, as a cross between the narrating voice and the character voice. Cognitive linguists have been interested in these narrative forms too (e.g. Sanders and Redeker 1996, Vandelanotte 2009). While such constructions of speech and thought representation are characterized by a rather orderly set of formal features (tense, pronouns), there are other issues. First of all, the text does not only represent what Michael thinks – it also represents what he reads. The terms used to describe symptoms of hypothermia (such as extreme ashen pallor, italicized in the original) come from the book on mountaineering Michael reads, and what we are seeing is Michael’s somewhat ridiculous recognition (the repeated Yes) of the embodied symptoms of a severe condition, potentially leading to death, as his own symptoms. The discomfort he feels signals his mental state, possibly guilt and regret, even if he feels it as embodied symptoms. When he comes to the final diagnosis – It was exposure – we realize that he is indeed suffering as a result of having been exposed as an immoral man, and the double meaning of the word expose allows the writer to frame it in this way. The example is thus more complex than a standard example of Free Indirect Discourse. It is based on the additional textual layer of the book Michael reads; it gives us his thoughts – mainly the silly recognition of his state as hypothermia, but also a good description of his embodied responses to shame – and it allows us also to recognize the humor of the situation, and Michael’s blind inability to admit his guilt. Perhaps most importantly, we are told that Michael feels ashamed, even though he does not admit being ashamed even to himself – instead, he sees himself as sick. All this is a necessary part of the reading, prompted, among other things, by the use of Free Indirect Discourse, but also by wordplay, reference to the text read, and the focus on embodied symptoms rather than psychological trauma. In its different way, the excerpt also relies on the Illness frame, using the concept of physical symptoms of illness as a vehicle to represent the character’s feelings – another instance of a broad use of figuration. The complexity of the meaning of
613
614
DANCYGIER
the excerpt cannot thus be reduced to the constructional nature of Free Indirect Discourse. What is also needed is a view of psychological states from an embodied perspective, and the understanding of the complex configuration of viewpoints: from the technical content of the book being read, to Michael’s sense of not feeling well, to the narrator’s way of telling the reader that Michael is still in denial (some cognitive linguistic basics of narrative viewpoint have been discussed in Dancygier 2005 and developed in Dancygier 2012a). The Accidental is structured through the point of view of several characters, as each of them in turn is the vehicle chosen by the narrating voice through a substantial part of the book. But all the events narrated revolve around another character, Amber, who influences everyone’s lives, but never becomes the onstage character. The narrative does not let us see the events of the story from her perspective. This is what Ali Smith often does – she creates a narrative completely driven by one character’s actions, while not giving us access to that character’s thoughts (at least, she did that also in There but for the and Hotel World). There are, however, brief sections written in the first person, representing Amber, even though they do not contribute to our understanding of the events. Example (8) shows a brief fragment. (8)
I sold flowers in Covent Garden. A posh geezer taught me how to speak proper and took me to the races, designed by Cecil Beaton, though they dubbed my voice in the end because the singing wasn’t good enough. (2005: 104)
The excerpt is embedded in a three-page section written in the same style, representing films, popular culture concepts, songs, and well-known events of the 1960s. It is written in the first person (as are all the brief and rare ‘Amber’ parts of the book). The text refers to the well-known musical and 1964 film My Fair Lady. The character, flower-girl Eliza Doolittle, does indeed learn how to speak proper (an expression that she might have used, in her original Cockney dialect, as in standard English speak properly or speak proper English would be chosen), by Professor Higgins (again, posh geezer is Eliza-speak). In the film (costumes were designed by Cecil Beaton) she is played by Audrey Hepburn (my voice is Audrey’s voice then), but the songs are performed by Julie Andrews. The movie frame is often the source of metonymic patterns whereby the names of actors and characters are interchangeable, and still refer to the film world (as in Throughout the movie, Robert Redford is alone on his boat). So the conflation of Eliza and Audrey is not unusual. But it is unusual for them to be indistinguishable and sharing an identity with a book character, from a book published in 2005. Also, actor/role metonymy typically remains within the film narrative, while here we also see the name of the costume designer, who is outside of the movie story. So the I in (8) places the character, Amber, in three different deictic set-ups (the book’s story, the
Cognitive Linguistics and the Study of Textual Meaning
film story, and the making of the film story). The I pronoun thus participates in a complex deictic blend. So what does the excerpt do? It starts with more-or-less ordinary frame metonymy (as described by Mental Space Theory analysts; see Fauconnier and Sweetser 1996), and uses it to blend the three story spaces/deictic centers to give the reader an idea of the 1960s background as characteristic of an enigmatic character we otherwise know nothing about. The first person is necessary to preserve the convention whereby other characters’ thoughts and identities are relayed through the narratorial voice (as in [7]), while this character can only speak for herself, not through another viewpoint, as nobody knows anything about her, and what she says or thinks does not matter to the story. The directness and self-sufficiency of the first person is what is being relied on here. Also, the contrast between the use of a deictic pronoun I and the use of anaphoric he, assuming a deictic viewpoint of another person, is exploited in an original way, to represent the position of the characters in the story. Several points can be made, then. Fictional narratives use grammatical resources in ways which are subject to some narrative conventions, but are also text-specific. On the one hand, it can typically be said that narrative fiction creates a set-up wherein there is a narrating voice (not representing anyone in particular, other than a layer of narrative structure containing the overall narrative viewpoint). There are narrative choices, such as the pronoun choice, which signal a certain deictic and viewpoint structure. On the other hand, these standard narrative choices can be further complicated by reliance on discourse and lexical choices (in the examples above we see description of symptoms, the double meaning of exposure, the use of dialectal forms, and the evocation of frames through strategic selection of proper names, such as Cecil Beaton), by reliance on metaphor (double meaning of the words to expose/exposure) and metonymy (actor/ role), and, last but not least, by evocation of embodied responses as tokens of psychological reactions. All these observations can be clarified in terms of conceptual patterns, but looking at them one-at-a-time is not sufficient, as narrative fiction relies on a range of conventional and non-conventional linguistic choices, combined in original and often innovative ways. We need a better understanding of how various phenomena described by cognitive linguistic tools can work together to create an original effect.
37.3 Interaction, Intersubjectivity, and Multimodality As was pointed out above, narrative fiction assumes an interesting communicative set-up. In the absence of an actual communicative setting, narratives function, in most cases, as if there were an actual ‘teller’ of the story, narrating it to the reader. The concept of a narrator emerged as a way to profile the subjectivity who is responsible for how the story is
615
616
DANCYGIER
told, from a deictic perspective different from the one that the reader inhabits. Even in the cases of first-person narrative, where the narrator is also a character who describes his or her experience, the narrator is often not confined to the here-and-now of the events being told. However, there is a text-type, a literary genre, which relies fully on the pretense of a shared deictic center and a conversation-like exchange – drama. Drama, like prose fiction, is a narrative form, but it is performed by actors representing characters, on the stage, within the duration of the play. All that is communicated is communicated on the stage, by a character speaking (most of the time) to another character. Occasionally, a character may address the audience, but it is not common. Essentially, the deictic center of the play is different from that of the situation in which the audience members find themselves – a play may be taking place in the past, and fictional characters speak to each other, while the audience is in fact not expected to say anything. However, all that happens on the stage has been prepared with the audience in mind – what they need to see and hear and what they need to know to understand the story. It is thus an intersubjective set-up, where speakers structure their discourse in the way which benefits other characters in the story and audience members at the same time. Discourse on the stage, then, is supposed to feel like a natural conversation, but it is not a private exchange – there are ‘overhearers’ (the audience), who are in fact the primary addressees. At the same time, the theater is a multimodal genre. Actors’ bodies participate in the exchanges on the stage in ways which parallel spontaneous communication – they direct their gaze to facilitate communication, use gesture, orient their bodies, express meaning through body posture, facial expression and tone of voice, etc. The actor’s craft, though, does not postulate a completely accurate imitation of natural interaction – for example, gestures are typically broader and more iconic, and articulation is somewhat overdone. To sum up, drama is an intersubjective and multimodal genre, and its text has to take account of the fact. It is a narrative genre which lacks a narrating voice (except in rare interventions of the chorus), so all the needs of the narrative have to be satisfied differently. There are many such needs, but I focus here on just one example of how the viewers learn about characters’ thoughts and feelings. From the linguistic perspective, one of the most interesting ways to represent emotions on the Early Modern stage is to have characters speak to objects and bodies – instead of other characters. This takes the form of a construction, which was described in earlier work by Dancygier (2012a) as the ‘Vocative-cum-Imperative’ construction (I will refer to it here as the VCI Construction). While often used on the stage of the period, it has not totally disappeared from contemporary English. When a speaker expresses impatience by making a faux address to God as in God, give me patience!, the same construction is used, suggesting an emotional state of losing one’s patience and inability to change that, along with
Cognitive Linguistics and the Study of Textual Meaning
the realization that patience is required. The meaning of the construction relies on both the vocative and the imperative component: evoking an absent addressee who can act in a way that satisfies the speaker’s needs suggests that the speaker herself does not have the ability to perform the act required, and, at the same time, suggesting that the act should be performed. In Shakespearean drama, the construction is used to represent emotion and experience. The choice of addressees falls into two categories: materially present addressees (body parts, bodies of dead characters, material objects), or imaginary addressees (powers not visible to viewers and present in the speaker’s mind). Many examples of the first kind can be found in the tomb scene of Romeo and Juliet. As Romeo prepares for his own death, he is parting with his wife. He says: Eyes, look your last! / Arms, take your last embrace! and lips, O you / The doors of breath, seal with a righteous kiss / A dateless bargain to engrossing death! The body parts addressed here are associated with behaviors that the imperative forms profile – looking one last time, embracing, and kissing, which are common actions in parting. They represent what Romeo is doing as he speaks, giving the intended meanings to his actions – the intersubjective role of such usage is very salient here, as all that is said benefits the audience, the never-addressed addressee. It is important to realize here that the ordinary viewer’s perspective on the scene does not necessarily allow them to appreciate the fact that Romeo is lovingly looking at his wife – the words clarify how Romeo feels about what happens in the scene. Later, Romeo speaks to the poison vial he will use (Come, bitter conduct) and then Juliet speaks to the dagger which she will kill herself with – O happy dagger! / This is thy sheath; there rust and let me die. The vocatives evoke instruments (not agents) of the acts to be performed – but in fact represent the speaker’s intended actions. Thus the roles of the speaker and the addressee in the basic onstage set-up of the theater are blended with the agent and instrument roles in the acts to be performed – so the dagger is an instrument, but it is presented as responsible for the act of killing. A more complex example of the use of the VCI construction is the final monologue in Marlowe’s Doctor Faustus. Faustus is alone, in the final hour of his life (until some Devils enter at the end), knowing that what awaits him is hell and damnation. He uses the VCI construction numerous times: (9)
a. Stand still, you ever-moving spheres of heaven, That time may cease, and midnight never come; b. Fair Nature’s eye, rise, rise again, and make Perpetual day; c. Mountains and hills, come, come, and fall on me, And hide me from the heavy wrath of God! d. Earth, gape! e. You stars that reign’d at my nativity,. . . Now draw up Faustus, like a foggy mist,
617
618
DANCYGIER
f. Now, body, turn to air, Or Lucifer will bear thee quick to hell! [Thunder and lightning.] g. O soul, be chang’d into little water-drops, And fall into the ocean, ne’er be found! h. [Enter Devils.] My God, my God, look not so fierce on me! Adders and serpents, let me breathe a while! Ugly hell, gape not! come not, Lucifer!
In all the a–h, VCI lines Faustus tells the audience what he feels by evoking actions which would fulfill his desires (so time would stand still, Faustus would disappear, be hidden, would not be taken to hell, etc.). The speaker is represented as a patient or experiencer of all these acts, and the evoked agents are not typical intentional humans – because what is being profiled is what the actions would do for Faustus. Embodiment is a crucial part of the scene, as viewers are prompted to imagine Faustus’s body disappearing, being hidden, being dragged to hell by merciless devils – something that would be very difficult to stage in a convincing manner, but that is powerfully represented through the use of the VCI construction. These examples profile well-defined types of addressees – non-human, not conscious, and/or absent. Also, while these are in a sense fictive commands (similar to Call me crazy discussed by Pascual and Oakley this volume Ch. 21), they are also situated in the conventions of a literary genre, and they play a well-defined role of representing emotions and desires. Most importantly, they can accompany actions (parting, killing oneself), using material instruments as addressees of the vocative form to communicate the embodied experience of a character which cannot be shown on the stage. A cognitively informed explanation of the phenomenon, then, focuses on the fact that these forms of discourse rely on the pattern of spoken interaction, with a speaker, a hearer, a time, and a place; however, they are used to represent inner thoughts, intentions, or desires. In literature, thoughts and feelings have to become part of what is being communicated, and they have to be relayed through the forms of the genre. In fact, I argue, literary discourse has gradually developed forms of representation of experience and consciousness, which are specific to genres. In prose, the forms such as Free Indirect Discourse started playing that role; in drama, the VCI construction represents an early attempt to render the inner voice, by relying on available linguistic resources.
37.4 Embodiment and Experience In the examples above, I presented language use which relies on mechanisms of meaning construction such as figuration, narrative constructions, and even genre-specific constructions. However, it is important to note as well that in most cases the meaning construction patterns chosen serve
Cognitive Linguistics and the Study of Textual Meaning
rather specific needs – giving a reader an understanding of the situation which appeals to their embodied sense of what a certain kind of experience means. The texts discussed achieve their meaning through representations of embodied or emotionally rich states such as being sick, being forcefully displaced, suffering from exposure, being subjected to pain, preparing to die, or wanting to stop the passage of time. They have not been selected to represent these specific experiences (and in the case of [8] the experience described is culturally salient, rather than embodied), but they do raise an important question regarding the role of representation of experience in the discourse types we looked at: fiction, but also personal and argumentative non-fiction. In this section, I look at some more specific examples, where the linguistic choices are directly targeting representation of emotion and experience. In To the Lighthouse (1927 [2004]), Woolf describes a conversational exchange involving Mr. Tansley – a young man, not feeling quite secure in the social circle he has been asked to join. The discussion is about potential hardships of the boat trip to the lighthouse, and Mr. Tansley is asked if he is a good sailor: (10)
Mr. Tansley raised a hammer: swung it high in air; but realizing, as it descended, that he could not smite that butterfly with such an instrument as this, said only that he had never been sick in his life, but in that one sentence lay compact, like gunpowder, that his grandfather was a fisherman, his father a chemist; that he worked his way up entirely himself; that he was proud of it; that he was Charles Tansley – a fact that nobody there seemed to realize; but one of these days every single person would know it. He scowled ahead of him. He could almost pity these mild cultivated people, who would be blown sky high, like bales of wood and barrels of apples, one of these days by the gunpowder that was in him. (1927 [2004]: 108)
Mr. Tansley’s response described in (10) – raising a hammer and swinging it in the air – builds on embodied and metaphorical concepts. Words, whether spoken or intended, are represented as material presences (hammer, butterfly, gunpowder). The heavy hammer stands for a weighty response (I M P O R T A N C E I S W E I G H T ), and swinging it in the air is the state of emotional agitation Tansley gets himself into in order to show his interlocutors that he is a serious man, not to be fooled with (I N T E N S I T Y O F F E E L I N G I S T H E I N T E N S I T Y O F A C T I O N ); he is ready to dismiss once and for all any suggestion of lack of stamina – until he realizes that the remark was benign (that butterfly; again I M P O R T A N C E I S W E I G H T ). The emotional build-up his inner response requires is correlated with the momentum of the imaginary hammer – in other words, the fictive experience of the show of bodily strength stands for the actual experience of emotional agitation. Of course, the humor of this exaggerated response is part of the interpretation. But then the paragraph goes on to represent the hidden meaning of the actual response Mr. Tansley gives – it is compared to gunpowder, hidden,
619
620
DANCYGIER
inconspicuous, but explosive (metaphors of anger often rely on the image of emotional pressure being violently released; see Ko¨vecses 1986, 2000 for an exhaustive discussion of emotion metaphors), and the next image is of all the innocent participants in the conversation blown up, paying for their lack of regard for Mr. Tansley’s inner power. This is the second violent image in the short fragment, used to represent not actions, but emotions – Mr. Tansley’s violent dislike of the people around him. Importantly, the overall image of hammer-swinging and then the exploding gunpowder constitute a rich representation of how Mr. Tansley feels – though the narrative does not use overt means of signaling that, such as he felt as if X. The directness with which his experience is represented in embodied terms is characteristic of some forms of literary style. Importantly, the effect does not just rely on the presence of metaphorical expressions – what literary style requires is a rich coherent image, rather than individual expressions. The role of such images calls for a sustained cognitive linguistic investigation. One of the most central aspects of embodied experience is sensory perception. Perception is also important in the study of literary texts, especially for the understanding of the concept of imagery. Here again, however, a deeper cognitive understanding of embodiment adds texture to the understanding of the role senses play in literature. Philip Larkin, a British poet known for his clear, sober style, often thought, and wrote, about death. Here, I will compare his depictions of death and sensory perception in two poems: ‘Going’ and ‘Aubade.’ In the former, Larkin describes a gradual shutting-down of the senses: dissolving vision (an evening . . . one never seen before, that lights no lamps), loss of touch (silken . . . (but) it brings no comfort . . . What is under my hands, that I cannot feel?), loss of bodily movement and control of motion (What loads my hands down?), loss of sense of spatial organization (Where has the tree gone that locked Earth to the sky?). The text is full of negative images (never seen, cannot feel, no lamps, no comfort) and uses wh-questions as ways of signaling loss of embodied sense and perception. Importantly, the questions are directed at what the poem’s persona experiences as the questions are asked in the moment. In many people’s reading, the poem is evocative of death, but one should perhaps be more specific here – the poem allows the reader to simulate the experience of the process of dying, through giving a vivid image of embodied experience of dissolving sensory perception. The poem relies essentially on primary metaphors, which is not surprising given its focus on representation of experience. Almost every line evokes a (primary) metaphor: the metaphoric title (‘Going,’ as in D E A T H I S D E P A R T U R E ); the described arrival of the lightless evening evokes L I F E T I M E I S A D A Y , D E A T H I S D A R K N E S S , and K N O W I N G I S S E E I N G ; C H A N G E I S M O T I O N underlies the image of the disappearing tree (gone), etc. These metaphorical forms support the experiential meaning and the role of perception in a consistent manner. Thus the overall effect of the poem lies in the
Cognitive Linguistics and the Study of Textual Meaning
sustained metaphorical evocation of experience and repeated use of negation and questions, signaling the loss of that experience. Clearly, various linguistic forms cooperate to create the full effect of the poem. Not all texts represent perception in this way, of course. In another poem, ‘Aubade,’ Larkin talks about his fear of death in much more sober, descriptive terms: this is what we fear – no sight, no sound, / No touch or taste or smell, nothing to think with, / Nothing to love or link with, / The anaesthetic from which none come round. The ambiguous rendering of dissolving sensory perception in ‘Going’ is given an almost technical, accurate description in ‘Aubade.’ Negation is used again, but there is a difference between an impersonal phrase no touch and an experientially rich sentence such as I cannot feel (what I’m touching). The nouns denoting senses (sight, sound, touch, taste, smell) create a viewpoint of someone not actually experiencing any of these deprivations – ‘Going’ is at the opposite end of the spectrum in this respect. Literary discourse can thus be soberly descriptive even when tragic embodied events are at stake, but it is at its most powerful when it attempts to render the raw experience of an emotionally charged state. I have described the contrast elsewhere (Dancygier 2014) as the contrast between two stylistic techniques: experience and description, also pointing out that only the former properly fits the literary term ‘imagery.’
37.5 Conclusions There are several broader points which emerge from the above discussion: 1. Textual analyses require that all of our cognitive linguistic tools be used – whether constructional or lexical. Also, focusing on individual expressions is not sufficient, as such linguistic choices often participate in broader textual meaning. A clear understanding of the role word- and sentence-level expressions play at the textual level is what cognitive linguistics should work toward. 2. For the purposes of understanding complex linguistic artifacts, cognitive linguists might complement the study of very specific forms, such as metaphor or simile, with broader discussion of strategies such as figuration. What looking at texts suggests is that a single frame (such as Illness) can yield various figurative and constructional forms, and many different mappings. The figurative potential of frames and the ways in which they are re-construed in conceptual metaphor or in constructions such as Think of X as Y or X is the new Y should be a subject of sustained investigation. 3. There may be no obvious linguistic distinction between the tools needed for a textual analysis of a non-fiction text and a literary text. But there are several important differences: literary texts use genre-conventions (such as a dialogue form in drama or Free Indirect Discourse in prose), and
621
622
DANCYGIER
many aspects of the meaning of a literary text depend crucially on what the conventions contribute. Also, literary discourse and non-fiction discourse may rely on similar meaning-construction means, such as metaphor, but the syntactic complexity and the richness of the experiential images is often significantly increased in literature. 4. Embodied and experiential meanings underlie much of extended discourse. These components are necessary for readers to be able to relate to the meanings emotionally. They take different forms, but they are central to textual meaning. 5. Emotional impact of texts is often achieved through some broadly construed strategies. The distinction between experience and description is one such discourse dimension, but the excerpts quoted above also suggest other potentially important strategies. On the basis of these examples we can talk about a talk-strategy and a do-strategy – each using very different linguistic resources. For example, the emotional effects of Larkin’s ‘Going,’ Smith’s exposure example, or Faustus’s soliloquy, are achieved primarily through forms which resemble spoken discourse (the talk-strategy) – though they are in fact intended to represent inner feelings, experiences, and thoughts. For comparison, Woolf ’s portrayal of Mr. Tansley’s response is through what he might do; similarly, Hitchens describes his experience of illness in terms of what other people do, which adds to the sense of helplessness he conveys, while the NPR excerpt and The Atlantic article depend on what the economy should do or what pathogens do to affect the political system. There are certainly other angles further analyses can reveal, but the above are some inherently worthwhile directions of study, aiming at uncovering general linguistic strategies used to express experiential meanings. Cognitive text analysis is important. It can offer inspiration to linguists, as it points out meaning phenomena which require clarification. It can also be useful to other scholars interested in texts – stylisticians, discourse analysts, narratologists, or literary scholars, as this detailed methodological and theoretical approach to analysis can help others focus on the role language plays in textual meaning-construction.
Part VI
Concepts and Approaches: Space and Time
38 Linguistic Patterns of Space and Time Vocabulary Eve Sweetser and Alice Gaby 38.1 Introduction Linguists have long observed that the dominant metaphors for time, in the vast majority of languages, are spatial ones. Unfortunately, it still remains the case that we do not have deep, solid semantic analyses of any domain, or any metaphoric structure, in nearly as many languages as those for which we have reliable data on word order, or consonant inventories. Even a good grammar is not usually sufficient to give us all we need to discuss a language’s system of spatiotemporal metaphors. However, compared to other domains of meaning, spatiotemporal metaphor has been systematically examined in a wide variety of unrelated languages, including both languages using a range of different spatial metaphors and languages with very different semantic systems for expressing the source domain of space. We are helped (thanks in great part to Stephen Levinson’s group at the Max Planck Institute at Nijmegen) by the fact that the cross-linguistic typology of spatial expressions (the Source Domain for spatiotemporal metaphor) is remarkably well examined, and also by the fact that spatiotemporal metaphor has been a major subject of interest to linguists recently, as well as to workers in cognitive science, gesture studies, and other related areas. This class of metaphors are therefore of very special interest to anyone working on semantic typology and universals, as well as to cognitive linguistics and cognitive science. This chapter gives a summary and synthesis of the linguistic side of the situation: what range of patterns of spatiotemporal metaphor do we observe in languages, and how do they correlate with linguistic spatial systems? In the next chapter, a broader range of evidence will be examined, showing how laboratory and field work, and gesture studies, bear on these linguistic analyses.
626
SWEETSER AND GABY
38.2 Patterns of Language Change and Polysemy One of the oldest observations in metaphor studies is that crosslinguistically, the single primary historical source for temporal vocabulary is spatial vocabulary. This is not just a tendency, but an overwhelmingly dominant one. The English words before and after once literally meant ‘in front of’ and ‘in back of,’ a pattern repeated again and again across the world’s languages. And in fact, before can still mean ‘in front of’ in some restricted frames, e.g. before my very eyes or before the judge. This is also typical, since historical change normally proceeds via patterns of polysemy – in this case, the same word referring to spatial and temporal senses. Crucially, the inverse direction of change is not observed: we do not find lots of linguistically conventionalized spatial phrases with origins in the semantic domain of time (instead their origins are in areas such as physical bodily structure, e.g. in back of includes the word back). This striking asymmetry is what we would expect of metaphoric mappings, which are known to be asymmetric in structure. As Cognitive Metaphor Theory has demonstrated, a metaphor is not a two-way comparison, but a cognitive framing of one domain or frame in terms of another (Lakoff and Johnson 1980a, 1999). The metaphor A N G E R I S H E A T A N D P R E S S U R E O F A C O N T A I N E D L I Q U I D , for example, maps our understanding of the metaphoric Source frame of Heat and Pressure onto the Target frame of Anger, including inferences such as the possibility of an explosion, which maps onto the possibility of a socially dangerous sudden rage (Lakoff 1987). This metaphor does not in the same way produce new inferences about Heat and Pressure based on our understanding of Anger. That metaphors are a basic component of historical linguistic change is well recognized (see Sweetser 1990 and its references). This is in itself a matter of major interest for semantics, since cognitive semantics (unlike formal logical semantic models) has developed a theory of meaning which includes and motivates the analysis of metaphoric polysemy patterns and meaning-change patterns. In this case, the overwhelming cross-cultural dominance of spatial metaphors for time suggests strong asymmetric cognitive bases in experience. But specifically, how are spatial semantic frames mapped onto temporal ones? What patterns are observed, and what variation is found?
38.3 The Range of Mappings 38.3.1 Moving-Time and Moving Ego First of all, many languages all around the world exhibit what we might call a family of T I M E I S ( R E L A T I V E M O T I O N I N ) S P A C E metaphors (Lakoff and Johnson 1980a, 1999). English abounds with metaphorical expressions of
Linguistic Patterns of Space and Time Vocabulary
this kind, which may be divided into two subtypes of space–time mapping. The first mapping (often known in the literature as Moving Ego) takes the viewpoint of a person moving along a temporal ‘landscape,’ as seen in expressions like She had left midterm time behind her, and was coming up on spring break. The second mapping, Moving Time, takes the viewpoint of a stationary observer or Ego, toward whom time moves from in front, as seen in Spring break was fast approaching. Ego is here understood to be some experiencer with a viewpoint on the situation; that viewpoint includes spatial experience of a location as Here and of front–back orientation, and also temporal experience of a Present time and its relationship to Past and Future times. Many other expressions are compatible with both mappings; if we say that Spring break is a long way off, we might invoke the Moving Ego frame to imagine ourselves (or some Ego) traveling over the temporal landscape toward the spring break, or we may invoke the Moving Time mapping to imagine spring break moving toward us from its currently distant position. This makes us note some shared aspects of these two metaphors’ mappings, namely T E M P O R A L S E Q U E N C E I S A L I N E A R S P A T I A L P A T H , and AMOUNT OF TIME BETWEEN EVENTS IS PHYSICAL DISTANCE BETWEEN L O C A T I O N S . But in other respects we must see different basic mappings: for the Moving Ego class of metaphors, Times are seen as Locations, and experiencing a time is being located at a location. This seems to inherit a basic structure from Lakoff and Johnson’s much-discussed general S T A T E S A R E L O C A T I O N S . That is, Times are understood to be a subclass of States, and Ego to be an experiencer of temporal states as well as other ones. But in the Moving Time mappings, things are somewhat more complicated: it is the Times which move past Ego’s (presumably fixed) location. Note that nowhere are Time and Ego understood metaphorically as two moving objects, moving (for example) in opposite directions; this is understandable, perhaps even predictable from the fact that linguistic spatial systems appear universally to treat spatial scenes primarily in terms of Figure–Ground relations; the located or moving entity is a Figure, relative to a stable Ground or Reference Point (cf. Talmy 2000a, 2000b, Levinson 2003); the same scene might thus be described as The bicycle (Figure) is behind the car (Ground) or The car (Figure) is in front of the bicycle (Ground). A Location or Motion Scene understanding of time thus needs to conceptualize either Ego or Time as Figure, and the other as Ground. This means that in any given conceptualization, Time and Ego cannot both be construed as moving. As Moore (2006, 2011a, 2014) has argued in detail for Wolof as well as English, and Nu´n˜ez and Sweetser (2006) concur, we should not really be talking only about Moving Ego or Moving Time – an even more basic question is whether the Reference Point (or Ground, in the spatial scenario) is Ego or Time. The Moving Time metaphors cited above are Egobased, in the sense that Spring break is approaching means specifically that
627
628
SWEETSER AND GABY
Ego is being ‘approached’ by (i.e will soon experience) spring break. There are also Time-based front-back Moving Time metaphors, such as the sequential relationship between times (Monday is always two days ahead of Wednesday) – here, although there is always some viewer, it is only the times which are metaphorically spatially related to each other: E A R L I E R T I M E S A R E A H E A D O F L A T E R T I M E S . This is different from seeing times as moving toward Ego, passing Ego, and then receding behind Ego (Monday will be here soon, Sunday is passing swiftly, Saturday has gone past). Although both involve Moving Time, this second one is T I M E S A R E O B J E C T S G O I N G P A S T E G O and F U T U R E I S I N F R O N T O F E G O , P A S T I S I N B A C K O F E G O as well as E A R L I E R T I M E S A R E A H E A D O F L A T E R . Both are distinct from Moving Ego metaphors, where Ego moves along a path in a temporal landscape: We’re getting close to exam time, We’ve left midterms behind us, which involve submetaphors T I M E S A R E L O C A T I O N S A L O N G A P A T H ; E G O I S A M O V E R O N T H E P A T H ; P A S T I S B E H I N D E G O ; F U T U R E I S I N F R O N T O F E G O ; and T I M E S A R E LINEARLY ARRANGED LOCATIONS. The contrast between Ego-based and Time-based metaphors, it should be noted, fits into a larger contrast between deictic and non-deictic linguistic models of Time. An Ego-based model necessarily has an Ego and an Ego’s Now as deictic centers of the spatial and the temporal frames mapped. Tense systems are another linguistic system which necessarily depends on some Now as a deictic center, relative to which Past and Future are earlier and later; so do expressions like tomorrow and next year. Other linguistic expressions, such as Tuesday or June 1925, or before and after, do not necessarily involve any Ego’s Now as a temporal deictic center: describing the review session as before the exam does not tell us whether these two events are seen as Past or Future. Thus, there are at least three major classes of time metaphors in English; and only two of these have been the focus of experimental investigation (to be discussed in the next chapter). All three are represented gesturally, however; the side-to-side timeline used by English speakers to show relative temporal sequence of events does not incorporate any placement of the gesturing self as Ego or Now, and therefore can be used to represent sequences of events which are all in the past, all in the future, or both. The front–back gestural timeline (in both English and ASL) necessarily incorporates Ego’s Now as the location right in front of the Speaker; so the front half of the timeline is necessarily Future, and the behind-Ego portion is Past. In the next chapter, the relationship between linguistic and gestural timelines will be discussed in more detail. All of the metaphors so far discussed are, as most investigators agree, Primary Metaphors stemming from very early and pervasive Primary Scenes in experience (as defined by Grady 1997a, Johnson 1997, Grady and Johnson 2002). It is impossible to move along a path, or to watch objects going past you, without experiencing the spatial sequence of locations as correlated with the temporal sequence of moments when you or the
Linguistic Patterns of Space and Time Vocabulary
moving object is at different locations. And both frames are understood as involving single-dimensional linear scales; a person moving along a path cannot get from point A to point D without traversing intermediate locations B and C, nor can an Ego now experiencing June 1 possibly experience June 4 without living through June 2 and 3. So there are parallel inferential structures in the two domains, readily mapped. Specifically, for an Ego moving forwards on a path, Locations ahead of Ego will be experienced at Future times, and Locations behind Ego are ones where Ego has already been at Past times: a very basic experiential correlation between Ego’s front–back axis and Ego’s experience of Past and Future. This might lead us to think that this family of space–time metaphors would be very common and basic indeed – and such is certainly the case; English space–time metaphor patterns turn out to be relatively unremarkable. However, they are not the only possibilities, though they are the dominant ones cross-linguistically.
38.3.2 What about Vertical Space? Mandarin Chinese, like English, construes the Past as behind Ego and the Future as in front of Ego; Yu (1998, 2012) described both the treatment of Past as behind a moving Ego, and future as being in front of a moving Ego, as well as the Time-based model where earlier times are in front of later ones. However, Mandarin speakers also regularly construe the Past as above Ego and the Future as below Ego: the Chinese phrases for ‘last week’ and ‘next week’ translate into English as UP- (or ABOVE-)WEEK and DOWN- (or BELOW-) WEEK respectively. Metaphors vary with the time-unit; ‘last year’ and ‘next year’ are FRONT YEAR and BACK YEAR (our standard Time-based metaphor, where E A R L I E R T I M E S A R E A H E A D O F L A T E R T I M E S ). Vertical metaphors for time are not unknown in other languages, though they are not as pervasive as the ones in Mandarin. French, for example, refers to what English historians call Late Antiquity (the late Roman and premedieval era) as basse Antiquite´ ‘low Antiquity’; and French haut Moyen Age (literally ‘high Middle Ages’) means ‘early Middle Ages.’ There are real correlations potentially underlying these metaphors – among them, the fact that in Mandarin, historically earlier writing systems were primarily from top to bottom rather than from left to right (so higher characters were earlier in the reader’s experience of the text) and that water flows downwards (an important motif in Chinese philosophy). Still, they do not seem to be as salient as correlations in early childhood human experience such as those relating motion along a path to temporal sequence, which are experienced any time there is any motion in the child’s environment. This may explain why the back–front spatial metaphors are so very much more pervasive across the world’s languages.
629
630
SWEETSER AND GABY
38.3.3 When the Future is behind You We have seen that there is an extremely broad tendency for humans to follow the mappings T H E P A S T I S B E H I N D E G O and T H E F U T U R E I S I N F R O N T O F E G O – that is, to map past times onto the local path behind the Moving Ego, and future times onto the area of a local path in front of Moving Ego. These mappings have a universally accessible correlational basis; and they also carry inferences that seem self-evidently useful. For a moving Ego on a path, we can infer that Ego will at a Future time be located at a location further ahead, and that Ego was located in the Past at locations which are (locally) behind Ego on the Path. Thus, we can infer that Future times will become the Ego’s Now and that Past times have been Now, and will not be Now again. (The mappings simply do not take into account the possibility of walking the opposite direction along a path, which makes sense since – barring science fiction scenarios – our experience of Time is entirely unidirectional.) However, this is by no means the only possible option. As Nu´n˜ez and Sweetser (2006) point out, the Aymara (and probably other neighboring high-Andean languages) appear to have a static time metaphor, where F U T U R E I S B E H I N D E G O and P A S T I S I N F R O N T O F E G O . At least part of the motivation for the Aymara mapping appears to be the very common (and Primary) metaphor K N O W L E D G E I S V I S I O N or K N O W I N G I S S E E I N G . We can see what is in front of us, not what is in back of us; and the Past is at least potentially Known, the Future is necessarily Unknown, hence mapped onto the area behind Ego. Note that this static metaphor has an extremely different inferential structure from the motion metaphors for time. The fact that something is currently in front of or in back of a static person does not tell us much about future or past co-location scenarios. If I am standing in a classroom, facing the class and the wall is behind them, that does not mean that I am going to be located at that back wall in the future; and if I turn to face the blackboard and not the class, that does not allow a viewer to infer that I was previously located at the back wall which is now behind me. So another Primary Metaphor, K N O W L E D G E I S V I S I O N , combines with moving and static time metaphors differently. We get F U T U R E I S B E H I N D E G O in a static construal of Ego in a Time ‘landscape’ (the area behind a static Ego is not visible), but a F U T U R E I S A H E A D O F E G O mapping is beautifully coherent with a construal where Ego moves dynamically along a temporal ‘path’ (since the part of the path ahead of a moving Ego, at least around the next curve, is not visible yet, or still unseen). We need to add here that Moore (2014) has thoroughly examined a range of earlier scholarly claims of languages in Europe and Asia where ‘P A S T I S I N F R O N T ,’ and shown that they generally reduce to E A R L I E R T I M E S / E V E N T S A R E I N F R O N T O F L A T E R T I M E S / E V E N T S (a time-based metaphor) rather than
Linguistic Patterns of Space and Time Vocabulary
to the ego-based P A S T I S I N F R O N T O F E G O . So Aymara and its neighbors are truly unusual in the world’s languages.
38.3.4 Absolute Spatial Languages and Time In relative spatial languages, spatial reference to relative location pervasively involves some projection of Ego’s directional coordinates; for example, English speakers will say My purse is to the left of the laptop, or The ball is in front of the tree. The first of these usages needs an interpreter to project some viewer’s left–right axis onto the scenario of the computer and the purse; the second, even more complex, involves a construal of the tree as a person facing the viewer, so that a ball between the viewer and the tree is in front of the tree. Relative spatial languages are far more common crosslinguistically than Absolute spatial languages (Levinson 2003), in which spatial location may be established primarily in terms of Geocentric structures (my laptop is east of my purse) or other geographic features (you have an ant on your ‘upriver’ leg). We have noted that the cross-linguistically dominant spatial models of time are all relative spatial ones: times are seen as above or below Ego, in front of or in back of Ego or other Times. Field and experimental work with relative spatial languages will be discussed in the next chapter. But crucially, a language cannot exploit spatial systems which it does not have, in building temporal language. Absolute spatial languages generally have egocentric deictic systems (‘here’ and ‘there’ relative to Speaker and/or Addressee), but not left–right or front–back egocentric linguistic expressions. So the kinds of temporal models we have seen as characteristic of relative spatial languages would not be possible in absolute languages. At present, it is not clear that there are any linguistic temporal systems of metaphor based entirely on absolute spatial language, although (as we shall see in the next chapter) speakers of absolute spatial languages do seem to construe time in absolute spatial terms, judging by other metrics such as gesture and spatial arrangement tasks.
38.4 Writing Systems and Time We mentioned above that writing systems do appear to influence metaphors for time; the fact that Chinese has vertical metaphors in particular could be motivated by the vertical writing direction dominant in earlier Chinese. However, an added puzzle arises here. Chinese is now written primarily left-to-right (and of course the sequence of left–right lines are read from the top down), as are European languages. Arabic and Hebrew are written from right to left. And yet no language seems, anywhere, to have a basic or even systematic left–right linguistic metaphor for temporal structure. We never find words for ‘right’ and ‘left’ coming to mean ‘after’
631
632
SWEETSER AND GABY
and ‘before’ (or the other way round, for Semitic languages) – even though visual scenes are certainly interpreted this way – for example, an English speaker reading a comic interprets panel A as representing an event before the event in Panel B, precisely if/because panel A is to the left of panel B. And, when English speakers gesture along a side-to-side timeline, they gesture left-to-right. One possible simple reason for this linguistic gap, proposed by Dancygier and Sweetser (2014), is that we do not in fact conceptualize the side-to-side timeline, or our left–right writing and comic strip layouts, as primarily being from our (the reader/gesturer’s) left to our right, but rather as forwards motion of a virtual or literal trajector (gaze, gesturing hand) along its trajectory. And this fits in reading very nicely with all the very pervasive back–front time metaphors which are so dominant in many languages. It is not that the writing system is the primary motivation for these metaphors; as we said earlier, motion along a path is the primary experiential motivation. But once that is there, it is easy to see how the writing system’s structure can be neatly incorporated into the front–back structure of the time metaphors. Our gaze or our hand moves along a trajectory, and it will reach farther-right locations later than farther-left ones; hence any location to the left of another is ‘forwards’ along this trajectory, and a location to the right of another is ‘farther back.’
38.5 Conclusions The linguistic data, therefore, seems to reflect a great many cognitive and cultural factors. The following generalizations from cross-linguistic comparison do, however, hold up. First, since the world’s languages are dominantly relative-spatial rather than absolute-spatial, time metaphors are also dominantly relative-spatial metaphors, involving Ego and Time as some combination of Figure and Ground. Second, within that range, front–back metaphors dominate over up–down ones, which may be primarily based in writing system structures and lack Primary Metaphor status. And this is very likely because of Primary Scene correlations which make these metaphors easily accessible cognitively. Third, it is motion metaphors rather than static ones which dominate (no surprise given Lakoff and Johnson’s more general concepts of S T A T E S A R E L O C A T I O N S and C H A N G E I S M O T I O N ); and other metaphors such as K N O W L E D G E I S V I S I O N interact differently with motion metaphors than with static ones. And fourth, writing systems may influence spatial metaphors for time; but in order to understand their influence on the linguistic metaphors, we may need to understand how the writing system’s structure is itself metaphorically construed as motion. The next big set of questions are of course (1) what are the cognitive bases for the wide range of less dominant spatial understandings of time
Linguistic Patterns of Space and Time Vocabulary
and (2) to what extent are these spatial understandings linguistically manifested (as opposed to gesturally or in physical space, for example)? Field and lab work have investigated some relevant languages in detail, and the next chapter will lay this out. But a great deal of complex metaphoric structure can be analyzed in the linguistic patterns themselves. We can still hope for much more cross-linguistic data on polysemy and historical change. But at least in this domain, some major generalizations have emerged. As was mentioned earlier, it would be extremely hard even to talk about (much less analyze) these generalizations about semantic systems without the resources of cognitive linguistics. We have other reasons to believe that inter-frame metaphoric mappings, based ultimately on experiential correlations, are generally crucial to human cognition and language. These frameworks allow us to lay out the structures of spatial metaphors for time in a systematic way, and compare them in detail, as well as examine motivations for the dominance of some models over others crosslinguistically. Spatial metaphors for time have interested cognitive scientists and cognitive linguists partly because they are one example of the ways in which humans apparently ‘bootstrap’ their abstract concepts on concrete aspects of experience. And it does seem clear that it would be impossible for any human to experience, in particular, motion in space without experiencing time; the two domains are inextricably correlated. Yet the observed patterns in spatiotemporal metaphors are not as simple as in the classic M O R E I S U P , L E S S I S D O W N , one of a few apparently actually universal metaphors, and plausibly hypothesized to be based on the experience of rising vertical level which correlates with adding more liquid to a container or more objects to a vertical stack. We see a range of preferred patterns, but nothing as fixed as M O R E I S U P or the perhaps almost equally universal M O R E I S L A R G E R (a greater quantity occupies more space). And we are still working on understanding the experiential motivations involved in spatiotemporal metaphors; it should be noted that just a couple of decades ago, some of these linguistic patterns had not yet been studied at all. And to what extent do culturally shaped aspects of physical experience help set the options for choices between different experiential motivations for metaphors? We know that Absolute spatial language speakers not only talk about space differently, but carry out spatial cognition tasks differently (Levinson 2003); although the motor experience of walking on a path is presumably shared, construal of spatial relations and navigation tasks are different. Within the larger universe of meaning, it is hard to say how typical or atypical this area of the metaphor map is. There are overall apparent patterns – such as that Target frames are less intersubjectively accessible than Source frames; this seems a more sustainable claim than that they are always more physically concrete (see Dancygier and Sweetser 2014),
633
634
SWEETSER AND GABY
since many metaphors map abstract social frames onto other abstract frames (e.g. A N A T I O N I S A F A M I L Y ) or concrete ones onto concrete ones. But we just do not have enough data on metaphor systems outside of European languages and a few other well-documented ones such as Chinese, Japanese, or (to a lesser extent) Arabic. And even in these relatively documented languages, we lack fully laid-out analyses of metaphors for Self, Action, Relationship, Morality, Social Structure, and the wide range of cognitive and cultural domains which are typically construed metaphorically. Do these domains also have a few major dominant metaphoric patterns, and a few other salient ones with different motivations? Or is there wider and more scattered variation, some of it perhaps based on culturally distinct experiences of both source and target frames? We are still really just starting the work on this area of cognitive linguistics.
39 Space–time Mappings beyond Language Alice Gaby and Eve Sweetser
39.1 Introduction Over the course of an average day, our experience of time is linked to spatial locations countless times. Every motion event – such as catching the train to work – involves a correspondence between a change in position and the passage of time (the train is at position α at time α, then position β at time β). The act of eating involves fine coordination of the timing of the mouth’s opening and closing with the position of the fork, and this is only scratching the surface. It is perhaps unsurprising, then, that the association between space and time is arguably a human universal, documented as preceding the development of language, both ontogenetically in human infants (Srinivasan and Carey 2010, de Hevia et al. 2014) and phylogenetically in other species of great ape (Merritt, Casasanto, and Brannon 2010). This space–time linkage is reflected in linguistic metaphors the world over (though see also section 39.3.4), and also in nonlinguistic artifacts, conventions, and other manifestations of human culture and cognition (the subject of the present chapter). Of these, the sundial offers an excellent example of how the natural and the conventional spatialization of time may combine; the position of the shadow, by naturally reflecting the position of the sun, conventionally represents the hour. Other spatializations of time are more arbitrary. For example, the English writing convention of transcribing earlier uttered sounds and words to the left of those uttered afterwards is simply a matter of cultural convention, as witnessed by right-to-left and top-to-bottom writing conventions for other languages (section 39.3.2.1). Similarly arbitrary spatializations of time include calendars, timelines, flow charts, and the like (see Gell 1992, Munn 1992 for a fuller discussion of these forms). This chapter is divided into two major parts. The first (section 39.2) considers a range of lab-based explorations of examinations of the extent and nature of the link between time and space in the mind. These studies
636
GABY AND SWEETSER
have focused upon populations co-located with the researchers’ universities (e.g. speakers of English, Mandarin, Hebrew, and Dutch). The fieldbased studies that are the focus of this chapter’s second part, meanwhile, have widened the empirical base in terms of the languages and cultures under consideration. While many lab-based studies have presumed backto-front and left-to-right (or, occasionally, right-to-left/top-to-bottom) timelines to be the only candidates for linguistic and conceptual construal, field-based studies have emphasized cross-cultural and cross-linguistic diversity in how temporal categories are mapped onto spatial categories, and have examined the extent to which thought, gesture, and nonlinguistic representations of time reflect the dominant modes of talking and thinking about both time and spatial relationships. Accordingly, section 39.3 is organized according to the major qualitative divisions in how time is spatialized.
39.2 Insights from the Lab Researchers from various disciplines and theoretical approaches have designed lab-based experiments to explore the relationship between time and space in the mind. The following sections present some of the most important findings of this research.
39.2.1 Time and Space in the Mind In this section, we review two important debates regarding the relationship between time and space in the mind. Section 39.2.1.1 considers the degree to which the specific T I M E I S S P A C E metaphors instantiated in various languages reflect (and/or construct) the cognitive construal of time. Section 39.2.1.2 reports on the debate over whether space and time are on an equal footing in the mind (e.g. as part of a more general system of reasoning about magnitude), or whether space plays a more fundamental role in human reasoning, supporting cognition about the more ephemeral time. 39.2.1.1 The Psychological Reality of Metaphors English furnishes its speakers with two major subtypes of the T I M E I S S P A C E metaphor (Lakoff and Johnson 1980a). The first – known as the egomoving frame – takes the viewpoint of a person moving along a trajectory, where points in time are landmarks along that trajectory (i.e. we are getting close to lunchtime). The second – known as the time-moving frame – takes the viewpoint of a stationary observer, toward whom points in time move from far ahead, to nearby, to co-located, to the behind-space (i.e. winter is fast approaching). Many metaphorical expressions of time are compatible with both of these frames; if we say that those days of leisure are far behind us,
Space–time Mappings Beyond Language
we may have traveled beyond our leisure time in an ego-moving frame, or our leisure time may have reached us and then passed into the behindspace within the time-moving frame. Many experimental studies have focused upon whether these metaphors reflect and/or construct how English-speakers conceptualize time. For example, whether a participant is or has recently been moving through space – or even imagining themselves moving through space – can influence whether they adopt an ego-moving or time-moving perspective when processing temporal metaphors (McGlone and Harding 1998, Boroditsky 2000, Boroditsky and Ramscar 2002). For example, someone waiting at an airport is more likely to adopt a time-moving perspective, which in turn means they will likely interpret a prompt such as (1) as indicating that the meeting has been moved to Monday. (1)
Next Wednesday’s meeting has been moved forward two days. What day is the meeting now that it has been rescheduled?
Someone arriving at that same airport after a flight, however, will more likely adopt an ego-moving perspective, indicating that the meeting has been rescheduled for Friday. From both the ego-moving and time-moving perspectives, the future is consistently in front of the viewpoint and the past behind. Torralbo, Santiago, and Lupian˜ez (2006) tested the psychological reality of these linguistic associations by presenting their participants with words with temporal connotations, placed at different positions with respect to a head in silhouette. They found participants were faster to judge the words as referring to the past or future “when the irrelevant word location was congruent with the back-past, front-future metaphoric mapping” (Torralbo, Santiago, and Lupian˜ez 2006: 745). Other such judgment tasks find the mental representation of the past to be associated with the left, and the future with the right, despite the absence of any left-to-right metaphorical timeline in spoken language (cf. section 39.3.2.1 and section 39.4.2). Santiago et al. (2007), for example, asked their participants to categorize words as referring to the past or the future by pressing keys with either their left or right hand. Participants were faster to press the correct key when the right-hand key indicated the future and the left-hand key the past, but also when the words referring to the past were presented on the left side of the screen and words referring to the future on the right. The psychological reality of the left-to-right timeline was supported by studies such as Ulrich and Maienborn (2010) and Flumini and Santiago (2013), which further suggest that such associations are nonautomatic. That is to say, the left-to-right conceptual timeline is only activated by temporal language when the task participants are performing explicitly involves temporal judgments. Some studies go further than simply identifying a correlation between linguistic and mental representations of time in terms of space. Casasanto
637
638
GABY AND SWEETSER
(2008b), for example, finds a telling difference between how English and Greek speakers conceptualize time, reflecting differences in the metaphorical description of time in the two languages. While English speakers describe time in terms of length, Greek speakers describe time in terms of amount. Correspondingly, the spatial extent (= length) of visual stimuli influenced the temporal judgments of English but not Greek speakers, whereas visual stimuli manipulating amount influenced the temporal judgments of Greek but not English speakers. The extent to which experimental studies reveal linguistic metaphors to correspond to, or even determine, nonlinguistic construals of time is considered in detail by Gijssels and Casasanto (this volume Ch. 40), and so will not be discussed further here, except with reference to the origins of spatiotemporal construals (in section 39.4).
39.2.1.2 The Cognitive Primacy of Space Given the strong conceptual relationship between time and space, a key theoretical question concerns whether this relationship is symmetrical or asymmetrical. ATOM (A Theory Of Magnitude) emphasizes symmetry, drawing on evidence that space, time, and number (as well as some other domains such as pitch and quantification) share neural structures, feeding both priming and interference in processing and reasoning in each of these domains (Walsh 2003). Conceptual Metaphor Theory (CMT), however, emphasizes the asymmetric influence of space (as the concrete source domain) in structuring the representation of abstract target domains such as time and number. A number of studies supporting an asymmetric relationship between space and time have shown spatial information to interfere with temporal judgments, but not the reverse (e.g. Boroditsky 2000, Bottini and Casasanto 2010). For instance, participants shown a dot moving along one dimension on a computer screen judge the time taken for the dot to reach its final position as longer for dots that move further, and shorter for dots moving shorter distances, whereas the length of time the dot spent moving does not strongly affect participants’ judgments of the distance travelled (Casasanto and Boroditsky 2008). Significantly, this bias extends to young children (Casasanto, Fotakopoulou, and Boroditsky 2010, Bottini and Casasanto 2013) but not to monkeys: “In monkeys, both spatial and temporal manipulations showed large bidirectional effects on judgments . . . human adults showed asymmetrical space–time interactions that were predicted by metaphor theory” (Merritt, Casasanto, and Brannon 2010: 191). Other studies, however, would seem to support ATOM at CMT’s expense. CMT, for example, predicts that target domains such as time and number should only be related to one another via their shared source domain, space. Yet several studies evidence a bidirectional relationship between judgments of number and time among both Arabic speakers (Xuan et al.
Space–time Mappings Beyond Language
2007, Roitman et al. 2007) and English speakers (Kiesel and Vierck 2009, Matlock et al. 2011; cf. Winter, Marghetis, and Matlock 2015: 214). In their review of the debate about whether ATOM or CMT best accounts for the interactions between space, time, number, and some other domains, Winter, Marghetis, and Matlock (2015) conclude that ATOM is best not interpreted in its strictest sense, that is, entailing symmetrical, bidirectional associations between (for our purposes) time and space. But they see a place for both (the more loosely defined) ATOM and CMT in explaining the relationship between time and space in the human mind: “The evolutionarily older magnitude system in parietal cortex posited by ATOM might be subject to neural reuse or recycling as a result of culture and experience (Anderson 2010; Dehaene and Cohen 2007), shaped throughout ontogeny by cultural artifacts and practices – including language and writing – to produce more directional, asymmetric mappings” (Winter, Marghetis, and Matlock 2015: 219).
39.2.2 Body Movements Researchers do not have direct access to the cognitive processes of their subjects. Instead, we must rely upon indirect evidence of what is going on ‘under the hood.’ Body movements of various kinds represent a valuable source of such indirect evidence. The following sections review experimental research into what hand gestures (39.2.2.1), eye movements (39.2.2.2), and leaning and body posture (39.2.2.3) reveal about how we mentally construe time in terms of space. 39.2.2.1 Hand Gestures Since McNeill’s pioneering work (e.g. 1992), gesture has featured prominently in psycholinguistic research as a window on cognition. Speakers of English and other European languages have a well-documented propensity for gesturally locating the past in the space behind their bodies and the future in front of them (e.g. de Jorio 1832 [2000], Calbris 1990, Cooperrider and Nu´n˜ez 2009). Additionally, the conventions of literacy, calendars, and other cultural artifacts provide these speakers with a well-exploited alternative: gestures that map the past to the left and the future to the right are also extremely well attested, despite the absence of equivalent metaphors in any spoken language (cf. Cienki 1998, Sweetser and Gaby this volume Ch. 40). Surprisingly, these lateral (left–right) temporal gestures frequently accompany sagittal (back–front) temporal metaphors in concurrent speech (Casasanto and Jasmin 2012). Clearly, spoken metaphors are not the only source for the space–time mappings seen in gesture (see section 39.4 for further discussion). Moreover, the particular mapping instantiated in a particular gesture has been shown to be sensitive to pragmatic context as well as the level of granularity of the time period indicated (see section 39.4.3). The various forms temporal gestures
639
640
GABY AND SWEETSER
take in the languages of the world are comprehensively reviewed in Cooperrider, Nu´ñez, and Sweetser (2014); some of the more unusual forms are discussed further in section 39.3. Before moving on, it must be acknowledged that gestural timelines are not a cultural universal. Recent research on Yucatec Mayan (spoken in Mexico; Le Guen and Pool Balam 2012) found the gestures of its speakers to make instead a simple “opposition between ‘current time’ (mapped on the ‘here’ space) and ‘remote time’ (mapped on the ‘remote/distant space’). Additionally, past and future are not contrasted” (Le Guen and Pool Balam 2012). These authors suggest that the prominence of geographically oriented gestures (where pointing gestures are overwhelmingly interpreted as indicating a specific location), “may to some extent preempt the use of gesture space for other domains like time” (Majid, Gaby, and Boroditsky 2013). While it is true that gesture is neither linguistic nor wholly consciously produced, it remains at least somewhat conventionalized, arbitrary, and culturally transmitted (see Kendon 1983). The following sections consider some other kinds of bodily movements which have been argued to reflect temporal cognition more directly.
39.2.2.2 Eye Movements Only relatively recently has eye-tracking technology been employed in investigating the mapping of time to space in nonlinguistic cognition. Stocker et al. (2015) measured German-speaking participants’ eye movements along a blank screen as they listened to stimuli involving past, future, and same-time related sentences. They found significantly more upward saccades were made by participants listening to future-related sentences than past-related ones, suggesting that “as we mentally represent time, our mind’s eye follows an upward – and possibly forward – progressing mental time line, and this is reflected in corresponding oculomotor correlates” (Stocker et al. 2015). These results are consistent with a conceptual timeline running outwards from the body along the sagittal axis (as instantiated by linguistic metaphors of the past behind and the future ahead) since, as Stocker et al. put it, “if participants were able to mentally project the future as extending sagittally out of their body, then the geometrical projection onto screen coordinates would lead to future locations higher on the screen than past location” (2015). Strikingly, the participants in this study did not show any evidence of a left-to-right mental timeline. This contrasts not only with many of the studies reported in sections 39.2.2.1, 39.3.2.1, and 39.4, but also with another eye-tracking study conducted by the same research team but with a different methodology. Instead of measuring eye-movements during the comprehension of (time-related) linguistic stimuli, Hartmann et al.’s (2014) participants were instructed to imagine themselves either one year in the past or in the future. Participants were found to gaze downwards and to the left under
Space–time Mappings Beyond Language
the past condition, and upwards and to the right under the future condition. Together, these eye-tracking studies suggest that German speakers access both sagittal (back-to-front) and lateral (left-to-right) mental timelines. Further research is required to tease apart any effects of timescale, and language comprehension versus imagination, but eye-tracking seems a fertile area for future study.
39.2.2.3 Leaning and Body Posture It might be argued that the saccades and fixation points measured in eyetracking studies are conditioned by habits of reading and attention, and do not directly reflect the conceptualization of time per se. Another way in which body movements have been exploited as a window on cognition circumvents this issue. Miles, Nind, and Macrae (2010) attached extremely sensitive motion sensors to their participants’ legs, designed to measure sway (changes in posture or leaning direction). With sensors attached, participants were instructed to recall a typical day in their lives four years previously, or to imagine what a typical day will be like for themselves four years into the future. They found that participants remembering the past leaned backwards, while those engaging in mental time travel into the future leaned forwards. Thus, as (Miles, Nind, and Macrae 2010: 223) put it, “the embodiment of time and space yields an overt behavioral marker of an otherwise invisible mental operation.”
39.3 Insights from the Field Most of the lab-based studies described in section 39.2 assume continuity in how time is described, and focus upon the questions of how these spatial representations of time in language relate to how time is conceptualized, as well as relationships with other domains such as numerosity, size, and pitch. Meanwhile, field linguists, psychologists, and anthropologists have emphasized cross-linguistic and cross-cultural diversity in these domains. First, they do so by finding dramatic variation in how time is described in terms of space linguistically (see Sweetser and Gaby this volume Ch. 38). Second, they do so by finding that the description of spatial relationships between objects varies dramatically from language to language, and that this variation corresponds with variation in spatial reasoning, memory, and more (see, e.g., Levinson 2003, Majid et al. 2004). These findings taken together raise an intriguing question: to what extent do cross-cultural differences in construing spatial relationships engender different construals of time? The studies discussed in this section attempt to answer that question by various means. Several of them draw on data elicited through the Temporal Representations Task introduced in section 39.3.1. Others consider evidence from gesture (cf. section 39.2.2.1), cultural artifacts,
641
642
GABY AND SWEETSER
and other devices. But rather than organizing this section according to the nature of the data analyzed (as in section 39.2), the discussion below is presented according to the nature of the timeline revealed. First of all, there is a cleavage between timelines arranged according to an individual’s perspective (the ‘ego-centric’ timelines, section 39.3.2), and those anchored by (features of) the broader environment that the individual is situated within (the ‘geo-centric’ timelines, section 39.3.3). The ego-centric timelines in turn divide into those extending along the sagittal axis (in front/behind, section 39.3.2.2) and those extending along the lateral axis (left/right section 39.3.2.1). Geo-centric timelines, meanwhile, may extend along an axis defined by the uphill/downhill slope (section 39.3.3.1), a river course (section 39.3.3.2) or the cardinal directions (north, south, east, and west, section 39.3.3.3). These categories do not distinguish the full array of analytical possibilities, however. Since there is not space to do justice to the proposals of, for example, Moore (2006, 2011a), Bender, Beller, and Bennardo (2010), and Tenbrink (2011), the interested reader is referred to those excellent works, as well as Tenbrink (this volume Ch. 41).
39.3.1 Temporal Representation Task Since many of the studies discussed in sections 39.3.2–39.3.4 draw upon evidence from a standard experimental task designed to elicit novel spatial representations of time, it is worth beginning with a brief overview of this task. The Temporal Representation Task (described in full by Boroditsky, Gaby, and Levinson 2008) consists of two components. In the first, participants are presented with a set of between four and six photo-cards. The photos of each set depict a particular event or process unfolding over time. The timescales involved vary from set to set, from the very short (e.g. an egg being dropped onto a table, the shell cracking, and the contents spilling out), to the long (e.g. a young boy aging to become an old man). Participants are instructed to lay out the cards in order, and are thus covertly required to choose a spatial layout for the passage of time. Having laid out half the sets, participants are re-seated at a 90° or 180° rotation from their original orientation in order to complete the remaining trials. This rotation allows for the ego-centric and geo-centric arrangements to be distinguished from one another. In the second component, the researcher indicates a point in front of the participant (either in the air, or by drawing a dot in the sand, or placing a counter on the ground or tabletop, depending on the experimental context and researcher involved). The participant is told that this point represents a particular deictic timepoint (e.g. today), and is then asked where they would place other deictic timepoints (e.g. tomorrow and yesterday).
Space–time Mappings Beyond Language
The various timelines produced in response to this task by participants around the world are discussed in the following subsections, along with insights garnered from observation, interview, elicitation, and other tasks.
39.3.2 Ego-centric Timelines This section considers the two ways in which time has been documented to be spatialized with respect to the speaker/thinker/gesturer’s body; along the lateral axis (left-to-right or right-to-left, section 3.2.1) or along the sagittal axis (in.front-to-in.back or in.back-to-in.front). 39.3.2.1 Lateral Timelines As we have seen, the lateral left-to-right timeline used by English speakers has been attributed to writing conventions and associated graphic representations (e.g. timelines, calendars and so forth; Tversky, Kugelmass, and Winter 1991; cf. section 39.4.2). However, not all languages are written from left-to-right. Where writing conventions vary, corresponding variation has been found in other nonlinguistic representations of time. For example, in the Temporal Representation Task (section 39.3.1), speakers of Hebrew arranged the stimulus cards to show the passage of time proceeding from right-to-left, conforming to the direction of the Hebrew script (Fuhrman and Boroditsky 2010). Particularly telling is the example of Chinese, which has been written top-to-bottom, right-to-left, and left-toright in different locations (e.g. mainland China versus Taiwan) and different eras. Both de Sousa (2012) and Bergen and Chan Lau (2012) find their participants’ performance in experimental tasks (including the Temporal Representation Task) to conform to the writing direction to which they have had most exposure (cf. Chan and Bergen 2005, see also Lum [in prep.] on the effects of text messaging and the use of a Romanized script on how time is represented by speakers of Dhivehi, which is conventionally written from right to left). These lateral timelines are of particular interest given the complete absence of lateral terms used to describe time in language (see section 39.4.3 for further discussion of this point). The influence of literacy on temporal representations is also considered further in section 39.4.2. 39.3.2.2 Sagittal Timelines The conceptualization of the past as behind us and the future ahead (along an ego-centric, sagittal timeline) is familiar to all speakers of English and most other languages (cf. section 39.2). More unusual, however, is the reversal of this timeline as by the Aymara. Living in the mountainous region bordering Bolivia, Peru, and Chile, Aymara speakers are now famous for gesturing in front of themselves to indicate the past, while pointing over their shoulders to the unknowable future (Nu´n˜ez and Sweetser 2006a). Like English speakers, though, the Aymara have access
643
644
GABY AND SWEETSER
to an alternative, lateral timeline running from left to right. In the Aymara case, the sagittal timeline is used with reference to deictic time (e.g. long ago, next year), while lateral gestures are used for sequence time (he baked the cake and then went for a walk) (see section 39.4.3 for further discussion of how topic, granularity and pragmatic context influence the choice of coordinate frame).
39.3.3 Geo-centric Timelines This section presents three ways in which time can be construed independently of any viewpoint, being instead anchored to the broader environment. Specifically, locating the future: uphill or upriver (39.3.3.1), downriver (39.3.3.2), or to the west (39.3.3.3). 39.3.3.1 The Future is Uphill (or Upriver) The past is associated with the downhill direction and the future with uphill in at least three languages: Tzeltal (spoken in Mexico; Brown 2012), Yupno and Nungon (both spoken in Papua New Guinea; Nu´n˜ez et al. 2012, Sarvasy 2014). Of these three, only Yupno shows strong evidence of these associations extending beyond language (though see Sweetser and Gaby this volume Ch. 38 for further discussion of the linguistic metaphors; see also section 39.4.1 for further discussion of Tzeltal). Nu´n˜ez et al. (2012) elicited temporal gestures in twenty-seven semistructured interviews with Yupno speakers. They found gestures indicating the present time point directly to the ground (at the speaker’s current position). Gestures for past and future, however, do not conform to a straight line. Instead, past gestures (in aggregate) point downhill toward the mouth of the Yupno river, while future gestures point upwards and/or uphill toward the river’s source (Nu´n˜ez et al. 2012: 30). As a result, the future is rotated 111 degrees from the past – rather than 180 – a fact which Nu´n˜ez et al. (2012) attribute to the fact that the village of Gua (where consultants were tested) lies off to one side of the linear axis connecting the river’s source and mouth. This begs the question of whether the Yupno timeline in fact runs from downriver (/past) to upriver (/future). A further intriguing complexity of Yupno spatiotemporal construals involves how this system is transposed from outside to indoors. Measurement of gestures produced inside three houses (with entryways oriented at 90°, 150°, and 345°) showed Yupno speakers to gesture toward the entryway in order to indicate the past, and away from the entryway for the future (Nu´n˜ez et al. 2012: 32). This aligns with linguistic description of spatial relationships such that objects on the entry side of the home are described as ‘downhill’ and those away from the entry ‘uphill,’ regardless of actual orientation (Nu´n˜ez et al. 2012: 33, Cooperrider, Slotta, and Nu´n˜ez 2016).
Space–time Mappings Beyond Language
39.3.3.2 The Future is Downriver A possible – albeit limited – reversal of the Yupno’s uphill passage of time is seen in Mian, an Ok language of Papua New Guinea. As with Kuuk Thaayorre (section 39.3.3.3) and Aymara (section 39.3.2.2), the use of Mian spatial terms to refer to temporal meanings is extremely limited. However, the presence of tab ‘down(river)’ in examples such as (2), may indicate that time is conceptualized as flowing down(river). (2)
am=o hebmamsaˆb tab tl-Ø-o=be time=N2 quickly down come.PFV-REAL-N2.SBJ=DECL ‘The time passed quickly’. (Fedden and Boroditsky 2012: 485)
In the Temporal Representation Task described in section 39.3.1, Mian speakers employed a variety of strategies, including the representation of time flowing from left-to-right (consistent with local literacy practices), toward the body, and along a landscape-based, ‘absolute’ axis. Since the rivers of the region flow to the WNW, it is difficult to distinguish arrangements according to the compass directions from those aligning with the river. However, the temporal usage of tab ‘down(river),’ together with the fact that “the absolute arrangements appear to be rotated slightly clockwise off of the east-west axis” (Fedden and Boroditsky 2012: 487) is suggestive of a construal of time in terms of the riverflow. This remains a tantalizing area for further investigation.
39.3.3.3 The Future is West Speakers of the Australian Aboriginal language Kuuk Thaayorre also produced geographically grounded arrangements of stimulus cards and dot points in response to the Temporal Representation Task (section 39.3.1). In this case, however, it was not the direction of the riverflow or slope that anchored the timeline, but the directional east–west axis (Boroditsky and Gaby 2010). Given the sun’s apparent trajectory across the sky over the course of a day, the mapping of earlier/past to the east and later/future to the west seems natural.
39.3.4 Nonlinear Timelines We have already seen that the Yupno timeline is structured according to a “bent geometry” (Nu´n˜ez et al. 2012: 30). Other timelines depart even more dramatically from the straight line model familiar to most English-speaking readers. Perhaps the most famous of these is the cyclical representation of time seen in Mayan and other Meso-American cultural artifacts, such as the traditional Long Count and other calendars (Gossen 1974, Tedlock 1982). Neither of the two Mayan populations that participated in the Temporal Representation Task (section 39.3.1) drew consistently on a particular coordinate frame to represent the passage of time spatially. For example, of the twelve Tzeltal-speakers who participated in the card-arrangement task,
645
646
GABY AND SWEETSER
seven produced inconsistent responses across trials, and each of the remaining five – who were internally consistent – employed a different coordinate frame from the others: one left-to-right, one right-to-left, one near-to-far, one south-to-north, one east-to-west (Brown 2012: 8). This inconsistency may be at least partly due to the availability of no fewer than five distinct schemata for conceptualizing time, as evidenced by temporal expressions in spoken Tzeltal. Speakers of Ye´lıˆ Dnye (a language isolate spoken on Rossel Island, PNG) were likewise found to employ a range of different strategies for representing time spatially in the Temporal Representation Task: “experimental evidence fails to show a single robust axis used for mapping time to space” (Levinson and Majid 2013: 1). Levinson and Majid suggest that the lack of systematic spatialization of time in this experimental task, along with the lack of spatial expressions for time in spoken Ye´lıˆ Dnye, may stem from the language’s abundance of dedicated terms for deictic time categories. These include six diurnal tenses, special nominals for n days from coding time, and special constructions for overlapping events.
39.4 Convergent Insights, Lingering Questions and Future Directions If conceptualizing time in terms of space is not universal, it is certainly extremely widespread. While section 39.2.1 considered lab-based investigations into the nature of the relationship between conceptual representations of time and space, in this section we consider the origins of this relationship. Section 39.4.1 presents evidence from both lab-based and cross-linguistic studies that temporal cognition does not always reflect temporal language. Clearly, then, language cannot be the (only) force shaping how humans think about time. An alternative model for conceptual representations of time are nonlinguistic conventions such as literacy, calendars, and other cultural artifacts. The influence of literacy on temporal cognition is considered in section 39.4.2. How competition between all of these models of representing time in terms of space plays out is considered in section 39.4.3, while section 39.4.4 considers some open questions for further research.
39.4.1 Spatiotemporal Thought without Spatiotemporal Language To some extent, the way that we conceptualize time mirrors the way we talk about it. Consider, for example, the contrast between Greek and English speakers’ associations of time with quantity and length respectively (section 39.2.1.1; cf. Gijssels and Casasanto this volume Ch. 40). But spatialized construals of time do not depend upon the presence of corresponding space–time metaphors in language. Among speakers of the
Space–time Mappings Beyond Language
Amazonian language Amondawa, for example, “even when entrenched, habitual, regular linguistic space–time mapping is absent, the cognitive capacity for construing temporal concepts in terms of spatial arrays is present” (Sinha et al. 2011: 164).1 Kuuk Thaayorre likewise possesses only two polysemous terms with spatiotemporal meanings (raak ‘place, earth, ground, time’ and kanpa ‘in front of, earlier’), in the absence of more extended spatial metaphors for time (Gaby in press). As discussed in section 39.3.3.3, however, speakers of Kuuk Thaayorre have been found to represent time as flowing from east to west in experimental tasks – a timeline with no linguistic analogue whatsoever. The converse situation, meanwhile, is illustrated by speakers of Tzeltal, for whom the “systematic and consistent use of spatial language in an absolute frame of reference does not necessarily transfer to consistent absolute time conceptualization in nonlinguistic tasks” (Brown 2012: 10), such as those described in section 39.3.1 (cf. section 39.3.4). Studies involving participants without language – such as infants and other great ape species – are another important source of evidence that linguistic metaphors of time in terms of space are not the (only) foundation of spatiotemporal cognition. Srinivasan and Carey (2010), for example, show that the use of metaphorical expressions such as a long time is not prerequisite to associating (spatial) length with duration. In their experiments, nine-month-old infants – like English-speaking adults – associate length and duration, but not, for example, length and tone amplitude (even after controlling for visual cues). A nonverbal study conducted with rhesus monkeys, meanwhile, showed spatial cues to interfere with the processing of temporal information, as well as the reverse (Merritt, Casasanto, and Brannon 2010).
39.4.2 Cultural Foundations of Conceptual Timelines All forms of motion offer a strong experiential basis for associating time and space. Specific subtypes of timeline may also have experiential, cultural, and historical bases. For example, the Yupno conceptualization of time flowing from downhill to uphill may have its roots in the Yupno’s origins at the coast (Nu´n˜ez et al. 2012: 34; cf. also the Yupno ‘entrance schema’ for indoor representations of time). The association of earlier times with the east and later times with the west (as among the Thaayorre, section 39.3.3.3) has a more immediate experiential basis in the daily arc of 1
Although Sinha et al. (2011: 161) claim that the Amondawa “do not employ linguistic space–time mapping constructions,” the polysemous term awo ‘here’ and ‘now’ (Sinha et al. 2011: 150) indicates that there is at least some basic-level association between deictic space and time. Stronger claims have been made regarding the Pirahã, who – per Everett (2005) – lack any linguistic resources for talking about reference-time (as opposed to utterance-time). Everett claims this to be a consequence of a cultural Immediacy of Experience Principle.
647
648
GABY AND SWEETSER
the sun’s (perceived) trajectory across the sky. Both the ego-moving and time-moving metaphors that underlie the sagittal timelines of, for example, English speakers (section 39.3.2.2) reflect our experience of travel and interaction with moving objects (Lakoff and Johnson 1980a). But perhaps the most powerful force-shaping nonlinguistic representations of time today is literacy. We have seen that in highly literate societies, the directionality of the writing system shapes everything from the processing of temporal language, to improvised physical representations of time, to temporal gestures. The effect of the written word on temporal representations is felt even in societies in which it is less than ubiquitous. In Fedden and Boroditsky’s study of how Mian speakers performed in the Temporal Representation Task (section 39.3.1, section 39.3.3.2), for example, “only the number of years of formal education emerged as a significant predictor of temporal arrangement type. Greater number of years of formal education positively predicted left to right arrangements [r(7) = 0.61, p < 0.05] and negatively predicted absolute spatial arrangements [r(7) = −0.65, p < 0.05]” (Fedden and Boroditsky 2012: 7). Ontogenetically, too, it seems that children represent time spatially in alignment with the writing direction they are in the process of acquiring, but only once they are actively producing the written form. Leembruggen, Kelly, and Gaby (in prep.), for example, find three- to four-year-old, English-speaking children to eschew spatial representations of time in the Temporal Representation Task (section 39.3.1). But by the time children are established in school and the literacy practices taught there (age 5.5–7), their representations of time are robustly organized from left to right.2 As we have seen, however, literacy is not the only force shaping spatial representations of time. Among ethnic Thaayorre in Australia’s Cape York Peninsula, for example, monolingual English-speaking participants in the Temporal Representation Task produced uniformly left-to-right timelines, while Kuuk Thaayorre speakers with equivalent levels of literacy produced geo-centric east-to-west and other timelines (in addition to some left-to-right; Gaby 2012). The competition between literacy and spoken metaphors in structuring spatial construals of time is revisited in section 39.4.4 below.
39.4.3 The Importance of Context One common theme emerging from lab-based and cross-linguistic studies alike is the role context plays in determining which of multiple available spatial construals of time is activated on a particular occasion. Context here should be understood in terms of both the particulars of the speech (or thought) event and the more fine-grained subtype of ‘time’ under 2
Cf. also Stites and Özçalişkan’s (2013) study of children’s relative comprehension of moving-time, moving-ego, and sequence-as-position metaphors.
Space–time Mappings Beyond Language
consideration. For example, deictic time (e.g. when an event occurred/ will occur in relation to the moment of speech/thought) is represented along a different axis from sequence time (when events occurred/will occur relative to one another) among signers/speakers of American Sign Language (Emmorey 2002), Aymara (Nu´n˜ez and Sweetser 2006a) and English, for whom signs/gestures along the sagittal axis were more common in referring to deictic time, and gestures along the lateral axis more common in referring to sequence time (cf. section 39.3.2.2). The time scale involved may also plays a role, as Nu´n˜ez and Cooperrider (2013: 225) observe: “the choice [between alternative spatial construals of time] could be modulated by the temporal granularity required – front-back for coarse-grained material and left-right for fine-grained.” The Yupno downhill-to-uphill timeline discussed in section 39.3.3.1, for instance, is for the most part restricted to large timescales (as befits its likely origin in Yupno migration). The Yupno case also illustrates the importance of speech/thought location, since the downhill-to-uphill timeline is supplanted by one locating the past in the direction of the entryway and the future away from it when the speech/thought event takes place indoors (Nu´n˜ez et al. 2012; cf. section 39.3.3.1).
39.4.4 Lingering Questions For all the advances in our understanding of how we humans construe time in spatial terms, there remain a number of open questions. One such puzzle is the absence of any linguistic correlate of the left-to-right timeline that so dominates English (and other) speakers’ representations of time in gesture, written and cultural artifacts, and problem-solving tasks. Yet in no language (of which we are aware) would an expression like two years leftwards be used to refer to two years in the past. One explanation that has been proposed is that while the impact of literacy on structuring timelines is profound, it is also relatively recent. Literacy has become the norm in technologically advanced cultures only within the last century or so, and Western graphical timelines only date back to the eighteenth century (Casasanto and Jasmin 2012, Nu´n˜ez and Cooperrider 2013: 224). Thus, it may be that time is not “metaphorized laterally in language because the cultural artifacts that provide the experiential basis for people’s implicit lateral timelines did not exist – or were not widely used – when our conventions for talking about time were developing” (Casasanto and Jasmin 2012: 669). If this reasoning is correct, we might expect to see left-to-right temporal metaphors emerge in future varieties of English. Alternatively, however, it may be that left-to-right temporal gestures do not really reflect a conceptualization of time as moving along the lateral axis with respect to the gesturer’s viewpoint. Instead, they might represent the forwards motion of some imagined trajector moving along a left-toright trajectory. Thus, a rightwards movement can still be construed as
649
650
GABY AND SWEETSER
forwards with respect to the trajector, depending on whether the motion is framed according to an external or internal perspective. Much as spoken English lacks lateral axis metaphors for time, spoken Kuuk Thaayorre lacks any absolute (east-to-west) temporal metaphors. No such cardinal direction-based temporal metaphors have been documented in a spoken language, despite the fact that for many Australian Aboriginal languages (Kuuk Thaayorre included), these direction terms are of extremely high frequency and cultural importance. This may of course simply reflect an accidental gap in the data, since we have much still to learn about the metaphorical and metonymic extensions of directional terms in Australian languages. Lastly, there is much still be learned by combining the two approaches outlined in this chapter, extending the empirical base of lab-based research to include a more culturally and typologically diverse set of languages and speakers. There are unquestionably financial and practical hurdles to be overcome in bringing the lab to the field, but as our technology advances both of these hurdles descend.
39.5 Conclusion The insights gleaned from the studies reviewed above paint a picture of both unity and diversity in how time is spatially construed. There is unity in the very fact of time being expressed using the vocabulary of space, and conceptualized in terms of space before this vocabulary is even acquired. But there is also considerable diversity in the particulars of how time is spatialized. There is diversity in which axis is recruited for gestural, graphical, and other timelines. There is diversity in which semantic and pragmatic factors condition the choice between multiple available timelines. And there is diversity in how time is differently spatialized in language as opposed to gesture as opposed to cultural artifacts as opposed to nonlinguistic cognition, as evidenced by eye movements, posture, and responses to problem-solving tasks.
40 Conceptualizing Time in Terms of Space: Experimental Evidence Tom Gijssels and Daniel Casasanto
40.1 Introduction Do space–time metaphors in language reflect how space and time are related in speakers’ minds? People often use spatial expressions to talk about time (H. Clark 1973, Lakoff and Johnson 1980a). For instance, you can fall behind on a deadline, go to sleep before midnight, or look forward to a long vacation. Based on these patterns in language, metaphor theorists have argued that space and time are metaphorically related in thought: people use space to talk about time because they use space to think about time (Lakoff and Johnson 1980a, 1999). Skeptics have pointed out, however, that linguistic data alone are not sufficient to support the claim that people think metaphorically (Murphy 1996) – only that they talk metaphorically. Linguistic analyses of space–time metaphors can help to identify possible ways in which people could think about time. Finding out whether people actually think about time in corresponding ways requires the use of nonlinguistic methods, motivating collaboration between linguists and researchers in allied areas of the cognitive sciences (Casasanto 2008b). Performing experimental tests of nonlinguistic space–time mappings in people’s minds is not just a formality to satisfy skeptics, nor an exercise in confirming what linguistic analyses have already made clear. Experiments allow researchers to discover relationships between space and time in the mind that cannot be predicted or explained by language. In this chapter we review evidence for four different relationships between people’s linguistic metaphors and their mental metaphors1 that link space and time: (i) some linguistic metaphors accurately reflect nonlinguistic mental 1
To disambiguate between metaphors in language and in thought, we use the term ‘linguistic metaphor’ to refer to metaphoric expressions in language and ‘mental metaphor’ to refer to associative mappings between nonlinguistic mental representations in metaphorical source and target domains (see Casasanto 2009, 2013 for discussion). This
652
G I J SS EL S A N D C A S A SA N TO
metaphors (section 40.2); (ii) additionally, in some cases, linguistic metaphors not only reflect the way people think but can also play a causal role in determining which mental metaphors people use (section 40.3); (iii) by contrast, people use some mental metaphors that are not reflected in spoken language at all (section 40.4); (iv) finally, people use some mental metaphors that directly contradict the conventional metaphors in their spoken languages (section 40.5). Together, this growing body of experimental research suggests that space–time metaphors in language provide a rich set of hypotheses about people’s spatial conceptions of time, but that understanding the full range of space–time mappings in our minds, and determining which mental metaphors people may be using at any moment, requires looking beyond language.
40.2 When Temporal Language Reflects Temporal Thinking 40.2.1 A Sagittal Mental Timeline Across many languages, temporal sequences are described as unfolding along the sagittal (front–back) axis. People look forward to their retirement, think back to their first kiss, or plan ahead for the arrival of a new baby. In each of these examples, time is described as a line that runs through the speaker’s body with the future being in front of the speaker and the past being in the back. On one proposal, this mapping arises from the universal experience of moving forward through space: when people walk somewhere, places they have already passed lie behind them, and places they have yet to reach lie in front of them (H. Clark 1973). The results from psycholinguistic studies suggest that when people process language about temporal sequences, they activate front–back spatial schemas, at least in certain contexts. In one study, people read sentences describing spatial and temporal sequences with two possible frames of reference. Some sentences described sequences of objects or events as if the reader was moving through space or time (e.g. The flower is in front of me; In March, May is ahead of us). Other sentences described sequences of objects or events moving relative to each other (e.g. The hatbox is in front of the Kleenex; March comes before May). Participants were faster to process temporal sentences after reading spatial sentences with matching as opposed to mismatching frames of reference (Boroditsky 2000; see also Boroditsky and Ramscar 2002, Duffy and Feist 2014). In another study, participants judged whether verbs described past or future events. Participants were faster to make ‘future’ judgments if the verb appeared in front of a silhouette of a face, rather than behind it, whereas ‘past’ distinction becomes particularly important here as we discuss multiple dissociations between linguistic metaphors and mental metaphors.
Conceptualizing Time in Terms of Space
judgments showed the opposite pattern (Torralbo, Santiago, and Lupia´n˜ez 2006, Sell and Kaschak 2011, Ulrich et al. 2012). Two studies from Miles and colleagues suggest that implicitly activating a sagittal mental timeline can affect people’s spontaneous thoughts and motor actions. In one study, participants were given the perceptual illusion of moving forwards or backwards through space by watching dots move toward or away from the center of the screen. Participants who experienced illusory forward motion were more likely to report daydreaming about future events whereas participants who experienced illusory backward motion were more likely to report daydreaming about the past (Miles et al. 2010). In a second study, participants stood in the middle of a room with a motion-tracking device attached to their knee. Half of the participants were asked to imagine a typical day in the past, and the other half were asked to imagine a typical day in the future. Participants who were assigned to think about the past tended to lean backwards, whereas participants assigned to think about the future tended to lean forwards (Miles, Nind, and Macrae 2010). These results suggest that people not only talk but also think about time as being mapped onto a front–back continuum. As we will describe below, however, this is not the only axis people use to think about time – it may not even be the dominant way in which English speakers think about time.
40.2.2 Asymmetric Use of Space to Think about Time Space and time are asymmetrically related in language: people use spatial expressions to talk about time more than vice versa. For instance, English speakers often talk about short careers or extended relationships, but using time to talk about space is much less common. Are nonlinguistic representations of space and time also asymmetrically related? 40.2.2.1 Space–time Asymmetry in Adults Adults use space to think about time, more so than vice versa, as shown by psycholinguistic studies (Boroditsky 2000, Bottini and Casasanto 2010) and nonlinguistic studies (Casasanto and Boroditsky 2008). In a series of nonlinguistic psychophysical experiments, people saw stimuli that varied in their spatial and temporal extents, and estimated either the spatial or temporal magnitude of each stimulus. For example, in some experiments participants saw lines of different lengths ‘growing’ across the screen for different durations. The durations and lengths of the lines were fully crossed, so their temporal and spatial properties were not correlated. On each trial, participants reproduced either the line’s length or its duration. When people reproduced the durations of the lines, they were consistently influenced by task-irrelevant spatial information. The farther the line traveled across the screen, the longer people estimated its duration to be, even though on average all lines had the same duration. However,
653
654
G I J SS EL S A N D C A S A SA N TO
people’s space estimates were not influenced by task-irrelevant temporal information: the time it took for a line to grow across the screen did not affect people’s length estimates. This pattern of asymmetric interference persisted across experiments that changed the attentional, mnemonic, and perceptual demands of the task: space influences time more than vice versa, even when people are not using language (Casasanto and Boroditsky 2008, Bottini and Casasanto 2010, Gijssels et al. 2013). In contrast to human adults, space and time appear to be symmetrically related in monkeys. In another psychophysical study, human subjects and rhesus macaques saw lines of different durations and lengths. Participants saw a line and then judged either its duration or its length. As predicted, human subjects showed the standard space – time asymmetry: when judging durations, people had more difficulty ignoring taskirrelevant spatial information than vice versa. Monkeys, however, showed a symmetrical pattern of interference: when judging durations, they were influenced by task-irrelevant variation in space, as much as vice versa (Merritt, Casasanto, and Brannon 2010). Do these results show that language is a prerequisite for having a space–time asymmetry? Not necessarily. Humans and monkeys differ in many more ways than just their linguistic abilities. Yet these results raise the possibility that the asymmetric mapping between spatial and temporal representations may be uniquely human (Merritt, Casasanto, and Brannon 2010). A space–time asymmetry has now been shown in numerous behavioral studies that were designed expressly to test for it, and to control for factors that could give rise to this pattern spuriously (Boroditsky 2000, Casasanto and Boroditsky 2008, Bottini and Casasanto, 2010, Casasanto, Fotakopolou, and Boroditsky 2010, Bottini and Casasanto 2013, Gijssels et al. 2013, Magnani, Oliveri, and Frassinetti 2014, Open Science Collaboration 2015). In all of these studies, participants judged visually presented stimuli. One study, however, used haptic (i.e. tactile) stimuli, and reported a reversal of the usual space–time asymmetry (Cai and Connell 2015). In one version of the task, based loosely on Casasanto and Boroditsky’s (2008), participants held sticks of varying lengths between their fingertips, while their hands were occluded from view. From the moment they touched the ends of the stick, the experimenter played a tone for a given duration, and participants were instructed to remove their fingers from the stick when the tone stopped. They then reproduced either the spatial length of the stick by holding their hands out and pushing them against a board, or reproduced the duration of the tone by holding down a button. Results showed that the duration of a tone influenced participants’ estimates of how long the stick was, but not vice versa. Although an exception to the usual space–time asymmetry could be of theoretical interest, there are several reasons to be cautious in interpreting these results. First, in Casasanto and Boroditsky’s (2008) tasks, participants were asked to estimate the spatial and temporal
Conceptualizing Time in Terms of Space
dimensions of a single stimulus: temporal and spatial information were conveyed by the same percept of the same physical event (e.g. a line). In Cai and Connell’s task, however, participants estimated the spatial and temporal dimensions of different stimuli (i.e. a stick, a tone), which are difficult, if not impossible, to equate perceptually. Second, in a valid test of a cross-domain asymmetry, the perceptible input should be matched across the domains (e.g. space, time) in every way possible; there should be no asymmetries built into the stimuli. This was not the case in Cai and Connell’s design. In the task described above, spatial information was available via one sensory modality: touch. By contrast, temporal information was available via two sensory modalities: touch and sound. As such, when participants were encoding the stimuli into memory there were two sources of temporal interference (haptic, auditory), but only one source of spatial interference (haptic). Likewise, when reproducing the spatial length of a stick, participants were susceptible to temporal interference from the duration for which they held the stick; but there was no equivalent source of spatial interference when they were reproducing the duration of the tone, since tones have no spatial extent. These asymmetries built into the stimuli and responses could be responsible for the observed ‘backward’ time–space asymmetry. Thus, it remains an open question whether the space–time asymmetry found in experiments with visual stimuli extends to stimuli in other sensory modalities.
40.2.2.2 Space–time Asymmetry in Children By kindergarten, children already show an asymmetric space–time mapping in the mind. In one study, four- to ten-year-olds saw two snails travel across the screen for different distances and durations. Children either judged which snail traveled farther in space (while ignoring duration) or which snail traveled for a longer time (while ignoring distance). Like adults, children showed an asymmetric pattern of space–time interference: children were better at making spatial judgments in the presence of irrelevant temporal information than they were at making temporal judgments in the presence of irrelevant spatial information (Casasanto, Fotakopoulou, and Boroditsky 2010; see Bottini and Casasanto 2013 for similar results in Dutch and Brazilian children). How are space and time related in children’s minds before they acquire language? Prelinguistic children may already have some expectations that spatial and temporal magnitudes should correspond. In two looking-time studies, infants differentiated between stimuli that were matched in size and duration and those that were not matched, suggesting that they expect bigger objects and longer durations to go together (Srinivasan and Carey 2010, de Hevia et al. 2014). Is the space–time mapping in these young children already asymmetric? One study raised this question in infants, and the data were interpreted as
655
656
G I J SS EL S A N D C A S A SA N TO
suggesting that space and time are symmetrically related in their minds. Children were trained to associate a visual feature of the stimuli (e.g. black versus white color) with either greater or lesser magnitude, in one domain (e.g. space). After training, they transferred the association between color and magnitude to another domain (e.g. time). The strength of this transfer effect was similar no matter whether the infants were trained on space and tested on time, or vice versa (Lourenco and Longo 2010). These results show that babies can generalize magnitude mappings from one domain to another. Yet they do not provide a clear answer as to whether space and time are asymmetrically related in infants’ minds for several reasons, the simplest of which is that the lack of a difference between conditions constitutes a null effect (see Bottini and Casasanto 2013 for further discussion). In sum, studies in preschoolers show the same space–time asymmetry as has been found in adults. Studies in infants are less conclusive. They show that space and time are already related in the infant mind, but these data do not provide a clear answer as to whether this mapping is symmetric or asymmetric. It is possible that humans’ earliest conceptions of space and time are symmetric, like monkeys’, and only become asymmetric over the course of cognitive development (Merritt, Casasanto, and Brannon 2010). Srinivasan and Carey (2010) have suggested that using metaphors in language gives rise to the conceptual asymmetry. However, this possibility is difficult to reconcile with the language acquisition data. In general, children appear to use spatial words before they use their temporal equivalents. For instance, they use the preposition in spatially (e.g. in a box) more frequently than temporally (e.g. in a minute; H. Clark 1973). Young children also sometimes misinterpret temporal expressions as having spatial meanings. After seeing scenarios like a toy doll petting a dog, children were asked temporal questions about the events. When asked questions like ‘When did the boy pat the dog?,’ some children responded by giving spatial answers, like ‘over there’ (E. Clark 1971). Finally, children produce spatial uses of words like ‘long’ and ‘short’ earlier than temporal uses of the same words, even though temporal uses are more frequent in the adult input that children receive (Casasanto 2016a, 2016b). The most natural explanation for these patterns is that space is conceptually more fundamental than time from early childhood, and this conceptual space–time asymmetry guides language acquisition.
40.2.2.3
Experimental Evidence for a Space–time Asymmetry: Summary and Implications Together, data from children and adults indicate that the space–time asymmetry found in linguistic metaphors is also found in people’s more basic, nonlinguistic mental representations: across many contexts, people use space to think about time, more than vice versa. This asymmetric relationship is consistent with metaphor theory, but conflicts with the
Conceptualizing Time in Terms of Space
predictions made by an influential neurocognitive model of spatial and temporal magnitude representation. According to A Theory of Magnitude (ATOM), space, time, and other prothetic domains (i.e. domains that can be experienced as ‘more’ or ‘less’ in magnitude) are represented in the mind by a common, analog magnitude metric (Walsh 2003). If this model is correct and space and time rely on the same representations, then there is no a priori reason to assume that these domains should affect each other asymmetrically (Casasanto, Fotakopoulou, and Boroditsky 2010). Yet the behavioral evidence reviewed here argues against this possibility: whereas monkeys show a symmetric space–time relationship that is consistent with ATOM, human children and adults do not. Rather, people mentally represent space and time asymmetrically, as predicted on the basis of their linguistic metaphors.
40.2.3 Cross-linguistic Differences in Linguistic Space–time Metaphors Predict Corresponding Differences in Mental Metaphors The studies reviewed so far suggest that people think about time the way they talk about it in their native language. Yet not all languages use the same kinds of space–time metaphors. Do people who use different space–time metaphors in language use correspondingly different mental metaphors for time? 40.2.3.1
Cross-linguistic Differences in Talking and Thinking about the Past and Future Even when languages agree that temporal sequences unfold along a sagittal axis, they may differ in how they map the past and future onto the front and back ends of the sagittal continuum. In contrast to English and many other Western languages, Aymara talks about the past as being in front of the speaker and the future as being behind: nayra mara ‘front year’ means last year, whereas qhipa marana ‘back year’ means next year. This mapping conflicts with the universal experience of moving forward through space, but corresponds to another universal experience: people can see what is in front of them but not what is behind them. Metaphorically, we can see what has happened in the past, but not what is going to happen in the future. Accordingly, the Aymara word nayra is used to mean both front and eye (Nu´n˜ez and Sweetser 2006b). Do the Aymara think about the past as being in front? The results from one gesture study suggest they do. Aymara speakers were more likely to produce forward gestures when talking about the past, and more likely to produce backward gestures when talking about the future (Nu´n˜ez and Sweetser 2006b; see also Nu´n˜ez et al. 2012). In addition to using the sagittal axis, some languages also use the vertical axis to talk about temporal sequences. Like English, Mandarin frequently
657
658
G I J SS EL S A N D C A S A SA N TO
maps temporal sequences onto a front–back axis, using words like qia´n ‘front’ and ho`u ‘back’ to refer to future and past events. However, Mandarin also commonly uses vertical spatial terms to talk about temporal sequences: earlier and later events are referred to as sha`ng ‘up’ and xia` ‘down’ (Scott 1989, Boroditsky 2001). To test whether English and Mandarin speakers also think about temporal sequences differently, one study primed speakers of these languages with sagittal and vertical spatial scenarios, and then asked participants to judge temporal sentences (e.g. August comes later than June). After seeing a sagittal spatial scenario, English speakers responded faster to sentences describing temporal events, whereas Mandarin speakers responded faster to the same sentences after seeing a vertical spatial scenario (Boroditsky 2001; see also Boroditsky, Fuhrman, and McCormick 2011; but see Chen 2007, January and Kako 2007). Similarly, Mandarin speakers (but not English speakers) were faster to judge whether pictures depicted earlier or later events by pressing a top button for “earlier” responses and a bottom button for “later” responses, rather than the opposite mapping (Boroditsky, Fuhrman, and McCormick 2011; see also Miles et al. 2011).
40.2.3.2
Cross-linguistic Differences in Talking and Thinking about Duration Previous cross-linguistic experiments either required participants to use language overtly or allowed them to use it covertly in ways that could, in principle, give rise to the observed behavioral results. Do people who talk about time in different ways still think about time differently when they are not using language? One study addressed this question by comparing how Greek and English speakers think about duration. Like many other languages, English tends to describe duration in terms of one-dimensional spatial length (e.g. a long time, like a long rope; Alverson 1994, Evans 2004). This unidimensional mapping has been assumed to be universal: a consequence of the unidirectional flight of time’s arrow, and of universal aspects of our bodily interactions with the environment (Clark 1973). In contrast with English speakers, however, Greek speakers tend to express duration in terms of three-dimensional volume or amount, such as poli ora ‘a lot of time’ and poli nero ‘a lot of water.’ Rather than ‘a long night,’ Greek speakers would say ‘a big night’ (megali nychta) to indicate that the night seemed to last a long time. Greek speakers can express duration in terms of linear extent, just as English speakers can make use of volume or amount expressions, but volume metaphors are more frequent and productive in Greek, whereas linear extent metaphors are more frequent and productive in English (Casasanto et al. 2004, Casasanto 2008b, 2010). To test whether Greek and English speakers think about duration differently, speakers of these languages performed a pair of psychophysical duration estimation tasks, with two types of spatial interference. In the distance interference condition, participants saw lines of various
Conceptualizing Time in Terms of Space
durations and lengths travel across the screen. After each line had disappeared, participants reproduced its duration (as in Casasanto and Boroditsky 2008; see section 40.2.2.1). In the amount interference condition, participants saw animations of an empty container ‘filling up’ with liquid to different levels for different durations. After the container disappeared, participants estimated its duration. The experimenters predicted that, if English and Greek speakers tend to think about duration the way they talk about it, using spatial schemas of different dimensionalities, then they should show different patterns of spatial interference in their time estimates: English speakers should show greater interference from irrelevant length information, whereas Greek speakers should show greater interference from irrelevant amount information. Importantly, in both tasks, the spatial extents or amounts were fully crossed with the temporal durations of the stimuli; that is, there was no correlation between the spatial and temporal aspects of the stimuli. This feature of the tasks is critical for supporting the inference that the predicted pattern is not due to participants verbally labeling the relevant dimension of the stimuli during the task: because the lengths and durations of the stimuli were uncorrelated, labeling long-duration lines as ‘long’ and short-duration lines as ‘short’ could only wipe out the predicted spatial interference effect, rather than creating or enhancing it. Results showed that task-irrelevant distance information interfered with English speakers’ duration estimates (more than Greek speakers’): the farther the line traveled across the screen, the longer English speakers estimated its duration to be. By contrast, task-irrelevant amount information interfered with Greek speakers’ duration estimates (more than English speakers’): the fuller a container, the longer Greek speakers estimated its duration to be. Greek and English speakers, who use different spatial metaphors to talk about duration, think about duration in correspondingly different ways, even when they are performing low-level, nonlinguistic tasks (Casasanto et al. 2004).
40.3 When Temporal Language Shapes Temporal Thinking If people who talk about time differently also think about it differently, using different spatial metaphors, does this mean that using space–time metaphors in language influences their mental metaphors for time? Not necessarily. In principle, nonlinguistic factors could cause members of different groups to conceptualize time differently – and therefore to talk about it differently. That is, cross-linguistic differences in verbal metaphors could be either a cause or an effect of cross-group differences in spatial conceptions of time.
659
660
G I J SS EL S A N D C A S A SA N TO
All of the cross-linguistic data reviewed so far come from quasiexperiments: studies in which participants were already members of one group or another (e.g. English versus Greek speakers), prior to the experiment. Quasi-experiments (which, perhaps surprisingly, constitute the majority of the ‘experiments’ in the cognitive sciences) are capable of showing correlations – for example, correlations between the way people talk and the way they think – but they are not capable of establishing causal relationships. In order to determine whether linguistic metaphors can have a causal influence on people’s mental metaphors, it is necessary to perform an experimental intervention in which participants from a single population are randomly assigned to different ‘treatment’ groups (i.e. a randomized-controlled trial [RCT] design). Casasanto and colleagues (2008b, 2010) followed up on their crosslinguistic tests of distance-duration and amount-duration metaphors in language and thought with an RCT experiment. English speakers were assigned to talk about duration using either distance or amount metaphors. Participants completed a fill-in-the-blank questionnaire comparing the durations of events using either distance words (e.g. a meeting is longer than a sneeze) or amount words (e.g. a meeting is more than a sneeze), and then performed the nonlinguistic filling container task from Casasanto et al. (2004). If using space–time metaphors in language is sufficient to influence how people think about time, then training English speakers to talk about duration like Greek speakers should also make them think about duration like Greek speakers. As predicted, English speakers who were trained to talk about duration in terms of amount were affected by task-irrelevant variation in a container’s fullness when estimating its duration, like native Greek speakers. After distance training, however, task-irrelevant amount information did not influence English speakers’ duration estimates (Casasanto 2008b, 2010; see Boroditsky 2001 for compatible results in a psycholinguistic task). Using space–time metaphors in language can influence which mental metaphors people use, causing them to activate spatial schemas of one dimensionality or another to conceptualize duration. These results raise the question: is language creating these space–time mappings, or is linguistic experience modifying pre-linguistic mental metaphors? Prelinguistic infants appear to intuit links between duration and distance (Srinivasan and Carey 2010) and between duration and size (Lourenco and Longo 2010). Thus, both the one-dimensional mapping that is evident in English and the three-dimensional mapping that is evident in Greek may be present in infants’ minds pre-linguistically. Using linguistic metaphors, then, appears to influence temporal thinking by strengthening one of these pre-linguistic space–time associations, and at the same time weakening the alternative association (though not extinguishing it). The presence (and persistence) of both mental metaphors in the same individuals may explain how people are able to switch between them
Conceptualizing Time in Terms of Space
rapidly, on the basis of brief but intense exposure to a dispreferred linguistic metaphor in the laboratory (Casasanto 2008b, 2014, Casasanto this volume Ch. 2, Casasanto and Bottini 2014). Somewhat paradoxically, this training study, which was designed to test whether language influences the way people conceptualize time when they are not using language, yields insights into the way people process verbal space–time metaphors when they are using language. In principle, what appear to be verbal space–time metaphors could instead be instances of polysemy. For instance, people could be processing phrases like long time or more time as having a purely temporal meaning, without activating any spatial representations. Yet the results from the training study rule out this possibility. For English speakers to show the predicted interference effect from spatial amount on their duration estimates, the linguistic stimuli must have caused readers to activate representations of three-dimensional space when comparing event durations, demonstrating that the spatial language was being interpreted metaphorically (i.e. causing sourcedomain representations to be activated online, during language use).
40.4 Space–time Mappings in the Mind That are Absent from Language: Lateral Mental Timelines No known spoken language talks about time as being organized laterally (left–right). English speakers can say that “July comes before August” but not “July comes to the left of August” (Cienki 1998). Yet people consistently represent temporal sequences on a lateral axis. When talking about the order of events, English speakers produce gestures that follow a mental timeline running from left to right: earlier events are more likely to be accompanied by leftwards gestures, whereas later events are accompanied by rightwards gestures (Cooperrider and Nu´n˜ez 2009, Casasanto and Jasmin 2012). This lateral space–time mapping may even be stronger than the sagittal one: English speakers’ spontaneous cospeech gestures mainly follow a left-to-right mapping, even when they are explicitly using sagittal (front–back) metaphors in language (e.g. gesturing left while saying ‘even farther back [in the past]’; Casasanto and Jasmin 2012). Even when people are not gesturing, they still represent the order of events along a left–right timeline. In several experiments, people were faster to judge the relative temporal order of stimuli if the left–right locations of the responses or stimuli was consistent with the direction of their mental timeline (Gevers, Reynvoet, and Fias 2003, Torralbo, Santiago, and Lupia´n˜ez 2006, Santiago et al. 2007). In one experiment, English speakers decided whether famous people became celebrities before or after participants were born. People were faster to respond ‘before’ and ‘after’ when responses were mapped to the left and right than with the opposite
661
662
G I J SS EL S A N D C A S A SA N TO
mapping, as predicted on the direction of their mental timeline (Weger and Pratt 2008).
40.4.1 What Gives Rise to the Lateral Space–time Mapping? Where does the lateral mapping of temporal sequences come from? Across cultures, the direction of people’s mental timeline correlates with the direction in which they read and write. In one experiment, children and adults recreated the order of temporal events (e.g. meals of the day) by placing stickers relative to a reference sticker (e.g. lunch). English speakers, who write from left to right, preferentially placed breakfast to the left of lunch and dinner to the right. Arabic speakers, who write from right to left, showed the opposite pattern: they placed breakfast to the right of lunch and dinner to the left (Tversky, Kugelmass, and Winter 1991). Reaction time studies comparing English and Hebrew speakers (Fuhrman and Boroditsky 2010) and Spanish and Hebrew speakers (Ouellet et al. 2010) find similar results: people who read from left to right map later events to the right, whereas people who read from right to left map later events to the left. Does writing direction play a causal role in shaping the mental timeline? The correlational evidence reviewed above does not support this inference. In principle, a writing system could emerge with one directionality or another as a consequence of culture-specific conceptions of time – not the other way around. Furthermore, groups who write from left to right also tend to spatialize time from left to right on calendars and graphs (Tversky, Kugelmass, and Winter 1991). This covariation leaves open a host of possible scenarios according to which orthography could either play a primary causal role, a mediating role, or no causal role at all in determining the direction of the mental timeline. To find out whether reading experience is sufficient to determine the direction of the mental timeline, Casasanto and Bottini (2014) trained Dutch participants to read in different directions while they judged whether phrases referred to past or future events. For participants in the standard orthography condition, all instructions and stimuli appeared in the usual left-to-right Dutch orthography. However, for participants in the mirror-orthography condition, all instructions and stimuli were mirrorreversed, forcing them to read from right to left. Whereas people in the standard orthography condition responded faster when the locations of the response keys matched a left-to-right mental timeline, this effect reversed for the people in the mirror-reversed condition. After about five minutes of training, people who had read from right to left responded faster when ‘earlier’ was mapped to the right button and ‘later’ to the left button, compared to the opposite mapping. Further experiments showed that Dutch speakers could be trained to conceptualize events in time as flowing upward (after reading upward orthography) or downward (after
Conceptualizing Time in Terms of Space
reading downward orthography). A few minutes of training with a new orthography can redirect the mental timeline, demonstrating a causal role for reading experience. Why does reading experience influence the organization of the mental timeline? For each line of text that people read or write, they move their attention gradually from the left to the right side of the page or the computer screen. Therefore, moving rightward in space is tightly correlated with ‘moving’ later in time, strengthening an association of earlier times with the left side of space and later times with the right (Casasanto and Bottini 2014). Although reading experience alone is sufficient to determine the direction of the mental timeline in the laboratory, beyond the lab it is likely that other culture-specific experiences (e.g. using calendars, lateral gestures for time) also contribute to the way people mentally spatialize sequences of events on the lateral axis.
40.4.2 Why is the Lateral Space–time Mapping Absent from Spoken Language? There is now abundant evidence that people think about sequences of events as unfolding along a lateral axis – so why do they not talk about them using lateral space–time metaphors in language? One possible reason is that spatial metaphors in languages like English developed generations before the age of mass literacy, and of ubiquitous calendars and graphs. Time may not be metaphorized laterally in language because the cultural artifacts that provide the experiential basis for people’s implicit lateral timelines did not exist – or were not widely used – when our conventions for talking about time were developing (Casasanto and Jasmin 2012). If this proposal is correct, then cultures that strongly rely on printed text and on graphical left–right representations of time should start talking about time in lateral spatial terms. One anecdotal example comes from the US army, where the daily schedules for soldiers are shown on a chart with twenty-four columns: one column for each hour of the day, starting with 0000 on the left and ending with 2300 on the right. Each day, the timing of a team’s shift (i.e. the period for which they are assigned to work) is indicated by a laterally oriented rectangle displayed on top of the chart (e.g. a six-hour shift is indicated by a rectangle covering six columns). If a shift gets moved, the rectangle is moved accordingly. One informant reports that when a shift is moved to an earlier slot, people often say they are being ‘shifted left,’ whereas when a shift gets moved to a later slot they are being ‘shifted right’ (Casasanto and Jasmin 2012). Now that many people within language communities share the same experience of reading and writing, and of lateral space–time mappings in graphic conventions, these lateral mappings could begin to infiltrate spoken language.
663
664
G I J SS EL S A N D C A S A SA N TO
40.5 Space–time Mappings in the Mind That Contradict Mappings in Language: The Temporal Focus Hypothesis The lateral mental timeline illustrates a mental metaphor that is not conventionalized in any known spoken language. A mental metaphor in speakers of Darija, a dialect of Modern Standard Arabic, departs even farther from expectations based on linguistic metaphors. In Arabic, as in many languages, the future is ahead and the past is behind. Yet a study by de la Fuente et al. (2014) suggests that some Arabic speakers conceptualize the past as ahead and the future as behind them. To probe Spanish and Moroccan Arabic speakers’ conceptions of time, participants were shown a diagram of a cartoon figure, viewed from the top, with one box in front of him and another box behind him, and asked to indicate which box corresponded to the future and which to the past. Even though both Arabic and Spanish speakers talk about the future as being in front, the groups showed opposite mappings in the diagram task: Spanish speakers tended to put the future event in front and the past event behind, whereas Moroccan Arabic speakers put the past event in front and the future event behind (de la Fuente et al. 2014). Why do Moroccan Arabs think about the past as in front and the future behind them? There is no plausible explanation based on metaphors in the Arabic language, which encodes the opposite space–time mapping. There is also no explanation based on the universal experience of moving forward through space and time as we walk (or ride, drive, etc.; H. Clark 1973). In the absence of any explanation based on language or bodily experience, de la Fuente et al. turned to aspects of Moroccan culture. Compared to many Europeans and Americans, Moroccans tend to focus more on past times and older generations, they are more observant of ancient rituals, and they place more value on tradition, rather than focusing on futurerelated concerns like technological development and globalization (Mateo 2010). De la Fuente et al. hypothesized that this cross-cultural difference in attitudes toward the past and future was responsible for the observed differences in spatial mappings of time. According to the Temporal Focus Hypothesis, people who ‘focus’ on the past metaphorically (i.e. who devote attention to it) should tend to place the past in front of them, in the location where they could focus on the past literally with their eyes if past events were physical objects that could be seen; by contrast, people who ‘focus’ on the future metaphorically should place the future in front of their eyes. De la Fuente et al. (2014) confirmed that Morocans and Spaniards tended to focus more on the past and future, respectively, as indicated by their agreement with past-focused statements (e.g. The young people must preserve the traditions) versus future-focused statements (e.g. Technological and economic advances are good for society) on a questionnaire. Individuals’ degree of
Conceptualizing Time in Terms of Space
past- or future-focus, as indexed by their responses on this questionnaire, predicted whether they would place the past or future in front of them in the cartoon diagram task. To determine whether temporal focus plays a causal role in people’s front–back mappings of time, de la Fuente et al. (2014) experimentally manipulated whether participants were focusing their attention on the past or future, using a writing exercise. People who were assigned to write about past events were subsequently more likely to place the past in front on the cartoon diagram task, whereas people who wrote about future events were more likely to place the future in front. These results show that cultural attitudes toward the past and future can determine how people spatialize time in their minds, and that mental metaphors for time can vary independently of linguistic metaphors.
40.6 Are There People Who Do Not Have Any Space–time Mappings in Language or Thought? In contrast to most other languages that have been attested, Amondawa does not use spatial metaphors to talk about temporal sequences. Even though this Western Amazonian language has a range of temporal expressions to refer to the past and future, to parts of the day, and to seasonal cycles, these expressions do not show any overlap with how Amondawa talks about space (Sinha et al. 2011). Does the absence of linguistic metaphors for temporal sequences mean that the Amondawa do not use space to conceptualize time? Not necessarily. Traditionally, the Amondawa people have not been exposed to cultural artifacts that influence space–time mappings in other populations, like calendars or a writing system. Yet, when asked to visually represent temporal sequences, they still produced a timeline, which was roughly lateral. In two separate tasks, Amondawa speakers were asked to arrange a set of plates on the ground to represent the different parts of the year and the day. For both tasks, participants arranged the plates on a slightly curved line that either ran from left-to-right or right-to-left (Sinha et al. 2011). Therefore, even though the Amondawa do not have any linguistic or cultural conventions for spatializing temporal sequences, they still appear to use space systematically to think about time – at least when prompted to create a spatial diagram. Turning from spatial mappings of temporal succession to mappings of duration, there is preliminary evidence that not all kinds of space–time metaphor are absent from the Amondawa language. In one elicitation task, members of the Amondawa tribe who were bilingual in Amondawa and Portuguese translated Portuguese phrases into Amondawa. The original Portuguese stimuli consisted of spatial phrases describing the length or size of objects (e.g. uma corda comprida ‘a long rope’), and temporal phrases
665
666
G I J SS EL S A N D C A S A SA N TO
describing the duration of events (e.g. uma cantiga comprida ‘a long song’). In their translations, these informants used the same Amondawa words to refer to both spatial length and duration. For instance, the word -puku ‘long’ was used to talk about a long rope as well as to talk about a long conversation (Bottini 2011). Although these data suggest that the Amondawa may use space–time metaphors to talk about duration, these findings should be treated as preliminary. As Bottini (2011) points out, the original Portuguese phrases contained space–time metaphors, which may have biased participants to use Amondawa words for spatial extent to talk about duration. Still, it is notable that when Amondawa informants were asked whether they could use literal translations of Portuguese spatial/ motion metaphors for temporal sequences (e.g. the coming year) in Amondawa, the informants rejected this possibility (Sinha et al. 2011). In sum, although the Amondawa appear not to use space–time metaphors to talk about temporal sequences, they produce roughly linear diagrams upon request, suggesting that they may conceptualize temporal sequences spatially. It remains an open question whether the Amondawa systematically talk and think about duration in terms of spatial length.
40.7 Separating Influences of Language and Culture on Temporal Cognition Researchers sometime question whether ‘language’ and ‘culture’ can ever truly be separated; indeed, language is one highly systematic aspect of human culture. Is it possible to know whether language or nonlinguistic cultural factors are influencing people’s thoughts? Although language and culture may appear inextricable in the world, the previous sections illustrate how these different strands of human experience can be disentangled in the laboratory, for the purpose of identifying the experiential determinants of our thoughts. One strategy for disentangling language and culture is to elicit behavior that is only consistent with one strand of experience, and not with the other. For instance, de la Fuente et al. (2014) showed that Moroccan participants’ spatial conceptions of past and future were incompatible with conventional space–time metaphors in their language. Similarly, numerous studies have shown that people conceptualize time according to a left–right mental timeline that is absent from language. Linguistic experience, therefore, is not a potential basis for these conceptions to time, but cultural attitudes and practices provide potential explanations. In order to determine whether such potential explanations are sufficient to account for a pattern of behavior, a second strategy is needed: performing true experimental interventions, in which participants are drawn from a single population and randomly assigned to ‘treatment’ conditions. For instance, Casasanto and Bottini (2014) manipulated reading direction to
Conceptualizing Time in Terms of Space
show that this cultural practice is sufficient to shape the direction of people’s lateral mental timeline, even if their language remains unchanged. Conversely, Boroditsky (2001) and Casasanto et al. (2004) exposed people to new verbal metaphors to show that linguistic experience can shape conceptions of time, while ruling out any nonlinguistic, cultural explanations. Note that this kind of novel metaphor treatment is different from randomly assigning bilinguals to use one of their languages or another; switching languages may also cause bilinguals to switch mindsets. Real, known languages are embedded in cultures, and using them may cause bilinguals to activate culture-specific ways of thinking, even if they have been randomly assigned to use one language or the other. By assigning people from the same population to novel linguistic or cultural treatments, however, it is possible to isolate the influence of factor of interest, and to rule out the contribution of other confounding factors.
40.8 Conclusions Space and time are closely related in both language and thought. Yet the space–time relationships found in linguistic metaphors are not always the same as those found in people’s nonlinguistic mental metaphors. The spatial metaphors people use to talk about time can reveal how they use space to think about time. In some cases, spatiotemporal language not only reflects the structure of people’s thoughts, but also actively shapes nonlinguistic mental representations of time, causing people to conceptualize temporal sequences using spatial schemas with different orientations (horizontal versus vertical), and to conceptualize duration using spatial schemas of different dimensionalities (one-dimensional versus three-dimensional). Yet language is not the only force that shapes mental metaphors. Cultural practices and artifacts can influence whether people conceptualize events as unfolding along left-to-right or right-to-left mental timelines, neither of which is conventionalized in any known spoken language. Cultural attitudes can influence whether people conceptualize the future as in front of them or as behind them, even though the latter mental metaphor may directly contradict the sagittal mapping of time encoded in a speaker’s language. The emerging body of experimental research on nonlinguistic space–time mappings has taken inspiration from the larger body of descriptive research on space–time metaphors within and across languages. Cognitive linguistics, in particular, has highlighted the centrality of space in language about time, and by inference in temporal thinking. Importantly, however, only some experiments confirm predictions based on linguistic analyses. Other experiments reveal space–time mappings in people’s minds that are absent
667
668
G I J SS EL S A N D C A S A SA N TO
from language, or at odds with their linguistic metaphors. This is good news. It suggests these experiments are serving their most important scientific function, which is not to confirm what we already believed, but rather to challenge those beliefs and generate new insights.
Acknowledgments Research supported by a James S. McDonnell Foundation Scholar Award (#220020236) and an NSF grant (BCS-1257101) to DC, and a PhD grant from the Scientific Research Fund of Flanders to TG.
41 Discovering Spatiotemporal Concepts in Discourse Thora Tenbrink 41.1 Introduction How do we think about space and time, and how is this related to different contexts and situations? To what extent are our thoughts represented in language – and what can we learn from what people say about their current mindset? This chapter addresses these questions by first reviewing findings on conceptual features reflected in spatial and temporal language, and then showing how these can be discovered and analyzed in natural discourse. Speakers’ linguistic choices in discourse reflect their underlying concepts in the given situation, and this enables the analyst to gain insights about relevant cognitive aspects through the analysis of language use. Cognitive linguistics research has produced a wide range of insights on how the linguistic system (across languages) reveals human concepts of space and time. These ubiquitous domains are intricately interrelated, and restricted to a limited range of conceptual patterns that have been subject to diversified analysis. Crucially for the focus of this chapter, both space and time are commonly represented throughout much of natural discourse, as speakers incorporate relevant aspects about these domains in much of their language use. Spatial relationships between people, objects, and locations are regularly conveyed through language, events are linguistically anchored in space and time, different times and events are represented in their relation to each other, and so on. Since there is no simple one-to-one mapping between concepts and linguistic expression, speakers need to choose from the available linguistic repertory to convey the concepts that are relevant in a particular discourse context. Languages differ in the linguistic patterns of space and time vocabulary (Sweetser and Gaby this volume Ch. 38), and also in the ways in which both domains interact and give rise to conceptual and metaphorical structures (Gijssels and Casasanto this volume Ch. 40), resulting in
670
TENBRINK
a specific (more or less extensive) repertory of spatiotemporal expressions. Discourse situations affect which of the available options is best suited for current purposes. Referring to an object by its location, as in it’s the one over there, for instance, is typically simpler than describing where an object is, as in it is on the right-hand side of the table (Vorwerg and Tenbrink 2007). The choice of different formulations for each of these discourse goals (out of the same general repertory) reveals the speakers’ different ways of conceptualizing the same scene for communication purposes. In spite of the fact that space and time are pervasively represented in natural discourse, speakers frequently leave central aspects of a spatiotemporal situation underspecified. Typically, the implied aspects can be inferred from the context; this works in fundamentally different ways for each of the two domains (Tenbrink 2007b). While temporal language use is often associated with notions of sequentiality and causality, spatial language use implies underlying conceptual reference frames that involve a relatum and a perspective, both of which are often not made explicit. These issues will be explored in more detail below. Whenever speakers express concepts of time and space in natural discourse, they reveal which aspects of the domains are relevant to them in a particular context (Tenbrink 2012). Patterns of usage reflecting significant concepts can be detected using systematic analysis methods such as those offered by Cognitive Discourse Analysis (CODA; Tenbrink 2015), which uses controlled empirical settings with experimental variation to elicit systematic differences in conceptualizations as reflected in discourse. This chapter reviews insights gained into spatiotemporal discourse in English by drawing on exploratory as well as controlled empirical studies in the past two decades, introduces CODA as a method, and demonstrates ways in which spatiotemporal concepts can be discovered in discourse.
41.2 Spatial Language in English Discourse In many cultures including English-speaking ones, everyday discourse is full of spatial expressions. On the one hand, we talk about what is happening at particular places, where things are, where we will go next, and the like. Being somewhere, and interacting with things and people around us, is central to our lives, and this determines our conversations. On the other hand, spatial language is also the basis for conceptual (metaphorical) transfer of the kind that is extensively discussed in Part IV in this book, as in expressions like the bad times are behind us. Here we will focus entirely on the literal use of spatial language, and in particular look at the use of terms indicating relative spatial location, such as on, in, to the left, in front, between, and the like. These and various other terms can be used to locate an object (here called locatum) relative to another object (here called
Discovering Spatiotemporal Concepts in Discourse
relatum). Frequently, such relationships are expressed through spatial prepositions (on, in, along) or adjectives (close), although other syntactic forms are possible, such as nouns (closeness, distance) or adverbs (closely), phrases such as to the left, and the occasional conjunction, as in I don’t know where I am (Halliday and Matthiessen 2014). As Talmy (2000a, 2000b) aptly noted, spatial locational terms are highly abstract and schematic. For instance, a term like along makes the schematic assumption that the relatum is linear, as in along a path, in contrast to all over which assumes a planar relatum as in all over the table, or throughout which assumes a volumar relatum as in throughout the aquarium. Our use of these terms in general therefore indicates how we think about space: not in terms of metric distances and exact positions, but rather in terms of relative location. And in specific situations of usage, our choice of these terms indicates how we conceptualize the situation. For instance, the phrase along the aquarium indicates a cognitive focus on the length of the aquarium, not its volume. Objects relate to each other in meaningful ways, and this is reflected not only in the schematic nature of spatial terms, but also in their contextual usage. A term like over indicates not only a geometrically vertical relationship but also a functional relationship between locatum and relatum, as in the umbrella over the man (Coventry, Prat-Sala, and Richards 2001). With above, in contrast, function plays a far less central role than the geometrical relation. Choosing over rather than above in a discourse context, then, points to the relevance of the functional relationship: the umbrella is over, but not above the man would make sense if the umbrella protects the man from the rain even though it is not above him (i.e. the rain comes from a diagonal angle). With another class of terms that are often referred to as projective, the underlying concepts are rather intricate. These include terms like left, right, in front, behind and some other related terms that involve the projection of a spatial direction on a scene. As we saw in the introductory example, using these terms presupposes an underlying perspective, which is not always made explicit. The combination of a perspective with a particular choice of relatum leads to a range of different options of reference systems. To understand the intended use of projective terms, it is necessary to grasp the underlying reference system – even though this is not directly made explicit in English. Levinson (2003) classified the main options for reference systems as intrinsic and relative (plus a third category, absolute, that relates to concepts such as compass directions, which do not involve projective terms). Relative reference systems use three different positions: locatum, relatum, and origin of perspective (typically the observer), as in to the left of the table from my point of view. In intrinsic reference systems, the origin of perspective is identical with the relatum, as in to my left (see Tenbrink 2011 for more options and details).
671
672
TENBRINK
Like other spatial terms, projective terms are schematic, and do not normally indicate a precise spatial location. Nevertheless, there are certain spatial limitations for using these terms. When a projective term is used, the locatum will be positioned within a region surrounding an axis (left/ right/front/back) with respect to the relatum, based on the conceptualization of a reference system (intrinsic/relative). The size of the region depends on contextual factors, but cannot be wider than a half plane. Locatum and relatum are either externally related to each other, or the locatum is inside the relatum. If one of the objects is in motion (or both), interpretation can become complex, but the main basic distinction between intrinsic and relative reference systems remains valid (Tenbrink 2011). Generally, any oriented object can serve as the origin of both intrinsic and relative reference systems; the orientation can come about by (potential) motion (as in a car) or by intrinsic features such as perception (as in living beings). Furthermore, similar effects arise by functional ordering relations such as those induced by a queue (or other kind of sequence). Projective terms can occur across a wide range of contexts. Besides denoting the location of objects in relation to one another in a simple configuration, they are equally prominent in route descriptions, where a goal location is described via reference to streets and landmarks that can be easily identified in the real world. In such a context, salience and dimensions of buildings are prominent in the choice of landmarks, and spatial relations are often sufficiently hinted at via simple and vague expressions. For instance, in turn right at the church the spatial relationship between the path and the church remains rather vague, but the traveler will typically be able to identify the church and the relevant road easily enough. In natural discourse, the reference system that is underlying an utterance can typically not be identified directly on the basis of the linguistic form, since there is no one-to-one correspondence between forms and reference systems. Moreover, spatial utterances often leave at least the perspective implicit, and frequently also the relatum (Tenbrink 2007b) – and both of these are necessary to identify the reference system (Tenbrink 2011). As a result, the same description is often compatible with various possible conceptual reference systems (Tenbrink, Coventry, and Andonova 2011). In summary, spatial terms in English are highly interesting linguistic devices that build on a rich conceptual basis whenever they are used. Discovering their distribution in English discourse highlights important aspects of human conceptualization of spatial relationships.
41.3 Temporal Language in English Discourse Similar to space, human concepts of time are implicitly represented in language. The linguistic means available to speakers, as well as the way
Discovering Spatiotemporal Concepts in Discourse
temporal language is used in natural communication, reflect speakers’ underlying conceptions of the relations between events. However, there is one important difference. While spatial information is completely optional in English, it is impossible to use English grammatically without providing some information about time. Temporal features are conveyed by the main verb through tense and aspect, constituting an indispensable part of any grammatical sentence in English. This allows for expressing intricate temporal relations through variations and combinations of tense and aspect; furthermore, temporal information can be implicitly conveyed via clause order (Dowty 1986, Halliday and Matthiessen 2014). Some nonliteral constructs of (originally) spatial relationships also convey temporal relationships, such as Christmas is ahead of us (Haspelmath 1997, Boroditsky 2000, Moore 2006). Since these have been extensively discussed elsewhere in cognitive linguistic approaches, we focus here on concepts conveyed by literal uses of temporal language in English. On the lexical level, the temporal relational terms before and after (used as prepositions in expressions like before Christmas, and as conjunctions combining clauses as in I woke up after having slept for ten hours) are arguably the most explicit terms available in the language for concepts of temporal relationships. Another frequent explicit term is then, which can convey sequentiality of events. While the repertory of prepositions and adjectives that express relationships directly thus appears rather limited in comparison to the spatial domain (see above), the English language offers a large variety of conjunctions or adverbials such as meanwhile, until, by the time, and so on (Halliday and Matthiessen 2014) that serve to indicate temporal extension and overlap, rather than relating two events to each other. These structural differences in the repertory of linguistic expressions for space as opposed to time reflect fundamentally different conceptualizations of these two domains (Tenbrink 2007b). Since temporal order, as such, can be communicated simply by clause order and other grammatical means, the question arises why speakers would want to make it explicit by using one of the few lexical markers for temporal relationships (before, after). One possibility is that the conceptualized relationship is more specific than a mere temporal precedence relation. Herweg (1991) suggested that the terms need to be understood relative to a contextually dependent proximal time frame, that is, the events need to be sufficiently close to each other in time. This would depend on the conceptually relevant level of granularity (Habel, Herweg, and Pribbenow 1993). Example (1) will be understood on a different level of granularity than example (2): (1) (2)
After lunch, James continued his work. After college, Jill started as a teacher.
In example (1), the proximal time frame of after is determined by the contextual cue lunch, which suggests a scale of minutes or, at best, hours.
673
674
TENBRINK
In contrast, in example (2) the inferred proximal time frame involves weeks or months. However, an increased level of granularity does not generally imply an increased duration of the involved events; they are merely considered and represented at a coarser scale. Apart from the proximal time frame, there may also be an association of immediate succession. In such a case, no other events at the same level of granularity are expected to take place between the two events in question (Miller and Johnson-Laird 1976, Herweg 1991). In example (1), this means that James did not engage in any other substantial activity between lunch and work, and in example (2), Jill did not study or work anywhere else before starting as a teacher. However, these implications can be canceled easily in an appropriate context, as shown in example (3): (3)
After college, Jill started as a teacher, but she had to work as a shop assistant for a couple of years before finding that job.
Another reason for including an explicit temporal term concerns relevance. If the temporal relationship of two known events, or the length of the time span, is in focus as such, it will need to be communicated clearly and explicitly rather than just being implied through clause order or other grammatical devices. Compare the following: (4) (5)
(Mother to son): You know you are not allowed sweets before dinner – only after. (Mother to son): You know you are only allowed sweets having had dinner.
The concept of conveying a focused temporal relationship is called ‘Regulation’ by Tenbrink and Schilder (2003), and it typically cancels the implication of immediate succession. In example (4), sweets presumably are not allowed within a reasonable amount of time before dinner, but will be allowed after dinner, regardless of other events happening in between. With this concept, the temporal term is often stressed in spoken language (as indicated here by italics). Further much-discussed aspects of temporal relational terms concern the association with presuppositional effects of different kinds (e.g. Miller and Johnson-Laird 1976, Lascarides and Oberlander 1993, Tenbrink 2007a). In example (6), the temporal clause is presupposed, that is, assumed as known from the discourse context, and it also remains factual if the main clause is negated, as shown in example (7). (6) (7)
Do you need a meal before you hit the road? No, I don’t need a meal before I hit the road.
Moreover, it has frequently been suggested that explicit temporal terms typically carry additional meanings conveying a sense of connection beyond the temporal relationship itself (e.g. Heina¨ma¨ki 1974, Schilder 2001). To gain a sense of this, consider the following utterances:
Discovering Spatiotemporal Concepts in Discourse (8) (9) (10)
He talked to her. She decided to leave the school. After he talked to her, she decided to leave the school. He talked to her before she decided to leave the school.
The fact that the two events happen in sequential order can be inferred in example (8) through the order of narrating, although other interpretations are possible depending on context. Example (9), however, appears to suggest not only that the events happened sequentially, but also that there might be a causal relation between them: She decided to leave the school because he talked to her. Example (10) seems to be less clear in this respect. One possible interpretation is that he attempted to talk her out of deciding to leave the school, which would be a kind of reverse causality. Similarly, a so-called non-veridical reading (Heina¨ma¨ki 1974, Miller and JohnsonLaird 1976) is conveyed in example (11): (11) (12)
The bomb exploded before hitting the target. The bomb hit the target after exploding.
In example (11) the inference is clear that the bomb did not hit the target – at least not in the expected way. Example (12) demonstrates that this reading is not available with after: despite common sense, the only available interpretation is that the bomb somehow still managed to hit the target after exploding. Another kind of reverse causality suggests itself in example (13), called ‘termination’ by Schilder (2001) because the event in the temporal clause terminates the event in the main clause: (13)
Peter waited for Sue for hours before she finally arrived.
As we have seen, the association of causal and other conceptual relationships seems to differ between the temporal relational terms, and it also seems to be a matter of degree rather than a directly licensed inference. Tenbrink and Schilder (2003) systematically identified the types of nontemporal information conveyed by before, after, and then in a corpus of natural discourse. According to their findings, before and after normally convey additional meanings beyond the temporal relationship, while then mostly conveys a sense of immediate succession – much more strongly so than previously suggested for before and after. The implied discourse relations between the clauses connected by the temporal terms are highly context dependent, although they can typically be readily inferred, as illustrated by the examples given so far. Many of the commonly inferred interpretations are causal in some sense. This fact reflects how humans think of time: not as an abstract domain of unrelated time points, but rather as a network of causally interrelated events (Carston 2002). Accordingly, it is only natural that our linguistic representations of temporal relations are closely related to notions of causality. However, why would speakers wish to use temporal terms to represent causality? In some situations, some amount of freedom of interpretation may be desirable. In example (9) above, the implication
675
676
TENBRINK
that she decided to leave the school because of his talking to her might not be an ‘official’ concept that can be communicated explicitly. Alternatively, speakers might be uncertain about this aspect and wish to leave it open so as not to be held responsible. In such cases, it is convenient to use a formulation that allows for more than one interpretation while still getting the main message concerning temporal order across clearly. Tenbrink (2007b) noted that such conceptual implications remain unparalleled in the spatial domain. While the conceptualized relationship between events often remains implicit in the temporal domain, spatial concepts are normally communicated explicitly. However, unlike the events involved in a verbalized temporal relation, spatial descriptions can leave some of the participants of a spatial relationship implicit, as in the book on the right: this would correspond to an incomplete temporal description such as the event before, which is rather rare in English. Moreover, the underlying conceptual reference system can only be inferred when all participants of the spatial relationship are known (locatum, relatum, origin). In this respect, both temporal and spatial descriptions leave important conceptual aspects implicit, but in very different ways.
41.4 Cognitive Discourse Analysis After having outlined some of the concepts that are conveyed in English discourse representing spatial and temporal relationships, we will now consider how their occurrence can be examined in natural language data. Cognitive Discourse Analysis (CODA, Tenbrink 2015) offers a systematic methodology for this purpose. The main tenet in CODA is that language offers a broad repertory of options from which speakers choose according to what they perceive as relevant and suitable in a discourse context. Therefore, identifying patterns in speakers’ verbalizations in a specific context reveals patterns in their thought and concepts that are relevant in this context. This becomes especially clear through contrast and comparison: changing the discourse context will change the patterns and features of verbalization in systematic ways. This idea is related to previous research in various areas. Much psycholinguistic experimentation, for instance, aims at identifying which of two (or more) available options are chosen under what kinds of circumstances. This kind of research has led to the identification of functional relationships underlying the use of spatial terms such as over (see above; Coventry, Prat-Sala, and Richards 2001), as well as cross-linguistic and cross-cultural differences between uses of reference systems (Levinson 2003). Further, many cognitive linguists investigate structures in language that reflect structures in the mind (e.g. Evans 2009), variously focusing on semantics, grammar, metaphorical transfer processes, or other aspects relevant for
Discovering Spatiotemporal Concepts in Discourse
the relationship between language and thought. What these diverse approaches have in common is the aim of identifying general principles underlying language production or structure. In contrast, CODA aims at highlighting the specific significance of a speaker’s choice in a particular situation in which thought plays a distinct role, showing how general cognitive principles are at work in natural discourse, relative to the situational demands. To achieve this goal, tasks designed to elicit data for CODA allow for unrestricted linguistic choices, aiming for language production in a situation that is as natural as possible – but still controlled with respect to the features of the discourse task, such that changes can be introduced while keeping other factors constant. The language data elicited from speakers in different conditions then exhibit patterns that systematically reflect conceptual differences that are relevant in the given context. Unlike psycholinguistic studies, which are typically more narrowly designed, in CODA-based studies it is generally not the case that specific linguistic choices can be predicted. The researcher elicits discourse in natural (but controlled) situations that are cognitively challenging or otherwise relevant to a research question, and examines the data with respect to conceptually meaningful distinctions and recurring patterns across speakers. In this respect, CODA resembles more discursively oriented analysis methodologies such as Conversation Analysis or Critical Discourse Analysis, where language data are approached with an open mind using a systematic analysis procedure. Similar to these methods, CODA aims at identifying particular features of discourse; in CODA the specific aim is to discover patterns of language that are particularly revealing concerning the ways in which speakers conceptualize a specific, cognitively interesting or challenging situation. Combinations with other modalities or representations of cognitive processes further highlight the kinds of concepts that are decisive for human cognition in the given situation. Since not all cognitive processes can be put into words, it needs to be clarified what kinds of experimental procedures are suitable for eliciting meaningful verbal data, including the tricky question of what counts as meaningful. Dealing primarily with complex problem-solving tasks, Ericsson and Simon (1984) offered insightful answers to these issues. The analysis procedures suggested in their (and related) work rely mainly on the content of verbal protocols, focusing on verbalizations of cognitive processes that the speakers are aware of. The content-based inspection of verbal reports, particularly if carried out by experts in the problem domain and set against a substantial theoretical background (Krippendorff 2004), often leads to well-founded specific hypotheses about the cognitive processes involved. This, in brief, is the state of the art in verbal protocol analysis in cognitive science research. These procedures have been applied in a broad range of areas by researchers across
677
678
TENBRINK
many disciplines concerned with the human mind. However, a close linguistic analysis as suggested in CODA is rarely attempted in this kind of research. The next section will spell out how CODA works in practice for the analysis of spatial and temporal concepts in discourse.
41.5 Discovering Spatial and Temporal Concepts in English Discourse The first step for any CODA approach is to identify a research question with a scope that can be meaningfully addressed by the analysis of language use (Tenbrink 2015). Generally, CODA is suitable to address questions that pertain to mental representation (the conceptualization of complex scenes, event perception, and the like), and complex cognitive processes (such as problem solving or decision making). Although spatial and temporal concepts are nearly ubiquitously represented in natural discourse (at least in English), not all types of language data are suitable for systematic linguistic analysis if no clear relationship to cognitive effects can be established. Research questions suitable for CODA typically aim at discovering particular types of patterns of spatiotemporal concepts that presuppose a certain kind of context in which they become relevant. The most promising approach is to elicit language in a controlled situation where spatial and/or temporal cognition plays a distinct role. In the domain of space, as shown above, speakers frequently leave basic elements of a spatial description implicit. To understand the intended meaning, the analyst needs access to the spatial situation in which the language was produced. Out of context, the reference system underlying a description like the box on the left cannot be identified, since it does not contain information about the perspective or the relatum. Likewise, the functional element of over cannot be detected without taking the actual spatial situation into account. In the domain of time, this is somewhat different since the temporal entities (events or times) represented in temporal language (or grammar) are typically not directly perceptually accessible. As a result, most of the information needed for interpretation is often present within the discourse itself, and speakers talk about temporal relationships in a different way than they talk about object relations (Tenbrink 2007b). Nevertheless, to address specific aspects about human concepts of time relative to a specific situation, it will typically be necessary to elicit data in a controlled way, so as to keep the situation and discourse context constant across speakers and identify patterns of usage that reveal patterns of thinking about time (or space). For instance, analyzing descriptions of (or dialogues about) perceived scenes and events reveals the speakers’ conceptualizations of spatial and temporal relationships within these scenes and events.
Discovering Spatiotemporal Concepts in Discourse
After identifying a suitable research question and scope, the next important step is to clarify the precise way in which data should be collected (Tenbrink 2015), since different kinds of linguistic data, such as verbal protocols, dialogue data, descriptions, or instructions can reveal cognitive aspects in fundamentally different ways. It is therefore worth examining which linguistic data source would most likely lead to most insights of relevance to the research question. Different conditions allow for establishing relevant contrasts, and can be enhanced by collecting more than one type of linguistic data. Note that CODA, in principle, can also be used to analyze existing language data, such as corpus data (e.g. Danino 2014, Egorova, Tenbrink, and Purves 2015); in such a case, data collection means identifying a suitable data set that can be meaningfully analyzed in this way. Relevant research in the domain of space has used situations in which participants were asked to describe object configurations of varying complexity (Ehrich and Koster 1983, Tenbrink, Coventry, and Andonova 2011) or maps (Taylor and Tversky 1996), to communicate object arrangements in dialogue (Schober 1995), to describe routes under various circumstances (Denis 1997, Ho¨lscher, Tenbrink, and Wiener 2011), to classify direction concepts (Mast et al. 2014), to provide a retrospective report of tour planning (Tenbrink and Seifert 2011), and more. In these studies, unrestricted language data were elicited in controlled settings. To exemplify, Tenbrink, Coventry, and Andonova (2011) addressed the main strategies speakers use for describing complex object arrangements. We speculated that strategies might differ depending on the extent to which objects were related to each other in a meaningful way. If functional object relationships mattered for spatial description, speakers would refer to a table with chairs around it in a different way than to a shower with nearby chairs. We furthermore aimed to identify the kinds of reference systems that speakers used in this situation. Previous research had mainly used far simpler configurations, leading to contradictory speculations about reference system preferences in English-speaking cultures. For instance, while Miller and Johnson-Laird (1976) proposed that intrinsic reference systems might generally be predominant, Levinson (2003) suggested that in Western cultures, relative reference systems are preferred. To address these questions, Tenbrink, Coventry, and Andonova (2011) used a controlled monological spoken discourse task. Participants were asked to describe an arrangement consisting of fifteen objects in such a way that a listener would be able to reconstruct the arrangement. In two main conditions, arrangements differed with respect to the availability of functionally meaningful object clusters. The elicited linguistic descriptions were recorded, transcribed, and segmented for purposes of systematic and thorough annotation. Following data collection, Tenbrink (2015) recommends a thorough content analysis as speakers may directly express concepts in language that
679
680
TENBRINK
are relevant to the research question. Content analysis can also lead to the identification of conceptual categories to pursue further, based on a closer linguistic analysis. Tenbrink, Coventry, and Andonova (2011) established through content analysis whether utterances contained reference to the location or the orientation of an object, none of these, or both. This analysis yielded interesting patterns as to the cognitive relevance of object orientation relative to object location. Particularly when orientation seemed to be self-evident in functional object clusters, this information tended to be omitted from description. The most central task step in CODA is linguistic feature analysis (Tenbrink 2015). Typically, this comes with a qualitative and a quantitative aspect. Central linguistic features are detected and analyzed with respect to their (known) relation to cognition, drawing on insights in cognitive linguistics such as those represented throughout this book (Part III and Part IV, and other chapters in the current Part VI), functional grammar (Halliday and Matthiessen 2014), psycholinguistics (Ellis 1985–87), and other resources that highlight the deeper meaning and cognitive significance of particular linguistic features. Systematic annotation then allows for an assessment of the significance of these features within the overall data set, identifying where the relevant features occur, and what characterizes their occurrence. Importantly, any type of data annotation must be reliable, that is, done in such a way that a different observer will come to the same conclusions. This can be ensured, on the one hand, by a suitable operationalization procedure while annotating, and on the other hand, by using one of various available inter-coder reliability test procedures (e.g. Krippendorff’s Alpha as described in Hayes and Krippendorff 2007). In Tenbrink, Coventry, and Andonova (2011), the main aim of systematic annotation concerned the identification of spatial reference frames. However, data inspection revealed frequent cases where speakers left the relatum or origin implicit, or both. For this reason, a thorough identification of all underlying reference systems turned out to be impossible – which may serve to explain why the frequently asked question of general reference system preferences is still not resolved. In order to address meaningfully the research question, each individual object description in the collected corpus was annotated with respect to the mention and identity of the relatum, as well as with respect to compatibility with the actual observer’s perspective on the scene. It was hypothesized that if speakers departed from this default perspective by using a cognitively effortful mental rotation, this would be related to conceptually meaningful aspects of the configuration (since no different addressee perspective was available that speakers might have used instead). Following systematic annotation, inter-coder reliability was established through applying Krippendorff’s Alpha. The next step following annotation concerns the identification of patterns of distribution of the annotated features within the collected data;
Discovering Spatiotemporal Concepts in Discourse
for example, relative to conditions or at certain decisive moments during the task process (e.g. beginning or end, along with an insight or strategy change). These patterns are the main results to be discussed and interpreted in light of an in-depth qualitative analysis of the linguistic features identified as crucial in this context, relative to the research question (Tenbrink 2015). In Tenbrink, Coventry, and Andonova (2011), results showed that speakers systematically preferred using either the environment as a whole (and its features) as a relatum, or other objects mentioned in directly preceding utterances (independent of condition). They normally used their actual (outside) perspective on the scene, except in specific situations where objects were functionally oriented toward each other, or otherwise oriented in a particular way that suggested a different perspective. For instance, one cupboard placed at the left-hand wall was frequently used as a relatum for another object next to it on the wall, using the terms left and right. This suggests that the observer imagined being inside the scene facing the cupboard, rather than using an outside perspective, in which the same objects would have been described as being in front of or behind the cupboard. Altogether, the analysis supported the conclusion that relative reference frames dominated over intrinsic ones in our setting, while at the same time suggesting that specific object configuration could induce other types of reference frames. This calls into question the assumption that a general, situation-independent preference for a reference frame can be identified for a culture. The analysis of linguistic data can often be meaningfully combined with other types of data, such as behavioral performance data, reaction times, and the like. In this final step, the CODA results are related to such other data, or possible extensions such as formalizations or modeling procedures are initiated (Tenbrink 2015). For the study reported in Tenbrink, Coventry, and Andonova (2011), a straightforward extension would have been to present another set of participants with the descriptions in order to establish the communicative success of specific strategies. This was, however, left for future investigation. In the temporal domain, research has mainly focused on the discussion of conceptual phenomena found in discourse examples, rather than eliciting natural language data to examine temporal concepts. This includes research on discourse relations including the phenomena discussed above (Lascarides and Oberlander 1993, Tenbrink and Schilder 2003, Tenbrink 2007a), space-to-time mappings (Haspelmath 1997, Boroditsky 2000, Moore 2006), temporal reference frames (Tenbrink 2011, Chilton 2013), and more. Typically in research of this kind, discourse examples are used to explore concepts and phenomena underlying language use in general. Closer to CODA, research involving elicited data has addressed, for instance, the acquisition of temporal structuring devices in second language learners (Bardovi-Harlig 2000), use of tense and modality in think-aloud protocols at different stages of a problem-solving
681
682
TENBRINK
process (Gralla 2013), and use of discourse markers indicating the temporal development of a problem-solving process (Caron-Pargue and Caron 1991, Tenbrink and Wiener 2009). Further relevant work addressed the ways in which events are structured and segmented in narratives, in relation to cognitive aspects (Zacks, Tversky, and Iyer 2001) and in bilinguals (Bylund 2011). However, to my knowledge the issues explored above concerning nontemporal conceptual aspects conveyed by temporal terms have not been addressed using specifically elicited natural language data. A number of predictions emerge from example-based findings in this area, which could be empirically tested. For instance, if (as hypothesized) temporal connectives are preferably used in situations where a causal connection is plausible but not evident, this should lead to systematic patterns in usage (depending on context) that reflect the extent to which a causal relationship is conceptualized by the speaker. Likewise, if after and before differ with respect to presuppositional patterns and associated conceptual relationships, their distribution in natural discourse should differ in ways going far beyond the abstract distinction with respect to temporal order, reflecting their distinct roles in representing temporal cognition. Further promising avenues of research concern the interplay of spatial and temporal concepts in natural discourse, as most empirical work in this area so far has focused on either one of the domains rather than explicitly targeting their interrelations. Tenbrink (2007b) pointed to a range of parallels and contrasts between general principles of language use in the spatial versus temporal domain, which would lend themselves to further empirical exploration. For instance, as exemplified above, each of the domains leaves central aspects underspecified in spatiotemporal discourse, namely the conceptualized relationship between the entities (events) in the temporal domain, as opposed to the entities themselves (relatum and origin) in the spatial domain. These effects are related to the different discourse functions played by spatial versus temporal terms that express relations between entities in each domain, and can ultimately be traced back to the fundamental differences in the linguistic repertory for each domain (in English). These effects should result in systematically different usage patterns and conceptualization effects in discourse where both types of relationships are equally relevant, or relevant in different ways. In the vast literature on route directions, both spatial and temporal concepts play a role to some extent, but the main focus is on spatial relationships. Arguably, this reflects the route givers’ conceptual focus on the spatial rather than the temporal domain in a motion setting that involves both. It would be interesting to explore scenarios in which the focus of attention shifts more flexibly between space and time, revealing different patterns of discourse to represent each domain.
Discovering Spatiotemporal Concepts in Discourse
41.6 Conclusion In this chapter, I have reviewed some ways in which spatial and temporal terms in their natural occurrence in discourse convey how humans conceptualize (and, as a result, verbalize) the domains of space and time. To discover the significance of their use in a specific situation context, it is useful to adopt a systematic approach to data collection and analysis from a cognitive perspective, as done in Cognitive Discourse Analysis (Tenbrink 2015). Research using this and related methods extends traditional discourse and content analysis methods by integrating cognitive linguistic findings about the conceptual significance of specific linguistic patterns. This leads to insights about human thought processes (as reflected in discourse) that the speakers themselves may not be consciously aware of, such as the perspective underlying a spatial description, or the non-temporal concepts conveyed when using temporal terms.
683
References
Achard, Michel. 2008. Teaching construal: cognitive pedagogical grammar. In R. Robinson and N. C. Ellis (eds.), Handbook of cognitive linguistics and second language acquisition, 432–55. London: Routledge. Ackerman, Farrell, James P. Blevins, and Robert Malouf. 2009. Parts and wholes: implicative patterns in inflectional paradigms. In Jim P. Blevins and Juliette Blevins (eds.), Analogy in grammar, 54–82. Oxford University Press. Ackerman, Joshua, Christopher C. Nocera, and John A. Bargh. 2010. Incidental haptic sensations influence social judgments and decisions. Science 328, 1712–15. Ahlers, Jocelyn. 1999. Proposal for the use of cognitive linguistics in Hupa language revitalization. PhD dissertation. University of California at Berkeley. Aijmer, Karin. 2005. Conversational routines in English: convention and creativity. London: Addison Wesley Longman. Aikhenvald, Alexandra Y. 2004. Evidentiality. Oxford University Press. 2008. Semi-direct speech: Manambu and beyond. Language Sciences 30, 383–422. Akatsuka, Noriko. 1999. Towards a theory of desirability in conditional reasoning. In A. Kamio and K. Takami (eds.), Function and structure: in honor of Susumu Kuno, 195–213. Amsterdam: John Benjamins. Akhtar, Nameera. 1999. Acquiring basic word order: evidence for data-driven learning of syntactic structure. Journal of Child Language 26, 339–56. Alexander, Marc, and Ellen Bramwell. 2014. Mapping metaphors of wealth and want: a digital approach. Studies in the Digital Humanities 1, 1–19. Alibali, Martha W., Dana C. Heath, and Heather J. Myers. 2001. Effects of visibility between speaker and listener on gesture production: some gestures are meant to be seen. Journal of Memory and Language 44, 169–88.
References
Alibali, Martha W., Miriam Bassok, Karen O. Solomon, Sharon E. Syc, and Susan Goldin-Meadow. 1999. Illuminating mental representations through speech and gesture. Psychological Science 10(4), 327–33. Alibali, Martha W., Robert C. Spencer, Lucy Knox, and Sotaro Kita. 2011. Spontaneous gestures influence strategy choices in problem solving. Psychological Science 22(9), 1138–44. Allan, Kathryn. 2008. Metaphor and metonymy: a diachronic approach. Oxford: Wiley-Blackwell. Al-Sharafi, Abdul. 2004. Textual metonymy: a semiotic approach. Basingstoke: Palgrave MacMillan. Alverson, Hoyt. 1994. Semantics and experience: universal metaphors of time in English, Mandarin, Hindi, and Sesotho. Baltimore: Johns Hopkins University Press. Amberger, Mengistu (ed.). 2007. The language of memory in a crosslinguistic perspective. Amsterdam: John Benjamins. Ambridge, Ben, and Adele E. Goldberg. 2008. The island status of clausal complements: evidence in favor of an information structure explanation. Cognitive Linguistics 19, 357–89. Ambridge, Ben, and C. F. Rowland. 2009. Predicting children’s errors with negative questions: testing a schema-combination account. Cognitive Linguistics 20(2), 225–66. Ambridge, Ben, E. Kidd, C. F. Rowland, and A. L. Theakston. 2015. The ubiquity of frequency effects in first language acquisition. Journal of Child Language 42(02), 239–73. Ambridge, Ben, Julian M. Pine, Caroline F. Rowland, Daniel Freudenthal, and Franklin Chang. 2012. Avoiding dative overgeneralisation errors: semantics, statistics, or both? Language and Cognitive Processes 29(2), 218–43. Ambridge, Ben, Julian M. Pine, C. F. Rowland, F. Chang, and A. Bidgood. 2013. The retreat from overgeneralization in child language acquisition: word learning, morphology, and verb argument structure. WIREs Cognitive Science 4(1), 47–62. Anderson, John R. 2000. Cognitive psychology and its implications. New York: W. H. Freeman. 2014. Cognitive psychology and its implications, 8th edn. New York: Worth. Anderson, Michael L. 2010. Neural reuse: a fundamental organizational principle of the brain. Behavioral and Brain Sciences 33, 245–313. Anderwald, Lieselotte, and Bernd Kortmann. 2013. Applying typological methods in dialectology. In Manfred Krug and Julia Schlu¨ter (eds.), Research methods in language variation and change, 313–33. Cambridge University Press. Andre´n, Mats. 2010. Children’s gestures between 18 and 30 months. Lund: Media Tryck. Anketa. 1997. Anketa aspektologicˇeskogo seminara filologicˇeskogo fakulteta MGU. Trudy aspektologicˇeskogo seminara filologicˇeskogo fakulteta MGU, vol. 2.
685
686
References
Anscombre, Jean-Claude, and Oswald Ducrot. 1983. L’argumentation dans la langue. Bruxelles: P. Mardaga. Antic´, Eugenia. 2012. Relative frequency effects in Russian morphology. In Stefan Th. Gries and Dagmar S. Divjak (eds.), Frequency effects in language learning and processing, 83–107. Berlin: De Gruyter Mouton. Antilla, Raimo. 2008. Analogy: the warp and woof of cognition. In Brian D. Joseph and Richard Janda (eds.), Blackwell handbook of historical linguistics, 425–40. Oxford: Blackwell. Archangeli, Diana, 1988. Aspects of Underspecification Theory. Phonology 5(2), 183–208. Arendholz, Jenny, Wolfram Bublitz, Monika Kirner, and Iris Zimmermann. 2013. Food for thought – or, what’s (in) a recipe? A diachronic analysis of cooking instructions. In Cornelia Gerhardt, Maximiliane Frobenius, and Susanne Ley (eds.) Culinary linguistics: the chef’s special, 119–37. Amsterdam: Benjamins. Armstrong, David F. 1983. Iconicity, arbitrariness, and duality of patterning in signed and spoken language: perspectives on language evolution. Sign Language Studies 38, 51–69. Armstrong, David F., and Sherman Wilcox. 2007. The gestural origin of language. Oxford University Press. Armstrong, David F., William C. Stokoe, and Sherman E. Wilcox. 1994. Signs of the origin of syntax. Current Anthropology 35(4), 349–68. 1995. Gesture and the nature of language. Cambridge University Press. Arppe, Antti. 2008. Univariate, bivariate and multivariate methods in corpus-based lexicography – a study of synonymy. PhD dissertation. University of Helsinki. Athanasiadou, Amgeliki, and Rene´ Dirven. 1997. Conditionality, hypotheticality, counterfactuality. In A. Athanasiadou and R. Dirven (eds.), On conditionals again, 61–96. Amsterdam: Philadelphia. Atkins, Sue, Michael Rundell, and Hiroaki Sato. 2003. The contribution of FrameNet to practical lexicography. International Journal of Lexicography 16(3), 333–57. Auer, Peter and Stefan Pfa¨nder (eds.). 2011. Constructions: emerging and emergent. Berlin: de Gruyter. Austin, John L. 1961. The meaning of a word. In J. O. Urmson and Geoffrey Warnock (eds.), Philosophical papers, 55–75. Oxford University Press. 1962. How to do things with words. Oxford: Clarendon Press. Aziz-Zadeh, Lisa, Stephen M. Wilson, Giacomo Rizzolatti, and Marco Iacoboni. 2006. Congruent embodied representations for visually presented actions and linguistic phrases describing actions. Current Biology 16(18), 1818–23. Baayen, R. Harald. 1992. Quantitative aspects of morphological productivity. In G. Booij and J. V. Marle (eds.), Yearbook of Morphology 1991, 109–50. Dordrecht: Kluwer.
References
2008. Analyzing linguistic data: a practical introduction to statistics using R. Cambridge University Press. 2009. Corpus linguistics in morphology: morphological productivity. In Anke Lu¨deling and Mirya Kyto¨ (eds.), Corpus linguistics: an international handbook, 900–17. Berlin: De Gruyter. 2011. Corpus linguistics and naive discriminative learning. Brazilian Journal of Applied Linguistics 11, 295–328. Baayen, R. Harald, Ton Dijkstra, and Robert Schreuder. 1997. Singulars and plurals in Dutch: evidence for a parallel dual route model. Journal of Memory and Language 36, 94–117. Baayen, R. Harald, P. Milin, D. Filipovic Durdjevic, P. Hendrix, and M. Marelli. 2011. An amorphous model for morphological processing in visual comprehension based on naive discriminative learning. Psychological Review 118, 438–82. Baayen, R. Harald, Anna Endresen, Laura A. Janda, Anastasia Makarova, and Tore Nesset. 2013. Making choices in Russian: pros and cons of statistical methods for rival forms. Space and time in Russian temporal expressions. Special issue of Russian Linguistics 37(3), 253–91. Backus, Ad, and Maria Mos. 2011. Islands of productivity in corpus data and acceptability judgments: contrasting two potentiality constructions in Dutch. In Doris Scho¨nefeld (ed.), Converging evidence: methodological and theoretical issues for linguistic research, 165–92. Amsterdam: John Benjamins. Badarneh, Muhammad A. 2003. The rhetorical question as a discursive and stylistic device in the Quran. PhD Dissertation. Arizona State University. Bakeman, Roger, and Lauren B. Adamson. 1984. Coordinating attention to people and objects in mother-infant and peer-infant interaction. Child Development 55(4), 1278–89. Baker, Collin. 2012. FrameNet, current collaborations and future goals. Language Resources and Evaluation 46, 269–86. 2014. FrameNet: a knowledge base for natural language processing. Proceedings of Frame Semantics in NLP: a workshop in honor of Chuck Fillmore at the 52nd Annual Meeting of the Association of Computational Linguistics (ACL 2014), 1–5, Baltimore, Maryland. Baker, Collin, and Josef Ruppenhofer. 2002. FrameNet’s frames vs. Levin’s verb classes. In J. Larson and M. Paster (eds.), Proceedings of the 28th Annual Meeting of the Berkeley Linguistics Society, 27–38. Berkeley, CA: University of California, Berkeley Linguistics Department. Baker, Collin, Charles J. Fillmore, and Beau Cronin. 2003. The structure of the FrameNet database. International Journal of Lexicography 16, 281–96. Bakhtin, Mikhail M. 1963 [1984]. Problems of Dostoevsky’s poetics. Trans. Caryl Emerson. Minneapolis: University of Minnesota Press.
687
688
References
1975 [1981]. The dialogic imagination. Ed. Michael Holquist, trans. Caryl Emerson and Michael Holquist. Austin: University of Texas Press. 1979 [1986]. Speech genres and other late essays. Trans. V. W. McGee. Austin: University of Texas Press. Bal, Mieke. 2009. Narratology: introduction to the theory of narrative, 3rd edn. University of Toronto Press. Banfield, Ann. 1982. Unspeakable sentences: narration and representation in the language of fiction. Boston: Routledge and Kegan Paul. Bangerter, Adrian. 2004. Using pointing and describing to achieve joint focus of attention in dialogue. Psychological Science 15, 415–19. Bannard, Colin, and Danielle Matthews. 2008. Stored word sequences in language learning: the effect of familiarity on children’s repetition of four-word combinations. Psychological Science 19(3), 241–48. Ba´nre´ti, Zolta´n. 2010. Recursion in aphasia. Clinical Linguistics and Phonetics 24(11), 906–14. Banville, John. 2005. The sea. London: Picador. Baraba´si, Albert-La´szlo´. 2003. Linked: how everything is connected to everything else and what it means. New York: Plume. Barcelona, Antonio (ed.). 2000. Metaphor and metonymy at the crossroads. Berlin: De Gruyter Mouton. 2003. The case for a metonymic basis of pragmatic inferencing: evidence from jokes and funny anecdotes. In Klaus-Uwe Panther and Linda Thornburg (eds.), Metonymy and pragmatic inferencing, 81–102. Amsterdam: John Benjamins. 2010. Metonymic inferencing and second language acquisition. AILA Review 23, 134–55. 2011. Reviewing the properties and prototype structure of metonymy. In Reka Benczes, Antonio Barcelona, and Francisco Ruiz de Mendoza (eds.), Defining metonymy in cognitive linguistics: towards a consensus view, 7–57. Amsterdam: John Benjamins. Barðdal, Jo´hanna. 2008. Productivity: evidence from case and argument structure in Icelandic. Amsterdam: John Benjamins. 2011. Lexical vs. structural case: a false dichotomy. Morphology 21(1), 619–54. 2013. Construction-based historical-comparative reconstruction. In Thomas Hoffmann and Graeme Trousdale (eds.), The Oxford handbook of construction grammar, 438–57. Oxford University Press. Barðdal, Jo´hanna, and Thorhallur Eytho´rsson. 2012. Reconstructing syntax: construction grammar and the comparative method. In Hans C. Boas and Ivan A. Sag (eds.), Sign-based construction grammar, 261–312. Stanford: CSLI Publications. Bardovi-Harlig, Kathleen. 2000. Tense and aspect in second language acquisition: form, meaning and use. Oxford: Blackwell. Barlow, Michael, and Suzanne Kemmer (eds.). 2000. Usage-based models of language. Stanford: CSLI Publications.
References
Barnden, John. 2010. Metaphor and metonymy: making their connections more slippery. Cognitive Linguistics 21(1), 1–34. Barr, Dale J., and Boaz Keysar. 2004. Making sense of how we make sense: the paradox of egocentrism in language use. In H. Colston and A. Katz (eds.), Figurative language processing: social and cultural influences, 21–41. Mahwah, NJ: Erlbaum. Barsalou, Lawrence W. 1999. Perceptual symbol systems. Behavioral and Brain Sciences 22, 577–660. 2008. Grounded cognition. Annual Review of Psychology 59, 617–45. 2009. Simulation, situated conceptualization and prediction. Philosophical Transactions of the Royal Society B 364, 1281–89. Barsalou, Lawrence W., and Katja Wiemer-Hastings. 2005. Situating abstract concepts. In D. Pecher and R. Zwaan (eds.), Grounding cognition: the role of perception and action in memory, language, and thought, 129–61. New York: Cambridge University Press. Barth, Danielle, and Vsevolod Kapatsinski. 2014. A multimodel inference approach to categorical variant choice: construction, priming and frequency effects on the choice between full and contracted forms of am, are and is. Corpus Linguistics and Linguistic Theory 2, 1–58. Bartlett, Frederic C. 1932 [1967]. Remembering: a study in experimental and social psychology. Cambridge University Press. Bates, E., L. Benigni, I. Bretherton, L. Camioni, and V. Volterra. 1979. The emergence of symbols: cognition and communication in infancy. New York: Academic Press. Bates, E., D. Thal, D. Trauner, J. Fenson, D. Aram, J. Eisele, and R. Nass. 1997. From first words to grammar in children with focal brain injury. Developmental Neuropsychology 13, 275–343. Bates, E., J. Reilly, B. Wulfeck, N. Dronkers, M. Opei, J. Fenson, et al. 2001. Differential effects of unilateral lesions on language production in children and adults. Brain and Language 79, 223–65. Battison, Robbin. 1978. Lexical borrowing in American Sign Language. Silver Spring, MD: Linkstok Press. Baudouin de Courtenay. 1972. An attempt at a theory of phonetic alternations. In Edward Stankiewicz (ed.), A Baudouin de Courtenay anthology: the beginnings of Structural Linguistics, 144–213. Bloomington: Indiana University Press. Bauer, Laurie, Rochelle Lieber, and Ingo Plag. 2013. The Oxford reference guide to English morphology. Oxford University Press. Bavelas, Janet B. 2005. The two solitudes: reconciling social psychology and social interaction. In K. L. Fitch and R. E. Sanders (eds.), Handbook of language and social interaction, 179–200. Mahwah, NJ: Lawrence Erlbaum Associates. Bavelas, Jane B., Linda Coates, and Trudy Johnson. 2002. Listener responses as a collaborative process: the role of gaze. International Communication Association 52, 566–80.
689
690
References
Bavelas, Janet B., Nicole Chovil, Douglas A. Lawrie, and Allan Wade. 1992. Interactive gestures. Discourse Processes 15, 469–89. Bavelas, Janet B., Jennifer Gerwing, Chantelle Sutton, and Danielle Prevost. 2008. Gesturing on the telephone: independent effects of dialogue and visibility. Journal of Memory and Language 58(2), 495–520. Bavelier, D., D. Corina, P. Jezzard, V. Clark, A. Karni, A. Lalwani, et al. 1998. Hemispheric specialization for English and ASL: left invariance-right variability. Neuroreport 9(7), 1537–42. Beattie, Geoffrey, and Heather Shovelton. 2002. An experimental investigation of some properties of individual iconic gestures that mediate their communicative power. British Journal of Psychology 93(2), 179–92. Beaudoin-Ryan, Leanne, and Susan Goldin-Meadow. 2014. Teaching moral reasoning through gesture. Developmental Science 17(6), 984–90. Bechtel, William, and Adele Abrahamsen. 2002. Connectionism and the mind: parallel processing, dynamics, and evolution in networks. Oxford: Blackwell. Beck, Sigrid. 1997. On the semantics of comparative conditionals. Linguistics and Philosophy 20(3), 229–71. Beckner, Clay, Richard Blythe, Joan Bybee, Morten H. Christiansen, William Croft, Nick C. Ellis, et al. 2009. Language is a complex adaptive system: position paper. Language Learning 59, 1–26. Beeke, Suzanne, Jane Maxim, and Ray Wilkinson. 2007. Using conversation analysis to assess and treat people with aphasia. Seminars in Speech and Language 28(2), 136–47. Beekhuizen, Barend, Rens Bod, and Willem Zuidema. 2013. Three design principles of language: the search for parsimony. Language and Speech 56, 265–90. Behrens, Heike. 2006. The input-output relationship in first language acquisition. Language and Cognitive Processes 21, 2–24. Beilock, Sian L., and Susan Goldin-Meadow. 2010. Gesture changes thought by grounding it in action. Psychological Science 21(11), 1605–10. Bell, Alan, Daniel Jurafsky, Eric Fosler-Lussier, Cynthia Girand, Michelle Gregory, and Daniel Gildea. 2003. Effects of disfluencies, predictability, and utterance position on word form variation in English conversation. Journal of the Acoustical Society of America 113(2), 1001–24. Bencini, Giulia M. L. 2013. Psycholinguistics. In Thomas Hoffmann and Graeme Trousdale (eds.), The Oxford handbook of construction grammar, 379–96. Oxford University Press. Bencini, Giulia M. L., and Adele E. Goldberg. 2000. The contribution of argument structure constructions to sentence meaning. Journal of Memory and Language 43, 640–51. Bencini, Giulia M. L., and Virginia Valian. 2008. Abstract sentence representations in 3-year-olds: evidence from comprehension and production. Journal of Memory and Language 59, 97–113.
References
Benczes, Re´ka, Antonio Barcelona, and Francisco Ruiz de Mendoza (eds.). 2011. Defining metonymy in cognitive linguistics: towards a consensus view. Amsterdam: John Benjamins. Bender, Andrea, Sieghard Beller, and Giovanni Bennardo. 2010. Temporal frames of reference: conceptual analysis and empirical evidence from German, English, Mandarin Chinese, and Tongan. Journal of Cognition and Culture 10, 283–307. Bennett, Jonathan. 2003. A philosophical guide to conditionals. Oxford University Press. Berber-Sardinha, Tony. 2010. A program for finding metaphor candidates in corpora. The Especialist (PUCSP) 31, 49–68. Bere´ndi, Marta, Szilvia Csa´bi, and Zolta´n Ko¨vecses. 2008. Using conceptual metaphors and metonymies in vocabulary teaching. In Frank Boers and Seth Lindstromberg (eds.), Cognitive linguistic approaches to teaching vocabulary and phraseology, 65–100. Berlin: De Gruyter Mouton. Berez, Andrea L., and Stefan Th. Gries. 2010. Correlates to middle marking in Dena’ina iterative verbs. International Journal of American Linguistics 76(1), 145–65. Bergen, Benjamin K. 2003. Towards morphology and agreement in Embodied Construction Grammar. www.cogsci.ucsd.edu/~bkbergen/ papers/ECGmorph.pdf. 2004. The psychological reality of phonaesthemes. Language 80(2), 290–311. 2007. Experimental methods for simulation semantics. In Monica Gonzalez-Marquez, Irene Mittelberg, Seana Coulson, and Michael J. Spivey (eds.), Methods in cognitive linguistics, 277–301. Amsterdam: John Benjamins. 2012. Louder than words: the new science of how the mind makes meaning. New York: Basic Books. Bergen, Benjamin, and Nancy C. Chang. 2005. Embodied construction gram¨ stman mar in simulation-based language understanding. In Jan-Ola O and Mirjam Fried (eds.), Construction grammars: cognitive and cross-language dimensions, 147–90. Amsterdam: John Benjamins. 2013. Embodied construction grammar. In T. Hoffman and G. Trousdale (eds.), Oxford handbook of construction grammar, 168–90. New York: Oxford University Press. Bergen, Benjamin, and T. T. Chan Lau. 2012. Writing direction affects how people map space onto time. Frontiers in Psychology 3, 109. Bergen, Benjamin, and Jerome Feldman. 2008. Embodied concept learning. In Paco Calvo and Toni Gomila (eds.), Handbook of cognitive science: an embodied approach, 313–31. San Diego: Elsevier. Bergen, Benjamin, and Madelaine Plauche´. 2001. Voila` voila`: extensions of deictic constructions in French. In Alan Cienki, Barbara Luka, and Michael Smith (eds.), Conceptual and discourse factors in linguistic structure, 238–49. Stanford: CSLI Publications.
691
692
References
Bergmann, Kirsten, and Stefan Kopp. 2012. Gestural alignment in natural dialogue. In R. P. Cooper, D. Peebles, and N. Miyake (eds.). Proceedings of the 34th Annual Conference of the Cognitive Science Society (CogSci 2012), 1326–31. Austin: Cognitive Science Society. Bergmann, Till, and Teenie Matlock. In prep. Fictive motion in the wild: discourse data from the TV News Archive in the wild. Bergs, Alexander. 2008. Shall and shan’t in Contemporary English. In Graeme Trousdale and Nick Gisbourne (eds.), Constructional explanations in English grammar, 113–44. Berlin and New York: De Gruyter Mouton. 2010. Expressions of futurity in contemporary English: a Construction Grammar perspective. English Language and Linguistics 14(2), 217–38. 2012a. Construction Grammar. In Alexander Bergs and Laurel Brinton (eds.), English historical linguistics, 1631–46. Vol 2. Berlin: de Gruyter Mouton. 2012b. The uniformitarian principle and the risk of anachronisms in language and social history. In Juan Manuel Herna´ndez-Campoy, and Juan Camilo Conde-Silvestre (eds.), The Blackwell handbook of historical sociolinguistics, 80–98. Oxford: Blackwell. Bergs, Alexander, and Gabriele Diewald (eds.). 2008. Constructions and language change. Berlin: De Gruyter Mouton. 2009. Contexts and constructions. Amsterdam: John Benjamins. Bergs, Alexander, and Meike Pentrel. 2015. Ælc þara þe þas min word gehierþ and þa wyrcþ. . .: Psycholinguistic perspectives on early English. In Michael Adams, Robert D. Fulk, and Laurel J. Brinton (eds.), Studies in the history of the Englishl VI: Evidence and method in histories of English, 249–78. Berlin: De Gruyter Mouton. Berko, Jean. 1958. The child’s learning of English morphology. WordJournal of the International Linguistic Association 14(2–3), 150–77. Berlin, Brent, and Paul Kay. 1969. Basic color terms. Berkeley: University of California Press. Berman, Ruth, and Dan Slobin. 1994. Relating events in narrative: a crosslinguistic developmental study. Hillsdale, NJ: Lawrence Erlbaum Associates. Bernolet, Sarah, and Timothy Colleman. In prep. Sense-based and lexemebased alternation biases in the Dutch dative alternation. In Jiyoung Yoon and Stefan Th. Gries (eds.), Corpus-based approaches to Construction Grammar. Amsterdam: John Benjamins. Bertoldi, Anderson, and Rove Luize de Oliveira Chishman. 2011. Developing a frame-based lexicon for the Brazilian legal language: the case of the Criminal_Process frame. AICOL, 256–70. Beukeboom, Camiel J., Catrin Finkenauer, and, Danie¨l H. J. Wigboldus. 2010. The negation bias: when negations signal stereotypic expectancies. Journal of Personality and Social Psychology 99(6), 978–92. Bex, Tony, Michael Burke, and Peter Stockwell. 2000. Contextualized stylistics: in honour of Peter Verdonk. Amsterdam: Rodopi.
References
Biber, Douglas. 1994. An analytical framework for register studies. In Douglas Biber and E. Finegan (eds.), Sociolinguistic perspectives on register, 31–56. Oxford University Press. Biber, Douglas, and Edward Finegan. 1994. Introduction: situating register in sociolinguistics. In Douglas Biber and E. Finegan (eds.), Sociolinguistic perspectives on register, 3–12. Oxford University Press. Bienfait, Frits, and Walter E. A. van Beek. 2001. Right and left as political categories: an exercise in ‘not-so-primitive’ classification. Anthropos 96, 169–78. Biernacka, Eva. 2013. The role of metonymy in political discourse. PhD dissertation. Open University, Milton Keynes. Binder, Jeffrey R., and Rutvik H. Desai. 2011. The neurobiology of semantic memory. Trends in cognitive sciences 15(11), 527–36. Binder, Jeffery R., Rutvik H. Desai, William W. Graves, and Lisa L. Conant. 2009. Where is the semantic system? A critical review and meta-analysis of 120 functional neuroimaging studies. Cerebral Cortex 19(12), 2767–96. Bitzer, Lloyd. 1969. The rhetorical situation. Philosophy and Rhetoric 1(1), 1–14. Blank, I., Z. Balewski, K. Mahowald, and E. Fedorenko. 2016. Syntactic processing is distributed across the language system. NeuroImage 127, 307–23. Blevins, James P. 2006. Word-based morphology. Journal of Linguistics 42, 531–73. Bleys, Joris, Kevin Stadler, and Joachim De Beule. 2011. Search in linguistic processing. In Luc Steels (ed.), Design patterns in fluid construction grammar, 149–80. Amsterdam: John Benjamins. Blomberg, Johan. 2015. The expression of non-actual motion in Swedish, French and Thai. Cognitive Linguistics 26(4), 657–96. Blomberg, Johan and Jordan Zlatev. 2014. Actual and non-actual motion: why experientialist semantics needs phenomenology (and vice versa). Phenomenology and the Cognitive Sciences 13(3), 395–418. 2015. Non-actual motion: phenomenological analysis and linguistic evidence. Cognitive Processing 16, 153–57. Blommaert, Jan. 2014. From mobility to complexity in sociolinguistic theory and method. Tilburg Papers in Culture Studies 103, 1–24. Blommaert, Jan, and Ben Rampton. 2011. Language and superdiversity. Diversities 13, 1–21. Bloom, Paul. 2000. How children learn the meanings of words. Cambridge, MA: MIT Press. Bloomfield, Leonard. 1933. Language. New York: Holt. Blumberg, Mark S. 2006. Basic instinct: the genesis of behavior. New York: Basic Books. Blumstein Sheila, and Dima Amso. 2013. Dynamic functional organization of language: insights from functional neuroimaging. Perspectives on Psychological Science 8(1), 44–48.
693
694
References
Boas, Hans C. 2001. Frame Semantics as a framework for describing polysemy and syntactic structures of English and German motion verbs in contrastive computational lexicography. In P. Rayson, A. Wilson, T. McEnery, A. Hardie, and S. Khoja (eds.), Proceedings of Corpus Linguistics 2001 Conference, 64–73. Lancaster University UCREL. 2002. Bilingual FrameNet dictionaries for machine translation. In M. Gonza´lez Rodrı´guez and C. Paz Sua´rez Araujo (eds.), Proceedings of the Third International Conference on Language Resources and Evaluation, Vol. IV, 1364–71, Las Palmas, Spain. 2003a. A constructional approach to resultatives. Stanford: CSLI. 2003b. A lexical-constructional account of the locative alternation. In L. Carmichael, C.-H. Huang, and V. Samiian (eds.), Proceedings of the 2001 Western Conference in Linguistics. Vol. 13, 27–42. 2005a. Determining the productivity of resultative constructions: a reply to Goldberg and Jackendoff. Language 81(2), 448–64. 2005b. From theory to practice: Frame Semantics and the design of FrameNet. In S. Langer and D. Schnorbusch (eds.), Semantik im Lexikon, 129–60. Tu¨bingen: Narr. 2005c. Semantic frames as interlingual representations for multilingual lexical databases. International Journal of Lexicography 18(4), 445–78. 2006. A frame-semantic approach to identifying syntactically relevant elements of meaning. In Petra Steiner, Hans C. Boas, and Stefan Schierholz (eds.), Contrastive studies and valency: studies in honor of Hans Ulrich Boas, 119–49. Frankfurt: Peter Lang. 2008a. Resolving Form-meaning Discrepancies in Construction Grammar. In J. Leino (ed.), Constructional Reorganization, 11–36. Amsterdam and Philadelphia: Benjamins. 2008b. Determining the structure of lexical entries and grammatical constructions in Construction Grammar. Annual Review of Cognitive Linguistics 6, 113–44. 2008c. Towards a frame-constructional approach to verb classification. In E. Sosa Acevedo and F. J. Corte´s Rodrı´guez (eds.), Grammar, Constructions, and Interfaces. Special issue of Revista Canaria de Estudios Ingleses 57, 17–48. (ed.). 2009. Multilingual FrameNets in computational lexicography. Berlin: De Gruyter Mouton. 2010a. Linguistically relevant meaning elements of English communication verbs. Belgian Journal of Linguistics 24, 54–82. 2010b. Comparing constructions across languages. In H. C. Boas (ed.), Contrastive Studies in Construction Grammar, 1–20. Amsterdam and Philadelphia: Benjamins. (ed.). 2010c. Contrastive studies in construction grammar. Amsterdam: John Benjamins.
References
2011a. Zum Abstraktionsgrad von Resultativkonstruktionen. In Stefan Engelberg, Anke Holler, and Kristel Proost (eds.), Sprachliches Wissen zwischen Lexikon und Grammatik, 37–69. Berlin: De Gruyter Mouton. 2011b. A frame-semantic approach to syntactic alternations with buildverbs. In P. Guerrero Medina (ed.), Morphosyntactic alternations in English, 207–34. London: Equinox. 2011c. Coercion and leaking argument structures in Construction Grammar. Linguistics 49(6), 1271–303. 2013a. Cognitive Construction Grammar. In Thomas Hoffmann and Graeme Trousdale (eds.), The Oxford handbook of construction grammar, 233–52. Oxford University Press. 2013b. Wie viel Wissen steckt in Wo¨rterbu¨chern? Eine frame-semantische Perspektive. Zeitschrift fu¨r Angewandte Linguistik 57, 75–97. 2013c. Frame Semantics and translation. In A. Rojo and I. IbarretxteAntunano (eds.), Cognitive Linguistics and Translation, 125–58. Berlin: De Gruyter Mouton. 2014. Zur Architektur einer konstruktionsbasierten Grammatik des Deutschen. In A. Ziem and A. Lasch (eds.), Grammatik als Inventar von Konstruktionen? Sprachliches Wissen im Fokus der Konstruktionsgrammatik. Berlin: de Gruyter. 2016. Frames and constructions for the study of oral poetics. In C. Paga´n Ca´novas and M. Antovic´ (eds.), Oral Poetics and Cognitive Science, 99–124. Berlin: De Gruyter Mouton. Boas, Hans C., and Ryan Dux. 2013. Semantic frames for foreign language education: towards a German frame-based dictionary. Veridas On-line. Special Issue on Frame Semantics and its Technological Applications 82–100. Boas, Hans C., and Ivan A. Sag (eds.). 2012. Sign-based construction grammar. Stanford: CSLI. Boas, Hans C., Ryan Dux, and Alexander Ziem. 2016. Frames and constructions in an online learner’s dictionary of German. In S. De Knop and G. Gilquin (eds.), Applied Construction Grammar, 303–326. Berlin: de Gruyter. Bock, Kathryn. 1986a. Meaning, sound and syntax: lexical priming in sentence production. Journal of Experimental Psychology: Learning, Memory and Cognition 12, 575–86. 1986b. Syntactic persistence in language production. Cognitive Psychology 18, 355–87. Bock, Kathryn, and Zenzi Griffin. 2000. The persistence of structural priming: transient activation or implicit learning? Journal of Experimental Psychology: General 129, 177–92. Bod, Rens. 2000. The storage vs. computation of three-word sentences. Paper presented at AMLaP-2000. Bod, Rens, Jennifer Hay, and Stefanie Jannedy. 2003. Probabilistic linguistics. Cambridge, MA: MIT Press.
695
696
References
Boers, Frank. 2000a. Metaphor awareness and vocabulary retention. Applied Linguistics, 21, 553–71. 2000b. Enhancing metaphoric awareness in specialized reading. English for Specific Purposes 19, 137–47. 2011. Cognitive semantic ways of teaching figurative phrases: an assessment. Review of Cognitive Linguistics 9, 227–61. 2013. Cognitive Linguistic approaches to second language vocabulary: assessment and integration. Language Teaching: Surveys and Studies 46(2), 208–24. Boers, Frank, and Seth Lindstromberg (eds.). 2008a. Cognitive linguistic approaches to teaching vocabulary and phraseology. Berlin: De Gruyter Mouton. Boers, Frank, and Seth Lindstromberg. 2008b. From empirical findings to pedagogical practice. In Frank Boers and Seth Lindstromberg (eds.), Cognitive linguistic approaches to teaching vocabulary and phraseology, 375–94. Berlin: De Gruyter Mouton. Boers, Frank, Seth Lindstromberg, Jeannette Littlemore, He´le`ne Stengers, and June Eyckmans. 2008. Variables in the mnemonic effectiveness of pictorial elucidation. In Frank Boers and Seth Lindstromberg (eds.), Cognitive linguistic approaches to teaching vocabulary and phraseology, 65–100. Berlin: De Gruyter Mouton. Bohnemeyer, Ju¨rgen. 2010. The language-specificity of conceptual structure: path, fictive motion, and time relations. In Barbara Malt and Philip Wolff (eds.), Words and the mind: how words capture human experience, 111–37. Oxford University Press. Bolinger, Dwight. 1949. The sign is not arbitrary. Boletı´n del Instituto Caro y Cuervo (= Thesaurus) 5, 52–62. 1985. The inherent iconism of intonation. In John Haiman (ed.), Iconicity in syntax, 97–108. Amsterdam: John Benjamins. Boogaart, Ronny. 2009. Semantics and pragmatics in construction grammar: the case of modal verbs. In Alexander Bergs and Gabriele Diewald (eds.), Contexts and constructions, 213–41. Amsterdam: John Benjamins. Boogaart, Ronny, and Egbert Fortuin. 2016. Modality and mood in cognitive linguistics and construction grammar(s). In Johan Van der Auwera and Jan Nuyts (eds.), The Oxford handbook of modality and mood, 514–34. Oxford University Press. Boogaart, Ronny, Timothy Colleman, and Gijsbert Rutten. 2014. Constructions all the way everywhere: four new directions in constructionist research. In Ronny Boogaart, Timothy Colleman, and Gijsbert Rutten (eds.), Extending the scope of Construction Grammar, 1–14. Berlin, Boston: De Gruyter. Booij, Geert. 2002. The morphology of Dutch. Oxford University Press. 2009. Lexical integrity as a morphological universal, a constructionist view. In Sergio Scalise, Elizabeta Magni, and Antonietta Bisetto (eds.),
References
Universals of language today, 83–100. Dordrecht: Springer Science and Business Media. 2010. Construction morphology. Oxford University Press. 2013. Morphology in construction grammar. In Thomas Hoffmann and Graeme Trousdale (eds.), The Oxford handbook of construction grammar, 255–73. Oxford University Press. 2014. Language use and the architecture of grammar. Suvremena lingvistika (Contemporary linguistics) 40, 192–212. 2015. The nominalization of Dutch particle verbs: schema unification and second order schemas. Nederlandse Taalkunde 20, 285–314. 2016. Inheritance and motivation in construction morphology. In Nik Gisborne and Andrew Hippisley (eds.), Default inheritance, n. pag. Oxford University Press. Booij, Geert, and Jenny Audring. 2016. Construction morphology and the parallel architecture of grammar. Cognitive Science. doi:10.1111/ cogs.12323. In prep. Category change in construction morphology. In Kristel Van Goethem, Muriel Norde, Evie Cousse´, and Gudrun Vanderbauwhede (eds.), Category change from a constructional perspective. Amsterdam: John Benjamins. Booij, Geert, and Matthias Hu¨ning. 2014. Affixoids and constructional idioms. In Ronny Boogaart, Timothy Colleman, and Gijsbert Rutten (eds.), Extending the scope of construction grammar, 77–105. Berlin: De Gruyter Mouton. Booij, Geert, and Francesca Masini. 2015. The role of second order schemas in word formation. In Laurie Bauer, Livia Ko˝rtve´lyessy, and Pavol Sˇtekauer (eds.), Semantics of complex words, 47–66. Cham: Springer. Borin, Lars, Dana Danne´lls, Markus Forsberg, Maria Toporowska Gronostaj, and Dimitrios Kokkinakis. 2009. Thinking green: toward Swedish FrameNet+. In FrameNet Masterclass and Workshop, n. pag. Milan: Universita` Cattolica del Sacro Cuore. 2010. The past meets the present in the Swedish FrameNet++. https://svn .spraakdata.gu.se/sb/fnplusplus/pub/SweFN_Euralex_extended.pdf. Borkent, Mike. 2010. Illusions of simplicity: a cognitive approach to visual poetry. English Text Construction 3(2), 145–64. Borkent, Mike, Barbara Dancygier, and Jennifer Hinnell (eds.). 2013. Language and the creative mind. University of Chicago Press. Boroditsky, Lera. 2000. Metaphoric structuring: understanding time through spatial metaphors. Cognition 75(1), 1–28. 2001. Does language shape thought? Mandarin and English speakers’ conceptions of time. Cognitive Psychology 43(1), 1–22. Boroditsky, Lera, and Alice Gaby. 2010. Remembrance of times East: absolute spatial representations of time in an Australian Aboriginal Community. Psychological Science 21(11), 1635–39. Boroditsky, Lera, and Michael Ramscar. 2002. The roles of body and mind in abstract thought. Psychological Science 13, 185–88.
697
698
References
Boroditsky, Lera, Orly Fuhrman, and Kelly McCormick. 2011. Do English and Mandarin speakers think about time differently? Cognition 118(2), 123–29. Boroditsky, Lera, Alice Gaby, and Stephen C. Levinson. 2007. Time in space. In A. Majid (ed.), Field manual volume 10, 59–80. Nijmegen: Max Planck Institute for Psycholinguistics. Borsley, Robert D. 2004. An approach to English comparative correlatives. In Stephan Mu¨ller (ed.), Proceedings of the 11th international conference on head-driven phrase structure grammar, 70–92. Stanford: CSLI. 2006. Syntactic and lexical approaches to unbounded dependencies. Essex Research Reports in Linguistics 49, 31–57. 2007. Hang on again! Are we ‘on the right track’? Martin Atkinson the minimalist muse. Essex Research Reports in Linguistics 53, 43–70. Bottini, Roberto. 2011. An interdisciplinary study of time in language and mind. PhD dissertation. Universita` degli Studi di Bergamo, Italy. Bottini, Roberto, and Daniel Casasanto. 2010. Implicit spatial length modulates time estimates, but not vice versa. In Stellan Ohlsson and Richard Catrambone (eds.), Proceedings of the 32nd Annual Conference of the Cognitive Science Society, 1348–53. Austin: Cognitive Science Society. 2013. Space and time in the child’s mind: metaphoric or ATOMic? Frontiers in Psychology 4, 1–9. Bourdieu, Pierre. 1990. The logic of practice. Stanford University Press. Bouveret, Myriam. 2012. GIVE frames and constructions in French. In M. Bouveret and D. Legallois (eds.), Constructions in French, 99–126. Amsterdam: John Benjamins. Bowdle, Brian F., and Deirdre Gentner. 2005. The career of metaphor. Psychological Review 112(1), 193–216. Bowerman, Melissa. 1982. Reorganizational processes in lexical and syntactic development. In E. Wanner and L. R. Gleitman (eds.), Language acquisition: the state of the art, 319–46. Cambridge University Press. 1988. The ‘no negative evidence’ problem: how do children avoid constructing an overgeneral grammar? In J. A. Hawkins (ed.), Explaining language universals, 73–101. Oxford: Blackwells. 1996. The origins of children’s spatial semantic categories: cognitive versus linguistic determinants. In J. J. Gumperz and S. C. Levinson (eds.), Rethinking linguistic relativity, 145–76. Cambridge University Press. Bowerman, Melissa, and Eric Pederson. 1992. Topological relations picture series. In Stephen C. Levinson (ed.), Space stimuli kit 1.2: November 1992, 51. Nijmegen: Max Planck Institute for Psycholinguistics. Boyd, William. 2009. Ordinary thunderstorms. London: Bloomsbury. Boye, Kasper, and Peter Harder. 2012. A usage-based theory of grammatical status and grammaticalization. Language 88, 1–44. Bradlow, Ann R., Lynne Nygaard, and David Pisoni. 1999. Effects of talker, rate and amplitude variation on recognition memory. Perception and Psychophysics 61, 206–19.
References
Braine, Martin D., and Patricia J. Brooks. 1995. Verb argument structure and the problem of avoiding an overgeneral grammar. In Michael Tomasello and William E. Merriman (eds.), Beyond names for things: young children’s acquisition of verbs, 352–76. Hillsdale, NJ: Erlbaum. Brandt, Line. 2008. A semiotic approach to fictive interaction as a representational strategy in communicative meaning construction. In Todd Oakley and Anders Hougaard (eds.), Mental spaces in discourse and interaction, 109–48. Amsterdam: John Benjamins. 2013. The communicative mind: a linguistic exploration of conceptual integration and meaning construction. Newcastle upon Tyne: Cambridge Scholars. Brandt, Line, and Per Aage Brandt. 2005. Making sense of a blend. A cognitive semiotics approach to metaphor. Annual Review of Cognitive Linguistics 3, 216–49. Brandt, Line, and Esther Pascual. 2016. ‘Say yes hello to this ad’: the persuasive rhetoric of fictive interaction in marketing. In Esther Pascual and Sergeiy Sandler (eds.), The conversation frame: forms and functions of fictive interaction, 303–22. Amsterdam: John Benjamins. Brandt, Per Aage. 2004. Spaces, domains, and meaning: essays in cognitive semiotics. Bern: Peter Lang. Brandt, Silke, E., Evan Kidd, Elena Lieven, and Michael Tomasello. 2009. The discourse bases of relativization: an investigation of young German and English-speaking children’s comprehension of relative clauses. Cognitive Linguistics 20(3), 539–70. Branigan, Holly, Martin J. Pickering, and Alexandra A. Cleland. 2000. Syntactic co-ordination in dialogue. Cognition 75, 13–25. Brasoveanu, Adrian. 2008. Comparative and equative correlatives as anaphora to differentials. Poster presented at Semantics and Linguistic Theory 18, University of Massachusetts Amherst, Massachusetts, and at the 9th Semfest, Stanford, California. Braº ten, Stein. 2006. Intersubjective communication and emotion in early ontogeny. Cambridge University Press. Brdar, Mario, and Rita Brdar-Szabo´. 2009. The (non) metonymic use of place names in English, German, Hungarian and Croatian. In KlausUwe Panther, Linda Thornburg, and Antonio Barcelona (eds.), Metonymy and metaphor in grammar, 229–57. Amsterdam: John Benjamins. Brdar, Mario, Stephan Gries, and Milena Zic Fuchs (eds.). 2011. Cognitive linguistics: convergence and expansion. Amsterdam: John Benjamins. Brdar-Szabo´, Rita, and Mario Brdar. 2011. What do metonymic chains reveal about the nature of metonymy? In Re´ka Benczes, Antonio Barcelona, and Francisco Ruiz de Mendoza (eds.), Defining metonymy in cognitive linguistics: towards a consensus view, 217–48. Amsterdam: John Benjamins. Brems, Lieselotte, and Sebastian Hoffmann. 2012. Grammaticalization. In Alexander Bergs and Laurel Brinton (eds.), English historical linguistics: vol. 2, 1558–76. Berlin: de Gruyter Mouton.
699
700
References
Brennan, Susan E., and Herbert H. Clark. 1996. Conceptual pacts and lexical choice in conversation. Journal of Experimental Psychology: Learning, Memory, and Cognition 22, 1482–93. Brentari, Diane. 1998. A prosodic model of sign language phonology. Cambridge, MA: MIT Press. Bressler, Steven L. and Vinod Menon. 2010 Large-scale brain networks in cognition: emerging methods and principles. Trends in Cognitive Sciences 14(6), 277–90. Brezina, Vaclav, Tony McEnery, and Stephen Wattam. 2015. Collocations in context: a new perspective on collocation networks. International Journal of Corpus Linguistics 20(2), 139–73. Brindo¨pke, Christel, J. Ha¨ger, Michaela Johanntokrax, Arno Pahde, Michael Schwalbe, and Britta Wrede. 1995. Darf ich Dich Marvin nennen? Instruktionsdialoge in einem Wizard-of-Oz Szenario: Szenario-design und auswertung. Universita¨t Bielefeld: SFB-Report ‘Situierte ku¨nstliche Kommunikatoren’ 95/16. Brinton, Laurel. 2012. Lexicalization. In Alexander Bergs and Laurel Brinton (eds.), English historical linguistics: vol. 2, 1577–98. Berlin: de Gruyter Mouton. Brinton, Laurel, and Elizabeth Closs Traugott. 2005. Lexicalization and language change. Cambridge University Press. Brisard, Frank. 2006. Logic, subjectivity and the semantics/pragmatics distinction. In Angeliki Anthanasiadou, Costas Canakis, and Bert Cornillie (eds.) Subjectification: various paths to subjectivity, 41–74. Berlin: de Gruyter Mouton. Broaders, Sarah C., Susan Wagner Cook, Zachary Mitchell, and Susan Goldin-Meadow. 2007. Making children gesture brings out implicit knowledge and leads to learning. Journal of Experimental Psychology: General 136(4), 539–50. Broccias, Cristiano. 2013. Cognitive grammar. In Thomas Hoffmann and Graeme Trousdale (eds.), The Oxford handbook of construction grammar, 191–210. Oxford University Press. Broˆne, Geert. 2008. Hyper- and misunderstanding in interactional humour. Journal of Pragmatics 40(12), 2027–61. 2010. Bedeutungskonstitution in verbalem Humor: ein kognitiv-linguistischer und diskurssemantischer Ansatz. Frankfurt am Main: Lang. 2012. Humour and irony in cognitive pragmatics. In Hans-Jo¨rg Schmid (ed.), Handbook of cognitive pragmatics, vol. 4: cognitive pragmatics, 463–504. Berlin: De Gruyter Mouton. Broˆne, Geert, and Seana Coulson. 2010. Processing deliberate ambiguity in newspaper headlines: double grounding. Discourse Processes 47(3), 212–36. Broˆne, Geert, and Kurt Feyaerts. 2005. Headlines and cartoons in the economic press: double grounding as a discourse supportive strategy.
References
In Guido Erreygers and Geert Jacobs (eds.), Language, communication and the economy, 73–99. Amsterdam: John Benjamins. Broˆne, Geert, and Bert Oben. 2013. Resonating humour: a corpus-based approach to creative parallelism in discourse. In Kurt Feyaerts, Tony Veale, and Charles Forceville (eds.), Creativity and the agile mind: a multi-disciplinary study of a multi-faceted phenomenon, 181–204. Berlin: De Gruyter Mouton. 2015. InSight Interaction: a multimodal and multifocal dialogue corpus. Language Resources and Evaluation 49, 195–214. ˆ Brone, Geert, and Jeroen Vandaele (eds.). 2009. Cognitive poetics: goals, gains, and gaps. New York: De Gruyter Mouton. Broˆne, Geert, and Elisabeth Zima. 2014. Towards a dialogic construction grammar: a corpus-based approach to ad hoc routines and resonance activation. Cognitive Linguistics 25(3), 457–95. Broˆne, Geert, Kurt Feyaerts, and Tony Veale (eds.). 2015. Cognitive linguistic humor research: current trends and new developments. Berlin: De Gruyter Mouton. Broˆne, Geert, Bert Oben, and Kurt Feyaerts. 2017. Eye gaze and viewpoint in multimodal interaction management. Special issue on Viewpoint phenomena in multimodal communication. Barbara Dancygier and Lieven Vandelanotte (eds.) Cognitive Linguistics 28(3). Brooks, Patricia, and Michael Tomasello. 1999a. Young children learn to produce passives with nonce verbs. Developmental Psychology 35, 29–44. 1999b. How children constrain their argument structure constructions. Language 75, 720–38. Brooks, Patricia, Michael Tomasello, Kelly Dodson, and Lawrence B. Lewis. 1999. Young children’s overgeneralizations with fixed transitivity verbs. Child Development 70, 1325–37. Brown, Amanda, and Jidong Chen. 2013. Construal of manner in speech and gesture in Mandarin, English, and Japanese. Cognitive Linguistics 24(4), 605–31. Brown, Penelope 2012. Time and space in Tzeltal: Is the future uphill? Frontiers in Psychology 3, 212. Brown, Penelope, and Stephen C. Levinson. 1987. Politeness: some universals in language usage. Cambridge University Press. Brugman, Claudia. 1983. The use of body-part terms as locatives in Chalcatongo Mixtec. Survey of California and Other Indian Languages 4, 235–90. Bruner, Jerome S. 1983. Child’s talk: learning to use language. New York: W. W. Norton. 1990. Acts of meaning. Cambridge, MA: Harvard University Press. Brunye´, Tad T., Aaron Gardony, Caroline R. Mahoney, and Holly A. Taylor. 2012. Body-specific representations of spatial location. Cognition 23(2), 229–39.
701
702
References
Brunye´, Tad T., Tali Ditman, Caroline R. Mahoney, Jason S. Augustyn, and Holly A. Taylor. 2009. When you and I share perspectives: pronouns modulate perspective taking during narrative comprehension. Psychological Science 20(1), 27–32. Bu¨chel, Christian, Cathy Price, R. S. Frackowiak, and Karl Friston. 1998. Different activation patterns in the visual cortex of late and congenitally blind subjects. Brain 121, 409–19. Bucholtz, Mary. 2010. White kids: language and white youth identities. Cambridge University Press. Bucholtz, Mary, and Kira Hall. 2005. Identity and interaction: a sociocultural linguistic approach. Discourse Studies 7, 585–614. Budwig, Nancy, Bhuvana Narasimhan, and Smita Srivastava. 2006. Interim solutions: the acquisition of early verb constructions in Hindi. In Eve V. Clark and Barbara F. Kelly (eds.), Constructions in acquisition, 163–83. Stanford: CSLI. Buescher, Kimberly, and Susan Strauss. In prep. Conceptual frameworks and L2 pedagogy: the case of French prepositions. In Andrea Tyler, Lourdes Ortega, Mariko Uno, and Hae In Park (eds.), Usage-inspired L2 instruction: researched pedagogy. Amsterdam: John Benjamins. Burchardt, Aljoscha, Katrin Erk, Anette Frank, Andrea Kowalski, Sebastian Pado´, and Manfred Pinkal. 2009. Using FrameNet for the semantic analysis of German: annotation, representation, and automation. In Hans C. Boas (ed.), Multilingual FrameNets: methods and applications, 209–44. Berlin and New York: De Gruyter Mouton. Burling, Robbins, 1966. The metrics of children’s verse: a cross-linguistic study, American Anthropologist 68(6), 1418–41. Burton, H, A. Z. Snyder, T. E. Conturo, E. Akbudak, J. M. Ollinger, and M. E. Raichle. 2002a. Adaptive changes in early and late blind: a fMRI study of Braille reading. Journal of Neurophysiology 87, 589–611. Burton, H., A. Z. Snyder, J. B. Diamond, and M. E. Raichle. 2002b. Adaptive changes in early and late blind: a FMRI study of verb generation to heard nouns. Journal of Neurophysiology 88(6), 3359–71. Busse, Dietrich. 2012. Frame-Semantik: Ein Kompendium. Berlin: De Gruyter Mouton. Butler, Christopher. 1985. Statistics in linguistics. Oxford: Blackwell. Butterworth, George, and Lesley Grover. 1990. Joint visual attention, manual pointing, and preverbal communication in human infancy. In M. Jeannerod (ed.), Attention and performance, vol. XIII, motor representation and control, 605–24. Hillsdale, NJ: Erlbaum. Bybee, Joan. 1985. Morphology. Amsterdam: John Benjamins. 1995. Regular morphology and the lexicon. Language and Cognitive Processes 10(5), 425–55. 2000. The phonology of the lexicon: evidence from lexical diffusion. Usage-based models of language, 65–85.
References
2001. Phonology and language use. Cambridge University Press. 2003. Mechanisms of change in grammaticization: the role of frequency. In Brian D. Joseph, and Richard D. Janda (eds.), The handbook of historical linguistics, 602–23. Oxford: Blackwell. 2006. From usage to grammar: the mind’s response to repetition. Language 82, 711–33. 2007. Frequency of use and the organization of language. Oxford University Press. 2010. Language, usage and cognition. Cambridge University Press. 2013. Usage-based theory and exemplar representations of constructions. In Thomas Hoffmann and Graeme Trousdale (eds.), The Oxford handbook of construction grammar, 49–69. Oxford University Press. Bybee, Joan, and Paul Hopper (eds.). 2001. Frequency and the emergence of linguistic structure. Amsterdam: John Benjamins. Bybee, Joan, and Joanne Scheibman. 1999. The effect of usage on degrees of constituency: the reduction of don’t in English. Linguistics 37, 575–96. Bylund, Emanuel. 2011. Segmentation and temporal structuring of events in early Spanish-Swedish bilinguals. International Journal of Bilingualism 15(1), 56–84. Caballero, Rosario. 2006. Re-viewing space: figurative language in architects’ assessment of built space. Berlin: Walter de Gruyter. Cabeza, Roberto. 2002. Hemispheric asymmetry reduction in older adults: the HAROLD model. Psychology and Aging 17(1), 85. Cabeza, Roberto, and Lars Nyberg. 2000. Imaging cognition II: an empirical review of 275 PET and fMRI studies. Journal of Cognitive Neuroscience 12(1), 1–47. Cadierno, Teresa. 2004. Expressing motion events in a second language: a cognitive typological approach. In Michel Achard and Suzanne Neimeier (eds.), Cognitive linguistics, second language acquisition and foreign language teaching, 13–49. Berlin: De Gruyter Mouton. 2008. Learning to talk about motion in a foreign language. In Peter Robinson and Nick C. Ellis (eds.) Handbook of cognitive linguistics and second language acquisition, 239–75. London: Routledge. Cadierno, Teresa, and Søren Wind Eskildsen. 2016. In Teresa Cadierno and Søren Wind Eskildsen (eds.), Usage-based perspectives on second language learning, 121–223. Berlin: De Gruyter Mouton. Cadierno, Teresa, Iraide Ibarretxe-Antun˜ano, and Alberto Hijazo-Gasco´n. 2016. Semantic categorization of placement verbs in L1 Danish and Spanish. Language Learning 66(1), 191–223. Cai, Zhenguang G., and Louise Connell. 2015. Space-time interdependence: evidence against asymmetric mapping between time and space. Cognition 136, 268–81. Calbris, Genevieve. 1990. The semiotics of French gesture. Bloomington: Indiana University Press.
703
704
References
Caldwell-Harris, Catherine, Jonathan Berant, and Shimon Edelman. 2012. Measuring mental entrenchment of phrases with perceptual identification, familiarity ratings, and corpus frequency statistics. In Dagmar S. Divjak and Stefan Th. Gries (eds.), Frequency effects in language representation, 165–94. Berlin: De Gruyter Mouton. Cameron, Lynne. 2011. Metaphor and reconciliation: the discourse dynamics of empathy in post–conflict conversations. London: Routledge. Cameron, Lynne, and Alice Deignan. 2003. Combining large and small corpora to investigate tuning devices around metaphor in spoken discourse. Metaphor and Symbol 18(3), 149–60. 2006. The emergence of metaphor in discourse. Applied Linguistics 27(4), 671–90. Cameron-Faulkner, Thea, Elena Lieven, and Anna Theakston. 2007. What part of no do children not understand? A usage-based account of multiword negation. Journal of Child Language 34(2), 251–82. Cantrall, William R. 1974. Viewpoint, reflexives, and the nature of noun phrases. The Hague: De Gruyter Mouton. Cappelle, Bert. 2011. The the. . . the. . . construction: meaning and readings. Journal of Pragmatics 43(1), 99–117. Cappelle, Bert, Yury Shtyrov, and Friedemann Pulvermu¨ller. 2010. Heating up or cooling up the brain? MEG evidence that phrasal verbs are lexical units. Brain and Language 115(3), 189–201. Carden, Guy, and Thomas G. Dieterich. 1980. Introspection, observation and experiment: an example where experiment pays off. Journal of the Philosophy of Science Association 2, 583–97. Cardin, Velia, Eleni Orfanidou, Jerker Ro¨nnberg, Cheryl M. Capek, Mary Rudner, and Bencie Woll. 2013. Dissociating cognitive and sensory neural plasticity in human superior temporal cortex. Nature Communications 4, 1473. Carlson, Richard A., Marios N. Avraamides, Melanie Cary, and Stephen Strasberg. 2007. What do the hands externalize in simple arithmetic? Journal of Experimental Psychology: Learning, Memory, and Cognition 33(4), 747–56. Caron-Pargue, Josiane and Jean Caron. 1991. Psychopragmatics vs. sociopragmatics: the function of pragmatic markers in thinkingaloud protocols. In Jef Verschueren (ed.), Pragmatics at issue: selected papers of the International Pragmatics Conference, 29–36. Amsterdam: Benjamins. Carpenter, Bob. 1992. The logic of typed feature structures. Cambridge University Press. Carpenter, Malinda, Nameera Akhtar, and Michael Tomasello. 1998. Fourteen-through 18-month-old infants differentially imitate intentional and accidental actions. Infant Behavior and Development 21(2), 315–30.
References
Carston, Robyn. 2002. Thoughts and utterances: the pragmatics of explicit communication. Oxford: Blackwell. Carter, Ronald. 2004. Language and creativity, the art of common talk. London: Routledge. Casad, Eugene. 1982. Cora locationals and structured imagery. PhD dissertation. University of California at San Diego. Casad, Eugene, and Ronald W. Langacker. 1985. ‘Inside’ and ‘outside’ in Cora grammar. IJAL 51, 247–81. Casad, Eugene, and Gary Palmer (eds.). 2003. Cognitive linguistics and non-Indo-European languages. Berlin: De Gruyter Mouton. Casasanto, Daniel. 2008a. Similarity and proximity: when does close in space mean close in mind? Memory and Cognition 36, 1047–56. 2008b. Who’s afraid of the big bad Whorf? Crosslinguistic differences in temporal language and thought. Language Learning 58(1), 63–79. 2009. Embodiment of abstract concepts: good and bad in right- and left-handers. Journal of Experimental Psychology: General 138(3), 351–67. 2010. Space for thinking. In Vyvyan Evans and Paul Chilton (eds.), Language, cognition, and space: state of the art and new directions, 453–78. London: Equinox. 2013. Experiential origins of mental metaphors: language, culture, and the body. In Mark J. Landau, Michael D. Robinson, and Brian P. Meier (eds.), The power of metaphor: examining its influence on social life, 249–68. Washington, DC: American Psychological Association Books. 2014. Development of metaphorical thinking: the role of language. In Mike Borkent, Barbara Dancygier, and Jennifer Hinnel (eds.), Language and the creative mind, 3–18. Stanford: CSLI. 2016a. A shared mechanism of linguistic, cultural, and bodily relativity. Language Learning 66(3), 714–30. 2016b. Relationships between space, time, and number: insights from language acquisition. Paper presented at the International Meeting of the Psychonomic Society, Granada, Spain. Casasanto, Daniel, and Lera Boroditsky. 2008. Time in the mind: using space to think about time. Cognition 106(2), 579–93. Casasanto, Daniel, and Roberto Bottini. 2014. Mirror reading can reverse the flow of time. Journal of Experimental Psychology: General 143(2), 473–79. Casasanto, Daniel, and Evangelia G. Chrysikou. 2011. When left is ‘right’: motor fluency shapes abstract concepts. Psychological Science 22(4), 419–22. Casasanto, Daniel, and Katinka Dijkstra. 2010. Motor action and emotional memory. Cognition 115, 179–85. Casasanto, Daniel, and Tom Gijssels. 2015. What makes a metaphor an embodied metaphor? Linguistics Vanguard 1(1), 327–37.
705
706
References
Casasanto, Daniel, and Tania Henetz. 2012. Handedness shapes children’s abstract concepts. Cognitive Science 36 359–72. Casasanto, Daniel, and Kyle Jasmin. 2010. Good and bad in the hands of politicians: Spontaneous gestures during positive and negative speech. PLoS ONE 5(7), e11805. 2012. The hands of time: temporal gestures in English speakers. Cognitive Linguistics 23(4), 643–74. Casasanto, Daniel, Olga Fotakopoulou, and Lera Boroditsky. 2010. Space and time in the child’s mind: evidence for a cross-dimensional asymmetry. Cognitive Science 34, 387–405. Casasanto, Daniel, Lera Boroditsky, Webb Phillips, Jesse Greene, Shima Goswami, Simon Bocanegra-Thiel, et al. 2004. How deep are effects of language on thought? Time estimation in speakers of English, Indonesian, Greek, and Spanish. In Kenneth Forbus, Dedre Gentner, and Terry Regier (eds.), Proceedings of the 26th annual meeting of the Cognitive Science Society, 186–91. Austin: Cognitive Science Society. Casenhiser, Devin, and Adele E. Goldberg. 2005. Fast mapping between a phrasal form and meaning. Developmental Science 8(6), 500–08. Cassell, Justine, David McNeill, and Karl-Erik McCullough. 1998. Speechgesture mismatches: evidence for one underlying representation of linguistic and nonlinguistic information. Pragmatics and Cognition 7(1), 1–33. Chaemsaithong, Krisda. 2013. Interaction in Early Modern news discourse: the case of English witchcraft pamphlets and their prefaces (1566–1621). Text and Talk 33(2), 167–88. 2014. Dramatic monologues: the grammaticalization of speaking roles in courtroom opening statements. Pragmatics 24(4), 757–84. 2015. Communicating with silent addressees: engagement features in the opening statement. Language and Communication 43, 35–46. 2016. Persuading and arguing with the reader: fictive interaction strategies in witchcraft pamphlet prefaces (1566–1621). In Esther Pascual and Sergeiy Sandler (eds.), The conversation frame: forms and functions of fictive interaction, 113–30. Amsterdam: John Benjamins. Chafe, Wallace. 1962. Phonetics, semantics, and language. Language 38, 335–44. 1965. Meaning in language. American Anthropologist 67 (5), 23–36. 1967. Language as symbolization. Language 43, 57–91. 1968. Idiomaticity as an anomaly in the Chomskyan paradigm. Foundations of Language 4, 109–127. 1970a. Meaning and the structure of language. University of Chicago Press. 1970b. A semantically based sketch of Onondaga. IJAL Memoir 25, supplement to vol. 36(2). 1987. Cognitive constraints on information flow. In Russell S. Tomlin (ed.), Coherence and grounding in discourse, 21–51. Amsterdam and Philadelphia: John Benjamins. 1994. Discourse, consciousness, and time: the flow and displacement of conscious experience in speaking and writing. University of Chicago Press.
References
2002. Searching for meaning in language: a memoir. Historiographia Linguistica 29, 245–61. Chakravarty, Auditi, and Bonnie Boehme. 2004. Grammar and usage for better writing. New York: Amsco School Publications. Chambers, Jack. 2001. Vernacular universals. In Josep M. Fontana, Louise McNally, M. Teresa Turell, and Enric Vallduvı´ (eds.), ICLaVE 1: Proceedings of the First International Conference on Language Variation in Europe, 52–60. Barcelona: Universitat Pompeu Fabra. Chan, Angel, Elena Lieven, and Michael Tomasello. 2009. Children’s understanding of the agent-patient relations in the transitive construction: cross-linguistic comparison between Cantonese, German and English. Cognitive Linguistics 20(2), 267–300. Chan, Ting Ting and Benjamin Bergen. 2005. Writing direction influences spatial cognition. In Proceedings of the Twenty-Seventh Annual Conference of the Cognitive Science Society, University of California, San Diego. Chang, Franklin. 2002. Symbolically speaking: a connectionist model of sentence production. Cognitive Science 26, 609–51. Chang, Franklin, J. Kathryn Bock, and Adele E. Goldberg. 2003. Can thematic roles leave traces of their places? Cognition 90, 29–49. Chang, Franklin, Gary S. Dell, J. Kathryn Bock, and Zenzi M. Griffin. 2000. Structural priming as implicit learning: a comparison of models of sentence production. Journal of Psycholinguistic Research 29, 217–29. Chang, Nancy. 2008. Constructing grammar: a computational model of the emergence of early constructions. PhD dissertation. University of California, Berkeley. Charteris-Black, Jonathan. 2001. Blood sweat and tears: a corpus based cognitive analysis of ‘blood’ in English phraseology. Studi Italiani di Linguistica Teorica e Applicata 30(2), 273–87. 2003. Speaking with forked tongue: a comparative study of metaphor and metonymy in English and Malay phraseology. Metaphor and Symbol, 14(4), 289–310. 2004. Corpus approaches to critical metaphor analysis. Basingstoke: Palgrave Macmillan. 2005. Politicians and rhetoric: the persuasive power of metaphor. Basingstoke: Palgrave Macmillan. 2012. Shattering the bell jar: metaphor, gender and depression. Metaphor and Symbol 27(3), 199–216. Chen, Jenn-Yeu. 2007. Do Chinese and English speakers think about time differently? Failure of replicating Boroditsky 2001. Cognition 104, 427–36. Chenu, Florence, and Harriet Jisa. 2006. Caused motion constructions and semantic generality in early acquisition of French. In Eve V. Clark and Barbara F. Kelly (eds.), Constructions in acquisition, 233–61. Stanford: CSLI.
707
708
References
Childers, Jane B., and Michael Tomasello. 2001. Two-year-olds learn novel nouns, verbs, and conventional actions from massed or distributed exposures. Developmental Psychology 37(6), 739–48. Chilton, Paul. 2013. Frames of reference and the linguistic conceptualization of time: present and future. In K. M. Jaszczolt and L. de Saussure (eds.), Time: language, cognition, and reality, 236–58. Oxford University Press. Chomsky, Noam. 1957. Syntactic structures. The Hague: Mouton and Co. 1959. A review of B. F. Skinner’s Verbal behavior. Language 35(1), 26–58. 1965. Aspects of the theory of syntax. Cambridge, MA: MIT Press. 1981. Lectures on government and binding. Dordrecht: Foris. 1986. Knowledge of language: its nature, origin, and use. New York: Praeger. 1995. The minimalist program. Cambridge, MA: MIT Press. 2000a. New horizons in the study of language and mind. Cambridge University Press. 2000b. Minimalist inquiries: the framework. In Roger Martin, David Michaels and Juan Uriagereka (eds.), Step by step: essays on minimalist syntax in honor of Howard Lasnik, 89–155. Cambridge, MA: MIT Press. 2011. Language and other cognitive systems: what is special about language? Language Learning and Development 7(4), 263–78. Chomsky, Noam, and Morris Halle. 1965. Some controversial issues in phonological theory, Journal of Linguistics 1, 97–138. 1968. The sound pattern of English. New York: Harper and Row. Christiansen, Morten H., and Nick Chater. 2015. The now-or-never bottleneck: a fundamental constraint on language. Behavioral and Brain Sciences 38, 1–52. Chu, Mingyuan, and Sotaro Kita. 2011. The nature of gestures’ beneficial role in spatial problem solving. Journal of Experimental Psychology: General 140(1), 102–16. Chui, Kawai. 2011. Conceptual metaphors in gesture. Cognitive Linguistics 22(3), 437–58. Chung, Siaw-Fong. 2008. Cross-linguistic comparison of the MARKET metaphors. Corpus Linguistics and Linguistic Theory 4(2), 141–75. Church, R. Breckinridge, and Susan Goldin-Meadow. 1986. The mismatch between gesture and speech as an index of transitional knowledge. Cognition 23, 43–71. Church, R. Breckinridge, Saba Ayman-Nolley, and Shahrzad Mahootian. 2004. The role of gesture in bilingual education: does gesture enhance learning? International Journal of Bilingual Education and Bilingualism 7(4), 303–19. Cienki, Alan. 1998. Metaphoric gestures and some of their relations to verbal metaphorical expressions. In Jean-Pierre Koenig (ed.), Discourse and cognition: bridging the gap, 189–204. Stanford: CSLI. 2013. Image schemas and mimetic schemas in cognitive linguistics and gesture studies. Review of Cognitive Linguistics 11(2), 417–32.
References
2015. Spoken language usage events. Language and Cognition 7(4), 499–514. Cienki, Alan, and Gianluca Giansante. 2014. Conversational framing in televised political discourse: a comparison from the 2008 elections in the United States and Italy. Journal of Language and Politics 13(2), 255–88. Cienki, Alan, and Cornelia Mu¨ller (eds.). 2008a. Metaphor and gesture. Amsterdam: John Benjamins. 2008b. Metaphor, gesture, and thought. In Raymond W. Gibbs, Jr. (ed.), The Cambridge handbook of metaphor and thought, 483–501. Cambridge University Press. Citron, Francesca, and Adele Goldberg. 2014. Social context modulates the effect of hot temperature on perceived interpersonal warmth: a study of embodied metaphors. Language and Cognition 6, 1–11. Clancy, Steven J. 2006. The topology of Slavic case: semantic maps and multidimensional scaling. Glossos 7, 1–28. Clark, Eve V. 1971. On the acquisition of the meaning of before and after. Journal of Verbal Learning and Verbal Behavior 10(3), 266–75. 1987. The principle of contrast: a constraint on language acquisition. In Brian MacWhinney (ed.), Mechanisms of language acquisition, 1–33. Hillsdale, NJ: Erlbaum. 1997. Conceptual perspective and lexical choice in acquisition. Cognition 64(1), 1–37. 2003. First language acquisition. Cambridge University Press. 2009. First language acquisition, 2nd edn. Cambridge University Press. Clark, Eve. V., and Clark, Herbert H. 1979. When nouns surface as verbs. Language 55(4), 767–811. Clark, Herbert H. 1973. Space, time, semantics, and the child. In Timothy Moore (ed.), Cognitive development and the acquisition of language, 27–63. New York: Academic Press. 1996. Using language. Cambridge University Press. 1999. How do real people communicate with virtual partners? In Proceedings of AAAI-99 Fall Symposium, Psychological Models of Communication in Collaborative Systems, November 5–7, North Falmouth, MA and Menlo Park, CA: AAAI Press. Clark, Herbert H., and Richard J. Gerrig. 1990. Quotation as demonstration. Language 66(4), 784–805. Clark, Lynn. 2007. Cognitive sociolinguistics: a viable approach to variation in linguistic theory. LACUS Forum 33, 105–18. Clark, Lynn, and Graeme Trousdale. 2009. The role of token frequency in phonological change: evidence from TH-fronting in east-central Scotland. English Language and Linguistics 13, 33–56. 2013. Using participant observation and social network analysis. In Manfred Krug and Julia Schlu¨ter (eds.), Research methods in language variation and change, 36–52. Cambridge University Press.
709
710
References
Clark, Lynn, and Kevin Watson. 2011. Testing claims of a usage-based phonology with Liverpool English t-to-r. English Language and Linguistics 15, 523–47. Clausner, Timothy, and William Croft. 1999. Domains and image schemas. Cognitive Linguistics 10, 1–31. Coegnarts, Maarten, and Peter Kravanja (eds.). 2015. Embodied cognition and cinema. Leuven University Press. Cohen, L. G., P. Celnik, A. Pascual-Leone, B. Corwell, L. Faiz, J. Dambrosia, et al. 1997. Functional relevance of cross-modal plasticity in blind humans. Nature 389(6647), 180–83. Cole, M. W., J. R. Reynolds, J. D. Power, G. Repovs, A. Anticevic, and T. S. Braver. 2013. Multi-task connectivity reveals flexible hubs for adaptive task control. Nature Neuroscience 16(9), 1348–55. Colleman, Timothy. 2009a. The semantic range of the Dutch double object construction: a collostructional perspective. Constructions and Frames 1, 190–221. 2009b. Verb disposition in argument structure alternations: a corpus study of the dative alternation in Dutch. Language Sciences 31, 593–611. Collins, Allan M., and Elizabeth F. Loftus. 1975. A spreading-activation theory of semantic processing. Psychological Review 82, 407–28. Collins, Belinda. 1998. Convergence of fundamental frequencies in conversation: if it happens, does it matter? Fifth International Conference on Spoken Language Processing, 579. Australian National University. Comrie, Bernard. 1986. Conditionals: a typology. In Elizabeth C. Traugott, Alice ter Meulen, Judy S. Reilly, and Charles A. Ferguson (eds.), On Conditionals, 77–99. Cambridge University Press. Cook, Guy. 2000. Language play, language learning. Oxford University Press. Cook, Kenneth. 1988. A cognitive analysis of grammatical relations, case, and transitivity in Samoan. PhD dissertation, University of California at San Diego. Cook, Susan Wagner, and Michael K. Tanenhaus. 2009. Embodied communication: speakers’ gestures affect listeners’ actions. Cognition 113(1), 98–104. Cook, Susan Wagner, Terina Kuangyi Yip, and Susan Goldin-Meadow. 2012. Gestures, but not meaningless movements, lighten working memory load when explaining math. Language and Cognitive Processes 27(4), 594–610. Cooperrider, Kensy. 2014. Body-directed gestures: pointing to the self and beyond. Journal of Pragmatics 71, 1–16. Cooperrider, Kensy, and Rafael Nu´n˜ez. 2009. Across time, across the body: transversal temporal gestures. Gesture 9(2), 181–206. Cooperrider, Kensy, Rafael Nu´n˜ez, and Eve Sweetser. 2014. The conceptualization of time in gesture. In C. Mu¨ller, A. Cienki, E. Fricke, S. Ladewig,
References
D. McNeill, and J. Bressem (eds.), Body-Language-Communication (vol. 2), 1781–88. New York: De Gruyter Mouton. Cooperrider, Kensy, James Slotta, and Rafael Nu´n˜ez. 2016. Uphill and downhill in a flat world: the conceptual topography of the Yupno house. Cognitive Science 1–32. Cooren, Franc¸ois. 2010. Action and agency in dialogue: passion, incarnation and ventriloquism. Amsterdam: John Benjamins. 2012. Communication theory at the center: ventriloquism and the communicative constitution of reality. Journal of Communication 62(1), 1–20. Cooren, Franc¸ois, and Sergeiy Sandler. 2014. Polyphony, ventriloquism and constitution: in dialogue with Bakhtin. Communication Theory 24(3), 225–44. Cormier, Kearsy, Adam Schembri, and Bencie Woll. 2013. Pronouns and pointing in sign languages. Lingua: International Review of General Linguistics 137, 230–47. Cornillie, Bert. 2004. The shift from lexical to subjective readings in Spanish prometer ‘to promise’ and amenazar ‘to threaten’: a corpusbased account. Pragmatics 14(1), 1–30. Cornips, Leonie, and Cecilia Poletto. 2005. On standardising syntactic elicitation techniques (part 1). Lingua: International Review of General Linguistics 115, 939–57. Cotter, Colleen. 1997. Claiming a piece of the pie: how the language of recipes defines community. In Anne L. Bower (ed.), Recipes for reading: community cookbooks, stories, histories, 51–72. Amherst: University of Massachusetts Press. Coulson, Seana. 1995. Analogic and metaphoric mapping in blended spaces: Menendez brothers virus. CRL Newsletter 9(1), n. pag. [2001] 2005. What’s so funny: cognitive semantics and jokes. Cognitive Psychopathology/Psicopatologia Cognitive 2(3), 67–78. 2005. Extemporaneous blending: conceptual integration in humorous discourse from talk radio. Style 39(2), 107–22. 2006a. Semantic leaps: frame-shifting and conceptual blending in meaning construction. Cambridge University Press. 2006b. Conceptual Blending in thought, rhetoric, and ideology. In Gitte Kristiansen, Michel Achard, Rene´ Dirven, and Francisco J. Ruiz de Mendoza Iba´n˜ez (eds.), Cognitive linguistics: current applications and future perspectives, 187–208. Berlin: De Gruyter Mouton. 2008. Metaphor comprehension and the brain. In Raymond W. Gibbs Jr. (ed.), Metaphor and thought, 3rd edn, 177–94. Cambridge University Press. 2012. Cognitive neuroscience of figurative language. In Michael Spivey et al. (eds.), The Cambridge handbook of psycholinguistics, 523–37. Cambridge University Press. Coulson, Seana, and Todd Oakley (eds.). 2000. Special issue on conceptual blending. Cognitive Linguistics 11(3/4). 2005. Special issue on conceptual blending. Journal of Pragmatics 37(10).
711
712
References
2006. Purple persuasion: Conceptual Blending and deliberative rhetoric. In June Luchjenbroers (ed.), Cognitive linguistics: investigations across languages, fields, and philosophical boundaries, 47–65. Amsterdam: John Benjamins. Coulson, Seana, and Esther Pascual. 2006. For the sake of argument: mourning the unborn and reviving the dead through conceptual blending. Annual Review of Cognitive Linguistics 4, 153–81. Coulson, Seana, and Cyma Van Petten. 2002. Conceptual integration and metaphor: an event-related potential study. Memory and Cognition 30(6), 958–68. Couper-Kuhlen, Elizabeth, and Margret Selting (eds.). 1996. Prosody in conversation: interactional studies. Cambridge University Press. 2001. Studies in interactional linguistics. Amsterdam: John Benjamins. Coventry, Kenny R., Merce Prat-Sala, and Lynn V Richards. 2001. The interplay between geometry and function in the comprehension of ‘over,’ ‘under,’ ‘above’ and ‘below.’ Journal of Memory and Language 44, 376–98. Cowart, Wayne. 1994. Anchoring and grammar effects in judgments of sentence acceptability. Perceptual and Motor Skills 79, 1171–82. 1997. Experimental syntax: applying objective methods to sentence judgments. Thousand Oaks: Sage. Craig, Colette (ed.). 1986. Noun classes and categorization. Amsterdam: John Benjamins. 1991. Ways to go in Rama: a case study in polygrammaticalization. In Elizabeth C. Traugott and Bernd Heine (eds.), Approaches to grammaticalization, 455–92. Amsterdam: John Benjamins. Croft, William. 1993. The role of domains in the interpretation of metaphors and metonymies. Cognitive Linguistics 4, 335–70. 1998. Event structure in argument linking. In Miriam Butt and Wilhelm Geuder (eds.), The projection of arguments: lexical and compositional factors, 21–63. Stanford: CSLI. 2001. Radical Construction Grammar: syntactic theory in typological perspective. Oxford University Press. 2003. Lexical rules vs. constructions: a false dichotomy. In H. Cuyckens, T. Berg, R. Dirven, and K.-U. Panther (eds.), Motivation in language: studies in honor of Gu¨nther Radden, 49–68. Amsterdam: John Benjamins. 2009a. Connecting frames and constructions: a case study of ‘eat’ and ‘feed.’ Constructions and Frames 1(1), 7–28. 2009b. Toward a social cognitive linguistics. In Vyvyan Evans and Ste´phanie Pourcel (eds.), New directions in cognitive linguistics, 395–420. Amsterdam: John Benjamins. 2010. The origins of grammaticalization in the verbalization of experience. Linguistics 48, 1–48. 2012. Verbs: aspect and causal structure. Oxford University Press.
References
2013. Radical Construction Grammar. In Thomas Hoffmann and Graeme Trousdale (eds.), The Oxford handbook of Construction Grammar, 211–32. Oxford University Press. Croft, William, and D. Alan Cruse. 2004. Cognitive linguistics. Cambridge University Press. Croft, William, and K. T. Poole. 2008. Inferring universals from grammatical variation: multidimensional scaling for typological analysis. Theoretical Linguistics 34, 1–37. Crowley, Terry. 1996. Inalienable possession in Paamese grammar. In Hilary Chappell and William McGregor (eds.), The grammar of inalienability, 383–464. Berlin: De Gruyter Mouton. Cruse, Alan D. 1986. Lexical semantics. Cambridge University Press. Crystal, David. 2001. Language and the Internet. Cambridge University Press. Csa´bi, Szilvia. 2004. A cognitive linguistic view of polysemy in English and its implications for teaching. In Michel Achard and Suzanne Niemeier (eds.), Cognitive linguistics, second language acquisition and foreign language teaching, 233–56. Berlin: De Gruyter Mouton. Culicover, Peter W., and Ray Jackendoff. 1999. The view from the periphery: the English comparative correlative. Linguistic Inquiry 30, 543–71. 2005. Simpler syntax. Oxford University Press. Cˇulo, Oliver. 2003. Constructions-and-Frames analysis of translations: the interplay of syntax and semantics in translations between English and German. Constructions and Frames 5(2), 143–67. Culy, Christopher. 1996. Null objects in English recipes. Language Variation and Change 8, 91–124. 1997. Logophoric pronouns and point of view. Linguistics 35(5), 845–59. Curry, Kaitlin. 2010. ¿Pero Para? ¿Por Que´? The application of the principled polysemy model to por and para. MA thesis. Georgetown University. Curtis, Richard, Ben Elton, John Lloyd, and Rowan Atkinson. 1998. Blackadder: the whole damn dynasty. London: Penguin. Da˛browska, Ewa. 2000. From formula to schema: the acquisition of English questions. Cognitive Linguistics 11, 83–102. 2004. Language, mind and brain: some psychological and neurological constraints on theories of grammar. Edinburgh University Press. 2010. Naive v. expert competence: an empirical study of speaker intuitions. The Linguistic Review 27, 1–23. 2012. Different speakers, different grammars: individual differences in native language attainment. Linguistic Approaches to Bilingualism 2, 219–53. 2015a. Language in the mind and in the community. In Jocelyne Daems, Eline Zenner, Kris Heylen, Dirk Speelmand, and Hubert Cuyckens (eds.), Change of paradigms: new paradoxes, 221–35. Berlin: Walter de Gruyter. 2015b. What exactly is Universal Grammar, and has anyone seen it? Frontiers of Psychology 6(23), 852.
713
714
References
Da˛browska, Ewa, and Dagmar Divjak (eds.). 2015. Handbook of cognitive linguistics. Berlin: De Gruyter Mouton. Da˛browska, Ewa, and Elena Lieven. 2005. Towards a lexically specific grammar of children’s question constructions. Cognitive Linguistics 16, 437–74. Da˛browska, Ewa, and Marcin Szczerbin´ski. 2006. Polish children’s productivity with case marking: the role of regularity, type frequency, and phonological diversity. Journal of Child Language 33(3), 559–97. Da˛browska, Ewa, Caroline Rowland, and Anna Theakston. 2009. The acquisition of questions with long-distance dependencies. Cognitive Linguistics 20, 571–98. Dachkovsky, Svetlana, and Wendy Sandler. 2009. Visual intonation in the prosody of a sign language. Language and Speech 52(2–3), 287–314. Dale, Rick, and Michael J. Spivey. 2006. Unraveling the dyad: using recurrence analysis to explore patterns of syntactic coordination between children and caregivers in conversation. Language Learning 56, 391–430. Dale, Rick, Riccardo Fusaroli, Nicholas Duran, and Daniel C. Richardson. 2014. The self-organization of human interaction. In Brian H. Ross (ed.), The psychology of learning and motivation, 43–95. New York: Academic Press. Dalton-Puffer, Christiane, and Ingo Plag. 2000. Category-wise, some compound-type morphemes seem to be rather suffix-like: on the status of -ful, -type, and -wise in present-day English. Folia Linguistica 34, 225–44. Damasio, Antonio. 2010. Self comes to mind. New York: Pantheon. Damasio, Antonio and Hanna Damasio. 1994. Cortical systems for retrieval of concrete knowledge: the convergence zone framework. In C. Koch (ed.), Large-scale neuronal theories of the brain, 61–74. Cambridge, MA: MIT Press. Dancygier, Barbara. 1998. Conditionals and prediction: time, knowledge, and causation in conditional constructions. Cambridge University Press. 2005. Blending and narrative viewpoint: Jonathan Raban’s travels through mental spaces. Language and Literature 14(2), 99–127. (ed.). 2006. Special issue on conceptual blending. Language and Literature 15(1). 2007. Narrative anchors and the processes of story construction: the case of Margaret Atwood’s The Blind Assassin. Style 41(2), 133–52. 2008a. Review of the book Constructions of Intersubjectivity by A. Verhagen. Linguistics 46(3), 651–77. 2008b. The text and the story: levels of blending in fictional narratives. In Todd Oakley and Anders Hougaard (eds.), Mental spaces in discourse and interaction, 51–48. Amsterdam: John Benjamins. 2009. Genitives and proper names in constructional blends. In Vyvyan Evans and Ste´phanie Pourcel (eds.), New directions in cognitive linguistics, 161–84. Amsterdam: John Benjamins.
References
2011. Modification and constructional blends in the use of proper names. Constructions and Frames 3(2), 208–35. 2012a. The language of stories: a cognitive approach. Cambridge University Press. 2012b. Negation, stance verbs, and intersubjectivity. In Barbara Dancygier and Eve Sweetser (eds.), Viewpoint in language: a multimodal perspective, 69–93. Cambridge University Press. 2014. Intensity and texture in imagery. In Peter Stockwell and Sarah Whiteley (eds.), The Cambridge handbook of stylistics, 212–27. Cambridge University Press. 2016a. Concluding remarks: why viewpoint matters. In Barbara Dancygier, Wei-lun Lu, and Arie Verhagen (eds.), Viewpoint and the fabric of meaning: form and use of viewpoint tools across language and modalities, 281–88. Berlin: Walter de Gruyter. 2016b. Un-walling the wall: viewpoint and multimodality. In Peter Garratt (ed.), The cognitive humanities: embodied mind in literature and culture, 55–70. London: Palgrave Macmillan. 2017a. Figurativeness, conceptual metaphor and blending. In Elena Semino and Zso´fia Demje´n (eds.), The Routledge handbook of metaphor and language, 28–41. London: Routledge. 2017b. Viewpoint phenomena in constructions and discourse. Glossa: a Journal of General Linguistics. Special Collection: Perspective Taking. Dancygier, Barbara, and Eve Sweetser. 2005. Mental spaces in grammar: conditional constructions. Cambridge University Press. (eds.). 2012. Viewpoint in language: a multimodal perspective. Cambridge University Press. 2014. Figurative language. Cambridge University Press. Dancygier, Barbara, and Lieven Vandelanotte. 2016. Discourse viewpoint as network. In Barbara Dancygier, Wei-lun Lu, and Arie Verhagen (eds.), Viewpoint and the fabric of meaning: form and use of viewpoint tools across languages and modalities, 13–40. Berlin: De Gruyter Mouton. (eds.). 2017a. Viewpoint phenomena in multimodal communication. Special issue of Cognitive Linguistics 28(3). 2017b. Internet memes as multimodal constructions. In Viewpoint phenomena in multimodal communication. Special issue of Cognitive Linguistics 28(3). Dancygier, Barbara, Wei-lun Lu, and Arie Verhagen (eds.). 2016. Viewpoint and the fabric of meaning: form and use of viewpoint tools across language and modalities. Berlin: Walter de Gruyter. Daniels, Peter T., and William Bright (eds.). 1996. The world’s writing systems. New York: Oxford University Press. Danino, Charlotte. 2014. Language production and meaning construction mechanisms in the discourse on an ongoing event: the case study of CNN’s live broadcast on 9/11. PhD dissertation. University of Poitiers, France.
715
716
References
Danziger, Eve, and Alan Rumsey (eds.). 2013. Intersubjectivity: cultural limits, extensions and construals. Special issue of Language and Communication 33(3), 247–343. David, Oana, George Lakoff, and Elise Stickles. 2016. Cascades in metaphor and grammar: a case study of metaphors in the gun debate. Constructions and Frames 8(2): 214–253. Davies, Mark. 2008. The corpus of contemporary American English (COCA). www.americancorpus.org. Davis, Barbara L., and Peter F. MacNeilage. 2004. The frame/content theory of speech evolution: from lip smacks to syllables. Primatologie 6, 305–28. Davis, George H. 2013. Visual metaphors for teaching 3D geological thinking and interpretation. In Robert Krantz and D. Boonstra (eds.), 3D structural geologic interpretation: earth, mind, and machine. AAPG Hedberg Research Conference, Reno, Nevada. De Clerck, Bernard, and Timothey Colleman. 2013. ‘Het was massa’s lekker!’ Massa and massa’s in Flemish varieties of Dutch. Language Sciences 36, 147–60. de Cuypere, Ludovis. 2008. Limiting the iconic. Amsterdam: John Benjamins. De Deyne, Simon, and Gert Storms. 2015. Word associations. In J. R. Taylor (ed.), The Oxford handbook of the word, 465–80. Oxford University Press. de Hevia, Maria D., Ve´ronique Izard, Aure´lie Coubart, Elizabeth S. Spelke, and Arlette Streri. 2014. Representations of space, time, and number in neonates. Proceedings of the National Academy of Sciences 111(13), 4809–13. de Jorio, A. 1832 [2000]. Gesture in Naples and gesture in Classical Antiquity. A translation of La mimica degli antichi investigata nel gestire napoletano with introduction and notes by Adam Kendon. Bloomington: Indiana University Press. De Knop, Sabine, and Gae¨tanelle Gilquin (eds.). 2016. Applied Construction Grammar. Berlin: De Gruyter Mouton. De Knop, Sabine, Frank Boers, and Teun De Rycker (eds.). 2010. Applications of cognitive linguistics: exploring the lexis-grammar continuum in second language pedagogy. Berlin: De Gruyter Mouton. de la Fuente, JuanMa, Daniel Casasanto, Antonio Roma´n, and Julio Santiago. 2015. Can culture influence body-specific associations between space and valence? Cognitive Science 39, 821–32. de la Fuente, JuanMa, Julio Santiago, Antonio Roma´n, Cristina Dumitrache, and Daniel Casasanto. 2014. When you think about it, your past is in front of you: how culture shapes spatial conceptions of time. Psychological Science 25(9), 1682–90. de la Vega, Irmgard, Carolin Dudschig, Mo´nica De Filippis, Martin Lachmair, and Barbara Kaup. 2013. Keep your hands crossed: the valence-by-left/right interaction is related to hand, not side, in an
References
incongruent hand–response key assignment. Acta Psychologica 142(2), 273–77. de la Vega, Irmgard, Mo´nica De Filippis, Martin Lachmair, Carolin Dudschig, and Barbara Kaup. 2012. Emotional valence and physical space: limits of interaction. Journal of Experimental Psychology: Human Perception and Performance 38, 375–85. De Looze, Ce´line, Stefan Scherer, Brian Vaughan, and Nick Campbell. 2014. Investigating automatic measurements of prosodic accommodation and its dynamics in social interaction. Speech Communication 58, 11–34. De Meulder, Maartje. 2015. The legal recognition of sign languages. Sign Language Studies 15(4), 498–506. de Saussure, Ferdinand. [1916] 2006. Course in general linguistics. Ed. Charles Bally and Albert Sechehaye, trans. Roy Harris. La Salle, IL: Open Court. (Orig. pub. Cours de linguistique ge´ne´rale. Paris: Payot.) De Smet, Hendrik, and Jean-Christophe Verstraete. 2006. Coming to terms with subjectivity. Cognitive Linguistics 17, 365–92. de Sousa, Hila´rio. 2012. Generational differences in the orientation of time in Cantonese speakers as a function of changes in the direction of Chinese writing. Frontiers in Psychology 3(255), n. pag. de Vos, Connie. 2014. The Kata Kolok pointing system: morphemization and syntactic Integration. Topics in Cognitive Science 7(10), 150–68. de Vos, Connie, and Roland Pfau. 2015. Sign language typology: the contribution of rural sign languages. Annual Review of Linguistics 1(1), n. pag. Deese, James. 1965. The structure of associations in language and thought. Baltimore: John Hopkins. Dehaene, Stanislas, and Laurent Cohen. 2007. Cultural recycling of cortical maps. Neuron 56(2), 384–98. Dehaene, Stanislas, Serge Bossini, and Pascal Giraux. 1993. The mental representation of parity and number magnitude. Journal of Experimental Psychology: General 122, 371–96. Deignan, Alice. 1995. Collins Cobuild guides to English 7: Metaphor. London: Harper Collins. Deignan, Alice, and Liz Potter. 2004. A corpus study of metaphors and metonyms in English and Italian. Journal of Pragmatics 36(7), 1231–52. Deignan, Alice, Jeannette Littlemore, and Elena Semino. 2013. Figurative language, genre and register. Cambridge University Press. 2005. Metaphor and corpus linguistics. Amsterdam: John Benjamins. Delbecque, Nicole. 1990. Word order as a reflection of alternate conceptual construals in French and Spanish: similarities and divergences in adjective position. Cognitive Linguistics 1, 349–416. DeLoache, Judy S. 2004. Becoming symbol-minded. Trends in Cognitive Sciences 8(2), 66–70.
717
718
References
DeMatteo, Asa. 1977. Visual imagery and visual analogues in American Sign Language. In Lynn A. Friedman (ed.), On the other hand: new perspectives on American Sign Language, 109–36. Cambridge, MA: Academic Press. Demeter, Gusztav. 2011. Explicit apologies in English and Romanian: a construction grammar approach. PhD dissertation. Oklahoma State University. 2016. On discourse-motivated ‘sorries’: fictive apologies in English, Hungarian, and Romanian. In Esther Pascual and Sergeiy Sandler (eds.), The conversation frame: forms and functions of fictive interaction, 151–68. Amsterdam: John Benjamins. Demje´n, Zso´fia. 2015. Sylvia Plath and the language of affective states: written discourse and the experience of depression. London: Bloomsbury. Demje´n, Zso´fia, and Elena Semino. 2017. Using metaphor in healthcare: physical health interventions. In Elena Semino and Zsofia Demje´n (eds.), The Routledge handbook of metaphor and language, 385–99. London: Routledge. Demmen, Jane, Elena Semino, Zso´fia Demje´n, Veronika Koller, Andrew Hardie, Paul Rayson, and Sheila Payne. 2015. A computerassisted study of the use of violence metaphors for cancer and end of life by patients, family carers and health professionals. International Journal of Corpus Linguistics 20(2), 205–31. Den Dikken, Marcel. 2005. Comparative correlatives comparatively. Linguistic Inquiry 36, 497–532. Denis, Michel. 1997. The description of routes: a cognitive approach to the production of spatial discourse. Cahiers de Psychologie Cognitive 16(4), 409–58. Denison, David. 2010. Log(ist)ic and simplistic S-curves. In Raymond Hickey (ed.), Motives for language change, 54–70. Cambridge University Press. Dennett, Daniel C. 1991. Consciousness explained. Boston: Little Brown. Denroche, Charles. 2015. Metonymy and language: a new theory of linguistic processing. London: Routledge. Deppermann, Arnulf. 2006. Construction grammar – eine Grammatik fu¨r die Interaktion? In Arnulf Deppermann, Reinhold Fiehler, and Thomas Spranz-Fogasy (eds.), Grammatik und Interaktion, 43–65. Radolfzell: Verlag fu¨r Gespra¨chsforschung. 2011. Konstruktionsgrammatik und Interaktionale Lingui-stik: Affinita¨ten, Komplementarita¨ten und Diskrepanzen. In Alexander Ziem and Alexander Lasch (eds.), Konstruktionsgrammatik III : Aktuelle Fragen und Lo¨sungsansa¨tze, 205–38. Tu¨bingen: Stauffenburg. 2013. Introduction: multimodal interaction from a conversation analytic perspective. Journal of Pragmatics 46, 1–7. Desagulier, Guillaume. 2014. Visualizing distances in a set of nearsynonyms: rather, quite, fairly, and pretty. In Dylan Glynn and Justyna
References
Robinson (eds.), Polysemy and synonymy: corpus methods and applications in Cognitive Linguistics, 145–78. Amsterdam: John Benjamins. Desai, R., J. Binder, L. Conant, Q. Mano, and M. Seidenberg. 2011. The neural career of sensory-motor metaphors. Journal of Cognitive Neuroscience 23, 2376–86. Descartes, Rene´. 1641 [1980]. Discourse on method and meditations on first philosophy. Trans. Donald A. Cress. Indianapolis: Hackett. Deshors, Sandra C. and Stefan Th. Gries. 2016. Profiling verb complementation constructions across New Englishes: a two-step random forests analysis to ing vs. to complements. International Journal of Corpus Linguistics 21(2), 192–218. Di Giuseppantonio Di Franco, Paola, Justin L. Matthews, and Teenie Matlock. 2015. Framing the past: how virtual experience affects bodily descriptions of artefacts. Journal of Cultural Heritage 17, 179–87. Diemer, Stefan. 2013. Recipes and food discourse in English – a historical menu. In Cornelia Gerhardt, Maximiliane Frobenius, and Susanne Ley (eds.), Culinary linguistics: the chef’s special, 139–56. Amsterdam: Benjamins. Diemer, Stefan, and Maximiliane Frobenius. 2013. When making a pie, all ingredients must be chilled. Including you: lexical, syntactic and interactive features of online discourse – a synchronic study of food blogs. In Cornelia Gerhardt, Maximiliane Frobenius, and Susanne Ley (eds.), Culinary linguistics: the chef’s special, 53–81. Amsterdam: Benjamins. Diessel, Holger. 2004. The acquisition of complex sentences. Cambridge University Press. 2008. Iconicity of sequence: a corpus-based analysis of the positioning of temporal adverbial clauses in English. Cognitive Linguistics 19, 465–90. 2009. On the role of frequency and similarity in the acquisition of subject and non-subject relative clauses. In Talmy Givo´n and Masayoshi Shibatani (eds.), Syntactic complexity, 251–76. Amsterdam: John Benjamins. 2013. Construction Grammar and first language acquisition. In Thomas Hoffmann and Graeme Trousdale (eds.), The Oxford handbook of Construction Grammar, 347–64. Oxford University Press. Diessel, Holger, and Michael Tomasello. 2001. The acquisition of finite complement clauses in English: a corpus-based analysis. Cognitive Linguistics 12(2), 97–142. Diewald, Gabriele. 2007. Abto¨nungspartikel. In Ludger Hoffmann (ed.), Handbuch der deutschen Wortarten, 117–41. Berlin: Walter de Gruyter. Dingemanse, Mark, and N. J. Enfield. 2015. Other-initiated repair across languages: towards a typology of conversational structures. Open Linguistics 1(1), n. pag. Dingemanse, Mark, S. G. Roberts, J. Baranova, J. Blythe, P. Drew, S. Floyd, et al. 2015. Universal principles in the repair of communication problems. PLoS ONE 10(9), e0136100.
719
720
References
Dirven, Rene´. 1993 [2003]. Metonymy and metaphor: different mental strategies of conceptualisation. In Rene´ Dirven and Ralf Po¨rings (eds.), Metaphor and metonymy in comparison and contrast, 75–112. Berlin: De Gruyter Mouton. Dittmar, Miriam, Kirsten Abbot-Smith, Elena Lieven, and Michael Tomasello. 2008. German children’s comprehension of word order and case marking in causative sentences. Child Development, 79, 1152–67. Divjak, Dagmar S. 2006. Ways of intending: delineating and structuring near synonyms. In Stefan Th. Gries and Anatol Stefanowitsch (eds.), Corpora in cognitive linguistics: corpus-based approaches to syntax and lexis, 19–56. Berlin: De Gruyter Mouton. Divjak, Dagmar S., and Antti Arppe. 2010. Extracting prototypes from corpus data: a distributional account of representing nearsynonymous verbs. Paper presented at the interdisciplinary workshop on verbs, ‘The identification and representation of verb features’. Pisa. Divjak, Dagmar S., and Stefan Th. Gries. 2006. Ways of trying in Russian: clustering behavioral profiles. Corpus Linguistics and Linguistic Theory 2(1), 23–60. 2008. Clusters in the mind? Converging evidence from near synonymy in Russian. The Mental Lexicon 3(2), 188–213. 2009. Corpus-based cognitive semantics: a contrastive study of phasal verbs in English and Russian. In Katarzyna Dziwirek and Barbara Lewandowska-Tomaszczyk (eds.), Studies in cognitive corpus linguistics, 273–96. Frankfurt am Main: Peter Lang. (eds.). 2012. Frequency effects in language representation. Berlin: De Gruyter Mouton. 2015. Four challenges for usage-based linguistics. In Jocelyne Daems, Eline Zenner, Kris Heylen, Dirk Speelman, and Hubert Cuyckens (eds.). Change of paradigms – new paradoxes: recontextualizing language and linguistics, 297–310. Berlin: De Gruyter Mouton. Dodge, Ellen. 2016. A deep semantic corpus-based approach to metaphor analysis: a case study of metaphoric conceptualizations of poverty. Constructions and Frames 8(2): 254–292. Dodge, Ellen, Jisup Hong, and Elise Stickles. 2015. MetaNet: deep semantic automatic metaphor analysis. NAACL HLT June 2015, 40. Dodwell, C. Reginald. 2000. Anglo-Saxon gestures and the Roman stage. Cambridge University Press. Dolbey, Andrew. 2009. BioFrameNet: a FrameNet extension to the domain of molecular biology. PhD dissertation. University of California, Berkeley. Dolscheid, Sarah, Sabine Hunnius, Daniel Casasanto, and Asifa Majid. 2014. Prelinguistic infants are sensitive to space-pitch associations found across cultures. Psychological Science 25(6), 1256–61.
References
Dolscheid, Sarah, Shakila Shayan, Asifa Majid, and Daniel Casasanto. 2013. The thickness of musical pitch: psychophysical evidence for linguistic relativity. Psychological Science 24(5), 613–21. Dominey, Peter, and Michael Hoen. 2006. Structure mapping and semantic integration in a construction-based neurolinguistic model of sentence processing. Cortex 42, 476–79. Donegan, Patricia J., and David Stampe. 2009. Hypotheses of natural phonology. Poznan´ Studies in Contemporary Linguistics 45(1), 1–31. 1979. The study of natural phonology. In Dan Dinnsen (ed.), Current approaches to phonological theory, 126–73. Bloomington: Indiana University Press. Dornelas, Aline, and Esther Pascual. 2016. Echolalia as communicative strategy: fictive interaction in the speech of children with autism. In Esther Pascual and Sergeiy Sandler (eds.), The conversation frame: forms and functions of fictive interaction, 343–61. Amsterdam: John Benjamins. Dorst, Aletta G. 2015. More or different metaphors in fiction? A quantitative cross-register comparison. Language and Literature 24(1), 3–22. Downing, Laura J. 2006. Canonical forms in prosodic morphology. Oxford University Press. Dowty, David R. 1986. The effect of aspectual class on the temporal structure of discourse: semantics or pragmatics? Linguistics and Philosophy 1(9), 37–62. Dryer, Matthew S., and Martin Haspelmath (eds.). 2013. The world atlas of language structures online. Leipzig: Max Planck Institute for Evolutionary Anthropology. Du Bois, John W. 1985. Competing motivations. In John Haiman (ed.), Iconicity in syntax, 343–65. Amsterdam: John Benjamins. 2007. The stance triangle. In Robert Englebretson (ed.), Stancetaking in discourse: subjectivity, evaluation, interaction, 139–82. Amsterdam: John Benjamins. 2014. Towards a dialogic syntax. Cognitive Linguistics 25, 359–410. Ducrot, Oswald. 1996. Slovenian lectures: argumentative semantics/confe´rences slove`nes: se´mantiques argumentatives. Ljubljana: ISH Insˇtitut za humanisticˇne sˇtudije Ljubljana. Dudis, Paul G. 2004. Body partitioning and real-space blends. Cognitive Linguistics 15(2), 223–38. Duffau, H., P. Gatignol, E. Mandonnet, P. Peruzzi, N. Tzourio-Mazoyer, and L. Capelle. 2005. New insights into the anatomo-functional connectivity of the semantic system: a study using cortico-subcortical stimulations. Brain 128, 797–810. Duffy, Sarah E., and Michele I. Feist. 2014. Individual differences in the interpretation of ambiguous statements about time. Cognitive Linguistics 25(1), 29–54.
721
722
References
Duijn, Max J. van, and Arie Verhagen. In press. Beyond triadic communication: a three-dimensional conceptual space for modeling intersubjectivity. In D. Glynn and K. Krawczak (eds.), Subjectivity and stance: usagebased studies in epistemic structuring. Amsterdam: John Benjamins. Dunning, Ted. 1993. Accurate methods for the statistics of surprise and coincidence. Computational Linguistics 19(1), 61–74. Earis, Helen, and Kearsy Cormier. 2013. Point of view in British Sign Language and spoken English narrative discourse: the example of ‘The Tortoise and the Hare.’ Language and Cognition 5(4), 313–43. Ebbinghaus, Hermann. 1885 [1913]. Memory: a contribution to experimental psychology New York: Teachers College, Columbia. Eckert, Penelope. 2012. Three waves of variation study: the emergence of meaning in the study of sociolinguistic variation. Annual Review of Anthropology 41, 87–100. Eckhoff, Hanne M., and Laura A. Janda. 2014. Grammatical profiles and aspect in Old Church Slavonic. Transactions of the Philological Society 112(2), 231–58. Egorova, Ekaterina, Thora Tenbrink, and Ross Purves. 2015. Where snow is a landmark: Route direction elements in alpine contexts. Spatial Information Theory: 12th International Conference (COSIT 2015), Santa Fe, October 12–16, 2015, 175–95. Berlin: Springer. Ehlich, Konrad. 1986. Interjektionen. Tu¨bingen: Niemeyer. Ehrich, Veronika, and Charlotte Koster. 1983. Discourse organization and sentence form: the structure of room descriptions in Dutch. Discourse Processes 6, 169–95. Ekman, Paul, and Wallace V. Friesen. 1969. The repertoire of nonverbal behavior: categories, origins, usage, and coding. Semiotica 1(1), 49–98. Ellis, Andrew W. 1985–87. Progress in the Psychology of Language, 3 vols. London: Lawrence Erlbaum. Ellis, Nick C. 2002. Frequency effects in language processing: a review with implications for theories of implicit and explicit language acquisition. Studies in Second Language Acquisition 24(2), 143–88. 2003. Constructions, chunking and connectionism: the emergence of second language structure. In Catherine Doughty and Michael H. Long (eds.), Handbook of second language acquisition, 63–103. Oxford: Blackwell. 2006a. Cognitive perspectives on SLA: the associative cognitive CREED. AILA Review 19: 100–21. 2006b. Language acquisition as rational contingency learning. Applied Linguistics 27(1). 1–24. 2008. Usage-based and form-focused language acquisition: the associative learning of constructions, learned attention and the limited L2 end state. In Peter Robinson and Nick C. Ellis (eds.) Handbook of
References
cognitive linguistics and second language acquisition. 372–406. London: Routledge. 2012. What can we count in language, and what counts in language acquisition, cognition, and use? In Stefan Th. Gries and Dagmar S. Divjak (eds.), Frequency effects in language learning and processing, 7–33. Berlin and New York: De Gruyter Mouton. 2013. Construction grammar and second language acquisition. In Thomas Hoffmann and Graeme Trousdale (eds.), The Oxford handbook of construction grammar, 365–78. Oxford University Press. Ellis, Nick C., and Fernando Ferreira-Junior. 2009. Constructions and their acquisition: islands and the distinctiveness of their occupancy. Annual Review of Cognitive Linguistics 7, 187–220. Ellis, Nick C., and Stefanie Wulff. 2015. Second language acquisition. In Ewa Da˛browska and Divjak Dagmar (eds.), Handbook of cognitive linguistics, 409–31. Berlin and New York: De Gruyter Mouton. Ellis, Nick C., and Teresa Cadierno. 2009. Constructing a second language: introduction to the special section. Annual Review of Cognitive Linguistics 7, 111–39. Ellis, Nick C., Matthew B. O’Donnell, and Ute Ro¨mer. 2014a. The processing of verb-argument constructions is sensitive to form, function, frequency, contingency, and prototypicality. Cognitive Linguistics 25(1). 55–98. 2014b. Second language verb-argument constructions are sensitive to form, function, frequency, contingency, and prototypicality. Linguistic Approaches to Bilingualism 4(4). 405–31. Ellsworth, Michael, Kyoko Ohara, Carlos Subirats, and Thomas Schmidt. 2006. Frame-semantic analysis of motion scenarios in English, German, Spanish, and Japanese. Paper presented at the Fourth International Conference on Construction Grammar, Tokyo, Japan. Available at http://jfn.st.hc.keio.ac.jp/publications/HoundICCG4.pdf. Elman, Jeffrey L., and James L. McClelland. 1984. Speech perception as a cognitive process: the interactive activation model. In Norman Lass (ed.), Speech and language, vol. 10, 337–74. New York: Academic Press. Emmorey, Karen. 2002. Language, cognition, and the brain: Insights from sign language research. Mahwah, NJ: Lawrence Erlbaum and Associates. Emmorey, Karen, and Shannon Casey. 2001. Gesture, thought and spatial language. Gesture, 1(1), 35–50. Emmorey, Karen, Barbara Tversky, and Holly A. Taylor. 2000. Using space to describe space: perspective in speech, sign, and gesture. Spatial Cognition and Computation 2, 157–80. Enfield, N. J. (ed.) 2002. Ethnosyntax: explorations in grammar and culture. Oxford University Press. 2003. Linguistic epidemiology: semantics and grammar of language contact in mainland Southeast Asia. Abingdon: RoutledgeCurzon.
723
724
References
2009. The anatomy of meaning: speech, gesture, and composite utterances. Cambridge University Press. 2013. Relationship thinking: agency, enchrony, and human Sociality. Oxford University Press. 2014. Natural causes of language: frames, biases, and cultural transmission. Berlin: Language Science Press. 2015. The utility of meaning: what words mean and why. Oxford University Press. Enfield, N. J., and Stephen C. Levinson. 2006a. Introduction: human sociality as a new interdisciplinary field. In N. J. Enfield and Stephen C. Levinson (eds.), Roots of human sociality: culture, cognition and interaction, 1–35. Oxford: Berg. (eds.) 2006b. Roots of human sociality: culture, cognition and interaction. Oxford: Berg. Engelberg, Stefan, Svenja Ko¨nig, Kristel Proost, and Edeltraud Winkler. 2011. Argumentstrukturmuster als Konstruktionen? In S. Engelberg, A. Holler, and K. Proost (eds.), Sprachliches Wissen zwischen Lexikon und Grammatik, 71–112. Berlin: De Gruyter. Ericsson, K. Anders, and Herbert A. Simon. 1984. Protocol analysis – verbal reports as data. Cambridge, MA: Bradford Books. Erman, Adolf. 1971. The literature of the ancient Egyptians: Poems, narratives, and manuals of instruction from the third and second millennia B.C. Aylward M Blackman (trans.). New York: Benjamin Blom. Erting, Carol and James Woodward. 1975. Synchronic variation and historical change in American Sign Language. Language Sciences, 37(Oct), 9–12. Eskildsen, Søren Wind. In prep. L2 constructions and interactional competence: subordination and coordination in English L2 learning. In Andrea Tyler, Lihong Huang and Hana Jan (eds.), What is applied cognitive linguistics? Answers from current SLA research. Berlin: De Gruyter Mouton. Eskildsen, Søren Wind, and Johannes Wagner. 2015. Embodied L2 construction learning. Language Learning 65, 419–48. Evans, Jonathan S. B. T. 2005. The social and communicative function of conditional statements. Mind and Society 4(1), 97–113. Evans, Nicholas. 1995. View with a view: towards a typology of multiple perspective constructions. In Rebecca T. Cover and Yuni Kim (eds.), Proceedings of the thirty-first annual meeting of the Berkeley Linguistics Society: general session and parasession on prosodic variation and change, 93–120. Berkeley: Berkeley Linguistics Society. 2010. Semantic typology. In Jae Jung Song (ed.), The Oxford handbook of typology, 504–33. Oxford University Press. Evans, Nicholas, and Stephen C. Levinson. 2009a. The myth of language universals: Language diversity and its importance for cognitive science. Behavioral and Brain Sciences 32(5), 429–92.
References
2009b. With diversity in mind: freeing the language sciences from Universal Grammar. Behavioral and Brain Sciences 32, 472–84. Evans, Nicholas, and David Wilkins. 2000. In the mind’s ear: the semantic extensions of perception verbs in Australian languages. Language 76(3), 546–92. Evans, Vyvyan. 2004. The structure of time: language, meaning and temporal cognition. Amsterdam: John Benjamins. 2009. How words mean: Lexical concepts, cognitive models and meaning construction. Oxford University Press. Evans, Vyvyan, and Melanie Green. 2006. Cognitive linguistics: an introduction. Lawrence Erlbaum Associates Publishers. Evans, Vyvyan, Benjamin K. Bergen, and Jo¨rg Zinken. 2007. The Cognitive Linguistics enterprise: an overview. In Vyvyan Evans, Benjamin K. Bergen, and Jo¨rg Zinken (eds.), The cognitive linguistics reader: advances in cognitive linguistics, 2–36. London: Equinox. Everett, Daniel L. 2003 [2008]. Wari’ intentional state constructions. In Robert Van Valin (ed.), Investigations of the syntax-semantics-pragmatics interface, 381–411. Amsterdam: John Benjamins. 2005. Cultural constraints on grammar and cognition in Piraha˜ (including commentary). Current Anthropology 46, 621–46. 2011. 26 September. Wari’. [Electronic mailing list message]. Retrieved from: http://listserv.linguistlist.org/cgi-bin/wa?A2=ind1109andL=FUNK NETandF=andS=andP=19496. Fairclough, Norman. 1994. Conversationalisation of public discourse and the authority of the consumer. In Russell Keat, Nigel Whiteley, and Nicholas Abercrombie (eds.), The authority of the consumer, 253–68. London: Routledge. Falck, Marlene Johansson, and Raymond W. Gibbs, Jr. 2012. Embodied motivations for metaphorical meanings. Cognitive Linguistics 23, 251–72. Fauconnier, Gilles. 1985. Mental spaces. Cambridge, MA: MIT Press. 1994. Mental spaces: aspects of meaning construction in natural languages. 2nd edn. Cambridge University Press. 1997. Mappings in thought and language. Cambridge University Press. Fauconnier, Gilles, and George Lakoff. 2013. On metaphor and blending. Journal of Cognitive Semiotics (2), 393–99. Fauconnier, Gilles, and Eve Sweetser (eds.). 1996. Spaces, worlds and grammar. Chicago: University of Chicago Press. Fauconnier, Gilles, and Mark Turner. 1996. Blending as a central process of grammar. In Adele E. Goldberg (ed.), Conceptual structure, discourse, and language. Stanford: Center for the Study of Language and Information Publications. 1998. Conceptual integration networks. Cognitive Science 2(1), 133–87. 2000. Compression and global insight. Cognitive Linguistics 11(3/4), 283–304. 2002. The way we think: Conceptual blending and the mind’s hidden complexities. Basic Books.
725
726
References
2008. Rethinking metaphor. In Raymond W. Gibbs, Jr. (ed.), Cambridge handbook of metaphor and thought, 53–66. Cambridge University Press. Faulhaber, Susen. 2011. Verb valency patterns. Berlin/New York: De Gruyter Mouton. Fay, Nicholas, Casey J. Lister, T. Mark Ellison, and Susan Goldin-Meadow. 2014. Creating a communication system from scratch: gesture beats vocalization hands down. Frontiers in Psychology 5, 1–12. Feagin, Crawford. 2013. Entering the community: fieldwork. In J. K. Chambers and Natalie Schilling (eds.), The handbook of language variation and change, 2nd edn, 20–39. Chichester: WileyBlackwell. Fedden, Sebastian and Lera Boroditsky. 2012. Spatialization of time in Mian. Frontiers in Psychology 3: 485. Feldman, Jerome. 2006. From molecules to metaphors. A neural theory of language. Cambridge, MA: MIT Press. Feldman, Jerome, and Srini Narayanan. 2004. Embodied meaning in a neural theory of language. Brain and Language 89(2), 385–92. Fellbaum, Christiane (ed.). 1998. WordNet: an electronic lexical database. Cambridge, MA: MIT Press. 2015. Lexical relations. In John R. Taylor (ed.), The Oxford handbook of the word, 350–63. Oxford University Press. Ferguson, Charles A. 1982. Simplified registers and linguistic theory. In L. Obler and L. Menn (eds.), Exceptional language and linguistics. New York: Academic Press. Ferrara, Lindsay, and Trevor Johnston. 2014. Elaborating who’s what: a study of constructed action and clause structure in Auslan (Australian Sign Language). Australian Journal of Linguistics 34(2), 193–215. Ferrari, Lillian and Eve Sweetser. 2012. Subjectivity and upwards projection in mental space structure. In Barbara Dancygier, Eve Sweetser (eds.), Viewpoint in language. A multimodal perspective, 47–68. Cambridge University Press. Ferrer i Cancho, Ramon, and Richard Sole´. 2001. The small-world of human language. SFI Working Paper 268(1482), 2261–65. Feyaerts, Kurt. 2013a. A cognitive grammar of creativity. In Tony Veale, Kurt Feyaerts, and Charles Forceville (eds.), Creativity and the agile mind: a multidisciplinary approach to a multifaceted phenomenon, 205–27. Berlin: De Gruyter Mouton. 2013b. Tackling the complexity of spontaneous humorous interaction: an integrated classroom-modeled corpus approach. In Leonor Ruiz Gurillo and M. Bele´n Alvarado Ortega (eds.), Irony and humor: from pragmatics to discourse, 243–68. John Benjamins. Feyaerts, Kurt, and Geert Broˆne. 2005. Expressivity and metonymic inferencing: stylistic variation in non-literary language use. Style 39(1), 12–35.
References
Feyaerts, Kurt, and Bert Oben. 2014. Tracing down schadenfreude in spontaneous interaction: evidence from corpus linguistics. In Wilco van Dijk and Jaap Ouwerkerk (eds.), Schadenfreude: understanding pleasure at the misfortune of others, 275–91. Cambridge University Press. Feynman, Richard. 1992. Surely you’re joking, Mr. Feynman! London: Vintage. Field, Andy. 2013. Discovering statistics using IBM SPSS statistics. London: Sage. Fillmore, Charles J. 1968. The case for case. In E. Bach and R. Harms (eds.), Universals in Linguistic Theory, 1–90. New York: Rinehart and Winston. 1975. An alternative to checklist theories of meaning. Proceedings of the First Annual Meeting of the Berkeley Linguistic Society (BLS), 123–31. 1976. Frame semantics and the nature of language. Annals of the New York Academy of Sciences: conference on the Origin and Development of Language and Speech 280, 20–32. 1977a. The case for case reopened. In Peter Cole (ed.), Grammatical relations, syntax and semantics 8, 59–81. New York: Academic Press. 1977b. Topics in Lexical Semantics. In Peter Cole (ed.), Current issues in linguistic theory, 76–136. Bloomington: Indiana University Press. 1977c. Scenes-and-frames semantics. In A. Zampolli (ed.), Fundamental Studies in Computer Science, 55–88. Dordrecht: North Holland Publishing. 1978. On the organization of semantic information in the lexicon. Papers from the Parasession on the Lexicon, Chicago Linguistic Society, 148–73. 1979. Innocence: a second idealization for linguistics. Proceedings of the Fifth Annual Meeting of the Berkeley Linguistic Society (BLS), 63–76. 1982. Frame semantics. In Linguistic Society of Korea (ed.), Linguistics in the morning calm, 111–38. Seoul: Hanshin. 1985a. Syntactic intrusions and the notion of grammatical construction. Berkeley Linguistic Society 11, 73–86. 1985b. Frames and the semantics of understanding. Quaderni di Semantica 6(2), 222–54. 1986. Pragmatically controlled zero anaphora. Proceedings of the Berkeley Linguistics Society, 95–107. 1988. The mechanisms of construction grammar. Proceedings of the Annual Meeting of the Berkeley Linguistics Society 14, 35–55. 1990. Epistemic stance and grammatical form in English conditional sentences. In Michael Ziolkowski, Manuela Noske, and Karen Deaton (eds.), Papers from the 26th Regional Meeting of the Chicago Linguistic Society, 137–62. Chicago Linguistic Society. 2008. Border Conflicts: FrameNet meets Construction Grammar. Proceedings of the XIII EURALEX International Congress, 49–68. Barcelona, 15–19 July 2008. 2013. Berkeley construction grammar. In Thomas Hoffmann and Graeme Trousdale (eds.), The Oxford handbook of construction grammar, 111–32. Oxford University Press.
727
728
References
Fillmore, Charles J., and Beryl T. S. Atkins. 2000. Describing polysemy: the case of ‘crawl.’ In Yael Ravin and Claudia Leacock (eds.), Polysemy: theoretical and computational approaches, 89–110. Oxford University Press. 1992. Toward a frame-based lexicon: the semantics of RISK and its neighbors. In A. Lehrer and E. Kittay (eds.), Frames, Fields and Contrasts: New Essays in Semantic and Lexical Organization, 75–102. Hillsdale: Erlbaum. Fillmore, Charles J., and Colin Baker. 2010. A frames approach to semantic analysis. In Bernd Heine and Heiko Narrog (eds.), The Oxford handbook of linguistic analysis, 313–40. Oxford University Press. Fillmore, Charles J., and Paul Kay. 1993. Construction grammar coursebook. Unpublished manuscript. University of California, Berkeley. 1995. Construction grammar. Unpublished manuscript. University of California, Berkeley. 1999. Grammatical constructions and linguistic generalizations: the what’s X doing Y? construction. Language 75, 1–33. Fillmore, Charles J., Chris Johnson, and Miriam Petruck. 2003. Background to FrameNet. International Journal of Lexicography 16, 235–51. Fillmore, Charles J., Paul Kay, and Mary Catherine O’Connor. 1988. Regularity and idiomaticity in grammatical constructions: the case of let alone. Language 64, 501–38. Fillmore, Charles J., Russell Lee-Goldman, and Russell Rhomieux. 2012. The FrameNet Constructicon. In Hans C. Boas and Ivan Sag (eds.), Signbased Construction Grammar, 309–72. Stanford: CSLI Publications. Filppula, Markku, Juhani Klemola, and Heli Paulasto (eds.). 2009. Vernacular universals and language contacts: evidence from varieties of English and beyond. London: Routledge. Fink, Gernot A., Michaela Johanntokrax, and Brigitte Schaffranietz. 1995. A flexible formal language for the orthographic transcription of spontaneous spoken dialogues. Proceedings of the European Conference on Speech Communication and Technology, 871–74. Madrid. Fintel, Kai. von. 2011. Conditionals. In K. von Heusinger, C. Maienborn, and P. Portner (eds.), Handbooks of linguistics and communication science, 1515–38. Berlin: De Gruyter. Firth, John R. 1948. Sounds and prosodies. Transactions of the Philological Society 47(1), 127–52. 1957. Papers in linguistics 1934–1951. Oxford University Press. Fischer, Kerstin. 2000a. From cognitive semantics to lexical pragmatics: the functional polysemy of discourse particles. Berlin: De Gruyter Mouton. 2000b. Discourse particles, turn-taking, and the semantics–pragmatics interface. Revue de Se´mantique et Pragmatique 8, 111–37. 2002. The interactive constitution of word meaning: participant strategies in defining Cognitive and Cognitive Linguistics. Revue de Se´mantique et Pragmatique 12, 145–76.
References
2006. What computer talk is and isn’t. Saarbru¨cken: AQ. 2010. Beyond the sentence: constructions, frames and spoken interaction. Constructions and Frames 2, 185–207. 2013. The addressee in the recipe. In Cornelia Gerhardt, Maximiliane Frobenius, and Susanne Ley (eds.), Culinary linguistics: the chef’s special, 103–17. Amsterdam: John Benjamins. 2015a. Conversation, construction grammar, and cognition. Language and Cognition 7, 563–88. 2015b. Situation in grammar or in frames? Evidence from the so-called baby talk register. Constructions and Frames 7(2), 258–88. 2016. Designing speech for a recipient: the roles of partner modeling, alignment and feedback in so-called ‘simplified registers’. Amsterdam: John Benjamins. 2017. The situatedness of a pragmatic act: explaining a lamp to a robot. In A. Capone (ed.). Pragmemes: Festschrift for Jacob Mey. Amsterdam: John Benjamins. Fischer, Kerstin, Katrin S. Lohan, Joe Saunders, Chrystopher Nehaniv, Britta Wrede, and Katharina Rohlfing. 2013. The impact of the contingency of robot feedback on HRI. Cooperative Technological Systems, San Diego, May 20–24. Fischer, Kerstin, Katrin S. Lohan, Katharina Rohlfing, and Kilian Foth. 2014. Partner orientation in asymmetric communication: evidence from contingent robot response. HRI ’14 Workshop on Humans and Robots in Asymmetric Interactions, March 3rd, 2014, Bielefeld, Germany. Fischer, Martin H. 2008. Finger counting habits modulate spatial-numerical associations. Cortex 44(4), 386–92. Fischer, Martin H., Richard A. Mills, and Samuel Shaki. 2010. How to cook a SNARC: number placement in text rapidly changes spatial– numerical associations. Brain and Cognition 72(3), 333–336. Fischer, Olga. 2003. The development of the modals in English. Radical versus gradual changes. In David Hart (ed.), English modality in context, 16–32. Bern: Peter Lang. Fisher, Mary F. K. 1983. The Anatomy of a Recipe. In With Bold Knife and Fork, p 13–24. Paragon. Fishman, Joshua A. 1991. Putting the ‘socio’ back into the sociolinguistic enterprise. International Journal of the Sociology of Language 92, 127–38. Fludernik, Monika. 1993. The fictions of language and the languages of fiction: the linguistic representation of speech and consciousness. London: Routledge. Flumini, Andrea, and Julio Santiago. 2013. Time (also) flies from left to right . . . if it is needed! In M. Knauff, M. Pauen, N. Sebanz, and I. Wachsmuth (eds.), Proceedings of the 35th Annual Conference of the Cognitive Science Society, 2315–20. Austin: Cognitive Science Society. Fodor, Jerry. 1983. The modularity of mind. Cambridge, MA: MIT Press. Fontenelle, Thierry. 1997. Using a bilingual dictionary to create semantic networks. International Journal of Lexicography 10.4. 275–303.
729
730
References
Forceville, Charles J. 2002. The identification of target and source in pictorial metaphors. Journal of Pragmatics 34(1), 1–14. Forceville, Charles J., and Eduardo Urios-Aparisi (eds.). 2009. Multimodal metaphor. Berlin: De Gruyter Mouton. Ford, Cecilia E. 1993. Grammar in interaction: adverbial clauses in American English conversations. Cambridge University Press. 2003. Social interaction and grammar. In Michael Tomasello (ed.), The new psychology of language: cognitive and functional approaches to language structure, 119–43. Mahwah: Lawrence Erlbaum. 2004. Contingency and units in interaction. Discourse Studies 6(1), 27–52. Ford, Cecilia E., Barbara Fox, and Sandra A. Thompson. 1996. Practices in the Construction of Turns: the ‘TCU’ revisited. Pragmatics 6, 3, 427–454. Forsbert, Markus, Richard Johansson, Linnea Ba¨ckstro¨m, Lars Borin, Benjamin Lyngfelt, Joel Olofsson, and Julia Prentice. 2014. From construction candidates to constructicon entries. Constructions and Frames 6.1, 114–135. Fortescue, Michael. 2001. Thoughts about thought. Cognitive Linguistics 12, 15–46. Fortuin, Egbert, and Boogaart, Ronny. 2009. Imperative as conditional: from constructional to compositional semantics. Cognitive Linguistics 20(4), 641–73. Fowler, Carol A., Julie M. Brown, Laura Sabadini, and Jeffrey Weihing. 2003. Rapid access to speech gestures in perception: evidence from choice and simple response time tasks. Journal of Memory and Language 49, 396–413. Fox, Barbara A. 2007. Principles shaping grammatical practices: an exploration. Discourse Studies 9, 299–318. FrameNet. 2015. FrameNet Release 1.6. Freely available as a download from the FrameNet project. Updated data available at https://framenet.icsi .berkeley.edu/fndrupal/home (accessed July 2015). Francis, Elaine J., and Laura A. Michaelis. 2003. Mismatch: a crucible for linguistic theory. In Elaine J. Francis and Laura A. Michaelis (eds.), Mismatch: form-function incongruity and the architecture of grammar, 1–30. Stanford: CSLI. Frank, Mark G., and Thomas Gilovich. 1988. The dark side of self- and social perception: black uniforms and aggression in professional sports. Journal of Personality and Social Psychology 54(1), 74–85. Frayman, Bruce J., and William L. Dawson. 1981. The effect of object shape and mode of presentation of judgments of apparent volume. Perception and Psychophysics 29(1), 56–62. Freeman, Margaret H. 2002. The body in the word: a cognitive approach to the shape of a poetic text. In Elena Semino and Jonathan Culpeper (eds.). Cognitive stylistics: language and cognition in text analysis, 23–47. Amsterdam: John Benjamins.
References
2009. Minding: feeling, form, and meaning in the creation of poetic iconicity. In Geert Broˆne and Jeroen Vandaele (eds.), 2009. Cognitive poetics: goals, gains, and gaps, 169–96. New York: de Gruyter. Freudenthal, Daniel, Julian M. Pine, Javier Aguado-Orea, and Fernand Gobet. 2007. Modelling the developmental pattern of finiteness marking in English, Dutch, German and Spanish using MOSAIC. Cognitive Science 31 (2), 311–41. Fried, Mirjam. 2008. Constructions and constructs: mapping a shift between predication and attribution. In Alexander Bergs and Gabriele Diewald (eds.), Constructions and language change, 47–79. Berlin: De Gruyter Mouton. 2013. Principles of constructional change. In Thomas Hoffmann and Graeme Trousdale (eds.), The Oxford handbook of construction grammar, 419–37. Oxford University Press. ¨ stman. 2005. Construction grammar and Fried, Mirjam, and Jan-Ola O spoken language: the case of pragmatic particles. Journal of Pragmatics 37(11), 1752–78. Friedrich, Paul. 1969. On the meaning of the Tarascan suffixes of space. IJAL Memoir 23, 5–48. 1971. The Tarascan suffixes of locative space: meaning and morphotactics. Indiana University Language Science Monographs, no. 9. Bloomington: Indiana University. Friedrich, Paul, and Dale Pesmen. 2014. A conversation with Paul Friedrich. Annual Review of Anthropology 43, 15–26. Frishberg, Nancy. 1975. Arbitrariness and iconicity: historical change in American Sign Language. Language 51, 676–710. Frisson, Steven, and Martin Pickering. 1999. The processing of metonymy: evidence from eye movements. Journal of Experimental Psychology: Learning, Memory, and Cognition 25, 1366–83. Frith, Uta. 1998. Editorial: literally changing the brain. Brain 121, 1051–52. Frost, Douglas O., and Christine Metin. 1985. Induction of function retinal projections to the somatosensory system. Nature 317, 162–64. Fuchs, Thomas. 2012. The phenomenology of body memory. Body Memory, Metaphor and Movement 84, 9–22. Fuhrman, Orly, and Lera Boroditsky. 2010. Cross-cultural differences in mental representations of time: evidence from an implicit nonlinguistic task. Cognitive Science 34(8), 1430–51. Fujii, Akiko. 2008. Meaning construction in humorous discourse: context and incongruities in conceptual blending. In Andrea E. Tyler, Mari Takada, Yiyoung Kim, and Diana Marinova (eds.), Language in the context of use: discourse and cognitive approaches to language, 183–98. Berlin: De Gruyter Mouton. Fujii, Seiko. 2004. Lexically (un)filled constructional schemes and construction types: the case of Japanese modal conditional constructions. In Mirjam Fried and Jan-Ola O¨stman (eds.), Construction grammar in a cross-language perspective, 121–56. Amsterdam: John Benjamins.
731
732
References
Fusaroli, Riccardo, and Kristian Tyle´n. 2015. Investigating conversational dynamics: interactive alignment, interpersonal synergy, and collective task performance. Cognitive Science 40(1), 1–27. Fusaroli, Riccardo, Ivana Konvalinka, and Sebastian Wallot. 2014. Analyzing social interactions: the promises and challenges of using cross recurrence quantification analysis. Springer Proceedings in Mathematics and Statistics 103, 137–55. Fusaroli, Riccardo, Bahador Bahrami, Karsten Olsen, Andreas Roepstorff, Geraint Rees, Chris Frith, and Kristian Tyle´n. 2012. Coming to terms: quantifying the benefits of linguistic coordination. Psychological Science 23, 931–39. Gaby, Alice. 2012. The Thaayorre think of time like they talk of space. Frontiers in Psychology 3, 300. In press. A Grammar of Kuuk Thaayorre. Berlin: De Gruyter Mouton. Galbraith, Mary. 1995. Deictic shift theory and the poetics of involvement in narrative. In Judith F. Duchan, Gail A. Bruder, and Lynne E. Hewitt (eds.), Deixis in narrative: a cognitive science perspective, 19–59. Hillsdale: Lawrence Erlbaum. Gallagher, Shaun. 2005. How the body shapes the mind. Oxford University Press. Gallagher, Shaun, and David Zahavi. 2008. The phenomenological mind: an introduction to philosophy of mind and cognitive science. London: Routledge. Gallese, Vittorio. 2003. The manifold nature of interpersonal relations: The quest for a common mechanism. Philosophical Transactions of the Royal Society of London, B 358, 517–28. Gallese, Vittorio, and George Lakoff. 2005. The brain’s concepts: the role of sensory-motor system in conceptual knowledge. Cognitive Neuropsychology 22, 455–79. Gallese, Vittorio, and Corrado Sinigaglia. 2011. What is so special about embodied simulation? Trends in Cognitive Sciences 15, 512–19. Galton, Francis. 1879. Psychometric experiments. Brain 2: 149–62. ¨ Boya. 2009. The effect of Garber, Lawrence L. Jr., Eva M. Hyatt, and U¨nal O package shape on apparent volume: an exploratory study with implications for package design. Journal of Marketing Theory and Practice 17(3), 215–34. Garber, Philip, and Susan Goldin-Meadow. 2002. Gesture offers insight into problem-solving in adults and children. Cognitive Science, 26, 817–831. Ga¨rdenfors, Peter. 2000. Conceptual spaces: the geometry of thought. Cambridge, MA; London: MIT Press. Gardner, Rod. 2001. When listeners talk. Amsterdam: John Benjamins. Garrod, Simon, and Anthony Anderson. 1987. Saying what you mean in dialogue: a study in conceptual and semantic co-ordination. Cognition 27, 181–218. Gavins, Joanna, and Gerard Steen (eds.). 2003. Cognitive poetics in practice. London: Routledge.
References
Geeraerts, Dirk. 2003. ‘Usage-based’ implies ‘variational’: on the inevitability of cognitive sociolinguistics. Plenary lecture presented at the 8th International Cognitive Linguistics Conference. Logron˜o. Geeraerts, Dirk, and Hubert Cuyckens (eds.). 2010. The Oxford handbook of cognitive linguistics. Oxford University Press. Geeraerts, Dirk, Stefan Grondelaers, and Peter Bakema. 1994. The structure of lexical variation. Berlin: De Gruyter Mouton. Geertz, Clifford. 1973. The interpretation of cultures: selected essays. New York: Basic Books. Gell, Alfred. 1992. The anthropology of time: cultural constructions of temporal maps and images. Oxford: Berg Publishers. Gemmell, Maggie. 2015. Semantic role alignment in metaphor: A frame semantic approach to metaphoric meaning. PhD dissertation. The University of Texas at Austin. Genette, Ge´rard. 1980. Narrative discourse: an essay in method. Oxford: Blackwell. 1988. Narrative discourse revisited. Ithaca: Cornell University Press. Gentner, Dedre. 1983. Structure-mapping: a theoretical framework for analogy. Cognitive Science 7(2), 155–70. Gentner, Dedre, and Brian Bowdle. 2001. Convention, form, and figurative language processing. Metaphor and Symbol 16(3/4), 223–47. 2008. Metaphor as structure-mapping. In R. W. Gibbs (ed.), The Cambridge handbook of metaphor and thought, 109–28. Cambridge University Press. Gentner, Dedre, Brian Bowdle, Philip Wolff, and Consuelo Boronat. 2001. Metaphor is like analogy. In Dedre Gentner, Keith J. Holyoak, and Boicho N. Kokinov (eds.), The analogical mind: perspectives from cognitive science 199–253. Cambridge, MA: MIT Press. Gertner, Yael, Cynthia Fisher and Julie Eisengart. 2006. Learning words and rules; abstract knowledge of word order in early sentence comprehension. Psychological Science 17(8), 684–91. Gevers, Wim, Bert Reynvoet, and Wim Fias. 2003. The mental representation of ordinal sequences is spatially organized. Cognition 87(3), 87–95. Gibbs, Raymond W. Jr. 1990. Psycholinguistic studies on the conceptual basis of idiomaticity. Cognitive Linguistics 1, 417–52. 1994. The poetics of mind: figurative thought, language and understanding. Cambridge University Press. 1996. What’s cognitive about cognitive linguistics? In Eugene H. Casad (ed.), Cognitive linguistics in the redwoods: the expansion of a new paradigm in linguistics, 27–53. Berlin: De Gruyter Mouton. 2005a. Embodiment and cognitive science. Cambridge University Press. 2005b. The psychological status of image schemas. In Beate Hampe (ed.), From perception to meaning: image schemas in cognitive linguistics, 113–36. Berlin: Mouton.
733
734
References
2006. Introspection and cognitive linguistics: should we trust our own intuitions? Annual Review of Cognitive Linguistics 4, 135–51. 2007. Why cognitive linguists should care more about empirical methods. In Monica Gonzalez-Marquez, Irene Mittelberg, Seana Coulson, and Michael J. Spivey (eds.), Methods in cognitive linguistics, 2–18. Amsterdam: John Benjamins. 2013. Walking the walk while thinking about the talk: embodied interpretation of metaphorical narratives. Journal of Psycholinguistic Research 42, 363–78. 2014. Embodied metaphor. In Jeanette Littlemore and John R. Taylor (eds.), The Bloomsbury companion to cognitive linguistics, 167–84. London: Bloomsbury. 2016. Metaphor wars: conceptual metaphor in human life. New York: Cambridge University Press. 2017. Metaphor, language and dynamical systems. In Elena Semino and Zso´fia Demje´n (eds.), The Routledge handbook of metaphor and language, 56–70. London: Routledge. Gibbs, Raymond W. Jr., and N. Blackwell. 2012. Climbing the stairs to literary heaven: a case study of allegorical interpretation of fiction. Scientific Study of Literature 2, 197–217. Gibbs, Raymond W. Jr., and Herbert Colston. 1995. The cognitive psychological reality of image schemas and their transformations. Cognitive Linguistics 6, 347–78. 2012. Interpreting figurative meaning. New York: Cambridge University Press. Gibbs, Raymond W. Jr., and Teenie Matlock. 2001. Psycholinguistic perspectives on polysemy. In Hubert Cuyckens and Britta Zawada (eds.), Polysemy in cognitive linguistics, 213–39. Amsterdam and Philadelphia: John Benjamins. 2008. Metaphor, imagination, and simulation: psycholinguistic evidence. In Raymond. W. Gibbs (ed.), Cambridge handbook of metaphor and thought, 161–76. New York: Cambridge University Press. Gibbs, Raymond. W. Jr., and Solange Nascimento. 1996. How we talk about love: metaphorical concepts and understanding love poetry. In Roger Kreuz and Mary Sue MacNealy (eds.), Empirical approaches to literature and aesthetics, 291–308. Norwood, NJ: Ablex. Gibson, James J. 1979. The ecological approach to visual perception. Boston: Houghton Mifflin. Gigerenzer, Gerd, Ralph Hertwig, and Thorsten Pachur, eds. 2011. Heuristics: the foundations of adaptive behavior. New York: Oxford University Press. Gijssels, Tom, Roberto Bottini, Shirley-Ann Rueschemeyer, and Daniel Casasanto. 2013. Space and time in the parietal cortex: fMRI evidence for a neural asymmetry. In Markus Knauff, Michael Pauen, Natalie Sebanz, and Ipke Wachsmuth (eds.), Proceedings of the 35th Annual
References
Conference of the Cognitive Science Society, 495–500. Austin: Cognitive Science Society. Giles, Howard, Nikolas Coupland, and Justine Coupland. 1991. Accommodation theory: communication, context, and consequence. In Howard Giles, Justine Coupland, and Nikolas Coupland (eds.), Contexts of accommodation, 1–68. Cambridge University Press. Gilquin, Gae¨tanelle. 2006. The verb slot in causative constructions: finding the best fit. Constructions SV 1–3/2006. Ginzburg, Jonathan, and Ivan A. Sag. 2000. Interrogative investigations: the form, meaning and use of English interrogatives. Stanford: CSLI Publications. Gipper, Sonja. 2011. Evidentiality and intersubjectivity in Yurakare´: an interactional account. PhD dissertation. Radboud University and Max Planck Institute Nijmegen. Givo´n, Talmy. 1979. From discourse to syntax: grammar as a processing strategy. In Talmy Givo´n (ed.), Discourse and syntax, 81–112. New York: Academic Press. 1985. Iconicity, isomorphism and non-arbitrary coding in syntax. In John Haiman (ed.), Iconicity in syntax, 187–220. Amsterdam: John Benjamins Publishing Company. 2005. Context as other minds: the pragmatics of sociality, cognition and communication. Amsterdam: John Benjamins. Glenberg, Arthur M. 1997. What memory is for: Creating meaning in the service of action. Behavioral and Brain Sciences 20(01), 41–50. 2015. Few believe the world is flat: how embodiment is changing the scientific understanding of cognition. Canadian Journal of Experimental Psychology/Revue canadienne de psychologie expe´rimentale 69(2), 165. Glenberg, Arthur M., and Vittorio Gallese. 2012. Action-based language: a theory of language acquisition, comprehension and production. Cortex 48(7), 905–22. Glenberg, Arthur M., and Michael P. Kaschak. 2002. Grounding language in action. Psychonomic Bulletin and Review 9, 558–65. Glynn, Dylan. 2010a. Corpus-driven cognitive semantics: introduction to the field. In Dylan Glynn and Kerstin Fischer (eds.), Quantitative methods in cognitive semantics: corpus-driven approaches, 1–42. Berlin: De Gruyter Mouton. 2010b. Testing the hypothesis: objectivity and verification in usagebased Cognitive Semantics. In Dylan Glynn and Kerstin Fischer (eds.), Quantitative methods in cognitive semantics: corpus-driven approaches, 239–629. Berlin: De Gruyter Mouton. 2014a. Polysemy and synonymy: cognitive theory and corpus method. In Dylan Glynn and Justyna A. Robinson (eds.), Corpus methods for semantics: quantitative studies in polysemy and synonymy, 7–38. Amsterdam: John Benjamins.
735
736
References
2014b. The many uses of run: corpus methods and socio-cognitive semantics. In Dylan Glynn and Justyna Robinson (eds.), Corpus methods for semantics: quantitative studies in polysemy and synonymy, 117–44. Amsterdam: John Benjamins. 2015. Semasiology and onomasiology. In Jocelyne Daems, Eline Zenner, Kris Heylen, Dirk Speelman, and Hubert Cuyckens (eds), Change of paradigms – new paradoxes, 47–79. Berlin: De Gruyter Mouton. Glynn, Dylan, and Kerstin Fischer (eds.). 2010. Quantitative methods in cognitive semantics: corpus-driven approaches. Berlin: De Gruyter Mouton. Glynn, Dylan and Justyna Robinson (eds.). 2014. Corpus methods for semantics: quantitative studies in polysemy and synonymy. Amsterdam & Philadelphia: John Benjamins. Goatly, Andrew. 1997. The language of metaphors: literal metaphorical. London: Routledge. Goddard, Cliff. 1998. Semantic analysis: a practical introduction. Oxford University Press. Goddard, Cliff, and Anna Wierzbicka. 2014. Words and meanings. Oxford University Press. Goffman, Erving. 1974. Frame analysis: an essay on the organization of experience. New York: Harper and Row. Gogate, Lakshmi J., Laura H. Bolzani, and Eugene A. Betancourt. 2006. Attention to maternal multimodal naming by 6- to 8-month-old infants and learning of word-object relations. Infancy 9, 259–88. Goldberg, Adele E. 1995. Construction grammar: a construction grammar approach to argument structure. Chicago: The University Of Chicago Press. 2002. Surface generalizations: an alternative to alternations. Cognitive Linguistics 13, 327–56. 2003. Constructions: a new theoretical approach to language. Trends in cognitive sciences 7(5), 219–24. 2006. Constructions at work: the nature of generalization in language. Oxford University Press. 2011. Corpus evidence of the viability of statistical preemption. Cognitive Linguistics 22, 131–53. 2013. Constructionist approaches. In Thomas Hoffmann and Graeme Trousdale (eds.), The Oxford handbook of construction grammar, 15–31. Oxford University Press. 2016. Subtle implicit language facts emerge from the functions of constructions. Frontiers in Psychology, n. pag. Goldberg, Adele E., and Devin Casenhiser. 2006. English constructions. In Bas Aarts and April McMahon (eds.), The handbook of English linguistics, 343–55. Malden, MA: Wiley-Blackwell. Goldberg, Adele E., and Ray Jackendoff. 2004. The English resultative as a family of constructions. Language 80, 532–68.
References
2014. How gesture helps children learn language. In I. Arnon, M. Tice, C. Kurumada, and B. Estigarribia (eds.), Language in interaction: studies in honor of Eve V. Clark, 157–71. Amsterdam: John Benjamins. Goldin-Meadow, Susan. 2003. Hearing gesture: how our hands help us think. Cambridge, MA: Harvard University Press. Goldin-Meadow, Susan, and Martha W. Alibali. 2013. Gesture’s role in speaking, learning, and creating language. Annual Review of Psychology 64, 257–83. Goldin-Meadow, Susan, and Diane Brentari. In press. Gesture, sign and language: the coming of age of sign language and gesture studies. Behavioral and Brain Sciences. Goldin-Meadow, Susan, Susan Wagner Cook, and Zachary A. Mitchell. 2009. Gesturing gives children new ideas about math. Psychological Science 20(3), 267–72. Goldin-Meadow, Susan, Howard C. Nusbaum, Spencer D. Kelly, and Susan Wagner. 2001. Explaining math: Gesturing lightens the load. Psychological Science 12, 516–22. Goldin-Meadow, Susan, Aaron Shield, Daniel Lenzen, Melissa Herzig, and Carol Padden. 2012. The gestures ASL signers use tell us when they are ready to learn math. Cognition 123, 448–53. Goldsmith, John. 1976. Autosegmental Phonology. PhD dissertation. MIT. Bloomington: Indiana University Linguistics Club. Go´mez, Rebecca L. 2002. Variability and detection of invariant structure. Psychological Science 13, 431–36. Gonzalez, J., A. Barros-Loscertales, F. Pulvermu¨ller, V. Meseguer, A. Sanjuan, V. Belloch, and C. Avila. 2006. Reading cinnamon activates olfactory brain regions. NeuroImage 32, 906–12. Gonzalez-Marquez, Monica, Raymond B. Becker, and James E. Cutting. 2007. An introduction to experimental methods for language researchers. In Monica Gonzalez-Marquez, Irene Mittelberg, Seana Coulson and Michael J. Spivey (eds.), Methods in cognitive linguistics, 53–86. Amsterdam: John Benjamins. Gonza´lvez-Garcı´a, Francisco. 2010. Contrasting constructions in English and Spanish: the influence of semantic, pragmatic and discourse factors. In Hans C. Boas (ed.), Contrastive studies in construction grammar, 43–86. Amsterdam: John Benjamins. Goodglass, Harold. 1976. Agrammatism. Studies in Neurolinguistics 1, 237–60. Goodwin, Charles. 1995. The negotiation of coherence within conversation. In Morton Ann Gernsbacher and Talmy Givo´n (eds.), Coherence in spontaneous text, 117–37. Amsterdam: John Benjamins. 2007. Environmentally coupled gestures. In S. Duncan, J. Cassell, and E. Levy (eds.), Gesture and the dynamic dimensions of language, 195–212. Amsterdam: John Benjamins.
737
738
References
Goossens, Louis. 1990. Metaphtonymy: the interaction of metaphor and metonymy in expressions for linguistic action. Cognitive Linguistics 1, 323–42. 2003. Metaphtonymy: the interaction of metaphor and metonymy in expressions for linguistic action. In Rene´ Dirven and Ralf Po¨rings (eds.), Metaphor and metonymy in comparison and contrast, 349–78. Berlin: De Gruyter Mouton. Gossen, Gary H. 1974. Chamulas in the world of the sun: time and space in a Maya oral tradition. Cambridge, MA: Harvard University Press. Goswami, Usha. 2012. Entraining the brain: applications to language research and links to musical entrainment. Empirical Musicology Review 7(1–2), 57–63. Goswami, Usha, and Victoria Leong. 2013. Speech rhythm and temporal structure: converging perspectives? Laboratory Phonology 4(1), 67–92. Gradecˇak-Erdeljic´, Tania. 2004. Euphemisms in the language of politics or how metonymy opens one door but closes the other. In P. Cap (ed.), Pragmatics today, 287–99. Frankfurt: Peter Lang. Grady, Joseph E. 1997a. Foundations of meaning: primary metaphors and primary scenes. PhD dissertation. University of California, Berkeley. 1997b. THEORIES ARE BUILDINGS revisited. Cognitive Linguistics 8, 267–90. 2005. Primary metaphors as inputs to conceptual integration. Journal of Pragmatics 37(10), 1595–14. Grady, Joseph E., and Christopher Johnson. 2002. Converging evidence for the notions of subscene and primary scene. In Rene´ Dirven and Ralf Po¨rings (eds.), Metaphor and metonymy in comparison and contrast, 533–54. Berlin: De Gruyter Mouton. Grady, Joseph E., Todd Oakley, and Seana Coulson. 1999. Conceptual blending and metaphor. In Raymond Gibbs (ed.), Metaphor in cognitive linguistics, 101–24. Amsterdam: John Benjamins. Graff, David, and Christopher Cieri. 2003. English Gigaword LDC2003T05. Philadelphia: Linguistic Data Consortium. Gralla, Linn. 2013. Linguistic representation of problem solving processes in unaided object assembly. PhD dissertation. University of Bremen, Germany. Green, Jennifer. 2014. Drawn from the ground: sound, sign and inscription in Central Australian sand stories. Cambridge University Press. Greenberg, Joseph H. 1966. Language universals: with special reference to feature hierarchies. The Hague: Mouton. Gregory, Stanford W., and Brian R. Hoyt. 1982. Conversation partner mutual adaptation as demonstrated by Fourier series analysis. Journal of Psychological Research 11, 35–46. Gregory, Stanford W., and Stephen Webster. 1996. A nonverbal signal in voices of interview partners effectively predicts communication
References
accommodation and social status perceptions. Journal of Personality and Social Psychology 70, 1231–40. Gries, Stefan Th. 1999. Particle movement: a cognitive and functional approach. Cognitive Linguistics 10, 105–45. 2000. Towards multifactorial analyses of syntactic variation: The case of Particle Placement. PhD dissertation, University of Hamburg. 2003a. Multifactorial analysis in corpus linguistics: a study of Particle Placement. London: Continuum. 2003b. Towards a corpus-based identification of prototypical instances of constructions. Annual Review of Cognitive Linguistics 1, 1–27. 2005. Syntactic priming: a corpus-based approach. Journal of Psycholinguistic Research 34, 365–99. 2006. Corpus-based methods and cognitive semantics: the many senses of to run. In Stefan Th. Gries and Anatol Stefanowitsch (eds.), Corpora in cognitive linguistics: corpus-based approaches to syntax and lexis, 57–99. Berlin: De Gruyter Mouton. 2008. Dispersions and adjusted frequencies in corpora. International Journal of Corpus Linguistics 13(4), 403–37. 2009a. Statistics for linguistics with R: a practical introduction, 1st edn. Berlin: De Gruyter Mouton. 2009b. What is corpus linguistics? Language and Linguistics Compass 3, 1–17. 2010a. Dispersions and adjusted frequencies in corpora: further explorations. In Stefan Th. Gries, Stefanie Wulff, and Mark Davies (eds.), Corpus linguistic applications: current studies, new directions, 197–212. Amsterdam: Rodopi. 2010b. Behavioral profiles: a fine-grained and quantitative approach in corpus-based lexical semantics. The Mental Lexicon 5(3), 323–46. 2011. Phonological similarity in multi-word units. Cognitive Linguistics 22, 491–510. 2012. Frequencies, probabilities, association measures in usage-/exemplar-based linguistics: some necessary clarifications. Studies in Language 36(3), 477–510. 2013a. 50-something years of work on collocations: what is or should be next . . . International Journal of Corpus Linguistics 18(1), 137–65. 2013b. Sources of variability relevant to the cognitive sociolinguist, and corpus – as well as psycholinguistic methods and notions to handle them. Journal of Pragmatics 52, 5–16. 2013c. Statistics for linguistics with R: a practical introduction, 2nd edn. Berlin: De Gruyter Mouton. 2014a. Corpus and quantitative methods. In Jeanette Littlemore and John R. Taylor (eds.), The Bloomsbury companion to cognitive linguistics, 279–300. London: Bloomsbury. 2014b. Frequencies, probabilities, and association measures in usage-/ exemplar–based linguistics: some necessary clarifications. In Nikolas
739
740
References
Gisborne and Willem B. Hollmann (eds.), Theory and data in cognitive linguistics, 15–49. Amsterdam: John Benjamins. 2015a. More (old and new) misunderstandings of collostructional analysis: on Schmid and Ku¨chenhoff 2013. Cognitive Linguistics 26(3), 505–36. 2015b. The role of quantitative methods in Cognitive Linguistics: corpus and experimental data on (relative) frequency and contingency of words and constructions. In Jocelyne Daems, Eline Zenner, Kris Heylen, Dirk Speelman, and Hubert Cuyckens (eds.), Change of paradigms – new paradoxes: recontextualizing language and linguistics, 311–25. Berlin: De Gruyter Mouton. Gries, Stefan Th. and Sandra C. Deshors. 2014. Using regressions to explore deviations between corpus data and a standard/target: two suggestions. Corpora 9(1), 109–36. Gries, Stefan Th., and Nick C. Ellis. 2015. Statistical measures for usage-based linguistics. Language Learning 65 (Supplement 1), 1–28. Gries, Stefan Th., and Martin Hilpert. 2008. The identification of stages in diachronic data: variability-based Neighbour Clustering. Corpora 3(1), 59–81. 2010. Modeling diachronic change in the third person singular: a multifactorial, verb- and author-specific exploratory approach. English Language and Linguistics 14(3), 293–320. Gries, Stefan Th., and Anatol Stefanowitsch. 2004a. Extending collostructional analysis: a corpus-based perspectives on ‘alternations.’ International Journal of Corpus Linguistics 9(1), 97–129. 2004b. Co-varying collexemes in the into-causative. In Michel Achard and Suzanne Kemmer (eds.), Language, culture and mind, 225–36. Stanford: CSLI. 2010. Cluster analysis and the identification of collexeme classes. In Sally Rice and John Newman (eds.), Empirical and experimental methods in cognitive/functional research, 73–90. Stanford: CSLI. Gries, Stefan Th., and Stefanie Wulff. 2005. Do foreign language learners also have constructions? Evidence from priming, sorting and corpora. Annual Review of Cognitive Linguistics 3, 182–200. 2009. Psycholinguistic and corpus-linguistic evidence for L2 constructions. Annual Review of Cognitive Linguistics 7(1), 163–86. Gries, Stefan Th., and Dagmar S. Divjak (eds.). 2012. Frequency effects in language learning and processing. Berlin and New York: De Gruyter Mouton. Gries, Stefan Th., and Anatol Stefanowitsch (eds.). 2006. Corpora in cognitive linguistics: corpus-based approaches to syntax and lexis. Berlin and New York: De Gruyter Mouton. Gries, Stefan Th., Beate Hampe, and Doris Scho¨nefeld. 2005. Converging evidence: bringing together experimental and corpus data on the association of verbs and constructions. Cognitive Linguistics 16(4), 635–76.
References
2010. Converging evidence II: more on the association of verbs and constructions. In Sally Rice and John Newman (eds.), Empirical and experimental methods in cognitive/functional research, 59–72. Stanford: CSLI. Grodzinsky, Y. 2000. The neurology of syntax: language use without Broca’s area. Behavioral and Brain Sciences 23(01), 1–21. Grondelaers, Stefan. 2000. De distributie van niet anaforisch er buiten de eerste zinplaats: sociolexicologische, functionele en psycholinguı¨stische aspected van er’s status als presentatief signaal. PhD dissertation. Katholieke Universiteit Leuven. Grondelaers, Stefan, Dirk Speelman, and Dirk Geeraerts. 2007. A case for a cognitive corpus linguistics. In Monica Gonzalez-Marquez, Irene Mittelberg, Seana Coulson, and Michael J. Spivey (eds.), Methods in cognitive linguistics, 149–69. Amsterdam: John Benjamins. 2008. National variation in the use of er ‘there’: regional and diachronic constraints on cognitive explanations. In Gitte Kristiansen and Rene´ Dirven (eds.), Cognitive sociolinguistics: language variation, cultural models, social systems, 153–203. Berlin: De Gruyter Mouton. Gu¨hlich, Elisabeth. 1970. Makrosyntax der Gliederungssignale im gesprochenen Franzo¨sisch. Mu¨nchen: Fink. Guilbeault, Douglas. 2017. How politicians express different viewpoints in gesture and speech simultaneously. Special Issue on Viewpoint phenomena in multimodal communication. Cognitive Linguistics 28(3). Gu¨ldeman, Tom, and Manfred von Roncador (eds.). 2002. Reported discourse: a meeting ground for different linguistic domains. Amsterdam: John Benjamins. Gullberg, Marianne, and Bhuvana Narasimhan. 2010. What gestures reveal about how semantic distinctions develop in Dutch children’s placement verbs. Cognitive Linguistics 21(2), 239–62. Gumperz, John. 1962. Types of linguistic communities. Anthropological Linguistics 4, 28–40. 1964. Linguistic and social interaction in two communities. American Anthropologist 66, 137–53. Gumperz, John, and Dell Hymes (eds.). 1972. Directions in sociolinguistics: the ethnography of communication. New York: Holt, Rinehart and Winston. Gumperz, John J., and Stephen C. Levinson (eds.). 1996. Rethinking linguistic relativity. Cambridge University Press. 1991. Rethinking linguistic relativity. Current Anthropology 32, 613–23. Gu¨nter, Franziska. 2014. Form, meaning and cognition: language- and speaker-specific variation in linguistic and non-linguistic forms of interaction with spatial scenes. PhD dissertation. LudwigMaximilians University. Gu¨nthner, Susanne, and Jo¨rg Bu¨cker. 2009. Grammatik im Gespra¨ch: Konstruktionen der Selbst- und Fremd-positionierung. Berlin: De Gruyter Mouton.
741
742
References
Gu¨nthner, Susanne, and Wolfgang Imo (eds.). 2006. Konstruktionen in der Interaktion. Berlin: De Gruyter Mouton. Gussenhoven, Carlos, and Haike Jacobs. 2005. Understanding phonology: understanding language series. London: Hoddor Arnold. Guy, Gregory R. 2013. The cognitive coherence of sociolects: how do speakers handle multiple sociolinguistic variables? Journal of Pragmatics 52, 63–71. Habel, Christopher, Michael Herweg, and Simone Pribbenow. 1993. Wissen u¨ber Raum und Zeit [Knowledge about space and time]. In Gu¨nther Go¨rz (ed.), Einfu¨hrung in die ku¨nstliche Intelligenz, 139–204. Bonn: Addison-Wesley. Haddington, Pentti. 2007. Positioning and alignment as activities of stancetaking in news interviews. In Robert Englebretson (ed). Stancetaking in discourse: subjectivity, evaluation, interaction, 283–317. Amsterdam: John Benjamins. Haigh, Matthew, Andrew J. Stewart, Jeffrey S. Wood, and Louise Connell 2011. Conditional advice and inducements: are readers sensitive to implicit speech acts during comprehension? Acta Psychologica 136(3), 419–24. Haiman, John. 1978. Conditionals are topics. Language 54(3), 564–89. 1980. Dictionaries and encyclopedias. Lingua 50, 329–57. 1983. Iconic and economic motivation. Language 59(4), 781–819. 1985a. Iconicity in syntax. Amsterdam: John Benjamins. 1985b. Natural syntax. Cambridge University Press. 1998. Talk is cheap: Sarcasm, alienation, and the evolution of language. Oxford University Press. Hakulinen, Auli, and Margret Selting (eds.). 2005. Syntax and lexis in conversation. Amsterdam: John Benjamins. Halle, Morris. 1962. Phonology in generative grammar. Word 18, 54–72. 1964. On the bases of phonology. In J. A. Fodor and J. Katz (eds.), The structure of language, 324–33. New York: Prentice Hall. Halliday, Michael A. K., and Christian M. I. M. Matthiessen. 2014. Halliday’s Introduction to Functional Grammar, 4th edn. London: Routledge. 2004. An introduction to functional grammar, 3rd ed. London: Hodder Arnold. Hampe, Beate (ed.). 2005. From perception to meaning: image schemas in cognitive linguistics. Berlin: Mouton. Handl, Sandra. 2011. The conventionality of figurative language: a usage-based study. Tubingen: Narr Verlag. Handl, Sandra, and Hans-Jo¨rg Schmid (eds.) 2011. Windows to the mind: metaphor, metonymy and conceptual blending. Berlin: De Gruyter Mouton. Hanks, William F. 1992. The indexical ground of deictic reference. In Alessandro Duranti and Charles Goodwin (eds.), Rethinking context: language as an interactive phenomenon, 46–76. Cambridge University Press.
References
Harder, Peter. 1996. Functional semantics: a theory of meaning, structure and tense in English. Berlin: De Gruyter Mouton. 2010. Meaning in mind and society: a functional contribution to the social turn in cognitive linguistics. Berlin: Walter de Gruyter. Harris, Zellig S. 1970. Papers in structural and transformational linguistics. Dordrecht: Reidel. Hartmann, Matthias, Corinna S. Martarelli, Fred W. Mast, and Kurt Stocker. 2014. Eye movements during mental time travel follow a diagonal line. Consciousness and Cognition 30, 201–09. Hasegawa, Yoko, Russell Lee-Goldman, Kyoko Hirose Ohara, Seiko Fuji, and Charles J. Fillmore. 2010. On expressing measurement and comparison in English and Japanese. In H. C. Boas (ed.), Contrastive studies in construction grammar, 169–200. Amsterdam: John Benjamins. Haspelmath, Martin, and Andrea D. Sims. 2010. Understanding morphology, 2nd edn. London: Routledge. Haspelmath, Martin. 1997. From space to time: temporal adverbials in the world’s languages. Mu¨nchen: Lincom. 2005. Iconicity versus frequency in explaining grammatical asymmetries. Handout for a presentation at Jena University 13 April. www .staff.eva.mpg.de/~haspelmt/Jena05.pdf. 2006. Explaining alienability contrasts in adnominal possession: economy vs. iconicity. Paper presented at the 2nd Conference on the Syntax of the World’s Languages. Lancaster University. 2008. Parametric versus functional explanation of syntactic universals. In Theresa Biberauer (ed.), The limits of syntactic variation, 75–107. Amsterdam: Benjamins. Haser, Verena. 2003. Metaphor in semantic change. In Antonio Barcelona (ed). Metaphor and metonymy at the crossroads, 171–94. Berlin: De Gruyter Mouton. Hasson, Uri, Janice Chen, and Christopher J. Honey. 2015. Hierarchical process memory: memory as an integral component of information processing.Trends in cognitive sciences 19(6), 304–13. Hauser, Marc D., Noam Chomsky, and W. Tecumseh Fitch. 2002. The faculty of language: what is it, who has it, and how did it evolve. Science 298, 1569–79. Haviland, John B. 1993. Anchoring, iconicity, and orientation in Guugu Yimithirr pointing gestures. Journal of Linguistic Anthropology 3(1), 3–45. Hay, Jennifer B., and R. Harald Baayen. 2003. Phonotactics, parsing and productivity. Italian Journal of Linguistics 1, 99–130. 2005. Shifting paradigms: gradient structure in morphology. Trends in Cognitive Science 9(7), 342–48. Hay, Jennifer, Paul Warren, and Katie Drager. 2006. Factors influencing speech perception in the context of a merger-in-progress. Journal of Phonetics 34, 458–84.
743
744
References
Hay, Jessica. S., and R. L. Diehl. 2007. Perception of rhythmic grouping: testing the iambic/trochaic law. Perception and Psychophysics 69(1), 113–22. Hay, Jessica F., and Jenny R. Saffran. 2012. Rhythmic grouping biases constrain infant statistical learning. Infancy 17(6), 610–41. Hayashi, Makoto, Geoffrey Raymond, and Jack Sidnell (eds.). 2013. Conversational repair and human understanding. Cambridge University Press. Hayes, Andrew F., and Klaus Krippendorff. 2007. Answering the call for a standard reliability measure for coding data. Communication Methods and Measures 1, 77–89. Hayes, Bruce. 1995. Metrical stress theory. University of Chicago Press. 2009. Introductory phonology. Blackwell textbooks in linguistics. Oxford: Wiley-Blackwell. Hearst, Marti. 1992. Automatic acquisition of hyponyms from large text corpora. Proceedings of the Fourteenth International Conference on Computational Linguistics, 539–45. Nantes. Heid, Ulrich. 1996. Creating a multilingual data collection for bilingual lexicography from parallel monolingual lexicons. In Proceedings of the VIIth EURALEX International Congress, Gothenburg, 559–73. Heim, Irene, and Aangelika Kratzer. 1998. Semantics in generative grammar. Oxford: Blackwell. Hein, Grit, and Robert T. Knight. 2008. Superior temporal sulcus – it’s my area: or is it? Journal of Cognitive Neuroscience 20, 2125–36. Heina¨ma¨ki, Orvokki. 1974. Semantics of English temporal connectives. PhD dissertation. University of Texas, Austin. Heine, Bernd. 1995. Conceptual grammaticalization and prediction. In John Taylor and Robert MacLaury (eds.) Language and the cognitive construal of the world, 119–35. Berlin: De Gruyter Mouton. 1997. Cognitive foundations of grammar. Oxford University Press. Heine, Bernd, and Tania Kuteva. 2002. World lexicon of grammaticalization. Cambridge University Press. Heine, Bernd, Ulrike Claudi, and Friederike Hu¨nnemeyer. 1991. Grammaticalization: a conceptual framework. University of Chicago Press. Herbst, Thomas. 2011. The status of generalizations: valency and argument structure constructions. In Thomas Herbst and Anatol Stefanowitsch (eds.), Argument structure: valency and/or construction? Special issue of Zeitschrift fu¨r Anglistik und Amerikanistik 59(4), 347–67. 2014. The valency approach to argument structure constructions. In Thomas Herbst, Hans-Jo¨rg Schmid, and Susen Faulhaber (eds.), Constructions – collocations – patterns. Berlin: de Gruyter Mouton. Heritage, John. 1984. A change-of-state token and aspects of its sequential placement. In J. Maxwell Atkinson and John Heritage (eds.), Structures of social action: studies in conversation analysis. Cambridge University Press, 299–345.
References
1998. Oh-prefaced responses to inquiry. Language in Society 29, 291–334. 2002. Oh-prefaced responses to assessments: a method of modifying agreement/disagreement. In Cecila Ford, Barbara Fox., and Sandra Thompson (eds.), The language of turn and sequence, 196–224. New York: Oxford University Press. Herman, David (ed.). 2003. Narrative theory and the cognitive sciences. Stanford: CSLI. 2009. Beyond voice and vision: Cognitive Grammar and focalization theory. In Peter Hu¨hn, John Pier, Wolf Schmid, and Jo¨rg Scho¨nert (eds.), Handbook of narratology, 119–42. Berlin: Walter de Gruyter. 2013. Storytelling and the sciences of the mind. Cambridge, MA: Massachusetts Institute of Technology. Herrmann, Esther, Josep Call, Marı´a Victoria Herna´ndez-Lloreda, Brian Hare, and Michael Tomasello. 2007. Humans have evolved specialized skills of social cognition: the cultural intelligence hypothesis. Science 317, 1360–66. Herweg, Michael. 1991. Temporale Konjunktionen und Aspekt. Der sprachliche Ausdruck von Zeitrelationen zwischen Situationen. Kognitionswissenschaft 2, 51–90. Hickok, G. 2009. Eight problems for the mirror neuron theory of action understanding in monkeys and humans. Journal of Cognitive Neuroscience 7, 1229–43. Hickok, Gregory, and David Poeppel. 2004. Dorsal and ventral streams: a framework for understanding aspects of the functional anatomy of language. Cognition 92(1), 67–99. 2007. The cortical organization of speech processing. Nature Reviews Neuroscience 8, 393–402. Hickok, Gregory, Ursula Bellugi, and Edward S. Klima. 1998. The neural organization of language: evidence from sign language aphasia. Trends in Cognitive Sciences 2(4), 129–36. Hilpert, Martin. 2006a. Keeping an eye on the data: metonymies and their patterns. In Anatol Stefanowitsch and Stefan Th. Gries (eds.), Corpus-based approaches to metaphor and metonymy, 123–51. Berlin: De Gruyter Mouton. 2006b. Distinctive collexeme analysis and diachrony. Corpus Linguistics and Linguistic Theory 2(2), 243–57. 2007. Chained metonymies in lexicon and grammar: a cross-linguistic perspective on body part terms. In Gu¨nther Radden, Klaus-Michael Ko¨pcke, Thomas Berg, and Peter Siemund (eds.), Aspects of meaning construction, 77–98. Amsterdam: John Benjamins. 2008. Germanic future constructions: a usage-based approach to language change. Amsterdam and Philadelphia: John Benjamins. 2009. The German mit-predicative construction. Constructions and Frames 1, 29–55. 2013a. Constructional change in English: developments in allomorphy, word formation, and syntax. Cambridge University Press.
745
746
References
2013b. Corpus-based approaches to constructional change. In Thomas Hoffmann and Graeme Trousdale (eds.), The Oxford handbook of construction grammar, 458–77. Oxford University Press. 2014. Construction grammar and its application to English. Edinburgh University Press. Himmelmann, Nikolaus P. 2004. Lexicalization or grammaticalization: opposite or orthogonal? In Walter Bisang, Nikolaus P. Himmelmann, and Bjo¨rn Wiemer (eds.), What makes grammaticalization? A look from its fringes and its components, 21–42. Berlin: De Gruyter Mouton. Hinzen, Wolfram, and Michiel van Lambalgen. 2008. Explaining intersubjectivity: a comment on Arie Verhagen, Constructions of Intersubjectivity. Cognitive Linguistics 19(1), 107–23. Hobson, Peter. 2004. The cradle of thought: exploring the origins of thinking. London: Pan Macmillan. Hock, Hans Heinrich. 1991. Principles of historical linguistics. Berlin: de Gruyter. Hockett, Charles. 1978. In search of Jove’s brow. American Speech 53, 243–313. 1982. The origin of speech. In William S.-Y. Wang (ed.), Human communication: language and its psychobiological bases, 5–12. San Francisco: W. H. Freeman. Hoey, Michael. 2005. Lexical priming: a new theory of words and language. London: Routledge. 2015. Words and their neighbours. In John R. Taylor (ed.), The Oxford handbook of the word, 141–53. Oxford University Press. Hoffmann, Thomas. 2011. Preposition placement in English: a usage-based approach. Cambridge University Press. 2013a. Abstract phrasal and clausal constructions. In Thomas Hoffmann and Graeme Trousdale (eds.), The Oxford handbook of construction grammar, 307–28. Oxford University Press. 2013b. Obtaining introspective acceptability judgements. In Manfred Krug and Julia Schlu¨ter (eds.), Research methods in language variation and change, 99–118. Cambridge University Press. 2014. The cognitive evolution of Englishes: the role of constructions in the dynamic model. In Sarah Buschfeld, Thomas Hoffmann, Magnus Huber, and Alexander Kautzsch (eds.), The evolution of Englishes: the dynamic model and beyond, varieties of English around the world G49, 160–80. Amsterdam: John Benjamins. Hoffmann, Thomas, and Graeme Trousdale (eds.). 2013. The Oxford handbook of construction grammar. Oxford University Press. 2011. Variation, change and constructions in English. Cognitive Linguistics 22(1), 1–24. Hoftstadter, Douglas, and Emmanuel Sander. 2013. Surfaces and essences: analogy as the fuel and fire of thinking. New York: Basic Books.
References
Holleman, Bregje C., and Henk L. W. Pander Maat. 2009. The pragmatics of profiling: framing effects in text interpretation and text production. Journal of Pragmatics 41(11), 2204–21. Hollmann, Willem. 2013. Constructions in cognitive sociolinguistics. In Thomas Hoffmann and Graeme Trousdale (eds.), The Oxford handbook of construction grammar, 491–509. Oxford University Press. Hollmann, Willem, and Anna Siewierska. 2006. Corpora and (the need for) other methods in a study of Lancashire dialect. Zeitschrift fu¨r Anglistik and Amerikanistik 54, 203–16. 2007. A construction grammar account of possessive constructions in Lancashire dialect: some advantages and challenges. English Language and Linguistics 11: 407–24. 2011. The status of frequency, schemas and identity in cognitive sociolinguistics: a case study on definite article reduction. Cognitive Linguistics 22, 25–54. Holme, Randal. 2009. Cognitive linguistics and language teaching. New York: Palgrave MacMillan. Ho¨lscher, Christoph, Thora Tenbrink, and Jan Wiener. 2011. Would you follow your own route description? Cognition 121, 228–47. Hong, Jisup. 2016. Automatic metaphor detection using constructions and frames. Constructions and Frames 8(2): 293–320. Hopper, Paul J. 1998. Emergent grammar. In Michael Tomasello (ed.), The new psychology of language: cognitive and functional approaches to language structure, 155–75. Mahwah: Lawrence Erlbaum. 2004. The openness of grammatical constructions. Papers from the 40th Regional Meeting of the Chicago Linguistic Society 40, 239–56. Hopper, Paul J., and Elizabeth Closs Traugott. 2003. Grammaticalization. Cambridge University Press. Horn, Laurence. R. 2010. The expression of negation. Berlin: Walter de Gruyter. Hostetter, Autumn B. 2011. When do gestures communicate? A meta-analysis. Psychological Bulletin 137(2), 297–315. Hougaard, Anders. 2005. Conceptual disintegration and blending in interactional sequences. Journal of Pragmatics 37(10), 1653–85. Hougaard, Anders, and Todd Oakley (eds.). 2008. Mental spaces in discourse and interaction. Amsterdam: John Benjamins. Howes, Christine, Patrick G. T. Healey, and Matthew Purver. 2010. Tracking lexical and syntactic alignment in conversation. Proceedings of the Twentyfifth Annual Conference of the Cognitive Science Society, 2004–09. Huang, Ping-Yu, David Wible, and Hwa-Wei Ko. 2012. Frequency effects and transitional probabilities in L1 and L2 speakers’ processing of multiword expressions. In Gries, Stefan Th. and Dagmar S. Divjak (eds.), Frequency effects in language learning and processing, 145–75. Berlin and New York: De Gruyter Mouton. Huddleston, Rodney and Geoffrey K. Pullum. 2002. The Cambridge grammar of the English language. Cambridge University Press.
747
748
References
Hudson, Richard A. 1991. English word grammar. Oxford: Blackwell. 1992. Review of Ronald W. Langacker, Concept, image, and symbol: the cognitive basis of grammar. Journal of Linguistics 28, 506–09. 1996. Sociolinguistics, 2nd edn. Cambridge University Press. 2007. Language networks: the new word grammar. Oxford University Press. 2010. An introduction to word grammar. Cambridge University Press. Hu¨hn, Peter, John Pier, Wolf Schmid, and Jo¨rg Scho¨nert (eds.). 2009a. Handbook of narratology. Berlin: Walter de Gruyter. Huffman, K. J., Z. Molna´r, A. Van Dellen, D. M. Kahn, C. Blakemore, and L. Krubitzer. 1999. Formation of cortical fields on a reduced cortical sheet. The Journal of Neuroscience 19(22), 9939–52. Hu¨hn, Peter, Wolf Schmid, and Jo¨rg Scho¨nert (eds.). 2009b. Point of view, perspective, and focalization: modeling mediation in narrative. Berlin: Walter de Gruyter. Hu¨ning, Matthias, and Geert Booij. 2014. From compounding to derivation: the emergence of derivational affixes through ‘constructionalization.’ Folia Linguistica 48, 579–604. Hunston, Susan, and Gill Francis. 2000. Pattern grammar: a corpus-driven approach to the lexical grammar of English. Amsterdam: John Benjamins. Hurford, James R. 2007. The origins of meaning: language in the light of evolution. Oxford University Press. 2011. The origins of grammar: language in the light of evolution. Oxford University Press. Husserl, Edmund. 1936 [1970]. The crisis of European sciences and transcendental phenomenology. Evanston, IL: Northwestern University Press. Hutchby, Ian, and Robin Wooffitt. 1998. Conversation analysis: principles, practices and applications. Cambridge: Polity Press. Hutchins, Edwin. 2003 [2005]. Material anchors for conceptual blends. Journal of Pragmatics 37(10), 1555–77. Hutto, Daniel. 2008. Folk psychological narratives: the socio-cultural basis of understanding reasons. Cambridge, MA: MIT Press. Huumo, Tuomas. 1999. Path settings, subjective motion, and the Finnish partitive subject. In Shin Ja J. Hwang and Arle L. Lommel (eds.), LACUS Forum XX, 363–74. Houston, Tx: The Linguistic Association of Canada and the United States. Ibbotson, Paul, Elena Lieven, and Michael Tomasello. 2013. The attention-grammar interface: eye-gaze cues structural choice in children and adults. Cognitive Linguistics 24(3), 457–81. Ibbotson, Paul, Anna Theakston, Elena Lieven, and Michael Tomasello. 2012. Prototypes in the transitive construction. Cognitive Science 36, 1268–88. Imo, Wolfgang. 2007. Construction Grammar und Gesprochene SpracheForschung. Tu¨bingen: Max Niemeyer. Inkelas, Sharon. 2014. The interplay of morphology and phonology. Oxford University Press.
References
Irvine, Judith. 2001. Style as distinctiveness: the culture and ideology of linguistic differentiation. In Penelope Eckert and John Rickford (eds.), Stylistic variation in language, 21–43. Cambridge University Press. Itkonen, Esa 2003. What is language? A study in the philosophy of linguistics. Turku University Press. Iverson, Jana M., and Susan Goldin-Meadow. 1998. Why people gesture when they speak. Nature 396(6708), 228. Iwata, Seizi. 2005. Locative alternation and two levels of verb meaning. Cognitive Linguistics 16, 355–407. 2008. Locative alternation: a lexical-constructional approach. Amsterdam: John Benjamins. Jackendoff, Ray. 2002. Foundations of language: brain, meaning, grammar, evolution. Oxford University Press. 2011. What is the human language faculty? Two views. Language 87, 586–624. 2013. Constructions in the parallel architecture. In Thomas Hoffmann and Graeme Trousdale (eds.), The Oxford handbook of construction grammar, 70–92. Oxford University Press. Jacobsen, Natalia D. 2012. Applying cognitive linguistics and tasksupported language teaching to instruction of English conditional phrases. PhD dissertation. Georgetown University. 2015. A cognitive linguistic (CL) analysis of English conditionals in English for Academic Purposes (EAP) instruction: implications from Sociocultural Theory (SCT). In Kyoko Masuda, Carlee Arnett, and Angela Labarca (eds.), Cognitive linguistics and sociocultural theory in second and foreign language teaching, 103–25. Berlin: De Gruyter Mouton. Jakobson, Roman. 1972. Motor signs for ‘yes’ and no.’ Language in Society 1(1), 91–96. Jakobson, Roman, and Morris Halle. 1956. Fundamentals of language. The Hague: Mouton. Janda, Laura A. (ed.). 2009. What is the role of semantic maps in cognitive linguistics? In Piotr Stalmaszczyk and Wiesław Oleksy (eds.), Cognitive approaches to language and linguistic data.: studies in honor of Barbara Lewandowska–Tomaszczyk, 105–24. Frankfurt am Main: Peter Lang. 2013a. Cognitive Linguistics: the quantitative turn. Berlin: De Gruyter Mouton. 2013b. Quantitative methods in Cognitive Linguistics. In Laura A. Janda (ed.), Cognitive linguistics: the quantitative turn – the essential reader, 1–32. Berlin: De Gruyter Mouton. Janda, Laura A., and Valery D. Solovyev. 2009. What constructional profiles reveal about synonymy: a case study of Russian words for S A D N E S S and H A P P I N E S S . Cognitive Linguistics 20, 367–93. Janda, Laura A., Tore Nesset, and R. Harald Baayen. 2010. Capturing correlational structure in Russian paradigms: a case study in
749
750
References
logistic mixed-effects modeling. Corpus Linguistics and Linguistic Theory 6, 29–48. Janssen, Theo A. J. M. 2007. A speaker/hearer-based grammar: the case of possessives and compounds. In Mike Hannay and Gerard J. Steen (eds.), Structural-functional studies in English grammar, 353–87. Amsterdam: Benjamins. January, David, and Edward Kako. 2007. Re-evaluating evidence for linguistic relativity: reply to Boroditsky 2001. Cognition 104, 417–26. Janzen, Terry. 1999. The grammaticization of topics in American Sign Language. Studies in Language 23(2), 271–306. 2004. Space rotation, perspective shift, and verb morphology in ASL. Cognitive Linguistics 15(2), 149–74. 2012a. Lexicalization and grammaticalization. In Martin Steinbach, Roland Pfau, and Bencie Woll (eds.), Handbook of sign languages, 816–40. Berlin: De Gruyter Mouton. 2012b. Two ways of conceptualizing space: motivating the use of static and rotated vantage point space in ASL discourse. In Barbara Dancygier and Eve Sweetser (eds.), Viewpoint in language: a multimodal perspective, 156–74. Cambridge University Press. Janzen, Terry, and Barbara Shaffer. 2002. Gesture as the substrate in the process of ASL grammaticization. In Richard Meier, David Quinto, and Kearsy Cormier (eds.), Modality and structure in signed and spoken languages, 199–223. Cambridge University Press. Jarque, Maria Josep. 2016. What about? Fictive question-answer pairs for non-information-seeking functions across signed languages. In Esther Pascual and Sergeiy Sandler (eds.), The conversation frame: forms and functions of fictive interaction, 171–92. Amsterdam: John Benjamins. Jarque, Maria Josep, and Esther Pascual. 2015. Direct discourse expressing evidential values in Catalan Sign Language. Approaches to Evidentiality in Romance, special issue of eHumanista/IVITRA 8, 421–45. 2016. Mixed viewpoints in factual and fictive discourse in Catalan Sign Language narratives. In Barbara Dancygier, Wei-lun Lu, and Arie Verhagen (eds.), Viewpoint and the fabric of meaning: form and use of viewpoint tools across languages and modalities, 259–80. Berlin: Walter de Gruyter. Jensen, Janni. 2016. ‘Crank your oven to 400 F’: a multimodal case study of recipes in cookbooks and food blogs. Master’s thesis, University of Southern Denmark. Jensen, Kim Ebensgaard. 2015. Review of Pu¨tz et al. (eds.), Cognitive sociolinguistics: social and cultural variation in cognition and language use 2014. LINGUIST List 26(1176), n. pag. Johansson Falck, Marlene, and Raymond W. Gibbs Jr. 2012. Embodied motivations for metaphorical meanings. Cognitive Linguistics 23(2), 251–72.
References
Johnson, Christopher. 1997. Metaphor vs. conflation in the acquisition of polysemy: the case of SEE. In Masako K. Hiraga, Chris Sinha, and Sherman Wilcox (eds.), Cultural, typological and psychological issues in cognitive linguistics, 155–69. Amsterdam: John Benjamins. Johnson, Edward D. 1991. The handbook of good English: revised and updated. New York: Facts on File. Johnson, Keith. 2008. Quantitative methods in linguistics. Malden, MA: Blackwell. Johnson, Mark. 1987. The body in mind: the bodily basis of meaning, imagination and reason. University of Chicago Press. 2010. Metaphor and cognition. In Shaun Gallagher and Daniel Schmicking (eds.), Handbook of phenomenology and cognitive science, 401–14. Berlin: Springer. Johnson, Mark, and George Lakoff. 2002. Why cognitive linguistics requires embodied realism. Cognitive Linguistics 13(3), 245–64. Johnson, Samuel. 1755. A dictionary of the English language. London: J. and P. Knapton. Johnstone, Barbara. 2007. Discourse analysis, 2nd edn. New York: WileyBlackwell. Johnstone, Barbara, and Scott E. Kiesling. 2008. Indexicality and experience: exploring the meanings of /aw/-monophthongization in Pittsburgh. Journal of Sociolinguistics 12, 5–33. Jones, Mark. 2002. The origin of definite article reduction in northern English dialects: evidence from dialect allomorphy. English Language and Linguistics 6, 325–45. Joseph, Brian D. 2004. On change in language and change in language. Language 80, 381–83. Joy, Annamma, John F. Sherry Jr., and Jonathan Deschenes. 2009. Conceptual blending in advertising. Journal of Business Research 62(1), 39–49. Jurafsky, Daniel. 1992. An on-line computational model of human sentence interpretation. In American Association for Artificial Intelligence (eds.), Proceedings of the National Conference on Artificial Intelligence (AAAI-92), 302–08. Cambridge, MA: MIT Press. Kable, Joseph W., Irene P. Kan, Ashely Wilson, Sharon L. Thompson-Schill, and Anjan Chatterjee. 2005. Conceptual representations of action in lateral temporal cortex. Journal of Cognitive Neuroscience 17, 855–70. Kapatsinski, Vsevolod. 2013. Conspiring to mean: experimental and computational evidence for a usage–based harmonic approach to morphophonology. Language 89, 110–48. Kappelhoff, Hermann, and Cornelia Mu¨ller. 2011. Embodied meaning construction: multimodal metaphor and expressive movement in speech, gesture, and feature film. Metaphor and the Social World 1, 121–53. Ka¨rkka¨inen, Elise. 2006. Stance-taking in conversation: from subjectivity to intersubjectivity. Text and Talk 26(6), 699–731.
751
752
References
Kay, Paul. 1973 On the form of dictionary entries: English kinship semantics. In Roger Shuy and Charles-James N. Bailey (eds.) Toward tomorrow‘s linguistics, 120–38. Georgetown University Press. 2003. Pragmatic aspects of grammatical constructions. In L. Horn and G. Ward (eds.), Handbook of Pragmatics, 675–700. Oxford: Blackwell. 2013. The limits of (construction) grammar. In Thomas Hoffmann and Graeme Trousdale (eds.), The Oxford handbook of construction grammar, 32–48. Oxford University Press. Kay, Paul, and Charles J. Fillmore. 1999. Grammatical constructions and linguistic generalizations: the ‘What’s X doing Y?’ construction. Language 75, 1–33. Kay, Paul and Ivan Sag. 2012. Cleaning up the big mess: discontinuous dependencies and complex determiners. In H. C. Boas and I. Sag (eds.), Sign-Based Construction Grammar, 229–56. Stanford: CSLI. Keating, Patricia. 1984. Phonetic and phonological representation of stop consonant voicing. Language 60(2), 286–319. Keating, Patricia, Wendy Linker, and Marie Huffman. 1983. Patterns of allophone distribution for voiced and voiceless stops. Journal of Phonetics 11, 277–90. Keller, Rudi. 1994. Sprachwandel: Von der unsichtbaren Hand in der Sprache. Tu¨bingen: Francke. Kelly, Kevin. 2010. What technology wants. New York: Viking Press. ¨ zyu¨rek, and Eric Maris. 2010. Two sides of the same Kelly, Spencer D., Asli O coin: speech and gesture mutually interact to enhance comprehension. Psychological Science 21(2), 260–67. Kemmer, Suzanne, and Michael Barlow. 2000. Introduction: a usage-based conception of language. In Michael Barlow and Suzanne Kemmer (eds.), Usage-based models of language, vii–xxviii. Stanford: CSLI. Kendon, Adam. 1983. Gesture and speech: how they interact. J. M. Weimann and R. P. Harrison (eds.), Nonverbal interaction, 13–45. Beverly Hills: Sage. 1994. Do gestures communicate? A review. Research on Language and Social Interaction 27(3), 175–200. 1995. Gestures as illocutionary and discourse structure markers in Southern Italian conversation. Journal of Pragmatics 23(3), 247–79. 2002. Some uses of the head shake. Gesture 2(2), 147–82. 2004. Gesture: visible action as utterance. Cambridge University Press. Kiefer, Markus, Eun-Jin Sim, Ba¨rbel Herrnberger, Jo Grothe, and Klaus Hoenig. 2008. The sound of concepts: four markers for a link between auditory and conceptual brain systems. Journal of Neuroscience 28, 12224–30. Kiesel, Andrea, and Esther Vierck. 2009. SNARC-like congruency based on number magnitude and response duration. Journal of Experimental Psychology Learning Memory and Cognition 35, 275–79.
References
Kilgarriff, Adam. 2005. Language is never, ever, ever random. Corpus Linguistics and Linguistic Theory 1, 263–76. Kim, Jong-Bok. 2011. English comparative correlative construction: interactions between lexicon and constructions. Korean Journal of Linguistics 36(2), 307–36. Kim, Nuri, Jon Krosnick, and Daniel Casasanto. 2015. Moderators of candidate name-order effects in elections: an experiment. Political Psychology 36(5), 525–42. Kimbara, Irene. 2008. Gesture form convergence in joint description. Journal of Nonverbal Behavior 32, 123–31. Kimmel, Michael. 2005. Culture regained: situated and compound image schemas. In Beate Hampe (ed.), From perception to meaning: image schemas in cognitive linguistics, 205–12. Berlin: Mouton. 2010. Why we mix metaphors (and mix them well): discourse coherence, conceptual metaphor, and beyond. Journal of Pragmatics 42(1), 97–115. King, Brian. 1989. The conceptual structure of emotional experience in Chinese. PhD dissertation. Ohio State University. Kiparsky, Paul. 1982. Lexical phonology and morphology. In-Seok Yang (ed.), Linguistics in the Morning Calm, 3–91. Seoul: Hanshin. Kirjavainen, Minna, Anna Theakston, and Elena Lieven. 2009. Can input explain children’s me-for-I errors? Journal of Child Language 36, 1091–114. Kishner, Jeffrey M., and Raymond W. Gibbs Jr. 1996. How just gets its meanings: polysemy and context in psychological semantics. Language and Speech 39(1), 19–36. Kissling, Elizabeth, Andrea Tyler, Lisa Warren, and Lauren Negrete. 2013. Reexamining por and para in the Spanish foreign language intermediate classroom: a usage-based, cognitive linguistic approach. Paper presented at International Cognitive Linguistics Conference. Edmonton, AB. July. Kita, Sotaro. 2000. How representational gestures help speaking. In David McNeill (ed.), Language and Gesture, 162–85. Cambridge University Press. 2009. Cross-cultural variation of speech-accompanying gesture: a review. Language and Cognitive Processes 24(2), 145–67. Kita, Sotaro (ed.). 2003. Pointing: where language, culture, and cognition meet. Mahwah, NJ: Lawrence Erlbaum. Kita, Sotaro, and James Essegbey. 2001. Pointing left in Ghana: how a taboo on the use of the left hand influences gestural practice. Gesture 1(1), 73–95. ¨ zyu¨rek. 2003. What does cross-linguistic variation Kita, Sotaro, and Asli O in semantic coordination of speech and gesture reveal? Evidence for an interface representation of spatial thinking and speaking. Journal of Memory and Language 48(1), 16–32. Kita, Sotaro, Ingeborg Van Gijn, and Harry van der Hulst. 1998. Movement phase in signs and co-speech gestures, and their transcriptions by
753
754
References
human coders. Gesture and Sign Language in Human-Computer Interaction. Bielefeld Gesture Workshop. September 17–19, 23–35. Klima, Edward, and Ursula Bellugi. 1979. The signs of language. Cambridge, MA: Harvard University Press. Koch, Sabine C., Thomas Fuchs, Michaela Summa, and Cornelia Mu¨ller. 2012. Body memory, metaphor and movement. Amsterdam: Benjamins. Koenig, Jean-Pierre. 1999. Lexical relations. Stanford: CSLI. Kok, Kasper I., and Alan Cienki. 2014. Taking simulation semantics out of the laboratory: towards an interactive and multimodal reappraisal of embodied language comprehension. Language and Cognition. https://doi .org/10.1017/langcog.2014.25. 2016. Cognitive grammar and gesture: points of convergence, advances and challenges. Cognitive Linguistics 27(1), 67–100. Koller, Veronika. 2004. Metaphor and gender in business media discourse: a critical cognitive study. Basingstoke: Palgrave Macmillan. 2014. Cognitive linguistics and ideology. In Jeanette Littlemore and John R. Taylor (eds.), The Bloomsbury companion to cognitive linguistics, 234–52. London: Bloomsbury. Koller, Veronika, Andrew Hardie, Paul Rayson, and Elena Semino. 2008. Using a semantic annotation tool for the analysis of metaphor in discourse. Metaphorik.de 15, 141–60. Kolter, Astrid, Silva H. Ladewig, Michela Summa, Cornelia Mu¨ller, Sabine C. Koch, and Thomas Fuchs. 2012. Body memory and the emergence of metaphor in movement and speech. In Sabine C. Koch, Thomas Fuchs, Michaela Summa, and Cornelia Mu¨ller (eds.), Body memory, metaphor and movement, 201–26. Amsterdam: Benjamins. Kong, Feng. 2013. Space–valence associations depend on handedness: evidence from a bimanual output task. Psychological Research 77(6), 773–79. Konopka, Agnieszka E., and Kathryn Bock. 2008. Lexical or syntactic control of sentence formulation? Structural generalizations from idiom production. Cognitive Psychology 58, 68–101. Koptjevskaja-Tamm, Maria. 2008. Approaching lexical typology. In Martine Vanhove (ed.), From polysemy to semantic change, 3–52. Amsterdam: John Benjamins. Koptjevskaja-Tamm, Maria, and Martine Vanhove (eds.). 2012. New directions in lexical typology. Special issue of Linguistics 50 (3), n. pag. Kortmann, Bernd. 1999. Typology and dialectology. In Bernard Caron (ed.), Proceedings of the 16th International Congress of Linguists, Paris 1997. Amsterdam: Elsevier Science. Kortmann, Bernd, and Kerstin Lunkenheimer (eds.). 2013. The electronic world atlas of varieties of English. Leipzig: Max Planck Institute for Evolutionary Anthropology. Kortmann, Bernd (ed.). 2003. Dialectology meets typology: dialect grammar from a cross-linguistic perspective. Berlin: De Gruyter Mouton.
References
Kosslyn, Stephen Michael. 1980. Image and mind. Cambridge, MA: Harvard University Press. Kousidis, Spyros, David Dorran, Ciaran Mcdonnell, and Eugene Coyle. 2009. Convergence in human dialogues time series analysis of acoustic features. Proceedings of SPECOM 2009, n. pag. Ko¨vecses, Zolta´n. 1986. Metaphors of anger, pride and love: a lexical approach to the study of concepts. Amsterdam: John Benjamins. 1995. Anger: its language, conceptualization, and physiology in the light of cross-cultural evidence. In John R. Taylor and Robert E. MacLaury (eds.), Language and the cognitive construal of the world, 181–96. Berlin: Mouton de Gruyter. 2000. Metaphor and emotion: language, culture, and body in human feeling. Cambridge University Press. 2002. Metaphor: a practical introduction. Oxford University Press. 2005. Metaphor in culture: universality and variation. Cambridge University Press. 2008. The conceptual structure of happiness and pain. In Chryssoula Lascaratou, Anna Despotopoulou, and Elly Ifantidou (eds.), Reconstructing pain and joy: linguistic, literary and cultural perspectives, 17–33. Cambridge Scholars Publishing. 2010. Metaphor: a practical introduction. New York: Oxford University Press. 2013. The metaphor-metonymy relationship: correlations metaphors are based on metonymy. Metaphor and Symbol 28, 75–88. 2015. Where metaphors come from: reconsidering context in metaphor. Oxford University Press. Ko¨vecses, Zolta´n, and Gu¨nter Radden. 1998. Metonymy: developing a cognitive linguistic view. Cognitive Linguistics 9, 37–77. Kraska-Szlenk, Iwona, and Marzena Z˙ygis. 2012. Phonetic and lexical gradience in Polish prefixed words. Cognitive Linguistics 23, 317–66. Kratzer, Angelika. 2012. Modals and conditionals: new and revised perspectives. Oxford University Press. Krauss, Michael. 1992. The world’s languages in crisis. Language 68, 4–10. Krauss, Robert M. 1998. Why do we gesture when we speak? Current Directions in Psychological Science 7(2), 54–60. Krauss, Robert M., Palmer Morrel-Samuels, and Christina Colasante. 1991. Do conversational hand gestures communicate? Journal of Personality and Social Psychology 61(5), 743–54. Kress, Gunther, and Theo Van Leeuwen. 2001. Multimodal discourse: the modes and media of contemporary communication. London: Arnold. Krikmann, Arvo. 2009. On the similarity and distinguishability of humour and figurative speech. Trames 13(1), 14–40. Krippendorff, Klaus. 2004. Content analysis: an introduction to its methodology, 2nd edn. London: Sage. Kristiansen, Gitte. 2003. How to do things with allophones: linguistic stereotypes as cognitive reference points in social cognition.
755
756
References
In Rene´ Dirven, Roslyn Frank, and Martin Pu¨tz (eds.), Cognitive models in language and thought, 69–120. Berlin: De Gruyter Mouton. 2006. Towards a usage-based cognitive phonology. International Journal of English Studies 6, 107–40. 2007. Towards a usage-based cognitive phonology. International Journal of English Studies 6(2), 107–40. Murcia: Universidad de Murcia. 2008. Style-shifting and shifting styles: a socio-cognitive approach to lectal variation. In Gitte Kristiansen and Rene´ Dirven (eds.), Cognitive sociolinguistics, 45–88. De Gruyter, Berlin. ´ Krolak, Emilia. 2008. Fictive interaction: its functions and usage in discourse. PhD dissertation. University of Warsaw. 2016. Polish nominal construction involving fictive interaction: its scope and functions in discourse. In Esther Pascual and Sergeiy Sandler (eds.), The conversation frame: forms and functions of fictive interaction, 235–53. Amsterdam: John Benjamins. Ku¨chenhoff, Helmut, and Hans-Jo¨rg Schmid. 2015. Reply to ‘More (old and new) misunderstandings of collostructional analysis: on Schmid and Ku¨chenhoff’ by Stefan Th. Gries. Cognitive Linguistics 26, 537–47. Kujala, Teija, Kimmo Alho, and Risto Na¨a¨ta¨nen. 2000. Cross-modal reorganization of human cortical functions. Trends in Neurosciences 23(3), 115–20. Kujala, Teijo, Kimmo Ahlo, M. Huotilainen, R. J. Ilmoniemi, A. Lehtokoski, A. Leinonen, et al. 1997. Electrophysiological evidence for cross-modal plasticity in humans with early- and late-onset blindness. Psychophysiology 34, 213–16. Kumashiro, Toshiyuki, and Ronald W. Langacker. 2003. Double-subject and complex-predicate constructions. Cognitive Linguistics 14, 1–45. Kuno, Susumo. 1987. Functional syntax: anaphora, discourse, and empathy. University of Chicago Press. Kuperman, Victor, and Joan Bresnan. 2012. The effects of construction probability on word durations during spontaneous incremental sentence production. Journal of Memory and Language 66(4), 588–611. Kussmaul, Paul. 2010. Verstehen und U¨bersetzen. Tu¨bingen: Narr. Kusters, Annelies. 2012. Adamorobe: a demographic, sociolinguistic and sociocultural profile. In Ulrike Zeshan and Connie De Vos (eds.), Sign languages in village communities: anthropological and linguistic insights, 347–51. Berlin: De Gruyter Mouton. L’Hoˆte, Emilie. 2014. Identity, narrative and metaphor: a corpus-based cognitive analysis of New Labour discourse. Basingstoke: Palgrave Macmillan. Labov, William. 1963. The social motivation of a sound change. Word 18, 1–42. 1966. The social stratification of English in New York City. Washington, DC: Center for Applied Linguistics. 1973. The boundaries of words and their meanings. In C.-J. Bailey and R. Shuy (eds.), New ways of analysing variation in English, 340–73. Washington, DC: Georgetown University Press.
References
2010. Principles of linguistic change, vol. 3: cognitive and cultural factors. Oxford: Wiley. 2014. What is to be learned: the community as the focus of social cognition. In Martin Pu¨tz, Justyna A. Robinson, and Monika Reif (eds.), Cognitive sociolinguistics: social and cultural variation and language use, 23–52. Amsterdam: John Benjamins. Labov, William, Sharon Ash, and Charles Boberg. 2006. Atlas of North American English: phonology and sound change. Berlin: De Gruyter Mouton. Labov, William, Sharon Ash, Maciej Baranowski, Naomi Nagy, Maya Ravindranath, and Tracey Weldon. 2006. Listeners’ sensitivity to the frequency of sociolinguistic variables. University of Pennsylvania Working Papers in Linguistics 12(2), 105–29. Labov, William, Sharon Ash, Maya Ravindranath, Tracey Weldon, Maciej Baranowski, and Naomi Nagye. 2011. Properties of the sociolinguistic monitor. Journal of Sociolinguistics 15, 431–63. Lacey, Simon, Randall Stilla, and K. Sathian. 2012. Metaphorically feeling: comprehending textural metaphors activates somatosensory cortex. Brain and Language 120, 416–21. Ladd, D. Robert, Sea´n G. Roberts, and Dan Dediu. 2015. Correlational studies in typological and historical linguistics. Annual Review of Linguistics 1, 221–41. Ladefoged, Peter, and Ian Maddieson. 1996. The sounds of the world’s languages. Oxford: Blackwell. Lakoff, George. 1987. Women, fire, and dangerous things: what categories reveal about the mind. University of Chicago Press. 1990. The Invariance Hypothesis: is abstract reason based on image schemas? Cognitive Linguistics 1(1). 39–74. 1993. The contemporary theory of metaphor. In Andrew Ortony (ed.), Metaphor and thought, 202–51. Cambridge University Press. 2008. The neural theory of metaphor. In R. Gibbs (ed.), Cambridge handbook of metaphor and thought. New York: Cambridge University Press. 2012. Explaining embodied cognition results. Topics in Cognitive Science 4(4), 773–85. 2014. Mapping the brain’s metaphor circuitry: metaphorical thought in everyday reason. Frontiers in Human Neuroscience 8, 958–87. Lakoff, George, and Mark Johnson. 1980a. Metaphors we live by. University of Chicago Press. 1980b. The metaphorical structure of the human conceptual system. Cognitive Science 4(2), 195–208. 1999. Philosophy in the flesh: the embodied mind and its challenge to Western thought. New York: Basic Books. Lakoff, George, and Srini Narayanan. In prep. Conceptual science: the embodiment of thought and language.
757
758
References
Lakoff, George, and Rafael Nu´n˜ez. 2002. Where mathematics comes from: how the embodied mind brings mathematics into being. New York: Basic Books. Lakoff, George, and Mark Turner. 1989. More than cool reason: a field guide to poetic metaphor. University of Chicago Press. Lam, Yvonne. 2009. Applying cognitive linguistics to teaching Spanish prepositions por and para. Language Awareness 18, 2–18. Lambrecht, Knud, and Kevin Lemoine. 2005. Definite null objects in (spoken) French. In Mirjam Fried and Hans C. Boas (eds.), Grammatical constructions – back to the roots, 13–56. Amsterdam: John Benjamins. Landau, Mark, Michael Robinson, and Brian Meier (eds.). 2015. The power of metaphor: examining the influence on social life. Washington, DC: APA Books. Lane, Harlan. 1984. When the mind hears: a history of the deaf. New York: Random House. Lane, Harlan, and Franc¸ois Grosjean (eds.). 1980. Recent perspectives on American sign language. Hillsdale, NJ: Lawrence Erlbaum Associates. Lane, Wardlow Liane, and Victor S. Ferreira. 2010. Abstract syntax in sentence production: evidence from stem-exchange errors. Journal of Memory and Language 62, 151–65. Langacker, Ronald W. 1972a. Review of Meaning and the structure of language by Wallace Chafe. Language 48, 134–61. 1972b. Possessives in Classical Nahuatl. IJAL 38, 173–86. 1972c. Fundamentals of linguistic analysis. New York: Harcourt Brace Jovanovich. 1975a. Relative clauses in Classical Nahuatl. IJAL 41, 46–68. 1975b. Semantic representations and the linguistic relativity hypothesis. Foundations of Language 14, 307–57. 1986. Abstract motion. Proceedings of the twelfth annual meeting of the Berkeley Linguistics Society, 455–71. Berkeley Linguistics Society. 1987. Foundations of cognitive grammar, vol. 1: theoretical prerequisites. Stanford University Press. 1988. A usage-based model. In Brygida Rudzka-Ostyn (ed.), Topics in Cognitive Linguistics, 127–61. Amsterdam: Benjamins. 1990a. Concept, image, and symbol: the cognitive basis of grammar. Berlin: De Gruyter Mouton. 1990b. Subjectification. Cognitive Linguistics 1(1), 5–38. 1991. Foundations of cognitive grammar, vol. 2: descriptive application. Stanford University Press. 1993. Reference point constructions. Cognitive Linguistics 4, 1–38. 1995. Viewing in cognition and grammar. In Philip W. Davis (ed.), Alternative linguistics: descriptive and theoretical models, 153–212. Amsterdam: John Benjamins. 1997. Constituency, dependency, and conceptual grouping. Cognitive Linguistics 8, 1–32. 1999a. Grammar and conceptualization. Berlin: De Gruyter Mouton.
References
1999b. Virtual reality. Studies in the Linguistic Sciences 29, 77–103. 2000. A Dynamic Usage-Based Model. In Michael Barlow and Suzanne Kemmer (eds.), Usage-based models of language, 1–63. Stanford: CSLI. 2001. Discourse in cognitive grammar. Cognitive Linguistics 12, 143–88. 2002. Concept, image and symbol: the cognitive basis of grammar. New York: De Gruyter Mouton. 2005a. Construction grammars: cognitive, radical, and less so. In Francisco Ruiz de Mendoza and Sandra Pen˜a Cervel (eds.), Cognitive linguistics: internal dynamics and interdisciplinary interaction, 101–59. Berlin: De Gruyter Mouton. 2005b. Dynamicity, fictivity, and scanning: the imaginative basis of logic and linguistic meaning. In Diane Pecher and Rolf A. Zwaan (eds.), Grounding cognition: the role of perception and action in memory, language, and thinking, 164–97. New York: Cambridge University Press. 2006. Subjectification, grammaticization, and conceptual archetypes. In A. Athansiadou, C. Canakis, and B. Cornillie (eds.), Subjectification: various paths to subjectivity, 17–40. Berlin: De Gruyter. 2008a. Cognitive grammar: a basic introduction. New York: Oxford University Press. 2008b. Metaphoric gesture and cognitive linguistics. In Alan Cienki and Cornelia Mu¨ller (eds.), Metaphor and gesture, 249–51, Amsterdam: John Benjamins. 2008c. Sequential and summary scanning: a reply. Cognitive Linguistics 19, 571–84. 2009. Investigations in cognitive grammar. Berlin: De Gruyter Mouton. 2012a. Interactive cognition: toward a unified account of structure, processing, and discourse. International Journal of Cognitive Linguistics 3, 95–125. 2012b. Elliptic coordination. Cognitive Linguistics 23, 555–99. 2013. Essentials of cognitive grammar: a basic introduction. Oxford University Press. 2015a. Baseline and elaboration. Plenary Lecture, International Cognitive Linguistics Conference. Northumbria University, UK. 2015b. Construal. In Ewa Da˛browska and Dagmar Divjak (eds.), Handbook of cognitive linguistics, 120–43. Berlin: De Gruyter Mouton. 2015c. On grammatical categories. Journal of Cognitive Linguistics 1, 44–79. 2015d. How to build an English clause. Journal of Foreign Language Teaching and Applied Linguistics 2(2), n. pag. In prep. Entrenchment in Cognitive Grammar. Langton, Stephen R., and Vicki Bruce. 2000. You must see the point: automatic processing of cues to the direction of social attention. Journal of Experimental Psychology: Human Perception and Performance 26(2), 747–57. Lantolf, James P. 2011. Integrating sociocultural theory and cognitive linguistics in the second language classroom. In Eli Hinkel (ed.),
759
760
References
Handbook of research in second language teaching and learning, vol. II. London: Routledge, 303–18. Lantolf, James P., and Matthew Poehner. 2013. Sociocultural theory and the pedagogical imperative in L2 education: Vyogtskian praxis and the theory/ research divide. London: Routledge. Larson-Hall, Jenifer. 2010. A guide to doing statistics in second language research using SPSS. New York: Routledge. Lascaratou, Chryssoula. 2007. The language of pain: expression or description, Amsterdam: John Benjamins. Lascarides, Alex, and Jon Oberlander. 1993. Temporal coherence and defeasible knowledge. Theoretical Linguistics 19(1), 1–35. Lass, Roger. 1997. Historical linguistics and language change. Cambridge University Press. Lavenda, Robert H., and Emily A. Schultz. 2014. Anthropology: what does it mean to be human? 3rd edn. Oxford University Press. Le Guen, Olivier. 2011. Speech and gesture in spatial language and cognition among the Yucatec Mayas. Cognitive Science 35, 905–38. Le Guen, Olivier, and L. I. Pool Balam. 2012. No metaphorical timeline in gesture and cognition among Yucatec Mayas. Frontiers in Psychology 3, 271. Lee-Goldman, Russell, and Miriam Petruck. In prep. The FrameNet Constructicon in action. In B. Lyngfelt, L. Borin, K. Ohara, and T. Torrent (eds.), Constructicography: construction development across languages. Leembruggen, Linda, Barbara Kelly, and Alice Gaby. In prep. How children talk and think about time. Lehmann, Christian. 2008. Information structure and grammaticalization. In Marı´a Jose´ Lo´pez-Couso and Elena Seoane Posse (eds.), Theoretical and empirical issues in grammaticalization, 207–29. Amsterdam: John Benjamins. 2015. Thoughts on grammaticalization. Berlin: Language Science Press. Lehmann, Thomas. 1989. A grammar of Modern Tamil. Pondicherry Institute of Linguistics and Culture. Leino, Jaakko. 2010. Results, cases, and constructions: argument structure constructions in English and Finnish. In H. C. Boas (ed.), Contrastive studies in construction grammar, 103–36. Amsterdam: Benjamins. ¨ stman. 2005. Constructions and variability. Leino, Jaakko, and Jan-Ola O In Mirjam Fried and Hans C. Boas (eds.), Grammatical constructions: back to the roots, 191–213. Amsterdam: John Benjamins. ¨ stman. 2008. Language change, variability and Leino, Pentti, and Jan-Ola O functional load: Finnish genericity from a constructional point of view. In Jaakko Leino (ed.), Constructional reorganization, 37–54. Amsterdam: John Benjamins. Lemke, Shannon, Antoine Tremblay, and Benjamin V. Tucker. 2009. Function words of lexical bundles: the relation of frequency and reduction. Proceedings of Meetings on Acoustics 6, 060009.
References
Lemmens, Maarten, and Julien Perrez. In prep. French onions and Dutch trains: typological perspectives on learners’ descriptions of spatial scenes. In Andrea Tyler, Lihong Huang and Hana Jan (eds.), What is applied cognitive linguistics? Answers from current SLA research. Berlin: De Gruyter Mouton. Leva¨nen, S., V. Jousmaki, and R. Hari. 1998. Vibration-induced auditorycortex activation in a congenitally deaf adult. Current Biology 8, 869–72. Levelt, Willem J. M. 1989. Speaking: from intention to articulation. Cambridge, MA: MIT Press. Levin, Beth. 1993. English verb class and alternations. University of Chicago Press. Levin, Beth, and Malka Rappaport Hovav. 2005. Argument realization. Cambridge University Press. Levinson, Stephen C. 1983. Pragmatics. Cambridge University Press. 1997. From outer to inner space: linguistic categories and non-linguistic thinking. In Jan Nuyts and Eric Pederson (eds.), Language and conceptualization, 13–45. Cambridge University Press. 2003. Space in language and cognition. Cambridge University Press. 2006. On the human ‘interaction engine.’ In N. J. Enfield and Stephen C. Levinson (eds.), Roots of human sociality: culture, cognition and interaction, 39–69. Oxford: Berg. Levinson, Stephen C., and Penelope Brown. 1994. Immanuel Kant among the Tenejapans: anthropology as empirical philosophy. Ethos 22 (1), 3–41. Levinson, Stephen C., and Nicholas Evans. 2010. Time for a sea-change in linguistics: response to comments on ‘The myth of language universals.’ Lingua 120, 2733–58. Levinson, Stephen C., and Asifa Majid. 2013. The island of time: Ye´lıˆ Dnye, the language of Rossel Island. Frontiers in Psychology 4, 61. Levinson, Stephen C., and Francisco Torreira. 2015. Timing in turn-taking and its implications for processing models of language. Frontiers in Psychology 12, n. pag. Levinson, Stephen C., and David Wilkins (eds.). 2006. Grammars of space. Cambridge University Press. Levshina, Natalia. 2015. How to do linguistics with R. Amsterdam: John Benjamins. Levshina, Natalia, Dirk Geeraerts, and Dirk Speelman. 2013. Mapping constructional spaces: a contrastive analysis of English and Dutch analytic causatives. Linguistics 51(4). 825–54. 2014. Dutch causative constructions: quantification of meaning and meaning of quantification. In Dylan Glynn and Justyna Robinson (eds.), Corpus methods for semantics: quantitative studies in polysemy and synonymy, 205–21. Amsterdam: John Benjamins. Lewandowski, Natalie. 2012. Talent in nonnative phonetic convergence. PhD dissertation. Stuttgart University.
761
762
References
Lewis, C. S. 1936. The allegory of love. Oxford University Press. 1960. Studies in words. Cambridge University Press. Li, Charles N., and Sandra A. Thompson. 1976. Subject and topic: a new typology of language. In Charles N. Li (ed.), Subject and topic, 457–89. New York: Academic Press. Li, Peiwen, Søren Wind Eskildsen, and Teresa Cadierno. 2014. Tracing an L2 learner’s motion constructions over time: a usage-based classroom investigation. Modern Language Journal 98, 612–28. Liberman, Mark. 1979. The intonational system of English. Outstanding dissertations in linguistics series. New York: Garland. Lichtenberk, Frantisek. 1991. Semantic change and heterosemy in grammaticalization. Language 67, 475–509. Liddell, Scott K. 1984. THINK and BELIEVE: sequentiality in American Sign Language. Language 60(2), 372–99. 1995. Real, surrogate and token space: grammatical consequences in ASL. In Karen Emmorey and Judy Reilly (eds.), Language, gesture and space, 19–41. New Jersey: Lawrence Erlbaum Associates. 1998. Grounded blends, gestures, and conceptual shifts. Cognitive Linguistics 9, 283–314. 2000. Blended spaces and deixis in sign language discourse. In David McNeill (ed.), Language and gesture, 331–57. Cambridge University Press. 2003. Grammar, gesture, and meaning in American Sign Language. New York: Cambridge University Press. Lieven, Elena, Julian Pine, and Gillian Baldwin. 1997. Lexically-based learning and the development of grammar in early multi-word speech. Journal of Child Language 24(1), 187–219. Lieven, Elena, Dorothe´ Salomo, and Michael Tomasello. 2009. Two-yearold children’s production of multiword utterances: a usage-based analysis. Cognitive Linguistics 20(3), 481–507. Lieven, Elena. 2016. Usage-based approaches to language development: where do we go from here? Special Issue on usage-based approaches to language and language issues. Language and Cognition 1, 1–23. Lindstromberg, Seth. 2010. English prepositions explained, rev. edn. Amsterdam: John Benjamins. Lindstromberg, Seth, and Frank Boers. 2005. From movement to metaphor with manner-of-movement verbs. Applied Linguistics 26, 241–61. Linell, Per. 2009a. Grammatical constructions in dialogue. In Alexander Bergs and Gabriele Diewald (eds.) Contexts and constructions, 97–110. Amsterdam: John Benjamins. 2009b. Rethinking language, mind, and world dialogically: interactional and contextual theories of human sense-making. Charlotte, NC: Information Age. Littlemore, Jeannette. 2009. Applying cognitive linguistics to second language learning and teaching. London: Palgrave MacMillan.
References
Littlemore, Jeannette, and Graham Low. 2006. Metaphoric competence, second language learning, and communicative language ability. Applied Linguistics 27, 268–94. Littlemore, Jeannette, and Fiona MacArthur. 2012. Figurative extensions of word meaning: how do corpus data and intuition match up? In Stefan Th. Gries and Dagmar S. Divjak (eds.), Frequency effects in language learning and processing, 195–233. Berlin: De Gruyter Mouton. Littlemore, Jeannette, and Caroline Tagg. 2016. Metonymy and text messaging: a framework for understanding creative uses of metonymy. Applied Linguistics, 1–28. Littlemore, Jeannette, and John R. Taylor (eds.). 2014. Bloomsbury companion to cognitive linguistics. London: Bloomsbury. Littlemore, Jeannette, Satomi Arizono, and Alice May. To appear. The interpretation of metonymy by Japanese learners of English. Review of Cognitive Linguistics. 2015. Metonymy: hidden shortcuts in language, thought and communication. Cambridge University Press. Local, John. 1996. Conversational phonetics: some aspects of news receipts in everyday talk. In E. Couper-Kuhlen and M. Selting. (eds.), Prosody in conversation: interactional studies, 177–230. Cambridge University Press. Locke, John. 1979 [1690]. An essay concerning human understanding. New York: Oxford University Press. Loetscher, Tobias, Christopher J. Bockisch, Michael E. R. Nicholls, and Peter Brugger. 2010. Eye position predicts what number you have in mind. Current Biology 20(6), R264–R265. Lohan, Katrin S. 2011. A model of contingency detection to spot tutoring behavior and to respond to ostensive cues in human-robot interaction. PhD dissertation. Bielefeld University. Lohan, Katrin S., Katharina J. Rohlfing, Karola Pitsch, Joe Saunders, Hagen Lehmann, Chrystopher L. Nehaniv, et al. 2012. Tutor spotter: proposing a feature set and evaluating it in a robotic system. International Journal of Social Robotics 4(2), 131–46. Los, Bettelou, Corrien Blom, Geert Booij, Marion Elenbaas, and Ans Van Kemenade. 2012. Morphosyntactic change: a comparative study of particles and prefixes. Cambridge University Press. Lourenco, Stella F., and Matthew R. Longo. 2010. General magnitude representation in human infants. Psychological Science 21(6), 873–81. Louwerse, Max M., Rick Dale, Ellen G. Bard, and Patrick Jeuniaux. 2012. Behavior matching in multimodal communication is synchronized. Cognitive Science 36, 1404–26. Lucas, Ceil. 2001. The sociolinguistics of sign languages. Cambridge University Press. Lu¨cking, Andy, Kirsten Bergman, Florian Hahn, Stefan Kopp, and Hannes Rieser. 2013. Data-based analysis of speech and gesture: the Bielefeld Speech and Gesture Alignment Corpus (SaGA) and its applications. Journal on Multimodal User Interfaces 7(1), 5–18.
763
764
References
Lucy, John A. 1992a. Language diversity and thought: a reformulation of the linguistic relativity hypothesis. Cambridge University Press. 1992b. Grammatical categories and cognition: a case study of the linguistic relativity hypothesis. Cambridge University Press. 1997. Linguistic relativity. Annual Review of Anthropology 26, 291–312. Lum, Jonathon. In prep. Frames of spatial reference in Dhivehi language and cognition. PhD dissertation, Melbourne: Monash University. Lundmark, Carita. 2005. Metaphor and creativity in British magazine advertising. PhD dissertation. Luleaº University of Technology. Lyngfelt, Benjamin. 2012. Re-thinking FNI: on null instantiation and control in Construction Grammar. Constructions and Frames 4(1), 1–23. Lyons, Christopher. 1999. Definiteness. Cambridge University Press. Lyons, John. 1968. Introduction to theoretical linguistics. Cambridge University Press. MacArthur, Fiona, and Jeannette Littlemore. 2008. A discovery approach using corpora in the foreign language classroom. In Frank Boers and Seth Lindstromberg (eds.), Cognitive linguistic approaches to teaching vocabulary and phraseology, 159–88. Berlin: De Gruyter Mouton. Macaulay, Monica. 1982. Verbs of motion and arrival in Mixtec. Proceedings of eighth annual meeting of the Berkeley Linguistic Society, 414–26. Berkeley: BLS. 1985. On the semantics of ‘come,’ ‘go,’ and ‘arrive’ in Otomanguean languages. Kansas Working Papers in Linguistics 10(2), 56–84. Macdonald, Helen. 2014. H is for hawk. London: Vintage Books. Macionis, John J. 2013. Sociology, 15th edn. Boston: Pearson. MacLaury, Robert E. 1976. Ayoquesco Zapotec words for shapes and parts of objects: the use of human body and body part prototypes. Manuscript, Berkeley, CA. 1989. Zapotec body-part locatives. IJAL 55, 119–54. 1995. Vantage theory. In J. R. Taylor and R. MacLaury (eds.), Language and the cognitive construal of the world, 231–76. Berlin: De Gruyter Mouton. MacNeilage, Peter. 2008. The origin of speech. Oxford University Press. MacWhinney, Brian. 1987a. Applying the competition model to bilingualism. Applied Psycholinguistics 8(4), 315–27. 1987b. The competition model. In Brian MacWhinney (ed.), Mechanisms of language acquisition, 249–308. Hillsdale, NJ: Erlbaum. Magnani, Barbara, Massimiliano Oliveri, and Francesca Frassinetti. 2014. Exploring the reciprocal modulation of time and space in dancers and non-dancers. Experimental Brain Research 232(10), 3191–99. Mahon, Bradford and Alfonso Caramazza. 2008. A critical look at the embodied cognition hypothesis and a new proposal for grounding conceptual content. Journal of Physiology–Paris 102, 59–70. Mahpeykar, Narges, and Andrea Tyler. 2015. A principled cognitive linguistics account of English phrasal verbs with up and out. Language and Cognition 7, 1–35.
References
Mahpeykar, Narges, Hana Jan, Andrea Tyler, and Yuka Akiyama. 2015. The multiple meanings of English phrasal verbs: applying cognitive linguistics in the L2 classroom. Paper presented at American Association of Applied Linguists. Toronto, CA. March. Majid, Asifa, and Melissa Bowerman (eds.). 2007. Cutting and breaking events: a crosslinguistic perspective. Special issue of Cognitive Linguistics 18(2), 133–52. Majid, Asifa, Alice Gaby, and Lera Boroditsky. 2013. Time in terms of space. Frontiers in Psychology 4, 554. Majid, Asifa, N. J. Enfield, and Miriam van Staden. 2006. Parts of the body: cross-linguistic categorisation. Special issue of Language Sciences 28(2–3), 137–360. Majid, Asifa, Melissa Bowerman, Sotaro Kita, D. B. M. Haun, and Stephen C. Levinson. 2004. Can language restructure cognition? The case for space. Trends in Cognitive Science, 8, 108–14. Majid, Asifa, Melissa Bowerman, Miriam van Staden, and James S. Boster. 2007. The semantic categories of cutting and breaking events: a crosslinguistic perspective. Cognitive Linguistics 18, 133–52. Mallinson, Christine. 2009. Sociolinguistics and sociology: current directions, future partnerships. Language and Linguistics Compass 3, 1034–51. Malt, Barbara. 2015. Words as names for objects, actions, relations, and properties. In John R. Taylor (ed.), The Oxford handbook of the word, 320–33. Oxford University Press. Mandel, Mark. 1977. Iconic devices in American Sign Language. In Lynn A. Friedman (ed.), On the other hand: new perspectives on American Sign Language, 57–108. New York: Academic Press. Mandelblit, Nili. 1997. Grammatical blending: creative and schematic aspects in sentence processing and translation. PhD dissertation. University of California, San Diego. Mandler, Jean M., and Cristo´bal Paga´n Ca´novas. 2014. On defining image schemas. Language and Cognition 6(4), 510–32. Mangelschots, Katinka, Annelies Jehoul, Steven Schoonjans, and Kurt Feyaerts. In prep. Multimodal marking of obviousness in German and Dutch. CogniTextes. Manson, Joseph H., Gregory A. Bryant, Matthew M. Gervais, and Michelle A. Kline. 2013. Convergence of speech rate in conversation predicts cooperation. Evolution and Human Behavior 34, 419–26. Marghetis, Tyler. 2015. Every number in its place: the spatial foundations of calculation and conceptualization. PhD dissertation. University of California, San Diego Marghetis, Tyler and Rafael Nu´n˜ez. 2013. The motion behind the symbols: a vital role for dynamism in the conceptualization of limits and continuity in expert mathematics. Topics in Cognitive Science 5(2), 299–316.
765
766
References
Martin, James H. 2006. A corpus-based analysis of context effects on metaphor comprehension. In Anatol Stefanowitsch and Stefan Th. Gries (eds.), Corpus-based approaches to metaphor and metonymy, 214–36. Berlin: De Gruyter Mouton. Maslen, Robert J. C., Anna L. Theakston, Elena V. Lieven, and Michael Tomasello. 2004. A dense corpus study of past tense and plural overregularization in English. Journal of Speech, Language, and Hearing Research, 47(6), 1319–33. Mason, Zachary J. 2004. CorMet: a computational, corpus-based conventional metaphor extraction system. Computational Linguistics 30(1), 23–44. Mast, Vivien, Diedrich Wolter, Alexander Klippel, Jan Oliver Wallgru¨n, and Thora Tenbrink. 2014. Boundaries and prototypes in categorizing direction. In Christian Freksa, Bernhard Nebel, Mary Hegarty, and Thomas Barkowsky (eds.), Spatial Cognition 2014, 92–107. Mateo, Josep. 2010. Salud y ritual en Marruecos. Concepciones del cuerpo y pra´cticas de curacio´n. [Health and ritual in Morocco. Conceptions of the body and healing practices.]. Barcelona: Bellaterra. Materna, Jirˇı´. 2010. Building FrameNet in Czech. PhD dissertation. Masaryk University. Matlock, Teenie. 2001. How real is fictive motion? PhD dissertation. University of California, Santa Cruz. 2004a. The conceptual motivation of fictive motion. In Gunter Radden and Rene´ Dirven (eds.), Motivation in Grammar, 221–48. Amsterdam: John Benjamins. 2004b. Fictive motion as cognitive simulation. Memory and Cognition 32, 1389–1400. 2006. Depicting fictive motion in drawings. In June Luchenbroers (ed.), Cognitive Linguistics: investigations across languages, fields, and philosophical boundaries, 67–85. Amsterdam: John H. Benjamins. 2010. Abstract motion is no longer abstract. Language and Cognition 2(2), 243–60. 2013. Motion metaphors in political races. In Michael Borkent, Barbara Dancygier, and Jennifer Hinnell (eds.), Language and the Creative Mind, 193–201. Stanford: CSLI. Matlock, Teenie, and Till Bergman. 2014. Fictive motion. In Ewa Da˛browska and Dagmar Divjak (eds.), Handbook of cognitive linguistics, 771–90. Berlin: DeGruyter Mouton. Matlock, Teenie, and Daniel C. Richardson. 2004. Do eye movements go with fictive motion? Proceedings of the 26th Annual Conference of the Cognitive Science Society, 909–14. Chicago. Matlock, Teenie, and Bodo Winter. 2015. Experimental semantics. In Bernd Heine and Heiko Narrog (eds.), Oxford handbook of linguistic analysis, 771–90. Oxford University Press.
References
Matlock, Teenie, Michael Ramscar, and Lera Boroditsky. 2003. The experiential basis of meaning. In Proceedings of the Twenty-fifth Annual Conference of the Cognitive Science Society, 792–97. Mahwah, NJ: Lawrence Erlbaum. 2005. The experiential link between spatial and temporal language. Cognitive Science 29, 655–64. Matlock, Teenie, Kevin J. Holmes, Mahesh Srinivasan, and Michael Ramscar. 2011. Even abstract motion influences the understanding of time. Metaphor and Symbol 26, 260–71. Matlock, Teenie, Spencer C. Castro, Morgan Fleming, Timothy Gann and Paul Maglio. 2014. Spatial metaphors in web use. Spatial Cognition and Computation 14, 306–20. Matsuki, Kazunaga, Victor Kuperman, and Julie A. Van Dyke. 2016. The Random Forests statistical technique: an examination of its value for the study of reading. Scientific Studies of Reading 20(1), 20–33. Matsuki, Keiko. 1995. Metaphors of anger in Japanese. In John R. Taylor and Robert E. MacLaury (eds), Language and the cognitive construal of the world, 137–51. Berlin: De Gruyter Mouton. Matsumoto, Yo. 1996. Subjective motion and English and Japanese verbs. Cognitive Linguistics 7, 183–226. Matsumoto, Yoshiko. 2010. Interactional frames and grammatical description. Constructions and Frames 2(2), 135–57. 2015. Partnerships between grammatical construction and interactional frame: the stand-alone noun modifying construction in invocatory discourse. Constructions and Frames 7(2), 289–314. Matthews, Danielle, and Colin Bannard. 2010. Children’s production of unfamiliar word sequences is predicted by positional variability and latent classes in a large sample of child-directed speech. Cognitive Science 34(3), 465–88. Matthews, Danielle, and Anna L. Theakston. 2006. Errors of omission in English-speaking children’s production of plurals and the past tense: the effects of frequency, phonology, and competition. Cognitive Science 30(6), 1027–52. Mayberry, Rachel, and Joselynne Jaques. 2000. Gesture production during stuttered speech: insights into the nature of gesture-speech integration. In D. McNeill (ed.), Language and gesture, 99–213. Cambridge University Press. McCarthy, John J. (ed). 2004. Optimality theory in phonology: a reader. Malden, MA: Blackwell. 1982. Prosodic structure and expletive infixation. Language 58(3), 574–90. McCawley, James D. 1988. The comparative conditional construction in English, German and Mandarin Chinese. Berkeley Linguistics Society 14, 176–87. McClave, Evelyn Z. 1994. Gestural beats: the rhythm hypothesis. Journal of Psycholinguistic Research 23(1), 45–66.
767
768
References
2000. Linguistic functions of head movements in the context of speech. Journal of Pragmatics 32(7), 855–78. McCleary, Leland and Evani Viotti. 2009. Sign-gesture symbiosis in Brazilian Sign Language Narrative. In Fey Parrill, Vera Tobin, and Mark Turner (eds.), Meaning, form, and body, 181–201. Stanford, CA: CSLI. McClelland, James L., and David Rumelhart. 1988. Explorations in parallel distributed processing: a handbook of models, programs, and exercises. Boston, MA: MIT Press. McCune, L. 2008. How children learn how to learn language. Oxford University Press. McDonough, Kim. 2006. Interaction and syntactic priming: English L2 speakers’ production of dative constructions. Studies in Second Language Acquisition 28, 179–207. McDonough, Kim, and Alison Mackey. 2006. Responses to recasts: repetitions, primed production and linguistic development. Language Learning 56, 693–720. McDonough, Kim, and Pavel Trofimovich. 2008. Using priming methods in second language research. London: Routledge. McEnery, Tony, and Andrew Hardie. 2012. Corpus linguistics: methods, theory and practice. Cambridge University Press. McGlone, Matthew S., and Jennifer L. Harding. 1998. Back (or forward?) to the future: the role of perspective in temporal language comprehension. Journal of Experimental Psychology: Learning Memory and Cognition 24(5), 1211–23. McKee, Rachel, Adam Schembri, David McKee, and Trevor Johnston. 2011. Variable ‘subject’ presence in Australian Sign Language and New Zealand Sign Language. Language Variation and Change 23(3), 375–98. McNeil, Nicole M., and Martha W. Alibali. 2002. A strong schema can interfere with learning: the case of children’s typical addition schema. In Wayne D. Gray and Christian D. Schunn (eds.), Proceedings of the Twenty-Fourth Annual Conference of the Cognitive Science Society, 661–66. Mahwah, NJ: Erlbaum. McNeill, David. 1992. Hand and mind: what gestures reveal about thought. University of Chicago Press. 2005. Gesture and thought. University of Chicago Press. 2006. Gesture, gaze, and ground. In Steve Renals and Sammy Bengio (eds.), Proceedings of machine learning for multimodal interaction: second international workshop 2005, 1–14. Berlin: Springer Verlag. 2012. How language began: gesture and speech in human evolution. Cambridge University Press. Meier, Brian P., and Michael D. Robinson. 2004. Why the sunny side is up: associations between affect and vertical position. Psychological Science 15(4), 243–47.
References
Meier, Brian P., Michael D. Robinson, and Gerald L. Clore. 2004. Why good guys wear white: automatic inferences about stimulus valence based on brightness. Psychological Science 15(2), 82–87. Meier, Richard P. 1980. Icons and morphemes: models of the acquisition of verb agreement in ASL. Papers and Reports on Child Language Development, 20, 92–99. Meier, Richard P., and Diane Lillo-Martin. 2013. The points of language. Journal of Philosophical Studies 24, 151–76. Meir, Irit, Wendy Sandler, Carol Padden, and Mark Aronoff. 2010. Emerging sign languages. Oxford handbook of deaf studies, language, and education 2, 267–80. Melser, Derek. 2004. The act of thinking. Cambridge, MA: MIT Press. Melzack, Ronald. 1975. The McGill pain questionnaire: major properties and scoring method. Pain 1, 277–99. Mendoza-Denton, Norma, Jennifer Hay, and Stefanie Jannedy. 2003. Probabilistic sociolinguistics. In Rens Bod, Jennifer Hay, and Stefanie Jannedy (eds.), Probabilistic linguistics, 97–138. Cambridge, MA: MIT Press. Mercier, Hugo and Dan Sperber. 2011. Why do humans reason? Arguments for an argumentative theory. Behavioral and Brain Sciences 34(2), 57–74. Merleau-Ponty, Maurice. 1962. Phenomenology of perception. London: Routledge. Merritt, Dustin J., Daniel Casasanto, and Elizabeth M. Brannon. 2010. Do monkeys think in metaphors? Representations of space and time in monkeys and humans. Cognition 117, 191–202. Meteyard, Lotte, Sara R. Cuadrado, Bahador Bahrami, and Gabriella Vigliocco. 2012. Coming of age: a review of embodiment and the neuroscience of semantics. Cortex 48(7), 788–804. Metta, Giorgio, Lorenzo Natale, Francesco Nori, Giulio Sandini, David Vernon, Luciano Fadiga, et al. 2010. The iCub humanoid robot: an open-systems platform for research in cognitive development. Neural Networks 23(8), 1125–34. Meunier, David, Emmanuel A. Stamatakis, and Lorraine K. Tyler. 2014. Age-related functional reorganization, structural changes, and preserved cognition. Neurobiology of Aging 35(1), 42–54. Meurers, Walt Detmar. 2001. On expressing lexical generalizations in HPSG. Nordic Journal of Linguistics 24(2), 161–217. Meurmann-Solin, Anneli, Bettelou Los, and Maria Jose´ Lo´pez-Couso (eds.). 2012. Information structure and syntactic change. New York: Oxford University Press. Mey, Jacob L. 1999. When voices clash: a study in literary pragmatics. Berlin: De Gruyter Mouton. Meyerhoff, Miriam, and James N. Stanford. 2015. ‘Tings change, all tings change’: the changing face of sociolinguistics with a global perspective. In Dick Smakman and Patrick Heinrich (eds.),
769
770
References
Globalising sociolinguistics: challenging and expanding theory, 1–15. Abingdon: Routledge. Michaelis, Laura A. 1994. A case of constructional polysemy in Latin. Studies in Language 18, 45–70. 2010. Sign-Based Construction Grammar. In Bernd Heine and Heiko Narrog (eds.), The Oxford handbook of linguistic analysis, 155–76. Oxford University Press. 2013. Sign-based construction grammar. In Thomas Hoffmann and Graeme Trousdale (eds.), The Oxford handbook of construction grammar, 133–52. Oxford University Press. Michaelis, Laura A., and Knud Lambrecht. 1996. Toward a construction-based model of language function: the case of nominal extraposition. Language 72, 215–47. Michaelis, Laura A., and Josef Ruppenhofer. 2001. Beyond alternations: a constructional model of the German applicative pattern. Stanford: CSLI. Mielke, Jeff. 2008. The emergence of distinctive features. Oxford University Press. Miles, Lynden K., Louise K. Nind, and C. Neil Macrae. 2010. Moving through time. Psychological Science 21(2), 222–23. Miles, Lynden K., Katarzyna Karpin´ska, Joanne Lumsden, and C. Neil Macrae. 2010. The meandering mind: vection and mental time travel. PLoS One 5(5), 1–5. Miles, Lynden K., Lucy Tan, Grant D. Noble, Joanne Lumsden, and C. Neil Macrae. 2011. Can a mind have two time lines? Exploring space-time mapping in Mandarin and English speakers. Psychonomic Bulletin and Review 18(3), 598–604. Miller, Cynthia L. 1996 [2003]. The representation of speech in Biblical Hebrew narrative: a linguistic analysis. Winona Lake, IN: Eisenbrauns. Miller, George A., and Philip N. Johnson-Laird. 1976. Language and perception. Cambridge University Press. Milroy, Lesley. 1980. Language and social networks. Oxford: Blackwell. Minsky, Marvin. 1977. Frame-based theory. In P. N. Johnson-Laird and P. C. Watson (eds.), Thinking and reading in cognitive science, 355–76. Cambridge University Press. Mishkin, Mortimer, Leslie G. Ungerleider, and Kathleen A. Macko. 1983. Object vision and spatial vision: two cortical pathways. Trends in Neurosciences, 6, 414–17. Mishra, Ramesh K., and Niharika Singh. 2010. Online fictive motion understanding: an eye-movement study with Hindi. Metaphor and Symbol 25 (3), 144–61. Mittelberg, Irene. 2007. Methodology for multimodality: one way of working with speech and gesture data. In Monica GonzalezMarquez, Irene Mittelberg, Seana Coulson, and Michael J. Spivey (eds.), Methods in cognitive linguistics, 225–48. Amsterdam: John Benjamins.
References
2014. Multimodal existential constructions in German and English. Paper presented at the 6th International Conference of the German Cognitive Linguistics Association (DGKL6). Erlangen, 30 September– 2 October. Moder, Carol. 2008. It’s like making a soup: metaphors and similes in spoken news discourse. In Andrea Tyler, Yiyoung Kim, and Mari Takada (eds.), Language in the context of use: discourse and cognitive approaches to language, 301–20. Berlin: De Gruyter Mouton. 2010. Two puzzle pieces: fitting discourse context and constructions into Cognitive Metaphor Theory. English Text Construction 3(2) 294–320. Mohanan, K. P. 1986. The theory of lexical phonology. Dordrecht: Reidel. Mompean, Jose. 2004. Category overlap and neutralization: the importance of speakers’ classification in phonology. Cognitive Linguistics 15, 429–69. 2014. Phonology. In Jeanette Littlemore and John R. Taylor (eds.), The Bloomsbury companion to Cognitive Linguistics, 253–76. London: Bloomsbury. Mondada, Lorenza. 2007. Multimodal resources for turn-taking: pointing and emergence of next speakers. Discourse Studies 9(2), 194–225. Moore, Kevin Ezra. 2006. Space to time mappings and temporal concepts. Cognitive Linguistics 17(2), 199–244. 2011a. Ego-perspective and field-based frames of reference: temporal meanings of FRONT in Japanese, Wolof, and Aymara. Journal of Pragmatics 43, 759–76. 2011b. Frames and the experiential basis of the Moving Time metaphor. Constructions and Frames 24, 80–103. 2014. The spatial language of time: metaphor, metonymy, and frames of reference. Amsterdam: John Benjamins. Moore, Randi, Katharine Donelson, Alyson Eggleston, and Juergen Bohnemeyer. 2015. Semantic typology: new approaches to crosslinguistic variation in language and cognition. Linguistics Vanguard 1(1), 189–200. Morrel-Samuels, Palmer, and Robert M. Krauss. 1992. Word familiarity predicts temporal asynchrony of hand gestures and speech. Journal of Experimental Psychology: Learning, Memory, and Cognition 18(3), 615–22. Morris, Desmond, Peter Collet, Peter Marsh, and Marie O’Shaughnessy. 1979. Gestures, their origins and distribution. New York: Stein and Day. Mo¨tto¨nen, Tapani. 2016. Construal in expression: intersubjective approach to cognitive grammar. University of Helsinki. Mueller Gathercole, Virginia C. 2007. Miami and North Wales, so far and yet so near: a constructivist account of morphosyntactic development in bilingual children. International Journal of Bilingual Education and Bilingualism 10(3), 224–47. Mukherjee, Joybrato, and Stefan Th. Gries. 2009. Collostructional nativisation in New Englishes: verb-construction associations in the International Corpus of English. English World-Wide 30(1), 27–51.
771
772
References
Mu¨ller, Cornelia. 2007. Metaphors dead and alive, sleeping and waking: a dynamics view. University of Chicago Press. Mu¨ller, Ralph-Axel, and Surina Basho. 2004. Are nonlinguistic functions in ‘Broca’s area’ prerequisites for language acquisition? fMRI findings from an ontogenetic viewpoint. Brain and Language 89(2), 329–36. Mu¨ller, Stefan. 2006. Phrasal or lexical constructions? Language 82(4), 850–83. 2015. Open review: grammatical theory: from transformational grammar to constraint-based approaches, Textbooks in Language Sciences 1. Berlin: Language Science Press. Mummery, C. J., K. Patterson, C. J. Price, J. Ashburner, R. S. J. Frackowiak, and J. R. Hodges. 2000. A voxel-based morphometry study of semantic dementia: relationship between temporal lobe atrophy and semantic memory. Annals of Neurology 47, 36–45. Munn, Nancy D. 1992. The cultural anthropology of time: a critical essay. Annual Review of Anthropology 21, 93–123. Murphy, Gregory L. 1996. On metaphoric representations. Cognition 60, 173–204. 2002. The big book of concepts. Cambridge, MA: MIT Press. 2004. The big book of concepts, 2nd edn. Cambridge, MA: MIT Press. Murtha, Susan, Howard Chertkow, Mario Beauregard, and Alan Evans. 1999. The neural substrate of picture naming. Journal of Cognitive Neuroscience 11, 399–423. Musolff, Andreas. 2004. Metaphor and political discourse: analogical reasoning in debates about Europe. Basingstoke: Palgrave Macmillan. 2006. Metaphor scenarios in public discourse. Metaphor and Symbol 21(1), 23–38. Næss, Åshild. 2007. Prototypical transitivity. Typological studies in language 72. Amsterdam: John Benjamins. Naigles, Letitia R., Erika Hoff, Donna Vear, Michael Tomasello, Silke Brandt, Sandra R. Waxman, Jane B. Childers, and W. Andrew Collins. 2009. Flexibility in early verb use: evidence from a multiple-n diary study. Monographs of the Society for Research in Child Development. Oxford: John Wiley. Namiki, Takayasu. 2010. Morphological variation in Japanese compounds: the case of hoodai and the notion of ‘compound-specific meaning.’ Lingua 120, 2367–87. Namiki, Takayasu, and Taro Kageyama 2016. Word structure and headedness. In Taro Kageyama and Hideki Kishimoto (eds.), Handbook of Japanese lexicon and word formation, 201–35. Berlin: De Gruyter. Narayan, Shweta. 1997. Knowledge-based action representations for metaphor and aspect (KARMA). PhD dissertation. University of California, Berkeley. 2000. Mappings in art and language: conceptual mappings in Neil Gaiman’s Sandman. Unpublished senior honors thesis in Cognitive Science. University of California, Berkeley.
References
2012. ‘Maybe what it means is he actually got the spot’: physical and cognitive viewpoint in a gesture study. In Barbara Dancygier and Eve Sweetser (eds.), Viewpoint in language: a multimodal perspective, 113–35. Cambridge University Press. Nathan, Geoffrey S. 1986. Phonemes as mental categories. Proceedings of the 12th Annual Meeting of the Berkeley Linguistics Society 12, 212–24. 2007a. Is the phoneme usage-based? – Some issues. International Journal of English Studies 6(2), 173–95. 2007b. Phonology. In Dirk Geeraerts and Hubert Cuykens (eds.), The Oxford handbook of cognitive linguistics, 611–31. Oxford University Press. 2008. Phonology: a Cognitive Grammar introduction. Amsterdam: Benjamins. 2009. Where is the natural phonology phoneme in 2009? Poznan´ Studies in Contemporary Linguistics 45(1), 141–48. 2015. Phonology. In Ewa Da˛browska and Dagmar Divjak (eds), Handbook of cognitive linguistics, 253–73. Berlin: De Gruyter Mouton. Nayak, Naomi P., and Raymond W. Gibbs, Jr. 1990. Conceptual knowledge in the interpretation of idioms. Journal of Experimental Psychology 119(3), 315–30. Nemoto, Noriko. 1998. On the polysemy of ditransitive save: the role of frame semantics in construction grammar. English Linguistics 15, 219–42. 2005. Verbal polysemy and Frame Semantics in Construction Grammar: some observations about the locative alternation. In Mirjam Fried and Hans C. Boas (eds.), Grammatical constructions: back to the roots, 119–38. Amsterdam: John Benjamins. Nesset, Tore. 2008. Abstract phonology in a concrete model: cognitive linguistics and the morphology-phonology interface. Berlin: De Gruyter Mouton. Nesset, Tore, and Laura A. Janda. 2010. Paradigm structure: evidence from Russian suffix shift. Cognitive Linguistics 21, 699–725. Nesset, Tore, Anna Endresen, Laura A. Janda, Anastasia Makarova, Francis Steen, and Mark Turner. 2013. How here and now in Russian and English establish joint attention in TV news broadcasts. Russian Linguistics 37, 229–51. Neuman, Yair, Dan Assaf, Yohai Cohen, Mark Last, Shlomo Argamon, Netwon Howard, and Ophir Frieder. 2013. Metaphor identification in large texts corpora. PloS ONE 8(4), 1–9. Neville, Helen. 1995. Developmental specificity in neurocognitive development in humans. In Michael S. Gazzaniga (ed.), The cognitive neurosciences, 219–31, Cambridge, MA: MIT Press. Newell, Karl M., and P. Vernon McDonald. 1997. The development of grip patterns in infancy. In Kevin Connolly and Hans Forssberg (eds.). Neurophysiology and neuropsychology, 232–56. London: McKeith Press. Newman, Aaron J., Ted Supalla, Nina Fernandez, Elissa L. Newport, and Daphne Bavelier. 2015. Neural systems supporting linguistic structure, linguistic experience, and symbolic communication in sign language and gesture. Proceedings of the National Academy of Sciences 112(37), 11684–89.
773
774
References
Newman, John. 1996. Give: a cognitive linguistic study. Berlin: De Gruyter Mouton. (ed.). 1998. The linguistics of giving. Amsterdam: John Benjamins. (ed.). 2002. The linguistics of sitting, standing, and lying. Amsterdam: John Benjamins. (ed.). 2009. The linguistics of eating and drinking. Amsterdam: John Benjamins. Newport, Elissa L. 2016. Statistical language learning: computational, maturational and linguistic constraints. Special issue: usage-based approaches to language and language issues. Language and Cognition 8, 3. Nichols, Johanna. 1984. Functionalist theories of grammar. Annual Review of Anthropology 13, 97–117. 1988. On alienable and inalienable possession. In William Shipley (ed.), In honor of Mary Haas: from the Haas Festival Conference on Native American Linguistics, 557–609. Berlin: De Gruyter Mouton. Nickels, Ernest. 2007. Good guys wear black: uniform color and citizen impressions of police. Policing: An International Journal of Police Strategies and Management 31(1), 77–92. Niederhoff, Burkhard. 2009a. Focalization. In Peter Hu¨hn, John Pier, Wolf Schmid, and Jo¨rg Scho¨nert (eds.), Handbook of narratology, 115–23. Berlin: Walter de Gruyter. 2009b. Perspective/point of view. In Peter Hu¨hn, John Pier, Wolf Schmid, and Jo¨rg Scho¨nert (eds.), Handbook of narratology, 384–97. Berlin: Walter de Gruyter. Niemeier, Susanne. 2008. The notion of boundedness/unboundedness in the foreign language classroom. In Frank Boers and Seth Lindstromberg (eds.), Cognitive linguistic approaches to teaching vocabulary and phraseology, 309–27. Berlin: De Gruyter Mouton. Nikiforidou, Kiki. 2010. Viewpoint and construction grammar: the case of past + now. Language and Literature 19(3), 265–84. 2012. The constructional underpinnings of viewpoint blends: the past + now in language and literature. In Barbara Dancygier and Eve Sweetser (eds.), Viewpoint and perspective in language and gesture, 177–97. Cambridge University Press. 2015. Grammatical constructions and cross-text generalizations: Empathetic narration as genre. Constructions and Frames 7(2), 181–217. Nikiforidou, Kiki, and Kerstin Fischer. 2015. On the interaction of constructions with register and genre: Introduction to the special issue. Constructions and Frames 7(2), 137–47. Nikiforidou, Kiki, Sophia Marmaridou, and George K. Mikros. 2014. What’s in a dialogic construction? A constructional approach to polysemy and the grammar of challenge. Cognitive Linguistics 25(4), 655–99. Noble, Claire, Faria Iqbal, Elena Lieven, and Anna Theakston. 2015. Converging and competing cues in the acquisition of syntactic structures: the conjoined agent intransitive. Journal of Child Language 1, 1–32.
References
Noe¨l, Dirk. 2007a. Diachronic construction grammar and grammaticalization theory. Functions of Language 14(2), 177–202. 2007b. Verb valency patterns, constructions and grammaticalization. In Thomas Herbst and Karin Ko¨tz-Votteler (eds.), Valency: theoretical, descriptive and cognitive issues. Trends in Linguistics, Studies and Monographs 187, 67–83. Berlin: De Gruyter Mouton. 2008. The nominative and infinitive in Late Modern English: a diachronic constructionist approach. Journal of English Linguistics 36(4), 314–40. Noe¨l, Dirk, and Timothy Colleman. 2009. The nominative and infinitive in English and Dutch: an exercise in contrastive diachronic construction grammar. Languages in Contrast 9, 144–81. Norde, Muriel, and Kristel Van Goethem. 2014. Bleaching, productivity and debonding of prefixoids: a corpus-based analysis of ‘giant’ in German and Swedish. Linguisticae Investigationes 37, 256–74. Norrick, Neal R. 2011. Conversational recipe telling. Journal of Pragmatics 43, 2740–61. Novack, Miriam A., Eliza L. Congdon, Naureen Hemani-Lopez, and Susan Goldin-Meadow. 2014. From action to abstraction: using the hands to learn math. Psychological Science 25 (4), 903–10. Nunberg, Geoffrey, Ivan A. Sag, and Thomas Wasow. 1994. Idioms. Language 70, 491–538. Nu´n˜ez, Rafael, and Kensy Cooperrider. 2013. The tangle of space and time in human cognition. Trends in Cognitive Sciences 17(5), 220–29. Nu´n˜ez, Rafael and Eve Sweetser. 2006a. Aymara, where the future is behind you: convergent evidence from language and gesture in the crosslinguistic comparison of spatial realizations of time. Cognitive Science 30, 410–50. 2006b. With the future behind them: convergent evidence from Aymara language and gesture in the crosslinguistic comparison of spatial construals of time. Cognitive Science 30(3), 401–50. ´ Nun˜ez, Rafael, Benjamin A. Motz, and Ursina Teuscher. 2006. Time after time: the psychological reality of the ego- and time-references-point distinction in metaphorical construals of time. Metaphor and Symbol 21, 133–46. Nu´n˜ez, Rafael, Kensy Cooperrider, D. Doan, and Ju¨rg Wassmann. 2012. Contours of time: topographic construals of past, present, and future in the Yupno valley of Papua New Guinea. Cognition 124(1), 25–35. Nuyts, Jan. 2014. Notions of (inter)subjectivity. In Lieselotte Brems, Lobke Ghesquie`re, and Freek Van de Velde (eds.), Intersubjectivity and intersubjectification in grammar and discourse, 53–76. Amsterdam: John Benjamins. Nycz, Jennifer. 2013. Changing words or changing rules? Second dialect acquisition and phonological representation. Journal of Pragmatics 52, 49–62. O’Farrell, Maggie. 2013. Instructions for a heatwave. London: Tinderpress.
775
776
References
O’Halloran, Kieran. 2007. Critical discourse analysis and the corpus-informed interpretation of metaphor at the register level. Applied Linguistics 28(1), 1–24. O’Leary, D. D., and Y. Nakagawa. 2002. Patterning centers, regulatory genes and extrinsic mechanisms controlling arealization of the neocortex. Current Opinion in Neurobiology 12, 14–25. Oakley, Todd. 1998. Conceptual blending, narrative discourse, and rhetoric. Cognitive Linguistics 9(4), 320–60. 2007. Image schemas. In Dirk Geeraerts and Hubert Cuyckens (eds.), Oxford handbook of cognitive linguistics, 214–35. New York: Oxford University Press. 2009. From attention to meaning: explorations in semiotics, linguistics and rhetoric. Bern: Lang Verlag. ¨ stman and Jef Verschueren 2012. Conceptual Integration. In Jan-Ola O (eds.), Handbook of pragmatics, vol. 6, 1–25. Amsterdam: John Benjamins. 2014. Semantic domains in the Dream of the Rood. International Journal for Language and Communication 40, 331–52. Oakley, Todd, and Per Aage Brandt. 2009. Hypotyposis: meta-representation, mind-reading, and fictive interaction. In Wolfgang Wildgen and Barend van Heusden (eds.), Metarepresentation, self-organization and art, 115–36. Berlin: Peter Lang. Oakley, Todd, and Seana Coulson. 2008. Connecting the dots: mental spaces and metaphoric language in discourse. In Todd Oakley and Anders Hougaard (eds.), Mental spaces in discourse and interaction, 27–50. Amsterdam: John Benjamins. Oakley, Todd, and Anders Hougaard. 2008. Mental spaces in discourse and interaction. Amsterdam: John Benjamins. Oakley, Todd, and David Kaufer. 2008. Experience by design: three layers of analysis in clinical reports. In Todd Oakley and Anders Hougaard (eds.), Mental spaces in discourse and interaction, 149–78. Amsterdam: John Benjamins. Oakley, Todd, and Vera Tobin. 2014. The whole is sometimes less than the sum of its parts: towards a theory of document acts. Language and Cognition 6(1), 79–110. Oben, Bert. 2015. Modelling interactive alignment: a multimodal and temporal account. PhD dissertation. University of Leuven. Oben, Bert, and Geert Broˆne. 2015. What you see is what you do: on the relation between gaze and gesture in multimodal alignment. Language and Cognition 7, 546–62. Obhi, S. Sukhvinder, and Natalie Sebanz. 2011. Moving together: toward understanding the mechanisms of joint action. Experimental Brain Research 211, 329–36. Occhino, Corrine. 2016. A cognitive approach to phonology: evidence from signed languages. PhD dissertation. University of New Mexico, Albuquerque.
References
Ogura, Mieko. 2012. The timing of language change. In Juan Manuel Herna´ndez-Campoy and Juan Camilo Conde-Silvestre (eds.), The Blackwell handbook of historical sociolinguistics, 427–50. Oxford: Blackwell. Ohara, Kyoko. 2009. Frame-based contrastive lexical semantics in Japanese FrameNet: the case of risk and kakeru. In Hans C. Boas (ed.), Multilingual FrameNets: methods and applications, 163–82. Berlin: De Gruyter Mouton. Ohara, Kyoko Hirose, Seiko Fujii, Toshio Ohori, Ryoko Suzuki, Hiroaki Saito, and Shun Ishizaki. 2004. The Japanese FrameNet Project: an introduction. Proceedings of the workshop on building lexical resources from semantically annotated corpora, 9–12. Ohm, Eyvind, and Valerie A. Thompson. 2004. Everyday reasoning with inducements and advice. Thinking and Reasoning 10(3), 241–72. Ojemann, George A. 1983. Brain organization for language from the perspective of electrical stimulation mapping. Behavioral and Brain Sciences 6, 189–230. Olofsson, Joel. 2014. Argument structure constructions and syntactic productivity – the case of Swedish Motion Constructions. Constructions 1 (7), 1–17. Open Science Collaboration. 2015. Estimating the reproducibility of psychological science. Science 349(6251), aac4716 1–8. Oppenheimer, Daniel M., and Thomas E. Trail. 2010. When leaning to the left makes you lean to the left. Social Cognition 28, 651–61. Orban, Guy A., David Van Essen, and Wim Vanduffel. 2004 Comparative mapping of higher visual areas in monkeys and humans. Trends in Cognitive Science 8, 315–24. Ortega, Lourdes, Andrea Tyler, Hae In Park, and Mariko Uno (eds.). 2016. The usage-based study of language learning and multilingualism. Georgetown University Press. Osborne, Timothy, and Thomas Gross. 2012. Constructions are catenae: construction grammar meets dependency grammar. Cognitive Linguistics 23, 165–216. Osthoff, Hermann, and Karl Brugmann. 1887. Morphologische Untersuchungen auf dem Gebiete der indogermanischen Sprachen. Vol. 1. Leipzig: S. Hirzel. ¨ stman, Jan-Ola. 2002. Sulvan kansan wellerismit konstruktiona. O In Ilona Herlin, Jyrki Kalliokoski, Lari Kotilainen, and Tiina Onikki Rantaja¨a¨sko¨. (eds.), A¨idinkielen merkitykset, 75–97. Helsinki: Finnish Literature Society. ¨ stman and 2005. Construction discourse: a prolegomenon. In Jan-Ola O Mirjam Fried (eds.), Construction grammars: cognitive grounding and theoretical extensions, 121–44. Amsterdam: John Benjamins. ¨ stman, Jan-Ola, and Graeme Trousdale. 2013. Dialects, discourse, and O construction grammar. In Thomas Hoffmann and Graeme Trousdale (eds.), The Oxford handbook of construction grammar, 476–90. Oxford University Press.
777
778
References
Ouellet, Marc, Julio Santiago, Ziv Israeli, and Shai Gabay. 2010. Is the future the right time? Experimental Psychology 57(4), 308–14. Palmer, Gary. 1996. Toward a theory of cultural linguistics. Austin: University of Texas Press. Palmer, Gary, Cliff Goddard, and Penny Lee (eds.). 2003. Talking about thinking across languages. Special issue of Cognitive Linguistics 14 (2/3). Pang, Kam-yiu S. 2005. ‘This is the linguist in me speaking’: constructions for talking about the self talking. Functions of Language 12(1), 1–38. Panther, Klaus-Uwe. 2005. The role of conceptual metonymy in meaning construction. In Francisco J. Ruiz de Mendoza Iba´n˜ez and M. Sandra Pen˜a Cervel (eds.), Cognitive linguistics: internal dynamics and interdisciplinary interaction, 353–86. Berlin: De Gruyter Mouton. Panther, Klaus-Uwe, and Gu¨nter Radden (eds.). 1999. Metonymy in language and thought. Amsterdam: John Benjamins. Panther, Klaus-Uwe, and Linda Thornburg. 1998. A cognitive approach to inferencing in conversation. Journal of Pragmatics 30, 755–69. (eds.). 2003. Metonymy and pragmatic inferencing. Amsterdam: John Benjamins. 2007. Metonymy. In Dirk Geeraerts and Hubert Cuykens (eds.), The Oxford handbook of cognitive linguistics, 236–62. Oxford University Press. 2009. On figuration in grammar. In Klaus-Uwe Panther, Linda Thornburg, and Antonio Barcelona (eds.), Metonymy and metaphor in grammar, 1–44. Amsterdam: John Benjamins. Panther, Klaus-Uwe, Linda Thornburg, and Antonio Barcelona (eds.). 2009. Metonymy and metaphor in grammar. Amsterdam: John Benjamins. Paradis, Johanne, Elena Nicoladis, Martha Crago, and Fred Genesee. 2011. Bilingual children’s acquisition of the past tense: a usage-based approach. Journal of Child Language 38(3), 554–78. Pardo, Jennifer S. 2006. On phonetic convergence during conversational interaction. The Journal of the Acoustical Society of America 119, 2382. Parise, Cesare V., Katharina Knorre, and Marc O. Ernst. 2014. Natural auditory scene statistics shapes human spatial hearing. Proceedings of the National Academy of Sciences 111(16), 6104–08. Parrill, Fey. 2008. Subjects in the hands of speakers: an experimental study of syntactic subject and speech-gesture integration. Cognitive Linguistics 19(2), 283–99. 2009. Dual viewpoint gestures. Gesture 9(3), 271–89. 2010. Viewpoint in speech-gesture integration: linguistic structure, discourse structure, and event structure. Language and Cognitive Processes 25(5), 650–68. 2012. Interactions between discourse status and viewpoint in cospeech gesture. In Barbara Dancygier and Eve Sweetser (eds.), Viewpoint in language: a multimodal perspective, 97–112. Cambridge University Press.
References
Parrill, Fey, and Eve Sweetser. 2004. What we mean by meaning: Conceptual Integration in gesture analysis and transcription. Gesture 4(2), 197–219. Parrill, Fey, Benjamin K. Bergen, and Patricia V. Lichtenstein. 2013. Grammatical aspect, gesture, and conceptualization: using co-speech gesture to reveal event representations. Cognitive Linguistics 24(1), 135–58. Parrill, Fey, Vera Tobin, and Mark Turner (eds.). 2010. Meaning, form, and body. University of Chicago Press. Partington, Alan, Alison Duguid, and Charlotte Taylor. 2013. Patterns and meanings in discourse: theory and practice in corpus-assisted discourse studies. Amsterdam: John Benjamins. Pascual, Esther. 2002. Imaginary trialogues: conceptual Blending and fictive interaction in criminal courts. Utrecht: LOT. 2006a. Questions in legal monologues: fictive interaction as argumentative strategy in a murder trial. Text and Talk 26(3), 383–402. 2006b. Fictive interaction within the sentence: a communicative type of fictivity in grammar. Cognitive Linguistics 17(2), 245–267. 2008a. Fictive interaction blends in everyday life and courtroom settings. In Todd Oakley and Anders Hougaard (eds.), Mental spaces in discourse and interaction, 79–107. Amsterdam: John Benjamins. 2008b. Text for context, trial for trialogue: an ethnographic study of a fictive interaction blend. Annual Review of Cognitive Linguistics 6, 50–82. 2014. Fictive interaction: the conversation frame in thought, language, and discourse. Amsterdam: John Benjamins. Pascual, Esther, and Line Brandt. 2015. Embodied fictive interaction metaphors: the primacy of the Conversation Frame in dance discourse. In Sander Lestrade, Peter de Swart, and Lotte Hogeweg (eds.), Addenda. Artikelen voor Ad Foolen, 321–34. Nijmegen: Radboud Repository. Pascual, Esther, and Sergeiy Sandler (eds.). 2016. The conversation frame: forms and functions of fictive interaction. Amsterdam: John Benjamins. Pascual, Esther, Emilia Kro´lak, and Theo A. J. M. Janssen. 2013. Direct speech compounds: evoking sociocultural scenarios through fictive interaction. Cognitive Linguistics 24(2), 345–66. Patten, Amanda. 2010. Grammaticalization and the it-cleft construction. In Graeme Trausdale and Elizabeth C. Traugott (eds.), Gradience, gradualness and grammaticalization, 221–43. Amsterdam: John Benjamins. Paxton, Alexandra, Drew H. Abney, Christopher T. Kello, and Rick Dale. 2014. Network analysis of multimodal, multiscale coordination in dyadic problem solving. Proceedings of the 36th Annual Meeting of the Cognitive Science Society, 2735–40. Pearson, Richard G. 1961. Judgment of volume from two-dimensional representations of complex irregular shapes. PhD dissertation. Carnegie Institute of Technology.
779
780
References
1964. Judgment of volume from photographs of complex shapes. Perceptual and Motor Skills 18, 889–900. Pederson, Eric, Eve Danziger, David Wilkins, Stephen Levinson, Sotaro Kita, and Gunter Senft. 1998. Semantic typology and spatial conceptualization. Language 74, 557–89. Peirce, Charles Sanders. 1932. Collected writings, 2: Elements of logic. Cambridge, MA: Harvard University Press. Pen˜a, Jorge, Jeffrey T. Hancock, and Nicholas A. Merola. 2009. The priming effects of avatars in virtual settings. Communication Research 36(6), 838–56. Pentrel, Meike. 2015. The position of adverbial clauses in the diary of Samuel Pepys (1660–1669): a cognitive historical study. PhD dissertation. Osnabru¨ck University. Perdue, Clive (ed.) 1993. Adult Language Acquisition. Vol 1: Field Methods. Cambridge University Press. Perek, Florent. 2014. Rethinking constructional polysemy: the case of the English conative construction. In Dylan Glynn and Justyna Robinson (eds.), Corpus methods for semantics: quantitative studies in polysemy and synonymy, 61–85. Amsterdam: John Benjamins. 2015. Argument structure in usage-based construction grammar: experimental and corpus-based perspectives. Amsterdam: John Benjamins. Perry, Michelle, R. Breckinridge Church, and Susan Goldin-Meadow. 1988. Transitional knowledge in the acquisition of concepts. Cognitive Development 3(4), 359–400. Petitto, L. A., R. J. Zatorre, K. Gauna, E. J. Nikelski, D. Dostie, and A. C. Evans. 2000. Speech-like cerebral activity in profoundly deaf people processing signed languages: implications for the neural basis of human language. Proceedings of the National Academy of Sciences, 97(25), 13961–66. Petre´, Peter. 2014. Constructions and environments: copular, passive and related constructions in Old and Middle English: Oxford Studies in the History of English 4. Oxford University Press. ´ Petre, Peter, and Hubert Cuyckens. 2008. Bedusted, yet not beheaded: the role of be-’s constructional properties in its conservation. In Alexander Bergs and Gabriele Diewald (eds.). Constructions and language change, 133–70. Berlin: De Gruyter Mouton. Petruck, Miriam R. L. 1996. Frame semantics. In Jef Verschueren, Jan-Ola ¨ stman, Jan Blommaert, and Chris Bulcaen (eds.), Handbook of pragO matics, 1–11. Philadelphia: John Benjamins. 2009. Typological considerations in constructing a Hebrew FrameNet. In Hans C. Boas (ed.), Multilingual FrameNets: methods and applications, 183–208. Berlin: De Gruyter Mouton. 2013. Advances in frame semantics. In Mirjam Fried and Kiki Nikiforidou (eds.), Advances in frame semantics, 1–12. Amsterdam: John Benjamins. Petruck, Miriam R. L., and Hans C. Boas. 2003. All in a day’s week. In E. Hajicova, A. Kotesovcova and Jiri Mirovsky (eds.), Proceedings of CIL 17, CD-ROM. Prague: Matfyzpress.
References
Petruck, Miriam R. L., Charles J. Fillmore, Collin Baker, Michael Ellsworth, and Josef Ruppenhofer. 2004. Reframing FrameNet data. Proceedings of the 11th EURALEX International Congress, Lorient, France, 405–16. Pfau, Roland and Martin Steinbach. 2006. Modality independent and modality specific aspects of grammaticalization in Sign Languages. Potsdam: Universita¨tsverlag Potsdam. Piaget, Jean. 1962. Play, dreams, and imitation in childhood. New York: W. W. Norton. 1972. The psychology of the child. New York: Basic Books. 1990. The child’s conception of the world. New York: Littlefield Adams. Pickering, Martin J., and Simon Garrod. 2004. Towards a mechanistic psychology of dialogue. Behavioural and Brain Sciences 27, 169–225. 2006. Alignment as the basis for successful communication. Research on Language and Communication 4, 203–88. Pierrehumbert, Janet B. 2001. Exemplar dynamics: word frequency, lenition, and contrast. In Joan Bybee and Paul J. Hopper (eds.), Frequency and the emergence of linguistic structure, 137–57. Amsterdam: John Benjamins. 2002. Word-specific phonetics. In Carlos Gussenhoven and Natasha Warner (eds.), Laboratory phonology 7, 101–39. Berlin: De Gruyter Mouton. 2003a. Probabilistic phonology: discrimination and robustness. In Rens Bod, Jennifer Hay, and Stefanie Jannedy (eds.), Probabilistic linguistics, 177–228. Cambridge, MA: MIT Press. 2003b. Phonetic diversity, statistical learning, and acquisition of phonology. Language and Speech 46(2–3), 115–54. 2006. The next toolkit. Journal of Phonetics 34, 516–30. Pine, Karen J., Nicola Lufkin, and David Messer. 2004. More gestures than answers: children learning about balance. Developmental Psychology 40(6), 1059–67. Ping, Raedy, and Susan Goldin-Meadow. 2008. Hands in the air: using ungrounded iconic gestures to teach children conservation of quantity. Developmental Psychology 44 (5), 1277–87. Ping, Raedy, Susan Goldin-Meadow, and Sian L. Beilock. 2014. Understanding gesture: is the listener’s motor system involved? Journal of Experimental Psychology: General, 143(1), 195–204. Pinker, Steven. 1989. Learnability and cognition. Cambridge, MA: MIT Press. 1997. How the mind works. New York: W. W. Norton. 2007. The stuff of thought. New York: Basic Books. Pitt, Benjamin, and Daniel Casasanto. 2014. Experiential origins of the mental number line. In Paul Bello, Marcello Guarini, Marjorie McShane, and Brian Scassellati (eds.), Proceedings of the 36th Annual Conference of the Cognitive Science Society, 1174–79. Austin: Cognitive Science Society. 2016. Reading experience shapes the mental timeline but not the mental number line. In Daniel Grodner, Daniel Mirman, Anna Papafragou,
781
782
References
and John Trueswell (eds.), Proceedings of the 38th Annual Conference of the Cognitive Science Society, 2753–58. Austin: Cognitive Science Society. Pizzuto, Elena, and Sherman Wilcox. 2001. A study of modal verbs, subjectivity, and gesture in Italian Sign Language: final report, Italian National Research Council, Short Term Mobility Grant Programme. Plag, Ingo, Christiane Dalton-Puffer, and R. Harald Baayen. 1999. Productivity and register. English Language and Linguistics 3, 209–28. Plank, Frans. 1984. The modals story retold. Studies in Language 8, 306–64. Poeppel, David, William Idsardi, and Virginie van Wassenhove. 2008. Speech perception at the interface of neurobiology and linguistics. Philosophical Transactions of the Royal Society 363, 1071–86. Pollard, Carl, and Ivan A. Sag. 1994. Head-driven phrase structure grammar. University of Chicago Press. Port, Robert. 2007. How are words stored in memory? Beyond phones and phonemes. New Ideas in Psychology 25, 143–70. 2010. Language as a social institution: why phonemes and words do not live in the brain. Ecological Psychology 22, 304–26. Po¨rtner, Paul H., and Barbara H. Partee. 2002. Formal semantics: the essential readings. Oxford: John Wiley and Sons. Poser, William. 1990. Foot structure in Japanese. Language 66(1), 78–105. Pragglejaz Group. 2007. MIP: a method for identifying metaphorically used words in discourse. Metaphor and Symbol 22(1), 1–39. Pratt, Carroll C. 1930. The spatial character of high and low tones. Journal of Experimental Psychology 13, 278–85. Preston, Dennis. 2013. The influence of regard on language variation and change. Journal of Pragmatics 52, 93–104. Price, Cathy J. 2012. A review and synthesis of the first 20 years of PET and fMRI studies of heard speech, spoken language and reading. Neuroimage 62(2), 816–47. Pullum, Geoffrey K. 1991. The great Eskimo vocabulary hoax and other irreverent essays on the study of language. University of Chicago Press. Pulvermu¨ller, Friedemann. 1993. On connecting syntax and the brain. In Ad Aertsen (ed.), Brain theory: spatio-temporal aspects of brain function, 131–45. New York: Elsevier. 2003. The neuroscience of language. Cambridge University Press. 2010. Brain embodiment of syntax and grammar: discrete combinatorial mechanisms spelt out in neuronal circuits. Brain and Language 112(3), 167–79. Pulvermu¨ller, Friedemann, and Andreas Knoblauch. 2009. Discrete combinatorial circuits emerging in neural networks: a mechanism for rules of grammar in the human brain? Neural Networks 22(2), 161–72. Pulvermu¨ller, Friedemann, Bert Cappelle, and Yury Shtyrov. 2013. Brain basis of meaning, words, constructions, and grammar. In Thomas Hoffmann and Graeme Trousdale (eds.), The Oxford handbook of construction grammar, 397–416. Oxford University Press.
References
Pulvermu¨ller, Friedemann, Yury Shtyrov, and Bert Cappelle. 2013. Brain basis of meaning, words, constructions, and grammar. In Thomas Hoffmann and Graeme Trousdale (eds.), The Oxford handbook of construction grammar, 397–416. Oxford University Press. Pulvermuller, Friedemann, Y. Shtyrov, and R. Ilmoniemi. 2005. Brain signatures of meaning access in action word recognition. Journal of Cognitive Neuroscience 17, 884–92. Pu¨tz, Martin, and Marjolijn H. Verspoor (eds.). 2000. Explorations in linguistic relativity. Amsterdam: John Benjamins. Pu¨tz, Martin, Justyna A. Robinson, and Monika Reif (eds.). 2014. Cognitive sociolinguistics: social and cultural variation in cognition and language use. Amsterdam: John Benjamins. Quine, Willard van Orman. 1951. Two dogmas of empiricism. The Philosophical Review 60 (1), 20–43. 1960. Word and object. Cambridge, MA: MIT Press. Quinto-Pozos, David, and Fey Parrill. 2015. Signers and co-speech gesturers adopt similar strategies for portraying viewpoint in narratives. Topics in Cognitive Science 7(1), 12–35. R Development Core Team. 2010. R: A Language and Environment for Statistical Computing. Vienna, Austria: R Foundation for Statistical Computing. Radden, Gu¨nter P. 1997. Time is space. In Birgit Smieja and Meike Tasch (eds.), Human Contact through Language and Linguistics, 147–66. Frankfurt: Peter Lang. Radden, Gu¨nther P., and Zoltan Ko¨vecses. 1999. Towards a theory of metonymy. In Klaus-Uwe Panther and Gu¨nther Radden (eds.), Metonymy in language and thought, 17–59. Amsterdam: John Benjamins. Radden, Gu¨nther P., Klaus-Michael Ko¨pcke, Thomas Berg, and Peter Siemund (eds.). 2007. Aspects of meaning construction. Amsterdam: John Benjamins. Rainer, Franz. 2013. Morphological metaphysics: virtual, potential and actual words. Word Structure 5, 165–812. Ralli, Angela. 2013. Compounding in Modern Greek. Dordrecht: Springer. Ramscar, Michael, Teenie Matlock, and Melody Dye. 2010. Running down the clock: the role of expectation in our understanding of time and motion. Language and Cognitive Processes 25, 589–615. Rapp, Alexander, Michael Erb, Wolfgang Grodd, Mathias Bartels, and Katja Markert. 2011. Neurological correlates of metonymy resolution. Brain and Language 119(3), 196–205. Rappaport Hovav, Malka, and Beth Levin. 1998. Building verb meaning. In Miriam Butt and Wilhelm Geuder (eds.), The Projection of Arguments, 97–134. Stanford: CSLI. Rauscher, Francis H., Robert M. Krauss, and Yihsiu Chen. 1996. Gesture, speech, and lexical access: the role of lexical movements in speech production. Psychological Science 7(4), 226–31. Raymond, William D., and Esther L. Brown. 2012. Are effects of word frequency effects of context of use? An analysis of initial fricative
783
784
References
reduction in Spanish. In Stefan Th. Gries and Dagmar S. Divjak (eds.), Frequency effects in language learning and processing, 35–52. Berlin: De Gruyter Mouton. Rayson, Paul. 2008. From key words to key semantic domains. International Journal of Corpus Linguistics 13(4), 519–49. Reagan, Timothy G. 2010. Language policy and planning for sign languages. Washington, DC: Gallaudet University Press. Redcay, E., F. Haist, and E. Courchesne. 2008. Functional neuro-imaging of speech during a pivotal period in language acquisition. Developmental Science 11, 237–52. Reddy, Michael J. 1979. The conduit metaphor: a case of frame conflict in our language about language. In Andrew Ortony (ed.). Metaphor and thought, 284–324. Cambridge University Press. Reitter, David, Johanna D. Moore, and Frank Keller. 2006. Priming of syntactic rules in task-oriented dialogue and spontaneous conversation. Proceedings of the 28th Annual Conference of the Cognitive Science Society, 685–90. Renouf, Antoinette. 2001. Lexical signals of word relations. In Mike Scott and Geoff Thompson (eds.), Patterns of text: in honour of Michael Hoey, 35–54. Amsterdam: Benjamins. Reuneker, Alex. 2016. Conditional use of prepositional phrases in Dutch: the case of zonder (‘without’). In Jenny Audring and Sander Lestrade (eds.), Linguistics in the Netherlands 33. 121–34. Amsterdam: John Benjamins. Rhee, Seongha. 2004. From discourse to grammar: grammaticalization and lexicalization of rhetorical questions in Korean. LACUS Forum 30, 413–23. Rhenius, Charles and Theophilus Ewald. 1836. A grammar of the Tamil language. Madras: Church Mission Press. Rice, Sally. 2011. Applied field linguistics: delivering linguistic training to speakers of endangered languages. Language and Education 25, 319–38. 2017. Phraseology and polysynthesis. In Nicholas Evans, Marianne Mithun, and Michael Fortescue (eds.), The Oxford handbook of polysynthesis. Oxford University Press. Rice, Sally, and John Newman (eds.). 2010. Empirical and experimental methods in cognitive/functional research. Stanford: CSLI. Richards, A. 1936. Metaphor, a lecture delivered in 1936. The Philosophy of Rhetoric. New York: Oxford University Press. Richardson, Daniel C., and Teenie Matlock. 2007. The integration of figurative language and static depictions: an eye movement study of fictive motion. Cognition 102(1). 129–38. Richardson, Daniel C., Rick Dale, and Michael J. Spivey. 2007. Eye movements in language and cognition. In Monica Gonzalez-Marquez, Irene Mittelberg, Seana Coulson, and Michael J. Spivey (eds.), Empirical Methods in Cognitive Linguistics, 323–44. Amsterdam: John Benjamins.
References
Richardson, Daniel C., Michael Spivey, Lawrence Barsalou, and Ken McCrae. 2003. Spatial representations activated during real-time comprehension of verbs. Cognitive Science 27, 767–80. Riemer, Nick. 2015. Word meanings. In J. R. Taylor (ed.), The Oxford handbook of the word, 305–19. Oxford University Press. Ritchie, Graeme. 2006. Reinterpretation and viewpoints. Humor 19(3), 251–70. Rizzolatti, Giacomo, Leonardo Fogassi, and Vittorio Gallese. 2001. Neurophysiological mechanisms underlying the understanding and imitation of action. Nature Neuroscience Reviews 2(9), 661–70. Robenalt, Clarice, and Adele E. Goldberg. 2015. Judgment evidence for statistical preemption: it is relatively better to vanish than to disappear a rabbit, but a lifeguard can equally well backstroke or swim children to shore. Cognitive Linguistics 26(3), 467–504. Roberts, Sea´n, and James Winters. 2013. Linguistic diversity and traffic accidents: lessons from statistical studies of cultural traits. PLOSone 8 (8), e70902. Roberts Sea´n G., Francisco Torreira, and Stephen C. Levinson. 2015. The effects of processing and sequence organization on the timing of turn taking: a corpus study. Frontiers in Psychology 6, 509. Robinson, Peter, and Nick C. Ellis (eds.). 2008. Handbook of cognitive linguistics and second language acquisition. London: Routledge. Rocci, Andrea. 2008. Modality and its conversational backgrounds in the reconstruction of argumentation. Argumentation 22(2), 165–89. Rochat, Philippe. 2011. Primordial sense of embodied self-unity. In Virginia Slaughter and Celia A. Brownell (eds.), Early development of body representations, 3–18. Cambridge University Press. Ro¨der, Brigitte, Oliver Stock, Siegfried Bien, Helen Neville, and Frank Ro¨sler. 2002. Speech processing activates visual cortex in congenitally blind humans. European Journal of Neuroscience 16(5), 930–36. Roe, Anna W., Preston E. Garragthy, Manuel Esguerra, and Mriganka Sur. 1993. Experimentally induced visual projections to the auditory thalamus in ferrets: evidence for a W cell pathway. Journal of Computiational. Neuroscience 334, 263–80. Roffler, Suzanne K., and Robert A. Butler. 1968. Localization of tonal stimuli in the vertical plane. Journal of the Acoustical Society of America 43, 1260–65. Rohlfing, Katharina J., Jannik Fritsch, Britta Wrede, and Tanja Jungmann. 2006. How can multimodal cues from child-directed interaction reduce learning complexity in robots? Advanced Robotics 20(10), 1183–99. Rohrer, Tim. 2007. The body in space: dimensions of embodiment. In Tom Ziemke, Jordan Zlatev, and Roslyn Frank (eds.), Body, language and mind, vol. 1, 339–78. Berlin: De Gruyter Mouton. Roitman, Jamie D., Elizabeth M. Brannon, Jessica R. Andrews, and Michael L. Platt. 2007. Nonverbal representation of time and number in adults. Acta Psychologica, 124, 296–318.
785
786
References
Rojo, Ana. 2002. Frame semantics and the translation of humor. Babel: International Journal of Translation 48(1), 34–77. Rojo, Ana and Javier, Valenzuela. 2003. Fictive motion in English and Spanish. International Journal of English Studies 3, 123–150. Romero Lauro, L. J., Giulia Mattavelli, Costanza Papagno, and Marco Tettamanti. 2013. She runs, the road runs, my mind runs, bad blood runs between us: literal and figurative motion verbs: an fMRI study. Neuroimage 83, 361–71. Rosch, Eleanor. 1973. Natural categories. Cognitive Psychology 4, 328–50. 1975. Universals and cultural specifics in human categorization. In Richard W. Brislin, Stephen Bochner, and Walter Josepth Lonner (eds.), Cross-cultural perspectives on learning, 177–206. New York: Wiley. 1978. Principles of categorization. In E. Rosch and B. Lloyd (eds.), Cognition and categorization, 27–48. Hillsdale, NJ: Lawrence Erlbaum. Rosch, Eleanor, and Carolyn B. Mervis. 1975. Family resemblances: studies in the internal structure of categories. Cognitive Psychology 7, 573–605. Rousse-Malpat, Audrey, and Marjolijn Verspoor. In prep. Instruction and assessment from a dynamic usage-based (DUB) perspective. In Andrea Tyler, Lourdes Ortega, Mariko Uno, and Hae In Park (eds.), Usage-inspired L2 instruction: Researched pedagogy. Amsterdam: John Benjamins. Rowland, Caroline F. 2007. Explaining errors in children’s questions: auxiliary DO and modal auxiliaries. Cognition 104, 106–34. Rowland, Caroline F., and Julian Pine. 2000. Subject-auxiliary inversion errors and wh-question acquisition: ‘What children do know?’ Journal of Child Language 27(1), 157–81. Roy, Arundhati. 1997 [2009]. The God of small things. London: Harper Collins. Ruiz de Mendoza Iba´n˜ez, Francisco Jose´. 2008. Cross-linguistic analysis, second language teaching and cognitive semantics: the case of Spanish diminutives and reflexive constructions. In S. De Knop and T. De Rycker (eds.), Cognitive approaches to pedagogical grammar: A volume in honor of Rene Dirven, 121–55. Berlin: Mouton de Gruyter. 1998. On the nature of blending as a cognitive phenomenon. Journal of Pragmatics 30, 259–74. 2000. The role of mappings and domains in understanding metonymy. In Antonio Barcelona (ed.). Metaphor and metonymy at the crossroads, 109–32. Berlin: De Gruyter Mouton. Ruiz de Mendoza Iba´n˜ez, Francisco, and Jose´ Luis Otal Campo. 2002. Metonymy, grammar and communication. Granada: Comares Ruiz de Mendoza Iba´n˜ez, Francisco, and Olga Diez Velasco. 2003. Patterns of conceptual interaction, In Rene´ Dirven and Ralf Po¨rings (eds.), Metaphor and metonymy in comparison and contrast, 489–532. Berlin: De Gruyter Mouton. Ruiz de Mendoza Iba´n˜ez, Francisco, and Lorena Pe´rez Herna´ndez. 2003. Cognitive operations and pragmatic implication, In Klaus-Uwe
References
Panther and Linda Thornburg (eds.), Metonymy and pragmatic inferencing, 23–50. Amsterdam: John Benjamins. Ruiz de Mendoza Iba´n˜ez, Francisco, and Ricardo Mairal Uson. 2007. High level metaphor and metonymy in meaning construction. In Gu¨nther Radden, Klaus-Michael Ko¨pcke, Thomas Berg, and Peter Siemund (eds.), Aspects of meaning construction, 33–49. Amsterdam: John Benjamins. Rundblad, Gabriella, and Dagmara Annaz. 2010. Metaphor and metonymy comprehension: receptive vocabulary and conceptual knowledge. British Journal of Developmental Psychology 28, 547–63. Ruppenhofer, Josef, Hans C. Boas, and Collin Baker. 2013. The FrameNet approach to relating syntax and semantics. In R. H. Gouws, U. Heid, W. Schweickhard, and H. E. Wiegand (eds.), Dictionaries: an international encyclopedia of lexicography, 1320–29. Berlin: Mouton. Ruppenhofer, Josef, and Laura A. Michaelis. 2010. A constructional account of genre-based argument omissions. Constructions and Frames 2, 158–84. Ruppenhofer, Josef, Michael Ellsworth, Miriam R. L. Petruck, Christopher R. Johnson, and Jan Scheffczyk. 2016. FrameNet II: extended theory and practice. Berkeley: International Computer Science Institute. Russo, Tommaso. 2005. A crosslinguistic, cross-cultural analysis of metaphors in two Italian Sign Language (LIS) Registers. Sign Language Studies 5(3), 333–59. Sacks, Harvey, Emanuel A. Schegloff, and Gail Jefferson. 1974. A simplest systematics for the organization of turn taking for conversation. Language 50(4), 696–735. Sag, Ivan A. 1997. English relative clause constructions. Journal of Linguistics 33, 431–84. 2010. English filler-gap constructions. Language 86(3), 486–545. 2012. Sign-based Construction Grammar: an informal synopsis. In H. C. Boas and I. A. Sag (eds.), Sign-based Construction Grammar, 69–202. Stanford: CSLI. Sag, Ivan A., Hans C. Boas, and Paul Kay. 2012. Introducing sign-based construction grammar. In Hans C. Boas and Ivan A. Sag (eds.), Signbased construction grammar, 1–30. Stanford: CSLI. Sag, Ivan A., Thomas Wasow, and Emily M. Bender. 2003. Syntactic theory: a formal introduction. Stanford: CSLI. Sagerer Gerhard, Hans-Ju¨rgen Eikmeyer, and Gert Rickheit. 1994. ‘Wir bauen jetzt ein Flugzeug’Ł Konstruieren im Dialog. Arbeitsmaterialien. Technical Report SFB 360, Bielefeld University. Sahin, N. T., S. Pinker, S. S. Cash, D. Schomer, and E. Halgren 2009. Sequential processing of lexical, grammatical, and phonological information within Broca’s area. Science 326(5951), 445–49. Saloma˜o, Maria, Margarida Martins, Tiago Timponi Torrent, and Thais Fernandes Sampaio. 2013. A linguı´stica de corpus encontra a linguı´stica computacional: notı´cias do projeto FrameNet Brasil. Cadernos de Estudos Linguı´sticos 55(1), 7–34.
787
788
References
Sampson, Geoffrey. 2015. Writing systems. Sheffield: Equinox. Sanders, Jose´. 2010. Intertwined voices: journalists’ modes of representing source information in journalistic subgenres. English Text Construction 3 (2), 226–49. Sanders, Jose´, and Gisela Redeker. 1996. Perspective and the representation of speech and thought in narrative discourse. In Gilles Fauconnier and Eve Sweetser (eds.), Spaces, worlds, and grammar, 290–317. University of Chicago Press. Sanders, Ted J. M., Jose´ Sanders, and Eve Sweetser. 2009. Causality, cognition and communication: a mental space analysis of subjectivity in causal connectives. In Ted Sanders and Eve Sweetser (eds.), Causal categories in discourse and cognition, 19–60. Berlin: De Gruyter Mouton. Sandler, Wendy. 1999. A prosodic model of sign language phonology. Phonology 16(3), 443–47. 2009. Symbiotic symbolization by hand and mouth in sign language. Semiotica: Journal of the International Association for Semiotic Studies/Revue de l’Association Internationale de Se´miotique 174, 241–75. Sandra, Dominiek, and Sally Rice. 1995. Network analyses of prepositional meaning: mirroring whose mind – the linguist’s or the language user’s? Cognitive Linguistics 6(1), 89–130. Sanford, Anthony, and Catherine Emmott. 2012. Mind, brain, and narrative. Cambridge University Press. Santiago, Julio, Juan Lupia´n˜ez, Elvira Pe´rez, and Marı´a J. Funes. 2007. Time (also) flies from left to right. Psychonomic Bulletin and Review 14(3), 512–16. Sapir, Edward. 1972. The psychological reality of phonemes [La re´alite´ psychologique des phone`mes]. In Valerie Becker Makkai (ed.), Phonological theory: evolution and current practice, 22–31. New York: Holt Rinehart and Winston. Trans. in and reprinted from the Journal de psychologie normale et pathologique 30, 247–65. 1921. Language: An introduction to the study of speech. New York: Harcourt. Sarvasy, Hannah. 2014. A grammar of Nungon: a Papuan language of the Morobe Province, Papua New Guinea. PhD dissertation. James Cook University, Cairns. Sasse, Hans-Ju¨rgen. 1993. Syntactic categories and subcategories. In Joachim Jacobs, Arnim von Stechow, Wolfgang Sternefeld, and Theo Vennemann (eds.), Syntax. Ein internationales Handbuch zeitgeno¨ssischer Forschung, 646–86. Berlin: Mouton de Gruyter. Sato, Hiroaki. 2008. New functions of FrameSQL for multilingual FrameNets. Proceedings of the Sixth International Language Resources and Evaluation Conference (LREC), 758–62, Marrakech, Morocco. Saur, Dorothee, Ru¨diger Lange, Annette Baumgaertner, Valeska Schraknepper, Klaus Willmes, Michel Rijntjes, and Cornelius Weiller. 2006. Dynamics of language reorganization after stroke. Brain 129, 1371–84.
References
Savage, Ceri, Elena Lieven, Anna Theakston, and Michael Tomasello. 2006. Structural priming as learning in language acquisition: the persistence of lexical and structural priming in 4-year-olds. Language Learning and Development 2, 27–49. Saygin, Ayse P., Stephen McCullough, Morana Alac, and Karen Emmorey. 2010. Modulation of BOLD response in motion-sensitive lateral temporal cortex by real and fictive motion sentences. Journal of Cognitive Neuroscience 22(11), 2480–90. Scarry, Elaine. 1987. The body in pain: the making and unmaking of the world. Oxford University Press. Schane, Sanford A. 1968. French phonology and morphology. Cambridge, MA: MIT Press. Schank, Roger, and Robert Abelson. 1977. Scripts, plans, goals and understanding: an inquiry into human knowledge structures. Hillsdale: Lawrence Erlbaum. Schegloff, Emanuel A. 1968. Sequencing in conversational openings. American Anthropologist 70(6), 1075–95. 1972. Notes on a conversational practice: formulating place. In David N. Sudnow (ed.), Studies in social interaction, 75–119. New York: MacMillan, Free Press. 1984. On some gestures’ relation to talk. In J. Maxwell Atkinson and John Heritage (eds.), Structures of social action, 266–98. Cambridge University Press. 2007. Sequence organization in interaction: a primer in conversation analysis, volume 1. Cambridge University Press. Schegloff, Emanuel A., Gail Jefferson, and Harvey Sacks. 1977. The preference for self-correction in the organization of repair in conversation. Language 53(2), 361–82. Schembri, Adam, and Trevor Johnston. 2012. Sociolinguistic aspects of variation and change. In Roland Pfau, Martin Steinbach, and Bencie Woll (eds.), Sign language: an international handbook. Berlin: De Gruyter Mouton. Schilder, Frank. 2001. Presupposition triggered by temporal connectives. In Miriam Bras and Laure Vieu (eds.), Semantic and pragmatic issues in discourse and dialogue: experimenting with current dynamic theories, 85–108. Amsterdam: Elsevier. Schilling-Estes, Natalie. 2002. Investigating stylistic variation. In Jack K. Chambers, Peter Trudgill, and Natalie Schilling-Estes (eds.), The handbook of language variation and change, 375–401. Malden, MA: Blackwell. Schmid, Hans-Jo¨rg. 1993. Cottage and co., idea, start vs. begin. Tu¨bingen: Max Niemeyer. Schmid, Hans-Jo¨rg, and Helmut Ku¨chenhoff. 2013. Collostructional analysis and other ways of measuring lexicogrammatical attraction: theoretical premises, practical problems and cognitive underpinnings. Cognitive Linguistics 24, 531–77.
789
790
References
Schmidt, Thomas. 2009. The Kicktionary – a multilingual lexical resource of football language. In Hans C. Boas (ed.), Multilingual FrameNets: methods and applications, 101–34. Berlin: De Gruyter Mouton. Schmitt, Reinhold. 2005. Zur multimodalen Struktur von turn-taking. Gespra¨chsforschung 6, 17–61. Schober, Michael F. 1995. Speakers, addressees, and frames of reference: whose effort is minimized in conversations about location? Discourse Processes 20(2), 219–47. Scho¨nefeld, Doris. 2006. From conceptualization to linguistic expression: where languages diversify. In Stefan Th. Gries and Anatol Stefanowitsch (eds.), Corpora in cognitive linguistics: the syntax-lexis interface, 297–344. Berlin: De Gruyter Mouton. (ed.). 2011. Converging evidence: methodological and theoretical issues for linguistic research. Amsterdam: John Benjamins. Schoonjans, Steven. 2013a. The distribution of downtoning gestures: a pilot study. TiGeR 2013, 1–4. 2013b. Is gesture subject to grammaticalization? Studies van de BKL 8, 30–43. 2014a. Modalpartikeln als multimodale Konstruktionen: eine korpusbasierte Kookkurrenzanalyse von Modalpartikeln und Gestik im Deutschen. PhD dissertation. University of Leuven. 2014b. Gestische Modalpartikeln oder Modalpartikelgesten? Zur Kookkurrenz von Modalpartikeln und Gestikmustern im Deutschen. Bavarian Working Papers in Linguistics 3, 91–106. Schoonjans, Steven, and Elisabeth Zima. 2014. The frequency issue in multimodal construction grammar. ICCG 8, Osnabru¨ck, 3–6 September 2014, n. pag. Schoonjans, Steven, Geert Broˆne, and Kurt Feyaerts. 2015. Multimodalita¨t in der Konstruktionsgrammatik: eine kritische Betrachtung illustriert anhand einer Gestikanalyse der Partikel einfach. In Jo¨rg Bu¨cker, Susanne Gu¨nthner, and Wolfgang Imo (eds.), Konstruktionsgrammatik V: Konstruktionen im Spannungsfeld von sequenziellen Mustern, kommunikativen Gattungen und Textsorten, 291–308. Tu¨bingen: Stauffenburg. Schoonjans, Steven, Paul Sambre, Geert Broˆne, and Kurt Feyaerts. 2016. Vers une analyse multimodale du sens: perspectives constructionnelles sur la gestualite´ co-grammaticale. Langages 201, 33–49. Schott, G. D. 2004. Communicating the experience of pain: the role of analogy. Pain 108(3), 209–12. Schreier, Daniel. 2013. Collecting ethnographic and sociolinguistic data. In Manfred Krug and Julia Schlu¨ter (eds.), Research methods in language variation and change, 17–35. Cambridge University Press. Schriver, Karen. 1997. Dynamics in document design. New York: John Wiley. Schutz, Alfred. 1966. Collected papers III: studies in phenomenological philosophy. The Hague: Martinus Nijhoff.
References
Schwarz, Wolf, and Inge M. Keus. 2004. Moving the eyes along the mental number line: comparing SNARC effects with saccadic and manual responses. Perception and Psychophysics 66(4), 651–64. Scott, Amanda. 1989. The vertical dimension and time in Mandarin. Australian Journal of Linguistics 9(2), 295–314. Seabright, Paul. 2010. The company of strangers: a natural history of economic life. Princeton University Press. Searle, John. 1980. The background of meaning. In John Searle, Ferenc Kiefer, and Manfred Bierwisch (eds.), Speech-act theory and pragmatics, 221–32. Dordrecht: Reidel. Selkirk, Elisabeth O. 1984. Phonology and syntax: the relation between sound and structure: current studies in linguistics series. Cambridge, MA: MIT Press. Sell, Andrea J., and Michael P. Kaschak. 2011. Processing time shifts affects the execution of motor responses. Brain and Language 117(1), 39–44. Selting, Margret. 1996. On the interplay of syntax and prosody in the constitution of turn-constructional units and turns in conversation. Pragmatics 6(3), 357–88. Semino, Elena. 2008. Metaphor in discourse. Cambridge University Press. 2010. Descriptions of pain, metaphor and embodied simulation. Metaphor and Symbol 25(4), 205–26. In prep. A corpus-based study of ‘mixed metaphor’ as a metalinguistic comment. In Raymond W. Gibbs, Jr. (ed.), Mixing metaphor. Amsterdam: John Benjamins. Semino, Elena, and Jonathan Culpeper (eds.). 2002. Cognitive stylistics: language and cognition in text analysis. Amsterdam: John Benjamins. Semino, Elena, and Zso´fia Demje´n (eds.). 2017. The Routledge handbook of metaphor and language. London: Routledge. Semino, Elena, Zso´fia Demje´n, Jane Demmen, Veronika Koller, Sheila Payne, et al. 2015. The online use of Violence and Journey metaphors by patients with cancer, as compared with health professionals: a mixed methods study. BMJ Supportive and Palliative Care online 1–7. Seyfeddinipur, Mandana, and Sotaro Kita. 2001. Gestures and selfmonitoring in speech production. Proceedings of the 27th Annual Meeting of the Berkeley Linguistics Society, 457–64. Shaffer, Barbara, Maria Josep Jarque, and Sherman Wilcox. 2011. The expression of modality: conversational data from two signed languages. In M. T. Nogueira and M. F. V. Lopes (eds.), Modo e modalidade: grama´tica, discurso e interac¸a˜o, 11–39. Fortaleza: Edic¸o˜es UFC. Shafto, Meredith A., and Lorraine K. Tyler. 2014. Language in the aging brain: the network dynamics of cognitive decline and preservation. Science 346(6209), 583–87. Shank, Christopher, Koen Plevoets, and Julie Van Bogaert. 2016. A multifactorial analysis of that/zero alternation: the diachronic
791
792
References
development of the zero complementizer with think, guess and understand. In Jiyoung Yoon and Stefan Th. Gries (eds.), Corpus-based approaches to Construction Grammar, 201–40. Amsterdam and Philadelphia: John Benjamins. Sharifian, Farzad, and Gary Palmer (eds.). 2007. Applied cultural linguistics: implications for second language learning and intercultural communication. Amsterdam: John Benjamins. Shayan, Shakila, Ozge Ozturk, and Mark A. Sicoli. 2011. The thickness of pitch: crossmodal metaphors in Farsi, Turkish, and Zapotec. The Senses and Society 6(1), 96–105. Sherzer, Joel. 1972. Verbal and nonverbal deixis: the pointed lip gesture among the San Blas Cuna. Language in Society 2, 117–31. Shockley, Kevin, Daniel C. Richardson, and Rick Dale. 2009. Conversation and coordinative structures. Topics in Cognitive Science 1(2), 305–19. Shutova, Ekaterina. 2010. Models of metaphor in NLP. Proceedings of the 48th Annual Meeting of the Association for Computational Linguistics (July), 688–97. Shutova, Ekaterina, Lin Sun, and Anna Korhonen. 2010. Metaphor identification using verb and noun clustering. Proceedings of the 23rd International Conference on Computational Linguistics, 1002–10. Shutova, Ekaterina, Simone Teufel, and Anna Korhonen. 2012. Statistical metaphor processing. Proceedings of the 48th Annual Meeting of the Association for Computational Linguistics (July), 1–92. Siemund, Peter. 2002. Animate pronouns for inanimate objects: pronominal gender in English regional varieties. In Dieter Kastovsky, Gunther Kaltenbo¨ck, and Susanne Reichl (eds.), Anglistentag 2001 Vienna: Proceedings, 19–34. Trier: Wissenschaftlicher Verlag. 2008. Pronominal gender in English: a study of English varieties from a crosslinguistic perspective. London: Routledge. Siewierska, Anna, and Willem B. Hollmann. 2007. Ditransitive clauses in English with special reference to Lancashire dialect. In Mike Hannay and Gerard J. Steen (eds.), Structural-functional studies in English grammar, 83–102. Amsterdam: John Benjamins. Simmons, William K., Vimal Ramjee, Michael S. Beauchamp, K. McRae, A. Martin, and Lawrence W. Barsalou. 2007. A common neural substrate for perceiving and knowing about color. Neuropsychologia 45, 2802–10. Simo´, Judit. 2011. Metaphors of blood in American English and Hungarian: a cross-linguistic corpus investigation. Journal of Pragmatics 43(12), 2897–910. Simone, Raffaele (ed.) 1995. Iconicity in language. Amsterdam: John Benjamins. Sinclair, John M. 1991. Corpus, concordance, collocation. Oxford University Press. 2004. Trust the text: language, corpus and discourse. London: Routledge.
References
Singer, Melissa A., and Susan Goldin-Meadow. 2005. Children learn when their teachers’ gestures and speech differ. Psychological Science 16, 85–89. Sinha, Chris. 1999. Grounding, mapping, and acts of meaning. In Theo Janssen and Gisela Redeker (eds.), Cognitive linguistics: foundations, scope and methodology, 223–55. Berlin: De Gruyter Mouton. 2004. The evolution of language: from signals to symbols to system. In D. Kimbrough Oller and Ulrike Griebel (eds.), Evolution of communication systems: a comparative approach, 217–35. Cambridge, MA: MIT Press. 2007. Cognitive linguistics, psychology and cognitive science. In Dirk Geeraerts and Hubert Cuyckens (eds.), Handbook of cognitive linguistics, 1266–94. Oxford University Press. 2009. Language as a biocultural niche and social institution. In Vyvyan Evans and Ste´phanie Pourcel (eds.), New directions in cognitive linguistics, 289–309. Amsterdam: John Benjamins. Sinha, Chris, Vera da Silva Sinha, Jo¨rg Zinken, and Wany Sampaio. 2011. When time is not space: the social and linguistic construction of time intervals and temporal event relations in an Amazonian culture. Language and Cognition 3(1), 137–69. Skipper, Jeremy I., Susan Goldin-Meadow, Howard C. Nusbaum, and Steven L. Small. 2009. Gestures orchestrate brain networks for language understanding. Current Biology 19(8), 661–67. Skorczynska, Hanna. 2010. A corpus-based evaluation of metaphors in a business English textbook. English for Specific Purposes 29(1), 30–42. Skorczynska, Hanna, and Alice Deignan. 2006. A comparison of metaphor vehicles and functions in scientific and popular business corpora. Metaphor and Symbol 21(2), 87–104. Slabakova, Roumyana, Jennifer Amaro, and Sang-Kyun Kang. 2013. Regular and novel metonymy in native Korean, Spanish and English: experimental evidence for various acceptability. Metaphor and Symbol 28, 275–93. Slobin, Dan. 1996. Two ways to travel: verbs of motion in English and Spanish. In Masayoshi Shibatani and Sandra Thompson (eds.), Grammatical constructions: their form and meaning, 195–220. Oxford University Press. 2004. The many ways to search for a frog: linguistic typology and the expression of motion events. In Sven Stro¨mqvist and Ludo Verhoeven (eds.), Relating events in narrative: vol. 2, Typological and contextual perspectives, 219–57. Mahwah, NJ: Lawrence Erlbaum. Smith, Neil. 2010. Acquiring phonology: a cross-generational case-study. Cambridge University Press. Snider, Neal, and Inbal Arnon. 2012. A unified lexicon and grammar? Compositional and non-compositional phrases in the lexicon. In Dagmar S. Divjak and Stefan Th. Gries (eds.), Frequency effects in language representation, 127–63. Berlin: De Gruyter Mouton.
793
794
References
Sokolova, Svetlana, Olga Lyashevskaya, and Laura A. Janda. 2012. The Locative Alternation and the Russian ‘empty’ prefixes: a case study of the verb gruzit’ ‘load’. In Dagmar S. Divjak and Stefan Th. Gries (eds.), Frequency effects in language representation, 51–85. Berlin and New York: De Gruyter Mouton. Sontag, Susan. 1979. Illness as metaphor. New York: Farrar, Straus and Giroux. Soriano, Cristina, and Javier Valenzuela 2009. Are conceptual metaphors accessible on-line? A psycholinguistic exploration of the CONTROL IS UP metaphor. In Javier Valenzuela, Ana Rojo, and Cristina Soriano (eds.), Trends in cognitive linguistics, 29–49. Peter Lang: Frankfurt. Sotirova, Violeta. 2004. Connectives in free indirect style: continuity or shift? Language and Literature 13(3), 216–34. Soukup, Barbara. 2013. Austrian dialect as a metonymic device: a cognitive sociolinguistic investigation of Speaker Design and its perceptual implications. Journal of Pragmatics 52, 73–82. Spencer, Andrew. 2013. Lexical relatedness: a paradigm-based model. Oxford University Press. Spivey, Michael J., and Joy Geng. 2001. Oculomotor mechanisms activated by imagery and memory: eye movements to absent objects. Psychological Research 65, 235–41. Spivey, Michael J., Daniel C. Richardson, and Monica Gonzalez-Marquez. 2005. On the perceptual- motor and image-schematic infrastructure of language. In Diane Pecher and Rolf A. Zwaan (eds.), Grounding cognition: the role of perception and action in memory, language, and thinking, 246–81. Cambridge University Press. Spreckels, Janet. 2009. Ich hab einfach gedacht – Stellungnahme und Positionierung durch einfach in Erkla¨rinteraktionen. In Susanne Gu¨nthner and Jo¨rg Bu¨cker (eds.), Grammatik im Gespra¨ch: Konstruktionen der Selbst- und Fremdpositionierung, 115–46. Berlin: Walter de Gruyter. Srinivasan, Mahesh, and Susan Carey. 2010. The long and the short of it: on the nature and origin of functional overlap between representations of space and time. Cognition 116(2), 217–41. Spronck, Stef. 2016. Evidential fictive interaction (in Ungarinyin and Russian). In Esther Pascual and Sergeiy Sandler (eds.), The conversation frame: forms and functions of fictive interaction, 255–75. Amsterdam: John Benjamins. Stampe, David. 1987. On phonological representation. In Wolfgang U. Dressler, Hans C. Luschu¨tzky, Oskar Pfeiffer, and John R. Rennison (eds.), Phonologica 1984, 287–300. Cambridge University Press. Stanfield, Robert A., and Rolf A. Zwaan. 2001. The effect of implied orientation derived from verbal context on picture recognition. Psychological Science 12(2), 153–56.
References
Stanford, W. Gregory Jr., and Stephen Webster. 1996. A nonverbal signal in voices of interview partners effectively predicts communication accommodation and social status perceptions. Journal of Personality and Social Psychology 70(6), 1231–40. Stec, Kashmiri. 2012. Meaningful shifts: a review of viewpoint markers in co-speech gesture and sign language. Gesture 12(3), 327–60. Stec, Kashmiri, and Eve Sweetser. 2013. Borobudur and Chartres: religious spaces as performative real-space blends. In Rosario Caballero and Javier E. Diaz Vera (eds.), Sensuous cognition, 265–91. Berlin: De Gruyter Mouton. Steels, Luc. 2011a. Introducing fluid construction grammar. In Luc Steels (ed.), Design patterns in fluid construction grammar, 3–30. Amsterdam: John Benjamins. 2011b. A design pattern for phrasal constructions. In Luc Steels (ed.), Design patterns in fluid construction grammar, 71–114. Amsterdam: John Benjamins. (ed.). 2011c. Design patterns in fluid construction grammar. Amsterdam: John Benjamins. 2013. Fluid construction grammar. In Thomas Hoffmann and Graeme Trousdale (eds.), The Oxford handbook of construction grammar, 152–67. Oxford University Press. Steels, Luc, and Joachim De Beule. 2006. Unify and merge in fluid construction grammar. In Paul Vogt, Yuuga Sugita, Elio Tuci, and Chrystopher Nehaniv (eds.), Symbol grounding and beyond. Proceedings of the third international workshop on the emergence and evolution of linguistic communication, LNAI 4211. Berlin: Springer, 197–223. Steen, Francis, and Mark Turner. 2013. Multimodal construction grammar. In Michael Borkent, Barbara Dancygier, and Jennifer Hinnell (eds.), Language and the creative mind, 255–274. Stanford: CSLI. Steen, Gerard J. 2007. Finding metaphor in grammar and usage: a methodological analysis of theory and research. Amsterdam: John Benjamins. Steen, Gerard J., Aletta G. Dorst, J. Berenike Herrmann, Anna A. Kaal, Tina Krennmayr, and Trijntije Pasma. 2010. A method for linguistic metaphor identification: from MIP to MIPVU. Amsterdam: John Benjamins. Stefanowitsch, Anatol. 2004. HAPPINESS in English and German: a metaphorical-pattern analysis. In Michel Achard and Suzanne Kemmer (eds.), Language, culture, and mind, 137–49. Stanford: CSLI. 2006a. Corpus-based approaches to metaphor and metonymy. In Anatol Stefanowitsch and Stephan Th. Gries (eds.), Corpus-based approaches to metaphor and metonymy, 1–16. Berlin: De Gruyter Mouton. 2006b. Words and their metaphors: a corpus-based approach. In Anatol Stefanowitsch and Stephan Th. Gries (eds.), Corpus-based approaches to metaphor and metonymy, 63–105. Berlin: De Gruyter Mouton. 2006c. Distinctive collexeme analysis and diachrony: a comment. Corpus Linguistics and Linguistic Theory 2(2), 257–62.
795
796
References
2010. Cognitive linguistics meets the corpus. In Mario Brdar, Stephan Th. Gries and Milena Zˇic Fuchs (eds.), Cognitive lingusitics: convergence and expansion, 257–90. Amsterdam: John Benjamins. 2011. Constructional preemption by contextual mismatch: a corpus-linguistic investigation. Cognitive Linguistics 22, 107–29. 2013. Collostructional analysis. In Thomas Hoffmann and Graeme Trousdale (eds.), The Oxford handbook of construction grammar, 290–306. Oxford University Press. Stefanowitsch, Anatol, and Stefan Th. Gries. 2003. Collostructions: investigating the interaction between words and constructions. International Journal of Corpus Linguistics 8(2), 209–43. 2005. Covarying collexemes. Corpus Linguistics and Linguistic Theory 1(1), 1–43. (eds.). 2006. Corpus-based approaches to metaphor and metonymy. Berlin: De Gruyter Mouton. 2008. Channel and constructional meaning: a collostructional case study. In Gitte Kristiansen and Rene´ Dirven (eds.), Cognitive sociolinguistics: language variation, cultural models, social systems, 129–52. Berlin: De Gruyter Mouton. Steinkrauss, Rasmus. 2011. The interaction of function and input frequency in L1-acquisition: the case of was . . . fu¨r ‘what kind of . . .’ questions in German. In Doris Scho¨nefeld (ed.), Converging evidence: methodological and theoretical issues for linguistic research, 249–71. Amsterdam: John Benjamins. Stern, Daniel N. 2000. The interpersonal world of the infant: a view from psychoanalysis and developmental psychology. New York: Basic Books. Stevens, Kenneth. 1989. On the quantal nature of speech. Journal of Phonetics 17(1–2), 3–46. 1998. Acoustic phonetics. Cambridge, MA: MIT Press. Stickles, Elise, Oana David, and Eve Sweetser. 2014. Frame role type constraints and frame metonymy in metaphoric interpretation. In Proceedings of the 11th High Desert Linguistics Society Conference, 317– 45. University of New Mexico. Stickles, Elise, Ellen Dodge, Oana David, and Jisup Hong. 2016. Formalizing contemporary conceptual metaphor theory: a structured repository for metaphor analysis. Constructions and Frames 8(2), 166–213. Stiles, Joan. 2008. The fundamentals of brain development: Integrating nature and nurture. Cambridge, MA: Harvard University Press. Stiles, Joan, and Terry L. Jernigan. 2010. The basics of brain development. Neuropsychology review, 20(4), 327–48. Stiles, Joan, C. Stern, D. Trauner, and R. Nass. 1996. Developmental change in spatial grouping activity among children with early focal brain injury: evidence from a modeling task. Brain and Cognition 31, 46–62. Stiles, Joan, Judy Reilly, Brianna Paul, and Pamela Moses. 2005. Cognitive development following early brain injury: evidence for neural adaptation. Trends in Cognitive Sciences 9(3), 136–43.
References
¨ zc¸alıs¸kan. 2013. Children’s developing underStites, Lauren, and S¸eyda O standing of different spatial metaphors for time. Journal of Child Language 40(5), 1123–37. Stocker, Kurt, Matthias Hartmann, Corinna S. Martarelli, and Fred W. Mast. 2015. Eye movements reveal mental looking through time. Cognitive Science 40(7) n.pag. Stockwell, Peter. 2002. Cognitive poetics: an introduction. London: Routledge. Stockwell, Peter, and Sarah Whiteley (eds). 2014. The Cambridge handbook of stylistics. Cambridge University Press. Stokoe, William C. 1960. Sign language structure. Department of Anthropology and Linguistics, University of Buffalo. 1974. Motor signs as the first form of language. In Roger W. Wescott and Gordon W. Hewes (eds.), Language origins, 35–50. Silver Spring, MD: Linstok Press. Stokoe, William C., Dorothy Casterline, and Carl Croneberg. 1965. A dictionary of American Sign Language on linguistic principles. Washington, DC: Gallaudet College Press. Stosic, Dejan, and Laure Sarda. 2009. The many ways to be located in French and Serbian: the role of fictive motion in the expression of static location. In M. Brala-Vukanovic´ and L. Gruic´-Grmusˇa (eds.), Space and time in language and literature, 39–59. Newcastle: Cambridge Scholars. Streeck, Ju¨rgen. 2002. Grammars, words, and embodied meanings. Journal of Communication 52(3), 581–96. 2009. Gesturecraft: the manufacture of meaning. Amsterdam: John Benjamins. Street, James A., and Ewa Da˛browska. 2010. More individual differences in language attainment: how much do adult native speakers of English know about passives and quantifiers? Lingua 120(8), 2080–94. Strobl, Carolin, Gerhard Tutz, and James Malley. 2009. An introduction to recursive partitioning: rationale, application, and characteristics of classification and regression trees, bagging, and random forests. Psychological Methods 14, 323–48. Stump, Gregory S., and Raphael Finkel. 2013. Morphological typology: from word to paradigm. Cambridge University Press. Sturtevant, Edgar H. 1947. An introduction to linguistic science. New Haven, NJ: Yale University Press. Subirats, Carlos. 2009. Spanish FrameNet: A frame-semantic analysis of the Spanish lexicon. In Hans C. Boas (ed.), Multilingual FrameNets: methods and applications, 135–62. Berlin: De Gruyter Mouton. Subirats, Carlos, and Miriam Petruck. 2003. Surprise: Spanish FrameNet! Proceedings of CIL 17. Prague: Matfyzpress. Sullivan, Karen S. 2006a. Frame-based constraints on lexical choice in metaphor. Proceedings of the 32nd Annual Meeting of the Berkeley Linguistics Society, 387–400, Berkeley, CA.
797
798
References
2006b. How does art ‘speak’ and what does it ‘say’: conceptual metaphor theory as a tool for understanding the artistic process. In David E. Boyes and Frances B. Cogan (eds.), Thought tools for a new generation: essays on thought, ideas, and the power of expression, 81–89. Eugene: Robert D. Clark Honors College. 2007. Grammar in metaphor: a Construction Grammar account of metaphoric language. PhD dissertation. University of California Berkeley. 2009. The languages of art: how representational and abstract painters conceptualize their work in terms of language. Poetics Today 30(3), 517–60. 2013a. Frames and constructions in metaphoric language. Amsterdam: John Benjamins. 2013b. A tangled knot of target domains: assessing I N T E L L I G E N C E I S B R I G H T N E S S and G O O D N E S S I S B R I G H T N E S S in an image rating task. Metaphorik.de 23, 7–18. 2015. Judging a book by its cover (and its background). Visual Communication 14(1), 3–14. 2016. Silent abstractions versus ‘Look at me’ drawings: corpus evidence that artworks’ subject matter affects their fictive speech. In Esther Pascual and Sergeiy Sandler (eds.), The conversation frame: forms and functions of fictive interaction, 87–109. Amsterdam: John Benjamins. Sullivan, Karen S., and Eve Sweetser. 2009. Is ‘Generic is Specific’ a metaphor? In Fey Parrill, Vera Tobin, and Mark Turner (eds.), Meaning, form, and body, 309–28. Stanford: CSLI. Summa, Michaela. 2012. Body memory and the genesis of meaning. In Sabine C. Koch, Thomas Fuchs, Michaela Summa, and Cornelia Mu¨ller (eds.), Body memory, metaphor and movement, 23–42. Amsterdam: Benjamins. Supalla, Ted. 1978. Morphology of verbs of motion and location in American Sign Language. In Frank Caccamise (ed.), American Sign Language in a bilingual, bicultural context. Proceedings of the National Symposium on Sign Language Research and Teaching, 27–45. Silver Spring, MD: National Association of the Deaf. Svorou, Soteria. 1994. The grammar of space. Amsterdam: John Benjamins. Swann, Joan. 2006. The art of the everyday. In Janet Maybin and Joan Swann (eds.), The art of English: everyday creativity, 3–53. Basingstoke: Palgrave Macmillan. Sweetser, Eve. 1990. From etymology to pragmatics: metaphorical and cultural aspects of semantic structure. Cambridge University Press. 1999. Compositionality and blending: semantic composition in a cognitively realistic framework. In Theo Janssen and Gisela Redeker (eds.), Cognitive linguistics: foundations, scope, and methodology, 129–62. Berlin: De Gruyter Mouton. 2004. ‘The suburbs of your good pleasure’: cognition, culture and the bases of metaphoric structure. In Graham Bradshaw, Tom Bishop and
References
Mark Turner (eds.), The Shakespearean International Yearbook 4. Aldershot: Ashgate. 2007. Looking at space to study mental spaces: co-speech gestures as a crucial data source in cognitive linguistics. In Monica GonzalezMarquez, Irene Mittelberg, Seana Coulson, and Michael J. Spivey (eds.), Methods in Cognitive Linguistics, 203–26. Amsterdam: John Benjamins. 2012. Introduction: viewpoint and perspective in language and gesture, from the ground down. In Barbara Dancygier and Eve Sweetser (eds.), Viewpoint in Language, 1–22. Cambridge University Press. 2013. Creativity across modalities in viewpoint construction. In Mike Borkent, Barbara Dancygier, and Jennifer Hinnell (eds.), Language and the creative mind, 239–54. Stanford: CSLI. Sweetser, Eve, and Kashmiri Stec. 2016. Maintaining multiple viewpoints with gaze. In Barbara Dancygier, Wei-Lun Lu, and Arie Verhagen (eds.), Viewpoint and the fabric of meaning: form and use of viewpoint tools across languages and modalities, 237–57. Berlin: De Gruyter Mouton. Sweetser, Eve, and Karen Sullivan. 2012. Minimalist metaphors. English Text Construction 5(2), 153–73. Szczepek Reed, Beatrice. 2010. Prosody and alignment: a sequential perspective. Cultural Studies of Science Education 5, 859–67. Szmrecsanyi, Benedikt. 2005. Language users as creatures of habit: a corpus-linguistic analysis of persistence in spoken English. Corpus Linguistics and Linguistic Theory 1, 113–50. 2006. Morphosyntactic persistence in spoken English. Berlin: De Gruyter Mouton. Szymanek, Bogdan. 2010. A panorama of Polish word-formation. Lublin: Wydawnictwo KUL. Tabacaru, Sabina. 2014. Humorous implications and meanings: a multimodal approach to sarcasm in interactional humor. PhD dissertation. Universite´ Lille III/KU Leuven. Tagliamonte, Sali A. 2006. Analysing sociolinguistic variation. Cambridge University Press. Takahashi, Kiyoko. 2001. Access path expressions in Thai. In Alan Cienki, Barbara Luka, and Michael Smith (eds.), Conceptual structure and discourse factors in linguistic structure, 237–52. Stanford: CSLI. Talmy, Leonard. 1972. Semantic structure in English and Atsugewi. PhD dissertation. University of California at Berkeley. 1978a. Figure and ground in complex sentences. In Joseph H. Greenberg (ed.), Universals of human language, vol. 4: syntax, 625–49. Stanford University Press. 1978b. The relation of grammar to cognition: a synopsis. Proceedings of the 1978 workshop on theoretical issues in natural language processing, 14–24. Association for Computational Linguistics.
799
800
References
1983. How language structures space. In Herbert Pick and Linda Acredolo (eds.), Spatial orientation: theory, research, and application, 225–82. New York: Plenum. 1985. Lexicalization patterns: semantic structure in lexical forms. In Timothy Shopen (ed.), Language typology and syntactic description, vol. III: grammatical categories and the lexicon, 57–149. Cambridge University Press. 1988a. Force dynamics in language and cognition. Cognitive Science 12, 49–100. 1988b. The relation of grammar to cognition. In Brygida Rudzka–Ostyn (ed.), Topics in cognitive linguistics, 165–205. Amsterdam: John Benjamins. 1996 [2000]. Fictive motion in language and ‘ception.’ In Toward a cognitive semantics: concept structuring systems, vol. 1, 99–175. Cambridge, MA: MIT Press. 1996. Fictive motion in language and ‘ception.’ In Paul Bloom, Mary A. Peterson, Lynn Nadel, and Merrill F. Garrett (eds.), Language and Space, 211–76. Cambridge, MA: MIT Press. 2000a. Toward a cognitive semantics, vol. 1: concept structuring systems. Cambridge, MA: MIT Press. 2000b. Toward a cognitive semantics, vol. 2: typology and process in concept structuring. Cambridge, MA: MIT Press. Tanenhaus, Michael K., and Michael J. Spivey-Knowlton. 1996. Eyetracking. In Language and cognitive processes: a guide to spoken word recognition paradigms 11, 583–88. London: Routledge. Tannen, Deborah. 1986. Introducing constructed dialogue in Greek and American conversational and literary narratives. In Florian Coulmas (ed.), Direct and indirect speech, 311–22. Berlin: De Gruyter Mouton. Taremaa, Piia. 2013. Fictive and actual motion in Estonian: encoding space. SKY Journal of Linguistics 26, 151–183. Taub, Sarah. 2001. Language in the body: iconicity and metaphor in American Sign Language. Cambridge University Press. Taylor, Holly A., and Barbara Tversky. 1996. Perspective in spatial descriptions. Journal of Memory and Language 35, 371–91. Taylor, John R. 2002a. Cognitive grammar. Oxford University Press. 2002b. Near synonyms as co-extensive categories: ‘high’ and ‘tall’ revisited. Language Sciences 25: 263–84. 2003. Linguistic categorization. Oxford University Press. 2006. Where do Phonemes come from? A view from the bottom. International Journal of English 6(2), 19–54 2012. The mental corpus: how language is represented in the mind. Oxford University Press. 2015. Word formation in cognitive grammar: word-formation. In Peter O. Mu¨ller, Inge Ohnheiser, Susan Olsen, and Franz Rainer (eds.). An international handbook of the languages of Europe, 145–58. Berlin: De Gruyter Mouton.
References
Teddiman, Laura. 2012. Conversion and the lexicon: comparing evidence from corpora and experimentation. In Dagmar S. Divjak and Stefan Th. Gries (eds.), Frequency effects in language representation, 235–54. Berlin: De Gruyter Mouton. Tedlock, Barbara. 1982. Time and the Highland Maya. Albuquerque: University of New Mexico Press. Tenbrink, Thora. 2007a. Imposing common ground by using temporal connectives: the pragmatics of before and after. In Anita Fetzer and Kerstin Fischer (eds.), Lexical markers of common grounds, 113–39. Amsterdam: Elsevier. 2007b. Space, time, and the use of language: an investigation of relationships. Berlin: De Gruyter Mouton. 2011. Reference frames of space and time in language. Journal of Pragmatics 43, 704–22. 2012. Relevance in spatial navigation and communication. In Cyrill Stachniss, Kerstin Schill, and David Uttal (eds.), Spatial cognition 2012, 358–77. Heidelberg: Springer. 2015. Cognitive Discourse Analysis: accessing cognitive representations and processes through language data. Language and Cognition 7(1), 98–137. Tenbrink, Thora, and Frank Schilder. 2003. (Non)temporal concepts conveyed by before, after, and then in dialogue. In Peter Ku¨hnlein, Hannes Rieser, and Henk Zeevat (eds.), Perspectives on dialogue in the new millennium, 353–80. Amsterdam: John Benjamins. Tenbrink, Thora, and Inessa Seifert. 2011. Conceptual layers and strategies in tour planning. Cognitive Processing 12(1), 109–25. Tenbrink, Thora, and Jan Wiener. 2009. The verbalization of multiple strategies in a variant of the traveling salesperson problem. Cognitive Processing 10(2), 143–61. Tenbrink, Thora, Kenny R. Coventry, and Elena Andonova. 2011. Spatial strategies in the description of complex configurations. Discourse Processes 48, 237–66. Teuscher, Ursina, Marguerite McQuire, Jennifer Collins, and Seana Coulson. 2008. Congruity effects in time and space: behavioral and ERP measures. Cognitive Science 32(3), 563–78. Theakston, Anna L., and Elena Lieven. 2008. The influence of discourse context on children’s provision of auxiliary BE. Journal of Child Language 35, 129–58. Theakston, Anna L., Elena Lieven, Julian Pine, and Caroline Rowland. 2001. The role of performance limitations in the acquisition of verb-argument structure. Journal of Child Language 28, 127–52. Theakston, Anna L., Paul Ibbotson, Daniel Freudenthal, Elena V. Lieven, and Michael Tomasello. 2015. Productivity of noun slots in verb frames. Cognitive Science 39, 1369–95. Theakston, Anna L., Robert Maslen, Elena V. M. Lieven, and Michael Tomasello. 2012. The acquisition of the active transitive
801
802
References
construction in English: a detailed case study. Cognitive Linguistics 23, 91–128. Theijssen, Daphne, Louis ten Bosch, Lou Boves, Bert Cranen, and Hans van Halteren. 2013. Choosing alternatives: using Bayesian networks and memory-based learning to study the dative alternation. Corpus Linguistics and Linguistic Theory 9(2), 227–62. Thibodeau, Paul H. and Lera Boroditsky. 2011. Metaphors we think with: the role of metaphor in reasoning. PLoS ONE 6, e16782. Thompson, Cynthia K. 2000. Neuroplasticity: evidence from aphasia. Journal of Communication Disorders 33, 357–66. Thompson, D’Arcy Wentworth W. 1917. On growth and form. Cambridge University Press. Thompson, Sandra A. 2002. Object complements and conversation: towards a realistic account. Studies in Language 26(1), 125–64. Thompson, Sandra A., and Yuka Koide. 1987. Iconicity and ‘indirect objects’ in English. Journal of Pragmatics 11(3), 309–406. Thurmair, Maria. 1989. Modalpartikeln und ihre Kombinationen. Tu¨bingen: Niemeyer. Timyam, Napasri, and Benjamin Bergen. 2010. A contrastive study of the caused-motion and ditransitive constructions in English and Thai: semantic and pragmatic constraints. In Hans C. Boas (ed.), Contrastive studies in construction grammar, 137–68. Amsterdam: John Benjamins. Tissari, Heli. 2001. Metaphors we love by: on the cognitive metaphors of love from the 15th century to the present. Studia Anglica Posnaniensia 36, 217–42. Tobin, Vera, and Michael Israel. 2012. Irony as a viewpoint phenomenon. In Barbara Dancygier and Eve Sweetser (eds.), 25–46. Viewpoint in language: a multimodal perspective. Cambridge University Press. Tollefsen, Deborah, and Rick Dale. 2012. Naturalizing joint action: a process-based approach. Philosophical Psychology 25, 385–407. Tomasello, Michael. 1999. The cultural origins of human cognition. Cambridge, MA: Harvard University Press. 2000. First steps toward a usage-based theory of language acquisition. Cognitive Linguistics 11, 61–82. 2003. Constructing a language: a usage-based theory of language acquisition. Cambridge, MA: Harvard University Press. 2009. Universal grammar is dead. Behavioral and Brain Sciences 32(5), 470–71. 2010. Cognitive linguistics and first language acquisition. In Dirk Geeraerts and Hubert Cuyckens (eds.), The Oxford handbook of cognitive linguistics, 1092–1112. Oxford University Press. Tomasello, Michael, and Patricia Brooks. 1998. Young children’s earliest transitive and intransitive constructions. Cognitive Linguistics 9, 379–95.
References
Tomasello, Michael, and Hannes Rakoczy. 2003. What makes human cognition unique? From individual to shared to collective intentionality. Mind and Language 18, 121–47. Tomasello, Michael, Malinda Carpenter, Joseph Call, Tanya Behne, and Henrik Moll. 2005. Understanding and sharing intentions: the origins of cultural cognition. Behavioral and Brain Science 28(5), 675–91. Torralbo, Ana, Julio Santiago, and Juan Lupia´n˜ez. 2006. Flexible conceptual projection of time onto spatial frames of reference. Cognitive Science 30, 745–57. Traugott, Elizabeth C. 1972. A history of English syntax. New York: Holt, Rinehardt, Winston. 1982. From propositional to textual and expressive meanings: some semantic-pragmatic aspects of grammaticalization. In Winfred P. Lehmann and Yakov Malkiel (eds.), Perspectives on historical linguistics, 245–71. Amsterdam: John Benjamins. 1989. On the rise of epistemic meanings in English: an example of subjectification in semantic change. Language 65(1), 31–55. 1995. Subjectification and grammaticalization. In Dieter Stein and Susan Wright (eds.), Subjectivity and subjectification, 31–45. Cambridge University Press. 1997. Subjectification and the development of epistemic meaning: the case of promise and threaten. In Toril Swan and Olaf Jansen Westvik (eds.), Modality in Germanic languages: historical and comparative perspectives, 185–210. Berlin: De Gruyter. 2003. From subjectification to intersubjectification. In Raymond Hickey (ed.), Motives for language change, 124–40. Cambridge University Press. 2008a. The grammaticalization of NP of NP patterns. In Alexander Bergs and Gabriele Diewald (eds.), Constructions and language change, 23–45. Berlin: De Gruyter Mouton. 2008b. ‘All That he Endeavoured to Prove was. . . ’: on the emergence of grammatical constructions in dialogual contexts. In Ruth Kempson and Robin Cooper (eds.), Language change and evolution, 143–77. London: Kings College. 2008c. Grammaticalization, constructions and the incremental development of language: suggestions from the development of degree modifiers in English. In Regine Eckardt, Gerhard Ja¨ger, and Tonjes Veenstra (eds.), Variation, selection, development: probing the evolutionary model of language change, 219–50. Berlin: De Gruyter Mouton. 2010. (Inter)subjectivity and (inter)subjectification: a reassessment. In Kristin Davidse, Lieven Vandelanotte and Hubert Cuyckens (eds.), Subjectification, intersubjectification and grammaticalization, 29–71. Berlin: De Gruyter Mouton. 2015. Toward a coherent account of grammatical constructionalization. In Jo´hanna Barðdal, Elena Smirnova, Lotte Sommerer, and
803
804
References
Spike Gildea (eds.), Diachronic construction grammar, 51–80. Amsterdam: John Benjamins. Traugott, Elizabeth C., and Bernd Heine (eds.). 1991. Approaches to grammaticalization, vol. I and II. Amsterdam: John Benjamins. Traugott, Elizabeth C., and Ekkehard Ko¨nig. 1991. The semanticspragmatics of grammaticalization revisited. In Elizabeth C. Traugott and Bernd Heine (eds.), Approaches to grammaticalization, 189–218. Amsterdam: John Benjamins. Traugott, Elizabeth C., and Graeme Trousdale. 2010. Gradience, gradualness and grammaticalization: how do they intersect? In Elizabeth Closs Traugott and Graeme Trousdale (eds.), Gradience, gradualness and grammaticalization: typological Studies in Language 90, 19–44. Amsterdam: John Benjamins. 2013. Constructionalization and constructional changes. Oxford University Press. Trevarthen, Colwyn. 1979. Communication and cooperation in early infancy: a description of primary intersubjectivity. In Margaret Bullowa (ed.), Before speech: the beginning of interpersonal communication, 321–47. Cambridge University Press. 2011. What is it like to be a person who knows nothing? Defining the active intersubjective mind of a newborn human being. Infant and Child Development 20(1), 119–35. Trousdale, Graeme. 2008a. Words and constructions in grammaticalization: the end of the English Impersonal Construction. In Susan M. Fitzmaurice and Donka Minkova (eds.), Studies in the history of the English language IV: empirical and analytical advances in the study of English language change, 301–26. Berlin: De Gruyter Mouton. 2008b. A constructional approach to lexicalization processes in the history of English: evidence from possessive constructions. Word Structure 1, 156–77. 2008c. Constructions in grammaticalization and lexicalization: evidence from the history of a composite predicate construction in the history of English. In Graeme Trousdale and Nikolas Gisborne (eds.), Constructional approaches to English grammar: topics in English linguistics 57, 33–67. Berlin: De Gruyter Mouton. 2010. Issues in constructional approaches to grammaticalization. In Katerina Stathi, Elke Gehweiler, and Ekkehard Ko¨nig (eds.), Grammaticalization: current views and issues, 51–72. Amsterdam: John Benjamins. 2012. Theory and data in diachronic construction grammar: the case of the what with construction. Studies in Language 36(3), 576–602. Trousdale, Graeme, and Nikolas Gisborne (eds.). 2008. Constructional approaches to English grammar: topics in English Linguistics 57. Berlin: De Gruyter Mouton.
References
Trousdale, Graeme, and Elizabeth Closs Traugott. 2013. Constructionalization and constructional changes. Oxford University Press. Trubetzkoy, Nicolai. S. 1939 [1969]. Gru¨ndzu¨ge der Phonologie [Principles of Phonology]. Trans. and ed. Christine A. M. Baltaxe. Prague [Los Angeles]: Travaux du cercle linguistique de Prague [University of California Press]. Tsai, Peggy Wei-lun and Shelley Ching-yu Hsieh. 2013. Fictive motion in Chinese and English tourist guidebooks. Canadian Social Science 9(2), 1–6. Tsur, Reuven. 1992. Toward a theory of Cognitive Poetics. Amsterdam: Elsevier Science Publishers. Tuggy, David. 1981. The transitivity-related morphology of Tetelcingo Na´huatl: an exploration in Space Grammar. PhD dissertation. University of California at San Diego. Turkeltaub, P. E., S. Messing, C. Norise, and R. H. Hamilton. 2011. Are networks for residual language function and recovery consistent across aphasic patients? Neurology 76, 1726–34. Turner, Mark. 1989. Death is the mother of beauty: mind, metaphor, criticism. University of Chicago Press. 1991. Reading minds: the study of English in the age of cognitive science. Princeton University Press. 1996. The literary mind: the origins of language and thought. Oxford University Press. 2010. Ten lectures on mind and language. Beijing: Foreign Language Teaching and Research Press. 2015. Blending in language and communication. In Ewa Da˛browska and Dagmar Divjak (eds.), Handbook of cognitive linguistics, 211–32. Berlin: De Gruyter Mouton. Tutton, Mark. 2013. A new approach to analysing static locative expressions. Language and Cognition 5(1), 25–60. Tversky, Barbara, Sol Kugelmass, and Atalia Winter. 1991. Cross-cultural and developmental trends in graphic productions. Cognitive Psychology 23(4), 515–57. Tversky, Barbara, Julie Heiser, Paul Lee, and Marie-Paule Daniel. 2009. Explanations in gesture, diagram, and word. In Kenny R. Coventry, Thora Tenbrink, and John Bateman (eds.), Spatial language and dialogue, 119–31. Oxford University Press. Tyler, Andrea. 2008. Cognitive linguistics and second language instruction. In P. Robinson and N. C. Ellis (eds.), Handbook of cognitive linguistics and second language acquisition, 456–88. London: Routledge. 2012. Applying cognitive linguistics to second language learning: theoretical basics and experimental evidence. London: Routledge. Tyler, Andrea (ed.). In prep. What is applied cognitive linguistics? Answers from current SLA research. Berlin: De Gruyter Mouton.
805
806
References
Tyler, Andrea, and Vyvyan Evans. 2003. The semantics of English prepositions: spatial scenes, embodied meaning and cognition. Cambridge University Press. Tyler, Andrea, Vu Ho, and Charles Mueller. 2011. Using Cognitive Linguistics to teach the double object construction. Paper presented at AAAL annual conference. Chicago, IL (March 2011). Reported in Tyler 2012. Tyler, Andrea, Yiyoung Kim, and Mari Takada (eds.), Language in the context of use: discourse and cognitive approaches to language. Berlin: De Gruyter Mouton. Tyler, Andrea, Charles Mueller, and Vu Ho. 2010. Applying cognitive linguistics to instructed L2 learning: the English modals. AILA Review, 23, 30–49. 2011. Applying cognitive linguistics to learning the semantics of English to, for and at: an experimental investigation. Vigo International Journal of Applied Linguistics 8, 181–205. Ulrich, Rolf, and Claudia Maienborn. 2010. Left-right coding of past and future in language: the mental timeline during sentence processing. Cognition, 117(2), 126–38. Ulrich, Rolf, Verena Eikmeier, Irmgard de la Vega, Susana Ruiz Ferna´ndez, Simone Alex-Ruf, and Claudia Maienborn. 2012. With the past behind and the future ahead: back-to-front representation of past and future sentences. Memory and Cognition 40(3), 483–95. Ungerer, Friedrich. 2007. Word formation. In Dirk Gereraerts and Hubert Cuyckens (eds.), The Oxford handbook of cognitive linguistics, 650–75. Oxford University Press. Ungerer, Friedrich, and Hans-Jo¨rg Schmid. 2006. An introduction to cognitive linguistics. 2nd edn. London: Pearson Education. Valenzeno, Laura, Martha W. Alibali, and Roberta Klatzky. 2003. Teachers’ gestures facilitate students’ learning: a lesson in symmetry. Contemporary Educational Psychology 28, 187–204. Valenzuela, Javier, and Christina Soriano. 2008. Is friendship more important than money? A psycholinguistic exploration of the important is big metaphor. VI AELCO Conference. Castello´n. van Elk, Michiel, Hein T. van Schie, and Harold Bekkering. 2010. From left to right: processing acronyms referring to names of political parties activates spatial associations. Quarterly Journal of Experimental Psychology 63, 22202–19. Van Goethem, Kristel. 2008. Oud-leerling versus ancien e´le`ve: a comparative study of adjectives grammaticalizing into prefixes in Dutch and French. Morphology 18, 27–49. 2010. The French construction nouveau and past participle revisited: arguments in favour of a prefixoid analysis of nouveau. Folia Linguistica 44, 163–78. Van Heugten, Marieke, and Elizabeth K. Johnson. 2011. Gender-marked determiners help Dutch learners’ word recognition when gender information itself does not. Journal of Child Language 38(1), 87–100.
References
van Hoek, Karen. 1995. Conceptual reference points: a cognitive grammar account of pronominal anaphora constraints. Language 71, 310–40. 1997. Anaphora and conceptual structure. University of Chicago Press. Van Krieken, Kobie, Jose´ Sanders, and Hans Hoeken. 2016. Blended viewpoints, mediated witnesses: a cognitive-linguistic approach to news narratives. In Barbara Dancygier, Wei-lun Lu, and Arie Verhagen (eds.), Viewpoint and the fabric of meaning: form and use of viewpoint tools across languages and modalities, 145–68. Berlin: De Gruyter Mouton. Van Trijp, Remi. 2013. A comparison between Fluid Construction Grammar and Sign-based Construction Grammar. Constructions and Frames 5(1), 88–116. 2014. Cognitive vs. generative construction grammar: the case of coercion and argument structure. Cognitive Linguistics 26(4), 613–32. Van Valin, Robert D. Jr., and David P. Wilkins. 1996. The case for ‘effector’: case roles, agents, and agency revisited. In Masayoshi Shibatani and Sandra Thompson (eds.), Grammatical constructions, 289–322. Oxford University Press. Vandelanotte, Lieven. 2004. Deixis and grounding in speech and thought representation. Journal of Pragmatics 36(3), 489–520. 2009. Speech and thought representation in English: a cognitive-functional approach. Berlin: De Gruyter Mouton. 2012. ‘Wait till you got started’: how to submerge another’s discourse in your own. In Barbara Dancygier and Eve Sweetser (eds.), Viewpoint in language: a multimodal perspective, 198–218. Cambridge University Press. 2015. ‘More than one way at once’: simultaneous viewpoints in text and image. In Andra´s Benedek and Kristo´f Nyı´ri (eds.), Beyond words: pictures, parables, paradoxes, 75–81. Frankfurt: Peter Lang. Vandelanotte, Lieven, and Barbara Dancygier (eds.) In prep. Multimodal artefacts and the texture of viewpoint. Special issue of Journal of Pragmatics. Vannerem, Mia and Mary Snell-Hornby. 1986. Die Szene hinter dem Text: scenes-and-frame Semantics in der U¨bersetzung. In M. Snell-Hornby (ed.), U¨bersetzungswissenschaft – Eine Neuorientierung, 184–205. Tu¨bingen: Francke. Vater, Heinz. 2010. Sprachspiele: kreativer Umgang mit Sprache. Linguistische Berichte 221, 3–36. Vaughan, Brian. 2011. Prosodic synchrony in co-operative task-based dialogues: a measure of agreement and disagreement. Proceedings of Interspeech, 1865–68. Florence, Italy, August 28–31. Veale, Tony. 2012. Exploding the creativity myth: the computational foundations of linguistic creativity. London: Bloomsbury. Veale, Tony, Kurt Feyaerts, and Geert Broˆne. 2006. The cognitive mechanisms of adversarial humor. Humor: The International Journal of Humor Research 19(3), 305–38.
807
808
References
Veale, Tony, Kurt Feyaerts, and Charles Forceville (eds.). 2013. Creativity and the agile mind: a multidisciplinary approach to a multifaceted phenomenon. Berlin: De Gruyter Mouton. Vennemann, Theo. 1974. Topics, subjects, and word order: from SXV to SVX via TVX. In John Anderson and Charles Jones (eds.), Historical linguistics: proceedings of the first International Congress of Historical Linguistics, Edinburgh, September 1973. Vol. II, 339–76. Amsterdam: North-Holland. Verhagen, Arie. 2000. ‘The girl that promised to become something’: an exploration into diachronic subjectification in Dutch. In T. F. Shannon and J. P. Snapper (eds.), The Berkeley conference on Dutch linguistics 1997: the Dutch language at the millennium, 197–208. Lanham, MD: University Press of America. 2005. Constructions of intersubjectivity: discourse, syntax, and cognition. Oxford University Press. 2008a. Intersubjectivity and explanation in linguistics: a reply to Hinzen and Van Lambalgen. Cognitive Linguistics 19(1), 125–43. 2008b. Intersubjectivity and the architecture of the language system. In Jordan Zlatev, Timothy P. Racine, Chris Sinha, and Esa Itkonen (eds.), The shared mind: perspectives on intersubjectivity, 307–31. Amsterdam: Jolm Benjamins. 2015. Grammar and cooperative communication. In Ewa Da˛browska and Dagmar Divjak (eds.), Handbook of cognitive linguistics, 232–52. Berlin: De Gruyter Mouton. Versluis, Christine, and Lou-Ann Kleppa. [2008] 2016. The use of interactive structures as communicative strategy in Dutch and Portuguese aphasic speakers. In Esther Pascual and Sergeiy Sandler (eds.), The conversation frame: forms and functions of fictive interaction, 323–42. Amsterdam: John Benjamins. Verspoor, Marjolijn H. 2008. Cognitive linguistics and its applications to second language teaching. In J. Cenoz, and Nancy H. Hornberger (eds.), Encyclopedia of language and education, vol. 6. 2nd edn. New York: Springer, 1843–54. Verspoor, Marjolijn H., and N. T. P. Hong. 2008. Cognitive grammar and teaching English articles to Asian students. In Jean-Re´mi Lapaire (ed), From grammar to mind: grammar as cognition, 249–68. Presses Universitaires de Bordeaux. 2013. A dynamic usage-based approach to communicative language teaching. European Journal of Applied Linguistics 1(1), 22–54. Verspoor, Marjolijn H., and Wander Lowie. 2003. Making sense of polysemous words. Language Learning 53, 547–86. Viberg, Åke. 1983. The verbs of perception: a typological study. Linguistics 21, 123–62. Vis, Kirsten, Jose´ Sanders, and Wilbert Spooren. 2012. Diachronic changes in subjectivity and stance: a corpus linguistic study of Dutch news texts. Discourse, Context and Media 1, 95–102.
References
Voloshinov, Valentin N. [1929] 1986. Marxism and the philosophy of language. Trans. Ladislav Matejka and I. R. Titunik. Cambridge, MA: Harvard University Press. Von Heusinger, Klaus, and Christoph Schwarze. 2013. Italian V+N compounds, inflectional features and conceptual structure. Morphology 23, 325–50. von Stockert, Theodor R., and Louisa Bader. 1976. Some relations of grammar and lexicon in aphasia. Cortex 12(1), 49–60. Voort, Hein, van der. 2002. The quotative construction in Kwaza and its (de)grammaticalisation. In Mily Crevels, Simon van de Kerke, Sergio Meira, and Hein van der Voort (eds.), Current studies on South American languages: indigenous languages of Latin America (ILLA) 3, 307–28. 2009. Reduplication and repetition of person markers in Guapore´ isolates. Morphology 19(2), 263–86. 2013. Fala fictı´cia fossilizada: o tempo futuro em Aikana˜. Boletim do Museu Paraense Emı´lio Goeldi 8(2), 359–77. 2016. Recursive inflection and grammaticalised fictive interaction in the southwestern Amazon. In Esther Pascual and Sergeiy Sandler (eds.), The conversation frame: forms and functions of fictive interaction, 277–99. Amsterdam: John Benjamins. Vorwerg, Constanze, and Thora Tenbrink. 2007. Discourse factors influencing spatial descriptions in English and German. In Thomas Barkowsky, Markus Knauff, Ge´rard Ligozat, and Dan Montello (eds.), Spatial cognition V: reasoning, action, interaction, 470–88. Berlin: Springer. Vries, Lourens J. de. 2010. Direct speech, fictive interaction, and bible translation. The Bible Translator 61(1), 31–40. Vygotsky, Lev S. 1934 [1962]. Thought and language. Cambridge, MA: MIT Press. 1978. Mind in society. Cambridge, MA: Harvard University Press. Wagner, Petra, Zofia Malisz, and Stefan Kopp. 2014. Gesture and speech in interaction: an overview. Speech Communication 57, 209–32. Wa¨lchli, Bernard, and Michael Cysouw. 2012. Lexical typology through similarity semantics: toward a semantic map of motion verbs. Linguistics 50(3), 671–710. Walker, Esther, and Kensy Cooperrider. 2016. The continuity of metaphor: evidence from temporal gestures. Cognitive Science 40, 481–95. Wallentin, Mikkel, Torben E. Lund, Svend Østergaard, Leif Østergaard, and Andreas Roepstorff. 2005. Motion verb sentences activate left posterior middle temporal cortex despite static context. NeuroReport 16, 649–52. Walsh, Vincent. 2003. A theory of magnitude: common cortical metrics of time, space and quantity. Trends in Cognitive Sciences 7(11), 483–88. Waltereit, Richard. 2006. The rise of discourse markers in Italian: a specific type of language change. In Kerstin Fischer (ed.), Approaches to discourse particles, 61–67. Leiden: Brill.
809
810
References
Warren, Beatrice. 2006. Referential metonymy. Lund, Sweden: Scripta Minora. Waxman, Nach. 2004. Recipes. In Smith, Andrew (ed.), The Oxford encyclopedia of food and drink in America, 247–50. Oxford University Press. Webb, James T. 1972. Interview synchrony: an investigation of two speech rate measures in an automated standardized interview. In Aron W. Siegman and Benjamin Pope (eds.), Studies in dyadic communication, 115–33. New York: Pergamon Press. Weger, Ulrich W., and Jay Pratt. 2008. Time flies like an arrow: space-time compatibility effects suggest the use of a mental timeline. Psychonomic Bulletin and Review 15(2), 426–30. Weidhaas, Thomas, and Hans-Jo¨rg Schmid. 2015. Diminutive verbs in German: semantic analysis and theoretical implications. Morphology 25, 183–228. Weinreich, Uriel, William Labov, and Marvin Herzog. 1968. Empirical foundations for a theory of language change. In Winfred P. Lehman and Yakov Malkiel (eds.), Directions for historical linguistics, 95–189. Austin: University of Texas Press. Welke, Klaus. 2009. Valenztheorie und Konstruktionsgrammatik. Zeitschrift fu¨r Germanistische Linguistik 37, 81–124. Werner, Heinz, and Bernard Kaplan. 1963. Symbol formation: an organismicdevelopmental approach to language and the expression of thought. New York: Wiley. Wernicke, Carl. 1977. Der aphasische symptomencomplex: eine psychologische studie auf anatomischer basis. In G. H. Eggert (ed.), Wernicke’s works on aphasia: a sourcebook and review, 91–145. The Hague: Mouton. Whiten, Andrew. 1991. Natural theories of mind: evolution, development and simulation of everyday mindreading. Oxford: Blackwell. Whorf, Benjamin Lee. 1956. Language, thought, and reality: selected writings of Benjamin Lee Whorf. Ed. John B. Carroll. Cambridge, MA: MIT Press. Wide, Camilla. 2009. Interactional construction grammar: contextual features of determination in dialectal Swedish. In Alexander Bergs and Gabriele Diewald (eds.), Contexts and constructions, 111–41. Amsterdam: John Benjamins. Wieling, Martijn, John Nerbonne, and R. Harald Baayen 2011. Quantitative social dialectology: explaining linguistic variation geographically and socially. PLOS ONE 6(9), e23613. Wierzbicka, Anna. 1987. Boys will be boys. Language 63(1), 95–114. 1988. The semantics of grammar. Amsterdam: John Benjamins. 1995. Dictionaries vs. encyclopaedias: how to draw the line. In Philip W. Davis (ed.), Alternative linguistics: descriptive and theoretical modes, 289–315. Amsterdam: John Benjamins. 1996. Semantics: primes and universals. Oxford University Press. Wilbur, Ronnie B. 1987. American Sign Language: linguistic and applied dimensions. Boston, MA: College-Hill Press.
References
2013. The point of agreement: changing how we think about sign language, gesture, and agreement. Sign Language and Linguistics 16(2), 221–58. Wilcox, Phyllis P. 2004. A cognitive key: metonymic and metaphorical mappings in ASL. Cognitive Linguistics 15(2), 197–222. Wilcox, Sherman. 2004a. Cognitive iconicity: conceptual spaces, meaning, and gesture in signed language. Cognitive Linguistics 15(2), 119–47. 2004b. Gesture and language: cross-linguistic and historical data from signed languages. Gesture 4(1), 43–75. 2007. Routes from gesture to language. In Elena Pizzuto, Paola Pietrandrea, and Raffaele Simone (eds.), Verbal and signed languages: comparing structures, constructs and methodologies, 107–31. Berlin: De Gruyter Mouton. 2009. Symbol and symptom: routes from gesture to signed language. Annual Review of Cognitive Linguistics 7, 89–110. 2014. Moving beyond structuralism: usage-based signed language linguistics. Lingua de Sen˜as e Interpretacio´n 5, 97–126. Wilcox, Sherman, and Jill P. Morford. 2007. Empirical methods in signed language research. In Monica Gonzalez-Marquez, Irene Mittelberg, Seana Coulson, and Michael J. Spivey (eds.), Methods in cognitive linguistics, 173–202. Amsterdam: Benjamins. Wilcox, Sherman, and Corrine Occhino. 2016. Constructing signs: place as a symbolic structure in signed languages. Cognitive Linguistics 27(3), 371–404. Wilcox, Sherman, and Barbara Shaffer. 2006. Modality in American Sign Language. In William Frawley (ed.), The expression of modality, 207–37. Berlin: De Gruyter Mouton. Wilcox, Sherman, and Phyllis Perrin Wilcox. 1995. The gestural expression of modality in American Sign Language. In Joan Bybee and Suzanne Fleischman (eds.), Modality in grammar and discourse, 135–62. Amsterdam: John Benjamins. 2013. Cognitive linguistics and signed languages. International Journal of Cognitive Linguistics 3(2), 127–51. Wilkins, David P. 1996. Natural tendencies of semantic change and the search for cognates. In Mark Durie and Malcolm Ross (eds.), The comparative method reviewed: regularity and irregularity in language change, 264–304. Oxford University Press. Willems, Dominique. 2012. Verb typology: between construction and lexicon. In M. Bouveret and D. Legallois (eds.), Constructions in French, 23–48. Amsterdam: John Benjamins. Willems, Roe, and Arthur Jacobs. 2016. Caring about Dostoyevsky: the untapped potential of studying literature. Trends in Cognitive Sciences 20(4), 243–45. Williams, Lawrence, and John Bargh. 2008. Experiencing physical warm influences interpersonal warmth. Science 322, 606–07.
811
812
References
Williams, Robert F. 2004. Making meaning from a clock: material artifacts and conceptual blending in time-telling instruction. PhD dissertation. University of California, San Diego. Willkop, Eva-Maria. 1988. Gliederungspartikeln im Dialog. Mu¨nchen: Iudicium. Wilson, Nicole, and Raymond W. Gibbs Jr. 2007. Real and imagined body movement primes metaphor comprehension. Cognitive Science 31, 721–31. Winter, Bodo, Tyler Marghetis, and Teenie Matlock. 2015. Of magnitudes and metaphors: explaining cognitive interactions between space, time, and number. Cortex, 64, 209–24. Winter, Bodo, Marcus Perlman, and Teenie Matlock. 2013. Using space to talk and gesture about numbers: evidence from the TV News Archive. Gesture 13(3), 377–408. Winters, Margaret, Heli Tissari, and Kathryn Allan (eds.). 2010. Historical cognitive linguistics. Berlin: de Gruyter. Wittgenstein, Ludwig. 1953 Philosophical investigations. Oxford: Basil Blackwell. 1978. Philosophical investigations. Trans. G. E. M. Anscombe. Oxford: Blackwell. Wood, Guilherme, Klaus Willmes, Hans-Christoph Nuerk, and Martin H. Fischer. 2008. On the cognitive link between space and number: a meta-analysis of the SNARC effect. Psychology Science Quarterly 50(4), 489–525. Woodward, James. 1974. Implicational variation in American Sign Language: negative incorporation. Sign Language Studies 5, 20–30. 1976a. Black Southern signing. Language in Society 5(2), 211–18. 1976b. Signs of change: historical variation in American Sign Language. Sign Language Studies 10, 81–94. 1978. Historical bases of American Sign Language. In Patricia Siple (ed.), Understanding language through sign language research, 333–48. New York: Academic Press. Woolf, Virginia. 1927 [2004]. To the lighthouse. London: Collector’s Library Wray, Alison. 2015. Why are we so sure we know what a word is? In John R. Taylor (ed.), The Oxford handbook of the word, 725–50. Oxford University Press. Wu, Ying C., and Seana Coulson. 2015. Iconic gestures facilitate discourse comprehension in individuals with superior immediate memory for body configurations. Psychological Science 26(11), 1717–27. Wulf, Alyssa, and Paul Dudis. 2005. Body partitioning in ASL metaphorical blends. Sign Language Studies 5(3), 317–32. Wulff, Stefanie. 2008. Rethinking idiomaticity: A usage-based approach. London and New York: Continuum. 2013. Words and idioms. In Thomas Hoffmann and Graeme Trousdale (eds.), The Oxford handbook of construction grammar, 274–89. Oxford University Press.
References
Wulff, Stefanie, Stefan T. Gries, and Nicholas Lester. In prep. Optional that in complementation by German and Spanish learners. In Andrea Tyler, Lihong Huang and Hana Jan (eds.), What is applied cognitive linguistics? Answers from current SLA research. Berlin: De Gruyter Mouton. Wulff, Stefanie, Nick C. Ellis, Ute Ro¨mer, Kathleen Bardovi-Harlig, and Chelsea LeBlanc. 2009. The acquisition of tense-aspect: converging evidence from corpora, cognition and learner constructions. Modern Language Journal 93, 354–69. Xavier, Andre´o Nogueira, and Sherman Wilcox. 2014. Necessity and possibility modals in Brazilian Sign Language (Libras). Linguistic Typology 18(3), 449–88. Xia, M., J. Wang, and Y. He. 2013 BrainNet viewer: a network visualization tool for human brain connectomics. PLoS ONE 8, e68910. Xiang, Mingjian. 2016. Real, imaginary, or fictive? Philosophical dialogues in an early Daoist text and its pictorial version. In Esther Pascual and Sergeiy Sandler (eds.), The conversation frame: forms and functions of fictive interaction, 63–86. Amsterdam: John Benjamins. Xiang, Mingjian, and Esther Pascual. 2016. Debate with Zhuangzi: fictive interaction blends in ancient Chinese philosophy. Pragmatics 26(1), 137–62. Xuan, B., D. Zhang, S. He, and X. Chen. 2007. Larger stimuli are judged to last longer. Journal of Vision 7(10), 1–5. Yang, Tangfeng 2016. Image schemas in verb-particle constructions: evidence from a behavioral experiment. Journal of Psycholinguistic Research 45, 379–93. Yaxley, Richard H., and Rolf A. Zwaan. 2007. Simulating visibility during language comprehension. Cognition 105(1), 229–36. Yu, Ning. 1998. The contemporary theory of metaphor: a perspective from Chinese. Amsterdam: John Benjamins. 2012. The metaphorical orientation of time in Chinese. Journal of Pragmatics 44, 1335–54. Zacks, Jeff M., Barbara Tversky, and Gowri Iyer. 2001. Perceiving, remembering, and communicating structure in events. Journal of Experimental Psychology: General 130, 29–58. Zahavi, David. 2001. Beyond empathy: phenomenological approaches to intersubjectivity. Journal of Consciousness Studies 8(5–6), 151–67. 2003. Husserl’s phenomenology. Stanford University Press. Zatorre, Robert J., Ernst Meyer, Albert Gjedde, and Alan C. Evans. 1996. PET studies of phonetic processing of speech: review, replication, and reanalysis. Cerebral Cortex 6, 21–30. Zbikowski, Lawrence. 2009. Music, language and multimodal metaphor. In Charles Forceville and Eduardo Urios-Aparisi (eds.), Multimodal metaphor, 359–82. Berlin: De Gruyter Mouton.
813
814
References
Zelazo, Philip D. 2004. The development of conscious control in childhood. Trends in Cognitive Sciences 8(1), 12–17. Zeldes, Amir. 2012. Productivity in argument selection: a usage-based approach to lexical choice in syntactic slots. PhD dissertation. Humboldt University Berlin. Zenner, Eline, Dirk Speelman, and Dirk Geeraerts. 2012. Cognitive sociolinguistics meets loanword research: measuring variation in the success of anglicisms in Dutch. Cognitive Linguistics 23, 749–92. Zenner, Eline, Gitte Kristiansen, Laura Janda, and Arie Verhagen. 2015. Introduction. In Daems, Jocelyne, Eline Zenner, Kris Heylen, Dirk Speelman, and Hubert Cuyckens (eds.), Change of paradigms – new paradoxes: recontextualizing language and linguistics, 1–14. Berlin: De Gruyter Mouton. Zeschel, Arne. 2012. Incipient productivity: a construction-based approach to linguistic creativity. Berlin: De Gruyter Mouton. Zeshan, Ulrike. 2004. Interrogative constructions in signed languages: crosslinguistic perspectives. Language 80(1), 7–39. Zeshan, Ulrike, and Connie De Vos. 2012. Sign languages in village communities: anthropological and linguistic insights. Berlin: Walter de Gruyter. Zhang, Weiwei, Dirk Speelman, and Dirk Geraarts. 2011. Variation in the (non)metonymic capital names in mainland Chinese and Taiwanese Chinese. Metaphor and the Social World 1(1), 90–112. Zhao, Helen, and Man Ho Wong. 2015. Applying cognitive linguistics to teaching English prepositions: an experimental-CALL study. Paper presented at the 13th International Cognitive Linguistics Conference. Northumbria University, Newcastle, UK. Zhong, Chen-Bo, and Geoffrey J. Leonardelli. 2008. Cold and lonely: does social exclusion literally feel cold? Psychological Science 19(9), 838–42. Zhong, Chen-Bo, and Katie Liljenquist. 2006. Washing away your sins: threatened morality and physical cleansing. Science 313, 1451–52. Ziegler, Johannes C., and Usha Goswami. 2005. Reading acquisition, developmental dyslexia, and skilled reading across languages: a psycholinguistic grain size theory. Psychological Bulletin 131(1), 3–29. Ziem, Alexander. 2008. Frames und sprachliches Wissen. Berlin: De Gruyter Mouton. Ziem, Alexander, and Alexander Lasch. 2013. Konstruktionsgrammatik: konzepte und Grundlagen gebrauchsbasierter Ansa¨tze, Germanistische Arbeitshefte 44. Berlin: de Gruyter. Zima, Elisabeth. 2013. Kognition in der Interaktion: aine kognitiv-linguistische und gespra¨chsanalytische Studie dialogischer Resonanz in o¨sterreichischen Parlamentsdebatten. Heidelberg: Universita¨tsverlag Winter. 2014. Gibt es multimodale Konstruktionen? Eine Studie zu [V(motion) in circles] und [all the way from X PREP Y]. Gespra¨chsforschung – OnlineZeitschrift zur verbalen Interaktion 15, 1–48.
References
Zima, Elisabeth, and Geert Broˆne. 2015. Cognitive linguistics and interactional discourse: time to enter into dialogue. Language and Cognition 7 (4), 485–98. Zinken, Jo¨rg. 2007. Discourse metaphors: the link between figurative language and habitual analogies. Cognitive Linguistics 18(3), 445–66. Zipf, George K. 1949. Human behaviour and the principle of least effort. Cambridge, MA: Addison-Wesley. Zlatev, Jordan. 1997. Situated embodiment: studies in the emergence of spatial meaning. Stockholm: Gotab. 2005. What’s in a schema? Bodily mimesis and the grounding of language. In Beate Hampe (ed.), From perception to meaning: image schemas in cognitive linguistics, 313–43. Berlin: De Gruyter Mouton. 2007a. Intersubjectivity, mimetic schemas and the emergence of language. Intelectica 2–3(46–47), 123–52. 2007b. Language, embodiment and mimesis. In Tom Ziemke, Jordan Zlatev, and Roslyn Frank (eds.), Body, language, mind, vol 1: embodiment, 297–337. Berlin: De Gruyter. 2008. The co-evolution of intersubjectivity and bodily mimesis. In Jordan Zlatev, Timothy P. Racine, Chris Sinha, and Esa ltkonen (eds.), The shared mind: perspectives on intersubjectivity, 215–44. Amsterdam: John Benjamins. 2010. Phenomenology and cognitive linguistics. In Shaun Gallagher and Daniel Schmicking (eds.), Handbook of phenomenology and cognitive science, 415–43. Berlin: Springer. 2011. From cognitive to integral linguistics and back again. Intellectica 56 (2), 125–47. 2013. The mimesis hierarchy of semiotic development: five stages of intersubjectivity in children. Public Journal of Semiotics 4, 47–70. 2014. Image schemas, mimetic schemas, and children’s gestures. Cognitive Semiotics 7(1), 3–29. 2016. Turning back to experience in cognitive linguistics via phenomenology. Cognitive Linguistics 27(4), 559–72. Zlatev, Jordan and Johan Blomberg. 2016. Embodied intersubjectivity, nonactual motion expressions and the sedimentation of meaning. Nordic Journal of Linguistics, 39(2): 1–24. Zlatev, Jordan, Johan Blomberg, and Ulf Magnusson. 2012. Metaphors and subjective experience: motion-emotion metaphors in English, Swedish, Bulgarian and Thai. In Ad Foolen, Ulrike Lu¨dtke, Timothy Racine, and Jordan Zlatev (eds.) Moving ourselves, moving others: motion and emotion in consiousness, intersubjectivity and language, 423–50. Amsterdam: Benjamins. Zlatev, Jordan, Timothy P. Racine, Chris Sinha, and Esa Itkonen (eds.). 2008. The shared mind: perspectives on intersubjectivity. Amsterdam: John Benjamins.
815
816
References
Zlatev, Jordan, Elainie Madsen, Sara Lenninger, Tomas Persson, Susan Sayehli, Go¨ran Sonesson, and Joost van der Weijer. 2013. Understanding communicative intentions and semiotic vehicles by children and chimpanzees. Cognitive Development 28, 312–29. Zribi-Hertz, Anne. 1989. Anaphor binding and narrative point of view: English reflexive pronouns in sentence and discourse. Language 65 (4), 695–727. Zwaan, Rolf A. 2004. The immersed experiencer: toward an embodied theory of language comprehension. In Brian H. Ross (ed.), The psychology of learning and motivation, 35–62. New York: Academic Press. 2009. Mental simulation in language comprehension and social cognition. European Journal of Social Psychology 39, 1142–50. Zwaan, Rolf A., and Lawrence J. Taylor. 2006. Seeing, acting, understanding: motor resonance in language comprehension. Journal of Experimental Psychology: General 135, 1–11. Zwaan, Rolf A., Robert A. Stanfield, and Richard H. Yaxley. 2002. Language comprehenders mentally represent the shapes of objects. Psychological Science 13(2), 168–71.
Index
absolute spatial languages, 631 abstract vs. concrete meanings, 20–21, 48, 65, 120, 132, 180, 256, 260, 321, 397, 471, 634 affordances, 17, 176, 184–87, 346 agent, 62–63, 67, 165, 215, 235–36, 296–309, 354, 400–2, 404, 549, 571, 617–18 alignment, 135–37, 139–41, 142–43, 148–50, 158, 171, 186, 265, 271, 279–81, 424, 648 allophone, 219–24 American structuralism, 105 Amondawa, 647, 665–66 analogy, 8, 14, 15, 59, 63, 67–68, 123, 181, 183, 224, 234–35, 361–75, 427–34, 452 anthropology, 535, 543–44, 548 aphasia, 352, 356, 521–22, 524 Arabic, 32, 34, 218, 631, 634, 638, 662, 664 argument role, 297–306 argumentative (nature of communication), 192–205 assembly (of symbolic structures), 267, 269, 272–79 atlas of languages, 368 eWAVE, 548 ATOM (A Theory of Magnitude), 638–39 Atsugewi, 41, 45–6 autism, 71, 356, 359 Aymara, 630–31, 643–45, 649, 657 basic communicative spaces network, 439 Behavioral Profile, 595–96 behaviorism, 495 bilingualism, 469, 665, 667, 682 blending. See Conceptual Blending Theory body movements. See gesture, eye-movements body partitioning, 109, See also signed language, gesture body-part terms, 43, 49, 52, 54, 421 bounded versus unbounded construal, 79–81 brain. See language and the brain brain imaging, 161, 414 British National Corpus, 465–70, 474, 553, 595, 599 Canadian Indigenous Languages and Literacy Development Institiute (CILLDI), 56–58 Canadian raising, 541
categorization, 75, 85, 115–16, 220, 267, 271–73, 322 basic level, 248 Chalcatongo Mixtec, 49 Cherokee, 216 Chinese, 79, 216, 350–51, 356, 372, 388–89, 416, 629–34 CODA (Cognitive Discourse Analysis), 670, 676–81 coercion, 233 cognition, 6, 40, 52, 78, 176, 186–87, 223, 347–48, 375, 380–81, 472, 517–18, 528, 534–37, 578–81 “more general cognition”, 13, 14 and gesture, 100–2, 108, 115–16, 131–33 cognitive load, 131–32 grounded, 518 language versus cognition, 2, 4, 13–16, 17, 19, 21, 37, 46–48, 100, 106, 118–19, 131–34, 385–86, 406, 449–50, 633, 635–50, 666–67, 677–82 social versus non-social, 14–16 cognitive commitment, 2, 10, 155–56, 590 cognitive coordination, 8, 188, 192–93, 204 Cognitive Grammar (CG), 211, 262–83, 326 Cognitive Linguistics, 1–6, 10, 172, 209–13, 330–46, 493–97, 607–12, 621–22, See also methods in Cognitive Linguistics basic tenets, 13–17, 20, 36, 95, 137, 188, 379–84, 533–34, 542–44 history of, 1–4, 19, 38–41, 42, 153–56, 162–69 qualitative methods, 495–97 quantitative turn, 498–514, 590, 598–601, 641–46 role of introspection/interpretation, 209, 225, 246, 495, 500–11, 575, 590 social turn, 191, 493–97 theory versus methodology, 493 versus formal linguistics, 493–94 Cognitive Poetics, 212, 607, 608 cognitive psychology, 7, 367–69, 453, 495, 534, 543–44, 548 cognitive sociolinguistics, 2, 9, 154, 533–48, 603, 604, See also sociolinguistics coherence, 235, 316, 420–21
818
Index cohesion, 417–18 colligation, 252–53, 259, 261, 591, 594, 598 collocation, 57, 83, 252–53, 259, 261, 468, 575, 579, 582, 585–88, 591, 594, 598 collostructional analysis, 504, 598 color, 248–49, 398, 455, 656 common ground, 136, 138–40, 192, 204, 439, 488 communicative strategy, 355–59 complementation constructions, 196–99 Complex Dynamic Systems Theory (CDST), 87 complex metaphor. See Conceptual Metaphor Theory compounds, 232–45, 356, 372, 373, 413 computational linguistics, 141, 502, 573, See also FrameNet, MetaNet Conceptual Blending Theory (CBT), 108, 423–48, 455 blend, 81–82, 97, 162 blends in signed language, 108 completion, 428–29, 434, 442 composition, 373, 428–29, 443–45 compression, 426, 427–28 decompression, 427–28 double grounding, 426, 435–36 double-scope network, 429–34 elaboration, 428–29, 433, 434–35, 440, 442, 445 generic space, 372, 438 input spaces, 81, 427–38, 443 metaphoric blends, 381 mirror network, 430 optimality principles, 434–38 selective projection, 444 simplex network, 429–30 single-scope network, 434 vital relations, 426, 427–28, 433 conceptual mappings, 4–5, 8, 260, 379–84, 434–38, 469–70, 471, 576, See also Conceptual Metaphor Theory, Conceptual Blending Theory, metonymy Conceptual Metaphor Theory (CMT), 181–83, 380, 392–94, 464, See also abstract versus concrete meanings complex metaphor, 381, 398, 405–6, 455, 581, 609, 633 crosslinguistic study of metaphor, 380, 625, 626, 628–34 image mappings, 383 metaphoric mappings, 4–6, 8–9, 20–37, 42, 49, 60, 63–70, 74, 94, 106, 162, 181–87, 260, 320, 364, 379–84, 392–406, 423–48, 449–62, 463–76, 477–89, 626–34, 635–50, 651–68 metaphors of emotion, 76 primary metaphor, 76, 380–82, 397–98, 405–6, 413, 456, 457–60, 464, 576–87, 609–11, 620, 628–30, 632 projection, 380, 428, 443–44 source domain, 8, 22, 364, 392–95, 399, 400, 405, 420, 452–56, 466–69, 470–72, 576 source-target asymmetry, 21–22 systematicity of metaphorical mappings, 384 target domain, 364, 399, 400, 405, 413, 420, 452–56, 466, 469, 478 conceptualization (and language), 2, 7, 10, 189–92, 196–205, 263, 347–49, 392, 641, 643, 647–48, 649, 672
object of/subject of, 189–92, 193–203, 265, 278 concordances, 466–67, 470–75, 592–98 conditional constructions, 81–82, 171, 188, 201–4, 294 in L2 teaching, 81–90 Conflation Hypothesis, 395–96 connotation, 211, 249, 416, 421, 446, 471, 637 consciousness, 44, 174–76, 179, 451, 618 construal, 2, 5, 8, 46–48, 52, 56, 74–87, 111, 124–27, 136–45, 157–58, 165–70, 173, 181, 184–86, 188–204, 212, 257, 263–74, 296, 327, 413, 448, 462, 478–88, 537, 612, 630–34, 636–41, 644–49 configuration, 192, 200, 204 subjective versus objective, 175 constructed dialogue, 348–49 Constructicon, 9, 141, 146, 284, 295, 310–16, 320, 374, 549–73 Construction Grammar (CxG), 1, 4, 8, 19, 76–77, 135, 141–43, 145–47, 212–13, 232, 233, 263, 284–88, 303–9, 310–29, 330–31, 373–74, 562, 565–73, 600–1, See also constructions, Constructicon, Construction grammars Construction grammars Berkeley Construction Grammar (BCG), 310, 315, 321, 323–24, 572 Cognitive Construction Grammar (CCG), 310, 315, 321, 326–29 Embodied Construction Grammar (ECG), 4, 212, 310, 321 Fluid Construction Grammar (FCG), 310, 321, 324–26 Head-Driven Phrase-Structure Grammar (HPSG), 323 Radical Construction Grammar (RCG), 315, 321, 326 Sign-Based Construction Grammar (SBCG), 307, 315, 323 constructional compositionality, 163 constructional links, 319 instance links, 320 metaphor links, 76, 320 polysemy links, 320 subpart links, 320 constructionalization, 361, 373–75 constructions Argument Structure Constructions, 160, 295–309, 562–71, 604 Causative with construction, 300–3 Caused Motion construction, 76, 300–3, 320 Ditransitive construction, 300, 303, 304, 307, 308, 313, 315, 319, 320, 327 Intransitive Motion construction, 299, 302, 320 Resultative construction, 69, 285, 296–99, 305–7, 312, 320, 571 Transitive construction, 65–68, 299–305, 307–9, 313–16, 319–20, 326–27 Way construction, 300, 305, 598 Comparative Correlative construction, 294, 295 Covariational-Conditional construction, 294 idiomatic constructions, 565–66 in First Language Acquisition, 59–72 item-based, 61 partially productive, 61–64 slot-and-frame, 61, 67, 70
Index
intersubjective, 7, 370 Intransitive construction, 65, 66, 326 Let-alone construction, 566 macro-construction, 235, 315 meso-construction, 235, 316 micro-construction, 235, 315, 319, 374 modification constructions, 609, 610 multimodal constructions, 147 Negation construction, 566 Passive construction, 66, 306–9 Subject-Predicate construction, 566 Subject-Verb Agreement construction, 325 Tense-Aspect construction, 316 Think of X as Y, 609–12, 621 Transitive construction, 299, 326 Verb-Class-Specific construction, 304 Verb-Specific construction, 319 Vocative-cum-Imperative construction, 616 WH-interrogative construction, 293 X is like Y, 609 X is the new Y, 610–12 constructivist approach, 59–61, 62, 70–72, See also First Language Acquisition, generativist versus constructionist approach continuum of symbolic structures, 8, 267 conventionalization, 16–17, 133, 146, 236, 259 conventional gestures, 640 linguistic convention, 646 convergence zones. See language in the brain conversation frame, 358, See also fictive interaction Conversational Analysis (CA), 75, 88 co-occurrence, 597 Cora, 50 corpora (language corpora), 499, 501–2, 513 examples of, 473–76 types of corpora, 465 corpus linguistics, 3, 6, 9, 463–76, 590–606 keywords, 467 study of metaphor, 464–73 correlation, 1 co-speech gesture. See gesture Critical Discourse Analysis, 677 cross-linguistic comparison, 83–86, 321, 386–89, 564–65, 641–50, 658–61 cultural relativity, 471 culture. See language and culture data Big Data, 95–96, 513 collection methods, 2, 5, 90, 533, 536, 679–80, 683 elicitation, 185, 254, 501, 546, 643, 665 from minority languages, 40, 41, 51, 53 in FrameNet, 555–64 linguistic data, 2, 19, 54–55, 217, 502, 508, 511–14, 575, 588, 590, 594, 651, 679, 681 naturalness of, 591 nonlinguistic data, 19 public archiving of, 512 representativeness of, 546 decomposability of words, 210 deixis, 97, 121, 146, 162, 166–71, 265, 439, 551, 614–16, 628, 631, 642–49 Dena’ina, 602 Descartes, 100–2 dialogic syntax, 142, 282
dialogue, 17, 136, 138, 139, 142, 147–48, 339, 348–52, 359, 621, 678–79 dictionary definition (of word meaning), 551, 553 Direct Speech. See Speech and Thought Representation constructions discourse, 158–59, 262–83, 425–26, 439, 470 discourse analysis, 2–4, 330–31, 607, 676–83 discourse community, 75, 83, 87 discourse organization, 75, 348 discourse particles, 58, 331, 335–41, 346 Distributed Little Red Hen Lab, 96 diversity. See linguistic diversity documentation. See language documentation duration, 653–59, 660–61, 665–68 Dutch, 25–29, 65, 200–1, 231–44, 436–38, 507, 603, 636, 662 Dyirbal, 51 dynamicity of language, 272 elaboration site, 269 ellipsis, 282 embodied cognition, 155, 160, 186, 330, 518 embodied metaphor. See metaphor embodied simulation. See simulation embodiment, 4, 7, 77–78, 159–60, 172–87, 225, 360, 449–62, 517–20, 608, 618–21 and gesture, 125 body memory, 175–87 body schema, 174–75, 177 definition of, 225, 450–51 embodied experience, 29–30, 78, 221, 247, 260, 394–95, 405–6, 452–60, 470, 575–76, 611, 618–21 emotional valence. See space in metaphors emotions, 22, 43, 107, 136–37, 173–74, 182, 466, 576, 595, 612, 616–20 empirical methods, 154, 155, 513, 574–77, 592, 670, 682 encyclopedic view of meaning, 263, 335, 372 endangered languages of North America, 38–51 entrenchment, 48, 52, 57, 68–69, 74, 83, 271–74, 315–22, 548, 590, 593, 601, 604 evidentiality, 44, 51, 162, 355, 357 experience, 1, 20, 106, 107, 116, 348, 394–95, 398, 400, 406, 618–21, 622, 633, 662 experimental evidence, 22, 413, 651–68 Extended Invariance Principle, 404–5, See also Invariance Principle, Conceptual Metaphor Theory eye gaze, 4, 61, 145, 148, 159, 342 eye-movements. See eye-tracking eye-tracking, 148, 414, 483, 601, 640–41 fictive direct speech, 355–57 fictive interaction, 8, 9, 347–60, 426, 430, 438 fictive motion, 182–86, 477–89 fictive speech act, 355–60 fictivity, 100, 212, 348, 360 field linguistics versus Cognitive Linguistics, 50, 55 figurative language, 47, 164, 379, 417, 474–76, 479, 487–88, 576, 610 Figure/Ground, 45–46, 56, 85–86, 190–92, 627–28, 632 Finnish, 300, 479, 597 First Language Acquisition, 59 competition between cues and schemas, 67, 70 cues, 62 errors (in language acquisition), 67–70
819
820
Index First Language Acquisition (cont.) argument structure errors, 68–69 auxiliary doubling errors, 69 case marking errors, 70 negation errors, 69 overregularization errors, 60–64, 68 generativist versus constructionist approach, 59, 60 input (to children), 59 novel words versus regular words, 67 word order in acquisition, 62, 70 focalization, 157, 162, 169–70, 351–52, 360 force dynamics, 77, 257 form-meaning mismatch, 238 form-meaning pairs, 4–5, 60, 83, 86, 89, 141–47, 210–13, 237–40, 262–67, 284–92, 309, 312–13, 322–26, 329, 337, 341, 374, 470, 566 frame, 331–46, 379–84, 429–32, 438–39, See also metonymy definition of, 551 event frames, 564 frame roles and fillers, 401 hierarchy, 563 in metaphor, 400–6, 633 relations, 553, 557, 563, 567, 571, 572, 581 versus domain, 582, 585 Frame Semantics, 297, 324–29, 345, 382, 549–65, 571 FrameNet, 96, 324, 341–42, 402, 549–73, 577–84 annotation, 554 as a linguistic tool, 563–64 German FrameNet, 564 Japanese FrameNet, 564 Spanish FrameNet, 564 Swedish FrameNet, 564 Free Indirect Discourse. See Speech and Thought Representation constructions frequency, 8, 252–53, 313–16, 367–69, 591–605 absolute frequency, 66 frequency effects in L1 acquisition, 65–67 frequency effects in L2 learning, 83–88 relative frequency, 66–67, 510, 596, 602 token frequency, 65–67, 315, 327, 374, 541, 593–96 type frequency, 65–66, 235, 315–20, 374, 540–41, 592–93, 604–5 functional relationships, 676 future, 627–31, 637–46, 652–53, 657–58, 662–68 gaze. See eye gaze gender, 51, 101, 210, 218, 535–37 generative grammar, 230, 384 generative grammar versus Cognitive Linguistics, 38, 42, 215 genitive, 65, 163–64, 171, 241–44, 252, 566, See also possessive form as a constructional marker, 244 as a viewpoint marker, 164, 171 genre, 346, 417, 465, 472 German, 62, 84–85, 146, 150, 173, 232–35, 240, 244, 257, 284, 305–6, 364–68, 373, 469–70, 494, 564, 571, 577, 595, 606, 640–41 gesture, 3, 4, 7, 95, 111–12, 118–34, 145–51, 158–61, 165, 383, 661 and Peirce’s classification of signs, 120 and spatial concepts, 120
beat gestures, 123, 130 cognitive functions of, 133 communicative functions of, 121, 133 dominant versus nondominant hand, 35 emblems, 120–22 gestural theory of language origin, 111 gesture-speech integration, 129 head and face gestures, 119 iconic gesture, 121–23, 130, 180 metaphoric gesture, 32–35, 120, 121, 127–28, 625–32 pointing, 121, 127, 129, 130, 148, 159, 178 versus signed language, 99, 100–2 gradient morphology, 115, 210, 540 grammaticalization, 8, 42, 44, 48–50, 53–56, 108, 112, 356, 358, 361, 369–71, 375, 539 in signed language, 112–14 Greek, 226, 232, 237, 475, 638, 646, 658–59, 660–61 ground. See Figure/Ground grounding, 439–40 experiential grounding, 186, 381–82, 608, 618–21 grounding spaces, 426, 439, 446–47 headshake. See gesture Hebrew, 631, 643, 662 Hopi, 41, 47 humor, 135, 138–39, 354, 407, 415, 416, 419, 433–38, 467, 472, 611, 613, 619 Hungarian, 388 Hupa, 57 iconicity, 366–67 cognitive iconicity, 109–10 in signed language, 103 Idealized Cognitive Models (ICMs), 257 idiomaticity, 47, 600 image schema, 14, 50, 157, 162, 165, 179–81, 186, 214, 398–400, 404–6, 452–54, 460–62, 610–12, See also Invariance Principle imagery, 15, 46–50, 56, 77–78, 129, 265, 406, 470, 620–21 imperfective aspect, 81, 481 indigenous languages (ILs), 38–58 Indirect Speech. See Speech and Thought Representation constructions inflectional paradigm, 243 information structure, 44, 75, 277, 282, 306, 351, 358, 367 interaction. See also multimodality, multimodal interaction face-to-face interaction, 51, 135–38, 141, 143, 145, 348 human-computer interaction, 338–39 human-robot interaction, 342–46 interactional alignment, 135, 139–41 parent-child interaction, 344 interdisciplinarity, 3, 9–10, 155, 183, 497, 606 intersubjectivity, 7, 172–87, 188–205 constructions of, 193–204 embodied, 176–79 intersubjective coordination, 201 Invariance Principle, 399–400, See also Conceptual Metaphor Theory, Extended Invariance Principle irony, 96, 138–39, 168, 407, 415, 472
Index
Japanese, 388, 410, 419, 479, 564, 577, 634 joint action hypothesis, 136, 138 joint attention, 97, 161 Korean, 216 Kuuk Thaayorre, 40, 645–50 language and culture, 2, 666–67 language and the brain, 2, 515–32 brain activation, 484 brain anatomy and language, 16, 179 cerebral hemispheres, 516–28 left hemisphere, 516–28 right hemisphere, 516, 527–28 convergence zones, 519–20, 526 dual stream models, 524 modular approach, 522 language change, 24, 48–50, 56, 236, 283, 361–75, 526–27, 540, 626, See also grammaticalization, lexicalization, constructionalization language corpora. See corpora, corpus linguistics language documentation, 54–55 language learning, 59–72, 73–90, 505–10, See also First Language Acquisition, Second Language Acquisition statistical learning, 65, 71, 228, 603 language variation, 367 Belgian and Netherlandic Dutch, 538 Canadian English, 541 Lancashire English, 539–48 Liverpool English, 540 New York English, 535 North of England English, 547–48 social variables, 534–39, 546 Sri Lankan English, 548 Yorkshire English, 547 language, design features of, 102, 235 arbitrariness, 101–3 discreteness, 102, 115, 598 duality of patterning, 102–3 lexical bundles, 600 lexical entry, 289, 556–64 lexical semantics, 246–61 network approach, 255 word versus world approaches, 247–50 words versus concepts, 254–60 lexical-constructional approach, 562 lexicalization, 8, 41, 46, 49, 56–57, 112, 257, 315, 361, 372–73, 375 lexicography, 258, See also FrameNet, lexicographic databases, 553–73, 577 lexicon, 8, 41, 50, 53–55 hierarchical lexicon, 233–37 lexicon-grammar continuity, 312–13, 372–73 linguistic, 22–37 linguistic diversity, 7, 40, 54, 506, 636 linguistic relativity, 41, 46, 47–48, 54 literacy, 216, 639, 643, 645, 646, 647–48, 649, 663 literal versus non-literal, 89, 481, 484, 486, 576, 673 Luisen˜o, 46 material anchor, 426–29, 435 mathematical concepts, 126, 129–33 meaning emergence processes, 4 language versus gesture, 95
meme, Internet, 158, 161, 167, 574 mental simulation. See simulation Mental Spaces Theory, 81–82, 615 alternative spaces, 195, 205 in signed language, 100, 105, 108–9 MetaNet, 9, 570, 574–89 metaphor, 385–406 and gesture, 145 conceptual metaphor. See Conceptual Metaphor Theory embodied metaphor, 451–52 for pain, 473–76 in discourse, 575 in signed language, 106–8 mental metaphor versus linguistic metaphor, 28 metaphor identification, 465, 575–89 metaphoric cascade, 585 metaphoric entailment, 578, 581 metaphtonomy, 413 methods in Cognitive Linguistics computational methods, 5, 575 corpus linguistics, 591–94 experimental methods, 5, 59, 534 qualitative methods, 495–97 quantitative methods, 503–11 metonymy, 407–22 categorial metonymy, 382, 428 conventional metonymy, 414–15 experimental studies of, 413 frame metonymy, 382, 615 functions of, 415–21 in language teaching, 409 in signed language, 107, 108 types of, 411–14 versus metaphor, 408–14 mimesis, 137, 176–81 mimetic schema, 173, 178–87 modal verbs, 76–77, 199–201, 365 modality, 103, 188, 199–201 modular approach. See language in the brain monkeys, 161, 638, 647, 654–56 monologue, 348, 350, 617 morphology Construction Morphology (CxM), 231, 236, 243–45 construction-dependent, 241–43 morphological schemas as constructional schemas, 8, 229–45 morphology-syntax boundary, 242 non-concatenative morphology, 239, 245 productivity of schemas, 234–42 storage of morphological schemas, 244 morphosyntactic features, 17, 231, 237 motion, 185–86, 518, 523–25 fictive motion, 477–89 metaphorical motion, 477–79, 486, 488 motivation, 180–87 moving window metaphor, 279–80 multimodality, 3, 4–6, 7, 58, 135–56, 615–18 multimodal construction grammar, 145–47 multimodal interaction, 135–56 multimodal usage events, 116 Multiple Word Units (MWU’s), 601, See also lexical bundles Na´huatl, 46, 50 narrative, 108, 157, 165–66, 170, 179, 615–16, 618–20, 682
821
822
Index narrator, 613–15 nativism, 54, 529–31 negation, 188, 193–96, 439–40 network, 162–69, 429–34 multiple inheritance network, 316 of concepts, 255 of mental spaces, 427 of viewpoints, 166 taxonomic network, 310, 312, 313–20, 322 neural plasticity, 529–31 neurons, 161, 364, 516–17, 520, 524–30 non-actual motion. See fictive motion number. See space in metaphors Onondaga, 41, 44 orality, 40, 41, 57, 347, 350–51, 356, 539–40, 570 paradigmatic relations, 238–39, 243–45, 250–52 participant role, 264, 297–303, 562, 570 past, 627–31, 637–45, 649, 652–53, 657–59, 664–68 path, 477–81, 487–89 perspective. See viewpoint phenomenology, 173–76 phonaesthemes, 602–3 phoneme and distinctive features, 220–22 as a prototype category, 214, 221 phonology, 214–28 cognitive phonology, 110–11 exemplar theory of, 221 phonological rules versus processes, 215 stored units, 216 pitch. See space in metaphors polysemy, 4, 78–79, 258–59 polysynthetic languages, 38, 46, 57 Portuguese, 358, 359, 577, 665 possessive, 179, 252, 303, 429, 539–40, 544, 568 alienable versus inalienable possession, 544–45 pragmatics, 5, 8 pragmatics in Cognitive Linguistics, 3, 14, 330–46 primary metaphor. See Conceptual Metaphor Theory primary scenes, 165, 380, 397–99, 628, See also primary metaphor, Conceptual Metaphor Theory primates, 360, 517, 519 productivity, 235, 236, 242 profiling, 7, 137–39, 173, 265–71, 277, 301, 403–4, 562–65, 570 pronouns, 69, 104, 157, 182, 252, 268, 358, 545, 613 prosody, 44, 58, 93, 95–96, 104, 114, 143–44, 148–51, 231, 274–82, 336, 447, 461 protolanguage, 177 prototype, prototypicality effects, 220 psycholinguistics, 14, 52, 139, 311, 453, 590, 606, 680 radial category network, 97 reaction time, 29, 32, 33, 35, 160, 414, 436, 505, 593, 662, 681 recipes, 331–35 reduction, 539–45, 547, 593, 601 reference, 42, 247, 379, 424
reference point, 56, 212, 265, 538, 627–28 register, 236, 332, 417, 465, 470–72, 508, 577 regression analysis, 506, 536, 597, 601 relative spatial languages, 631 relevance, 671, 680 rhythm, 121, 131, 225–27 role-value mapping, 435 Russian, 97, 226, 507–11, 595–96 scanning, 184, 185, 212, 268, 274, 480, 484–88 science cognitive science, 1, 5, 7, 8, 37, 449, 456–58, 462 natural sciences versus human sciences, 494–97 neuroscience, 9, 456–58 Second Cognitive Revolution, 493 Second Language Acquisition/Learning, 73–90 as re-constructing a language, 82 categorization and re-categorization, 75 comparative corpus analysis in L2 teaching, 82–86 role of embodiment, 77–78 role of grammar instruction, 74 role of input, 75 semantic memory, 178, 515, 517–19 semantic ontology, 577 semantic primes, 258 semantic relations, 211 semantic roles, 549–52, 563 semantics, 188–95 truth-conditional semantics, 193, 550 semantics versus pragmatics, 3, See also pragmatics in Cognitive Linguistics Seneca, 41, 44 shared mind approach, 176 signed language, 99–117 articulators, 104, 109–10 as a natural language, 99–100 emergence from gesture, 100–2 phonology, 102–5 signs as language-gesture fusions, 115–16 sociolinguistics of sign language, 104–5 signed languages American Sign Language, 99, 106, 530, 649 British Sign Language, 99 Catalan Sign Language, 112 French Sign Language, 99, 105 simile, 465, 466–67, 472–76, 608–12, 621 simulation, embodied, 157, 160, 458–60, 461–62, 470, 474, 486 Sinhala, 548 situatedness, 331, 345 sociolinguistics, 105 sociolinguistic ‘monitor’, 542–43 versus cognitive linguistics, 534–43 wave model, 535 socio-pragmatic skills, 60, 71 space in language, 625–34, See also absolute spatial languages, relative spatial languages conceptualization of, 672 locatum versus relatum, 670–72 projective spatial terms, 671–72 spatial reference systems absolute, 671 intrinsic, 670–72
Index
relative, 671 space in metaphors, 20–31, See also timeline in signed language, 106–8 left-right, 22, 28–32, 35–37 number, 22–23, 638 pitch, 24–25, 26–27 time, 22 space–time mappings, 635–50, See also timeline asymmetry, 653–57 Ego-based versus Time-based, 627–31 experimental evidence, 646, 656–57 in discourse, 669–83 in gesture, 639–40 Moving Ego versus Moving Time, 626–31 Spanish, 79, 84–90, 201, 387, 409, 479, 488, 564, 571, 602, 662, 664 Speech and Thought Representation constructions, 157, 162–63, 199, 613 Direct Speech, 162, 168, 170, 348, 355–59, 440 Free Indirect Discourse, 162, 163, 355–59 Indirect Speech, 192, 355–59, 421 stance, 51, 58, 82, 93–95, 135–37, 150–52, 164, 171, 194–99, 203, 496–97, See also viewpoint statistical models, 498, 502–11, 512, 513 statistical software, 96, 498, 501–2 storage, 215 style, 94, 175, 461, 537, 539, 620 stylistic strategies, 621–22 subjectification, 200, 370 subjectivity, 188–93, 365 Sumerian, 216 Swedish, 577 syllable, 215, 225–27 syntagmatic relations, 250, 252–53 syntax-lexis interface, 600 Tamil, 40, 548 Tarascan, 41–43, 48 Temporal Focus Hypothesis, 664–65 temporal terms (non-metaphorical), 674–76, 682–83 temporal adverbials, 507 temporal versus causal meanings, 661 tense past tense, 65, 67, 146, 163, 267 past tense in First Language Acquisition, 59 text analysis, 608, 622 theory of mind, 137, 161, 176, 443, 523 time metaphors. See space–time mappings time, concept of, 672, See also space–time mappings, Temporal Focus Hypothesis static versus dynamic models of time, 632 timeline, 32–33, 37, 127, 632, 635, 637, 645–50, 652–53, 661–68 ego-centric versus geo-centric, 642
front-to-back timeline, 627–32, 639, 643–44, 649, 652, 658, 661, See also sagittal timeline lateral timeline, 32, 643, 649 left-to-right timeline, 637, 643, 648–49, See also lateral timeline non-linear, 645–46 sagittal timeline, 643–44, 648 trajector versus landmark, 265, 270, 282, 372, 632 translation, 249, 258 typology, 539, 544–48 typological universals, 544–46 Tzeltal, 39, 644–47 uniqueness of human language, 15 universal grammar, 217, 227, 515 universals, 214–18, 225, 227 usage event, 115–16, 135–37, 143–47, 152–55, 271–72, 276, 500–1 usage-based approach, 2, 7, 13, 17, 48, 73, 88, 105–6, 110, 211–13, 235, 271, 310, 315, 318, 326–27, 505, 531, 590, 601 diachronic frame, 14, 17 enchronic frame, 14, 17 microgenetic frame, 14, 17 ontogenetic frame, 14, 17 phylogenetic frame, 17 Uto-Aztecan languages, 41, 42, 46–47, 50 vantage point, 124, 264, 440 variation. See language variation viewing arrangement, 45, 157, 189, 191, 264 viewpoint, 157–71, 627 and gesture, 124 character versus observer viewpoint (in gesture), 124, 158 conceptual viewpoint, 171 decompression for viewpoint, 166 deictic viewpoint, 166, 170–71, 615 discourse viewpoint, 166–69 emotional viewpoint, 171 epistemic viewpoint, 171 experiential viewpoint, 171 viewpoint compression, 165, 166 viewpoint network, 157, 169 visual metaphor, 389–91 Whorfian hypothesis. See linguistic relativity, cultural relativity Wolof, 40, 627 word association, 254 writing systems, 216, 225, 356, 629–33, 648, 662, 665 Ye´lıˆ Dnye, 40, 646 Yucatec Mayan, 125, 640 Yupno, 644–49
823