Sie sind auf Seite 1von 418

Cognitive Linguistics in Action

Applications of Cognitive Linguistics 14

Editors Gitte Kristiansen Francisco J. Ruiz de Mendoza Ibanez Honorary editor Rene Dirven

De Gruyter Mouton

Cognitive Linguistics in Action

From Theory to Application and Back

Edited by

Elzbieta Tabakowska Micha Choinski ukasz Wiraszka

De Gruyter Mouton

ISBN 978-3-11-020581-7 e-ISBN 978-3-11-022609-6 ISSN 1861-4078

Library of Congress Cataloging-in-Publication Data Cognitive linguistics in action : from theory to application and back / edited by Elzbieta Tabakowska, Micha Choinski, ukasz Wiraszka. p. cm. Includes bibliographical references and index. ISBN 978-3-11-020581-7 (alk. paper) 1. Cognitive grammar. 2. Language and languages Study and teaching. I. Tabakowska, Elzbieta. II. Choinski, Micha, 1983 III. Wiraszka, ukasz, 1982 P165.C644 2010 371.3315 dc22 2010032908

Bibliographic information published by the Deutsche Nationalbibliothek The Deutsche Nationalbibliothek lists this publication in the Deutsche Nationalbibliografie; detailed bibliographic data are available in the Internet at 2010 Walter de Gruyter GmbH & Co. KG, Berlin/New York Typesetting: RoyalStandard, Hong Kong Printing: Hubert & Co. GmbH & Co. KG, Gttingen Printed on acid-free paper Printed in Germany

Table of Contents
Introduction Theory informing applications, applications informing theory . . . . . Elzbieta Tabakowska, Micha Choinski and ukasz Wiraszka Part I. From loop to cycle Looking back at thirty years of Cognitive Linguistics . . . . . . . . . . . Rene Dirven and Francisco Jose Ruiz de Mendoza Ibanez Recontextualizing Grammar: Underlying trends in thirty years of Cognitive Linguistics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Dirk Geeraerts How theory informs application and how application informs theory. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Laura Janda Part II. The context for prototypes Why not? Prototypes and blocking of sux shift in Russian verbs. . . Tore Nesset Prototype-based taxonomy of idiomatic expressions . . . . . . . . . . . . Esa Penttila Part III. Contexts for Cognitive Grammar Control and the mind/body duality: Knowing vs. eecting . . . . . . . Ronald W. Langacker A discourse perspective to nominal reference-point constructions. . . Peter Willemse Lexicalizing indirect path: Focus on Finnish motion verbs . . . . . . . Jari Sivonen A cognitive approach to parenthetical speech . . . . . . . . . . . . . . . . . Jaakko Leino Part IV. The pragmatic context Using RST to analyze subjectivity in text and talk . . . . . . . . . . . . . Kirsten Vis, Wilbert Spooren and Jose Sanders 291 163 209 241 273 123 145 11 1




Table of contents

Typology meets witness narratives and memory: Theory and practice entwined in Cognitive Linguistics . . . . . . . . . . . . . . . . . . . . . . . . . . 317 Luna Filipovic Part V. The social and cultural contexts for CMT and CI Cross-cultural variation in idiomatic expression: Insights from Cognitive Metaphor Theory and implication for Translation Studies. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Diane Ponterotto Mundane Transcendence? Conceptualizations of faith in prosperity theology. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Magorzata Pasicka



From god is a father to god is a friend: Conceptual integration in metaphors for God in Christian discourse . . . . . . . . . . . . . . . . . . . 387 Aleksander Gomola Subject index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 409

Introduction: Theory informing applications, applications informing theory Elzbieta Tabakowska, Micha Choinski and ukasz Wiraszka
The present volume brings together a selection of papers delivered during the 10th International Cognitive Linguistics Conference, which was held in Krakow (Poland), from July 15 to July 20, 2007. This was, indeed, a jubilee year for Cognitive Linguistics: thirty years after various strands of linguistics focusing upon the relation between language and mind began to develop their theoretical frameworks in a cognitive direction, building up during the 1980s uniform cognitive foundations for linguistic description, and twenty years after the appearance of the rst two milestones: Lako s Women, Fire and Dangerous Things and Langackers Foundations of Cognitive Grammar. Eighteen years after the rst conference, organized by Rene Dirven in Duisburg in 1989, the event in Poland was attended by over ve hundred scholars from 42 countries. The conference was organized by the Polish Cognitive Linguistics Association in cooperation with the Jagiellonian University of Krakow. Its leading theme, Cognitive Linguistics in Action: From Theory to Application and Back, gave the title to the present volume. During the ve days of the Conference the participants presented results of their research devoted to a wide array of linguistic phenomena of interest to Cognitive Linguistics. Despite the great variety of scholarly interests and the problems addressed, all the researchers whether the six plenary speakers (Laura Janda, Leonard Talmy, Ronald W. Langacker, Francisco J. Ruiz de Mendoza Ibanez, Chris Sinha and Dirk Geeraerts), the hundreds of speakers in general and theme sessions, or the forty authors of poster presentations shared at least one common objective: to illustrate, as Laura Janda puts it in her chapter below, how theory informs application and how application informs theory. Traditionally, jubilee was a year of celebration and forgiveness. The present volume was conceived as part of the celebration, bearing testimony to the event and its key theme. But at the same time the editors feel obliged to ask their readers forgiveness: because of space limitation, the book could not encompass all the quality papers, lectures and presen-

Elzbieta Tabakowska, Micha Choinski and ukasz Wiraszka

tations oered as contributions to the 10th ICLC. A choice had to be made so as to present the readers with a more or less coherent volume, which would at the same time provide a varied and stimulating reading. As a result, it was decided to shift the focus towards such aspects of the relation between theory and application in Cognitive Linguistics that have not been comprehensively dealt with in other recent publications. Therefore the articles oered here do not specically address such topics as second/foreign language learning and pedagogy (Putz, Niemeier, and Dirven 2001; Boers and Linstromberg 2008), language acquisition (Tomasello 2003; Dabrowska 2004), sign language (Emmorey 2002; Janzen 2005), nonverbal communication (McNeill 1992, 2000, 2005) or computational and corpus-based linguistics (Gries and Stefanowitsch 2006; Stefanowitsch and Gries 2006). Naturally, individual papers in the book address a wide array of topics, dealing with various elds of application, as dened in the introduction to the rst volume in the ACL series (Kristiansen, Achard, Dirven and Ruiz de Mendoza Ibanez 2006). Some more general in scope indicate new directions and open new perspectives, others originally presented at the Conference as section papers dealing with ner-grained issues demonstrate ways in which application of the theory to new data and using new methodologies lead to renement, further development or modication of the theoretical framework. All provide illustration for the recent empirical turn in Cognitive Linguistics, which is marked by the consideration and description of large amounts of authentic data, facilitated due to increased use of electronic corpora, as well as by the increasingly interdisciplinary approach. The ill-famed linguistic artifacts, which linguists of various persuasions often used to create to illustrate and support their hypotheses about what language is and how it works, are being replaced by large linguistic corpora, setting new standards for empirical research. The impact that the theoretical framework of Cognitive Linguistics has been making on the neighbouring elds of study is evident; it is often reected by the very titles of books written within new interdisciplinary disciplines, where the adjective cognitive comes to implicate signicant modications that a given eld has been undergoing, following the adaptation of CL theoretical models, methodologies and principles of practical application. Thus we have witnessed the emergence of Cognitive Sociolinguistics (Kristiansen and Dirven 2008), Cognitive Poetics (Stockwell 2002; Gavins and Steen 2003; Brone and Vandaele 2009), Cognitive Stylistics (Semino and Culpeper 2002) and Cognitive Translation Theory (Tabakowska 1993; Gutt 2000). Cognitive principles of categorization as reected in linguistic

Introduction: Theory informing applications, applications informing theory

labels put on entities and categories, as well as the cognitive model of grammar of the natural language, are now used, and developed, in elds seemingly very remote from that of linguistics proper, such as archaeology, cultural anthropology or theology. Many further vistas lie open. By bringing up a collection of papers which point out these new directions the editors hope to provide scholars working in the eld of linguistics, as well as in other elds within the humanities, with a survey of recent trends which constitute the empirical turn in the Cognitive Linguistics of today. The contributions in the rst section of the volume, written by eminent scholars, whose position within the eld of Cognitive Linguistics has been rmly established, touch upon issues of continuing relevance to the discipline and introduce thematic areas covered in the next four sections. There, the contributions come mainly from young scholars, whose research illustrates various ways leading to the implementation of the cycle through dierent forms of contextualization, either presenting descriptive analyses that lead to theoretical adjustments and amendments, or widening the eld of possible applications, e.g. to include theological or metaphysical discourse. Part I of the volume (From loop to cycle) begins with a comprehen sive survey by Rene Dirven and Francisco Jose Ruiz de Mendoza Ibanez, who review relevant developments that took place over the thirty years of the history of Cognitive Linguistics, putting special emphasis on the relation of grammar to cognition and presenting the distinctive features of CL, in particular the non-modular conception of language and conceptual prominence in grammar. The authors discuss the notion of embodiment, showing the philosophical grounding of embodied realism in phenomenology, and demonstrate the role of categorization in the emergence of the lexicon-grammar continuum, focusing upon the crucial notions of construal, perspective, and compositionality. In the contrastive sections of the chapter, Dirven and Mendoza Ibanez elaborate on the areas of overlap and contrast between Cognitive Grammar and the various versions of construction grammars. In the second part of their survey, they critically review the issue of polysemy, arguing for a solution based on complementing qualitative and quantitative analytical techniques. Finally, they discuss blending as a pervasive cognitive phenomenon and a psychologically plausible explanation for non-compositional aspects of grammar, pragmatics and discourse. In Dirk Geeraerts chapter, historical trends in the history of Cognitive Linguistics are also a central issue. Arguing for recontextualization in

Elzbieta Tabakowska, Micha Choinski and ukasz Wiraszka

CL methodology, the author looks for patterns underlying the developments in CL descriptions over the last thirty years, claiming that their discovery and re-thinking will have signicant implications for the directions and ways which CL should follow in the future. Posing the question about the common trend underlying the development of various strands within CL, the author argues that what keeps together the vast breadth of cognitive approaches in linguistic description is their tendency to recontextualize the study of grammar. Laura Jandas contribution demonstrates the working of the full cycle: from theory through application to further renement of theory. The author presents an example of this cycle of mutual interaction by showing how the ndings of CL studies of case in Slavic languages may be applied to language pedagogy and how this application in turn inuences the theory. Motivated by the need for teaching materials, Jandas paper gives an illustration of a scholarly agenda in which practical needs lead to a comprehensive investigation of a linguistic phenomenon and, ultimately, to uncovering new territory for theoretical exploration. Part II (The context for prototypes) focuses on the fundamental notion of prototypicality. It presents two contextualizing applications of prototypes in linguistic analysis. Tore Nesset demonstrates how the notions of prototypical and peripheral members of radial categories may be applied to explain certain changes in Russian verb morphology. He investigates the process which results in Russian verbs with the sux /a/ shifting to the group with the sux /aj/ due to the interaction of the prototypes of source and target verb categories. Esa Penttila demonstrates how a prototype-based approach can be used to investigate characteristics of idiomatic expressions and to create a taxonomy which takes into account the entire complexity of the problem. In his analysis, the author highlights various relationships between dierent subclasses of idioms, arguing for viewing the phenomenon of idiomaticity in terms of a continuum rather than in terms of categorial notions. Part III (Contexts for Cognitive Grammar) opens with a contribution by Ronald W. Langacker, who explores the problem of the mind/ body duality and its linguistic manifestations in the domain of English complementation, arguing that a more unied account of complementation in language can be achieved by further rening the concept. Langacker claims that the body/mind duality is a natural product of embodied cognition and thus has a rm experiential basis. As he demonstrates, it is this duality, which is systematically manifested in linguistic expressions, that underlies the distinction between what he denes

Introduction: Theory informing applications, applications informing theory

as eective and epistemic relationships on the conceptual plane, with the former involving the occurrence of events, and the latter pertaining to the validity of knowledge. The progression from eective to epistemic relationships is shown to be correlated with dierent degrees of structural complexity of linguistic expressions, from lexical predicates, to direct complementation, to innitival and clausal complementation. Moreover, the author suggests that the opposition between eective and epistemic relationships can also be manifested within each of these levels, in dierent ways, depending on the inherent properties of the particular level. In conclusion, Langacker contends that the recognition of the systematicity of the opposition between the two types of relationships allows for signicant unication of the account of dierent types of complementation. Peter Willemse takes up the problem of the description of nominal reference-point constructions, paying specic attention to the way in which such constructions are embedded referentially in the discourse context, and the way in which the reference-point relation may interact with information in the surrounding discourse. Following a corpus study of English possessive noun phrases, which as the author claims should be studied in extensive discourse contexts, Willemse argues for distinguishing a continuum of discourse statuses of these phrases rather than adopting binary distinctions. The paper concludes with implications for both the description itself and the theory of English nominal reference-point constructions. The author of the next contribution, Jari Sivonen, focuses on an analysis of path as part of the semantics of Finnish motion verbs. An essay in Cognitive Semantics, Sivonens paper presents the results of his investigation of fteen lexemes: the most frequently used motion verbs which prole indirect path. He distinguishes between ve groups of verbs with respect to their implied path shapes (Paths of Irregular Shape, Paths of Regular Shape, Paths of a Single Turn, the Back and Forth Paths and the Crossing Shape Paths). Once again, Sivonens paper demonstrates that recognizing the systematicity of seemingly disparate linguistic phenomena enhances uniformity of linguistic descriptions, and at the same time provides renement of existing theories (in the case of motion verbs, Talmys theory of verb-framed and satellite-framed languages). Jakko Leino demonstrates that Construction Grammar is a useful tool for a researcher aiming at bringing together issues of spoken language and syntactic theory. To this end, the article presents an analysis of two parenthetical expression types characteristic of spoken Finnish and elaborates on the role of expression types in shaping the notion of construction, the organization of speech, and the relation between the two. In conclusion,

Elzbieta Tabakowska, Micha Choinski and ukasz Wiraszka

the author claims that models of grammatical constructions are just as useful for analyzing the syntax of spoken language (which is less frequently addressed in the linguistic enterprise than the more orderly written language) as they are for the written variety. Moreover, the notion of construction may actually prove to be more useful than the traditional notions of clause or sentence. Part IV (The pragmatic context) oers two illustrations of the signicance of pragmatic context in linguistic research. Kirsten Vis, Wilbert Spooren and Jose Sanders investigate the issue of conversationalization of public discourse. Analyzing three Dutch newspaper corpora: two corpora including texts from national newspapers from 1950 and 2002 and a corpus consisting of spontaneous conversations recorded c. 2002, the authors use tools provided by Rhetorical Structure Theory to compare various levels of discourse coherence; the content as well as epistemic and speech act relations between clauses are investigated to establish the degree of subjectivity in the investigated texts. Recognizing the level of discourse as a major challenge to Cognitive Linguistics, they demonstrate how RST allows to provide explicit links between the level of linguistic information and the level of conceptual representation. Vis, Spooren and Sanderss chapter demonstrates the eectiveness of the method itself rather than nal results of the study. Luna Filipovic analyses the content of witnesses testimonies: transcripts of police interviews with native speakers of Spanish and their attested translations into English. Focusing upon descriptions of motion events, she investigates the dierences between lexicalizations of the experiential domains in the two languages. In her work, these dierences (e.g. reference to the manner of the motion verb), which result from Talmys (1985) linguistic typology, are used to help to explain the modications of the content of the testimonies, which take place when they are translated from Spanish into English. The study demonstrates how systematic research of lexicalization patterns within particular cognitive domains might reveal signicant typological contrasts between languages and thus contribute to our general knowledge of language typology. On the pragmatic plane, the implications of Filipovics study are equally signicant: the description and the following awareness of typological dierences can allow for a better understanding of language-specic bias in legal contexts and provide signicant clues for training language professionals e.g. interpreters or translators. Filipovics chapter provides a mental link with the nal part of the volume (The social and cultural contexts for CMT and CI), devoted to

Introduction: Theory informing applications, applications informing theory

demonstrating social and cultural contexts for Cognitive Metaphor and Conceptual Integration Theories. Diane Ponterotto demonstrates how Cognitive Metaphor Theory can be applied to cross-linguistic analyses, ultimately contributing to the understanding of translation problems regarding idiomaticity. By oering a contrastive analysis of a selection of English and Italian idiomatic expressions, her study explores similarities and dierences in cross-language renderings of such expressions, focusing on the issue of conceptual metaphor identication. Ponterotto ends her study with some practical implications for translation studies and translator training by pointing out the importance of the awareness of constraints imposed on translation practice by cross-cultural variation of metaphorical conceptualizations. Emphasis on cross-cultural dierences, as manifested in particular instantiations of universal conceptual metaphors, suggests a possibility of completing the theory-application-theory cycle: investigation of attested translations produced by practising translators might obviously contribute to our understanding of universal vs. language- and culturespecic metaphorization. The nal two chapters oer CL analyses of two key notions in Christian religious discourse. In her chapter on metaphorical conceptualizations Magorzata Pasicka oers an analysis of expressions used by followers of Faith Movement, a branch of charismatic Christianity, which she denes as prosperity theology. She discusses conceptual blends and metaphorical conceptualizations of faith, which, as she claims, lie at the very heart of the Faith Movement. She demonstrates how the blending theory, seen as an enrichment of the conceptual metaphor theory, oers a better possibility of interpreting complex and novel metaphors, which are characteristic of the positive confession parlance of the Faith Movement. Moreover, she indicates and explores certain innovative syntactic structures which appear in the Faith Movement discourse, and concludes by showing how such syntactic innovations co-operate with metaphors and conceptual blends to create an overall picture of faith as something almost tangible, something mundane rather than metaphysical. The direction for further investigation is clear-cut: although the hypothesis of linguistic relativism is not explicitly mentioned, with the pragmatic goals of FM in view, psychological and sociocultural implications of Pasickas analysis seem evident. Last but not least, Aleksander Gomola investigates the concept of God that underlies the metaphor God is Friend, often used by process theology and feminist theology. An analysis of the metaphor, carried out in the framework of Conceptual Blending Theory, leads to the second part

Elzbieta Tabakowska, Micha Choinski and ukasz Wiraszka

of the chapter, which brings out the semantic contrast between the God is Friend metaphor and some other, well established metaphors that occur in Christian discourse, e.g. God is Father. Gomola elaborates on the origin of the new concept as well as on social and religious consequences of its wider application. It is generally assumed that man cannot talk about God in plain, i.e. non-metaphorical language; Gomolas analysis shows that a linguist can protably use plain language to talk about the metaphors that men use to talk about God. Theory informs application, application informs theory. They feed each other, and at the same time provide food for thought for anyone keen on trying to understand the nature of complex bonds that tie human language and human mind together. References
Boers, Frank and Seth Lindstromberg 2008 Cognitive Linguistic Approaches to Teaching Vocabulary and Phraseology. (Applications of Cognitive Linguistics 6.) Berlin/ New York: Mouton de Gruyter. Brone, Geert and Jeroen Vandaele 2009 Cognitive Poetics: Goals, Gains and Gaps. (Applications of Cognitive Linguistics 10.) Berlin/New York: Mouton de Gruyter. Dabrowska, Ewa 2004 Language, Mind and Brain: Some Psychological and Neurological Constraints on Theories of Grammar. Edinburgh: Edinburgh University Press. Emmorey, Karen 2002 Language, Cognition, and the Brain: Insights from Sign Language Research. Mahwah, NJ: Lawrence Erlbaum and Associates. Gavins, Joanna and Gerard Steen (eds.) 2003 Cognitive Poetics in Practice. London: Routledge. Gries, Stefan Thomas and Anatol Stefanowitsch (eds.) 2006 Corpora in Cognitive Linguistics: Corpus-based Approaches to Syntax and Lexis. Berlin/New York: Mouton de Gruyter. Gutt, Ernst-August 1991/2000 Translation and Relevance, Cognition and Context. Manchester: St Jerome Publishing. Janzen, Terry 2005 Topics in Signed Language Interpreting. Amsterdam/Philadelphia: John Benjamins. Kristiansen, Gitte, Michel Achard, Rene Dirven and Francisco J. Ruiz de Mendoza Ibanez 2006 Cognitive Linguistics: Current Applications and Future Perspectives. Berlin/New York: Mouton de Gruyter.

Introduction: Theory informing applications, applications informing theory

Kristiansen, Gitte and Rene Dirven (eds.) 2008 Cognitive Sociolinguistics. Language Variation, Cultural Models, Social Systems. (Cognitive Linguistics Research 39.) Berlin/New York: Mouton de Gruyter. Lako, George and Mark Johnson 1980 Metaphors We Live By. Chicago: University of Chicago Press. Langacker, Ronald W. 1993 Reference point constructions. Cognitive Linguistics 4: 138. Langacker, Ronald W. 1995 Possession and possessive constructions. In: John R. Taylor and Robert E. MacLaury (eds.), Language Language and the Cognitive Construal of the World, 5179. Berlin/New York: Mouton de Gruyter. McNeill, David 1992 Hand and Mind: What Gestures Reveal about Thought. Chicago: The University of Chicago Press. McNeill, David (ed.) 2000 Language and Gesture. (Language Culture and Cognition.) Cambridge: Cambridge University Press. McNeill, David 2005 Gesture and Thought. Chicago: University of Chicago Press. Putz, Martin, Susanne Niemeier and Rene Dirven (eds.) 2001 Applied Cognitive Linguistics I: Theory and Language Acquisition. (Cognitive Linguistics Research 19.) Berlin/New York: Mouton de Gruyter. Semino, Elena and Jonathan Culpeper (eds.) 2002 Cognitive Stylistics: Language and Cognition in Text Analysis. Amsterdam/Philadelphia: John Benjamins. Stefanowitsch, Anatol and Stefan Thomas Gries (eds.) 2006 Corpus-based Approches to Metaphor and Metonymy. Berlin/ New York: Mouton de Gruyter. Stockwell, Peter 2002 Cognitive Poetics: An Introduction. London: Routledge. Talmy, Leonard 1985 Lexicalization patterns: semantic structure in lexical forms. In: Timothy Shopen (ed.), Language Typology and Syntactic Description (Vol. 3), 57149. Cambridge: Cambridge University Press. Tabakowska, Elzbieta 1993 Cognitive Linguistics and Poetics of Translation. Tu bingen: Gunter Narr Verlag. Taylor, John R. 1996 Possessives in English. Oxford: Oxford University Press. Tomasello, Michael 2003 Constructing a Language: A Usage-Based Theory of Language Acquisition. Cambridge, MA: Harvard University Press.

Part I.

From loop to cycle

Looking back at 30 years of Cognitive Linguistics1 Rene Dirven and Francisco Jose Ruiz de Mendoza Ibanez
1. Introduction: Looking back in pride After 30 years of inner development and steady growth, Cognitive Linguistics stands rm, both qualitatively and quantitatively. As a symbol of this self-assured presence in the linguistic landscape rises Geeraerts and Cuyckenss (2007) monumental Oxford Handbook of Cognitive Linguistics,2 which even boasts on its blurb: In the past decade, Cognitive Linguistics has developed into one of the most dynamic and attractive frameworks within theoretical and descriptive linguistics. What has made Cognitive Linguistics so dynamic and attractive? The dynamicity of Cognitive Linguistics (CL) is largely due to the fact that it is not a single persons enterprise. Rather it arises from the combi1. This paper is a heavily revised and expanded version of previous work by Dirven (2002a, 2005). We want to thank Gunter Radden, Ad Foolen and Verena Haser for their valuable comments on this earlier work. Financial support for part of this research has been provided by the Spanish Ministry of Education and Science, grants HUM2004-05947-C02-01/FILO and HUM2007-65755/FILO, which are co-nanced through FEDER funds. 2. Cognitive Linguistics now has a whole range of introductions, surveys and readers. There is the shorter type of introduction such as Taylor (1995a), Ungerer and Schmid ( 22006), Dirven and Verspoor ( 22004), or Lee (2001), and the longer type of an advanced introduction such as Evans and Green (2006). Then there is the more specialized type of introduction such as Taylor (2002) and Langacker (2008) for Cognitive Grammar, and Croft and Cruse (2004) for Radical Construction Grammar. In contrast, Geeraerts and Cuyckens (2007) is a highly specialized handbook, covering all strands, linguistic elds and even a number of neighbouring disciplines. Readers are: Evans, Bergen and Zinken (2006) and Geeraerts (2006b), and Kristiansen, Achard, Dirven and Ruiz de Mendoza (2006). An introduction to a two-model conception of CL, with a strict separation between the semantic and the conceptual world, is Schwarz (1996). Finally, for a comprehensive overview of the contrasts between the various versions of Construction Grammar and some functional models of language, see Gonzalvez and Butler (2006).


Rene Dirven and Francisco Jose Ruiz de Mendoza Ibanez

nation of various pioneering ideas that, acting as separable strands of one whole, have drawn together to give rise to a unied paradigm. Central to the whole paradigm is the cognitive commitment, i.e. the conviction that there is fundamental unity and interaction among all cognitive faculties including perception, attention, categorization, conceptualization, aect, memory, reasoning, and language (see Lako 1990; Talmy 1997). The attractiveness of CL may be due to the fact that it has largely given up the traditional axioms that reduce language to a self-sucient system, especially dichotomies such as syntax vs. semantics, lexis vs. grammar, semantics vs. pragmatics, langue vs. parole, competence vs. performance, and synchrony vs. diachrony. The belief in the arbitrariness of the linguistic sign has given way to a search for the motivation of linguistic organization on the basis of cognitive principles. This change of direction stands in sharp contrast to the central claim of generative linguistics that language is an autonomous system, detached from any other type of knowledge, including encyclopedic or world knowledge. In contrast, CL holds that there is no clear-cut distinction between linguistic and encyclopedic knowledge (Haiman 1980; Langacker 1987). Additionally, in CL, the view of the interwovenness of language and other cognitive faculties has generally been synonymous with an extreme anti-modularity view, which is in line with a general anti-generative mind-set. However, a milder approach in all these matters is now growing. Whereas language itself is still seen as an internally non-modular system (e.g. by Glynn 2007), from an anatomic perspective, there are some who would not deny that the various cognitive faculties may be organized on the basis of functional modularity, i.e. the encapsulation of information in distributed areas of the brain, for which there is a growing body of evidence in neuroscience (e.g. Calabretta, Di Fernando, Wagner and Parisi 2003).3 In spite of this non-modular conception of language and the belief in the continuity of grammar and lexicon, all the main strands in CL have
3. Functional modularity rejects the naive notion of modularity as information encapsulation in specic parts of the brain. Similarly, in the case of language, the evidence strongly suggests that there is no domain specicity. For example, there are brain areas involved in language that are also related to motor control (Heiser et al. 2003). There is additional evidence that understanding the meaning of some linguistic expressions often requires that listeners become involved in an embodied simulation of the described situation (Gallese and Lako 2005; Gallese 2005). Embodied simulations of actual object manipulation are thus used in reasoning processes such as understanding the metaphorical meaning of grasp an idea (Bergen 2005; Feldman and Narayanan 2004).

Looking back at 30 years of Cognitive Linguistics


emphasized either the grammatical units of language or else its lexicalsemantic units. In line with this fact, the present retrospective overview will rst concentrate on the more grammatical strands, and next on the lexical-semantic strands, not losing sight of further discourse orientations. We will also outline some developments within these strands, highlight internal problems, if any, and obstacles on the path to their integration. The following strands, or topics related to them, will be discussed: (i) the relation of grammar to cognition (Section 2); (ii) Cognitive Grammar (CG) (Section 3); (iii) dierences of opinion between CG and the various versions of construction grammar (Section 4); (iv) Construction Grammar (CxG) (Section 5); (v) Radical Construction Grammar (RCG) (Section 6); (vi) the philosophical basis of CL (Section 7); (vii) Conceptual Metaphor and Metonymy Theory (Section 8); (viii) Lexical Semantics (Section 9); (ix) Discourse: Mental Space and Blending Theory (Section 10).

2. The relation of grammar to cognition The most important inroads into the relation between grammar and cognition have been made by, amongst others, the following scholars: Haiman (1985a, 1985b) on iconicity; Fillmore (1977) on frames and scenes; Talmy (1975, 1978, 1988a, 1988b, 2000a, 2000b) on gure/ground alignment, and on attention and (per)ception phenomena; Langacker (1987, 1990, 1991, 1999) on construal, perspective, subjectivity and mental scanning; Fillmore (1990), Fillmore, Kay, and OConnor (1988), Fillmore et al. (2004), Goldberg (1995, 2002, 2005), Croft (2001), and Bergen and Chang (2005) on constructional patterns; Lako and Johnson (1980, 1999), Lako (1993) and Lako and Turner (1989) on metaphor; Johnson (1987) on image schemas; Lako (1987) on categorization, and Fauconnier and Turner (1996, 2002) on blending.4 From a historical point of view, CL belongs to the functionalist tradition in linguistics. Although Saussure ([1916] 1974) saw linguistics as part of semiology or semiotics, he mainly emphasized one of the three main modes of semiotic reference, i.e. symbolicity, as the organizing principle of linguistic structure. In the more balanced semiotic view of language taken by Haiman (1985b), the other two main semiotic principles, i.e.
4. Since Fillmores work, although generally compatible with the main assumptions of CL, does not subscribe to the cognitive commitment, we will only make occasional reference to it where it is relevant for our discussion.


Rene Dirven and Francisco Jose Ruiz de Mendoza Ibanez

iconicity and indexicality, are shown to be highly relevant, too. They are, moreover, more perceptually and experientially based, and as such fully in line with the cognitive view of language. As a direct manifestation of the interaction between perception and language, the principle of iconicity becomes visible in three sub-principles of linguistic organization, i.e. sequential order, proximity/distance, and quantity. The sequential order in the perceived or conceived world is reected in syntactic word order as in the advertising slogan Eye it, try it, buy it. Syntactic proximity and distance reect conceptual closeness or dierence, e.g. in the position within adjective sequences as in a large purple satin coverlet: here the conceptual closeness between the linen (coverlet) and the material that it is made of (satin) determines the primacy of the material adjective; color is next in the intrinsic nature of artifacts, and size is the least intrinsic feature and stands at the greatest distance from the noun. The principle of quantity relates to the pairing of form and meaning: a more complex form usually carries a greater amount of meaning, and form is motivated by functional factors such as politeness, demands of informativeness, and rhetoric, among others (e.g. compare the wording of No smoking in a public place with that of Customers are kindly requested to refrain from smoking, if they can in the dining room of a chic restaurant). Thus iconicity clearly reveals that extra-linguistic cognitive factors can have a direct bearing on linguistic form, or that linguistic form is not so arbitrary as many have claimed since Saussure.5 One of the greatest merits of Talmy is to have incorporated Gestalt Psychology into the cognitive exploration of language. Another major asset of Talmys work is its concern with both grammar and lexicon, and with their subtly dierent relation to cognition. As a highly abstract symbolic system of generalized or schematic conceptualizations, the grammar of a language is even more intimately linked with, and subject to, general cognitive principles and processes than the lexical system, which is an inventory of more specic and singular forms of conceptualization. Whereas iconicity is a receptive cognitive principle in that it reects metonymic links between parts (single event or property) and wholes
5. This claim arises from an extreme interpretation of Saussures real views. As Radden and Panther (2004) have noted, Saussure was well aware of the dangers of applying the principle of arbitrariness without restriction and was consequently in favor of relative motivation, a view that is not incompatible with most CL claims on the relationship between form and meaning (cf. Langackers 1987 treatment of the relationship as symbolic assemblies).

Looking back at 30 years of Cognitive Linguistics


(total of events or objects), the imposition of gestalts on the experiential world is a far more creative achievement. That the structure of grammar is directly related to principles of gestalt perception was rst shown by Talmy (1975; see also 1978, 2000a, ch. 5). One of these is the principle of gure/ground alignment, according to which the perception of an overall shape comes about by dividing the perceptual eld into a more prominent part, the gure, and a less salient part called the ground. It is against this ground that the gure moves, is moved or stands out. Talmy furthermore applies this perceptual principle to complex sentences thus showing that the main clause has the function of the gure and the subordinate clause that of the ground. Probing more generally into the relation of grammar to cognition, Talmy (1988a, 2000a, ch. 1) deals with the relations between lexicon, grammar and cognition in terms of a building metaphor. Whereas the lexicon can be compared to the single bricks of a building, the grammar is the conceptual framework or, imagistically, a skeletal structure or scaolding for the conceptual material that is lexically specied (Talmy 1988a: 165). The lexicon contains content words and reects the tens of thousands of individual phenomena as single, conceptual categories, whereas the grammar develops more abstract, schematic categories, applying over wide ranges of phenomena. Such a schematic category of meaning, e.g. that of the plural morpheme, is one that applies to all possible relevant contexts. Thus the schematic meaning of the plural is the notion of more than one, technically called multiplexity, which is found not only with count nouns (cups), but also with abstract nouns ( fears, misgivings), uncountable nouns (ashes, waters), or event nouns (the silences between the two lovers). Furthermore, the concept multiplex is not limited to nouns and plurals, but is also found with iterative verb forms as in He was hitting her. Thus, whereas the lexicon diversies the conceptual world more and more, the grammar synthesizes, under one common denominator, quite dierent manifestations of multiplexity, be it concrete or abstract, countable or uncountable, things or events. It is also in this sense that grammar creates gestalts: [grammatical] structuring is necessary for a disparate quantity of contentful material to be able to cohere in any sensible way and hence to be amenable to simultaneous cognizing as a Gestalt (Talmy 1988a: 196). Still, lexical and grammatical specications are to be seen along a continuum ranging from content categories to schematic categories. An example of closed-class notional category that impinges on grammar is force dynamics (Talmy 1988b, 2000a). Linguistic expressions can


Rene Dirven and Francisco Jose Ruiz de Mendoza Ibanez

have a force-dynamic grounding where either of two forces, an agonist and an antagonist, one stronger than the other, can be in focus. We can have basic steady-state force-dynamic patterns with a tendency to motion (e.g. The ball kept rolling) or to rest (e.g. The fence kept standing in spite of the strong wind ) or change-of-state patterns which also exhibit a tendency to motion (e.g. The water made the re die down) or to rest (e.g. The ball made the lamp fall ). There can be internal psychodynamics as in He forced himself to keep working (motion) and He refrained from smoking in the lab (rest). As Deane (1996) has noted the elements of force dynamics are basic to human cognition since they are grounded in natural motor activities. A related area of Talmys research where the cognitive role of perception is crucial for grammar is the study of motion. Talmy (2000a) distinguishes three types of motion: factive, ctive, and metaphorical. Factive motion, concerns real motion as in The boy ran from the tree to the fence. Fictive motion consists in depicting a situation where there is no actual motion, but it is conceived as if it involved real motion, as in The road ran from the river to the valley; this is possible since our perceptual mechanisms give us the false impression that there is motion (we move our eyes from one point to another and interpret the static scene as if it were dynamic). Finally, with metaphorical motion we see a non-physical entity as if it were a moving object, as exemplied by The smell came into the room. In general, Talmys own grammatical description of a number of single constructions such as those for motion events in various languages around the world tends to be typological in orientation (Talmy 2000b). Thus, he notes that there are satellite-oriented languages like English that represent the path of the motion in a prepositional phrase or satellite (e.g. The child crawled into the room) and verb-oriented languages like Spanish and Japanese that represent the path in the verb root (e.g. El nino entro en la habitacion gateando lit. The child came into the room crawling). In the latter, the co-event (i.e. gateando) is secondary to the motion event itself (entro ); it describes the manner of motion and is typically expressed by a gerund. Thus there is a double opposition: Spanish expresses path as a rst event and manner as a secondary event; English expresses path as a satellite and manner as the sole event. Thus, the English path expression by means of a satellite is a major carrier of the notion of motion in the sentence. In contrast to other founding fathers of CL, Talmy has never developed a specic grammar model. But, stimulated by Talmy and also by Fillmore, CL has seen the development of various models of cognitive grammar (the

Looking back at 30 years of Cognitive Linguistics


use of lower case throughout this paper suggests a generic use of a term and the upper case a specic model). In the following sections we discuss what we feel are the most relevant contributions of the various approaches, which include Langackers Cognitive Grammar, several versions of construction grammar, and Fauconnier and Turners Blending Theory.

3. Cognitive Grammar: Construal, perspective, and compositionality Historically speaking, Langackers accomplishment is to have realized the enormous potential of Talmys gestalt-based approach to language and to have developed it into a fully-edged theory, which at rst he labeled Space Grammar and later Cognitive Grammar. Langackers main inroads into cognition in grammar are the application of the notion of construal to explain various linguistic phenomena, his elaboration of Talmys notion of perspective, and the generalization of the gestalt principle of gure/ground alignment to all levels of linguistic structure. The notion of construal is needed to mark the cognitive operation transforming a conceptual scene into a linguistic description of it. Whereas a conceptual scene is in principle not limited in the number of participants and relations, its linguistic expression always imposes a construal on the scene in terms of scope and perspective. Scope, which relates to how much of the scene is to be included in the description and which elements are to be excluded, is roughly equivalent to Fillmores (1977) notion of frame. Thus the well-known scene of a commercial transaction comprises participants and relations.6 Among the participants we have a buyer, a seller, goods, a price, and the quality of the goods or the extent of the sales. Then, there are two main relations: the transfer of goods and of money. Of all this rich conceptual content, the construal operation never allows all elements within its scope. The absolute maximum is to express all of the participants, but only one direction in the transfer: I paid her 200 dollars for the ring or else She sold me the ring for 200 dollars. It is of
6. The example of the commercial transaction scene stems from Fillmore (1982). His notions of scene and frame (Fillmore 1977) are comparable to Langackers notions of conceived situation and scope. Semantic frames are schematic representations of situation types described in terms of participants and their roles, which matches the elements of a scene taken into the speakers scope. Associated with the frame there are a number of lexical units and inheritance relations with other frames.


Rene Dirven and Francisco Jose Ruiz de Mendoza Ibanez

course possible to express the double transfer by means of the verb exchange (We exchanged the ring for 200 dollars), but here the buyer/seller relation is kept out of sight so that the commercial character of the transaction itself is out of the scope. One can exchange anything for anything else, e.g. stamps for horse pictures. Other forms of construal of the commercial transaction focus on the goods and the price (The ring cost 200 dollars), on the seller and the price (She charged 200 dollars for the ring), on the buyer and the price (I bought the ring for 200 dollars, I paid 200 dollars for the ring). All in all, there is a limited number of ways of construing the same content. Construal is determined by the perspective that speakers impose on the scene to be described. In the commercial transaction only one of the four participants can serve as the point of departure: the buyer can only buy, pay a price and receive the goods; the seller can only sell, charge a price and transfer the goods; the goods cost a certain amount; and money can buy them. Langackers main contribution in this respect is that he has not just left the notion of perspective at a fairly vague level, as exemplied by the various ways of construing the commercial transaction scene, but that he has further elaborated it into more discrete components. Thus, the perspective adopted on a given scene or situation can involve three further components: a vantage point, an objective vs. subjective view, and a given direction in the mental scanning of the scene. First, construal can involve the choice of a vantage point from which one looks at the situation. This can be the sentential subject as the gure selected for the scene, as illustrated above for each of the four dierent participants in the commercial transaction, whereby the other participants then form the ground (see further down below). But the vantage point can also be either of the participants in the speech situation, i.e. the speakers or the hearers position as in the use of come: here the speaker can take either his own perspective or that of the hearer as in the contrast between Shall I come to your place? or Will you come to my place? In the former the coming is seen from the hearers perspective, in the latter from the speakers. Second, perspective can involve the choice between an objective and a subjective construal. These two terms are used in a very specic sense, based on their everyday meanings. An objective construal presents an explicit temporal or spatial setting of the scene relative to speech time or to the speakers position. Thus, by using the adverb before now the speaker denes the time reference point objectively as the time of the speech act (now); and by saying She was sitting across the table from me the speaker puts himself or herself objectively on the stage as the spatial reference point. Subjec-

Looking back at 30 years of Cognitive Linguistics


tivity in such cases means that the speaker leaves out the speech time (She was sitting here) or the speakers position (She was sitting across the table) such that these vantage points are merely implied, explicitly or objectively put on the stage. That is, a subjective construal as She was sitting across the table is precisely subjective because it implies an o-stage, speaker-dependent reference point. As a third component, perspective can involve the choice of a direction of the motion in real motion events or in ctive motion events. Fictive motion is found in scenes such as The roof slopes, where the mind simulates motion in processing the sentence (Matlock 2004a, 2004b). Also in such mental scanning the speaker has to decide on the direction of the motion by choosing between The roof slopes steeply upward and The roof slopes steeply downward, where the former suggests motion away from, and the latter motion towards, the speaker. After selecting a perspective, the speaker assembles the various elements chosen from the scene into larger composite wholes, linguistically expressed as phrases, clauses, sentences, or text. Next to these renements in viewing processes, Langacker has also considerably extended the notion of gure/ground alignment to all linguistic levels.7 The structuring of a scene into a gure and a ground does not only apply to the relation of main clause and subclause, as shown by Talmy (1975, 1978), but also to other levels such as a lexical unit, a compound, a phrase (verb phrase, prepositional phrase, adjective-noun phrase), a clause, and a speech act. Before going into some of these levels, it is necessary to point out that, according to Langacker (1987, 1995), all linguistic units and their meanings can be reduced to two kinds of conceptual entities: things and relations. Things are independent conceptual entities such as book or linguistics, whereas relations are conceptual entities constituting conceptual links between things such as the temporal relation know or the atemporal relation about. Conceptual entities are joined to one another by assembling them into increasingly larger and more complex units. Thus book is assembled with about linguistics into the relationship book about linguistics; this initial assembly is in its turn combined with good into good book about linguistics; these are all atemporal relations, which are further assembled with the temporal relation I know into I know a good book about linguistics. This principle of composing and integrating smaller units into larger higher-order ones is known as compositionality, to which we will come back in Section 4. But let us rst see

7. For a survey of the uses of gure/ground, see Grundy and Jiang (2001: 115).


Rene Dirven and Francisco Jose Ruiz de Mendoza Ibanez

how the gure/ground relationship applies to various conceptual units and their expression. Each conceptual entity, e.g. book or strawberry, consists of a prole and a base: the gure is what is proled and the base is what the gure is proled against. For strawberry the base is the domain of plants, in this case a strawberry plant with roots, leaves and fruit, and the noun strawberry proles the fruit. A relationship like the strawberry on the plate consists of the relational link (or relation) on and two things as participants, i.e. strawberry and plate. The relation on proles contact or support with a surface in the domain of space. The gure/ground relationship holds between the rst participant strawberry as the gure and the second participant, plate, as the ground. In order to have a more specic term for gure/ground relationships, Langacker (1987: 217220) has introduced the spatial terms trajector which does not necessarily identify a moving entity and landmark. Langacker sees the trajector/landmark conguration, which further species the notion of gure/ground alignment, as applicable to those conceptual relationships that are linguistically expressed as phrases, clauses, and complex sentences. In the structure of a simple sentence or clause, the trajector is the subject, and the landmark is the (direct) object or the oblique complement. Furthermore, each trajector/ landmark relation in a nite clause or rather the situation or scene described in it is simultaneously grounded in time and reality by means of the tense system or in irreality (i.e. potential reality or nonreality) by the modality system. A clause uttered in discourse functions in the interaction between speaker and hearer as a speech act (see Searle 1979). Figure/ground relationships also hold in this case, which is in our view Langackers most innovative extension of the notion in question. That is, the speech act itself functions as the gure and the whole speech situation is the ground. The speech situation comprises the speech act participants (speaker and hearer), the speech act time and location, and the dierent elements of background knowledge. The speech situation is thus the whole context or the ground against which the linguistic expression communicated makes sense as the gure. Langacker has also developed outspoken views on three issues that have caused dierences of opinion within CL: the notion of compositionality, the number of conceptual-linguistic units, and the universal status of grammatical categories such as noun, verb, subject, and direct object. Though not apparent at rst sight, these issues may well be linked to one another, which we will do in the context of construction grammar models. Thus, as will be evident in the ensuing discussion, Langackers

Looking back at 30 years of Cognitive Linguistics


schemas which may range from specic to abstract are but conventionalized patterns that give rise to more creative constructions.

4. Dierences between Cognitive Grammar and construction grammars Although Langackers Cognitive Grammar and the various construction grammar models share the basic tenets of cognitive linguistics, they also dier fundamentally on the principle of compositionality, the inventory of linguistic units, and the status of grammatical categories. Construction grammar claims that compositionality cannot account for all linguistic phenomena and that acceptance of a third conceptual-linguistic unit is unavoidable: in addition to things and relations, they also argue for the existence of constructions, i.e. xed patterns in the combination of relations and things, not built up compositionally, but available as whole constructional units with their own constructional meanings (see e.g. Michaelis 2003). Thus, the linguistic expression I painted myself in a corner can have two dierent interpretations and hence also two constructional paths. The compositional build-up accounts for the literal meaning that the speaker made a picture of himself/herself and did so in a corner. In this interpretation the object (myself ) is assembled to the verb ( paint) into a verb phrase that is then assembled into a relationship with the subject (I ), thus yielding a new composite structure that is further assembled to the adverbial phrase (in a corner). The non-compositional, idiomatic and constructional meaning is quite dierent, i.e. I expressed an opinion, which committed me to things which I subsequently regretted. This gurative meaning is based on the gurative interpretation of the global construction whereby the speaker does something that brings him/her into an unpleasant state. In order to distinguish this case from the caused-motion construction, as in The player kicked it into the goal, we will call it a caused-change construction. In such a construction there is no assembly of the verb ( paint) and a direct object (myself ), but only between the verb ( paint) and the subject (I ), on the one hand, and between the pronominal noun phrase (myself ) and the adverbial phrase (in a corner), on the other. But the link between those two relationships is not borne out by a linguistic form, which is the very reason why construction grammar claims that it is the construction as a whole that expresses the notion of caused change. This in turn entails that there are not just conceptual entities like things and relations, but also more abstract patterns or strings the ambiguous and hence infelicitous term chosen for them


Rene Dirven and Francisco Jose Ruiz de Mendoza Ibanez

is constructions which just like things and relations also function as conceptual units.8 In such constructional units or non-compositional constructions the interpretation process is claimed to follow a topdown direction, whereas in compositional constructions the interpretation results from a bottom-up assembly of the various lexical units (I, paint, myself, in, corner) into increasingly larger phrases and a clause. In the gurative interpretation of I painted myself in a corner it is especially the relationship myself in a corner that seems to lack any conceptual and syntactic status in the clause as a whole. It is important to realize that it is not the gurative use of the caused-change construction that makes the construction special. This non-compositional caused-change construction can also occur with many other verbs, also when used literally. Consider the verb drink. Here the caused-change construction can be used to express a resultant state, with adjectives (He drank himself poor/silly/senseless), with the goal prepositions to or into (He drank himself to death/into a coma), and with the source preposition out of (He drank himself out of his practice/out of his career). In Section 5 we will discuss the solution Construction Grammar oers for this, then evaluate Langackers view of it, and suggest other solutions. But rst we must refer to the other major point of disagreement between Langackers Cognitive Grammar and construction grammar models, especially Radical Construction Grammar, which is the status of grammatical categories like noun, verb, subject, and direct object. For Langacker, there are only two basic conceptual entities, i.e. things as somehow independent entities, and relations as establishing links between things. Linguistic expressions that prole things are, prototypically, nouns, pronouns, determiners, and higher-order expressions such as full noun phrases. Linguistic expressions that prole relations are verbs, prepositions, conjunctions, and adjectives, among others. There are two types of relations: nite verbs prole temporal relations (i.e. processes); non-nite verbs, prepositions, conjunctions, and adjectives prole atemporal relations. For Langacker, the two main categories discussed here, nouns and verbs are not just meta-theoretical constructs, useful to talk about language, but they are an intrinsic part of grammar and thus have a universal status. The same claim is made for the functional categories of subject and object (Langacker 2005: 128129). Radical Construction Grammar rejects this view on the grounds that most gram8. Langacker uses the term construction in a much looser sense than constructionists in CL do. For Langacker the notion of construction is even compatible with that of compositionality, as will be shown below.

Looking back at 30 years of Cognitive Linguistics


matical categories dier across languages and can only be dened from a language-specic perspective (see Section 6).

5. Constructions and Construction Grammar (CxG) Construction grammars dier from Langackers Cognitive Grammar in that they reject compositionality as the main principle governing the grammar of a language, and claim that languages have numerous xed or idiomatic expressions or constructions, i.e. grammatical units in themselves. There are four main constructionist approaches. The rst of all has been Fillmores construction grammar, also called Berkeley Construction Grammar (BCG), which was developed partly within a generative frame of thought by Fillmore, Kay and OConnor (1988), Fillmore (1990), Fillmore and Atkins (2000), Kay and Fillmore (1999) and Fillmore, Kay, Michaelis and Sag (2004). Still, Fillmore incorporates quite a number of assumptions from CL in his approaches to the lexicon and grammar, known as Frame Semantics and Construction Grammar respectively. But typical for the more generative outlook, in this respect, is the separation between lexicon and grammar. According to Lako (cf. Pires de Olivera 2001: 26), Fillmore does not consider himself a part of the Cognitive Linguistics paradigm, mainly because of his non-commitment to the interconnection between linguistic capacity and other cognitive faculties. Besides Fillmores approach, there are three fully cognitive versions of construction grammar: Goldbergs (1995) Construction Grammar (CxG), Crofts (2001) Radical Construction Grammar (RCG), and Bergen and Changs (2005) Embodied Construction Grammar (ECG).9 According to Langacker (1991: 8), the dierence between CG and CxG can be characterized as follows: whereas CG considers constructions to be reducible to symbolic relationships, CxG assumes that grammatical classes and other constructs are still thought of as a separate level of organization. But more is at stake than a separate level of organi9. ECG is mainly concerned with the embodied aspects of online language processing, which involves accessing and simulating conceptual representations or embodied schemas. ECG stands out for its emphasis on the simulation stage of processing, the inferences this process brings about, and its overall goal to achieve computational adequacy. However, ECG has not explicitly addressed issues of compositionality, the inventory of linguistic units, and the status of grammatical categories, which are the focus of this section.


Rene Dirven and Francisco Jose Ruiz de Mendoza Ibanez

zation; it is equally a question of constructional meaning, as has been pointed out by diverse constructivists such as Lako (1987: 467, 538), Fillmore (1990), Goldberg (1995, 2006), Kay and Fillmore (1999), Michaelis (2003) and many others. That is, construction grammar starts from the existence of gestalt-like patterns or established congurations that are both simpler to produce and have meaning relations between the composing parts above their ad hoc compositional meanings. According to Goldberg (1995: 4) such patterns or constructions carry meanings independently of the words in the sentence, as in the parade example of an extended caused-motion construction Fred sneezed the napkin right o the table (onto the oor). The verb sneeze, which has no causative meaning, is intransitive and cannot have a direct object. Since the causative meaning of the sentence as a whole is, according to Goldberg (1995: 4), not strictly predictable from the constructions component parts or from other previously established constructions, she proposes to derive it from a constructional pattern such as cause-move <cause theme source/ goal>. Of these thematic or semantic roles, Fred is the cause, the napkin the theme, o the table the source, and onto the oor the goal. Thanks to this higher semantic schema the verb form sneeze inherits a causative meaning from the abstract predicate cause-move and functions as an ordinary caused-motion verb such as to put in He put the napkin on the table. However, there is a major dierence between the verbs to put and to sneeze: whereas to put does command a direct object and an obligatory resulting location or state, to sneeze does not command either of these. This appears from the passive test, which works with verbs such as to put (The napkin was put on the table), but not with to sneeze (*The napkin was sneezed o the table). The solution that Construction Grammar has chosen is, in fact, to accept an abstract verb cause-move, which in our opinion is not necessarily the only possible, and perhaps not even the best solution. The rst problem is the choice of an abstract predicate, which is a solution comparable to accepting a zero-morpheme in order to account for a causative meaning for expressions without a causative marker. Now already 20 years ago and in a dierent context, Langacker (1987: 452) discussed causative expressions that have no overt causative marker and said that he is not in favor of positing causative zeromorphemes. Langacker (2005: 153) tends to a solution whereby verbs such as to sneeze in the parade example might gradually, through its frequent use, assume a causative meaning in the long run. However, this solution does not solve all problems. Although the various caused-change

Looking back at 30 years of Cognitive Linguistics


expressions with drink such as drink oneself to death are relatively frequent, it is unclear whether it is possible to think of this use of the verb to drink as involving causation. On top of all this uncertainty about the causative status of to sneeze or to drink, there is the non-object status of the napkin and oneself in such intransitive causative sentences. The incompatibility of the verb and the noun phrase is even clearer in drink oneself to death. Here to drink is a verb that has already incorporated its direct object (excessive amounts of alcohol ) and, moreover, if there was a direct object, it could not be human, but only non-human. This fact makes the explanation in terms of three arguments proposed by Goldbergs Construction Grammar for the extended caused-motion and caused-change constructions highly suspect. The same holds for Langackers acceptance of a potential gradual meaning shift in to sneeze or, for that matter, in to drink. The essential problem is that at the conceptual level there cannot be a predicate-argument relation between drink and himself in He drank himself to death, but there can only be one within each of the two substructures He drank and himself to death. The second problem with Goldbergs abstract predicate cause-move is that this does not account for the absolute necessity of a source/goal in the extended caused-motion construction with intransitive verbs such as sneeze, drink or others (e.g. laugh in The audience laughed the actor o the stage). In these extended uses of the caused-motion construction the notion of motion is by and large expressed or carried by the dynamic prepositional phrase with o, out of, to, into, etc. This fact cannot but be related to Talmys insight (discussed in Section 2) that English is a satellite-oriented type of language, using the motion verb for secondary aspects such as the manner of motion, but expressing the very motion path by satellites as in The thief must have escaped by the backdoor through the garden and over the garden wall. In the extended caused-motion construction, the path component (o the table, o the stage, out of his job) is the sole component to conceptualize and express the motion, which also explains why it is a conceptually necessary component. The intransitive verb itself (sneeze, laugh, drink) reects the action that serves as the direct or indirect cause triggering o the motion, but this does not have to imply that the verb itself assumes a causative potential. It thus turns out that the extended causedmotion construction is far more complex in conceptual content than a single predicate-argument schema like 3cause-move cause theme source/ goal4 can suggest. The further question is then: is there any other way to account for


Rene Dirven and Francisco Jose Ruiz de Mendoza Ibanez

the causative relation between the two substructures in [(X sneeze) (the napkin o the table)]? For a possible answer to this question, we turn to Langackers (1987) discussion of compositionality again. Langacker does not invoke compositionality for all usage events in language. Rather, he makes a distinction between fully compositional, partially compositional, and non-compositional constructions. As an example of the latter, Langacker (1987: 454457) discusses the adjective-noun combination a patriotic pole-climber. Here he analyzes a putative event of boosting the feelings of patriotism amongst the crowds of spectators before American football matches by having a cowgirl climbing a agpole in the stadium and kissing the American ag. The expression for such a putative event might be patriotic ( ag)pole-climber, which, if repeated regularly, might become an entrenched expression. Still, it would not have the compositional meaning of a patriotic person who climbs the agpole, but the non-compositional meaning associating it with the football context above. Langacker (1987: 453) sees such composite structures as evoking a knowledge system to which neither of its components [i.e. patriotic and pole-climber] provides direct access, [but] the component structures motivate and highlight selected facets of the composite meaning. In terms of gure/ground alignment, we can see the selected expressions and their meanings as the gures which here evoke two (back)ground domains, i.e. patriotism as associated with the American ag, and sports as associated with pole-climbing, combining both in an act of patriotic love amidst crowds of sports spectators. This means that metonymically referring expressions can evoke much wider scenes, and even implicitly highlight the conceptual relations potentially holding between elements in those scenes. That is, the relation between patriotic and pole-climber is largely contextually determined. Once the expression is repeated often enough, it acquires an increasingly de-contextualized meaning, which in turn becomes entrenched. As an alternative to Goldbergs schema cause-move <cause theme source/goal>, it is possible to propose an approach like that of Langackers above for the caused-motion or caused-change construction as found in paint oneself in a corner and drink oneself to death. A good reason to do so is that CG does not regard lexicon and grammar as dierent systems. According to Langacker (1987: 456), both lexicon and grammar may, though in dierent proportions, show all the various degrees of compositionality, i.e. full compositionality, partial compositionality, and non-compositionality. Therefore we will apply this approach to non-

Looking back at 30 years of Cognitive Linguistics


Figure 1. Perception of Causality

compositional constructions such as the extended caused-motion construction with to sneeze and the caused-change construction with to drink. It is part of the speakers knowledge system that the force-dynamic event (see Talmy 1988b) involved in sneezing may cause lighter objects to y away and that excessive alcohol drinking may lead to a persons corporeal or professional ruin. In each case two conceptual relationships are juxtaposed as follows: [(Fred sneeze) ! (the napkin is o the table)] and [(Fred drink) ! (Fred comes to death)]. The causal link between the two relationships in each construction is based on our general perception and knowledge system. As has been empirically shown in experimental psychology (see Selg 1966, quoted by Radden 1985: 186), naive observers spontaneously establish causal links between two successive events, scenes or situations, even when no such links hold. Radden (1985) describes an experiment where subjects were shown a lm of drawn triangles. One of the triangles approached the other and reached it; then, this second triangle started to move away in the same direction while the rst one remained in the same place. The naive observer sees triangle A moving towards triangle B (phase I), and then hitting triangle B (phase II) thus causing B to move away (phase III). However, in the lm the picture sequence was not one of causality. Given this strong propensity for imposing a chain of causal links onto observed or conceived successive events, it is quite legitimate to accept that speakers-conceptualizers do the same in language use. The causal link in to sneeze a napkin o the table and to drink oneself to death would then not have to follow from an abstract predicate cause-move or cause-change, but could simply follow from a perceptual-conceptual propensity of naive human observers to construe causal links between two successive events or states. Just as Langacker invokes contextual factors to motivate the symbolic links between the components in patriotic poleclimber, we could propose an empirically-based and contextually-imposed


Rene Dirven and Francisco Jose Ruiz de Mendoza Ibanez

causal link for the two constructions under discussion. For a more linguistic representation of this construction, see the end of Section 10.10 Solutions like this one are not needed for very frequent constructions such as the transitive, intransitive, passive, and ditransitive constructions (Goldberg 1992). Therefore we can limit the discussion to another test case and pick out one construction from the various less frequent cases discussed in the literature such as the incredulity response construction (What? Him write a novel?! ), the let-alone construction (Fillmore, Kay and OConnor 1988), or the middle construction (Yoshimura 1998; Heyvaert 2003; Ruiz de Mendoza 2008). We will concentrate on this last one. The middle construction as in The door opened is a special case of the intransitive construction as in The book fell down. The middle construction with sell (e.g. This book sells well ) is even more special because it has a transitive counterpart, which, as shown in Section 3, operates in a fully compositional construction. Used in a non-compositional middle construction, sell can no longer focus on all the four participants of the commercial transaction scene (that is, neither on the seller, the buyer, nor the money), but the scope is mainly constrained to the goods. Still, the use of sell in a middle construction allows the speaker to construe a
10. Ruiz de Mendoza and Mairal (2007, 2008) have explored the metaphorical grounding of some uses of the caused-motion construction, as in The audience laughed the actor o the stage. Typically, the predicate laugh is either intransitive (They laughed and laughed ) or it may take a prepositional object (They laughed at the poor actor). However, in laugh someone o somewhere, laugh seems to take a direct object, very much like any prototypical transitive predicate (cf. They kicked the actor o the stage). Both types of sentence can be passivized: He was laughed at and He was laughed o the stage. We suggest that this kind of motion is comparable to Talmys notion of ctive motion. That is, we do not actually use direct force as in the case of kick, but we see the laughing as a psychological force ctively causing motion. In a case like He drank himself out of his job we then have to deal with metaphorical motion, which in fact does not involve motion, but only a change of state. This use is the result of high-level metaphor whereby the action of excessive drinking is seen as an indirect force moving the man out of his job. High-level metaphor acts as a licensing factor for the participation of the predicate drink in the caused-motion construction. This account is compatible with one where the hearer is expected to impose a contextually-grounded causal link between the event of people laughing at someone and the emotional reaction of leaving the place, which would act as a further pragmatic constraint on the possibility of a caused-motion use of laugh.

Looking back at 30 years of Cognitive Linguistics


conguration that goes far beyond a classical predicate-argument structure. Although there is no agent-seller, there is an agent-like property of the goods that has an enabling force11 allowing the goods to make good sales. The utterance This book sells well is not just a neutral statement, but highlights a less prominent component of the commercial transaction scene, i.e. the quality of the goods and its relation to the manner or extent of the process of selling. Whereas in compositional constructions with sell this manner or extent component is optional, in the non-compositional middle construction it is obligatory, or at least implicitly present.12 Another important component in the sell middle construction is the fact that, unlike the intransitive construction, which may take all possible tenses, this use of the middle construction prototypically occurs in the simple present, suggesting a general situation. Again, this component of the middle construction does not stand on its own, but is intimately linked with the properties of goods and the manner or extent component. Each of these three components more or less conditions the other two components. Conceptually there is a focus on the quality of the selling process enabled by the goods properties, allowing a high sales output, and presupposing a whole series of selling events. It is this specic contextual background that is caught by the sell middle construction, which totally focuses on one participant in the commercial transaction event, i.e. the quality of the goods and sales. The special characteristic of a non-compositional construction is then that it can overrule the normal compositional paths of languages and create subjects or direct objects that at the conceptual level have no equivalent roles as agents or patients. This view of what non-compositional constructions do seems to go into a somewhat dierent direction from that of Taylor (1998), who still seems
11. Radden and Dirven (2007) have observed that the subject of middle constructions in English has an enabling function such that the subject takes on a causal, agent-like value, as in Our new stadium seats 80,000. The presence of causality has also been noted by Heyvaert (2003: 132), who argues that the middle construction foregrounds the fact that the Subject-entity has properties which inuence the occurrence of a particular process. 12. This formulation diers from that by Yoshimura (1998: 279), who says that the subject book goes beyond the semantic value of a non-agentive intransitive in that it has some special properties that enable what is denoted by the predicate, sell well. What is still lacking here is the link between the goods properties and the manner or extent component in the selling process. This lack in the overall conguration of the sell middle construction is somehow felt in speaking of a predicate sell well, where in fact only sell is the predicate.


Rene Dirven and Francisco Jose Ruiz de Mendoza Ibanez

to see a bottom-up compositional build-up in constructions. According to this author, constructions such as the middle construction are schemata which have to be characterized by criteria such as the conguration of the parts, the contribution of the parts to the overall meaning of the construction, and the semantic, pragmatic, and discourse value of the construction (the middle construction is especially favored in advertising). In our view, the semantic component of enabling property does not come from the assembly of book with sell, but it originates in specic facets such as certain factors in commercial transactions favoring good sales. In this sense, constructions are gestalts in which the construction is, as said before, the gure, and the commercial context the ground of the construction as a whole. This wider view also allows other factors than what would be a direct object in a transitive construction (sell a book) to become the more salient factor in the success of the sales process. Thus also the location of the goods can, though marginally, function as the point of departure in the middle construction. Yoshimura (1998) gives the following example of a booksellers exchange: Where shall we put the new travel book? Well, the corner shop window sells very well. Obviously, we can observe prototypicality eects in this construction too, demonstrating that we witness the impact of the same very general cognitive principles at all levels of description.

6. Radical Construction Grammar Compared with Goldbergs CxG, Crofts (2001) Radical Construction Grammar (RCG) is based on both Fillmores construction grammar and Langackers Cognitive Grammar (CG) and is, in fact, closer to Langackers CG than to Goldbergs CxG. RCG is sometimes even called a notational variant of CG, which is certainly not the case. But both RCG and CG are alike in that they do not start from Goldbergs putative abstract predicates such as cause-move as the predicate in argument structures. RCG is radical in comparison with both Goldbergs CxG and Langackers CG for various reasons. First, there is a dierence over the basic linguistic units. Whereas CxG originally used the term constructions only for non-compositional or unpredictable expressions, but not for compositional or predictable ones,13 RCG sees constructions as the only basic
13. Goldberg now subsumes, just like Langacker, all complex linguistic expressions under constructions.

Looking back at 30 years of Cognitive Linguistics


linguistic unit, or primitive as Croft calls them. This applies to both lexicon and grammar (or syntax, as Croft prefers to say). The lexicon contains atomic substantive constructions, whereas the syntax or syntactic rules are complex schematic constructions. Unlike Langacker, Croft thus avoids having to make conceptual distinctions between things and relations, and their lexical expression as nouns and verbs, respectively. Similarly, Croft avoids having to distinguish between temporal relations (verbs) and non-temporal relations (adjectives, prepositions, conjunctions), to which we will come back below. A second major dierence between the three CL grammar models concerns the number of levels in the formmeaning pairings of language. Langacker (2005: 105) observes that CG only posits two levels or poles, i.e. semantic structure and phonological structure, which are linked as symbolic structure, but that CxG and RCG posit three levels: semantic structure, grammatical form, and phonological structure. That is, for CxG the verb give has the semantic structure of cause-receive <agt rec pat>, where agt stands for agent, rec for receiver, and pat for patient, plus the syntactic form V SUBJ OBJ OBJ 2 and its phonological structure. For RCG the semantic structure of give is give (donor, gift, recipient) and the syntactic structure is NP1 < Verb < NP2 < NP3 . Although Langacker uses such labels for descriptive analysis, too, he claims that they are semantic notions, not syntactic ones, as Goldberg and Croft are purported to claim. But maybe Langacker misinterprets Croft here. For this purpose, let us see how Langacker (2005: 106) presents Goldbergs and Crofts analyses of the ditransitive construction with give. Whereas Goldbergs CxG (Fig. 2a) explicitly claims a syntactic level with a subject, an indirect object and a direct object between the semantic and phonological levels, Crofts RCG (Fig. 2b) only uses categories at phrase level (NP, V), which are directly linked by the dotted lines to the corresponding component units in the semantic structure. Croft (2001: 2)

Figure 2. Contrast between Construction Grammar and Radical Construction Grammar


Rene Dirven and Francisco Jose Ruiz de Mendoza Ibanez

claims that there are no syntactic relations between elements in a construction, which may be the reason why he avoids terms such as subject and object and only uses phrasal units whose mutual relations derive from the relations between their corresponding semantic relations, as suggested by the dotted lines. Thus, he argues that syntactic relations are not determined by semantic roles such as agent or instrument, but by the causal relations among participants in the action. This view is supported by the dierent construal operations that are possible with the verb sell: the subject can be the agent (He sells the book), the patient (The book sells well ), or the location, (This shop window sells well ). In each case, it is not the semantic role as such that determines the subject, but the whole causal conguration of the construction. Additionally, Croft (2001: 3545) thinks that grammatical categories such as subject, object, and oblique NP can only be dened languagespecically, because constructions are language-specic themselves. Croft concludes that such grammatical categories can have no universal denition and no meaningful relations to one another. Here, Langacker (2005: 111) defends the opposite view, dening subject and object in terms of focal prominence. The participant that receives primary focus becomes the subject and the participant that receives secondary focus is the object or the oblique NP. However, there need not be a contradiction between Langackers and Crofts views. In the case of sell, each of the three subjects (seller, book, shop window) exerts a certain causal force thus attracting the primary focus. Since in Langackers view the subject/ object conguration is an instantiation of the trajector/landmark conguration, which is itself a specication of gure/ground alignment, here the circle has become full: gure/ground alignment is a ubiquitous phenomenon. Our brief overview of some of the central assumptions of CG, CxG, and RCG has evidenced the complementariness of the three approaches, despite some apparently marked dierences, which are more a natural side eect of dierences of focus. Thus, CG focuses on compositional constructions, CxG on non-compositional constructions and on construction networks and alternations, and RCG on the semantization of syntax. We now turn to the more lexical explorations, including the creative processes of metaphor and metonymy. Whereas the grammatical innovations of CL are rather based on psychological foundations, lexical-conceptual explorations are rather faced with philosophical problems. Therefore the next section will discuss the philosophical roots of the CL enterprise.

Looking back at 30 years of Cognitive Linguistics


7. The philosophical basis of CL: embodied realism or phenomenology? From its inception, CL dened itself in strong opposition to its dominant structuralist and generative predecessors, which are both still alive and vital. Unlike the rival theories, CL saw language not as a self-sucient system as had been generally accepted since Saussure, but as a cognitive faculty interacting with other cognitive abilities such as perception, attention, memory, imagination, emotion, and reasoning. The main focus of CL is not like in many structuralist or generative approaches on form only, nor like in traditional semantics on meaning only, but on the pairing of form and meaning. From Saussure to Chomsky and beyond, linguistic meaning had silently been taken for granted as a semiotic system of its own or explicitly conceived of as an autonomous system in the mind, i.e. as a module separated from the human conceptual world, or at best as linked to it by a conceptual-semantic interface (Bierwisch and Schreuder 1992; Pustejovski 1995; Jackendo 1996, 2002; Schwarz 1996). But in CL meanings are part and parcel of our conceptual world. Meanings are concepts or combinations of concepts that have been laid down in language. By embracing Eleanor Roschs Prototype Theory CL has broken away from the traditional Aristotelian treatment of categories and accepts Roschs view of the internal layering of categories with central or prototypical members, less central members, and peripheral members, which also entails the partial overlapping of categories. That is, categories or concepts are man-made distinctions in the experience and perception of the human internal or external world and do not just reect the basic facts14 of reality in themselves. CL thus ventures into philosophical territory, but unfortunately it has not been suciently cautious in its search for a philosophical basis. Instead of formulating its own experiential (or embodied) realism as developed by Lako (1987) and Lako and Johnson (1980, 1999), CL might also have opted straight away for the philosophical roots of experiential realism, known as phenomenology, as has repeatedly been advocated by Geeraerts (1985, 1993). The phenomenologist revolution in philosophy was launched by philosophers such as Husserl (1960, 1970a, 1970b), Heidegger (1962, 1982), and Merleau14. This term is used by Searle (2005) to denote the physical facts in the world, which he sees as independent from consciousness. This is the opposite of what Rosch proposes. In addition to physical facts, Searle also accepts social (or institutional) facts and cultural facts.


Rene Dirven and Francisco Jose Ruiz de Mendoza Ibanez

Ponty (1945, 1962, 1979), and it has been actualized by Dreyfus and Dreyfus (1999), and Dreyfus (2002). Lako and Johnson (1999: 97) do not make much emphasis on the roots of CL notions in phenomenology and the connection is virtually absent in the authors previous discussions of experientialism versus objectivism (e.g. Lako and Johnson 1980; Lako 1987). Phenomenologists, in contrast, have devoted quite some attention to the points they share with cognitive science, such as the embodied mind and the important role of pre-conceptual experience in language and logic (cf. Wiggins 1994). In any event, Lako and Johnson (1999: 97) point out that what distinguishes CL from other approaches to embodied realism is the use of empirical evidence coming from neuroscience and cognitive science. According to Lako and Johnsons experiential realism, all individuals have access to the world by their embodied experience and perception of that world (experientialism), and they can all have the same experiences and perceptions, because they all share the same bodily constitution in contact with that world (realism). From a phenomenological point of view there is nothing wrong with this view, although it does need rening. The problem is that Lako and Johnson go far beyond this (implicitly) phenomenological claim in that they see experiential realism as a challenge to traditional Western thought, ranging from Aristotle to Descartes and to many present-day thinkers, whom they lump together under the label of objectivism or objectivist realism. For Lako (1987: 183), in objectivist realism true knowledge of the external world can only be achieved if the system of symbols we use in thinking can accurately represent the external world. The alternative view, i.e. experiential realism, holds that human language and thought are structured by, and bound to, an embodied experience (Lako and Johnson 1999: 233). Perceptual, especially spatial, experience is said to pave the way for categorization of the world phenomena, and these categories in concrete domains are mapped metaphorically onto more abstract domains such as time, emotion, and event structure. One objection to Lako and Johnsons notion of embodiment and embodied realism is found in the notion of a situated body as explored in the collective volumes Body, Language, and Mind. In the rst volume, edited by Ziemke, Zlatev and Frank (2007), the main reservation is that the conception of the same bodily constitution cannot merely be a mass of single human beings oating around in the air, but that they are situated bodies linked to their natural environment and to one another. Its twin volume, edited by Frank et al. (2008), concentrates

Looking back at 30 years of Cognitive Linguistics


on the social dimension of the body/minds situatedness, which is given shape in language and communication in human communities with their own cultural beliefs and institutions.15 The sharpest objection to Lako and Johnsons notions of experiential realism and objectivism is formulated in Haser (2005), who criticizes Lako and Johnsons presentation of and quotations from the so-called objectivist literature as being very scarce and ambiguous. Up until now there has not been any serious refutation of Hasers criticism. But can and must this criticism be refuted? Would it not be better to search for the very roots of experiential realism? In fact, this route was already proposed by Geeraerts (1985, 1993), who places the deeper roots of CL in the phenomenologist revolution in philosophy, especially as presented by Merleau-Ponty (1945, 1962). This phenomenologist stresses consciousness and intentionality in the joint interaction of the body and the mind with the environment. Embodiment is not a matter of the body alone, but of the mind through the body. Merleau-Pontys thesis is that consciousness is present in the corporeal experience of the world (Geeraerts 1985: 355). Here we nd all the basic epistemological tenets of CL: its realism, its experientialism, and its assumption of embodiment in the sense of an embodied mind, or as Merleau-Ponty puts it, an em-mind-ed body. These insights of Merleau-Pontys are taken up again by Violi (2008), who combines Merleau-Ponty with Peircean semiotics:
Through perception the subject meets the world in the rst place and begins to give meaning to it. Phenomenological and perceptive meaning is transformed into linguistic meaning through the corp propre16 which founds, at one and the same time, the subjectivity of consciousness and the exteriority of the world. Here we can see another possible compatibility with Peirces philosophy: in Merleau-Pontys phenomenology, too, external and internal world are not separate and in opposition with one another, but related to each other via the mediation of the corp propre that operates, in a way, as translator of perceptually constructed meaning into linguistic and conceptual meaning.

15. Yu (2009) is an impressive illustration of the role of a communitys cultural conceptualization of the world. He shows how the Chinese understanding of the body/mind as a microcosm in the macrocosm of the universe has always been the core of traditional Chinese philosophy and medicine and is still part and parcel of present-day literary and non-literary Chinese discourse. 16. The English translation for this somewhat opaque term is lived body.


Rene Dirven and Francisco Jose Ruiz de Mendoza Ibanez But the body is also the place where aect and emotion are rooted, as Freud and psychoanalysis have taught us, reminding us that the Ego is rst and foremost a corporeal Ego. Recent developments in semiotic theory are insistent on the fundamental role emotions play on the very deep level of sense structuring.

Violi has also re-introduced the concept of aordances, which are possibilities for action oered by the local environment to a particular type of embodied agent, equipped with specic bodily features. In this way perception is always contextualized and constructed: the world is essentially perceived by some given organism endowed with its own intentions in some given context, and is seen as aording opportunities for goaldirected actions. Perception is therefore always connected to action, and both perception and action are always connected to cognition (Violi 2004: 216).17 These observations call for a serious revision of the philosophical underpinnings of CL. If the paradigm is not adequately contextualized, it is likely to keep ignoring the connections between the notion of embodiment and the social and cultural dimensions of language, and to pursue in the oversimplication of regarding most of Western thought as imbued with an objectivistic vision of science and knowledge. If, on the contrary, CL takes on the challenge of exploring more deeply the roots of experientialism, the theory will gain broader philosophical support in the domains of body/mind unity and interaction, the connection between inner and exterior perceptions, and the nature of consciousness and intentionality, among other issues. Undoubtedly, CL has a privileged position for this kind of epistemological venture in comparison with other approaches to language: CL is a exible framework that is already backed up by serious empirical work in psychology and neuroscience. The mutual feedback between CL and research paradigms such as phenomenology is more than promising if the philosophical reection is sensitive to the cognitive commitment. 8. Conceptual metaphor and metonymy: How conceptual and how linguistic? To a large extent, CL owes its breakthrough to the innovative work by Lako and Johnson (1980), which launched Conceptual Metaphor Theory (CMT). This breakthrough is remarkable if one takes into account that, from early on, CMT met with severe criticism from psycholinguists (e.g.
17. Violi (2004) is an earlier version of Violi (2008).

Looking back at 30 years of Cognitive Linguistics


Camac and Glucksberg 1984; Glucksberg, Brown and McGlone 1993; Cacciari and Glucksberg 1994; Glucksberg 1998, 2001; Vervaeke and Kennedy 1996; Vervaeke and Green 1997). But the success of CMT has been overwhelming, both in CL (e.g. for ideology research) and outside it, even in many other sciences such as therapy and social studies.18 A positive evolution is found in the development of Conceptual Metonymy Theory (CMyT), which is now an equivalent partner of CMT. In CMT metaphor and metonymy are regarded as processes of thought, and not just a question of language. In other words, metaphors and metonymies are seen as basically conceptual processes, which subsequently nd expression in a number of concrete forms, both verbal and non-verbal. From a linguistic viewpoint, however, the problem is that a given linguistic form is usually linked to various conceptual categories such that the form seems to trigger o or create the conceptual links between these categories, which might not be there without the linguistic form. In theory, one cannot fully ignore the impact of the linguistic form on the coherence between diverse metaphorical conceptualizations. At least, this holds for verbal metaphor, but perhaps far less so, or even not at all, for non-verbal metaphors. Before going into a more detailed analysis of CMT, it may be appropriate to deal rst with some of the criticism on the basic CMT claims. The priority attributed by CMT to the conceptual level has been criticized on the grounds that one rst needs a literal linguistic expression and its semantic-conceptual domain, e.g. the fruit of a plant, before it can be mapped onto a dierent domain, e.g. the fruit of ones hard work (Bartsch 2002; Warren 2002; Haser 2005). However, this criticism is rather vacuous, since metaphor can be expressed, not only in language, but also in visual, audio or sensory-motor media such as drawing, painting, architecture, music, dance and gesture. The mere existence of non-verbal or multimodal metaphors (c.f. Forceville 2006) is a clear counterargument, since there is no reason why verbal or linguistic metaphors should be submitted to dierent conditions than non-verbal ones.19
18. To quote just a few examples of ideology research: see Dirven, Hawkins and Sandikcioglu (2000), Dirven, Frank and Ilie (2001), Dirven, Frank and Putz (2003), Grundy and Jiang (2001), Morgan (2001), and Verboven (2003); for therapy and social studies, see Lawley and Tompkins (2000, 2006) and Schmitt (2005). 19. We should bear in mind that linguistic expressions are not the carriers of meaning but simply pointers to meaning. This idea, which is widespread in CL, lies at the basis of other approaches to communication, especially Relevance Theory in the domain of inferential pragmatics (Sperber and Wilson 21995; Ruiz de Mendoza and Perez 2003).


Rene Dirven and Francisco Jose Ruiz de Mendoza Ibanez

The question is then rather: what exactly is the conceptual element in verbal or non-verbal metaphors according to CMT? Lako and Johnson (1980) start from the assumption, fully spelled out in Johnson (1987), that metaphor is ultimately grounded in the human perceptual system and experiential world. The perceptual system is based on a number of pre-conceptual, most of all spatial, congurations which allow humans to react to, and manipulate, the world around them. This assumption is basically in agreement with the phenomenologist views propounded by Merleau-Ponty and Violi, as discussed in Section 7 above. The perceiver imposes structure on the things perceived, as is clearly expressed in the following quotation taken from Violi (2008: 57):
According to Merleau-Ponty, meaning is in the rst place articulated in our body, through perception. Also for the French philosopher perception is not merely the simple and passive record of an external world, already structured and pre-given in its conguration; perception is rather the active construction of a world already endowed with meaning and intentionality.20

The label Lako and Johnson coined for such perceptual congurations is image schemas (see Hampe 2005: 1), i.e. topological, pre-conceptual structures that encompass sensory-motor and visual schemas such as motion (animate, inanimate, caused or self motion), containment/container, path (source-path-goal), link, part-whole, center-periphery, balance, contact, support, blockage, verticality (up-down), horizontality (front-back), and proximity-distance.21 As the human mind and language develop, image schemas serve as the basis for categorizing phenomena in the physical domain and, by means of metaphor, phenomena in concrete domains are used to come to terms with experiences in the more abstract mental domains. Thus we tend to conceptualize the emotion of anger as a heated substance via the conceptual metaphor (hence in upper case), love is a hot fluid in a container, which may be expressed in various linguistic metaphors such as My blood was boiling, He was seething with anger, and

20. For some more detailed, though brief information on Merleau-Pontys views, see the Stanford Encyclopedia of Philosophy at entries/merleau-ponty/ 21. For a full list, see Hampe (2005: 2); for a detailed corpus-based description of each schema in Johnsons list and its network relationships with other schemas, see Pena (2003).

Looking back at 30 years of Cognitive Linguistics


He blew his top. Time is envisaged as motion in space, either as inanimate motion via time is a moving object (The years ow by) or as self motion of an observer via time is an end point for a moving observer (We are coming up to Christmas). Events and states are captured in a complex event structure metaphor consisting of various subtypes such as states, changes of state, causes, actions, purposes, means, and diculties. All of these are conceptualized in terms of spatial image schemata: states are locations (be in doubt), change of state is change of location (get into trouble), action is self-propelled motion, purposes (of action) are destinations, means are paths (to destinations), and difficulties are impediments to motion or blockage. Lako and Johnson claim that such basic conceptual metaphors may well be universal since human bodily experience is basically the same all over the world. Research on the universality of conceptual metaphor has been burgeoning within CL. Although rst research results, for instance, on the spatialization of time seemed to conrm this universal tendency, it was soon found out that far more contrastive research was required: Marn (1996) for English and Spanish; Alverson (1994) for English, Mandarin Chinese, Hindi, and Sesotho (criticized in Tyler 1995). Even if the spatialization of time occurs in many cultures, there may be huge dierences. Thus Boroditsky (2001) shows in empirical research that the Western concept of time as a horizontal time line and the Chinese as a vertical one inuences the location of months before or after one another. Another remarkable dierence in metaphorical conceptualization was found by Ning Yu (1998) for the domain of emotion: whereas, according to Lako and Kovecses, English sees anger as a hot uid in a whole body con tainer (cf. Geeraerts and Grondelaers 1995 for an alternative analysis), Chinese sees it as hot gases in smaller body-part containers called belly re, kidney re, liver re, etc. Here traditional Chinese philosophy and medical practice leads to a far more concrete, bodily self-experience. The research by Boroditsky and Yu thus oers evidence both for the adequacy of Lako and Johnsons universal claims at a fairly abstract level and for the great, colorful variation of culture-specic realizations of these putative universal conceptual metaphors. Although Lako and Johnson (1980) did not neglect metonymy altogether, they dened it in terms of metaphor: metaphor was seen as a mapping, i.e. a set of correspondences across dierent conceptual domains, and metonymy as a mapping within one domain. However, their metaphor revolution received virtually all of the attention and it took almost another twenty years to fully redress the balance between meta-


Rene Dirven and Francisco Jose Ruiz de Mendoza Ibanez

phor and metonymy, a position that had already been defended some time before by Jakobson (1971). The new metonymy upsurge of the late nineties culminated in programmatic papers by Kovecses and Radden (1998), Radden and Kovecses (1999), Radden (2000), a book by Ruiz de Mendoza and Otal (2002), and in collective volumes by Panther and Radden (1999), Barcelona (2000), and Dirven and Porings (2002). For the sake of economy, we will concentrate here on Ruiz de Mendoza and Otal (2002: 2742, 5056), because, as already noted in Dirven (2005), they were the rst to dene metonymy fully independently from metaphor. Lako, Johnson, Turner, Radden, and others, see metonymy either as having a referential function, whereby one entity stands for another entity within the same cognitive domain, or else as serving as a reference point that oers access to the other entity. These approaches cannot, however, account for constraints on these referential functions. While it is not denied that these referential functions are often served by metonymies, it is not their essence. Ruiz de Mendoza and Otal (2002) conclude this from the existence of predicate metonymies such as Mary is a pretty face, in which a pretty face does not stand for the person of Mary, but is rather a very salient feature that even serves to sum up the whole person. The authors do agree with Lako and Johnsons domain criterion: in the case of a metaphor like John is a pig there are always at least two dierent domains involved (people are animals), whereas in metonymy only one domain is involved, usually a domain and a subdomain (people and their faces). Just like in metaphor source features are mapped onto target structures, also in metonymy we are dealing with mappings, i.e. conceived and imposed connections or correspondences between features in the source and the target. In metonymy, the salient features of the whole domain are mapped onto the subdomain or salient features of the subdomain are mapped onto the whole domain as in the above example: Mary not only has a pretty face but she is a pretty face. By saying this, the speaker identies Mary as a pretty face: the beautiful and attractive features of the face dominate the persons whole being. Unlike in CMT with its claim to one-directionality, i.e. from source to target, CMyT has, because of the domain/subdomain relationship, one more possibility. Thus, the pretty face example is a source-in-target metonymy: the source, pretty face, is part of and maps onto the target, Mary. The target is the most encompassing conceptual structure, or matrix domain, and, as such, it is the only domain available for further reference, as in The pretty face is paying her bill. This remarkable potential is called domain avail-

Looking back at 30 years of Cognitive Linguistics


ability.22 The opposite case is a target-in-source metonymy like Bush attacked Iraq, where Bush stands for army. Although traditionally examples like this have been seen as cases of part-for-part metonymy, Ruiz de Mendoza and Otal (2002: 55) claim that this example is a wholepart metonymy. Our encyclopedic knowledge about Bush is such that it includes his being the president of his country, the commander-in-chief of the army, the leader of the Republican Party, etc. But the target, i.e. the referent that is meant in the sentence about attacking Iraq, is only the U.S. army. So the target (army) is a subdomain of the source (Bush), which now serves as the matrix domain. Bush as the source is mapped onto the army as target. In this target-in-source metonymy, the target is no longer available for reference, because it is encapsulated in the source and, hence, no anaphoric reference to the intended referent is possible: Soon after Bush attacked Iraq, *it ( the U.S. army) took Bagdad. In other words, the target in a metonymy is only referentially accessible if it is not encapsulated in some other subdomain. The newer CL approach to metonymy not only re-analyzes metonymy itself, but has also rethought some central aspects of pragmatics in a cognitive sense. A pioneering eort in this respect is Thornburg and Panthers (1997) metonymic analysis of indirect speech acts. In their view, indirect speech acts of the types that Searle (1979) calls directive, commissive, and expressive such as promises, oers, requests, suggestions, etc., can be described in terms of metonymic models. In other words, just as a speaker can refer to a person by mentioning a salient feature of that person, a speaker can perform a speech act by calling up an attribute of that speech act, e.g. the speech act participants ability (Can you pass the salt?), their volition (Would you please sign here?), or the reason to perform the action (Why dont we stop here?). This approach works not only for overt speech acts, but also for other constructions such as hedged performatives (May I ask if youre married?) and sentences expressing sense perceptions (You can hear the motorway from here) and mental activities (I am thinking aloud now), all of which corroborate the hypothesis that speech act metonymies are just special applications of more general
22. The Domain Availability Principle, rst formulated in Ruiz de Mendoza (2000), states that only the matrix domain of a metonymic mapping is available for anaphoric reference. It does not work alone but in combination with other factors of domain prominence and conceptual chaining that have been explored in detail in Ruiz de Mendoza and Dez (2004).


Rene Dirven and Francisco Jose Ruiz de Mendoza Ibanez

metonymic principles. In Panther and Thornburg (2003) this insight is variously applied to pragmatic inferencing. The contributions to that volume aim to bridge the gap between CL and inferential pragmatics. Topics not only include the role of metonymically-based inferences in speech act and discourse interpretation, but also grammar itself. Further questions dealt with are: what is the pragmatic meaning of grammatical constructions, what is the impact of metonymic mappings on grammatical structure, and what is the role of inferencing in linguistic change? It may thus be useful to refer to all these connections with the label cognitive pragmatics, which is intended to capture all attempts to explain most, if not all, major pragmatic phenomena in terms of the standard conceptual tools provided by CL. The relevance-theoretic distinction between explicatures and implicatures is a clear case in point. Ruiz de Mendoza and Perez (2003) have made this inferential pragmatics distinction compatible with Lakos notion of idealized cognitive model. The explicatures of an utterance are the result of a number of cognitive operations like expansion or reduction, whereby, due to metonymy, the utterance meaning can be associated with more meaning (expansion) or with less meaning (reduction). Implicatures are enabled by the processes of correlation or resemblance, which are forms of metaphor. Another example of the power of the explanatory tools of CL is provided by Ruiz de Mendoza and Baicchis (2007) study of illocutionary constructions in terms of formmeaning pairings where the meaning part consists of high-level situational scenarios like the Cost-Benet cognitive model and a specication of the conditions that allow metonymic access to such scenarios. Exploring this avenue further, cognitive pragmatics is likely to become a more central issue for the core of the CL paradigm.

9. Lexical semantics: Meaning selection or meaning construction? Over the last 30 years, views of categorization have changed dramatically. This has aected lexical semantics in three ways: (i) prototype theory has largely replaced feature theory and the classical denition of categories; (ii) the structure of word meaning is understood as a semantic network resulting from semantic extensions via the gurative processes of metaphor and metonymy, and the non-gurative processes of specication, and generalization (the latter two will not be gone into in much detail); and (iii) lexical and phrasal meaning, which is often highly idiomatic, is seen as contextually determined.

Looking back at 30 years of Cognitive Linguistics


First, we will look back at prototype theory vs. classical denitions. Early CL had already given up the belief in the Aristotelian or classical denition of categories, which makes the following rigid assumptions: all members of a category, e.g. fruit, share some essential feature(s); these features are necessary and sucient; all category members have equivalent status as members, and category boundaries are clear-cut. However, no linguist has ever managed to set up a list of such basic features.23 Even if one selects some putative features, they never apply to all members of the category. Suppose that for the category fruit characteristics such as sweet, soft and having seeds are necessary and sucient features. In this case several types of fruit would be outside the category: lemons, because they are not sweet, avocados, because they are not necessarily soft, and bananas, because they have no seeds. Strawberries are more like rhubarb because both grow on the ground, not on bushes or trees. Are they fruits? Why is a strawberry a fruit, while rhubarb is a vegetable? It is this fuzziness within or between categories that has suggested the necessity of a dierent approach to categorization, such as the prototype view (Rosch 1977, 1978; Berlin and Kay 1969; Lako 1987; Geeraerts 1989). On this view, categories do not just reect feature assemblies, but are rather manmade constructs consisting of prototypical members such as apples, pears and oranges for the category fruit, and more marginal members such as avocados, lemons and strawberries. Members of a category do not have equivalent status, and category boundaries are not clear-cut (nuts grow on trees, but do not share any of the three basic features). Categories are
23. There are, however, two possible exceptions that come from neighboring research enterprises. One is Wierzbickas (1996) Natural Semantic Metalanguage, which consists of around 70 primitive concepts, which are claimed to be universal. However, such primitives do not lead to classical denitions, but to exhaustive explications of the conceptual components of lexical and grammatical units. Mairal and Faber (2007) and Ruiz de Mendoza and Mairal (2008), within a cognitive-functional orientation, have incorporated similar primitives to the metalanguage used to dene lexical constructions, which they call lexical templates. These authors have obtained their semantic primitives through extensive and systematic factorization of usage-based meaning denitions, which they have grouped hierarchically on the basis of standard lexematic procedures (cf. Faber and Mairal, 1999). Ruiz de Mendoza and Mairal (2008) provide a survey of the conditions for the incorporation of lexical templates into constructional templates (i.e. argument constructions) within the context of their Lexical Constructional Model. For reasons of space, we cannot go into these strands, which are compatible with the essentials of CL.


Rene Dirven and Francisco Jose Ruiz de Mendoza Ibanez

to some extent also based on family resemblances as shown by Wittgenstein (1953) for the German category Spiele games, which contains such diverse members as a football match, a theatre play, or gambling. Rosch (1978) also provided psychological evidence for prototype eects in categorization; statements about central members of a category are processed far more quickly than statements about marginal members, and reasoning about any category is based on what is known about good examples of the category. Thus, for Spiel game, the feature of entertainment or amusement is most central. The second insight in lexical semantics is not really new, but is new in view of generative linguistics, which postulates the analytical preponderance of monosemy, while CL has generally considered polysemy to be the rule. However, recently a number of cognitive linguists have moved towards the monosemy view. In the monosemist views of meaning, as stated by Bierwisch and Schreuder (1992), words have only one basic meaning, which may however get dierent applications to entities in the world. This is managed via an interface between language and thought (cf. Taylor 1995b). The objection to this solution is that it may work nicely for words denoting man-made entities such as artifacts or institutions, like the word university, but not for words designating natural categories and their metaphorical and metonymic extensions, such as fruit in fruit and vegetables, the fruit of the womb, the fruits of the earth, and the fruits of his work. But even in the case of words like university, the appeal to an interface is unnecessary. Thus, this word can refer to a building, an institution for learning and research, a period of time in a persons life, an academic qualication for a certain profession, etc. These senses need not be accounted for separately, since they are all linked to each other by what Dirven (2002b: 94) calls conjunctive metonymy, which relates to complex entities consisting of many components and in which a part of the complex entity may metonymically stand for the whole and vice versa. Given this basic metonymic link, an interface would not be needed at all. In the same way, words denoting relations such as the preposition or particle over or the verb to run are extremely diversied and hence less manageable by a putative interface. Admittedly, the reaction of early CL to monosemism may have been too extreme and researchers may have overreacted by trying to nd too many dierent senses and explaining the motivation for the links among them. In a more critical approach, the main problem for CL is to nd a good balance between the number of dierent uses or senses (indicated by subscripts) and the motivation for them. The item fruit is a good case in point. In its prototypical use in a given cultural context, fruit1 refers

Looking back at 30 years of Cognitive Linguistics


to something such as an apple, banana, or strawberry that grows on a tree or other plant and tastes sweet (Longman Dictionary of Contemporary English; LDCE ). In this sense we can oppose fruit1 to vegetables, e.g. fresh fruit and vegetables. But in a technical sense, fruit2 is the part of a plant, bush, or tree that contains the seeds (LDCE ).24 In this narrower sense, potatoes and all other root crop are fruits. Obviously, these two senses of one word form are mutually exclusive. Fruit2 is the result of specialization, which is, just like metonymy, one of the four dierent cognitive processes of meaning extension, the two others being generalization and metaphor. An instance of generalization is the meaning of fruit3 , i.e. all the natural things that the earth produces such as fruit, vegetables or minerals (LDCE ). Metaphorical extension has applied to fruit4 as in the fruits of ones work, meaning the good results from working very hard (LDCE ). Although the four conceptual categories designated by the same word form are independent of one another, the human conceptualizer may see similarities and extend the use of this form from one category to others. Such extensions are neatly recorded in dictionaries as conventionalized senses, but this should not trap us into thinking that words do indeed have clearly dened dierent meanings. What is at stake is rather that a given word form is applied in a number of dierent contexts of use and thereby exploits one or several of the four processes of semantic extension. Although the various senses of a word may be clearly interrelated, the question is whether they are conceptually so strongly linked as was thought in traditional semantics and also in early Cognitive Semantics, which posited a dierent sense for each use in a dierent context, trying to motivate the links between them. Thus, Lako (1987: 416461) saw no less than 24 dierent senses in the English preposition over (see also Brugman and Lako 1988). In a critical evaluation of Lako s analysis, Tyler and Evans (2001) managed to show that the senses of over can be reduced to half that number. The various senses of over presuppose the image of a full-arc trajectory (ABC), suggesting the motion from ground level A to a higher level B and back to ground level C at the other end. Though the arc trajectory is presupposed, it is not part of the meaning of over. This proto-image is exemplied in example (0) below, but the basic meaning of over, also applicable to static contexts, is only that of position B in the trajectory, i.e. higher than and approximate to. Apart from this proto-sense, the senses of over and several metaphorical extensions can be grouped into ve clusters, illustrated in (1)(5) below: (1) the other side
24. It is this sense that is exploited by the Latin-based metaphor fruit of the womb.


Rene Dirven and Francisco Jose Ruiz de Mendoza Ibanez

Figure 3. The semantic network for over

or end of the ABC trajectory cluster; (2) the Coverage schema; (3) the Verticality cluster; (4) the Up cluster; and (5) the Reexive cluster. Tyler and Evanss semantic network representing the links between the protosense, the 5 clusters and the 12 senses is presented below in Figure 3, where the senses marked as (aa) have a closer meaning relationship to those marked (a) than to the rest of the senses. (0) He jumped over the wall. (1) a. Arlington is over the Potomac River from Georgetown. b. The lm is over. c. Sally turned the keys of the oce over to the janitor. (2) The tablecloth is over the table. (3) a. Mary looked over the manuscript quite carefully. b. The little boy cried over his broken toy. (4) a. Jerome found over forty kinds of shells on the beach. aa. The heavy rain caused the river to ow over its banks. b. She has a strange power over me. c. I would prefer tea over coee. (5) a. The fence fell over. aa. After the false start they started the race over.

Looking back at 30 years of Cognitive Linguistics


The temporal or causal extensions within each cluster of senses are fully motivated by the particular spatial sense given rst. Further, the coherence of the clusters of senses within the whole semantic network of over is accounted for. Also the criteria for distinct senses are justied. Thus there is no distinct sense if the use of the meaning purely results from the context or from encyclopedic knowledge, which is contradicted by Gries (2006), as we will see later on in this section. Still, in doing all this, Tyler and Evans have convincingly shown the usefulness of the very notion of semantic network. However, criticism of the validity and value of semantic networks had already been expressed in the mid nineties. Taylor (1995b) has shed doubts on semantic networks as a descriptive tool, since dierent authors had come to dierent results for the same lexical item. Sandra and Rice (1995) have questioned the psychological reality of semantic networks. On the basis of empirical work, Rice (2003) has claimed that the various uses of polysemous items are acquired and stored in a relatively piecemeal, idiomatic and construction-based fashion. This suggests that semantic networks are linguists representations of motivated extensions of form-meaning pairings, but not mental representations of a words polysemy by language users. The semantic network may thus remain useful as a summary of the motivated links between the various senses of a polysemous lexical item, and for the same reason, of a grammatical construction too. The third new insight into lexical semantics concerns the contextuallybased determination of lexical meaning. Here we can distinguish between a purely theoretical approach with a strong monosemist orientation and a fully empirical approach with a polysemist outlook, based on corpus studies. The monosemist approach is exemplied by the papers by Allwood, Zlatev and Janssen in Cuyckens, Dirven and Taylor (2003), which are neither extremely monosemist nor extremely polysemist, but rather regard the lexicon as strongly context-adaptive or, seen from the hearers viewpoint, as strongly context-dependent. Assuming that a theory of meaning is more operational than representational, these authors suggest that some very few basic senses of an expression (in the limiting case, just one) serve as prompts for meaning construction in on-line discourse, combining with contextual interpretation to yield the necessary inferences about distinct semantic functions of an expression in distinct environments. Instead of trying to specify a xed set of senses for a word, Allwood (2003: 43) introduces the notion of meaning potential, which is the union of its individually or collectively remembered uses. Wright (2007: 571572) has criticized the wide gap between this denition of meaning potential and the absence of any procedure to specify it:


Rene Dirven and Francisco Jose Ruiz de Mendoza Ibanez [. . .] if lexical senses are mapped one-to-one into monosemous meaning potentials. . . , then specifying the meaning of any given word is just to specify the union of all information which has ever been conveyed in using it; and since no language user or linguist has access to all such information, the meaning potential of an expression can at most be approximated but never be specied.

Zlatev (2003, 2007) also speaks in terms of use potential, which he deals with on the basis of connectionist networks. Wright (2007: 573) criticizes Zlatevs understanding of use potentials as minimally dierentiated language games, and as an amalgam of ideas from Merleau-Ponty, Wittgenstein, Searle, Dreyfus, Langacker, Talmy, and many others. In the same monosemist vein, Janssen (2003) considers words as having undivided highly abstract meanings, which he illustrates by means of the behavior of Dutch possessives. Queller (2003) rejects sense-tosense mapping operations (e.g. Lakos image-schema transformations or metaphorical and metonymic extensions) in favor of the emergence of new, freestanding senses via an abductive reanalysis of usage events. All in all, however challenging these new orientations may be, the fact is that little or no empirical work has been brought forward to support the hypothesis of non-polysemy. Consequently, these theoretical discussions can only remain hypotheses. A totally dierent perspective has been opened up in the last ten years by the marriage between lexical semantics and corpus linguistics, which by denition looks at language in context, with the sentential context as the lower limit. Two successful CL approaches in this direction are the Leuven group around Dirk Geeraerts and the German group around Anatol Stefanowitsch and Stefan Gries. As a good representative illustration we will concentrate on Griess (2006) analysis of to run, for which he has selected 815 sentences from the International Corpus of English (ICE) for British English and the Brown Corpus for American English. The 815 sentences exhibit about fty dierent uses (or senses), including uses with prepositions or particles such as run into somebody, i.e. meet. In order to be able to handle the corpus material quantitatively and statistically, all sentences are rst coded for a variety of linguistic parameters (so-called ID tags) such as word class, syntactic analysis in terms of subject and object, verb form, etc., which is done partly automatically, but to a large extent manually, because there are no fewer than 252 dierent attributes or tags to choose from. This analysis of all examples yields a complete syntactic-semantic picture or behavioral prole of this verb (see Table 1 below). The quantitative method provides objective empirical evidence for questions of the prototypical meaning of to run, the (degree of ) sense

Looking back at 30 years of Cognitive Linguistics


distinctness, the clustering of senses, the structure of a possible semantic network and the possibility of automatic sense identication. The most frequent use of to run is that of fast pedestrian motion, which applies in 25 per cent of the corpus, or about 200 out of the 815 sentences. Other quantitative indicators of prototypicality are the sense with the greatest variety of prepositions or particles, the sense used most frequently in the nominal derivation of the verb, the sense used rst in child language acquisition and the sense rst encountered in diachronic perspective. For all these criteria the sense of fast pedestrian motion comes on top. The distinctiveness of dierent senses is much more dicult to decide. Thus, in its sense of to ow the verb to run always has a liquid as its subject, but according to two criteria by Tyler and Evans (2001) this use of to run with the meaning to ow would not be a distinctive sense. Indeed, it is a sense that follows from the context (the subject is a liquid) and, on top of that, it follows from our encyclopedic knowledge (only liquids can ow). So on the ground of those two criteria, the sense of to ow would not be a distinctive sense of to run, which is certainly counterintuitive. Gries attaches far more importance to the intra-clustering of senses in a cluster than to sense distinctiveness and to their inter-clustering and their eventual conguration in a semantic network. Thus the sense of (non-pedestrian) motion in There are three boats that run from the mainland to the Island is clearly more distinct than the dierences between fast and non-fast pedestrian motion. Or again, all motion senses are more distinct from the sense of functioning in The machine is running than they are to one another. A more general conclusion is the astonishing correlation between syntactic and lexical-semantic distinctions. Thus in all the senses discussed so far to run functions as an intransitive verb, but in its transitive use its most prototypical sense is that of to manage. The contrast between these intransitive and transitive uses, and in fact a summarizing illustration, of the corpus approach may appear from the following table (Gries 2006: 86). A relatively dierent approach is provided by Geeraerts and his Leuven team is basically the same, but it also diers in several respects. Thus in Geeraerts, Grondelaers and Bakema (1994), which is an investigation of Dutch clothing terms, the authors more strongly put the emphasis on lexical variation, not only nding out the various senses of words (the semasiological dimension), but also the various terms used for a given object such as the various terms referring to a jacket (the onomasiological dimension). This wider scope guarantees the embedding of cognitive linguistics in the wider tradition of 19th and 20th century semantics, as practiced in linguistics and philology. The quantitative statistical methods


Rene Dirven and Francisco Jose Ruiz de Mendoza Ibanez

Table 1. Complementary distribution of selected ID tags for the two most frequent senses of run fast pedestrian motion verb form clause type subject preposition of following PP ran runs intransitive transitive main clause, imperative clause (zero) relative clause, interrogative clause human, animate concrete objects, organization/institution towards, for, down, after, up to manage run (past part.) ran transitive intransitive zero relative clause, zero subordinate clause main clause organization/institution 0


as practiced in sociolinguistic research have always been part and parcel of Geeraerts work, but a major dierence with the recent turn in CL towards corpus linguistics is that they have also been applied both diachronically and sociolinguistically. Thus, for example, Geeraerts, Grondelaers and Speelman (1999) continue and extend the research on clothing terms by Geeraerts, Grondelaers and Bakema (1994), but they now investigate both clothing and soccer terms and cover three periods in the development of their use, both in Belgian Dutch and in Netherlandic Dutch. All these diachronic and sociolinguistic variation studies are also theoretically and programmatically explored in Geeraerts (2005, 2006a), which all stress the inevitability of a cognitive sociolinguistics. It is only to be hoped that these diverse CL approaches to lexical semantics will somehow integrate their research on prototype eects, semantic exibility and variation into a usage/corpus-based and quantitativequalitative synthesis. 10. Discourse: Mental Space Theory and Conceptual Blending Theory Cognitive Linguistics is not only a lexico-grammatical theory of language; it also embraces the whole of language functions and structure, including

Looking back at 30 years of Cognitive Linguistics


both pragmatic aspects, as shown in Section 8, and the discourse dimension, which is to be discussed now. In discourse, various knowledge systems, linguistic and non-linguistic ones, are invoked, which Fauconnier (1985/21994) calls mental spaces. Each utterance, even each content word, in discourse reects and evokes a mental representation of some situation and/or some entities taking part in it. For the encoding and interpretation of mental representations speakers and hearers draw, not only on the linguistic expression as the gure, which mainly gives hints for interpretation, but also on the speech situation, and on their encyclopedic or world knowledge, which, in Langackers view (see Section 3), all together form the ground. Each utterance is grounded in a basic mental space which is primarily the speakers perspective, but which comes to be shared by other participants in the speech event. This speech situation is the base space (space 0). In the base space we can open new spaces as illustrated in an often quoted and much discussed example: I dreamt I was Marilyn Monroe and kissed me. How can the obligatory use of the phrase I kissed me instead of I kissed myself be explained here? In the example, the phrase I dreamt is part of the base space, and the verb dream is a space-builder opening a new space (space 1) of an imagined world in which the second I (was Marilyn Monroe) is no longer identical with the rst I (dreamt) in the base space, but is part of a new knowledge frame in which Marilyn Monroe is not kissing herself, but the speaker, i.e. the I in the base space. Mental Space Theory (MST) initially started out as a cognitive alternative to traditional theories of reference and managed to solve many of the referential problems left unsolved by logic-oriented trends in generative linguistics. Gradually MST has, in the work of Fauconnier and Sweetser (1996) and Fauconnier (1997), developed into an encompassing cognitive theory of discourse and discourse management (see also Oakley and Hougaard 2008). In the development of the ongoing discourse, speaker(s) and hearer(s) have to keep track of all the mental spaces opened up for referents and they can at any time go back to any of them to elaborate them further. Still, MST is not limited to reference or discourse management, but also explores grammatical problems such as the tense-aspectmodality system serving the grammatical functions of perspective, viewpoint, epistemic distance, and grounding. This concept of grounding, which is Langackers term for anchoring the speech act and the whole discourse in the actual reality of the speech act situation at speech time (see Section 3), can directly be linked to Fauconniers notion of base space. Furthermore, MST is also at the basis of Turner and Fauconniers (1995) metaphor theory, which is one of the applications of Fauconniers Conceptual Blending Theory (CBT). In their view, the source domain of


Rene Dirven and Francisco Jose Ruiz de Mendoza Ibanez

a metaphorical expression is not just mapped onto the target domain as proposed in Lako and Johnsons Conceptual Metaphor Theory, but both the source and target domains are input spaces, which, via the generic space containing their common elements, are blended or integrated into the blend, which may also contain new emergent meaning, not present in the input spaces. Comparing the divergences between the two cognitive metaphor theories and those between the various grammar theories, it is not dicult to see that there is far less opposition between the former than between the latter. CBT is not a rival of Conceptual Metaphor Theory and certainly does not intend to replace it. The two models of metaphorical mapping are to be seen as complementary to each other rather than as rivals, as claimed by Grady, Oakley and Coulson (1999: 101): blending theorists are said to look at the particulars of single metaphorical expressions, whereas conceptual metaphor theorists are more interested in generalizations over conventional patterns across metaphors. Also here the desirability of a stronger integration of the various CL strands into a more synthetic view is strongly felt. This line of thought is also followed by Langacker (2003a, 2003b), who is making various attempts to come closer to and integrate views from both Fauconniers MST and Goldbergs CxG into his Cognitive Grammar model. That a harmonic integration of the various CL strands is feasible at the descriptive level has now been proven in Radden and Dirvens (2007) Cognitive English Grammar, from which an analysis in terms of conceptual blending for the extended caused-motion construction in Fred sneezed the napkin right o the table is presented here (Figure 4). This blending analysis allows one to unite CBT with either Langackers Cognitive Grammar or Goldbergs CxG. In fact, one input space is still lacking here, i.e. the as-path-specied motion that the napkin gets o the table. Lets call it input 3.25 We can interpret the causal link represented by input space 2 either as an element of encyclopedic knowledge, uniting two successive events in perception as one causing the other. This would be a possible representation of our own extension of Langackers line of
25. Blending Theory postulates inconsistencies and asymmetries between input spaces that are resolved in the blend. This thesis has been disputed in Ruiz de Mendoza (1998) and Ruiz de Mendoza and Pena (2005), who basically claim that inconsistencies and non-correspondences only arise when not enough input spaces have been activated. The blend in this alternative view becomes a mere repository of structure projected in a principled, non-arbitrary way from multiple input spaces, but not the creator of new, emergent structure.

Looking back at 30 years of Cognitive Linguistics


Figure 4. Blending in Fred sneezed the napkin o the table

reasoning for the phrase patriotic pole-climber. But input 2 can also be invoked as the spelled-out representation of Goldbergs predicate-argument structure with the abstract predicate cause-move and the arguments <cause goal theme>. The great advantage of the Conceptual Blending approach is that one does not have to pin down oneself on either a purely-conceptual-perceptual interpretation of the causal link or else a conceptual-linguistic one. Would it be conceivable that CL is going to look for viable solutions to its richly dierent approaches along these or similar lines? 11. Conclusion This necessarily brief survey has attempted to highlight the main areas of interest of the CL enterprise from the point of view of its unity despite apparent diversity and of its ability to interact in a fruitful way with dierent disciplines related to language, mind, and the social context. It has also put into focus the strong descriptive and explanatory potential of the dierent strands of CL, which, far from being mutually contradictory accounts of language, can be regarded as highly complementary. The dierences are thus a matter of shifts of emphasis and can be reconciled, as we have tried to show, and integrated into a broad, ambitious picture


Rene Dirven and Francisco Jose Ruiz de Mendoza Ibanez

of language use in relation to the embodied mind and specic cultural exploitations of its universal features. The necessarily unnished picture that 30 years of theoretical growth of CL presents is suggestive, colorful, and dynamic, with each and every element of the overall approach gradually taking shape in its appropriate place. References
Allwood, Jens 2003

Meaning potentials and context: Some consequences for the analysis of variation in meaning. In: Hubert Cuyckens, Rene Dirven and John Taylor (eds.), Cognitive Approaches to Lexical Semantics, 2965. Berlin/New York: Mouton de Gruyter.

Alverson, Hoyt 1994 Semantics and Experience: Universal Metaphors of Time in English. Baltimore/London: John Hopkins University Press. Barcelona, Antonio (ed.) 2000 Metaphor and Metonymy at the Crossroads: A Cognitive Perspective. Topics in English Linguistics 30. Berlin/New York: Mouton de Gruyter. Bartsch, Renate 2002 Generating polysemy: Metaphor and metonymy. In: Rene Dirven and Ralf Po rings (eds.), Metaphor and Metonymy in Comparison and Contrast, 4974. Berlin/New York: Mouton de Gruyter. Bergen, Benjamin 2005 Mental simulation in literal and gurative language understanding. In: Seana Coulson and Barabara Lewandowska-Tomaszczyk (eds.), The Literal and Nonliteral in Language and Thought, 255 280. Frankfurt: Peter Lang. Bergen, Benjamin K. and Nancy Chang 2005 Embodied construction grammar in simulation-based language understanding. In: Jan-Ola Ostman and Mirjam Fried (eds.), Construction Grammars: Cognitive Grounding and Theoretical Extensions, 147190. Amsterdam/Philadelphia: John Benjamins. Berlin, Brent and Paul Kay 1969 Basic Color Terms. Their Universality and Evolution. Berkeley/ Los Angeles: University of California Press. Bierwisch, Manfred and Robert Schreuder 1992 From concepts to lexical items. Cognition 42: 2360. Boroditsky, Lera 2000 Metaphoric structuring: Understanding time through spatial metaphors. Cognition 75: 128. Boroditsky, Lera 2001 Does language shape thought? Mandarin and English speakers conceptions of time. Cognitive Psychology 43: 122.

Looking back at 30 years of Cognitive Linguistics


Boroditsky, Lera and Michael Ramscar 2002 The roles of body and mind in abstract thought. Psychological Science 13: 185189. Brugman, Claudia and George Lako 1988 Cognitive topology and lexical networks. In Lexical Ambiguity Resolution. Perspectives from Psycholinguistics, Neuropsychology, and Articial Intelligence, Steven L. Small. Garrison W. Cottrell, and Michael K. Tannenhaus (eds.), 477507. San Mateo, Cal.: Morgan Kaufmann Publishers. Cacciari, Cristina and Sam Glucksberg 1994 Understanding gurative language. In: Morton Ann Gernsbacher (ed.), Handbook of Psycholinguistics, 447477. San Diego, Calif.: Academic Press. Calabretta, Raaele, Andrea Di Ferdinando, Gunter P. Wagner and Domenico Parisi 2003 What does it take to evolve behaviorally complex organisms? BioSystems 69: 245262. Camac, Mary K. and Sam Glucksberg 1984 Metaphors do not use associations between concepts, they are used to create them. Journal of Psycholinguistic Research 13: 443455. Croft, William 2001 Radical Construction Grammar: Syntactic Theory in Typological Perspective. Oxford: Oxford University Press. Croft, William and D. Alan Cruse 2004 Cognitive Linguistics. Cambridge: Cambridge University Press. Csab, Szilvia 2004 A cognitive linguistic view of polysemy in English and its implications for teaching. In: Michel Achard and Susanne Niemeier (eds.), Cognitive Linguistics, Second Language Acquisition, and Foreign Language Teaching, 233256. Berlin/New York: Mouton de Gruyter. Cuyckens, Hubert, Rene Dirven and John Taylor (eds.) 2003 Cognitive Approaches to Lexical Semantics. Berlin/New York: Mouton de Gruyter. Deane, Paul D. 1996 On Jackendo s conceptual semantics. Cognitive Linguistics 7 (1): 3591. Dirven, Rene 2002a Cognitive linguistics. In: Kirsten Malmkjaer (ed.), The Linguistics Encyclopedia. Second Edition, 7682. London/New York: Routledge. Dirven, Rene 2002b Metonymy and metaphor: Dierent mental strategies of concep tualisation. In: Rene Dirven and Ralf Porings (eds.), Metaphor


Rene Dirven and Francisco Jose Ruiz de Mendoza Ibanez and Metonymy in Comparison and Contrast, 75111. Berlin/New York: Mouton de Gruyter.

Major strands in cognitive linguistics. In: Francisco J. Ruiz de Mendoza Ibanez and Sandra Pena Cervel (eds.), Cognitive Lin guistics: Internal Dynamics and Interdisciplinary Interaction, 69 100. Berlin/New York: Mouton de Gruyter. Dirven, Rene, Bruce Hawkins and Esra Sandikcioglu (eds.) 2000 Language and Ideology. Vol. I: Theoretical Cognitive Approaches. Current Issues in Linguistic Theory 204. Amsterdam/Philadelphia: John Benjamins. Dirven, Rene, Roslyn Frank and Cornelia Ilie (eds.) 2001 Language and Ideology. Vol. II: Descriptive Cognitive Approaches. Current Issues in Linguistic Theory 205. Amsterdam/Philadelphia: John Benjamins. Dirven, Rene and Ralf Porings (eds.) 2002 Metaphor and Metonymy in Comparison and Contrast. Berlin/ New York: Mouton de Gruyter. Dirven, Rene, Roslyn M. Frank and Martin Putz (eds.) 2003 Cognitive Models in Language and Thought: Ideologies, Metaphors, and Meanings, 123157. Berlin/New York: Mouton de Gruyter. Dirven, Rene, and Marjolin Verspoor 22004 Cognitive Exploration of Language and Linguistics. Amsterdam/ Philadelphia: John Benjamins. Dreyfus, Hubert L. 2002 Intelligence without representation: Merleau-Pontys critique of mental representation. Phenomenology and the Cognitive Sciences 1: 367383. Dreyfus, Hubert L. and Stuart E. Dreyfus 1999 The challenge of Merleau-Pontys phenomenology of embodiment for cognitive science. In: Gail Weiss and Honi Fern Haber (eds.), Perspectives on Embodiment: The Intersections of Nature and Culture, 103120. New York: Routledge. Evans, Vyvyan and Melanie Green 2006 Cognitive Linguistics: An Introduction. Edinburgh: Edinburgh University Press. Evans, Vyvyan, Benjamin K. Bergen and Jorg Zinken (eds.) 2006 The Cognitive Linguistics Reader. Hereford, U.K.: Equinox Publishing. Faber, Pamela and Ricardo Mairal 1999 Constructing a Lexicon of English Verbs. Berlin/New York: Mouton de Gruyter. Fauconnier, Gilles 1985 Mental Spaces: Aspects of Meaning Construction in Natural Language. Cambridge, Mass.: The MIT Press.

Dirven, Rene 2005

Looking back at 30 years of Cognitive Linguistics


Fauconnier, Gilles 21994 Mental Spaces: Aspects of Meaning Construction in Natural Language. Cambridge: Cambridge University Press. Fauconnier, Gilles 1997 Mappings in Thought and Language. Cambridge: Cambridge University Press. Fauconnier, Gilles and Eve Sweetser 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 Goldberg (ed.), Conceptual Structure, Discourse and Language, 113130. Stanford, CA: CSLI Publications. Fauconnier, Gilles and Mark Turner 2002 The Way We Think. Conceptual Blending and Minds Hidden Complexities. New York: Basic Books. Feldman, Jerome and Srini Narayanan 2004 Embodied Meaning in a Neural Theory of Language. Brain and Language 89: 385392. Fillmore, Charles J. 1977 Scenes-and-frames semantics. In: Antonio Zampolli (ed.), Linguistic Structures Processing, 5581. Amsterdam/New York: NorthHolland. Fillmore, Charles J. 1982 Frame semantics. In: The Linguistic Society of Korea (ed.), Linguistics in the Morning Calm, 111137. Seoul: Hanshin Pub. Co. Fillmore, Charles J. 1990 Construction grammar. Course reader for linguistics 120A. Ms. University of California. Fillmore, Charles J., Paul Kay and Catherine OConnor 1988 Regularity and idiomaticity in grammatical construction: The case of let alone. Language 64: 501538. 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, 91110. Oxford: Oxford University Press. Fillmore, Charles J., Paul Kay, Laura A. Michaelis and Ivan Sag 2004 Construction Grammar. Stanford, Calif.: CSLI Publications. Forceville, Charles 2006 Non-verbal and multimodal metaphor in a cognitivist framework: Agendas for research. In: Gitte Kristiansen, Michel Achard, Rene Dirven and Francisco Ruiz de Mendoza (eds.), Cognitive Linguistics: Current Applications and Future Perspectives, 379 402. Berlin/New York: Mouton de Gruyter.


Rene Dirven and Francisco Jose Ruiz de Mendoza Ibanez

rdez (eds.) Frank, Roslyn M., Rene Dirven, Tom Ziemke and Enrique Berna 2008 Body, Language and Mind, Volume 2: Sociocultural Situatedness. Cognitive Linguistics Research 35.2. Berlin/New York: Mouton de Gruyter. Gallese, Vitorio 2005 Embodied simulation: From neurons to phenomenal experience. Phenomenology and the cognitive sciences 4: 2348. Gallese, Vitorio and Gorge Lako 2005 The brains concepts: The role of sensory-motor system in conceptual knowledge. Cognitive Neuropsychology 22: 455479. Geeraerts, Dirk 1985 Paradigm and Paradox. Explorations into a Paradigmatic Theory of Meaning and its Epistemological Background. Leuven: Leuven University Press. Geeraerts, Dirk 1989 Introduction: Prospects and problems of prototype theory. Linguistics 27: 587612. Geeraerts, Dirk 1993 Cognitive semantics and the history of philosophical epistemology. In: Richard A. Geiger and Brygida Rudzka-Ostyn (eds.), Conceptualizations and Mental Processing in Language. A Selection of Papers from the First International Cognitive Linguistics Conference in Duisburg, 1989, 5379. Berlin/New York: Mouton de Gruyter. Geeraerts, Dirk 2005 Lectal variation and empirical data in Cognitive Linguistics. In: Francisco J. Ruiz de Mendoza Ibanez and Sandra Pena Cervel (eds.), Cognitive Linguistics: Internal Dynamics and Interdisciplinary Interaction, 163189. Berlin/New York: Mouton de Gruyter. Geeraerts, Dirk 2006a Methodology in Cognitive Linguistics. In: Gitte Kristiansen, Michel Achard, Rene Dirven and Francisco Ruiz de Mendoza (eds.), Cognitive Linguistics: Current Applications and Future Perspectives, 5180. Berlin/New York: Mouton de Gruyter. Geeraerts, Dirk 2006b Cognitive Linguistics: Basic Readings. (Cognitive Linguistics Research 34.) Berlin/New York: Mouton de Gruyter. Geeraerts, Dirk and Hubert Cuyckens (eds.) 2007 The Oxford Handbook of Cognitive Linguistics. New York: Oxford University Press. Geeraerts, Dirk, and Stef Grondelaers 1995 Looking back at anger: Cultural traditions and metaphorical patterns. In Trends in Linguistics, John Taylor, and Robert MacLaury (eds.), 153179. Berlin/New York: Mouton de Gruyter.

Looking back at 30 years of Cognitive Linguistics


Geeraerts, Dirk, Stef Grondelaers and Peter Bakema 1994 The Structure of Lexical Variation: Meaning, Naming, and Context. Berlin; New York: Mouton de Gruyter. Geeraerts, Dirk, Stefan Grondelaers and Dirk Speelman 1999 Convergentie en Divergentie in de Nederlandse Woordenschat. Een Onderzoek naar Kleding- en Voetbaltermen. ( Convergence and Divergence in Dutch vocabulary. An investigation into clothing and soccer terms). Amsterdam: Meertens Instituut. Glucksberg, Sam 1998 Understanding metaphors. Current Directions in Psychological Science 7: 3943. Glucksberg, Sam 2001 Understanding Figurative Language: From Metaphors to Idioms. (With a contribution by Matthew S. McGlone). Oxford: Oxford University Press. Glucksberg, Sam, Mary Brown and Matthew S. McGlone 1993 Conceptual metaphors are not automatically accessed during idiom comprehension. Memory and Cognition 21: 711719. Glynn, Dylan 2007 Iconicity and the grammar-lexis interface. In: Elzbieta Tabakow ska, Christina Ljungberg and Olga Fischer (eds.), Insistent Images, 269288. Amsterdam/Philadelphia: John Benjamins. Goldberg, Adele E. 1992 The Inherent Semantics of Argument Structure: The Case of the English Ditransitive Construction. Cognitive Linguistics 3: 37 74. Goldberg, Adele E. 1995 Constructions: A Construction Grammar Approach to Argument Structure. Chicago: University of Chicago Press. Goldberg, Adele E. 2002 Surface generalizations: An alternative to alternations. Cognitive Linguistics 13: 327356. Goldberg, Adele E. 2006 Constructions at Work: The Nature of Generalization in Language. Oxford: Oxford University Press. Gonzalvez, Francisco and Christopher S. Butler 2006 Mapping functional-cognitive space. Annual Review of Cognitive Linguistics 6: 3996. Grady, Joseph 1997 theories are buildings revisited. Cognitive Linguistics 8(4): 267290. Grady, Joseph, Todd Oakley and Seana Coulson 1999 Conceptual blending and metaphor. In: Raymond W. Gibbs, Jr. and Gerard J. Steen (eds.), Metaphor in Cognitive Linguistics, 101124. Amsterdam/Philadelphia: John Benjamins.


Rene Dirven and Francisco Jose Ruiz de Mendoza Ibanez

Gries, Stephan 2006

Corpus-based methods and cognitive semantics: The many senses of to run. In: Stefan Gries, and Anatol Stefanowitsch (eds.), Corpora in Cognitive Linguistics. Corpus-Based Approaches to Syntax and Lexis, 5799. Berlin/New York: Mouton de Gruyter. Grundy, Peter and Yan Jiang 2001 The bare past as an ideological construction in Hong Kong dis course. In: Rene Dirven, Roslyn Frank and Cornelia Ilie (eds.), Language and Ideology. Volume II: Descriptive Cognitive Approaches, 117134. Amsterdam/Philadelphia: John Benjamins. Haiman, John 1980 Dictionaries and encyclopedias. Lingua 50: 329357. Haiman, John (ed.) 1985a Iconicity in Syntax. Typological Studies in Language 6. Amsterdam/Philadelphia: John Benjamins. Haiman, John 1985b Natural Syntax: Iconicity and Erosion: Cambridge University Press. Hampe, Beate (In cooperation with Joseph E. Grady) (eds.) 2005 From Perception to Meaning: Image Schemas in Cognitive Linguistics. Berlin/New York: Mouton de Gruyter. Haser, Verena 2005 Metaphor, Metonymy, and Experientialist Philosophy: Challenging Cognitive Semantics. Berlin/New York: Mouton de Gruyter. Heidegger, Martin 1962 Being and Time. New York: Harper and Row. Heidegger, Martin 1982 The Basic Problems of Phenomenology. Bloomington, Ind.: Indiana University Press. Heiser, Marc, Marco Iacoboni, Fumiko Maeda, Jake Marcus and John C. Mazziotta 2003 The essential role of Brocas area in imitation. European Journal of Neuroscience 17: 11231128. Heyvaert, Liesbet 2003 A Cognitive-Functional Approach to Nominalization in English. Berlin/New York: Mouton de Gruyter. Husserl, Edmund 1960 Cartesian Meditations. The Hague: Martinns Nijho. Husserl, Edmund 1970a Logical Investigations, 2 vols. New York: Humanities Press. Husserl, Edmund 1970b The Crisis of the European Sciences and Transcendental Phenomenology. Evanston: Northwestern University Press.

Looking back at 30 years of Cognitive Linguistics


Jackendo, Ray 1996 Conceptual semantics and cognitive linguistics. Cognitive Linguistics 7: 93130. Jackendo, Ray 2002 Foundations of Language: Brain, Meaning, Grammar, Evolution. Oxford: Oxford University Press. Jakobson, Roman 1971 The metaphoric and metonymic poles. In: Roman Jakobson and Maurice Halle (eds.), Fundamentals of Language, 5482. The Hague: Mouton. Again in: Dirven and Porings (eds.), 2002, Metaphor and Metonymy in Comparison and Contrast, 4147. Berlin/New York: Mouton de Gruyter. Janssen, Theo A. J. M 2003 Monosemy versus polysemy. In: Hubert Cuyckens, Rene Dirven and John Taylor (eds.), Cognitive Approaches to Lexical Semantics, 93122. Berlin/New York: Mouton de Gruyter. Johnson, Mark 1987 The Body in the Mind: The Bodily Basis of Meaning, Reason and Imagination. Chicago: University of Chicago Press. Kay, Paul and Charles J. Fillmore 1999 Grammatical constructions and linguistic generalizations: The Whats X doing Y? construction. Language 75: 133. Kovecses, Zoltan and Gunter Radden 1998 Metonymy: Developing a cognitive linguistic view. Cognitive Linguistics 9: 3777. Kristiansen, Gitte, Michel Achard, Rene Dirven and Francisco Ruiz de Mendoza (eds.) 2006 Cognitive Linguistics: Current Applications and Future Perspectives. Berlin/New York: Mouton de Gruyter. Lako, George 1987 Women, Fire and Dangerous Things: What Categories Reveal about the Mind. Chicago: University of Chicago Press. Lako, George 1990 The Invariance Hypothesis: Is Abstract Reason Based on ImageSchemas? Cognitive Linguistics 1(1): 3974. Lako, George 1993 A contemporary theory of metaphor. In: Andrew Ortony (ed.), Metaphor and Thought, 202251. Cambridge: Cambridge University Press. Lako, George and Mark Johnson 1980 Metaphors We Live by. Chicago: University of Chicago Press. Lako, George and Mark Johnson 1999 Philosophy in the Flesh: The Embodied Mind and its Challenges to Western Thought. Chicago: University of Chicago Press.


Rene Dirven and Francisco Jose Ruiz de Mendoza Ibanez

Lako, George and Mark Turner 1989 More than Cool Reason: A Field Guide to Poetic Metaphor. Chicago: University of Chicago Press. Langacker, Ronald W. 1987 Foundations of Cognitive Grammar: Volume 1: Theoretical Prerequisites. Stanford, Cal.: Stanford University Press. Langacker, Ronald W. 1990 Concept, Image, and Symbol: The Cognitive Basis of Grammar. Berlin/New York: Mouton de Gruyter. Langacker, Ronald W. 1991 Foundations of Cognitive Grammar. Vol. 2: Descriptive Application. Stanford, Cal.: Stanford University Press. Langacker, Ronald W. 1995 The symbolic alternative. In: Georges Ludi and Claude-Anne ` Zuber (eds.), Linguistique et Modeles Cognitifs, 5176. Basel: Romanisches Seminar der Universitat Basel. Langacker, Ronald W. 1999 Grammar and Conceptualization. Berlin/New York: Mouton de Gruyter. Langacker, Ronald W. 2003a One any. Korean Linguistics 18: 65105. Langacker, Ronald W. 2003b Constructions in Cognitive Grammar. English Linguistics 20: 4183. Langacker, Ronald W. 2005 Construction Grammars: cognitive, radical, and less so. In: Francisco J. Ruiz de Mendoza Ibanez and Sandra Pena Cervel (eds.), Cognitive Linguistics. Internal Dynamics and Interdisciplinary Interaction, 101159. Berlin/New York: Mouton de Gruyter. Langacker, Ronald W. 2008 Cognitive Grammar: A Basic Introduction. Oxford: Oxford University Press. Lawley, James and Penny Tompkins 2000 Metaphors in Mind: Transformation through Symbolic Modelling. London: Developing Company Press. Lawley, James and Penny Tompkins 2006 What is therapeutic modelling? ReSource Magazine, Issue 8, April 2006. Lee, David 2001 Cognitive Linguistics: An Introduction. Oxford: Oxford University Press. Levin, Beth 1993 English Verb Classes and Alternations. A Preliminary Investigation. Chicago/London: The University of Chicago Press.

Looking back at 30 years of Cognitive Linguistics


Mairal, Ricardo and Pamela Faber 2007 Lexical templates within a functional cognitive theory of meaning. Annual Review of Cognitive Linguistics 5: 137172. Marn, Juana I. 1996 To die, to sleep: a contrastive study of metaphors for death and dying in English and Spanish. Language Sciences 18: 3752. Matlock, Teenie 2004a Fictive motion as cognitive simulation. Memory & Cognition 32: 13891400. Matlock, Teenie 2004b The conceptual motivation of ctive motion. In: Gunter Radden and Klaus-Uwe Panther (eds.), Studies in Linguistic Motivation, 221248. Berlin/New York: Mouton de Gruyter. Merleau-Ponty, Maurice 1945 Phenomenologie de la Perception. Paris: Gallimard. Merleau-Ponty, Maurice 1962 Phenomenology of Perception. London: Routledge. Merleau-Ponty, Maurice 1979 Consciousness and the Acquisition of Language. Evanston, Ill.: Northwestern University Press. Michaelis, Laura A. 2003 Word meaning, sentence meaning, and syntactic meaning. In: Hubert Cuyckens, Rene Dirven and John Taylor (eds.), Cognitive Approaches to Lexical Semantics, 163209. Berlin/New York: Mouton de Gruyter. Morgan, Pamela S. 2001 The semantics of an impeachment. In: Rene Dirven, Roslyn M. Frank and Cornelia Ilie (eds.), Language and Ideology. Volume II: Descriptive Cognitive Approaches, 77106. Amsterdam/ Philadelphia: John Benjamins. Oakley, Todd and Anders Hougaard 2008 Mental Spaces in Discourse and Interaction. Amsterdam/ Philadelphia: John Benjamins. Panther, Klaus-Uwe and Gunter Radden (eds.) 1999 Metonymy in Language and Thought. Amsterdam/Philadelphia: John Benjamins. Panther, Klaus-Uwe and Linda Thornburg (eds.) 2003 Metonymy and Pragmatic Inferencing. Amsterdam/Philadelphia: John Benjamins. Pena, Sandra 2003 Topology and Cognition: What Image-Schemas Reveal about the Metaphorical Language of Emotions. lincom europa: lincom Studies in Cognitive Linguistics. Pires de Olivera, Roberta 2001 Language and ideology: An interview with George Lako. In: Rene Dirven, Bruce Hawkins and Esra Sandikcioglu (eds.),


Rene Dirven and Francisco Jose Ruiz de Mendoza Ibanez

Language and Ideology. Volume 1: Theoretical Cognitive Approaches, 2347. Amsterdam/Philadelphia: John Benjamins. Pustejovsky, James 1995 The Generative Lexicon. Cambridge, Mass.: The MIT Press. Queller, Kurt 2003 Metonymic sense shift: Its origins in hearers abductive construal of usage in context. In: Hubert Cuyckens, Rene Dirven and John Taylor (eds.), Cognitive Approaches to Lexical Semantics, 211 241. Berlin/New York: Mouton de Gruyter. Radden, Gunter 1985 Spatial metaphors underlying propositions of causality. In: Wolf Paprotte and Rene Dirven (eds.), The Ubiquity of Metaphor: Metaphor in Language and Thought, 177207. Amsterdam/ Philadelphia: John Benjamins. Radden, Gunter 2000 How metonymic are metaphors? In: Antonio Barcelona (ed.), Metaphor and Metonymy at the Crossroads: A Cognitive Perspective, 93108. Berlin/New York: Mouton de Gruyter. Radden, Gunter and Zoltan Kovecses 1999 Towards a theory of metonymy. In: Klaus-Uwe Panther and Gunter Radden (eds.), Metonymy in Language and Thought, 1759. Amsterdam/Philadelphia: John Benjamins. Radden, Gunter and Klaus-Uwe Panther 2004 Introduction: Reections on motivation. In: Gunter Radden and Klaus-Uwe Panther (eds.), Studies in Linguistic Motivation, 1 46. Berlin/New York: Mouton de Gruyter. Radden, Gunter and Rene Dirven 2007 Cognitive English Grammar. Amsterdam: John Benjamins. Rice, Sally 2003 Growth of a lexical network: Nine English prepositions in acqui sition. In: Hubert Cuyckens, Rene Dirven and John Taylor (eds.), Cognitive Approaches to Lexical Semantics, 243260. Berlin/New York: Mouton de Gruyter. Rosch, Eleanor 1977 Human categorization. In: Neil Warren (ed.), Studies in Crosscultural Psychology, 149. New York/London: Academic Press. Rosch, Eleanor 1978 Principles of categorization. In: Eleanor Rosch and Barbara B. Lloyd (eds.), Cognition and Categorization, 2748. Hillsdale, N.J.: Lawrence Erlbaum. Ruiz de Mendoza Ibanez, Francisco J. 1998 On the nature of blending as a cognitive phenomenon. Journal of Pragmatics 30/3: 259274. Ruiz de Mendoza Ibanez, Francisco J. 2000 The role of mappings and domains in understanding metonymy. In: Antonio Barcelona (ed.), Metaphor and Metonymy at the

Looking back at 30 years of Cognitive Linguistics


Crossroads. A Cognitive Perspective, 109132. Berlin/New York: Mouton de Gruyter. Ruiz de Mendoza Ibanez, Francisco J. 2008 Cross-linguistic analysis, second language teaching and cognitive semantics: the case of Spanish diminutives and reexive constructions. In: Sabine De Knop and Teun De Rycker (eds.), Cognitive Approaches to Pedagogical Grammar Volume in Honor of Rene Dirven, 121152. Berlin/New York: Mouton de Gruyter. Ruiz de Mendoza Ibanez, Francisco J. and Jose Luis Otal Campo 2002 Metonymy, Grammar, and Communication. Albolote, Granada: Editorial Comares. Ruiz de Mendoza Ibanez, Francisco J. and Lorena Perez Hernandez 2003 Cognitive operations and pragmatic implication. In: Klaus-Uwe Panther and Linda Thornburg (eds.), Metonymy and Pragmatic, 293320. Berlin/New York: Mouton de Gruyter. Ruiz de Mendoza Ibanez, Francisco J. and Olga I. Dez Velasco 2004 Metonymic motivation in anaphoric reference. In Gu nter Radden and Klaus-Uwe Panther (eds.), Studies in Linguistic Motivation, 2350. Amsterdam/Philadelphia: John Benjamins. Ruiz de Mendoza Ibanez, Francisco J. and Sandra Pena Cervel 2005 Conceptual interaction, cognitive operations and projection spaces. In: Francisco J. Ruiz de Mendoza Ibanez and Sandra Pena Cervel (eds.), Cognitive Linguistics: Internal Dynamics and Interdisciplinary Interaction, 254280. Berlin/New York: Mouton de Gruyter. Ruiz de Mendoza Ibanez, Francisco J. and Annalisa Baicchi 2007 Illocutionary Constructions: Cognitive Motivation and Linguistic Realization. In: Istvan Kecskes and Laurence Horn (eds.), Explorations in Pragmatics: Linguistic, Cognitive, and Intercultural Aspects, 95128. Berlin/New York: Mouton de Gruyter. Ruiz de Mendoza Ibanez, Francisco J. and Ricardo Mairal 2007 High-level metaphor and metonymy in meaning construction. In: Gunter Radden, Klaus-Michael Kopcke, Thomas Berg and Peter Siemund (eds.), Aspects of Meaning Construction in Lexicon and Grammar, 3349. Amsterdam/Philadelphia: John Benjamins. Ruiz de Mendoza Ibanez, Francisco J. and Ricardo Mairal 2008 Levels of description and constraining factors in meaning construction: an introduction to the Lexical Constructional Model. Folia Linguistica 42/2: 355400. Sandra, Dominiek and Sally Rice 1995 Network analyses of prepositional meaning: mirroring whose mind the linguists or the language users? Cognitive Linguistics 6(1): 89130.


Rene Dirven and Francisco Jose Ruiz de Mendoza Ibanez

Saussure, Ferdinand de 1974 Course in General Linguistics (trans. Wade Baskin). London: Fontana/Collins. First published [1916]. Schmitt, Rudolf 2005 Systematic metaphor analysis as a method of qualitative research. The Qualitative Report 10: 358394. Schwarz, Monika 1996 Einfuhrung in die Kognitive Linguistik. 2. Uberarbeitete und Aktualisierte Auage. Tubingen: Franke Verlag. Searle, John R. 1979 A taxonomy of illocutionary acts. In: John R. Searle, Expression and Meaning 129. Cambridge: Cambridge University Press. Searle, John R. 2005 The phenomenological illusion. In: Maria E. Reicher and Johann C. Marek (eds.), Experience and Analysis. Proceedings of the 27th International Wittgenstein-Symposium, Kirchberg am Wechsel, August 2004, 1736. Vienna: Obvehpt. Selg, H. 1966 Einfuhrung in die Experimentelle Psychologie. Stuttgart: Kohl hammer. Sperber, Dan, and Deirdre Wilson 1995 Relevance. Communication and Cognition. Oxford: Basil Blackwell. Talmy, Leonard 1975 Figure and ground in complex sentences. BLS 1: 419430 Berkeley, Cal.: Berkeley Linguistics Society. Talmy, Leonard 1978 Figure and ground in complex sentences. In: Joseph H. Greenberg (ed.), Universals of Human Language. Volume 4: Syntax, 625649. Stanford, Cal.: Stanford University Press. Talmy, Leonard 1988a The relation of grammar to cognition. In: Brygida RudzkaOstyn (ed.), Topics in Cognitive Linguistics, 165205. Amsterdam/ Philadelphia: John Benjamins. Talmy, Leonard 1988b Force dynamics in language and cognition. Cognitive Science 12: 49100. Talmy, Leonard 1997 Relating language to other cognitive systems. Ms. Talmy, Leonard 2000a Toward a Cognitive Semantics, Vol. I: Concept Structuring Systems. Cambridge, Mass.: The MIT Press. Talmy, Leonard 2000b Toward a Cognitive Semantics, Vol. II: Typology and Process in Concept Structuring. Cambridge, Mass.: The MIT Press.

Looking back at 30 years of Cognitive Linguistics


Taylor, John R. 1995a Linguistic Categorization: Prototypes in Linguistic Theory. 2nd Revised Edition. Oxford: Clarendon Press. Taylor, John R. 1995b Models of word meaning in comparison: The two-level model (Manfred Bierwisch) and the network model (Ronald Lan gacker). In: Rene Dirven and Johan Vanparys (eds.), Current Approaches to the Lexicon, 326. Frankfurt: Peter Lang. Taylor, John R. 1998 Syntactic constructions as prototype categories. In: Michael Tomasello (ed.), The New Psychology of Language: Cognitive and Functional Approaches to Language Structure, 177202. Mahwah, N.J.: Lawrence Erlbaum. Taylor, John R. 2002 Cognitive Grammar. Oxford: Oxford University Press. Thornburg, Linda and Klaus-Uwe Panther 1997 Speech act metonymies. In: Wolf-Andreas Liebert, Gisela Redeker and Linda Waugh (eds.), Discourse and Perspective in Cognitive Linguistics, 205219. Amsterdam/Philadelphia: John Benjamins. Turner, Mark and Gilles Fauconnier 1995 Conceptual integration and formal expression. Metaphor and Symbolic Activity 10: 183204. Tyler, Andrea and Vyvyan Evans 2001 Reconsidering prepositional polysemy networks: The case of over. Language 77: 724765. Tyler, Stephen 1995 The semantics of time and space. American Anthropologist 97: 567592. Ungerer, Friedrich and Hans-Jorg Schmid 22006 An Introduction to Cognitive Linguistics. London/New York: Longman. Verboven, Hans 2003 Die Metapher als Ideologie: Eine kognitiv-semantische Analyse der Kriegsmetaphorik im Fruhwerk Ernst Jungers. Heidelberg: CarlWinter. Vervaeke, John and John M. Kennedy 1996 Metaphors in language and thought: Falsication and multiple meanings. Metaphor and Symbolic Activity 11: 273284. Vervaeke, John and Christopher D. Green 1997 Women, Fire, and Dangerous Theories: A Critique of Lako s Theory of Categorization. Metaphor and Symbol 12: 5980. Violi, Patrizia 2004 Embodiment at the crossroads between cognition and semiosis. Recherches en Communication 19: 199217.


Rene Dirven and Francisco Jose Ruiz de Mendoza Ibanez

Violi, Patrizia 2008

Beyond the body: towards a full embodied semiosis. In: M. Frank, Rene Dirven, Tom Ziemke, Enrique Bernardez (eds.), Body, Language and Mind. Volume 2: Sociocultural Situatedness, 5376. Roslyn Berlin/New York: Mouton de Gruyter.

Warren, Beatrice 2002 An alternative account of the interpretation of referential metonymy and metaphor. In: Rene Dirven and Ralf Porings (eds.), Metaphor and Metonymy in Comparison and Contrast, 113130. Berlin/New York: Mouton de Gruyter. Wierzbicka, Anna 1996 Semantics: Primes and Universals. Oxford: Oxford University Press. Wiggins, Osborne 1994 Phenomenology and Cognitive Science. In: Mano Daniel and Lester Embree (eds.), Phenomenology of the Cultural Disciplines, 6783. Dordrecht: Springer. Wittgenstein, Ludwig 1991 Philosophical Investigations. Oxford: Blackwell. First published [1953]. Wright, Cory D. 2007 Review of Hubert Cuyckens, Rene Dirven and John R. Taylor (eds.), Cognitive Approaches to Lexical Semantics. Berlin/New York: Mouton de Gruyter. Cognitive Linguistics 18(4): 570579. Yoshimura, Kimihiro 1998 The Middle Construction in English: A Cognitive Linguistic Analysis. Department of Linguistics, University of Otago: Ph.D. dissertation. Yu, Ning 1998 The Contemporary Theory of Metaphor. A Perspective from Chinese: John Benjamins. Yu, Ning 2009 The Chinese HEART in a Cognitive Perspective: Culture, Body, and Language. Berlin/New York: Mouton de Gruyter. Ziemke, Tom, Jordan Zlatev and Roslyn M. Frank (eds.) 2007 Body, Language, and Mind. Volume 1: Embodiment. Berlin/New York: Mouton de Gruyter. Zlatev, Jordan 2003 Polysemy or generality? Mu. In: Hubert Cuyckens, Rene Dirven and John Taylor (eds.), Cognitive Approaches to Lexical Semantics, 447494. Berlin/New York: Mouton de Gruyter. Zlatev, Jordan 2007 Embodiment, language, and mimesis. In: Tom Ziemke, Jordan Zlatev and Roslyn M. Frank (eds.), Body, Language, and Mind. Volume 1: Embodiment, 297337. Berlin/New York: Mouton de Gruyter.

Recontextualizing Grammar: Underlying trends in thirty years of Cognitive Linguistics Dirk Geeraerts
1. Cognitive Linguistics: radial set or schematic network? Cognitive Linguistics is a success: from the relatively marginal position that it originally occupied in the linguistic landscape, it has developed into one of the mainstream trends in current linguistics. A search on the strings cognitive linguistics and cognitive grammar, in contrast with generative linguistics and generative grammar in the LLBA bibliography, broken down over four periods of ve years starting in 1988, yields the gures presented in Table 1. Although the search terms could be extended to get a more complete picture, the present gures already indicate with sucient force that Cognitive Linguistics seems to be overtaking the generativist enterprise in terms of scholarly productivity and appeal. This conspicuous success raises the question whether there is a common denominator in the expansion of Cognitive Linguistics? A cursory inspection would seem to suggest that the internal evolution of Cognitive Linguistics is of the radial network type (to put the matter in an appropriate terminology). A number of core ideas like grammatical construal, prototypicality, radial networks, ICMs, conceptual integration, constructions are pursued and developed rather independently of each other, to the extent that each of them constitutes a mini research programme of its own. But what is it that keeps these separate approaches together if anything?

Table 1. The presence of Cognitive Linguistics and Generative Grammar in the LLBA from 1998 to 2007 (gures as of October 15, 2007) 8892 generative grammar / linguistics cognitive grammar / linguistics 304 81 9397 538 337 9802 337 376 0307 296 916


Dirk Geeraerts

In Geeraerts 2006a, written as an introduction to a collection of basic readings in Cognitive Linguistics, two principles of cohesion were mentioned. On the one hand, the central concepts of Cognitive Linguistics are mutually related because they derive from a common set of underlying fundamentals: the principle that language is all about meaning, in the broadest possible sense, in combination with four specic assumptions about the nature of linguistic meaning that meaning is exible and dynamic, that it is encyclopedic and non-autonomous, that it is based on usage and experience, and that it is perspectival in nature. On the other hand, repeating a point originally formulated in Geeraerts (2003), it was suggested that the various strands of Cognitive Linguistics belong together because they exhibit various ways of recontextualizing the study of language. In this sense (once again applying the models of categorization developed in Cognitive Linguistics), we can not only say that the various branches of Cognitive Linguistics constitute a radial network, but also that there is a schematic commonality over and above the radial structure. In the present paper, the recontextualizing nature of Cognitive Linguistics is further analyzed: the paper will present the dierent parts of the recontextualization in more detail, and it will show that the internal development of the framework and the chronological steps in the gradual elaboration of the radial set structure are guided by the recontextualization programme. This analysis will lead to a brief consideration of a further question that arises from the success of Cognitive Linguistics: to what extent is the manifest success of Cognitive Linguistics also a threat? Under which conditions could the very wealth of the research undermine the unity of the approach? Or conversely, would the conceptual unity that we may discern if we look at the evolution of Cognitive Linguistics be sucient to maintain its success? As a practical introductory remark, it should be noted that the present paper is a highly synthetic one that assumes familiarity with the basic concepts of Cognitive Linguistics. The purpose is to reveal an underlying but largely unheeded pattern in the rich variety of approaches in Cognitive Linguistics, not to present these approaches as such. Discovering such a pattern, to the extent that it is successful, will be important for the decision how to further elaborate Cognitive Linguistics. But it also serves a more immediate, didactic purpose: it may help to introduce the full breadth of research in Cognitive Linguistics within a synthetic and systematic framework.

Underlying trends in thirty years of Cognitive Linguistics


2. Decontextualization and recontextualization in 20th century linguistics If we are interested in what drives the development of Cognitive Linguistics and keeps its various branches together, we may start by trying to determine what distinguishes Cognitive Linguistics from other approaches in modern linguistics. How does Cognitive Linguistics t into the development of theoretical linguistics? An answer to that question requires an insight into the basic lines of evolution of modern linguistics. So, if we were to present the history of 20th century linguistics in a nutshell, what would be the main lines of its development? Would it at all be possible to synthesize a century of theoretical development into a few pages? The following pages will try to identify some of the main lines in the development of 20th century linguistics, arguing that this development is characterized by a succession of a decontextualising and a recontextualizing movement. Obviously, there is a price to pay for the attempt to cover such a vast domain. For one thing, we will be able to focus only on the mainstream developments in the international scene of linguistics, disregarding local traditions, isolated individual achievements, avant-gardes and rearguards. For another, a historiographical programme of this type basically takes the form of a logical reconstruction: can we retrospectively nd a perspective that brings order into the apparent chaos? To what extent does the development lend itself to a rational reconstruction, where the dierent steps in the development are interpreted as an elaboration of a basic research question? Such a rational reconstruction implies that hardly any attention can be given to the actual biographical factors and the sociological interactions. Also, the analysis presented here does not claim to be the only possible one. We will take our starting-point in the dierences between the Saussurean dichotomy of langue and parole and the Chomskyan dichotomy of competence and performance, but that does not rule out the possibility that the story could be told with a dierent point of departure and from a dierent point of view. 2.1. Gaps in the system: Saussure and Chomsky The Saussurean dichotomy between langue and parole creates an internally divided grammar, a conception of language with, so to speak, a hole in the middle. On the one hand, langue is dened as a social system, a set of collective conventions, a common code shared by a community:


Dirk Geeraerts ` Mais quest-ce que la langue? . . . Cest a la fois un produit social de la faculte du langage et un ensemble de conventions necessaires, adoptees par le corps social pour permettre lexercice de cette faculte chez les individus [But what is language? . . . It is at the same time a social product of the language faculty and a collection of necessary conventions adopted by a community to allow individuals to use that faculty] (1976: 25).

On the other hand, parole is an individual, psychological activity that consists of producing specic combinations from the elements that are present in the code:
La parole est au contraire un acte individuel de volonte et dintelligence, dans lequel il convient de distinguer 1) les combinaisons par lesquelles le sujet parlant utilise le code de la langue en vue dexprimer sa pensee per sonelle 2) le mecanisme psycho-physique qui lui permet dexterioriser ces combinaisons [Speech on the other hand is a deliberate and intelligent individual act, in which we can distinguish, rst, the combinations by means of which the individual subject uses the code of the language to express his personal thought, and second, the psycho-physical mechanism that allows him to exteriorize those combinations] (1967: 30).

When langue and parole are dened in this way, there is a gap between both: what is the mediating factor that bridges the distance between the social and the psychological, between the community and the individual, between the system and the application of the system, between the code and the actual use of the code? The Chomskyan distinction between competence and performance formulates the fundamental answer to this question: the missing link between the social code and the individual usage is the individuals knowledge of the code. Performance is basically equivalent with parole, but competence interiorizes the notion of linguistic system: competence is the internal grammar of the language user, the knowledge that the language user has of the linguistic system and that he puts to use in actual performance. Remarkably, however, Chomsky introduces a new gap into the system. Rather than the trichotomy that one might expect, he restricts his conception of language to a new dichotomy: the social aspects of language are largely ignored. In comparison with a ternary distinction distinguishing between langue, competence, and parole/performance (between social system, individual knowledge of the system, and actual use of the system), the binary distinction between competence and performance creates a new empty slot, leaving the social aspects of language largely out of sight. Figure 1 schematically summarizes the Saussurean and the Chomskyan positions, highlighting the systematic relationship between both. The ques-

Underlying trends in thirty years of Cognitive Linguistics


Figure 1. Schematic representation of the gaps in the Saussurean (upper picture) and Chomskyan (lower picture) conception of language

tion marks indicate those aspects of language that disappear in either of the two approaches. Relegating the social nature of language to the background, in the Chomskyan approach, correlates with a switch towards the phylogenetic universality of language. The Chomskyan emphasis on the genetic nature of natural language links up rationally with his apparent lack of interest for language as a social semiotic. Where, in particular, does the individual knowledge of the language come from? If the source of linguistic knowledge is not social, what else can it be than an innate and universal endowment? If the language is not learned through acculturation in a linguistic community (given that a language is not primarily a social code), what other source could there be for linguistic knowledge except genetics? 2.2. The decontextualization of grammar The link between the Chomskyan genetic perspective and the absence of any fundamental interest in language as a social phenomenon engenders a stepping-stone development, leading by an internal logic to an isolation of the grammar. Let us go through the argument in the form of the following chain of (deliberately succinct and somewhat simplistic) propositions. First, if natural language is not primarily social, it has to be genetic. This is the basic proposition that was described in the previous paragraph. The relationship could of course be construed in the other direction as well. As presented above, the Chomskyan predilection for a genetic per-


Dirk Geeraerts

spective in linguistics follows from his lack of interest for the social side of language. But in actual historical fact, Chomskys preference for a genetic conception of language probably grew more from his discussion with behaviorist learning theory (Skinner in particular) than from a confrontation with Saussure. Because the amazing ability of young children to acquire language cannot be explained on the basis of a stimulus-response theory so the argument goes an innate knowledge of language has to be assumed. But if one of the major features of language is its genetic nature, then of course the social aspects of language are epiphenomenal. Regardless of the direction in which the link is construed, however, the eects are clear. Second, if natural language is primarily a genetic entity, semantics or the lexicon cannot be part of the core of language (or the core business of linguistics). Meanings constitute the variable, contextual, cultural aspects of language par excellence. Because social interaction, the exchange of ideas, changing conceptions of the world are primarily mediated through the meaning of linguistic expressions, it is unlikely that the genetic aspects of language will be found in the realm of meaning. Further, if the lexicon is the main repository of linguistically encoded meaning, studying the lexicon is of secondary importance. Here as before, though, it should be pointed out that the actual historical development is less straightforward than the reconstruction might suggest. The desemanticization of the grammar did not happen at once (nor was it absolute, for that matter). Triggered by the introduction of meaning in the standard model of generative grammar (Chomsky 1965), the Linguistic Wars (see Harris 1995) of the late 1960s that opposed Generative Semantics and Interpretive Semantics basically involved the demarcation of grammar with regard to semantics. The answer that Chomsky ultimately favoured implied a restrictive stance with regard to the introduction of meaning into the grammar, but this position was certainly not reached in one step; it was prepared by severe debates in the generativist community. Third, if semantics or the lexicon cannot be part of the core of linguistics, linguistics will focus on formal rule systems. The preference for formal syntax that characterizes generative grammar follows from the fact that generative grammar links up with cognitive science as it originated in the late 1950s and the 1960s, which sees knowledge as a set of internal symbolic representations, and mental processing as internal symbol manipulation. The Chomskyan emphasis on the formal syntax ts in with the symbol manipulation paradigm that was dominant in cognitive

Underlying trends in thirty years of Cognitive Linguistics


science. But it also follows by elimination from generative grammars genetic orientation: formality is required to restrict the impact of meaning, and studying syntax (or more generally, the rule-based aspects of language) correlates with the diminished interest in the lexicon. It should be added that the focus on rules is not only determined by a negative attitude with regard to meanings, but also by a focus on the innity of language: language as an innite set of sentences requires a rule system that can generate an innity of entities. (At this point, an additional undercurrent in the history of 20th century linguistics crops up: the relationship between linguistics and logic. Chomsky, in fact, got the inspiration for his conception of linguistic rule systems from the architecture of logical proof theory. This is not a line to be pursued here, though.) Finally, if linguistics focuses on formal rule systems, the application of the rule systems in actual usage is relatively uninteresting. If the rules dene the grammar, it is hard to see what added value could be derived from studying the way in which the rules are actually put to use. The study of performance, in other words, is just as secondary as research into the lexicon. This chain of consequences leads to a decontextualisation of the grammar. It embodies a restrictive strategy that separates the autonomous grammatical module from dierent forms of context. Without further consideration of the interrelationship between the various aspects of the decontextualising drift, the main eects can be summarized as follows. First, through the basic Chomskyan shift from langue to competence, linguistics is separated from the social context of language as a social code. Second, through the focus on the genetic aspects of the language, linguistics is separated from the cognitive context that shows up in the semantic side of the language. Third, through the focus on formal rule systems, linguistics is separated from the situational context of actual language use. And fourth, these two last features tend to favour the study of formal syntactic rules over the study of the lexicon. 2.3. Trends towards recontextualization In terms of the subdisciplines covered by linguistics, this means that the core of linguistics in Chomskyan terms disfavours the study of language in its social context, any model of language in which semantics or the lexicon is at the heart of the grammar, and the investigation of language in actual usage. This does not mean, however, that these areas of research,


Dirk Geeraerts

which would be considered peripheral from the generativist point of view, were totally non-existent in the heyday of generative grammar. In fact, the generativist era witnessed the birth of approaches that autonomously developed the aspects that were rejected or downplayed by generative grammar. A brief overview may suce to establish that the period from the late 1960s to the 1980s constitutes a crucial period for the development of sociolinguistics, pragmatics, and semantics: if we look at what would now be considered foundational publications for the approaches in question, we see that they are situated precisely in the period from the mid 1960s to the mid 1980s. (If the main factors that were dispelled by generative grammar are meaning, the lexicon, language use, and the social context of language, then the approaches mentioned here cover three out of four factors. Lexicology, in fact, did not boom as a separate discipline in the way the other three did.) Sociolinguistics in its present form (including the sociology of language, the ethnography of speaking, and sociohistorical linguistics, next to sociolinguistics in the narrow, Labovian sense) came into existence with works such as Labov (1972), Haugen (1966), Weinreich, Labov and Herzog (1968), Gumperz and Dell Hymes (1972). Pragmatics as a separate discipline started o in the wake of Grice (1975) and Austin (1962). In seminal works such as Stalnaker (1974) or Gazdar (1979), a Gricean, logically inspired form of pragmatics saw the light, focusing on questions of presupposition, conversational implicature, and contextual interpretation. In the same period, Searle (1969) developed Austins approach into speech act theory. Somewhat later, a broader type of discourse linguistics and conversational analysis was introduced by Coulthard (1977), Gumperz (1982), Brown and Yule (1983), or Tannen (1984), to name just a few of the early works. Semantics received a major impetus through the development of formal semantics, as in Montague (1974), Partee (1979), Dowty (1979). Building on the achievements of formal logic, formal semantics is a type of meaning-based grammar, but the conception of meaning that lies at the basis of formal semantics is restricted to the referential, truth-theoretical aspects of meaning. (In this sense, it is a more or less restricted form of semantics. Later developments like Cognitive Linguistics will take a less restrictive approach to meaning, as we will see.) In short, the decontextualising, autonomist attitude of generative grammar was to some extent compensated by the development of disciplines that explore the aspects of language that are relegated to the background

Underlying trends in thirty years of Cognitive Linguistics


by generative grammar. The works quoted above are indicative of the initial stages in the development of these disciplines: anyone familiar with the history of contemporary linguistics will recognize that each of them ourished afterwards. With the exception of formal semantics (specically in its association with categorial grammar), these approaches are not models of the grammar, if we think of the grammar as the description of the internal structure of the language. As separate disciplines (or, at best, as separate modules in a modular conception of language) sociolinguistics and pragmatics developed alongside grammatical theory rather than interacting with it intensively. This suggests that the recuperation of the contextual aspects rejected by generative grammar can be carried one step further, if the study of context does not take the form of a set of separate disciplines but if context features are introduced into the heart of the grammar. This is exactly what is happening in a number of more recent trends in linguistics. From roughly 1980 onwards, a number of developments in linguistics appear to link the grammar more closely to the contextual aspects that were severed by generative theorizing. The peripheral aspects meaning, the lexicon, language use, the social context of language that were being developed largely separately and autonomously, are now being linked up more narrowly with the grammar itself. The following overview of the relevant tendencies will again be brief and schematic. Its main purpose is to point out the existence of the trends, not to describe them in much detail. This also means, for instance, that no attention will be paid to the overlap that may exist between dierent tendencies. The reintroduction of the lexicon into the grammar is probably the most widespread of the tendencies to be mentioned here; it is, in fact, relatively clear within generative grammar itself. This lexicalist tendency in grammatical theory is triggered by the recognition that describing grammatical rules appears to imply describing the lexical sets that the rules apply to. Reversing the descriptive perspective then leads to a description of the valency of the lexical items (i.e. the structures that an item can appear in). The lexicalist tendency appears in various forms in the more formal approaches to grammar: one may think of the projections and theta-roles of generative grammar, of the central role of the lexicon in Lexical Functional Grammar (Bresnan 2001), and of the lexically driven grammar developed in the framework of Head-driven Phrase Structure Grammar (Sag, Wasow and Bender 2003). Hudsons Word Grammar (1991) is a functionally oriented type of lexical grammar.


Dirk Geeraerts

Approaches that give meaning a major, if not dominant, role in linguistics were not restricted to formal semantics. As we mentioned already, within generative grammar, the Generative Semantics movement had to give way to the supremacy of the semantically restricted Chomskyan approach. But parallel to generative grammar, a cluster of functionalist approaches to grammatical description gave pride of place to meaning. This holds true for Functional Grammar as dened by Dik (1989), for the functional-typological approaches developed by Givon (1979), and for Hallidays Systemic Functional Linguistics (1994), to mention only the most important representatives. Typically, these approaches take a broad view of meaning, i.e. they include pragmatics: in all forms of functional linguistics, discursive and interactional communicative functions are seen as essential features of natural language. For instance, a number of functionalist approaches try to nd (potentially universal) discourse motivations for grammatical constructs. Discourse is then no longer the mere application of grammatical rules, but the grammatical rules themselves are motivated by the discourse functions that the grammar has to fulll. The existence of passives in a given language, for instance, is explained as a topicalization mechanism: grammars contain passives because topicalizing direct objects is a useful function in discourse. Seminal publications within this approach include Givon (1979), Hopper and Thompson (1980), Hopper (1987). The communicative aspect inherent in this family of functional approaches also means that the social aspects of language are explicitly recognized. If language is primarily seen as an instrument for communicative interaction, a social conception of language is automatically implied. Within the group of functionalist frameworks, Systemic Functional Linguistics is the one that most distinctly follows up on this social conception of language. Thinking about language in social, interactional terms suggests that the systemic descriptive and theoretical framework might be particularly suited for socially oriented types of linguistic investigation. In practice, this shows up in the many studies in Systemic Functional Linguistics that are geared towards the analysis of text and discourse from a social perspective (see e.g. Eggins 1994; Thompson 1994). Not surprisingly, the methodology of Systemic Functional Linguistics has also been embraced by Critical Discourse Analysis (Fairclough 1995). At the same time, this social perspective in Systemic Functional Linguistics seems to be restricted primarily to the study of text types and register dierences, without a lot of attention for language-internal variation of a lectal or sociolinguistic kind.

Underlying trends in thirty years of Cognitive Linguistics


3. Cognitive Linguistics as a recontextualizing approach It is not dicult to establish that the four elements of context that we mentioned earlier (meaning, the lexicon, discourse and use, and the social context) receive particular attention in Cognitive Linguistics. Meaning Cognitive Linguistics constitutes an outspoken attempt to give meaning a central position in the architecture of the grammar. The basic vocabulary of the cognitive framework involves semantics: the notion of prototype, schematic network, conceptual metaphor, metonymy, conceptual integration, idealized cognitive models, frames, and all sorts of construal mechanisms are semantic notions. Crucially, these semantic concepts involve a contextualized view of semantics. There are several ways of making that clear. To begin with, if we compare Cognitive Linguistics with formal semantics, it is obvious that the conception of meaning that lies at the basis of the cognitive approach is not restricted to a referential, truth-functional type of meaning. Linguistic structures are thought to express conceptualizations, i.e. conceptualization is central for linguistic structure and conceptualization goes further than mere reference. It involves imagery in the broadest sense of the word: ways of making sense, of imposing meaning. Further, if we come down from this very general level and look more closely into the semantic concepts in question, we can see that they systematically refer to various contextualized forms of meaning. Cognitive Linguistics embodies a fully contextualized conception of meaning in that its central semantic concepts describe dierent ways in which the conceptualizations that are expressed in the language have an experiential basis: the type of relevant context is dierent for various central concepts. Prototypicality eects and the various aspects of categorial polysemy, including conceptual mechanisms like metonymy and metaphor, derive from the fact that new knowledge is constituted against the background of existing cognitive and linguistic categories: existing categories provide a context for the development of new nuances and extended meanings. Frames in the Fillmorean sense, Idealized Cognitive Models, and mental spaces in the sense of conceptual integration theory (a.k.a. blending) represent the idea that linguistically relevant knowledge is structured knowledge of the world: language has to be seen in the context of encyclopedic cognition, and not as an autonomous realm of the mind. Mechanisms of grammatical construal like gure/ground perspectivization implement the idea that linguistic meaning has to be studied in corre-


Dirk Geeraerts

lation with general cognitive mechanisms, like the gestalt features of knowledge in the case of gure/ground alignment. The neural theory of language developed by Lako shows that the embodied nature of language is not only considered in a psychological sense (focusing on individual experience in a phenomenological sense), but also in the most literal sense possible: the neural embodiment of the mind in the brain constitutes the material context of natural language semantics. Lexicon From very early on, treating grammatical categories according to the model provided by the lexicon was a natural thing to do in Cognitive Linguistics: if meaning description is the focus of Cognitive Linguistics, and if the models for the description of meaning were primarily developed in the realm of lexical semantics, it is no surprise to nd that notions of prototypicality and polysemy were applied to grammatical categories. But it is only with the rise of Construction Grammar (Goldberg 1995; Croft 2001; Langacker 2005) that the lexicalization of the grammar becomes outspoken, because it is only at that point that the lexicon begins to play a role on the form side of the grammatical description. There are two aspects to be mentioned. First, the concept of a construction introduces lexical material into the notion of a grammatical entity. Constructions of the let alone type (Fillmore, Kay and OConnor 1988) constitute combinations of specic words and abstract patterns. Second, while not all constructions need be of this mixed type (e.g. the ditransitive construction does not include specic lexical materials), there is no a priori dividing line between the lexically specic and the abstract patterns. Between fully lexicalized formal units (words and idioms) and patterns that can only be described in terms of abstract grammatical categories (like traditional immediate constituent analyses), there are intermediate entities at dierent levels of abstraction. In this sense, lexicon and grammar are integrated, as dierent levels of abstraction in an inventory of constructional patterns. Language use and discourse The most immediate type of discourserelated investigation in Cognitive Linguistics is the study of all kinds of pragmatic and discursive phenomena, like discourse particles and pragmatic markers (Fischer 2000), information structure (Sanders and Spooren 2007), grounding (Brisard 2002), Current Discourse Space (Langacker 2001), etc. the references are indicative only. A recent development along these lines is the investigation of blending phenomena as on-line meaning construction, as in Coulson (2006). However, there is a more fundamental link between Cognitive Linguistics and the study of performance: more and more, Cognitive Linguistics

Underlying trends in thirty years of Cognitive Linguistics


conceives of itself as a usage-based approach to language. According to a number of programmatic accounts of usage-based linguistics (Bybee and Hopper 2001; Kemmer and Barlow 2000; Langacker 1999; Tomasello 2000; Verhagen and Van de Weijer 2003), the essential idea of a usagebased linguistics is the dialectic nature of the relation between language use and the language system. The grammar does not only constitute a knowledge repository to be employed in language use, but it is also itself the product of language use. The former perspective considers usage events as specic, actual instantiations of the language system. According to this view, one can gain insight into the language system by analyzing the usage events that instantiate it. This is a strong motivation for empirical research: the usage data constitute the empirical foundation from which general patterns can be abstracted. The latter perspective considers usage events as the empirical source of the system. From this point of view, usage events dene and continuously redene the language system in a dynamic way. As a result, every usage event may slightly redene a persons internal language system. The consequences of such a position are both thematic and methodological. Thematically speaking, a usage-based approach fosters interest in specic topics and elds of investigation. For instance, it follows from the dialectic relationship between structure and use that the analysis of linguistic change (Bybee 2007) is a natural domain of application for any usage-based approach. Similarly, literary analysis from a cognitive point of view (Freeman 2007; Brone and Vandaele 2009) ensues naturally from an interest in language use: if the analysis of discourse is a legitimate (and, in fact, important) goal for Cognitive Linguistics, cognitive poetics is likely to emerge as the study of the very specic type of discourse represented by literary texts. This holds more generally for cognitive stylistics in its various forms (Semino and Culpeper 2002). Further, interesting perspectives for language acquisition research open up: the usage-based approach holds the promise of answering the acquisition problem that looms large in the Chomskyan delimitation of linguistics. In the work done by Tomasello and his group (2003), an alternative is presented for the Chomskyan genetic argument. These researchers develop a model of language acquisition in which each successive stage is (co)determined by the actual knowledge and use of the child at a given stage, i.e. language acquisition is described as a series of step by step usage-based extensions of the childs grammar. The grammar, so to speak, emerges from the childs interactive performance. At the same time, there are methodological consequences: you cannot


Dirk Geeraerts

have a usage-based linguistics unless you study actual usage as it appears in an online and elicited form in experimental settings or as it appears in its most natural form in corpora in the shape of spontaneous, non-elicited language data. We can indeed see that the interest in corpus-based and experimental studies is growing, but it would be an exaggeration to say that it has become the standard approach in Cognitive Linguistics (cf. Geeraerts 2006b; Tummers, Heylen and Geeraerts 2005). Social context There are four, more or less hierarchically ordered levels at which Cognitive Linguistics pays explicit attention to the social nature of language. The rst level is that of language as such: the denition and the basic architecture of language are recognized as involving not just cognition, but socially and culturally situated cognition. The type of work produced from this perspective emphasizes and analyzes the way in which the emergence of language as such and the presence of specic features in a language can only be adequately conceived of if one takes into account the socially interactive nature of linguistic communication. Examples of this strand of research include Sinha (2000, 2007) on language as an epigenetic system, Zlatev (2005) on situated embodiment, Itkonen (2003) on the social nature of the linguistic system, Verhagen (2005) on the central role of intersubjectivity in language, and Harder (2003) on the socio-functional background of language. Taking one step towards a more specic approach, the next level is that of variation among languages and cultures. This is the oldest form of a social perspective in Cognitive Linguistics: the notion of cultural model played a signicant role in the emergence of the new framework. However, a tension existed between a more universalist approach and a more culturally oriented approach. A typical case in point is the discussion between Geeraerts and Grondelaers (1995) on the one hand and Kovecses (1995) on the other regarding the nature of anger is heat metaphors: while the former emphasized the culturally specic and historically contingent nature of such metaphorical patterns, the latter defended a universalist, physiologically grounded position. In recent years, however, the socio-cultural perspective has been gaining ground: see Palmer (1996) and, very explicitly, Kovecses (2005). In practical terms, this type of socio cultural investigation takes the form of a historical investigation into changing conceptualizations within a given language or culture, or of cultural and anthropological comparisons. Within the latter group, we can also place the ourishing research tradition of cross-linguistic and crosscultural investigation into the relationship between language and thought,

Underlying trends in thirty years of Cognitive Linguistics


as illustrated by works like Slobin (1996), Boroditsky (2000) and Levinson (2003). The third level considers not variation between languages, but variation within languages: to what extent do the phenomena that we typically focus on in Cognitive Linguistics exhibit variation within the same linguistic community? The research conducted within this approach links up with the research traditions of sociolinguistics, dialectology, and stylistic analysis, using the same type of meticulous empirical research methods: see Kristiansen and Dirven (2008). This attempt to bring grammatical analysis and variationist research closer together cannot be disentangled from the usage-based perspective mentioned above. Usage-based and meaning-based models of grammar introduce more variation into the grammar than a rule-based approach tends to do: the language-internal or discourse-related factors that inuence the use of a particular construction may be manifold, and the presence or absence of a construction is not an all-or-none matter. In the analysis of this type of variation, it often appears that the variation is co-determined by external, sociolinguistic factors: the variation that appears in actual usage (as attested in corpora) may be determined simultaneously by grammatical, discursive, and sociolinguistic factors. Disentangling those dierent factors, then, becomes one methodological endeavour: in the actual practice of a usage-based enquiry, grammatical analysis and variationist analysis will go hand in hand. At this level, there is also a less descriptive and more critical form of Cognitive Linguistics to be mentioned a form of analysis, in other words, that not only intends to describe but that also takes an evaluative stance. The best known example is Lakos study of metaphorical models of the family exploited in a political context (1996). This work led to his practical interest in the framing of public issues (and to his active involvement in US political debate). Lakos work, though highly visible, is not the only one in this domain. There is a growing body of research on ideology (see Dirven, Polzenhagen and Wolf 2007), and in works such as Chilton (2004), Musol (2006), Hart and Lukes (2007), we witness an exciting convergence between critical Cognitive Linguistics and the older British tradition of Critical Discourse Analysis. The nal and most specic level overlaps with the discourse related approach mentioned above: this is the level where actual conversations and communicative exchanges are analyzed from a socially interactionist point of view. In the context of socially oriented linguistics, this perspec-


Dirk Geeraerts

Figure 2. The reintroduction of context features into Cognitive Linguistics

tive links up with interactionist sociolinguistics and ethnomethodology rather than with variationist sociolinguistics (as was the case on the third level of analysis). In Cognitive Linguistics circles, representatives of this approach tend to take some of their inspiration from Clark (1996): see Croft (2009), and most elaborately, Tomasello (2003) on the interactionist nature of language acquisition. Figure 2 oers a schematic representation of the reintroduction of the four crucial context factors into the grammar as conceived of by Cognitive Linguistics. The external boxes in the gure indicate the four elements as we discussed them above, and the labelled arrows represent the reintroductory movements by means of keywords. 3.2. Stages in the expansion of Cognitive Linguistics and the chronological recovery of context The previous section demonstrates that Cognitive Linguistics indeed embodies the recontextualizing trends in contemporary linguistics in a singular way: all the aspects of recontextualization that we identied in our overview of the history of 20th century linguistics are substantially covered by Cognitive Linguistics; even more importantly, they determine core features of Cognitive Linguistics, like the importance of meaning for grammatical description and the choice for a usage-based approach. Now, if recontextualization is indeed a core feature, would it also be the case that the internal development of Cognitive Linguistics is characterized by a gradual expansion towards the various forms of recontextualization that we distinguished? Does recontextualization not only portray Cognitive

Underlying trends in thirty years of Cognitive Linguistics


Linguistics at large, against the broad canvas of modern linguistics, but does it also characterize its internal growth? Does the internal evolution of Cognitive Linguistics take the form of a progressive recovery of the various types of context? To answer that question, we rst need to have a look at the history of Cognitive Linguistics from the point of view of the sociology of science: if we look at it from an external point of view (not with a focus on the development of ideas and theories, but with a focus on the people it mobilizes, the public appeal it exerts, the organizational and institutional entrenchment it achieves), what are the main periods in its development? The seeds of Cognitive Linguistics were planted some thirty years ago: Len Talmy published his gure and ground paper in 1975, Ron Langacker started working on his Cognitive Grammar in 1976, and George Lako published his linguistic gestalts article in 1977. But the real public life and the international expansion of Cognitive Linguistics started about ten years later. The year 1987 is an outspoken landmark, with the publication of Lako s Women, Fire, and Dangerous Things, and the rst volume of Langackers Foundations of Cognitive Grammar. In 1988, Brygida Rudzka edited the seminal Topics in Cognitive Linguistics volume, and in 1989, John Taylor published his Linguistic Categorization, which is still one of the best readable introductions to Cognitive Linguistics. 1989 is important in another sense as well: it is the year that the rst International Cognitive Linguistics Conference (ICLC) took place in Duisburg, Germany. It was one of the so-called LAUD symposia (where LAUD stands for Linguistic Agency of the University of Duisburg) that had been organized by Rene Dirven since 1977 and where some of the worlds most distinguished linguists were invited to present their work. So in 1989, Rene Dirven (whose role in the international expansion of Cognitive Linguistics can hardly be underestimated) invited Lako and Langacker for a Symposium on Cognitive Linguistics. The Duisburg conference was of crucial importance for the institutionalization and the international expansion of Cognitive Linguistics: it was there and then that the International Cognitive Linguistics Association was founded (the conference was accordingly rebaptized as the First International Cognitive Linguistics Conference), that plans were made to launch the journal Cognitive Linguistics, and that the monograph series Cognitive Linguistics Research was announced. The next twenty years can be roughly divided into two more periods of ten years. The rst ten years leading to the present situation roughly, up to 1997 were years of international consolidation. The whole approach


Dirk Geeraerts

demonstrated its viability on the international forum, in a number of senses: the newly founded journal was able to attract high-quality contributions, and the successful series of biannual ICLC conferences proved that there was a broad interest in the framework. The years from, say, 1997 up to now were marked by the international institutionalization of Cognitive Linguistics; the whole framework got rmly entrenched in the international fabric of linguistic studies. This is most clear if you look at the emergence of national ICLA aliates. These are ICLA branches dened by region or country (and occasionally by language). The rst one to be founded was the Spanish Cognitive Linguistics Association (1998), whose aliation was formally approved at the 1999 ICLC. The year 2001 saw the aliation of the Finnish, the Polish, and the Slavic Cognitive Linguistics Associations. Further aliates include the Russian Association of Cognitive Linguists (2004), the German Cognitive Linguistics Association (2005), the Discourse and Cognitive Linguistics Association of Korea (2005), the Association Francaise de Linguistique Cognitive (2005), the Japanese Cognitive Linguistics Association (2005), the Conceptual Structure, Discourse and Language Association (2005), the UK Cognitive Linguistics Association (2006), the Chinese Association for Cognitive Linguistics (2006), and BeNeCla, the Cognitive Linguistics association of Belgium and The Netherlands (2008). Another unmistakable sign of the institutionalization of Cognitive Linguistics is the number of introductory books and reference works published in the last decade: Ungerer and Schmid was rst published in 1996, Dirven and Verspoor in 1998, Violi in 2001, Croft and Cruse in 2004, Evans and Green in 2006, the Basic Readings collection (Geeraerts 2006c) and the companion volume edited by Kristiansen, Achard, Dirven and Ruiz de Mendoza in 2006, Geeraerts and Cuyckens in 2007. Next to Cognitive Linguistics, the Annual Review of Cognitive Linguistics provides a second journal specically dedicated to publications in Cognitive Linguistics. Given this classication into three periods of more or less ten years each (a foundational one, a period of expansion, and a period of institutionalization), is there any correlation with the internal development, in terms of ideas and theoretical constructs? In very general terms, it would seem to be the case that the major steps in the recovery of context are situated in the past decade. After the foundational rst decade, the second decade, up to the middle of the 1990s, focused on the basic notions of cognitive semantics: prototypes, radial networks, conceptual metaphor, image schemas, and the various aspects of Langackers and Talmys construal-based grammars.

Underlying trends in thirty years of Cognitive Linguistics Table 2. Nine topical items in the Cognitive Linguistics Bibliography, divided over ve-year periods 7781 No. of articles Meaning Semantics Construction Usage-based Discourse Pragmatic(s) Sociolinguistic(s) Social Cultural 247 23 15 9 0 13 6 2 7 2 8286 591 66 53 28 0 23 19 0 13 5 8791 1093 137 128 66 4 62 37 4 38 38 9296 1647 275 201 150 2 141 98 3 61 87 9701 2416 458 281 295 16 255 162 15 127 159


0206 2557 603 300 439 48 394 209 42 176 240

In the third decade, however, we witness a broadening of the type of context that is deemed relevant: the majority of the publications that were mentioned above under the rubric of lexicon, usage, and social context are situated in the second half of the 1990s and the beginning of the new millennium. The initial insight of Cognitive Linguistics is that the semantics of natural language expressions needs to be studied in a broader cognitive context: individual meanings are part of polysemous structures, of frames, of Idealized Cognitive Models whatever structure the experiential basis of language takes. Moving beyond this initial cognitive context, the previous decade has introduced the context of use in its various forms: in the attention for the performative usage level as the dialectic basis of grammar, in the attention for the lexicon as the embodiment of abstract grammatical patterns, in the attention for the social context of language. That the wave of types of research contextualizing grammar came after a rst wave of research exploring the semantics of language also becomes clear when we have a look at the presence of certain topics in the Cogni tive Linguistics Bibliography compiled by Rene Dirven (The Cognitive Linguistics Bibliography is made available annually as an electronic addition to the journal Cognitive Linguistics.). Table 2 lists the frequencies of topical items in the bibliography, divided over ve-year periods. The items


Dirk Geeraerts

Figure 3. The development of recontextualizing themes in the Bibliography of Cognitive Linguistics

may be grouped into four groups corresponding to the four types of contextualization that we distinguished earlier: meaning and semantics under the label semantics, construction under the label construct(ion grammar), usage-based, discourse, and pragmatic(s) under the label usage, and sociolinguistic(s), social and cultural under the label social. We then calculate the proportion of each item with regard to number of articles in each period, and average over the items in each group. If we then trace the development of the four groups over time, as is done in Figure 3, we can indeed conrm that the interest in meaning comes rst in the development of Cognitive Linguistics, while the interest in constructions, in usage-based models, in the social context of language takes o later. Also, while the attention for meaning has reached a stable level since the beginning of the 1990s, the focus on the other three forms of recontextualization is still increasing. (It should be noted that, while it is straightforward to have a look at the temporal development of the lines,

Underlying trends in thirty years of Cognitive Linguistics


we should be careful with a direct comparison of the level of the lines among each other. The gures give the proportion of articles that contain one of the topical keywords, calculated against the total number of articles in a given period. However, because dierent keywords may occur together within the same article, any given article may be referred to more than once at a certain point in time.)

4. Cognitive Linguistics in Context So far, we have established two crucial things. First, thinking about Cognitive Linguistics in terms of the recontextualization of grammar adequately captures the singular position of Cognitive Linguistics in the context of contemporary linguistics. If post-Chomskyan linguistics is characterized by an emerging tendency to reintroduce into the grammar those aspects of context that were discarded as irrelevant by hardcore generativism, then Cognitive Linguistics may be seen to embody that tendency in an outspoken way. All the relevant features meaning, the lexicon, the level of performance and language use, and the social context are saliently present in the theoretical and descriptive apparatus of Cognitive Linguistics. And some of them, like the importance of meaning and the usage-based character of linguistics, even belong to the core beliefs of Cognitive Linguistics. Second, thinking about Cognitive Linguistics in terms of the recontextualization of grammar not only helps to determine its external position in the context of linguistics at large, it also sheds a revealing light on the internal development of the framework. Parallel to the sociological expansion of Cognitive Linguistics as an international movement, we observe an internal expansion that is driven by the gradual recovery of context. The conceptual drift of Cognitive Linguistics (if we may use that term) seems to reside precisely in the recontextualizing movement. It appears, in other words, that recontextualization is a common denominator that keeps together at least conceptually speaking the various branches of Cognitive Linguistics that are being developed in an era of massive international expansion. But if, coming back to the introduction to this article, recontextualization can indeed be seen as the schematic node overarching the radial network of cognitive approaches, does that theoretical common denominator suce to keep Cognitive Linguistics together on a practical and sociological level? Will the centripetal force of an underlying trend be stronger than the centrifugal forces that are


Dirk Geeraerts

inherent in a radial development, where each separate topic of interest may start to develop into a domain of investigation and a sociological network of its own? 4.1. Cognitive Linguistics as a scientic paradigm To see more clearly why there might be a diculty, we may have a look at what the sociology of science has to say about the lifecycle of theories. Indeed, if we see Cognitive Linguistics as a paradigm in the sense of Kuhns theory of science, and if we then apply a sociological perspective to the development of the paradigm, we inevitably come across one of the most intriguing questions in the sociology of science: what determines the lifecycle of scientic paradigms, (sub)disciplines and specialties? There are various lifecycle models in the sociology of science, but for the present purposes, it may suce to have a look at the one that is presented in De Mey (1992). It distinguishes between four stages. In the pioneering stage, the paradigm is formulated. The methodological and rhetorical justication focuses on the originality of the approach. The organization of the eld is informal at most. In the building stage, the paradigm works according to the normal science model described by Kuhn: the basic tenets are accepted without questioning by the adherents; research is geared towards productively showing the applicability of the approach. At the same time, an organizational structure is set up, in the form of journals and conferences. The third stage, that of internal criticism, has a two-sided character. On the one hand, it is a period of conceptual organization: the ndings that were reached somewhat disorderly during the building stage have to receive a place within the overall framework; you might say: the body of knowledge accumulated in the building stage needs to be consolidated within the theoretical fabric of the paradigm. This is typically the stage, then, in which textbooks and reference works appear. On the other hand, the third stage is also the period in which the internal anomalies of the paradigm become apparent. The initial conrmation has been carried through massively in the building stage, so that the attention may now shift to the resistant problems. Those problems may inspire new research and new ideas within the paradigm, but they may also lead some people to question the overall approach and perhaps abandon the framework. In the nal stage, that of external criticism, the internal problems have become so important that the paradigm loses its attractiveness. Although

Underlying trends in thirty years of Cognitive Linguistics


the existing institutional organization may ensure a prolonged existence of the framework, it has stopped to grow, and it loses terrain with regard to newer, more appealing alternatives. Now, if we accept this classication, it is obvious that Cognitive Linguistics has moved beyond the pioneering stage (the rst decade) and the building stage (the second and third decade). The current situation, at the end of the third decade of its existence, denitely exhibits at least one side of the third stage mentioned by De Mey: we are living in a period of consolidation, in the form of textbooks and reference works. In fact, a retrospective article like the present one is probably typical of this stage in the development: one aspect of the consolidating movement is the eort to provide clear overviews of the history of the discipline. 4.2. Cognitive Linguistics within cognitive science If this is correct, we also need to ask the question whether the other aspects of the third stage of development a growing awareness of outstanding problems, increasing tensions between potentially rival approaches, fragmentation into independent subdisciplines are part of the current situation in Cognitive Linguistics. If recontextualization is at the center of Cognitive Linguistics, is there any indication that the centre may not hold? To round o this discussion of the basic tendencies within Cognitive Linguistics, we may try to identify a number of factors that could either constitute a threat or a safeguard for the unity of Cognitive Linguistics. On the negative side, we may observe that the thematic unity of Cognitive Linguistics as a recontextualizing approach to grammar is as yet a highly schematic one, which is not accompanied by an outspoken tendency towards theatrical unication or the development of appropriate methods. First, building a unied theory has never been a prominent feature of Cognitive Linguistics. Cognitive linguists work with a number of key notions that are motivated by general assumptions about language and cognition, but the exact relationship between many concepts and approaches within Cognitive Linguistics is still unclear. How exactly do a Langacker-type approach and a Talmy-type approach to grammar relate to each other? On which points are they notational variants, on which points are they compatible, on which points are they in opposition to each other? And if there is an incompatibility, what kind of evidence could decide between the two? Similar questions may be asked about Concep-


Dirk Geeraerts

tual Metaphor Theory and a Mental Spaces approach: in some respects, the latter is an extension of the former, but could it completely replace Conceptual Metaphor Theory or not? Further, what about the dierent crucial components of recontextualization (like the dierent forms of socially oriented Cognitive Linguistics)? In short, there are lots of theoretical issues in Cognitive Linguistics that need further clarication. Progress in Cognitive Linguistics now basically takes the form of developing and applying one or the other central concept or well established approach. Now that Cognitive Linguistics is reaching a mature age, systematic attempts at theory formation are called for. Second, if we can agree that Cognitive Linguistics is essentially characterized by a recontextualizing drift, there are specic consequences with regard to the observational basis and the analytical method of linguistic research. If one takes a usage-based model of the grammar seriously, one will have to study actual language use. In terms of the observational basis of Cognitive Linguistics, this suggests a shift from introspective conceptual analysis to the study of non-elicited language use, as epitomized by corpus linguistics, and to the study of on line processes, as epitomized by experimental research. Further, if one wishes to investigate how diverse factors like meaning, structure, discourse and lectal variation interact, the sheer complexity of the phenomena calls for appropriate analytic methods, i.e. for a shift towards quantitative testing of hypotheses. The complexity of a fully recontextualized grammar requires a methodology that goes beyond the traditional reliance on introspection: disentangling the eect of the various contextual factors that enter into the constitution of actual language use requires an advanced quantitative analysis that is able to capture the multivariate nature of linguistic usage. Now, while there are various indications for such an empirical turn in Cognitive Linguistics, the tendency is not dominant, and explicit defenses of an empirical approach are countered by equally outspoken defenses of introspection (cf., for instance, Geeraerts 2006b and Talmy 2007). Such a methodenstreit, if it continues, would obviously not contribute to the unity of the framework. In contrast with the absence of outspoken internal tendencies towards theoretical and methodological unication, there is an external factor that is likely to contribute strongly to the integrity of Cognitive Linguistics: the recontextualizing trend that we have identied as the unifying factor within the theoretical and methodological pluralism of Cognitive Linguistics is not restricted to linguistics. It is, in fact, a growing tendency within the cognitive sciences at large. In an overview article charting the

Underlying trends in thirty years of Cognitive Linguistics


development of cognitive psychology, Wagemans (2005) characterizes the recent evolution of cognitive science with two well-chosen phrases: downward into the brain and outward into the environment. By thinking of the human mind as a machine algorithmically manipulating symbols, the traditional paradigm of cognitive science (as it arose in the 1960s) isolates the mind both from the brain and the environment. By contrast, in the contemporary developments within cognitive psychology, intelligence became embodied again and cognition became situated in its context again (Wagemans 2005: 359). The parallelism with Cognitive Linguistics is striking, and ts entirely into Sinhas (2007) argument that Cognitive Linguistics is part of second generation cognitive science, just like generative linguistics (the culmination in the decontextualization of grammar) belonged to the cognitive revolution of the late 1950s and 1960s. The recontextualizing tendency in cognitive science at large will undoubtedly buttress the recontextualizing drift in Cognitive Linguistics. But at the same time, the question concerning the internal cohesion of Cognitive Linguistics receives a dierent interpretation when we consider Cognitive Linguistics in the context of cognitive science: what is important from this perspective is not primarily internal cohesion as such, but rather the way in which Cognitive Linguistics may optimally contribute to the interdisciplinary development of a contextualized cognitive science. But that, of course, is a task for the next thirty years.

Austin, John 1962

How to Do Things with Words: The William James Lectures Delivered at Harvard University in 1955. Oxford: The Clarendon Press.

Boroditsky, Lera 2000 Metaphorical structuring: Understanding time through spatial metaphors. Cognition 75: 128. Bresnan, Joan 2001 Lexical-Functional Syntax. Oxford: Blackwell Publishers. Brisard, Frank (ed.) 2002 Grounding: The Epistemic Footing of Deixis and Reference. Berlin/ New York: Mouton de Gruyter. Brown, Gillian and George Yule 1983 Discourse analysis. Cambridge: CUP. Brone, Geert and Jeroen Vandaele (eds.) 2009 Cognitive Poetics. Goals, Gains and Gaps. (Applications of Cognitive Linguistics 10.) Berlin/New York: Mouton de Gruyter.


Dirk Geeraerts

Diachronic linguistics; In: Dirk Geeraerts and Hubert Cuyckens (eds.), Handbook of Cognitive Linguistics, 945987. New York: Oxford University Press. Bybee, Joan and Paul Hopper (eds.) 2001 Frequency and the Emergence of Linguistic Structure. Amsterdam: John Benjamins. Chilton, Paul 2004 Analysing Political Discourse: Theory and Practice. London/ New York: Routledge (Taylor and Francis). Chomsky, Noam 1965 Aspects of the Theory of Syntax. Cambridge, Mass.: MIT. Clark, Herbert H. 1996 Using Language. Cambridge and New York: Cambridge University Press. Coulson, Seana 2006 Conceptual blending in thought, rhetoric and ideology. In: Gitte Kristiansen, Michel Achard, Rene Dirven and Francisco Ruiz de Mendoza Ibanez (eds.), Cognitive Linguistics: Current Applications and Future Perspectives, 187208. Berlin/New York: Mouton de Gruyter. Coulthard, Malcolm 1977 An Introduction to Discourse Analysis. London: Longman. Croft, William 2001 Radical Construction Grammar: Syntactic Theory from a Typological Perspective. Oxford: Oxford University Press. Croft, William 2009 Toward a Social Cognitive Linguistics. In: Vyvyan Evans and Stephanie Pourcel (eds.), New Directions in Cognitive Linguistics. Amsterdam: John Benjamins. Croft, William and D. Allan Cruse 2004 Cognitive Linguistics. Cambridge: Cambridge University Press. De Mey, Marc 1992 The Cognitive Paradigm. Chicago: The University of Chicago Press. Dik, Simon 1989 The Theory of Functional Grammar. Dordrecht: Foris. Dirven, Rene and Marjolein Verspoor (eds.) 1998 Cognitive Exploration of Language and Linguistics. Cognitive Linguistics in Practice 1. Amsterdam/Philadelphia: John Benjamins. Dirven, Rene, Frank Polzenhagen and Hans-Georg Wolf 2007 Cognitive linguistics, ideology, and critical discourse analysis. In: Dirk Geeraerts and Hubert Cuyckens (eds.), Handbook of Cognitive Linguistics, 12221240. New York, Oxford University Press. Dowty, David 1979 Word Meaning and Montague Grammar. Dordrecht: Reidel.

Bybee, Joan 2007

Underlying trends in thirty years of Cognitive Linguistics


Eggins, Susanne 1994 An Introduction to Systemic Functional Linguistics. London: Pinter. Evans, Vyvian and Melanie Green 2006 Cognitive Linguistics: An Introduction. Mawhaw, N.J.; Edinburgh, U.K.: Lawrence Erlbaum; Edinburgh University Press. Fairclough, Norman 1995 Critical Discourse Analysis: the Critical Study of Language. London: Longman. Fillmore, Charles J., Paul Kay and Catherine OConnor 1988 Regularity and idiomaticity in grammatical construction: The case of let alone . Language 64: 501538. Fischer, Kerstin 2000 From Cognitive Semantics to Lexical Pragmatics: The Functional Polysemy of Discourse Particles. Berlin/New York: Mouton de Gruyter. Freeman, Margaret 2007 Cognitive linguistic approaches to literary studies: state of the art in cognitive poetics. In: Dirk Geeraerts and Hubert Cuyckens (eds.), Handbook of Cognitive Linguistics, 11751202. New York: Oxford University Press. Gazdar, Gerald 1979 Pragmatics: Implicature, Presupposition, and Logical Form. New York: Academic Press. Geeraerts, Dirk 2003 Decontextualizing and recontextualizing tendencies in 20thcentury linguistics and literary theory. In: Ewald Mengel, HansJoerg Schmid and Michael Steppat (eds.), Anglistentag 2002 Bayreuth. 369379. Trier: Wissenschaftlicher Verlag. Geeraerts, Dirk 2006a A rough guide to Cognitive Linguistics. In: Dirk Geeraerts (ed.), Cognitive Linguistics: Basic Readings, 128. Berlin/New York: Mouton de Gruyter. Geeraerts, Dirk 2006b Methodology in Cognitive Linguistics. In: Gitte Kristiansen, Michel Achard, Rene. Dirven and Francisco Ruiz de Mendoza Ibanez (eds.), Cognitive Linguistics: Current Applications and Future Perspectives, 2149. Berlin/New York: Mouton De Gruyter. Geeraerts, Dirk (ed.) 2006c Cognitive Linguistics: Basic Readings. Cognitive Linguistics Research. Berlin/New York, Mouton de Gruyter. Geeraerts, Dirk and Stefan Grondelaers 1995 Looking back at anger: Cultural traditions and looking back at anger: Cultural traditions and metaphorical patterns. In: John R. Taylor and Robert E. MacLaury (eds.), Language and the


Dirk Geeraerts

Cognitive Construal of the World, 153179. Berlin/New York, Mouton de Gruyter. Geeraerts, Dirk and Hubert Cuyckens (eds.) 2007 The Oxford Handbook of Cognitive Linguistics. New York: Oxford University Press. Givon, Talmy 1979 On Understanding Grammar. New York: Academic Press. Goldberg, Adele E. 1995 Constructions: A Construction Grammar Approach to Argument Structure. Chicago: University of Chicago Press. Grice, Paul 1975 Logic and conversation. In: Peter Cole and Jerry L. Morgan (eds.), Speech Acts, 4158. New York: Academic Press. Gumperz, John (ed.) 1982 Language and Social Identity. Cambridge: CUP. Gumperz, John and Dell Hymes (eds.) 1972 Directions in Sociolinguistics: the Ethnography of Communication. New York: Holt, Rinehart and Winston. Halliday, Michael. A.K. 1994 Introduction to Functional Grammar. Second Edition. London: Edward Arnold. Harder, Peter 2003 The status of linguistic facts: Rethinking the relation between cognition, social institution and utterance from a functional point of view. Mind and Language 18: 5276. Harris, Randy A. 1995 The Linguistic Wars. Oxford: Oxford University Press. Hart, Christopher and Dominik Lukes (eds.) 2007 Cognitive Linguistics in Critical Discourse Analysis: Application and Theory. Cambridge: Cambridge Scholars Publishing. Haugen, Einar 1966 Language Conict and Languages Planning: the Case of Modern Norwegian. Cambridge, Mass.: Harvard University Press. Hopper, Paul 1987 Emergent grammar. Berkeley Linguistic Society 13: 139157. Hopper, Paul and Sandra Thompson 1980 Transitivity in grammar and discourse. Language 56: 251299. Hudson, Richard A. 1991 English Word Grammar. Oxford, Blackwell. Itkonen, Esa 2003 What is Language? A Study in the Philosophy of Linguistics. Turku: Abo Akademis tryckeri. Kemmer, Suzanne and Barlow, Michael 2000 Introduction. In: Usage-Based Models of Language, Michael Barlow and Suzanne Kemmer (eds.), viixxviii. Stanford: CSLI Publications.

Underlying trends in thirty years of Cognitive Linguistics


Kovecses, Zoltan 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, 181196, Berlin: Mouton de Gruyter. Kovecses, Zoltan 2005 Metaphor in Culture: Universality and Variation. Cambridge/ New York: Cambridge University Press. Kristiansen, Gitte, Michel Achard, Rene Dirven and Francisco Ruiz de Mendoza Ibanez (eds.) 2006 Cognitive Linguistics: Current Applications and Future Perspectives. Applications of Cognitive Linguistics 1. Berlin/New York: Mouton de Gruyter. Kristiansen, Gitte and Rene Dirven (eds.) 2008 Cognitive Sociolinguistics: Language Variation, Cultural Models, Social Systems. Berlin/New York, Mouton de Gruyter. Labov, William 1972 Sociolinguistic Patterns. Philadelphia: University of Pennsylvania Press. Lako, George 1977 Linguistic Gestalts. Papers from the Thirteenth Regional Meeting of the Chicago Linguistic Society. Chicago, Ill., Chicago Linguistic Society: 236287. Lako, George 1987 Women, Fire and Dangerous Things: What Categories Reveal about the Mind. Chicago: University of Chicago Press. Lako, George 1996 Moral Politics: What Conservatives Know that Liberals Dont. Chicago, Ill.: University of Chicago Press. Langacker, Ronald W. 1987 Foundations of Cognitive Grammar. Vol. 1.: Theoretical Prerequisites. Stanford, Calif.: Stanford University Press. Langacker, Ronald W. 1999 A dynamic usage-based model. In: Ronald W. Langacker, Grammar and Conceptualization, 91145. Berlin/New York, Mouton de Gruyter. Langacker, Ronald W. 2001 Discourse in cognitive grammar. Cognitive Linguistics 12: 143188. Langacker, Ronald W. 2005 Construction grammars: Cognitive, radical, and less so. In: Francisco Ruiz de Mendoza Ibanez and Sandra Pena Cervel (eds.), Cognitive Linguistics: Internal Dynamics and Interdisciplinary Interaction, 101159. Berlin/New York: Mouton de Gruyter. Levinson, Stephen C. 2003 Space in Language and Cognition: Explorations in Cognitive Diversity. Cambridge: Cambridge University Press.


Dirk Geeraerts

Montague, Richard 1974 Formal Philosophy. Selected papers of Richard Montague. New Haven: Yale University Press. Musol, Andreas 2006 Metaphor scenarios in public discourse. Metaphor and Symbol 21: 2338. Palmer, Gary B. 1996 Toward a Theory of Cultural Linguistics. Austin, Tex.: University of Texas Press. Partee, Barbara H. 1979 Semanticsmathematics or psychology? In: Rainer Baurle, Urs Egli and Arnim von Stechow (eds.), Semantics from Dierent Points of View, 114. Berlin: Springer. Rudzka-Ostyn, Brygida (ed.) 1988 Topics in Cognitive Linguistics. Current Issues in Linguistic Theory 50. Amsterdam/Philadelphia: John Benjamins. Sag, Ivan, Thomas Wasow and Emily Bender 2003 Syntactic Theory. A Formal Introduction. 2nd edition. Stanford: CSLI Publications. Sanders, Ted J. M. and Wilbert Spooren 2007 Discourse and text structure. In: Dirk Geeraerts and Hubert Cuyckens (eds.), Handbook of Cognitive Linguistics, 916941. New York: Oxford University Press. Saussure, Ferdinand de 1967 Cours de Linguistique Generale. Paris: Payot. First published [1916]. Searle, John 1969 Speech Acts: An Essay in the Philosophy of Language. Cambridge: CUP. Semino, Elena and John Culpeper (eds.) 2002 Cognitive Stylistics: Language and Cognition in Text Analysis. (Linguistic Approaches to Literature 1.) Amsterdam/Philadelphia: John Benjamins. Sinha, Chris 2000 Culture, language and the emergence of subjectivity. Culture and Psychology 6: 197207. Sinha, Chris 2007 Cognitive linguistics, psychology and cognitive science. In: Dirk Geeraerts and Hubert Cuyckens, (eds.), Handbook of Cognitive Linguistics, 12661294. New York: Oxford University Press. Slobin, Dan I. 1996 From thought and language to thinking for speaking. In: John J. Gumperz and Stephen C. Levinson (eds.), Rethinking Linguistic Relativity, 7096, Cambridge: Cambridge University Press.

Underlying trends in thirty years of Cognitive Linguistics


Stalnaker, Robert 1974 Pragmatic presuppositions. In: Milton K. Munitz and Peter Unger (eds.), Semantics and Philosophy, 197214. New York: New York University Press. Talmy, Leonard 1975 Figure and ground in complex sentences. In: Cathy Cogen, Henry Thompson, Graham Thurgood, Kenneth Whistler and James Wright (eds.), Proceedings of the First Annual Meeting of the Berkeley Linguistics Society, 419430. Berkeley, Calif.: Berkeley Linguistics Society. Talmy, Leonard 2007 Introspection as a Methodology in Linguistics. Plenary lecture presented at the 10th ICLC conference, Krakow, July 2007. Tannen, Deborah 1984 Coherence in Spoken and Written Discourse. Norwood NJ: Ablex. Taylor, John R. 1989 Linguistic Categorization: Prototypes in Linguistic Theory. Oxford: Clarendon Press. Thompson, Geo 1994 Introducing Functional Grammar. London: Arnold. Tomasello, Michael 2000 First Steps toward a Usage-based Theory of Language Acquisition. Cognitive Linguistics 11(12): 6182. Tomasello, Michael 2003 Constructing a Language: A Usage-Based Theory of Language Acquisition. Cambridge, Mass.: Harvard University Press. Tummers, Jose, Kris Heylen and Dirk Geeraerts 2005 Usage-based approaches in Cognitive Linguistics: A technical state of the art. Corpus Linguistics and Linguistic Theory 1: 225261. Ungerer, Friedrich and Hans-Jorg Schmid 1996 An Introduction to Cognitive Linguistics. London/New York: Longman. Verhagen, Arie 2005 Constructions of Intersubjectivity: Discourse, Syntax, and Cognition. Oxford: Oxford University Press. Verhagen, Arie and Jeroen Van de Weijer 2003 Introduction. In: Arie Verhagen and Jeroen van de Weijer (eds.), Usage-Based Approaches to Dutch, 16. Utrecht: LOT. Violi, Patrizia 2001 Meaning and Experience. Bloomington, Ind.: Indiana University Press. Wagemans, Johan 2005 Cognitive Psychology. In: Kimberly Kempf-Leonard (ed.), Encyclopedia of Social Measurement 1, 351359. Amsterdam: Elsevier.


Dirk Geeraerts

Weinreich, Uriel, William Labov and Marvin Herzog 1968 Empirical foundations for a theory of language change. In: Winfred P. Lehmann and Yakov Malkiel (eds.), Directions for Historical Linguistics, 98127, 150151, 183195. Austin: University of Texas Press. Zlatev, Jordan 2005 Whats in a schema? Bodily Mimesis and the grounding of language. In: Beate Hampe (ed.), From Perception to Meaning: Image Schemas in Cognitive Linguistics, 313342. Berlin/New York: Mouton de Gruyter.

How theory informs teaching and how teaching informs theory* Laura A. Janda
1. Introduction One of my best friends and favorite colleagues in the eld of Slavic linguistics is a generativist whom I have known for over thirty years now. A few years ago I asked him what applications there might be for generative linguistics in the language classroom. His answer was: Uh, none. That was a real non-starter and it was clear that I needed to move the conversation quickly to another topic. For my friend, linguistic theory is something akin to a pure science like mathematics, and application is unnecessary to justify its existence. Indeed, even the suggestion that one might undertake such applications was apparently insulting. There is a general tendency to overlook and underestimate the value of application in linguistics, and this tendency is supported by the hiring, tenure and promotion processes at universities, as well as by the reviewing and ranking of scholarly publications. I do not mean to imply that there is anything wrong with these structures, but they do have an eect on the relative prestige attached to achievements in theory in comparison with those in application. Nor do I wish to imply that either theory or application should be considered superior, one over the other. The intent is instead to show that theory and application can inform each other in a cyclic way as suggested by the title of this article. This article begins with a brief philosophical discussion on the roles of theory and application in linguistics in general and in cognitive linguistics in particular (section 2). Taking language pedagogy as an example of a domain where applications can be developed (cf. de Knop & de Rycker 2008), I describe the cycle of how theoretical endeavors are translated into language teaching materials, which then provoke further theoretical inquiry (section 3). I have gone through this cycle three times thus far (section 4), and can provide concrete examples of theoretical and pedagogical contributions relevant to case meaning (section 4.1), aspect meaning
* I would like to thank my employer, the University of Troms.


Laura A. Janda

(section 4.2), and aspectual clusters of verbs (section 4.3). I conclude (section 5) that the development of pedagogical applications can reveal further research opportunities, thus enhancing ones scholarly prole. Many of the pedagogical applications are available online via links from my homepage (; these are noted as follows: (OL).

2. What is the role of linguistics? This is a question I have often posed to both myself and colleagues. In terms of theory and application, we might suggest that there are two answers, but it is important to keep in mind that these two answers are not mutually exclusive. The rst answer is that linguistics is a theoretical enterprise and this justies the pursuit of science for sciences sake. There is nothing wrong with this. Indeed, if we did not have visionary theorists, we would not have our science at all. So science for sciences sake is important. The other answer is that linguistics can contribute to applications that are valuable for people other than linguists. Some of these applications might include teaching materials and language resources, computer technologies for dealing with real-language interfaces, primary research on documenting languages and dialects, and consulting on programs to revitalize smaller languages and get them needed political protections. Cognitive linguistics has a tradition of being accountable to other disciplines, like psychology, neurobiology, etc. It seems logical that this tradition of accountability should extend also to society by making our theoretical achievements accessible in applications that are useful to others. Fortunately results achieved in cognitive linguistics can be transferred fairly easily to applications without requiring users to master a theoretical linguistic artice because cognitive linguistics does not make any assumptions beyond those necessary and common to all cognition. In this article I focus on applications that can be used in the language classroom. All language learners have bodies, and they can use their embodied physical experience to make sense of the metaphors that underlie the grammars of foreign languages. Cognitive linguistics is utterly transparent in these applications, which do not require learners to have any linguistic expertise in order to access state-of-the-art analyses of linguistic subsystems of Slavic languages. I have often been warned that real linguists do not write language

How theory informs teaching and how teaching informs theory


textbooks, presumably because such activity would detract from their scholarly output and the results would not be comprehensible to the average layman. The aim of this article is to demonstrate that cognitive linguists can make contributions to language pedagogy while boosting their scholarly production at the same time.

3. How theory can inspire application and application can inspire theory This section contrasts the goals and audiences of theoretical contributions which are typically tightly focused, oering a narrow scope perspective on phenomena, with the goals and audiences of language teaching applications, which cover entire subsystems of languages and thus oer a broad scope perspective on phenomena. These two approaches can be complementary and mutually supportive. The narrow scope approach can make it possible to pinpoint a problem and work out a model, which can then be extended into a broader approach. The very process of extension inevitably uncovers some previously unnoticed wrinkles in the model, leading to more narrow scope investigations. Our scholarly publications tend to oer narrowly specied theoretical contributions aimed at a relatively small group of colleagues. This is partly due to the nature of science: linguistic research is hard to do, and the horizons of our knowledge are gradually pushed forward in tiny increments. It is also partly due to the peer-review process, which makes it easier to get narrow contributions published, and disadvantages broader works. A work that takes on a broader issue constitutes a bigger target for anonymous peer reviewers, and this problem is compounded if a scholar wants to suggest a truly original analysis of a broader issue, bringing into question traditional assumptions that most peer-reviewers are heavily invested in. It is thus often narrower works that receive the prestige of publication in scholarly journals. A consequence is that there is often less prestige and less recognition by our universities attached to broader applications. The audience of broader pedagogical applications is a potentially unlimited population of learners, whose goals include passing an exam at the end of the semester or learning to speak a language well. Learners are typically interested in macroscopic issues. They do not want part of an answer. They want an explanation that will apply as broadly and exhaustively as possible. The advantage is that they can force us as scholars to


Laura A. Janda

take a comprehensive look at a phenomenon and connect all the dots of a given system, often giving us new perspectives on language phenomena. As I have experienced it, the interaction between theory and application can be cyclic. Usually I start with a narrowly focused theoretical model developed on the basis of a limited dataset. This makes it possible for me to pinpoint a problem without excess noise and gure out what the relevant parameters are and how they interact. Once I understand the mechanics and have a model, I can extend it to account for an entire subsystem of a language and take it into the noisy environment of the language classroom. Thus a narrowly focused theoretical model is translated into a broadly focused pedagogical application. Usually the nice tidy solution suggested by a linguistic model turns out to need some readjustment when it is stretched to cover the needs of the learner, and these readjustments are opportunities for new research. The new patterns and issues revealed in this process thus inspire more narrowly focused research. Occasionally, this process also leads in the direction of a more philosophical discussion of linguistics (Janda 2008a). I nd that the end product of this cycle tends to open up entirely new directions that send me back through the cycle again with a new topic. As detailed below in section 4, work on the rst set of issues (case meaning) actually led to work on aspect meaning, which then led to work on aspectual clusters, so in a sense the process can be understood as a spiral.

4. Case studies of how theory and application inform each other This section presents three examples of cycles I have experienced in my work. In each example, I began with a theoretical investigation which led to a model designed to account for specic data. I was then tempted to try my model out in the classroom, and this led to the creation of pedagogical materials with a broader scope. However, in the course of creating pedagogical materials to account for entire subsystems in the grammars of Slavic languages, I discovered new patterns and phenomena that I might not have noticed otherwise. These discoveries launched further, more theory-oriented research. The pedagogical materials include interactive components, using sophisticated programming, graphics, and audio, and all of them are available in part or in their entirety over the Internet. These materials required teamwork to produce and have been supported by various grants; funding sources and collaborators are listed under Works Cited, Part 2.

How theory informs teaching and how teaching informs theory


4.1. Case meaning With the exceptions of Macedonian and Bulgarian, which retain only vestiges of a case system restricted to pronouns, in the remaining Slavic languages all noun phrases are case marked, in a system with six or seven cases: Nominative, Accusative, Dative, Genitive, Locative, Instrumental, and in some languages also Vocative. Case use, when explained at all in traditional grammars and textbooks, is usually described by listing typical syntactic roles, prepositions, and verbs associated with given cases. This information is atomistic and unsatisfactory for both linguistic description and language pedagogy. The research agenda described in this section was inspired by my own frustrations as a learner of Slavic languages. Often even though I knew all the words in a sentence and could parse it accurately, I still did not know what the sentence meant because my textbooks had told me about only a small portion of the uses of case. Later on, when I was able to consult academy grammars and linguistic publications I discovered that though they provided more detail, they lacked coherence. 4.1.1. Primary research on case meaning The initial goal of my theoretical research was to work out the relationships between the various submeanings in each grammatical case. From the perspective of cognitive linguistics, it was apparent that the grammatical cases were examples of polysemy, because each case had multiple meanings. As a cognitivist, I was then inspired to look for the prototypical meanings that motivated each case. I could further assume that the prototypical meanings were grounded in concrete physical experiences that served as the source domain for extensions to abstract metaphorical meanings, and that all of the meanings of any given case were related to each other in a radial category. Here is an example to illustrate the kind of problems I was faced with as both a language learner and a linguist. The Genitive case in Russian presents at least eighteen submeanings governed by a rather baing array of constructions, involving over one hundred prepositions, a couple dozen verbs, and a wide variety of quantiers. But when viewed from the perspective of a radial category, it is easier to see clear well-motivated patterns. One of the core prototypical meanings of the Genitive involves withdrawal from, most often associated with prepositions, as in this example:


Laura A. Janda

(1) Doc prisla iz skoly. [Daughter-N came from school-G] My daughter has come from school. This meaning of physical withdrawal serves as the source domain for other, metaphorical, types of withdrawal, thus explaining the association of verbs like bojatsja be afraid and styditsja be ashamed with the Genitive case, as in this example: (2) Doc bojalas/stydilas bednosti. [Daughter-N feared/was ashamed poverty-G] My daughter was afraid/ashamed of poverty. Similar explanations can be found throughout the landscape of Slavic case. The research agenda on case meaning unfolded gradually, tackling one case at a time and building up the theoretical model over the course of twelve years. It began with some peculiar uses of the Dative case in Czech that are grammatically superuous (not governed, but optional) and are used to assert authority or solidarity. These uses were explained in terms of a mapping of meaning into the pragmatic domain (Janda 1988). A larger study (Janda 1993) compared the full extent of two cases, the Dative and the Instrumental, in Czech and Russian. It was several more years before I managed to make sense of the Genitive (Janda 1999) and the Accusative (Janda 2000). 4.1.2. Pedagogical applications for case meaning I never actually nished the entire case system in developing the theoretical model, since I did not cover the Nominative and the Locative. But I had to make up for that deciency when Steven Clancy and I took up the creation of pedagogical materials, which were complete descriptions of the case systems for Russian (Janda and Clancy 2002) and Czech (Janda and Clancy 2006), with a third book in the series devoted to Polish soon to be completed. The case system is also featured in a reference grammar of Czech (Janda and Townsend 2000). Scholars continue to cite my 1993 book on the dative and instrumental, overlooking the case books, where the analysis is more complete and mature. No doubt this is due to the fact that most people are not looking for theoretical contributions in pedagogical works. Though cognitive linguistics is the basis for all these applications, it is

How theory informs teaching and how teaching informs theory


never mentioned by name, and indeed, we specically avoid any unnecessary linguistic terminology in the case books, restricting ourselves only to words like: noun, preposition, verb. These books prove that it is possible to translate a complex linguistic analysis into straightforward, useful language teaching materials without compromising on detail or depth. Sample chapters of the case books available online (OL) illustrate the manner of presentation, though they lack the interactive features of the full packages (which provide audio for all examples, with choice of male or female voice, plus instant navigation between indexes, table of contents, and text, making it possible to click and link to any meaning or use with any preposition, verb, etc.). The full menus of exercises can be accessed online (OL). All of the examples in the texts and exercises represent real language data, collected from various corpora, not sanitized textbook examples. Also, these materials are designed to cover all meanings of case in the given languages, which means that hundreds of examples are used, in both text and interactive exercises. A learner who has worked through these materials can expect to fully master the meanings of the cases, mimicking the prociency of a native speaker, who can interpret and use case meaning even in novel situations. One advantage of writing material for non-linguists is that ones audience is potentially unlimited, and this can lead to unexpected uses for such applications. Recently, for example, I was informed that The Case Book for Czech (Janda and Clancy 2006) is being used at Charles University in Prague to teach Czech to hearing-impaired students. For speakers of Czech Sign Language, the case system of Czech is just as foreign and mysterious as it is for speakers of non-Slavic languages, and colleagues in Prague have long been struggling to nd a way to help their deaf students over this barrier to achieving literacy in Czech. The system of connected meanings grounded in physical experiences shared by all human beings, whether deaf or hearing, is accessible to these students. Thus The Case Book for Czech is now being used to teach Czech in the Czech Republic too. 4.1.3. Further research on case meaning inspired by applications Work on the Case Books inspired multiple new theoretical inquiries. Within a given Slavic language there are often border zones in the case system where a nearly synonymous idea can be expressed by more than one case. These instances of case competition made for some ne-tuned analyses (Janda 2002a, 2002c, 2004b, 2004c). Furthermore, there are


Laura A. Janda

signicant dierences in the use of the same cases across the various Slavic languages. This led me to undertake some syntactic dialect geography, which yielded results mostly parallel to dialect geography of Slavic in terms of phonology (Janda 2002b, 2002e). Furthermore, when I looked at what kinds of cross-linguistic dierences in case marking there were within Slavic, I noticed that the domain that showed largest dierences was that of time. The time domain alone accounts for nearly a third of all variation in case marking among Slavic languages (Janda 2002d, 2002f ). Expressions of time were and continue to be a fascination, and this carried over into the later investigations of aspect, which are the topic of the remaining two case studies. I was not the only one to be fascinated by comparing case usage across Slavic languages my case book co-author, Steven Clancy, has undertaken a more sophisticated comparison supported by multi-dimensional scale modeling (Clancy 2006). Case meaning continues to oer me new opportunities. Goldbergs Constructions (1995) and Constructions at Work (2006) gave me the idea of piecing together a comprehensive construction grammar for a case language. A rst draft accounts for the range of transitivity constructions facilitated by case usage in Russian, presenting them as a radial category of constructions (Janda 2008c). This approach has been further extended in two dierent directions, both involving case use and construal. A striking feature of Russian as opposed to English is the use of the Dative case in impersonal expressions to express ideas that would occasion a personal construction in English. In other words, whereas in English we would say I was cold, I was having a hard time, I had to leave, in Russian the equivalent expressions are Mne bylo xolodno, mne bylo trudno, mne pris los ujti. All of the Russian expressions begin with a Dative rst person singular pronoun, and their literal meanings are to me it was cold, to me it was dicult, to me it arrived/was necessary to go. This tendency to avoid expression of agency in Russian has its roots in the meanings of the Dative case and the relationship of the Dative to other cases in closely related constructions. This is an issue that I have pursued together with Dagmar Divjak (Divjak and Janda 2008). Wierzbicka (1999) and Kovecses (2001) both claim that the metapho rical understanding of given emotions are language-specic, though the strategies may be universal. Russian has six terms typically translated as sadness: grust, pecal, toska, unynie, melanxolija, xandra. Together with Valery Solovyev, I have analyzed the relative corpus frequencies of the case constructions that these nouns appear in (Janda and Solovyev 2009). This type of analysis makes it possible to make precise distinctions

How theory informs teaching and how teaching informs theory


among the six synonyms, and also to examine what types of metaphors are used to understand the various types of sadness, revealing that only a portion of the phenomena are explainable via extension from the container source domain. Overall, case meaning is a treasure trove that I am still mining out. Creating pedagogical materials forced me to undertake a comprehensive description of case systems, but these in turn have yielded a wealth of ideas that I might never have had otherwise, because I might not have seen the patterns if I was not forced to connect all the dots. 4.2. Aspect meaning In the Slavic languages, a given verb, throughout its paradigm, is either Perfective or Imperfective, regardless of whatever other categories it might express. Like case meaning, aspect use was traditionally described in terms of long lists of uses, many of which appeared contradictory and unmotivated. 4.2.1. Primary research on aspect meaning Because I am a cognitive linguist, I suspected that there must be a metaphor motivating Slavic aspect, and that there must be an orderly polysemy behind the apparently chaotic inventories of usage. I went in search of a source domain that would account for the observed phenomena in a coherent way, and discovered that physical matter served this purpose (Janda 2004a). Physical matter, understood as a distinction between hard, solid objects vs. uid substances, is a rich source domain because human beings have many experiences relating to this distinction, and these experiences are isomorphic to the distinctions in the Slavic aspect system. Perfective events are understood as metaphorical discrete solid objects, which have clear boundaries and are unique and countable, as opposed to Imperfective events which are understood as metaphorical uid substances and therefore lack inherent boundaries and shapes, and are neither unique nor countable, but can be spread about. Furthermore, many of the seeming contradictions in use of aspect can be cleared up when we realize that this metaphor is applied at three dierent levels, and that subsequent levels can trump prior ones. At the rst level the metaphor applies to the inherent structure of events. At the discourse level, the metaphor applies to how events interact, which can motivate a dierent construal. And lastly at the pragmatic level this metaphor invokes dierences in satisfaction, comfort, and danger associated with physical objects, again moti-


Laura A. Janda

vating construals that can concur with or override those at the event and discourse level. Here is an illustration of how the rich domain of embodied experience we all have with matter motivates Russian aspect. First lets take a discrete solid object like an apple. It has a shape, it has edges, it has a distinct identity (no other apple can also be this apple, making this one unique), and I cannot put two apples into the exact same location the best I can do is to set one next to the other one. These are just a few of the things that we know about discrete solid objects thanks to our embodied experience. Like an apple, a Perfective event has a denite shape, which means that we know that it had a discrete beginning and/or end, that we are talking about a single, unique event, and when I have more than one such event (discourse level), the normal interpretation is that they are not at the same temporal location, but next to each other, and therefore sequenced. Here is a sentence with two Perfective events, each of which has clear boundaries and is unique. (3) Oleg sel v masinu i poexal v restoran. Oleg got into the car and drove to the restaurant. The normal interpretation of this sentence is as a sequence. This experience of discrete solid objects motivating Perfective aspect can be contrasted with that of substances motivating Imperfective events. A uid substance like sand has no inherent shape or boundaries and cannot be counted unless it is put in buckets. Furthermore sand is just sand, without unique identity, and it can be spread around. If there are two piles of sand, they can be blended together in the same place. Like sand, an Imperfective event is not understood to have a clear beginning or end, it does not have to be unique, and if there are two such events, they can easily occupy the same temporal location, making them simultaneous. Here is a sentence with two Imperfective events, neither of which tell us anything about temporal boundaries or uniqueness, and which are understood to be simultaneous. (4) Oleg nosil galstuk i ezdil na sportivnoj masine. Oleg wore a tie and drove a sportscar. Note that example (4) has the same syntactic structure as example (3), the only dierence being that the two past tense verbs joined by the conjunction in (3) are Perfective, whereas the verbs in (4) are Imperfective. These examples illustrate just a few of the meanings of Perfective and Imper-

How theory informs teaching and how teaching informs theory


fective that are isomorphic to parameters of physical matter; a fuller inventory can be found in Janda 2004a. 4.2.2. Pedagogical applications for aspect meaning In order to make the metaphorical model of Russian aspect accessible to teachers and learners of Russian I wrote an article for pedagogues (Janda 2003), and I got funding from the National Science Foundation to create an interactive media module (OL; note that starred items are under construction, completed items are the Introduction and, under Chapter 2, Module 1, sections headed shape and convertibility). The Aspect in Russian media module combines text, audio, graphics, and animations to show users the relevant parameters of physical matter and how they correspond to the behaviors of Perfective and Imperfective verbs. Users can conduct virtual experiments on solid objects and uid substances and compare the results with the uses of aspect in Russian. Key concepts can be reviewed, various interactive activities exercise the concepts, and all of the materials are illustrated with authentic natural language examples, with options to view translations and a choice of native speakers (male and female) oering models for pronunciation. The site encourages users to search the internet for further examples for analysis. The Aspect in Russian media module has been integrated into the Russian language curriculum at over two dozen institutions across the US and Europe. 4.2.3. Further research on aspect meaning inspired by applications As we saw with case meaning, work on a comprehensive pedagogical presentation of the model led me back to basic research and ultimately led me to start a new cycle too. Among the new things I worked on was an outline of how the matter metaphor diers in its extension across Slavic (Janda 2006). I also noticed that verbs showed various behaviors in terms of morphological derivation of Perfectives from Imperfectives and viceversa, and this initiated the idea of aspectual clusters. Furthermore, I began to realize that some similar metaphors were at work in determining what kinds of aspectual relations there were between verbs within a cluster (Janda 2008b). Thus my next move was to explore aspectual relationships among verbs built from the same lexical item, what I call clusters. 4.3. Aspectual clusters of verbs Traditionally it has been assumed in Slavic that the relationship between Perfective and Imperfective verbs built from the same lexical item was


Laura A. Janda

that of an aspectual pair containing a single Imperfective verb and its Perfective aspectual partner. The process of collecting data and illustrative examples for the metaphorical model of aspect gave me the insight that there could be an alternative to the pair model, a model that would comport better with the messy reality of aspectual relationships among Russian verbs. This is how the cluster model came into being. Given that there are aspectual relationships between Perfective and Imperfective verbs, there is no necessity to assume that they are paired in a one-to-one relationship. Other relationships, namely one-to-many relationships, could also account for the phenomena traditionally described as pairs, and I found that Russian actually has dierent types of Perfectives. My proposal is that aspectual pairs represent only a portion of a system where an Imperfective verb can be related to a number of Perfectives. 4.3.1. Primary research on aspectual clusters I rst worked out the cluster model by taking a multiply stratied sample of the morphological types of Russian verbs. This means that I included all verbs from all non-productive classes, plus samples of productive classes. In other words, I used linguistic criteria to build this database, in order to assure that I had accounted for all morphological types of verbs. This meant that all paradigm types were included regardless of their type frequency. My database contains 283 clusters and approximately 2,000 verbs. Research on this database made it possible to discover the four types of Perfectives and the implicational hierarchy that determines how these elements can be combined in aspectual clusters of Russian verbs (Janda 2007). It is possible to distinguish four types of Perfectives in Russian. There is a Natural Perfective which describes the culmination of a completable activity, and it is usually this Perfective that is considered the aspectual partner in the pair model. So Russian has two verbs for write: one, pisat, is Imperfective and describes the activity, and one, napisat, that describes the completion of a document. In addition, there are Specialized Perfectives that give a specic path and goal to the action, and thus provide enough new lexical information to motivate the derivation of secondary imperfectives. Adding a prex to pisat write gives a Specialized Perfective such as perepisat rewrite which can furthermore be suxed to give a secondary Imperfective, perepisyvat with the same meaning used to describe a process or repeated action. There are also Complex Act Perfec-

How theory informs teaching and how teaching informs theory


tives which take an atelic action and give it temporal boundaries, usually expressing action that lasts a certain time or begins or ends, in all cases without result. An example of a Complex Act Perfective is popisat, write for a while without result. Finally, with some verbs it is possible to form a Single Act Perfective which removes a single cycle from a repeated atelic action, such as cixat sneeze, which has a Single Act Perfective cixnut used to describe a single sneeze. An aspectual cluster can contain zero, one, two, three or all four types of Perfectives, but there are strict constraints on what combinations are possible. In fact most theoretically possible combinations are not attested, and those that are attested follow an implicational hierarchy, described in detail in Janda 2007. The cluster model oers several advantages over the traditional pair model. Firstly, the cluster model accounts for more aspectual relations among verbs, giving a more accurate picture of the aspectual system in Russian. Secondly, the pair model is often used to incorrectly identify Complex Act and Single Act Perfectives as the Perfective partners of Imperfective verbs, especially in clusters that lack Natural Perfectives. 4.3.2. Pedagogical applications for aspectual clusters The pair model is just as inadequate in language teaching as it is as a linguistic model. To remedy this situation, John Korba and I built a second database with the aim of providing a resource for instructors and learners. This database contains the 266 clusters of the verbs listed in the vocabularies for a rst-year and a second-year textbook of Russian. This pedagogical database was designed to represent high-frequency verbs most useful for learners, regardless of morphological class (i.e., token frequency). There was some overlap in the linguistic and pedagogical databases and the databases were approximately the same size. The results of the pedagogical project were much more interesting than I had anticipated. On the one hand, the cluster model was perfectly conrmed by the pedagogical database. The pedagogical database showed the same four types of Perfectives, and the same implicational hierarchy, giving the same range of possible cluster structures. However the distribution of the cluster types was not identical in the two databases. Figure 1 compares the frequencies of the four most important cluster structures in the two databases. Whereas the linguistic database gave an order of cluster structures that seems rather arbitrary (A NP SP CA then A NP SP then A NP SP CA SA, then A NP), the pedagogical database gives


Laura A. Janda

Figure 1.

an order that directly follows the implicational hierarchy, namely A NP then A NP SP then A NP SP CA, then A NP SP CA SA. This means that high-frequency verbs showed that the implicational hierarchy was more than just a predictor of possible structures in the system, but also a predictor of how frequent they were. This discovery has a valuable pedagogical implication, since given this distribution it makes sense to teach students the implicational hierarchy, so that they can predict cluster structures and variants. On the basis of this discovery, we composed an article oering suggested instructional strategies and exercises (Janda and Korba 2008). We also published to a website the pedagogical database (OL), which makes it possible for instructors and learners to look up the aspectual clusters of given verbs or to nd groupings of verbs according to cluster structure. 4.3.3. Further research on aspectual clusters inspired by applications Several further projects have been inspired by the work done on pedagogical applications of the cluster model. These projects involve groups

How theory informs teaching and how teaching informs theory


of verbs traditionally considered to be aspectually anomalous and the so-called empty prexes. There are two types of verbs that are often considered to be aspectually anomalous, the rst are the motion verbs, and the second are the biaspectual verbs. The motion verbs make an additional distinction within Imperfective between travel to a destination and other kinds of motion, and are notoriously hard to learn. Work on the pedagogical database conrmed that the motion verbs are not anomalous, but actually prototypical, for they serve as the metaphorical motivation for the types of Perfectives that can be formed (Janda 2008b). All completable (telic) verbs are understood as metaphorical directed motion verbs, where an activity is leading to a result, and all non-completable (atelic) verbs are understood as metaphorical non-directed motion verbs, where activity is not leading to a result. Furthermore, motion verbs display the maximal cluster structure; all other cluster structures can be arrived at by removing types of Perfectives from the structure associated with motion verbs. The biaspectual verbs use only one form to express both aspects (always disambiguated by context). These verbs were considered anomalous because they violated the one-to-one correspondence expectations of the aspectual pair model. However, within the cluster model we nd many form-meaning correspondences other than one-to-one within aspectual clusters, and indeed the biaspectual verbs are not so unusual after all (Janda forthcoming). The cluster model also made a prediction about the cluster structures that would be possible for bi-aspectual verbs, namely that bi-aspectual verbs should be negatively correlated with the formation of Complex Act Perfectives, a prediction that was conrmed in an empirical study (Janda 2007b). It has been traditionally assumed that the prexes used to form Natural Perfectives are semantically empty. There are numerous theoretical problems with the notion of the empty prex, among them the fact that there are over a dozen such empty prexes, so why would Russian need dierent ones for dierent verbs, and also the fact that the same prexes can be used to form the other kinds of Perfectives, in which instances (especially in the case of Specialized Perfectives) it is clear that they have semantic content, so why would the prexes be sometimes empty and sometimes not? In a future project I hope to prove that what we have is conceptual overlap, not semantic emptiness. The cluster model oers a principled way to distinguish among the various types of Perfectives, and there may be a correlation between cluster structure, verb semantics, and prexal semantics.


Laura A. Janda

5. Conclusions The bibliography usually plays only a supporting role in an article, but I would like to use it to make a point. The bibliography that follows has been arranged under several subheadings: Primary Research, Applications, Research Inspired by Applications, and Other Works Cited. Comparison of the volume of output shows that the bulk of publications came in round three, the Research Inspired by Applications, when the wealth of both the primary research and the applications provided a knowledge base to build upon. When I look at this distribution, I realize that building applications for instructors and learners is not just something I did to be altruistic. It turns out that the resources that I created for others also forced me to take a comprehensive, big-picture look at phenomena, and that has brought very tangible benets to my own research agenda.

Works Cited, Part 1: Primary Research

Janda, Laura A. 1988 Pragmatic vs. semantic uses of case. In: Diane Brentari, Gary Larson and Lynne MacLeod (eds.), Chicago Linguistic Society 24-I: Papers from the Twenty-Fourth Regional Meeting, 189 202. Chicago: University of Chicago Press. Janda, Laura A. 1993 A Geography of Case Semantics: The Czech Dative and the Russian Instrumental. (Cognitive Linguistics Research 4). Berlin: Mouton de Gruyter. Janda, Laura A. 1999 Peircean semiotics and cognitive linguistics: a case study of the Russian genitive. In: Michael Shapiro (ed.), The Peirce Seminar Papers, 441466. New York/Oxford: Berghahn Books. Janda, Laura A. 2000 A cognitive model of the Russian accusative case. In: Rodmonga K. Potapova, Valery D. Solovev and Vladimir N. Poljakov (eds.), Trudy mezdunarodnoj konferencii Kognitivnoe modelirovanie, No. 4, part I, 2043. Moscow: MISIS. Janda, Laura A. 2004a A metaphor in search of a source domain: the categories of Slavic aspect. In: Cognitive Linguistics 15(4): 471527. Janda, Laura A. 2007a Aspectual clusters of Russian verbs. In: Studies in Language 31(3): 607648.

How theory informs teaching and how teaching informs theory


Works Cited, Part 2: Applications publications

Janda, Laura A. 2003 A user-friendly conceptualization of Aspect. In: Slavic and East European Journal 47(2): 251281. Janda, Laura A. and Steven J. Clancy 2002 The Case Book for Russian. Bloomington, IN: Slavica. Funding sources: Chancellors Award for Instructional Technology, Title VI Dept of Education Grant for the Joint Duke-UNC Slavic and East European Language Resource Center. Janda, Laura A. and Steven J. Clancy 2006 The Case Book for Czech. Bloomington, IN: Slavica. Funding source: Title VI Dept of Education Grant for the Joint DukeUNC Slavic and East European Language Resource Center. Janda, Laura A. and John J. Korba 2008 Beyond the pair: Aspectual clusters for learners of Russian. Slavic and East European Journal 25(2): 254270. Janda, Laura A. and Charles E. Townsend 2000 Czech ( Languages of the World/Materials 125. Munich/ Newcastle: lincom europa.

Internet resources
Aspect in Russian Media Module: Funding sources: Title VI Dept of Education Grant for the Joint Duke-UNC Slavic and East European Language Resource Center, NSF Proposal #0341628 for Curriculum, Laboratory and Instructional Material Development, NSF Proposal #0550129 supplemental award; collaborators: Catherine Macallister, Donald Loand, Kerry OSullivan, Eleonora Magomedova, Yuri Panov. Cluster Types for Russian Verbs: Funding source: Title VI Dept of Education Grant for the Joint Duke-UNC Slavic and East European Language Resource Center; collaborators: Miroslav Styblo, John J. Korba.

Works Cited, Part 3: Research inspired by applications

Clancy, Steven J. 2006 The topology of Slavic case: semantic maps and multidimensional scaling. In: Glossos 6, at Divjak, Dagmar and Laura A. Janda 2008 Ways of attenuating agency in Russian. In: Anna Siewierska (ed.), Impersonal Constructions in Grammatical Theory, a special issue of Transactions of the Philological Society (v. 106), 138 179.


Laura A. Janda

Janda, Laura A. 2002a Semantika padu v cestine. In: Alena Krausova, Marketa Sleza n kova and Zdenka Svobodova (eds.), Setka s Cestinou, 2935. Prague: Ustav pro jazyk cesky. Janda, Laura A. 2002b Cases in collision, cases in collusion: the semantic space of case in Czech and Russian. In: Laura A. Janda, Steven Franks and Ronald Feldstein (eds.), Where Ones Tongue Rules Well: A Festschrift for Charles E. Townsend, 4361. Columbus, Ohio: Slavica. Janda, Laura A. 2002c Cognitive hot spots in the Russian case system. In: Michael Shapiro (ed.), Peircean Semiotics: The State of the Art ( The Peirce Seminar Papers 5), 165188. New York: Berghahn Books. Janda, Laura A. 2002d The conceptualization of events and their relationship to time in Russian. In: Glossos 2 at Janda, Laura A. 2002e The case for competing conceptual systems. In: Barbara Lewandowska-Tomaszczyk and Kamila Turewicz (eds.), Cognitive Linguistics Today ( odz Studies in Language 6), 355374. Frankfurt: Peter Lang. Janda, Laura A. 2002f Concepts of case and time in Slavic. In: Glossos 3 at http:// Janda, Laura A. 2004b Border zones in the Russian case system. In: Ju. D. Apresjan (ed.), Sokrovennye Smysly (a festschrift for Nina D. Arutjunova), 378398. Moscow: Jazyki slavjanskoj kultury. Janda, Laura A. 2004c The dative case in Czech: what it means and how si ts in. In: the published proceedings of the annual meeting of the Spolecnost pro vedy a umen 2003, published in 2004 at: Janda, Laura A. 2006 A metaphor for aspect in Slavic. In: Henrik Birnbaum in Memoriam ( International Journal of Slavic Linguistics and Poetics), 4445, 249260. Janda, Laura A. 2007b What makes Russian bi-aspectual verbs Special. In: Dagmar Divjak and Agata Kochanska (eds.), Cognitive Paths into the Slavic Domain. (Cognitive Linguistics Research). Berlin/New York: Mouton de Gruyter, 83109. Janda, Laura A. 2008a From Cognitive Linguistics to Cultural Linguistics. Slovo a smysl/Word and Sense 8, 4868.

How theory informs teaching and how teaching informs theory


Janda, Laura A. 2008b Semantic motivations for aspectual clusters of Russian verbs. In: Christina Y. Bethin, ed. American Contributions to the 14th International Congress of Slavists, Ohrid, September 2008. Bloomington, IN: Slavica Publishers, 181196. Janda, Laura A. 2008c Transitivity in Russian from a Cognitive Perspective. In: Galina Kustova (ed.), Dinamiceskie Modeli: Slovo. Predlozenie. Tekst. Sbornik statej v cest E. V. Paducevoj Moscow: Jazyki slavjanskoj kultury, 970988. Janda, Laura A. 2009 Totally normal chaos: The aspectual behavior of Russian motion verbs. In: a festschrift for Michael S. Flier (Harvard Ukrainian Studies 28), 183193. Janda, Laura A. Forthcoming Mesto dvuvidovyx glagolov v modeli vidovyx gnezd. In: Marina Ju. Certkova (ed.). Moscow State University. Janda, Laura A. and Valery D. Solovyev 2009 What Constructional Proles Reveal About Synonymy: A Case Study of Russian Words for sadness and happiness. Cognitive Linguistics 20: 2, 367393.

Works Cited, Part 4: Other Works Cited

Sabine de Knop and Teun de Rycker (eds.) 2008 Cognitive Approaches to Pedagogical Grammar. (Applications of Cognitive Linguistics 9.) Berlin/New York: Mouton de Gruyter. Goldberg, Adele 1995 Constructions: A Construction Grammar Approach to Argument Structure. Chicago: Chicago University Press. Goldberg, Adele 2006 Constructions at Work: The Nature of Generalizations in Language. Oxford: Oxford U Press. Kovecses, Zoltan 2001 Metaphor and Emotion: Language, Culture, and Body in Human Feeling. Cambridge: Cambridge University Press. Wierzbicka, Anna 1999 Emotions across Languages and Cultures: Diversity and Universals. Cambridge: Cambridge University Press.

Part II.

The context for prototypes

Why not? Prototypes and blocking of language change in Russian verbs* Tore Nesset
This paper explores the relevance of prototypes for morphological change. Traditionally, historical linguistics has described what happens in language change and tried to explain why. However, what about the changes that do not take place? Can we explain them? Are prototypes relevant? I will discuss three cases where an ongoing language change in Russian verbs that I call sux shift has been blocked. Three factors pertaining to prototypes are shown to be relevant: simplication, target consistency and product-oriented generalizations. Although when viewed in isolation, none of the factors is sucient to explain why sux shift is blocked, I will argue that in concert they provide a reasonable account of why sux shift did not take place. In addition to accounting for blocking, this paper has implications for future change. Although sux shift has been blocked for fricative-nal roots so far, verbs of this type are shown to be in a more vulnerable position than the other verbs that display blocking in present-day Russian. If sux shift spreads to new classes of verbs, the proposed analysis predicts that fricative-nal roots are most likely to be aected, and that these roots will undergo sux shift rst. The implications of the present study go beyond Russian and Slavic linguistics. First, it is shown that blocking is a fruitful area of research in historical linguistics and that simplication, target consistency and product-oriented generalizations are relevant for blocking of language change in other languages. Second, this study emphasizes the importance of the target category in language change; all the hypotheses in this paper refer to the target category. Third, the investigation of target consistency demonstrates how important the sizes of morphological classes are for language change. More specically, the proposed analysis suggests that a category needs to be about three times as large as a competing category
* I would like to thank my employer, the University of Troms, for nancial support and L. Janda, V. Plungjan, E. Romanova, P. Iosad, E. Tabakowska and audiences in Krakow and Copenhagen for comments on earlier versions of this article.


Tore Nesset

in order to steal members from it. Last but not least, the account of blocked sux shift in this article provides evidence in favor of linguistic frameworks like Cognitive Linguistics, which accommodate productoriented generalizations and prototypes. Section 1 oers an overview of sux shift in Russian verbs. In sections 2 through 4, I consider simplication, target consistency and productoriented generalizations for prototypes, before the contribution of the paper is summarized in section 5.

1. Sux shift in Russian verbs In the overwhelming majority of Russian verbs, the root is followed by a verbal sux. Five of these suxes, /aj/, /ej/, /ova/, /i/ and semelfactive /nu/, form productive verb classes, while the remaining patterns are nonproductive.1 In situations like this, the historical linguist would expect migration of verbs from non-productive to productive patterns. This prediction is indeed borne out. As is well known in Slavic linguistics, many verbs replace the non-productive verbal sux /a/ with the productive /aj/. It is not hard to understand why. The two suxes are homophonous in the innitive and past tense. In these forms, where the verbal sux is followed by a consonant-initial inectional ending, both suxes have the shape /a/. On the basis of innitives like delat do and kapat drip and past tense forms like delal and kapal, there is no way to infer that delat has the /aj/ sux, whereas kapat belongs to the non-productive pattern /a/-verbs. Apparently, the homophony in the innitive and past tense has formed the basis for a diachronic change, whereby the productive sux supplants its non-productive competitor in the remaining verb forms. In the terminology of Andersen (1973), this is an example of abductive change. Based on the ambiguous evidence of the innitive and past tense forms, the speakers make the wrong inference about the sux in the present tense and imperative. As a result of this, sux shift takes place.

1. According to standard practice, I use slashes for examples in phonemic transcription. For the convenience of the reader, I place a hyphen between a root and a derivational sux and a plus sign before an inectional ending: /del-aj ut/ (they) do. When the sound shape of an example is not essential the example is cited in transliterated orthography, indicated by italics.

Prototypes and blocking of language change in Russian verbs


By way of illustration, consider the following examples both representing 3 pl present tense forms of the verb kapat drip:2 (1) Slezy kapljut odna za drugoj na klavisi. (Goncarov 1859, with /a/ sux) Tears are dripping one after another onto the keyboard. (2) Slezy v sci kapajut. (Bitov 1969, with /aj/ sux) Tears are dripping into the cabbage soup. The form kapljut in (1) represents the original, non-productive pattern. According to the (morpho-)phonological rules of Russian, the /a/ sux is truncated before a vowel-initial ending so as to avoid hiatus. Example (2), on the other hand, includes the innovative /aj/ sux. We are dealing with ongoing language change that has created a situation where the /aj/ and /a/ suxes compete for a number of verbs. As observed by Henning Andersen, who provides a detailed study of sux shift, the synchronic situation of the relevant verbs has all the earmarks of a change in progress (Andersen 1980: 297). In the case of kapat drip, Zaliznjak (1977) considers both the conservative forms in (1) and the innovative forms in (2) acceptable in Contemporary Standard Russian. However, when kapat is used as a transitive verb, Zaliznjak (1977) accepts only the /aj/ sux. Sux shift in Russian verbs is well attested from numerous studies of language change, language acquisition, psycholinguistics, stylistic variation, sociolinguistics and dialectology (cf., e.g., Andersen 1980; Gagarina 2003; Gor and Chernigovskaya 2001, 2003a, 2003b, 2003c; Kiebzak-Mandera, Smoczynska and Protassova 1997; Krysin 1974; Tkachenko and Chernigovskaya 2006 and references therein). In order to shed light on the phenomenon, I examined all the verbs with the /a/ sux listed in the Russian Academy Grammar (Svedova 1980). I elicited all the examples with these verbs in the Russian National Corpus and created a database of about 20,000 examples. This database makes it possible to investigate the factors that facilitate sux shift, but also the factors that inhibit or block sux shift, which is the topic of the present article. It is interesting to notice that blocking has been outside the focus of previous research on sux shift in Russian verbs; the emphasis has been on factors that facilitate
2. The examples are elicited from the Russian National Corpus available at


Tore Nesset

Table 1. Sux shift in Russian /a/-verbs 3 Root type a) Syllabic, nal velar b) Syllabic, nal labial c) Syllabic, nal dental plosive d) Syllabic, nal dental fricative e) Syllabic, nal /j/ f ) Non-syllabic Sux shift 2 (17) of 20 verbs 4 (7) of 9 verbs 2 (10) of 31 verbs not attested: 0 (0) of 12 verbs not attested: 0 (0) of 20 verbs not attested: 0 (0) of 13 verbs Example /dvig-a t/ move /kap-a t/ drip /prat-a t/ hide /maz-a t/ smear /laj-a t/ bark /br-a t/ take

sux shift. Table 1, which reports some of the results in the database, shows that the shape of the verb root is relevant for blocking. For syllabic roots, sux shift is frequent if the root ends in a labial or velar consonant, or if the root-nal consonant is a dental plosive. However, for verbs with a root-nal dental fricative or /j/, sux shift is blocked. Blocking furthermore occurs for non-syllabic roots. The question is why. Why is the /aj/ sux incompatible with categories (d)(f ) in Table 1, which, for ease of reference, I will refer to as fricative-nal, /j/-nal and non-syllabic? Why are conceivable verb forms like */maz-aj ut/ from mazat smear unattested? Why do /j/-nal roots not combine with /aj/ to form conceivable, but unattested, present tense forms like */laj-aj ut/ from lajat bark? Why do present tense forms like */br-aj ut/ not compete with

3. The investigation was based on the verbs listed in the Russian Academy Grammar (Svedova 1980: 653655). The numbers in the table include only non-prexed verbs without the postx -sja. The numbers show how many purported /a/-verbs in each category display sux shift in actual usage (as reected in my database). Statements on the form X (Y) of Z verbs indicate that X of a total of Y verbs display sux shift in 50% or more of the elicited examples. The number in parentheses reects the number of verbs where sux shift was attested at all. Notice that the two isolated verbs stonat moan and sosat suck were not included in the investigation. Also omitted were the verbs kloxtat cluck, krapat spatter, scepat chip and svistat whistle, for which there were less than 10 examples each available in the Russian National Corpus (

Prototypes and blocking of language change in Russian verbs


/ber ut/ for verbs with non-syllabic roots like brat take? In sections 2 through 4, I will discuss three hypotheses. As we will see, they all relate to prototype theory.

2. Blocking and simplication The fact that fricative-nal, /j/-nal and non-syllabic roots do not undergo sux shift suggests that verbs with such roots lack the factors that motivate this diachronic process. What would this motivation be? A possible answer would be that language change involves simplication. Following this line of reasoning, we would not expect language change to take place unless the result is simplication: (3) Simplication hypothesis: Sux shift occurs only if it involves simplication. It is probably not the case that language change always leads to simplication of (a component of ) the grammar (cf. e.g. Lass 1997: 252257), and it is furthermore not clear exactly what counts as simple. However, in-depth discussion of these questions is beyond the scope of the present study. For our purposes it is sucient to notice that a very widespread way for language change to simplify grammars is analogical leveling of morphophonological alternations. This is sometimes referred to as Manczaks second tendency, because it was stated explicitly in an inuential paper by the Polish linguist Witold Manczak (1963): (4) Manczaks second tendency: Alternations within paradigms are more often abolished than introduced (cited after McMahon 1994: 79).4 Characteristic for the inection of /a/-verbs is a family of morphophonological alternations traditionally labeled softening.5 In the present tense and imperative forms, where the inectional ending is vowel-initial, the /a/ sux is truncated, and the root-nal consonant undergoes softening. For instance, kapat drip displays the softening alternation /p/ P /pl/
4. Manczaks (1963: 22) original statement was: Lalternance du radical est plus souvent abolie quintroduite. 5. A full overview of softening in Russian is beyond the scope of this study. The reader is referred to grammars (e.g. Svedova 1980; Isacenko 1982; Timberlake 2004). For a detailed study with particular emphasis on the paradigmatic dimension of softening, see Andersen 1995.


Tore Nesset

(cf. 3 pl present tense /kapl ut/), and murlykat purr has an alternation /k/ P /c/ (cf. 3 pl present tense /murlic ut/). Verbs with the /aj/ sux, on the other hand, do not display softening. The verbal sux is not truncated before vowel-initial endings, and the root-nal consonant does not alternate, as shown by innovative present tense forms like /kap-aj ut/ and /murlik-aj ut/ where the roots are intact. In other words, sux shift eliminates the softening alternation, and in that sense this process involves a simplication of Russian conjugation. However, there is one exception. The /j/-nal roots do not have a softening alternation, since /j/ does not alternate. In lajat bark, for instance, the root is /laj/ in past tense forms like /laj-a l/ and present tense forms like /laj ut/. If we adopt hypothesis (3), we would expect sux shift to take place only to the extent that it involves simplication. For /j/-roots we expect blocking, because these verbs do not display the softening alternation and hence there is no potential for simplication in sux shift. This prediction is borne out by the facts. As we saw in section 2, sux shift is consistently blocked for /j/-nal roots. What about non-syllabic roots and fricative-nal roots? Verbs with the /a/ sux and non-syllabic or fricative-nal roots display softening alternations, so sux shift would indeed simplify their conjugation. The conceivable, but unattested present tense forms */br-aj ot/ (they) take and */maz-aj ot/ (they) smear do not involve softening and are therefore simpler than the corresponding attested forms /ber ot/ and /maz ot/. Nevertheless, sux shift is not attested. Since this cannot be explained by the simplication hypothesis in (3), we are left with two options. Either we can revise the hypothesis so as to cover non-syllabic and fricative-nal roots, or we can come up with additional explanations for these groups of verbs. I propose pursuing both strategies. In the remainder of this section, I will discuss a relativized version of the simplication hypothesis, before we turn to additional hypotheses in later sections. The idea I pursue is simple. The softening alternations dier in complexity, and this may have an impact on sux shift. Consider the relativized hypothesis, which is sensitive to dierent degrees of complexity: (5) Relativized simplication hypothesis: Sux shift is more likely to occur if it involves a high degree of simplication. The hypothesis in (5) is a relativized version of (3), insofar as (5) acknowledges dierent degrees of simplication. In order to evaluate the implications of (5), we need to consider some subtypes of the softening alternation in Russian verbs. There are two main types, which are traditionally known

Prototypes and blocking of language change in Russian verbs


as substitutive (or transitive) and plain softening. Examples of the substitutive type are /t/ P /c/ and /s/ P /s /, while /t/ P /t/ and /s/ P /s/ illustrate plain softening. It seems fair to say that plain softening is simpler than substitutive softening. Plain softening involves only the addition of a secondary place of articulation. For instance /s/ and /s/ are identical except that the latter is palatalized, i.e. is articulated with a secondary palatal place of articulation in addition to the dental (alveolar) primary place. For substitutive softening, the change of the consonant is more drastic insofar as the primary place is changed. More features are changed, and hence the alternation is more complex. However, not all versions of substitutive softening are equally complex. While alternations like /s/ P /s/ only involve change of the place of articulation, /t/ P /c / also concerns manner features, insofar as a plosive alternates with an aricate.6 In traditional terms, /s/ P /s / involves palatalization only, while /t/ P /c / represents a combination of palatalization and lenition.7 I submit that alternations involving only palatalization are less complex than alternations that combine palatalization with lenition. I propose the following hierarchy of complexity: Before we discuss the ramications of this hierarchy, a remark on labial sounds is in order. As can be seen from Table 2, I have placed them at the highest level of complexity, i.e. in the same slot as plosives that display palatalization and lenition. The claim that alternations like /p/ P /pl/ involve palatalization does not need substantiation, since /l/ is palatalized. But is /p/ P /pl/ an example of lenition? In part, at least, this is a matter of denition. However, the similarities are evident. The sequence /pl/ resembles aricates like /c/ in that it starts with a plosive and ends with a continuant. In other words, alternations like /p/ P /pl/ involve changes in both place and manner features, and are therefore placed at the top of the hierarchy.

6. Voiced lingual plosives alternate with fricatives, e.g. /d/ P /z/ and /g/ P /z/. Since fricatives are more dierent from plosives than aricates are, voiced plosives arguably display more complex alternations than voiceless plosives. However, for the purpose of the present study, it is not necessary to distinguish between the softening alternations of voiced and voiceless plosives. 7. I use palatalization in a wide sense to denote alternations with a palatal, palatalized or post-alveolar consonant. Lenition (weakening) may be dened as a process, which increases the permeability of the vocal tract to airow (Lass 1984: 177). Notice that I consider arication an example of lenition. For evidence, see Bybee (2001: 80).


Tore Nesset

Table 2. Degrees of complexity Complexity 3. 2. 1. 0. Softening type substitutive softening (plosives/labials) substitutive softening (fricatives) plain softening no softening alternation Feature change primary place/manner primary place secondary place none Examples t P c, p P pl sPs s P s jPj

The hierarchy in Table 2 enables us to evaluate the implications of the relativized simplication hypothesis in (5), which correlates sux shift and a high degree of complexity. This means that we would expect sux shift to take place at the top end of the hierarchy, but not at the bottom. This prediction is indeed borne out by the facts. Sux shift is attested for verbs with complexity degree 3, but blocked for fricative-nal roots (complexity degree 2), for monosyllabic roots (complexity degree 1), and for /j/-nal roots with complexity degree 0 (no alternation at all). How does the relativized simplication hypothesis relate to prototypes? Since the majority of Russian verbs do not display consonant alternations, the absence of such alternations is prototypical. In terms of the relativized simplication hypothesis, the less complex the alternation, the more prototypical the verb is. Viewed from this angle, the relativized simplication hypothesis is a special case of the general hypothesis that peripheral subcategories are prone to analogical leveling, while prototypes tend to resist change. To the extent the data surveyed in this section support the relativized simplication hypothesis, they lend support to prototype theory.

3. Blocking and Target Consistency Since sux shift transforms /a/-verbs into /aj/-verbs, we may call /a/-verbs the source category and /aj/-verbs the target of sux shift. The previous section illustrated the importance of the target category insofar as sux shift occurs whenever the target category involves a simplication compared to the source category. In this section, we turn to an additional hypothesis which also focuses on the target category:

Prototypes and blocking of language change in Russian verbs


(6) Target consistency hypothesis: Sux shift occurs only if it is consistent with the target category. The idea in (6) is that sux shift only takes place as long as it creates verbs that t into the target category. In other words, sux shift is blocked whenever it would produce verbs of a type that is not attested in the /aj/-verb category. Let us consider non-syllabic roots rst. As mentioned in section 1, /aj/verbs represent a productive pattern that literally comprises thousands of verbs. However, the /aj/-verbs have syllabic roots. For non-syllabic roots, sux shift is therefore not target consistent. In other words, the hypothesis in (6) correctly predicts blocking of sux shift for non-syllabic roots. What about /j/-nal roots and target consistency? Although the /aj/ category contains a large number of verbs, verbs with root-nal /j/ are very few. Zaliznjak (1977) lists only the following: (7) /aj/-verbs with root-nal /j/: obajat fascinate, vajat sculpture, pajat solder, zijat gape, vlijat inuence, sijat shine, obujat seize Notice that for all these verbs the stress is on the verbal sux. The combination of /j/-nal root and root stress is not attested in the target category. In the source category, on the other hand, 17 out of 20 verbs have root stress: (8) /a/-verbs with root-nal /j/: bajat talk, blejat bleat, vejat winnow, grajat call (of crows), zatejat undertake, kajatsja repent, lajat bark, lelejat coddle, majatsja toil, nadejatsja hope, rejat soar, sejat sow, smejatsja laugh, tajat melt, xajat play down, cajat think, cujat scent, vozlijat pour, make libations, vopijat cry out, dejatsja happen. For verbs with root stress, sux shift would create an unattested pattern, viz. /aj/-verbs with /j/-nal root and stress on the root. In other words, for root-stressed verbs sux shift is target inconsistent. The hypothesis in (6) correctly predicts blocking of sux shift for stressed /j/-nal roots, but leaves the three verbs smejatsja, vozlijat and vopijat unaccounted for. As it stands, the hypothesis in (6) gives correct predictions for nonsyllabic roots and is partially successful in explaining the blocking for /j/nal roots. However, target consistency does not help us explain why fricative-nal roots do not undergo sux shift. Fricative-nal roots are not very common among /aj/-verbs, but Zaliznjak (1977) lists 28 verbs,


Tore Nesset

so we cannot consider sux-shift to be inconsistent with the target. Is it possible to revise the hypothesis? Consider the following: (9) Relativized target consistency hypothesis: Sux shift is the more likely to occur when a root type has more strength in the target category than in the source category. Behind this somewhat cumbersome statement is a simple idea. Instead of only considering the target category, we compare the target and source categories. If a given root type is strong in the target category and weak in the source category, sux shift is likely to take place. If, on the other hand, a root type is weak in the target and strong in the source category, sux shift is likely to be blocked. I propose measuring strength as type frequency. In other words, classes that contain many verbs are strong, while classes with few members are weak. The relevance of type frequency for morphological categories is fairly uncontroversial (cf., e.g., Wurzel 1984, 1989; Bybee 1985, 2001). What about token frequency? It seems likely that verbs with very low token frequency contribute little to the strength of a category, since such verbs are not often activated in the speakers mental grammar. However, it is also well known that high token frequency correlates with rote learning. Verbs that are rote-learned as unanalyzed wholes are activated directly in the mental grammar, i.e. without invoking a category. In the following, I will not consider token frequency. As we will see, type frequency provides a suciently precise picture. Table 3 compares the strength of various root types in the source and target categories. The numbers were arrived at in the following way. I elicited all forms from the Russian National Corpus (http:// of all /a/-verbs listed in the Russian Academy Gram-

Table 3. Relative strength in source and target categories Root type Syllabic, nal velar Syllabic, nal labial Syllabic, nal dental plosive Syllabic, nal dental fricative Syllabic, nal /j/ Non-syllabic /aj/-verbs 195 53 92 28 6 1 /a/-verbs 18 5 29 12 20 13 /aj/ : /a/ 10.8 10.6 3.2 2,3 0,3 0,1

Prototypes and blocking of language change in Russian verbs


Figure 1. Relative strength ratio for dierent root types

mar (Svedova 1980: 653655). As mentioned in section 1, this resulted in a database of approximately 20,000 examples. For the purposes of table 3, verbs with less than 50% /aj/-forms in the database were classied as /a/-verbs. The numbers of /a/-verbs of various root types are given in the second column from the right. Verbs that showed 50% /aj/-forms or more in the database were classied as /aj/-verbs. The column for /aj/-verbs in Table 3 contains all these verbs plus all the /aj/-verbs listed in Zaliznjak (1977). The rightmost column represents the proportion target: source category, i.e. the type frequency for the target category divided by the type frequency for the source category. Let us call this the relative strength ratio. If the ratio is high, the prediction from hypothesis (9) is that sux shift is likely to occur. If, on the other hand, the ratio is low, blocking of sux shift is a likely outcome. The reader can compare the ratios for the root types that show blocking with those that display sux shift in Figure 1. Table 3 and Figure 1 show a fairly good correlation between sux shift and relative strength ratio. For syllabic roots in labials and velars, the ratio is high. We therefore correctly predict sux shift. For non-syllabic and /j/-nal roots, the ratio is very low, so here we predict blocking once again correctly. Fricative-nal roots have the third lowest ratio, so in sum the table oers support for the idea that relativized target consistency facilitates blocking of sux shift.


Tore Nesset

A closer look at syllabic roots in dental plosives and fricatives is required. A priori, one might think that a target category would start attracting new members if it is larger than the source category. However, the data in Table 3 suggest that the critical mass is closer to three times the size of the source category. To be sure, the numbers in Table 3 are so small that it would be premature to draw rm conclusions at this point, especially since my data material does not allow us to control for the impact of other factors such as simplication discussed in section 2. These qualications notwithstanding, my investigation of relative strength in the source and target categories yields a clear implication for future research. In order to occasion language change, target categories need to be about three times larger than the source category. Another observation deserves mention at this point. As can be seen from the table, the relative strength ratio for the fricative-nal roots is quite close to that of the roots in dental plosives (which show sux shift). At the same time, fricative-nal roots have a ratio that is much higher than that of the other root types that show blocking, i.e. /j/-nal and monosyllabic roots. A prediction from this is that sux shift is more likely to spread to fricative-nal roots than to /j/-nal and monosyllabic roots. Linguists of the future will have the opportunity to test this prediction. At this point, we can only conclude that sux shift in Russian verbs oers some evidence for the relevance of (relativized) target consistency for language change. The target consistency hypothesis furthermore yields implications for future research. Before we close the discussion of target consistency and relative strength, we must ask how these concepts relate to prototype theory. The strength of a subcategory can be considered a measure of this subcategorys degree of prototypicality; the stronger it is, the closer it is to the prototype. Relative strength reects the interplay between the prototypes of the source and target categories. If a verb class has a high relative strength ratio, this means that it is close to the prototype of the target category, but not so close to the prototype of the source category. As we have seen, this state of aairs correlates with sux shift. If, on the other hand, the relative strength ratio is low, this indicates proximity to the prototype in the source category, but a lesser degree of prototypicality in the target category. The data cited in this section indicate that this state of aairs yields blocking. In summary, this section shows that prototype eects do not always involve the relationship to one prototype. The case discussed above involves two prototypes, and it is certainly possible to imagine more complex cases involving the interaction of several prototypes.

Prototypes and blocking of language change in Russian verbs


4. Blocking and Product-Oriented Generalizations for Prototypes The notion of product-oriented generalization is related to target consistency insofar as the focus is on the target category. In order to explain what product-oriented generalizations are, it is useful to take a traditional linguistic rule of the form A ! B as our point of departure. This rule picks out an input (the source) with the property A and changes A to B. Rules of this sort capture source-oriented generalizations. Product-oriented generalizations, on the other hand, characterize the output, i.e. B, without specifying whether it has been created from A or not. Bybee and Slobin (1982) show that product-oriented generalizations are relevant for language change. English has a class of strong verbs of the type spin spun and string strung, which has attracted quite a few new members since Old English. A case in point is hang which has developed the innovative past tense form hung. This verb was clearly not attracted to the strong verbs on the basis of the present tense form (the source), because hang has a dierent vowel in the present tense than spin and string. Instead, Bybee and Slobin (1982) argue that speakers entertain a product-oriented generalization about prototypical past tense forms specifying that the relevant class of strong verbs have a certain vowel followed by a velar and/or nasal consonant. It is possible to formulate product-oriented generalizations for Russian present tense forms, which are the target category in the context of the present study: (10) Russian present tense forms have the structure /. . .Vj V. . ./. The capital V represents any vowel and the morpheme boundary between stem and inectional ending, so the product-oriented generalization in (10) says that the forms in question have a vowel-j sequence that is followed by a vowel-initial ending. Although this schema does not cover all Russian verbs, it accounts for an impressive range of verbs including three of the ve productive classes. The statement in (10) therefore represents the prototypical present tense form of Russian verbs. Verbs with the productive /aj/, /ej/ and /ova/ suxes form present tense forms with the specications in (10). This is shown in Table 4 where the relevant string of segments is boldfaced. However, Table 4 furthermore shows that the product-oriented generalization in (10) covers /j/-nal roots with the /a/ sux as well. Verbs like lajat bark have present tense forms like /laj ut/, which display a vowel-j sequence followed by a vowel-initial inectional ending. This fact


Tore Nesset

Table 4. Verb classes with /. . .Vj V. . ./ in the present tense Sux /aj/ /ej/ /ova/ /a/ (/j/-nal root) Innitive /del-a t/ /krasn-e t/ /obraz-ova-t/ /laj-a t/ 3 pl present tense /del-aj ut/ /krasn-ej ut/ /obraz-uj ut/ /laj ut/ Gloss do redden form bark

may help us explain why sux shift is blocked for /j/-nal roots. Consider the following hypothesis: (11) Product-oriented hypothesis: Sux shift occurs only if a productoriented generalization about the target category is not already satised. This is a version of the general principle If it aint broke, dont x it. In other words, dont add a /j/ if you already have one. Since, as we have seen, /j/-nal roots like lajat already have a vowel plus /j/ before a vowel-initial ending, there is no need to undergo sux shift. The hypothesis in (11) predicts that sux shift will be blocked for /j/-nal roots a prediction that is borne out by the facts. At this point we must ask whether the hypothesis in (11) is relevant for non-syllabic and fricative-nal roots too. Attested present tense forms like berut (they) take and lgut (they) lie from non-syllabic brat and lgat do not conform to the product-oriented generalization in (10), so there is nothing to prevent sux shift from taking place for non-syllabic roots. In the same way, present tense forms like mazut (they) smear of fricativenal mazat are at variance with the prototype in (10). I hasten to add that this does not indicate that we must reject the hypothesis in (11). It only means that in the case of monosyllabic and fricative-nal roots sux shift must be blocked for other reasons than product-oriented generalizations. The hypothesis in (11) has implications for linguistic theory insofar as not all theoretical frameworks accommodate product-oriented generalizations. A classical example is the SPE model of Chomsky and Halle (1968), which employs rewrite rules of the A ! B type. As pointed out above, such rules capture source-oriented generalizations. They pick out a source and characterize procedures that apply to this source.

Prototypes and blocking of language change in Russian verbs


Figure 2. Category network for present tense forms

A framework that facilitates a principled account of product-oriented generalizations is Optimality Theory (Prince and Smolensky 2004). Markedness constraints like Onset, i.e. the requirement that syllables begin with a consonant, represent wellformedness conditions on surface structures. Constraints of this type characterize an output (a product) without specifying how it can be formed from a source. Another theory that provides a straightforward account of productoriented generalizations is Cognitive Linguistics. Consider Figure 1, which is based on the variety of cognitive linguistics that Langacker (1987, 1991) refers to as Cognitive Grammar. The nodes in the network are schemas, which represent product-oriented generalizations about the present tense forms of Russian verbs. The topmost schema captures the generalization in (10) and therefore represents the prototypical present tense form for Russian verbs. The schemas further down are fully compatible with the topmost schema, but are more specic in that they indicate which vowel precedes /j/. There are three schemas in the second row from the top, since the three vowels /a, e, u/ are attested in the relevant context. Notice that this information could not be incorporated in the topmost schema, since /a, e, u/ do not constitute a natural segment class. The gure enables us to consider an implication concerning /a/-verbs with /j/-nal roots. If the product-oriented generalization in (10) is relevant for verbs of this type, we would expect such verbs to display the same vowels before /j/ as other verbs, i.e. /a, e, u/. This prediction is borne out by the facts, as the reader can check for himself/herself by going back to (8). In the lower portion of Figure 1, I provide an example verb for each of the productive suxes


Tore Nesset

/aj/, /ej/ and /ova/ (delajut (they) do, krasnejut (they) redden and obrazujut (they) form), as well as three examples of /a/-verbs with /j/-nal roots and the same vowels before /j/, viz. lajut (they) bark, sejut (they) sow and cujut (they) scent.

5. Conclusion In this paper, we have seen that Russian /a/-verbs tend to undergo sux shift, but that sux shift is consistently blocked for /j/-nal, monosyllabic and fricative-nal roots. How can we account for blocking? We have considered three hypotheses: simplication, target consistency and productoriented generalizations for prototypes. These hypotheses are not mutually exclusive, but rather complement each other. If we adopt one of them, this does not force us to reject the others. For instance, we have seen that the simplication and product-oriented hypotheses account for /j/-nal roots, while the target consistency accommodates monosyllabic roots. In conclusion, we need all three hypotheses in order to explain the blocking of sux shift in Russian verbs. In order to account for the fricative-nal roots, we needed to invoke relativized versions of simplication and target consistency, i.e. consider degrees of simplication and target consistency. We have seen that with regard to both hypotheses the evidence for blocking is less strong for fricative-nal roots than for /j/-nal and monosyllabic roots. This result suggests that fricative-nal roots are in a more vulnerable position compared to /j/-nal and monosyllabic roots. If sux shift continues to spread to new verbs, the analysis presented in this paper predicts that fricativenal verbs will be more likely to undergo sux shift, and that this class of verbs will undergo sux shift before /j/-nal and monosyllabic roots. The present paper has implications beyond Russian and Slavic linguistics. First of all, it has been shown that the systematic study of blocking is a rewarding area of historical linguistics that allows us to formulate and test hypotheses about language change. Second, this study emphasizes the importance of the target category in language change insofar as all the hypotheses discussed in this paper refer to the target category. Third, the proposed analysis of target consistency suggests that a target category needs to be two to three times larger than the source category in order to attract members from the source category. This prediction deserves to be tested against larger sets of data where the impact of other factors can be controlled for. Finally, we have seen that all factors explored in this

Prototypes and blocking of language change in Russian verbs


study involve prototypes. Prototype theory helps us understand language change as well as the blocking of language change.

Andersen, Henning 1973 Abductive and deductive change. Language 49: 765793. Andersen, Henning 1980 Russian conjugation: acquisition and evolutive change. In: Elizabeth C. Traugott (ed.), Papers from the 4th International Conference on Historical Linguistics, 285301. Amsterdam/ Philadelphia: John Benjamins. Andersen, Henning 1995 Consonant reduction in Russian. In: Henrik Birnbaum and Michael S. Flier (eds.), The Language and Verse of Russia. In Honor of Dean S. Worth on his Sixty-Fifth Birthday, 1930. Moscow: Vostocnaja Literatura. Bybee, Joan 1985 Morphology. Amsterdam/Philadelphia: John Benjamins. Bybee, Joan 2001 Phonology and Language Use. Cambridge: Cambridge University Press. Bybee, Joan and Dan Slobin 1982 Rules and schemas in the development and use of the English past tense. Language 58: 265289. Chomsky, Noam and Morris Halle 1968 The Sound Pattern of English. New York: Harper and Row. Gagarina, Natalia 2003 The early verb development and demarcation of stages in three Russian-speaking children. In: Dagmar Bittner, Wolfgang U. Dressler and Marianne Kilani-Schoch (eds.), Development of Verb Inection in First Language Acquisition: A Cross-Linguistic Perspective, 131170. Berlin/New York: Mouton de Gruyter. Gor, Kira and Tatiana Chernigovskaya 2001 Rules in the processing of Russian verbal morphology. In: Gerhild Zybatow, Uwe Junghanns, Grit Melhorn and Luka Szucsich (eds.), Current Issues in Formal Slavic Linguistics, 528 536. Frankfurt am Main: Peter Lang. Gor, Kira and Tatiana Chernigovskaya 2003a Formal instruction and the acquisition of verbal morphology. In: Alex Housen and Michel Pierrard (eds.), Current Issues in Instructed Second Language Learning, 103136. Berlin/New York: Mouton de Gruyter.


Tore Nesset

Gor, Kira and Tatiana Chernigovskaya 2003b Mental lexicon structure in L1 and L2 acquisition: Russian evidence. GLOSSOS 4: 131. Gor, Kira and Tatiana Chernigovskaya 2003c Generation of Complex Verbal Morphology in First and Second Language Acquisition: Evidence from Russian. Nordlyd 31.6: 819833. Isacenko, Aleksandr V. 1982 Die Russische Sprache der Gegenwart. Formenlehre. Munich: Max HueberVerlag. Kiebzak-Mandera, Dorota, Magdalena Smoczynska and Ekaterina Protassova 1997 Acquisition of Russian verb morphology: the early stages. In: Wolfgang U. Dressler (ed.), Studies in Pre- and Protomorphology, 101114. Wien: Verlag der o sterreichischen Akademie der Wissenschaften. Krysin, Leonid Petrovic (ed.) 1974 Russkij Jazyk po Dannym Massovogo Obsledovanija. Moscow: Nauka. Langacker, Ronald W. 1987 Foundations of Cognitive Grammar. Vol. 1. Theoretical Prerequisites. Stanford: Stanford University Press. Langacker, Ronald W. 1991 Foundations of Cognitive Grammar. Vol. 2. Descriptive Application. Stanford: Stanford University Press. Lass, Roger 1984 Phonology. Cambridge: Cambridge University Press. Lass, Roger 1997 Historical Linguistics and Language Change. Cambridge: Cambridge University Press. McMahon, April 1994 Understanding Language Change. Cambridge: Cambridge University Press. Manczak, Witold 1963 Tendences generales du developpement morphologique. Lingua 12: 1938. Nesset, Tore 2000 Iconicity and prototypes: a new perspective on Russian verbs of motion. Scando-Slavica 46: 105119. Prince, Alan and Paul Smolensky 2004 [1993] Optimality Theory. Constraint Interaction in Generative Grammar. Oxford: Blackwell. Svedova, Natalija Julevna (ed.) 1980 Russkaja Grammatika 1. Moscow: Nauka. Timberlake, Alan 2004 A Reference Grammar of Russian. Cambridge: Cambridge University Press.

Prototypes and blocking of language change in Russian verbs


Tkachenko, Elena and Tatiana Chernigovskaya 2006 Focus on Form in the Acquisition of Inectional Morphology by L2 Learners: Evidence from Norwegian and Russian. Paper presented at The Second Biennial Conference on Cognitive Science, St. Petersburg, June 913, 2006. Wurzel, Wolfgang Ulrich 1984 Flexionsmorphologie und Naturlichkeit. Berlin: Akademie-Verlag. Wurzel, Wolfgang Ulrich 1989 Inectional Morphology and Naturalness. Dordrecht, Boston and London: Kluwer Academic Publishers. Zaliznjak, Andrej A. 1977 Grammaticeskij Slovar Russkogo Jazyka. Moscow: Russkij jazyk.

A prototype-based taxonomy of idiomatic expressions1 Esa Penttila

1. Introduction Idiomaticity is a complex, multifaceted and ubiquitous phenomenon, and expressions that can be regarded as idiomatic carry numerous properties, which are not equally weighed and which cluster in various ways. Some of the variation that is possible in idioms can be seen in (1). (1) a. b. The ordinary working-man and his kith and kin, unless he was in trouble, he didnt want to know you. (BNC2 B24 276) He wanted to conde in OHara, to get him on their side, but he didnt want every Tom, Dick and Harry knowing their business. (BNC2 FNU 2718) Not familiar with the maxim that pride goeth before a fall, Feargal? (BNC2 HGY 3472) How do you do, Zambia said silkily. (BNC2 AD9 1592) This grammar took advantage of the limited vocabulary (127 words) and constructions permitted in the domain. (BNC2 HGR 548) Its a doddle you said well sell up, move down here and do a Peter Mayle you said! (BNC2 BNP 1317) Therefore, in the bleak aftermath of war, he lived a hand-tomouth existence in the less attractive areas of London. (BNC2 CDE 2279) 2

c. d. e.

f. g.

The binomial noun phrase kith and kin (1a) contains archaic isolates rarely used outside idiomatic contexts. The trinomial noun phrase every Tom, Dick and Harry (1b) consists of a determiner and three coordinated proper
1. I would like to thank Elzbieta Tabakowska and two anonymous reviewers for their valuable comments on earlier versions of this paper. The aws that still remain are all mine. 2. The examples are taken from the second version of the British National Corpus, BNC2, version 1.00. The idioms in the examples are marked in bold.


Esa Penttila

names. Pride goeth before a fall (1c) is a familiar proverb and a quotation from the Bible. How do you do? (1d) is a pragmatically restricted formula used for greetings. Take advantage of (1e) is a verb phrase, which allows variation in tense and aspect; do a Peter Mayle (1f ) is also a verb phrase, but a special one, since it instantiates a syntactic pattern that allows variation in its object element. Hand-to-mouth (1g) is a restricted adjectival expression and may even be regarded as a complex lexeme. Several of the phrases are also guratively motivated. These are just some of the attributes that can be related to idioms, but they demonstrate how versatile the phenomenon is. Because of this multiplicity, idioms can be classied in various ways, and idiom researchers have often created their own special taxonomies (see, e.g., Smith 1925; McMordie 1954; Healey 1968; Makkai 1972; Rose 1978; Fernando 1996; Moon 1998). Many of these taxonomies only concern a certain subgroup of idioms or, if they deal with the whole class of idiomatic expressions, appear somewhat random as regards the criteria they are based on. Moreover, the classications are usually based on principles of classical categorization; in other words, they are based on necessary and sucient features which are binary and which produce classes with clear boundaries, where each category member has a more or less equal status (see, e.g., Taylor 2003: 21).3 As a consequence, the resulting taxonomies consist of categories, each of which contains dierent types of expression, but the degree of idiomaticity in each category is the same, and there is no clear indication of how or whether the categories relate to each other. This does not seem to do justice to the phenomenon. For some time already, it has been acknowledged that idiomaticity involves degrees and multiple characteristics which are not necessarily interdependent (see, e.g., Fernando and Flavell 1981; Akimoto 1983; Wood 1986; Nunberg, Sag and Wasow 1994). With the introduction of large corpora, this multiplicity has become even more evident (see, e.g., Jackendo 1997b; Moon

3. Although these principles are often referred to as Aristotelian, strictly speaking they are not such. Aristotle (1984: 1011) does not regard the members of all categories as equally related to one another. For example, of relatives, which is a category of things that are viewed as part of something else or with respect to something else (e.g. a large mountain is large because it is compared to something smaller than itself ), he says that they seem to admit of a more or less. Other categories with degrees include quality and doing and being-aected (Aristotle 1984: 1718).

A prototype-based taxonomy of idiomatic expressions


1998; Stubbs 2001). This would suggest that it would be at least natural, if not necessary, to provide a classication based on prototype theory. This type of classication would dier from classical categorization in a few main respects: rst, it is not based on a single set of necessary and sucient criteria, but on combinations of attributes that produce categories that are structured according to family resemblances; moreover, it does not regard all category members as equally representative but views membership in a category as having dierent degrees; and, nally, it views categories as fuzzy and blurred at the edges. It would also be able to account for the constant change and development of the category of idiomatic expressions, since the storage of formulaic expressions is an ever-evolving entity as, as Wray (2002: 101) points out. In a way, this type of approach to categorization would regard it as natural, or even necessary, to approach classication from a usage-based perspective that is so crucial for the framework of cognitive linguistics. Although recent empirical ndings would suggest that a prototypeoriented approach would be fruitful for discussing the phenomenon, there has been surprisingly little research on idiomaticity carried out from a prototype-based perspective. Akimoto (1992, 1994), Taylor (1998) and Nenonen (2007) discuss the prototypical properties of idioms, but a detailed and systematic prototype-based classication of idiomatic expressions is still lacking (see also Glaser 1998). There is an attempt to this eect made by Barkema (1996), whose multidimensional model of idiom classication is closely related to the prototypebased approach, although he does not even once mention the word prototype in his text. His model classies idioms on the basis of four criteria (institutionalization, collocational restriction, idiomatic meaning and transformational deciency) and thus approaches the problem from various perspectives. The criteria are not necessary and sucient, but they are graded; the model makes explicit the relations between dierent idiom characteristics and shows some relationships within the internal structure of the category. It can be applied to each expression separately, which means that it is able to reveal dierences in the representativeness of the expressions in the category. However, Barkemas model does not take into account additional properties that, in an essential way, relate to dierent idioms and therefore deserve to be recognized in the classication, although some of them are mentioned in his text (see also Glaser 1998). The present study deals with some of Barkemas (1996) basic intuitions and develops them further.


Esa Penttila

2. Prototypical characteristics of idiomatic expressions A suitable starting-point for a prototype-based classication of idiomatic expressions is to dene the characteristics of a prototypical idiom. At this point, one has to keep in mind that there is a distinction between the notions of idiom and idiomatic expression (see, e.g., Fernando 1996; Stubbs 2001; Wray 2002). The latter is a wider concept covering the full range of expressions that cannot be regarded as idioms in the strict sense but are idiomatic to some extent; they, for example, follow Pawley and Syders (1983: 191) idiom choice principle. Since this paper mainly deals with the nature of idiomatic language at large, the proposed taxonomy aims at covering the whole range of idiomatic expressions from lexicalized idioms to proverbs, quotations and idiomatic constructions. Previous studies provide various lists of typical idiom characteristics, the basic denition simply stating that an idiom is an expression whose meaning cannot be deduced from the meanings of its parts (see, e.g., Hockett 1958; Katz and Postal 1963; Weinreich 1969; Wood 1986). However, this seems a much too limited denition for the phenomenon and therefore various other criteria have been added to the denition. Nunberg, Sag and Wasow (1994: 492493) list six basic idiom properties: conventionality, inexibility, guration, proverbiality, informality, and aect. Of these, conventionality 4 is the only one that applies to all idioms obligatorily. Fernando (1996: 3), on the other hand, lists three basic properties: compositeness,5 institutionalization, and semantic opacity. Barkemas (1996) denition, as mentioned previously, contains four characteristics: institutionalization, collocational restriction, idiomatic meaning and transformational deciency. As this brief overview indicates, there is no generally accepted list of attributes that are commonly related to idioms, except for the basic idea that idioms are somehow semantically and syntactically special. Because of this, I have provided my own denition based on empirical data and the discussions in research literature (see Penttila 2006). The denition contains four central criteria (based on syntactic, semantic, pragmatic and sociolinguistic factors) that essentially relate to idioms and an additional meta-condition, which emphasizes the graded nature of the phe4. By conventionality Nunberg, Sag and Wasow (1994: 492) refer to the noncompositional meaning of idiomatic expressions, which cannot be predicted, although one would know the meanings of constituent parts in isolation from one another. 5. Compositeness in this context means multiwordiness.

A prototype-based taxonomy of idiomatic expressions


nomenon and implies that a prototype-based approach is applicable to its description. The denition reads as follows: An idiom is an (I) (i) institutionalized, (ii) noncompositional, (iii) syntactically restricted, (iv) multiword expression.

(II) Moreover, idioms and non-idioms form a continuum in which idiomaticity is a matter of degree rather than a dichotomous notion. This notion of scalarity concerns both the entire class and each idiomaticity criterion separately. The basic criteria in this denition more or less coincide with the basic idiom properties discussed by Barkema (1996). Institutionalization refers to the conventional status that an expression must have in a language community in order to be considered an idiom. According to Bauer (1983: 48), this happens when a word string becomes accepted and recognized as a known lexical item. An idiom is often, although not necessarily, codied and occurs frequently enough so that it is known to an average native speaker. Institutionalization also helps to distinguish idioms from novel metaphors (see, e.g., Lattey 1986: 219). Noncompositionality refers to the fact that the meaning of an idiom is not a compositional function of the transparent 6 meanings of its constituents; it can be either total or partial. With kick the bucket, for example, noncompositionality is total, since none of the components carry their transparent meanings, but with lose ones cool, it is partial, since the verb lose in the idiom can be regarded as carrying the same meaning that it has in nonidiomatic contexts. Here it should be pointed out that, although idioms can be regarded as noncompositional in principle, psycholinguistic studies have shown that the meanings of idiom parts still play a signicant role in their interpretation (see, e.g., Cacciari and Tabossi 1993; Gibbs 1994; Everaert et al. 1995). Syntactic restrictedness refers to the fact that idioms are not syntactically as exible as nonidiomatic word combinations (see, e.g., Fraser 1970; Newmeyer 1974; Akimoto 1983; OGrady 1998; Nenonen 2002). However, this restrictedness is not as strict as is usually claimed (see, e.g., Akimoto 1983; Moon 1998; Nenonen 2002; Fellbaum 2007). An example of the possible exibility can be seen in (2).
6. Here, transparent refers to the basic meaning that an idiom component has in other, nonidiomatic, contexts.


Esa Penttila

(2) Most popes also sought to control the senate by making the key ocials their liegemen and placing reliable men (often their relatives but sometimes kith rather than kin) in key positions. (BNC2 HPW 279) Multiwordiness is a criterion generally attached to idioms; because of their compositeness, they need to consist of more than one meaning-bearing element. Of course a multimorphemic expression would also count as composite; for example, Katz and Postal (1963) and Weinreich (1969) dene idioms as multimorphemic. In the present discussion, multiwordiness is preferred because it helps to distinguish idioms from compounds. Meta-condition (II) denes the nature of each basic criterion of idiomaticity and combines the denition with a prototype-based approach, in which no clear divisions within and without dierent classes and subclasses are made. In consequence, since idiom attributes vary independently of one another, the category of idioms is a multidimensional phraseological space with several areas of high degrees of idiomaticity (prototypical idiom peaks), which cannot be predicted. This denition gives the description of the prototypical idiom and provides a working hypothesis for idiomaticity in general. Expressions that completely fulll the basic criteria of idiomaticity, such as kick the bucket or by and large, can justiably be regarded as prototypical idioms, while expressions that fulll the criteria only partly or fulll only some of them are less prototypical, although they may still be idiomatic. For example, the formulas speaking (used when answering the phone) or cheers (used for toasting) are highly idiomatic and may even be regarded as idioms, although they are not multi-word (see, e.g., Warren 2005: 39). One of the main points of the meta-condition is to emphasize the continuum-like nature of idiomaticity; there are no strict boundaries between dierent types of idiomatic expressions. Although there are only four basic criteria in the denition, the attributes related to idioms are not restricted to these properties. On the contrary, there are various other properties which characterize idioms. They include, for example, pragmatic constraints, idiosyncratic syntax, polysemy, semantic constraints, stylistic constraints, guration, and hypothesized origin. Idioms may also contain isolates, or cranberry forms, which do not occur elsewhere in the language, such as fro in to and fro or shrift in short shrift. Moreover, idiomatic proverbs, in particular, often involve advisory quality or a moral, and aim to educate the audience. Most of such additional properties, however, are fairly sporadic and by no means central to the class. As a consequence, they do not characterize every idiom and are

A prototype-based taxonomy of idiomatic expressions


therefore not regarded as essential for the general denition. However, they do contribute to the rich texture that forms the class of idiomatic expressions by adding further layers to the taxonomy and some of them are so crucial for certain subclasses of idiomatic expressions that they deserve to be included in their dening properties. 3. Taxonomy: some examples Since the characteristics of prototypical idioms, which score high on the four basic criteria of idiomaticity, have been discussed above, the discussion in this section will be directed at some of the less prototypical subcategories in the taxonomy of idiomatic expressions. Although the discussion is not a detailed analysis of various types of idiomatic expressions, it should give an idea of how a prototype-based approach helps in classifying them into subclasses. 3.1. Idiomatic formulas There is a particular class of expressions, called idiomatic formulas, which resemble idioms but are usually not regarded as such. They are preconstructed multi-word phrases which follow the so-called idiom principle in that they are restricted in their choice of words but can still be analyzed into segments on the basis of the so-called open choice principle (see Sinclair 1991: 109115). Expressions in this class include, for example, certain irreversible binomials,7 most of the sememic idioms discussed by Makkai (1972), many of the recurrent word-combinations discussed by Altenberg (1998) and prefabs discussed by Erman and Warren (2000). An example of such an expression is How do you do? in (1d), and a few further examples are given in (3). (3) a. Few witnesses in court refuse to tell the truth, the whole truth and nothing but the truth on the grounds that they dont know what the truth is or how to tell it. (BNC2 FBD 586) If youre determined to be so pigheadedly stubborn and blind to the truth, so be it. (BNC2 JY5 1882) Well, well, well, this is long time no see, soul! (BNC2 AD9 3262) And as a matter of fact, she isnt Harrys sister. (BNC2 FS1 1686)

b. c. d.

7. For a thorough study on English binomials, see Gustafsson (1975).


Esa Penttila

Because of their formulaic nature, such expressions can be referred to as idiomatic formulas. They fulll three of the basic idiomaticity criteria. First of all, they are institutionalized expressions known to every native speaker of English. Syntactically, they are restricted, although often not totally frozen, and of course they are multiword. However, semantically these expressions are usually compositional; because of this, they cause no problems to understanding (i.e. they are idioms of encoding, according to Makkais (1972) denition). This makes them slightly less idiomatic than prototypical idioms. In addition to their basic idiom attributes, idiomatic formulas may also carry additional properties. The most obvious of these extra features is the fact that they are often pragmatically constrained. This could even be regarded as one of the central properties of idiomatic formulas, and thus something that needs to be part of their denition. For example, the truth, the whole truth and nothing but the truth (3a) is restricted to, or at least has connotations with, legal contexts. Some of the formulas are ungrammatical, such as long time no see (3c), but this clearly seems to be a more peripheral characteristic, which is not involved in idiomatic formulas to any larger extent than it is in other idiomatic expressions. 3.2. Quotations and proverbs Example (1c) contains a phrase which is often regarded as idiomatic, although not necessarily an idiom in the strict sense of the word: pride goeth before a fall. Further examples of this type are given in (4). (4) a. Thus his conduct very well illustrates the maxim Look before you leap; and if his nal decision was made when the trend of circumstances was making it harder to leave home, he might well have recalled He who hesitates is lost. (BNC2 CB1 818) That Shakespeare fellow was right when he said: All the worlds a stage And all the men and women merely players. (BNC2 BN3 876) This is how careers end, Adjudicator: not with a bang, but with a whimper. (BNC2 G1M 563)



Pride goeth before a fall is at the same time both a quotation and a proverb, but as the sentences in (4) indicate, these two types do not necessarily coincide. However, in this classication quotations and proverbs are included in the same subclass because of their similar characteristics; they both usually consist of full sentences and they both also rate similarly with

A prototype-based taxonomy of idiomatic expressions


respect to the four basic criteria of idiomaticity. In terms of institutionalization, syntactic restrictedness, and multi-wordiness, quotations and proverbs are highly idiomatic, whereas in terms of semantics they are not very idiomatic but rather compositional and fairly transparent although one has to admit that the guration commonly involved in them often adds an aspect of ambiguity to their meaning. When compared with idiomatic formulas, quotations and proverbs appear no dierent, at least in terms of the basic criteria of idiomaticity. However, when additional properties are taken into consideration, dierences appear. Both quotations and proverbs are usually based on guration and consist of a full sentence. This is the case with all examples in (4) as well. These two properties are so important that they seem to be crucial for the denition of the class and distinguish it from, for example, idiomatic formulas. Quotations and proverbs also carry some additional attributes which apply to them in various ways. Proverbs, in particular, carry a moral. For example, look before you leap warns the recipient to avoid unnecessary danger and begs for reconsideration of the risks. Quotations, on the other hand, can usually be traced back to a hypothesized origin. For example, All the worlds a stage And all the men and women merely players comes from Shakespeares play As you like it and not with a bang, but with a whimper comes from T. S. Eliots poem The Hollow Men. Such additional properties show that, although quotations and proverbs can be regarded as a single category within idiomatic language, they each have certain special properties which help to extend this category, on the one hand, towards more quotation-like expressions and, on the other, towards more proverb-like expressions. As a consequence, it is possible to represent the combined class of proverbs and quotations as in Figure 1. 3.3. Idiomatic constructions Idiomatic expressions which have recently received attention in research literature since the seminal paper by Fillmore, Kay and OConnor (1988) are the so-called idiomatic constructions, which employ a syntactic pattern that is partly lexically determined and partly lexically open. Do a Peter Mayle in (1g) illustrates one such construction, and the sentences in (5) illustrate a few more, which are discussed in the literature (see Clark and Gerrig 1983; Michaelis and Lambrecht 1996; Jackendo 1997b; Kay and Fillmore 1999; Penttila 2006).


Esa Penttila

Figure 1. The category of quotations and proverbs with sample sentences. Although there are a number of expressions of either proverbial or quotational character, there are also various expressions which can be classied as both

(5) a.




but what I really wanted to ask you, said Patrick OSullivan, putting his hands in his jacket pockets, is what is a woman like you doing in a place like this? (BNC2 CMJ 2260) Then, unerringly, he found his way to Modigliani, wherever he was, and the two of them would drink the night away until they were unconscious, taken to the police station and beaten up. (BNC2 ANF 1045) Ive never seen the sort of breathtaking squalor I witnessed there but give them a day when their neighbourhood is shown to be worth the world running through it, gratefully accepting water and candy from them at the roadside, and its amazing the way mutual mistrust can disappear. (BNC2 K97 18228) He didnt think that the US would do a Suez in the Middle East. The more likely prospect was our declining inuence. (BNC2 B0H 1033)

Such constructions resemble prototypical idioms in many respects. They are institutionalized to the extent that some instances can even be found in dictionaries, for example, the more the merrier in Ammers (1997) The American Heritage Dictionary of Idioms, and do a Bertie and do a Garbo in Partridges (1984) A Dictionary of Slang and Unconventional English. Their meaning is noncompositional in the same way as the meaning of prototypical idioms. Syntactically, they are restricted and, since they con-

A prototype-based taxonomy of idiomatic expressions


sist of complex syntactic wholes, they are multiword. However, in addition to these attributes, there is an additional attribute which is essential to idiomatic constructions and which separates them from pure prototypical idioms; as opposed to prototypical idioms, idiomatic constructions are always lexically partially open. For example, do a Suez realizes the pattern do indenite article a proper noun. In this pattern the position of the proper noun can be taken by any proper noun whatsoever (see Penttila 2006). Some of this lexical variation can be seen in the following phrases. (6) a. A Science Ministry spokesman said: We are not doing a Dianagate there is no question of anybody listening in to conversations. It is good housekeeping. We want to see who is making most calls and how costs can be cut. (BNC2 CH6 5586)8 Before you think Oh, its only Tarrant trying to do an Orson Welles War Of The Worlds on us, I have to tell you its true. (BNC2 CH2 2042) I hope nobody does an Argentina and number their players alphabetically. (BNC2 J1E 506) May the scarf unravel and do an Isadora Duncan on the wretch. (BNC2 G0A 2323)


c. d.

So, although idiomatic constructions resemble prototypical idioms with respect to prototypical idiom attributes, their partial lexical openness can be regarded as an additional crucial property, which justies their inclusion into a separate subcategory. After all, lexical openness is something that is not available to prototypical lexicalized idioms, at least not to the same extent, although recent corpus-based studies have shown that also prototypical idioms allow much more structural and lexical variation than has traditionally been acknowledged (see, e.g., Fellbaum 2007). In addition to the crucial extra feature, idiomatic constructions may also carry some further attributes, but these are more peripheral and can also be found in prototypical idioms, so they cannot be regarded as distinctive features between these two categories. Such extra attributes include guration (e.g. the do an Nprop construction is gurative), ungrammaticality (e.g. the so-called Mad Magazine construction illustrated by
8. The phrase do a Dianagate is another do a(n) Nprop phrase in the BNC, in addition to do a Watergate, which productively mixes the eponymous do a(n) Nprop construction and the -gate ending deriving from the Watergate scandal.


Esa Penttila

Him, a doctor? is clearly ungrammatical), and even a hypothesized origin (e.g. the above-mentioned Mad Magazine construction).

4. Conclusion Although the discussion above only deals with some subclasses within the whole range of idiomatic expressions, it shows how a prototype-based approach helps us view the vague and multifarious class of idiomatic expressions and create a taxonomy that takes into account the whole multiplicity and the true nature of the phenomenon. At the same time, it helps us to become aware of the varying relationships between the dierent subclasses and to weigh the dierent attributes within the subclasses (see, e.g., Geeraerts 1989). A very general illustration of how this works is given in Table 1.

Table 1. Main classes of idiomatic expressions and the attributes related to them.
Prototypical idiom attributes Institutionalization Noncompositionality Syntactic restrictedness Multiwordiness Additional idiom attributes

Prototypical idioms





i) ii) iii) iv) v) vi)

guration ungrammaticality pragmatic constraints polysemy isolate elements hypothesized origin

Idiomatic formulas Quotations and proverbs





i) pragmatic constraints ii) ungrammaticality i) ii) iii) iv) i) ii) iii) iv) guration sentence-length hypothesized origin moral partial lexical openness guration ungrammaticality hypothesized origin





Idiomatic constructions





A prototype-based taxonomy of idiomatic expressions


The four basic criteria of idiomaticity are listed as individual properties in Table 1, and they are each given a value of high or low depending on how well a particular subclass of idiomatic expressions fullls that criterion. Additional idiom properties related to each subclass are listed separately on the right. Properties crucial for the particular subclass are marked in bold, the others contain no special marking and are not found in all, or even in most, of the representatives of that category. Consequently, Table 1 shows that there are two types of attributes which are essential for each subclass: basic idiom attributes, with which the class scores high, and crucial additional attributes, with respect to which they also score high; all the other attributes are more peripheral and not necessary for the denition of the subclass. It would be easy enough to add further subclasses to this classication. The advantages of this type of multifaceted, prototype-based approach to idiom classication are obvious. It allows us to see the true nature of idiomaticity. Instead of viewing the phenomenon in terms of categorial notions, it helps us see idiomaticity in terms of a continuum, or even several continua, where certain expressions are more or less prototypical of either the whole category or of a certain subcategory. We can also see how dierent categories overlap and merge into one another, thus reecting the internal scalarity of the phenomenon. As a consequence, the whole system of idiomatic expressions can be regarded as radially structured. We can also see more clearly the similarities and the dierences between the dierent subcategories of idiomatic expressions. In particular, close similarity between prototypical idioms and idiomatic constructions becomes obvious when these classes are analyzed according to these graded criteria. Both classes fulll the four basic criteria of idiomaticity, but they dier from one another in terms of their additional properties; prototypical lexical idioms do not carry any crucial additional properties at all, while it is a crucial property of idiomatic constructions that they are partially open lexically. However, with respect to the prototypical idioms attributes, both classes score high and therefore they can both be located at the idiomatic extreme of the general continuum of idiomaticity, while some of the other subclasses are clearly lower on that scale. The prototype-based approach also indicates that the area of language that can be regarded as non-idiomatic expressions is not necessarily as wide as has been thought traditionally; this in turn suggests that idiomaticity in our language could be a fairly ubiquitous phenomenon, which, in fact, has been indicated by various studies (see, e.g., Pawley and Syder 1983; Nattinger and DeCarrico 1992; Jackendo 1997a; Erman and Warren


Esa Penttila

2000). Moreover, the ubiquity of the phenomenon ties in with Langackers (1987) notion of entrenchment, suggesting that frequency of occurrence, with its obvious links to prototypicality, could also be of help in classifying idiomatic expressions. The similarity between prototypical idioms and idiomatic constructions provides a challenge for both dictionary writing and language pedagogy. As mentioned in section 3.3, some instances of idiomatic constructions are listed in dictionaries, but the actual constructions themselves are mainly neglected. This is understandable, since the partial lexical openness of idiomatic constructions makes them appear dierent from traditional idioms; however, since they obviously form a natural part of any native speakers idiomatic usage, they deserve to be noticed in the dictionaries as well. When the actual resemblance between prototypical idioms and idiomatic constructions is shown with the help of prototype-based analysis, it should encourage dictionary writers in the future to include idiomatic constructions in dictionaries, which would then cover the phenomenon of idiomaticity more extensively. At the same time as the role of idiomatic constructions is acknowledged in idiom dictionaries, it should be reconsidered in foreign language teaching as well. After all, idiomatic language is constantly gaining more importance in foreign language instruction, while the methods of teaching it are being studied and developed further (see, e.g., Kovecses and Szabo 1996; Schmitt and Carter 2004; Boers and Lindstroemberg 2007). When the general signicance of idiomatic language is recognized and the position of idiomatic constructions in the framework of language acknowledged, the role of idiomatic constructions in language education will certainly be enhanced as well.

Akimoto, Minoji 1983 Idiomaticity. Tokyo: Shinozaki Shorin. Akimoto, Minoji 1992 Prototypicality and idiomaticity. The Eighteenth LACUS Forum 1991: 235245. Akimoto, Minoji 1994 A typological approach to idiomaticity. The Twentieth LACUS Forum 1993: 459468. Altenberg, Bengt 1998 On the phraseology of spoken English: The evidence of recurrent

A prototype-based taxonomy of idiomatic expressions


word-combinations. In: Anthony Paul Cowie (ed.), Phraseology: Theory, Analysis, and Applications, 101122. Clarendon Press: Oxford. Ammer, Christine 1997 The American Heritage Dictionary of Idioms. Boston and New York: Houghton Miin. Aristotle 1984 The Complete Works of Aristotle: Volume One. The Revised Oxford Translation. Edited by Jonathan Barnes. Princeton: Princeton University Press. Barkema, Henk 1996 Idiomaticity and terminology: A multi-dimensional descriptive model. Studia Linguistica 50: 125160. Bauer, Laurie 1983 English Word-Formation. Cambridge: Cambridge University Press. Boers, Frank and Seth Lindsroemberg 2007 Cognitive linguistic applications in second or foreign language instruction: rationale, proposals, and evaluation. In: Gitte Kris tiansen, Michel Achard, Rene Dirven, and Francisco J. Ruiz de Mendoza Ibanez (eds.), Cognitive Linguistics: Current Applications and Future Perspectives, 305355. Berlin/New York: Mouton. Cacciari, Cristina and Sam Glucksberg 1991 Understanding idiomatic expressions: The contribution of word meanings. In: Greg B. Simpson (ed.), Understanding Word and Sentence, 217240. Amsterdam: North-Holland. Cacciari, Cristina and Patrizia Tabossi (eds.) 1993 Idioms: Processing, Structure, and Interpretation. Hillsdale: Lawrence Erlbaum. Clark, Herbert H. and Richard J. Gerrig 1983 Understanding old words with new meanings. Journal of Verbal Learning and Verbal Behavior 22: 591608. Erman, Britt and Beatrice Warren 2000 The idiom principle and the open choice principle. Text 20: 29 62. Evereart, Martin, Erik-Jan van der Linden, Andre Schenk and Rob Schreuder (eds.) 1995 Idioms: Structural and Psychological Perspectives. Hillsdale: Lawrence Erlbaum. Fellbaum, Christiane (ed.) 2007 Idioms and Collocations: Corpus-based Linguistic and Lexicographic Studies. London: Continuum. Fernando, Chitra 1996 Idioms and Idiomaticity. Oxford: Oxford University Press.


Esa Penttila

Fernando, Chitra and Roger Flavell 1981 On Idiom: Critical Views and Perspectives. Exeter: University of Exeter. Fillmore, Charles J., Paul Kay and Mary Catherine OConnor 1988 Regularity and idiomaticity in grammatical constructions: The case of let alone. Language 64: 501538. Fraser, Bruce 1970 Idioms within a transformational grammar. Foundations of Language 6: 2242. Geeraerts, Dirk 1989 Prospects and problems of prototype theory. Linguistics 27 (4): 587612. Reprinted in: Dirk Geerarts (ed.) (2006), Cognitive Linguistics: Basic Readings. Berlin/New York: Mouton. 141 165. Gibbs, Raymond W. Jr. 1994 The Poetics of Mind: Figurative Thought, Language, and Understanding. Cambridge: Cambridge University Press. Glaser, Rosemarie 1998 The stylistic potential of phraseological units in the light of genre analysis. In: Anthony Paul Cowie (ed.), Phraseology: Theory, Analysis, and Applications, 125143. Oxford: Clarendon Press. Gustafsson, Marita 1975 Binomial expressions in present-day English: a syntactic and semantic study. Annales Universitatis Turkuensis, Ser. B, Tom. 136. Turku: University of Turku. Healey, Alan 1968 English idioms. Kivung 1: 71108. Hockett, Charles F. 1958 A Course in Modern Linguistics. New York: MacMillan. Jackendo, Ray 1997a The Architecture of the Language Faculty. Cambridge, MA: The MIT Press. Jackendo, Ray 1997b Twistin the night away. Language 73: 534559. Katz, Jerrold J. and Paul M. Postal 1963 Semantic interpretation of idioms and sentences containing them. MIT Quarterly Progress Report 70: 275282. Kay, Paul and Charles J. Fillmore 1999 Grammatical constructions and linguistic generalizations: The Whats X doing Y? construction. Language 75: 133. Kovecses, Zoltan and Peter Szabo 1996 Idioms: A view from cognitive semantics. Applied Linguistics 70: 326355. Langacker, Ronald 1987 Foundations of Cognitive Grammar, Volume 1: Theoretical Prerequisites. Stanford: Stanford University Press.

A prototype-based taxonomy of idiomatic expressions Lattey, Elsa 1986


Pragmatic classication of idioms as an aid for the language learner. IRAL 24: 217233.

Makkai, Adam 1972 Idiom Structure in English. The Hague: Mouton. McMordie, William 1954 English Idioms and How to Use Them. Revised by Raymond C. Gon. Third edition. London: Oxford University Press. Michaelis, Laura A. and Knud Lambrecht 1996 Toward a construction-based theory of language function: The case of nominal extraposition. Language 71: 215247. Moon, Rosamund 1998 Fixed Expressions and Idioms in English: A Corpus-Based Approach. Oxford: Clarendon. Nattinger, James R. and Jeanette S. DeCarrico 1992 Lexical Phrases and language Teaching. Oxford: Oxford University Press. Nenonen, Marja 2002 Idiomit ja Leksikko: Lausekeidiomien Syntaktisia, Semanttisia ja Morfologisia Piirteita Suomen Kielessa [Idioms and the Lexicon: Syntactic, Semantic and Morphological Properties of Phrasal Idioms in Finnish]. Joensuu: Joensuu University Press. Nenonen, Marja 2007 Prototypical idioms: evidence from Finnish. SKY Journal of Linguistics, Vol. 20: 309330. Newmeyer, Frederick J. 1974 The regularity of idiom behavior. Lingua 34: 327342. Nunberg, Georey, Ivan A. Sag and Thomas Wasow 1994 Idioms. Language 70: 491538. OGrady, William 1998 The syntax of idioms. Natural Language and Linguistic Theory 16: 279312. Partridge, Eric 1984 A Dictionary of Slang and Unconventional English: Colloquialisms and Catch-phrases, Fossilised Jokes and Puns, General Nicknames, Vulgarisms and such Americanisms as have been naturalized. Edited by Paul Beale. Eight Edition. London. Routledge. Pawley, Andrew and Frances Hodgetts Syder 1983 Two puzzles for linguistic theory: Nativelike selection and nativelike uency. In: Jack C. Richards and Richard W. Schmidt (eds.), Language and Communication. London: Longman. 191226. Penttila, Esa 2006 It takes an age to do a Chomsky: Idiomaticity and verb phrase constructions in English. Unpublished PhD Thesis. University of Joensuu, Department of Foreign Languages.


Esa Penttila

Rose, James H. 1978 Types of idioms. Linguistics 203: 5562. Schmitt, Norbert and Ronald Carter 2004 Formulaic sequences in action: An introduction. In: Norbert Schmitt (ed.), Formulaic Sequences: Acquisition, Processing, and Use. Philadelphia: John Benjamins. 122. Sinclair, John 1991 Corpus, Concordance, Collocation. Oxford: Oxford University Press. Smith, Logan Pearsall 1925 Words and Idioms: Studies in the English Language. London: Constable & Company. Stubbs, Michael 2001 Words and Phrases: Corpus Studies of Lexical Semantics. Oxford: Blackwell. Taylor, John R. 1998 Syntactic constructions as prototype categories. In: Michael Tomasello (ed.), The New Psychology of Language, 177202. Mahwah, NJ: Lawrence Erlbaum. Taylor, John R. 2003 Linguistic Categorization. Third edition. Oxford: Oxford University Press. Warren, Beatrice 2005 A model for idiomaticity. Nordic Journal of English Studies 4: 3554. Weinreich, Uriel 1969 Problems in the analysis of idioms. In: Jaan Puhvel (ed.), Substance and Structure of Language: Lectures Delivered Before the Linguistic Institute of the Linguistic Society of America, University of California, Los Angeles, June 17August 12, 1966, 2381. Berkeley: University of California Press. Wood, Mary M. 1986 A Denition of Idiom. Bloomington: Indiana University Linguistics Club. Wray, Alison 2002 Formulaic Language and the Lexicon. Cambridge: Cambridge University Press.

Part III.

Contexts for Cognitive Grammar

Control and the mind/body duality: Knowing vs. eecting Ronald W. Langacker
Introduction Cognition is taken by cognitive linguists to be embodied (Johnson 1987; Lako 1987). While there are many ways of interpreting this notion (Hampe 2005), one common denominator is the rejection of the classic mind/body duality. In this respect cognitive linguistics is non-Cartesian (cf. Chomsky 1966; Lako and Johnson 1999). This is quite consistent, of course, with the mind/body dualism being part of the conceptual system that speakers employ in talking and thinking.1 On the face of it, a dualistic conception is perfectly justied. Mental activity is quite dierent from physical activity. Not only do they have a dierent locus (the head vs. the body as a whole), but mental eort is qualitatively very dierent from physical exertion (thus we speak of brain vs. brawn). They also dier in their eect on the outside world (thinking vs. doing). In and of itself, thinking has no eect: if we want a box to go on a shelf, we cannot just think about it being there; we actually have to get our body in gear and do something. On the other hand, the mind unencumbered by physical constraints can take us far more places than our body can. Ironically, then, the mind/body dualism has a rm experiential basis. It is a natural product of embodied cognition. The duality has many linguistic manifestations, the most obvious being lexical items pertaining to both facets. Some of these, in fact, form opposing pairs with the status of familiar collocations: mind vs. body, mental vs. physical, brain vs. brawn, thought vs. action, knowing vs. doing. Less obvious is the contrast discussed by Sweetser (1990: ch. 4) regarding the use of causal and adversative conjunctions. Thus in (1) the causation coded by because occurs at two dierent levels. In (1)(a), the chairs being poorly

1. Analogously, even physicists use nave physics in casual thought and even technical writing (Talmy 1988).


Ronald W. Langacker

made is physically responsible for its breaking. By contrast, in (1)(b) the chairs breaking is responsible for the mental act of concluding that it must have been poorly made. (1) (a) (b) The chair broke because it was poorly made. [eective relationship] The chair was poorly made, because it broke. [epistemic relationship]

I will describe the rst relationship as being eective (EF), and the second epistemic (EP). Precisely what I mean by these terms will become apparent (if it does at all) only in the course of the exposition. Basically it reects the unfortunate fact of life, mentioned previously, that thinking does not itself have any eect on the outside world. The labels thus allude to a basic experiential correlation of physical activity with the causation of events, on the one hand, and of mental activity with knowledge of events, on the other hand. More generally, I will use these labels for a systematic opposition apparent at multiple levels of structure, ranging from single predicates to complex sentences with nite-clause complements. Depending on the level, one of the two component oppositions either physical vs. mental or causation vs. knowledge may fail to be fully manifested. In particular, let me note a basic dierence between the lowest and the highest levels. Predicates describe events, which occur and can be caused as well as apprehended.2 By contrast, nite clauses represent propositions, which while they pertain to events do not themselves occur and cannot be caused; they can only be apprehended and assessed for validity. The actual form assumed by the eective/epistemic opposition is therefore malleable. However, the variation follows in a principled manner from the level of organization. When this is properly taken into account, the opposition emerges as a signicant cross-level parallelism.

2. For convenience, I will use the term event in a very general way, to encompass both events (in the narrow sense) and states or situations (i.e. both perfective and imperfective processes). Also, for the role of Event2 in Figure 1(b), I will normally presume the canonical case of physical occurrences. Obviously, though, the events we instigate or apprehend can be of any sort: concrete or abstract; physical, mental, or social; etc. As the ultimate basis of embodied cognition, physical events have a special status and reveal most clearly the rationale for certain aspects of language structure.

Control and the mind/body duality: Knowing vs. eecting


Figure 1.

In (1), the eective and epistemic relationships expressed by because hold between two potentially separate events, the respective proles of the clauses it connects. This is roughly sketched in Figure 1(a). The focus here will be on the slightly dierent conguration in diagram (b), where one event consists in an eective or epistemic relationship directed at the other. Proling varies with level of organization. Direct complements My basic order of presentation will be to work up from the smallest structures (single predicates) to the largest. However, I am going to make a brief detour, starting from sentences like those in (2), which have the advantage of showing most fully and explicitly the congurations in Figure 1(b). I refer to these as direct complement constructions because the matrix and complement predicates are directly adjacent, with no intervening complementizer like to or that. (2) (a) She {made/let/had} him close the window. [eective] (b) She {saw/heard/watched} him close the window. [epistemic] (c) *He was {made/let/had} close the window. [cf. It was made very crudely.] (d) *He was {seen/heard/watched} close the window. [cf. He was seen there.]

In (2)(a), the matrix verbs describe causation, which may be physical but in this case is more likely eected verbally through social force. In (2)(b),


Ronald W. Langacker

Figure 2.

the perceptual activity does nothing to eect the closing of the window, but it does allow the matrix subject to know about it. Despite pertaining to dierent realms of activity, these expressions exhibit a number of semantic and formal similarities. They are of course precisely analogous in terms of their overt form, including word order. Interestingly, they both resist the passive, even though verbs like make and see are transitive and readily passivize in simple sentences. One factor in their non-passivizability may be a common semantic property, indicated by the diagrams in Figure 2.3 Verbs like make and see are ambivalent as to whether the proled relationship is directed at Event2 the caused or perceived occurrence or at its trajector. In (2)(a), we can think of the eective force as being directed at the events realization, or else at the person who carries out the action. Likewise, in (2)(b) she perceives both the action and the person engaged in it. In the diagrams, therefore, both the action and its trajector are shown in bold to indicate their joint status as landmark of the matrix predicate. The absence of a single, clear-cut landmark may contribute to the infelicity of a passive.4 The parallelism of these two kinds of expressions is also observed in French, where an additional grammatical property makes it even more evident. As shown in (3)(b), the construction admits of a variant in which the matrix and complement predicates are adjacent (Langacker 1966). In a detailed analysis, Achard (1996) has argued that together they constitute a complex verb construed as designating a single event, with the following nominal as its object.
3. This is not a full explanation, since alternative expressions like the following would seem to represent the same conguration: He was {made/seen} to close the window; He was seen closing the window. 4. Note that I am streamlining diagrams by not including labels for trajector and landmark (primary and secondary focal participants). They can be identied ` by their position vis-a-vis the arrows representing relationships.

Control and the mind/body duality: Knowing vs. eecting


(3) (a) (b)

Elle a {laisse /entendu} les enfants pleurer. She {let/heard} the children cry. Elle a {laisse /entendu} pleurer les enfants. She {let/heard} the children cry.

Since causation and perception are very dierent, I have long wondered why these constructions are so similar. I am not sure I fully understand it yet, but at least I now know that it represents a special case of a systematic opposition between eective and epistemic relationships. Of course, we only speak of an opposition between linguistic elements when they are basically the same. We recognize hot and cold as opposites because they have so much in common (as compared, say, to hot and short).

Complex events The sentences in (3)(b) are intermediate between a complement construction and single-clause expressions. Though it consists morphologically of separate verbs, the composite clausal predicate proles a complex process construed as single event. The construction thus accomplishes by grammatical means what many lexical predicates achieve directly. The verb kill, for example, proles a complex occurrence involving both causation and the resultant process of dying. Examples like those in (4) have been used to demonstrate that with this lexical verb the causing and the dying constitute a single event, in contrast to an innitival complement construction, where cause to die involves a sequence of events (Fodor 1970; Wierzbicka 1975).5 (4) (a) By poisoning his tea on Monday, she caused him to die on Friday. (b) *By poisoning his tea on Monday, she killed him on Friday. A verb like kill is sketched in Figure 3(a). It proles what is conceived as a single event subsuming both causation and the process thereby induced. Predicates dier in the extent to which that process is capable of occurring
5. Observe that the direct complement construction with make is intermediate in this regard: ??By poisoning his tea on Monday, she made him die on Friday. As it stands, diagram Figure 2(a) does not capture this subtle contrast with innitival complements. Nor does Figure 3(a) capture the nuance distinguishing tuer kill from faire mourir make die.


Ronald W. Langacker

Figure 3.

independently or is recognized as a separate process type. In the case of kill the connection with dying is clearly evident despite the lexemes total lack of morphological analyzability. With many verbs, for instance drop, the process is clearly recognized because the same verb is used intransitively with a non-causative meaning. Thus diagrams (a) and (b) correspond respectively to (5)(a)(b). And with some predicates tickle, for example the portion corresponding to the inner box is not an independently established concept. (5) (a) (b) (c) (d) He inadvertently dropped his keys in the sewer. His keys dropped into the sewer. He carefully dropped his keys into the sewer. When he heard the shot, he dropped to the oor.

As shown in Figure 3(a), a transitive verb like kill or drop incorporates causation as part of the proled process. The complex event it designates ` includes an eective relationship vis-a-vis the simplex event represented by the inner box.6 In view of my general thesis, it is reasonable to ask whether there are any comparable predicates which instead incorporate an epistemic relationship, e.g. a verb analogous to kill except that, instead of meaning (roughly) cause to die, it means something like see die or know to die. While I would not rule this out, examples do not come to mind. Here is a place, evidently, where the analogy between eective and epistemic relationships breaks down. This is not particularly worrisome, for I am not claiming that they are analogous in every conceivable respect my goal is merely to explore the extent of their parallelism. In the spirit of exploration, suppose we assume that there should indeed be some manifestation at the lexical level of the eective/epistemic opposition. I have in mind something beyond the mere existence of both physical and mental predicates, e.g. die vs. fear or kill vs. frighten. The question is rather whether we can nd an epistemic counterpart for the causal component of the latter pair. What might it be? A plausible candidate, I suggest,

6. I refer to the latter as a thematic process (Langacker 1993).

Control and the mind/body duality: Knowing vs. eecting


is what is usually referred to as volition. But what does this actually mean? What do we imply by describing an action as volitional? Even when done on purpose, the act of killing someone is not considered epistemic in any narrow sense. That is, it is not an act of knowing or acquiring knowledge. It is however a knowing act if we do something under volitional control, we do it knowingly, and know that we are doing it. So in a broader sense, volition does have an epistemic component. To simplify a complex issue, we might characterize volitional actions as being under mental control. There are several facets to this notion. First, the actor has to be aware of what is happening (know what is going on). Second, the actor must intend for the action to occur. And third, this apprehension of the action must serve to instigate and guide its execution.7 I will use a dashed arrow to represent this notion of mental control. Thus the diagrams in Figure 3 correspond respectively to the matrix verbs in (5). The transitive drop can be construed as either inadvertent, as in diagram (a), or volitional, as in diagram (c).8 There is also a volitional intransitive sense, as in (d), which is not however obtained just by adding a dashed arrow to diagram (b). It further includes the physical causation inherent in the transitive senses, but is intransitive because the agent and the mover are conated in a single individual (as indicated by the dotted correspondence line). Volitional control can thus be seen as the analog of (canonically) physical causation at this level of organization. Although it is better described as mental than as epistemic, it does involve a certain kind of knowing. This adjustment to the import of epistemic can actually be seen as a consequence of its application to the lexical level: the term knowledge, in its classic sense, is not really applicable to the internal structure of a single physical action.9 In eect, verbs designating volitional actions compress into one lexical package all the elements shown in Figure 2: the
7. This last factor is meant to capture the willful nature of volitional action. The requisite notion of volition does not include desire life is such that we often have to carry out actions we really do not want to perform. 8. Diagrams (c) and (d) are misleading by virtue of suggesting that the mental control is directed specically at the thematic participant (in this case a mover). Cognizance of the theme is only one facet of the control relationship, which is more accurately portrayed as being directed at the causal process as a whole. 9. It does of course apply to facets of mental or communicative predicates (e.g. learn, inform, or know itself ). There, however, the knowledge has a dierent role, functioning as the analog of the physical action per se (as opposed to its mental control).


Ronald W. Langacker

eected occurrence (e.g. dying in the case of kill ), its causation (double arrow), and its apprehension (dashed arrow). Moreover, all the participants shown in Figure 2 collapse into a single individual, corresponding to the agent in Figure 3(d), and all the component relationships constitute a single complex event. Hence the type of knowledge involved corresponding to the dashed arrow in Figure 2(b) is the type we can expect to nd in this compacted environment. And in this context the analogy between the eective and epistemic relationships is even stronger than in Figure 2. Since the latter involves mental control, both relationships are causal in nature.

Grounding Even when describing a series of events, discourse seldom consists in just a sequence of verbs. The basic unit of discourse (if we have to choose one) is a clause, and in particular a nite clause, which has the potential to stand alone as full (non-elliptic) sentence. By itself, a verb (say kill ) does not tell us anything very useful about the world. On the other hand, a nite clause (e.g. Alice killed a rattlesnake) conveys a bit of potentially useful knowledge. One factor enabling it to do so is the specication of event participants (Alice and a rattlesnake). Also essential is an indication of the events status in regard to time and reality whether and when it occurred. A highly grammaticized means of providing this information is what I refer to as clausal grounding (Langacker 1991: ch. 6). Pivotal to English grounding is what I will call the Reality Model. This is the notion that, in our world (which is not just physical, but also has mental and social aspects), things have developed in a particular way, out of all those ways conceivable. There is a certain course of events, whereby some events have occurred, while countless others have not. Reality (R) consists of the events and situations that have occurred up through the present moment. This course of events cannot be changed what has happened has happened. Reality is thus the established course of events. The future, however, has yet to be determined. As reality continues to evolve through time, there are many future paths it can take, some more likely than others given the worlds essential nature and what has happened so far. Another aspect of this basic cognitive model is that our knowledge of reality is partial and imperfect. Moreover, as shown in Figure 4, every individual has a dierent view of it, a dierent take on reality. As living

Control and the mind/body duality: Knowing vs. eecting


Figure 4.

and sentient creatures, we are constantly engaged in building up a conception of reality (RC), what we ourselves accept as real. And one thing we accept as real is that there are other conceptualizers engaged in this task, each with their own reality conception. While they overlap, these reality conceptions diverge in what they cover and dier in specics, hence they are subject to negotiation. This is one major reason why we talk. Of course, the Reality Model species that reality is the way it is regardless of whether anybody knows it, i.e. it is independent of particular reality conceptions. As indicated in Figure 4, the Reality Model itself incorporates an opposition reasonably labeled eective vs. epistemic. The eective level pertains to the actual occurrence of events, which constitutes reality independently of whether we know about them. Our presumed knowledge of these events constitutes the epistemic level. Another facet of the model is that C is part of RC, for a given conceptualizer. That is, one thing C accepts as real is Cs own existence.10 At its core, the English grounding system consists of two binary oppositions, each with a zero member. The rst opposition is what is usually described as present vs. past, but in general is better characterized as immediate vs. non-immediate. Except for third-person singular, where it shows up as -s, the immediate form is zero; the non-immediate form (past tense) of course is variable. The second opposition is the absence (zero) vs. the presence of a modal (may, can, will, shall, or must). Together
10. In the words of Descartes: Je pense, donc je suis.


Ronald W. Langacker

Figure 5.

these specications serve to locate the proled clausal process (p) with respect to the ground, i.e. the speech event and the interlocutors. The interlocutors are also conceptualizers who apprehend the proled process in producing or understanding a nite clause. So in diagrams I will use the label C to represent them in particular the speaker in their conceptualizing role. The absence of a modal indicates that the proled process p is part of Cs conception of reality. Reality being the history of what has happened up through the present moment, in terms of time RC encompasses the past and the present (but not the future). The present, of course, is Cs current temporal location. Immediacy is thus interpreted as immediacy in time (t), so as shown in Figure 5(a), immediate vs. non-immediate translates into present vs. past. But whether present or past, C accepts p as being real. The presence of a modal situates p outside RC, as shown in Figure 5(b). That is, p has not yet been realized not yet incorporated in Cs conception of reality. Each modal makes a dierent kind of epistemic assessment, pertaining to its prospects for realization. For example, may indicates the possibility of ps realization, whereas will predicts it. For all the modals, ps incorporation in RC its acceptance as being real lies in the future. However p itself does not have any particular temporal location; a sentence like She may be upset describes either a future or a present situation. Modals do not locate p with respect to time, but rather in relation to reality (RC). With modals, therefore, immediacy vs. nonimmediacy does not have temporal import (except as a special case). Instead, it is a matter of whether RC functions as the immediate basis for modal assessment. An immediate form (like may or will ) takes RC itself as

Control and the mind/body duality: Knowing vs. eecting


the basis for projection. If I say She will be upset, I am basing my prediction on my current conception of reality. By contrast, non-immediate forms (like would and might) indicate that RC does not support the modal assessment the prediction or the possibility only holds with respect to another, imagined conception of reality (labeled RC 0 ) that in some respect is dierent from RC. For instance, both occur in counterfactual conditionals, as in (6)(e). (6) (a) (b) (c) (d) (e) She is upset. She was upset. [immediate reality] [non-immediate reality]

She {may/will} be upset. [immediate non-reality (future or present)] She might be upset. [non-immediate non-reality (future or present)] If you forgot her birthday, she {would/might} be upset.

What most concerns us here is that English modals have both root and epistemic senses. It should be quite apparent that root modals are the eective counterparts of epistemic modals. I have thus far considered only epistemic modals, which pertain to what the speaker purports to know (accept as real). On the other hand, root modal uses are aimed at having some eect they are meant to inuence what happens at the level of reality itself (R), not just the knowledge or conception of reality (RC). Typically the inuence is social in nature. Often the speaker directs this force at the hearer, who may be expressed as clausal subject, as in (7)(a). But this is only a special case, as the other examples indicate. In general, all one can say is that a root modal describes some force tending toward the realization of p at the eective level. (7) Examples of root modals (eective level): (a) You must apologize to her. (I demand it.) (b) People should always be polite. (That is a social expectation.) (c) Alcohol may be served at the party. (Permission has been granted.) For the most part, we cannot describe the root/epistemic opposition as physical vs. mental in and of itself, social inuence is exerted via the mind rather than the body. Though abstract, both are viewed by cognitive linguists as being force-dynamic (Sweetser 1982; Talmy 1988; Langacker 1991: 6.3). One way to characterize the dierence is that root modals are interactive, whereas epistemic modals express an individual assessment.


Ronald W. Langacker

Figure 6.

Their interactive nature gives root modals the potential to have an eect on what actually happens. With varying degrees of force, they tend to direct the evolution of R (the course of events) along a certain path. And since only its future path can be inuenced (reality at any moment being the established course of events), they normally pertain to future occurrences.11 By contrast, the individual nature of epistemic modals renders them noncausal. The modal force inheres in the evolution of RC Cs knowledge of reality not that of R itself. It is the force that C experiences in mentally extrapolating RC with the aim of encompassing the proled process (p). And since only knowledge is at issue, p is not limited to future occurrences.12 In Figure 6, a double dashed arrow represents the modal force, which in all cases is directed at the realization of p. This realization can occur at either the epistemic or the eective level. With an epistemic modal, p is realized by being incorporated in RC, i.e. accepted by C as part of Cs conception of reality. p can either be external to R (usually lying in the future) or a part of R that C does not yet know about (usually a present situation). With a root modal, p is realized by actually occurring, and the modal force is aimed at this result. Something that occurs thereby belongs to reality, which is simply the history of occurrences.
11. An exception is a sentence like You should have been more considerate, where social force is directed at a counterfactual occurrence whose chance for realization has passed. Rather than action, the eect of the force is therefore guilt. 12. For past occurrences, English uses the perfect construction, e.g. She may have been upset. In this case p is actually the present situation of her having been upset (i.e. the trace in the present of the past occurrence).

Control and the mind/body duality: Knowing vs. eecting


I can mention only in passing what I take to be a manifestation of the eective/epistemic opposition in the domain of nominal grounding (Langacker 2009a). In the nominal realm, the primary issue is not the reality of events, but rather the identication of individuals. As with modals, eective grounding is interactive: it is aimed at directing the hearers attention to the intended nominal referent. The various grounding elements (e.g. demonstratives, articles, quantiers) are alternate means of achieving this objective of joint attentional control, where the referent is momentarily the focus of attention for both interlocutors. If successful, all nominal grounding has this outcome. Roughly, then, the focus of attention is the analog of R in Figure 6(c), with the nominal referent being the analog of p. At the same time, nominal grounding has an epistemic aspect involving the referents position in relation to prior knowledge (whether this is preexisting or established through the discourse itself ). This is manifested in the various degrees of deniteness. It is a matter of whether on the basis of the nominal itself (independently of the clause containing it) the referent has a specic place in established knowledge and can thus be identied in terms of it. Degrees of deniteness are a reection of whether this is so for both interlocutors (denite), for the speaker alone (specic indenite), or for neither (non-specic indenite). Established knowledge is thus the analog of RC in Figure 6(a)(b).

Complementation As dened for English, grounding elements prole the grounded thing or process, not the ground or the grounding relationship (Langacker 2002). For instance in (7)(a), You must apologize to her, the proled relationship is apologize, not the modal must. This is shown in Figure 6, where p is in bold to indicate its status as clausal prole. Despite their importance, both the conceptualizer (C) and the modal force (be it epistemic or eective) are ostage and unproled (subjectively construed). In terms of proling, the situation is therefore dierent from Figure 2, where a verb like make or see proles an eective or epistemic relationship. It also contrasts with Figure 3, where both relations are internal to the proled process. My central point, of course, is that a comparable opposition is manifested at multiple levels of organization, with dierences being attributable to the level. As we move to higher levels of structural organization, a question arises concerning C, the conceptualizer invoked by clausal grounding. I have so far equated C with the interlocutors, and with the speaker in


Ronald W. Langacker

particular. That, however, is only valid as a rst approximation. It holds for the default-case situation where a nite clause is intended as a statement expressing the actual view of the current speaker. If I say Syntax is autonomous and intend this as a true statement of my actual belief, then I am the conceptualizer whose epistemic judgment is conveyed by the absence of a modal; it is my conception of reality which incorporates the proled relationship of syntax being autonomous. But obviously, there are many uses of a nite clause, exemplied in (8), where the speaker does not subscribe to the judgment indicated by clausal grounding. Even a qualifying adverb, like perhaps, insulates the speaker from the epistemic judgment. The proposition in question may be explicitly denied, or ascribed to a conceptualizer other than the current speaker. Indeed, there need not be anybody who actually subscribes to it. (8) (a) (b) (c) (d) (e) Perhaps syntax is autonomous. Its not the case that syntax is autonomous. Many linguists believe that syntax is autonomous. Syntax is autonomous? No way! No one is foolish enough to believe that syntax is autonomous.

In general, then, we have to say that the conceptualizer invoked by a grounding element has no particular identity. C is rather a generalized or virtual conceptualizer, i.e. one that is simply imagined as a way of representing the idea of a conceptualizer assessing p with respect to Cs conception of reality (Langacker 1999a). It is natural for the speaker to assume the role of C; this happens when the speaker wishes to describe his own, actual view of reality. But C can also be identied with the hearer, with somebody else, or with no one at all. So as shown in Figure 7(a), use of a nite clause involves at least two conceptualizers and two levels of conception. There is rst the actual speaker, the initial conceptualizer, which I will represent as C0. The dashed arrow indicates that the speaker apprehends the clause. As part of

Figure 7.

Control and the mind/body duality: Knowing vs. eecting


its meaning, this clause invokes a conceptualizer, C, who apprehends the proled occurrence, p, and adopts some epistemic stance in regard to it; the diagram shows the case where C accepts p as belonging to Cs conception of reality, RC. This much holds for all uses of a nite clause, including those in (8). As a special case, the actual speaker (C0) may identify with C, assuming the role of C in making the judgment expressed by the clausal grounding (Langacker 2004). Showing this identication is the dotted correspondence line in diagram (b). As a result, C0 occupies the role of C in the grounding relationship, so that RC is the reality conception of the actual speaker, labeled RC0 in the third diagram. This happens when the speaker uses the nite clause to make a statement that honestly reects her own view of reality. I say that a nite clause expresses a proposition (P), dened as a grounded process: P [G a p]. It consists of both a process (p) and its epistemic assessment by a conceptualizer, as indicated by the grounding element. So when a speaker, C0, uses and apprehends a nite clause, the speaker is conceptualizing not only the proled process, p, but also the apprehension of that process by another conceptualizer, as shown in Figure 7(a). We may not be aware that there are two conceptualizer roles, either because the speaker identies with C, as shown in diagrams (b) and (c), or else because C is only a virtual conceptualizer and is left implicit. But there are two roles: only as a special case is the conceptualizer implied by clausal grounding identied with the actual speaker. We can use a nite clause for something other than the direct expression of our own actual view because we recognize the existence of other conceptualizers and other conceptions of reality, as shown in Figure 4. Not only do we recognize their existence, to some extent we apprehend the content of their conceptualization and their epistemic stance in regard to it. Furthermore, since we are able to conceptualize propositions without assuming the role of C, they have a kind of autonomy, being independent of any particular conceptualizer. Thus it is possible for dierent conceptualizers to entertain the same proposition, each apprehending it from their own vantage point. One way to deal with this linguistically is by means of predicates of propositional attitude taking nite complements. In (9), for example, both Sally and the current speaker adopt an epistemic stance in regard to the proposition he will lose his job. (9) I doubt that hell lose his job, even though Sally is sure he will. Complementation is a vast topic that I cannot begin to deal with adequately in the space available. Certain points of analysis pertinent to the


Ronald W. Langacker

eective/epistemic opposition can only be sketched here, not fully justied. As a very general point, I suggest that complementation primarily reects the interplay of multiple conceptualizers and levels of conception. Eective and epistemic relationships have dierent manifestations in the various kinds of complement constructions. We can distinguish three basic types of complementation, which with respect to the total range of possibilities have a function roughly analogous to that of cardinal vowels. I will refer to these as direct, innitival, and nite complements, exemplied in (10). In purely formal terms, the predicate of a direct complement bears no marking, whereas an innitival complement is marked by to. A nite complement diers from both by virtue of being grounded (as a matter of denition) and is often introduced by that. What matters here, though, is the conceptual import of these distinctions. In each case the complement participates in the relationship proled by the matrix predicate. How it participates how it interacts with the matrix subject depends on its semantic nature. So when the same predicate occurs with dierent complement types, as in (10), it has subtly dierent values which either permit or reect this usage (cf. Borkin 1973; Postal 1974; Newman 1981; Sweetser 1990). (10) (a) (b) (c) I saw him be abusive toward his wife. [direct complement] I saw him to be abusive toward his wife. [innitival complement] I saw that he was abusive toward his wife. [nite complement]

With a direct complement, the entity participating in the matrix relationship is the proled process itself (p). That is the import or eect of zero marking. We saw in (2) and Figure 2 that the matrix predicates involve either causation or perception, for these are precisely the kinds of relationships that bear directly on events per se. By contrast, I propose that an innitival complement represents an apprehended process (C a p), where the apprehension in question is not just perceptual but mental in a restricted sense.13 In addition to p, it thus invokes a conceptualizer (C) who apprehends this process, and since conception cannot be directly perceived, the matrix predicate is non-perceptual. So although the judgment
13. This is a non-obvious claim that I cannot justify here in any detail. The following discussion should at least make it plausible, however.

Control and the mind/body duality: Knowing vs. eecting


Figure 8.

in (10)(b) may well be based on perceptual evidence, see is used there primarily as a mental predicate. The connection with vision is more tenuous in (10)(c), where see is essentially just a predicate of propositional attitude, roughly equivalent to ascertain or know. A nite complement expresses a proposition, which is not just an apprehended process but a grounded process (G a p). The dierence is that the conceptualizer invoked by grounding, despite its virtual or generalized nature, is still conceived as an interlocutor in an imagined speech event. The complement thus ex` presses an epistemic stance vis-a-vis the proled process, and denes a virtual deictic center which is actualized when the clause is used by an actual speaker in an actual speech event. I show these dierences in Figure 8, where the outer box represents the composite semantic structure of the complex sentence, and the heavy-line box represents the complement. Within the complement, I use C vs. G to indicate the dierence between mere apprehension of p and the particular kind of apprehension implied by the grounding of p. C and G are internal to the complement. In a complement construction, the complement itself is apprehended by an external conceptualizer, given as C1. Often (but not invariably) its apprehension by C1 constitutes (or at least is part of ) the relationship proled by the matrix predicate. The diagrams correspond to the special case where this is so and where C1 functions as the matrix subject, as in (10). I am claiming, then, that both innitival and nite complementation imply the apprehension of p at the mental not just the perceptual level. This is not always obvious, since the conceptualizer is often not explicit or even identied with a particular individual (instead being virtual or construed in generalized fashion). Since grounding involves an epistemic stance, the claim that p is mentally apprehended seems fairly straightforward in the case of nite complements. It is less so in the case of innitival complements, if only because its apprehension has no separate realization. Note the correspondence lines in Figure 8(b). With innitival complements, it is usual for C and Cs apprehension of p to be identied


Ronald W. Langacker

with C1 and the matrix relationship. Thus in (10)(b), C and Cs apprehension of his being abusive are simply those expressed by the matrix subject and predicate (I saw). This overlap between the matrix and complement clauses is the basic reason why an innitival clause as opposed to a nite clause cannot stand alone as a complete and independent sentence. Unlike grounding, innitival to implies that ps epistemic status has to be determined in the context of a larger construction. With respect to conceptualizers and conceptualizer roles, the three kinds of complementation dier as exemplied in (11). In each case we can distinguish between the speaker, C0, and a conceptualizing matrix subject, C1; only as a special case do they happen to be the same (e.g. in (10). For direct complements these are the two basic conceptualizing roles. The other two types involve a third conceptualizer, C, the one who apprehends the complement process. With innitives C and C1 are the same, so there is only one conceptualizer other than the speaker. Cs apprehension of the bomb sitting in the attic is simply that of Harvey imagining it. But with nite complements the three conceptualizers are in principle distinct, and in (11)(c) only C the virtual conceptualizer implied by clausal grounding accepts the complement process as real. Harvey merely suspects that a bomb is sitting in the attic (he does not know it), and the speaker takes no position on the matter. (11) (a) Harvey saw a bomb sitting in his attic. [C0 A C1] (b) Harvey imagines a bomb to be sitting in his attic. [C0 A C1 C] (c) Harvey suspects that a bomb is sitting in his attic. [C0 A C1 A C]

The progression in Figure 8 leads from something potentially physical (an event) to something necessarily abstract (a proposition). Thus in global terms it entails a progression from the eective level, which primarily involves the occurrence of events, to the epistemic level, which primarily involves the knowledge of propositions. Events occur; they can be caused. Strictly speaking, however, a proposition cannot be caused it can only be apprehended and assessed for its validity. Hence this global progression is one manifestation of the eective/epistemic opposition. We have seen, though, that direct complementation itself manifests an eective/epistemic opposition, hinging on whether the matrix predicate is causal or perceptual. Local oppositions of this sort can also be discerned with both inni-

Control and the mind/body duality: Knowing vs. eecting


tival and nite complementation. These will be our main concern in what follows. The overall progression correlates with degrees of directness or immediacy in how C1 interacts with p via the matrix relationship. We can relate this to the meanings of the markings borne by the complement, usually referred to as complementizers. These are often considered meaningless, but we of course know better. This is not however the place for comprehensive semantic descriptions or extensive justication for adopting them. I will merely give a brief sketch of what I take their meanings to be, as useful background for the following discussion. With zero marking of the complement, the matrix relationship can either be causal or perceptual. If causal, it may be purely physical, as in (12)(a). More commonly, though, the causation is eected through social or verbal means, as in (12)(b). So even when the resulting action is physical, the matrix predicate generally has a mental component. Perception, of course, has an intermediate status. The receptor organs are external to the brain, and they are only sensitive to physical stimulation. At the same time, the brain is responsible for perceptual processing and interpretation, which result in mental awareness. Consequently, with either type of direct complementation there is usually a conceptualizer (C1) who apprehends the complement process (p), as shown in Figure 9(a). (12) (a) (b) (c) (d) The weight of the boxes made the table sag. She made them take their shoes o. She saw them take their shoes o. She saw them taking their shoes o.

As one facet of its directness, direct complementation involves temporal immediacy between the matrix and complement events. With perception predicates this immediacy translates into temporal coincidence. Given the nature of perception, the time of perceiving an event coincides exactly

Figure 9.


Ronald W. Langacker

with the time of the event itself. Their coincidence is represented in Figure 9(a) by the two boxes (for the matrix and complement predicates) being coextensive along the time arrow. This corresponds to a sentence like (12)(c) she can only see them take their shoes o as they take their shoes o. With causal predicates, temporal immediacy is interpreted more exibly. It may involve temporal coincidence, as in (12)(a), but since cause precedes eect (in functional terms), there may be some temporal oset between the matrix and complement processes. But while the order to take their shoes o precedes the actual event in (12)(b), the causation and the action may still be coincident in the sense that the social force is in eect (and continues to be eective) throughout the actions execution. An alternative available with perception predicates is for the complement to be marked by -ing, as in (12)(d). In this case the temporal relationship is still direct (immediate) but not precisely coincident. Instead, as shown in Figure 9(b), the perception subtends only a portion of the complement event (Kirsner and Thompson 1976; Langacker 1995a).14 The /-ing contrast is not just one of timing, however. In some small measure, using -ing gives greater play to C1s activity at the mental level. It does so because, even though just a portion of p is actually perceived, it is nonetheless apprehended as a portion of p (cf. Langacker: 2009b). Hence the scope of C1s conception encompasses more than what is perceptually experienced. In terms of the overall progression from physical/eective relationships to mental/epistemic relations, -ing represents one small step in the direction of the latter. Complementation with -ing is still direct, but less so than with zero, since only a portion of p is directly apprehended by C1. On its other ank, -ing alternates with innitival to. The to/-ing alternation is of course a much studied problem of great complexity (cf. Wierzbicka 1988: ch. 1; Egan 2003). I cannot even attempt to deal with it here. I mention -ing only by way of indicating its position in the complement system, as intermediate between direct and innitival complementation. At least prototypically, -ing and to exhibit the temporal contrast shown in Figure 9: whereas -ing implies temporal overlap between the matrix and complement processes, with to there is temporal separation. It most typically appears with matrix predicates involving C1s apprehension of a future occurrence: want, intend, expect, plan, decide, etc. We thus have
14. Thus temporal coincidence is indicated by zero marking, as in the case of the English present tense, while -ing has essentially the same import as in the progressive.

Control and the mind/body duality: Knowing vs. eecting


the contrast in (13)(a)(b), where try picking up the baby implies that the baby was indeed picked up, while try to pick up the baby implicates that it wasnt.15 But this temporal separation is just the most obvious manifestation of a certain indirectness in how C1 connects with the complement process via the matrix relationship. As shown in Figure 8(b), it is not the direct connection aorded by perception. Rather, since the complement represents an apprehended process, C1s connection with p is mediated by the mental act of apprehension if we distinguish between the two, perception is a more direct relationship than conception. Another way to put it is that the mental nature of the relationship serves to insulate C1 from direct contact with p. For instance, in contrast to try picking up the baby, which focuses on the experience engendered by picking it up, try to pick up the baby instead emphasizes the intention of picking it up, a mental act prior to actual execution. Similarly, the contrast in (13)(c) between like to run and like running is one of mental attitude vs. direct experience in regard to the complement process. (13) (a) (b) (c) She tried picking up the baby, but that didnt stop its crying. She tried to pick up the baby, but she couldnt manage to lift it. I like to run a marathon each month, but I dont like actually running them.

I am not claiming that predicates taking -ing complements are never mental, nor that mental activity is always central to predicates taking innitival complements. We are dealing here with complex categories comprising elaborate networks of variants established by conventional usage (Langacker 2000). It is rather a matter of these networks being anchored by dierent prototypes, and a strong tendency for minimal pairs like (13)(a)(b) to contrast in ways reecting these dierent centers of gravity. I suspect that innitival complements can indeed be characterized schematically as representing apprehended processes, but the specic import of such a claim, for the full range of cases, is less than obvious. Here I am only claiming that innitival complement constructions occupy an intermediate position with respect to the global progression from physical/eective to mental/epistemic relationships.

15. At least I take these to be the default interpretations. By extension, try Ving is also used for situations where the eort (or in this case the baby) never gets o the ground.


Ronald W. Langacker

Being ungrounded, an innitival complement lacks a deictic anchor (even a virtual one). Thus it does not have the potential to stand alone as a proposition making a useful specication about the world. It indicates only that p is apprehended, leaving any details about its apprehension and its status to the larger grammatical context. It is thus dependent on the matrix clause both grammatically and for its full semantic interpretation.16 By contrast, a nite complement has greater autonomy, expressing a proposition which can be apprehended by dierent conceptualizers, each with their own vantage point and epistemic stance in regard to it. The eect of including that, I suggest, is precisely to underscore a complements autonomous status as an object of conception independent of any particular conceptualizer or conceptualizing relationship. In Figure 9(d), Ps role as an object of conception is indicated by the large circle containing it. The appearance of that is thus encouraged by factors serving to distance P from C1 or C0 (the speaker) factors like complexity, formality, etc. I note in this regard that it occurs obligatorily when the complement accompanies a noun like fact, claim, idea, or theory, which explicitly portrays it as an abstract object, as in (14)(a). That is also obligatory for complements which function as matrix subjects, as in (14)(b). In my view, this reects the role of a subject complement as matrix trajector, i.e. primary focal participant in the proled relationship. As the onstage focus of attention, its status as an object of conception is maximized. (14) (a) (b) Bush disputes the { fact/claim/idea/theory} that birds evolved from dinosaurs. That birds evolved from dinosaurs is indisputable.

Viewed in global terms, the distinction between innitival and nite complements is one manifestation of the eective/epistemic opposition. With innitival complements, what is primarily at issue is the occurrence of events. Among the predicates taking innitival complements are causal predicates like force, induce, and cause itself, where the matrix subject eects the complement events occurrence. But as noted previously, a proposition per se cannot be caused it can only be apprehended or assessed for validity. Finite complements thus occupy the epistemic end
16. Typically, C is identied with the matrix conceptualizer, C1, and Cs apprehension of p with the matrix process. So in (11)(b), Cs apprehension of p is nothing other than Harveys imagining it. It is also possible for C to remain virtual as a generalized conceptualizer (Its always bad to tell a lie), to be contextually identied (It was necessary to tell a lie), or as a special case of this to be identied with the speaker (To be young again!).

Control and the mind/body duality: Knowing vs. eecting


of the scale, knowledge of propositions being primarily at issue. We have seen, though, that the eective/epistemic opposition is malleable, with varied manifestations depending on level of organization. In the following sections, I propose that it is also manifested locally, within each class of complement constructions. As in the case of direct complements, with both innitival and nite complement constructions we can discern a contrast reasonably described as eective vs. epistemic.

Innitival complements I have oered two basic characterizations of innitival complement constructions: rst, that they are broadly eective in nature; and second, that an innitival complement represents an apprehended process. By eective, I mean that they are primarily concerned with the realization of events and situations.17 Thus they are not limited to causation in any narrow sense, as in the case of direct complementation. Nor does apprehension of the complement process imply that the matrix predicate proles a specically mental relationship, or that the conceptualizer has to be a specic, explicitly mentioned individual. Obviously, we are dealing with an immense family of constructions that dier greatly with respect to numerous parameters. Their center of gravity does however appear to lie with matrix predicates describing sentient states or activities that somehow contribute to the complement events realization. In (15), I describe and illustrate what might be called the main sequence for eective matrix predicates where C1 is the subject. The seven phases represent a natural sequence, though clearly some could be grouped together or further divided. The lists of predicates in (16) do not purport to be complete. While they do not all invariably take personal or sentient subjects, this is certainly the norm for most of them. Even the aspectual predicates (start, begin, continue, cease) com monly refer to willful acts (Perlmutter 1970).18 (15) (a) potential > inclination > decision > intention > preparation > execution > result (b) He {knew how/wanted/decided/intended/was ready/tried/ managed} to cook a soue.

17. As shorthand locutions, I speak of the occurrence of events, or just occurrences. 18. Naturally, many predicates have multiple senses and can thus be listed in more than one category.


Ronald W. Langacker

(16) (a) (b) (c) (d) (e) (f ) (g)

Potential: know how, able, qualied, equipped Inclination: want, hope, willing, inclined, eager, prefer, would like, afraid, hesitant Decision: decide, opt, resolve Intention: intend, expect, plan Preparation: ready, set, prepared, primed Execution: try, attempt, remember, forget, start, begin, continue, cease Result: manage, able, fail

Also part of the main sequence are transitive predicates where the object generally functions as C1. The phases in (17) represent the outcomes of the proled actions. For instance, if you teach someone to do something, then that person knows how to do it. If you persuade someone to do something, then that person decides and intends to do it. (17) Eective with respect to the object (e.g. She taught him to cook a soue ): (a) Potential: teach, enable, qualify, equip (b) Inclination: tell, order, beg, exhort, encourage (c) Decision/Intention: persuade, convince (d) Preparation: get . . . ready, prepare, prime (e) Execution/Result: cause, force, allow, permit, induce, get, help, remind, prompt We can further note impersonal constructions, where the subject is either it or the innitival clause itself. Here the conceptualizer is usually left implicit and construed in generalized fashion, though it can be expressed periphrastically. The basic import of these constructions is that something about the overall circumstances, or the nature of the complement event itself, has an inuence on the events realization. (18) Impersonals (e.g. Its possible [for him] to cook a soue ): (a) Potential: possible, impossible, feasible, permissible, legal, allowable (b) Inclination: important, essential, crucial, advisable, proper, appropriate (c) Execution: easy, hard, dicult, tough, a breeze

Control and the mind/body duality: Knowing vs. eecting


At the very least, predicates like these have a substantial presence among those taking innitival complements. While only a few are specically causal, they are all eective in the narrow sense that the relationship they designate has some role in the realization of the complement event in one way or another it subserves the function of eecting events.19 But this is not the only kind of relationship that pertains to events per se (as opposed to propositions concerning them). Many other relations are eective in the broader sense of involving the realization of events even if they do not contribute to it. For example, numerous predicates designate emotive or aective rela` tionships vis-a-vis the innitival process.20 These can be either personal or impersonal. The aective reaction can be prompted by either the events occurrence or else the prospect of its occurrence. Irrespective of these dierences, all these predicates invoke a conceptualizer (if only in generalized fashion) who apprehends the complement process and has a certain aective experience in regard to it. Hence they do involve a kind of knowledge concerning the complement event the experiencer knows of the event in the sense of being aware of either its actual or prospective realization. But since these predicates merely presuppose this awareness, rather than proling it, they are not essentially epistemic in nature. Moreover, the knowledge and aect in question pertain to the event itself, not to a full-edged proposition (event cum epistemic stance) concerning it. This is the crucial feature distinguishing complementation at the eective and epistemic levels. (19) Emotive/Aective relationships (e.g. Im afraid to watch; Its amusing to watch them): (a) Personal: happy, sad, glad, delighted, sorry, afraid, anxious, reluctant, hate, like, love, surprised, shocked, amazed, astounded, amused, embarrassed (b) Impersonal: good, bad, nice, sad, tragic, terrible, horrible, fun, encouraging, daunting, frightening, surprising, shocking, amazing, astounding, amusing, embarrassing What other kinds of relationships engage events? One kind consists of purely aspectual relationships, when a verb like begin or continue occurs
19. Of course, some relations are eective in a negative way: forget, fail, impossible, dicult, etc. 20. Naturally, there is some overlap with the set of more narrowly eective predicates. We are not dealing with discrete or clear-cut categories.


Ronald W. Langacker

with a non-sentient subject. A number of impersonal predicates give indications of normativity, with respect to either general expectations or personal impression. Some predicates, like pretend or imagine, intimate that the complement process is other than real. Verbs of commitment, like promise, should perhaps be grouped with eective predicates of decision or intention. Still, they are slightly dierent in that the obligation established need not have any actual bearing on the events realization (e.g. the promise may be false). But whether or not they contribute to its realization, these various sorts of relationships all concern its realization in one way or another. (20) (a) (b) Aspectual: start, begin, continue, cease (The ice began to melt.) Normative: normal, common, usual, typical, funny, bizarre, strange, odd, peculiar (Its usual to tip the waiter; Its strange to be here.) Non-reality: pretend, purport, claim, imagine, fancy (She pretended to sleep; He imagines Martians to be living in his attic.) Commitment: promise, threaten, vow, commit (He promised to stay sober.)



This partial survey should at least make plausible and tangible the notion that predicates taking innitival complements pertain primarily to event occurrence, while varying as to their degree of eectiveness in this regard. Of special concern here are some which lie at the non-eective end of the spectrum. In the relationships they designate, the conceptualizer engages the complement process only in the minimal sense of knowing about it or projecting its realization. These predicates are, in a word, epistemic. They accomplish through lexical means something comparable to what is done in grammar through the presence or absence of a modal. Though relatively small in number, these predicates are not infrequent or insignicant. In conjunction with predicates representing the main sequence, they amount to an eective/epistemic opposition at the level of innitival complementation. One aspect of the eective/epistemic opposition is substantial parallelism between the two poles. These epistemic constructions are indeed parallel to their eective counterparts in several respects. First, some of the same predicates function in both (albeit with slightly dierent senses), including promise, threaten, expect, and if we include them the aspec-

Control and the mind/body duality: Knowing vs. eecting


tual predicates. Second, the epistemic predicates involve various phases analogous to those in the eective main sequence. And third, they show a distinction analogous to that between eective predicates centered on the subject and on the object. The predicates in (21) are epistemic with respect to the subject in the sense that their subject is the central participant in the occurrence whose realization is at issue. For example, in the sentence The children are likely to be disappointed, the children are the ones whose disappointment is anticipated. In other words, the matrix subject functions as a kind of topic for the epistemic assessment. The phases indicated in (21) all correspond to phases in (16). Likely is an inclination predicate, i.e. it species a certain force tending toward realization of the complement occurrence. It is not however an eective force neither the subject nor the speaker is portrayed as the source of any impetus toward bringing it about. As with the epistemic modals, the force inheres instead in the very process of epistemic judgment, experienced as a conception of reality (RC) is mentally extrapolated to encompass the innitival process (p). In some cases, the force is such that this reality conception is denitely projected as reaching and incorporating the occurrence. I use the term projection for predicates of this sort. I see these predicates sure, certain, bound, etc. as being analogous to the intention predicates in (16). Likewise, in their nonvolitional construal the aspectual predicates can be seen as counterparts to their eective uses, where they represent the execution phase. Finally, verbs like prove and turn out (e.g. He turned out to be incompetent) are analogous to eective result-phase predicates like manage and able. (21) Epistemic with respect to subject (e.g. The children are likely to be disappointed ): (a) Inclination: likely, apt, liable, seem, appear, promise, threaten (b) Projection: sure, certain, bound, destined, guaranteed (c) (d) Execution: start, begin, continue, cease Result: prove, turn out

The predicates in (22) are epistemic with respect to the object in the sense that their object is the central participant in the occurrence whose realization is at issue. In She expects the children to be disappointed, the children this time expressed as matrix object are once more the unfortunates whose disappointment is anticipated. With these predicates the matrix subject functions as conceptualizer (C1). Most of these can be classed as inclination predicates; know represents the result phase. Thus


Ronald W. Langacker

in (23)(b) the situation of his being abusive is not denitely accepted as being real, whereas in (23)(c) it is. Along another axis, we can distinguish expect from all the others on the basis of temporal orientation. For the most part, expect pertains to future events; (23)(a) would normally relate to a particular future encounter.21 By contrast, the other predicates relate to present situations. As a special case, this situation can be the current relevance of a past event, as expressed by the perfect in (23)(d). (22) Epistemic with respect to the object (e.g. She expects the children to be disappointed ): (a) Inclination: expect, believe, imagine, suppose, suspect (b) Result: know (23) (a) I expect him to be abusive. (b) I believe him to be abusive. [inclination; future-oriented] [inclination; present-oriented]

(c) I know him to be abusive. [result; present-oriented] (d) I know him to have been abusive. [result; present relevance of past event] I mentioned that these predicates accomplish lexically what epistemic modals accomplish grammatically. The analogy with modals is in fact multifaceted. My main point, of course, is that both modals and predicates taking innitival complements exhibit an eective/epistemic opposition. Moreover, epistemic modals are like the epistemic predicates in being either future- or present-oriented. When used in regard to past events, the modals as well do so via the perfect construction, which describes a current situation: (24) (a) (b) (c) He might be abusive so be careful. [future event] He may be abusive were not sure. [present situation] He may have been abusive. [present relevance of past event]

In fact, the diagrams in Figure 6, representing the dierent kinds of modals, apply as well to eective and epistemic predicates at this level, the only dierence being that the predicates prole the eective or epistemic relationship, whereas the modals prole the target process (p).

21. Through metonymy, the anticipated event can also be that of nding out about the ongoing situation of his being abusive.

Control and the mind/body duality: Knowing vs. eecting


A further similarity is that modals and the epistemic predicates both exhibit the phenomenon commonly known as raising, in which a nominal in the main clause appears to belong both semantically and grammatically to the complement. In the transformational era, a rule was posited which raised the subject of the complement into its surface position in the matrix clause (Postal 1974). Thus in (25)(a)(b), the existential there which appears in the matrix is part of the there be construction of the complement. As shown in (25)(c), raising applies to modals as well, on the assumption that modals originate as higher-level verbs. (25) (a) (b) (c) I expect [there be a riot] % I expect there [to be a riot] [Subject-to-Object Raising] [there be a riot] is likely % there is likely [to be a riot] [Subject-to-Subject Raising] could [there be a riot] % there could [be a riot] [Subject-to-Subject Raising]

From the standpoint of Cognitive Grammar, there is neither any need nor any possibility of handling such expressions through derivations from underlying structures (Langacker 1995b). These constructions are simply a matter of the raised nominal having an extremely limited semantic role in the matrix relationship. Instead of interacting with another matrix participant, or having some eective inuence with respect to the complement process, the nominal referent serves only to specify the topic to which the matrix relationship pertains; in The children are likely to be disappointed, the likelihood pertains to the children. Another way to put it is that the nominal stands metonymically for the complement, providing mental access to the complement process, which is what participates directly in the matrix relation (e.g. what is directly assessed for likelihood is the situation of the children being disappointed).22 It is therefore no accident that the raising predicates turn out to be the ones most reasonably characterized as epistemic: they designate relationships in which the nominal referent has no eective role. Often this results from a diachronic process of semantic bleaching or progressive attenuation in degree of agentive control (Traugott 1993; Langacker 1999b; Verhagen 2000). In its basic use, for example, the verb promise is eective in the sense that the subject makes a commitment to carry out
22. Hence the complement process is the active zone for the nominals participation in the matrix relationship (Langacker 1984).


Ronald W. Langacker

the complement process. But in other uses, like (26)(b), the subjects role is largely reduced to it being the topic for a predicted occurrence. Taken to the extreme, this development yields a raising predicate, as seen in (26)(c). (26) (a) (b) (c) She promised to study hard. [eective (commitment)] She promises to become a world-famous scientist. [epistemic (prediction)] There promises to be a lot of controversy on this issue. [epistemic (raising)]

The epistemic nature of certain predicates taking innitival complements indicates that not all knowledge is propositional knowledge. We can be cognizant of an events occurrence, or even project its occurrence, without explicitly formulating a proposition concerning it. When we see an event occur, for example, we have direct knowledge of its realization. With grounding as well, the entity apprehended by C and located with respect to Cs conception of reality is the proled occurrence itself (p), not a proposition. It is only through grounding that a proposition concerning p emerges in the rst place. The resulting proposition (P) is more complex and more abstract, for it incorporates an epistemic stance in regard to p that a particular conceptualizer may or may not subscribe to. Once formulated, a proposition can thus be assessed for validity or negotiated among dierent conceptualizers. The so-called predicates of propositional attitude prole various facets of this assessment or negotiation. In this respect, innitival complementation is intermediate between direct and nite complementation. We observed this in (10), with see as the matrix predicate: I saw him be abusive vs. I saw him to be abusive vs. I saw that he was abusive. As a perceptual predicate, see proles a relationship that engages the complement process directly. With an innitival complement, see is mental rather than perceptual, but still engages the process itself.23 When see takes a nite complement the relationship is also mental, but in this case it engages an autonomous proposition, with its own time specication and epistemic stance in regard to the complement process. I take this to be a consistent conceptual dierence between innitival and nite complementation. They are semantically distinct, not just alter23. The apprehension of this process, indicated by to, is identied with that of the matrix predicate, so it is not included in what is seen.

Control and the mind/body duality: Knowing vs. eecting


nate grammatical forms for expressing the same conceptual content. The dierence is usually not noticed (in fact, it is usually denied) because the constructions very often seem interchangeable. For example, the sentence pairs in (27) appear to be synonymous: (27) (a) (i) (ii) (b) (i) (ii) (c) (i) (ii) (d) (i) (ii) I know her to be dependable. I know that she is dependable. She claims to be dependable. She claims that she is dependable. They expect her to be on time. They expect that she will be on time. She is likely to be on time. It is likely that she will be on time.

But despite their being equivalent in practical terms, I would argue that the members of each pair are slightly dierent in meaning. There is a subtle but real semantic distinction between, on the one hand, assessing the reality or likelihood of an occurrence, and on the other hand, judging the validity of a proposition which incorporates such an assessment. In (27)(a)(i), for instance, the verb know indicates that C1 accepts as real the situation of her being dependable. By contrast, in sentence (ii) know indicates that C1 accepts the complement proposition as being valid (i.e. as conforming to reality). However, since that proposition (through its grounding) expresses the reality of her being dependable, the practical implications are identical: in both cases, C1 accepts as real the situation of her being dependable. The dierence is that sentence (i) describes this acceptance directly, whereas sentence (ii) arrives at it indirectly. It does so by indicating the validity of a proposition where the situations acceptance as real is specied by clausal grounding (in particular, the absence of a modal). The matrix clause I know implies that C1 who happens to be the speaker identies with the virtual conceptualizer (C) invoked by the complements grounding. As a consequence, C1s conception of reality (hence the speakers as well) is identied with that of C. And Cs conception of reality includes her being dependable. Similarly, both sentences in (27)(d) specify as being probable the event of her arriving on time. Sentence (i) does so directly the event itself is assessed as being likely. In projecting the future evolution of reality, the implicit conceptualizer likes the chances of the event being incorporated. Sentence (ii) also indicates the probability of an occurrence, but not the complement process per se. Instead, what is projected as being


Ronald W. Langacker

realized is a situation associated with the complement process: the situation where the complement proposition is accepted as valid. Through a kind of metonymic shift, the assessment of probability applies to the proposition (P) rather than the process (p) it pertains to. Still, there is no dierence in the practical consequences. Projecting the events realization is functionally equivalent to projecting the validity of the proposition describing it.24

Finite complements Viewed in global terms, innitival and nite complementation contrast as eective vs. epistemic: the former concerns the occurrence of events; and the latter, the validity of propositions. Internally, innitival complementation manifests a local eective/epistemic opposition, since one kind of relationship that engages events consists in knowing of them or projecting their realization. It thus seems reasonable to expect that nite complementation might also exhibit a local eective/epistemic contrast. But since it lies at one extreme of the spectrum based on the mind/body duality in the mental task of negotiating propositions, its hard to see that the body as such has any role it could well be that eective relationships are either absent at this level or assume a form rather dierent from notions of causation.25 To begin with the global contrast, we can rst observe the eect of using the same matrix predicate with an innitival vs. a nite complement. The dierence is sometimes quite subtle, as we saw in (27). Often, though, it produces a clear semantic contrast of precisely the sort predicted. The

24. The proposition has to be distinguished from the specic form of the proposition. Right now, when Ps validity is still at issue, it is marked as future because the event has not yet occurred: she will be on time. When the event does occur, so that Ps validity is established, it will be expressed as she is on time. The dierence in tense reects the dierence in vantage point. A basic characteristic of propositions is that they can be apprehended by dierent conceptualizers and from dierent vantage points, yet still count as the same proposition. 25. This would be analogous to physical action predicates at the other extreme of the spectrum. There the role of the mind volitional control is not the sort of thing usually described as knowing.

Control and the mind/body duality: Knowing vs. eecting


sentences in (28) evidence a consistent pattern: with an innitival complement, the matrix relationship pertains to an events occurrence; with a nite complement, it pertains instead to the knowledge and validity of propositions. In (28)(b), for instance, the result of persuasion is intention in sentence (i), but belief in sentence (ii). There is a comparable dierence even when the alternatives have the same practical import, as in (28)(d) (e). For instance, both sentences in (d) describe an attempt to get the speaker to behave, but the attempts are made in very dierent ways. The innitival construction in sentence (i) describes a directive act (some kind of order) aimed at eecting the complement event per se. On the other hand, the act of saying in sentence (ii) is aimed at getting the speaker to accept a proposition. It is through the acceptance of this proposition that they hope to modify the speakers behavior. The directive force comes from the proposition itself, inhering in the modal. By accepting the proposition I should behave, the speaker will presumably acknowledge the social force conveyed by the eective modal should. (28) (a) (b) (c) (d) (e) (i) (ii) (i) (ii) (i) (ii) (i) (ii) (i) (ii) She forgot to lock the door. [EF] She forgot that I dont eat sh. [EP] I persuaded him to hire me. [EF] I persuaded him that I was right. [EP] Its possible to buy a Ph.D. [EF] Its possible that she complained. [EP] They told me to behave. [EF] They told me that I should behave. [EP] He decided to resign. [EF] He decided that he would resign. [EP]

Another indication of the contrast is that only nite complements provide a means of directly referring to past events. Like modals, innitival complements are either future- or present-oriented; past occurrences are described indirectly, via their reection in the present, as in (23)(d): I know him to have been abusive. But nite complements do not force this detour through the present: instead of resorting to the perfect, we can use the simple past: I know that he was abusive. The ultimate source of this dierence, I suggest, is that innitival and nite complementation are respectively concerned with eecting events and knowledge of events. With the former, the past is of lesser concern, since the previous course of


Ronald W. Langacker

Figure 10.

events has already been established only the future remains to be determined.26 But we can perfectly well have knowledge of the past. Let me note one further, crucial dierence between innitival and nite complementation. Both levels are reasonably described in terms of conceptions of reality and the process of realization, i.e. something becoming real. We will see that there is in fact a strong analogy between them. They dier, however, as to what constitutes reality and hence in the nature of realization. The dierence is actually captured by two conventional uses of the English verb realize, as in (29): (29) (a) (b) He realized his ambition to become rich. [eective realization] He realized that he would never become rich. [epistemic realization]

With innitival complements (as well as grounding), reality is dened as the history of occurrences up through the present moment. An event is therefore realized (becomes real) when it happens, and a situation when it obtains.27 Moreover, as shown in Figure 10(a), the reality conception (RC) is just that: Cs conception of reality, so dened. Thus, from Cs
26. Of course, eecting events is just the prototype. The formal restrictions it motivates carry over as the construction is extended to situations and to epistemic relationships. 27. Thus a result phase predicate, like manage, is coextensive with the complement events occurrence.

Control and the mind/body duality: Knowing vs. eecting


own standpoint, a process (p) incorporated in RC belongs to R as a matter of denition. Finite complementation represents a higher level of conceptual organization, where the target of assessment is not a process (p) per se, but rather a proposition (P) concerning p. At this level a reality conception (RC) consists of a set of propositions. Incorporating a proposition in RC accepting it as valid does not imply that p itself is real. We can accept as valid, for instance, a proposition to the eect that p did not occur, that it could never occur, that it might occur in the future, etc. Diagram (b) depicts the situation of a conceptualizer (C1) accepting as valid a proposition of this sort, e.g. in (29)(b). The proposition, to the eect that he would never become rich, is part of C1s reality conception (RC1), even though the event in question (his becoming rich) is specically viewed as non-real (external to R).28 In contrast to an event, a proposition is realized when a conceptualizer comes to know it. And what we supposedly know includes not only the established course of events, but how the path it has followed relates to other conceivable paths (both past and future). As in the case of innitival complementation, with nite complements we can also recognize a core set of matrix predicates as constituting a main sequence with various phases. Here, though, the nal stage is not an events occurrence (eective realization), but rather the acceptance of a proposition as something known (epistemic realization). For our purposes, it suces to indicate ve basic phases, exemplied by the predicates in (30). (30) (a) (b) (c) Potential: possible, conceivable, plausible, impossible, inconceivable Assessment: wonder, ponder, consider, ask, unsure, uncertain

Inclination: think, believe, suspect, suppose, imagine, expect, gure, reckon, doubt, seem, appear, likely, probable, dubious, unlikely, improbable (d) Action: learn, decide, realize, gure out, ascertain, recognize, notice, see, discover, observe, nd out, calculate, determine (e) Result: know, sure, certain, convinced, persuaded, obvious, clear, denite, true, valid, correct, demonstrable, undeniable, false, incorrect The epistemic potential phase is simply a matter of the (usually tacit) conceptualizer being aware of a proposition as something which cannot be
28. Note the correspondence line: because C1 accepts the proposition, he identies with the virtual conceptualizer (C) invoked by the grounding of the complement.


Ronald W. Langacker

dismissed out of hand. Recognizing P as being possible or conceivable sets the stage for its actual assessment. Of course, this preliminary phase also includes the negative alternative of P being impossible or inconceivable, i.e. it cannot be valid, so there is no need for further evaluation. In the assessment phase, P is under consideration, but this is prior to any conclusion. A characteristic of this phase, indicative of its nature, is that the complement is marked by whether or another question word. The inclination phase involves a preliminary judgment, whereby C1 is inclined to accept P as valid (or else to reject it, as with doubt) but has not yet actually done so. The action phase is the counterpart of the execution phase with eective predicates. The action consists in admitting P into C1, i.e. accepting it as valid. Finally, the result phase represents the outcome of this overall process of evaluation: P is included in RC1 (or specically excluded). As with innitival complementation, there are also matrix predicates beyond those constituting the main sequence. I will mention only the large class of predicates describing emotive/aective relationships, exemplied in (31). These correspond to the result phase, since they accept the validity of the complement proposition. However, instead of proling this acceptance they merely presuppose it. (31) Emotive/Aective: happy, sad, elated, thrilled, disappointed, devastated, regret, pleased, displeased, annoyed, disturbed, oended, appalled, disappointed, shocked, surprised, devastated, like it, love it, hate it, resent it, dislike it, detest it The question remains of whether any predicates taking nite complements might be recognized as eective rather than epistemic. Possible candidates are those in (32), where the matrix relationship clearly bears on the complement events realization. Observe, however, that the complements are not really nite, despite being marked by that. What makes a clause nite is not that, but rather its grounding by the tense-modal complex. The complement verb in (32) is not grounded in this fashion. What occurs instead is just the bare, uninected stem (be). If this be grounding and I am not averse to thinking about it in this manner it is grounding of a rather different sort. In lieu of being grounded clause-internally, the status of the complement process (p) in regard to reality is specied by the matrix predicate and the overall construction. There is an evident anity to imperatives, which likewise use the bare stem, as well as to eective (i.e. root) modals. But we can also see this as an alternative kind of innitival construction, where the subject is overt and freely chosen. We need not feel obliged to choose among these options. In a construction-based frame-

Control and the mind/body duality: Knowing vs. eecting


work, we can view this particular construction as drawing motivation from all the others mentioned. I incline to place it at the juncture between nite and innitival complementation (a limiting case of each). (32) (a) She {demanded/required/asked/requested} that I be more cooperative. (b) It is {important/essential/crucial/advisable} that you be more cooperative. The construction in (32) is clearly eective but quasi-nite. Certain other constructions are clearly nite but quasi-eective. Broadly speaking, the predicates in (33) are eective, since they have to do with ensuring the occurrence of events. The relation, however, is indirect the matrix predicates merely indicate a commitment to bring about the complement event, not any actual eort to do so (let alone causation or success). They contrast in this respect with narrowly eective predicates like make, cause, and try. But neither are verbs like promise, etc. epistemic in the narrow sense of pertaining to propositional knowledge. Observe that the complement, though nite, is not a fully autonomous proposition. For the most part, it can only describe future occurrences that the subject is able to control. With a past event, as in (33)(b), what the subject promises or guarantees is that future knowledge will include the complement occurrence (it will turn out to be the case that she got the award). In this case the predicates do indeed have epistemic import. (33) (a) (b) I {promise/guarantee/vow/am determined} that she will get the award. I {?promise/?guarantee/*vow/*am determined} that she got the award.

A number of other predicates, like those in (34), are biased toward future events. To the extent that the complement is constrained to being future, it does not represent a fully autonomous proposition. This correlates with there being less emphasis on knowledge and more on event occurrence. Certainly this is so for fear and afraid what is actually feared is the occurrence itself, not the situation of the proposition being valid.29 Even with expect and anticipate, it is more the event than the propositions validity that is anticipated. All four predicates do allow a past-tense complement, as in (34)(b), although this usage seems less natural than (34)(a). With the past tense the issue of the events occurrence has already been resolved one way or the other in actual practice. As a consequence, the
29. The two, of course, are metonymically related. The event is the propositions active zone for its participation in the matrix relationship (Langacker 1984).


Ronald W. Langacker

interpretations are more clearly epistemic. What is expected or anticipated is not the event of the judge nding him guilty, but rather that the speaker will come to know of this decision. In the case of fear and afraid, the most likely interpretation involves an additional step in the epistemic direction: what the speaker fears is not the event or even the proposition proving valid, but rather the prospect of having to inform the listener of this news. (34) (a) (b) I {expect/anticipate/fear/am afraid} that the judge will nd him guilty. I {expect/?anticipate/?fear/am afraid} that the judge found him guilty.

In short, there are various cases which approximate the situation of nite complementation having eective value, but as expected they are nontypical and less than clear-cut. Control The evident prototype for nite complements is an autonomous proposition capable of being apprehended and independently assessed by multiple conceptualizers. Even when viewed from dierent deictic centers, it can be recognized as being the same proposition, hence subject to communication and negotiation. Metaphorically, propositions are reied as objects that can be examined, passed around, and accepted or rejected. English deploys a rich array of matrix predicates describing the many nuanced ways we have of dealing with them. But what exactly does it mean to say that a proposition is autonomous? From the linguistic standpoint, we can speak of a complement proposition being autonomous to the extent that one can determine what the proposition is independently of the matrix clause (or conversely, that the matrix clause does not impose signicant constraints on it). In this respect, there is a clear contrast between nite and innitival complementation. It is normal in a nite complement for all the central participants to be overtly specied, with no requirement that they be the same as any matrix participant. On the other hand, it is usual for innitival complements to lack an overt subject, precisely because a matrix participant fullls the trajector role. They also lack tense. So, whereas in (35)(a) we know from the complement alone that the emotional state of being upset is predicated of Alice with respect to some earlier time, in (35)(b) we learn these crucial specications only from the matrix clause. Considered in isolation, the innitival expression to be upset is not specic enough to be useful or checked for validity.

Control and the mind/body duality: Knowing vs. eecting


(35) (a) (b)

It appeared that Alice was upset. Alice appeared to be upset.

There are of course intermediate cases. An innitival subject can be specied periphrastically with a for-phrase: I would like for her to get a good job. The bare-verb complements in (32) have specied subjects but are comparable to innitival complements in regard to time reference. With a matrix verb like vow in (33)(a), the complement is nite and expresses a full proposition. However, it is less than fully autonomous because vow imposes certain constraints, notably that the complement event has to be future relative to the time the vow is made (*I vow that she got the award ). With a verb like know or say there is no such restriction. Despite these qualications, the general pattern is clear. An innitival clause represents an apprehended process but does not, in and of itself, express an independently accessible proposition. Certain key elements that would make it an autonomous proposition are instead determined by the matrix, in particular the temporal location of the proled occurrence and its central participant (the trajector). We can say that the matrix controls the complement with respect to these factors. And since these factors relate to the complement event itself, being essential to its characterization, we can describe this as eective control. As it pertains to the complement subject, control is a much studied topic. The problem is usually formulated in terms of predicting which matrix participant is understood as being the controller (i.e. the one that functions semantically as complement trajector). With the verb persuade, for example, the object is the controller [in (36)(a) it is Bill who takes the vacation], whereas with promise it is usually the subject [in sentence (b) it is Alice who takes the vacation]. Control predicates are distinguished from raising predicates, like expect in (36)(c) and appear in (35)(b), where the matrix nominal does not control the complement subject but actually is the complement subject in underlying structure. I consider this putative distinction between control and raising to be articial (Newman 1981; Langacker 1995b), and I do not consider the issue of predicting the controller to be very interesting. I cannot pursue these matters here, but merely point out that this notion of control is quite familiar. (36) (a) (b) (c) Alice persuaded Bill to take a vacation. Alice promised Bill to take a vacation. Alice expected Bill to take a vacation. [object control] [subject control] [raising]

I have called it eective control because it pertains to the complement event per se. Might there be something we can reasonably describe as


Ronald W. Langacker

epistemic control, pertaining to propositions rather than events? I believe there is. Epistemic control is what I previously described as identication, in which some external conceptualizer assumes the role of the virtual conceptualizer (C) invoked by clausal grounding. This external conceptualizer can be the current speaker (C0), as shown in Figure 7(c). C0 identies with C when a nite clause is oered as a genuine statement reecting what the speaker actually believes. More relevant here is the case where the external conceptualizer is the one implied by the matrix predicate, given as C1 in Figure 10(b). In particular, C1 assumes the role C with result phase predicates (know, sure, convinced, etc.), which specify that C1 accepts the complement proposition as valid (hence part of RC1).30 In short, just as eective control species a participant in the complement event, epistemic control species the conceptualizer of the complement proposition. And if we view an innitival complement as corresponding to a proposition, eective control by the matrix serves to indicate what the proposition is. With a nite complement, which species the proposition independently, epistemic control serves instead to indicate its epistemic status, i.e. its standing in relation to established knowledge. Conclusion Complementation is extraordinarily complex. In this presentation I have followed a single thread through a number of structural levels. Of course, the threads are not independent but interwoven in a fabric which I have sketched only in the briefest terms. Nonetheless, I hope to have made the case that the eective/epistemic opposition is a basic organizational feature with numerous manifestations in this domain.31 Recognizing this cross-level systematicity makes possible a more unied account of complementation and related phenomena. I see the eective/epistemic opposition as being grounded in the mind/body duality, which can itself be seen (ironically) as a product of embodied cognition. It is hardly surprising that such a fundamental aspect of experience as we actually experience it should be pervasively reected in language structure.

30. Recall that C1 is not invariably an overt matrix participant. It can also be a generalized conceptualizer or the speaker. In the case of factive predicates (Kiparsky and Kiparsky 1970), the speaker identies with C directly, in addition to whatever epistemic stance is ascribed to C1 (Langacker 2004). 31. It is one linguistic example of fractal organization, where the same pattern repeats itself at successive levels.

Control and the mind/body duality: Knowing vs. eecting


Achard, Michel 1996 Two causation/perception constructions in French. Cognitive Linguistics 7: 315357. Borkin, Ann 1973 To be and not to be. Papers from the Regional Meeting of the Chicago Linguistic Society 9: 4456. Chomsky, Noam 1966 Cartesian Linguistics: A Chapter in the History of Rationalist Thought. New York/London: Harper & Row. Egan, Thomas 2003 Distance and Direction: A Usage-Based Study of Innitive and -ing Complement Clauses in English. (Acta Humaniora 156.) Oslo: Faculty of Arts, University of Oslo. Fodor, Jerry A. 1970 Three reasons for not deriving kill from cause to die. Linguistic Inquiry 1: 429438. Hampe, Beate (ed.) 2005 From Perception to Meaning: Image Schemas in Cognitive Linguistics. (Cognitive Linguistics Research 29.) Berlin/New York: Mouton de Gruyter. Johnson, Mark 1987 The Body in the Mind: The Bodily Basis of Meaning, Imagination, and Reason. Chicago/London: University of Chicago Press. Kiparsky, Paul and Carol Kiparsky 1970 Fact. In: Manfred Bierwisch and Karl Erich Heidolph (eds.), Progress in Linguistics, 143173. The Hague: Mouton. Kirsner, Robert S. and Sandra A. Thompson 1976 The role of pragmatic inference in semantics: a study of sensory verb complements in English. Glossa 10: 200240. Lako, George 1987 Women, Fire, and Dangerous Things: What Categories Reveal About the Mind. Chicago/London: University of Chicago Press. Lako, George and Mark Johnson 1999 Philosophy in the Flesh: The Embodied Mind and Its Challenge to Western Thought. New York: Basic Books. Langacker, Ronald W. 1966 Les verbes faire, laisser, voir, etc. Langages 3: 7289. Langacker, Ronald W. 1984 Active zones. Proceedings of the Annual Meeting of the Berkeley Linguistics Society 10: 172188.


Ronald W. Langacker

Langacker, Ronald W. 1991 Foundations of Cognitive Grammar, vol. 2, Descriptive Application. Stanford: Stanford University Press. Langacker, Ronald W. 1993 Clause structure in cognitive grammar. Studi Italiani di Linguistica Teorica e Applicata 22: 465508. Langacker, Ronald W. 1995a Viewing in cognition and grammar. In: Philip W. Davis (ed.), Alternative Linguistics: Descriptive and Theoretical Modes (Current Issues in Linguistic Theory 102.), 153212. Amsterdam/ Philadelphia: John Benjamins. Langacker, Ronald W. 1995b Raising and transparency. Language 71: 162. Langacker, Ronald W. 1999a Virtual reality. Studies in the Linguistic Sciences 29 (2): 77103. Langacker, Ronald W. 1999b Losing control: grammaticization, subjectication, and transparency. In: Andreas Blank and Peter Koch (eds.), Historical Semantics and Cognition (Cognitive Linguistics Research 13), 147175. Berlin/New York: Mouton de Gruyter. Langacker, Ronald W. 2000 A dynamic usage-based model. In: Michael Barlow and Suzanne Kemmer (eds.), Usage-Based Models of Language, 163. Stanford: CSLI Publications. Langacker, Ronald W. 2002 Deixis and subjectivity. In: Frank Brisard (ed.), Grounding: The Epistemic Footing of Deixis and Reference (Cognitive Linguistics Research 21), 128. Berlin/New York: Mouton de Gruyter. Langacker, Ronald W. 2004 Aspects of the grammar of nite clauses. In: Michel Achard and Suzanne Kemmer (eds.), Language, Culture and Mind, 535577. Stanford: CSLI Publications. Langacker, Ronald W. 2009a Enunciating the parallelism of nominal and clausal grounding. In: Ronald W. Langacker, Investigations in Cognitive Grammar, 148184. Berlin/New York: Mouton de Gruyter. Langacker, Ronald W. 2009b Constructions and constructional meaning. In: Vyvyan Evans and Stephanie Purcel (eds.), New Directions in Cognitive Linguistics, 225267. Amsterdam/Philadelphia: John Benjamins. Newman, John 1981 The Semantics of Raising Constructions. Doctoral dissertation. San Diego: University of California. Perlmutter, David M. 1970 The two verbs begin. In: Roderick A. Jacobs and Peter S. Rosen-

Control and the mind/body duality: Knowing vs. eecting


baum (eds.), Readings in English Transformational Grammar, 107119. Waltham, MA: Ginn. Postal, Paul M. 1974 On Raising: One Rule of English Grammar and Its Theoretical Implications. (Current Studies in Linguistics 5.) Cambridge, MA/London: MIT Press. Sweetser, Eve E. 1982 Root and epistemic modals: causality in two worlds. Proceedings of the Annual Meeting of the Berkeley Linguistics Society 8: 484 507. Sweetser, Eve E. 1990 From Etymology to Pragmatics: Metaphorical and Cultural Aspects of Semantic Structure. (Cambridge Studies in Linguistics 54.) Cambridge: Cambridge University Press. Talmy, Leonard 1988 Force dynamics in language and cognition. Cognitive Science 12: 49100. Traugott, Elizabeth C. 1993 The conict promises/threatens to escalate into war. Proceedings of the Annual Meeting of the Berkeley Linguistics Society 19: 348358. Verhagen, Arie 2000 The girl that promised to become something: an exploration into diachronic subjectication in Dutch. In: Thomas F. Shannon and Johan P. Snapper (eds.), The Berkeley Conference on Dutch Linguistics 1997: The Dutch Language at the Millenium (Publications of the American Association for Netherlandic Studies 12), 197208. Lanham, NY/Oxford: University Press of America. Wierzbicka, Anna 1975 Why kill does not mean cause to die: the semantics of action sentences. Foundations of Language 13: 491528. Wierzbicka, Anna 1988 The Semantics of Grammar. (Studies in Language Companion Series 18.) Amsterdam/Philadelphia: John Benjamins.

A discourse perspective to nominal reference-point constructions* Peter Willemse

1. Introduction This paper deals with an aspect of the description of nominal referencepoint constructions that has so far received limited attention in the literature, i.e. the way in which these constructions are embedded referentially in the discourse context, and the way in which the reference-point relation may interact with information in the surrounding discourse. The focus will be on the English possessive NP, which was the rst construction type described as a reference-point construction by Langacker (1990 and onwards) and was also the focus of Taylors (1996) work. Both NPs with a possessive determiner (e.g. their room, its banks) and NPs with a genitive (e.g. the jurys verdict, the castles moat) will be taken into account. This paper will zoom in on one specic, so far neglected, issue in the description of possessive NPs, i.e. the discourse status of possessee referents. Possessive NPs, when used in discourse, refer to two discourse referents: a possessor (the referent of the possessive determiner or the genitive) and a possessee (the referent of the whole possessive NP, which is anchored to the possessor). Whereas the possessor has direct denite or indenite grounding reecting its status as given or new in the discourse, the possessee is grounded indirectly, via the possessor. Hence, its discourse status is a more complex issue, which needs to be studied in extensive discourse

* This paper is a slightly adapted version of a paper presented at the 10th International Cognitive Linguistics Conference, which was held at the Jagiellonian University of Krakow, Poland, from 1520 July 2007. I gratefully acknowledge the nancial support of the Research Council of the University of Leuven (postdoctoral grant PDM/06/078) and of the research unit FEST (Linguistics Department, University of Leuven) for my participation in this conference. The corpus study presented in this article is based on a chapter from my doctoral dissertation (Willemse 2005: 59136), which was supervised by Kristin Davidse and Liesbet Heyvaert, both of whom I wish to thank for useful comments and suggestions.


Peter Willemse

contexts. Taylor (1996), who opens up a discourse perspective on possessive NPs as reference-point constructions, derives the prototypical discourse status of possessor and possessee referents from the inherent logic of the reference-point relation. It will be shown, on the basis of an analysis of possessive NPs in authentic, extensive discourse contexts, that Taylors (1996) predictions about the discourse status of the possessee need to be rened. The structure of this paper is as follows. In section 2, I will review some of the existing claims and assumptions about nominal reference-point constructions. In section 3, I will present a corpus study of the discourse status of possessee referents of possessive NPs, which shows that possessee referents may be embedded in the discourse context in various ways and may enter into various relations with information in the preceding discourse. Finally, I will draw some conclusions concerning the implications of this study for the analysis of nominal reference-point constructions.

2. Nominal reference-point constructions: previous claims and assumptions The concept of a reference-point construction was rst dened in relation to nominal possessive determiner constructions by Langacker (1990, 1991, 1993). In general, the notion of a reference point involves identication of one entity (typically a less salient one) through its relation to another (typically more salient) entity. The latter functions in this way as a mental reference point for the former, which is referred to as the target. The relation between a reference point and a target is conceived of as one of sequenced mental access (Langacker 2003: 92): mental contact is rst made with the reference point (due to its relative cognitive salience) and subsequently with the target that is being linked to it, via the reference point.1

1. The linguistic notion of reference point in this sense reects a phenomenon that is ubiquitous in our daily experience (Langacker 1993: 5), viz. the fact that it is easier to locate and identify an object in relation to another, more salient, object. For instance, if one is enjoying a panoramic view of a city and wants to locate (or point out to someone else) a specic building, one might resort to locating a more easily visible or recognizable building rst and using it as a reference point, by tracing a path from it to the target building.

A discourse perspective to nominal reference-point constructions


Langacker (1990 and onwards) proposes that possessive NPs should be analyzed as constructions motivated by the reference-point relation. More specically, within a possessive NP, the possessor (i.e. the referent of the genitive or possessive determiner) functions as a reference point for the possessee (i.e. the referent of the whole possessive NP), which is the target that ultimately needs to be identied. Langacker (1993: 89) argues that the relation types that dene the prototypes of the category of possession, i.e. ownership relations (e.g. my house), kinship relations (e.g. Johns cousin) and part-whole (especially body part) relations (e.g. her hand ) are conceptual archetypes, each of which saliently incorporates a referencepoint relationship (Langacker 1993: 8; 2003: 9192). In each case, that is, the entity functioning as the possessor is a natural reference point for the possessee for instance, one normally identies a body part in relation to its owner and a person as a cousin in relation to a certain relative. When a possessive NP is processed, the hearer enters into mental contact with the possessee via the possessor, which (s)he has previously made mental contact with. For example, upon processing the possessive NP Greggs suitcase, the hearer will rst make mental contact with the possessor, Gregg, and subsequently identify the suitcase which the whole possessive NP ultimately refers to as the one associated with Gregg (e.g. the one he owns or is using). The reference-point analysis can account for the grounding function of genitives and possessive determiners. Langacker (1991: 175) analyzes them as atypical grounding predications that ground an instance indirectly: possessive NPs establish mental contact with the possessee instance via the possessor, which itself has direct (denite or indenite) grounding. In this way, the possessor functions as a kind of anchor grounding the possessee.2 An entity is chosen as a reference point for another when it has a certain cognitive salience, either intrinsic or contextually determined (Langacker 1993: 6). The reference-point relation is therefore an asymmetrical one: the reference point starts out as a cognitively salient and accessible entity and provides access to initially less salient and accessible entities. Furthermore, the reference-point relation is dynamic: the reference point is activated (i.e. mental contact is made with it) rst, but

2. For a more general analysis of possessive NPs as anchoring constructions see, for instance, Rosenbach (2002) and Ariel (2004).


Peter Willemse

once the target has been singled out by the hearer, it in turn becomes the focus of the hearers attention, while the original reference point recedes into the background. Once activated, the target may even come to function as a reference point for other entities.3 The inherent logic of the reference-point relation as an asymmetrical construal has led to a number of predictions concerning prototypical properties of reference points and targets in possessive NPs. Langackers analysis of possessive NPs as reference-point constructions is strongly focused on the NP-internal relation between reference point and target as a relation of identication. It elucidates the exact way in which possessive NPs anchor entities by linking a less salient and less easily accessible target entity to a cognitively salient and easily accessible reference point. On the basis of this analysis, Langacker (1991: 175) predicts that reference points (i.e. possessor NPs) will overwhelmingly have denite grounding. Although grounding through an indenite possessor is possible, Langacker argues that it is far from optimal (1991: 175), since indenitely grounded possessors do not involve pre-established mental contact with a unique instance of [a] category (1991: 175). Denite possessors, on the other hand, do involve pre-established mental contact with a specic instance, which is consequently eminently suitable for reference-point function, which is to give access to another entity. Because of the focus on the NP-internal reference-point relation, the embedding of (nominal) reference-point constructions in discourse is not explored systematically in Langackers account. Taylor (1996), who further develops Langackers (1993, 1995) reference-point analysis, specically in relation to pre-nominal possessives (i.e. NPs with a genitive or possessive determiner), also analyzes prototypical discourse properties of possessive NPs. The possessor (reference point) and possessee (target) are thus also considered as discourse referents, embedded in the discourse context in which the possessive NP is used. Taylor (1996: 210218) derives a number of prototypical characteristics of possessive NPs from the inherent logic of the reference-point relation. He summarizes this inherent logic as follows:

3. This may even lead to reference-point chains, in which each target in turn functions as a reference point for the next target, as in the possessive NP Toms girlfriends cousins mother (Langacker 1993: 27; see also Taylor 1996: 230232).

A discourse perspective to nominal reference-point constructions


On the reference-point analysis, the possessor is invoked to facilitate identication of the possessee. It is evident that the reference point, by denition almost, needs to be more easily accessible than the target. Were the target as easily accessible as the reference point, there would be no point in using the reference point for its identication. And if the reference point were less accessible than the target, it would be perverse indeed to invoke a less accessible entity to aid the identication of a more accessible entity. (Taylor 1996: 210)

Taylor further argues (with reference to Chafe 1976) that one way in which the reference point may be an accessible entity, is by being a given entity in the discourse. Discourse-given entities are highly suitable for the reference point function, since they are accessible in the hearers consciousness and thus available as anchoring points. Therefore, entities recently mentioned in the preceding discourse will make good possessors. Discourse-givenness is further associated with other characteristics, such as deniteness, pronominal reference and topicality, which leads Taylor (1996: 212) to predict the following characteristics for possessor NPs: (a) Possessor nominals will refer to entities mentioned in recently preceding text. (b) Discourse, or text topics, will tend to be amongst the preferred referents of possessor nominals. (c) Possessor nominals will be overwhelmingly denite. (d) Pronominal forms will be frequent. (Taylor 1996: 212) On the other hand, Taylor predicts that possessee referents are maximally dierentiated from possessors (1996: 218) with regard to their prototypical discourse status: they are said to overwhelmingly introduce new, previously unnamed entities into the discourse (1996: 217). This also follows from the inherent logic of the reference-point relation, according to which it only makes sense to anchor an entity to a reference point if that entity is not by itself fully accessible. Taylors (1996) account is interesting because it opens up a discourse perspective on nominal reference-point constructions by addressing the question of how referents of possessive NPs are (typically) embedded in the discourse context. This discourse perspective is missing in many accounts of possessive NPs, which merely analyze them as denite NPs and leave it at that (cf., e.g., Quirk et al. 1985: 326; Biber et al. 1999: 271; Lyons 1999: 2326; Huddleston and Pullum 2002: 467). However, Taylor (1996) derives the prototypical characteristics rather aprioristically from the inherent asymmetrical logic of the reference-point relation,


Peter Willemse

which dictates that the reference point is normally a mentally accessible entity that serves to give mental access to an initially less accessible entity. This logic is then reinterpreted in relation to the status of the reference point and the target as referents in the discourse. It is predicted that possessors are typically discourse-given (hence more accessible) referents that anchor typically discourse-new (hence less accessible) possessee referents. While this prediction is largely borne out for possessor referents, which are indeed frequently discourse-given and hence referred to by a denite possessor NP (see Taylor 1991, Ariel 2004; cf. also Martin 1992),4 the claim is much more problematic for possessee referents. An analysis of possessive NPs in extensive discourse contexts reveals that the possessee referent is in the majority of cases not a completely new referent, but rather a referent which has a certain degree of givenness or identiability in the discourse. In the next section, I will therefore present the results of a corpus study5 of the discourse status of possessee referents, based on the analysis of possessive NPs in extensive discourse contexts. This analysis reveals that the possessee referent, rather than being by default a new referent, may be embedded in the discourse in a number of dierent ways and frequently maintains a signicant relation with (an) element(s) in the preceding discourse context that contributes to its identiability in a given context.

3. A discourse perspective on possessive NPs Since the possessee referent does not receive direct denite or indenite grounding, but is only indirectly grounded by the genitive or possessive determiner, its discourse status is a complex matter. In order to uncover it, possessive NPs must be studied in extensive discourse contexts, which make it possible to trace the possessee referents givenness or newness throughout the discourse context preceding the possessive NP. For the study presented here, 400 instances of possessive NPs were studied in discourse contexts of around 500 words preceding the mention of the possessive NP. Table 1 details the composition of the data set of 400 instances of possessive NPs that was used in the analysis:
4. At the same time, it has to be recognized that indenite possessor NPs (i.e. indenite genitives) are by no means anomalous; see Willemse (2005 and 2007b). 5. The discussion in section 3 is a summary of the study which is presented more elaborately in Willemse (2005) and Willemse, Davidse and Heyvaert (forthcoming).

A discourse perspective to nominal reference-point constructions Table 1. Data set NP type Genitive N Total Genitive its N their N my N her N Total Possessive Det. Total 50 [30] [20] 50 50 # of instances 200 Source COBUILD Bank of English (CB)6 200


Lancaster-Oslo-Bergen corpus (LOB)7 [LOB] [CB] CB CB 200 400

The aim of the analysis was to examine the referential relations of the possessee referent to the preceding discourse context. This involved analyzing the way in which the possessee referent is embedded in the discourse context as it were independently of being linked up to a possessor within the possessive NP. For example, the possessive NP the jurys verdict in (1) refers to two discourse referents: the possessor, jury, and the possessee, (the jurys) verdict. (1) Mr Ashby, a former name who suered substantial losses at Lloyds, had sought damages over an article in January last year alleging that he shared a double bed with another man on holiday in Goa. And when the jury found against him in a majority verdict, he put his head in his hands and wept. [. . .] After the verdict, the newspapers solicitor, Alistair Brett, said he would expect the present editor, John Witherow, to see the case as a tragic family problem and be sensible about what to do now. [. . .] Senior Tories expressed their determination to help him after the verdict and launched a campaign to keep him buoyant that was immediately evident in his reception in the Commons. [. . .] He has 28 days to appeal against the jurys verdict and it is then likely to take up to nine months for his costs to be determined by taxation proceedings. (CB)
6. The COBUILD Bank of English corpus is jointly owned by HarperCollins Publishers and the University of Birmingham. Examples taken from this corpus will be marked (CB) throughout the text. They are reproduced with the kind permission of HarperCollins Publishers Ltd. 7. Examples taken from the Lancaster-Oslo-Bergen corpus will be marked (LOB) throughout the text.


Peter Willemse

The possessor referent has been mentioned in the preceding discourse context and is hence referred to with the denite genitive the jurys. The possessee verdict, i.e. the referent of the whole possessive NP, is linked to this possessor, which serves as a reference point for its identication, within the possessive NP. However, this referent is itself also given in the discourse: it is introduced rst by the indenite NP (in bold) a majority verdict, subsequently taken up again by the denite NP the verdict, and is then referred back to by the underlined possessive NP. Thus, the possessive NP in (1) does not conform to the prototype predicted by Taylor (1996): rather than linking a new possessee to a given possessor, it refers back to a given possessee (and links it to a given possessor). In other words, the possessive NP in (1) is coreferential with another NP in the preceding discourse. As will become clear from the discussion below, in the majority of the cases, the possessee has a certain degree of discourse-givenness, either because it has been mentioned explicitly in the preceding context or because it can be inferred. The possessee is strictly new, i.e. introduced for the rst time by the possessive NP, in a relatively small portion of the data. In the cases where it is not strictly new, it is not always fully given, as was the case in example (1). Rather, it may link up with the preceding discourse in a number of dierent ways, which involve dierent degrees of givenness or identiability. Thus, the analysis of the data shows not only, against Taylors (1996) claims, that possessee referents may be given in discourse, but also that it is important to recognize a variety of discourse statuses in between fully discourse-given (coreferential possessive NP) and fully discourse-new (possessive NP introducing a new referent).8 In other words, the analysis will explore the grey area of the cline of discourse statuses between given and new. In total, ve categories are distinguished which form a continuum from discourse-given to discourse-new; Table 2 lists these ve categories. In the remaining part of this section, I will discuss each category in detail and give the quantitative results.9

8. As opposed to the black-and-white distinction between given and new which is sometimes found in the literature (cf., e.g., Barker 2000). 9. A small number of tokens in my sets of data were left out of the classication, since the possessive NP in these cases functioned as part of a xed expression (often a prepositional phrase). This had as a consequence that the referential potential of the possessive NP has not been (fully) actualized, and that the

A discourse perspective to nominal reference-point constructions Table 2. Discourse statuses of possessee referents coreferential text reference bridging the possessee referent has been mentioned in the preceding discourse and is referred back to the possessee referent is a text referent which is construed on the basis of the preceding discourse the possessee referent is identiable based on its association with a preceding referent or with a scenario in the preceding context


anchoring new

the possessee referent is anchored to (an) element(s) in the preceding discourse, which reduces its newness the possessee referent is newly introduced by the possessive NP

3.1. The possessive NP is COREFERENTIAL with another NP in the preceding discourse This rst category was already illustrated by example (1) above, in which the possessive NP the jurys verdict refers back to a referent which had been introduced in the preceding discourse by another NP, a majority verdict, with which it is, consequently, coreferential. The occurrence of a relatively small, but certainly not insignicant number of such cases clearly indicates that Taylors (1996) claim that possessee referents are overwhelmingly new referents can only have validity as a prediction of a tendency. Another example is the following, in which the relation between possessor and possessee is explicitly established in the preceding context (Lawyers acting for dr. Al-Masari) before the two referents are linked in the possessive NP:
entities designated by the NP did not participate in the referential relations that are set up in the discourse. Examples are the following: (i) If Laurie noticed she chose not to comment, but she was curious in her way. (CB) (ii) In a stern reply, he reminded him that Our troops are elated and condent; those on the enemys side cannot but be depressed. (CB) For each category, the number of data left out of the classication was as follows: genitive (8); its (2); their (4); my (4); her (4). The percentages in the tables are calculated on the basis of the number of tokens included in the classication.


Peter Willemse

(2) Britains leading arms manufacturers secretly liaised with ministers, civil servants and the CIA on ways to silence the Saudi opposition leader Muhammed al-Masari, it was claimed last night. [. . .] Lawyers acting for Dr. al-Masari were preparing yesterday to appeal against his removal. [. . .] Dr. al-Masaris lawyers also allege that the home secretary decided to push ahead with the removal of the dissident in spite of a written pledge by the Home Oce that his application to stay in Britain would be considered substantively. (CB) When the possessive NP is coreferential with a preceding NP, the possessee is, in Chafes (1994) terminology, an active referent, i.e. a referent which is given in the discourse and therefore readily available in the addressees consciousness. In addition to the possessor referent, which, in the case of a denite genitive or a possessive pronoun, is explicitly marked as being retrievable10, the possessee may thus be retrievable from the preceding discourse context as well. The possessive NP may even form part of a longer reference chain (Martin 1992) in the discourse, i.e. it may be one NP in a sequence of dierent NPs referring to the same referent, as was illustrated by example (1) above. Table 3 shows the proportion of coreferential or given possessee referents in my data:
Table 3. Results for the category coreferential data set Genitive [CB] Total Genitive its N [LOB] their N [LOB] my N [CB] her N [CB] Total Poss. Det. Grand total # of tokens 28/192 28/192 1/48 4/46 8/46 3/46 16/186 44/378 percentage 14.58% 14.58% 2.08% 8.70% 17.39% 6.52% 8.60% 11.64%

10. Cf. the analysis of denite NPs as phoric in Halliday and Hasans (1976) and Martins (1992) work.

A discourse perspective to nominal reference-point constructions


Coreferential possessives do not constitute a very large category; they account for only about 10% of the data. Still, in a comprehensive account of the discourse status of possessive NPs, this coreferential type is a nonnegligible category which evidences the potential of possessive NPs to refer to given referents in the discourse. Most importantly, the fact that possessive NPs may be used coreferentially in discourse has important implications for the reference-point model in its application to possessive constructions, and underlines the importance of studying the discourse context in which possessive NPs, as reference-point constructions, occur, rather than treating them as isolated syntagms. While the reference-point relation motivates the construction at a schematic level, it does not give the entire picture. In actual usage, the reference-point mechanism can be adapted to specic discourse needs and used for particular rhetorical purposes. Thus, the use of a possessive NP (i.e. a reference-point construction) to refer to a given referent will usually be motivated by the desire to ensure non-ambiguity of the referent in contexts where there is potential confusion. For example, in (2), the possessive construal makes it clear that reference is to those lawyers acting for al-Masari, and not, for instance, to lawyers working for the opposition. Another type of rhetorical use of the reference-point mechanism of possessive NPs is found in coreferential possessives which employ a dierent type specication to refer to a given referent. This may be a synonymous type specication or it may (and often does) entail a more or less drastic recategorization of the referent. Such cases are classied as coreferential because, despite the use of a dierent type specication (dierent lexical realization), they set up a referential identity relation with the surrounding discourse, since the same referent is referred to several times.11 It is interesting to note that when a dierent type specication is used, it often incorporates additional contextually specied information about the referent. For instance, in (3) the underwear being referred to is rst described as knickers whereas in the possessive NP, the type specication silk panties is used, which incorporates the information that the garment in question is made of silk, as indicated in the preceding context.

` 11. Blanche-Benveniste and Chervel (1966) describe such cases as anaphore indele ` (unfaithful anaphora), as opposed to anaphore dele (faithful anaphora), the latter being restricted to cases in which the referent is not recategorized.


Peter Willemse

(3) On Monday, Christies in New York is to sell Greta Garbos knickers. They are described with proper dignity. A pair of silk, cream-coloured ladies briefs (. . .) In fact they are a souvenir of what Christies delicately call a night of romance between Garbo and the Mexican star Roland Gilbert. (. . .) When they parted, Roland gave Garbo the gold ring he was wearing and was given the actresss silk panties in exchange. (CB) Such incorporation of contextual information conrms the idea that reference is not only about referring back to the previous textual mention of a referent, but rather about activating a mental representation of the referent, which naturally evolves and is enriched as the discourse progresses and new information is added.12 Additionally, the use of a dierent type specication to refer to a referent which is already present in the discourse can also be used to categorize that referent according to the subjective evaluation of the speaker. This is the case in the following example, where the referent Bond is classied as a cardboard booby: (4) With the Bond books, as his friend Ernest L. Cuneo wrote, Flemings objective was the making of money and he succeeded. But it wasnt until after his death in August 1964 that Bondmania erupted and the money really began to ow. Film was responsible. Goldnger, the third Bond movie, was released in December of that year, and with it was founded an industry that would turn Flemings cardboard booby into a product that 30 years later rivals Mickey Mouse in terms of global penetration. (CB) This kind of subjective recategorization by the speaker makes it particularly clear that the reference-point mechanism present in possessive NPs interacts with the surrounding discourse to categorize referents and can serve a variety of rhetorical purposes. 3.2. The possessee referent is a TEXT REFERENT construed from the preceding discourse context Text reference has been described by Halliday and Hasan (1976: 5253, 6667) as a special type of reference which can be realized by the pronoun it and by the demonstratives this and that. Text reference involves these
12. See, among others, Emmott (1997: chapter 7) and Brown and Yule (1983: 201204).

A discourse perspective to nominal reference-point constructions


elements referring to any identiable portion of text rather than to an entity that is encoded linguistically as a participant (Halliday and Hasan 1976: 52). More specically, it concerns cases such as the following: (5) [The Queen said:] Curtsey while youre thinking what to say. It saves time. Alice wondered a little at this, but she was too much in awe of the Queen to disbelieve it. (qtd. in Halliday and Hasan 1976: 52) The rst pronoun it refers back to curtsey[ing] while youre thinking what to say, while the second it refers to [the fact] that curtsey[ing] while youre thinking what to say saves time. Both pronouns refer to a discourse referent which has been built up over a stretch of preceding text rather than to a previous nominal mention of a referent; the referent of the rst it is a phenomenon in the extralinguistic world, whereas that of the second it is metatextual13 in the sense that it is a linguistically processed version of this phenomenon (as a fact). In contrast to Halliday and Hasans (1976) strict denition of text reference, I use the term to denote any kind of reference in which an NP designates a referent which has been built up over a stretch of text in the preceding discourse context. Table 4 shows the quantitative results for this category. The number of tokens for this category is relatively small in my corpus. Still, the cases which occur shed an interesting light on possessive NPs, since they reveal their potential to categorize complex referents built up over longer stretches of text. Again, this demonstrates the versatility of possessive NPs as reference-point constructions in interaction with the surrounding discourse. Two types of text reference can be distinguished. On the one hand, there are cases in which the possessive NP contains a metatextual type specication, i.e. in which the possessive NP summarizes and categorizes something which has been described over a stretch of text in the preceding discourse as a semiotic, i.e. a linguistically processed, phenomenon. Possessive NPs containing a metatextual type specication often (but not necessarily) contain typical semiotic nouns, i.e. nouns which designate linguistically processed phenomena, such as decision, claim, remark, etc. An example of this is the following:
13. In the terminology of Halliday and Hasan (1976), reference to a linguistically processed phenomenon or metaphenomenon is extended reference, while the term text reference is limited to reference to extralinguistic phenomena. However, we will use the term text reference in a broad sense, to cover both types of reference.


Peter Willemse

Table 4. Results for the category text reference # of tokens data set Genitive [CB] Total Genitive its N [LOB] their N [LOB] my N [CB] her N [CB] Total Poss. Det. Grand total metatextual T.S. 22/192 22/192 0/48 4/46 1/46 5/46 6/186 16/378 summative T.S. 12/192 12/192 0/48 3/46 0/46 1/46 4/186 16/378 Percentage metatextual T.S. 5.21% 5.21% 0% 2.17% 2.17% 8.70% 3.22% 4.23% summative T.S. 6.25% 6.25% 0% 6.52% 0% 2.17% 2.15% 4.23%

(6) Government cuts in housing grants will lead to huge rent rises, causing thousands of Black pensioners to be worse o, claims a Labour MP. The Government has called on housing associations to raise more of their funds through the private sector and announced that housing association grants will be cut next year. John Battle, Labours Shadow Housing Minister, warned that Black pensioners will be badly hit by increasing housing association rents if Government grant cuts go ahead. (. . .) If this report is taken together with the possible threat to funding of smaller housing associations meeting the specic needs of those in the Black community, the eect on Black elders in housing association homes will be catastrophic, Mr Battle said. Mr Battle, who addressed the National Conference on Housing Black and Minority Ethnic Elders last week, said added rents will soar, hitting pensioners already facing huge hikes in their fuel bills as a result of the VAT increases already announced in the Budget. A National Federation of Black Housing Organisations spokeswoman echoed Mr Battles sentiments. (CB) The noun sentiments in the possessive NP Mr. Battles sentiments refers back to the opinions and ideas of John Battle, which have been repre-

A discourse perspective to nominal reference-point constructions


sented explicitly through indirect speech in the preceding discourse. Note that the possessive NP does not only refer back to what has been stated in the text, but also categorizes (parts of ) the preceding text as a text referent. Thus, a possessive NP realizing text reference not only refers back to a preceding stretch of discourse, but also involves the active construal of a text referent in the sense that it categorizes and labels the text to which it refers. On the other hand, there are cases in which the possessive NP summarizes and categorizes something which has been described in the preceding discourse as a non-semiotic phenomenon, i.e. as a phenomenon (event, state, activity, etc.) in reality that has not been linguistically processed. In such cases I will speak of a summative type specication (cf. Takahashi 1997: 63). Possessive NPs containing a summative type specication generally require some inferencing on the part of the addressee. An example is the following: (7) One BAE insider said last week that there could be no formal discussions until Daimler and the Dutch government had resolved the problems of Fokker, the ailing short-haul aircraft maker that will collapse unless it receives an emergency cash injection of almost 1.4 billion. (. . .) But Fokkers crisis is only one contributor to the problems of Daimler-Benz Aerospace, which will this year see a big increase in losses which totalled almost 200 m last year. (CB) The underlined possessive NP in this example refers to a state of aairs in the extralinguistic reality (Fokkers crisis), not to a metatextual label. It summarizes and categorizes what has been described in the preceding discourse. The information that Fokker needs a vital cash injection in order not to collapse is categorized as a crisis (which requires the addressee to make the inference that for a business, such a situation qualies as a crisis). As has become clear from the examples, text reference involves a fair amount of inferencing on the part of the addressee. It is therefore a separate category in the classication of discourse statuses, distinct from coreferential possessives in which the possessee refers back to a preceding nominal realization of a referent. However, on the continuum of discourse statuses, text reference is right next to coreferential possessives, since it involves a possessee referent corresponding directly (in terms of identication) to a stretch of text in the preceding discourse.


Peter Willemse

3.3. The possessee referent is identiable through BRIDGING The notion of bridging which I want to use as a starting point to describe this particular discourse status needs some preliminary clarication. It is widely discussed in the literature, where it has received a number of dierent interpretations and denitions. It is generally characterized, however, in terms of indirect referent retrieval. Martin (1992: 124), for instance, applies the term bridging14 to NPs which presume information that is implied rather than directly retrievable. Normally, bridging involves denite reference to a referent which is being introduced (i.e. has not been mentioned before) in the discourse, based on the relation with another referent or element in the preceding discourse context. Bridging is therefore an indirect type of reference: the referent of a bridging NP is identiable through its specic relation with another element. That other element is usually another referent realized by a preceding NP, as in (8): (8) Peter has bought a new car. There is much more room in the boot than there was in his old car. In this example, the referent boot is realized by a denite NP, the boot, and is hence coded as being identiable. However, assuming that this is a context in which the referent is being mentioned for the rst time and cannot be identied in the extralinguistic situation, the identiability stems from its relation with another referent in the preceding discourse context, viz. the referent of the indenite NP a new car. In this case, the relation between these two referents is a part-whole relationship (cars generally have boots). Importantly, it is a general (or stereotypical; Charolles and Kleiber 1999: 307; Charolles 1999: 314315; Kleiber 2003: 46; Epstein 1999: 55) conceptual relationship between the two entities that motivates bridging: the part-whole relation between boot and car applies not only to the two specic referents intended here, but holds for boots and cars in general. The processing of bridging reference thus requires some

14. The term bridging was coined by Haviland and Clark (1974). Dierent authors have used dierent terms to describe it, e.g. associative anaphora (Hawkins 1978; Cornish 1986; Kleiber 1999, 2003), inferrables (Prince 1981; Gundel, Hedberg and Zacharski 1993; Gundel 1996), indirect anaphora (Quirk et al. 1985; Erku and Gundel 1987; Epstein 1999) and missing links (Brown and Yule 1983).

A discourse perspective to nominal reference-point constructions


inferencing15 on the part of the addressee, who needs to retrieve the conceptual relation which forms the basis for the identication of the referent from background knowledge (cf. Ariel 1990: 185), while the role of the immediate discourse context is to provide an element which activates this inference (a trigger, Hawkins 1978: 123). When a bridging relation holds between two referents, it may be motivated by one of several types of conceptual relations. Generally speaking, the basis for bridging is some sort of strong associational relationship between entities, one of which is strong enough to allow identication of one entity on the basis of an earlier mention of the other, associated entity. Part-whole relationships form a strong basis for referent identication, as evidenced by their frequent occurrence as the motivating relation for bridging (see Martin 1992; Kleiber 1996), as illustrated by example (8). However, more general relations of a strong and habitual association between two entities may also motivate bridging (cf., e.g., Du Bois 1980: 236; Brown and Yule 1983: 257; Fraurud 1996: 7172). In general, it is possible to retrieve the identity of an entity which is closely associated with another, previously mentioned entity, even when the association does not amount to an actual part-whole relationship sensu stricto, as in the following example: (9) He was very interested in buying that old house up the road, but the owner wouldnt sell. Although an owner is of course not a part of a house, it is an entity typically associated with a house, which can hence be introduced with the denite NP the owner. The application of the notion of bridging to possessive NPs is an extension of its traditional use, since bridging has traditionally been limited to denite reference by means of an NP grounded by a denite article. In the latter case a clear bridging inference is involved: the denite article signals that the referent of the NP which it grounds is presumed to be identiable to the addressee, and for the addressee to identify it, the link with the indirect antecedent needs to be retrieved. The possessive realiza15. Note, however, that the notion of bridging as we use it here is restricted to reference relations in which the (indirect) identiability of a referent is at stake. It is thus narrower than the more general concept of pragmatic bridging inferences which is found in the literature on inferencing (e.g. Van de Velde 1992).


Peter Willemse

tion, by contrast, makes explicit that there is a link between two entities (i.e. the possessor and the possessee), and is thus in a sense an alternative for bridging reference.16 The notion of bridging can nonetheless be said to be relevant to the analysis of possessives, but in a very specic way, viz. with respect to the way in which the possessee referent may relate to elements outside of the possessive NP, i.e. to other elements in the preceding discourse context. As in non-possessive (i.e. classic) instances of bridging, a possessive may refer to an entity which is conceptually closely connected with an element in the preceding context and is thus inferable from this element, i.e. identiable on the basis of the existing conceptual relation. Using Chafes (1994) terminology concerning activation states, such referents can be considered semiactive (i.e. neither fully active nor inactive) and accessible to the addressee: a referent may be accessible because it is directly associated with information that is or was fully active (Chafe 1994: 87). The information that is or was fully active is in the case of bridging the information which forms the basis for the inference and which is explicitly stated in the preceding discourse. Consider the following example: (10) The Leviathan of Parsonstown, said to be the worlds largest telescope, was built in 1845 by the 3rd Earl of Rosse at Birr Castle, Co Oaly. It will form the centrepiece of Irelands new Historic Science Centre which will be based at the castle. (. . .) The rest of the scientic centre, which will include an exhibition of the astronomical work of the 3rd and 4th Earls and scientic galleries in the castles moat, will be completed by the end of the century. (CB) The possessee referent of the possessive NP, moat, maintains a relation of strong association with the referent castle which is mentioned in the preceding context (a castle is very often surrounded by a moat). On the basis of this conceptual relation, it is conceivable that at the moment when the possessive NP refers to moat, this referent is immediately identiable, in the sense that it can be inferred from the referent castle on the basis of general knowledge of the relation between the two entities. It

16. I am not claiming that there is free variation between the two possible realizations of a referent. Dierent semantic and pragmatic restrictions apply to possessives and bridging reference, i.e. many instances of bridging have no possessive alternative and vice versa (see also Willemse 2007a).

A discourse perspective to nominal reference-point constructions


is even conceivable that the referent moat would be referred to with a denite NP rather than a possessive: (10) (. . .) an exhibition of the astronomical work of the 3rd and 4th Earls and scientic galleries in the moat (. . .)

The latter point is crucial to the analysis proposed here. I consider deniteness to be a criterion for bridging in general, i.e. for an NP to have bridging reference, it must be denite. The denite article signals that the instance referred to is assumed to be identiable for the addressee. Indenite NPs, by contrast, do not code the identity of their referent as recoverable (cf. Langacker 2001: 167). Given that bridging is considered here as a type of reference in which the referent is retrievable from the discourse context (i.e. as a phoric relation; cf. Martin 1992: 124), only denite NPs can be analyzed in terms of bridging. An analysis in terms of bridging of a possessee referent is, then, justied when there is reason to believe that this referent can be inferred from an element in the preceding discourse, based on a conceptual connection with that preceding element. For example, in (10) above there is reason to believe that the referent moat is identiable based on the previous mention of the referent castle, because it is an entity typically associated with the latter. However, the ultimate proof that the referent in question is identiable lies in the possibility to refer to it by means of an NP grounded with the denite article, since this is normally possible for fully inferable referents. Thus, the criterion used to categorize a possessee referent as identiable based on bridging boils down to the question: can the referent also be introduced with an NP grounded by a denite article? In the example of the castles moat, an analysis in terms of bridging is plausible, since the referent moat could also be referred to by means of a denite articleNP (the moat) on account of it being inferable from the referent castle in the preceding context. Table 5 shows the quantitative results for this category in the data. A question which immediately springs to mind in this connection is why the speaker chooses to use a possessive NP, rather than an NP grounded by the denite article, to refer to a referent which is inferable from the discourse context. In other words, why choose an explicit anchoring of the possessee to its possessor (by means of a possessive NP) when the latter has been mentioned already and in principle allows the possessee to be inferred? There are a number of possible pragmatic factors which motivate the use of a possessive in such contexts. In general, these have to do with the speaker aiming at greater clarity: the inferential


Peter Willemse

Table 5. Results for the category bridging data set Genitive [CB] Total Genitive its N [LOB] their N [LOB] my N [CB] her N [CB] Total Poss. Det. Grand total # of tokens 27/192 27/192 23/48 4/46 3/46 1/46 31/186 58/378 percentage 14.06% 14.06% 47.92% 8.70% 6.52% 2.17% 16.67% 15.34%

relation or bridge that is there in the discourse is made explicit by the possessive NP (cf. Willemse (2007a) for a detailed discussion). Speakers thus seem to choose the possessive in order to avoid potential confusion about the antecedent, as illustrated by (10), in which there are four clauses between the nearest mention of the referent castle and the introduction of the referent moat, so that the link between moat and castle might not be suciently salient at the moment when moat is mentioned. However, the data show that a clear motivation for the use of a possessive NP to refer to an entity which could have been referred to by bridging is certainly not always present. Consider, for instance, the following example: (11) But even this hardly prepared one for the spectacle that the house itself presented on closer view. It stood, as it were, knee-deep in weeds like some forlorn prehistoric creature in an edible pasture. Its grey surfaces were aked and cracked; its woodwork was denuded of paint; many of the lower windows showed tattered curtains pulled awry, and some of the upper ones lacked entire panes of glass. (. . .) If challenged to date it, Appleby would have said 1718; if challenged to the name of the builder, he would have said James Gibbs. But now it spoke either of madness which, indeed, was what was attributed to its owner or of penury. (LOB) All of the underlined NPs can be analyzed as realizing bridging reference, since they name parts (surfaces, woodwork, windows) or associated

A discourse perspective to nominal reference-point constructions


human referents (builder, owner). However, two of them are anaphorically bridged, i.e. referred to with an NP grounded with the denite article, whereas the others are realized by means of a possessive NP. The antecedent is clear for all of these referents, and no obvious other dierences between e.g. lower windows (referred to by means of bridging) and woodwork (referred to with a possessive NP), or between builder (bridging) and owner (possessive), can account for their dierent realization. It seems, then, that in many cases the choice of one construal or the other is a matter of personal choice on the part of the speaker. What is the import of the analysis proposed here to the overall analysis of the discourse status of possessee referents? I argue that when the possessee referent is identiable via bridging, it cannot be analyzed as entirely new to the discourse. For instance, although strictly speaking, the referent moat in (10) is being introduced into the discourse (i.e. mentioned for the rst time) by the possessive NP, it is identiable based on its conceptual (part-whole) relation with castle, which is explicitly mentioned earlier on in the discourse. However, since it has no direct antecedent (i.e. the referent moat has itself not been explicitly mentioned before) but only an indirect antecedent, it is not strictly discourse-given (i.e. coreferential) either. Moreover, in the hierarchy of discourse statuses it is below text reference (i.e. closer to newness than text reference). The latter type of reference, as was remarked above, involves a referent which directly corresponds to and categorizes a preceding stretch of text. Bridging, by contrast, is essentially indirect: it involves inference of a conceptual relation which holds between two dierent referents, between an event or activity and a referent which is involved in it, or between a scenario or frame and a referent which fulls a role in it. However, since text reference as well as bridging involve some inferencing on the part of the addressee, in order to identify the referent concerned, the boundary between the two categories should not be seen as absolute. Still, core cases of text reference (e.g. metatextual type specications) are quite dierent from core cases of bridging (e.g. bridging based on part-whole relations between entity referents). The former involve direct correspondence in terms of identication; the latter, by contrast, involve only an indirect link with the preceding discourse, via the conceptual relation which forms the basis for bridging (e.g. the part-whole relation). In any case, the use of possessive NPs to refer to bridged entities again shows that the reference-point mechanism is used in interaction with the surrounding discourse and may full dierent functions adapted to the specic requirements of the discourse context.


Peter Willemse

3.4. The possessee referent maintains an ANCHORING relation with the preceding discourse This fourth category forms a transitional category between category 3.3 of possessee referents identiable by bridging and the category 3.5 of new possessee referents that are being introduced in the discourse. It contains cases in which there is partial non-newness of the possessee referent based on an inferential, but non-bridging, relationship with the preceding discourse. In these cases, the use of the denite article to introduce the possessee into the discourse is impossible or only marginally possible, which indicates that the referent cannot be identied to the same extent as for a bridging relationship. Despite this, however, there is undeniably an inferential link with the preceding discourse, so that the possessee referents in these cases cannot be classied as entirely new to the discourse. Specically, two kinds of inferential relations reduce the newness of the referent based on an earlier mention of another referent: on the one hand, body part-relations with human possessors and on the other hand, kinship and other interhuman relations. Table 6 gives an overview of the number of cases in which the possessee maintains a relation of anchoring with the discourse context. Firstly, human body parts cannot normally be introduced with a denite article-NP based on the preceding mention of the possessor (cf. Du Bois

Table 6. Results for the category anchoring relation

# of tokens data set body part 10/192 10/192 0/48 8/46 9/46 15/46 32/186 42/378 inter human 11/192 11/192 0/48 2/46 13/46 8/46 23/186 34/378 other 26/192 26/192 14/48 9/46 5/46 3/46 31/186 57/378 body part 5.21% 5.21% 0% 17.39% 19.57% 32.61% 17.20% 11.11% percentage inter human 5.73% 5.73% 0% 4.34% 28.26% 17.39% 12.37% 8.99% other 13.54% 13.54% 29.17% 19.57% 10.87% 6.52% 16.67% 15.08%

Genitive [CB] Total Genitive its N [LOB] their N [LOB] my N [CB] her N [CB] Total Poss. Det. Grand total

A discourse perspective to nominal reference-point constructions


1980: 241); realization with a possessive NP is obligatory. This is illustrated by the following example: (12) (a) (b) She is on sick leave because she has broken her leg. *She is on sick leave because she has broken the leg.

However, certain constructions do allow for a body part to be introduced with a denite NP: (13) The burglar hit John on the head with a baseball bat. Of course, in this example, reference to the body part by means of a denite NP involves a considerable degree of idiomaticity. Still, such examples indicate that body part-relations allow for certain inferences to be made. This is not very surprising, since body part-relations are a type of part-whole relation, and part-whole relations frequently motivate inference in bridging reference. Another indication of the inferential potential of body part-relations is that when the possessor is non-human, bridging of a body part is possible, as illustrated by the following example: (14) When hounds had run a stag to a standstill, the stag would turn to defend itself. And few dogs fancied rst toss on the antlers, so would stand baying for the huntsmen to come in for the kill. (CB) I have therefore analyzed human possessors and possessive NPs involving body parts as involving a possessee referent which maintains an anchored relation with the preceding discourse, e.g. (15). On the other hand, I have classied NPs designating body parts of non-human possessors in which realization with a denite NP seems possible as involving a possessee referent which is identiable through bridging (category 3.3), as in (16). (15) Claires ngers tingled. She looked down at her hands and saw the raw red bands around her wrists. (CB) (16) To my great regret, it was not Midas who had laid a nger on the ock, it was the ram who had been touching them up. (. . .) His annual visit to the ewes has more the atmosphere of a slightly randy Saga holiday than the frenzied Club Med experience. (. . .) The theory is that when he covers the ewes, his crayon leaves a mark on the ewes rump (. . .). (CB) Kinship and similar interhuman relations are another type of relation which reduces the newness of a referent on its rst mention, based on an


Peter Willemse

earlier mention of an associated referent to which it can be anchored. They are evoked by certain nouns with an inherently relational meaning (see Barker 1991, 2000; Fraurud 1996). Taylor (1989: 675) gives a list of nouns which in addition to kinship nouns invoke, in their semantic structure, various other kinds of interpersonal relationships, including friend, ancee, colleague, guest, fellow student, competitor, condant, etc. Taylor (1989) discusses such relational nouns in terms of how they steer the interpretation of the relationship between possessor and possessee by evoking an unproled relationship in their semantic structure, an element of which is elaborated by the possessor nominal. Another interesting aspect of such relational nouns which we observed in our data is their potential to be used in dierent ways. Either (and most frequently) they are used in a possessive NP in which the possessor nominal elaborates an element of the relation which they evoke, or (and this is a more marginal use) they are grounded by a denite article, when preceded by a mention of the related human referent. The latter use is informal, and mainly found in spoken, colloquial language. An example of this is the following: (17) <ZF1> H h <ZF0> how do you get to the Fab Club? <M01> Well FX c erm <ZF1> the d <ZF0> the daughter comes and picks us up or we get taxis. <F01> Yes. <M01> They order taxis for us one week or we get the bus <F01> Yes. <M01> and er thats how <tc text pause> Im on the bus one week or sometimes Im on two weeks so it all depending on FX and the daughter like organize it the bus and then weve got a transport manager they call him MX. (CB)17 This possibility of realization by means of a denite article-NP is indicative of the anchoring potential of kinship nouns and other nouns denoting interhuman relations. 3.5. The possessee referent is a NEW referent The quantitative results for the category of new possessee referents in my data are given in Table 7.

17. The codes in this example are part of the transcription of spoken language used in the COBUILD corpus. <F01> and <M01> indicate the two participants in the dialogue; <ZF1> and <ZF0> indicate a repetition of (parts of ) words, and FX and MX replace, for privacy reasons, two proper names mentioned in the dialogue.

A discourse perspective to nominal reference-point constructions Table 7. Results for the category new data set Genitive [CB] Total Genitive its N [LOB] their N [LOB] my N [CB] her N [CB] Total Poss. Det. Grand total # of tokens 68/192 68/192 10/48 15/46 7/46 11/46 43/186 111/378 percentage 35.42% 35.42% 20.83% 32.61% 15.22% 23.92% 23.12% 29.38%


The results of the corpus analysis show that although it is true that possessive NPs newly introduce referents in the discourse in a considerable number of cases (30% of the data set), new referents certainly do not represent the majority of possessee referents. This last category forms one end of the continuum of discourse statuses proposed in this study. This implies not only that the referent in question has not been mentioned as such in the preceding discourse, but also that it cannot be inferred on the basis of elements in the preceding discourse context. In many of these cases, the possessee is realized by a nominalization, as in example (18), or by a noun denoting an action or activity, as in (19): (18) His Turnable Emergency Non-capsizable Triangular System (Tents) will inate as it is launched, leaving a gap for four people to crawl in before the entrance is sealed by a zip. It can survive punctures in two of its surfaces and still remain aoat. Hunter has produced two prototypes and is in talks with a lifeboat manufacturer that could lead to the systems launch in the spring of next year. (CB) (19) Chaueur held in socialite slaying were the headlines and they contained all the nuances of sin and sex that readers ate up. It was the kind of case the papers loved. There were people in high places, a beautiful and almost naked woman, and the possibility that under the bright light of police investigation all sorts of scandals would be uncovered. It was the dream case and editors had spared no pains in their eort to give colour to the facts. (LOB)


Peter Willemse

In these cases, the possessor nominal designates the subject of the action described by the nominalized or deverbal head noun. While the possessor referent typically maintains an anaphoric relation with the preceding discourse, the action designated by the possessee noun is not inferable from the preceding text. In some of these cases, there is some sort of a general semantic relation between the possessee referent and the preceding discourse context, but this relation is not strong enough to confer (even partial) givenness on the referent. This is most clearly the case in examples such as (20), in which the possessee referent ts in with the scenario that is being evoked in the context, without being a predictable part of this scenario. That no givenness of the possessee referent is present in such cases is conrmed by the fact that an introduction of the referent by means of an NP with the denite article is excluded. (20) Queensland cricket captain Stuart Law yesterday warmed up for this weeks Sheeld Shield battle against New South Wales with a mighty double century for Valley in a XXXX Brisbane club cricket game yesterday. (. . .) In a punishing display, Law smacked 32 boundaries including two sixes, and faced 266 balls. He came to the wicket when the score was eight after opener Michael Ephraims departed, and immediately went on the attack. (. . .) There was little respite for Universitys bowling attack which included Sheeld Shield bowlers Michael Kasprowicz and Peter Jackson. (CB) In (20), the context evokes the scenario of a cricket game, in which a bowling attack is possible, though by no means necessary; consequently, the scenario does not lead to identiability of the referent bowling attack, which is being introduced (for the sake of clarity, the bowling attack referred to is not the attack implied earlier (Law . . . immediately went on the attack), since it is being executed by the team playing against Laws team). Note that the possessive construction, via the reference-point mechanism, leads to (a certain degree of ) identiability of the possessee referent. The possessee is being linked as a target to a reference point facilitating mental access to it. However, this is local, constructioninternal identiability, which is not relevant to the external discourse status of possessee referents as analyzed here. Only when the possessee referent maintains a relation (of coreferentiality, text reference, independently of the possessive construction, bridging or anchoring) with the surrounding (particularly the preceding) discourse context is its discourse status aected in the sense that it is not (entirely) new.

A discourse perspective to nominal reference-point constructions


4. Conclusions and further research The aim of this paper was to shed some light on the way in which possessive NPs as reference-point constructions are embedded in the discourse context. More specically, this paper has focused on the underresearched issue of the discourse status of the possessee referent. A corpus study of possessive NPs in extensive discourse contexts has been presented of the way in which possessee referents may interact with previously given information and participate in various referential relations with elements in the preceding context. It has been shown that a binary distinction between (fully) discourse-given and (fully) discourse-new does not suce to capture the status of the possessee referent. Instead, a continuum of discourse statuses, ranging from discourse-given (coreferential) to discourse-new over a number of in-between statuses in which the possessee is given to a certain degree, has been proposed. This analysis has a number of important implications for the description and theory of English nominal reference-point constructions. Reasoning on from the reference-point analysis of possessive NPs to their discourse behaviour, Taylor (1996) had predicted that possessee/target referents are overwhelmingly discourse-new and anchored to a typically discourse-given possessor/reference point. Against this view, it has been shown that in fact fully discourse-new referents represent only a relatively small portion (about 30%) of the data. In the majority of the cases, the possessee has a certain degree of discourse-givenness, either by being given explicitly in the preceding discourse context or by being inferable from it to some extent. It may be retrievable from extended descriptions, quotes or reports in the preceding discourse (text reference), it may be identiable through bridging from a given referent in the preceding context, or it may be anchored to an element in the preceding text, which reduces its newness but does not make it identiable in a strict sense. Finally, in a small but signicant number of cases (about 11% of the data), the possessee was fully given, in the sense of being coreferential with a given discourse referent. Moreover, the study presented in this paper adds a systematic discourse perspective to the theory of nominal reference-point constructions. Claims about the discourse status of reference-point constructions cannot be derived aprioristically from the reference-point model, and possessive NPs, as reference-point constructions, cannot be studied as isolated syntagms, without taking into account the way in which they function and


Peter Willemse

are embedded in the discourse context in which they occur. A systematic study of possessive NPs in extensive discourse contexts reveals that the reference-point relation, i.e. one facilitating mental access to and identication of an entity by anchoring it to another entity, may be employed in dierent ways in discourse, adapted to specic discourse purposes. While it is true that possessive NPs are always anchoring constructions, the use of the reference-point mechanism is not limited to the local, NPinternal identication of a referent. Instead, the NP-internal referencepoint relation is used in interaction with explicitly stated and inferable information in the surrounding discourse. The main discourse functions of the possessive NP as a reference-point construction can then be summarized as follows. The possessive NP anchors a target to a reference point in order to facilitate mental access to the target. The target may be a completely discourse-new entity, which is being introduced into the discourse via a more accessible reference point. On the other hand, the target may be an entity having a certain degree of givenness or even a fully given entity in the discourse. When this is the case, the function of the reference point seems to guarantee referential clarity in the unfolding discourse (see also Willemse 2007a). For example, when the possessive NP is coreferential with (an)other NP(s) in the discourse, a possessive NP may be chosen (rather than, say, an NP grounded by the denite article) when there is potential confusion about the antecedent. Anchoring the referent in question to a reference point will in such cases make sure that the hearer picks out the right referent from the preceding discourse. Similarly, when the possessee referent is in principle retrievable through a relation of bridging from a previously given referent, tying it to a reference point in a possessive NP will support the hearer in making the correct inference and identifying the intended referent. In the case of text reference, the possessive NP labels for the rst time a referent which is given through an extended description in the preceding discourse, and the use of a reference-point construction will support the inference that needs to be drawn by the hearer in order to make the right connection. All of these discourse functions need to be studied in more detail. A particularly interesting question concerns the motivation behind the choice of a possessive NP (i.e. a reference-point construction) to refer to a largely or fully given or identiable referent. In other words, why do speakers choose to explicitly anchor such a referent to a possessor/reference point, when it should in principle be identiable by itself, given the information from the preceding discourse?

A discourse perspective to nominal reference-point constructions


Ariel, Mira 1990 Ariel, Mira 2004 Barker, Chris 1991 [1995] Barker, Chris 2000

Accessing Noun-phrase Antecedents. London/New York: Routledge. Accessibility marking: discourse functions, discourse proles and processing cues. Discourse Processes 37(2): 91116. Possessive Descriptions. PhD dissertation. University of California Santa Cruz [published, Stanford: CSLI].

Denite possessives and discourse novelty. Theoretical Linguistics 26: 211227. Biber, Douglas, Stig Johansson, Georey Leech and Randolph Quirk 1999 Longman Grammar of Spoken and Written English. London: Longman. Blanche-Benveniste, Claire and Andre Chervel 1966 Recherches sur le syntagme substantif. Cahiers de Lexicologie IX.2: 333. Brown, Gillian and George Yule 1983 Discourse Analysis. Cambridge: Cambridge University Press. Chafe, Wallace L. 1976 Givenness, contrastiveness, deniteness, subjects and topics. In: Charles N. Li (ed.), Subject and topic, 2555. New York: Academic Press. Chafe, Wallace L. 1994 Discourse, Consciousness, and Time. Chicago/London: The University of Chicago Press. Charolles, Michel 1999 Associative anaphora and its interpretation. Journal of Pragmatics 31: 311326. Charolles, Michel and Georges Kleiber 1999 Introduction. Journal of Pragmatics 31: 307310. Cornish, Francis 1986 Anaphoric Relations in English and French: A Discourse Perspective. London: Croom Helm. Du Bois, John W. 1980 Beyond deniteness: the trace of identity in discourse. In: Wallace L. Chafe (ed.), The Pear Stories: cognitive, cultural and linguistic aspects of narrative production, 203274. Norwood: Ablex. Emmott, Catherine 1997 Narrative Comprehension: A Discourse Perspective. Oxford: Oxford University Press.


Peter Willemse

Epstein, Richard 1999 Roles, frames and deniteness. In: Karen van Hoek, Andrej A. Kibrik and Leo Noordman (eds.), Discourse Studies in Cognitive Linguistics, 5374. Amsterdam/Philadelphia: John Benjamins. Erku, Feride and Jeannette K. Gundel 1987 The pragmatics of indirect anaphors. In: Jef Verschueren and Marcella Bertuccelli-Papi (eds.), The Pragmatic Perspective. Selected Papers from the 1985 International Pragmatics Conference, 533545. Amsterdam/Philadelphia: John Benjamins. Fraurud, Kari 1996 Cognitive ontology and NP form. In: Thorstein Fretheim and Jeannette K. Gundel (eds.), Reference and Referent Accessibility, 6587. Amsterdam/Phildelphia: John Benjamins. Gundel, Jeannette K. 1996 Relevance theory meets the givenness hierarchy. An account of inferrables. In: Thorstein Fretheim and Jeanette K. Gundel (eds.), Reference and Referent Accessibility (Pragmatics & Beyond New Series 38), 141153. Amsterdam/Philadelphia: John Benjamins. Gundel, Jeannette K., Nancy Hedberg and Ron Zacharski 1993 Cognitive status and the form of referring expressions in discourse. Language 69: 274307. Halliday, Michael A.K. and Ruqaiya Hasan 1976 Cohesion in English. London: Longman. Haviland, Susan E. and Herbert H. Clark 1974 Whats new? Acquiring new information as a process in comprehension. Journal of Verbal Learning and Verbal Behavior 13: 512521. Hawkins, John A. 1978 Deniteness and Indeniteness: A Study in Reference and Grammaticality Prediction. London: Croom Helm. Huddleston, Rodney and Georey K. Pullum 2002 The Cambridge Grammar of the English Language. Cambridge: Cambridge University Press. Kleiber, Georges 1996 Anaphores associatives meronimiques: denitions et proprietes. In: Peter Blumenthal, Giovanni Rovere and Christoph Schwarze (eds.), Lexikalische Analyse romanischer Sprachen, 5162. Tu bingen: Niemeyer. Kleiber, Georges 1999 Associative anaphora and the part-whole relationship: the condition of alienation and the principle of ontological congruence. Journal of Pragmatics 31: 339362. Kleiber, Georges 2003 The possessive via associative anaphor. In: Martine Coene and Yves Dhulst (eds.), From NP to DP, 4371. Amsterdam/ Philadelphia: John Benjamins.

A discourse perspective to nominal reference-point constructions


Langacker, Ronald W. 1990 Concept, Image, and Symbol. Berlin: Mouton de Gruyter. Langacker, Ronald W. 1991 Foundations of Cognitive Grammar. Volume II: Descriptive Application. Stanford: Stanford University Press. Langacker, Ronald W. 1993 Reference-point constructions. Cognitive Linguistics 4: 138. Langacker, Ronald W. 1995 Possession and possessive constructions. In: John R. Taylor and Richard E. MacLaury (eds.), Language and the Cognitive Construal of the World, 5179. Berlin/New York: Mouton de Gruyter. Langacker, Ronald W. 2001 Discourse in cognitive grammar. Cognitive Linguistics 12(2): 143188. Langacker, Ronald W. 2003 Conceptual overlap in reference point constructions. In: Masamoto Ukaji, Masayuki Ike-Uchi and Yoshiki Nishimura (eds.), Current Issues in English Linguistics, 87117. Tokyo: Kaitakusha. Lyons, Christopher 1999 Deniteness. Cambridge: Cambridge University Press. Martin, James R. 1992 English Text: System and Structure. Amsterdam/Philadelphia: John Benjamins. Prince, Ellen F. 1981 Towards a taxonomy of given-new information. In: Peter Cole (ed.), Radical Pragmatics, 223255. New York: Academic Press. Quirk, Randolph, Sidney Greenbaum, Georey Leech and David Crystal 1985 A Comprehensive Grammar of the English Language. London: Longman. Rosenbach, Anette 2002 Genitive Variation in English. Berlin/New York: Mouton de Gruyter. Takahashi, Hidemitsu 1997 Indirect anaphors: deniteness and inference. Leuvense BijdragenLeuven Contributions in Linguistics and Philology 86: 5380. Taylor, John R. 1989 Possessive genitives in English. Linguistics 27: 663686. Taylor, John R. 1991 Possessive genitives in English: a discourse perspective. South African Journal of Linguistics 9.3: 5963. Taylor, John R. 1996 Possessives in English. Oxford: Oxford University Press.


Peter Willemse

Van de Velde, Roger G. 1992 Text and Thinking. On Some Roles of Thinking in Text Interpretation. Berlin/New York: Mouton de Gruyter. Willemse, Peter 2005 Nominal Reference-point Constructions: Possessive and Esphoric NPs in English. Ph.D. dissertation, University of Leuven. Willemse, Peter 2007a Direct and indirect anaphora and the possessee referent of pos sessive NPs in English. In: Antonio Branco, Tony McEnery, Ruslan Mitkov and Fatima Silva (eds.), Proceedings of DAARC 2007, 151158. Porto: CLUP. Willemse, Peter 2007b Indenite possessive NPs and the distinction between determining and non-determining genitives in English. English Language and Linguistics 11(3). Willemse, Peter, Kristin Davidse and Liesbet Heyvaert forthcoming English possessives as reference-point constructions and their function in the discourse. In: William B. McGregor (ed.), The Expression of Possession. Berlin: Mouton de Gruyter.

Lexicalizing indirect path: Focus on Finnish motion verbs Jari Sivonen

1. Introduction Motion is a phenomenon that is highly important for human existence and cognition. Therefore, it is not surprising that, particularly within the framework of Cognitive Semantics, the importance of linguistic analyses of motion events has been recognized. As Langacker (1987: 166) argues, the motion of physical objects through space is fundamental to our experience, so an explicit analysis of its conceptualization is important for linguistic semantics. Consequently, there exist numerous cognitive semantic studies and typologies of motion verbs. These studies have generally looked at the linguistic coding of motion events, namely the Trajector (TR), the Landmark (LM), the path, the manner of the motion and the predicated event itself. In addition to this, commonly found semantic extensions of motion verbs, such as abstract and ctive motion, are another rather well-surveyed area. However, the shape of the path itself has usually gained less attention,1 although it plays an important role in the motion event. There are also too few systematic studies of lexicalized path shapes, although the analysis of Tzeltal by Brown (2006) provides an interesting exception. This is not to say that motion verbs proling shaped-paths are uninteresting. On the contrary, as captured by Slobin (2004: 238), the path is an obligatory component of motion-event expressions: in fact no motion can be produced

1. One reason why the paths shape has been largely neglected is probably the fact that the great majority of the well-analyzed motion verbs, with a possible exception of Spanish, comes from the so called Satellite-Framed Languages (Talmy 1985). These languages mostly encode the path into the accompanying particle and not into the verb lexeme itself. The verbs in these languages do not usually specify the shape of the path at all but focus on the manner (e.g. run, walk, limp etc.). However, even then, the path is normally understood as existing and being direct by its shape. This will be discussed in greater detail in section 5.


Jari Sivonen

without a moving entity following a path. Especially, in the spirit of Cognitive Semantics, which underlines the importance of studying a wide range of linguistic phenomena and basing linguistic statements and generalizations on them, as Langacker (1987: 46, 1991: 512514) notes, it is worth investigating what kind of semantic features are emphasized when the path is actually highlighted, and furthermore, when it has unexpectedly an indirect shape. In this paper, by indirect I refer to a path that has one or more explicit lateral movements with regard to the main direction of motion. For example, the English verb outank, when referring to objective motion, proles a path that involves marked lateral movement, compared to straight, rectilinear progression. With respect to the lexicon of motion verbs, if a verb proles the path in general, it species an indirect path shape. In Standard Finnish, there are no single verbs which focus on other aspects of the path than the indirectness of its shape.2 This paper supplements previous research in Cognitive Semantics by focusing explicitly on Finnish verbs that lexicalize indirect paths. Moreover, the study sheds light on a little-studied area by providing a systematic and comprehensive analysis of the most frequent verbs that prole an indirect path. As a result, it suggests a typology of ways in which indirect paths have been lexicalized in verbs in Modern Standard Finnish. I have three objectives. First is to present an analysis of these verbs (section 3). It will be shown that Finnish verbs of indirect path (henceforth VIPs) may be classied into ve separate subtypes according to their path shapes. My second goal is to explore the cognitive aspects of these path types and to provide answers to such questions as why these particular path shapes should be found in the lexicon (section 4). The suggested answer is anthropocentrism: language encodes the external world from the human perspective and through human conceptualization processes. My third goal is more theoretical. Based on the analysis of Finnish indirect paths, I will return from practice to theory and try to point out how the broadly applied theory of lexicalization patterns, Verb-Framed and Satellite-Framed Languages, as introduced and developed by Leonard Talmy (1985, 1991), could benet from a more detailed analysis of indirect paths (section 5). It will be demonstrated that the Satellite-Framed Finnish includes a few Verb-Framed motion verbs which prole an indirect path. In a prototypical case, the shape of these paths is based on a mental scanning process
2. The verb suoria, which roughly means to proceed via a straight path, is the best candidate, but according to recently published dictionary (KS s.v. suoria), it is usually considered to be rare or stylistically coloured.

Lexicalizing indirect path: Focus on Finnish motion verbs


followed by a metaphorical conceptualization of a verb roots referent (e.g. risteilla criss-cross) or, more rarely, on a metaphorization of the verb root referents motion (e.g. sahata move back and forth as a saw does). The Finnish verb lexemes analyzed in this paper are kaarrella curve, wheel, kaartaa encircle, curve, corner, kiemurrella wriggle, wind, curl, coil, kierrella prowl, tour, swan about, circle, kiertaa bypass, circle, circulate, orbit, revolve, skirt, tour, koukata outank, luikerrella wriggle, luovia tack, mutkitella meander, puikkelehtia weave, pujotella do slalom skiing, zigzag, manoeuvre, snake, pyortaa curl back to the starting point, risteilla criss-cross, sahata move back and fort like a saw does and sukkuloida shuttle. These verbs have been chosen, rst of all, because they are the most frequently used verbs that in Modern Standard Finnish express motion along an indirect path. The other selection criterion is that the above verbs are stylistically neutral and unmarked, in contrast to verbs like siksakata (zigzag), which are considered not to be part of standard Finnish.

2. Indirectness of path Since my focus is on the indirect path, the denition of path deserves a closer discussion. In Cognitive Semantics, one of the most inuential researchers on motion verbs is undoubtedly Leonard Talmy. According to Talmys (1975: 181) early work,3 the path can be dened as the respect in which one object is considered as moving or located to another object in a motion situation. In other words, the path can be understood as a route along which the TR is moving towards, into, in or with respect to the LM. Talmy (2000a: 149) even asserts that one can concretely see the path of a moving object as having Euclidean specics, such as an exact shape and size, while the object is in motion. Even if one does not share such a radical statement, since a path is not usually visible in the same sense as ordinary objects, there is no doubt that a path can be extrapolated, or sensed, as Talmy (2000a: 149) puts it, in a motion event. Thus, from a cognitive perspective, the path denitely exists. As Svorou (1994: 29) points out, conceptualization of the path also presumes a priori
3. Later on Talmy (2000b: 53) postulated three dierent components in the path: the Vector, the Conformation and the Deictic. However, this division is ignored here as not necessary for the purposes of the present analysis.


Jari Sivonen

that there is a starting point of the motion as well as an ending point, even though these need not to be proled by a motion verb or explicated in a sentence referring to a motion event. Indirectness, on the other hand, can be understood as the opposite of directness, since any route or shape which is not direct must be more or less indirect by denition. A common semantic feature of all VIPs is that they do not allow an adverbial modier expressing an explicit linear path. Therefore it is quite possible to say Vene mutkitteli jarvella A scull was meandering on the lake, but the same sentence with an adverbial modier proling the TRs path as direct, is semantically contradictory and ineligible (*Vene mutkitteli suoraviivaisesti jarvella A scull was meandering straightforwardly on the lake). Theoretically, there is an unlimited number of dierent types of indirect paths. What is important, however, is to focus on those indirect shapes of paths that are lexicalized and thus have their own representative verbs in the lexicon of a given language.4 As pointed out by Janda (2008: 3), human sensory perception organs lter out part of the information from our observable environment. In a motion event, only certain aspects of the path are conceptually categorized and stored in the lexicon. These particular forms can be considered cognitively salient in human categorization. The path of an entity moving in a three-dimensional space can naturally be observed from various angles. Dewell (2007: 398403) discusses the English preposition around and concludes that in a state of aairs referred to by around one of four dierent vantage points may be chosen for conceptualization: global scope, local scope, LMs viewpoint and TRs viewpoint. This applies, in principle, to verbs of motion too. Vantage plays a role in such verbs as kiertaa circle, orbit, revolve, which resem bles semantically the preposition around, since both prole paths that vary from semi-circle to full circle and thus allow dierent vantage points. However, dierent vantage points become less signicant in the discussion of other Finnish VIPs, because circularity is just one type of indirectness of paths lexicalized into verbs of motion. For example, the verb risteilla criss-cross is clearly dierent, because the shape of the path it proles

4. It should be mentioned, of course, that especially in a Satellite-Framed Language like Finnish, the indirectness of the path is mostly expressed by numerous periphrastic constructions, where a verb of motion is accompanied by a path-expressing adverbial modier, for example menna edestakaisin go back and forth.

Lexicalizing indirect path: Focus on Finnish motion verbs


cannot be easily perceived from the plane of the path, for example from the TRs or LMs vantage point, but needs to be observed from above. If it is not observed from above, the sense of the paths particular shape is lost, because if the path is viewed from the same plane, there is little evidence that would support the paths self-cutting shape. As observed by Dewell (2007: 404), some vantage points reveal information that others suppress. In order to be able to describe indirect paths proled by Finnish motion verbs, we need to select a vantage point that is broad enough to cover all lexicalized cases, including paths proled by verbs like risteilla. It is quite plausible to presume, as acknowledged by Dewell (2007: 410), that the canonical vantage point of conceiving the path is from above, and that it is the vantage point that should be assumed when making a linguistic description of indirect paths. Therefore, the path shapes are described in this paper as seen by an external conceptualizer: from above of the motion situation taking place in a twodimensional space. In other words, the virtual position from which the motion event and the TRs movement are viewed, which is the vantage point also adopted in Cognitive Grammar (Langacker 1987: 123), is a birds-eye view. When we describe indirect paths, it is important to notice that none of these verbs actually determines the shape of the TRs path in great detail. This means that rather than placing exact restrictions upon the precise shape of the path, each verb provides relatively loose limits and boundaries to the paths prole. This is illustrated by the verb kaartaa to curve in example 1 below.5 (1) Auto kaartaa pihalle. car-nom yard-allat A car is curving into the yard.6

5. In order to keep the focus on indirect paths, the examples are simplied and usually only the necessary elements (e.g. the TR, the verb, the LM) are included. In a natural context, the sentences may be more complex. 6. The following abbreviations are used in the glosses. acc accusative elat elative iness inessive pl plural adess adessive gen genitive nom nominative sg singular allat allative illat illative part partitive imperf imperfect (preterite)


Jari Sivonen

Figure 1. Dierent paths proled by the verb kaartaa curve

If we consider the TRs path in example 1, it is obvious that during the process the TR moves into the LM. This predication species that the basic conceptual relation between the TR (auto car) and the LM ( piha yard) changes from separation to inclusion. TR also moves along a path that has a kind of sweeping prole and the path begins outside the LM and ends inside it. However, it is important to notice that not all possible aspects of the path are specied in the prole of the predication of kaartaa. For example, neither the distance of the TR from the LM nor the degree of the paths actual bending is lexicalized. Nor does kaartaa specify the direction from which the TR approaches the LM, and therefore it remains neutral as to the direction of the curve. Figure 1 above illustrates some possible paths proled by the verb kaartaa when viewed from above (apart from the one represented in example 1). The stick-man represents the TR, the square the LM and the arrow line the TRs path. As Figure 1 shows, the TR may enter the LM either from the left or from the right (compare paths A vs. B and C). Although Figure 1 does not show this, the TR can move either clockwise or counterclockwise. It is also quite possible for the TR to enter the LM from the front or from the rear (when observed from above in a two-dimensional space), though these paths are not shown in Figure 1. According to Dewell (2007: 411412), path expressions characteristically [in two-dimensional space] make a distinction in the up from down dimension, and not in the right-to-left or the left-to-right dimension. As shown in Figure 1,

Lexicalizing indirect path: Focus on Finnish motion verbs


this is also true about the path proled by kaartaa, since it allows a free choice with respect to lateral movement. However, it would not be natural to use the verb kaartaa if the TR was conceived of as moving in a three-dimensional space and if it entered the LM from the top or from the bottom. There are no specic motion verbs in Finnish that conceptualize such paths. The best candidates are perhaps verbs ylittaa to y/go over and alittaa to y/go under, but they do not specify the shape of the path and imply that the TRs path does not end within the connes of the LM, as the verb kaartaa does, but rather that it continues beyond it. Figure 1 also demonstrates the role of the observers orientation in the path analysis. Paths A and B are identical. Only the orientation or the observers own positioning in the situation, is opposite.7 In other words, the shape of the path itself remains the same regardless of the viewers own position, and only the mutual relation between the TR and the LM changes according to the viewers orientation. So, as is shown in Figure 1 (A), the TRs path goes from left to right. However, if the observer makes a 180-degree turn and thus there is a shift in orientation, the paths appearance will also change from right to left. Nevertheless, the actual shape of the path itself remains the same. This suggests that in this case the role of the viewer is not signicant. However, comparison of paths B and C in Figure 1 reveals the verbs exibility in proling the path, as mentioned earlier. Path B bends more emphatically into the LM, whereas path C is gentler. Both paths still meet the limitations and specications proled by the predication of kaartaa and the verb is undoubtedly suitable for referring to both motion events. To sum up the observations made so far, we can argue that VIPs allow certain variation in the exact shape of the path. They also prole certain characteristics which restrict the appearance of actual paths. Langacker (1999: 23, 271) argues that a lexical verb names only the process type while a nite form proles a (grounded) elaboration of that type. As regards verbs that prole an indirect path, this means that only the concrete instance of a verb, e.g. an utterance where it is used, ultimately species the accurate shape of the path. This is illustrated in examples 2 and 3 below. (2) Janis puikkelehtii metsassa. hare-nom weave-3sg forest-iness There is a hare weaving in the forest.
7. The point at which the TR enters the LM is ignored here.


Jari Sivonen

(3) Taskuvaras puikkelehtii vakijoukossa. pickpocket-nom weave-3sg crowd-iness There is a pickpocket weaving in the crowd. Although 2 and 3 contain the same verb puikkelehtia to weave, the precise shape of the TRs path proled is likely to be dierent. This is due to the physical characteristics of the TR and the LM as well as to other attributes attached to them. One fundamental tenet of the Cognitive Semantic theory is that it acknowledges the importance of encyclopaedic knowledge in the interpretation of linguistic expressions. In the most extreme case this means that everything one knows about a concept forms part of its meaning (see Langacker 1987: 155158). Thus, linguistic semantics becomes a study of commonsense human experience. Encyclopaedic knowledge about the TR and the LM is therefore an important factor when we consider the semantics of an indirect path, since it guides the way in which the paths shape is understood. For example in 2, it is possible to imagine that the hares general path, despite the sudden to-and-fro manoeuvres, which is characteristic of hares, proceeds in a relatively straightforward manner. The accurate path of the pickpocket in example 3, however, might be more localized since the TR is now an intentional human being, more likely to remain in the same place in order to commit the theft. Moreover, forest provides a broader setting for movement than a concourse of people, which also has an inuence on the TRs path. The path of the thief may therefore have deeper lateral movements with respect to the main direction of motion. However, it should be underlined that the paths of both the hare and the pick-pocket are similar enough to be categorized as the same type of motion and, consequently, to be referred to by the same verb. As examples 2 and 3 suggest, the encyclopaedic knowledge attached to the TR and the LM of a verb predication has important consequences for the description of paths. Consequently, it is not possible, even in theory, to provide detailed descriptions of paths that are proled by particular verbs. Thus, in the remainder of the paper, I will analyse the most prototypical appearances of paths that characterize and distinguish each verb and base generalizations about the semantics of VIPs on these analyses.

3. Lexical facts: Finnish verbs of indirect path In this section, my aim is to present a typology of indirect paths proled by Finnish verbs. It builds upon my previous work within the framework

Lexicalizing indirect path: Focus on Finnish motion verbs


of Cognitive Grammar (Sivonen 2005b). Further down I will analyse the fteen most important and most frequently used Finnish verbs that prole an indirect path. The data come mainly from newspapers published in the 1990s and they were drawn from the large Modern Standard Finnish corpus called Language Bank of Finland. They consist of almost 20,000 verb tokens in their sentential context. In Modern Standard Finnish, the fteen most frequent VIPs can be divided into ve distinct categories according to the shape of the path.8 These types are, in order of the number of verb lexemes they contain, Paths of Irregular Shape (3.1.), Paths of Regular Shape (3.2.), Paths of Single Turn (3.3.), Paths of Back and Forth Movement (3.4.), and nally Paths of Crossing Movement (3.5.). It should be noted that these fteen lexemes are highly polysemous and they prole several distinct process types in all kinds of motion (objective, abstract and ctive).9 As a result, some VIPs prole processes which fall into dierent path types. For example, the transitive use of the verb kiertaa can have both the meaning of avoidance, in the sense of avoiding an obstacle (cf. example 4), and the meaning of moving inside a region (cf. example 5). (4) Kierramme aidan ja astelemme ovelle. circle-1pl fence-gen.acc and pace-1pl door-allat We will go past the fence and pace towards the door. (5) Kesalla kierramme Lapissa. summer-add circle-1pl Lapland-iness During the summer, we travel in Lapland. Thus, example 4 proles Path of Single Turn and example 5 Path of Regular Shape. The verb sukkuloida to shuttle has some similar characteristics. One process type expressed by sukkuloida belongs to the category Paths of Back and Forth (e.g. Myyntiedustaja sukkuloi Helsingin ja Turun valilla The sales agent shuttled to and fro between Helsinki and Turku), but sukkuloida can also mean that the TR is moving inside a certain region (e.g. Diplomaatti sukkuloi Afrikassa The diplomat shuttles in
8. Various Finnish dialects are likely to have more verbs of indirect path. However, it would be surprising if some of those proled a dierent kind of path than the ve types discussed here. 9. Dierent types of motion are discussed in detail for example by Langacker (1987: 170173, 1990: 157160) and Talmy (2000a: 99175); for metaphorical extensions of motion verbs see for example Jakel (2003), for detailed analysis of the polysemy of Finnish verb kierrella see Sivonen (2005a). ,


Jari Sivonen

Africa). The latter example provides a strong image of several locations scattered inside the LM (Africa), which are visited by the TR. Hence the TRs path in this case provides a typical example of a crossing movement. Motion verbs are not only polysemous, but the meanings of VIPs overlap in many respects. The verbs in the rst two groups, Paths of Irregular Shape and Paths of Regular Shape, resemble each other with respect to the indirect and the direct phase in the path. In both groups, the TRs path contains both rounded (indirect) and linear (direct) parts, when observed from above in a two-dimensional space. However, in the latter group the paths contain more rounded elements, whereas in the rst group, circular as well as straight phases are allowed and, therefore, the overall shape of the path is more irregular and ragged. The other three groups are dierentiated more easily. In the third group (Paths of Single Turn), the verbs prole a path that contains only a single indirect phase, which forms the overall TRs route. The verbs in the fourth group (Paths of Back and Forth Movement) prole a path in which the TR moves back and forth, and therefore form a path type of their own. Also the verbs in the nal group (Paths of Crossing Movement) prole paths that can be distinguished from all the others, because they emphasize the TRs iterative crossing of its own route. As is broadly accepted in Cognitive Grammar (cf. Langacker 1988: 1011), these kinds of lexemes are not semantically identical, i.e. synonymous, even though they may be used as functional counterparts in certain contexts. However, they all have one common denominator: they prole indirect shapes of the path. 3.1. Irregularly shaped paths, verb roots and derivation Three main characteristics describe the Finnish verbs proling paths of Irregular Shape. First, the paths contain various kinds of round or arclike phases. Second, these phases are situated in the TRs path randomly, not in any regular order. Third, the indirect phases do not form any observable or predictable pattern. The verbs which fall into this path type are kaarrella, kierrella, kiemurrella, puikkelehtia and mutkitella. I will begin by taking a closer look the verb kierrella, which is the most fre quent and prototypical verb proling irregular shape. Consider example 6. (6) Kiertelin viime syksyna last autumn-ess I circled in the bear areas last autumn. karhualueilla. bear area-pl.adess

In example 6 above, the TR moves inside the LM along a twisted path. The LM is conceived of as an open spatial setting and the TRs path is

Lexicalizing indirect path: Focus on Finnish motion verbs


located inside it. The shape of the path is motivated by the meanings of the verb stem and the derivative axes. The verb kierrella has the same root as the adjective kiero, which means indirect, wrong, twisted (Finnish Etymological Dictionary SSA s.v. kiertaa).10 As pointed out by Pajunen (2001: 78), verb roots are often used to express descriptive aspects of events, such as the manner or the path of motion. This is clear in many VIPs which indeed manifest the shape of the paths in the conventional imagery attached to the verbs root. For example, since the verb kierrella has an adjective root kiero, the shape of its path is bound to contain at least some phases that can be perceived as being indirect, wrong or twisted. Thus, if a certain path does not have any twisted-like parts, a motion along it certainly cannot be referred to by the verb kierrella. Kierrella is only one example, but other VIPs, for instance verbs pyortaa bend, curl and risteilla criss-cross, manifest the same strategy. What, then, links together the verb roots meaning and the shape of the TRs path? My suggestion is that there is a mental scanning process and metaphorical understanding involved in the events conceptualization. As Janda (2008: 8) points out, in Cognitive Linguistics metaphor is seen as a pervasive and necessary component of linguistics meaning. I suggest that in the conceptualization of the VIP the conceptualizer rst mentally traces along the TRs path (scans through it). After that, the conceptualizer discovers that the path shape is similar to the physical shape expressed by the verb roots referent (or similar to the mental images associated to the roots meaning). This second part of the conceptualization process fulls the criteria of metaphorical comparison. For instance, in the case of kierrella, the shape of the TRs path is compared to the meanings indirect, wrong, twisted, i.e. to the semantics of the verb root. This then provides the understanding of the indirectness of the paths shape. Hence, the shape of the Finnish VIPs path is basically inuenced by a metaphorical conceptualization of the verb roots referent. If the mental scanning process with subsequent metaphorical conceptualization was not presumed, it would be dicult to explain why the roots of these motion verbs represent the same semantic type. These verbs are prototypically derived from nouns and their roots refer to entities with particular shapes. I presume that the scanning procedure is similar to mental processes involved in subjective (ctive) motion, as in the sentence The fence goes from the plateau to the valley (Talmy 2000a: 99). Of course, the above

10. This is true even though kierrella has been derived from the verb kierta and a not directly from the original root kiero.


Jari Sivonen

cases depict an overtly objective indirect path of motion, but the conceptualization strategy may nevertheless be the same. However, the verb root is not the only semantically important element here. Finnish has a very complicated and productive derivation system, and many of its axes are polysemous. Finnish derivational axes appear in dierent allomorphs due to the consonant gradation system and other numerous sound alternations. Moreover, the basic form of the Finnish verb, the 1st innitive, is not necessarily identical to the root to which derivational axes are attached, nor is it necessarily identical to the verb stem, the form to which inectional axes are attached. As KangasmaaMinn (1993: 16) argues, aspectual meanings, particularly Aktionsart, are often expressed by derivational means in Finnish. This is why derivation greatly inuences the shapes of VIPs. Basically, the stem of a VIP consists of a root which is accompanied by a derivational ax or by a successive or compound derivational ax. The shape of the VIPs path then emerges from the semantic contribution of both the verb root and the derivational morphemes. For example, the verb kierrella contains the derivative ax -ele-, which has a frequentative meaning in Finnish (Hakulinen 1979: 258). It was suggested by Wiik (1975: 155157) and later by Kytomaki (1992: 247) that derivative axes which have this kind of meaning imply randomness. In this case, what follows is that the meaning of the verb root together with its derivative ax constitutes a verb stem which semantically motivates the shape of the path. In other words, the path contains indirect and twisted phases which, in the mental scanning process of the conceptualizer, are conceived of as following each other randomly. This randomness aects the path in two ways. First, short straight (direct) phases are allowed. In other words, the frequentativity expressed by the derivative ax inuences the roots semantic content so that indirectness is not a constant feature of the path, but rather a recurring but irregular feature of it. This means that the path does not need to curve all the time. Second, the path may also cross itself, but this is only a possibility that the verb allows, not a compulsory characteristic, as is the case with verbs proling Paths of Crossing Movement (3.5.). Consequently, at least some paths suitable for kierrella might be characterized as looped paths, a term used by Talmy (2000a: 164). Similar features apply to the other verbs expressing Paths of Irregular Shape. As regards the semantic motivation of the path shape, the verb kaarrella is even more transparent than kierrella. Kaarrella is a denominal verb derived from the noun kaari arc, arch. The verb includes two derivative axes, -t- and -ele-. The ax -t- (originally -ta-) has a meaning of

Lexicalizing indirect path: Focus on Finnish motion verbs


factivity, a term used by Finnish scholars for derivational axes that roughly mean that the compound stem of a verb proles an entity expressed by the noun root (Hakulinen 1979: 258, 287288). In the case of the kaari-based kaarrella, the ax adds to the noun root (kaari) the meaning of making arcs into the TRs path. The frequentative ax -ele-, once again, expresses the fact that these arcs are placed randomly and iteratively. The verb kiemurrella has the same deritative axes (-t- -ele-) and the root kiemura, which has a meaning of convolution, twist (SSA s.v. kiemura). What is dierent here compared to kierrella, however, is that the root kiemura inuences the derived motion verb, so that no straight phases are likely to appear in the TRs path. Etymological explanations suggest that puikkelehtia is derived from the verb stem pujoa, which roughly means to go through a tight interstice by wriggling ones body forward (SSA s.v. puikkia, pujoa). The verb has a combined derivative (reexive) ax -elehti-, which refers to momentarity and frequentativity (Hakulinen 1979: 274). The joint semantic impact of the verbs root and derivative elements is that puikkelehtia contains randomly lateral laps, which appear as if they were suddenly poking out when compared to linear progression. The denominal verb mutkitella also has the frequentative ax -ele-, which is attached to the ax -i- expressing continuity and the causative ax -tt- (originally -tta-) (Hakulinen 1979: 260, 268269, 277). The noun root mutka turn, bend, curve has a major contribution to the shape of the TRs path. When associated with the derivative axes, it is quite understandable that the verb proles progression which includes a series of curves. However, these curves are placed randomly, and their size and steepness vary a lot. In addition, it is also possible or even likely that the TRs path contains short direct phases, which are followed by the turns. Because mutkitella allows this much variation in the TRs path, while at the same time proling a path that includes observable and distinct indirect phases, turns and corners, it can be considered to be a prototype of all Finnish VIPs. Mutkitella is also suitable for illustrating the appearance of Finnish Paths of Irregular shape, as shown in Figure 2. The path illustrated in Figure 2 is also more or less similar to the paths proled by the verbs kaarrella, kiemurrella, puikkelehtia and kierrella. However, it must be emphasized that these paths are not semantically identical. Rather, each verb proles a dierent path so that each path has a distinctive feature of its own. For example, the verb kiemurrella emphasizes that there are small randomly placed convolutions in the path, whereas the verb puikkelehtia proles sharper lateral transitions.


Jari Sivonen

Figure 2. The prototypical path of Irregular Shape in Finnish. The picture illustrates characteristics of the path proled by the verb mutkitella meander

3.2. Regularly shaped paths, verb roots and derivation Paths of Regular Shape have three main characteristics. First, they contain successive round or arc-like phases. Second, the TRs lateral movements create a certain observable pattern. And nally, this pattern appears cyclically in the path. In Standard Modern Finnish the verbs kiertaa, luikerrella, luovia and pujotella prole this kind of indirectness. I will rst consider the verb luikerrella to wriggle. Consider example 7. (7) Kaarme luikertelee ruohikolla. snake-nom lawn-adess There is a snake wriggling on the lawn. The general path represented in example 7 includes numerous indirect phases that follow one another. This shape is not directly motivated by the meanings of the formal elements of the verb, because the root of the verb luikerrella is neither transparent nor well-known. However, evidence from old Finnish folk dialects supports the idea that the component luik refers to optimal skiing conditions or generally to fast motion. The prototypical path proled by luikerrella forms a distinct, observable shape, which may also be called circular in Hawkings (1985: 85) terminology. Figure 3 illustrates the prototypical path proled by luikerrella. In this case, the paths shape cannot be explained by the meaning of the verb root. But another motivation can be found. The verb luikerrella is

Lexicalizing indirect path: Focus on Finnish motion verbs


Figure 3. The prototypical path of Regular Shape in Finnish. The picture illustrates characteristics of the path proled by the verb luikerrella wriggle

often used for referring to a snakes progression on the ground; this is an old meaning of the verb that can be traced to the Baltic-Finnish proto-language that was spoken on the east coast of the Baltic sea approximately two thousand years ago (SSA s.v. luikkia). The circular shape of the path corresponds to the average mental image of, for example, a wriggling snake moving forward on the ground. Consequently, the shape of the TRs path resembles a typical snakes path, even though in Modern Finnish the TR does not need to refer to a snake or any other reptile. The verbs derivative axes, -t(a)- and -ele-, express factivity and continuity (Hakulinen 1979: 258), which can be metaphorically conceptualized as a path with semicircular shapes, similar to a snakes progression. The other three verbs in this group have similar, but not identical paths. This is because each lexeme species its path in a slightly dierent way. For instance, when compared to kierrella (discussed in the previous section), the verb kiertaa has a dierent path in the sense that it is com posed of semicircles which are attached to each other and which partly overlap. What follows is that the TRs path does not lead from one place to another as in the paths proled by the other verbs in this group, but remains in the location expressed by the LM. Certain paths proled by kiertaa can also be categorized as circumferential, as in the transitive sentence Kierran kaappia ja tutkin sita I am going around the closet and inspecting it. Circumferentiality is a concept used by Svorou (1994: 152)


Jari Sivonen

to refer to situations where the LM is treated as a spherical object, either because of its shape or because of a path a moving entity follows when it starts from a specic point and moves along the boundaries of the LM till it reaches the same point. However, Svorous view can perhaps be broadened by suggesting that a path is also circumferential when there is a separation between the TR and the LM, and when the TR makes more than one round. According to this view, a circumferential path requires only that the TR starts and nishes (or passes) its route in the same location, and the LM is an entity which is thus surrounded by the TRs path. However, during the process the TR does not need to be in constant contact with the LMs boundaries. With this specication, the concept of circumferential path is also more in line with the idea of a Closed Path proposed by Talmy (2000a: 268). The third lexeme proling Path of Regular Shape in Modern Finnish is luovia. It is an old Indo-European loan word, probably based on the Scandinavian stem lof, which means roughly to sail windward (SSA s.v. luovia). The fact that the stem is part of the shipping and navigation terminology accounts for the path it proles. The verb has a derivative ax -i- containing the meaning of continuity (Hakulinen 1979: 261). If we compare it to luikerrella, we notice that although their paths resemble each other, luikerrella diers in one respect: as is shown in Figure 3, it contains more circular and lateral curves of roughly the same size, which occur one after another. In other words, luovia allows more variation into the TRs path, and hence, the shape of the path it proles may contain gentler turns, while the length and the angle of the lateral movements are more random. However, the path still forms an observable and predictable pattern, even though accurate shapes of individual curves are not xed. The last verb in this group is pujotella, a deverbal verb that has a root pujoa, assumed to be the same as in the verb puikkelehtia. Pujotella has a causative derivation ax -tta- (Hakulinen 1979: 277), which is attached to the reexive-frequentative ax -ele- (Hakulinen 1979: 260). The ax brings about the meaning of progressing steadily through tight interstices. These interstices are conceptualized as ctive gates through which the TR jams or makes its way. What is then relevant in the state of aairs referred to by verb pujotella is the presence of entities that can plausibly be conceptualized as barriers, or real or ctive gateposts. It is also necessary that these barriers are placed so that the TR needs to take lateral shifts in order to be able to approach each barrier. A prototypical situation which meets these criteria is slalom skiing. A slalom slope provides the TR with

Lexicalizing indirect path: Focus on Finnish motion verbs


the conditions where progression via a path proled by pujotella is quite natural. However, it is important to notice that even though slalom skiing is a typical context for the verb pujotella in Modern Finnish, it is certainly not the only one. The TR may, for instance, move in a room lled with chairs or other objects in a way that makes the use of pujotella appropriate. A good example is Mies pujotteli tuolien valista A man navigated his way round the chairs. 3.3. Scope of attention: Focus on global path as a single turn The main characteristics of the paths in this group include two features. First, the paths form only one indirect line. Second, this form comprises the whole path when observed from above in a two-dimensional space. Thus, these paths are inherently dierent from all the other path types, since they constitute a sort of global curvature, a term used by Bohnemeyer (2003: 108109.) The Finnish verbs kaartaa, kiertaa, koukata and pyortaa belong to this group. Since kaartaa and kiertaa have already been discussed, I take a closer look at the verbs koukata and pyortaa. Consider examples 8 and 9. (8) Partio koukkasi vihollisen selustaan. patrol-nom enemy-gen rear-illat The patrol outanked into the enemy rear. (9) Tulijat pyorsivat takaisin kaytavalle. comer-pl.nom go back corridor-allat The comers went back to the corridor. The verb koukata is denominal, and its root is the noun koukku hook. The verb has also a derivative ax -a-, which has a translative meaning in Finnish (to make into what is expressed by the root) (Hakulinen 1979: 281). Therefore, the verb koukata can be understood as having the meaning transform into a hook, which in this case is metaphorically conceptualized as make a path that resembles a shape of a hook. However, the verb does not specify the details of this virtual hook, which explains why the distinct paths expressed by koukata vary a lot. Nevertheless, a common feature of all the paths of koukata is that the TR makes one sudden and sharp lateral turn and penetrates into a location expressed by the LM. In Cognitive Semantics it is widely acknowledged that conventionalized imagery plays an important role in the emergence of linguistic meanings.


Jari Sivonen

Langacker (1987: 39, 1991: 294) argues that linguistic structure embodies conventional imagery. This imagery is involved in grammatical structure (e.g. which participant is chosen to act as a syntactic subject) and, it is tempting to say, even more in lexical structure. This explains why the shape of the path proled by koukata might be considered sharper than any other paths expressed by Finnish VIPs. This is due to the conventional imagery attached to the root koukku hook. Hooks typically have some sort of a sharp turn in their shape, and it is precisely this shape that makes a hook. It is then quite plausible that the path proled by the denominal verb koukata gains its shape from the mental image of a concrete, physical hook, and therefore includes this sharp turn. Consequently, koukata can be seen as the most radical Finnish verb of indirect path. This becomes clear when koukata is compared to the verb kaartaa (see above). It is more than likely that the physical shape of any arc is gentler than the shape of any concrete hook. Thus, it makes sense to argue that the shape of the path proled by hook-based verb koukata is conceived of as being sharper than that of kaartaa. The same is true when koukata is compared to other Finnish VIPs, which supports the status of koukata amongst Finnish VIPs as a verb proling the sharpest turn. The other verb to be discussed here is pyortaa, which is particularly interesting because it seems to have no direct equivalent in English. The meaning of pyortaa is roughly curl back to the starting point. Like koukata, pyortaa is a denominal verb with the root pyora wheel. The verb has a derivative ax -ta-, which expresses factivity in Finnish (Hakulinen 1979: 288), and in this case, proles a wheel-like shape of the TRs path. In contrast to koukata, which as we saw proles the most precipitous path, pyortaa seems to prole the gentlest path. This is because the images attached to its path are based on the shape of a wheel. In Figure 4, the verb pyortaa functions as an example of verbs proling a Single Turn. As is shown in Figure 4 by means of the straight broken line, in the process of pyortaa, there is a strong sense that the TRs original aim is to progress linearly. It is worth noting that even though forward motion is generally presupposed to be rectilinear and that the meaning of all VIPs can to some extent be associated with this connotation, the prole of the verb pyortaa expresses such presumed forward motion most saliently. Another important thing to bear in mind is that the TR may turn towards the LM, which expresses the TRs initial starting point, from both sides, and not necessarily from the right side, as shown in Figure 4. In other words, what is specied then, is only the characteristic round-shape of the

Lexicalizing indirect path: Focus on Finnish motion verbs


Figure 4. The prototypical path of Single Turn in Finnish. The picture illustrates characteristics of the path proled by the verb pyortaa curl back to the starting point

TRs path. Similar features characterize the other verbs of this type. In the processes conceptualized as koukata and kiertaa, for example, the TR may turn either left or right. 3.4. Metaphors based on the verb root referents movement: Paths of back and forth The paths which exhibit the TRs back-and-forth movement have three characteristic features. First, they are based on the TRs motion between two locations. Second, the motion includes several successive and individual legs that are placed between the turning points. Finally, the route between these turning points is more or less identical during all journeys. In Modern Standard Finnish, verbs sahata and sukkuloida specify their paths in this way. Prototypical examples are given in 10 and 11. (10) Mies sahaa makea ylos alas. man-nom hill-part up down There is a man running up and down the hill. (11) Myyntiedustaja sukkuloi Oulun ja Raahen valilla. sales agent -nom Oulu-gen and Raahe-gen gap-adess A sales agent is shuttling between Oulu and Raahe. The root of the verb sahata is the noun saha saw. The ax -a- has an instrumental meaning (Hakulinen 1979: 287, 290), and the combination of the root and the derivative ax give the meaning to saw. However,


Jari Sivonen

Figure 5. The prototypical path of Back and Forth in Finnish. The picture illustrates characteristics of the path proled by the verb sahata move back and fort like a saw

as a motion verb, it has a dierent meaning. For instance, the situation described in example 10 has the TR moving back and forth on a hill between certain locations, which act as the TRs turning points. Even though there is no need to explicate the turning points (as in example 10), they are no doubt an important part of the predication of the verb sahata. Figure 5 illustrates a TR path that is prototypical for this subtype, and the squares represent the TRs turning locations. What is interesting here is the conceptualization strategy. As we saw earlier, in many cases, the shape of the TRs path is motivated by metaphorical conceptualization of the verb root referents shape. In other words, the shapes of the paths resemble the shapes of the verb roots referents. Here the paths shape is not based on that of any physical saw, but rather on the metaphorical conceptualization of the prototypical motion of a saw, when it is used as a tool for cutting. The TRs path then resembles the prototypical back-and-forth motion in a process of concrete sawing of, for example, timber. The other verb in this path type is sukkuloida; it has been derived from the noun root sukkula shuttle and has the derivative ax -i-, which has the meaning of essentiality. The concept of essentiality is used in Fennistics to imply that the TR of a verb process acts as an entity that is expressed in the verbs noun root. (Hakulinen 1979: 282283). In this case, this means that the TR acts as a shuttle. Here again the shape of the TRs path can be understood as a metaphor where the TRs path is conceived of as motion of the verb roots referent. This particular meaning of sukkuloida proles a process where the TR moves between two locations (for example the cities of Oulu and Raahe in example 11), and therefore it belongs to this group of path types.

Lexicalizing indirect path: Focus on Finnish motion verbs


3.5. Criss-cross and shuttle: Paths of crossing movement The nal group of Finnish VIPs includes verbs risteilla criss-cross and 11 These paths have the following three main characsukkuloida shuttle. teristics. First, the path describes the TRs motion, prototypically between several locations. Second, these locations are situated at dierent places inside the LM. Finally, and most importantly, the TRs motion forms a jagged path (Talmy 2000a: 164), which crosses itself at least in one junction, but prototypically at many points. Example 12 below illustrates the use of the verb risteilla, which I analyse in more detail. (12) Vene risteilee jarvella. scull-nom lake-adess There is a scull cruising on the lake. The verb risteilla has the noun root risti cross and the -ile- derivative ax, which has a factive-frequentative meaning (Hakulinen 1979: 294). As regards motion verbs in particular, factive-frequentativity is a concept which means that the verb proles a process where the TRs path not only resembles the shape of the entity that the verbs nominal root expresses but is repeated several times one after the other. In other words, as regards the motion verb risteilla in particular, the TR guratively makes several crosses so that the path includes several cross-shaped phases. Figure 6 illustrates a prototypical path of Grossing Shape proled by risteilla.

Figure 6. The prototypical path of Crossing Movement in Finnish. The picture illustrates characteristics of the path proled by the verb risteilla to criss-cross 11. The verb sukkuloida is polysemous and proles both path of Back and Forth and path of Crossing Movement.


Jari Sivonen

In Figure 6, the dotted squares stand for the TRs turning points. These intermediate stopping points are visited by the TR and specied in the predication of the verb, even though they do not need to be explicated in the verbal context in which risteilla is used, as is the case in example 12, where only the lake is mentioned. However, in order to be able to use the verb risteilla, these turning points must be located inside the LM in such a manner that the TRs progression from one turning point to another forms a path that crosses itself. Furthermore, in a prototypical case, the path is conceived of as crossing itself at a very tight angle, almost 90 degrees, which is due to the conventional imagery attached to the meaning of the verb root risti cross. What is necessary for any cross is precisely the shape where two bars or sticks are placed more or less perpendicularly to each other. So, the shape of the path proled by risteilla is once again motivated by metaphorical conceptualization of the verb root referents shape. Another feature of the path of the Crossing Movement is that it is possible, but by no means necessary, that the TR nally returns to its original starting point. For example, if the verb risteilla is used with refer ence to the act of cruising at sea, it is quite possible that the path has a circumferential shape in the sense that it ends in the starting location. However, in the sentence Parvi lintuja risteilee taivaalla (lit.) A ock of birds is criss-crossing in the sky no return to the initial starting point is proled. This is illustrated in Figure 6 by means of the broken arrow line. The verb sukkuloida also falls into this subtype when it refers to motion between several locations, as in Diplomaatti sukkuloi Euroopassa The diplomat is shuttling in Europe. However, the metaphor that motivates this kind of usage is dierent. Here the TR moves from one place to another in a location specied by the LM. Now the shape of the TRs path is not based on the verb root referents physical shape, but rather on the path of the root referents (sukkula shuttle) motion.

4. Cognitive implications Having presented the indirect path types lexicalized into Finnish motion verbs, it is perhaps useful to focus on their cognitive aspects. From the cognitive point of view, it can be argued that things which are salient in culture and in human interaction are also coded in language, both in the lexicon and in the grammar. It is therefore only natural to presume that paths lexicalized into motion verbs can reveal how humans conceptualize the motion event and how they experience it (see also Svorou 1994: 27).

Lexicalizing indirect path: Focus on Finnish motion verbs


A description of indirect paths characterizes this experience and human conceptualization of the motion event. Languages systematically pay attention to some aspects of the human existence and ignore others (Janda 2008: 26). Consequently, it is important to focus on those facets of the path which are systematically highlighted by lexicalization, because they provide an insight into the way in which we conceive of the world. If we utilize Slobins (e.g. 1991: 12) well-known idea of thinking for speaking, it can be argued that the speakers of Finnish choose such features of indirect-shaped motion that correspond to any of the ve dierent path shapes lexicalized into Finnish verbs, and only afterwards choose the appropriate verb lexeme. It is interesting to notice that, as the analysis of Finnish VIPs shows, the paths tend to have relatively gentle turns. In other words, sharp curves are rarely lexicalized. There are only three verbs in Standard Finnish (koukata, risteilla, sahata) which prole sharp turns. Another noteworthy observation is that none of the verbs proles indirectness in the third dimension, i.e. depth; only 2-dimensional space is considered. I suggest that if there is one single factor which can explain the above observations, it is anthropocentrism. The indirect paths lexicalized into Finnish motion verbs reveal the way in which language encodes the external world from the human perspective. Langacker (1987: 123) concludes that most entities are seen from the canonical viewpoint, i.e. through the eyes of the human observer. The canonical viewpoint is involved not only in the conceptualization of physical objects, such as houses, cars or snakes, but it is also important in conceiving paths. Indirect paths lexicalized into verbs of motion reect canonical paths of a human TR. I suggest that angular paths are less frequent than gentle paths simply because lexicalized paths are typical for human motion in general. Compared to sh or birds, for example, humans do not tend to make sharp turns when they move forward. However, if the Path contains sharp turns, the TR is likely to refer to a non-human entity. Consider examples 13 and 14 below.12 (13) Muurahaiset risteilevat. ant-pl.nom criss-cross-3rd-pl Ants are criss-crossing.

12. Examples 13 and 14 are strongly reduced since only the TR of the process is explicated. However, they form an interesting minimal pair which enables us to focus on the semantic dierence based solely on the meanings of the TRs.


Jari Sivonen

(14) Pojat risteilevat. boy-pl.nom criss-cross-3rd-pl Boys are cruising. As is shown in example 13, in Finnish it is possible to say that ants are criss-crossing. The unmarked reading would be that we are referring, for instance, to a forest where ants are moving along a path which crosses itself and thus forms a shape of a stereotypical cross. Example 13 would thus have a literal reading describing the motion of the ants. However, if the TR refers to humans, as in example 14, the sentence is bound to have a gurative reading. In other words, the TR (boys) is understood metonymically and it refers, not literally to the boys themselves, but to a boat or a ship on which they are travelling as passengers. In other words, encyclopaedic knowledge enters into the understanding that ants are likely to follow paths of Crossing Movement, whereas for human TRs a crisscrossing path would not be prototypical and therefore 14 gets a metonymical reading. The fact that only 2-dimensional space is needed for the description of the indirect path is also based on anthropocentrism. For humans, movement forward on a surface is the most natural and unmarked way of progression, rather than movement below or above the surface. Possible lateral movement is also performed on a surface, and it is this movement that is seen as the lexicalized path type. As was argued above, shapes of indirect paths basically reect the canonical movements of a human TR. An important thing to notice is that all shapes of indirect paths in Finnish are lexicalized not from the human point of view, but rather from above, from a bird eyes view. Still, it seems to be a fact that humans typically experience paths from the ground level and not from a bird eyes perspective. The explanation must lie in the human cognitive capacity that enables humans to mentally imagine the TRs path from above. The reason for this kind of conceptualization is likely to be functional. Dewell (2007: 410) suggests three advantages of the canonical viewpoint from above the path. According to him this perspective provides a neutral standard which is equally accessible to any conceptualizer, regardless of his/her actual location. It also allows the conceptualizer to observe other possible elements of the scene. Finally, the birds eye view enables summary-scanning of the situation and thereby an overall view on the whole path. The last reason seems most explanatory. It is very dicult to get an overall picture of the shape of a path if one is moving along it (TRs viewpoint) or if one is observing the path from the vantage point

Lexicalizing indirect path: Focus on Finnish motion verbs


of the location or from the vantage point of the object of the motion (LMs viewpoint). On the other hand, it is much easier to get a general perspective of the whole path if one is able to imagine it from above. In Finnish, there are two ways in which indirect paths can be conceived of from the birds-eye view. As Talmy (2006: 543544) argues, the two strategies involve dierent ways of locating the point at which one places ones mental eye in relation to the represented scene. One common way to lexicalize the TRs path is by using a local path. Such a path enables the conceptualizer to focus the mental eye on and thus prole only some phases of the path at a time. In other words, this strategy results in observing the TRs path from a moving proximal perspective point and with a local scope of attention (Talmy 2000a: 269). The local path pattern is used in those Finnish verbs which prole paths of both Irregular and Regular Shape (kaarrella curve, kiemurrella wriggle, kierrella circle, luikerrella wriggle, luovia tack, mutkitella meander, puikkelehtia weave, pujotella zigzag). The second strategy is to use a global path; according to Talmy (2000a: 269, 2006: 544), it involves those cases in which the path is represented from a stationary viewpoint and with a global scope of attention, because in order to discern the shape of the path it needs to be conceived as a whole. The Finnish verbs proling paths of Single Turn (kaartaa encircle, kiertaa orbit, koukata outank, pyortaa curl back to the starting point), of Back and Forth (sahata move back and fort like a saw does, sukkuloida shuttle) and of Crossing Movement (risteilla criss-cross, sukkuloida shuttle) follow this strategy. Another issue that is often emphasised in Cognitive Semantics is the relationship between the lexicon and culture. Evidence from Finnish VIPs suggests that a verb with a particularly-shaped path is introduced into the language if it is needed by the culture. For example, the Finnish sukkuloida (shuttle) is a relatively new verb and it is becoming more and more frequent as the world diminishes and the amount of travel increases: a verb that proles movement in which the path crosses itself is needed. Moreover, in addition to the path shape, sukkuloida often proles fast and very purpose-oriented motion, as in the example Diplomaatti sukkuloi Euroopassa The diplomat is shuttling in Europe. Thus, sukkuloida is suitable for referring to motion which blends crossing path shape, fast speed, and purpose-oriented motion. The need for sukkuloida can easily be seen in Finnish dictionaries. For instance, Nykysuomen sanakirja (A Dictionary of Modern Finnish), a large dictionary from the late 1950s, does not recognize sukkuloida at all, while the most recent dic-


Jari Sivonen

tionary from 2004 lists the verb and gives the meaning movement along an indirect path (KS s.v. sukkuloida). Yet it is important to bear in mind that it is not only the human cognitive way of perceiving the world that aects the lexicalized path. Also the cultural environment has an impact on the understanding and conceptualization of lexicalized paths. For instance, we could imagine that the more people live in big cities with straight streets that cut each other at sharp angels (as in square-shaped street plans), the more likely a language spoken in such an environment includes verbs proling paths that contain sharp curves and crossing shape. 5. What can Cognitive Linguistics learn from indirect paths? It is obvious that the most utilized model to analyze verb lexicalization in Cognitive Semantics is the theory introduced and developed by Leonard Talmy (1985, 1991). In this theory, Talmy presents a typology of lexicalization patterns, which has also inuenced path analysis. The main principle in this tenet is that the so-called Satellite-Framed Languages (S-languages) lexicalize Path into a participle attached to the verb and the Verb-Framed Languages (V-languages) encode the Path into the verb root. As Talmy (2000b: 222) claims:
Languages that characteristically map the core schema into the verb will be said to have a framing verb and to be verb-framed languages. On the other hand, languages that characteristically map the core schema onto the satellite will be said to have a framing satellite and to be satellite-framed languages.

Talmy (1985: 62) presents several subtypes within this dichotomy and emphasizes that any language uses only one of these types for the verb in its most characteristic expression of Motion. Finnish and other Finno-Ugric languages are usually considered to be S-languages (see, e.g., Talmy 1991: 486; Slobin 2004: 249). This is not to say that Finnish does not have verbs (or verb roots) encoding path. As pointed out by Levinson and Wilkins (2006: 18), many languages allow both types of packaging. So it is not surprising that Finnish has some verb lexemes which saliently lexicalize the TRs path and can be characterized as Motion Path-verbs in Talmys (2000a: 49) taxonomy. Recently also other languages, such as French (Kopecka 2006: 99) and German (Gehrke 2007: 258), were found to utilise both lexicalization patterns. Furthermore, Pajunen (2001: 79) points out that in a language such

Lexicalizing indirect path: Focus on Finnish motion verbs


as Finnish, no lexicalization pattern is more emphatic than the other. We can accept that the dichotomy of V- and S-languages is not clear-cut. For example, in English, which is an S-language, some verbs, such as enter, behave like V-verbs. This is explained to be caused by their Latinate roots (see Talmy 2000b: 118). However, there is no etymological dierence between Finnish verbs that follow the S-strategy and the V-strategy. Moreover, verbs of both types may be original Finnish words. What Pajunen seems to be missing though is the role of the frequency factor, which is emphasized in Talmys (1985: 62) original theory. Regardless of the variation in the lexicalization types of Path expressions in Finnish, in terms of frequency, Finnish has more S-type than V-type path expressions. According to the Frequency Dictionary of Finnish (Saukkonen et al. 1979), the most common motion verbs in Finnish are tulla (come), number 16 on the frequency list of all lexemes, menna (go), number 68, kayda (come), number 87, lahtea (leave), number 106, and nally kulkea (stroll), number 269. These verbs are of course polysemous and have meanings other than motion. Nevertheless, the point is that when they are used to express motion, they encode the path in the satellite. On the other hand, verbs which prole the path and fall into the V-pattern category are relatively rare. The most common is kiertaa (bypass, circle, circulate etc.), number 763 on the frequency lists. Thus, on the basis of frequency, Talmys insight seems to be right, since Finnish is more an S- than a V-language. Nevertheless, as the Finnish examples suggest, it is interesting that an S-language also includes some relatively rare V-verbs which encode the shape of the TRs path and, furthermore, specify it as indirect. Consider examples 15 and 16 below which are very typical in Finnish. (15) Poika menee pihalle. boy-nom go-3sg yard-allat. A boy is going out. (16) Lokki seagull-nom kiertelee circle-2sg taivaalla. sky-adess

There is a seagull circling at the sky. Example 15 shows the use of the S-strategy in Finnish: the verb menna to go is used to express motion, and the path is expressed by the accompanying adverbial ( pihalle). The example illustrates the use of the SatelliteFramed tactic for expressing the TRs path. The path is proled by the LMs case suxes, usually by the illative (15) or the allative, which are


Jari Sivonen

appropriately called the movement towards cases (cf. Karlsson 1987: 99). However, the construction in example 15 does not actually make a claim about the shape of the TRs path, and only the direction (or destination) of the motion is encoded in the satellite expressing the path. In example 16, on the other hand, the V-pattern is used and the verb expresses motion as well as the TRs path. In addition, the verb species a particular kind of indirect path. The LM is in the local adessive case and proles only the location of the TRs movement and not the path. Another interesting characteristic of some Finnish motion verbs is that the shape of the path in the V-verbs is motivated by meanings of the verb roots and derivational axes. Then what happens is that the verb root lexicalizes the main element for the shape of the path and the derivative ax species the location of this elementary element in the overall path. It also seems to be the case that when the shape of the path is lexicalized into the verb root, the verb uses the global path strategy. Those verbs which encode repetition of path elements, which is expressed by derivational axes, utilize the local path strategy. For example, in the verb kiertaa, the indirect shape of the path is due to the semantics of the verbs original root kiero (indirect, wrong, twisted), and the scope of attention is global, whereas in the derived verb kierrella the same root is accompanied by the derivative ax ele, which expresses frequentative meaning. This combination makes the shape of TRs path random, as if a number of separate indirect or twisted phases occurred in a succession along the path. This makes use of the local path strategy. Based on the evidence from Finnish, I argue that the S- and the Vlanguage models can be elaborated by analyzing in more detail verb lexicalization patterns of path in those languages which predominantly rely on the S-strategy, but also make use of the V-strategy for expressing the shape of the TRs path. The semantic role of both the verb root and the derivational axes, as well the global and local path strategies, also require more detailed analysis.

6. Conclusions In this paper I have presented a cognitive semantic description of verbs that code an indirect path. Although there exist grammatical analyses of the motion event, there are few systematic studies of the paths of motion verbs. I have tried to shed some light on the cognitive aspects of these verbs by arguing that anthropocentrism plays a central, cognitive role in

Lexicalizing indirect path: Focus on Finnish motion verbs


the lexicalization of indirect paths. Finally, I have suggested that Talmys theory of Verb-Framed and Satellite-Framed Languages could benet from a more detailed semantic analysis of VIPs. The present paper focuses on motion verbs in Finnish, and obviously more careful analyses of the shapes of the lexicalized paths in Finnish are required. Moreover, this type of research could be carried out for other languages and from a cross-linguistic perspective. Finally, what still needs to be done, is to investigate in more depth how the analysis of VIPs could enrich the cognitive approach to language, which, I am sure, it would be able to do.

Bohnemeyer, Jurgen 2003 The unique vector constraint: The impact of direction changes on the linguistic segmentation of motion events. In: Emile van der Zee and Jon Slack (eds.), Representing Direction in Language and Space, 86110. Oxford: Oxford University Press. Brown, Penelope 2006 A sketch of the grammar of space in Tzeltal. In: Stephen C. Levinson and David P. Wilkins (eds.), Grammars of Space. Explorations in Cognitive Diversity, 230272. Cambridge: Cambridge University Press. Dewell, Robert B. 2007 Moving around: The role of the conceptualizer in semantic interpretation. Cognitive Linguistics 18(3): 382415. Gehrke, Berit 2007 Putting path in place. In: Estela Puig-Waldmuller (ed.), Proceed ings of Sinn und Bedeutung 11: 244260. Barcelona: Universitat Pompeu Fabra. Hakulinen, Lauri 1979 Suomen Kielen Rakenne ja Kehitys. Neljas, Korjattu ja Lisatty Painos. Helsinki: Otava. Hawkins, Bruce Wayne 1985 The Semantics of English Spatial Prepositions. Dissertation. Series A Paper no. 142. San Diego: University of California. Janda, Laura 2008 From cognitive linguistics to cultural linguistics. To appear in: Slovo a smysl/Word and Sense (ISSN 12147915). http://hum. Read 1.10.2007. Jakel, Olaf 2003 Motion metaphorized in the economic domain. In: Hubert Cuyckens, Thomas Berg, Rene Dirven and Klaus-Uwe Panther


Jari Sivonen

(eds.), Motivation in Language. Studies in Honor of Gunter Radden, 297318. Amsterdam: John Benjamins Publishing Company. Kangasmaa-Minn, Eeva 1993 Aspektista ja sen sukulaisilmioista suomalais-ugrilaisissa kielissa . Studia Comparativa Linguarum Orbis Maris Baltici 1. Tutkimuksia Syntaksin ja Pragmasyntaksin Alalta (toim. Valma Yli-Vakkuri), 1323. Turku: TYSYKLJ 43. Karlsson, Fred 1987 Finnish Grammar. Second edition (translated by Andrew Chesterman). Porvoo: WSOY. Kielitoimiston sanakirja ( KS) 2004 Kotimaisten kielten tutkimuskeskus ja Kielikone Oy. Kotimaisten kielten tutkimuskeskuksen julkaisuja 132. Helsinki. Kopecka, Anette 2006 The semantic structure of motion verbs in French. In: Maya Hickman and Stephane Robert (eds.), Space in Languages. Linguistic Systems and Cognitive Categories, 83101. Amsterdam/ Philadelphia: John Benjamins Publishing Company. Kytoma Leena ki, 1992 Suomen Verbiderivaation Kuvaaminen 1600-luvulta nykypaiviin. Turku: TYSYKLJ 40. Langacker, Ronald W. 1987 Foundations of Cognitive Grammar. Vol. 1. Theoretical Prerequisites. Stanford: Stanford University Press. Langacker, Ronald W. 1988 An overview of cognitive grammar. In: Brygida Rudzka-Ostyn (ed.), Topics in Cognitive Linguistics, 348. Amsterdam: John Benjamins Publishing Company. Langacker, Ronald W. 1990 Subjectication. Cognitive Linguistics 1: 538. Langacker, Ronald W. 1991 Foundations of Cognitive Grammar Vol. II: Descriptive Application. Stanford: Stanford University Press. Langacker, Ronald W. 1999 Grammar and Conceptualization. (Cognitive Linguistics Research 14.) Berlin: Mouton de Gruyter. Levinson, Stephen C. and Wilkins, David P. 2006 The background to the study of the language of space. In: Stephen C. Levinson and David P. Wilkins (eds.), Grammars of Space. Explorations in Cognitive Diversity, 123. Cambridge: Cambridge University Press. Nykysuomen sanakirja ( NS) 19621964 Nykysuomen sanakirja IVI. Valtion toimeksiannosta teettanyt Suomalaisen Kirjallisuuden Seura. Porvoo: WSOY.

Lexicalizing indirect path: Focus on Finnish motion verbs


Pajunen, Anneli 2001 Argumenttirakenne. Asiaintilojen luokitus ja verbien kayttayty minen suomen kielessa. Suomi 187. Helsinki: SKS. Saukkonen, Pauli, Marjatta Haipus, Antero Niemikorpi, and Helena Sulkala 1979 Suomen kielen taajuussanasto. A frequency dictionary of Finnish. Juva: WSOY. Sivonen, Jari 2005a An Exercise in cognitive lexical semantics: the case of the Finnish motion verb kiertaa. In: SKY Journal of Linguistics 18 (2005): 311340. Sivonen, Jari 2005b Mutkia Matkassa. Nykysuomen Epasuoraa Reittia Ilmaisevien Verbien Kognitiivista Semantiikkaa. Helsinki: SKS. Suomen Sanojen Alkupera ( SSA) 19922000 Suomen sanojen alkupera [The Origin of Finnish Words]: etymologinen sanakirja IIII. SKST 556. Helsinki: Kotimaisten kielten tutkimuskeskuksen julkaisuja 62. Slobin Dan I. 1991 Learning to think for speaking: Native language, cognition, and rhetorical style. Pragmatics 1: 726. Slobin Dan I. 2004 The many ways to search for a frog. linguistic typology and the expressions of motion events. In: Sven Stromqvist and Ludo Verhoeven (eds.), Relating Events in Narrative: Typological and Contextual Perspectives. Mahwah: Lawrence Erlbaum. Svorou, Soteria 1994 The Grammar of Space. (Typological Studies in Language 25.) Amsterdam/Philadelphia: John Benjamins Publishing Company. Talmy, Leonard 1975 Semantics and Syntax of Motion. In: John P. Kimball (ed.), Syntax and Semantics 4, 181238. New York: Academic Press. Talmy, Leonard 1985 Lexicalization patterns: semantic structure in lexical forms. In: Timothy Shopen (ed.), Language typology and syntactic description III. Grammatical categories and the lexicon, 57149. Cambridge: Cambridge University Press. Talmy, Leonard 1991 Path to realization: a typology of event conation. In: Laurel A. Sutton, Christopher Johnson, and Ruth Shields (eds.), Proceedings of the Seventeenth Annual meeting of the Berkeley Linguistics Society, 480519. Berkeley: Berkeley Linguistics Society. Talmy, Leonard 2000a Toward a Cognitive Semantics. Volume I: Concept Structuring Systems. Cambridge: The MIT Press.


Jari Sivonen

Talmy, Leonard 2000b Toward a Cognitive Semantics. Volume II: Typology and Process in Concept Structuring. Cambridge: The MIT Press. Talmy, Leonard 2006 Cognitive linguistics. In: Keith Brown (ed.), The Encyclopedia of Language and Linguistics. 2nd edition. Vol. 2, 542545. Amsterdam: Elsevier Science. TYSYKLJ Turun yliopiston suomalaisen ja yleisen kielitieteen laitoksen julkaisuja. Wiik, Kalevi 1975 Suomen frekventatiivi ja kontinuatiivi. Vir (Virittaja. Kotikielen Seuran aikakauslehti.) 79: 153167.

A cognitive approach to parenthetical speech Jaakko Leino

1. Background Generally speaking, there are few propositions in linguistics which may rightfully be considered axiomatic. One candidate for an axiomatic statement in language sciences is that spoken language should be considered primary with regard to written language. This position manifests itself in many ways: spoken language existed rst, it is learned rst by infants, and so on. Therefore, it is generally accepted without further question that linguistics should take spoken language as the primary object of study (cf., e.g., Saussure 1916). However, in practice, many (if not most) studies, especially in the eld of syntax, end up using written language as primary data for their studies, and nearly all theoretical frameworks within the study of syntax and morphosyntax are tuned in to the description of written language, to the extent that they often seem unable to account for spoken data in a fruitful and comprehensive manner. At the same time, the study of dialects (and spoken language more generally) has, for the past hundred years, been practically limited to the study of phonological, morphological, and lexical phenomena. Dialect syntax has been a largely neglected area. Recently, however, interest in dialect syntax has grown rapidly, notably among generativist syntacticians, on the one hand, and language typologists, on the other (cf., e.g., Barbiers, Cornips and van der Kleij 2002; Kortmann 2004a, 2004b). An important reason for this development is the criticism, directed equally at both of these otherwise quite dierent approaches to language, which states that also non-standard varieties of languages need to be accounted for in order for the analyses to be fully adequate. What I wish to show in this paper is that cognitive linguistics, and in particular Construction Grammar (henceforth CxG, e.g. Fillmore and Kay 1996; Fried and Ostman 2005; Goldberg 1995, 2006; Croft 2001), provides useful tools for tackling the problem of bringing together spoken language and syntactic theory. To provide a concrete case in point, or a window to the theoretical ideas in question, I shall analyze (in section


Jaakko Leino

4) two parenthetical expression types which frequently occur in Finnish dialect data. Central to this discussion are not these particular expression types, but rather the notion of construction, the organization of speech, and the role of the former in the latter. 2. Constructions (and how they emerge) The trademark characteristic of CxG as originally developed consists in the insight that language is a repertoire of more or less complex patterns constructions that integrate form and meaning in conventionalized and often non-compositional (or non-analyzable) ways. Form in constructions may refer to any combination of syntactic, morphological, or prosodic patterns, and meaning is understood in a broad sense that includes lexical semantics, pragmatics, and discourse structure. In this view, a grammar consists of intricate networks of overlapping and complementary patterns that serve as blueprints for encoding and decoding linguistic expressions of all types. CxG is widely seen as a part of the Cognitive Linguistics tradition (cf., e.g., Ostman and Fried 2005) which subscribes to the usage-based model of language description and acquisition (cf., e.g., Langacker 1988; Barlow and Kemmer 2000). The basic tenet of this model is that linguistic units are shaped, in terms of both form and meaning, specically by their use and by the contexts in which they are used. In other words, language like quite a number of other skills which human beings acquire is learned by making generalizations of situations and phenomena which resemble one another. As an illustration, let us assume that a child is learning Finnish, and hears the following sentences (among others):1

1. I use the following abbreviations in the glosses: pst past tense (somewhat misleadingly called the imperfect tense in traditional Finnish grammars), 1sg 1st person singular, 2pl 2nd person plural (etc.), gen genitive case, acc accusative case, par partitive case, ill Illative case, ela elative case, ine inessive case, abl ablative case, ade adessive case, tra translative case, pass passive voice (or impersonal; for a discussion of the interpretation of the form as passive, see Leino and Ostman 2008; Ostman 1981; Shore 1986).

A cognitive approach to parenthetical speech


(1) a.

isa lammitti saunan kuumaksi father warmed sauna-acc hot-tra father warmed the sauna hot Tom Sawyer maalasi aidan valkoiseksi Tom Sawyer painted fence-acc white-tra Tom Sawyer painted the fence white puhui metsat puhtahiksi blew forest-pl-acc clean-pl-tra jaati jarvet kantaviksi froze lake-pl-acc bearing-pl-tra [the winter chill] blew forests clean, [and] froze lakes solid [i.e. such that the ice bears] (from a childrens song)



Based on these sentences, the child can make the following generalization: in the Finnish language, it is conventional to express a certain (and recognizable) type of situation or event in this manner. By comparing the sentences and the situations that they describe, it is possible to make a rather detailed generalization of the properties of this expression type: The expression type is composed of four parts or elements which correspond to dierent elements of the situation type being described: the rst element of the expression type, the subject, expresses the actor; the second element, the verb, is an expression of the deed itself; the third element, the object, expresses the undergoer or patient; and the fourth element expresses the end result of the situation. In all of the sentences in (1), the second and the fourth element have the same form, respectively: the expression of the deed ends in -i, and the expression of the end result ends in -ksi. In the third element, there are also formal similarities: in two instances, the third element ends in -n, and in two instances in -t. The expressions themselves vary, however, and none of the four elements of the expression type is constant. These properties characterize this conventional expression type of Finnish, which is known as the resultative construction. This construction was studied within the CxG framework by Palsi (2000), and its counterparts


Jaakko Leino

in other languages by, e.g., Goldberg (1995: 180198) and Boas (2003). We may present schematically in the following manner:2

Figure 1. The Finnish resultative construction

Figure 1 illustrates a rough characterization of what CxG calls a construction: a model or formula, learned by generalization, which is used as a template for novel expressions. Cognitive Grammar (Langacker 1987, 1991) calls such generalizations constructional schemas, and in more traditional approaches to grammar they are often called e.g. expression types, sentence types, basic sentence types, etc. The basic idea itself is by all means very well-established in the study of language and grammar. One crucial property of the formula in Figure 1 is that it represents certain elements in a certain constellation with one another. Yet, none of those elements is constant (aside from the translative case ending): they all vary. They are not lexically specied, but the language speaker still has some idea of what does and what does not t in any given slot in the formula. Therefore, the formula can be used both productively and comprehensively: it is productive to the extent that it allows for variation in its lexical instantiations; and it is comprehensible due to the fact that it includes information of what can be expected of each part of the whole, what the relations between the parts are, and what type of situation the structure expresses as a whole.

2. A proper analysis of the Finnish resultative construction would naturally require a signicantly more complex description. For a detailed discussion (including an attribute value matrix description), see Palsi (2000). For the pur poses of this paper, however, the characterization in Figure 1 will suce, and a detailed attribute value matrix would be more confusing than helpful.

A cognitive approach to parenthetical speech


3. The spoken language issue As pointed out at the beginning, speech is primary with regard to writing. However, the study of syntax has operated mainly with written language data and invented examples. The ow of speech seems too chaotic and unorganized to be eectively analyzed with theoretical tools created for written language. What I wish to show in what follows is that the notion of construction is useful in this respect, and that the basic mechanisms of CxG are not only applicable to spoken data, but that they also solve what I call the delimitation problem of spoken language syntax. I use the term delimitation problem to refer to the fact that delimiting the object of investigation, i.e. pointing out the relevant units of description and discussion, is often a central problem in the syntactic study of speech. This state of aairs is closely connected to the fact that whenever syntactic theories encounter unrestricted spoken language data, they also tend to encounter problems. One of the central methodological challenges in the study of speech in general and spoken language syntax in particular is delimiting the object of study, or the basic unit of description: is it to be a sentence, an utterance, a phrase, a turn, or perhaps something else? If we look at the examples in (1) and the corresponding Figure 1 from the point of view of spoken language, we are instantly confronted with the question as to how actual spoken examples can be tted together with such expression types or constructions. In written language syntax, one normally takes the clause or sentence to be the basic unit of description. As it happens, the uncomfortable question as to how we can apply syntactic rules to spoken language data (or how we can apply constructions to speech) often emerges in connection with another question related to the delimitation problem: can we use the term sentence to refer to units of spoken language? Roughly speaking, there are two types of standard answers to this question: the yes, but. . . type and the no, but. . . type. What I have chosen to do within the limits of this paper is to dismiss the yes or no part of the answer which I feel is arbitrary, and concentrate on the but part, which I feel is very interesting. I shall illustrate the problem with further examples. I claim that the following utterance contains a resultative structure very similar to that in example (1), but it is dicult to observe and point out in the ow of speech and it would certainly sound far-fetched to claim that what


Jaakko Leino

we have in (2) is a sentence which instantiates the Finnish resultative construction: (2)
se, se imellettiin se, se re. . . ruis, ruis, rukiit teht. . . tehtiim maltaiks it it malt-pass-pst it it re- rye rye rye-pl dodo-pass-pst malt-pl-tra semmottiksi ni- etta ne idatettiij ja jam. . . ste, loylytettiin etta such-pl-tra so that they germinate-pass and and then heat-pass that tuli maltaita ja sittej jauvettii ja ja sitte sittet come-pst-3sg malt-pl-par and then ground-pass-pst and and then then tehtii semmosta sita sita sita, varoa sittes semmosta sittes siitta , do-pass-pst such-ptv it-par it-par it-par wait then such-ptv then it-ill siitta rukiin idatesjauhosta sitt- etta. it-ill rye-gen germinated-our-ela then that. it was malted, that rye, then, by germinating, and then heated so that it became malts, and then it was ground and then such sort of, wait, such, was made out of the our of germinated rye.

Example (2) illustrates a problem which is frustrating for the syntactician interested in spoken language: it does not follow the rules, there are several repetitions, hesitations, etc. which undoubtedly reveal interesting details about how the sentence is processed by the speaker, but which also blur any syntactic structure be it in terms of constructions, in terms of constituent structure, or whatever which the expression may be based on. However, suitably arranged, this expression can be organized in such a manner that its essential structure becomes quite recognizable. This manner of arranging the utterance is inspired by the work of the GARS (Groupe Axois de Recherches en Syntaxe) research team headed by Claire Blanche-Benveniste, based at the University of Provence (cf. Blanche-Benveniste 1997; for an application to Finnish, cf. Duvallon 2006). The basic idea which underlies this approach is simple: all words which correspond to the same position in the overall structure are placed in the same vertical column. As a result, even such rather complex, apparently unstructured and not-very-straightforwardly proceeding expressions as that in (2) organize themselves into wholes which are surprisingly similar to constructions that have been postulated for written language. It thus appears that a Construction Grammar approach to spoken language syntax bridges, or at least greatly narrows, the apparent gap

A cognitive approach to parenthetical speech


Figure 2. Column-by-column representation of example (2)

between spoken and written language: even in very obviously nonstandard dialect speech, we nd either the same or at least very similar morphosyntactic structures as those which we nd in written language. The main dierence is that in speech their syntagmatic structure does not proceed in the same unidirectional and straightforward way as it does in written language. Speech contains repetitions, hesitations, and other processingrelated phenomena which do not show up in writing to an equal degree, but this is not a fundamental qualitative dierence of structure. Rather, it is a natural consequence of the fact that speech is necessarily processed in a here-and-now manner whereas writing allows for more planning and processing time. While it is clear that the interface between spoken and written language syntax is something that should be looked at in more detail, one can say that it should not be mystied as it often has been. The relationship between spoken and written language syntax is much more straightforward than it may seem, and we can indeed analyze speech with essentially the same machinery that we use for written language, as long as we are not mislead by the repetitions etc. in speech (and, it must be added, as long as we do not try to explain everything in terms of constituent structure).


Jaakko Leino

4. Parenthesis In section 3, I pointed out that we may or may not accept the notion of sentence as an applicable unit of spoken language, but that it is in fact arbitrary to either accept or reject the sentence as a unit of speech. What is crucial for the present discussion is not the yes or no type of answer. Rather, what is of importance is the fact that we should keep apart the notions of construction and sentence type: while all conventional sentence types may at the same time be conventional constructions, not all constructions are sentence types. In order to illustrate this, let us now look at a slightly dierent set of constructions the ones that the title of the present paper refers to. As a starting point for this discussion, it should be noted that the constructions (or expression types) that we encounter in spoken language often include not only syntactic and semantic information but also information of a pragmatic nature: information about the discourse function of the expression type or, more broadly, information about what the expression type is used for. The following example shows an expression type which somewhat resembles the Finnish resultative construction discussed above, and which is pretty frequent especially in dialect interview data. We may call it the [X] was called Y construction.3 (3) se it siell oli, there be-pst-3sg semmonev vanha such old vanaha nuottamies old There was, in Onkamo, he was called old Antti, an old seine sherman. (LA, Salla)

Onkamossa Onkamo-ine

3. The construction is especially frequent in dialect interviews because the interviews tend to contain lengthy passages concerning traditional habits, past events, and the like, in which a lot of traditional names for old tools, habits, and people come up. For the same reason, the construction almost invariably occurs in past tense: it normally refers to names or concepts relevant to past events.

A cognitive approach to parenthetical speech


As example (3) shows, this construction is a piece of explanatory information inserted in the middle of another expression: there was, in a place called Onkamo, they called him old Antti, an old sherman. This construction is typically used in a dictionary-like manner for explaining names or words assumed not to be familiar to the hearer. Thus, it has a clear function. Yet, it is seldom a separate utterance or sentence-like unit; rather, it is typically embedded in another expression. Thus, it is a parenthetical structure similar to, for example, those studied by Duvallon and Routarinne (2005).4 At the same time, this expression type is well conventionalized and quite xed: it is hardly ever broken apart in speech, and very few hesitations or other processing phenomena occur within it. There is little doubt that what example (3) illustrates is a full-edged construction. However, this construction dees the supposedly normal organization of messages into sentences (not that this makes it by any means unique, though). Still, it is a quite typical construction in the sense that it brings together a certain formal structure, a specic meaning, and a conventionalized discourse function, which all combine to form a recognizable pattern or gestalt: this expression type is really very easy to recognize in spoken data. Let us take two more examples of the same construction. Examples (4) and (5) show that, despite the compactness and conventionality of the construction, there is some variation involved.5 First, the verb varies: in most cases, it is the Finnish equivalent of the verb say, i.e. sanoa, but several other communication verbs are used as well. Secondly, the verb usually occurs in the passive form, but this is not always the case. And, thirdly, the number of overt arguments of the verb varies. The object argument is sometimes present in the parenthetical expression, e.g. sita in both exam ple (4) and (5), but more often it is left implicit. And, when the verb is in the passive voice, the subject argument is left out, while in the active voice it may be present, e.g. ne in example (5).

4. In a more traditional line of thought, structures of this kind were also studied by Ravila (1945). His pioneering work on this and related phenomena was a rare case of Finnish spoken language syntax in the mid-20th century, but it lead to little results largely due to lack of a suitable methodological apparatus for pursuing his ideas. 5. For suggestions on how to cope with variation in the CxG framework, see Leino and Ostman (2005).


Jaakko Leino

(4) niin sita sellaahella puulo. . . , so it-par such-ade

jo. . . s- oli, kolomella jalaalla se loukkuhoito ja jo- it be-pst-3sg three-ade leg-ade it brake.thing and sittes siina karentila ja, sittes silla kannella paskitetahan then there and then it-ade lid-ade slam-pass ninku oli nain tallaaset like be-pst-3sg like.this such-pl niinku huulet siina loukus like lip-pl there brake-ine So one would, with such a wood br-, they called it a brake, it stood on three legs that brake thing, and then there was a handle and then one would slam with the lid, it was like this, it had sort of like lips, the brake. (LA, Peraseinajoki) (5) taekka se tul tuolta or it come-pst-3sg there-abl Etela-Suomesta, pae se, Tiitine South-Finland-ela around it Tiitinen kun when

enne aekaa ku se ol siella Vesjarvella paen. before time-ill when it be-pst-3sga there-ade Vesijarvi-ade around Or [perhaps] it came from Southern Finland, this Tiitinen, whom they called the old man from Vesijarvi in the old days since it [or he] was around Vesijarvi. (LA, Mikkelin mlk)

6. The Finnish word loukku means both trap and ax brake (for processing linen).

A cognitive approach to parenthetical speech


We may conclude that the [X] was called Y construction is a parenthetical construction whose function is normally to explicate some part of the utterance, typically a dialect word, a proper name, or some other concept or detail assumed not to be familiar to the hearer. Signicantly, it does not neatly correspond to such traditional notions as NP, clause, sentence, etc. In this respect, it is by no means unique as a construction: there are plenty of similar parenthetical expression types in speech, and it is possible to nd dierent kinds of groups among such expressions. One characteristic group of parenthetical expression types are explicative parenthetical constructions, which explain, oer further information, etc. i.e. constructions by means of which the speaker provides the hearer with an explanation or comment concerning the state of aairs under discussion, or constructions by means of which the speaker more generally realigns himself with regard to the hearer or temporarily shifts attention from the universe of discourse to the speech setting. The [X] was called Y construction is naturally one member of this group, but there are other ones as well. Let us take another example. In (6), there is a sequence marked by the word nimittain namely: (6) oha be-3sg-clit siina kaloja, there sh-pl-par nyt tassa now here

joka on, nyt on iso jarvi tassa. which be-3sg now be-3sg big lake here. There are sh here now all right, namely in the lake, which is a big lake, here. (LA: Joutsa) In Finnish grammars, the word nimittain is traditionally classied as a subordinating conjunction (if it is considered worth mentioning). On the other hand, Ikola, Palomaki and Koitto (1989: 52) specically argue that it should not be considered a conjunction however, they do not say what exactly it should be called.7 Be that as it may, if we look at how the word
7. Historically, the word is a converb form of the verb nimitta call, roughly a meaning by calling [someone/something by a name]. Synchronically, however, it is more properly analyzed as a grammaticalized connective.


Jaakko Leino

nimittain is used in speech, we nd a close connection with the [X] was called Y construction, and we note that nimittain is used in a rather similar manner as sanottiin: it typically begins an explanatory parenthetical section. Example (6) also shows a typical property of parentheticals: the word now is repeated both before and after the parenthetical insertion. This phenomenon repeating some element of the surrounding structure is characteristic of parentheticals, and it may also be found in the [X] was called Y examples above. In fact, these two constructions are so similar that they occasionally even occur together as explanatory parenthetical expressions with both of these markers (i.e. sanottiin and nimittain) being present:8 (7) kasamaks, sanottiin kasama-tra say-pass-pst tata viarree puuta, this-par twisted-par wood-par

tialla nimitettii nimittai. here call-pass-pst nimittain They called this twisted wood kasama, namely they called it that around here. (LA: Kiihtelysvaara) (8) ennev before vanhaan pieti niiko old-ill keep-pass-pst like ja, jota and which-par luttonem luttonen

sanovat luttoseks nimittae. say-3pl luttonen-tra nimittain In the old times people used to have a luttonen [a device for processing linen], namely they called it luttonen. (LA: Konginkangas) Thus, we have two constructions which closely resemble each other. Neither one of the two constitutes a complete sentence (at least not necessarily) or even a sentence-like structure. They are both inserted in the middle of another, more sentence-like structure, and they both provide some sort of an explanation or clarication for some element in the surrounding
8. The co-occurrence of [X] was called Y and nimittain may be thought of as redundant: if they both function as markers of explicative parentheticals, why use both? It may be the case that nimittain in these cases serves to strengthen the marking of the parenthetical. This may be due to the fact that the verb form sanottiin was said/called is a rather weak marker of parenthesis in terms of cue validity, since the verb sanoa say is very frequently used, even in passive voice, also as a regular verb of communication.

piettii keep-pass-pst

A cognitive approach to parenthetical speech


structure. And, all in all, they both show a number of properties which are typically observed in parenthetical expressions. One crucial consequence of the existence of such conventional parenthetical constructions is that we must keep apart the notions of construction and sentence type, as was suggested above: we may with full condence call the X was called y expression type a construction, but it is not really a sentence type.

5. Constructions and spoken language Let us now consider the consequences of this discussion for the CxG framework and the applicability of the notion of construction. One thing that I have tried to show is that constructions are indeed well suited for the syntactic description of spoken language for several reasons: They are exible enough to be practical, i.e. to t together with actual speech data. Since constructions combine form, meaning, and function, they are patterns suciently recognizable to be credible and plausible be it intuitively, psychologically, or in some other sense. They are suciently explicit to be delimitable and describable in the ow of speech. Importantly for research on the syntax of spoken language, constructions oer a tool for delimiting the object of study and the basic unit of description, as I pointed out. Furthermore, they also provide means for pointing out the same thing, so to say, in two dierent language varieties or even two dierent languages such as the resultative construction in, say, Finnish and English, or in standard Finnish and in dierent dialects. On the other hand, spoken language is a perfectly plausible object of study for Construction Grammar: we often nd the same, or at least essentially similar, constructions in speech as we do in written language (although this may not seem to be the case at rst glance). One just needs to acknowledge that, with regard to constructions, speech proceeds through overlaps and repetitions. Another thing that may be pointed out is that the present discussion tells us something about the way that language is processed as we speak. It does not necessarily proceed straightforwardly, one sentence-sized package at a time. Rather, speech winds and turns around and back and forth as is necessary; it contains parentheses and other additions, repetitions and


Jaakko Leino

hesitations, etc. And, of course, one important conclusion is that the way we describe syntax should be able to cope with the fact that this is the way language works by nature: if we accept the claim that speech is primary with regard to writing, then spoken language syntax cannot be a derivative of a more primary written language syntax, and syntactic theories cannot be limited to written language phenomena.

6. Theoretical and methodological implications Let us conclude by briey considering the role of constructions in speech processing from a dierent point of view. What can be said following the above discussion is, of course, merely a set of assumptions and preliminary hypotheses to be looked at in more detail in subsequent research. However, it does seem obvious that constructions, or elements that very closely resemble constructions, play a central role in speech processing. When we start to talk, we apparently have an idea of the overall pattern of the utterance that we are going to produce: a sentence type, a corresponding construction, etc. We then proceed to ll up the open slots in that construction with suitable smaller elements: words and phrases. As was shown in section 3, the same slot can be lled over and over again: this results in overlaps in the same column, as the examples showed. And we can also interrupt the overall pattern with other constructions which do not t into the original one: these show up as parenthetical expressions, as discussed in section 4. Essentially the same observation was of course made before in other contexts. The idea of the speaker starting with a plan of the overall pattern that they are going to produce is not very dierent from the notion of projection in conversation analysis (CA; cf. e.g. Sacks 1992 a, b; Goodwin and Heritage 1990; Heritage 1995). As Scheglo (1987: 71) puts it, the units that turn construction employs (e.g. lexical, phrasal, clausal, sentential constructions) [ ] project, from their beginnings, aspects of their planned shape and type. CA uses a dierent terminology (instead of constructions, CA speaks of turn-constructional units) and approaches the phenomenon from a somewhat dierent (notably more interactional) point of view, but the main idea is indeed very similar. With regard to cognitive linguistics in general, and CxG in particular, the crucial conclusion is that the online process of generating utterances is organized by constructions (or constructional schemas): conventional, experience-based models of what kinds of expressions and combinations

A cognitive approach to parenthetical speech


we tend to use in our language, and of what kind of meanings we habitually express with a given kind of expression; in other words, what a given construction means and what it is used for. These models, grammatical constructions, are equally useful and signicant for the analysis of spoken language as they are for investigating written language. And, to conclude, we may eventually answer the question posed in section 3 by saying that although the notions of clause and sentence may or may not be useful in analyzing the syntactic structure of spoken language, the notion of construction certainly is. References
Barbiers, Sjef, Leonie Cornips and Susanne van der Kleij (eds.) 2002 Syntactic Micro-variation. Amsterdam: SAND. (Available in electronic form at Barlow, Michael and Suzanne Kemmer (eds.) 2000 Usage-Based Models of Language. Stanford: Center for the Study of Language and Information. Blanche-Benveniste, Claire 1997 Approches de la Langue Parlee en Francais. Paris: Ophrys. Boas, Hans C. 2003 A Constructional Approach to Resultatives. Stanford: Center for the Study of Language and Information. Croft, William 2001 Radical Construction Grammar. Oxford: Oxford University Press. Duvallon, Outi 2006 Le Pronom Anaphorique et lArchitecture de lOral en Finnois et en Francais. Paris: ADEFO/LHarmattan. Duvallon, Outi and Sara Routarinne 2005 Parenthesis as a resource in the grammar of conversation. In: Hakulinen, Auli and Margret Selting (eds.), Syntax and Lexis in Conversation. Studies on the Use of Linguistic Resources in Talkin-Interaction, 4574. Amsterdam: John Benjamins. Fillmore, Charles and Paul Kay 1996 Construction Grammar. CSLI Lecture Notes. Manuscript. Stanford: Center for the Study of Language and Information. Fried, Mirjam and Jan-Ola Ostman 2005 Construction grammar: A thumbnail sketch. In: Fried, Mirjam and Jan-Ola Ostman (eds.), Construction Grammar in a CrossLanguage Perspective, 1186. (Constructional Approaches to Language 2.) Amsterdam: John Benjamins. Goldberg, Adele E. 1995 Constructions. A Construction Grammar Approach to Argument Structure. Chicago: The University of Chicago Press.


Jaakko Leino

Goldberg, Adele E. 2006 Constructions at Work. The Nature of Generalization in Language. Oxford: Oxford University Press. Goodwin, Charles and John Heritage 1990 Conversation analysis. Annual Review of Anthropology 19: 283 307. Heritage, John 1995 Conversation analysis: Methodological aspects. In: Quastho, Uta M. (ed.), Aspects of Oral Communication, 391418. Berlin: Walter de Gruyter. Ikola, Osmo, Ulla Palomaki and Anna-Kaisa Koitto 1989 Suomen Murteiden Lauseoppia ja Tekstikielioppia. Helsinki: SKS. Kortmann, Bernd 2004a Why dialect grammar matters. The European English Messenger XIII: 2429. Kortmann, Bernd (ed.) 2004b Dialectology Meets Typology. Dialect Grammar from a Crosslinguistic Perspective. (Trends in linguistics. Studies and monographs. Vol. 153). Berlin: Mouton de Gruyter. LA Collections of the syntax archive. Computer corpus. Turku: University of Turku, department of Finnish and General Linguistics/Helsinki: Research Institute for the Languages of Finland. Langacker, Ronald W. 1987 Foundations of Cognitive Grammar I: Theoretical Prerequisites. Stanford: Stanford University Press. Langacker, Ronald W. 1988 A Usage-Based Model. In: Rudzka-Ostyn, Brygida (ed.), 1988: Topics in Cognitive Linguistics, 127161. (Current issues in Linguistic Theory 50.) Amsterdam: John Benjamins. Langacker, Ronald W. 1991 Foundations of Cognitive Grammar II: Descriptive Application. Stanford: Stanford University Press. Leino, Jaakko and Jan-Ola Ostman 2005 Constructions and variability. In: Fried, Mirjam and Hans Boas (eds.): Grammatical Constructions: Back to the Roots, 191213 (Constructional Approaches to Language 4.) Amsterdam: John Benjamins. Leino, Pentti and Jan-Ola Ostman 2008 Language change, variability, and functional load. Finnish genericity from a constructional point of view. In: Leino, Jaakko (ed.), Constructional Reorganization, 3754. (Constructional Approaches to Language 5.) Amsterdam: John Benjamins. Ostman, Jan-Ola 1981 The Finnish passive and Relational Grammar. CLS 17: 286 294.

A cognitive approach to parenthetical speech


Ostman, Jan-Ola and Mirjam Fried (ed.) 2005 Construction Grammars. Cognitive grounding and theoretical extensions. (Constructional approaches to language 3.) Amsterdam: John Benjamins. Palsi, Marja 2000 Finnish resultative sentences. SKY Journal of Linguistics 13: 211250. Ravila, Paavo 1945 Lauseeseen liittyneet irralliset ainekset. Virittaja 49: 116. Sacks, Harvey 1992a,b Lectures on Conversation. Vol. 1 and 2. Edited by Gail Jeerson. Oxford: Blackwell. Saussure, Ferdinand de 1916 Cours de Linguistique Generale. Lausanne: Payot. Scheglo, Emanuel A. 1987 Recycled turn beginnings: A precise repair mechanism in conversations turn-taking organisation. In: Graham Button and John R.E. Lee (eds.), Talk and social organisation, 7085. Clevedon, England: Multilingual Matters. Shore, Susanna 1986 Onko Suomessa Passiivia [Is there a passive in Finnish?]. Helsinki: SKS.

Part IV.

The pragmatic context

Using RST to analyze subjectivity in text and talk Kirsten Vis, Wilbert Spooren and Jose Sanders
1. Introduction Rhetorical Structure Theory (RST; Mann and Thompson 1988) is an elegant and much used model for the analysis of coherence relations in discourse. It is one of the rst approaches to systematically distinguish between object versus subject oriented coherence relations, highlighting the intersubjective character of discourse by dening subjective relations as presentational, i.e. expressing the particular relation between Speaker and Addressee. In presentational relations, such as Claim-Evidence, the Speaker is foregrounded as the conceptualizing subject in the SpeakerHearer relationship. In this article, we apply this inherently intersubjective RST-model to look for evidence for the conversationalization hypothesis (Fairclough 1994; Fairclough and Wodak 1997) in public discourse. To this end, we compare news texts with spoken discourse using RST to categorize and compare the proportions and characterization of presentational relations in each text type. We start by outlining the conversationalization hypothesis as brought forth by Fairclough c.s. and discussing earlier attempts to provide empirical evidence for this hypothesis. Subsequently we discuss Rhetorical Structure Theory and describe in more detail how the presentational relations in this model provide excellent conceptualizations of subjectivity as the central notion of conversationalization. Then, the RSTmodel is used in a corpus analysis that categorizes and diachronically compares recent news texts and conversations with fty-year-old news texts. Before presenting our results, we discuss the applicability of RST in the analysis of subjectivity in written and spoken discourse. 2. Conversationalization It has been suggested that a major change in discursive practices aecting many public institutions in contemporary society is the conversationalization of public discourse [. . .] (Fairclough and Wodak 1997: 265). This change involves the modelling of public discourse upon the discursive practices of ordinary life, conversational practices in a broad sense


Kirsten Vis, Wilbert Spooren and Jose Sanders

(Fairclough 1994: 253). It entails a change in the style and register of many genres. Conversationalization is an important linguistic mechanism for inuencing audiences in public discourse. And even though it may be suggestive of a concomitant shift in authority in favour of the public, Fairclough (1994) argues that the phenomenon is more ambivalent. Conversationalization as a trend in inuencing audiences may be seen as a reection of more equal, democratic discourse relations, which may be the result of cultural changes in the previous century; but it may also be regarded as a reection of more manipulated (or engineered) relations of communication, which are the result of an increasing technologization of discourse. There is anecdotal evidence of conversationalization for a number of public discourse genres, including doctor-patient interactions, university prospectuses, and media interviews (Fairclough 1992, 1994). Some smallscale studies suggest that there are eminent opportunities for making the notion more precise and productive for the description of linguistic devices used in public communication to inuence audiences. A case study of the linguistic reection of conversationalization as a historical trend in public communication has been performed by Steen (2003) upon a corpus of 85,000 words. A sample of 120 editorials from The Times between 1950 and 2000 exhibited an increase in involvement and persuasion and a decrease in narration. This may be interpreted as a stylistic shift in the direction of conversations, which are typically highly involved and persuasive and relatively unmarked regarding narration (Biber 1988). Similarly, in an analysis of a 37,000-word corpus of 32 party-election broadcasts from 1966 to 1997, Pearce (2005) found an increase in conversationalization, reected in such linguistic features as a decrease in the use of nominalizations, and an increase in the use of 1st and 2nd person pronouns, and in the use of that-complement clauses. Besides conversationalization, Pearce distinguishes personalization: the construction of a personal relationship between the producers and receivers of public discourse. He regards conversationalization and personalization as interrelated subprocesses of the more general process of informalization. It seems then that there is some quantitative evidence for the conversationalization claim at the lexical level. For Dutch, a small quantitative lexical analysis was reported by Van Rooij (2007). In a replication of Pearces (2005) research on informalization, Van Rooij investigated the occurrence of a large number of lexical elements signalling conversationalization and personalization in a small corpus of Dutch newspaper texts from 1950 and 2002; she found little evidence for informalization. The research presented here intends to possibly converge evidence by

Using RST to analyze subjectivity in text and talk


studying the conversationalization in Dutch at the text coherence level of discourse. Our study is part of a larger research project, Text, Cognition, and Communication: Mechanisms of Public Discourse, investigating conversationalization in Dutch public discourse. This particular study deals with conversationalization in subjective features, which is analyzed at three levels: the text, the sentence, and the word (Vis, in preparation). This article discusses the rst part of the corpus study, the analysis of subjectivity at the text coherence level. In the analysis presented here we compare the frequency of coherence phenomena that may reect conversationalization in three dierent subcorpora: Dutch national newspapers from 1950, Dutch national newspapers from 2002, and Dutch spontaneous conversations, recorded around the year 2002. The analysis will identify rst, to what extent newspaper texts from 2002 reect present day conversations, and second, whether and how newspaper texts have changed over the years, compared to 1950. 3. Subjectivity When investigating whether newspaper texts are becoming more conversational, the best place to look for changes is in features that are dierent between newspapers and conversations. An obvious dierence between newspapers and conversations is that the latter have a very clear interactional function. Both sender and receiver are present and the interactive nature of the communication between the two is salient. Speakers address individual listeners who are immediately present and personal. The interactional function of the language of conversation has also been described as expressive, emotive, social expressive or, in terms of systemic functional linguistics, interpersonal. Halliday (e.g. 1985), developer of systemic functional linguistics, describes language as simultaneously realizing three functions of meaning: the ideational (the expression of content or the experiential aspect of meaning), the interpersonal (how the message expresses the social relationships between the relevant interlocutors), and the textual (which realizes meanings via the structure and organization of the message itself ). In spontaneous conversations, the interpersonal function of language predominates, while in newspaper texts the ideational function comes to the fore. The aim of this study is to examine whether, and in what way, characteristics that are typical of spontaneous conversations have spread from conversations to the genre of news texts. If the conversationalization hypothesis is correct, one would expect newspaper texts from 2002 to have more interpersonal characteristics than newspaper texts from 1950.


Kirsten Vis, Wilbert Spooren and Jose Sanders

Many of the features that have been described as being typically used to express the interpersonal function are lexical elements. Halliday (1985) mentions, amongst others, modal adjuncts, pronouns, and connotative meanings of lexical items. Such lexical items have often been described in terms of subjectivity. From here on, instead of speaking of the interpersonal function of language we will use the term subjectivity. Subjectivity has been analyzed in a number of recent studies, at dierent levels: textual, grammatical and lexical. At the textual level, subjectivity can be studied in the relations between sentences, so-called coherence relations. According to many theories of discourse connectedness, coherence relations account for the coherence in the mental representation of a text that a reader constructs whilst processing the text (cf. Hobbs 1979; Mann and Thompson 1986; Sanders and Noordman 2000; Sanders, Spooren and Noordman 1992, 1993). Coherence relations are established by the reader by relating the dierent information units in the text. Hence, coherence is a mental phenomenon, rather than a characteristic of the text, spoken or written, itself (Gernsbacher and Givon 1995). The dierence in coherence between examples (1) and (2) has been described in terms of subjectivity (Spooren, Sanders, Huiskes, and Degand, in press): (1) Jan is niet thuis omdat hij weg moest. Jan is not at home because he had to leave. (2) Jan is zeker niet thuis, want zijn auto is weg. Jan is denitely not at home, because his car is gone. Example (1) describes the reason for Johns absence: the speaker reports on a causal relation between a reason and its consequence in the real world. Example (2), however, presents a conclusion drawn by the speaker on the basis of an argument; there is no real-world causal relation, but a causal relation originating in the conceptualizers (speakers) mind. Following Spooren et al. (in press), it can be said, then, that (2) is more subjective than (1). Subjectivity is a notion that is central in cognitive linguistics (Langacker, 1990). As has been illustrated by examples (1) and (2), subjectivity is an important notion with regards to coherence relations as well. In recent years, subjectivity and coherence have been studied intensively by several researchers. Pander Maat and Degand (2001) analyze dierences between coherence relations and their linguistic expressions in the form of connectives, in terms of the degree to which an active conceptualizer is responsible for the (causal) connection that these relations express. In a recent overview, Sanders and Spooren (2009) show how causal relations that are

Using RST to analyze subjectivity in text and talk


expressed with the Dutch causal connective WANT (as in example (2)) are typically processed and learned dierently than causal relations that are expressed with the Dutch causal connective OMDAT (as in example (1)). They connect this with the distinction that Sweetser (1990) makes between content relations on the one hand, and epistemic and speech act relations on the other. Sanders, Sanders and Sweetser (2009) elaborate on this in their model of causal relations within the Basic Communicative Spaces Network. Content causal relations are construed by an explicit subject of consciousness, for example an actor in the discourse, whereas epistemic and speech act causal relations are construed by the implicit subject of consciousness, i.e. the Speaker, in the deictic center of the communicative situation in which the Addressee is also represented. For the distinction between content and epistemic/speech act relations, in this paper we adopt the distinction that Mann and Thompson (1988) make between subject matter (content) and presentational (epistemic/ speech act) relations in their Rhetorical Structure Theory.

4. Rhetorical Structure Theory At the textual level we will study coherence relations in our corpus, using Rhetorical Structure Theory (Mann and Thompson 1988). RST is a descriptive theory of the structural organization of texts. It characterizes the structure of a text in terms of labelled relations that hold between parts of that text, such as Elaboration, Evaluation, and Evidence. The result of the RST analysis of a coherent text is a tree structure, where the leaves of the tree are the elementary discourse units (typically nite clauses). Most of the RST relations are binary and hypotactic, that is, they consist of two parts where one is considered to be more central (the nucleus), and the other more peripheral (the satellite). Some RST relations are multinuclear; they coordinate structures of equal importance, e.g. Contrast and List. Figure 1 presents an example of a hypotactic relation, Evidence, taken from Mann and Thompson (1988). It concerns the paradigm case of a claim followed by the evidence for the claim. The claim is in this case made up by text span 1, the evidence by spans 2 and 3. Span 1 is more essential to the text than spans 23, and therefore it is the nucleus in the relation. This is signalled by a vertical line descending down to span 1. Spans 23 together form the satellite, which is shown by an arc pointing from this span to the nucleus. RST is the most explicit and most widely used system of coherence


Kirsten Vis, Wilbert Spooren and Jose Sanders

Figure 1. Example of an Evidence relation (from Mann and Thompson 1988: 251)

relations. The main reasons for using RST in this study are that RST provides a fairly exhaustive list of well-dened relations and that it yields relatively reliable analyses (see Den Ouden 2004; Marcu, Romera and Amorrortu 1999a, 1999b). In RST a distinction is made between subject matter relations and presentational relations. The distinction is based on the eect intended by the text producer. In subject matter relations the text producers intention is that the reader recognizes the relation. The focus here is on providing information; compare the ideational (Halliday 1985) or content (Sweetser 1990) relations discussed in section 2. In presentational relations the intended eect is to increase some inclination on the part of the reader (positive regard, belief, or acceptance of the nucleus) (Taboada and Mann 2006a). This distinction is especially useful for the research presented here, because it is approximately equivalent to the distinction between the ideational and interpersonal function (Taboada and Mann 2006a) and between content and pragmatic or speech act/epistemic relations (Sanders, Spooren and Noordman 1992). Hence, to nd the interpersonal meaning expressed on a textual level, one needs only to analyze the corpus for the occurrence of presentational relations. However, in our view the relations that have been classied as presentational by Mann and Thompson (1988) do not present us with a satisfactory list. Their class of presentational relations consists of the relations Antithesis, Background, Concession, Enablement, Evidence, Justify and Motivation. Yet, Background does not seem to be presentational in that it does not intend to increase an inclination on the part of the reader. Potter (2007: 123) seems to share this view; he points out that Background and Elaboration are similar, the principal distinction being that Background precedes the nucleus and Elaboration follows it. According to Mann and Thompson (1988), Elaboration is a subject matter relation. For these reasons, we conclude that Background is not a presentational

Using RST to analyze subjectivity in text and talk Table 1. The subjective relations. Subjective relations Antithesis Description Situation presented in the nucleus comes in contrast with the situation presented in the satellite Situation indicated in the nucleus is contrary to expectation in the light of the information presented in the satellite Satellite and nucleus lead to opposite conclusions Action presented in the satellite increases the chances of the situation in the nucleus being realized Text spans are ordered in a list of arguments Situation presented in satellite assesses the situation presented in the nucleus Satellite provides evidence for the situation presented in the nucleus Situation presented in satellite interprets the situation presented in the nucleus Information in satellite provides writers right to express the nucleus Information in satellite increases desire to perform action in nucleus


Intended eect Increase positive regard Increase positive regard Increase positive regard Increase ability


Concessive opposition Enablement

Enumeration Evaluation Evidence Interpretation Justify Motivation

Increase belief Increase belief Increase belief Increase belief Increase acceptance Increase desire

relation. Furthermore, there are relations that are categorized as subject matter relations by Mann and Thompson, while we feel they are presentational. This applies to the relations Evaluation and Interpretation, because they are used to increase a positive belief in the nucleus. Finally, we added two relations to our list because we felt Mann and Thompsons list was not exhaustive: Concessive (or Argumentative) opposition (following Spooren 1989) and Enumeration. The resulting list of 10 subjective relations, with a short description and their intended eect, is shown in table 1. The dierence in subjectivity between example (1) and (2) can now be shown in their RST analyses. The RST-analysis of example (1) shows a


Kirsten Vis, Wilbert Spooren and Jose Sanders

Figure 2. RST analyses of examples (1) and (2)

Cause-Consequence relation, the analysis of example (2) an Evidence relation. Cause-Consequence is not one of our subjective relations, whereas Evidence is. If newspaper texts are in fact becoming more conversationalized and more subjective, one would expect that the relative amount of subjective relations has increased over time. To test this, we analyze the corpus for the occurrence of subjective relations. 5. Corpus The aim of the research project this study is involved in is to investigate conversationalization in Dutch public discourse. For this purpose, a corpus was compiled, comprising three subcorpora. As public discourse, newspaper texts were chosen from two years: 1950 and 2002. Microlm copies of the newspapers from 1950 are present at the Royal Library in The Hague and the library of the VU University in Amsterdam, where photocopying facilities are also available. The newspapers from 2002 are available digitally via the library of the VU University in Amsterdam. To identify to what extent texts from 2002 reect present day conversations, Dutch spontaneous conversations were selected from the Corpus of Spoken Dutch (CGN; Oostdijk 2000). The corpus consists of: a 50,000-word subcorpus of texts taken from Dutch national newspapers from 2002; 100 texts of 500 words taken from ve national newspapers, and spread over dierent sections (front page news, domestic and foreign news, cultural news, nancial news, sports, science, opinion); a 30,000-word subcorpus of texts taken from Dutch national newspapers from 1950; 60 texts of 500 words, selected on the basis of the same criteria. The same newspapers were used as for the newspaper

Using RST to analyze subjectivity in text and talk


subcorpus from 2002. As one of the newspapers was not available for 1950, the texts were taken from only four of the ve newspapers. The texts were copied from microlm and typed out by assistants; a 50,000-word subcorpus of 29 samples of transcribed spontaneous conversations, selected from the Corpus of Spoken Dutch, collected around the year 2002. Selection criteria were sex, age, geographical region, socio-economic class, level of education, number of speakers, and type of relation between the speakers. The samples include the recorded conversations, transcriptions and additional annotations such as part-of-speech tags and lemma information. Several considerations were taken into account when compiling the corpus. Firstly, since RST is devised to explain the organization of a text as a whole, it was necessary to use complete texts. Secondly, it was required that the texts be suciently extensive for subjective relations to be present in a substantive amount. Therefore, news ashes (newspaper articles of less than 200 words) were excluded. Thirdly, for the texts to be representative of Dutch newspaper texts on the one hand and Dutch spontaneous conversations on the other, there had to be sucient spreading over genres within newspapers and types of conversations and speakers. Both news subcorpora are enriched with part-of-speech and lemma information using the tagger and lemmatiser that were used for the conversations in the CGN (Van den Bosch, Busser, Daelemans, and Canisius 2007). The texts are all available in XML format.

6. Method The RST analysis of the three subcorpora took place in two stages: the segmentation into elementary discourse units and the assignment of RST relations. The three subcorpora were segmented following Marcus instructions for annotating RST structures (Marcu 1999). The conversations had already been segmented by the CGN builders. We preserved this segmentation, and sometimes made more segments, when required by Marcus instructions. Some adjustments were made to Marcus instructions if otherwise possible sites for subjective relations would have been missed. These adjustments were included in a protocol for the annotation of subjective relations. The segmentation stage was followed by the stage of assignment of the RST relations. This stage also includes the decision on nuclearity. The


Kirsten Vis, Wilbert Spooren and Jose Sanders

assignment of RST relations to a text can be done in two ways: locally or holistically. In the local approach, the annotator tries to build the tree structure in an incremental way as much as possible. In the holistic approach, the annotator builds a structure starting with the text as a whole. Following Marcus instructions, we opted for the local approach. With this approach it is possible to draw up a precise protocol for the assignment of RST relations and hence for the analysis of subjective relations. A precise protocol improves the interannotator agreement and yields more reliable analyses. For some texts the local approach did not work. We found that it was not possible to understand these texts when reading and analyzing one sentence at a time. Most often, these apparently complicated texts originated from the subcorpus from 1950. In such cases, it was necessary to read the entire text to understand the writers intentions. We then determined the large textual segments holistically and subsequently analyzed these segments locally.

7. RST and conversations RST was originally devised for the analysis of written texts. Its application to the analysis of written texts is still by far the most frequent (see Taboada and Mann 2006b, for an overview). Yet, there have been attempts to use RST to analyze other forms of communication. In her cross-linguistic study of Spanish and English task-related dialogues, Taboada (2004) concludes that it is feasible to use RST to capture the coherence of conversations. In an analysis of the coherence of newsgroups discussions, Potter (2007) reaches similar conclusions. He concludes that a slightly modied form of RST can be used to analyze these discussions. He suggests that of the four structural constraints on RST (uniqueness, adjacency, completeness, and connectedness), only the adjacency condition needs to be relaxed. By implication, Potter claims that a coherent discussion is one that is analyzable using RST (2007: 169). The question is to what degree these ndings can be generalized to other forms of conversations. Levinson (1979) notes that in dierent communicative settings the role of language can vary. In some settings the use of language is constitutive of what Levinson calls the activity type (examples are telephone conversations, job interviews), in other settings language plays only a secondary role (dancing, a football game). The

Using RST to analyze subjectivity in text and talk


types of communication for which an RST analysis seems possible are of the former type. In discourse genres like task-related dialogues and Internet discussions language plays a dominant role. However, in communicative settings where the role of language is diminished, it is much more dicult to come up with a coherent RST analysis. The conversations from the CGN corpus vary enormously in the role that language plays, from dinner conversations in which the use of language is manifest, to conversations accompanying board games, where the role of language is much less clear. Transcript (1) is a fragment from a conversation between a mother and a daughter talking about what bus to take.
Transcript (1). Transcription of conversation 1. A mother (1005, 56 years) and daughter (1004, 29 years), talking about what bus to take. 1 1004 (a) en dan gaat (b) moet je daar bus wat is het nou (c) negen (d) klopt dat nou negen (e) ja volgens mij wel (f ) bus negen hebben naar De Laar West (a) and then [it] goes (b) do you have to [take] there bus what is it again (c) nine (d) is that correct nine (e) yes I think it is (f) bus nine to De Laar West 2 1005 (a) ja (b) De Laar West (c) dadelijk even noteren hoor (d) want dat onthoud ik niet (a) yes (b) De Laar West (c) write down immediately (d) because I wont remember that 3 1004 (a) ja (b) want je hebt een negen die gaat naar Geitenkamp en een negen die gaat naar De Laar (a) yes (b) because there is a nine going to Geitenkamp and a nine going to De Laar 4 1005 oh oh 5 1004 (a) en dus je moet die negen naar De Laar hebben (b) en dan moet je uitstappen bij de halte Winkelcentrum Elderhof (a) and so you must take the nine to De Laar (b) and then you must get o at the stop Shopping Centre Elderhof 6 1005 ja yes


Kirsten Vis, Wilbert Spooren and Jose Sanders 1004 en nou ja waarschijnlijk herken je het wel als als ie als ie daar is maar je kunt voor de zekerheid even aan aan de chaueur vragen of die een seintje geeft and well yes probably you will recognize it if if it gets there but just to be on the safe side you can ask the driver to give you a sign


(a) ja (b) ja (a) yes (b) yes


dat is gewoon daar bij die SNS-Bank zeg maar die halte that is just near that SNS bank you know, that stop



(a) ja (b) ja (c) oh dat s ligt dichtbij (a) yes (b) yes (c) oh thats close



ja yes



maar dat zie ik wel hoor dat zijn daar toch al een aantal keren geweest but I will see that have been there already a couple of times havent we



ja daarom maar goed als je niet weet wanneer je moet drukken of zo maar ja dr staa stappen daar altijd mensen uit yes thats why but nevertheless if you dont know when to push or something like that but yes there are always lots of people getting o there



oh oh



het is t is zeg maar de derde halte op op die lange weg op die weg die helemaal door de wijk loopt en daar is t de derde halte van it is so to speak the third stop on on that long road on that road that runs all through the quarter and its the third stop of that



(a) ja (b) nja (a) yes (b) well yes want dr zitten vrij veel haltes dicht bij elkaar because there are quite of stops really close together





(a) hum (b) oh (a) hum (b) oh

Using RST to analyze subjectivity in text and talk


Figure 3. RST analysis of transcript (1)


Kirsten Vis, Wilbert Spooren and Jose Sanders

The analysis shows that it is possible to come up with a more or less complete analysis of the conversation. The basic structure of the fragment is that of sequence: rst the mother has to take bus 9 to De Laar West and then she has to get o at a particular stop. The analysis is more or less complete: only two segments are not included in the tree (segments 4 and 14) and these contain backchannel signals (oh). As a consequence, all segments apart from these two are connected to other segments. The analysis is unique in the sense that each segment is involved in one and only one relation (again with the exception of 4 and 14). And, nally, all satellites are adjacent to their nuclei. The following transcript comes from a conversation among three friends (two men and a woman) playing a board game.
Transcript 2. Transcription of conversation 2. Three friends, one woman (1016, 28 years) and two men (1017, 31 years; 1018, 29 years), playing a board game. 1 1018 (a) ee nadenken inderdaad (b) dit wordt natuurlijk (c) dit is naadje pit is dit gewoon (d) xxx (a) let me think indeed (b) this is gonna be of course (c) this is crummy this is really (d) xxx 2 1017 (a) hoe (b) ggg (a) how (b) ggg 3 1018 (a) naadje pit (b) dit is niks (c) dit is prut (d) dit is (e) ggg (f ) ik ben groen (a) crummy (b) this is nothing (c) this is lousy (d) this is (e) ggg (f) Im green 4 1017 ja je bent groen yes you are green ja je bent helemaal niet rood yes youre not red at all ik ik zi I Im thi 7 1016 en ook niet geel and not yellow either 8 1018 (a) k zit de hele te de ik moet hier ook nog heen (b) ggg (a) all the ti Im thinking tha I have to get here, too (b) ggg



Using RST to analyze subjectivity in text and talk 9 1016


(a) ggg (b) maar waarom denk je dat xxx (c) je hebt niet eens een transportche neergelegd (a) ggg (b) but why do you think that xxx (c) you didnt even put down a transport token



nee maar i ik zit nog te zoeken waar wil ik heen no but I Im still searching where do I want to go



waarheen leidt de weg whereto leads the road [sings a line from a well-known Dutch song]



(a) leidt de weg die we moeten gaan (b) t slaat allemaal nergens op (a) leads the road that we have to go [continues the song] (b) all of this doesnt make any sense



ja yes



nee het slaat allemaal nergens op no it doesnt make any sense at all ja ik weet het goed gemaakt yes I know what to do





xxx xxx



(a) ik ga (b) ja (c) ook dat slaat natuurlijk helemaal nergens op (a) Im going (b) yes (c) that also doesnt make any sense at all



nou dit slaat echt nergens op zeg well this really does not make any sense at all



de wolk the cloud



(a) de wolk (b) hatsiekiedee (a) the cloud (b) there we go



` he dat is een goede wolk he hey thats a good cloud isnt it ja die wolken die is dat is toch wel yes those clouds that is that really is




Kirsten Vis, Wilbert Spooren and Jose Sanders

Figure 4. RST analysis of transcript (2)

Using RST to analyze subjectivity in text and talk


Our analysis (gure 4) shows a completely dierent picture for this fragment. The analysis is incomplete: a substantial number of segments are not related to other segments. Examples are the segments 1a, 1d, 2a, 3e, 15, and 16. The connectedness requirement also does not hold. It does not seem to be possible to connect the four substructures (fragment 1b 3e; 3f10bc; 1114; and 1722) to one another. There is at least one case where the uniqueness condition seems to be violated: segment 6,8a is (in our analysis) in a Justify-relation with segment 3f and at the same time in a Concession-relation with 9a,b. Note also the absence of adjacency: on occasions segments in a relation are not adjacent to each other (the List-relation between 5 and 7 is a case in point). This happens for example when there is an interruption by a third party. Of course, interruptions occur in all types of conversations, whether language plays a dominant role or not, which is a reason for Potter (2007) to suggest that the adjacency condition needs to be relaxed. What we see in this type of conversation is that some of the moves in the conversation do not come from the words of the participants but from the game they are involved in. Consider utterance 15. Here after a substantial amount of thinking, speaker 1018 comes up with the conclusion of his thinking, namely that he has decided what move to make. Since RST is essentially capturing the rhetoric of a discourse in terms of relational propositions, it seems not suited to deal with situations in which part of the moves in the discourse are explicitly non-verbal (and hence non-rhetorical). RST does not seem to be suitable to apply to our particular type of object related conversations. This means that we cannot make the comparison between newspaper texts from 2002 and conversations to see to what extent newspaper texts from 2002 reect present day conversations. We can only compare the discourse coherence in the newspaper texts from 1950 with the discourse coherence in the newspaper texts from 2002 to nd out whether and how newspaper texts have changed over the years.

8. Results In our data analysis we compared the number of subjective relations between the two periods, which we related to the size of the subcorpora. For the analysis at the textual level we used a somewhat smaller corpus: 21,033 words from 1950 have been analyzed, and 51,468 words from 2002. In the 1950 corpus we found 226 subjective relations, compared to


Kirsten Vis, Wilbert Spooren and Jose Sanders

Table 2. Subjective relations in Dutch newspaper texts. Year Size subcorpus 21,033 words 16 60 27 0 9 21 55 34 4 0 226 1950 per 10,000 words 8 29 13 0 4 10 26 16 2 0 50 51,468 words 35 151 42 8 28 33 195 33 1 0 526 2002 per 10,000 words 7 29 8 2 5 6 38 6 0 0 44

Antithesis Concession Concessive opposition Enablement Enumeration Evaluation Evidence Interpretation Justify Motivation Total

526 subjective relations in the texts from 2002 (w2 (1) 0.40, n.s.). Table 2 shows the absolute number of occurrences for each relation in the two subcorpora, and the number of occurrences per 10,000 words. At the level of individual subjective relations, two important changes were found, one supporting the conversationalization hypothesis: the proportion of Evidence relations is signicantly higher in recent texts (1950: 26 per 10,000 words; 2002: 38 per 10,000 words; w2 (1) 5.97, p < .05). By contrast, the proportion of Interpretation relations has decreased signicantly from 1950 to 2002 (1950: 16 per 10,000 words; 2002: 6 per 10,000 words; w2 (1) 15.37, p < .0001).

9. Discussion From our attempt to use RST to analyze coherence in spontaneous conversations, it has become clear that, although RST has been used to analyze other types of conversations successfully, it is not well-suited for this task. Some of the constraints on RST are too strict to capture the

Using RST to analyze subjectivity in text and talk


coherence in face-to-face spontaneous relations, especially in situations in which part of the moves in the discourse are explicitly non-verbal, as is the case in many of the object-oriented conversations in our corpus. The results of our analysis of the newspapers show some, albeit not much evidence for the conversationalization hypothesis: the total number of subjective relations has not changed. This seems to t in with the results that Van Rooij (2007) reports on a lexical analysis of a small part of our newspaper texts from 1950 and 2002. However, we did nd signicant changes for two individual relations: Evidence relations have increased, whereas Interpretation relations have decreased. In several studies on coherence relations (for example Pander Maat and Degand 2001; Pander Maat and Sanders 2001; Pit 2003; Sanders and Spooren 2009; Spooren et al., in press), the Evidence relation is considered as the typical case for subjective relations. The fact that this ultimately subjective relation shows an increase over time can be seen as important evidence for the conversationalization hypothesis. But how can we explain the decrease in Interpretation relations? In their outline of the Rhetorical Structure Theory, Mann and Thompson (1988) did not consider the Interpretation relation as a presentational relation. We decided to include Interpretation in the set of subjective relations because we considered it related to the Evaluation relation, which is clearly subjective in its foregrounding of the Speakers subjective judgment. Consider examples (3) and (4) from our corpus, which were coded as Evaluation relations (the satellites in the relations are in bold). (3) Maar Mozart durfde hij nog wel aan, en terecht. But he still felt up to [playing] Mozart, and rightfully so. (NRC Handelsblad, section: culture, July 5, 2002) (4) De laatste tijd heersen er merkwaardige misvattingen over de democratie. Riskant, als ze ook nog eens politiek correct worden. Lately, there have been strange misconceptions about the democracy. Risky, when they even become politically correct. (Trouw, section: opinion, September 6, 2002) In both examples the satellites function to increase the readers belief that the situation in the nucleus is evaluated correctly. By contrast, if we look at the examples of Interpretation relations in our corpus, then their subjective nature may not be so straightforward. Consider examples (5) and (6) from our corpus, which were coded as Interpretation relations. In (5), the implication of the statement in the


Kirsten Vis, Wilbert Spooren and Jose Sanders

note (satellite) is that the prices should not rise too much (nucleus); in (6), the situation in the rst part of the sentence (nucleus) is more precisely dened in the interpretative part (satellite). In both cases, the satellite phrases implications of the nucleus, and thereby mainly seems to function to add detail to the description. (5) In de kabelnota van oud-staatssecretaris Van der Ploeg staat dat tot de consument vrij kan kiezen voor een kabelaanbieder het huidige niveau van consumentenbescherming gehandhaafd dient te blijven. Ofwel: de prijzen mogen niet te fors stijgen. The note on cable television by former State Secretary Van der Ploeg states that until the consumer is able to freely choose between cable providers, the current level of consumer protection must be preserved. In other words: prices should not rise too sharply. (Trouw, section: domestic news, September 6, 2002) (6) Dat de Zweden in deze ploeg zich daarbij volkomen hebben aangepast, ja, daarbij zelfs de toon aangeven, hebben we vandaag gezien. That the Swedes in this team are completely integrated, indeed, even have taken the lead, is what we have seen today. (Trouw, section: sports, March 6, 1950) Comparing examples (3)(4) with (5)(6) it appears that the Interpretation relation is more strongly oriented on the content (subject matter) of the coherence than is the case with Evaluation relation, where the presentation of the relation is most important. In other words, the Interpretation relation might not be truly subjective after all. Obviously, this needs further research. The general conclusion from our study is that conversationalization, in the sense of an increase in subjective relations, is not found across the board. Our results show no change over the years in the frequency of occurrence of most subjective relations: coherence relations in which the relation is not present in the real world, but is construed by an implicit subject of consciousness, i.e. the Speaker. Several explanations come to mind. A rst explanation is that there is a process of conversationalization in public discourse, but that it occurs only in a restricted way in newspapers texts. The communicative goals of newspapers may be too rigid to show changes in subjective coherence relations. The main goal in news is reporting on what is happening in the real world and not so much focusing on

Using RST to analyze subjectivity in text and talk


the speakers, i.e. the journalists, conclusions and evaluations of these real-world events. This goal has not changed between 1950 and 2002, and this could explain why subjective relations generally do not occur more often in present day newspaper texts. It may also explain why the Evidence relation did increase, since in this specic relation the Speaker as foregrounded conceptualizer does not so much evaluate a particular object of event, as posit it in a particular alignment in relation to another object or event, thus presenting merely their mutual relationship rather than the Speakers personal attitude towards either one. In general, the genre of newspapers may be too rigidly regulated by journalist educators and editors for a change towards a more conversational style. There are certain rules that news reporters must adhere to when writing a news article. Possibly, these rules have remained strict over the years. The validity of this explanation is of course is a matter for further investigation. Another explanation for this lack of conversationalization is that maybe in 2002 journalists have other ways of invoking positive regard and belief. For instance, citing sources in news texts is a method to increase the belief in and acceptance of news facts. Vis, Spooren, and Sanders (2009) show that in Dutch newspapers journalists use more direct citations in 2002 than in 1950. A third possibility is that only particular subgenres of newspaper texts are subject to conversationalization. So far, we have looked at all texts from dierent sections together. For instance, it could very well be that there is a dierence in conversationalization between opinion news texts or background articles on the one hand and hard news texts from the front page on the other. Sanders (to appear) reports more subjective perspectivization phenomena such as free indirect thought in longer background news articles, which seem to ctionalize the reported news facts. The texts in the present study were selected from various sections in the newspaper, hard and softer subgenres alike. At present we have not enough material to analyze dierent sections from the newspapers separately. In further research (Vis, in preparation) we will collect more texts from the two periods in order to be able to analyze dierences between sections. References
Biber, Douglas 1988

Variation across speech and writing. Cambridge: Cambridge University Press.


Kirsten Vis, Wilbert Spooren and Jose Sanders

Bosch, Antal van den, Bertjan Busser, Walter Daelemans and Sander Canisius 2007 An ecient memory-based morphosyntactic tagger and parser for Dutch. In: Frank van Eynde, Peter Dirix, Ineke Schuurman and Vincent Vandeghinste (eds.), Selected Papers of the 17th Computational Linguistics in the Netherlands Meeting, 99114. Leuven. Fairclough, Norman 1992 Discourse and social change. Cambridge: Polity Press. Fairclough, Norman 1994 Conversationalization of public discourse and the authority of the consumer. In: Russell Keat, Nigel Whitely and Nicholas Abercrombie (eds.), The authority of the consumer, 253268. London: Routledge. Fairclough, Norman, and Ruth Wodak 1997 Critical discourse analysis. In: Teun van Dijk (ed.), Discourse Studies: A Multidisciplinary Introduction. Volume 2: Discourse as Social Interaction, 258284. London: Sage. Gernsbacher, Morton Ann, and Talmy Givon (eds.) 1995 Coherence in spontaneous text. Amsterdam: John Benjamins. Halliday, Michael 1985 An Introduction to Functional Grammar. London: Edward Arnold. Hobbs, Jerry 1979 Coherence and coreference. Cognitive Science 3: 6790. Langacker, Ronald 1990 Subjectication. Cognitive Linguistics 1: 538. Levinson, Stephen 1979 Activity types and language. Linguistics 17: 365399. Mann, William and Sandra Thompson 1986 Relational propositions in discourse. Discourse Processes 9: 57 90. Mann, William and Sandra Thompson 1988 Rhetorical Structure Theory: Toward a functional theory of text organization. Text 8(3): 243281. Marcu, Daniel 1999 Instructions for manually annotating the discourse structures of texts. Available at Marcu, Daniel, Magdalena Romera and Estibaliz Amorrortu 1999a Experiments in constructing a corpus of discourse trees. In: Proceedings of ACL Workshop on Standards and Tools for Discourse Tagging, 4857. College Park, Maryland. Marcu, Daniel, Magdalena Romera and Estibaliz Amorrortu 1999b Experiments in constructing a corpus of discourse trees: Problems, annotation choices, issues. In: Workshop on Levels of Representation in Discourse, 7178. Edinburgh, UK. Oostdijk, Nelleke 2000 The Spoken Dutch Corpus Project. The ELRA Newsletter 5(2): 48.

Using RST to analyze subjectivity in text and talk


Ouden, Hanny den 2004 Prosodic realizations of text structure. Dissertation, University of Tilburg, Tilburg, The Netherlands. Pander Maat, Henk and Liesbeth Degand 2001 Scaling causal relations and connectives in terms of speaker involvement. Cognitive Linguistics 12(3): 211245. Pander Maat, Henk and Ted Sanders 2001 Subjectivity in causal connectives: An empirical study of language in use. Cognitive Linguistics 12(3): 247273. Pearce, Michael 2005 Informalization in UK party election broadcasts: 196697. Language & Literature 14(1): 6590. Pit, Mirna 2003 How to Express Yourself with a Causal Connective: Subjectivity and Causal Connectives in Dutch, German and French. Dissertation, Utrecht University. Amsterdam: Rodopi. Potter, Andrew 2007 An investigation of interactional coherence in asynchronous learning environments. Unpublished PhD-dissertation. Graduate School of Computer and Information Sciences, Nova Southeastern University. Available at Rooij, Ingrid van 2007 Informalisatie in Nieuwsberichten. Een Lingustische, Kwantita tieve Vergelijking Tussen 1950 en 2002. [Informalization in newspaper texts. A linguistic, quantitative comparison between 1950 and 2002]. MA Thesis, VU University Amsterdam. Sanders, Jose to appear Intertwined voices. Journalists representation modes of source information in journalistic subgenres. English Text Construction. Sanders, Ted and Leo Noordman 2000 The role of coherence relations and their linguistic markers in text processing. Discourse Processes 29: 3760. Sanders, Ted, 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.), Linguistic Categories of Causality in Discourse, 2160. Berlin/New York: Mouton de Gruyter. Sanders, Ted and Wilbert Spooren 2009 Causal categories in discourse Converging evidence from language use. In: Ted Sanders and Eve Sweetser (eds.), Linguistic Categories of Causality in Discourse, 205246. Berlin/New York: Mouton de Gruyter. Sanders, Ted, Wilbert Spooren and Leo Noordman 1992 Toward a taxonomy of coherence relations. Discourse Processes 15: 135.


Kirsten Vis, Wilbert Spooren and Jose Sanders

Sanders, Ted, Wilbert Spooren and Leo Noordman 1993 Coherence relations in a cognitive theory of discourse representation. Cognitive Linguistics 4: 93133. Spooren, Wilbert 1989 Some Aspects of the Form and Interpretation of Global Contrastive Coherence Relations. Dissertation, K.U. Nijmegen. Spooren, Wilbert, Ted Sanders, Mike Huiskes and Liesbeth Degand in press Subjectivity and causality: A corpus study of spoken language. In: Sally Rice and John Newman (eds.), Empirical and Experimental Methods in Cognitive/Functional Research. Stanford, CA: CSLI/University of Chicago Press. Steen, Gerard 2003 Conversationalization in discourse: Stylistic changes in editorials of The Times between 1950 and 2000. In: Luuk Lagerwerf, Wilbert Spooren and Liesbeth Degand (eds.), Determination of Information and Tenor in Texts: Multidisciplinary Approaches to Discourse 2003, 115124. Muenster: Nodus Publikationen. Sweetser, Eve 1990 From Etymology to Pragmatics. Cambridge: Cambridge University Press. Taboada, Maite 2004 Building Coherence and Cohesion: Task-Oriented Dialogue in English and Spanish. Amsterdam/Philadelphia: John Benjamins. Taboada, Maite and William Mann 2006a Rhetorical Structure Theory: Looking back and moving ahead. Discourse Studies 8(3): 423459. Taboada, Maite and William Mann 2006b Applications of Rhetorical Structure Theory. Discourse Studies 8(4): 567588. Vis, Kirsten in preparation Diachronic changes in subjectivity in Dutch news texts. Dissertation, VU University, Amsterdam, the Netherlands. Vis, Kirsten, Jose Sanders and Wilbert Spooren 2009 Subjectiviteit door de jaren heen: conversationalisatie in journalistieke teksten [Subjectivity over the years: conversationalization in journalistic texts]. In: Wilbert Spooren, Margreet Onrust, and Jose Sanders (eds.) Studies in Taalbeheersing 3 [Studies in Language Use 3], 405418. Assen: Van Gorcum.

Typology meets witness narratives and memory: Theory and practice entwined in Cognitive Linguistics1 Luna Filipovic
1. Introduction This paper illustrates how ndings from Cognitive Linguistics in general and a cognitive linguistic typology in particular can be used in dierent contexts in order to throw a new light onto potential consequences of linguistic diversity in dierent communicative situations. In return, theoretical hypotheses can be assessed with respect to their role in applied domains of linguistics and related disciplines while being moulded along the way. The result of this process is making necessary adjustments in order to enhance the predictive power of the theory. I shall explain why approaching translation, interpreting and language-mediated memory of events from a typological angle used in this paper provides novel insights into the interplay between universal and language-specic factors in habitual language use. The necessary exchange of information between linguistic theory and practice is deemed to be of central importance, and this paper is an attempt to advocate further communication between the two, as well as to illustrate the closely-knit relationship between Cognitive Linguistics and related disciplines. The principal aim is to show how typologically dierent ways of describing motion events in English and Spanish aect information content in narratives of eyewitnesses and suspects. The central issue of eective meaning transfer in translation is explored in a specic language-related social context (police interviews) in order to check for eects of language
1. The research presented here was funded by the Economic and Social Sciences Research Council (ESRC), award number: RES-000-27-0143, for which I am grateful. I would also like to express my thanks to my UCL colleagues, Gabriella Vigliocco, David Vinson, Sharon Geva and Sasha Dixon, for advice and assistance in my experimental work discussed here. I would like to use this opportunity to express my utmost gratitude to Dan Slobin for providing insightful comments and generous support over many years.


Luna Filipovic

that may go beyond mere dierences in the structure of linguistic expressions of events. Upon contrasting descriptions of motion events in original transcripts of verbal interaction between police and witnesses and suspects in Spanish and their ocial translation in English I detected discrepancies that could cause the absence of certain information details from the original text and their presence in the translation. Another consequence is the possibility to interpret events dierently due to dierent lexicalization patterns used to express them. I argue that the reason for this is not necessarily inadequate or imprecise translation but certain crucial typological dierences that underlie lexicalization of experiential domains and induce the most natural and spontaneous expressions in the native language of witnesses and interpreters respectively. I present the relevant theoretical prerequisites (section 2) that come in the form of a typology rst put forward by Talmy (1985) and then analyze the problems that may arise in the process of translating witness interviews (section 3). Section 4 provides experimental data based on mimicking witness experience of events in laboratory settings in order to assess the role of language in the witnesses memory of events. Finally, section 5 provides a discussion of the results of both naturalistic and experimental data and suggests the ways to explore these topics further. 2. Typological issues The typological assumptions that provide the backdrop for this study justify the choice of the languages we contrast here, Spanish and English, because in the latest linguistic typology (Talmy 1985) the two languages represent two polar opposites. Lexicalization of motion events was one of the central features of Talmys semantic typology of languages (1985, 2000), because the universality and omnipresence of motion in language and experience represents a solid testing ground for predictions regarding both universal and language-specic features.2 The present analysis also takes into consideration subsequent research that tested the typological hypothesis put forward by Talmy (e.g. Aske 1989; Slobin 1994, 1996, 1997, 2000, 2003; Naigles et al. 1998; Filipovic 2001, 2002, 2006, 2007a, 2008a; Gennari et al. 2002; Malt et al. 2003). Furthermore, relevant typological insights are also combined with the view on constructions expressed in Thompson (2001). Namely, certain habitual patterns used to
2. Lexicalization means rendering experiential data into a language (cf. Talmy 1985).

Theory and practice entwined in Cognitive Linguistics


express meanings relevant for a cognitive domain should be regarded as language-specic, possibly lexically highly skewed units, whose frequency and use are conrmed in attested conversational data. This understanding of constructions makes it possible to look for typical and habitual combination of elements into units favoured by native speakers in the expressions of events within a domain (Thompson 2001). In a nutshell, a motion event is dened as a change of location. Talmy singled out the main components of a motion event and the means to express them in languages. In the example The ball rolled down the hill from Talmy (1985), the moving Figure (the ball) moved in a certain Manner (rolled), along a certain Path (down) and against a certain Ground (the hill). Talmys typology divides all the languages of the world essentially into two groups: those that express the Manner component in the verb and Path in a particle (and/or prex/preposition) and those that express the Path in the verb and Manner outside the verb (e.g. in an adjunct). English belongs to the former type, termed satellite-framed languages and Spanish to the latter, verb-framed type. English speakers use one verb (usually a manner verb) with a number of particles in a clause, while Spanish speakers use one verb one preposition pattern per each section of the Path, the verb in Spanish being usually directional. The dierence in the two typological patterns conditions the typical way of referring to motion events, the English pattern being a manner verb and a directional particle and the Spanish pattern a directional verb with an optional manner adjunct. The lexicalization of the key component in an expression of motion, i.e. Path, is underlined in the following examples from Spanish and English: (1) Ella entro en el hotel tambaleandose ayer noche. She enter-PFV.3SG in the hotel staggering REFL yesterday night She staggered into the hotel last night. Spanish (and other languages of the same type) seems to operate under a constraint blocking the use of manner verbs with prepositions that express crossing a boundary (cf. Aske 1989; Slobin 1996, 1997). When there is no boundary to be crossed, Spanish speakers may use manner verbs, as in (2): (2) Corrieron para la calle, para la esquina, para la parada del autobus. . . They ran down the street, round the corner, past the bus stop. . . Slobin makes a point of saying that, even though there is an option to express manner easily in non-boundary-crossing, on the whole Spanish


Luna Filipovic

speakers, as well as speakers of other verb-framed languages, provide signicantly fewer and less varied references to manner than English speakers (cf. Slobin 2000, 2006). As a result of habitual patterning, English speakers frequently express manner of motion in their accounts of events whereas Spanish speakers provide scarce reference to manner, usually outside the verb, or omit it altogether (cf. Slobin 1996, 1997, 2000, 2003). As argued in Slobin (1996, 1997), Spanish speakers seem to have dramatically fewer types (and consequently provide fewer tokens) of manner verbs available for expression of motion. The issue at hand is to what extent the source language can accommodate itself to the target language or assimilate the original to its patterns (cf. Slobin 2000). Slobin (2000) found that only 51% of English manner verbs got translated into Spanish. The other half was translated by non-manner verbs or simply omitted. Slobin (2006) also quotes a number of other areas where Spanish and other languages of the same verb-framed type (e.g. Turkish, French, Korean) exhibit dierences when compared to languages of the satellite-framed type like English. Slobin claims that satellite-framed languages exhibit stronger manner salience than verb-framed languages on the basis of lexical access, conversational use, child language acquisition, use in creative ction and translation, metaphoric extension and imagery and understanding of manner verbs (Slobin 2006: 7071). Most of the research in this eld has focused on the verb that is used in describing motion events. The reasons for this are justiable, as pointed out in Hohenstein and Naigles (1999). The verb is generally a central categorizing factor and it would be more salient in inuencing an individuals thoughts about motion events than other aspects of the sentence. Syntactically, the verb is also of crucial importance, since it is a central element in the sentence that may be modied by adverbs, gerunds, or prepositional phrases (e.g., walks slowly or enters running) (Hohenstein and Naigles 1999.). Moreover, the verb is also a semantic focus of the sentence because it provides the listener with a way of quickly identifying the gist of the speakers intention (e.g., He gallops in versus He saunters in) (Hohenstein and Naigles 1999). The most recent research in cross-linguistic contrasts is shifting from earlier verbocentricity and takes into account information that is realized elsewhere (cf. Slobin 2006). The present research revolves around verbs for data collection, but it also includes manner gerunds and information about the Manner component expressed on the construction level where motion verbs (expressing Path or Manner) are present.

Theory and practice entwined in Cognitive Linguistics


3. Expressing and translating motion events in witness interviews 3.1. Data and methodology The ndings reported in Slobin (1996, 1997, 2000) provided the incentive to explore the relevant parameters in the context of verbal interactions between police and witnesses, where the consequences could be potentially more profound, reaching beyond the text itself. The role of language in witnesses accounts of events has been examined before from a number of angles.3 Ours is a typological one, and it has not been used in this context before. The starting point in the data analysis was to specify what counts as one motion event in the material. Perceptually and cognitively, if one wants to individuate an event from a chain of events, one is faced with a dicult task. This issue has been the subject of much discussion in psychology, linguistics and philosophy (cf. Croft 1990; Bohnemeyer 1999; Higginbotham 2000; Filipovic 2007a). For example, we can say that one motion event begins with the rst reference to the Figures motion and ends when the Figure stops moving or when the last reference to its motion is given in a description. Here, however, I dened motion events linguistically, based on the presence of verbs Ground expression(s), so what counts as one event is one verb Ground element(s) related to it.4 According to this denition, a sentence containing information on Figure, Ground and Path with/without Manner is considered to be an expression of one motion event. In practice, this meant that the number of verb phrases (VPs) equals the number of motion events expressed. The database consists of 123 les of transcribed police interviews with witnesses and suspects; the length of the les varies, ranging from approximately 20 to 120 pages, comprising over 74,000 words. They were collected in a number of courts in the counties of San Francisco and San Jose, California, USA. 3.2. Main results I found that in original Spanish texts the number of manner verbs per motion event compared to that of the directional verbs stands in the relation: 21% vs. 79%. The dierence in the quantity of manner vs. directional
3. Cf. Berk-Seligson (1990), Morris (1993), Gibbons, (2001), Hale (2004). 4. This denition corresponds to Slobins verbalized events (cf. Slobin 2000).


Luna Filipovic

verbs becomes more conspicuous if we take into consideration that the events described are frequently those referring to very dynamic situations of muggings, robberies, domestic violence, manslaughter, etc., where one could expect a variety of manners of motion. In fact, in great many cases it is the interpreter who oers a manner verb rst and only then does the witness use, for instance, the verb correr (run) instead of pasar (pass), entrar caminando (enter walking) as opposed to just entrar (enter), or saltar del coche (jump out of the car) instead of just bajarse del coche (get out of the car). Out of the 457 motion verbs used in the data, only 5 types (caminar walk, correr run, bailar dance, brincar skip, saltar jump) and 96 tokens of manner verbs were found, opposed to 8 types (irse go, venir come, entrar enter, salir exit, bajar go down, subir go up, pasar pass, cruzar cross) and 361 tokens of directional verbs. Directional verbs are used freely and equally frequently in boundarycrossing and non-boundary crossing situations. Information on manner in general (i.e. verbs and gerunds) is mainly present in non-boundarycrossing situations, while it is conspicuously absent from expressions of boundary-crossing. Out of 96 manner tokens, boundary-crossing has 5 tokens (manner gerunds), non-boundary-crossing 72 tokens (manner verbs manner gerunds), salir corriendo phrase 19 tokens.5 Thus, once more the scarcity of manner detail in accounts of motion events in Spanish is noticed, not only in the context of literary translations provided by Slobin (1996, 1997), but also in the novel context of witness interviews. Before I turn to discussing the consequences of this nding for witness testimonies, I illustrate typical witness testimony descriptions of a motion event, providing Path but not providing the Manner in which the movements occurred. Interestingly, both excerpts in (3) and (4) come from two very dynamic cases of sexual harassment and knife attack respectively. In both cases information on manner is conspicuously absent throughout the description of events: (3) Me sal de la ocina y me fui. Y el se fue atras de mi, se fue, pero el se fue para alla y yo me vine para alla. I got out of the oce and I left. And he went after me, he left, but he went over there and I came over there.
5. The reason this construction is singled out is because it is grammaticalized on those occasions: salir and corriendo do not refer to exit or running respectively, but taken together they form a construction meaning abrupt and speedy movement away from the speaker/scene.

Theory and practice entwined in Cognitive Linguistics


(4) . . . y yo le ca atras, lo vi que traa la, la, la navaja y le ca atras y cuando le ca atras, muchos le caimos atras para agarrar al que agredio el muchacho . . . . . . and I took o after him, I saw that he had the, the, the knife and I took o after him and when I went after him a lot of us went after him to grab the guy who had attacked the guy . . . In practical terms, being aware of pattern dierences could lead to the speakers of languages that leave Manner unexpressed (like Spanish) being explicitly encouraged to provide that piece of information during interrogation. On some occasions, the interrogators insist and elicit a response but generally, in continuous, uninterrupted speech of a witness/suspect, a potentially crucial Manner detail remains absent. The descriptions of the two situations in (3) and (4) also seem to lack dynamicity and intensity, which undoubtedly accompanied the reported events. While such descriptions may sound strange in English, they are the most natural and preferred option for Spanish speakers. Just as in the translations of novels from Spanish into English reported in Slobin (1996, 1997, 2000), there are occasions where manner of motion is absent from the Spanish original text but added in the English translation. The consequence of this is an image of the Figure moving in a certain way, whereas the original testimony did not provide that piece of information. The most frequent occurrences are translations of directional verbs in Spanish with an unmarked verb of manner of motion in English (walk), such as: entrar (enter) translated as walk into, or salir (exit) as walk out/away. Since walking is the most neutral manner of motion, this addition is not particularly problematic. On the whole, adding manner in English translations is noticeably less frequent than the large percentage of manner additions that Slobin found in literary translations from Spanish into English. This is most likely due to the fact that the context of translating witness interviews keeps the interpreters/translators alert and tuned to more precise translation. However, additions still occur. Spanish witnesses would give one manner item at the beginning of the description of the situation they witnessed and then continue the description with directional verbs. It is true that sometimes the manner of motion could be deduced or implicitly understood because it had been mentioned by the witness earlier in the interview, but the fact that the witness did not specify it on all occasions throughout the description of events still remains. For example, one can imagine a situation where the authorities are looking for witnesses, asking if anyone has seen the suspect run into or out of the


Luna Filipovic

restaurant, as in example (6), while the suspect in question may have been running before entering the restaurant but not when entering or exiting (e.g. he started walking to avoid suspicion). In this particular case, the witness mentioned the verb correr (run) only once at the beginning of his description and then continued describing the events for the rest of the 120 pages of the transcribed narrative using only directional verbs. The interpreter translated the motion information in this particular extract cf. (6) as ran into, but this, in eect, is not what the witness had said. Similarly in (5) the additional manner detail implies that the movement was a sneaky one, which the witness did not describe as such. In the case of harassment, from which example (5) was taken, adding this particular manner detail could engender all sorts of further implications as to the intensions of the suspect: (5) El se metio adelante de mi y se cerro la puerta. He REFL put-PFV.3SG in front of me and REFL close-PFV.3SG the door Literal translation: He put himself in front of me (i.e. he entered before me) and the door closed. Transcript translation: He slipped in before me and the door closed. (6) Pero. . . salio por la puerta detras But. . . exit-PFV.3SG via the door back Literal translation: But . . . he went out through the back door. Transcript translation: But he. . . ran out via the back door. Consecutive interpreting in particular (i.e. re-telling the witnesss narrative in larger chunks) contains more additional manner information compared to the original testimonies in Spanish. Such occurrences are possibly due to the narrative style (cf. Slobin 1994, 1996, 1997, 2000) that underlies the accounts of events in English, favouring the more dynamic construction of events in the narrative as opposed to the more static scene-setting in Spanish. In other words, Spanish texts locate protagonists at scenes but do not specify how the locations were reached, whereas in English it is continuous motion that is habitually described (i.e. how a location was reached). This dierence in narrative style was noticed by Slobin (1996, 1997, 2000) in the context of literary texts and elicited narratives based on a picture story, as well as by Barbara Tversky (personal communication), following an analysis of scripts for Latin American soap operas. This dierence was ascribed to typological contrasts between lexicalization patterns in the two languages. For example, the most frequent construc-

Theory and practice entwined in Cognitive Linguistics


tion in the Spanish descriptions that I analyzed was se metio, me met, literally meaning he put himself, I put myself (somewhere), and it is rarely specied how the location was reached. This narrative style would sound extremely unnatural if exercised in English, and as a result, dynamic descriptions of motion are used, frequently involving manner verbs. If we just imagine an account of a very dynamic motion event in English, about a hundred pages long, containing plenty of directional verbs but no mention of the manner of motion, it would seem a very odd account indeed. In Spanish, it is the most natural option. A further issue relevant for the points made here is that of agentivity in motion expressions and construction meaning.6 The issue of agentivity in motion expressions bears relevance to the understanding of motion events described by a witness or a suspect. I noticed that descriptions in the Spanish original contained some non-agentive constructions which posed particular diculty in translation into English. Pountain (2003) conrms that impersonal, non-agentive constructions are the genus of the Spanish language, and we do indeed have a conrmation of this in the original texts of the witness interviews. It may be the case that suspects were using these constructions to distance themselves purposefully from responsibility (cf. Gibbons 2001), but it may also be the case that the suspect was just using the most typical way of phrasing in his mother tongue applicable to what actually happened. One particular suspect in a homicide case kept repeating se me cayo, meaning (literally) to-me-it-happened-that-she fell throughout the interview. The interpreter was constantly translating this phrase as I dropped her, which could be interpreted as either agentive or non-agentive. In fact, I dropped her was oered as an equivalent for the explicitly non-agentive construction se me cayo that the defendant kept using throughout and (once) as an equivalent of the clearly agen tive construction la boto, which was introduced in the following way (Q-question, T-translation, S-suspect):
(7) Q: Did she fall or you dropped her on the stairs. . . T: Usted les dijo antes de que ella se cayo. . . o la boto en las gradas? (You said earlier that she REFL fell or her you threw on the stairs?) S: . . . s, s se me cayo. (Yes, yes, REFL me-ACC. fell-3SG) T:. . . yes, yes, I dropped her.

As a result, the suspects statement was that he dropped her both accidentally and on purpose, judging from the transcript translation in this
6. These points are discussed in more detail in Filipovic (2007b).


Luna Filipovic

particular case. In the Spanish original, however, the suspect only expressed his non-agentivity, i.e. that his dropping of the victim was unintentional. It may be the case that the speaker was not telling the truth, but this is not the issue at hand here: it is his words that are in focus not the truth or falsity of his statement. And the content of his statement was aected in translation. More research in this area should show the extent to which these dierences and diculties are pervasive in cognitive domains other than motion events. Unfortunately, it was not possible to obtain les that would contain original witness testimonies in English and their translations into Spanish, due to the lack of a systematic transcript-producing practice in Spanishspeaking countries (as well as in most other countries apart from the United States). It is necessary to check whether and how often information on manner of motion is omitted in Spanish translations of original witness interviews in English. In this way we would be able to contrast the situation with that in translations of novels, reported in Slobin (1996, 1997), whereby a signicant amount of manner information is simply lost in translation from English into Spanish, which, as Slobin argues, is due to the crucial typological dierences in lexicalization patterns. This research prospect would represent a natural continuation of the study presented in this paper, which would hopefully throw new light on the actual relevance of the dierences in typological patterns and habitual use in various contexts as well as in dierent social and interactive environments that inherently involve language. Stephan and Stephan (1986) observe that the increased number of guilty verdicts for defendants who did not testify in English appeared to be due to prejudice and language ethnocentrism, which in essence reects the belief on the part of the jury that defendants in U.S. courts should speak English. This attitude, together with the fact that even in the US court transcripts (unlike transcripts of police interviews used in this study) are only produced in English, indicate that the number of potential sources of diculties can increase drastically when combined with disparities in translation due to lexicalization patterns that condition habitual presence or absence of information. 4. Remembering with and without speaking: experimental check 4.1. Brief overview of previous related work Slobin (2000) introduced thinking-for-remembering and thinking-fortranslating as an extension of his thinking-for-speaking hypothesis. What Slobin was referring to was cognition on-line, as it were, assuming

Theory and practice entwined in Cognitive Linguistics


that the underlying cognitive processes mirror the language patterns of speakers while they are engaged in a language-related activity (e.g. speaking, writing, translating, etc.). Any preemptive language eect on memory is still controversial. Given the nature of memory in general, eyewitness testimony can never be completely accurate and reliable (cf. Cassel 2000). However, it is not unreasonable to assume that certain eects on how events are packaged in memory may come from the ways they are packaged in the expressions used to describe them. For example, Loftus and Palmer (1974) show how the witnesses testimonies were inuenced by the choice of words by the interrogators. Witnesses estimated the speed of vehicles in a car crash to be higher if the question contained the word smash instead of hit. The use of particular words or phrases is known to inuence the understanding of events described, but the issue at hand here is potentially of a more controversial kind: the possibility to use this kind of applied linguistic study in further tests of witness memory. The hypothesis is that Spanish speaking witnesses may provide (and possibly remember) detailed accounts less nely grained manner-wise than their English-speaking peers because of their lexicalization pattern. Even though language eects on cognition in general are the bone of contention still vivaciously discussed in the literature (cf. Gentner and Goldin-Meadow 2003), it seems less controversial to accept the principle of language-as-strategy (cf. Gennari et al. 2002). Eects of language-specic lexicalization patterns on cognitive processing of events (e.g. in memory and perceptual similarity) have been studied before (e.g. Finkbeiner et al. 2002; Gennari et al. 2002; Malt et al. 2003). According to Gennari et al. (2002), the eects of language on cognition are said to be limited and largely task-dependent. They found language eects only in a similarity-judgment task that involved prior verbal encoding. Spanish speakers, consistent with their lexicalization pattern, made more Path choices in the naming-rst condition (Gennari et al. 2002: 70). Thus, perceptual similarity seems to be manipulated by the presence of linguistic labels (Gennari et al. 2002). Hohenstein and Naigles (1999) present experimental data that support a version of linguistic relativity, because speakers of dierent languages behaved dierently towards nonlinguistic stimuli in ways that could be predicted by their respective languages. English and Spanish speakers watched a number of videos that depicted motion events. They were shown alternative video clips afterwards and were asked to judge the similarity of those videos with the original ones. The alternative video clips diered from the original ones in either Manner or Path of motion. English speakers pointed to the alternative video clips that matched the original videos in Manner but


Luna Filipovic

not in Path signicantly more frequently (70% of the time) than Spanish speakers did. Spanish speakers did not show any statistically signicant preference for either same-Manner or same-Path pairs of video clips. Similarly, Naigles and Terrazas (1998) showed that English speakers have stronger propensity than Spanish speakers to interpret novel motion verbs as manner verbs and mention the Manner of motion when describing motion events (Naigles and Terrazas 1998: 369). In the eld of language acquisition, Hohenstein (2001) found out that the dierence in cognitive salience of Manner and Path of motion for English and Spanish speaking children appears when they are aged 7 and above and not in younger children, which indicates that the disparity appears once the linguistic patterns are well and rmly established. We can say that language aects performance in tasks that involve categorization, which need not be controversial. Manner appears to be more salient for English speakers in some tasks, which the typology predicts. The issue at stake here was the possible eect of habitual language practices (such as providing more manner detail by English speakers) on speakers memory of witnessed events. On the basis of typological dierences between English and Spanish I hypothesized that dierent lexicalization patterns may induce speakers of these two languages respectively to pay more attention to those aspects of events that are habitually expressed in their languages. In the case of English speakers, both the Manner and the Path components are lexicalized habitually, so there should be no preferential treatment of one component over the other, whereas Spanish speakers were expected to pay more attention to the Path of motion, since Manner is not obligatorily expressed and frequently omitted from lexicalized motion events (cf. Slobin 1996, 1997, 2000, 2003). The novelty of the current study lies in going beyond the broadly sketched concepts of Manner and Path and establishing potentially universal components below the molar level of lexical concepts in the interaction of verbalization and memory of motion events.7 4.2. Experimental methodology, stimuli and results 4.2.1 Experiment one: similarity judgment In the rst experiment, I used a similarity judgment task whereby monolingual native speakers of Spanish and English (20 for each language)
7. In terms of Levinson (2003), the molar level is the level at which the atomic components of meaning combine into complex units and form the meaning of lexical items. The language-specic eects can then be traced at the molar level, where the atomic level is the one at which universality could be demonstrated.

Theory and practice entwined in Cognitive Linguistics


Figure 1. Manner vs. Path in Delay vs. No delay

judged the similarity between the original event (model video) and the modied original event (diering from the original video in either Manner or Path component, but never in both). For example, the original video depicted a man walking into a car park, the Path variation video depicted the same man walking out of the car park, and the Manner variation video depicted a man limping into the car park. All the experimental items were seen only once by each participant and were presented in random order together with an equal number of ller videos that contained non-motion activities (e.g. leang through a book or putting pearls on a string). The same number of target items was witnessed the same number of times in both delay and non-delay conditions across participants. The delay was 60 seconds, during which time the participants were performing a distractor task, which consisted of nding words in a 25 25 grid of randomized letters. The results for the two experimental conditions (Manner vs. Path and Delay vs. No-delay) are illustrated in Figure 1. I did not nd any language eect. Namely, both English and Spanish speakers judged the events that shared the same Path as being more similar than those that shared the same Manner. In other words, the only signicant dierence found was that between similarity ratings for Manner and Path variants (2 2 2 4 ANOVA; p 0.006), which meant that


Luna Filipovic

Figure 2. Manner-change and Path-change in boundary-crossing and nonboundary-crossing

in both languages the videos with the same Path were rated as more similar than those with the same Manner, regardless of the speakers native tongue. Since the delay vs. no delay condition did not generate statistically signicant dierences in responses, the data from the two conditions were collapsed according to the ratings of boundary-crossing vs. non-boundarycrossing in both Manner and Path conditions. Each subjects responses were converted to the values of the mean of 0 and standard deviation of 1 to account for inter-subject dierences in the way they used the scale (e.g. tending to use one or the other extreme). There is a signicant dierence between Manner and Path ratings. Manner changes in the variation videos were judged to be more similar to the original videos than Path changes, just as in the previous case (2 2 2 ANOVA: F 10.748, p 0.002). Again, there was no main eect of language (F 2.702; p 0.108), but the main eect of boundary approached signicance (F 3.921, p 0.055). Figure 2 indicates that there may be dierences between the eects of boundary-crossing within the Manner and Path conditions. The results in Figure 2 represent the similarity ratings when Manner and Path in the variation videos were dierent from the ones in the original videos.

Theory and practice entwined in Cognitive Linguistics


The eect of boundary-crossing in Manner-change conditions (MCBC vs. MCNBC) was not signicant (t 0.695). In Path-change condition, however, there was a signicant eect of boundary-crossing (PCBC vs. PCNBC; t 2.413). This result indicates that the dierences between a model and its variation video were judged to be more substantial in nonboundary-crossing pairs than in boundary-crossing pairs in the Pathchange condition. This correlation is due to the fact that PNBC condition, which elicited the lowest scale ratings of similarity for both speaker groups, contained videos that involved a change of vector, such as moving left to right in the original video vs. moving away from the observer in the variation video. This drastically lowered the similarity score within this category, and as a result, the overall score for PCNBC situations. It is an indication of the importance that a nely grained analysis like the one employed here has in the process of detecting atomic components of Manner and Path that have signicant bearing on the overall rating of those components. Although some previous experiments involved videos with a moving object (e.g. Finkbeiner et al. 2002), and some others used human gures (cf. Malt et al. 2003), they did not discuss the dierences that concern the vector change of Path. To my knowledge, the importance of vector change in similarity judgments has not been explicitly recognized. Some previous studies contrasted Path depictions along the same vector (e.g. in vs. out or up vs. down) and as a consequence this could have rendered the Path component less interesting and less striking in similarity judgment tasks. What this experiment showed is that, in human motion, perception is centered around the position of the body, which in the case of motion events involving a faceless object such as a ball is not (or is less) problematic from the point of view of the vectors of motion, especially where a more striking dierence in Manner is presented. When a human gure is moving however, the vectors of motion are of crucial importance to the analysis. If the vector of motion is changed the speaker/ viewer sees the back of the head in the original screen, for example, and then the prole in the variation screen. The reason why Manner was more salient overall in one of the experiments in Finkbeiner et al. (2002) may be that its change was just more striking and salient than the Path of the faceless object. The change with respect to which parts of the face and body are seen in the model and the variation video respectively takes precedence over both Manner of motion and same-vector Path of motion (i.e. left-to-right vs. right-to-left) in that it diminishes the similarity between the model and the variation videos. Preserving the vector of motion provided higher judgments of similarity even though the Manner of motion was saliently dierent. It is true that observing the Path of


Luna Filipovic

motion in its entirety is a task more demanding than observing the Manner, as Gennari et al. (2002) rightfully point out, because it takes longer compared to just a few frames that are sucient to perceive Manner. However, the processing demand of Path obviously did not sway the similarity judgments in our task towards Manner preference even in English speakers, who habitually express Manner in the central position of an expression (i.e. the verb). In summary, regardless of lexicalizing preferences, there is a cline of event similarity that this experiment produced depending on the (sub)component that two events dier in from least similar to most similar: Path vector change Path direction change Manner change (cf. Filipovic and Geva (in preparation) for a detailed analysis) 4.2.2. Experiment two: recognition task In another experiment I carried out for the purpose of this study I wanted to give initial advantage to English speakers by depicting events in pairs that can be individuated by dierent verbs in English but not in Spanish. The contrasting pairs were: running/jogging, walking/strolling, strutting/ striding, scurrying/sprinting, mincing/trudging and skipping/galloping. The participants viewed each pair of videos, whereby one event was depicted in the original video (e.g. walking) and the other member of the pair was the variation video (e.g. strolling). In the recognition stage the participants were asked to state whether or not they had witnessed the same event before (i.e. in the encoding stage). I included verbalization at dierent points in this experiment: during encoding only, during recognition only, both at encoding and at recognition, neither at encoding nor at recognition. I counted the number of false alarms (FAs) for both speaker groups. A FA occurs when a speaker says that an event in the variation screen had been witnessed previously, which in eect was not the case. There was no main eect of language: F 0.09; p > 0.05 or condition: F 0.492; p > 0.05. However, the items analysis revealed an interesting trend. A by-items analysis was conducted on the data and the GLM repeated measures ANOVA revealed signicant dierences between the items for English: F 14.55; p < 0.001 and Spanish speakers: F 9.42: p < 0.001. In a nutshell, the run/jog manner variant generated the highest mean for FAs for both language groups. The sprint/scurry manner variant produced the smallest number for the English speakers while it was the walk/stroll and skip/gallop variants which produced the fewest FAs for the Spanish speakers. For the English speakers the walk/ stroll and skip/gallop manner variants produced the same mean for FAs

Theory and practice entwined in Cognitive Linguistics Table 1. Mean number (and standard deviation) of false alarms per item for English and Spanish speakers
run/jog walk/ stroll 1 (s 0) 0.75 (s 0.5) strut/ stride 2 (s 0.82) 2.5 (s 1.29) sprint/ scurry 0 (s 0) 1.5 (s 1.73) mince/ trudge 3.75 (s 1.26) 4 (s 1.41)


skip/ gallop 1 (s 1.15) 0.75 (s 0.96)


4.5 (s 1) 4 (s 0.82)


and for the Spanish speakers this happened with the walk/stroll and skip/ gallop videos (cf. Table 1). Further inspection of the data also revealed that failure to make a verbal distinction in manner of motion during verbalization did not reect recognition memory performance. I observed that, in accordance with the typological prediction, English speakers provided more information on Manner (both in and outside the verb) than Spanish speakers. That is not controversial, and similar ndings were reported before.8 The raw scores (32 participants, 8 per condition for each language) are given in Table 2. A pre-test had been conducted to verify that the videos used were indeed dierent: a small group of native speaker judges had named each video (randomly) presented, using a single verb. The fact that all judges used dierent verbs for each member of the pair indicated that there was a suciently noticeable dierence between the manners of motion depicted. The videos were also presented in pairs and another group of judges were asked to say what the dierences between the videos were. Naming was accepted to be correct at a threshold of 75% accuracy; all judgments satised this criterion, and in some cases exceeded this level of accuracy. We can try to account for the reasons why certain pairs generated signicantly more FAs in both English and Spanish than some others. For example, speakers of both languages had signicantly more FAs when they had to distinguish between run and jog (18 FAs for English and 16 FAs for Spanish speakers out of the possible total of 32 FA) or mince and trudge than when they had to distinguish between skip and gallop or walk and stroll (4 FAs for English and 3 FAs for Spanish speakers). The dierence between run and jog is rather contextual and
8. Cf. Naigles and Terrazas (1998), Slobin (2000), Gennari et al. (2002).


Luna Filipovic

Table 2. Number of FAs per condition and per item english speakers Item V/NV NV/V V/V NV/NV Total run/ jog 5 5 3 5 18 walk/ stroll 1 1 1 1 4 strut/ stride 1 2 2 3 8 sprint/ scurry 0 0 0 0 0 mince/ trudge 4 4 5 2 15 skip/ gallop 0 2 2 0 4 Total 11 14 13 11 49/192

spanish speakers Item V/NV NV/V V/V NV/NV Total run/ jog 3 4 4 5 16 walk/ stroll 1 0 1 1 3 strut/ stride 1 2 3 4 10 sprint/ scurry 1 1 0 4 6 mince/ trudge 3 3 6 4 16 skip/ gallop 1 0 0 2 3 Total 10 10 14 18 54/192

also perceptually less salient. They were depicted as movement at the same speed but with dierent knee movement (more pronounced in the case of jogging, which looked like less relaxed running, as it were). I have to emphasise that run and jog can be said to be conceptually distinguishable primarily by the intentions of the actor rather than by specic dierences in the motion itself. Thus, run may be generally associated with some purpose or goal (as in running a race, running away or running toward) whereas jog may be seen to be a goal in itself (as in jogging for exercise). Further, it might even be said that run is a conceptual superordinate of jog, in which case jog is understood to be a type of run. This would explain the nding that both groups of participants had greater diculty distinguishing the run/jog or mince/trudge variants, which are very similar and lack more salient dierences such as rhythm or speed. In the case of the latter pair (mince vs. trudge), the movements diered in neither speed nor rhythm but rather in hip movement (coquettish in the case of mince and clumsy and heavy in the case of trudge), and light steps in the case of

Theory and practice entwined in Cognitive Linguistics


mince and heavy steps in the case of trudge. This was much more dicult to pin down than the dierence in speed, as in walk vs. stroll, or rhythm, as in skipping vs. galloping. It can be claimed that speakers found it easier to distinguish between the two events in a pair if they could clearly detect a component on which they can base their dierentiating, such as speed (as in walk vs. stroll ), rhythm (skip vs. gallop), step size (sprint vs. scurry), etc. Thus, these basic components of motion events seem to matter more than the molar, lexical concepts that speakers may or may not resort to, if their native language has them in the rst place. This brings us to the potentially essential subcomponents or atomic components of motion events, which perhaps better reects how events are stored in memory than the view that sees events decomposed into Manner and Path dened in broad terms. The fact that in verbalization conditions English speakers performed similarly to their Spanish peers, using simpler manner verbs and additional paraphrase rather than semantically complex verbs, shows that spontaneous, ordinary language use under the experimental condition of free encoding in this experiment motivates comparable linguistic behaviour, eliciting similar verb choices and frequencies, not conditioned by language-specic or typological restrictions. I did not coerce the participants to provide as much information as possible in the simplest way available (as Gennari et al. (2002) did). Thus we are presented with the intriguing ground where we can search for both language-specic (perhaps contextually-driven) and potentially universal underlying principles of linguistic behaviour. I demonstrated that our search for language eects actually provided us with possible candidates for universal principles in perception and memory in the domain of motion events. This raises an interesting question concerning the validity of the typological dierences and the specic circumstances under which they are manifested. It may be that a frequency eect was operating, in that participants chose to use more common English verbs rather than less frequent ones. Certainly, there is a frequency dierence between verbs such as walk/run and strut/stride. This eect may have therefore made it less likely that participants would generate these latter verbs spontaneously. Similarly, if words such as strut and stride are less common in everyday usage, it may be that individuals do not possess particularly strong conceptual representations of these manners of motion. At this point we have to make an important remark: in this study, I tested the memory for simple, single motion events. The need to individuate every portion of the Path by using separate directional verbs may aect memory of events in situations where a number of Paths and


Luna Filipovic

Manners of motion are depicted in sequence, whereby a motion event is experienced in terms of a journey that Slobin talks about (Slobin 1997). In other words, the diculty for Spanish speakers may be possible to detect if more complex continuous motion is being described. This is the occasion where those speakers tend to omit information about Manner, because they do not specify it for each portion of the Path. In Spanish, one cannot use a construction whereby the Manner of motion for the whole Path is specied they way it is done in English (e.g. He limped out of the house, across the street, into the garden, etc.). This is where typological contrasts between the two languages are the most conspicuous in translation from English, for example, whereby each portion of Path is also specied with regard to Manner (cf. Slobin 1997; Filipovic 2007a). It is possible to express Manner for each portion of Path in Spanish with a sentence-initial gerund or by means of a paraphrase that follows the verb, however, in general this is not done, because it would work against language economy, stylistic acceptability and linguistic habits in Spanish. Furthermore, if one has to describe a motion event that consists of a number of Paths and Manners (e.g. He stumbled out of the shop, limped across the street and staggered into the park), the question to ask is how Spanish speakers would cope with expressing such complex events (or journeys) and whether they would register manner of motion for each event segment in verbalization and memory with more diculty than English speakers? Filipovic (2008b) provides experimental data that demonstrate the same memory performance by English and Spanish speakers in recognition tasks that require speakers to distinguish between events that are not lexicalizable in a single verb in either language (e.g. limp with a bent leg vs. limp with a sti leg). The assumption was that English speakers may remember events better due to general manner salience in that language even thought they cannot lexicalize the Manner of motion in a single verb. This was not the case. The performance of the two speaker groups was also the same in another recognition task where Manners of motion were lexicalizable in both languages (such as run or skip) even though Spanish speakers express manner comparatively rarely (Filipovic 2008b). However, this experiment (as well as others presented in the recent literature on memory of events) tackled memory of a single change of location in a single manner. Future experimental work should focus on those aspects of communicative situations where lexicalization contrasts are at their most conspicuous (namely, events containing multiple Paths single Manner and multiple Paths multiple Manners) in order to check for typological dierences beyond single instantiations of single Path single

Theory and practice entwined in Cognitive Linguistics


Manner (Filipovic in progress). Initial evidence indicates that languagespecic eects on memory can be detected with such an increase in memory and lexicalization load whereby speakers of English seem to benet from the ease of lexicalization of the Manner component (Filipovic in progress).

5. Conclusions I demonstrated how systematic research into lexicalization of cognitive domains can help highlight relevant typological contrasts that aect information processing in general and giving and translating witness statements in particular. Typologically motivated lexicalization patterns appear to be entrenched and they pervade many aspects of language use. Awareness of typological dierences could help us understand possible prejudices that may underlie judgments and interpretations of events in legal contexts. My study of witness interviews and their translation enabled us to look at language use from a novel perspective. These ndings could be further implemented in education and training of language professionals such as interpreters, translators, language teachers and researchers in the areas of cross-cultural and cross-linguistic communication. I did not nd any language-specic eects on memory of motion events in the experiments presented here, which does not have to mean that they may not be found under dierent circumstances (cf. Filipovic in progress). What appears to be uncovered on this occasion is a potentially universal mechanism of event identication and decomposition of event components into provisionally termed atomic elements. They are positioned below the molar level of lexicalized concepts of manners of motion. A closer and more detailed look into lexicalized concepts of Manner and Path of motion should reveal more nely grained information about the parameters that matter when people speak and think of motion events, as well as of events in other domains. For example, we may nd out that speaker may activate both molar and atomic concepts but in dierent instances. This is one prospective route for future experimental investigation. The contrast between Spanish and English stemming from dierences in habitual patterns for expressing events in one domain (motion) could be seen as a template for all-encompassing research within other experiential domains (e.g. causation, location, posture, etc.), including the cognitive eects involved and practical use in various social contexts where language is inherently present (e.g. courtroom practices, education,


Luna Filipovic

language acquisition, development of communication skills, cross-linguistic information processing, bilingualism, etc.). I hope that this study will inspire further applied cross-linguistic and interdisciplinary research, whereby theory and practice will be mutually informative and advance side by side.

Aske, Jon 1989

Path predicates in English and Spanish: A closer look. In: Proceedings of the Fifteenth Annual Meeting of the Berkeley Linguistics Society, 114 Berkeley: Berkeley Linguistics Society. Berk-Seligson, Susan 1990 The Bilingual Courtroom: Court Interpreters in Judicial Process. Chicago: University of Chicago Press. Bohnemeyer, Jurgen 1999 Event representation: Some primordial soup for the evolution of a research project on event representation in language and cognition (revised draft 1999). At: profession/bonhem.html. Cassel, Elaine 2000 Behavioral science research leads to Department of Justice guidelines for eyewitness evidence. At: publications/valawyer/feb00/cassel.pdf. Croft, William 1990 Possible verbs and the structure of events. In: Savas L. Tsohatzidis, (ed.), Meanings and Prototypes: Studies in Linguistic Categorisation, 4873. London: Routledge. Filipovic, Luna 2001 Language-specic expression of motion and its use in narrative texts. In: Alexandre Veiga, Vctor M. Longa and JoDee Anderson ` (eds.), El Verbo entre el Lexico y la Gramatica, 5362. Lugo: Tris Tram. Filipovic, Luna 2002 Verbs in Motion Expressions: Structural Perspectives. PhD dissertation. University of Cambridge, UK. Filipovic, Luna 2006 Weaving the web of meaning. Languages in Contrast 6(1): 151 175. Filipovic, Luna 2007a Talking about Motion: A Cross-linguistic Investigation of Lexicalization Patterns. Amsterdam: John Benjamins.

Theory and practice entwined in Cognitive Linguistics


Filipovic, Luna 2007b Language as a witness: Insights from cognitive linguistics. Speech, Language and the Law 4(2): 245267. Filipovic, Luna 2008a Typology in action: Applying insights from typological contrasts. International Journal of Applied Linguistics 18(1): 4261. Filipovic, Luna 2008b Language eects on memory of events, ms. Department of Psychology, University College London. Filipovic, Luna in progress Bilingual witness report and translation, ms. Department of Linguistics, University of Cambridge. Filipovic, Luna and Sharon Geva in preparation To see, say and remember: Motion events in perception, language and memory. Finkbeiner, Metthew, Janet Nicol, Delia Greth, and Kumiko Nakamura 2002 The role of language in memory for actions. Journal of Psycholinguistic Research 31(5): 447457. Gennari, Silivia P., Steven A. Sloman, Barbara C. Malt, and W. Tecumseh Fitch 2002 Motion events in language and cognition. Cognition 83: 4979. Gentner, Dedre and Susan Goldin-Meadow 2003 Language in Mind. Cambridge MA: MIT Press. Gibbons, John 2001 Legal transformation in Spanish: An audencia in Chile. Forensic Linguistics 8(2): 13501771. Hale, Sandra B. 2004 The Discourse of Court Interpreting: Discourse Practices of the Law, the Witness and the Interpreter. Amsterdam: John Benjamins. Higginbotham, James 2000 On events in linguistic semantic. In: James Higginbotham, Fabio Pianesi and Achille C. Varzi (eds.), Speaking of Events, 4981. Oxford: Clarendon Press. Hohenstein, Jill 2001 Motion Event Similarities in English- and Spanish-speaking Children. PhD dissertation. Yale University, USA. Hohenstein, Jill and Letitia Naigles 1999 The development of linguistically-inuenced thoughts. Poster presented at the Biennial Meeting of the Society for Research on Child Development. Albuquerque, NM. Levinson, Stephen C. 2003 Space in Language and Cognition. Cambridge: Cambridge University Press. Loftus, Elizabeth F. and John C. Palmer 1974 Reconstruction of automobile destruction: an example of the interaction between language and memory. Journal of Verbal Learning and Verbal Behaviour 13: 585589.


Luna Filipovic

Malt, Barbara C., Steven A. Sloman and Siliva P. Gennari 2003 Speaking versus thinking about objects and actions. In: Dedre Gentner and Susan Goldin-Meadow (eds.), Language in mind: Advances in the Study of Language and Thought, 81111. Cambridge, MA: MIT Press. Morris, Ruth 1993 Images of the Interpreter: Studies of Language-switching in the Legal Process. PhD dissertation, Lancaster University. Naigles, Letitia and Paula Terrazas 1998 Motion-verb generalizations in English and Spanish. Psychological Science 9: 363369. Naigles, Letitia R., Ann R. Eisenberg, Edward T. Kako, Melissa Highter and Nancy McGraw 1998 Speaking of motion: Verb use in English and Spanish. Language and Cognitive Processes 13: 521549. Pountain, Christopher 2003 Exploring the Spanish Language. London: Hodder Arnold. Slobin, Dan I. 1994 Crosslinguistic Aspects of Child Language Acquisition (Sophia Linguistica 35). Tokyo: Sophia University. Slobin, Dan I. 1996 Two ways to travel: verbs of motion in English and Spanish. In: Masayoshi Shibatani and Sandra A. Thompson (eds.), Grammatical Constructions Their Form and Meaning, 195219. Oxford: Clarendon Press. Slobin, Dan I. 1997 Mind, code, and text. In: Joan Bybee, John Haiman and Sandra Thompson (eds.), Essays on Language Function and Language Type, 437467. Amsterdam: John Benjamins. Slobin, Dan I. 2000 Verbalised events: a dynamic approach to linguistic relativity and determinism. In: Susanne Niemeier and Rene Dirven (eds.), Evidence for Linguistic Relativity, 107138. Amsterdam: John Benjamins. Slobin, Dan I. 2003 Language and thought online: Cognitive consequences of linguistic relativity. In: Dedre Gentner and Susan Goldin-Meadow (eds.), Language in Mind, 157191. Cambridge, MA: MIT Press. Slobin, Dan I. 2006 What makes manner of motion salient? Explorations in linguistic typology, discourse and cognition. In: Maya Hickmann and Stephane Robert (eds.), Space in Languages: Linguistic Systems and Cognitive Categories, 5981. Amsterdam: John Benjamins. Stephan, Cookie W. and Walter G. Stephan 1986 Habla Ingles? The eects of language translation on simulated juror decisions. Journal of Applied Psychology 16(7): 15591816.

Theory and practice entwined in Cognitive Linguistics


Talmy, Leonard 1985 Lexicalization patterns: semantic structure in lexical forms. In: Timothy Shopen (ed.), Language Typology and Syntactic Description. Vol. 3, 57149. Cambridge: CUP. Talmy, Leonard 2000 Toward a Cognitive Semantics. Vol. 2. Cambridge, MA: MIT Press. Thompson, Sandra 2001 Construction and conversation. Plenary talk at the International Cognitive Linguistics Conference, Santa Barbara, California.

Part V.

The social and cultural contexts for CMT and CI

Cross-cultural variation in idiomatic expression: Insights from Conceptual Metaphor Theory and implications for Translation Studies Diane Ponterotto
1. Introduction This paper addresses an applied aspect of Cognitive Linguistics and will explore the contribution that a knowledge of Conceptual Metaphor Theory1 can make to cross-linguistic analysis and eventually also to translation practice. Through contrastive observation of two languages, English and Italian, the study will explore problems involved in cross-language rendering of idiomatic expressions and concomitantly focus on similarities/dierences of their metaphorical conceptualization. It will then draw some practical conclusions for translation strategies. Following initial insights given in Hiraga (1991) concerning cross-cultural similarities/dierences in metaphorical expressions and following suggestions given in Barcelona (1997) regarding methodological procedures for the identication of conceptual metaphors, the study will oer guidelines thought to be useful when determining gurative equivalences for contrastive analysis. It will also consider constraints imposed on translation practice both by cross-cultural variation in metaphorical conceptualization and by language-specic morphosyntactic structure, text-typology and discourse strategy conventions.

2. The problem The specic problem posed in this paper is How can we express the same idiomatic meaning in two dierent languages? Can cognitive linguistics, and especially conceptual metaphor theory, provide viable solutions?
1. In this paper the category Conceptual Metaphor is dened according to the theory rst given in Lako and Johnson (1980) and extensively developed in subsequent research. See Evans (2007: 3335 and 136138) for a concise denition.


Diane Ponterotto

As Kovecses (2003: 314) notes, for cognitive linguists, gurative mean ing cannot be expressed in literal ways. According to Gibbs (1994) to blow ones stack cannot be rendered by the expression to get very angry, or one of its equivalents, in English or in other languages. Thus, as an introduction to the problem, we would like to repeat comments made in a previous study (Ponterotto 2007), which posed the question of transfer of idiomatic meaning from English to Italian. Let us take the idiomatic expressions (IE) He kicked the bucket and I ipped my lid. The translation problem derives from the necessity to preserve the gurative nature of the original expression. A translator wishing to render He kicked the bucket into Italian has two options: 1. to use ` a literal expression: Allimprovviso e morto Suddenly, he died, or 2. to search for a comparable equivalent within the repertoire of synonymous options available in Italian. The translator could then decide to choose ` e crepato, He cracked, He burst or He split, which conveys the connotation of sudden death as well as that of lack of sympathy on the part of the speaker, as in the English expression.2 However, if we look at the other example, I ipped my lid, the solution is not as simple. The translator could opt for a literal utterance mi sono arrabbiato moltissimo I became very angry or search for another gurative expression, but in that case, the iconic eect of the original idiom is lost. It is precisely the image of a lid ipping open that conveys the meaning of extreme anger. In fact, the expression I ipped my lid has such a strong iconic coding that, in comic strips, the uncontrollable anger of the character is often accompanied by a visual image of the top of the head ying o (cf. Forceville and Urios-Aparisi 2009). Now, according to Cognitive Linguistics, idiomatic meaning is based on human conceptualization of experience, which is manifest in cognitive mechanisms like metaphor, metonymy and conventionalized shared knowledge (cf. Kovecses 1990: 211). This is also supported by psycholinguistic research which explores the conceptual and psychological reality of idiomatic knowledge (cf. Gibbs 1990; Gibbs and OBrian 1990). Thus, in the light of this perspective on idiomatic meaning, the expression I ipped my lid could be said to emerge from a conceptual metaphor (CM) (Lako and Johnson 1980; Kovecses 1990), which represents emotions as being in a body that is a container (CMs the body is a
2. In fact, the bilingual dictionary (Garzanti/Hazon 2001) gives, as a gurative meaning for crepare, to croak, or to snu it.

Insights from Conceptual Metaphor Theory and implications


container; emotions are fluids in a container). In this metaphorical frame, when the emotions are excessive, pressure is produced inside the container, causing its lid to pop up or ip o. This would be the conceptual motivation for the IE I ipped my lid. It is interpretable because Anglophone speakers conceptualize the emotion according to the cognitive model of a container (Lako and Johnson 1980; Kovecses 1990): the body is a container the body is the seat of emotions emotions are liquids in a container too much liquid causes the pressure in the container to rise when the liquid rises, the container explodes (the lid pops up or flips open; the liquid escapes from the container).

In English, this cognitive model for anger seems to be very productive, since there are many corresponding IEs: I ipped my lid. I I I I blew my top. blew my stack. hit the ceiling. went through the roof.

In Italian, on the other hand, extreme anger does not seem to be equally represented by this CM. In Italian, a person who is extremely angry is usually described as: E imbestialito. E nero. E fuori di se. He is beastly. He is black. He is out of himself.

It would seem that for the emotion of anger Italian speakers often use idiomatic expressions from the domain of animals, colours or madness.3 Common CMs for anger in Italian would then be:

3. Only one of the Italian informants suggested an expression assignable to the container metaphor: Sono esploso, I exploded. The others preferred the other options and the beast analogy was the most frequent response. Moreover, an Internet search on the Italian Google site for beast expression variants


Diane Ponterotto

an angry person is a wild animal an angry person is black an angry person is out of control It should be noted that in English along with the container metaphor for anger, anger is a hot fluid in a container, we also nd the animal metaphor anger is a dangerous animal (Kovecses 1986: 23). Moreover, we could also suggest that both cognitive models, container and animal conceptualize anger as the state of being out of control. Thus, translators would have recourse to both cognitive models, container and animal. However, the fact that similar expressions may be available in both languages does not mean that they share the same level of perceptual saliency in speakers conceptualizations.4 That is why the closest equivalent for I ipped my lid could rightly be suggested to be Mi sono imbestialito I became beastly, or Sono diventato una belva I became a wild beast.5
gave 104,000 hits for sono diventato una bestia (I became a beast), 134,000 hits for mi sono imbestialito (I became beastly/a beast) and 176,000 hits for sono diventato una belva (I became a wild beast). All three expressions have only a gurative decoding related to the emotion of anger. Another gurative option in Italian for extreme anger could be: Ha perso le stae (He lost the stirrups). The word stae (stirrups) also refers, although indirectly, to the semantic eld of animals. 4. Again when the Italian informants were asked why they preferred one idio` matic expression over another, the explanation often given was perche ha ` piu immediatezza (it has more immediacy). This native speaker intuition would lead us to reect more attentively on the role of metaphor vividness in contrastive and translation analysis. In other words, the question arises as to whether options are selected from similar metaphorical domains or on the basis of either communicative eect or strength of iconic coding (or both). On the role of intuition in cognitive linguistics research, see Geeraerts and Cuyckens (2007: 1718). 5. I would like to thank the anonymous reviewer who called my attention to the fact that an animal is also a container. Although this may be true, the native speakers of Italian whom I consulted on this point denied conceptualizing the beast as a container from which anger escapes. The conceptualization was, on the other hand, running wild, and, moreover, considered by the informants to be a permanent state of the beast, having little to do with the sudden explosion of anger characteristic of the container metaphor in English. What is more, I have found that Italian students of L2 English have diculty in processing idiomatic expressions motivated by the container metaphor. Only if they are familiar with Conceptual Metaphor Theory are they able to understand what ipping a lid has to do with anger.

Insights from Conceptual Metaphor Theory and implications


3. Identifying similarities and dierences in conceptual metaphor across languages Following insights given in Hiraga (1991), in Ponterotto (2007) examples were presented of three types of identity relation across languages (similarity, dierence or partial similarity) for three conceptual metaphors: time is money; mood is weather and life is a journey, applied to Italian and English. Another example of this type of analysis follows below, applied to the CM: life is gambling, with possible Italian translations: 3.1. CM: LIFE IS GAMBLING: Similarities
English IEs 1. 2. I wouldnt bet on it if I were you. Thank goodness, youve always got an ace up your sleeve. If you play your cards, right, you can do it. Italian translations Se fosse in te, non ci scommetterei. Grazie a Dio, hai sempre un asso nella manica. Se giochi bene le tue carte, ce la farai. English glosses of Italian translations If I were you, I wouldnt bet on it. Thank God you always have an ace in your sleeve. If you play your cards right, you will do it.


We can see that the idiomatic expressions coincide almost totally. For the communicative situations indicated, both languages refer to the CM: life is gambling, and use similar source domain elements. 3.2. CM: LIFE IS GAMBLING: Dierences
English IEs 4. When the chips were down, she realized that she didnt love him. Do you realize that your job is at stake? The Labour candidate is holding all the aces Italian translations Nel momento cruciale, ` si e resa conto che non lo amava. Ti rendi conti che il tuo ` posto e in pericolo? Il candidato laburista tiene il coltello dalla parte del manico. English glosses of Italian translations In the crucial moment, she realized that she didnt love him. Do you realize that your job is in danger? The Labour candidate is holding the knife by the end of the handle

5. 6.


Diane Ponterotto

As we can see, source domain elements in the English IEs, chips are down, holding all the aces, at stake, which derive mainly from the card game of poker, do not appear in the Italian equivalents. In examples 4 and 5, a literal paraphrase is suggested. In example 6, a change of metaphor is a possibility. In Italian, Tenere il coltello dalla parte del manico Holding the knife at the end of the handle stems from the CM: life is war, implying the advantage of an opponent who is wielding a knife rather than receiving the thrust.6 3.3. CM: LIFE IS GAMBLING: Partial similarities
English IEs 7. Hes been dealt a lousy hand. Its the luck of the draw Italian translations La fortuna ha smesso di girare dalla sua parte. ` La fortuna e cieca. English glosses of Italian translations Luck has stopped turning his way. Fortune is blind


These last two examples are considered partial similarities, because even if the Italian language selects items which do not refer directly to card games, as in the English expressions luck of the draw or a lousy hand, it may select aspects from the semantic eld of chance and fortune. In example 7, the Italian expression refers to the wheel of fortune which turns and favors people at random. In example 8, it refers to the mythological goddess of fortune, who is blindfold and therefore distributes luck randomly.

6. I thank the anonymous reviewer who commented that the Italian knife/handle metaphor would not necessarily be the only or most frequent equivalent of holding all the aces. The point here is that for the concept holding all the aces, speakers of Italian do not seem to activate the gambling game metaphor. There may be other options, but the knife/handle Italian expression is so common in everyday speech that its suitability for the context (being too strong to fear opposition) is immediately recognized and suggested. The question perhaps is just what is meant by equivalent. Again, similar options may be available in dierent languages, but they may not be equally salient. As will be discussed in subsequent paragraphs, idiomatic choice is contextspecic and constrained by numerous formal and functional variables.

Insights from Conceptual Metaphor Theory and implications


This type of analysis of translation equivalence is based on a search for similar source domains. In fact, of our eight examples for English idiomatic expressions deriving from the CM: life is gambling, we nd three examples having similar source domain elements chosen in English and Italian, three examples which do not have the same source domain elements and two examples which are partially similar, as summarized in the following table.

Source domain word of English IE 1 2 3 4 5 6 bet ace up your sleeve play your cards right chips are down job at stake holding all the aces a lousy hand the luck of the draw

Translation of English source domain word scommettere asso nella manica giocare bene le carte nel momento cruciale posto in pericolo tenere il coltello dalla parte del manico fortuna/girare fortuna/cieca

English gloss of Italian translation of English source domain word bet ace in the sleeve play your cards right in the crucial moment job in danger holding the knife at the end of the handle fortune/turns fortune/blind

Similar (S) Dierent (D) or Partially Similar (PS) S S S D D D

7 8


4. Other perspectives Firstly, the question arises as to what is exactly meant by partial similarity. Lako and Johnson (1980) and Kovecses (2002) both refer to the partial nature of metaphorical structuring. With reference to the CM: theories are buidings, Lako and Johnson (1980: 52) explain that parts of the concept building that are used to represent the concept theory are


Diane Ponterotto

the foundation and the outer shell. This explains expressions like His theory stands on solid ground or His position has no foundation. Other parts of buildings, like the roof, the staircases and the hallways, are not used for the mapping between the two domains, theories and buildings. Similarly, using the example of arguments are buildings, Kovecses (2002) notes that the metaphor does not select other options within the source domain of buildings like doors, windows or corridors, nor does it activate functional aspects of buildings related to their social, commercial and governmental roles. Thus, two languages can have the same metaphor but select dierent source domain elements to represent dierent aspects of the target domain. The selection can derive from the possibility of activating other source domain elements within the same semantic eld, for example to bring grist to the mill in English and portare lacqua al mulino to bring water to the mill in Italian. It can also depend on cultural specicities within the same common conceptual metaphor. Using another example of the CM: life is gambling, we nd I hit the jackpot in English, but Ho fatto tredici I made thirteen in Italian, to represent the concept of victory. The Italian idiomatic expression derives from the lottery, famous in Italy, based on weekly football matches. If you rightly guess the results of thirteen matches, you can win a very big lottery sum. Secondly, Kovecses (2003: 313) proposes a slightly dierent and more complex comparative taxonomy for cross-cultural comparison of gurative meaning, which he discusses in relation to English and Hungarian.


Word form

Literal meaning

Figurative meaning

Conceptual metaphor

Conceptual metaphor and metonymy (s/d) (s/d) (s/d) (s/d) no metaphor and metonymy

Conceptual metonymy

1 2 3 4 5

dierent dierent dierent dierent dierent

same dierent dierent dierent dierent

same same same dierent same by means of literal meaning

same same dierent dierent no metaphor

(s/d) (s/d) (s/d) (s/d) no metonymy

To illustrate the applicability of this taxonomy for an English/Italian comparison, we could suggest the following examples:

Insights from Conceptual Metaphor Theory and implications


Example 1 English: Ive invested a lot of time in this relationship. Italian: Ho investito molto tempo in questo rapporto. I have invested a lot of time in this relationship. Both languages activate the CM: time is money in connection with personal relationships.
Comparison of English and Italian Word form Literal meaning Figurative meaning Conceptual metaphor Same (English: invested / Italian: investito invested ) Same Same Same (time is money)

Example 2 English: What a problem! I dont know where to start. Italian: Che problema! Non so dove mettere le mani. What a problem! I dont know where to put my hands. Here there is a distinct dierence between the two languages. The English language guratively represents the concept of diculty in resolving a problem with the CM: a problem is a journey and its entailment: resolving a problem of life requires knowing where to begin the journey. The Italian option on the other hand expresses the same concept with the CM: a problem is an object and its entailment: resolving a problem of life requires knowing where to grasp the object.
Comparison of English and Italian Word form Literal meaning Figurative meaning Conceptual metaphor dierent (English: start / Italian: mani hands) same same Dierent (English: a problem is a journey / Italian: a problem is an object)

Kovecses (2003) comments on an inroad to cross-cultural comparison which starts from a potential source domain, a concrete noun like hand


Diane Ponterotto

or heart for example, identies literal expressions for the word in both languages, and then searches for their gurative uses. In his comparative studies of Malay and English, Charteris-Black (2001b, 2002) nds that the same literal expression carries dierent gurative meanings in the two languages. As an example for our language comparison, we could suggest beginning with the word mouth (English: mouth / Italian bocca mouth). We could then identify a literal expression in both English and Italian like to have a big mouth in English and its counterpart in Italian avere la bocca grande to have a big mouth. In English the expression means to talk too much; in Italian, however, the same literal expression carries a very dierent gurative meaning, which is to eat too much. Ko vecses (2003: 315) also notes that two languages may apparently have similar conceptual metaphors but reveal subtle dierences in the culturalideological background in which the conceptual metaphor functions. This has been suggested very forcefully by Charteris-Black (2000, 2001a, 2001b, 2002), who has documented cross-linguistic and cross-cultural dierences based on both overt and covert ideological underpinnings.

5. Guidelines for applied comparative CM analysis In Ponterotto (2007), some guidelines were proposed for determining equivalency solutions for gurative expressions in an eort to help applied linguists (foreign language teachers, translators etc.) with decoding and transfer of messages hinging on the gurative plane of the expression. In other words, a reasoning strategy of the following type was proposed: 1. 2. 3. 4. Is the idiomatic expression motivated by a conceptual metaphor (CM) and its potential entailments? Does the CM exist in the other language? Does the CM motivate a similar gurative expression in the other language (i.e. with the same source domain terms)? If it does not, is there an alternative within the same CM?

We can now show how this translation strategy can lead to viable solutions: Example 1 English IE: We need to help each other; we are all in the same boat. Q. Is the idiomatic expression motivated by a CM or its potential entailments? A. yes, life is a journey

Insights from Conceptual Metaphor Theory and implications


Q. Does that CM exist in Italian? A. Yes Q. Does the CM motivate a similar gurative expression in Italian (i.e. with the same source domain term boat)? A. Yes. Dobbiamo aiutarci; siamo tutti sulla stessa barca. We have to help each other; we are all on the same boat. Example 2 English IE: Thats beside the point. Q. Is the idiomatic expression motivated by a conceptual metaphor and its potential entailments? A. Yes. an argument is a journey a journey defines a path a well-planned argument follows a straight path Q. Does the CM exist in Italian? A. Yes Q. Does it motivate a similar gurative expression in Italian (i.e. with the same source domain terms: beside/point)? A. No Q. If it does not, is there a good alternative within the same CM? A. Perhaps. Quello non centra niente that does not enter at all, which can be said to be motivated by the following CMs: an argument is a journey a journey defines a path a well-planned argument enters the right path Q. Can the gurative eect be better preserved in Italian by selecting a dierent idiomatic expression? ` A. Perhaps, in this case: Quello e come mangiare cavolo a merenda. Thats like eating cabbage at snack time.7
7. As pointed out in Ponterotto (2007), in Italy a great deal of importance is given to which foods are eaten at which times of the day. Italians do not eat ` vegetables at snack times (merenda). That is why an expression like Quello e come mangiare cavolo a merenda Thats like eating cabbage at snack time carries a strong emphasis of the meaning: things or ideas that do not go together.


Diane Ponterotto

Q. Does the Italian expression refer to the same conceptual metaphor? A. No, the Italian IE is motivated by the CM: ideas are foods. The strategy outlined in the previous section delineates metaphorical frames from the point of view of the source domains. However, another possibility would be to focus on the target domain. This would entail starting from a target domain and then searching for similarities in source domains in the two languages. For example, starting from a target domain like happiness, if we look for a gurative expression to describe a happy person, we nd in English a sunny personality and similarly in Italian una ` personalita solare a sunny personality. Thus, in order to understand how moods are represented guratively across the two languages under discussion, we could begin the search from the target domains.

The target domain: moods Courage

English IE

Italian translation

He threw fear to the wind and decided to jump.

Happiness Anger Aection Sadness Confusion

What a sunny personality! He came storming into the room They were given a warm welcome. Why are you so gloomy? I havent answered his message yet. Im a bit hazy about what to do.

Ha gettato la paura al vento e ha deciso di saltare. He threw fear to the wind and decided to jump. Che carattere solare! What a sunny personality! E entrato nella stanza come una furia. He entered the room like a fury. Hanno avuto unaccoglienza calorosa. They received a warm welcome. ` Perche sei cos cupo? Why are you so dark? Ancora non ho risposto al suo messaggio. Oggi sono un po confuso sul da farsi. I havent answered his message yet. Im a bit confused about what to do.

Thus, with this inverse procedure, nding a correspondence of target and source domains is possible.

Insights from Conceptual Metaphor Theory and implications


The target domain: emotion

English source domain elements

Corresponding element in Italian translation

English gloss of Italian translation of English source domain elements

Target/Source mapping: S similarities D dierent PS partially similar S S PS S PS D

Courage Happiness Anger Aection Sadness Confusion

to the wind sunny storming warm gloomy hazy

al vento solare furia caloroso cupo confuso

to the wind sunny fury warm dark confused

This would mean that the reasoning strategy guiding the translator follows a dierent pattern. For example, when translating the English IE: She breezed into the room, starting from the target domain, the following questioning strategy could be used: Example 3 English IE: She breezed into the room. Q. What target domain does this idiomatic expression represent? A. Happiness Q. What conceptual metaphor mapping is implied in this IE? A. mood weather (CM: mood is weather) Q. What source domain element is present in the IE? A. breeze Q. What is then the specic target/source domain mapping for the English language? A. happiness breeze Q. Are there IEs in Italian to convey happiness? A. Yes, many


Diane Ponterotto

Italian idiomatic expressions Sentirsi su Avere il morale alto Non toccare terra dalla gioia Toccare il cielo con un dito Salire alle stelle Andare in orbita

English gloss To feel up To have ones morale high Not to touch the earth from joy To touch the sky with a nger To go up to the stars To go into orbit

Q. From which conceptual metaphors are the Italian IEs derived? A. They all seem to derive from the CM: happy is up. Q. Should the translation choice in Italian search then within the CM: happy is up, rather than the CM behind the English expression, mood is weather? A. Given the productivity of the CM: happy is up in Italian, the answer would seem to be Yes. Q. Can one of the Italian expressions be said to be the best solution? A: Probably not. All are potential translation choices. 6. Further considerations: Its not as simple as all that! Now a serious problem arises in the face of idiomatic expressions which admit transformations. Let us take the case of the IE: to speak ones mind. To speak ones mind is motivated by the combination of the container metaphor ideas are in the mind plus the conduit metaphor ideas are sent via words (from one person to another). If we consider the concept mind as a cognitive frame, we can suggest that the frame includes a set of conceptual metaphors and their entailments.8
8. The term frame is used here according to the denition given in Evans (2007: 8586): a schematisation of experience (a knowledge structure) which is represented at the conceptual level and is held in long term memory and which relates elements and entities with a particular culturally embedded scene, situation or event from human experience. If, according to Lako and Johnson (1980), metaphor is constitutive of the way we conceptualize and represent the world, then we can see how conceptual metaphors can contribute to the formation of knowledge structures or frame-like schematisations of human experiences.

Insights from Conceptual Metaphor Theory and implications


conceptual metaphor (container): ideas are in the mind conceptual metaphor (conduit): ideas are transmitted via words conceptual metaphor entailment (container conduit): Ideas are objects in the mind which are sent via words from one person to another. (cf. Lako and Johnson 1980; Reddy 1979): So, in English we nd: to have something in mind to have an empty mind to put something out of your mind to stick in your mind to be full of ideas to be bursting with ideas in my mind its all in the mind to send word to someone to send a message to send your love to give someone ideas Let me point out something to you Ill give you an idea He got that idea across very well The message doesnt seem to be getting across The Italian translation of the expression to speak ones mind (according to dictionary entries) is parlare chiaro to speak clearly. However, the Italian expression parlare chiaro to speak clearly can be said to be motivated by a dierent CM: understanding is seeing ideas are light-sources discourse is a light-medium (cf: Lako and Johnson 1980: 48) Now, the idiomatic expression, to speak ones mind can undergo transformations, as in the following cases, cited by Glucksberg (1993: 16):


Diane Ponterotto

1. When drugs are involved its time to speak your parental mind. 2. Did he nally speak his mind? 3. As a diverse but purposeful group, you should speak your minds. 4. He spoke his mind 5. The tenants association nally spoke their collective minds.

Adjectival modication Adverbial modication Quantication Tense marking All of the above

The following equivalents in Italian can be suggested:9

English 1. When drugs are involved its time to speak your parental mind. 2. Did he nally speak his mind? 3. As a diverse but purposeful group, you should speak your minds. 4. He spoke his mind 5. The tenants association nally spoke their collective minds. Italian Quando si arriva alla droga, ` e necessario che i genitori prendano una posizione molto chiara e decisa. Ha nalmente parlato chiaro? Essendo un gruppo eterogeneo e risoluto, dovreste parlare tutti in modo chiaro. Ha parlato chiaro. I membri dellassociazione degli inquilini nalmente hanno parlato tutti in unisono e in modo molto chiaro. English gloss of Italian translation When its about drugs, it is necessary for parents to assume a very clear and rm position. Did he nally speak clearly? Being a heterogeneous and determined group, you should all speak clearly. He spoke clearly. The members of the tenants association nally all spoke in unison and very clearly.

9. Professional translators, teachers of translation and Italian/English bilingual speakers were often consulted about these suggestions. It is interesting to note that where gurative expressions were concerned, rarely did they oer the same solutions and rarely were they entirely certain of their suggestions. I would like to take the opportunity to thank Dr. Laura Bocci, Prof. Roberta Falcone and Prof. Jeanne Clegg for their helpful comments.

Insights from Conceptual Metaphor Theory and implications


Now we know that semantically productive operations serve communicative functions. Thus, depending on the context and text-type (cf. MerliniBarberesi 2002; Masi 2007; Tabakowska 2005), a translator may want to (or have to) preserve the communicative force produced by the idiomatic characteristics. And actually he can only do so by imitating the type of transformation involved. Although this is possible for some cases as for (2) Did he nally speak his mind? or for (4) He spoke his mind, it is impossible for (1) When drugs are involved its time to speak your parental mind, or for (3) As a diverse but purposeful group, you should speak your minds, given the dierences in grammatical structure between the two languages. In (1), the grammatical structure of the English utterance includes a VP (V NP) speak your parental mind, whereas the Italian VP structure is verb adverb ( parlare chiaro to speak clearly). The English structure then permits adjectival modication of the noun, giving parental mind. In Italian, the connotation of parental duty can be maintained only by adding the literal paraphrase ( prendere una posizione molto chiara e decisa to assume a very clear and rm position). In (3), where quantication is involved, the mark of plurality in English is highlighted in the noun of the VP: speak your minds, whereas in Italian it appears inectionally in the verb (2nd person plural) with the addition of a quantier tutti all. As for example (5), which includes all the transformations, the translation requires nding a way to convey the notion of collectivity, collective minds, and a solution could be the addition of an adverb triggering a music metaphor, in unisono in unison and further adverbial modication, in modo molto chiaro very clearly. Thus, whereas the assumption of a conceptual metaphor perspective can be said to be a useful tool for translation practice, as shown above, it may also be the case that the interaction of grammatical patterns with conceptual metaphor will determine a linguistic, contextual and cultural specicity which may be very dicult to transfer over to another language. As the last point, I would like to further explore the complexities involved in translating idiomatic expressions triggered by CMs in conversational exchanges. Below is an example taken from Glucksberg (1993: 23): Ken: Dont worry. Ill cross that bridge when I come to it. Ann: By that time they will have burnt it down. Obviously the response of Ann implies a complex cognitive operation, which I would dene as meta-metaphorical, metalinguistic and metacogni-


Diane Ponterotto

tive. She does not respond to Kens message (Im going to worry about the problem later). She responds to the source domain word, bridge, and by so doing: acknowledges the metaphorical status of Kens utterance recodies which CM is being activated literalizes the tropological function then in a cooperative mode, re-metaphorizes it in order to remain within the discursive strategy chosen by Ken and adds an evaluation of Kens behaviour through a trope of irony. This is a case where a literal paraphrase will not do, as has been suggested by Gibbs (1994), because that would mean losing too much linguistic content: semantic: loss of the connotative meaning gurative: loss of the tropological form functional: loss of the communicative intent. Thus a gurative solution seems to be necessary. And in Italian, this is possible. Since the Italian language also encodes the life is a journey metaphor, the translation could be: ` Ken: Quando arrivo al ponte, lo attraversero. When I arrive at the bridge, I will cross it. ` Ann: A quel punto, lavranno gia bruciato. By that point, they will have already burned it. However, in the case of mind metaphors, where the Italian language does not present a comparable conceptual/idiomatic pattern, the translation strategy is more dicult. To stick to our earlier example to speak ones mind, let me use an exchange from a spontaneous conversation between two speakers of American English. A: Im going to give him a piece of my mind. B: By the looks of you, thats going to be a very big piece. Here the metaphorical map regarding mind, which would be: ideas are in the mind ideas are transmitted via words (from one person to another) seems also to acquire:

Insights from Conceptual Metaphor Theory and implications


the mind is an object (which can be divided into parts) Again, the response of interlocutor B implies a complex cognitive operation: meta-metaphorical, metalinguistic and metacognitive. B does not respond to the message of A (Im going to tell him bluntly, exactly and unsparingly what I think). B responds to the source domain word, piece, and again, by so doing: acknowledges the metaphorical status of As utterance recodies which CM is being activated literalizes the metaphorical function then in a cooperative mode, re-metaphorizes it in order to remain within the same discursive strategy chosen by A and adds an evaluation of As behaviour through a trope of irony Now here it is very dicult to convey the communicative intent of the English exchange in the absence of a comparable metaphor in Italian. The question of the most viable solutions for this exchange was posed to professional translators and teachers of translation. I would like to discuss two of the suggestions given to me: Suggestion 1 A: English: Im going to give him a piece of my mind. ` Italian: Gli diro esattamente quel che penso. I am going to tell him exactly what I think. B: English: By the looks of you, thats going to be a very big piece. ` Italian: Dallo stato in cui sei, non sara una cosa da poco. Considering the state you are in, its not going to be something small. With this rst solution, there is no transfer of the iconic image triggered by the source domain term piece. Thus, the metacognitive operation along with the tropological passage from metaphor to irony is lost, as is, consequently, the discursive emphasis. Suggestion 2 A: English: Im going to give him a piece of my mind. ` Italian: Gliene diro quattro. I am going to tell him four.


Diane Ponterotto

B: English: By the looks of you, thats going to be a very big piece. ` Italian: Dallo stato in cui sei, sembrerebbe piu come otto. Considering the state you are in, it would seem to be more like eight. The second solution seems to preserve the gurative strength of the original IE. In Italian, the IE: dirgli quattro to tell him four carries an intention very similar to the English IE: to give him a piece of my mind. We can also note that, both in Italian and in English, to have a word with someone means the desire to express disagreement or even slight anger (in English we say, I want to have a word with you; in Italian, we also say avere una parola to have a word). In Italian, then, to use four words rather than one expresses the intention of communicating anger forcefully. We could hypothesize that the Italian language encodes the conduit metaphor in the concept of transmitting words, and implicitly these words represent ideas in the mind. It does not present, however, the CM: the mind is an object (which can be divided into parts), which accounts for the English a piece of my mind. On the other hand, the Italian language adds the CM: more is stronger (i.e. more words means more anger; more numbers means more strength). With this solution, let us describe exactly what happens. The two languages coincide partially in that their textual representations activate some but not all the possibilities of the conceptual metaphor frame. The following table illustrates these similarities and dierences.

English IE Interlocutor A Interlocutor B Im going to give him a piece of my mind. By the looks of you, thats going to be a very big piece.

Italian ` Gliene diro quattro I am going to tell him four Dallo stato in cui sei, ` sembrerebbe piu come otto. Considering the state you are in, it would seem to be more like eight.

Same (S) or Dierent (D) CM S ideas are in the mind ideas are in the mind

Insights from Conceptual Metaphor Theory and implications CM S


ideas are transmitted via words from one person to another

ideas are transmitted via words from one person to another more is stronger


the mind is an object (which can be divided into parts) piece

Source domain elements: Interlocutor A Source domain elements: Interlocutor B

the number four

a very big piece

the number eight

` By choosing this option in Italian (Gliene diro quattro Im going to tell him four), the translator is able to partially preserve the metaphorical strength of interlocutor As comment by rendering interlocutor Bs response with a greater number, otto eight and producing Dallo stato in cui sei, ` sembrerebbe piu come otto Considering the state you are in, it would seem to be more like eight. Although the iconic coding of the English source domain (pieces of an idea) to express anger is lost, an equally gurative eect of intense anger remains (numerical quantity of ideas). The communicative force of interlocutor Bs response in commenting on interlocutor As agitated state is better preserved because the translation solution manages to respect the shift from the tropological plane of metaphor to that of irony. This is a prime example of how conceptual frames in two languages manifest some similarities and some dierences. It is also an example of how, within the possibilities of a conceptual metaphor, translators choices are constrained not only by intercultural variation in conceptualization but also by language form, text-type and communicative intentions.


Diane Ponterotto

7. Conclusions On the basis of the assumption that metaphor derives from human experience and conceptualization, it is thought to be characterized by similarity across cultures and sometimes claimed to be universal. Research on gurativity has revealed, on the other hand, a distinctive presence of crosslinguistic dierences (cf. Boers 2003; Kristiansen and Dirven 2008). Research on idiomaticity from the cognitive linguistics viewpoint has identied these dierences in the variation of cross-cultural conceptualizations (cf. Kovecses 2002, 2004). This type of research has generated studies in numerous languages representing widely dierent linguistic typologies and cultures (cf. for example Berendt 2008). This fact has signicant implications for many elds of application, such as intercultural communication, foreign language learning/teaching and contrastive analysis and translation practice (cf. Kristiansen, Achard, Dirven and Ruiz de Mendoza 2006). As the nal point we can stress, along with Deignan and Potter (2004), that, although there has been innovative research on the question of similarity versus dierence across languages and cultures, more work needs to be done on the extent to which languages coincide and diverge in their use of conceptual metaphors. It should be emphasized that careful weighting of the quantity and quality of cross-linguistic and cross-cultural dierences is an important direction for future research, especially in applied cognitive linguistics. Such a direction would imply study of the formal constraints that individual languages impose on apparently common conceptual metaphors. Moreover, it would imply paying closer attention to how the many and complex variables lying behind metaphorical conceptualization interact to yield idiomatic expressions. Although conceptual metaphor has in a certain sense a universal aspect, being grounded in human embodied experience, it is subject to a wide range of variation stemming from cultural dierence, language specicity, text-typology and message context. Only through an understanding of the interaction of such variables is it possible to compare conceptual metaphors across cultures. References
Barcelona, Antonio 1997 Clarifying and applying the notions of metaphor and metonymy within cognitive linguistics. Atlantis 19(1): 2148. Berendt, Eric (ed.) 2008 Metaphors for Learning: Cross-cultural Perspectives. Amsterdam/ Philadelphia: John Benjamins.

Insights from Conceptual Metaphor Theory and implications Boers, Frank 2003


Applied linguistics perspectives on cross-cultural variation in conceptual metaphor. Metaphor and Symbol 18(4): 231238. Charteris-Black, Jonathan 2000 Figuration, lexis and cultural resonance: A corpus based study of Malay. Pragmatics 10: 281300. Charteris-Black, Jonathan 2001a A comparative study of metaphor in Spanish and English nancial reporting. English for Specic Purposes 20: 249266. Charteris-Black, Jonathan 2001b Cultural resonance in English and Malay gurative phrases: The case of hand. In: Janet Cotterill and Anne Ife (eds.), Language Across Boundaries (British Studies in Applied Linguistics series), 151170. London: Continuum. Charteris-Black, Jonathan 2002 Second language gurative prociency: A comparative study of Malay and English. Applied Linguistics 23: 104133. Deignan, Alice and Liz Potter 2004 A corpus study of metaphors and metonyms in English and Italian. Journal of Pragmatics 36(7): 12311252. Evans, Vyvyan 2007 A Glossary of Cognitive Linguistics. Edinburgh: Edinburgh University Press. Forceville, Charles J. and Eduardo Urios-Aparisi (eds.) 2009 Multimodal Metaphor. (Applications of Cognitive Linguistics 11) Berlin/New York: Mouton de Gruyter. Geeraerts, Dirk and Hubert Cuyckens 2007 Handbook of Cognitive Linguistics. Oxford: Oxford University Press. Garzanti/Hazon 2001 Dizionario Inglese-Italiano/Italiano Inglese. Milan: Garzanti. Gibbs, Raymond W. 1990 Psycholinguistic studies on the conceptual basis of idiomaticity. Cognitive Linguistics 1: 417451. Gibbs, Raymond W. 1994 The Poetics of Mind: Figurative Thought, Language and Understanding. Cambridge: Cambridge University Press. Gibbs, Raymond W. and Jennifer OBrien 1990 Idioms and mental imagery: The metaphorical motivation of idiomatic meaning. Cognition 36: 3568. Glucksberg, Sam 1993 Idiom Meanings and allusional content. In: Cristina Cacciari and Patrizia Tabossi (eds.), Idioms: Processing, Structure and Interpretation, 326. Hillsdale: Lawrence Erlbaum.


Diane Ponterotto

Hiraga, Masako 1991 Metaphor and comparative cultures. In: Paul G. Fendos (ed.), Cross-cultural Communication: East and West, vol. II, 149166. Taiwan: National Cheng-Kung University. Kovecses, Zoltan 1986 Metaphors of Anger, Pride and Love: A Lexical Approach to the Structure of Concepts. Amsterdam/Philadelphia: John Benjamins. Kovecses, Zoltan 1990 Emotion Concepts. New York: Springer. Kovecses, Zoltan 2002 Metaphor: A Practical Introduction. Oxford: Oxford University Press. Kovecses, Zoltan 2003 Language, gurative thought and cross-cultural comparison. Metaphor and Symbol 18(4): 311320. Kovecses, Zoltan 2004 Metaphor in Culture: Universality and Variation. Cambridge, UK: Cambridge University Press. Kristiansen Gitte, Michel Achard, Rene Dirven and Francisco J. Ruiz de Mendoza (eds.) 2006 Cognitive Linguistics: Current Applications and Future Perspectives. Berlin/New York: Mouton De Gruyter. Kristiansen, Gitte and Rene Dirven (eds.) 2008 Sociolinguistics: Language Variation, Cultural Models, Social Systems. Berlin/New York: Mouton De Gruyter. Lako, George and Mark Johnson 1980 Metaphors We Live By. Chicago: University of Chicago Press. Masi, Silvia 2007 English and Italian dialogic descriptors across text-types. In: Marcella Bertuccelli-Papi, Gloria Cappelli and Silvia Masi (eds.), Lexical Complexity: Theoretical Assessment and Translational Perspectives, 157175. Pisa: Plus. Merlini-Barbaresi, Lavinia 2002 Text linguistics and literary translation. In: Alessandra Riccardi (ed.), Translation Studies: Perspectives on an Emerging Discipline, 120132. Cambridge, UK/New York: Cambridge University Press. Ponterotto, Diane 2007 Translating idiomaticity: a cognitive linguistic viewpoint. In: Annalisa Baicchi (ed.), Voices in Translation: Linguistic, Multimedia and Cognitive Perspectives. R.I.L.A. (Rassegna Italiana di Linguistica Applicata), 317335. Rome: Bulzoni. Reddy, Michael J. 1979 The Conduit Metaphor: A case of frame conict in our language about language. In: Andrew Ortony (ed.), Metaphor and Thought, 164201. Cambridge: Cambridge University Press.

Insights from Conceptual Metaphor Theory and implications


Tabakowska, Elzbieta 2005 Point of view in languages, texts and translations. In: Annalisa ` Baicchi, Cristiano Broccias and Andrea Sanso (eds.), Modelling Thought and Constructing Meaning: Cognitive Models in Interaction, 198212. Milan: Franco Angeli.

Mundane Transcendence? Conceptualisations of faith in prosperity theology Magorzata Pasicka

1. Introduction The purpose of the present paper is to apply some essential elements of the cognitive approach to an analysis of conceptualisations that lie at the very heart of the doctrine known as Faith Movement (FM), a pragmatically oriented branch of Protestantism (briey discussed in section 2). While Pentecostal and charismatic Christianity in general has received some attention from linguists (cf. e.g. Coleman 1980), FM discourse has not yet to the best of my knowledge been subject to linguistic investigation. By deconstructing some syntactically novel structures and metaphorical mappings pervading the discourse, I hope to attract some attention to the potential it oers to a linguist interested in the interplay between language and theology, and in the ways a pragmatic shift in doctrine evinces itself in God-talk. The analysis will mostly draw on blending theory (BT), involving the mental spaces model proposed by Turner and Fauconnier (cf. Turner and Fauconnier 1995; Fauconnier and Turner 1998), and on the conceptual metaphor theory (CMT), involving source domain (SD) to target domain (TD) mappings (cf. Lako and Johnson 1980; Lako and Turner 1989). The mental space model applied in BT has proved useful in analysing syntactic structures as reections of conceptual constructs (cf. Fauconnier and Turner 1998), and as such it will be applied in the discussion of the syntactic innovations characteristic of FM discourse about faith. An enrichment of CMT, the BT has opened up new possibilities of interpreting complex and novel metaphors, which can hardly be accounted for by simple oneto-one mappings. Researchers working on the application of BT to mental constructs have noted that whereas CMT has been primarily concerned with identifying regular, conventional patterns of metaphorical conceptualisation, BT has often explicitly addressed itself to novel and unique examples (Grady, Oakley and Coulson 1999: 5). Since the metaphorical imagery conjured up by FM evangelists is often rather conventional (though outside the theological framework the convention seems totally


Magorzata Pasicka

unconventional), the application of CMT in our analysis seems justied. Our analysis of the linguistic facet of metaphorical mappings of faith will not inch from certain uneasy questions some of those images may provoke when viewed in close-up (section 5). Throughout the paper, we will consider examples of conceptual integration and conceptual metaphor drawn from texts published in mainstream FM monthly magazines, Adventures in Faith (AF), Believers Voice of Victory (BVV), and The Word of Faith (WF), published in the last few years.

2. Doctrinal background Faith Movement emerged within charismatic/Pentecostal Christianity in America in the 1960s. Its various names, such as Health and Wealth, Prosperity theology or Positive confession theology, reect the extreme pragmatism of FM: apart from the tenets shared by all charismatics (such as stressing the importance of being born again and experiencing Spirit baptism along with its physical manifestations), its adherents hold a set of quite distinctive beliefs, the main one being that health and prosperity are in fact integral parts of salvation: Christs sacrice annulled not only sin, but every eect of sin, such as poverty and illness. Health and wealth are among the blessings enumerated in the Bible and are there for Christians to claim. The Bible is in fact viewed as a set of positive, i.e. practical, proposals or oers made by God to cover every human need. What you have to do is to nd scriptures that cover your case (Word of Faith Oct. 2000) and profess your belief in them coming true. The tenet has come to be known as positive confession the phrase itself literally referring to bringing into existence what we state with our mouth, since faith is a confession (Burgess and McGee 1988, s.v. Positive Confession Theology). How is this made possible? Once a man is born again, his frame of mind changes in accordance with the Word of God, he himself acquires divine nature and gains the power to constitute a new reality by uttering words, i.e. scriptural passages, with faith, i.e. in FM terms by confessing. Persistence in confessing ones faith in receiving what one hopes to receive (calling those things which be not as though they were, Rom. 4: 17) is usually seen as the key to success. The doctrine of positive confession involves a concept of faith that is not only pragmatic, but mechanistic: no active divine participation seems to be needed for man to draw blessings from the spiritual realm into the

Conceptualisations of faith in prosperity theology


physical world by the power of faith alone. Gods role seems limited to an automatic response to mans confession of faith.

3. Conceptual blends and syntactic innovation The extreme pragmatism of prosperity theology can be found on the syntactic level in the peculiar use of the verb believe in a distinctive novel collocation: to believe (God) for sth as illustrated by the example below (cf. Pasicka 2006 for a discussion of the expression in the argument structure framework): (1) In my rst year as a Bible student at Rhema Bible training Centre, my wife, Phyllis, and I really needed a good used car that would run well and get us from point A to B. So we believed God for it, He worked in response to what we were able to believe Him for, and we got one (BVV). The structure of the phrase to believe (God) for sth is actually analysable into two segments of meaning, reecting the two steps in the act of faith, where believing God is a prerequisite of believing God for sth. In order to get hold of the desired spiritual or material good, one must rst believe in God and believe (i.e. trust) Him and His promises, and only then direct ones faith towards the promises concerning the particular good. Since God has accomplished His part of the task, and His Word contains practical oers for those who believe it, mans task consists in nding the right scriptural passages: (2) Whatever youre praying and endeavouring to believe God for, nd scriptures that cover your case. And then stand rm (WF). The above-mentioned segmentation of the concept <believe God for sth> prompts an analysis in terms of a conceptual blend. Notice that the syntax resembles that of a prototypical verb expressing a wish directed to the potential agent able to fulll the wish (the agent appearing in the role of direct object), such as to ask sb for sth, while carrying over the FM-specic semantics of believing God, which includes both the conviction that whatever is being expected will be granted and the Calvinist belief that material wellbeing is a sign of divine election. The novelty of the structure results from the blending of two concepts: asking God for sth and believing God. The conceptual integration process postulated here can be represented as follows:


Magorzata Pasicka

Figure 1.

The problem remains how to account for the frequent omission of God in the structure, as in the example below: (3) If youre believing for restoration in your nances, then become engrossed with Phillipians 4:19: But my God shall supply all your need according to his riches in glory by Jesus Christ (AF). Seeing that the believe-structure inherits its syntax from the input space ask sb for sth, we can account for the ellipsis of the direct object in the believe-structure exactly as we do in the case of the ask-structure. Of the two participants involved in the event of asking/believing, the addressee (here: God) need not be specied and may be left implicit. Syntax allows

Conceptualisations of faith in prosperity theology


for the direct object in the active construction not to have a linguistic exponent. Such lack of specication is most likely to occur if the participant acting as the direct object is assumed to be known and/or is considered unimportant (cf. Schlesinger 1995: 59). In cognitive terms, we might say that frequent omission of God from the prole is likely to lead to removing the concept <God> from the base as well. Besides the ubiquitous believing (God) for sth, another novel syntactic structure appearing in the context of positive confession at work is to believe sth in. The meaning deduced from linguistic data is to obtain whatever one believes for, as in the sentence below: (4) [God to evangelist:] They asked you how did you get this [i.e. car] and you said you believed for it. But whose faith was it that you believed the car in with? (BVV) Again, the novelty of the structure reects a FM-specic conceptual novelty. The verb believe in this collocation refers to the nal stage of the positive confession procedure, whereby faith-lled words [uttered by man M.P.] draw things out of the spiritual realm, where they do exist [. . .] into the natural realm, where they do not exist yet (from a letter sent out by Joyce Meyer Ministries, April 2002). The whole faith process encompassed in the believe sth in construction involves an action (here: believing for a car), which brings about a reaction: the desired good (here: a car) materialises. The collocation to believe sth in (with the use of faith as an instrument) strongly resembles the non-prototypical cases of caused-motion constructions analysed by Adele Goldberg (1995) and later discussed by Fauconnier and Turner (1998). All three authors deal with semantically disturbing cases of verbs that do not themselves specify caused motion, but can be used in caused-motion constructions. On Fauconnier and Turners analysis, the [NP VP NP PP (directional)] caused-motion structure exemplied by a sentence coined by Goldberg: Frank sneezed the napkin o the table can be viewed as a blend integrating the syntax of prototypical caused-motion constructions (e.g. He threw the napkin o the table) with the conceptually decomposed (i.e. unintegrated) causal sequence: someone sneezes, napkin moves o the table (as a result). The verb sneeze in the blend highlights the causal agents action, but verbs in caused-motion constructions may also highlight the objects motion (e.g. Sarge sped the toy car around the Christmas tree) or the causation itself (Sarge let the tank into the compound) (cf. Fauconnier and Turner 1998 on-line version: 5). The believe sth in structure is actually analogous to the sneeze sth o structure: in both cases it


Magorzata Pasicka

Figure 2.

is the causal agents action that is being mapped. The mechanism of the conceptual integration involved here can be represented as follows (I use the abbreviations introduced in Fauconnier and Turner 1998, where a stands for agent, o for object, e for causal action and dm for direction of movement). 4. Metaphorical mappings Apart from such innovative syntactic constructions, the mechanistic approach to faith in FM is reected in metaphorical mappings. In conjuring up images of positive confession at work, FM evangelists often resort to the realm of technology, drawing on the readers basic technical knowledge. The following passage, referring to the mechanism of faith healing, contains one of the most striking images:

Conceptualisations of faith in prosperity theology


(5) But when they dont see results right away, they turn o the switch of faith, and the healing power quits owing and the recovery stops. Lets allow Gods healing power to continue by keeping the switch of faith on as we pray according to the Word of God for the healing of the brethren (WF). Expressions such as switch (of faith) and ow (of the healing power) encourage immediate association with an electrical device, where pressing a control button triggers the ow of electricity, which in turn puts the whole mechanism in motion. On the face of it, the image exploited in (5) seems clear: the ow of Gods healing power, conceptualised as electricity, can be activated by switching on faith. On closer examination, however, an uninitiated reader may be struck by the ambiguity of the expression the switch of faith. Is faith itself the control button triggering the ow of healing in the electrical circuit, or is it part of the current owing in a more complex circuit, where faith owing in one direction (i.e. towards God) is automatically converted into healing power owing in the opposite direction (i.e. towards man)? The following theological explication leaves us in no doubt about how to read the metaphor: (6) Faith is something you have but until you release it, it is as though you have none. Faith has to come out of you. It has to become an act of believing. Every kind of power known to man must have a point of contact to release it. With electricity, for example, you ip the light switch to send the power into the bulb which in turn lights the room. In the same way, your point of contact releases (sends) your faith to God which releases His power back into your life (Roberts, undated: 45). The image is thus of an electrical device (here: lamp) which requires electricity ow in order to give light and/or warmth in return. The image involves two conceptual metaphors: faith is electricity and healing power is light/warmth, plus a metonymy wherein <electricity> stands for <electrical device>: turning on faith is analogous to turning on an electrical device. The switch of faith in (5) refers to the point of contact in (6) releasing faith and thus triggering the ow of the healing power in return. FM theology tells us that anything can perform the role of this switch (be it prayer, the laying on of hands, Holy Communion, or simply a meaningful encounter), provided that it is accompanied by a decision to profess ones faith. Recovery begins the moment one consciously starts confessing ones belief in healing. However, the circuit


Magorzata Pasicka

only works as long as the switch is turned on. Therefore, in order to receive a full recovery one must keep believing. Though the electricity metaphor may be prevalent in the pragmatically oriented texts, it is by no means the only metaphorical conceptualisation of the mechanistic kind. The mental images conjured up by the following sentence may vary: (7) So were going to look this month at how to activate faith, or how to turn your faith loose (WF). Though the idea of activating faith is consistent with the image of a switch operating the ow of electricity in a circuit, the verb to activate, with its meaning to cause something to start working, has a more general application and will evoke an image of any kind of machinery. The perception of positive confession in terms of a mechanism controlled by man evinces itself in lexical choices referring to its eciency, especially in the frequent use of the verb to work in connection with faith. Excerpt (8) exemplies the appearance of the same verb with reference to divine actions, which makes the reader aware of the same mechanistic approach being applied to God himself (cf. example (1), where the VP to work in response reects the idea of an automatic divine action triggered by a specic action performed by man1). Notice that in example (9) and in the last use in (8) the verb to work applies not to God himself, but to the whole system of man acting and God reacting: (8) But my God shall supply all your need according to his riches in glory by Jesus Christ (Phil. 4: 19). Since he supplies all my needs, I dont have any! You see, believing Gods Word like that brings Him to work on my behalf. This [i.e. positive confession] isnt something that works just in an emergency. This is a life style. This is the way youre to live (WF). (9) Im not telling you something I made up. This [i.e. positive confession applied to prayer] works! I have had more prayers answered in my life since 1999 in such a short time frame, and I know that I have tapped into something from God (AF).
1. The approach to faith inherent in the FM doctrine may actually be viewed as a magical procedure. According to some denitions, magic is based on a belief that certain actions performed by man have an immediate eect on the spiritual realm and automatically evoke a predictable response from the deity.

Conceptualisations of faith in prosperity theology


Other manifestations of the mechanistic approach to faith include the perception of the positive confession doctrine in terms of a formula to be applied whenever the situation calls for it. Hence the tendency to stress productivity as well as eectiveness, both of which can be exactly predicted and measured, as illustrated by example (11) below. What we encounter here is the positive confession is a formula metaphor, on the linguistic level rendered directly as principle of faith in (10). Since the common usage of the verb to work covers not only the successful application of mechanical devices (e.g. I shook the calculator and it works now) but mental operations as well (e.g. I tried the dierential calculus several times before it nally worked and I got the result given in the key), lexical choices such as work (V) and results (N) as manifestations of the SD structure are in compliance with the Invariance Principle: (10) The Bible says, . . . whatsoever shall say and shall no doubt in his heart, but shall believe that those things which he saith shall come to pass; he shall have WHATSOEVER HE SAITH (Mark 11: 23). [. . .] For instance, notice the same principle of faith in operation in Mark 5: 2534 in the case of the woman with the issue of blood (WF). (11) Have you ever prayed for something over and over and still you didnt get results? Have you prayed the Prayer of Faith and every other kind of prayer that you knew to pray and it [i.e. positive confession applied to prayer M.P.] didnt work? [. . .] This is the secret I want to tell you. Your faith will produce, proportionately speaking, no more than the intensity of your desire (AF). (12) Begin to confess with your mouth the scriptures concerning your situation. There is a verse for everything youre going through. God covered it all when He wrote the Bible [. . .] your intensity [in confessing M.P.] will produce results (AF). Example (11) above, apart from assuring the reader of the formulaic character of positive confession, draws on his transactional associations, with the return (here: blessings) being commensurate with the investment (here: faith). It is interesting to note that the same transactional approach is typical of FMs way of talking about the Bible. The eectiveness of the Bible in problem solving largely depends on how accurately the formula is applied:


Magorzata Pasicka

(13) If we go for 100 percent of His Word, we get 100 percent results. If we go for 50 percent of His Word, we get 50 percent results, and so on (BVV). In order to illustrate the input-output relation, FM evangelists often resort to the image of a farmer sowing seeds and reaping crops, where sowing is an investment in the future harvest season. The following excerpt refers to the necessity of practising the Word of God in order to achieve material success guaranteed by the divine covenant. The concepts of <sowing> and <reaping> (enjoying the harvest season in (14) below) in the harvest metaphor function as source domains mapped onto the target domains <practising the Word> and <having results> (here: material success), respectively. By analogy to growing crops, practising the Word of God is a long-term investment that does not yield results in an instant, but requires patience. Hence the stress on reaping in season as a prerequisite of a really fruitful harvest: (14) Everyone is always talking about sowing and reaping in their nances and in other areas. But no one wants to wait for the harvest season. Many people want to sow today and reap tomorrow. But it doesnt work that way. If youre going to learn how to reap in season if youre going to learn how to enjoy the harvest season you will have to start practising the Word (WF). The remaining two passages contain a slight modication of the harvest metaphor, wherein faith comes in as an additional element of the investment: (15) Youve got to practise the Word. Someone might say, Im believing God for my nances. But when the oering plate comes by, he doesnt put anything in. He is not putting into practice Gods laws of giving and receiving set forth in His Word. Another man who has only a quarter puts that quarter in the oering bucket when it comes by and says, Lord, this is all Ive got, but Im giving. Im expecting a return according to the Word of God. Praise the Lord! Im practising the Word! The Word says, Give and it will be given to me [Luke 6: 38]. And its not very long until this man comes out on top nancially (WF).

Conceptualisations of faith in prosperity theology


(16) Then he got hold of the principle of sowing. He laid hold of the principles of prosperity taught by Kenneth Copeland ministries and others. He started sowing and believing, sowing and believing and God brought him out, totally. God brought him from the bottom to the top. He abounded in the provision of God and in anointings to do the work of God (BVV). The metaphorical structure of (15) and (16) can be rendered in the form of the following series of mappings:
<sowing> (SD) <believing> > <reaping> (SD) <practising the Word> (TD) <believing> > <having desired results> (TD) <giving> (specication) (TD) <believing> > <receiving> (specication) (TD)

Another mundane specication of the harvest metaphor can be found in the following excerpt from a praise report, i.e. testimony sent by believers in appreciation of evangelists: (17) After talking to my wife, we sowed a $35 seed into your ministry. Right after we sowed our seed, the interviewer called to interview me over the phone. Shortly after that, I received a call informing me that I had gotten the job. We are extremely excited. [. . .] We want to thank you and your ministry from the bottom of our hearts for receiving our gift and for being such great Godly ground to plant into and for your prayers (AF). The transactional metaphorical mapping coined by the pragmaticallyminded evangelists involves conceptualising faith as such in terms of nancial investment. The passage below is a good instance of the Invariance Principle at work: (18) You need to build your faith. You see, when you were born again, you were given the measure of faith. We have all been given the measure of faith, not a measure, but the measure. What you do with your faith depends on how much your measure grows. [. . .] Its like when you get paid. For example, when you get paid on Friday, you take your check to the bank and make a deposit. Right? During the week, you make withdrawals on what youve deposited. If you only put $200 in your account, and you make a withdrawal of $150, you only have $50 left. If you write a check for $75, then you are overdrawn by $25, and there is a decit in your account. You dont have enough. So what do you do? You have to make more deposits to have enough to support


Magorzata Pasicka

you. Its the same way with your faith. You have to make deposits every day. Why? Because you make withdrawals all day long on your faith. How do you make deposits with your faith? By hearing the Word over and over. When you do that, your spirit man will get so built up that you become intense about your breakthrough. (AF) To someone acquainted with the prosperity doctrine any association with money in the context of faith will seem most natural. The image in (18), involving the faith is money concep