Studies in Language and Cognition [1 ed.] 9781443803151, 9781443801744

Citation preview

Studies in Language and Cognition

Studies in Language and Cognition

Edited by

Jordan Zlatev, Mats Andrén, Marlene Johansson Falck and Carita Lundmark

Studies in Language and Cognition, Edited by Jordan Zlatev, Mats Andrén, Marlene Johansson Falck and Carita Lundmark This book first published 2009 Cambridge Scholars Publishing 12 Back Chapman Street, Newcastle upon Tyne, NE6 2XX, UK British Library Cataloguing in Publication Data A catalogue record for this book is available from the British Library Copyright © 2009 by Jordan Zlatev, Mats Andrén, Marlene Johansson Falck and Carita Lundmark and contributors All rights for this book reserved. No part of this book may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, electronic, mechanical, photocopying, recording or otherwise, without the prior permission of the copyright owner. ISBN (10): 1-4438-0174-7, ISBN (13): 978-1-4438-0174-4

TABLE OF CONTENTS

1. Bringing language and cognition back together again............................ 1 Jordan Zlatev, Mats Andrén, Marlene Johansson Falck and Carita Lundmark I. Linguistic Meta-Theory and General Issues 2. The true nature of typological linguistics .............................................. 19 Esa Itkonen 3. Reassessing the project of linguistics .................................................... 30 Alexander Kravchenko 4. Linguistic relativity, mediation and the categorization of motion ......... 46 Johan Blomberg and Jordan Zlatev 5. Can we tell what we said when we hear ourselves saying something else? ........................................................................................................... 62 Andreas Lind, Lars Hall, Petter Johansson and Sverker Sikström 6. Tests of true pictorial competence in chimpanzees: The case of drawings ................................................................................. 76 Tomas Persson II. Lexical Meaning 7. The construal of entity permanence....................................................... 97 Jan Svanlund 8. On the prototypicality of dimensional adjectives ................................ 111 Elena Tribushinina 9. Basic and non-basic color terms in Hungarian .................................... 129 Kornélia Papp

vi

Table of Contents

10. Everyday experience in word meaning: How an associative experiment reveals it................................................................................ 142 Svitlana Martinek 11. What can self-organizing maps reveal about the structure of emotion concepts? ................................................................................................. 160 Toomas Kirt and Ene Vainik 12. A three-dimentional format for modelling lexixal organization ........ 171 Michael Fortescue III. Metaphor 13. More than a metaphor: Metaphoric strategies in dynamic processes of meaning construction .......................................................................... 193 Claudio Vandi and Riccardo Fusaroli 14. Going towards the unknown: Expressions for dying in European languages ................................................................................................. 209 Anna Vogel 15. The Swedish adjective varm (‘warm’): A comparison between lexical meanings and associations ........................................................... 218 Annika Bergström 16. Domain knowledge in children’s understanding of metaphors: Do source domains have any impact? ..................................................... 228 Aivars Glaznieks IV. Grammar 17. Lexical and constructional organization of argument structure: A contrastive analysis.............................................................................. 241 Johan Pedersen 18. Impaired attention to phonological input, dysfunctional lexical “frequency counters” and impaired grammatical morphology: A possible causal change......................................................................... 257 Jussi Niemi, Alexandre Nikolaev and Kenneth Hugdahl

Studies in Language and Cognition

vii

19. What can child language tell us about prepositions? ......................... 272 Aliyah Morgenstern and Martine Sekali 20. Animacy and canonical word order ................................................... 287 Espen A. Larsen and Christer Johansson V. Pragmatics and Discourse 21. Complex anaphors and their referents: The impact of ontology, meaning, and conceptual knowledge ....................................................... 299 Manfred Consten, Mareile Knees and Monika Schwarz-Friesel 22. Imperative frames and the psychology of indirect speech acts.......... 317 Per Durst-Andersen 23. Talking in and about conflicts: How children display knowledge about discourse structure ......................................................................... 335 Aivars Glaznieks and Anna Komor 24. The father and the son: Vantages in Bruce Springsteen’s lyrics........ 348 Adam Gáaz VI. Gesture and Bodily Communication 25. Gesture’s role in creating and learning language............................... 363 Susan Goldin-Meadow 26. Stages and transitions in children’s semiotic development ............... 380 Jordan Zlatev and Mats Andrén 27. Pointing gestures, vocalizations and gaze: Two case studies ............ 402 Marie Leroy, Emmanuelle Mathiot and Aliyah Morgenstern 28. The expression of negation through grammar and gesture ................ 421 Simon Harrison 29. Gesture in the brain: A Multi-tasking experiment ............................. 436 Nicla Rossini

viii

Table of Contents

30. The hand is quicker than the mind..................................................... 454 Richard Hirsch 31. Towards a conceptualized model of bodily communication: A study of hand-expressions in Estonian .............................................................. 468 Ene Vainik VII. Historical Linguistics 32. Force-dynamics in the history of Swedish modals ............................ 487 Peter Andersson 33. On the evolutionary history of ‘yes’ and ‘no’.................................... 504 Junichi Toyota 34. Semantic motivation and syntactic memorability in Old Norse kenningar................................................................................................. 518 Bryan Weston Wyly Authors’ addresses................................................................................... 536 Index........................................................................................................ 541

BRINGING LANGUAGE AND COGNITION BACK TOGETHER AGAIN JORDAN ZLATEV, MATS ANDRÉN, MARLENE JOHANSSON FALCK AND CARITA LUNDMARK

Introduction For most of the two millennia or so of reflexive thinking in the West, language and cognition – both broadly conceived – have been considered to be intimately related. At the same time, the nature of this relationship has been rife with controversy. Following the tradition of Aristotle, many philosophers, and more recently psychologists and linguists, have privileged cognition, seeking to explain language as being more or less determined by the nature of the human mind. Since the latter can be argued to owe much to the nature of the human body – under both phenomenological and empirical interpretations (cf. Gallagher 2005) – it follows that language should be considerably determined by human corporeal existence. Today, we find this approach to unravelling the language-cognition nexus most clearly represented within the school of Cognitive Linguistics (cf. Geeraerts and Cuyckens 2007a). However, also more or less from the start, there has been a realization that language itself plays a substantial role in shaping human thinking. Over the past century this realization has become increasingly prominent. To this effect, a number of factors have contributed: the systematic study of radically different languages in linguistics (e.g. Sapir 1956), the “linguistic turn” in philosophy in the mid 20 century (cf. Rorty 1992), and the socio-cultural approach in psychology (e.g. Vygotsky 1978). After a period of disrepute, the work of Whorf (1956) is once again influential, and various interpretations of “the thesis of linguistic relativity” are constantly being debated (cf. Blomberg and Zlatev, this volume). Even more importantly, they are submitted to empirical scrutiny much more thoroughly than before.

2

Chapter One: Bringing Language and Cognition Back Together Again

The only clear consensus in these sometimes heated debates is that language and cognition are tightly connected. It is not a little ironic, therefore, that when the study of the mind first became “truly scientific” in the 1960s with the rise of cognitive science, the “essence” of language – grammar – and the rest of human cognition were considered separate. Language was declared to be a “mental organ”, but one governed by “rules and representations” distinct from structures and processes of consciousnesses, memory and communicative needs (e.g. Chomsky 1986). Psycholinguistics studying the mechanisms of the so-called “language faculty” and computational linguistics, attempting to formalize and implement aspects of it in machines quite literally, were the two most “cognitive” approaches to language for a few decades. But around the time when the Berlin Wall came down, so did the monopoly of those professing that language needs to be studied separately from cognition. Consequently, new constellations of researchers across traditional departmental boundaries have been formed, and new institutions have arisen to support the study of language and cognition. The Swedish Association for Language and Cognition (SALC) is one such institution. It was formed on June 16, 2006 at an international conference at Umeå University in Sweden involving mostly linguists, but also psychologists, semioticians and cognitive scientists working on a variety of topics and utilizing different methodologies. In brief, the aim of SALC is to promote high quality international research in which language is not treated in isolation (e.g. as a “module”), but as based on structures and processes of cognition, while at the same time affecting such structures and processes. Between November 29 and December 1 2007, The First International Conference of SALC was held at Lund University, with 133 participants from 18 countries, offering 93 scientific presentations. All but two chapters in the present volume, those by Kravchenko and Gáaz, originate from presentations at this conference. The chapters were subsequently developed and chosen for publication on the basis of editorial and peer reviews and several bouts of revision from the authors, for whose cooperation we are most grateful. In the variety of their topics, the authors’ academic backgrounds and methods employed, the chapters of this book give a fair representation of both the spirit of SALC and the field of language and cognition as a whole.

Zlatev Andrén, Johansson Falck and Lundmark

3

The organization of the book The book is organized in 7 parts, corresponding to some of the major fields in language research today: (I) linguistic meta-theory and general issues, (II) lexical meaning, (III) metaphor, (IV) grammar, (V) pragmatics, (VI) gesture and bodily communication, and (VII) historical linguistics. At the same time, the non-modular approach to language adopted by all authors is reflected by the fact that there are no strict boundaries between the parts in terms of the topics addressed in the various chapters. This is also obvious from the brief summaries we provide in this section.

I. Linguistic meta-theory and general issues This part deals with general, overarching issues and begins with two important chapters by Itkonen and Kravchenko, which constitute an implicit debate on the nature of language, and correspondingly on the appropriate methodologies for studying it. On the one hand, Itkonen argues that (typological) linguistics, “is based on notions such as analogy, empathy, agent’s knowledge, and normativity, making it a fundamentally human science, rather than a natural science such as biology” (p. 19). On the other hand, Kravchenko describes such ideas in linguistics as falling prey to Cartesian dualism, and thus regards them as “myths”. Instead, he seeks an ontological and explanatory foundation for language precisely in biology, and more specifically in the theory of autopoiesis as developed by Umberto Maturana and colleagues. These two positions appear at first glance completely opposed, but there are also points of agreement, particularly on rejecting the generativist-modular tradition. The two chapters may be considered as the different poles of a continuum of possible meta-theoretical stances, with most of the contributions to the book falling somewhere in between. Blomberg and Zlatev first take a bird’s eye view on the field of language and cognition, and present a classification of theories proposing different forms of linguistic influence on cognition. They discuss how several of these apply to the perceptual/conceptual domain of motion, and review experiments of others and of their own which may help evaluate these different theories. Lind et al. focus on a notion that is often discussed and sometimes presumed, but rarely studied empirically, namely pre-verbal intentions. In an ingenious experiment, they let subjects hear their voices (through earphones) say “something else” than what they actually said by playing back what was said on a previous trial. The results paint a complex

4

Chapter One: Bringing Language and Cognition Back Together Again

picture, somewhat intermediary between “pro-intention” theories such as that of Levelt (1989) and “anti-intention” theories such as that suggested by Dennett (1991). In the final chapter of this part, Persson critically reviews studies concerning something that may seem peripheral: the ability of apes to comprehend pictures as true signs, i.e. as standing for whatever it is that they represent. The relevance of this question becomes obvious when one considers how difficult it is for both preverbal children and non-language trained children to demonstrate such competence. As Persson points out, it is possible that the sign function (Piaget 1945; Sonesson 2007) needs to be mastered through a conventional semiotic system such as language, before the special type iconic signs that we know as representational pictures can be understood as such, i.e. as picture signs. But only future studies can show whether this conjecture is true.

II. Lexical meaning This part includes chapters dealing with what has rightly been called the foundation of language: words (cf. Burling 2005). For without an “inventory” of shared lexical signs, none of the more complex forms of language – grammar, metaphors, discourse, indirect speech acts – represented in the following parts of this book, would be possible. The chapter of Svanlund deals with a classical issue in nominal semantics – individuation and (object) permanence – from a fresh perspective, focusing on a particular type of “collective human institutions”: sports teams. Based on Swedish data, the author shows how these are construed as being both relatively permanent and flexible entities. Tribushinina compares dimensional adjectives (such as ‘big’) and colour adjectives in Russian, and on the basis of corpus and survey data, she demonstrates that it is not the case, as previously claimed, that the former do not display prototype effects. Furthermore, the author distinguishes between two types of prototypicality – as head-nouns and as best exemplars – and argues that different kinds of words may display one type more than the other. Papp looks further at colour terms, focusing on Hungarian and the 11-12 “basic colour terms” previously proposed for the language. Using the criteria originally suggested by Berlin and Kay (1969), and a quantitative analysis of corpus data, the author shows that there are only 10 such terms, excluding two earlier candidates, and instead adding a rather unusual member to the list: vörös ‘dark red’, “problematic for a strict, universalist interpretation of the Berlin and Kay hierarchy” (p. 139).

Zlatev Andrén, Johansson Falck and Lundmark

5

The chapter by Martinek continues the investigation of colour semantics, moving to Ukrainian, and focusing on a single colour category, GREY, but using different linguistic forms (adjectives, nouns and verbs) conceptualizing it as an attribute, abstract entity or process. Methodologically, the author employs an association experiment and discusses how it can complement the more traditional methods of introspection and intuition to explain how lexical meaning is motivated by everyday experience. Kirt and Vainik provide a case-study of emotion-concepts in Estonian. The authors assume three levels of representation – symbolic, conceptual and sub-conceptual – inspired by Gärdenfors’ (2000) theory of Conceptual Spaces. Two different experiments provide information concerning the first and the third of these levels, respectively. The authors then plot the two sets of data onto Self-Organizing Maps (SOMs), which turn out to be non-isomorphic. The authors probe deeper and uncover a certain common structure, which they hypothesize to reflect the intermediate “conceptual” level. In the process they illustrate how semantic analysis, experimentation and computational modelling can be combined. Finally, in one of the most ambitious chapters of the book, Fortescue presents the outlines of a new “templatic” format for modelling the mental lexicon, intended to serve as an interface of sorts between semantic analysis and neural representation. Inspired by a neuroscientific model proposed by Burnod (1990), the author demonstrates how nouns, verbs, adjectives and composite expressions can be represented as templates “each reflecting several interconnected cortical columns, with the component elements widely distributed but located in specifiable cortical areas” (p. 187). The representational format takes into account recent neuroscientific findings based on dissociations in aphasia and neuroimaging and makes a number of testable predictions.

III. Metaphor This part focuses on metaphor, which is treated by the authors as both a linguistic and a cognitive phenomenon. In the influential conceptual metaphor theory (CMT) (Lakoff and Johnson 1980, 1999; Grady 1997) metaphors are primarily regarded as conceptual cross-domain mappings, reflected only secondarily by linguistic expressions such as “our spirits are falling along with the stock market”. While influenced by CMT, however, all four chapters of this part show a growing need to either develop this

6

Chapter One: Bringing Language and Cognition Back Together Again

framework, or to go beyond it, focusing more on factors such as language (use) and culture. Vandi and Fusaroli discuss advantages and problems of CMT and blending theory (Fauconnier and Turner 2002) as theories of metaphor and argue for “a dynamic perspective on embodied cognition, where sedimented conceptual structures are locally deployed as resources and not as determining types” (p. 193). They provide a case-study of the metaphorical expression “that man is a caiman” (referring to the Italian prime-minister Berlusconi), showing how its force derives from a combination of cultural, discursive, lexical and co-textual factors. The chapter by Vogel discusses metaphorical expressions for dying in five European languages, Swedish, Norwegian, Danish, German and French, and finds considerable similarities, such as using highly agentive verbs for the act of dying, locations for the “afterlife” and various euphemisms. While not stated explicitly by the author, the findings seem to point more to commonalities in European cultural models of death, than (only) to universal conceptual metaphors deriving from bodily experience. Bergström analyzes the meaning of the Swedish adjective varm (‘warm’) along a model for temperature semantics based on 5 dimensions, and distinguishes between 24 “meaning variants” (or senses), 14 of which are literal and 10 metaphorical. When eliciting spontaneous associations to varm from a group of university students, the author found associations corresponding to 9 of the 14 literal senses, but only to 3 of the metaphorical ones. Assuming that these associations reflect “what is essential to people, in this case what is central about warmth, in both language and mind” (p. 226), the author suggests that most of the metaphorical senses do not reflect immediate experience, or “primary metaphors” in the sense of Grady (1997), but linguistic conventions typical for Swedish written discourse. Finally, Glaznieks presents a study in which German-speaking children performed comprehension tests on metaphorical expressions for fear and anger. The results showed that “the understanding of metaphorical expressions develops in a domain-specific way only with respect to target domains” (p. 229). On the other hand, knowledge of the source domains did not seem to have any impact on the understanding of the metaphorical expressions, thereby challenging a common assumption in both developmental psychology and conceptual metaphor theory.

IV. Grammar This part includes four chapters that deal, in part or in whole, with the organization of sentences or morphologically complex words, i.e.

Zlatev Andrén, Johansson Falck and Lundmark

7

grammar. Though using widely different approaches and methods, all the chapters contribute further evidence that grammar is not a “module” governed by autonomous rules, but is affected by factors such as lexical meaning and frequency. Pedersen presents a contrastive study of English and Spanish basic clausal constructions, and presents evidence for an analysis that is at variance with both universalist “parametric” models such as those of Chomsky, and with the influential typology of Talmy (2000) due to its too narrow focus on lexicalization patterns. Using the resources of Construction Grammar (CXG), it is suggested that English “tends to organize principal clausal information in schematic argument structure constructions, leaving secondary information for lexical (verbal) specification” (p. 241) while Spanish does more or less the converse. The chapter by Niemi, Nikolaev and Hugdahl investigates the mastery of Finnish morphology by speakers with Familial Language Impairment (FLI). The authors show that FLI speakers are much less sensitive to lexical frequencies than controls, and suggest a “possible causal chain” in the impairment: from attention to phonological input, through lexical frequency to morphology. While not entering the debate on the nature of such language impairment, these findings show how a fairly general cognitive impairment could have “specific” effects on linguistic processing. Morgenstern and Sekali focus on a particular form-class, prepositions, and ask a question that has been controversial for some time: since prepositions (e.g. ‘to’) often have both more spatial (‘He went to the bank’) and more grammatical meanings (‘He gave it to Mary’), are the first more primary in child language acquisition? Based on longitudinal data from English and French, the authors support the view that this is not necessarily the case, but rather that grammatical meanings are no less basic, since they “can express and organize social interaction, and are acquired by children thanks to the mediation of adults” (p. 272). Finally, Larsen and Johansson describe an experiment involving anaphora resolution in a set of three Norwegian sentences, with the first sentence having either an SVO or an OVS word order, and the subject noun denoting either a human being, an animal or an inanimate object. The results show a significant effect of “animacy (and/or expectancies from new or given information) on human sentence processing” (p. 287), and no evidence for a purely syntactically driven first stage of parsing.

8

Chapter One: Bringing Language and Cognition Back Together Again

V. Pragmatics and discourse This part of the book extends naturally the scope of investigation by addressing linguistic phenomena that necessarily go beyond the sentence boundary, involving context and co-text, falling traditionally within the fields of discourse analysis and pragmatics. Consten, Knees and Schwarz-Friesel investigate complex anaphors, which refer to situations and similar complex entities, and “establish them as condensed and stable discourse entities” (p. 299), on the basis of a German newspaper corpus. Using an original model of discourse representation, text-world model theory, the authors show how the interpretation of the referents of complex anaphors are co-determined by a combination of bottom-up factors relating to lexical meaning and co-text, and top-down processes such as world knowledge and discourse coherence. Durst-Andersen presents a taxonomy of 8 imperative frames, each consisting of four sub-structures. These frames are claimed to be universal, but realized through various means in different languages. For example, while English predominantly uses indirect speech acts, Russian relies on different aspectual forms and Danish on modal particles, but all (these) languages can ultimately achieve “the same goal by verbalizing one of the four possible places in the structure” (p. 317) of an imperative frame. After providing an original explanation for the more common use of indirect imperatives in English than in Russian, and the role of the modal particles in Danish, which are highlighted in the analysis, the author concludes that “pragmatics seems to be more structure-driven than rule- or principle-driven” (p. 317). In their contribution, Glaznieks and Komor investigate the speech of 5 and 8-year-old German-speaking children for evidence of awareness of discourse structures, in particular with respect to conflicts. Two studies are reported, the first based on interviews about conflict situations, and the second on authentic antagonistic interactions between peers. The results show evidence for mostly implicit knowledge about conflicts in the younger group, and much more explicit knowledge in the older group. Finally, Gáaz presents a case study in Cognitive Poetics, as “an approach to literary text whose ambition is to bridge the gap between the study of language and the study of literature” (p. 348). The author uses a particular theoretical model, Vantage Theory to analyze the shifting perspectives in a poetic text by Bruce Springsteen. The study illustrates both features of the theoretical model, and clarifies some subtle intuitive aspects of the lyrics.

Zlatev Andrén, Johansson Falck and Lundmark

9

VI. Gesture and bodily communication This part contains chapters dealing with the relationship between language and gesture, reflecting a growing interest in the topic. While there is a consensus on a close relationship between the two, there is an ongoing debate on the more exact nature of this relationship: do they constitute a “unified system” (McNeill 2005; de Ruiter 2007), or two closely integrated but still distinct semiotic capacities, supported by (partially) distinct cognitive mechanisms (Kita and Özyürek 2003)? While there is no consensus on this question, the chapters provide important clarifications through different perspectives and methods. Goldin-Meadow presents an overview of her group’s well-known studies on deaf children who spontaneously develop a language-like system of communication: homesign, as well as ongoing research attempting to determine the roots of the “resilient properties of language”. Interestingly, when hearing adults need to communicate without speaking, some (but not all) of these properties emerge, including an ordering of the gestures that cuts across cultures and does not resemble the basic orders of the speakers’ languages. Finally, the author suggests reasons why gesture may also be a “stepping stone” to the development of language even among hearing children. Zlatev and Andrén deal with bodily communication and its relation to language from a developmental perspective, and employ the notion of developmental stage, proposing a “rehabilitation” of this notion. Using a stage-based model, the mimesis hierarchy (Zlatev 2008), the authors provide a developmental analysis of “acts of bodily communication” in two groups of three children, one in Sweden and the other in Thailand. Despite some cultural differences, the chapter reports major similarities and general developmental patterns, the most significant of which is a rather sharp transition around 20 months, which can be interpreted as a “convention insight”, a key step in language acquisition. Leroy, Mathiot and Morgenstern analyse “pointing events” in a longitudinal corpus of multimodal data from two French children. The authors argue persuasively that since pointing constitutes a socialcognitive “tool” for the child, it should not be analyzed in isolation, but together with “features” such as gaze, the prosody of the vocalization, and the position of the child, addressee and the object pointed to. The interrelation of these features is found to be complex, but prosody appears more significant than gaze for determining the meaning of the pointing gesture, which contradicts some common assumptions in the literature. In his contribution, Harrison investigates the interaction between grammar (or rather, speech) and gesture in the expression of negation.

10

Chapter One: Bringing Language and Cognition Back Together Again

Through the careful study of elicited “negative speech acts”, gestures such as head-shake and two different hand gestures were found to “intensify, render more precise, and more explicit the negation expressed by grammar” (p. 433) and even to complement the meaning of the oral expression in subtle ways with respect to referential content and discourse functions. The full meaning of the acts thus derived from a combination of semiotic resources. The chapter by Rossini addresses the question of the dependency between language, gesture and motor action. A multi-tasking experiment is described in which participants were asked to read Italian prose and poetry texts, while repeating a rhythmic beat which was designed to be asynchronous with the reading rhythm. The results showed a general failure in multitasking, and are interpreted by the authors as evidence for a “strong connection between manual action and language” (p. 450), and against (neural) modularity. Hirsch uses the interaction between gesture and speech in order to argue for a rather radical position, consistent with the perspective of Kravchenko in Part I: that rather than originating from individual minds, meaning (and “mind”) is “embodied and distributed and emerges in the field created by interacting brains-and-bodies” (p. 454). This thesis is supported by detailed analysis of several specific examples of communicative interaction, demonstrating several different collaborative processes of sense-making. Finally, Vainik deals with the theme of this section from the perspective of lexical semantics, and analyzes systematically the semantics of “hand-expressions” in Estonian. The study reveals several features of a particular folk model concerning “bodily communication” in Estonian culture, including different types on the basis of parameters such as distance, co-operative behaviour, and the status of the interlocutors. In all cases, the meaning of the expressions analyzed is argued to be metonymic with respect to the situational frame to which they apply.

VII. Historical linguistics This part concludes the book by going back in time, using methods and addressing questions of historical linguistics, while still preserving a cognitive perspective. Andersson contributes to a central topic within historical linguistics: the emergence of epistemic modal meanings in expressions originally expressing “root modality”, e.g. capacity and obligation. This is done by investigating the development of the Swedish modal må ‘may’, and using

Zlatev Andrén, Johansson Falck and Lundmark

11

central cognitive-linguistic concepts such as force-dynamics and imageschemas. The author proposes a partial re-interpretation of these concepts, and a general image-schema consisting of forces and barriers which “underlies and motivates the root as well as the non-root meanings” (p. 487). Using historical evidence, Andersson suggests possible “bridging contexts” for the meaning transfer from the physical domain to the epistemic mental domain, and thus also connecting to the topic of metaphor of Part III. Toyota traces a possible path for the emergence of ‘yes-no’ words in the languages of the world, and proposes the hypothesis that ‘no’-words initially emerged from a negation marker, and then an affirmative counterpart was required in order to fulfill the requirement of binary opposition. In support of this, it is pointed out that many (human) cognitive and linguistic structures are organized in binary pairs, and that words for ‘yes’ seem to be a relatively late historical development. This part, and the book as a whole, concludes with the contribution by Wyly, which in a highly original manner combines data – a corpus of Old Norse scaldic verse characterized by extreme degrees of ambiguity – and methods from historical linguistics with recent results from psycholinguistics, showing different capacities and interpretative strategies in coping with ambiguity. To make these widely different traditions cohere, the author provides both a description of Old Norse verse and the social context in which it was used, and a theoretical model of the interrelation between linguistic competence and performance, consisting of four interrelated dimensions, inspired by Tomasello’s (2003) usagebased model language acquisition. Finally, a detailed case-study of the initial stanza of the poem Berodrápa is presented within the proposed historical-psycholinguistic framework. Wyly’s chapter illustrates nicely the kind of synergism that becomes possible once language and cognition are brought together.

Methodologies, models and languages The overview provided in the preceding section shows the richness in methodology, theoretical models and empirical material employed by the authors of the book, which as stated earlier is indicative of the field of language and cognition. Table 1 attempts to provide a (partial) specification of the methods employed in the studies described in the chapters of the book. Note that several chapters figure in more than one row of the table.

12

Chapter One: Bringing Language and Cognition Back Together Again

Table 1. Methods employed by the authors of the different chapters, with several chapters using multiple methods, by order of appearance Method employed Meta-analysis Intuition-based semantic/grammatical analysis Psychological experiments Synchronic corpus analysis Computational modelling Neuroscientific evidence Developmental studies Discourse analysis Historical analysis Interaction analysis

Chapter 2, 3, 4 4, 7, 12, 13, 14, 15, 17, 21, 22, 31, 32 4, 5, 6, 9, 11, 15, 16, 18, 20, 25, 29 8, 10, 31 11 12, 29 16, 19, 23, 25, 26, 27 23, 24 32, 33, 34 26, 27, 28, 30

While the presentation in Table 1 should be considered a rough approximation, it indicates that analyses based on (native) speaker’s intuitions and psychological experimentation appear to dominate. While some present this as a “subjective vs. objective” dilemma for e.g. Cognitive Linguistics (Geeraerts and Cuyckens 2007b), i.e. reliance on subjective intuition or on objective data, we would claim that this is a false dilemma, and that the multilevel phenomenon of language (and cognition) ultimately requires both these kinds of methods, as well as others (Zlatev 2007, in press; Itkonen 2008). Meanings and other linguistic norms are not natural phenomena like physical force and mass, and intuition will always remain a basic source of knowledge, to greater or lesser extent, depending on the degree to which what is investigated is of conceptual or causal character. Concerning models and frameworks, it is stated on the homesite of SALC that “[t]he association is intended to be a forum for cooperation and exchange of ideas between disciplines, fields of study and theoretical frameworks.” (www.salc-sssk.org) Table 2 lists again a partial list of the theoretical models employed in some of chapters. Finally, the book presents original data from a wealth of (mostly European) languages, again summarized in table form (Table 3).

Zlatev Andrén, Johansson Falck and Lundmark

13

Table 2. A list of theoretical models used as explanatory frameworks in the book, by order of appearance Theoretical model Prototype theory Basic-level theory Conceptual Spaces Template format Conceptual metaphor theory Construction Grammar Text-world model theory Imperative frame theory Vantage theory Mimesis hierarchy model Distributed language Cognitive Semantics Competence-performance dimensions

Chapter 8 10 11 12 13, 16, 31 17 21 22 24 26 30 32 34

Table 3. Primary data from different languages quoted in the chapters, by order of appearance Language Swedish Russian Ukrainian Hungarian Estonian Norwegian Danish German French Spanish English Finnish Thai Italian Celtic etc. Old Norse

Chapter 7, 14, 15, 26, 30, 32 8 9 10 11, 31 14, 20, 21, 23 14, 22 14, 16 14, 19, 27, 28 17 17, 24 18 26 29 33 34

Table 3 represents a good variety of languages, though for natural reasons, biased towards (Northern) Europe. Still, given that nearly all the chapters derive from presentations at The First International Swedish Association for Language and Cognition, it is noteworthy that an all too great focus on Swedish and the other Scandinavian language is absent.

14

Chapter One: Bringing Language and Cognition Back Together Again

Conclusions The purpose of this introductory chapter was to provide a context for and an overview of the contents of the present volume. Its title is somewhat rhetorical – neither this book, nor SALC or similar institutions will by themselves succeed in “bringing language and cognition back together again”. However, they reflect a general intellectual tendency that is nowadays pervasive, since the road previously taken leads (in ours and many others’ perception) to a dead end. If there is one thing we wish to emphasize, it is our belief that what is needed now is the kind of openminded methodological and theoretical pluralism reflected in this volume. Furthermore, we hold that this is essential, if linguistics is to emerge as a phoenix from the ashes of dogma and institutional stagnation that have plagued it for some decades. We leave it to the reader to determine if this approach is worth pursuing.

References Berlin, B. and Kay, P. 1969. Basic color terms: Their universality and evolution. Berkeley: University of California Press. Burling, R. 2005. The talking ape: How language evolved. Oxford: OUP. Burnod, Y. 1990. An adaptive neural network: The cerebral cortex. London: Prentice Hall, Paris: Masson. Chomsky, N. 1986. Knowledge of language. New York: Praeger. Dennett, D. 1991. Consciousness explained. Boston, Mass: Little Brown. Fauconnier, G. and Turner, M. 2002. The way we think: Conceptual blending and the mind’s hidden complexities. New York: Basic Books. Gallagher, S. 2005. How the body shapes the mind. Oxford: OUP. Geeraert, D. and Cuyckens, H. 2007a. The Oxford Handbook of Cognitive Linguistics. Oxford: OUP. Geeraerts, D. and Cuyckens, H. 2007b. Introducing cognitive linguistics. In D. Geeraerts, and H. Cuyckens (Eds.), The Oxford Handbook of Cognitive Linguistics, 3-21. Oxford: OUP. Grady, J. 1997. Foundations of meaning: Primary metaphors and primary scenes. Ph.D. dissertation, University of California, Berkeley. Gärdenfors, P. 2000. Conceptual spaces: The geometry of thought. London: MIT Press. Itkonen, E. 2008. The central role of normativity for language and linguistics. In J. Zlatev, T. Racine, E. Sinha, and E. Itkonen (Eds.), The shared mind: Perspectives on intersubjectivity, 279-306. Amsterdam/Philadelphia: Benjamins.

Zlatev Andrén, Johansson Falck and Lundmark

15

Kita, S. and Özyürek, A. 2003. What does cross-linguistic variation in semantic coordination of speech and gesture reveal? Evidence for an interface representation of spatial thinking and speaking. Journal of Memory and Language 48: 16-32. Lakoff, G. and Johnson, M. 1980. Metaphors we live by. Chicago: UCP. Lakoff, G. and Johnson, M. 1999. Philosophy in the flesh: The embodied mind and its challenge to Western thought. New York: Basic Books. Levelt, W. 1989. Speaking: From intention to articulation. Cambridge, Mass: MIT Press. McNeill, D. 2005. Gesture and thought. Chicago: UCP. Piaget, J. 1945. La formation du symbole chez l'enfant. Neuchâtel: Delachaux et Niestlé. Rorty, R. 1992 [1967]. The linguistic turn. Chicago: UCP. de Ruiter, J.P. 2007. Postcards from the mind: the relationship between speech, imagistic gesture and thought. Gesture 7(1): 21-38. Sapir, E. 1956. Culture, language and society: Selected essays. D.G. Mandelbaum (Ed.). Berkeley, CA: University of California Press. Sonesson, G. 2007. From the meaning of embodiment to the embodiment of meaning: A study in phenomenological semiotics. In T. Ziemke, J. Zlatev and R. Frank (Eds.), Body, language and mind. Vol 1: Embodiment. Berlin: Mouton de Gruyter. Talmy, L. 2000. Toward a cognitive semantics, Vol 1 and Vol 2. Cambridge, Mass: MIT Press. Tomasello, M. 2003. A usage-based theory of language acquisition. Cambridge, Mass.: Harvard University Press. Vygotsky, L.S. 1978. Mind in society: The development of higer psychological processes. Cambridge, Mass.: Harvard University Press. Whorf, B.L. 1956. Language, thought and reality: Selected writings of Benjamin Lee Whorf. John B. Carroll (Ed.). Cambridge: MIT Press. Zlatev, J. 2007. Spatial semantics. In H. Cuyckens and D. Geeraerts (Eds.), The Oxford Handbook of Cognitive Linguistics, 318-350. Oxford: OCP. —. 2008. From proto-mimesis to language: Evidence from primatology and social neuroscience. Journal of Physiology – Paris 102: 137-152. —. in press. Cognitive linguistics and phenomenology. In S. Gallagher, and D. Schmicking (Eds.), Handbook of Phenomenology and Cognitive Science. Berlin: Springer.

I. LINGUISTIC META-THEORY AND GENERAL ISSUES

THE TRUE NATURE OF TYPOLOGICAL LINGUISTICS ESA ITKONEN

Abstract. In this chapter I consider the nature of different kinds of typological linguistics, and of typological linguistics as whole. Based on previous work, I argue that it is based on notions such as analogy, empathy, agent’s knowledge and normativity, making it a fundamentally human science, rather than a natural science such as biology. Only in this way can the proper role of functional explanation in typological linguistics be appropriately understood.

Types of linguistic typology Research in linguistic typology is traditionally based on the following assumption: “All languages deal with approximately the same set of universal concepts, but these are coded and combined in different ways” (Dixon 2002: 56). The semantics of each language contains a repertoire of grammatical meanings (or “functional domains”), which qua basic categories of thought overlap with the items in the “philosophical lexicon” given in Book V of Aristotle’s Metaphysics. Across languages, the repertoires of grammatical meanings are assumed to be similar to such an extent that their mutual differences may (at least initially) be disregarded. Thus, in principle, one and the same semantic content is expressed in different languages by dissimilar morpho-syntactic means: same meanings, different forms. However, this is not the rationale of typology tout court, but of morpho-syntactic typology (and of lexical typology, as this term is generally understood). Morpho-syntactic (and lexical) typology must be supplemented with semantic typology. This has been shown most dramatically by Levinson (2003), who (together with his research group) establishes the existence of three distinct spatial frames of reference, namely intrinsic, relative, and absolute. These constitute three dissimilar semantic systems without any possibility of direct translation from one into the other. From the statement “The ball is in front of the chair”, it is

20

Chapter Two: The True Nature of Typological Linguistics

impossible to recover the information that the ball is (e.g.) to north of the chair (and vice versa), which entails that one statement cannot be a genuine translation of the other (Levinson 2003: 57–59). Here the typological interest resides in the semantic content while its formal realization is irrelevant: “the frame-of-reference distinctions are as much lexical as grammatical” (p. 301). Hence, different meanings, indifferent forms. Psycholinguistic experiments have shown that distinct spatial frames of reference correlate with behavioral differences, which has been taken to support the Whorfian view that language determines thought. However, the same experimental evidence has been claimed by Palmer (2004) to admit of the more traditional interpretation that culture (as mediated by language) determines thought. Taken together, morpho-syntactic(-cum-lexical) typology and semantic typology constitute what might be tentatively called metatypology.

Morpho-syntactic typology is based on cross-linguistic analogy Analogy is defined as structural similarity (Itkonen 2005a: 1-3), and not as physical (or perceptual) similarity (or resemblance), as suggested by Chomsky (1966:4) and Evans (2004: 48). It follows that when one and the same sentence-meaning (like “I do not see it”) is expressed in different languages, whether spoken or signed, the resulting sentence-forms turn out to be analogous (cf. Itkonen 2005a: 6-7). By the same token, the possibility of translation is based on cross-linguistic analogy. It is self-evident, but too seldom explicitly pointed out, that the existence of massive cross-linguistic analogy refutes the structuralist dogma of où tout se tient, or the view that each language is a selfcontained system which exclusively defines its own categories.

The semantics vs. pragmatics divide in morpho-syntactic typology It is a well-known fact that the logical connectives and, or, and if – then are lacking in many languages (cf. Itkonen 2003a: 150-151, 2005a: 154155; Levinson 2003: 292–293). Yet the speakers of such languages behave in ways which, from the logical point of view, are similar to ours. Since this similarity cannot exist at the level of semantics, it must exist at the level of pragmatics.

Esa Itkonen

21

We reach the same conclusion, more generally, when we study how the notion of subordination is handled in different languages. As I interpret Austin’s (1980) data, Diyari has a converb system with three relative tenses (= A[nterior], S[imultaneous], P[osterior]) and switch reference (SS = same subject vs. DS = different subject), producing six converbs: AandSS, AandDS, SandSS, SandDS, PandSS, PandDS (cf. Itkonen 2005b: 28-33). These (temporally-based) distinctions do not correspond to the familiar distinctions between different types of subordinate clause: “It is important to note that the Diyari sentence is simply vague as to the semantic connection between the sentences” (Austin 1980: 150, emphasis in original). Yet it can be shown in detail that Diyari is able to convey the familiar subordinate-clause distinctions. Thus, correct understanding of sentences must be a matter of context-dependent interpretation which transforms semantic indefiniteness into pragmatic definiteness. An even more extreme example of semantic vagueness/indefiniteness is offered by Rembarrnga (cf. McKay 1988). The verb has the structure PAT-AG-V, e.g. barran-ba-V (= ‘them-they-V’), and the subordinate status of a clause is indicated by the change a o i, e.g. birrin-bi-V. According to McKay (1988), the superordinate vs. subordinate distinction equals that between Figure and Ground. Pragmatically, the subordinate clause has temporal, locational, conditional, causal, complement, and relative uses, but “in Rembarrnga the various “uses” are not differentiated [syntactico-semantically] at all from one another” (p. 8). The upshot of this section is to revitalize Givón’s (1979) distinction between “syntactic vs. pragmatic mode”.

Explanation in typological linguistics: Empathy How should one understand the concept of “explanation” in linguistics? However difficult it may feel at first, one must learn to resist the temptation to borrow this notion from elsewhere, e.g. from such disciplines as Newtonian mechanics, evolutionary biology, quantum physics or string theory. In the 70’s it was customary to apply the deductive-nomological (= D-N) model of explanation to linguistics. Itkonen (1974) contained a critique of the D-N model while Dahl (1975) presented a passionate defense of it; see also Itkonen (1980). Today it is customary to apply the Darwinist model to linguistics, in spite of its obvious inadequacy: linguistic change is produced by intelligent goaldirected processes (see immediately below), whereas biological change is produced by random mutations (cf. Itkonen 1999, 2005a: 186–189, 2005b: 279–284).

22

Chapter Two: The True Nature of Typological Linguistics

The right way to proceed is to observe what representatives of typological linguistics themselves mean by such expressions as “explaining” or “making understandable”. This, and this alone, guarantees a sufficient degree of authenticity. Over the years I have collected many examples of how these expressions are used. One of my standard examples is how Mithun (1988) explains why there are, cross-linguistically, so few examples of the construction N-and-N. Explaining this fact requires assuming that speakers have performed such mental processes as introducing (names of) new referents into discourse and re-identifying them with an optimal combination of efficiency and economy. In order to grasp such processes and such efficiency vs. economy considerations, the only recourse that the linguist has is to rely on empathy, i.e. on his/her ability to re-enact those processes and considerations (to use Collingwood’s 1946 term). The speakers were confronted with a problem of choosing (what they considered as) the best means of achieving their goal (e.g. re-identifying the previously introduced referents); and when the linguist recapitulates what the speakers did, (s)he makes use of rational explanation, exactly in the sense of Itkonen (1983). The same remarks apply to explaining facts of grammaticalization, e.g. to Paul’s (1975 [1880]) explanation, in terms of reanalysis and extension, of how the German demonstrative pronoun das became grammaticalized as the conjunction dass. The point is that we reject any proposed reanalysis and/or extension which is such that we cannot imagine performing them ourselves. This is confirmed by the fact, known to every student of grammaticalization, that when the change A > B seems incomprehensible, the first move is to try to postulate some intermediate stages C and D, which are such that we can imagine having performed ourselves each of the more specific changes A > C, C > D, and D > B. At a rather high level of abstraction, the preceding account is supported by Givón (2005): “A bio-organism’s first imperative is to understand. That is, to explain — by abductive reasoning — why entities behave the way they do. Agency is but an adaptive hypothesis” (p. 211). “The scientist merely recapitulates the bio-organism ...” (p. 204). Additional, and more direct, support is given by Croft’s (2003) view of typological explanations: Because implicational universals are clearly not enough (cf. Itkonen 1998), “deeper explanations” are needed; and on closer inspection, these turn out to rest on such empathy-based notions as “cultural expectedness or salience” (Croft 2003: 115–116), “high salience or topicality” (pp. 178179), and “cognitive salience” (pp. 181-183); for discussion, see Itkonen (2004).

Esa Itkonen

23

On reflection, it is rather self-evident that empathy is exactly what is required in the study of “exotic” languages and cultures. However, as will be seen empathy “starts at home”.

Empathy = Vicarious introspection It is necessary to distinguish between the following (potentially) conscious acts of knowledge, based on the differences between their respective objects: observation about spatio-temporal entities (or ‘world-1’), introspection about one’s own mental states (or ‘world-2’), and intuition about socially or intersubjectively valid norms (or ‘world-3’) (cf. Itkonen 1981: 127–128, 1983: 7–10; Itkonen 2008; Katz 1981: 194–196). All three coincide in acts of communication. When someone says to me Christmas is coming, I hear, i.e. observe, (an utterance of) this sentence, I intuitively know that it is correct English, and I introspect the ‘feeling-tone’ (Sapir 1921: 40) that I personally attach to Christmas. The relation between observation and intuition is the central topic of e.g. Itkonen (1978) and (1983), and it seems reasonably well understood today. The relation between introspection is less well understood, as shown by the fact that even insightful scholars like Chafe (1994, 2002) and Talmy (2000, 2007) do not distinguish between the two at all (cf. Itkonen 2008). The difference between intuition and introspection was already defined above, but it may be further highlighted as follows. Variation between objects of intuition is a matter of right or wrong, as shown by the difference between the correct sentence That mountain range goes from Canada to Mexico and the incorrect sentence *That mountain range goes from Canada in Mexico (or *That mountain range goes from Canada to boy, etc, etc). By contrast, variation between objects of introspection is not a matter of right and wrong. At the level of mental imagery, two opposite fictive motions are assumed to be connected with the (correct) sentences That mountain range goes from Canada to Mexico and That mountain range goes from Mexico to Canada (Talmy 2000: 104). But if, on either hearing or uttering one or both of these sentences, I fail to perform the typical fictive motion, I have not made a mistake (of English) . Let us stipulate that the generic term “feeling” applies to (the awareness of) emotions, motives, and reasons. Recently, I have proposed the following “logical reconstruction” of the relation between introspection and empathy (Itkonen 2008: 26):

24

Chapter Two: The True Nature of Typological Linguistics

i) I now feel (or am capable of feeling) X Î ii) I would have felt X if I had been in situation Y Î iii) I would have felt X if I had been person Z in situation Y (= Weber-type Verstehen, Collingwood-type re-enactment).

Intuition = Conventionalized empathy Similarly, I have also proposed the following logical reconstruction of the relation between introspection and (linguistic) intuition (Itkonen 2008: 26): i) I introspectively know that right now I mean ‘Y’ by X Î ii) I empathically know that also others can mean or have meant ‘Y’ by X Î iii) I intuitively know that X means ‘Y’ (i.e. that one ought to mean ‘Y’ by X). What we have here is a (schematic) account of the emergence of (linguistic) normativity. It follows that before the meaning ‘Y’ of X has become fully conventional, there is (and must be) a period when it is unclear whether ‘Y’ is known (still) introspectively or (already) intuitively. This just expresses the general nature of linguistic change, but in terms of acts of knowledge, rather than — what is much more common — in terms of objects of knowledge. The “ascent” introspection Î empathy Î intuition can be interpreted in two different ways. First, it may a logical reconstruction of the relation between introspection and intuition in the case of the well-established meaning ‘Y-1’ of X. Second, it may be an actual description of the emergence of some new meaning ‘Y-2’ of X, which is some sort of extension of ‘Y-1’. It is important to understand what is meant here by “logical reconstruction”. Human thought is ultimately intersubjective or social in character. Hence, when I speak of a “transition” from introspection to intuition (via empathy), I do not mean to say that there is first a period when all thinking pertains to subjective states of consciousness, and then a period when there emerges a type of thinking that pertains to intersubjective norms and concepts. Rather, given that human beings are born into and raised in a social world, every act of introspection already takes place within a social context, i.e. it is both supported by and formulated in concepts that are socially given and publicly available (which, to be sure, does not preclude the possibility of gradually expanding the realm of such concepts).

Esa Itkonen

25

Similarly, norms (of language, for instance) must be of social nature: “It is only in a language that I can mean something by something” (Wittgenstein 1958: 18). Thus, I strongly oppose any attempt to reduce conventional meanings to (non-normative) “speaker intentions” (cf. Itkonen 1983: 167–168), just as I oppose any attempt to reduce norms of language to (non-normative) “hearer beliefs/expectations” (cf. Itkonen 1978: 182–186). As conclusively demonstrated by Frege and Husserl, “ought” just cannot be reduced to “is” (cf. Itkonen 1991: 283-284).

“Agent’s knowledge” subsumes introspection and intuition There is a long tradition in Western philosophy which maintains that genuine knowledge is agent’s knowledge or our (“internal”) knowledge of our own actions, and not observer’s knowledge or our (“external”) knowledge of spatio-temporal events beyond our control. This tradition has been represented by Plato, Aristotle, Hobbes, Vico, Kant, Dilthey, Weber, Collingwood, and Schutz (cf. Itkonen 1978: 193–198, 2003b: Chap. 11). As subtypes of agent’s knowledge, intuition and introspection differ insofar as the former does, while the later does not, contain a normative element. Intuition is one’s knowledge of one’s own rulefollowing actions and their results. Terms like ‘intuition’ and ‘introspection’ tend to have rather static connotations. Therefore it is good to emphasize that there is also intuition and introspection in, and of, action.

Implications for the philosophy of science The distinction between observer’s knowledge and agent’s knowledge underlies a distinction between two dissimilar types of scientific discipline, namely those characterized, in von Wright’s (1971) terms, by “explanation” and “understanding”, respectively. Among the disciplines of the latter type, moreover, additional distinctions have to be made, depending on whether, and to what extent, the data is of normative character. Any such thesis is standardly criticized as trying to reintroduce the outdated and discredited dichotomy between physical vs. human sciences, but this is just a knee-jerk reaction. At each moment one has to consider the level of abstraction at which the argument is being conducted. The more the level of abstraction rises, the more each scientific (or academic) discipline starts to resemble each other.

26

Chapter Two: The True Nature of Typological Linguistics

Choosing the right analogy for typological linguistics: Not vision, but architecture How does the subject matter of typological linguistics exist, i.e. what is its ontology? Certainly every (spoken or signed) language exists in the same way as every other, but the cross-linguistic perspective is apt to highlight some aspects that might otherwise remain in the shadow. The progress of Cognitive Linguistics (= CL) has been hampered by the fact that CL is a direct descendant of Generative Linguistics (= GL). Similarly, the self-understanding of typological linguistics (= TL) has been hampered by the fact that TL has taken the “philosophical” (ultimately GL-type) commitments of CL too seriously. According to GL (and CL), “language” is a mental and cognitive capacity, in exactly (or at least as much) the same way as vision. There is, however, this obvious dis-analogy that while vision is just a matter of perception, language is a matter of perception and production. Language qua (inter)action produces forms that clearly serve an instrumental function. Now, if language is taken to be comparable to vision, it follows (or seems to follow) that language too is a cognitive (subjective) capacity, and not easily amenable to anything like functional explanation. Instead of comparing language to vision, the alternative that I propose is to consider the world’s languages on a par with architectural styles, embodied in concrete buildings, around the world: Sooner or later you will move from describing the buildings to explaining why they are such as they are in fact. Your explanations will be couched in functional terms, and the functions which you will refer to are bound to be socially shared, in spite of the fact that some buildings are more clearly products of individual creation than others. (Itkonen 2005a: 199–200)

Individual innovation (in language and architecture) results in genuine change only if it is followed by social acceptance (ibidem: 110-111). What, then, is the ontology (not of “language” per se but) of languages? Just like architectural styles, they exist as social institutions (cf. Trubetzkoy 1958 [1939]: 15, 41). Qua functionally motivated entities, sentences and buildings are instruments. Because there are no instrumentlike counterparts in vision, vision is a false analogy for linguistics, to begin with. Thus, the answer to the question: “To what extent is the term “functional-typological linguistics justified?” is the following: Only to the extent that the thesis advanced in this section is accepted, which eo ipso entails recognizing the true nature of (functional-)typological linguistics.

Esa Itkonen

27

Concluding remark As was already mentioned, a given language inhabits all three “worlds”, i.e. social, psychological, and physical. In this chapter, I have been concerned with the social dimension (= “world-3”) and with that part of the psychological dimension (= “world-2”), which is accessible to consciousness (more precisely, introspection). My concern has not been with the unconscious part of the “world-2”, which is traditionally taken to be the realm of psycholinguistics. Trubetzkoy (1958 [1939]: 15) was intelligent enough to realize that, as a social institution, a given language does not belong to the “world of empirical phenomena”, which means that, unlike acts of speech, it cannot be perceived (i.e. heard or seen). A language is, rather, accessible to intuition. Now, in typological research one generally relies on the intuition of others, in a twofold sense. First, one makes use of grammars written by others, as I, for instance, have done in Itkonen (2001) and (2005b). Second, the data of these grammars are to a smaller or greater extent the result of elicitation (cf. Haiman 1980: xi).

References Austin, P. 1980. A grammar of Diyari, South Australia. Cambridge: Cambridge University Press. Chafe, W. 1994. Discourse, consciousness, and time. Chicago: University of Chicago Press. —. 2002. Putting grammaticalization in its place. In I. Wischer and G. Diewald (Eds.) New reflections on grammaticalization, 395-412. Amsterdam: Benjamins. Chomsky, N. 1966. Topics in the theory of generative grammar. The Hague: Mouton. Collingwood, R.G. 1946. The idea of history. Oxford: Clarendon Press. Croft, W. 2003. Typology and universals, 2nd ed. Cambridge: Cambridge University Press. Dahl, Ö. 1975. Is linguistics empirical? A critique of Esa Itkonen’s Linguistics and Metascience. Gothenburg Papers in Theoretical Linguistics 29. [Reprinted in Perry (1980)] Dixon, R.M.W. 2002. Australian languages. Cambridge: Cambridge University Press. Evans, V. 2004. The structure of time. Amsterdam: Benjamins. Givón, T. 1979. On understanding grammar. New York: Academic Press. —. 2005. Context as other minds. Amsterdam: Benjamins.

28

Chapter Two: The True Nature of Typological Linguistics

Haiman, J. 1980. Hua: A Papuan language of the Eastern Highlands of New Guinea. Amsterdam: Benjamins. Itkonen, E. 1974. Linguistics and metascience. Turku: Studia Philosophica Turkuensia II. —. 1978. Grammatical theory and metascience. Amsterdam: Benjamins. —. 1980. Reply to Dahl. In T. A. Perry (Ed.) Evidence and argumentation in linguistics, 146-151. Berlin: deGruyter —. 1981. The concept of linguistic intuition. In F. Coulmas (Ed.): A Festschrift for a native speaker, 127-140. The Hague: Mouton. —. 1983. Causality in linguistic theory. London: Croom Helm. —. 1991. Universal history of linguistics: India, China, Arabia, Europe. Amsterdam: Benjamins. —. 1998. Concerning the status of implicational universals. Sprachtypologie und Universalienforschung. 51 (2): 157-163. —. 1999. Functionalism yes, biologism no. Zeitschrift für Sprachwissenschaft 18 (2): 219-221. —. 2001. Maailman kielten erilaisuus ja samuus [= ‘The diversity and the unity of the world’s languages’, in Finnish], 2nd ed. University of Turku: Publications in General Linguistics 4. —. 2003a. Methods of formalization beside and inside both autonomous and non-autonomous linguistics. University of Turku: Publications in General Linguistics 6. —. 2003b. What is language? A study in the philosophy of linguistics. University of Turku: Publications in General Linguistics 8. —. 2004. Typological explanation and iconicity. Logos and Language 5 (1): 21-33. —. 2005a. Analogy as structure and process: Approaches in linguistics, cognitive psychology and philosophy of science. Amsterdam: Benjamins —. 2008. Concerning the role of consciousness in linguistics. Journal of Consciousness Studies 15/6: 15-33. —. 2005b. Ten non-European languages: An aid to the typologist. University of Turku: Publications in General Linguistics 9. Katz, J. 1981. Language and other abstract objects. Oxford: Blackwell. Levinson, S.C. 2003. Space in language and in cognition: Explorations in cognitive diversity. Cambridge: Cambridge University Press. McKay, G.R. 1988. Figure and ground in Rembarrnga complex sentences. In P. Austin (Ed.), Complex sentence constructions in Australian languages, 7-36. Amsterdam: Benjamins.

Esa Itkonen

29

Mithun, M. 1988. The grammaticization of coordination. In J. Haiman and S. Thompson (Eds.), Clause combining in grammar and discourse, 331-357. Amsterdam: Benjamins. Palmer, B. 2004. Standing Whorf on his head: Evidence from spatial reference on the relationship between language and thought. Paper distributed at LCM I, Portsmouth, July 2004. Paul, H. 1975 [1880]. Prinzipien der Sprachgeschichte. Tübingen: Niemeyer. Perry, T. A. (Ed.) 1980. Evidence and argumentation in linguistics. Berlin: deGruyter. Sapir, E. 1921. Language. New York: Harcourt. Talmy, L. 2000. Toward a cognitive semantics, Vol. I. Cambridge MA: The MIT Press. —. 2007. Introspection as methodology in linguistics. Paper distributed at ICLC X, Cracow, July 2007. Trubetzkoy, N.S. 1958 [1939]. Grundzüge der Phonologie. Göttingen: Vandenhoeck and Ruprecht. Von Wright, G. H. 1971. Explanation and understanding. London: Routledge and Kegan Paul. Wittgenstein, L. 1958. Philosophical investigations, 2nd ed. Oxford: Blackwell.

REASSESSING THE PROJECT OF LINGUISTICS ALEXANDER KRAVCHENKO

Abstract. It is argued that the foundational assumptions of orthodox linguistics may not be accepted as tenable since they are rooted in Cartesian dualism, which precludes a holistic understanding of cognition as a biological phenomenon. As a result, much of what orthodox linguistic thought holds as truths about language (and cognition) is nothing but myths that have never been (and may not be) empirically validated. A general ideological shift in contemporary cognitive science is identified, which consists in taking a holistic (bio-socio-cultural) stance toward language and cognition. The epistemological basis for this shift is provided by autopoiesis as the theory of the living, which opens new perspectives in the study of human cognitive powers. Keywords: language myths, holism, bio-social function of language.

The problem with orthodox linguistics Linguistics as the scientific study of language and communication has a long history. The study of language not only predates all other fields of study such as philosophy, mathematics, or physics, but it is in fact the primary science since language is a precondition for science, both historically and logically. Moreover, as Harris (2005) has convincingly argued, science itself is a construct of language, held together by means of an idiosyncratic semantics designed to give it credibility: science is a construct of language because scientists impose their language on what they assume is there to be named by that language. The role of language in the lives of human beings is of paramount importance, yet if we ask, “What notable effects has the 2-millenium-long study of language produced on human poiesis and teoria as distinguished from praxis1 (in the Aristotelian sense)?” the answer, for all practical

1 Teoria, poiesis and praxis were three basic activities of man distinguished by Aristotle: praxis is an essentially social, communicative activity of free subjects,

Alexander Kravchenko

31

purposes, will be, “None to speak of” – unlike, for example, in physics, chemistry, biology or computer science, to name but a few. Historically, the impetus to the study of language was given by two major practical tasks that emerged as a result of growing cross-cultural contacts – foreign language acquisition and translation. As G. Bruno is claimed to have said, “From Translation all science had its offspring” (Yates 1934: 89). However, to this day linguists have not offered an efficient methodology to deal with the first problem, and as for translation, it still remains a fine art that escapes algorithmisation. Machine translation as an aspiring project of first-generation cognitive science with its conceptual foundation expressed by the metaphor “thinking is computation”, continues to remain a project, and the best way to learn a foreign language, unsurpassed by any of the methods linguists might offer, is still direct cultural immersion. Another problem that has so far resisted solution is first language acquisition. As Jackendoff (1994: 26) observes, “an entire community of highly trained professionals, bringing to bear years of conscious attention and sharing of information, has been unable to duplicate the feat that every normal child accomplishes by the age or ten or so, unconsciously and unaided”.2 But why can’t “highly trained professionals” duplicate the child’s feat of first language acquisition? Could it be that the epistemological foundation of linguistics comes short of explaining the very nature of language? What if “highly trained professionals” have been trained to do something that has little to do with natural language as an empirical phenomenon and a lot to do with what they think language is? And what is language? Because we “do things with words”, we think of language as a kind of purposeful activity aimed at expressing thoughts: “The central role of language is to express thoughts. Derivatively, it has at least these two roles: explaining behavior and informing us about the world. Meanings are the properties that enable to play these roles” (Devitt and Sterelny 1999: 5). Meanings are the “properties” of words, so the study of meaning lies at the heart of linguistic science concerned with language as a sign system “for expressing and communicating thoughts”. As Jackendoff (2002: 267) puts it, meaning is the “holy grail” not only of linguistics, but also of philosophy, psychology, and neuroscience. However, the following observation made by Zlatev (2003: 253) is, to many, indisputably true: and in this it is distinguished from production (poiesis) and from knowledge (theoria) (cf. Oleksy 2002). 2 There is, however, another view: that the child learns his/her first language neither “unconsiously” nor “unaided” (see Zlatev 2008).

32

Chapter Three: Reassessing the Project of Linguistics

Our conception of meaning has become increasingly fragmented, along with much else in the increasing ‘postmodernization’ of our worldview. The trenches run deep between different kinds of meaning theories: mentalist, behaviorist, (neural) reductionist, (social) constructivist, functionalist, formalist, computationalist, deflationist… And they are so deep that a rational debate between the different camps seems impossible. The concept is treated not only differently but incommensurably within the different disciplines.

Now, isn’t it odd that the core discipline within the field of language sciences should be nothing more than a motley array of fancy theories causing feuds among their authors and followers? Couldn’t such situation be interpreted as an indication that there is simply no overarching theory? And if we concede to this assumption, can we really speak of linguistics as the “scientific study” of language? So far linguistics has not succeeded in bringing together in a noncontradictory fashion the two concepts of language, language as a sign system allegedly used for representing knowledge (“expressing thoughts”), and language as communicative activity allegedly used for “explaining behavior and informing about the world” (that is, for exchanging information). I say allegedly, because these two conceptions of language reflect commonly shared beliefs that have never been empirically validated. For example, in communication knowledge, especially socially and culturally shared knowledge, most of the time is taken for granted; it is presupposed and therefore not expressed in talk or text (Dijk 2006). We believe that linguistic signs “represent” something (such as knowledge) because they have meanings, yet we don’t know what meaning is or how signs ‘represent’ (that is, there is no unified theory explaining both meaning and representation). We believe that by making inventories of linguistic signs and their meanings (dictionaries) and rules for combining them (prescriptive grammars) we enable speakers of one language to learn another language by using appropriate textbooks – yet those with relevant experience will argue that, with rare exceptions, this is simply impossible. To make these rather mundane observations is to beg the question “What is wrong with linguistics as a science? Could it have been misled by erroneous beliefs about its object?” It is becoming more and more obvious that methodological inadequacies of traditional linguistics in general, and linguistic semiotics in particular, have been to a large degree responsible for the lack of serious progress in the study of natural language as an empirical phenomenon. The problem with linguistics in general, especially its once major branch

Alexander Kravchenko

33

known as generative grammar, is committing a classical Bayesian error of overlooking one’s priors in trying to establish the place of linguistics among other empirical sciences.3 In a special issue of The Linguistic Review dedicated to the status of linguistics as a “cognitive science”, for example, it has been claimed that “in the past 50 years linguistics has progressively established itself as a genuinely scientific discipline” (Boeckx and Piattelli-Palmarini 2005: 447). Furthermore, it was argued that, since language can be profitably studied as a natural object, “the study of language should share the developmental paths, the assumptions and the explanatory style of the most successful natural sciences, epitomized by theoretical physics” (ibidem: 462). However, even though the Chomskian revolution in the 1950s did in fact turn linguistics into a branch of cognitive science, it did not make generative grammar the one and only legitimate framework for the study of language. Cognitive science itself has since far outgrown generativist assumptions about the nature of cognition in general and language in particular, positing new foundational epistemological principles and using novel approaches in the empirical study of human cognitive abilities. There has been much talk among generativists about the role of linguistics in cognitive science whose ultimate objective is to understand how the mind works, or, as Pylyshin (1999) puts it, to understand “what’s in our mind”. However, the infamous generativist claims that “thinking is computation” and that the mind is a “syntactic engine” for carrying out such computation, that language “is in the mind” and, therefore, a property of the mind, a kind of mental organ (Anderson and Lightfoot 2002) that grows naturally in the biologically normal individual on the basis of limited external linguistic input, can hardly be taken seriously any more. As for the call for linguistics to follow the developmental paths, the assumptions and the explanatory style of theoretical physics, this sounds both misplaced and outdated. Misplaced, because if theoretical physics is best represented by quantum physics, then the explanatory style of the latter adopted by the language sciences should inevitably lead to the conclusion that nothing is determinate, Principles and Parameters or not, and everything is at the same time something else – while this is exactly what generativists have always argued to be ludicrous. Outdated, because contemporary research into the nature and function of human knowledge starts with an assumption that language is prior to science. Linguistics simply 3

For a relevant discussion see, for example, Itkonen (1978; 2003) who has argued that the core of linguistics is not strictly speaking an “empirical science”, since it is based on the explication of (native) speakers’ intuitions, and not on the observation of spatio-temporal events.

34

Chapter Three: Reassessing the Project of Linguistics

cannot follow the developmental paths of the natural sciences because these themselves developed with the development of language, particularly its written form which was brought to life, among other things, for the specific purpose of serving science in its quest for knowledge about the world. Now, what is the problem with linguistic semiotics? There is a strong tradition in semiotic studies to view signs as artificial, conventional, and arbitrary entities intentionally produced by humans for the purpose of communication understood as exchange of encoded meanings. These encoded meanings (or “mental content”, “thoughts”, “abstract knowledge”, “complex propositions”, “conceptual structure”, etc.) are assigned to special material entities (“sign forms” or “sign vehicles”) in order to be conveyed from the speaker’s head into the listener’s head via a given medium (speech, writing or signing) – hence the so-called “code-model” of communication based on the “conduit metaphor” (Reddy 1979). Communication as a dialogical physically grounded activity becomes, in this context, something superficial and irrelevant, and as a result linguistic behavior (languaging) is abstracted from the physical context of its occurrence in real time and space. It is, therefore, not surprising that both generative linguistics and classical representational theories of meaning have little to offer in the way of explaining real living language. To quote Harris (2004: 735), “the ‘circumstantial’ parameter of human communication […] has been scandalously neglected in the history of Western linguistics”. Thus the truly essential properties of language as a physically grounded biologically, socially and culturally determined joint orientational activity of human beings were presented, largely, as marginal. On the one hand, one could say about linguistics what has been said about biology: “In practice, most current research […] is not directed towards solving fundamental problems, but towards accumulating more detailed facts, and unfortunately this may cloud understanding as much as it improves it” (Cornish-Bowden et al. 2004: 715). On the other hand, there is a growing realization among language scientists that until recently linguistics has been busy not so much with the study of language as a natural phenomenon, but of a certain set of theoretical constructs built on a set of often intuitive and empirically unvalidated assumptions and premises. As a result, instead of the study of language as a kind of patterned cognitive activity characterized by complex dynamics, linguistics has largely become the study of theoretical models and conceptual frameworks most of which, one way or the other, lean heavily on the axioms of orthodox thinking about language epitomized in encyclopedias, dictionaries, school and college texts and the like. Because within the tra-

Alexander Kravchenko

35

ditional paradigm the empirical essence of natural language eludes a clearcut definition, the ideal project of linguistics as a science remains, strictly speaking, unspecified: what is the ultimate goal linguistic science sets out to achieve that, in fact, makes linguistics a science? Three questions seem to be invoked here, namely, What is language? What does language do for us? and How does it do it? Depending on how we answer the first one, we may be closer to, or farther away from, illuminated answers to the other two. Seeking to answer these questions, orthodox linguistics builds its “grand palace” on the following seemingly obvious but, nevertheless, intuitive assumptions about language and cognition which, overtly or covertly, imply that human beings are “input-output systems”: 1. Language is something “out there”, in the real world, and an important stage in the development of every human individual is “acquisition” of language. 2. Language is a system of material signs (a code) in which every sign form expresses a certain meaning (or a set of related meanings) attached to it. 3. Linguistic signs represent in the mind objects and things that exist independently in the real world, and our thoughts about these objects. 4. The meanings of linguistic signs are given in advance and may be identified and described as abstract invariants utilized by people in communication. 5. Since knowledge is the reflection in the mind of aspects of reality which itself is represented by linguistic signs, language is a symbolic system for knowledge representation. 6. The function of language is to transfer meanings (thoughts or mental content) from head to head, that is, “telementation”. 7. The transfer of thoughts by means of language lies at the core of human communication. 8. Communication is exchange of information (knowledge) between the sender and the receiver, and the system of linguistic signs is the conduit for such exchange. These basic assumptions underlie specific propositions about particular objects of study which characterize approaches taken within various linguistic fields – such as linguistic semiotics (assumptions 2 and 3), linguistic semantics and pragmatics (assumptions 4-6), computational linguistics (assumptions 2-6), and communication and discourse studies (assumptions 6-8). Various theories developed in the frameworks of specific linguistic

36

Chapter Three: Reassessing the Project of Linguistics

disciplines in orthodox science bear the “birthmark” of these initial axiomatic assumptions and speak unequivocally of their methodological foundation which is Cartesian dualism. Orthodox linguistics is permeated with dualism throughout, and regardless of its fleeting infatuations with one methodological trend or another over the past hundred years or so, it has shown remarkable tenacity in sustaining the two dualistic assumptions about language and communication as its cornerstones: (1) Language is a code and its function is that of representation, (2) Communication is telementation. Linguistics lives by these metaphors failing to see that they are nothing but myths.

Taking a holistic stance There is a curious situation in contemporary linguistics as far as language ontology is concerned. On the one hand, the term “natural” as routinely applied to human language is a commonplace label. On the other hand, it is hardly possible to cite an example of a single work where language has been consistently treated as a natural (empirical) object or phenomenon. Understandably, there is a good reason for that as such an approach would imply taking a holistic stand; that is, it would demand that we look at language as a whole rather than as an assortment of different functional (in the narrow linguistic sense of the term) features treated formally in the domains of syntax, semantics, or pragmatics. Speaking of interpretation as essentially a holistic enterprise, Putnam (1983: 149 ff) makes the following observation: To interpret a language… involves finding a translation scheme, an ‘analytical hypothesis’, which is capable of being learned, capable of yielding ready equivalents in our home language to the expressions being translated, and most important, which is such that when we interpret the speakers of the alien language as meaning what the translation scheme says they mean we are able to ‘understand’ their purposes, beliefs, and behavior. If this is right then the only criteria that we actually have for the ‘content’ of any signs, or sign-analogs, are our intuitive criteria of successful interpretation; and to formalize these would involve formalizing our entire conception of what it is to be human, of what it is to be intelligible in human terms.

In other words, a holistic approach to language is impossible without a coherent philosophical conception of language as a feature of the species

Alexander Kravchenko

37

Homo sapiens viewed empirically, that is, as an organism living in its phenomenal world. If one emphasizes the symbolic aspect of language as a system of signs used to “represent and process knowledge”, then the focus of attention shifts to the study of intrasystemic relationships between different types of signs and structures for representing knowledge (whatever those may be), with inevitable controversies about the nature and character of categories, concepts, representations, etc. Meaning in this case is understood as information, with linguistic structures as its vehicle. Hence, the theory of meaning developed within this framework is heavily burdened with logical formalizations aimed at providing an explicit well-defined theoretical (as opposed to empirical) model of semantics governed by sets of rules which, according to Chomsky (1986), ultimately constitute the so-called Ilanguage as a biologically based feature of the brain. If, to the contrary, one emphasizes the interactional features of language as a kind of joint orientational activity which consists in the continuous making of linguistic choices from a wide and unstable range of variable possibilities in a manner which is not rule-governed, but driven by highly flexible principles and strategies, it is only natural to ask how communication in this case can still be possible, particularly, if one subscribes to the claim that the meaning of a linguistic sign in this case is nothing but its use determined by a set of rules (see Verschueren 1999). It is, therefore, only appropriate to agree with Dirven and Verspoor (1998: 14) that “a more comprehensive view of language as a system of signs must also include the human “conceptualizer” and the world as it is experienced by him”. Over the past decade or so it has become evident that there is a dire need for the language sciences to coordinate their efforts in working out a more realistic framework for the study of natural language as a cognitive phenomenon. Assuming that language as species-specific joint coordinated behavior of humans is most appropriately viewed, analyzed, and described in biological terms, that is, as a kind of adaptive activity involving interactions in a linguistic domain constituted by the complex dynamics of communicative behavior in real time and space, such a framework should integrate various disciplines concerned with the study of different aspects of language, but above all, it should bring together linguistics, semiotics, and biology by placing them on a common epistemological foundation (Kravchenko 2007a). Recently, there have been a lot of criticisms of the core assumptions of orthodox linguistics and mainstream cognitive science (Lamb 2004; Love 2004; Millikan 2005 inter alia). Cognitivism as it was originally

38

Chapter Three: Reassessing the Project of Linguistics

conceived by the “founding fathers” at the dawn of the computer era, postulates that certain behaviors (e.g. language) can be explained only by appeal to “internal cognitive processes”. These internal processes consist of mental representations of information and a finite set of rules that operate on them. The mind’s capacity to automatically receive, store, manipulate, and output information depends on its receiving, storing, manipulating and outputting representations of that information (Eckardt 1993). On the internalist account of cognition, cognitive processes occur within the skin (Harnad and Dror 2006; for a critique see Kravchenko 2007b). By accepting an internalist account of mind and language, firstgeneration cognitive science fails to see the important methodological consequence of separating language and mind from the so-called “external” physical world. As argued by Rosch (1999: 75), “mind and world occur together in a succession of situations which are somewhat lawful and predictable” and, as pointed out by McGee (2005: 22), “experimental results challenge certain notions about the existence of an objective, timeless reality” (see also Maturana 1978; Imoto 2004). In the cognitive paradigm, linguistics as the study of language strives to answer the question about the nature of the relationship between language and mind (brain). However, viewed from this perspective, the cognitive enterprise remains largely disorganized in the sense that a synergistic integration of the various disciplines that examine the mind and brain and the role of language as a cognitive capacity of human beings has not been achieved, and the “true science of language” has not emerged. That is why the following suggestion seems to be well justified: Language should be investigated not so much as an interdisciplinary exercise, but rather as a single enterprise that incorporates theoretical and methodological approaches both across and within disciplines, including neuroscience, psychology, and theoretical linguistics (Walenski and Ullman 2005: 328).

At the same time caution should be taken when bringing together theoretical approaches in neuroscience and psychology on the one hand, and linguistics on the other. While the intrinsically biological nature of the former two has often been taken for granted, the latter, as represented by major schools of thought over the 20th century, appears to have had very little in common with biological science – except, of course, the generativist claim that language is a “mental organ”. True, what generativists say about the language faculty as a genetic endowment of humans may, after all, be not such a wild shot. Still, one cannot help being skeptical about their claims – which Postal (2004) describes as “wide off

Alexander Kravchenko

39

the mark” – that generative grammar studies biology on the grounds that language is a “mental organ”. As Everett (2005:160 ff) comments in his critique of Anderson and Lightfoot’s book The Language Organ (TLO), [I]f Anderson and Lightfoot are correct, Chomsky deserves the Nobel Prize in Medicine, for single-handedly re-opening the apparently moribund science of anatomy and discovering an organ of the human body which has been overlooked in the history of studies of the human body. Of course, the language organ is unusual in that it cannot actually be seen. What could they mean, you ask, by an invisible organ? Surprisingly, the nature of this ‘language organ’ never emerges clearly in TLO.

But this is only to be expected if the “biological perspective” on language advocated for by internalists doesn’t go beyond the anatomical and physiological make-up of humans. Looking for “mental organs” inside the brain which itself is an organ integrated in the nervous system whose function is to control different aspects of the functioning of the living organism, is hardly a gratifying endeavor. With regard to language as human species-specific behavior (“communication”), taking a biological perspective is tantamount to viewing it as a kind of cognitive activity with a distinct biological function. Here the focus should be not so much on the narrow notion of function as specific activity attributed to a particular organ as in the anatomical analysis of the body of an organism, but rather on the broad idea of function as a specifying feature of the living organism as a whole. In the biological theory of cognition proposed by Maturana (1970), living systems are unities of interactions which exist in an environment, they are characterized by circular organization, and this circular organization produces a homeostatic system whose function is to produce and sustain the circularity of organization. But the realization of this function depends on the components which determine it and the continuous production of which is sustained by the circular organization. All the particular features of various kinds of organisms overlay and support this fundamental circularity by sustaining its uninterrupted continuity in sequences of interactions with the constantly changing medium. So, if we really want to take a biological stance toward language, we should ask ourselves the following question: “What is the role of language in sustaining the uninterrupted continuity of a human organism in its sequences of interactions with the constantly changing medium?” In biology, the more complex the level at which one seeks to explain a living system, the greater the need to examine the network of interactions that lie behind the genome (Cornish-Bowden and Cárdenas 2001). As

40

Chapter Three: Reassessing the Project of Linguistics

emphasized by Cornish-Bowden et al. (2004: 716), “the fact that a complex network of interactions connect genes to phenotypes emphasizes the idea that only through the understanding of the whole can we understand the function of the parts”. Yet in theoretical linguistics such as structuralism or generativism there is a profound lack of understanding of language as a whole, and the question “What is language for?” does not seem to be a priority. Morris (1938: 32) was among the first to suggest that, for languaging human beings, “the response to things through the intermediacy of signs is biologically a continuation of the same process in which the distance senses have taken precedence over the contact senses in the control of conduct in higher animal forms”. Thus he explicitly addressed the biological function of language as one of “control of conduct” when the “process of taking account of a constantly more remote environment is simply continued in the complex processes of semiosis made possible by language, the object taken account of no longer needing to be perceptually present” (ibidem). As I have argued elsewhere (Kravchenko 2006), the advances made by contemporary science in the study of semiosis have been facilitated by an understanding that life is, essentially, a semiotic process. Zlatev’s (2003) unified theory of meaning defines meaning as the relationship between an organism and its environment, determined by the value which particular environmental aspects (falling into categories) hold for that organism. The categories into which these environmental aspects fall, are physical aspects, perceived via innate value systems, and cultural aspects evaluated on the basis of conventional value systems consisting (predominantly) of signs. Both innate and acquired value systems serve as control systems by directing and evaluating the organism’s behavior and its adaptation. Without taking into account the reality of these two systems which, in the case of languaging human beings, are integrated on the basis of the relationship of reciprocal causality between them as a result of the history of fine ontogenetic structural couplings between an organism and the environment – that is, without taking an integrationist approach to language as a bio-cultural phenomenon which supports the “fundamental circularity in the organization of the living system by sustaining its uninterrupted continuity in sequences of interactions with the constantly changing medium” – hopes for the “true science of language” to finally emerge will remain futile. Over the past fifteen years or so, new ideologies and approaches within cognitive science have emerged, such as the embodied mind (Varela, Thompson, Rosch 1991) and its ‘next-of-kin’ extended mind (Clark 1997), radical constructivism (Glasersfeld 1995), dialogism (Linell 1998), inte-

Alexander Kravchenko

41

grationism (Harris 1996; Cowley 2002), distributed cognition/language (Hutchins 1995; Cowley 2004; Cangelosi 2007) – to name but a few. Whatever the differences in the stances taken within each particular framework, they all share what seems to be a given, that is, the holistic, or bio-socio-cultural nature of language and cognition. Thus biologism (in a broad sense) appears to be a new general ideology with a very strong appeal to cognitive scientists.4 This naturally raises the issue of new epistemological foundations for cognitive science, as Cartesian philosophy appears to be strikingly at odds with the biological approach, precluding any meaningful discussion of the vast empirical data accumulated over the years by cognitive research. As has been argued elsewhere (Kravchenko 2003), such foundations have been laid out by the biological theory of cognition or autopoiesis (Maturana and Varela 1980). The epistemology of autopoiesis allows us to overcome the shortcomings of traditional philosophy of mind and language based on Cartesian logic with its ontological distinction between mind and body.

Conclusion The “tectonic” biological shift in contemporary cognitive science – and, particularly, in cognitive linguistics (see Kravchenko 2006) – indicates that the time has come to integrate linguistic research with autopoietic epistemology (Kravchenko 2002a). Such integration may lead to deeper insights into the nature of language and its function, relating it to the adaptive function of humans realized in the domain of cognitive/linguistic interactions with the environment. Although the basic theoretical tenets of autopoietic framework have not yet gained noticeable recognition in linguistic circles, particularly, in cognitive linguistics as an influential paradigm for the study of language, the unquestionably vast explanatory potential of autopoiesis lends it as a methodological stronghold for the effort to find solutions to major linguistic issues that heretofore have evaded more or less consistent and/or comprehensible explanation in the framework of the old traditional paradigm — such as the language/mind relationship and language acquisition (Maturana, Mpodozis and Letelier 1995; Cowley 2004; Kravchenko, to appear), linguistic semiosis 4 Of course, not everyone would agree with that. For example, Itkonen (1999) advocates for functionalism while categorically rejecting biologism, although how we can speak of functional linguistics without broaching the subject of the function of language as a species-specific biological feature of Homo sapiens, remains unclear to me.

42

Chapter Three: Reassessing the Project of Linguistics

(Kravchenko 2003), semantics (Kravchenko 2002b, c; 2004), pragmatics (Kravchenko 2001) and the nature of syntax (Kravchenko 2005, 2008).

References Anderson, S.R. and Lightfoot, D.W. 2002. The language organ: Linguistics as cognitive physiology. Cambridge: Cambridge University Press. Boeckx, C. and Piattelli-Palmarini, M. 2005. Language as a natural object – linguistics as a natural science. The Linguistic Review 22: 447-466. Cangelosi, A. 2007. Adaptive agent modeling of distributed language: investigations on the effects of cultural variation and internal action representations. Language Sciences 29: 633-649. Chomsky, N. 1986. Knowledge of language. New York: Praeger. Clark, A. 1997. Being there: Putting brain, body and world together again. Cambridge, MA: MIT Press. Cornish-Bowden, A. and Cárdenas, M.L. 2001. Complex networks of interactions connect genes to phenotypes. Trends in Biochemical Science 26: 463–465. Cornish-Bowden, A., Cárdenas, M.L. Letelier, J.-C., Soto-Andrade, J. and Abarzúa, F.G. 2004. Understanding the parts in terms of the whole. Biology of the Cell 96: 713–717. Cowley, S. 2002. Why brains matter: an integrational perspective on the Symbolic Species. Language Sciences 24: 73-95. —. 2004. Contextualizing bodies: human infants and distributed cognition. Language Sciences 26: 565-591. Devitt, M., and Sterelny, K. 1999. Language and reality. An introduction to the philosophy of language, Second edition. Cambridge, MA: The MIT Press. Dijk, T. van. 2006. Introduction: discourse, interaction and cognition. Discourse Studies 81: 5-7. Dirven, R. and Verspoor, M. (Eds.). 1998. Cognitive exploration of language and linguistics. Amsterdam, Philadelphia: John Benjamins. Eckardt, B. von. 1993. What is cognitive science? Cambridge, MA: The MIT Press. Everett, D.L. 2005. Biology and language: a consideration of alternatives. Journal of Linguistics 41: 157–175. Glasersfeld, E. von. 1995. Radical constructivism: A way of knowing and learning. London: Falmer Press. Harnad, S. and Dror, I. 2006. Distributed cognition: Special issue of Pragmatics and Cognition 14(2): 209-213.

Alexander Kravchenko

43

Harris, R. 1996. Signs, language and communication: Integrational and segregational approaches. London and New York: Routledge. —. 2004. Integrationism, language, mind and world. Language Sciences 26: 727-739 —. 2005. The semantics of science. Continuum International Publishing Group Ltd. Hutchins, E. 1995. Cognition in the wild. Cambridge, MA: MIT Press. Imoto, S. 2004. The philosophical nature of Maturana’s theory of perception. Cybernetics and Human Knowing 11(2): 12-20. Itkonen, E. 1978. Grammatical theory and metascience. Amsterdam: Benjamins. —. 1999. Functionalism yes, biologism no. Zeitschrift für Sprachwissenschaft 18 (2): 219-221. —. 2003. What is language? A study in the philosophy of linguistics. University of Turku: Publications in General Linguistics 8. Jackendoff, R. 1994. Patterns in the mind: Language and human nature. New York: Basic Books. —. 2002. Foundations of language. Oxford: Oxford University Press. Kravchenko, A.V. 2002a. Toward a bio-cognitive philosophy of language. Perspectives: Journal for Interdisciplinary Work in the Humanities, 15 (http://cogprints.org/4002/1/Cogphilosophy_Language.html!). —. 2002b. A cognitive account of tense and aspect: resurrecting "dead" metaphors. Anglophonia. French Journal of English Studies 12: 199—. 2002c. The cognitive roots of gender in Russian. Glossos, 3 (http://www.seelrc.org/glossos/issues/3/kravchenko.pdf) —. 2003. Sign, meaning, knowledge: An essay in the cognitive philosophy of language. Frankfurt/Main etc.: Peter Lang. —. 2004. A new cognitive framework for Russian aspect. F. Karlsson (Ed.), Proceedings of the 20th Scandinavian Conference of Linguistics. Helsinki, January 7-9, 2004. University of Helsinki, Department of General Linguistics, Publications No. 36 (http://www.ling.helsinki.fi/kielitiede/20scl/Kravchenko.pdf). —. 2005. Complex sentence as a structure for representing knowledge. In K. Turewicz (Ed.), Cognitive linguistics -- a user friendly approach, 49-63. Szczecin. —. 2006. Cognitive linguistics, biology of cognition, and biosemiotics: bridging the gaps. Language Sciences 28(1): 51-75. —. 2007a. Essential properties of language, or why language is not a code. Language Sciences 29(5): 650-671. —. 2007b. Whence the autonomy? A reply to Harnad and Dror. Pragmatics and Cognition 15(3): 587-597.

44

Chapter Three: Reassessing the Project of Linguistics

—. 2008. “Everything said is said by an observer”: the cognitive distinction between the infinitive/participle clausal arguments. In J.-R. Lapaire, G. Desagulier and J.-B. Guignard (Eds.), From gram to mind: Grammar as cognition. Vol. 1, 287-304. PUB- Presses Universitaires de Bordeaux. —. to appear. Language and mind: A bio-cognitive view. Proceedings of the 22nd Scandinavian Conference of Linguistics. 19-22 June, 2006. Aalborg, Denmark. Lamb, S. 2004. Language and reality. London; New York: Continuum. Linell, P. 1998. Approaching dialogue: Talk, interaction and contexts in dialogical perspectives. Amsterdam/Philadelphia: John Benjamins. Love, N. 2004. Cognition and the language myth. Language Sciences 26: 525-544. Maturana, H. R. 1970. Biology of cognition. BCL Report # 9.0. Urbana: University of Illinois. Maturana, H. 1978. Biology of language: The epistemology of reality. In G. Miller and E. Lenneberg (Eds.), Psychology and biology of language and thought, 28-62. New York: Academic Press. Maturana, H. and Varela, F. 1980. Autopoiesis and cognition: The realization of the living. Boston: D. Reidel. Maturana, H., Mpodozis J. and Letelier, J.C. 1995. Brain, language, and the origin of human mental functions. Biological Research 28: 15-26. McGee, K. 2005. Enactive cognitive science. Part 1: Background and research themes. Constructivist Foundations 1: 19-34. Millikan, R.G. 2005. Language: A biological model. Oxford: Clarendon Press. Morris, C.W. 1938. Foundations of the theory of signs. In O. Neurath, R. Carnap and C.W. Morris (Eds.), International Encyclopedia of Unified Science, Vol. 1, Part 2. Chicago. Oleksy, M. 2002. Cognition, language, and praxis: Is cognitive linguistics on the verge of the practical turn? In B. Lewandowska-Tomaszczyk and K. Turewicz (Eds.), Cognitive linguistics today, 55-64. Frankfurt/Main: Peter Lang. Putnam, H. 1983. Philosophical papers. Vol. 3: Realism and reason. Cambridge: Cambridge University Press. Postal, P.M. 2004. Skeptical linguistic essays. Oxford: Oxford University Press. Pylyshyn, Z. 1999. What’s in your mind? In E. Lepore and Z. Pylyshyn (Eds.), What is cognitive science?, 1-25. London: Blackwell. Reddy, M. 1979. The conduit metaphor. In A. Arthony (ed.) Metaphor and thought, 284-324. Cambridge: Cambridge University Press.

Alexander Kravchenko

45

Rosch, E. 1999. Reclaiming concepts. In R. Núñez and W. J. Freeman (Eds.), Reclaiming cognition: The primacy of action, intention, and emotion, 61-77. Thorverton, UK: Imprint Academic. Thibault, P.J. 2000. The dialogical integration of the brain in social semiosis: Edelman and the case for downward causation. Mind, Culture and Activity 7: 291-311. Varela, F. J., Thompson, E. and Rosch, E. 1991. The embodied mind: cognitive science and human experience. Cambridge, MA: The MIT Press. Verschueren, J. 1999. Understanding pragmatics. London, etc.: Arnold. Walenski, M. and Ullman, M.T. 2005. The science of language. The Linguistic Review 22: 327-346. Yates, F. A. 1934. John Florio: The life of an Italian in Shakespeare’s England. Cambridge: Cambridge University Press. Zlatev, J. 2003. Meaning = life (+ culture): An outline of a unified biocultural theory of meaning. Evolution of Communication 4: 253296. —. 2008. The dependence of language on consciousness. Journal of Consciousness Studies 15 (6): 34-62.

LINGUISTIC RELATIVITY, MEDIATION AND THE CATEGORIZATION OF MOTION JOHAN BLOMBERG AND JORDAN ZLATEV

Abstract. The relationship between language and thought has always been a debated question. In contemporary research one often addresses the question from the perspective of “linguistic relativity” and one of the semantic/cognitive domains often investigated is that of motion. As fruitful as this research has been in showing the possibility of linguistic influence on categorization, we need to widen our perspective beyond the notion of linguistic relativity as well as the often applied typological analysis of motion of Talmy (1991, 2000). This chapter presents a classification of different theories of linguistic influence in order to achieve an approach that is both broader and scientifically more valid. The outlines of an alternative analysis of motion to that of Talmy (1991) are sketched. This analysis of motion and linguistic influence is then explicitly employed in a psycholinguistic experiment with speakers of French and Swedish. The data are not easily given a unified interpretation. However, we seem to be able to rule out linguistic relativity as a factor in this particular case and to find support for the idea that linguistic influence is dependent foremost on the type of task, not on the specificities of the native tongue – thus supporting the notion of “linguistic mediation” as proposed by Vygotsky (1978).

Introduction Since the dawn of contemplation, man has wondered how language, taken in its broadest sense, is related to thought, also taken in its broadest sense (cf. Bhartrihari VƗkyapadƯya; Augustine Confessiones). After a long period of scientific mistrust, the debated Principle of Linguistic Relativity (Whorf 1956: 213) has been resurrected in contemporary research. Following the typological analysis of Talmy (1991, 2000) one, for example, predicts differences in the categorization of motion to accord with the typological features of the native tongue (Slobin 1996; Finkbeiner et al. 2002; Gennari et al. 2002; Papafragou, Massey and Geitman 2002).

Johan Blomberg and Jordan Zlatev

47

There are several difficulties with this research – some of a conceptual nature, others methodological. The theoretical difficulties are due to an all too narrow interpretation of linguistic influence. Language can in principle influence thought in ways that are not due to differences in the lexical and grammatical resources of the native tongue. In an attempt to overcome this problem, a theoretically open-minded perspective will be presented by which we are better prepared to present predictions and explanations. Another problem is the comprehension of the domain of study: motion. It must be questioned whether Talmy’s typological analysis should serve as one of the basic assumptions in research of this kind. The reasons are several: the analysis is not as exhaustive as one is led to believe. Further, we are faced with a number of conceptual problems. How are we to say that the linguistic expression of motion impinges on its categorization if we do not know “what is motion” and how it is perceived and conceived independent of the assumed linguistic influence? An attempt to overcome these problems has been at the centre of the empirical study described in the second half of this chapter. But first, we need to provide the outlines of a more general conceptual framework in the following two sections.

A classification of theories of linguistic influence This section offers a classification of models proposing a linguistic influence on thought. On the basis of this classification, the most plausible theories will be highlighted and serve for formulating hypotheses presented later on. Since there appears to be a consensus that language does in fact influence cognition at least in some respects, the question in need of an answer is how this impact is to be characterized. Together with the different and contradictory results of prior studies on the linguistic impact upon categorization, this leads us to propose a theoretically openminded approach to the relation between language and thought. Often one distinguishes between “strong” and “weak” versions of linguistic influence where “strong” is stereotyped as implying language as a “prison for the mind”, much like the maze of the Minotaur where not even wings of wax could set you free (cf. Wittgenstein’s famous dictum: “The limits of my language mean the limits of my world”). The “weak” version merely states that knowledge of an entity we cannot perceive is impossible without a corresponding term. An example would be the difficulty in grasping the notion of the basic constituent of matter without introducing a term denoting such a concept, i.e. ‘quark’. This weak version can be seen as both harmless and pointless since no conclusive implications

Chapter Four: Linguistic Relativity, Mediation and the Categorization of Motion

48

about the nature of linguistic impact follow. These two views can, ironically, be said to be due to the linguistic influence of the predicates “weak” and “strong” – these assert that the possibilities for linguistic impact on thought rest on a continuum, i.e. different “hypotheses” will differ only in degree along the same dimension. Instead of presenting them as quantitatively diverging, we suggest that a more fruitful analysis should consider various language-over-thought models as qualitatively different in some crucial dimensions. First, models of the linguistic impact on thought can be said to be either context-dependent or context-independent. The latter would characterize linguistic relativity: regardless of task, context or situation, language will influence thought. A context-dependent influence presupposes a higher degree of freedom for thought and implies a less ubiquitous mode of interaction between thought and language. This is the case with the notion of linguistic mediation (Vygotsky 1978). This parameter classifies theories with respect to the impact imposed by language. That is, how and when thought is affected by language. The other dimension regards the notion of different native tongues leading to differences on a conceptual level. That is, do (as the Whorfian tradition presupposes) differences between languages render differences in thought and categorization? If so, we have a relativistic (or languagespecific) theory. If not, we might label it non-relativistic (or languagegeneral), i.e. what is the effect of having a language as opposed to not having one? While the first parameter classifies theories with respect to their perspectives on thought (mental processes and cognition in general) the latter regards the analysis of language: what is language and which qualities of language lead to a linguistic impact? These two dimensions can be combined in any way, thereby creating four types of theories regarding the relation between language and thought, with corresponding examples: x x x x

Context-independent + Language-general: Language creates (the illusion of) conscious mental states (Dennett 1991). Context-independent + Language-specific: Language patterns affect thought (Whorf 1956). Context-dependent + Language-general: Language is used as a cognitive tool to mediate thought (Vygotsky 1978). Context-dependent + Language-specific: When in the act of speaking, thought is influenced by linguistic patterns, “thinking for speaking” (Slobin 1996).

Johan Blomberg and Jordan Zlatev

49

These models differ qualitatively in their analysis of language and thought; therefore we cannot find enough points of contact to order them on a continuum. The claim of language as the creator of consciousness will be left out in the following. This is so partly because the claim is highly doubtable – if true, infants and animals would merely be “automata” – and partly because it cannot be empirically tested within the methodology here employed, and to the extent that it can at all, the evidence for it are negative (e.g. Beshkar 2008).

The domain of Motion The domain of motion was brought to the attention of contemporary linguistic typology by Talmy (1985, 1991, 2000). Talmy’s approach is to classify motion through its lexicalized and grammaticalized semantic features, using motion event as the over-arching term for (most kinds of) motion. Ever since Slobin (1996), this typology has been widely employed in neo-Whorfian research. Talmy (1985, 2000) considers the “presence of motion”, along with the conceptual components (or cognitive categories) FIGURE, GROUND, PATH (which is regarded as the “core schema” of a motion event) and MANNER to constitute a “motion event”, exemplified in (1). (1)

Othello sailed

from

FIGURE MANNER PATH MOTION

Venice to

Cyprus

GROUND PATH

GROUND

Depending on the way these conceptual elements are mapped to linguistic form-classes, Talmy formulates the basis for his well-known typology. The main difference between an S-language and a V-language is in how PATH is mapped. A V-language typically maps PATH to the main verb of the clause, while MANNER is expressed by an optional constituent or in a separate clause. S-languages on the other hand conflate MANNER into the main verb, while PATH is expressed as a “satellite” (verb particle or prefix)1 or a preposition, as in (1). Furthermore, a V-language is said to follow the boundary-crossing constraint (Slobin and Hoiting 1994), stating that a V-language cannot have a Manner-verb in the representation of an event where a boundary is 1

A satellite is, according to Talmy (1985) an immediate sister to the verb root, i.e. semantically and syntactically the satellite stands in a close relation to the verb root.

50

Chapter Four: Linguistic Relativity, Mediation and the Categorization of Motion

crossed. On the other hand, in an S-language this constraint does not hold, e.g. “Desdemona ran into the room” is a perfect sentence in English. Recently, the validity of the typology has been questioned. It seems to be conceptually as well as empirically problematic – a number of languages, e.g. Tzeltal (Brown 2004), Russian (Smith 2003) and Thai (Zlatev and Yanklang 2004) do not fit easily into the classification. Closer analysis of Thai shows that it behaves as a V-language in certain respects, as an S-language in other, while differing from both in third respects. On a conceptual level, Talmy distinguished translational motion: “an object’s basic location shifts from one point to another in space”, from self-contained motion, where “an object keeps its basic, or ‘average’ location” (Talmy 2000: 35). Talmy’s linguistic typology only regards motion of the first kind, exemplified in (2), but not the second, illustrated in (3): (2) (3) (4)

Hamlet went to the throne room Romeo swam in the pool. Puck ran in the garden.

(Translational motion) (Self-contained motion)

In these examples it is fairly obvious why we should treat them as different kinds of motion, but Talmy’s theory is less clear about examples such as (4): Is the motion involved translational or not? For a more thorough discussion, see Blomberg (2007) and Zlatev, Blomberg and David (in press), which argue that the conceptual framework is in dire need of revision. Rather than using the typological analysis of Talmy, it would be preferable to analyze motion in terms that are more compatible with our experience of motion. Only one such experiential distinction will be highlighted: the boundedness of a motion situation.2 The difference is captured in examples (5) and (6): (5) (6)

Guildenstern strolled into the house. Rosencrantz walked uphill.

(Bounded motion) (Unbounded motion)

If a motion situation is bounded, it implies a state-transition (cf. Vendler 1967). This means that the figure will either depart from a source, pass through a route or reach a goal, or a combination of these. This is not the case for in unbounded motion, where the motion process can in 2

See Zlatev, Blomberg and David (in press) for a more complete explication of the experience of motion, and its relation to language.

Johan Blomberg and Jordan Zlatev

51

principle go on for all eternity, since the entire motion situation is not fixated, just the general direction of the trajectory, as in (6). A sentence representing bounded motion expresses PATH, while DIRECTION is expressed by a sentence representing unbounded motion (Zlatev 2007; Zlatev, Blomberg and David in press).3

Prior studies of linguistic effects on the categorization of motion The most influential studies of language effects in the domain of motion almost exclusively utilize non-linguistic tasks through forced-choice similarity judgements – thereby following Lucy’s (1992) methodological guidelines for research on linguistic impact, i.e. an explicit separation of linguistic categorization from non-linguistic (cognitive) categorization. Often these studies use video-clips, either animated or “real-life” sequences. The general method employed, with little or no modifications, has been the use of triads: a target situation is presented along with two alternatives, one differing from the target with respect to PATH (and in some cases what we call DIRECTION) and the other with respect to MANNER. The participant’s task is to decide which of the two “is the most similar” to the target. These options can be shown simultaneously, sequentially, together with or separate from the target. Sometimes the participants have been asked to give a linguistic description of the target prior to choosing, in order to test a hypothesis akin to thinking-forspeaking. The most common hypothesis claims that if language impinges on categorization, speakers of a V-language should be predisposed to prefer the same PATH rather than the same MANNER to a greater extent than speakers of S-languages, where both components are expressed equally easy. Thus, the binary typology presented by Talmy (1991, 2000) is an explicit assumption in these studies. Finkbeiner et al. (2002) tested speakers of English (S-language) along with Japanese and Spanish (V-languages). Their stimuli consisted of a number of 40-second long 3D-video animations where the alternatives were presented either simultaneously with or after the target. The motion situations presented were chosen deliberately so that they could not be easily expressed by simple verbs in any of the three languages. In terms of 3

Terms in “small caps” denote the supposedly non-linguistic conceptual categories (i.e. PATH), as in Talmy’s analysis. On the other hand, a single capital letter (i.e. Path) indicates (language-specific) semantic categories (cf. Zlatev 1997).

52

Chapter Four: Linguistic Relativity, Mediation and the Categorization of Motion

task-dependency, the fact that all information is perceptually available in the parallel presentation can be expected to counteract mediation through semiotic resources such as language. Their data showed a considerably stronger preference for MANNER in the English group when the alternatives were presented after the target, and thus support for some degree of linguistic relativity. When the three clips were presented simultaneously the MANNER-bias of the English group disappeared, leading the authors to conclude that “the apparently nonlinguistic task used in Experiment 1 [i.e. serially] actually encouraged the participants to encode the scenes linguistically” (ibid: 454). When the options were presented serially with respect to the target, the “manner prominence” (Slobin 1996) of English as an S-language became salient. Gennari et al. (2002) compared speakers of two prototypical S- and V-languages, English and Spanish, using 2D-animations with the targetand choice-part presented sequentially. Significant correlations between language and choice were found when the subjects were asked to describe the target-clip prior to viewing the options and making the choice, but not otherwise. Papafragou, Masely and Gleitman (2002) suggested an alternative interpretation of a linguistic effect: Since MANNER is often expressed in non-obligatory constituents in V-languages, when it is expressed it would be “foregrounded” and thus achieve more semantic salience (Talmy 1985) than in an S-language where MANNER and PATH are expressed with equal ease. They compared categorization preferences, using static pictures, between speakers of Greek (V-language) and speakers of English. Despite the differences in the linguistic descriptions that followed the predicted patterns (along the lines of Slobin’s research), they found no preference for either same- Path or same-Manner choices in either group. Pourcel (2005) conducted two neo-Whorfian studies on speakers of English and speakers of French. The first involved a categorization task with video clips of simple real life human motion and the second a storyretelling task with a clip from the Charlie Chaplin-movie City Lights as stimulus. In the first study Pourcel found no relativistic effect on categorization with or without linguistic description prior to judgment. Instead, the results showed an overall preference for same-PATH choice independent of language. An interesting finding was that two types of motion situations, corresponding to the distinction between bounded and unbounded motion described earlier, were categorized differently. There was a strong PATH-bias for the bounded cases (what Pourcel calls “telic Path”), but when the judgement task was preceded by a linguistic

Johan Blomberg and Jordan Zlatev

53

description this bias was nullified, and even shifted to a MANNER-bias for the unbounded cases (“atelic Path”) (ibid: 243-245). In the memory task some support for a “weak” Whorfian claim was found. The subjects were shown the clip and asked to retell it and answer some questions 24 hours later. In this task, the French speakers made greater use of Path in their descriptions and had difficulties remembering instances of MANNER, while the English speakers displayed the opposite tendency. Bohnemeyer et al. (ms.) conducted the most extensive neo-Whorfian research study on motion with participants of 17 typologically, genetically and aerially diverse languages. Their stimuli tool, The Event Triads elicitation tool, is the same as the one used in our study, albeit with some modifications in our version. The stimuli involved animated sequences of a tomato-like figure moving in different ways between some boundaries. Since the figure was spherical to its form, MANNER was limited to rolling, spinning, sliding, and bouncing. The trajectory was either: x x x

Out of a hut, across a field and into a cave (or vice versa). From a tree, across a field to a rock (or vice versa). Up or down a ramp.

After a 5-second long delay subjects were presented with two alternative clips: one same-PATH and one same-MANNER, but otherwise identical. The tool has been criticized by e.g. Pourcel (2005: 147) as unnatural and containing a limited kinds of MANNER. However, the design of the stimuli allows systematic contrasts only in the relevant dimensions of the motion situation, i.e. PATH (or DIRECTION) vs. MANNER and the task was not found “unnatural” by the speakers of the 17 languages, including those living in traditional cultures in South America and Asia. The authors found an overall same-MANNER bias, but with large differences between languages, from 85% (Polish) to 43% (Jalonke and Jukatek). The predicted Whorfian effect could not be established since the results did not coincide with typological belonging. The conclusion was that the binary typology of S- and V-languages is insufficient for predicting categorization preferences. Furthermore, the authors argued that a better conceptual and methodological framework is necessary to conduct neo-Whorfian studies. The study did however yield one interesting finding: across languages the scenes of the tomato moving up or down a ramp was predominantly categorized with a same-PATH (or DIRECTION) preference. The explanation suggested by Bohnemeyer et al. is that these scenes might be simpler since

54

Chapter Four: Linguistic Relativity, Mediation and the Categorization of Motion

they involved just one landmark, where the two other types of situations involved two. Another explanation is however possible: the situation was at least ambiguous between just moving upward or downward and moving to “the top” or “the bottom” of the ramp. That is, there might be a difference in cognitive categorization between PATH and DIRECTION. To sum up, the most significant conclusions from previous studies are that: x The nature of the stimuli, real life contra animated clips, static vs. dynamic seems to affect the categorization behaviour of the participants. x Linguistic description plays an important role: whether it is present or not as well as whether it is prior to choice (or not). x Motion situations differing between PATH and DIRECTION seem to differ in cognitive categorization. x It is preferable to compare more than two languages.

An empirical study of motion and language Following the general procedure of prior neo-Whorfian studies, we used a forced choice similarity judgments task with speakers of French and Swedish. The participants were asked to choose which of two options “were most similar to a target”. The two options differed from the target with respect to either PATH/DIRECTION or MANNER.

Figure 1. An example triad from the stimulus tool Event Triads.

In total, there were 12 triads of three different kinds, which were analyzed separately according to (i) choice (ii) correlations between choice and description. In order to study the pervasiveness of the linguistic effect, we split the French and Swedish participants in two groups (FG1, FG2 and SG1 SG2). Participants of FG1 and SG1 viewed the entire triad and made their choice. For FG2 and SG2 stimuli were modified to allow a break between

Johan Blomberg and Jordan Zlatev

55

target and option part where participants were asked to describe the targetpart prior to choice.

Predictions Due to the complex relation between language and thought, together with the contradictory results in prior empirical studies, a theoretically openminded approach seemed preferable. Such an approach can serve as the basis for the formulation of several alternative hypotheses. On a number of parameters, these models yield different predictions. These parameters primarily involve differences within language (i.e. F/SG1 vs. F/SG2) and differences between languages (i.e. FG1/2 vs. SG1/2). Linguistic Relativity Due to the ubiquitous effects of language on thought we should not find any significant differences between the two French and the two Swedish groups, i.e. between the groups of the same language. The judgment task will however unveil differences between speakers of typologically diverging languages (i.e. FG1/2 and SG1/2 will categorize differently). It might be that this effect is somewhat stronger for the G2-groups, since language is explicitly used. A strong correlation between description and choice are expected on all levels (i.e. triad, individual, and group). Linguistic Mediation We should find clear-cut differences in behaviour between F/SG1 and F/SG2. In the task of describing prior to choice the prediction is a preference towards the more complex categorization in terms of PATH than a simpler categorization in terms of MANNER (and DIRECTION). This difference in complexity is a reflection of the fact that PATH implies an analysis of the whole situation, while MANNER implies focus on a particular (local) aspect. There are at least three different reasons why these differences between the groups can be expected: x The description task (typically) involves utterances, which “parse” the situation into its components (“from X, to Y”), rather than focusing on particular aspects, where MANNER is one. x The temporal gap between the target and the options might yield an effect on short-time memory, where the linguistic description is used as a memory resource, to facilitate solving the task. x There are two tasks to solve for the participant: describing and

56

Chapter Four: Linguistic Relativity, Mediation and the Categorization of Motion

choosing. It could be the case that this is more cognitively demanding than doing just one at the time, and more difficult task tend to promote linguistic mediation (cf. Vygotsky 1978). Participants in F/SG1 will, on the other hand, categorize on the basis of less mediated processes, and thus tend to match the feature most perceptually salient in the target-part with its most similar alternative, i.e. MANNER. Thinking-for-speaking A “compromise” between the two previous predictions, follows from thinking-for-speaking. A linguistic impact is likely to appear when verbalizing is performed prior to choice. It differs from linguistic mediation insofar as differences will also depend on typological differences between languages, as predicted by linguistic relativity. For FG2 we will find a significant same-PATH preference, since French is a V-language, and for SG2, a same-MANNER bias, together with strong correlations between description and choice on a triad-by-triad basis. For FG1 and SG1 we cannot predict any (strong) linguistic effect on categorization preferences.

Results The overall results in the similarity judgment task for the French and Swedish groups are shown in Figure 2.

Figure 2. Total results of same-PATH/DIRECTION vs. same-MANNER preference for French (FG1) and Swedish (SG1) Group 1 (post-choice description) and French (FG2) and Swedish (SG2) Group 2 (pre-choice description). Total number of choices is 144 per group.

Johan Blomberg and Jordan Zlatev

57

These results are highly indicative. Whereas the similarity judgments for Group 1 (post-choice description) were similar to those in prior studies with Event Triads and practically identical for the two languages (chi2(1) = 0.14, p > 0.05), i.e. a preference for same-MANNER choices, the situation was completely reversed for Group 2 (pre-choice description), with a surprisingly strong bias for same-PATH/DIR choices. The difference between Group 1 and Group 2 was highly significant (p < 0.0001). Furthermore, there was a stronger PATH/DIR bias for SG2 than FG2, which was also significant (chi2(1) = 4.964, p=0.026). When we divided the 12 test triads according to the three types of situations (OUT/INTO, FROM/TO, VERTICAL), we noticed, however, also a difference between Group 1 in the case of VERTICAL: the MANNER-bias was neutralized for both the Swedish and the French speakers, see Figures 3a and 3b.

Figures 3a, 3b. The results for the two Swedish (3a) and the two French groups (3b), divided by the three different situation types: OUT/INTO, FROM/TO and VERTICAL. Total number of choices per situation type and group is 48.

58

Chapter Four: Linguistic Relativity, Mediation and the Categorization of Motion

We coded the linguistic descriptions for the presence of Manner expressions: Manner verbs such as hoppa and sautille (‘jump’) and adverbials such as snurrande or en roulant (‘rolling’), Path expressions such as från or de (‘from’) and till or á (‘to’) and Direction expressions such as upp (‘up’) or monte (‘climbs’) and ner (‘down’) or descend (‘descends’) and looked for correlations between the presence of these elements and the choices of the subjects (cf. Table 1). Table 1. Correlations of significant value (Pearson’s Correlation, significant at > ± .3 at the .05-level, two tailed) between elements in the descriptions and corresponding choice for the two groups of French speakers (FG1 and FG2) and the two groups of Swedish speakers (SG1 and SG2), divided by situation type (FROM/TO, OUT/INTO and VERTICAL) Non-existing or non-significant correlations are marked as “×”. Group

Type

FG1

FROM/TO OUT/INTO VERTICAL

FG2

FROM/TO OUT/INTO VERTICAL

SG1

FROM/TO OUT/INTO VERTICAL

SG2

FROM/TO OUT/INTO VERTICAL

Direction × × × × × × × -.338 -.302 × × +.674

Path × × +.308 × × × -.309 × × -.307 -.443 ×

Manner × × × × - 304 -.329 × × × × × ×

There were few positive correlations: for Path-VERTICAL (FG1) and for Direction-VERTICAL (SG2). In qualitative terms, this means that if a Swedish pre-choice verbalizing subject had used a Direction expression, he was more likely to make a same-DIRECTION choice than same-MANNER choice. We are not sure how to interpret the negative correlations for SG1, SG2 and FG2. On the face of it, it seems that, for instance, if a French speaker had used a Manner expression (in the pre-choice description group, FG2), he was less likely to make a same-MANNER choice. Also, it was surprising that the positive correlation for the French speakers was for FG1, the post-choice describing group, and it involved Path, rather than Direction expressions. In other words, the results do not lend themselves to an explanation in terms of the thinking-for-speaking hypothesis. According to the latter, and the classification of Swedish as an S-framed and of French as a V-framed language, one would have expected a

Johan Blomberg and Jordan Zlatev

59

Manner-MANNER correlation for SG2 and a Path-PATH correlation for FG2. In fact, the only really strong positive correlation was for SG2, and it involved DIRECTION rather than MANNER. By method of exclusion, the findings offered strongest support for the model of linguistic mediation.

Conclusions We began this chapter by arguing that at least 4 qualitatively different types of models of the relation between language and thought need to be distinguished, and classified three such models, currently used in motion research as belonging to three of these types: linguistic relativity (Whorf 1956), thinking-for-speaking (Slobin 1996) and linguistic mediation (Vygotsky 1978). The results from our empirical study within the domain of motion could be interpreted most straightforwardly as offering support for the latter (using speakers of French and Swedish, and the modified Event Triads tool). Most clearly we found that the categorization of motion situations changes when these are described in language, promoting a more analytic attitude (realized in the case of motion situations as a greater focus on PATH). Furthermore, the study offered empirical support for the conceptual distinction between PATH and DIRECTION (Zlatev 2007). On a more general level, the findings suggest that we need to move beyond the conceptual confines of present day neo-Whorfian research. Language can influence thought in different ways, without it being the case that different languages yield different “world-views”. Linguistic mediation implies task-dependency, as originally proposed by Vygotsky (1978). The preference for PATH in the more linguistically mediated task in the study is hardly surprising, if one considers that it involves the whole structure of the motion situation, the “core schema” as suggested by Talmy (2000). Finally, the parameter of boundedness, constituting the distinction between PATH and DIRECTION, is only one side of the conceptual issues involving situation types (Vendler 1967) and the semantics of Aktionsarten and grammatical aspect, which only recently have entered the discussions in the field of motion semantics (cf. Zlatev, Blomberg and David in press). However, we should be cautious not to over-generalize the conclusions from the study presented in this chapter. The language-though nexus needs to be much further explored, using different methodologies (e.g. using eye-tracking and memory tasks) as well as within other domains than space and motion.

60

Chapter Four: Linguistic Relativity, Mediation and the Categorization of Motion

References Beshkar, M. Animal consciousness. Journal of Consciousness Studies 15(3): 5-34. Blomberg, J. 2007. Linguistic relativity, mediation and the categorization of Motion. MA Thesis: Centre for Language and Literature, Lund University. Bohnemeyer, J., Eisenbeiss, S. and Naranhimsan, B. 2001. Event Triads. In S.C. Levinson and N. Enfield (Eds.), 'Manual' for the field season 2001, 100-114. Nijmegen: Max Planck Institute for Psycholinguistics. Bohnemeyer, J., Eisenbeiss, S., Naranhimsan, B. (ms). Ways to go: methodological considerations in Whorfian studies on motion events. Dennett, D.C. 1991. Consciousness explained. Toronto: Little Brown. Finkbeiner, M., Nicol, J. Greth, D. and Nakamura, K. 2002. The role of language in memory for actions. Journal of Psycholinguistic Research 31(5): 447-57. Gennari, S.P., Sloman, S.A., Malt, B.C. and Fitch, W.T. 2002. Motion Events in language and cognition. Cognition 83: 49-79. Lucy, J. 1992. Language diversity and thought: A reformulation of the linguistic relativity hypothesis. Cambridge: Cambridge University Press. Papafragou, A., Massey, C. and Gleitman, L. 2002. Shake, rattle n’ roll: the representations of Motion Events in language and cognition. Cognition 84: 184-219. Slobin, D.I. 1996. From “thought and language” to “thinking for speaking”. In J. Gumperz and S.C. Levinson (Eds.), Rethinking linguistic relativity: Studies in the social and cultural foundations of language, 70-96. Cambridge, MA: Cambridge University Press. Smith, V. 2003. Motion at the sugar factory. Is Russian a genuine Manner language? Proceedings of the 14th symposium on LSP, University of Surrey, August 18th- 22nd 2003. Talmy, L. 1985. Lexicalisation patterns: semantic structure in lexical form. In T. Shopen (Ed.) Language typology and syntactic descriptions, Vol. 3: 57-149. Cambridge: Cambridge University Press. —. 1991. Path to realization: a typology of event conflation. Proceedings of the seventeenth annual meeting of the Berkeley Linguistics Society: 480-520. —. 2000. Toward a cognitive semantics: Concept structuring Systems, 1. Cambridge, MA: MIT Press. Vendler, Z. 1967. Linguistics in philosophy. Ithaca: Cornell University Press.

Johan Blomberg and Jordan Zlatev

61

Vygotsky, L.S. 1978. Mind in society. Cambridge, MA: MIT Press. Whorf, B.L. 1956. Language, thought and reality. Selected writings of Benjamin Lee Whorf (Ed. J.B. Carroll). Cambridge, MA: MIT Press. Zlatev, J. 1997. Situated embodiment: Studies in the emergence of spatial meaning. Stockholm: Gotab. —. (2007) Spatial semantics. In H. Cuyckens and D. Geeraerts (Eds.) The Oxford Handbook of Cognitive Linguistics, 318-350. Oxford: Oxford University Press. Zlatev, J., Blomberg, J. and David, C. In press. Translocation, language and the categorization of experience. In P. Chilton and V. Evans (Eds.), Space in language and cognition: The state of the art. London: Equinox.

CAN WE TELL WHAT WE SAID WHEN WE HEAR OURSELVES SAYING SOMETHING ELSE? REAL-TIME SPEECH MANIPULATION AS A NEW INSTRUMENT FOR THE TESTING OF VERBAL INTENTIONS AND SELF MONITORING

ANDREAS LIND, LARS HALL, PETTER JOHANSSON AND SVERKER SIKSTRÖM

Abstract. It is commonly assumed that speech production is initiated and guided by clear pre-verbal intentions. These intentions are moreover supposed to function as a standard of accuracy against which actual performance can be measured. Alternative positions to such a centrally governed process have been offered, but very few empirical attempts have been made to explore the issue. We aimed to remedy this by investigating the role of auditory feedback of one’s own voice in the understanding of the meaning of self-produced speech. Participants performed a computerized Stroop test while hearing their own voice exclusively through earphones, and we covertly recorded three of their utterances. These words were then played back to the participants over the headphones while they were engaged in a different trial of the test, effectively creating a situation where the participants heard themselves saying something other than what they actually said (i.e. we exchanged what they said for something they had said earlier). The results showed that such manipulations were almost always retrospectively detected, but that a majority of the participants reported that they had experienced significant confusion as to the actual source of the manipulated feedback, and not being certain if it was produced by themselves or not. In addition, on a minority of manipulated trials, participants acted towards the manipulated feedback as if it was self-produced.

How can I tell what I think till I see what I say? —E.M. Forster, ‘Aspects of the Novel’ (1927)

Andreas Lind, Lars Hall, Petter Johansson and Sverker Sikström

63

Introduction Levelt (1989) formulated the most influential modern theory of speech production. In this model, first a preverbal message is formed in a processing system named the conceptualizer. The formation of the preverbal message is dependent on procedural, declarative and situational knowledge, and it involves forming an intention, selecting and ordering relevant information to express this intention, and keeping track of the things that have been said earlier in the conversation. All this work is performed by the conceptualizer. The preverbal message then serves as input for a system called the formulator, which translates the conceptual structure of the preverbal message into a grammatical and phonological code. This plan is then sent to the articulator, which produces audible speech by way of the articulatory apparatus. While differing in detail, there is an important point of agreement between Levelt’s model and other accounts of production (e.g. Dell 1986; Fromkin 1971) in that Levelt does not speculate about what happens before the content of a message is determined. He grants us a capacity for conceptualization but does not see it as his task to uncover what processes support it. Levelt states that “the mother of each speech act is a communicative intention” (1989:108). Thus, for all intents and purposes, the conceptualizer is what gets everything started. But he also makes it clear that “where intentions come from is not a concern of this book” (ibid: 59). And even though Levelt himself remarked that the conceptualizer was a simplification that would have to be unpacked in future investigations (ibid: 9), it has remained firmly in place in the model without being further developed. Avoiding the question of why a speaker says what is said, and focussing instead on how it is said (Dell 1986), can be seen as a pragmatic strategy. It clearly is motivated by the extreme difficulty of trying to get a grip of the processes involved in specifying the content of speech-acts. Such avoidance allows the researcher to focus on aspects of production where there is either existing empirical evidence or the potential for quantitative predictions. However, as a result of this strategy the delegation of the process of conceptualization to a conceptualizer has been solidified and is routinely not questioned. Today, many accounts of speech production begin with simply stating that all speaking starts with conceptualization, and do so without defining what goes into this conceptualization or what mechanisms support it (see e.g. Griffin 2004; Meyer 1992; Postma 2000 and textbooks such as Carroll 1999 and Harley 2001).

64

Chapter Five: Can we Tell What we Said When we Hear Ourselves Saying Something Else?

However, not all researchers endorse the strategy of positing central executives like the conceptualizer in psychological models. Dennett (1987, 1991) is one of the critics of this practice. He warns against relying too much on folk-psychological concepts, such as intention, when building scientific models of behaviour (1987) and he specifically questions Levelt’s reliance on the conceptualizer as both the point of origin and supervisor of the whole process of word-production (1991). In Levelt’s model the conceptualizer produces finished intentions that guide the speech-production process in a serial manner. But, Dennett cautions, postulating a “thought-thinker [that] begins with a determinate thought to be expressed” only moves the problem back one step (ibid: 241, emphasis in original). If we want to know how the content of speech-intentions is actually specified or fixed, we must not be tempted to leave the how up to this “thought-thinker”. Dennett (1991) offers some interesting suggestions for how the notion of a conceptualizer can be revised to open up for the possibility of empirically investigating the mechanisms specifying the content of speech-acts. He proposes a distributed, so-called “pandemonium model” of production where the content of speech is opportunistically determined in a “cobbled-together collection of specialist brain circuits” that have evolved independently, but now function jointly to much greater power. Within the conglomeration of these brain circuits, which Dennett nicknames “the Joycean machine”, there is a competitive process where no one circuit is ever in control for long. Rather, power is shifted around in a “quasi-evolutionary process” where the current mind-set of the speaker functions as a constraining mechanism that “judges” all the “suggestions” for speech-acts that float around in the Joycean machine at a specific time. In effect, several different variants of a proposed utterance are, often non-consciously but sometimes consciously, weighed against each other in a “collaboration of various subsystems none of which is capable on its own of performing – or ordering – a speech act” (ibid: 239). Eventually one of the proposed utterances gains strength and is spoken. The weighing is carried out on a scale of microseconds and, importantly, this means that the content of the speech act is not fully specified or fixed until it is actually spoken. If the situation demands it, and if we are hard pressed not to say the wrong thing, we can carry out a silent, conscious rehearsal of some of the propositions. But when speaking casually there is no time for this and we learn the specific content of our speech act at the same time as our listeners do. Furthermore, the process of conceptualization is not necessarily over once an utterance has been pronounced. What is spoken can work as a

Andreas Lind, Lars Hall, Petter Johansson and Sverker Sikström

65

confirmation of what we actually mean, and this can then be further reacted to, and acted upon in a continuous process of conceptualization. Dennett (1991) does not deny the existence of a process of conceptualization, only the existence of an “inner conceptualizer that is a proper part of the language-producing system”. Rather, he postulates a global conceptualizer which is equivalent to the whole person that does the uttering and “of which the language-producing system is itself a proper part” (ibid., p. 251). Dennett’s pandemonium model, while providing enthusiastic suggestions for testing uncharted aspects of conceptualization, has not received much interest in studies of speech production (though see Levelt, Roelofs and Meyer 1999 for a brief mention) and attempts at trying to test it empirically have been virtually nonexistent.

Real-time manipulation of voice feedback The present chapter describes a project which is an attempt to fill this empirical gap by studying the role of verbal feedback and self-monitoring during the process of conceptualization. Our goal is to create an artificial language loop outside the head of individuals, where it is possible to insert, remove, manipulate, or time-shift snippets of language use, and where we can determine whether this is noticed or not by the speakers, and what effects it might have on their thoughts, beliefs and preferences. We do this by fitting participants with sound isolating headsets and microphones. During pretest, or training trials, we record and extract words and utterances by the speaker which later can be inserted in realtime in the loop, effectively creating a situation where the participants hear themselves saying something other than what they are actually saying (i.e. we exchange what they say for something they have said earlier). In this way, we can create a contrast between the purported internal communicative intention, and the actual outcome of the utterance, and measure the reactions of the participants to this unexpected feedback. If, as Dennett contends, the process of conceptualization is not to be considered finished until after the utterance has actually been spoken, then it is not obvious that such manipulations will be noticed by the participants. On the other hand, from the perspective of Levelt (1989) self-monitoring is continually active both at the level of internal speech and the level of overt speech, and the influence of manipulated feedback on the participants’ understanding of what they have said should be negligible.

66

Chapter Five: Can we Tell What we Said When we Hear Ourselves Saying Something Else?

For a believable voice-exchange to occur we must know both when the participants will say the words we want to record and when they will say the words we want to exchange with the recorded ones. For these reasons, the Stroop test (Stroop 1935) was chosen for implementing the present experiment. In the Stroop test you are shown colour-words; the letters of these colour-words have a specific colour that does not always match the written word. The task is to always name the colour of the letters, while ignoring the spelled word. Apart from when participants make mistakes in the test, it is therefore fairly easy to predict what they will say and when they will say it. Our primary question was how will participants react to the manipulations? Would they immediately realize that the feedback did not match what they meant to say? Or would they take the manipulated feedback into consideration in the ongoing process of conceptualization and perhaps adjust their understanding of what the meaning of their utterance was?

Method Participants Forty-six participants (23 females) were drawn from a student population. Their mean age was 23.5 years (sd 2.4). Only native speakers of Swedish were included. All of the participants had normal hearing, and normal or corrected-to normal vision. Participants were unaware of the actual purpose of the experiment, but gave their informed consent after postexperiment debriefing and interviews.

Stimulus materials A computerized version of the Stroop test was employed (see e.g. Linnman et al. 2006). The colours blue, green, grey, red and brown were used, all written in their Swedish counterparts. The colours were modified so that the grey and green colours and the grey and blue colours were more similar, while still retaining their distinctiveness. The use of standard colours was avoided because it would increase the likelihood of participants getting too strong visual cues about what they had actually said. All words were preceded by 4 fixation points (****) and were presented one by one in lower case letters within a fixation-square centred on the computer screen. The words were shown for 200 ms and the interval between words was 1500 ms. The test consisted of 50 trials.

Andreas Lind, Lars Hall, Petter Johansson and Sverker Sikström

67

Three of these trials were manipulated so that the participants said one word but, through the earphones, heard themselves say another word. The 25 possible word/colour-combinations were randomly distributed in the trial sequence, each combination appearing twice. 10 of the trials were in a congruent condition (i.e. the colour of the letters matched the written word). Including congruent trials in the Stroop test has been established as a potent way of preventing participants from adopting strategies to exclude the semantic information (such as just glancing at the words, see MacLeod 1991). As a further measure against participants adopting such strategies, five fruit and vegetable words were inserted in the sequence, and the participants were requested to pronounce the fruitor vegetable word whenever they appeared, while at the same time ignoring the colour of the letters. An attempt to induce the distractor suppression effect (Neill 1977) was made in order to elevate the general frequency of mistakes, so that participants would not be able to reason from simple base-rates that manipulated trials were unlikely to be actual instances of mistakes. If the to-be-suppressed, spelled word on trial n-1 is the same as the to-be-named colour on trial n, then this will increase the interference on trial n (MacLeod 1991), possibly also increasing the chance of mistakes. Two such sequences were placed among the first 20 trials and then there was one between each manipulation. Manipulated trials were inserted on trials 21, 33 and 45. The colour/word-combinations for the manipulated trials were blue/grey, green/grey and grey/blue. This decision was made based on phonological similarity. In Swedish, blue is pronounced [bloʋ], grey is pronounced either [Ƞȸoʋ] or [Ƞɀoʋ], depending on dialect, and green is, similarly, pronounced either [Ƞȸøʋn] or [Ƞɀøʋn]. On manipulated trials, the written word always matched the manipulated feedback, thus avoiding a situation where the manipulated feedback would run counter to all visual cues, and provide additional sources of mismatch detection. A pre-test of 15-18 trials preceded the actual test. Here, incongruent, congruent and fruit/vegetable-trials were represented. While this pre-test was presented as a joint practice- and calibration session, its main purpose was to provide a chance to record the three words that were later used during the manipulated trials. Due to the audio processing demands of the software program used a low buzzing was present in the auditory feedback on all non-manipulated trials, but disappeared during the manipulated trials. During pilot testing, several participants remarked on this as a prime reason for detecting the manipulations. Therefore, this presence/absence of buzzing was masked

68

Chapter Five: Can we Tell What we Said When we Hear Ourselves Saying Something Else?

by playing back slightly louder homogenous white noise from a separate Mp3-player through the duration of the test. The Stroop test was presented on a 17” monitor connected to a laptop, on which the program was run. The computer, and the Mp3-player which outputted the white noise, were both connected to a mixer. Leading from the mixer were two sets of earphones (one set for the participant and one for the experimenter) and a cord connecting it to a portable recording device able to record on two separate channels. Also connected to the recording device was a stationary table-microphone placed in front of the stimulus-monitor.

Procedure The participants were seated in front of the stimulus screen and the experiment was presented as a common Stroop test – i.e. they were told that colour-words would appear one by one on the screen and that the letters of these words would have a specific colour that do not always match the actual word. The task was to name the colour of the letters. Participants were also told that, on a small number of trials, fruit- or vegetable-words would appear and that when they did, they were to read the word rather than name its colour. Furthermore, they were told that it was important that they would speak in a comparable voice and volume during the whole experiment. During pre-test, the participants did not wear headphones but solely spoke into the microphone. After the pre-test, participants were equipped with earphones and hearing protection. The hearing protection was used to minimise the natural feedback of their voice. They were also told that it is very common for people to make mistakes in this test, and that it was absolutely fine if they did so. In order to distinguish between the rivalling hypotheses about conceptualization, we instructed participants to signal the experimenter each time they made a mistake in the Stoop test, simply by saying “wrong!” as quickly as possible before moving on to the next word. A signal indicating that a mistake was made during a manipulated trial would be taken as a measure showing that the manipulated feedback exercised influence on the meaning eventually ascribed to the utterance by the participant. We also wanted to probe the experiential aspects of the voice exchange and this was done by asking the participants a series of questions directly after the experiment. They were first asked what they thought of the experiment, then if they had made any mistakes, and then if they had noticed anything strange with the feedback. If for either of

Andreas Lind, Lars Hall, Petter Johansson and Sverker Sikström

69

these questions the participants indicated that they had detected the manipulation(s), they were first asked how many times they had noticed them. Then they were asked in a neutral manner how they had experienced the manipulations, and were asked to describe in detail how they felt during the manipulations. If the participants stated that they had experienced significant confusion at the mismatched feedback, additional follow-up questions were posed about whether this was a general feeling of uncertainty or a particular doubt that they might have said the manipulated word or not, and also if this description applied to all the manipulated trials or only one or two of them. Finally, participants were asked if they had become suspicious about the feedback as the experiment progressed.

Data analysis The performance of each participant, complete with post-test interviews, was recorded and organized in separate two-channel audio files. One channel (C1) contained a recording of the experiment as it was actually performed by the participants, and the other channel (C2) contained a recording of the audio input that the participants received, including the manipulated feedback and the added noise. Analysis of the audio files was carried out in Audacity and Sound Forge by listening first to C1 in order to determine that participants did not make any mistakes of their own during the manipulated trials, which in itself would motivate them to vocalize “wrong!” after such a trial, and secondly, to count the number of mistakes made during the whole test. Then, C1 and C2 were analyzed jointly in order to measure if the manipulated feedback matched the actual vocal performances made by the participant. Two independent analyses were performed by members of the research team, and in the few occurrences where opinions differed, these differences were settled through discussion. Comparisons were made first based on word onset. The manipulated feedback was allowed to have its onset no more than 100ms (+/-20ms) after onset of the actually spoken word. If onset fell within this range, subsequent comparisons were made based on the strength, length and pitch of the relevant words. The average difference in onset was 46.4 ms. Due to an unidentified problem with the experiment software, feedback was sometimes turned off after a manipulated trial. These trials, and all subsequent trials for the specific participants for whom this happened were removed from further analysis. All in all, 20 participants were removed due to intolerable discrepancies in onset of the manipulations,

70

Chapter Five: Can we Tell What we Said When we Hear Ourselves Saying Something Else?

and/or problems with disappearing feedback. Among the remaining 26 participants, another 14 manipulated trials were removed from analysis due to either too large differences in onset, or due to the problem of the disappearing feedback. For the remaining 64 manipulated trials, all trials followed by a response that indicated that the manipulated feedback played a role in the participants’ understanding of the meaning of their utterance were noted. Such responses included the instructed pronouncements of “wrong!”, and also clear repetitions of the actually uttered word (which can be interpreted as a correction when compared to the manipulated feedback) and exclamations of “no!”. Finally, all trials where participants signalled a detection showing they understood the manipulated feedback was externally produced were labelled as concurrent detection. Next, the post-experimental interviews of the remaining participants were analyzed. Here, each participant was tagged for a number of characteristics: they were assigned the label retrospectively claimed detection if they either immediately after the experiment or when asked what they thought of the experiment said that they had noticed that the feedback had been manipulated. Based on their answer to the question of how they had experienced the manipulations they were tagged as either confusion or certain-that-manipulation-was-external. It was also noted if they claimed to have become more suspicious after the first manipulated trial.

Results The results are presented in two sections. The first section deals with the direct responses elicited during the test. The second section deals with the experiential aspects of the test and, specifically, of the manipulations which we attempted to probe in the post test interviews. Table 1 provides a summary of the results.

The Stroop test A total of 11 manipulated trials by 6 different participants were followed by either “wrong!”, “no!” or a repetition. No actual mistakes were made during these trials. No participants signalled that they had detected the feedback having been manipulated during the experiment. Instead, it was later in the post-experiment interviews that participants claimed that they had noticed something being wrong. Across all participants, out of 1222

Andreas Lind, Lars Hall, Petter Johansson and Sverker Sikström

71

non-manipulated trials, only 34 proper mistakes were made. This corresponds to an error-rate of 2.78%.

Post-experiment interviews All but one participant claimed that they had detected that the feedback had been manipulated. This one participant also responded to all three manipulations with repetitions (2 times) and a “no!”. The remaining 25 participants suggested, when asked how many times they had detected that the feedback had been manipulated, that between 2 to 5 trials had been manipulated. The follow-up questions revealed that 18 participants had experienced a clear feeling of confusion as to the actual source of the utterance during one or more of the manipulated trials. This confusion was present on a total of 32 manipulated trials and was described as an initial bewilderment followed by a “questioning” to the effect of “did I say that!?”, after which the participants “decided” they had not produced the manipulated feedback. There was some diversity regarding the number of trials on which the participants claimed that this confusion had been present; ten participants said they had experienced it on one trial, two participants said they had experienced it on two trials and six participants said they had experienced it on all three trials. 14 participants said they had become suspicious about the feedback after the first manipulated trial. 11 participants, on a total of 22 manipulated trials, claimed to have been absolutely certain that the feedback had been manipulated (these participants make up the category certain). They claimed to have ascribed the manipulations to either the computer or the experimenter. 4 of these participants also fall within the category of confusion. This is because they felt confusion on the first manipulated trial but then became more suspicious and felt certain that the two remaining manipulations were made by the apparatus or experimenter. Table 1. Summary of results; M refers to manipulated trials, % in parenthesis The Stroop test Exclamation or repetition Concurrent detection Post-test Retrospective detection interviews Confusion Certain detection Suspicion after first M Total

M trials 11 (17.2) 0 61 (95.3) 32 (50) 22 (34.4) 27 64

Participants 6 (23.1) 0 25 (96.2) 18 (69.2) 11 (42.3) 14 (53.8) 26

72

Chapter Five: Can we Tell What we Said When we Hear Ourselves Saying Something Else?

Discussion The most important aspects of the results are briefly recounted here. 6 participants, on a total of 11 trials, reacted to the manipulated feedback by saying either “wrong!”, “no!” or by repeating themselves. In the postexperiment interviews, 25 out of the 26 participants reported that they had detected that the feedback was manipulated on certain trials. At the same time, 18 of the 26 participants also reported that they had initially been very confused as to the source of the manipulated feedback. Two obvious and related questions that arise out of the results regards (i) the significance of the fact that certain participants behaved, on one or more occasions, towards the manipulated feedback as if it was they themselves who had produced it, and (ii) the significance of the state of confusion during one or more manipulated trials that was described by a majority of the participants. Even though all but one of the participants who acted towards the manipulated feedback as if it was self-produced later reported that they had detected that the feedback had been manipulated, it is important to stress that immediate reaction and retrospective reasoning must be considered separately. Our access to, and our ability to reason about, our own higher cognitive processes has repeatedly been shown to be fallible (Nisbett and Wilson 1977; Johansson et al. 2005; Johansson et al. 2006; Lind 2006). Using the concurrent measure we can see then how the manipulated feedback influenced participants understanding of what they had said on 17.2% of all manipulated trials. This is an interesting finding. It shows that we do rely, to some extent, on the feedback of our own voice when making inferences about what we have said. Furthermore, the observation of this influence invites speculation about its possible relation to the reported state of confusion, which was more widespread. Accepting the manipulated feedback as self-produced is not a matter of either-or. Rather, one ought to expect a continuum between minor confusion and a point where the feeling of confusion gains enough strength to provoke the participant to act upon the manipulated feedback as if it was selfproduced, only to later reinterpret the situation (and the meaning of their utterance) and decide that “something was wrong”. This interpretation is in line with Dennett’s (1991) pandemonium model of speech production. Another interpretation of the state of confusion could be that the participants had the firm speech intentions that Levelt’s (1989) model proposes, but the situation as such nevertheless provokes confusion. After all, it is very rare for us to say one thing but hear ourselves say another thing. So confusion may follow, and participants may then, more

Andreas Lind, Lars Hall, Petter Johansson and Sverker Sikström

73

or less consciously, start eliminating possible explanations to what just happened. After a while, and perhaps encouraged by the subsequent two manipulations, they come up with the plausible explanation that the feedback must have been manipulated. Here, the unexpectedly low number of self produced mistakes in the Stroop test may have contributed to delineating the manipulated trials as peculiar simply because they created “mistakes” that were otherwise very uncommon. Had the base mistake-frequency been higher, say between 15-25% instead of the observed 2.8%, this factor could have been ruled out. This explanation of the state of confusion, however, can not explain why it was observed that participants sometimes acted upon the manipulated feedback as if it was self-produced. Why would we act against a speech-intention that has set the whole speech-production system in motion and is the sole standard to which the output is supposed to be judged?

Conclusions and future studies The present study represents an attempt to empirically investigate aspects of speech production that previously have been dealt with exclusively in a theoretical manner. While the present results are interesting, it must be noted that they spring from a novel and primarily exploratory experimental design. In relation to the Levelt/Dennett dichotomy of serial versus pandemonium models of speech production, our results do not conclusively decide the issue. Sometimes the participants are confused of the origin of their utterance, which supports a pandemonium model. On the other hand, after three manipulated trials, almost all participants have some level of awareness that the feedback they received was sometimes manipulated, which supports a model with a higher degree of selfmonitoring and pre-defined speech intentions. In this first study of real-time voice feedback exchange we chose to use the Stroop test because it afforded us the opportunity to control what participants would say, and when they would say it. However, while technically manageable, this setup is less than ideal for providing the best contrast between models like those of Levelt (1989) and Dennett (1991). Whatever theoretical outlook one has, it is clear that we do not always pay the same general amount of attention to monitoring our output. Rather, our level of monitoring fluctuates depending on our needs, and the ideal experiment would capture a situation that neither exaggerates nor diminishes those tendencies (i.e. something that would be representative of language production in everyday life). For example, Pickering and Garrod (2004) argue that dialogue should be considered the

74

Chapter Five: Can we Tell What we Said When we Hear Ourselves Saying Something Else?

basic form of language use. Comparing this to the speed with which participants were speaking while performing the Stroop test, it is clear that in dialogue the processes of production usually operates at a much faster pace than it did in our experiment. Obviously, listening to yourself speaking through earphones also deviates from a normal conversational situation (except perhaps a voiceover-ip phone conversation). While the earphones are difficult to avoid in an experiment like this, for future studies it is important to strike up a balance between too little and too much monitoring. For purposes of measurement participants will have to monitor their speech enough to make some kind of response to the manipulated feedback. At the same time they should not monitor their speech to an unnaturally high degree, as this would test the research question only under the specific circumstances that such a high degree of self-monitoring would entail. As an example of this, in everyday speech, “instructions” for what to say are only very rarely displayed on a computer screen the way it was in our experiment. This entails that, in the present test, an objective standard of correctness existed, which means that participants could measure their behaviour not necessarily with respect to some inner model of conceptualization (as they are supposed to be doing in most models of speech production), but rather by simply comparing their speech performance to the visual memory of what was displayed on the screen. In future studies we aim to advance our experimental software platform to allow for feedback manipulations that more fully capture the dynamic of everyday speech, and to more forcefully open up the murky black box of the conceptualizer.

References Carroll, D.W. 1999. Psychology of language, Third Edition. Pacific Grove, Ca.: Brooks/Cole Publishing Company. Dell, G.S. 1986. A spreading-activation theory of retrieval in sentence production. Psychological review 93(3): 283-321. Dennett, D.C. 1987. The intentional stance. Cambridge: MIT Press. —. 1991. Consciousness explained. Boston: Little, Brown and Company. Fromkin, V.A. 1971. The non-anomalous nature of anomalous utterances. Language 47(1): 27-52. Griffin, Z.M. 2004. Why look? Reasons for eye movements related to language production. In J. M. Henderson and F. Ferreira (Eds.), The interface of language, vision, and action: Eye movements and the visual world, 213-248. New York: Psychology Press.

Andreas Lind, Lars Hall, Petter Johansson and Sverker Sikström

75

Harley, T. 2001. The psychology of language, Second Edition. New York: Psychology Press. Johansson, P., Hall, L., Sikström, S. and Olsson. A. 2005. Failure to detect mismatches between intention and outcome in a simple decision task. Science 310: 116-119. Johansson, P., Hall, L., Sikström, S., Tärning, B. and Lind. A. 2006. How something can be said about telling more than we can know: On Choice blindness and introspection. Consciousness and Cognition 15: 673-692. Levelt, W.J.M. 1989. Speaking – From intention to articulation. Boston: MIT Press. Levelt, W.J.M., Roelofs, A. and Meyer, A.S. 1999. A theory of lexical access in speech production. Behavioral and Brain Sciences 22: 1-75. Lind, A. 2006. On knowing and motivating one’s choices – Markers of uncertainty and cognitive load in manipulated choice-reports. Unpublished BA Thesis, Centre for Language and Literature, Lund University. Linnman, C., Carlbring, P., Åhman, Å., Andersson, H. and Andersson, G. 2006. The Stroop effect on the internet. Computers in Human Behavior 22: 448-455 MacLeod, C.M. 1991. Half a century of research on the Stroop effect: An integrative review. Psychological Bulletin 109(2): 163-203. Meyer, A.S. 1992. Investigation of phonological encoding through speech error analyses: Achievements, limitations, and alternatives. Cognition 42: 181-211. Neill, W.T. 1977. Inhibitory and facilitatory processes in selective attention. Journal of Experimental Psychology: Human Perception and Performance 3: 444-450. Nisbett, R.E., and DeCamp Wilson, T. 1977. Telling more than we can know: Verbal reports on mental processes. Psychological Review 84(3): 231-259. Pickering, M.J. and Garrod, S. 2004. Toward a mechanistic psychology of dialogue. Behavioural and Brain Sciences 27: 169-226. Postma, A. 2000. Detection of errors during speech production: A review of speech monitoring models. Cognition 77: 97-131. Stroop, J.R. 1935. Studies of interference in serial verbal reactions. Journal of Experimental Psychology 18: 643-662.

TESTS OF TRUE PICTORIAL COMPETENCE IN CHIMPANZEES: THE CASE OF DRAWINGS TOMAS PERSSON

Abstract. For us human beings, interpreting pictures is as much an act of imagination as one of recognition. Pictures are therefore a promising avenue for studying mental worlds. Can this be extended to the study of animal minds? True pictorial competence is here defined as a referential and differentiated apprehension of pictures, where reference is crucially dependent on visual resemblance. It is proposed that pictures can be appreciated in at least two other modes than the pictorial one, surface and reality modes, which have to be precluded in experimental investigations of true pictorial competence. One way to avoid the reality mode is to use non-photographic stimuli. The few experiments on chimpanzees (Pan troglodytes) that have taken this approach are critically reviewed in this chapter.

Introduction In science we have tried to understand ourselves by looking at the minds of nonhuman primates for well over a century. Today comparative cognitive studies from an evolutionary perspective usually take one of two forms. The first one can be called “cognitive cladistics”. Cladistics is the reconstruction of ancestral relations by grouping e.g. species according to shared features, which is contrasted to groups that do not share the same features. In this way ancestral trees can be built, where unique features for a given group are said to be “derived”, while features shared with ancestors are “primitive”. In order to find out whether a certain cognitive profile is derived or primitive, cognitive cladistics strives to find out which species can, and which cannot, perform a certain task in a certain way. However, the focus on contrasting averaged groups restricts this approach to being descriptive on a most general level and these studies seldom target cognitive processes.

Tomas Persson

77

Another way of doing comparative cognitive research is therefore not to focus on the typical, but on the potential of the ape mind. This can be studied on the level of individuals rather than species. A focus on typicality would for example not answer the question whether nonhumans can learn a human language in the positive. Typicality is interesting for many reasons, especially in combination with in-depth study of individual variation, but a single language-using chimpanzee can teach us more about language and language evolution than a hundred “typical” chimpanzees. A language-using chimpanzee can tell us that language is possible for a nonhuman and it provides us with necessary resources for understanding why this is so. Furthermore, it can help us guess whether these factors were active also in human prehistory. For the above reasons comparative cognitive research on something as “artificial” and “human” as pictorial competence is not misdirected. Apes are in fact prime subjects because there seem to be individuals that can, and individuals that cannot, decode pictures as the typical human does (see Persson 2008). In this chapter a framework for studying pictorial competence in nonhumans will be presented together with a critical review of central studies in the field.

Pictures and animals Pictures have been used as stimuli since the early days of animal cognition research, both as substitute for real-life objects, individuals and events, and as abstract stimuli. Pictures have been used to assess questions of spatial representation, social cognition, viewpoint consistency, and serial list learning (Fagot, Martin-Malivel and Dépy 2000), as well as in all sorts of discrimination and categorization tasks (Bovet and Vauclair 2000). Pictorial stimuli seem to have been used with such success that few scientists have asked the questions of why it works, and what it means. Since animals of all kinds readily accept pictures as examples of realworld objects, it must mean that pictures are simple and intuitive phenomena. This observation easily lends itself to the idea that there is only one way of viewing pictures: you either recognize them or you do not. This destiny was apparent already in the first reports on chimpanzee mentality: I made no further tests, as I consider it quite obvious that results are determined simply by the technical accuracy of the photographs and the difference of the objects they represent. Anyone who may take the trouble to experiment on other chimpanzees in the same way, will be able to demonstrate effectively and exactly, by means of larger and clearer

78

Chapter Six: Tests of True Pictorial Competence in Chimpanzees reproductions, that the animals recognize and differentiate between such photographs (Köhler 1957: 278).

The above quote is the closing paragraph of Köhler’s classic The Mentalities of Apes, published in 1925. Köhler was a Gestalt psychologist who conducted cognitive experiments and observations on chimpanzees in the 1910s, and he is considered one of the founding fathers of cognitive primatology. He chose to look at chimpanzees’ performance with pictures after having observed their reactions to stuffed toy animals, cardboard face masks, and mirrors. He observed that it seemed necessary for the toy animals to have some likeness to real animals, “nearness to life,” in order to invoke a response, i.e. fear. The stuffed animals invoked even stronger responses than did most real animals. Köhler concluded that the stuffed animals, not being fully real, played on the imagination in a way that real animals did not. Uncertainty as opposed to experience was the key factor. Köhler wondered whether the chimpanzees’ ability to recognize nearness to life in stuffed animals and mirrors would remain if the third dimension and colours were removed. He turned to black-and-white photographs. In the initial tests the chimpanzees were much interested in the pictures of themselves and other chimpanzees, but only one showed suggestive signs that he recognized their content: he made social responses towards it. Köhler then made a photograph of an empty crate and another photograph of a crate full of bananas and pasted these on two boxes baited with fruit. The star pupil of the previous test chose the box with the banana picture on 10 successive trials. By now Köhler wanted to control for rote learning and created two new photographs: one of bananas and one of a rock. The subject performed better with the new pictures than with the old ones. Köhler ascribed this to the superior quality and nearness to life in the second pair of pictures. When testing a second chimpanzee on the old pictures, Köhler could not establish a permanent good performance. A third subject tested with the old pictures performed much better, but lost the ability when exposed to the slightest distraction. When confronted with the second pair of photographs, the rock and the bananas, she made hardly any mistakes. Köhler ascribed these mixed results solely to the quality of the pictures. “I made no further tests […]” he reports (Köhler 1957: 278). Köhler’s brief investigation of picture perception in chimpanzees gives the impression that he assumed that the correspondence between a picture and the real world could only be of one kind. His conclusion was that the higher the similarity between picture and referent, the easier it is for a chimpanzee to solve an object choice task guided by picture information.

Tomas Persson

79

But why do we not see the same correlation in adult human pictorial competence? A distorted photograph is not necessarily more difficult to recognize than one that depicts reality more truthfully. We readily decode non-photographs that are far removed from the real perceptual world, like cartoons. We can even intend to see likeness where there is not supposed to be any, like faces in the clouds. Are Köhler’s chimpanzees and humans different points on one single competence continuum, with differences merely being the result of different experience of picture qualities? It will be argued below that it is not.

Three ways of looking at pictures Following previous work (Persson 2008), I propose that there are at least three ways, or modes, in which pictures can be related to the world. (1)

The first mode involves bypassing any estimation of what the picture might actually depict. What is perceived is rather the patterns, shapes, or colours on the surface of the picture, and it stays at that. Here this will be termed a surface mode of picture processing. Besides perceiving local elements, seeing motifs in the sense of global forms is in theory possible, but they have only a learned connection to the real world, if any. Through association, i.e. rote learning, of specific picture-object relations, or through generalization based on invariant features, one can judge correspondences while circumventing recognition on a categorical level. That is, one can e.g. sort pictures of apples on a level (e.g. colour) that does not involve realizing that it is apples that one sorts.

(2)

Pictures can also be meaningful due to their likeness to the real world, but without being sufficiently differentiated from this, leading to the perception of pictures as part of reality and not about reality. Although the photographic image is the typical example, the effect is not necessarily limited to stimuli that seem realistic from a human perspective. Critical features can likewise elicit reality-guided responses. In reality-based picture processing an object is not seen as being anywhere else but in the picture, albeit perhaps in a stranger than usual form. In this mode one can solve tasks that depend on categorization, but it is not qualitatively different from categorizing real instances of the depicted objects. If an object in a picture is e.g. matched with a

80

Chapter Six: Tests of True Pictorial Competence in Chimpanzees

similar object outside of the picture, it is an object relating to an object, not a picture of an object relating to an object. (3)

In the third mode pictures become meaningful through likeness (iconicity) to the real world (and other pictures), but are still sufficiently differentiated from it in order not to be confused with it. Reference rests on the specification of what the picture is similar to, and why it is not itself this thing. Such a stand-for relation implies two types of expectations. In pictures where the likeness is directly perceived in virtue of mirroring the real-life experiences of the perceiver, an expectation of separation between picture and reality is crucial. When it comes to pictures that require more of an interpretive stance, an expectation of likeness enhances actual likeness. Many pictures would not be perceived at all without such expectations. This type of picture use is distinguished from other referential instances by being called seeing pictures as pictures.

This three-part division is similar to that of Fagot et al. (2000) who propose independence, confusion, and equivalence modes of picture processing in a review of animal picture experiments. This is not surprising since “[t]he proposed classification into three modes of processing is intuitively obvious” (Fagot et al.: 297). Indeed, Premack (1976) makes a similar distinction between forms of picture competences in animals, as does Cabe (1980) indirectly. It also seems to follow naturally from the theory of (pictorial) semiotics of Sonesson (e.g. 1989, 2007). The modes can be seen as standing in a hierarchical relation to each other, but they should not be equated with a developmental trajectory. The modes are not general competences but depend on interactions with specific pictures in specific contexts. Switching between modes is not best described as reverting to a previous stage in development, but to a different way of approaching a certain visual display. Only the third of the three modes entails understanding pictures as pictures, i.e. a referential understanding of pictures. This type of processing is interesting for many reasons. For example, only when pictures are viewed in a referential way does an interpreter have a reason to assume that the picture is informative about things that are not in the picture itself, such as movement, where an object is hidden in the real world, and that the orange thing in Donald Duck’s face must be a beak. “Filling in” the content of a pictorial display is something akin to

Tomas Persson

81

imagination. One reason to expect such competence in nonhumans at all is data on other iconic competences that has received more attention, such as the understanding of model replicas (e.g. Premack and Premack 1983; Kuhlmeier, Boysen and Mukobi 1999) and mirrors (e.g. Gallup 1970; Menzel, Savage-Rumbaugh and Lawson 1985). The three modes of pictorial competence are applicable to all iconic media. When applying the above definitions on Köhler’s picture experiments it is clear that what he observed was a reality mode competence. The social responses and the performance with pictures judged high versus low in realism points to that. The chimpanzees saw the photographs as objects in themselves. Had they seen the photographs as views of objects they would not so easily have been confused by a slightly distorted photograph. Not making the distinction between different forms of competence with pictures is a mistake, especially when making claims that pertain to the pictorial mode. Many studies have indeed involved pictures (Bovet and Vauclair 2000; Cabe 1980; Fagot et al. 2000), especially photographs, and several studies give the impression of referential use of pictures. But on close scrutiny, processing of pictures in a non-differentiated manner, i.e. in reality mode, or in surface modes, can often not be excluded (Persson 2008). If we are interested in true pictorial competence we must look for experiments designed specifically for testing pictorial mode competence at the expense of reality and surface modes. Drawings, for example, can be used in experiments that normally use photographs, but can be designed to exclude reality-based responses. In addition, drawings often require an a priori expectation of a motif for recognition to occur. Such an expectation comes with the pictorial mode. In the remainder of this chapter I will review all studies up to now that have involved recognition of drawings by chimpanzees.

Recognising drawings When picture-naïve humans describe pictures, slow and stepwise recognition sometimes takes place. According to Deregowski (1976) such recognition can go something like: “that is a tail, this is a foot, that is a leg joint, those are horns… it is a waterbuck”. A parsimonious explanation for this process is one described by Gregory (1973, in Deregowski 1976) whereby the perception of a picture occurs in a series of “hypotheses.” (In experienced viewers the process is often too quick for introspection.) A set of properties in the picture is the basis for a hypothesis which is then verified against further properties. If necessary, the hypothesis might be modified and retested against the features until a stable identification has

82

Chapter Six: Tests of True Pictorial Competence in Chimpanzees

settled. Parts and wholes thereby define each other with continuous feedback. Some pictures might give only a small “whole” to start off the process, such as an eye or some other feature with high saliency from everyday life. Other times a larger but ill-defined whole can catch the eye. For example, one of the earlier recognitions that started the process of identifying the waterbuck might not have been a local feature, like its tail, but that it was some sort of animal. Only after this recognition, or perceptual hypothesis, could recognition of a tail, feet, and horns occur. This in turn, in their new configuration as a whole, led to the recognition of the animal as a waterbuck. Successive approximation is the reason that we can perceive novel pictures that have very little in common with real-life experiences of the world. The combination of features makes individual features meaningful, and these in turn influence the perception of the whole. Without this constructive process pictures that are not possible to interpret in a reality mode would fall flat. The reason that we apply successive approximation to pictures in the first place is that “pictures are not unique in being ambiguous and incomplete” (Hochberg 1980: 59). It is true also for objects in the world at large. At each momentary glance only parts of objects are informative to our brains. Identifying an object is a question of using attention electively to complete the picture. This means that eye and head movements are not random, but are dependent on the viewer’s “perceptual purpose” (Hochberg 1980). If we expect to see an array of lines and colours we will not see a waterbuck. What one needs in order to identify the content of non-realistic pictures is thus the intent to identify objects in a picture, and successive approximation will do the job. Let us now turn to those experiments that have specifically addressed recognition of drawings in chimpanzees.

Viki The chimpanzee Viki was raised in the home of psychologist Keith Hayes and Catherine Hayes with the purpose of studying an infant chimpanzee growing up in a human social and material environment. Since Viki was raised more or less as a human child she had early experience of pictures, but was never specifically trained to perceive them. She enjoyed picture books from the age of 6 months, but not until the age of 1.5 years did she spontaneously respond to pictures differentially (Hayes 1951). When Viki was 4 years old she readily pointed to pictures of beverages and led the addressed person to the refrigerator. This indicates that she

Tomas Persson

83

could relate pictures at least in some ways to the real world without confusing the two. But we need to know more about the context and generalisability of this behaviour before we can say that Viki knows that picture motifs are picture motifs and objects are objects. Viki could have discovered that she could barter those flat and flimsy special cases of drinks for more drinkable versions. If her performance was limited to certain categories, i.e. beverages, there is further reason to suspect that she had learned specific connections between pictures and their objects, rather than discovering the general nature of pictures and their communicative potential. It is possible that she used reality mode processing of pictures in a communicative context. This might in addition be limited to bartering situations. When 3.5 years old, Viki’s ability to imitate bodily actions from pictures was tested. Hayes and Hayes (1953) report that she did fairly well with stimuli ranging from movies, via black-and-white photographs, to “simple line drawings.” Her successful interpretation of line drawings speaks for a pictorial competence, but there is no information on the novelty of the pictures and rote learning can therefore not be ruled out in the present analysis. Besides the performance with line drawings it is noteworthy that imitation of dynamic actions depicted in static pictures requires abductions on behalf of the interpreter (Sonesson 1989). One must infer what happened just before the static view, and what will happen just after it, in order to read “clapping” and “patting” into the relations of body parts in a picture. This might not be possible when viewing a picture in reality mode. Unfortunately, without a detailed report on her responses we cannot know if Viki read clapping or patting into the pictures, or just “hands together” and “hand on head.” Viki, aged 4, did for example not learn how to solve problems when the solution was presented in pictures, but she did learn when human models demonstrated the solution in real life (Hayes and Nissen 1971). The ability to read dynamic content into pictures is a possible venue for exploring pictorial competence in future empirical work. At 5 years of age Viki was given a discrimination task using pictures of familiar objects like chairs, cars, dogs, flowers, etc. Cups in a tray were differentially baited with food, and pictures drawn from two categories of objects were placed on the cups. Viki was rewarded for choosing from the same category throughout a session. Importantly, a given picture was only used once, precluding rote learning. The sessions were divided into four successive groups of picture types. In the first sessions “naturalistic” colour pictures were used, and in the following a mix of realistic photographs and stylized drawings. The third group consisted of realistic

84

Chapter Six: Tests of True Pictorial Competence in Chimpanzees

colour as well as black-and-white photographs, and the last sessions involved only black-and-white line drawings. The pictures were a mix of magazine illustrations, photographs, and handmade line-drawings. Many of the drawings were freehand copies of photographs from earlier sessions, and can thus be criticized for lacking in novelty. Viki is reported to have been able to generalize discriminations to novel pictures of all the types mentioned above. Success rate on the four groups of problems were 85%, 75%, 82%, and 73% correct respectively. No data is given for the distribution within the second group (mixed), but the lower figures in this and the fourth group (line drawings) might be due to the abstractness of the stimuli. In the third group Viki was 95% correct on the colour photographs but only 68% on the black-and-white ones. Picture “realism” thus seems to have had an effect on Viki’s performance, but importantly she performed above chance on drawings. Note, however, that these stimuli were often traced from photographs. Furthermore, discrimination tasks are prone to responses based on invariant local features, bypassing recognition on a categorical level. Following the discrimination work Viki was given matching tasks. She was shown an object and was made to choose from two pictures the one that depicted an object similar to the one that was shown. This experiment was also divided into four conditions. In the first one Viki received trials with realistic photographs reused from the discrimination tasks. She was correct on 95% of these trials. In the second condition she received picture pairs from the former group, but now the matching picture had become the non-match and vice versa, and new sample objects were used. She was correct on 85% of these trials. The third group of trials utilized line drawings, some of which were reused from the discrimination tasks and were thus not novel. Her performance dropped somewhat to a still good 80%. The last group consisted of rearrangements of the line-drawing pairs from the previous group. Viki was 91% correct. This figure might not have been as high if novel line drawings had been used. Viki was also fond of drawing, but never seemed to make depicting pictures. However, she learned to connect multiple dots and twice she reacted to the shapes as if she recognized them. In response to the familiar words “get me one of these,” she fetched a stuffed dog after having connected a “rough approximation of a terrier,” and she spontaneously named a similarly drawn cup (one of the few words she could voice) and fetched one (Hayes and Nissen 1971). To other self-made drawings she was indifferent and, importantly, never tried to fetch an object through guesswork. The dog and the cup were thus not chance events. However, these observations lack control for contextual cueing. Self-made pictures

Tomas Persson

85

make performance through reality mode unlikely, but the numerous occasions where Viki did not fetch objects in response to the dot connecting exercise amounts to arguments against her understanding the depicting potential of self-made drawings.

Cross-modal matching By the 1970s the question regarding recognition of pictures by apes still seemed unresolved. Davenport and Rogers (1971) note with reference to the studies of Köhler (1957) and Hayes and Hayes (1953) that they “are unaware of any study which unequivocally demonstrates the ability of chimpanzees or any other organisms to perceive the representational character of photographs without specific training,” and Winner and Ettlinger (1979) criticize both studies for lacking controls for associative learning. In cross-modal matching e.g. a visual sample has to be matched to haptic comparison stimuli, or a haptic sample is to be matched to visual comparisons. The subject typically puts its hand behind an occluder and feels an object that cannot be seen. It then has to match what it feels to visually accessible objects or pictures of objects. The matching can be simultaneous or delayed. Davenport and Rogers (1971) report a cross-modal experiment with life-sized colour and black-and-white photographs of unfamiliar objects as visual stimuli. The subjects performed above chance and there was no difference between the two photograph conditions. In this study the subjects were naïve to pictures, and since the pictures were highly realistic and placed behind glass, a reality mode picture processing is the given candidate for the apes’ performance. Placement behind glass potentially reduces the two-dimensionality of the pictures and retains an illusion suitable for reality mode. The glass also served as response surface. A study by Winner and Ettlinger (1979) failed to replicate these findings, and this may be because their pictures were not instrumentally connected to either locus of response or to the food rewards (cf. Persson 2008), but perhaps also because the subjects could experience the two-dimensionality of the pictures by being allowed to investigate them directly. Davenport, Rogers and Russell (1975) introduced non-photographs in their paradigm. Five classes of pictures of unfamiliar objects were tested in a simultaneous cross-modal matching condition, given in the order full sized colour photographs, full sized black-and-white photographs, halfsized black-and-white photographs, full-sized silhouette pictures and, lastly, full-sized line drawings. In addition, colour photographs were used

86

Chapter Six: Tests of True Pictorial Competence in Chimpanzees

in a delayed version of the task. The silhouette pictures were created by increasing the contrast in black-and-white photographs until only the black mass of the depicted object (against a white background) was discernable. In order to control for learning, one-trial problems, completely novel stimuli, or novel combinations of stimuli were given for each picture condition. However, Davenport et al. (1975) seem to have neglected to control for order effects. At the time the subjects received the line drawing trials they had already had extensive experience with the previous types of pictures. The line-drawings might have been perceived as special cases of silhouette drawings, which in turn can be argued to be recognizable as shapes in a reality mode. For the simultaneous cross-modal matching problems four of the five chimpanzees performed above chance in the full-sized colour and blackand-white photograph conditions. All five were significantly above chance on the half-sized black-and-white photographs. Three of five were correct on the silhouette pictures, and four out of five passed the line drawing condition. In the delayed matching condition with colour photographs, four out of five subjects performed well, but in the critical tests with novel stimuli only two performed above chance. The fact that the testing was made in the context of cross-modal matching might very well have been a significant factor for the picture performance shown. Shape matching rather than identity matching might have taken place through comparison of the remembered sample shapes to the pictured shapes, which probably remained intact in all picture conditions. If the animals had hit upon this strategy in the silhouette condition transfer to line drawings might have been a simple task. The fact that most sample objects in the study were unknown to the subjects strengthens the advantage of matching based on shape similarity rather than object identity. That the pictures were behind glass further enhanced their aspects as shapes as opposed to pigment on two-dimensional surfaces. The poor performance with novel stimuli in the delayed condition suggests that the subjects could not reliably identify the objects and therefore had problems remembering them. If, instead, the task had been to match identifiable objects, memory demands might have been smaller. At the same time pictorial performance might have declined since the pictures would have to be interpreted with their identity taken into account. Line drawings might then fall short. The experiments by Davenport and colleagues therefore most likely involved combinations of picture processing in the surface and reality modes.

Tomas Persson

87

The Kyoto Studies Itakura (1994) tested black and white drawings on the female chimpanzee Ai at the Primate Research Institute in Kyoto, Japan. Ai was 12 years old at the time and versed in the use of the abstract visual symbols called lexigrams. Itakura (1994) does not report what previous experience Ai might have had with non-photographic pictures, but at least she had extensive training on discriminating figure from ground in drawing-like stimuli since e.g. her lexigrams are all black-and-white patterns (Matsuzawa 2003). Testing stimuli included pictures of three humans, three chimpanzees, and one orangutan that Ai could recognize in movies and photographs, and name with letters from the alphabet. Colour photographs and black-andwhite line drawings were used as representations of the individuals and were to be named with their respective letter. (Two extra distracter letters were included.) All line drawings had been constructed from photographic templates and were quite detailed. The presentation and apparatus used was computerized. In each trial, one of the seven individuals was the target and after the sample had been presented there was a two second delay until the letters came onscreen. A food item was dispensed each time Ai made a correct choice among the nine letters. It seems from the published descriptions that the matching was simultaneous, i.e. the sample picture remained in view during the response. If an incorrect choice was made, a signal sounded and the trial restarted after a brief timeout. Ai was allowed as many correction trials as she needed. Twenty-one pictures, three for each of the seven individuals, were used for seven sessions. Each session consisted of 84 trials. This means that each picture was used four times per session. One new drawing replaced a photograph in each session, starting with one individual in session one, and another one added in session two. Thus, in Session 7 all seven stimuli-individuals were represented by two photographs and one drawing. Photographs were changed for each session but the report does not say whether the drawings were also changed or remained the same throughout testing. That would have meant that in the 7th session the first individual’s drawing had been presented on four times seven trials, the second individual’s drawing four times six trials and so on, meaning that chances for rote learning were high for most drawings. Ai was correct on 100% of the photographs but not above chance on the first trial on any of the line drawings. She tended to choose the letter that had been reinforced in the previous trial, and responded randomly during correction trials. However, some of the drawings only needed to be

88

Chapter Six: Tests of True Pictorial Competence in Chimpanzees

rewarded once for Ai to perform correctly on the next trial that involved that picture, meaning that her memory for correct pairings was good. In a second phase of the study each stimulus-individual was represented by two photographs and one line drawing from the onset. Again 84 presentations (not counting correction trials) were given per session and novel line drawings and photographs were continuously introduced. But still, each picture was used as target four times within a session. Ai was much better in this phase and responded accurately to line drawings at first presentation for four of the seven individuals at a level of 60% correct, which is significantly above chance. As in phase one, her success rate improved during correction trials, suggesting learning rather than recognition. The four individuals that she seemed to be able to pinpoint were the three humans and the one orangutan. The three chimpanzees could thus not be identified in line drawings by Ai. (Nor could they by a human control group.) That the orangutan sticks out as stimulus is not surprising. It is not implausible that the orangutan drawings show some invariant features that the other drawings lack. However, such an analysis would have to be made from the complete drawing material, which is not included in the published data. The individual humans could possibly also be discriminated on surface features since, for example, only one of the two males wore glasses. The third person also wears glasses, but is a woman. Combining the glasses feature with e.g. the presence of much hair on the head, will suffice for mutually exclusive identification of the humans. It is in theory possible to appreciate these distinctive features and combinations even without attributing “glasses” and “hair” to them, and thus solve the problem completely in a surface mode. (This is also an alternative explanation that has to be checked against the full stimulus set.) Pattern recognition rather than categorical identification is therefore a possible explanation for Ai’s success rate, and might also account for her uneven performance. When looking at Ai’s errors one finds that it was more common for her to confuse individuals within the chimpanzee category than between species. The orangutan in turn was named as a human several times. The humans, however, were only named as other humans. Does this mean that chimpanzees look like chimpanzees and humans look like humans in these drawings, and that they are just a bit difficult to tell apart as individuals? Or does it rather mean that the visual patterns are recognizable as classes, i.e. “this is one of the human-patterns”? This latter alternative might explain the hardships of keeping individual patterns apart while still succeeding in keeping the species apart. The invariant features between

Tomas Persson

89

the humans as a group as opposed to the chimpanzees are easier to keep track of than the invariant features of the individuals, although that was apparently possible to a degree for four of them. This can all be done without actually seeing “humanness” or “chimpanzeeness” in the pictures themselves. Given Ai’s extensive experience with pictorial stimuli in the past in the form of photographs, and her ability in the present task to accurately “name” photographs, we must consider the possibility that she might have been very aware of the depicting function also of the drawings, but could not decode the transformations in most of them. However, we cannot be sure of what the names actually encode for Ai, especially after having subjected her to stimuli that unreliably match to the names, i.e. the drawings. One way to ease her task would be to request matching between drawings and photographs, doing away with naming altogether. The matching could also be made the other way around, so that she is given the sample name or photograph, and then require her to pick the matching drawing from an array of pictures. In this experimental format she could compare pictures with each other and pinpoint the drawing that is most like the referent. In this way there is no need to invoke an absolute resemblance, only a relative one. Although this would be a case of scaffolded recognition, it is still an iconic one, as well as referential and differentiated. For now we must unfortunately conclude that Ai’s performance leaves us with inconclusive evidence. A study that directly targets the questions raised in this chapter was recently published. Tanaka (2007) tested Ai, three other adult females, and three 4 to 5 year old juvenile chimpanzees at the Kyoto Primate Research Institute on a computerized generalization task involving photographs and drawings of various degree of realism. The subjects were required to choose three pictures in an array of 12. Correction trials were allowed and the subjects were food reinforced for every correct indication of a flower picture. Training went effortlessly and generalization to novel colour photographs was above chance for all but one subject. However, the subjects sometimes chose non-flower pictures that contained colourful patches. Whether they mistook these for flowers or did not choose in accordance with a flower concept is unknown. Next step was generalization to realistic colour sketches and less realistic colour cartoons, or computer clip-art. The criteria for “realism” are not given so we have to assume that it was the subjective judgment of the experimenter. 480 novel pictures of each type were tested, of which 120 in each group were flowers. (Non-targets were trees, leaves, branches, grasses, ground surfaces, and various everyday objects.) Ai was the only

90

Chapter Six: Tests of True Pictorial Competence in Chimpanzees

adult subject who could readily choose correctly above chance in both categories at first exposure. Chloe, another adult, managed for both categories as well, but sketches outdid cartoons/clip-art. The third adult performed equally low for both categories but still above chance. The fourth adult subject performed at chance level for all non-photographic stimuli. Only one subject, Chloe, showed a clear decline with decrease in “realism.” In contrast to the variable performance of the adults all three juveniles performed above chance from the very first trial on both types of pictures. The juveniles, together with Ai, also displayed strong learning effects with repeated exposures. To further abstract the pictures colour was removed from 48 novel line-drawn pictures, of which 12 were flowers. Now only Ai performed above chance in the adult group, but not until the third presentation. Two of the three juveniles were above chance from trial one on the black-andwhite line-drawings, and the third juvenile on second exposure. It is difficult to say whether Ai suddenly grasped, by the third exposure, that also the line-drawings were flowers, or whether she had formed a new category perhaps not remotely connected to flowers. She could also just have memorized the correct pictures, which were only 12. However, when tested with novel Kanji (Chinese) characters instead of novel pictures she did not learn the correct choices with repeated exposure. In fact none of the subjects learned the rewarded Kanji character for flower. Tanaka (2007) would be a more convincing case if possible generalization from photographs to drawings had been precluded, perhaps by testing the abstracted pictures before photographs. If transfer would occur in this direction the claim for recognition of drawings as flowers would be stronger. Also, repeated response to flower-like stimuli, regardless of level of abstraction, will not do as an indication of pictorial competence. Flowers are unfortunately a type of stimuli that take highly canonical forms in most cartoons and clipart. An analysis of those particular flower pictures that proved problematic for the subjects would be required in order to see if this was indeed the case. In order to make a study like this more convincing a requirement would be a conditional discrimination, as in a matching-to-sample paradigm, where flowers are sometimes correct and sometimes wrong, and where several categories of drawn objects are tested at the same time. Of course, critical trials with novel pictures are a necessity, and correct first trial performance needs to be demonstrated.

Tomas Persson

91

Conclusions This chapter has explored the experimental work that has so far been conducted with chimpanzees and drawings. It has been argued that drawings, in virtue of being less realistic than e.g. photographs, are a promising avenue for studying true pictorial competence in animals. True pictorial competence has been defined as a referential and differentiated apprehension of pictures, where reference is crucially dependent on visual resemblance. A necessity for demonstrating true pictorial competence is to control for associative learning of picture-object relations, as well as performance in “surface” and “reality” modes. The studies reviewed do not convince in this regard, unfortunately. Studying the interpretation of pictures when it takes place in a pictorial mode, in turn, supplies a tool for studying categorization and imagination in non-linguistic nonhumans. This has the potential not only to help us understand the minds of other species, but also to cast light on our own cognitive prehistory. Although the picture is a fairly recent cultural creation the pictorial mind might not be, and might not even be restricted to human brains. Of further interest are the common denominators that foster pictorial potential in minds. It has been suggested that a true pictorial competence is part of a more general “symbolic function” or a “sign function” (cf. Sonesson 2007) that cuts across domains such as play, imagination, verbal and gestural communication. Again, turning towards nonhumans seems a worthwhile endeavour. Comparing the pictorial abilities of apes that have a background in language projects to that of apes with a similar upbringing but without language-training, could reveal possible connections between pictures and language (see Persson 2008). Perhaps language-competent apes can decode pictures that language-naïve apes cannot, or perhaps there is no difference. Contrary to Köhler (1957: 278) the suggestion would therefore be that “anyone who may take the trouble to experiment on other chimpanzees” has a lot left to investigate. Some steps in this direction were taken in my dissertation (Persson 2008), but a lot remains to be done. The framework of different modes of apprehending pictorial stimuli described in this chapter seems to be a necessary prerequisite for such research.

92

Chapter Six: Tests of True Pictorial Competence in Chimpanzees

References Cabe, P.A. 1980. Picture perception in nonhuman subjects. In M.A. Hagen (Ed.), The Perception of pictures, Vol. 2, 305-343. New York: Academic Press, Inc. Bovet, D. and J. Vauclair. 2000. Picture recognition in animals and humans. Behavioural Brain Research 109(2): 143-165. Davenport, R.K. and Rogers, C.M. 1971. Perception of photographs by apes. Behaviour 39(2): 318-320. Davenport, R.K., Rogers, C.M. and Russell, I.S. 1975. Cross-modal perception in apes: altered visual cues and delay. Neuropsychologia 13: 229-235. Deregowski, J.B. 1976. On seeing a picture for the first time. Leonardo 9: 19-23. Fagot, J., Martin-Malivel, J. and Dépy, D. 2000. What is the evidence for an equivalence between objects and pictures in birds and nonhuman primates? In J. Fagot (Ed.), Picture perception in animals, 295-320. Hove: Psychology Press Ltd. Gallup, G.G., Jr. 1970. Chimpanzees: self-recognition. Science 167: 8687. Gregory, R.L. 1973. The confounded eye. In R.L. Gregory and E.H. Gombrich (Eds.), Illusion in nature and art. London: Duckworth. Hayes, C. 1951. The ape in our house. New York: Harper and Brothers. Hayes, K.J. and C. Hayes. 1953. Picture perception in a home-raised chimpanzee. Journal of Comparative and Physiological Psychology 46: 470-474. Hayes, K.J and Nissen, C.H. 1971. Higher mental functions of a homeraised chimpanzee. In A.M. Schrier and F. Stollnitz (Eds.), Behavior of Nonhuman Primates, Vol. 4, 59-115. New York and London: Academic Press. Hochberg, J. 1980. Pictorial functions and perceptual structures. In M.A. Hagen (Ed.), The perception of pictures, Vol. 2, 47-93. New York: Academic Press, Inc. Itakura, S. 1994. Recognition of line-drawing representations by a chimpanzee (Pan troglodytes). The Journal of General Psychology 121(3): 189-197. Kuhlmeier, V.A., Boysen, S.T. and Mukobi, K.L. 1999. Scale-model comprehension by chimpanzees (Pan troglodytes). Journal of Comparative Psychology 113(4): 396-402. Köhler, W. 1957. The mentality of apes. London and Tonbridge: Penguin Books.

Tomas Persson

93

Matsuzawa, T. 2003. The Ai project: historical and ecological contexts. Animal Cognition 6: 199-211. Menzel, E.W. Jr., Savage-Rumbaugh, E.S. and Lawson, J. 1985. Chimpanzee (Pan troglodytes) spatial problem solving with the use of mirrors and televised equivalents of mirrors. Journal of Comparative Psychology 99(2): 211-217. Persson, T. 2008. Pictorial primates: A search for iconic abilities in great apes. Lund: Lund University Cognitive Studies, 136. Premack, D. 1976. Intelligence in Ape and Man. New Jersey: Lawrence Erlbaum Associates, Inc. Premack, A.J. and Premack, D. 1972. Teaching language to an ape. Scientific American 227(4): 92-99. Premack, D. and Premack, A.J. 1983. The mind of an ape. New York: W.W. Norton and Company, Inc. Sonesson, G. 1989. Pictorial concepts. Inquiries into the semiotic heritage and its relevance for the analysis of the visual world. Lund: Aris/Lund University Press. —. 2007. From the meaning of embodiment to the embodiment of meaning: A study in phenomenological semiotics. In T. Ziemke, J. Zlatev and R. Frank (Eds.) Body, language and mind. Vol 1: Embodiment, 242-283. Berlin: Mouton de Gruyter. Tanaka, M. 2007. Recognition of pictorial representations by chimpanzees (Pan troglodytes). Animal Cognition 10: 169-179. Winner, E. and Ettlinger, G. 1979. Do chimpanzees recognize photographs as representations of objects? Neuropsychologia 17: 413-420.

II. LEXICAL MEANING

THE CONSTRUAL OF ENTITY PERMANENCE JAN SVANLUND

Abstract. This chapter concerns how we individuate entities more abstract than material objects and how we judge different instances as being instances of the same entity. Linguistic construals of sports teams in Swedish are examined as an example of how collective human institutions can be treated as permanent entities. Just like prototype categories, such entities must be conceived as rather flexible, and it is argued that they have a dual nature concerning the type vs. instance distinction.

Object permanence If a boy takes a ball and puts it behind his back, you will know that it is still there, even if you cannot see it. And when he brings it out at the other side of his body, you will know that it is the same ball as before. The ability to use this understanding is usually referred to as object permanence. However, it clearly involves inferences, and the boy could have fooled you by having two balls from the start. Consequently your “knowledge” would just have been a misleading conclusion. Magic tricks are often based precisely on playing games with object permanence. But mostly the robustness of this ability serves us well when we are dealing with physical reality. Object permanence has been much studied in connection with children’s cognitive development and language acquisition. We know now that children acquire this ability very early, at about 3 months of age (Carey 2001). In many ways this is a fundamental cognitive capacity: ”Studies of object permanence are, in part, studies of criteria for numerical identity, for they involve the capacity to establish a representation of an individual, and trace this individual through time and through loss of perceptual contact” (Carey 2001: 189). This means that object permanence presupposes cognitive processes like individuation, inferences of continuation and identity judgments. This is partly reflected linguistically in count noun properties like countability and pluralizability and in grounding devices such as definiteness. So, the definite singular (i.e. the ball) normally reflects that the speaker has a

98

Chapter Seven: The Construal of Entity Permanence

specific object individual in mind which the listener is intended to recognize, while a plural like balls presupposes that the speaker has been able to individuate several objects of the same kind. Children, as well as adults, use spatiotemporal continuity criteria to individuate material objects and judge if they are identical or not. These include the fact that an object cannot be in two places at the same time, and that you can track an object’s movements through continuous points in space and continuous points in time (Carey 2001; Bloom 2000; Spelke et al. 1995; Wagner and Carey 2003). Persons and material objects seem to be special in perception and cognition as well as in lexical acquisition, but they are certainly not the only kinds of entities that we individuate. Even small children can individuate, and count, things like sounds, patterns, collective entities, actions and so on. We also individuate jokes, thoughts, ideas and other entirely abstract phenomena (Bloom 2001). Now, the question is, could we speak of some kind of more generalized type of entity permanence, involving conceptualizations and assumptions we hold for entities in general, not only for material objects? This paper will discuss some non-object entities that are not entirely abstract. The term objects is only used about physical material objects here. As a superordinate term for all kinds of things that we individuate, individuals or—mostly—entities will be used as synonymous expressions. This means that individuals are not only people, and entities are only taken to mean individuated countable entities—not stuff or substances. Permanence does not imply eternal existence, but some sort of continued existence which does not preclude change. We think of people as permanent individuals, but we all know that people change over the years. How we conceptualize phenomena as entities is of course a huge question, and this paper will only offer some suggestions. The general conclusion is that we tend to make assumptions of continued existence and numerical identity for many other entities than objects. In a way, that goes with being an entity, and nouns are sometimes said to denote stable phenomena. But in order to attain this stability we seem to need a rather plastic conception of these permanent entities, and our construals of nonobject entities turn out to be quite flexible. The conclusions here are based entirely on linguistic analysis of authentic Swedish language use. The empirical foundation is mostly corpus data from the Bank of Swedish (http://spraakbanken.gu.se/lb/konk/) where the newspaper corpora have been used. All numbered examples are citations from these corpora. We will deal with only one special kind of non-object individuals: human collective institutions. But since judgments of the identity of an

Jan Svanlund

99

entity most often becomes relevant when we align instances from two different time periods, we will start by illustrating entity permanence with some constructions of temporal contrasts in Swedish.

Temporal contrasts There are several expressions in Swedish that primarily contrast phenomena in two or more time spaces. The Swedish words in Table 1 modify NPs and contrast entities in two time spaces. The English translations are somewhat rough, since the Swedish words are used in attributive position (except ex-) and often lack direct equivalents in English. To the left we have words that indicate an entity in an earlier time, a THEN space. To the right we have words marking that something applied to a later time space, which usually also includes the speaker’s NOW. However, that is not always the case, so this time space will schematically be labelled LATER. Table 1. Some expressions for temporal contrasts in Swedish SWEDISH DÅ dåvarande dåtida dåtidens den tidens förra ex-

ENGLISH SENARE nuvarande nu senare sedermera sedan dagens

THEN then at that time at that time of that time former ex-

LATER now now later later later today’s

These constructions are sometimes used pairwise—as in example 1 below—and they can then be combined rather freely, left to right. But most often they are not used in pairs—as in 2–4. Then some counterpoint in the other time space is left implicit. I have studied the THEN words, but here I will only comment on dåvarande, the most frequent of the THEN expressions besides förra ‘former’. (1) Dåvarande premiärministern och senare presidenten Milton Obote … ‘The then prime minister and later president Milton Obote …’ (2) Vi lärde känna varandra genom min dåvarande pojkvän och det blev kärlek vid första ögonkastet. ‘We got to know each other through my then boyfriend and it was love at first sight.’

100

Chapter Seven: The Construal of Entity Permanence

(3) För några år sedan upptäckte skidlandslagets dåvarande tränare Åke Jönsson att snön kom tidigt och låg kvar i Bruksvallarna. ‘A few years ago the then coach of the national skiing team Åke Jönsson discovered that the snow came early and stayed in Bruksvallarna.’ (4) 1989 bildade Charles Taylor rörelsen NPFL i uppror mot Liberias dåvarande regering. ‘In 1989 Charles Taylor formed the NPFL movement in a rebellion against the then government of Liberia.’ (Some of the English translations display unidiomatic ways of using then. This is meant to illustrate the structure in the Swedish sentences.) At the most schematic level dåvarande X implies a discrepancy between THEN and LATER concerning X—something applied THEN, but not LATER. That usually means that something has changed, and this something should be conceived of as a permanent entity. Change thereby also implies some kind of permanence. There are two main uses of dåvarande: change of role or role value and change of name. A) Change of role or role value (as in examples 1–3) concerns personal roles in the sense of Fauconnier’s (1994) theory of Mental Spaces. The focus is often on the role itself, which later got another value, another person filling the role (as in 2 and implicit in 3). These are normally interpreted as permanent roles, but their values change. Focus can also lie on the person (the value), who later got another role (as in 1). The person is naturally taken to have continued existence, but his roles change. The roles are quite specific. In example 1, the role is not prime minister, but prime minister of Uganda, and the general formula for a role is mostly THE X OF Y, even if Y is not always expressed and can be left implicit. In these examples we have succession roles, which imply that someone should occupy the role at a given time. So when one leaves, another will probably succeed him or her, and it is taken to be the same role in both time spaces. With boyfriend in 2, the implication of succession is not as strong, but in this specific case it seems safe to conclude that the former boyfriend was also succeeded. In example 4, government is a collective institution, and the role is a collective role with a collective value. We will come back to this kind of role. The source of inferences of role permanence and value successions is partly our knowledge of these kinds of roles—knowledge of the world—

Jan Svanlund

101

and partly linguistic knowledge of how these particular time-contrasting constructions are mostly used. B) Change of name (as in examples 5–7) concerns entities whose names have changed. In example 5 the entity is a company. (5) År 1947 anställdes hon i dåvarande Separator, sedermera Alfa Laval AB. ’In 1947 she was employed at then Separator, later Alfa Laval AB.’ (6) Vi vet inte mycket om vad han upplevt i sitt hemland, dåvarande Zaire, nu Kongo-Kinshasa. ‘We do not know much about what he experienced in his native country, then Zaire, now Congo-Kinshasa.’ (7) Det är nu 30 år sedan Storbritannien gick med i dåvarande EEC. ‘It is now 30 years since Great Britain joined the then EEC.’ Only permanent entities can change names, and they have to be conceived of as the same entity. But we can sometimes wonder if we really have the same entity in both spaces. There is probably a scale concerning the strength of equivalence, the identity, between entities in two spaces. This is because change of name also often involves some sort of transformation of the entity. You could argue that the European Union—the implicit counterpoint in 7—is something quite different from what the old EEC was. But still, it was exactly the EEC that developed into the EU, i.e. the same entity being transformed. It is not just an analogue. We have a temporal path connecting these two versions of the entity, a temporal continuity, a historical—and in some sense also a causal—connection between them. This seems to be central to entityhood and entity permanence, and it holds for the rest of the examples in this chapter. This is similar to object permanence, where temporal continuity also is one corner stone. But for more abstract entities, the temporal connection must be added or inferred by our world knowledge, we cannot check it perceptually. One could ask why we need constructions like these, anyway. It would seem obvious that if we are talking about THEN, what we say would naturally apply to THEN, not to LATER. But the reason for pointing out the THEN-status of some description is often to prevent unwarranted inferences of temporal stability. Some entities or facts could otherwise be interpreted as more permanent and in more parallel between THEN and LATER than they actually are. This suggests in itself that permanence is naturally taken for granted as default when we ascribe entity status to something. But

102

Chapter Seven: The Construal of Entity Permanence

entity permanence comes in degrees, which we shall see in the next section.

Team individuation There are relative differences in entity permanence. The entities that roles are representatives of—the Y’s in the role formula THE X OF Y above—are normally seen as more permanent than the role type or the role values. In example 3, the skiing team is taken to be more permanent than the coach, or at least than the value of the coach. National sports teams represent a kind of entity which we can call human collective institutions, which also include entities like governments, associations, and boards. They are constituted by a collection of members, persons. Our conceptions of these entities are naturally determined by our knowledge of how different kinds of institutions work. They are often partly constituted by laws or formal rules. This should make it easier for us to individuate them and pick out the relevant entities. On the website of The Swedish Football Association there is a picture with the header Herrlandslaget 2007 – ‘The men’s national team 2007’. It is a standard type of team picture with the players in three lines, everyone facing the photographer. In all there are 22 persons in the picture, including the two coaches and the two goal-keepers, who are dressed differently from the rest. There are also several lists of the players at the website. Some contain 41 names, others 43, etc. Obviously, then, the team is in some sense much larger than the picture indicates, and exactly who are included is maybe not self-evident. So the picture only shows one version of the team, and we can assume—in fact we know—that there have been several versions over the year. I will speak precisely of different team versions as different instances of the team. They are instances of an entity that is more schematic and idealized than we usually tend to imagine. We usually call this entity fotbollslandslaget ‘the national football team’, or metonymically just Sverige ‘Sweden’. But any instance, any team version, can be described by the same lexical items. In fact, for many authentic formulations it is quite hard to decide if they refer to the most schematic entity or to some more specific versions of it. The schematic entity is ascribed a great deal of permanence, which can be seen in examples 8í9. (8) Det svenska fotbollslandslaget har aldrig varit mer populärt. ‘The Swedish national team in football has never been more popular.’

Jan Svanlund

103

(9) På vägen mot den största framgången hittills, OS-guldet i London 1948, tog fotbollslandslaget också den hittills största segern i tävlingssammanhang. [12–0 mot Korea.] ‘On the way towards their greatest success so far, the Olympic gold in London in 1948, the national football team also took the greatest victory so far in competitions.’ [12–0 against Korea] These are implicit comparisons between team versions. They are construed as the “same” entity that varies over time. This would in Blending Theory be analyzed as a compression from several instances into one unique entity (Fauconnier and Turner 2002). The differences between instances are then construed as changes within one entity—rather than as differences between separate entities. Discrepancies over the years are then backgrounded, except for the very difference that the statement is about. This seems so natural to us, since it is a very conventionalized and entrenched kind of construal. But it is possible to construe the team versions explicitly as different entities and compare those separate entities instead. We have a choice of construal. (10) David Beckhams England har aldrig slagit Sverige i VM. ’David Beckham’s England has never beaten Sweden in the World Championships.’ (11) 1948 í 1958 í 1974 í 1994 í fyra förträffliga årgångar då allt stämde. Fyra framgångsrika landslag. ‘1948 í 1958 í 1974 í 1994 í four excellent [year classes] when everything worked. Four successful national teams.’ (12) Tidigt på måndagsmorgonen stegade han in på hotellet och anslöt sig till Tommy Svenssons fotbollslandslag. ‘Early on the Monday morning he walked into the hotel and joined Tommy Svensson’s national football team.’ Standard ways of separating team versions as individuated entities are by plural, by attaching them to a specific time period or to a specific coach or player, and in Swedish also by instance nouns such as årgång (‘year class’) and upplaga (‘edition’), which explicitly depict the entity as an instance of a more permanent and more schematic entity. And when someone says that a player is put outside the team, it does not mean that he or she is not a member of the team anymore. It means

104

Chapter Seven: The Construal of Entity Permanence

that he or she is not in the team actually playing at the moment. This is a more narrow conception of the team. An even more narrow conception is displayed by the common practice to construe exactly the same constellation of players as different teams, as when someone say that Sweden came out as a completely different team in the second half. Philosophers would say that this use is not about numerical identity but of qualitative identity. Still, it is an entrenched type of team construal. In these ways we can vary the degrees of compression. We can construe the team as a more narrow or more broad entity. We can zoom in and zoom out, and we can easily do this within the same discourse. Sports teams can be seen as collective roles, and they are usually representatives of some higher and more permanent entity, generally a club or a nation – once again, the Y’s of the role formula THE X OF Y. The maximum construal of the team is usually the Y which it represents – the entire club or the entire nation. This can be seen in jokes like “We won, but they lost”, alluding to common opportunistic formulations like “We beat England by 3–0”, said by someone who watched the game from the TV sofa, as opposed to “But then they lost to Germany”. While it is possible to conceive of they as referring to the current national team only, we must be seen as a much wider national entity (which, of course, is the point of the identification and dissociation processes alluded to by the joke). This tells us that the maximum construal need not be the most basic or most default construal. Most of the time we do not even notice the various entity zoomings in a text or a discourse. We generally think we are talking about exactly the same entity all the time; and in a sense, this is precisely what we are doing. It is just that our conceptions of single individual entities are much more flexible than we usually imagine. This is in line with statements in Blending Theory which say that the identity relation is mostly taken for granted as objectively given, although it is actually a mental construct: “Identity seems to be a primitive, an unanalyzable notion, but instead it is an achievement of the imagination.” (Fauconnier and Turner 2002: 115.) There is a parallel here to insights from categorization research, where our conceptions of categories have also been found to be flexible to certain degrees. They need to be, in order to meet the requirement that cognition should combine structural stability with flexible adaptability, which can be observed in so called prototype categories (cf. Geeraerts 2006: 42). Our conceptions of specific entities also seem to be flexible, probably for similar reasons, which is more rarely noted. At the perceptual level, we obviously need to be able to recognize for instance persons at different occasions under different appearances as being the same individuals, in

Jan Svanlund

105

spite of the fact that they might change considerably over time. We also need to acknowledge who these people are, in what relationships they stand to each other and to ourselves as observers. This can also change through time. For non-object entities, like teams, the perceptual appearance is not equally salient. But in order to conceive of the history of an entity, we need to recognize changes, i.e. different properties of different versions of the entity, and at the same time acknowledge precisely that they are versions of the same entity. Another parallel to prototype categories is that there seems to be basic level versions of an entity, which we tend to evoke as default in certain contexts (see the next section and the last one).

Plurals One aspect of individuation is the possibility of separating several entities of the same kind, mirrored by the pluralization of count nouns. Plurals can be used to contrast entities against one another. In the case of teams, this can be done in at least three dimensions. Consider the following examples. (13) Bättre svenska landslag än så här har haft svårt att föra matcher mot sämre motståndare. ‘Better Swedish national teams than this have had difficulties dominating matches against inferior opponents.’ (14) Sverige är ett av de tråkigaste fotbollslandslagen att kolla på. ‘Sweden is one of the most boring national football teams to watch.’ (15) Vi har åtta olika landslag i gång. Vi har inte mer pengar, vi är ett fattigt förbund. ‘We have eight different national teams running. We do not have more money, we are a poor association.’ The plural expression (X)-landslag í ‘national (X)-teams’ í where X stands for any sport designation, can be used to differentiate: A) entities in a national dimension, most often diachronic (when team versions from different times are compared, as in example 11 and implicitly in 13). B) entities in an international dimension, most often synchronic (comparing teams from different countries, as in example 14). C) entities which are subtypes: men’s, women’s, boys’, girls’ teams etc. (as in 15).

106

Chapter Seven: The Construal of Entity Permanence

These dimensions differ in quality, and they show that using plural is one thing, counting is partly another. It makes sense to count teams on dimensions B and C; we can ask how many national teams there were in the World Championships in 2006 (dimension B), and we can say that the Swedish Football Association has 14 national teams (dimension C). But if we ask similar questions at dimension A, they do not really make sense: “How many national football teams has Sweden had since 1948?” The question is odd, because team versions separated in this manner are more or less ad hoc-construals. There can be an infinite number of them. On the other hand, what constitutes individuated entity versions at the other dimensions is partly guided by formal rules and these versions are also more culturally entrenched. Dahl (1975) uses manifestations of individuals in a way very similar to the way versions of (team) entities are used in this article. His notion is intended to be very general and it includes quite narrow manifestations of human individuals as well, like John Smith today vs. John Smith yesterday. Dahl also prompts for a level between individuals and manifestations as the more salient level: subindividuals. A subindividual could be seen as a subset, or cluster, of manifestations of an individual, and the number of possible manifestations is infinite. It is uncertain how we can distinguish subindividuals from manifestations, since none of them are objectively given when it comes to non-object and non-human entities. They are all construed by human minds. It is perhaps more useful to say that we are able—and free—to construe manifestations/versions in all various degrees of compressions. But it would also be useful to distinguish some construals as more conventionalized, more entrenched or contextually more well-established ways of construing an entity. Such construals could perhaps be called subentities. Anyhow, even the most schematic entity must be interpreted as a kind of construal.

What is included? The last aspect of team semantics to comment upon here is rather the opposite of the previous one: What can be united into a single entity, who can be included? In Sweden, at least, we do not construe the national football team as comprizing both men’s and women’s teams. But in several other sports the national X-team is actually quite often construed as an entity of both male and female team versions united. These are sports like swimming, athletics/track-and-field, and skiing.

Jan Svanlund

107

(16) Engdahl ska coacha damlaget under helgens tävling. I oktober tar Engdahl sedan över ansvaret för hela det framgångsrika friidrottslandslaget. ‘Engdahl is coaching the ladies team during this weekend’s competition. In October Engdahl will then take over the full responsibility for the whole successful national athletics team.’ (17) Svenska friidrottslandslagets damer kom tvåa i Sevilla i för två veckor sedan och missade avancemang till Superligan. ‘The ladies of the Swedish national athletics team came second in Sevilla two weeks ago and missed the advancement to the Super League.’ (18) Anna-Karin Kammerling och Lars Frölander har gjort de kraftigaste magplasken än så länge i ett blekt svenskt simlandslag. ‘Anna-Karin Kammerling och Lars Frölander have made the greatest belly flops so far in a pale Swedish national swimming team.’ The examples show how the national X-team can include, and in these sports usually does include, both men and women í otherwise you could not refer to the women of this united team entity, as in 17; and in example 18 one female and one male swimmer are mentioned as representatives of the national swimming team. On the other hand, the national football team could not be construed in these ways. It refers either to the male or the female team versions, and the same goes for sports like handball and icehockey. We can imagine several reasons for these differences, some of them cultural, some of them more perceptual and conceptual in nature. The various sports can roughly be divided into two types, which are illustrated in Table 2. In “genuine” team sports, the athletes appear more clearly as a coherent perceptual unit, wearing the same outfit and primarily occupying a special part of the field etc. This is contrasted with the individual team sports, where this is not the case.

108

Chapter Seven: The Construal of Entity Permanence

Table 2. Differences between sports “Genuine” team sports (football, handball, hockey) performed collectively coherent appearance (i.e. as a team) competitions in separate events (males and females not competing together) different coach/manager for males and females less equal status for males and females

Individual team sports (athletics, swimming, skiing) primarily performed individually less coherent appearance (i.e. as individuals) competitions in joint events (males and females competing together) same coach/manager, joint organization more equal status for males and females

In short, we can say that there are fewer forces to conceive of genuine sports teams as united entities of men and women, while there are fewer constraints on conceiving individual sports teams in this way.

Type vs. instance From what has been said above, we can conclude that team entities seem to have a somewhat dual nature when it comes to the type vs. instance distinction. We can illustrate this with a lexical type hierarchy. In Figure 1, we find ever more specific subtypes of teams. lag (team) landslag (national team) fotbollslandslag (national football team) (svenska) fotbollslandslaget (the Swedish national football team) 1948

1958

1974

Figure 1. A lexical team hierarchy in Swedish

1994

Jan Svanlund

109

Near the bottom we find svenska fotbollslandslaget, ‘the Swedish national football team’, marked with italics. This would normally be seen as an instance of the subtype fotbollslandslag ‘national football team’. It is clearly an instantiation in the cognitive linguistic sense (cf. Langacker 1991), since it is grounded by the definite suffix article -et. At the same time, we have seen several construals of different kinds of instances of this entity, and some of them are suggested at the very bottom of the figure (however not with their normal designation, which would simply be precisely svenska fotbollslandslaget). This also makes the Swedish national football team some sort of type with several possible instantiations, in fact an infinite number of them. Collective entities of this sort, among others, generally seem to display this duality (unlike instances of object types like apple, cat or pencil). Perhaps this is connected to the fact that teams are collective roles, and roles have different values at different times, i.e., different instantiations. Langacker (1991) argues against an absolute distinction between elaboration (the relation between types and subtypes) and instantiation (the relation between (sub)types and instances). He considers instantiation as a special case of elaboration. Perhaps the duality of collective institutions can be seen as another argument against the absoluteness of this distinction. This kind of duality is however not unique. Other types of entities behave in similar ways. Symphony is an example sometimes mentioned in discussions of type-token distinctions, for instance by Jackendoff (1983) and Dahl (1975, 2008). Here we can speak of an instantiation hierarchy, with the type symphony at the top. A specific symphony—like Beethoven’s 9th symphony—is naturally a grounded instance of this lexical and categorical type. But this instance, which is attributed a great deal of permanence, is still only a schematic one, and various versions of it can differ substantially. Symphonies can be instantiated and materialized in several ways: x symphony (type) x Beethoven’s 9th Symphony, a specific symphony (= B9) x performances of B9 x recordings of (performances of) B9 x material representations of recordings (CD’s etc.) x playing of recordings We still think of Beethoven’s 9th Symphony as being the same entity each time we hear it, through all versions and instances. We do not just consider them as belonging to the same category or type. Note also the use

110

Chapter Seven: The Construal of Entity Permanence

of instance nouns here like performances and recordings. There is hardly any doubt as to which level is the most basic one in this hierarchy. If people were to answer the admittedly rather strange question “How many symphonies are there?”, I expect most would start counting like “Well, Beethoven wrote 9, Mozart 41 …” and so on. It would be very odd to start counting performances, recordings or records. In conclusion, when we talk about collective institutions and musical pieces (and other kinds of performed artistic work, at least), the basic entities seem to have this dual nature of instance and type, and they are naturally ascribed entity permanence.

References Bloom, P. 2000. How children learn the meanings of words. Cambridge, Mass: MIT Press —. 2001. Roots of word learning. In M. Bowerman and S.C. Levinson (Eds.), Language Acquisition and Conceptual Development, 159í181. Cambridge: Cambridge University Press Carey, S. 2001. Whorf versus continuity theorists: bringing data to bear on the debate. In M. Bowerman and S.C. Levinson (Eds.), Language acquisition and conceptual development, 85í214. Cambridge: Cambridge University Press Dahl, Ö. 1975. Individuals, subindividuals, and manifestations. Logical Grammar reports 15. Göteborg —. 2008. Animacy and egophoricity: Grammar, ontology and phylogeny. Lingua 118: 141–150. Fauconnier, G. and M. Turner, 2002. The way we think. Conceptual blending and the mind’s hidden complexities. New York: Basic Books —. 1994. Mental spaces: Aspects of meaning construction in natural language. Cambridge: Cambridge University Press. Geeraerts, D. 2006. Words and other wonders. Papers on lexical and semantic topics. Berlin, New York: Mouton de Gruyter Jackendoff, R. 1983. Semantics and cognition. Cambridge, Mass: MIT Press Langacker, R. 1991. Foundations of cognitive grammar. Vol II: Descriptive application. Stanford: Stanford University Press. Spelke, E.S., R. Kestenbaum, D.J. Simmons and D. Wein. 1995. Spatiotemporal continuity, smoothness of motion and object identity in infancy. British Journal of Developmental Psychology 13: 113–142. Wagner, L. and S. Carey 2003. Individuation of objects and events: a developmental study. Cognition 90: 163í191.

ON THE PROTOTYPICALITY OF DIMENSIONAL ADJECTIVES ELENA TRIBUSHININA

Abstract. This chapter critically assesses the view that the only substantial difference between colour terms and dimensional adjectives is prototypicality of the former and non-prototypicality of the latter. Using two sets of data – a survey and a corpus – I demonstrate that there are, at the very least, two different kinds of prototypicality effects relevant to the semantics of dimensional adjectives denoting vertical size. I term them prototypicality qua head-nouns and prototypicality qua best exemplars. Prototypicality qua head-nouns is manifested in the selective combinability of dimensional adjectives with noun heads, in the sense that some adjective-noun combinations are more prototypical than others. Prototypicality qua best exemplars involves the finding that some entities are considered to be prototypical instantiations (best exemplars) of the property denoted by the adjectives (e.g. towers and giraffes for tall, elephants for large). Importantly, in elicitation tests, subjects give fairly uniform judgments about what counts as a prototypically tall or a prototypically large entity in their culture. Moreover, there is a great deal of cross-linguistic uniformity. This study also shows that prototypicality is a matter of degree, in the sense that some concepts are more prototypical than others. Colour terms reveal more prototypicality effects qua best exemplars, whereas dimensional adjectives score higher on prototypicality qua head-nouns.

Introduction According to some modern semantic theories, the only substantial difference between colour terms, such as red, and dimensional adjectives, such as tall, is prototypicality of the former and non-prototypicality of the latter. To quote Kamp and Partee (1995: 176): As an example of a prototype-free vague concept we have given tall. Over tall’s vagueness there can hardly be any argument. We also think it is quite clear that tall has no prototype. This has to do with the fact that it can be applied to an indefinite variety of things and with the circumstance

112

Chapter Eight: On the Prototypicality of Dimensional Adjectives that there is in general no natural upper bound to how tall things can be. Other unbounded scalar concepts, such as heavy, big, wide, etc. also belong to this type.

In this chapter, I argue that the semantic analysis of tall along the lines suggested by Kamp and Partee (1995) is grossly inadequate. Firstly, as shown by Tribushinina (2008a), the assumption that dimensional adjectives are unbounded terms is not as uncontroversial as is often presented. Adjectival meanings interact with the meanings of their headnouns; and this interaction results in the establishment of categorical boundaries. Secondly, as demonstrated by Dirven and Taylor (1988), tall cannot be “applied to an indefinite variety of things”, since it has as its salient reference point a very specific kind of verticality, the one exhibited by human bodies. I will critically assess the third and central part of Kamp and Partee’s claim that dimensional adjectives are prototype-free. I will suggest that prototypicality is not a matter of yes-or-no distinction; rather, it is a matter of degree. Geeraerts, Grondelaers and Bakema (1994) term this phenomenon prototypicality of “prototypicality”, thereby emphasising that “some concepts are more typically prototypical than others, in the sense that they exhibit more of the ‘prototypical’ characteristics” (Geeraerts, Grondelaers and Bakema 1994: 54). It is not inconceivable that dimensional adjectives are to a lesser degree oriented to prototypes than, for example, colour terms. Furthermore, as noticed by Taylor (2003), prototypical redness can be represented without reference to entities displaying this colour, whereas we cannot “conceptualise ‘prototypical tallness’, or ‘focal tallness’, without at the same time picturing a tall kind of entity” (Taylor 2003: 279). However, the fact that prototypical tallness does not exist by itself and is always contingent on the entity exhibiting this property does not mean that prototypes of tallness do not exist. This category-dependence only shows that dimensional adjectives are more relative than colour terms. A number of studies of dimensional adjectives have demonstrated that there are prototypical instantiations of spatial properties. For instance, towers, trees and houses were shown to be prototypically tall entities (i.e. best exemplars of tallness), just like blood was shown to be a best exemplar of red. What is even more interesting, some of these prototypes were shown to be largely uniform across different languages (Dirven and Taylor 1988; Goy 2002; Taylor 2003; Vogel 2004; Weydt and SchliebenLange 1998).

Elena Tribushinina

113

Previous studies also suggest that there are at least two different manifestations of prototypicality relevant to dimensional adjectives. Firstly, there are certain topological types of objects that are prototypically described by means of particular adjectives. For instance, human beings are typically thought of in terms of tallness due to being dynamically growing entities with a canonical vertical orientation, standing out from the background by virtue of their maximally salient vertical dimension (Dirven and Taylor 1988; Taylor 2003). I will term this phenomenon prototypicality qua head-nouns. Secondly, entities whose dimensions are associated with the maximum on the relevant scale may function as best exemplars of the property. In this sense, towers and giraffes, but not people, are prototypically tall. The two sorts of prototypes may, but do not have to coincide. For instance, elephants may be considered as prototypically big objects, but exactly for this reason the noun elephant is rarely modified by the adjective big. The combination big elephant is redundant, unless we compare a particular elephant with its average conspecifics. With these considerations in mind, this chapter sets out to more systematically examine the role of prototypes in the semantics of dimensional adjectives. I will also compare prototypicality effects in dimensional adjectives with manifestations of prototypicality in colour terms. The chapter is structured as follows. I start by describing the methodology. I then turn to the discussion of prototypicality effects qua head-nouns. After that, I address prototypicality qua best exemplars. The last section summarises the main conclusions.

Data and methodology This study focuses on Russian adjectives denoting vertical size. Three dimensional adjectives will be targeted for analysis – a supra term vysokij ‘high/tall’ and two sub terms í nizkij ‘low/short’ and nevysokij ‘not.high’. (The terms supras and subs were borrowed from Croft and Cruse 2004). Notice that dimensional supras in Slavic languages, unlike their counterparts in Germanic languages, may have both lexical and morphological opposites. For a detailed study of differences between the two types of antonym, I refer the reader to Tribushinina (2008b). The principal method used in this study is a survey. The subjects were 174 undergraduate students (105 female, 69 male; age range: 17í35) attending either Kuzbas Technical University or Kemerovo State Medical Academy (Kemerovo, Russia). The survey consisted of four tasks and was

114

Chapter Eight: On the Prototypicality of Dimensional Adjectives

conducted as part of a larger study (Tribushinina 2008a). Only the tasks relevant to the present study will be discussed here. Two different tasks were designed in order to pinpoint two different kinds of prototypicality indicated above. Task 1 targeted prototypicality qua head-nouns; Task 2 was employed to study prototypicality qua best exemplars. The two tasks will be described in turn. Task 1 was used to elicit prototypical head-nouns of krasnyj ‘red’, vysokij ‘high/tall’, nizkij ‘low/short’, and nevysokij ‘not.high’. The subjects were told to give three nouns that they consider to go particularly well with a given adjective. This procedure was introduced by Weydt and Schlieben-Lange (1998) in their study of spatial adjectives in German. I modified the procedure in two ways. Firstly, I introduced nine distracters (dalekij ‘remote’, interesnyj ‘interesting’, sladkij ‘sweet’, krasivyj ‘beautiful’, þistyj ‘clean’, krasnyj ‘red’, dobryj ‘kind’, trudnyj ‘difficult’, and pustoj ‘empty’). Secondly, I made two versions of the test – one with nevysokij ‘not.high’ and one with its near-synonym nizkij ‘low’ – in order to avoid priming. In total, 87 subjects filled in version 1, and 87 subjects version 2 of the questionnaire. Only prototypical uses of the target adjectives (i.e. colour and vertical size) were considered in this study. All metonymical and metaphorical extensions were excluded from consideration. Task 2 addressed prototypicality effects qua best exemplars. The subjects were asked to continue three expressions: krasnyj kak ‘as red as’, vysokij kak ‘as high as’, and nizkij kak ‘as low as’. The adjective nevysokij ‘not.high’ was not included in the task, because it is uncommon in Russian to use negatively prefixed adjectives in this type of comparative construction. The elicited data was compared to and supplemented by non-elicited data from the Russian National Corpus (RNC). The data for the study were extracted in September 2006 when the corpus contained 120 million words. All extended uses of the adjectives were left out of consideration. The number of relevant uses was: 13,459 for krasnyj ‘red’, 6,385 for vysokij ‘high/tall’, 1,293 for nizkij ‘low/short’ and 1,494 for nevysokij ‘not.high’.

Prototypicality qua head-nouns There are three important questions that have to be answered in this part of the investigation. Firstly, are there any nouns whose frequency exceeds by far the frequency of other elicited noun heads? If all the elicited nouns have equally small frequencies in the survey, it would indicate that there

Elena Tribushinina

115

are no entities that are thought of as objects prototypically described by means of the above adjectives. If, on the other hand, the respondents consistently used the same nouns that, in their opinion, go particularly well with these adjectives, it would be evidence of prototypicality effects in the combinability of adjectives with a few prominent noun heads. Secondly, are there any prototypical referents systematically elicited across different languages? It is interesting to compare the findings from my survey with the results of similar surveys conducted for other languages. Thirdly, are dimensional adjectives less prototypical qua head-nouns than colour adjectives? It is necessary to compare the proportion of nouns frequently elicited for the dimensional adjectives with the salience of frequent head-nouns elicited for the colour term krasnyj ‘red’, since colour terms are usually cited as typical examples of prototypicality. Tables 1-3 show frequencies of head-nouns elicited for vysokij ‘high/tall’, nevysokij ‘not.high’ and nizkij ‘low/short’. Only nouns elicited with a frequency greater than 1 are listed in all the tables in this article. Table 1. Head-nouns of vysokij ‘high/tall’ WORD þelovek ‘man/human being’ dom ‘house’ derevo ‘tree’ gora ‘mountain’ trava ‘grass’ zdanie ‘building’ bašnja ‘tower’ rost ‘stature’ stolb ‘post’ devuška ‘girl’ dub ‘oak-tree’ zabor ‘fence’ stena ‘wall’

FREQ. 75 69 51 27 20 20 18 18 15 11 8 8 5

WORD lestnica ‘ladder/stairs’ mužþina ‘man’ el' ‘spruce’ ograda ‘hedge’ paren’ ‘lad’ skala ‘rock’ xolm ‘hill’ bereza ‘birch-tree’ maþta ‘mast’ most ‘bridge’ neboskreb ‘skyscraper’ sosna ‘pine-tree’ ženšþina ‘woman’

FREQ. 4 4 3 3 3 3 3 2 2 2 2 2 2

116

Chapter Eight: On the Prototypicality of Dimensional Adjectives

Table 2. Head-nouns of nevysokij ‘not.high’ WORD þelovek ‘man/human being’ dom ‘house’ derevo ‘tree’ rost ‘stature’ zdanie ‘building’ gora ‘mountain’ trava ‘grass’ zabor ‘fence’ devuška ‘girl’ kust ‘bush’ paren’ ‘lad’

FREQ. 40 32 25 15 12 9 9 6 5 4 4

WORD bašnja ‘tower’ stolb ‘post’ xolm ‘hill’ dver’ ‘door’ kabluk ‘heel’ lestnica ‘ladder/stairs’ mužþina ‘man’ ograda ‘hedge’ stol ‘table’ stul ‘chair’ teremok ‘tower-chamber’

FREQ. 3 3 3 2 2 2 2 2 2 2 2

Table 3. Head-nouns of nizkij ‘low/short’ WORD rost ‘stature’ dom ‘house’ trava ‘grass’ þelovek ‘man/human being’ stol ‘table’ derevo ‘tree’

FREQ. 22 15 15 12 11 10

WORD stul ‘chair’ zabor ‘fence’ krovat’ ‘bed’ skamejka ‘bench’ stolb ‘post’ zdanie ‘building’

FREQ. 10 8 3 2 2 2

It is clear from the figures in Tables 1-3 that there are a few headnouns that were with significant frequency elicited for each of the three adjectives. For vysokij ‘high/tall’ and nevysokij ‘not.high’, the three highly prominent head-nouns are þelovek ‘man/human being’, dom ‘house’ and derevo ‘tree’. This result suggests, in line with the findings reported in Tribushinina (2008b), that nevysokij ‘not.high’ inherits its salient referent categories from its source word vysokij ‘high/tall’. For nizkij ‘low/short’, the prototypical status of the most frequently elicited head-nouns – rost ‘stature’, trava ‘grass’, and dom ‘house’ – is somewhat less clear-cut, in the sense that, in terms of frequencies, there is no pronounced difference between these nouns and the items following them (see also Figure 1). Put differently, nizkij ‘low/short’ has a more even distribution of head-nouns than both vysokij ‘high/tall’ and nevysokij ‘not.high’. To return to the first question asked at the beginning of this section, if there are any nouns whose frequency exceeds by far the frequency of other elicited noun heads, the answer is a qualified yes. The results indicate that prototypicality effects with regard to head-nouns are much stronger for vysokij ‘high/tall’ and its morphologically related opposite nevysokij ‘not.high’ than for the lexical sub term – nizkij ‘low/short’. This finding is

Elena Tribushinina

117

consistent with the results of the corpus study, which has shown that nizkij ‘low/short’ displays a fairly even distribution with different head-nouns. In contrast, both vysokij ‘high/tall’ and nevysokij ‘not.high’ obviously prefer human referents to other types of head-nouns. Humans constitute 37.2% of the head-nouns of vysokij ‘high/tall’ and 54.6% of the head-nouns of nevysokij ‘not.high’ in the RNC. Other prominent referent categories of the two adjectives in the corpus were vegetation, buildings and eminences, which is very similar to the elicited data reported above. We are now in a position to compare these results with the results of similar studies reported for German (Weydt and Schlieben-Lange 1998), Swedish (Vogel 2004), and Italian (Goy 2002). As explained earlier, the method used in this task was originally designed by Weydt and SchliebenLange (1998) for German. Despite the minor methodological differences mentioned above, the results of the two studies are quite uniform. So, in the German study, the following nouns were elicited as the best noun heads of hoch ‘high’: Turm ‘tower’, Berg ‘mountain’, Haus ‘house’, Hochhaus ‘high-rise building’, Baum ‘tree’, and Wolkenkratzer ‘skyscraper’. Note that four of these entities (tower, mountain, house, and tree) are also prominent in the Russian data. As for niedrig ‘low’, the following nouns were most frequently provided by the German respondents: Tisch ‘table’, Decke ‘ceiling’, Haus ‘house’, Stuhl ‘chair’, and Tür ‘door’. Four of these nouns (‘table’, ‘ceiling’, ‘house’, and ‘chair’) were also elicited in my survey. Vogel (2004) used the same procedure as Weydt and Schlieben-Lange (1998) and elicited head-nouns for fourteen dimensional adjectives in Swedish. The subjects repeatedly responded that hög ‘high’ most felicitously describes houses, mountains, masts, buildings, towers, piles, and ladders. Yet again, we see considerable similarities to the Russian data presented in Table 1. For låg ‘low’, this elicitation test did not yield any head-nouns that would stand out from the rest, this probably because of the relatively small number of subjects participating in the study (17). Similar results were reported by Goy (2002, quoted in Vogel 2004) with respect to the Italian adjectives alto ‘high’ and basso ‘low’. Alto is prototypically used to describe such entities as a tower, a pyramid, a wall, a house, and a glass, whereas basso combines best of all with nouns automobile ‘car’ and divano ‘couch’. Thus, we observe a great deal of cross-linguistic uniformity in prototypicality judgments about head-nouns of dimensional adjectives. These results are remarkably similar to the findings from colour research, which laid the foundations of prototype theory. A major difference between the Russian data presented above and the results from German,

118

Chapter Eight: On the Prototypicality of Dimensional Adjectives

Italian and Swedish is that only in Russian human beings were elicited as prototypical referents of ‘high’ and ‘low’. Now let us turn to the third question: Are prototypicality effects qua head-nouns less clear-cut for dimensional adjectives than for colour terms? Table 4 lists head-nouns of krasnyj ‘red’ elicited on this task. Only relevant (i.e. colour) uses are included in the table. Table 4. Head-nouns of krasnyj ‘red’ WORD pomidor ‘tomato’ cvet ‘colour’ jabloko ‘apple’ solnce ‘sun’ zakat ‘sunset’ krov’ ‘blood’ kofta ‘jumper’ flag ‘flag’ mašina ‘car’ nos ‘nose’ cvetok ‘flower’ mak ‘poppy’ platje ‘dress’ rak ‘crayfish’ svet ‘light’ perec ‘paprika’ roza ‘rose’ avtomobil’ ‘automobile’ derevo ‘tree’ lico ‘face’ rjabina ‘ash-tree’ zarja ‘dawn’ zvezda ‘star’ karandaš ‘pencil’ odežda ‘clothes’ ruþka ‘pen’ šarf ‘scarf’

FREQ. 38 34 30 30 28 22 18 16 12 11 10 10 10 9 9 8 8 7 5 5 5 5 5 4 4 4 4

WORD serdce ‘heart’ glaz ‘eye’ guby ‘lips’ jubka ‘skirt’ morkovka ‘carrot’ apel’sin ‘orange’ Ded Moroz ‘Father Frost’ fonar’ ‘lantern’ ikra ‘caviar’ jagoda ‘berry’ kraska ‘paint’ list ‘leaf’ Mars mjaþ ‘ball’ pidžak ‘jacket’ plamja ‘flame’ platok ‘shawl’ pomada ‘lipstick’ rubaxa ‘shirt’ šapka ‘cap’ šar ‘sphere’ šþeki ‘cheeks’ šljapa ‘hat’ svetofor ‘traffic light’ tomat ‘tomato’ znak ‘sign’

FREQ. 4 3 3 3 3 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2

In order to compare prototypicality effects qua head-nouns in the colour term krasnyj ‘red’ and the dimensional adjectives under study, I calculated the proportion of the three most frequently elicited head-nouns to the total number of nouns elicited in the relevant sense (colour and vertical size). As is evident from Table 5, the nouns naming the prototypical referents of krasnyj ‘red’ were less frequently elicited than the prototypical head-nouns of the dimensional adjectives. The observation that krasnyj ‘red’ displays less prototypicality effects qua head-nouns is

Elena Tribushinina

119

also confirmed by the smoothly declining ‘red’ line in Figure 1, as compared to the steeper slopes of the dimensional adjectives (only the first ten elicitations are included in the graph). Table 5. Proportion of prototypical nouns with respect to all relevant uses ADJECTIVES Vysokij ‘high/tall’

NOUNS þelovek ‘man’ dom ‘house’ derevo ‘tree’ þelovek ‘man’ dom ‘house’ derevo ‘tree’ rost ‘stature’ trava ‘grass’ dom ‘house’ pomidor ‘tomato’ cvet ‘colour’ jabloko ‘apple’ solnce ‘sun’

Nevysokij ‘not.high’ Nizkij ‘low/short’ Krasnyj ‘red’

% 18.3 16.9 12.5 20 16 12.5 15.9 10.9 10.9 8.2 7.3 6.4 6.4

80 70 Frequency

60 Vysokij 'high/tall'

50

Nevysokij 'not.high'

40

Nizkij 'low/short'

30

Krasnyj 'red'

20 10 0 1

2

3

4

5

6

7

8

9

10

Head-nouns

Figure 1. Prototypical head-nouns: dimensional adjectives vs. colour terms

One possible reason for greater prototypicality qua head-nouns in dimensional adjectives is that almost all physical entities can have colour, but only a very specific topological type of entities may be said to have height. Prototypical head-nouns of adjectives, such as vysokij ‘high/tall’ and nizkij ‘low/short’, denote objects made of rigid material, having a canonical vertical orientation, a profound vertical extension, and a point of

120

Chapter Eight: On the Prototypicality of Dimensional Adjectives

attachment at the bottom (Apresjan 2000; Vogel 2004). The more of these properties an entity possesses, the more likely it is to be dubbed vysokij ‘high/tall’, nevysokij ‘not.high’ or nizkij ‘low/short’. And, conversely, if an object lacks the relevant properties, it is only marginally suitable for descriptions in terms of height. This finding bolsters the claim that the distinction between colour terms and dimensional adjectives based on the prototypicality of the former and non-prototypicality of the latter maintained by Kamp and Partee (1995) collapses on closer scrutiny.

Prototypicality qua best exemplars As indicated above, a second aspect of prototypicality applicable to adjectives is that some entities are given a status of the best exemplars of the property in question. For instance, blood is often treated as the best exemplar of the focal red colour (Wierzbicka 1990). Fire, ripe tomatoes, and fire-engines are sometimes also conceptualised as best exemplars of redness (Tribushinina 2006). In order to find out whether dimensional adjectives display a similar kind of prototypicality, the subjects of the survey were asked to continue three expressions: krasnyj kak … ‘as red as …’, vysokij kak … ‘as high as …’ and nizkij kak … ‘as low as …’. This construction was chosen, because it is strongly associated with the prototype-oriented meaning, i.e. the noun usually names an entity that is considered to be a best exemplar of the property denoted by the adjective. Tables 6 and 7 show the frequencies of nominals elicited on this task for the adjectives vysokij ‘high/tall’ and nizkij ‘low/short’. Perhaps the most remarkable observation to be made here is that in this type of prototypicality, nizkij ‘low/short’ has a more clear-cut set of prototypes than vysokij ‘high/tall’. So, in the construction nizkij kak X ‘as low as X’, the noun trava was elicited in almost one third of the cases (29%). The second most frequent prototype of lowness yielded by this task was a treestump (26%), followed by a dwarf (11%). These three prototypical entities together constitute 60% of the nouns elicited on this task.

Elena Tribushinina

121

Table 6. Best exemplars of vysokij ‘high/tall’ WORD stolb ‘post/pole’ bašnja ‘tower’ gora ‘mountain’ neboskreb ‘skyscraper’ dub ‘oak-tree’ žiraf ‘giraffe’ derevo ‘tree’ djadja Stepa ‘uncle Stepa’ kalanþa ‘watchtower’ dom ‘house’ skala ‘rock’

FREQ. 22 15 14 14 12 12 11 10 8 7 5

WORD nebo ‘sky’ papa ‘dad’ škaf ‘closet’ telebašnja ‘TV-tower’ zabor ‘fence’ brat ‘brother’ kiparis ‘cypress’ sosna ‘pine-tree’ stena ‘wall’ velikan ‘giant’

FREQ. 4 3 3 3 3 2 2 2 2 2

Table 7. Best exemplars of nizkij ‘low/short’ WORD trava ‘grass’ pen’ ‘stump’ karlik ‘dwarf’ gnom ‘gnome’ grib ‘mushroom’ plintus ‘plinth’

FREQ. 50 44 20 9 8 7

WORD stol ‘table’ lilliput ‘Lilliputian’ gazon ‘lawn’ kust ‘bush’ pol ‘floor’ stul ‘chair’

FREQ. 6 5 2 2 2 2

These results are consonant with the findings from the corpus study, in the sense that grass is the most prominent best exemplar of lowness attested in the RNC. The prototype status of grass in the domain of low size motivated the idiomatic expression tiše vody niže travy (lit. quieter than water, lower than grass) ‘meek as a lamb’; see example (1). This idiom is motivated by the idea that one cannot be quieter than the prototypically quiet entity – water; nor can one be lower than the prototypically low entity – grass. (1)

Veþno eternally

sidevšij v sitting-PST.SG.M.NOM in

pravitel’stve government-LOC

tiše vody quieter water-GEN

travy grass-GEN

glava head-NOM

Anvar Anvar-NOM

Šamuzafarov Shamuzafarov-NOM

niže lower

Gosstroja state.building-GEN vdrug suddenly

122

Chapter Eight: On the Prototypicality of Dimensional Adjectives

udostoilsja was.worthy-M.PFV

þesti honour-GEN

byt’ be-INF

raskritikovannym criticised-SG.M.INS

liþno personally

prezidentom. (RNC) president-INS

‘Anvar Shamuzafarov, the head of the State Department of Buildings and Facilities, who had always been as meek as a lamb, was honoured with the criticism directly from the president.’ The figures for vysokij ‘high/tall’ in Table 6 are not as pronounced as the results for nizkij ‘low/short’. This could be related to the fact that the greater salience of bigger sizes leads to a more elaborated series of prototypes. If we count different tree sorts and different instances of towers together, the percentages of the most prominent prototypes of vysokij ‘high/tall’ take the following form: trees – 17.9%, posts – 12.7%, towers – 11%, mountains – 8.7%, skyscrapers – 8%, and giraffes – 6.9%. Notice that two of these best exemplars – towers and mountains – were also attested in the RNC. Witness (2) and (3): (2)

Proplyvajut float-PRES.3.PL

vysokie, high-PL.NOM

priþeski coiffures-NOM

dam. (RNC) ladies-GEN

kak like

bašni, towers-NOM

‘Lady hairdos as high as towers are floating by my side.’ (3)

Pol’zujas’ using

sluþaem, occasion-INS

xoþu want-PRES.1.SG

poželat’, wish-INF.PFV

þtoby ơta so.that this-F

družba friendship-NOM

ostavalas’ stayed-F.IPFV

veþnoj, eternal-SG.F.INS

veþnozelenaja evergreen-SG.F.NOM

sosna, pine-NOM

kak like takoj such-SG.F.INS

Elena Tribushinina

že PCL

vysokoj, high-SG.F.INS

podobnoj similar-SG.F.INS

kak like

123

gory, mountains-NOM

reke, river-DAT

tekušþej flowing-SG.F.INS

daleko za gorizont. (RNC) far behind horizon-ACC ‘I would like to seize the opportunity and wish this friendship to live forever like evergreen pines, to stay as high as mountains, and to be like a river flowing far beyond the horizon.’ Further evidence of prototypicality qua best exemplars in the domain of dimensional adjectives comes from Tribushinina (2008c), a study demonstrating that best exemplars of “highness”, such as towers, are the most prominent referent categories of the English tall and the Dutch hoog ‘high’ both in child-directed speech and in child speech. This finding is in accord with a basic idea of prototype theory that prototypes are acquired before non-prototypical category members and constitute the semantic core of a word in adult language (Rosch 1971). Remarkably, towers are only rarely described by means of tall and hoog in adult language that is not directed to children, due to redundancy of this modification. For example, only in 66 cases out of 4,941 occurrences (i.e. 1.3%) in the British National Corpus tall is used to describe towers. Likewise, only in 2 cases out of 2,470 occurrences (i.e. 0.08%) in the Corpus of Spoken Dutch, hoog is used with reference to towers. However, when it comes to communication with children, adults frequently use adjectives such as tall and hoog to describe towers. Towerrelated uses constitute 32% of the instances of tall in the Brown corpus (Brown 1973) and Manchester corpus (Theakston et al 2001) from the CHILDES database (MacWhinney 2000). Similarly, tower-related dimensional uses of hoog in the Groningen corpus (Bol 1995) and Van Kampen corpus (Van Kampen 1994) account for 36% of all tokens of hoog in the child-directed speech. This finding strongly suggests that parents intentionally use their culturally shared knowledge of prototypically tall entities, such as towers, to explain the meaning of adjectives such as tall and hoog to their children. In this connection it is interesting to note that English-speaking subjects in the experiment reported in Dirven and Taylor (1988) judged towers as prototypically tall entities. This reinforces the conclusion that best exemplars from child

124

Chapter Eight: On the Prototypicality of Dimensional Adjectives

language are not erased later in development; rather the prototypical core persists in adult language and gets enriched by the periphery. Now let us compare the results obtained for the two dimensional adjectives with the figures for the colour term krasnyj ‘red’ (see Table 8). Table 8. Best exemplars of krasnyj ‘red’ WORD pomidor ‘tomato’ rak ‘crayfish’ mak ‘poppy’ krov’ ‘blood’ perec ‘paprika’ zakat ‘sunset’

FREQ. 67 32 20 12 6 6

WORD ogon’ ‘fire’ jabloko ‘apple’ arbuz ‘water-melon’ byk ‘bull’ roza ‘rose’ serdce ‘heart’

FREQ. 5 3 2 2 2 2

As evidenced by Table 8, krasnyj ‘red’ yielded a more clear-cut set of best exemplars than the dimensional adjectives vysokij ‘high/tall’ and nizkij ‘low/short’. This effect is clear from the sharp slope representing the transition from the most salient best exemplars of redness to less salient standards of comparison in Figure 2 (cf. smoothly declining lines for vysokij and nizkij). By far the most salient standard of redness given by 38.5% of the subjects was a tomato. Other entities repeatedly elicited on this task include crayfish (18.4%), poppies (11.5%), and blood (6.9%).

80 70 Frequency

60 50

Vysokij 'high/tall'

40

Nizkij 'low/short'

30

Krasnyj 'red'

20 10 0 1

2

3

4

5

6

7

8

9

10

Best exemplars

Figure 2. Best exemplars: dimensional adjectives vs. colour terms

Elena Tribushinina

125

The fact that there is more uniformity among subjects as to which objects count as prototypically red than as to which entities are prototypically high or low confirms the prediction that dimensional adjectives display a lesser degree of orientation to best exemplars than colour terms. At the same time, these results strongly suggest that the other extreme view positing that dimensional adjectives have zero prototypicality does not withstand criticism in the face of the above data. To be more precise, tomatoes were elicited as prototypes of redness in 38.5% of the cases, and grass was yielded as a prototype of lowness by 29% of the subjects. This difference is not statistically significant: Ȥ2 (1) = 3.4, p = 0.064. Further, it is not the case that there are no prominent prototypes of tallness at all. Simply, these prototypes are less salient than best exemplars of colour categories. In addition, there could be a broader range of prototypically high objects (i.e. the number of conspicuous entities with an extremely profound vertical extent), which could also reduce the overall frequency of each of them. Greater effects of best exemplars in the domain of colour terms, as compared to dimensional adjectives, are also manifested in the ability of colour terms to participate in what Ruzin (1994) calls denotative adjectives, i.e. words containing explicit reference to best exemplars of the property denoted by the adjective (e.g. blood-red, jet-black, rubine, skyblue, snow-white). In the domain of dimensional adjectives, this phenomenon is noticeably less ubiquitous than in colour terminology. For instance, no single denotative adjective derived from tall or short has been attested in the British National Corpus and in the dictionaries. The same goes for vysokij ‘high/tall’ and nizkij ‘low/short’. This could result from the relatively limited prototypicality qua best exemplars in the domain of dimensional terms, as compared to colour adjectives. However, the fact that prototype explication is not as frequent in the domain of dimensional adjectives does not mean that it does not exist at all. The adjective thin, for instance, participates in several compounds naming prototypically thin objects, such as paper-thin, pencil-thin, and wafer-thin. In summary: although there is a difference between colour adjectives and dimensional adjectives as to the prominence of best exemplars, this difference is not as big as predicted by Kamp and Partee (1995). Rather, the results strongly suggest, in line with Geeraerts, Grondelaers and Bakema (1994), that prototypicality is a matter of degree.

126

Chapter Eight: On the Prototypicality of Dimensional Adjectives

Conclusion I hope to have demonstrated that the view treating dimensional adjectives as prototype-free concepts is untenable. The analysis of two sets of data – a survey and a corpus í has shown that there are two kinds of prototypes applicable to dimensional adjectives. In the first place, an adjective may have a number of prototypical head-nouns, i.e. it can prototypically apply to a restricted number of entities. Put another way, an adjectival word cannot be equally felicitous with all head-nouns, since some entities are more likely to be described by means of a particular adjective than other (types of) entities. For instance, adjectives denoting vertical extent, such as vysokij ‘high/tall’ and nizkij ‘low/short’, prototypically describe entities made of rigid material, having a canonical vertical orientation, a profound vertical extent (which should be the maximal dimension for ‘high’), and a point of attachment at the bottom (Apresjan 2000; Vogel 2004). The more of these properties an entity possesses, the more likely it is to be dubbed vysokij ‘high/tall’ or nizkij ‘low/short’. And, conversely, the fewer of these properties it has, the less likely it is to be described by means of these adjectives. In addition, certain topological types of entities are better described by supra terms than by the corresponding sub terms, or the other way around. Another interesting finding is that the colour term krasnyj ‘red’ is to a lesser degree associated with this type of prototypicality. This is probably due to the fact that there are more combinatorial restrictions for dimensional adjectives, especially for dimensional hyponyms, such as ‘high’ and ‘low’ (thus, opposed to hyperonyms, such as ‘big’ and ‘little’), than for colour terms. The reason is that almost all physical entities can be attributed a certain colour, but only entities of a specific topological type can be said to have height. In the second place, there are certain objects that are given a status of the best exemplars of the property. These prototypes are strongly associated with the maximum on a gradual scale. It is this kind of prototypicality that is more characteristic of colour terms than of dimensional adjectives. However, counter to the claim made by Kamp and Partee (1995), dimensional adjectives score fairly high on this kind of prototypicality as well, though not as high as colour terms. All in all, the study reported in this paper has provided evidence that prototypicality is not a matter of yes-no distinction; there are degrees of prototypicality. For one, dimensional adjectives are more prototypical qua head-nouns than colour terms, whereas colour adjectives reveal more prototypicality effects qua best exemplars.

Elena Tribushinina

127

References Apresjan, J.D. 2000. Systematic lexicography. Oxford: Oxford University Press. Bol, G.W. 1995. Implicational scaling in child language acquisition: the order of production of Dutch verb constructions. In M. Verrips and F. Wijnen (Eds.), Papers from The Dutch-German Colloquium on language acquisition, 1-13. Amsterdam: Institute for General Linguistics. Brown, R. 1973. A first language: the early stages. Cambridge, MA: Harvard University Press. Croft, W. and Cruse, D.A. 2004. Cognitive linguistics. Cambridge: Cambridge University Press. Dirven, R. and Taylor, J. 1988. The conceptualisation of vertical space in English: the case of tall. In B. Rudzka-Ostyn (Ed.), Topics in cognitive linguistics, 379-402. Amsterdam and Philadelphia: John Benjamins. Geeraerts, D., Grondelaers, S. and Bakema, P. 1994. The structure of lexical variation: meaning, naming, and context. Berlin: Mouton de Gruyter. Goy, A. 2002. Grounding meaning in visual knowledge. In K.R. Coventry and P. Oliver (Eds.), Spatial language: cognitive and computational perspectives, 121-145. Dordrecht, etc.: Kluwer Academic. Kamp, H. and Partee, B. 1995. Prototype theory and compositionality. Cognition 57: 129-191. MacWhinney, B. 2000. The CHILDES project: tools for analyzing talk. Mahwah, NJ: Lawrence Erlbaum. Rosch Heider, E. 1971. “Focal” color areas and the development of color names. Developmental Psychology 4(3): 447-455. Ruzin, I.G. 1994. Kognitivnye strategii imenovanija: modusy percepcii i ix vyraženie v jazyke [Cognitive strategies in naming: perception modalitities and their expression in language]. Voprosy jazykoznanija, 6: 79-100. Taylor, J.R. 2003. Near synonyms as co-extensive categories: ‘high’ and ‘tall’ revisited. Language Sciences 25(3): 263-284. Theakston, A.L., Lieven, E.V.M., Pine, J.M. and Rowland, C.F. 2001. The role of performance limitations in the acquisition of verb-argument structure: an alternative account. Journal of Child Language 28: 127152. Tribushinina, E. 2006. Absolute and relative adjectives. Unpublished MPhil thesis, Vrije Universiteit Amsterdam.

128

Chapter Eight: On the Prototypicality of Dimensional Adjectives

—. 2008a. Cognitive reference points: semantics beyond the prototypes in adjectives of space and colour. Utrecht: LOT. —. 2008b. EGO as a cognitive reference point: the case of nevysokij and nizkij. Russian Linguistics 32(3): 159–183. —. 2008c. Prototypicality effects in the acquisition of dimensional adjectives: evidence from English and Dutch. Paper presented at the 18th Anéla Juniorendag, Tilburg (Netherlands), January 25, 2008. Van Kampen, N.J. 1994. The learnability of the left branch condition. In R. Bok-Bennema and C. Cremers (Eds.), Linguistics in the Netherlands, 83-94. Amsterdam: John Benjamins. Vogel, A. 2004. Swedish dimensional adjectives. Stockholm: Almqvist and Wiksell International. Weydt, H., Schlieben-Lange, B. 1998. The meaning of dimensional adjectives. Discovering the semantic process. Lexicology 4(2): 199236. Wierzbicka, A. 1990. The meaning of color terms: semantics, culture, and cognition. Cognitive Linguistics 1(1): 99–150.

BASIC AND NON-BASIC COLOUR TERMS IN HUNGARIAN KORNÉLIA PAPP

Abstract. A fundamentally perception-based cognitive process is investigated in this paper, namely, the prototypicality of colour terms. According to Berlin and Kay (1969), there are eleven or possibly twelve basic colour terms in Hungarian. They suggest that the basic colour terms in Hungarian are: fehér ‘white’, fekete ‘black’, piros ‘red’, kék ‘blue’, zöld ‘green’, sárga ‘yellow’, barna ‘brown’, rózsaszín ‘pink’, lila ‘purple’, narancssárga ‘orange’, szürke ‘grey’. This idea has so far been accepted in the linguistic literature. The present investigation of basic colour terms in Hungarian shows that instead of the previously assumed eleven or twelve basic colour expressions there are only ten basic terms in Hungarian. This is supported by: (i) rigorous application of the criteria given by Berlin and Kay for basic colour terms, (ii) quantitative investigation and (iii) analysis of the composed colour terms of the corpora.

Introduction The present chapter investigates the typicality of colour terms. Brent Berlin and Paul Kay argue in their classic 1969 study of world wide colour naming that colour terms can be organized into a coherent hierarchy in all languages, and that there are a limited number of universal basic colour terms that appear in a relatively fixed order in different cultures. Berlin and Kay base their analysis on the comparison of colour words in 20 languages from all over the world. Several publications have appeared on this topic since their 1969 monograph, (e.g. Corbett and Morgan 1988; Davies and Corbett 1994, 1995, 1998; Morgan 1993; Moss 1989, 1994; Moss et al. 1990; Paramei 2005). However, this chapter represents the first attempt to review the basic colour terms narancsárga ‘orange’ and rózsaszín ‘pink’ in Hungarian. According to Berlin and Kay (1969, henceforth BK), there are eleven or possibly twelve basic colour terms in Hungarian. They suggest that the basic colour terms in Hungarian are: fehér ‘white’, fekete ‘black’, piros

130

Chapter Nine: Basic and Non-basic Colour Terms in Hungarian

‘red’, kék ‘blue’, zöld ‘green’, sárga ‘yellow’, barna ‘brown’, rózsaszín ‘pink’, lila ‘purple’, narancssárga ‘orange’, szürke ‘grey’ and possibly vörös ‘dark red’. The list has been widely accepted and referred to in the linguistic literature (Pléh 1999, Kenesei 2004, Tolcsvai 2004). I propose that only ten of the above are really basic, i.e. prototypical colour terms.

Preliminaries Basic colour terms are a relatively well researched area of the lexicon and the studies on them cover a large number of languages in the world. BK argued against the Sapir-Whorf hypothesis suggested the existence of semantic universals in colour vocabulary (Kay 1997: 1; Ungerer and Schmid 1996: 3). This, however, has not remained undisputed (e.g. Roberson et al. 2006). As an important aspect, the theory also mapped out the evolutionary development of colour terms in all languages. BK formulated their theory on extensive experimental data collected from 20 languages representing samples of all major linguistic families. BK argued that every language possesses between 2 and 11 basic colour terms. They presented a hierarchy which specifies a limited number of evolutionary paths that a language can take when adding new colour categories (see Figure 1). All languages acquired their basic colour terms in a fixed sequence of evolutionary stages (1969: 14). white red black

blue green yellow

brown

pink purple orange grey

Figure 1. Paths in the development of colour terms, according to Berlin and Kay

According to BK, these stages of linguistic evolution are very strict. If a language encodes a particular colour, all colours from the previous stages have to be encoded (1969: 14). Some scholars like Barbara Saunders have challenged this theory claiming that it over-simplifies the relationship between language and colour (1992: 221). Despite the criticism, most scholars have accepted this theory making it a touchstone in colour studies that still provides the basic framework that recent studies build on (see, e.g., Özgen and Davies 1998; Sutrop 2001).

Kornélia Papp

131

BK used a set of strict criteria to define what is to be taken as a “basic colour term” in a language. A basic colour term has to exhibit the first four (i–iv) criteria. In uncertain cases, criteria (v-viii) were used to determine whether a colour term was in fact basic. The criteria defined by BK are as follows (1969:6–7): (i) It is monolexemic, i.e. its meaning is not predictable from the meaning of its parts. (ii) Its signification is not included in that of any other colour term. (iii) Its application must not be restricted to a narrow class of objects. (iv) It must be psychologically salient for informants. Indices of psychological salience include, among others, (1) a tendency to occur at the beginning of elicited lists of colour terms, (2) stability of reference across informants and across occasions of use, and (3) occurrence in the ideolects of all informants. (v) The doubtful form should have the same distributional potential as the previously established basic terms. (vi) Colour terms that are also the name of an object characteristically having that colour are suspect, for example, ‘gold’, ‘silver’, and ‘ash’. (vii) Recent foreign loan words may be suspect. (viii) In cases where lexemic status is difficult to assess [see criterion (i)], morphological complexity is given some weight as a secondary criterion (1969: 6).

Three-step analysis The basicness of two Hungarian colour names, narancssárga ‘orange’ and rózsaszín ‘pink’ is routinely accepted, but needs further study. The place of the colour term vörös ‘dark red’ among the basic colour terms in Hungarian has been questioned, because it would imply that two basic colour terms belong to the red domain. This question also needs clarification. Three separate tests were used to study the typicality characteristics of Hungarian colour terms. In Test 1, the BK criteria were applied to examine the typicality properties of twelve basic colour terms, including the disputed vörös ‘dark red’ (MacLaury, Almási and Kövecses 1997, Bogatkin-Uusküla and Sutrop 2005). In Test 2, the commonness of twelve colour names is measured using the data from the Hungarian National Corpus (HNC). Finally, in Test 3, the compositional features of the prototypical colour terms were examined in the compounds.

132

Chapter Nine: Basic and Non-basic Colour Terms in Hungarian

Test 1: Application of the BK criteria According to the original definition of Berlin and Kay, a colour term is basic if it is monolexemic, psychologically salient, not subordinate to other terms or restricted in usage. Nine of the colour names classified as basic: fehér ‘white’, fekete ‘black’, piros ‘red’, kék ‘blue’, zöld ‘green’, sárga ‘yellow’, barna ‘brown’, lila ‘purple’, szürke ‘grey’, meet these criteria, whereas both narancssárga ‘orange’ and rózsaszín ‘pink’ are morphologically complex and thus do not fulfil the prescribed requirements for basic colour terms. Narancssárga ‘orange’ is composed of the word narancs ‘orange’ and sárga ‘yellow’ and, thus, violates both the monolexemy and the hyponymy criteria. Furthermore, both narancs ‘orange’ and sárga ‘yellow’ are used to generate additional colour expressions. Similarly, rózsaszín ‘pink’, composed of rózsa ‘rose’ and szín ‘colour’, is not monolexemic either and rózsa ‘rose’ denotes the flower rose. Consequently, these colour terms fail to be classified as basic since they do not match the original criteria. On the other hand, vörös ‘dark red’ shows all the basic characteristics. It is monolexemic, its signification is not included in other basic terms, its distribution is not restricted in usage and it is psychologically salient for native speakers of Hungarian. Hence, it is essential to incorporate the colour term vörös ‘dark red’ in the analysis. In an additional test, therefore, the frequency of occurrence for narancssárga ‘orange’, rózsaszín ‘pink’ and vörös ‘dark red’ needs to be examined.

Test 2: Frequency analysis In the frequency analysis of the twelve basic colour term candidates, I have consulted the Hungarian National Corpus (HNC) of the Research Institute for Linguistics, Hungarian Academy of Sciences. The HNC is a general-aim corpus of present-day standard Hungarian. It contains approximately 187.6 million words. In the making of the frequency list for Hungarian colour words, the regional subcorpora were excluded from the research because the present study does not focus on the regional variances of colour naming. Table 1 shows the naming frequencies of the twelve examined colour terms.

Kornélia Papp

133

Table 1. Frequency of occurrence in HNC HUNGARIAN TERM FEKETE FEHÉR ZÖLD KÉK PIROS SÁRGA VÖRÖS BARNA SZÜRKE LILA RÓZSASZÍN NARANCSSÁRGA

ENGLISH TRANSLATION BLACK WHITE GREEN BLUE RED YELLOW DARK RED BROWN GREY PURPLE PINK ORANGE

FREQUENCY 18 360 16 713 12 287 8 999 8 632 7 408 6 135 4 434 3 095 2 020 1 942 491

The order of frequencies corresponds almost exactly to the BK hierarchy. The most frequent colour term is fekete ‘black’, followed by fehér ‘white’ as suggested by BK. Zöld ‘green’ and kék ‘blue’ precede most of the remaining colour terms both in the BK hierarchy and in the frequency list. Lila ‘purple’, barna ‘brown’, rózsaszín ‘pink’, narancssárga ‘orange’ and szürke ‘grey’ are the four least frequent colour names and they also come last in the BK hierarchy. One major difference, however, is that piros ‘red’ follows zöld ‘green’ and kék ‘blue’, being only the fifth most frequent in the list. In the BK hierarchy, piros ‘red’ is the third most frequent colour term followed by kék ‘blue’, zöld ‘green’ and sárga ‘yellow.’ I propose that this derangement results from the fact that piros ‘red’ and vörös ‘dark red’ share the domain of redness in Hungarian. Vörös ‘dark red’ occurs 6 135 times in the corpus while piros ‘red’ appears 8 632 times. This ranks vörös ‘dark red’ seventh among Hungarian colour names, outranking five other supposedly basic colour terms. This suggests that vörös ‘dark red’ should have a firm position among the basic colour terms. By looking at the occurrence frequencies, it appears conspicuous that narancssárga ‘orange’ and rózsaszín ‘pink’ are the least frequent colour terms with 1 942 and 491 occurrences, respectively. Therefore, they clearly represent a different degree of affiliation to the set of basic colour terms (Zadeh 1994: 78). Based on the results of Test 1, I suggest that narancssárga ‘orange’ and rózsaszín ‘pink’ should be categorised not as basic colour terms but as typical colour terms given the three following categories: (i) prototypical or basic colour terms, (ii) typical colour terms and

134

Chapter Nine: Basic and Non-basic Colour Terms in Hungarian

(iii) peripheral or untypical colour terms. Colour terms are prototypical when they match all BK criteria. Typical colour names are used often and for many objects, but can be morphologically complex and they can also contain a prototypical colour name. Untypical or peripheral colour names are restricted in usage, they refer to a concrete object and/or they are rarely used. Untypical colour terms can be compound or monolexemic. Thus, vörös ‘dark red’ is categorised as a prototypical colour name while narancssárga ‘orange’ and rósaszín ‘pink’ are typical colour terms. In the following, the behaviour of the twelve studied colour terms is examined in basic-term compounds.

Test 3: Compounding basic colour terms The compositional behaviour of the colour terms was examined in Test 3. The goal of the analysis was to discern how basic colour terms combine with each other. Table 2. Compositional features of basic colour terms in Hungarian

Kornélia Papp

135

Here again, the HNC has provided the necessary data to carry out the analysis. The figures for narancssárga ‘orange’ and rózsaszín ‘pink’ are not included in Table 2, because no compound colour terms are based on them: they have appeared less than 10 times in compound colour expressions. There are only 5 occurrences with narancssárga ‘orange’, and no more than 2 with rózsaszín ‘pink’ in HNC. The remaining 10 terms form compounds with each other as shown in Table 2. The total number of basic-term compounds is 1 072, the statistically relevant figures (>1%) are marked bold in the table. The colour terms in the columns are used as a left constituent; e.g. barna ‘brown’ in barnáspiros ‘brownish red’, and the colour names in the rows represent the right constituent of a compound term; e.g. barna ‘brown’ in szürkésbarna ‘greyish brown.’ The final row and column indicate the total occurrences in right and left positions, respectively. The distribution of the basic colour terms is illustrated in Figure 2 according to their position (left or right constituent).

Figure 2. Basic colour terms as left/right constituents in compound colour terms

Sárga ‘yellow’ is the most common left constituent, followed by szürke ‘grey’ and kék ‘blue.’ The least frequent colour terms in the left position are piros ‘red’ and fehér ‘white.’ As right constituents, barna ‘brown’ and zöld ‘green’ are the most common colour terms, while piros ‘red’ and lila ‘purple’ are the least frequent colour expressions occurring in compound colour terms. According to these indices, piros ‘red’ is the

136

Chapter Nine: Basic and Non-basic Colour Terms in Hungarian

least frequent basic colour term appearing among basic-term compounds. This again underlines that vörös ‘dark red’ and piros ‘red’ make joint use of the domain of redness, an observation, which is further supported by the fact that vörös ‘dark red’ and piros ‘red’ can compound as right constituents (i.e. as dominant constituents) with the same left constituents, e.g. tĦzpiros ‘fire red’ and tĦzvörös ‘fire dark red’, kárminpiros ‘carmine red’ and kárminvörös ‘carmine dark red’, meggypiros ‘sour cherry red’ and meggyvörös ‘sour cherry dark red’, karmazsinpiros ‘crimson red’ and karmazsinvörös ‘crimson dark red.’ The vörös ‘dark red’ variant habitually implies a darker shade in these types of compounds. Alike piros ‘red’ and vörös ‘dark red’, kék ‘blue’ and zöld ‘green’ share the green– blue domain in compounds, with kék ‘blue’ being the lighter version in compounds as opposed to zöld ‘green’, e.g. azúrkék ‘azure blue’ and azúrzöld ‘azure green’, tengerkék ‘sea blue’ and tengerzöld ‘sea green’ or türkizkék ‘turquoise blue’ and türkizzöld ‘turquoise green.’ There are, however, examples where the colour terms piros ‘red’ and vörös ‘dark red’ are not fully interchangeable or one of the expressions is far less frequent than its counterpart, e.g. tulipiros ‘tulip red’ and ?tulivörös ‘tulip dark red’, céklavörös ‘beetroot dark red’ and ?céklapiros ‘beet red.’ Although the basic colour name piros occurs less frequently in basicterm compounds, it is still more dominant than vörös. Piros appears 39 times out of 54 in the dominant right position, while vörös emerges 44 times out of 132. Vörös is a modifier in 67% of all cases. The same applies to zöld ‘green’ and kék ‘blue’, insofar as zöld ‘green’ is standing in the dominant right position in 55% of all occurrences, whereas kék ‘blue’ is the modifier in 60% of all occurrences. It is apparent that zöld and piros are also more frequent as basic colour terms. The similar perception of green and blue is not only characteristic of Hungarian, but it is fundamental in all languages, brought about by their proximity in the colour spectrum. Davies and Corbet (1995) have reported on several languages (e.g. xhosa, setswana, bothala, chichewa) where the green–blue domain was not separated into two colour names, in which case the resulting basic colour term is often translated into English as “grue”. By comparing sárga ‘yellow’ and barna ‘brown’, we obtain similar results. The darker colour brown is more often dominant, while yellow stands more frequently in the left position. Accordingly, the compound sárgásbarna ‘yellowish brown’ is 8 times more frequent than barnássárga ‘brownish yellow.’ This phenomenon is explained by the fact that neither white nor blue can express a lighter hue of the colour brown. Fehér ‘white’ rarely occurs as a left constituent, its modifying function is taken over by the adjective világos ‘light’ in compounds, while blue is

Kornélia Papp

137

perceptually not compatible with the colour brown. Their assemblage would most probably call for the colour term green instead of another shade of brown. Kékesbarna ‘bluish brown’ appears only 3 times in the corpora. The most common basic-term compounds are as follows: 1. 2. 3. 4.

vörös(es)barna ‘darkredish brown’: sárgá(s)zöld ‘yellowish green’: sárgá(s)barna, kék(es)zöld ‘yellowish brown, bluish green’: sárgá(s)fehér ‘yellowish white’:

7.36% 6.15% 5.13% 4.85%

The relative differences in occurrence (RDO) are also worth examining. The RDO of two colour terms shows how large the difference is between the frequencies of occurrence of the two colour names appearing in different positions (left or right constituent). In Table 3, the five highest and the five lowest relative differences are listed among Hungarian basic colour term compounds: Table 3. The highest relative differences in occurrence OCCURRENCE

RARE COMPOUND

OCCURRENCE

szürkésvörös ’greyish darkred’

13

vörös(es)szürke ’darkredish grey’

0

2nd

liláspiros ’purplish red’

10

piros(as)lila ’redish purple’

1

3rd

szürkésfehér ’greyish white’

16

fehér(es)szürke ’whitish grey’

3

4th

kékesfehér ’bluish white’

23

fehér(es)kék ’whitish blue’

3

5th

sárgásbarna ’yellowish brown’

55

barná(s)sárga ’brownish yellow’

7

RANK

1

st

FREQUENT COMPOUND

Based on Table 3, it is clear that szürke ‘grey’ and vörös ‘dark red’ are typically left constituents, thus they rarely form compounds with each other. When they do compound, vörös is always the dominant participant (see raw 1 in Table 3). The same can be observed in the case of the yellow–brown compounds (see raw 5 in Table 3), where barna ‘brown’ is the more prominent colour. The rare occurrence of pirosaslila ‘redish purple’ might be due to the closeness of the two colours in the spectrum, and red being more dominant as opposed to purple. The same can be

Chapter Nine: Basic and Non-basic Colour Terms in Hungarian

138

stated about fehéresszürke ‘whitish grey’, szürke ‘grey’ being whitish on its own compared to black. Further elaboration in a compound expression would be redundant in this case. The basic colour term fehér ‘white’ is rarely used in compounds as a left constituent. It only appears 26 times on the left, while 134 times in the right position. The similar behaviour of the colour names szürke ‘grey’ and kék ‘blue’ can be best observed in the lowest relative differences in occurrence. Table 4 summarises the five lowest relative differences. Table 4. The lowest relative differences in occurrence RANK

1st 2nd 3rd 4th 5th

FREQUENT COMPOUND kék(es)szürke ’bluish grey’ lilá(s)kék ’purplish blue’ kék(es)sárga ’bluish yellow’ szürké(s)zöld ’greyish green’ kék(es)zöld ’bluish green’

OCCURRENCE

42 22 14 37 55

RARE COMPOUND szürké(s)kék ’greyish blue’ kék(es)lila ’bluish purple’ sárgá(s)kék ’yellowish blue’ zöld(es)szürke ’greenish grey’ zöld(es)kék ’greenish blue’

OCCURRENCE

41 20 12 30 35

Table 4 shows that kék ‘blue’ and szürke ‘grey’ not only behave similarly as a left constituent, but they also occur with similar frequencies in the compounds built with the two colour terms (see raw 1 in Table 4). Similarly, colour names constructed of purple/blue, grey/green, blue/green are pairs whose constituents stand perceptually close to each other. However, there is one exception, as blue and yellow are paired (see row 3 in Table 4), even though a third basic colour, green, is obtained as the combination of the two. When the highest and the lowest relative differences are compared, colour dominance should be mentioned in the highest relative differences. The colour terms in the lowest relative differences are interchangeable, and the two terms stand for similar colours or they compose a third colour term (kék–sárga ‘blue–yellow’). The latter can hardly be expected for the colour terms with the highest relative differences in occurrence, which supports the high number of compounds in one direction, while it leads to a very low number of compounds in the other. The higher the relative difference is, the lower the possibility for the basic colour term to compound in both directions.

Kornélia Papp

139

In sum, a clear distinction can be made between the usage of basic and non-basic Hungarian colour terms based on how they appear in compounds. While vörös ‘dark red’ is used in compounds just like the other basic terms, narancssárga ‘orange’ and rószaszín ‘pink’ are rarely encountered. On the other hand, we have seen that although vörös ‘dark red’ is more frequent in compounds, piros ‘red’ is still the semantically dominant one.

Summary and conclusions It has been argued in this chapter that narancssárga ‘orange’ and rózsaszín ‘pink’ are not basic colour terms in Hungarian according to the “classical” BK criteria, whereas vörös ‘dark red’ (not previously classified a basic colour term) meets all the eligibility criteria. The frequencies of all potential basic colour terms have also been studied within the HNC. Vörös ‘dark red’ occurs as frequently as the other basic colour terms, whereas narancssárga ‘orange’ and rózsaszín ‘pink’ are less frequent. This leads to the conclusion that narancssárga ‘orange’ and rózsaszín ‘pink’ are not basic colour terms whereas vörös ‘dark red’ is one. The colour expressions composed of basic colour term candidates such as barnásvörös ‘brownish red’ or lilásrózsaszín ‘pinkish purple’ have also been analysed. The analysis reveals basic cognitive features in the Hungarian colour naming system. Corpus-based evidence was presented to show that a clear distinction can be made between the usage of basic and non-basic Hungarian colour terms in compounds. Furthermore, vörös ‘dark red’ is used in compounds the same way as the other basic terms are, while narancssárga ‘orange’ and rószaszín ‘pink’ are rarely encountered. It has this been shown that instead of the previously assumed eleven or twelve basic colour expressions there are only ten basic terms in Hungarian, namely fehér ‘white’, fekete ‘black’, piros ‘red’, kék ‘blue’, zöld ‘green’, sárga ‘yellow’, barna ‘brown’, lila ‘purple’, szürke ‘grey’ and vörös ‘dark red.’ Furthermore, the existence of two separate basic terms for the domain red, appears problematic for a strict, universalist interpretation of the Berlin and Key hierarchy, in line with the recent “rehabilitation” of the Sapir-Whorf hypothesis, even in the domain of colour (Roberson et al. 2005).

140

Chapter Nine: Basic and Non-basic Colour Terms in Hungarian

References Berlin, B., and Kay, P. 1969. Basic color terms: their universality and evolution. Berkeley: University of California Press. Bogatkin-Uusküla, M. and Urmas Sutrop. 2005. Kas ungari keeles on kaks punase värvi põhinime piros ja vörös? Emakeele Seltsi Aastaraamat 50: 93-110. Corbett, G. and Morgan, G. 1988. Colour terms in Russian: reflections of typological constraints in a single language. Journal of Linguistics 24: 31-64. Davies, I. and Corbett, G. 1994. The basic color terms of Russian. Linguistics 32: 65-89. Davies, I. and Corbett, G. 1995. A practical field method for identifying basic colour terms. Languages of the World 9(1): 25-36. Davies, I., G. Corbett. 1998. A cross-cultural study of colour grouping: tests of the perceptual-physiology account of color universals. Ethos 26(3): 338-360. Tolcsvai, N.G. 2004. A nyelvi kategorizáció kognitív nyelvészeti keretben. Manuscript. Budapest. Ungerer, F. and Schmid, H.J. 1996. An introduction to Cognitive Linguistics. London/New York: Longman. Kay, P., Berlin, B., Maffi, L., and Merrifield, W. 1997. Color naming across languages. In C.L. Hardin and L. Maffi (Eds.), Color categories in thought and language, 21–56. Cambridge: Cambridge University Press. Kay, P. 1997. Color categorization. In R. Wilson and F.C. Keil (Eds.), MIT Encyclopedia of the Cognitive Sciences, Cambridge: MIT Press. Kenesei I. 2004. A nyelv és a nyelvek. Budapest: Akadémiai Kiadó. MacLaury, R.E., Almási J. and Kövecses Z. 1997. Hungarian Piros and Vörös: Color from points of view. Semiotica 114: 67-81. Morgan, G. 1993. Basic colour terms: comparative results for French and Russian. French Language Studies 3: 1-17. Moss, A.E. 1989. Basic colour terms: problems and hypotheses. Lingua 78(3): 13–320. Moss, A.E., Davies, I., Corbett, G. and Laws, G. 1990. Mapping Russian colour terms using behavioural measures. Lingua 82: 313-332. Özgen, E. and Davies, I. 1998. Turkish color terms: tests of Berlin and Kay's theory of color universals and linguistic relativity. Linguistics 36: 919-956. Paramei, G.V. 2005. Singing the Russian blues: an argument for culturally basic color terms. Cross-Cultural Research 39(1): 10-34.

Kornélia Papp

141

Pléh Cs. 1999. Hozzájárulhatnak-e az empirikus pszichológiai kutatások a nyelv–gondolkodás viszony filozófiai problémájának megoldásához? In Neumer Katalin (Ed.), Nyelv, gondolkodás, relativizmus, 35-165. Budapest: Osiris Kiadó. Roberson, D., Davidoff, J., Davies, I. and Shapiro, L.R. 2005. Color categories: Evidence for the cultural relativity hypothesis. Cognitive Psychology 50: 378-411. Saunders, B.A.C. 1992. The Invention of Basic colour terms. Utrecht: Interdisciplinary Social Science Research Institute. Sutrop, U. 2001. List task and a cognitive salience index. Field Methods 13(3): 263-276. Zadeh, L.A. 1994. Fuzzy logic, neural networks, and soft computing. Communications of the ACM 37(3): 77-84.

EVERYDAY EXPERIENCE IN WORD MEANING: HOW IT IS REVEALED VIA AN ASSOCIATIVE EXPERIMENT SVITLANA MARTINEK

Abstract. The cognitive approach to studying languages creates new opportunities for investigating conceptual structures of human consciousness via their language manifestations. Until recently, most of the assumptions within the field have been based on the linguistic introspection and/or intuition of the researcher. Although these “phenomenological methods” are quite informative, nowadays many proponents of Cognitive Linguistics desire to expand their investigative repertory incorporating empirical methods. In particular, the psychological relevance of the linguistic analyses may be supported by the application of the associative experiment method as responses evoked by a certain stimulus can be seen as the reflection of corresponding conceptual structures. This chapter describes the results of an associative experiment that was conducted with Ukrainian native speakers. The responses evoked by the derivatives represented by the Ukrainian words siryj ADJ 'grey', sirist N 'grey-ness', sirity V IMPERF 'to grow grey', and posirity V PERF ‘to become grey’ which conceptualise colour sensation as an attribute, an abstract entity and a process respectively, discover specific domains and reveal distinctive prototypical reference points for each of the stimuli. These semantic modifications accompanying the changes in grammatical conceptualization of the grey colour in the form of an adjective, noun, and verbs of different aspects can be explained as having derived their origin from our everyday experience, namely from the knowledge of specificity of the signified colour sensations.

Introduction The cognitive trend in linguistics has opened new perspectives in studying linguistic phenomena and has enabled solving problems which could not be explained exhaustively within the traditional approaches. As it is in the paradigm of Cognitive Linguistics that mechanisms of language functioning are researched in a way closely interconnected with the

Svitlana Martinek

143

principles of cognitive organization, this approach makes it possible to penetrate peculiarities of conceptual categories ingrained in human consciousness via their language manifestations. Since the “set of grammatically specified concepts provides the basic schematic framework for conceptual organization within the cognitive system of language” (Talmy 2003: 21), it is necessary to bear in mind the demand of psychological relevance while studying the specific grammatical organization of language. However most of the assumptions within the field of Cognitive Linguistics have been so far based on linguistic introspection. Though it is a powerful form of empirical support, “the findings resulting from introspection must be correlated with those resulting from other methodologies” (Talmy 2003: 5) including the analysis of introspective reports by others, the analysis of discourse and corpora, cross-linguistic and diachronic analysis, the assessment of context and of cultural structure, the observational and experimental techniques of psycholinguistics, etc. Moreover, meaning is considered in Cognitive Linguistics as reflecting our overall biological, cultural and social experience as human beings. But “[m]embers of the same discourse community all have different biographies and life experiences, they may differ in age, gender, or ethnicity, they may have different political opinions” and thus culture is, in principle, heterogeneous (Kramsch 1998: 9–10). Therefore, the results obtained via introspection and intuition of a single person, even though s/he is a good expert in the field, are not representative enough.1 Nevertheless empirical methods are far from being widely exploited yet by researchers due to a certain bias against the acceptance of empirical methodology. In this regard Geeraerts and Cuyckens (2007: 18) point out that “[t]here seems to exist a tension… between a broad methodological tendency in Cognitive Linguistics that considers introspection the most or perhaps the only appropriate method for studying meaning and a marginal but increasing tendency to apply empirical methods that are customary in the other cognitive sciences”. This article purports to show that the data obtained via an associative experiment (AE), which meet the requirements of psychological relevance, allow us to reveal conceptual content of grammatical structures that cannot be explained exclusively based on linguistic grounds but is most probably motivated by our everyday experience. 1

See Itkonen (2009) for a discussion of the difference between introspection, which is subjective and intuition, which is intersubjective in being a form of “conventionalized empathy”.

144

Chapter Ten: Everyday Experience in Word Meaning

Justification of the associative experiment method It is generally accepted that language data are a unique source of information for the reconstruction of deeper conceptual structures. Fauconnier (1997: 1) claims that visible language is only “the tip of the iceberg of invisible meaning construction”. Therefore a need for “a theory of semantics that can take conceptual organization into account” (Sweetser 1996: 6) makes obvious the necessity of accumulating evidences from the experimental methods. The chapter describes the results obtained via an AE conducted with Ukrainian native speakers and presented in the Ukrainian Associative Thesaurus (Martinek 2007). The traditional way to conduct such an experiment is to show or say a word (stimulus) to respondents, and then ask them to say what other word (response) comes first to their minds after receiving the stimulus. With respect to the goal of the investigation, which is to reveal the peculiarities of grammatically specified concepts, the list of stimuli included the Ukrainian words sir-yj ADJ ‘grey’, sir’-ist’ N ‘greyness’, sir-ity V IMPERF ‘to grow grey’, and posir-ity V PERF ‘to become grey’, which conceptualise colour sensation as an attribute, an abstract entity and a process, respectively. These stimuli were presented to 200 persons aged between 18 and 60, representing both genders in equal numbers. The working hypothesis is that the associative network is not arbitrary but to a large extent is motivated as a reflection of hierarchical conceptual structures in the consciousness of speakers. As “there are relatively wellentrenched mental routines consisting of conventional pairing of form and meaning” (Evans 2007: 16), after a lexical sign is perceived, an appropriate fragment of the complicated conceptual structure with its specific features and associated emotions and evaluations becomes fully or partially activated. Thus, the responses evoked by a certain stimulus can be seen as the external manifestation of corresponding conceptual structures and their specific traits. In addition, since “predication typically invokes multiple domains, which characterize different aspects of the profiled entity” (Langacker 2002: 62), the AE allows us not only to reveal pertinent cognitive domains, but also to rank them according to their relative prominence for the speakers. However, despite the fact that the method of AE conforms to the requirement of psychological relevance and gives a secure and reliable foundation for the theoretical assumptions of the investigators, it would be a mistake to surmise that we can completely avoid the use of “phenomenological methods”. What we are dealing with is in fact that we

Svitlana Martinek

145

take turns using the experimental method to verify the results obtained via introspection and then undertaking new acts of introspection employing also empathy and intuition to interpret the data collected via experiment.

Siryj ‘grey’ Starting from the explanations proposed in Berlin and Kay (1969), and further in the earliest works of Rosch (Heider 1972; Rosch 1973), colour terms served as the ground for prototypical research. Conducting the field experiments, Rosch confirmed the fact that the borderline between different colours is fuzzy. Prototypicality as such is the result of the physiological structure of the perceptual apparatus which causes particular colours to be perceived as focal, because the human eye is more sensitive to certain light frequencies. But this explanation is probably fairly limited (see Geeraerts 2006: 28), in particular, because it is hard to imagine how information of this kind can be used in lexicographical definitions of colour terms. Language dictionaries usually give colour terms ostensive definitions enumerating the objects of a certain colour. Let us consider the following definitions given in a comprehensive Ukrainian dictionary for the colour term siryj ‘grey’ (Busel 2004: 1130): 1) “colour between black and white; colour of ash”; “pale, with a shade of such colour (about a face; about a person with such a face)”; “with fur which has a colour of ash (about animals)”; “an animal of grey colour (about an ox, a horse, a wolf, a hare etc.)”; “with grey hair (about hair)”; “not bright, dim”; 2) “cloudy, gloomy (about weather, morning, day etc.)”. First of all, we should pay attention to the fact that grey is a mixture of colours. Wierzbicka (1990) notes that in the case of mixed colours one of the components is more powerful, and as such grey is felt to be a version of black instead of white.2 As grey is a colour of black mixed with white, it should has some properties in common with the above-mentioned colours. On the one hand, Wierzbicka notes (1990) that the semantic structures of black and white reflect their interconnection with the concepts of ‘dark’ and ‘light’, and on the other hand, if there are standards of ‘darkness’ or ‘lightness’, it is necessary to search for them in the darkness of the night or in the daylight. Therefore, Wierzbicka claims that there is a close interconnection between ‘black’ and ‘dark’, and also between ‘white’ and ‘light’. It does not mean that people think of a day as “something white”, and of night as “something black”, but potential 2

In this regard, Tokarski (2004: 57) says that grey is a blackness that is less black.

146

Chapter Ten: Everyday Experience in Word Meaning

conceptual coherence is supposed. The associative connection between black and night is more obvious and transparent than between white and day (because it is possible to see many different colours in daytime), but interrelation between white and good visibility intuitively seems conclusive. In this regard, Tokarski distinguishes qualitative and quantitative aspects for black and white: this difference is less important for black because its both reference points are connected with night, but is more important for white, whose qualitative aspect is based on the colour of snow and quantitative aspect is connected with day. Further, Tokarski (2004: 59) draws the conclusion that this quantitative aspect of white understood as an amount of light is very important for grey, and as such, grey is a colour between the blackness of the night and the whiteness of the day. It is possible to verify the theoretical assumptions of these investigators and their conclusions on the basis of linguistic introspection using the data obtained via the AE. The responses evoked by the stimulus grey reveal its connection with black and dark (1a) as well as with white (1b), which is stronger in the case of black and dark: (1a) (1b)

chornyj ‘black’ 4%, temnyj ‘dark’ 2% bilyj ‘white’ 3%

6% 3%

Other responses that are metonymical in nature reveal day (obviously cloudy) as a reference point for grey (2a), but also periods of time between day and night (2b). Grey is also associated with clouds, fog or sky that is overcast with clouds (3a,b). (2a) (2b) (3a) (3b)

den’ ‘day’ 27%, nepohanyj den’ ‘not a bad day’ 1% ranok ‘morning’ 5%, vechir ‘evening’ 2% tuman ‘fog’ 5%, doshch ‘rain’ 2%, khmara ‘cloud’ 1%, pohana pohoda ‘bad weather’ 1%, pokhmuryj ‘gloomy’1% nebo ‘sky’ 2%, nebokraj ‘skyline’ 1%

28% 7% 10% 3%

Another important reference point is revealed by the responses (4) listing different animals: (4)

vovk ‘wolf’ 31%, zajets’ ‘hare’ 10%, kit ‘cat’ 6%, kotyk ‘cat (diminutive)’ 1%, mysha ‘mouse’ 3%, mysh ‘mouse’ 2%, myshka ‘mouse (diminutive)’ 2%, pes ‘dog’ 2%, kazhan ‘bat’ 1%, krolyk ‘rabbit’ 1%, slon ‘elephant’ 1%, zhuravel’ ‘crane’ 1%

61%

Svitlana Martinek

147

In her earliest work (Heider 1972), Rosch confirmed experimentally the fact that the borderline between different colours is fuzzy. Thus, the list of grey animals could be continued. In this regard, we should pay attention to the frequencies of the obtained responses. The frequencies mentioned here are frequencies of occurrence of words, evoked by a certain stimulus. As Geeraerts says (2006: 41), such “[f]requency of linguistic occurrence may be a heuristic tool in the pinpointing of prototypes, but it is not the source of prototypicality as meant in the statistical hypothesis”. So, there are two prototypical reference points for the Ukrainian siryj ‘grey’. The first one, which is quantitative grey in Tokarski’s terminology, is connected with tuman ‘fog’, doshch ‘rain’, khmara ‘cloud’, etc., as well as with den’ ‘day’, ranok ‘morning’, and vechir ‘evening’, in other words with circumstances that cause complicated visual perception. The second one, which is qualitative grey in Tokarski’s terminology, are revealed by the responses denoting the animals covered with grey fur, but among them we did not come across either ox or horse, which are proposed as reference points in the above-mentioned dictionary definition. The most frequent responses are vovk ‘wolf’, then zajets’ ‘hare’, kit ‘cat’, and mysha ‘mouse’. The respondents who had given those responses received coloured photos of a wolf, a hare, domestic cat and a mouse after the experiment. They had to answer the question of what colour those animals were. Though the animals in the photos were strictly speaking many-coloured, the respondents answered that those animals were of grey colour. In this case, we are dealing with the conventional epithet originating from fairy tales. As Geeraerts (2006: 47) notes, “[t]aking the cognitive, experiential, encyclopedic nature of linguistic signs seriously should not imply looking only for strictly conceptual explanations. Language is not just content: it is also form, and its formal side has an expressivity of its own, which does seem to create lexical configurations that can hardly be explained if we only take into account the conceptual expressivity of language”. As we can guess, this “folk” grey retains traces of the old colour system, which was less differentiated.3 It also sounds plausible that the usage of siryj ‘grey’ for signifying a barely defined colour is coherent with its connection with complicated visibility.

3

The Etymological Dictionary of the Ukrainian Language (Melnychuk 2006: 256) establishes a connection between Ukrainian siryj ‘grey’ and Indo-European *ʅoi-ro ‘dark, grey, brown’.

148

Chapter Ten: Everyday Experience in Word Meaning

Natural (5) as well as artificial objects (6) can also be of grey colour, but due to the low frequency of these responses none of them in an obvious way reveals any prototype of the category siryj ‘grey’. (5) (6)

zeml’a ‘earth, ground’ 1%, kamin’ ‘stone’ 1%, lis ‘forest’ 1%, more ‘sea’ 1% od’ah ‘clothes’ 3%, olivets’ ‘pencil’ 2%, asfal’t ‘asphalt’ 1%, chobit ‘boot’ 1%, doroha ‘road’ 1%, hlechyk ‘jug’ 1%, dim ‘house’ 1%, khlib ‘bread’ 1%, kost’um ‘suit’ 1%, pidzhak ‘coat, jacket’ 1%, pistolet ‘pistol’ 1%, platt’a ‘a dress’ 1%, trotuar ‘sidewalk’ 1%

6% 16%

The sense ‘pale, with a shade of such colour (about a face; about a person with such a face)’ was revealed by the response in (7) and indirectly by the responses in (8). There are also responses denoting not only skin but also eyes (9) or hair (10), although in the Ukrainian language there is a special word syvyj for grey hair which is more common; thus, in this sense both siryj and syvij can be considered synonyms. (7) (8) (9) (10)

blidyj ‘pale’ 1% mertvyj ‘dead’ 1%, vmyraju ‘(I am) dying’ 1% ochi ‘eyes’ 2%, moʀ ochi ‘my eyes’ 1% voloss’a ‘hair’ 1%

1% 2% 3% 1%

Another set of the responses allows us to reveal secondary meanings of siryj ‘grey’ described in (Busel 2004: 1130) as “nothing remarkable, inexpressive; deprived of novelty; monotonous” (11) and “deprived of expressiveness, brightness, originality” (12). This secondary meaning seems to be motivated by the mentioned connection between grey colour and bad, complicated visibility as well as with associated emotions and evaluations (13). (11) (12) (13)

budennyj ‘everyday’ 2%, buden’ ‘week-day’ 1%, shchodennyj ‘daily’ 1%, zvychajnyj ‘usual, ordinary’ 1% nepomitnyj ‘unnoticeable, imperceptible’ 3%, nijakyj ‘any, not good at all, characterless’ 2%, nevyraznyj ‘inexpressive, indistinct’ 1% nudnyj ‘dull, boring’ 2%, sumnyj ‘sad’ 2%, beznadija ‘hopelessness’ 1%, nastrij ‘mood, spirits’ 1%, nedobryj ‘not good, unkind’ 1%, nepryjemnyj ‘unpleasant’ 1%, pokhmuryj ‘gloomy’ 1%, skuka ‘boredom’ 1%, zlo ‘the evil’ 1%

5% 6% 11%

Svitlana Martinek

149

Another secondary meanings, motivated by the above-mentioned senses of the word siryj ‘grey’, are described in the dictionary (Busel 2004: 1130) in the following way: “the one who doesn’t belong to an exclusive class, a privileged social group”; “uneducated, lacking culture” (see responses in 14a,b). (14a) (14b)

bomzh ‘homeless person’ 1% cholovik ‘man, person’ 2%, puste misce literally ‘empty place’ (about a person) 1%, typ ‘fellow, character’ 1%

1% 4%

As Tokarski (2004: 57-58) notes in this respect, grey is the most colourless colour and that is why grey and colourless can be synonyms, and also can signify a lack of vivid intellectual or emotional qualities in a person or phenomenon; thus, Tokarski underscores: “Grey person is a synonym to an absolute mediocrity, absence of any prominent traits” (my translation. – S.M.).

Sirist’ ‘greyness’ Now let us consider the derivatives of the word grey. The Ukrainian dictionary describes the noun sirist’ ‘greyness’ (Busel 2004: 1130) as “the abstract noun for grey” in all its meanings. Nevertheless the responses evoked by the stimulus sirist’ ‘greyness’ demonstrate that the salience of the phenomena associated with sirist’ ‘greyness’ differ from those associated with siryj ‘grey’. Part of the obtained responses reveals its interconnection with other colours. Surprisingly, the responses white, whiteness (15a,b) are more frequent than the responses black, blackness, which could be explained by the fact that these responses are based on contrast (see also 16). (15a) (15b) (16)

bilist’ ‘whiteness’ 1,4%, bilyj ‘white’ 0,9% chorne ‘black’ 0,5%, chornist’ ‘blackness’ 0,5% jaskravist’ ‘brightness’ 0,5%, kontrast ‘contrast’ 0,5%

2,3% 1% 1%

A significant part of the evoked responses enumerates atmospheric phenomena (17a,b), which on the one hand are perceived as being grey and on the other hand interferes with visual perception (17c), as well as other accompanying occurrences (18). The responses also directly or indirectly reveal a connection between sirist’ ‘greyness’ and darkness (19a,b).

150

Chapter Ten: Everyday Experience in Word Meaning

The amount of the responses listing natural objects other than clouds and so on (20a), animals (20b), as well as artificial objects (20c), is significantly decreasing (compare to 4-6). (17a)

(17b) (17c) (18)

(19a) (19b) (20a) (20b) (20c)

tuman ‘mist, fog’ 4,1%, mr’aka ‘fog, drizzle’ 0,9%, pohana pohoda ‘bad weather’ 0,9%, tumannist’ ‘mist, fog’ 0,9%, dym ‘smoke’ 0,5%, imla ‘haze, mist’ 0,5%, khmary ‘clouds’ 0,5%, mr’achnist’ ‘fog, drizzle’ 0,5%, tumanna ‘misty, foggy’ 0,5% nebo ‘sky’ 0,5% neprohl’adna ‘impenetrable’ 0,5% doshch ‘rain’ 4,6%, syrist’ ‘dampness’ 0,9%, mokro ‘(it is) wet, wetly’ 0,9%, mokrota ‘wetness’ 0,9%, voloha ‘moisture’ 0,9%, boloto ‘marsh, bog; backwater’ 0,5%, doshchovyj ranok ‘rainy morning’ 0,5%, osin’ ‘autumn’ 0,5%, vohkist’ ‘humidity’ 0,5%, vohko ‘humidly, it is damp’ 0,5%; kholodno ‘coldly; it is cold’ 0,5%, voda ‘water’ 0,5% pidval ‘basement, cellar’ 0,9%, temnota ‘darkness’ 0,9%, t’m’anist’ ‘dimness’ 0,5% vechir ‘evening’ 0,9% zeml’a ‘earth, ground’ 1,4% myshka ‘mouse (diminutive)’ 0,5%, sirko ‘dog’ 0,5% asfal’t ‘asphalt’ 0,5%, pidloha ‘floor’ 0,5%, forma (vijs’kova) ‘uniform (service)’ 0,5%, stina ‘wall’ 0,5%

9,3%

0,5% 0,5% 11,2%

2,3% 0,9% 1,4% 1% 2%

A considerable part of the responses reveals some human characteristics. It can be the colour of eyes or skin (21a) but also moral, intellectual and other traits (21b). Moreover, if the former signify grey colour of the objects, the latter are used in a secondary meaning and are motivated by the indistinctness or imperceptibleness of sirist’ ‘greyness’ as revealed by the responses in (22). (21a) (21b)

(22)

ochej ‘of eyes’ 0,9%, blidist’ ‘paleness’ 0,5%, blidyj ‘pale’ 0,5% dusha ‘soul’ 0,5%, dushi ‘of (a) soul’ 0,5%, ja ‘I’ 0,5%, l’udyna ‘human being’ 0,5%, nyzist’ ‘baseness, meanness’ 0,5%, skromnist’ ‘modesty’ 0,5%, tykhist’ ‘quietness, gentleness’ 0,5% nevyraznist’ ‘indistinctness’ 1,4%, nepomitnist’ ‘imperceptibleness’, literally ‘unnoticeableness’ 0,5%, nevyraznyj ‘expressionless, indistinct’ 0,5%, obrazu ‘of (an) image’ 0,5%

1,9% 3,5%

2,9%

Svitlana Martinek

151

Some of the responses (23a,b) reveal a connection between sirist’ ‘greyness’ and ‘ignorance’, ‘uncertainty’. As Wierzbicka (1990) notes, the notion of vision is the key to the semantics of the words dark and light, and so far their prototypical use is connected with the environment. Taylor (2003: 33) claims that “much – perhaps most – of our knowledge of the outside world (for sighted people!) comes from vision”. Thus, we can assume that these responses are strongly motivated due to the fact that siryj ‘grey’ and consequently sirist’ ‘greyness’ is strongly associated with complicated visual perception (see also Martinek 2008). (23a)

(23b)

neznann’a ‘ignorance, absence of knowledge’ 0,9%, nejasnist’ ‘unclearness, ambiguity’ 0,5%, nevyznachenist’ ‘uncertainty, indefiniteness’ 0,5%, neutstvo ‘ignorance’ 0,5%, poserednist’ ‘mediocrity’ 0,5%, prymityv ‘primitive’ 0,5% hopy (ce taky l’udy) ‘hopypl4 (there are such people)’ 0,5%, hopnyky ‘hopnykypl3’ 0,5%

3,4%

1%

Another important property of greyness is its monotony (24). Thus, a significant number of the responses reveal the secondary meaning of sirist’ ‘greyness’ motivated by these properties (25). (24) (25)

odnomanitnist’ ‘monotony’ 0,9%, monotonnist’ ‘monotony’ 0,5% budennist’ ‘everyday dullness, routine’ 8,3%, budenna ‘everyday’ 1,9%, budni ‘weekdays’ 1,4%, budenshchyny ‘of everyday dullness’ 0,5%, of routine’, butt’a ‘existence/of existence’ 0,5%, moyi budni ‘my week days’ 0,5%, odnomanitne zhytt’a ‘monotonous life’ 0,5%, zavtra bude v mene ‘tomorrow I will have’ 0,5%, v spilkuvanni ‘in communication’ 0,5%, zhytt’a ‘life /of life’ 0,5%

1,4% 15,1%

The responses also reveal a connection between sirist’ ‘greyness’ and feelings (mostly negative) (26), illness (27), as well as old age (28). In addition, sirist’ ‘greyness’ is associated with a whole set of negatively evaluated notions (29). Part of the responses directly shows a negative attitude towards sirist’ ‘greyness’, most probably in its secondary meanings (30).

4

Rather intellectually primitive young hooligans and robbers who have their own dress code -.

152

(26)

(27) (28) (29)

(30)

Chapter Ten: Everyday Experience in Word Meaning

sum ‘grief, sadness, sorrow’ 3,2%, smutok ‘grief, sadness’ 1,4%, nudota ‘nausea; boredom’ 0,9%, pohanyj nastrij ‘bad mood’ 0,9%, bajduzhist’ ‘indifference’ 0,9%, pokhmurist’ ‘gloominess’ 0,9%, ml’avist’ ‘languidness’ 0,9%, motoroshnist’ ‘ghastliness, horribleness’ 0,5%, nud’ha ‘boredom’ 0,5%, spokij ‘calmness’ 0,5%, tuha ‘anguish’ 0,5%, uzhas ‘horror’ 0,5%, zlist’ ‘malice, fury’ 0,5% sukhota ‘consumption’ 0,5% starist’ ‘old age’ 1,4% nepryjemnist’ ‘unpleasantness’ 1,9%, bidnist’ ‘poverty’ 0,9%, bida ‘misfortune’ 0,5%, dol’a ‘fate’ 0,5%, neschast’a ‘unhappiness’ 0,5%, nevdacha ‘failure’ 0,5%, ubohist’ ‘poverty, mediocrity, wretchedness’ 0,5%, zlo ‘the evil’ 0,5% nepryjemno ‘unpleasant’ 0,9%, krasche vmerty ‘(it is) better to die’ 0,5%, nekomfortno ‘uncomfortably’ 0,5%, pohano ‘badly, it is bad’ 0,5%

12,1%

0,5% 1,4% 5,8%

2,4%

Sirity ‘to grow grey’ and posirity ‘to become grey’ As for the imperfective verb sirity ‘to grow grey’, the Ukrainian dictionary (Busel 2004: 1130) gives the following definitions: “to become, to turn grey, or to become more grey”, “to be distinguished with grey colour; to be seen (about something grey)”, “to lose the natural colour of the face, to turn pale from excitement, weariness, anger etc.”, “to dawn”. Let us consider the responses evoked by the derivative sirity IMPERF ‘to grow grey’. As in previous cases (1a,b, 15a,b), this verb evokes responses that reveal its connection with the notions of blackness and less prominent connection with white (31), darkness (the connection with light was not revealed) (32) and lack of sight (33) or bad visibility (34). (31) (32) (33) (34)

chornɿty IMPERF ‘to grow black’ 0,9%, bɿlɿty IMPERF ‘to grow white’ 0,5% temnɿty IMPERF ‘to grow (get) dark’ 2,8%, temno ‘(it is) dark’ 0,9%, t’ma ‘darkness’ 0,5% slɿpota ‘blindness’ 0,5% vdalynɿ ‘in the distance’ 1,9%, horyzont ‘horizon’ 0,5%, objekt vdalynɿ (daleko) ‘object in the distance (far)’ 0,5%; znykaty IMPERF postupovo to disappear gradually’ 0,5%

1,4% 4,2% 0,5% 3,4%

Other responses reveal the connection between sirity ‘to grow grey’ and the time between day and night (35a,b) or a day (again I think it is

Svitlana Martinek

153

cloudy) (36). The amount of the former sufficiently increases (compare 2b and 19b) in contrast to the latter (compare 2a). This fact can be explained by the difference in the conceptualization of colour sensations as an attribute and as a process, respectively, based on the speakers’ experience. (35a)

(35b)

(36)

vechɿr ‘evening’ 12,3%, nɿch ‘night’ 1,9%, nadvechɿr ‘towards the evening’ 0,5%, osɿnnɿj vechɿr ‘autumn evening’ 0,5%, sutɿnky ‘twilight’ 0,5%, vvecherɿ ‘in the evening’ 0,5%, vechorɿje literally ‘evening is coming on’ 0,5%, vechorɿty IMPERF literally ‘to be coming (about evening)’ 0,5% ranok ‘morning’ 5,2%, svɿtanok ‘day-break, dawn’ 1,4%, rano ‘early, early in the morning’ 0,5%, rozvydn’atys’a INF IMPERF literally ‘day is breaking’ 0,5%, svɿtaje literally ‘(it is) dawning’ 0,5%, vrancɿ ‘in the morning’ 0,5% den’ ‘day’ 2,4%, dnɿty IMPERF literally ‘to day, to grow day’ 0,5%

17,2%

8,6%

2,9%

The responses in (37) also reveal this connection between the stimulus sirity ‘to grow grey’ and the environment. (37)

nadvorɿ ‘outside’ 0,5%, na vulycɿ ‘in (the) street, outdoors’ 0,5%, vulyc’a ‘street’ 0,5%, za vɿknom ‘outside the window’ 0,5%

2%

As in previous cases (3a,b, 17a,b, 18), the verb sirity ‘to grow grey’ is also associated with the sky (38) and atmospheric phenomena. (38) (39)

nebo ‘sky’ 3,8%, obrɿj ‘skyline, overhead’ 0,5%, vhorɿ ‘above, overhead’ 0,5% tuman ‘fog’ 3,3%, dosch ‘rain’ 2,4%, bude dosch ‘(it)‘s going to rain’ 0,5%, hroza ‘thunderstorm’ 0,5%, khmara ‘cloud’ 0,5%, khmary ‘clouds’ 0,5%, osɿn’ ‘autumn’ 0,5%, pohana pohoda ‘bad weather’ 0,5%, pohoda ‘weather’ 0,5%, pryroda ‘nature’ 0,5%; syrɿty IMPERF ‘to become damp, wet’ 0,5%, ɿmla ‘haze, mist’ 0,5%, mr’aka ‘drizzle’ 0,5%

4,8% 6,7%

Only a few responses evoked by the stimulus sirity ‘to grow grey’ list natural objects, animals, and artefacts (40a,b,c). The amount of these responses sufficiently decreases (compare to compare to 4-6, 20a,b,c). Again, this fact can be explained by the different conceptualizations of colour sensations as an attribute, an abstract entity and as a process based on the speakers’ experience.

154 (40a) (40b) (40c)

Chapter Ten: Everyday Experience in Word Meaning zeml’a ‘earth, ground’ 0,9%, popɿl ‘ash’ 0,5% vovk ‘wolf’ 1,4%, zajec’ ‘hare’ 0,9%, jak zajec’ ‘as a hare’ 0,5%, kɿt ‘cat’ 0,5%, mysha ‘mouse’ 0,5%, myshɿ ‘mice’ 0,5% parkan ‘fence’ 0,5%

1,4% 4,3% 0,5%

A comparatively large number of the responses directly or indirectly (41a,b) reveals the above-mentioned meaning ‘to lose the natural colour of a face, to turn pale from excitement, weariness, anger etc.’ of the word sirity ‘to grow grey’ (see 7 and 21a) by signifying emotions which can cause these changes. (41a) (41b)

blɿdnɿty IMPERF ‘to turn, grow pale’ 0,5%, blɿdnuty IMPERF ‘to turn, grow pale’ 0,5%, bl’aknuty IMPERF ‘to fade’ 0,5% vɿd zlostɿ ‘with rage’ 0,9%, nɿjakovɿty IMPERF ‘to feel awkward, to be confused’ 0,5%, perezhyvaty ‘to endure, live through’ 0,5%, pokhmurnɿty IMPERF ‘to become gloomy’ 0,5%, stradaty IMPERF ‘to suffer’ 0,5%,, vɿd kohos’ ‘from somebody’ 0,5%, vɿd rozpachu ‘from despair’ 0,5%, vtrachaty IMPERF nastrɿj literally ‘to lose (good) mood; to get blue’ 0,5%, vtrachaty IMPERF shchos’ ‘to lose something’ 0,5%, z hor’a ‘with grief’ 0,5%, zlytys’ IMPERF ‘to be angry’ 0,5%

1,5% 5,9%

Additionally, the stimulus sirity ‘to grow grey’ evokes a large quantity of responses that reveal the meaning ‘to lose the natural colour of a face, to turn pale with age, from unsatisfactory physical condition or process of disintegration, damage’ (42a,b). (42a) (42b)

starɿty IMPERF ‘to grow old’ 2,4%, starɿst’ ‘old age’ 0,9%, vɿd starostɿ ‘from old age’ 0,9%, syvɿty IMPERF ‘to grey (about hair)’ 0,5% vmyraty IMPERF ‘to die’ 1,4%, pomyraty IMPERF ‘to die’ 0,9%, vɿd kholodu ‘from cold’ 0,9%, hnyty IMPERF ‘to rot, to decay’ 0,5%, sokhnuty IMPERF ‘to (get) dry’ 0,5%, tlɿty IMPERF ‘to smoulder, to decay’ 0,5%, trup ‘corpse, dead body’ 0,5%, vɿd pohanoho samopochutt’a ‘from feeling sick’ 0,5%, zanepadaty IMPERF ‘fall into decay’ 0,5 %

4,7% 6,2%

As with the previous stimuli (see 12, 22, 23a, 24, 25), the verb sirity ‘to grow grey’ also evokes responses that demonstrate its connection with monotony, routine (43a), something unnoticeable (43b), and mediocrity (43c), etc. but the amount of these responses is comparatively lower.

Svitlana Martinek (43a) (43b) (43c)

odnomanɿtnɿst’ ‘monotony’ 0,5%, rutynnyj ‘routine’ 0,5% nepomɿtnyj ‘unnoticeable, imperceptible’ 0,5%, nɿjakyj ‘not good at all, characterless’ 0,5% poserednɿst’ ‘mediocrity’ 0,5%, rehresɿja ‘regress’ 0,5%, svɿdomɿst’ ‘consciousness’ 0,5%

155 1% 1% 1,5%

Besides, the obtained responses demonstrate the connection between the stimulus sirity ‘to grow grey’ and a lack of individuality (44). (44)

masa ‘mass’ 0,5%, masa l’udej ‘mass of people’ 0,5%, stavaty IMPERF poseredynɿ ‘to stand in the middle’ 0,5%, zmɿshuvatys’ IMPERF ɿz natovpom literally ‘to mix with the crowd’ 0,5%

2%

Let us now consider the responses evoked by stimulus po-sirity ‘to become grey’. The dictionary of the Ukrainian language (Busel 2004: 892) gives the verb po-sirity PERF the following definition: “to become grey; to get a greyish shade, to brighten before a dawn (about the sky)”. The responses evoked by stimulus po-sirity ‘to become grey’ also reveal its coherence with black and white colours (45a,b), as well as with darkness (45c) and bad visibility (45d). Although the dictionary of Ukrainian pinpoints the meaning “to brighten before a dawn (about the sky)” responses demonstrate the connection between po-sirity ‘to become grey’ and not only the morning (46a) but also the evening (46b) which is stronger for the latter than for the former. (45a) (45b) (45c) (45d) (46a) (46b)

chornist’ ‘blackness’ 0,5%, chornyj ‘black’ 0,5%, pochornity PERF ‘to become black’ 0,5% pobility PERF ‘to become white’ 4,5% temno ‘(it is) dark’ 1,5%, potemnity PERF ‘to become dark’ 0,5%, temr’ava ‘darkness’ 0,5%,, sutenity IMPERF ‘to be getting dark’ 0,5% znyknuty ‘to disappear’ 0,5% vechir ‘evening’ 1,5%, vvecheri ‘in the evening’ 0,5% rano ‘early, early in the morning’ 0,5%

1,5% 4,5% 3% 0,5% 2% 0,5%

A significant part of the evoked responses show a connection between po-sirity ‘to become grey’ and the sky (47a) or atmospheric phenomena (47b). (47a)

nebo ‘sky’ 9,9%

9,9%

156 (47b)

Chapter Ten: Everyday Experience in Word Meaning khmary ‘clouds’ 1,5%, doshch ‘rain’ 1%, khmara ‘cloud’ 1%, osin’ ‘autumn’ 1%, pohoda ‘weather’ 1%, jak khmara ‘like a cloud’ 0,5%, jak osinn’e nebo ‘like autumn sky’ 0,5%, kholod ‘cold’ 0,5%, vid pohody ‘from the weather’ 0,5%

7,5%

The amount of the responses listing natural objects (48a), animals (48b) as well as artificial objects (48c) is comparatively small (see 40a,b,c). (48a) (48b) (48c)

zeml’a ‘earth, ground’ 0,5% mysha ‘mouse’ 1%, zajec’ ‘hare’ 1%, myshka ‘mouse (diminutive)’ 0,5%, kit ‘cat’ 0,5% stina ‘wall’ 1,5%, monitor ‘monitor’ 0,5%

0,5% 3% 2%

Although for posirity ‘to become grey’ the dictionary does not give the meaning “to lose the natural colour of a face, to turn pale from excitement, weariness, anger etc.”, a great number of responses directly (49a) or indirectly (49b) reveal this meaning by signifying negative emotions or other factors (49c), and the amount of this kind of responses is higher (see 41b). (49a) (49b)

(49c)

shkira ‘skin’ 0,5%, lyce ‘face’ 0,5%, na nosi ‘on (one’s) nose’ 0,5% vid zlosti ‘from malice’ 2%, strakh ‘fear, fright’ 1,5%, vid hor’a ‘from grief’ 1,5%, hore ‘grief’ 1%, sorom ‘shame’ 1%, sum ‘grief, sadness, sorrow’ 1%, khvyl’uvatys’ IMPERF ‘to worry, to be uneasy’ 0,5%, nudno ‘(it is) dull, tiresome’ 0,5%, vid hnivu ‘from anger’ 0,5%, vid smutku ‘from sadness’ 0,5%, vid stydu ‘from shame’ 0,5%, z hor’a ‘with grief’ 0,5%, z zhakhu ‘with horror’ 0,5%, zloba ‘malice, spite’ 0,5%, zl’akatys’ ‘to get scared’ 0,5% bida ‘misfortune, trouble’ 0,5%, pomylky ‘mistakes’ 0,5%, vid navchann’a ‘from studying’ 0,5%, vid utomy ‘from tiredness’ 0,5%, vid vtomy ‘from tiredness’ 0,5%, vtomytys’ PERF ‘to get tired’ 0,5%

1,5% 12,5%

3%

Furthermore, the dictionary of Ukrainian does not propose the meaning “to lose the natural colour of a face, to turn pale with age, from unsatisfactory physical condition or process of disintegration, damage” for the verb po-sirity PERF ‘to become grey’, but the responses (50a,b,c) evoked by the given stimulus demonstrate the importance of this meaning.

Svitlana Martinek (50a)

(50b) (50c)

posyvity PERF ‘to have become grey (about hair)’ 2%, syvyna ‘grey hair’ 1,5%, vid starosti ‘from old age’ 1,5%, postarity ‘to have become old’ 1%, starist’ ‘old age’ 1%, babc’a ‘granny’ 0,5%, syvyj did ‘grey grandfather, old man’ 0,5%, v starosti ‘in old age’ 0,5%, voloss’a ‘hair’ 0,5% kholod ‘hunger’ 0,5%, vava ‘sore (place)’ 0,5%, vid kholodu ‘from cold’ 0,5%, vid speky ‘from heat’ 0,5%, z holodu ‘with hunger’ 0,5%, zakhvority ‘to fall ill’ 0,5% smert’ ‘death’ 1%, pomerty PERF ‘to have died’ 0,5%, trup ‘corpse, dead body’ 0,5%, vmerty PERF ‘to have died’ 0,5%

157 9%

3% 2,5%

The amount of the responses that demonstrate the connection of verb po-sirity PERF ‘to become grey’ with routine (51a) or mediocrity (51b) is even lower than in previous cases (see 43a,b,c). (51a) (51b)

vid rutyny ‘from routine’ 0,5% v masi ‘in mass’ 0,5%

0,5% 0,5%

In addition, the responses evoked by the given stimulus allow us to discover the meaning ‘untidy, covered by dust or dirt’ (52), which is not described in the dictionary of Ukrainian. (52)

vid porokhu ‘from dust, gunpowder’ 1%, vid brudu ‘from dirt’ 0,5%, vid pyl’uky ‘from dust’ 0,5%

2%

Conclusions As shown in this chapter, the semantic modifications accompanying the changes in grammatical conceptualization cannot be explained exclusively on linguistic ground, but most probably they originate from our everyday experience, including the knowledge of the specificity of colour sensations. The responses evoked by the Ukrainian words siryj ‘grey’, sir’ist’ ‘grey-ness’, sirity IMPERF ‘to grow grey’, and posirity PERF ‘to become grey’, which conceptualise colour sensation as an attribute, an abstract entity or a process, reveal partially distinctive prototypical reference points for each of the stimuli. Moreover the frequency of the obtained responses allows us to rank them according to their relative salience for speakers. Furthermore, as “cognition is structured, not chaotic – and the apparently disorderly domain of linguistic meaning can often be shown to be structured around speakers’ understanding of a given cognitive domain” (Sweetser 1996: 12), the metaphorical use of sirity ‘grey’ and its derivatives can be explained via the peculiarities of their conceptualization and analyzed as forming an entire system.

158

Chapter Ten: Everyday Experience in Word Meaning

Thus, the method of the associative experiment conforms to the requirement of psychological relevance and may be employed to verify results obtained through introspection and intuition.

References Berlin, B. and Kay, P. 1969. Basic color terms: Their universality and evolution. Berkeley, CA: University of California Press. Busel, V.T. (Ed.). 2004. Velykyj tlumachnyj slovnyk suchasnoï ukraïns’koï movy. [Big Comprehensive Dictionary of Contemporary Ukrainian Language]. Kyiv: Irpin VTF “Perun”. Evans, V. 2007. Towards a cognitive compositional semantics: An overview of LCCM theory. In U. Magnusson, H. Kardela and A. Gáaz (Eds.), Further insights into semantics and lexicography, 11-42. Lublin: Wydawnictwo UMCS. Fauconnier, G. 1997. Mappings in thought and language. Cambridge: Cambridge University Press. Geeraerts, D. 2006. Words and other wonders. Papers on lexical and semantic topics. Berlin/New York: Mouton De Gruyter. Geeraerts, D. and Cuyckens. H. 2007. Introducing cognitive linguistics. In D. Geeraerts and H. Cuyckens (Eds.), The Oxford handbook of cognitive linguistics, 3-24. Oxford: Oxford University Press. Heider, E.R. 1972. Universals in color naming and memory. Journal of Experimental Psychology 93: 10-20. Itkonen, E. 2009. The true nature of typological linguistics. In Studies in language and cognition, J. Zlatev, M. Andrén, M. Johansson Falck, C. Lundmark (eds.), 16-26. Newcastle upon Tyne: Cambridge Scholars Publishing. Kramsch, C. 1998. Language and culture. Oxford: Oxford University Press. Langacker, R.W. 2002. Concept, image, and symbol: the cognitive basis of grammar, Second edition. Berlin/New York: Mouton de Gruyter. Martinek, S. 2007. Ukajins’kyj asociatyvnyj slovnyk [Ukrainian Associative Thesaurus]. In 2 vol. Lviv: Lviv University Press. —. 2008. A verb’s Aktionsart through the prism of the Associative Experiment. In J.-R. Lapaire, G. Desagulier and J.-B. Guignard (Eds.), Du fait grammatical au fait cognitif. From gram to mind: grammar as cognition, 83-101. Presses Universitaires de Bordeaux. Melnychuk, O.S. (Ed.) 2006. Etymologichnyj slovnyk ukraïnskoï movy [Etimological Dictionary of the Ukrainian Language]. Kyiv: Nauk. Dumka. Vol. 5.

Svitlana Martinek

159

Rosch, E. 1973. Natural categories. Cognitive Psychology 4: 328-350. Sweetser, E. 1996. From etymology to pragmatics: Metaphorical and cultural aspects of semantic structure. Cambridge: Cambridge University Press. Talmy, L. 2003. Foundations on conceptual structuring in language. In L. Talmy. Toward a cognitive semantics, Vol. 1: Concept structuring systems, 21-96. Cambridge Mass.: The MIT Press. —. 2003. Toward a cognitive semantics, Vol. 1: Concept structuring systems. Cambridge Mass.: The MIT Press. Taylor, J.R. 2003. Meaning and context. In H. Cuyckens, Th. Berg, R. Dirven and K.-U. Panther (Eds.), Motivation in language: Studies in honor of Guenter Radden, 27-48. Amsterdam: Benjamins. Tokarski, R. 2004. Semantyka barw we wspóáczesnej polszczyĨnie. Lublin: Wydawnictwo UMCS. Wierzbicka, A. 1990. The meaning of colour terms: semantics, culture, and cognition. Cognitive Linguistics 1(1): 99-150.

WHAT CAN SELF-ORGANIZING MAPS REVEAL ABOUT THE STRUCTURE OF EMOTION CONCEPTS? A CASE STUDY OF ESTONIAN TOOMAS KIRT AND ENE VAINIK

Abstract. Research into the structure of emotion concepts is complicated because concepts are part of our broader knowledge structure that is not directly accessible to consciousness. In the present chapter we report the results of an inquiry where the method of Self-Organizing Map (SOM) was used to visualize the hidden structure of the data. The inquiry consisted of two lexical tasks that addressed the concepts through symbolic and subconceptual levels according to Peter Gärdenfors’s theory of Conceptual Spaces. As both similarities and differences appeared on the two visual layouts resulting from the two tasks, an additional meta-level analysis was carried out to discover the coherent structure of emotion concepts.

Introduction The present chapter reports the results of a study into the structure of emotion concepts. Concepts are tools that help us to categorize and generalize our knowledge about the surrounding world. Concepts tie our past experience to our present interaction with the world, because they are connected to our larger knowledge structures (Murphy 2002). The theorists of symbol grounding also claim that a particular symbol system is used to carry our conceptual knowledge and that it is grounded into the sensory-motor interaction with the surrounding world (Harnad 1990). Other authors, however, say that such grounding has some difficulties in representing abstract concepts (Barsalou 2008) and that sensory-motor and conceptual processing are based on same system (Barsalou 2003). Therefore conceptual knowledge is claimed to be embodied and mapped within a sensory-motor system (Gallese and Lakoff 2005). Prinz (2005) has suggested that emotions are perceptions of our bodily states and, thus, concepts of emotion should be grounded in a sense of our bodily and mental states. Still, it remains unclear how concepts of emotion are interrelated and what kind of structure they form.

Toomas Kirt and Ene Vainik

161

Due to the inability to access emotion concepts directly they have to be approached through lexical tasks addressing emotional experience or verbal expression. In the field of psychology, a long tradition of lexical studies has led to partly controversial results and discussions on the topic of the structure of emotions vs. the structure of the emotion lexicon and their potential isomorphism (e.g., Cacioppo and Berntson 1999; Feldmann Barrett 2004; Larsen and Diener 1992; Ortony, Clore and Collins 1988). In this chapter we present a case study into the structure of the Estonian concepts of emotions, a survey was performed where participants assessed the semantics of 24 words in two different lexical tasks. The first task addressed the semantic interrelations of emotion terms (synonymy and antonymy) and the second addressed the qualities of emotional experience as measured on seven different scales. Using the theory of Conceptual Spaces (Gärdenfors 2000), which offers a geometrical model of assessing concept similarity based on spatial relations, the first task allowed us to gain access to the conceptual knowledge from the symbolic level and the second task from the sub-conceptual level. The structure of data gathered in both tasks was visualized by the method of SelfOrganizing Map (SOM) (Kohonen 2000), which has some analogies with the theoretical framework. As the results revealed two visually different topological layouts where coherence of the conceptual structure was recognizable only in general terms, another kind of analysis was necessary to discover the common conceptual structure present in the results of both tasks. The SelfOrganizing Map preserves local neighbourhood relations of similar samples from the input space and based on this assumption a method to measure the isomorphism between maps was used (Kirt, Vainik and Võhandu 2007), showing an isomorphic sub-graph of the two selforganizing layouts.

Theoretical background Our theoretical framework is based on the theory of Conceptual Spaces (Gärdenfors 2000), which distinguishes three non-competing levels of representations. In this model the most abstract level is the symbolic level on which the conceptual system is described by means of some symbol system (e.g., language). The second level is the conceptual level on which concepts are treated as points or an area in the conceptual space. The third level is the sub-conceptual level which consists mostly of inputs from sensory receptors and which forms so-called “quality dimensions” of the conceptual space.

162

Chapter Eleven: What Can Self-Organizing Maps Reveal about the Structure of Emotion Concepts?

When applied to the domain of emotions, this would mean that on the most abstract level we have emotion terms, on the conceptual level we have emotion concepts and on the sub-conceptual level we have (e.g., perceptual) qualities of emotions (Figure 1).

SYMBOLIC

CONCEPTUAL

SUBCONCEPTUAL

EMOTION NAMES

EMOTION CONCEPTS

QUALITIES OF EMOTIONS

Figure 1. Application of the three-level representational model to the domain of emotions

The Conceptual Spaces theory offers a geometrical model of assessing concept similarity based on spatial relations. Such a model where the similarity of concepts is represented by their spatial closeness is analogous with the method of SOM in which the topological layout and neighborhood relations are expected to represent homology of the input data. SOM (Kohonen 2000) is an artificial neural network that uses an unsupervised learning algorithm which does not need any prior knowledge of how a system’s input and output are connected. SOM is claimed to partly simulate the self-organizing learning processes that take place in the human brain (Kohonen 2000: 104). The SOM algorithm projects nonlinear relationships between high-dimensional input data into a two-dimensional output map. SOM is visualized by using a “U-matrix” which reveals the clustering structure of the data through the use of colours. A light colour on the map corresponds to a small distance between two map units and a dark colour indicates a bigger difference between the map units. The points on the output map that lie in the light area belong to the same group or cluster, while the dark area shows the borders between the clusters. As Gärdenfors (2000) has suggested there is an analogy between his notion of Conceptual Spaces and SOM. During the self-organizing process the points in high-dimensional space are mapped onto a two-dimensional output map that can be identified as a conceptual space. SOM is one way of modelling how the geometric structure of concepts can be created.

Toomas Kirt and Ene Vainik

163

Estonian concepts of emotions Relying on the theoretical framework of three-level representations and the claimed suitability of SOM to detect conceptual structures (Gärdenfors 2000) we hypothesized that the conceptual structure of emotions, if any, should be coherent and being possible to extract by SOM from data gathered either on the symbolic or sub-conceptual levels of representation (i.e., the SOM-representations on those levels should be largely isomorphic and comparable). To test this hypothesis we carried out a case study of 24 Estonian emotion concepts (see Table 1) that were accessed through two lexical tasks and the data analysis was carried out by the method of SOM. Table 1. 24 Estonian concepts of emotion and their glosses Term armastus erutus hirm häbi iha kaastunne kadedus kirg kurbus lõbu masendus mure mõnu pettumus põlgus raev rõõm süü uhkus vaimustus viha õnn ärevus üllatus

Gloss ‘love’ ‘excitement, arousal’ ‘fear’ ‘shame’ ‘lust, desire’ ‘pity, sympathy, compassion’ ‘envy’ ‘passion, heat’ ‘sadness’ ‘pleasure, fun’ ‘oppression’ ‘concern, worry’ ‘pleasure’ ‘disappointment’ ‘contempt’ ‘rage, fury’ ‘joy, gladness’ ‘guilt’ ‘pride’ ‘enthusiasm’ ‘anger, hate’ ‘happiness, luck’ ‘anxiety’ ‘surprise’

One hundred participants (50 male and 50 female), age range 14–76 (M = 40.2, SD = 18.61), all native speakers of Estonian, assessed the semantics of the test words in two different tasks that provided access to the concepts from two different levels of representation of knowledge. The first task addressed the semantic interrelations of emotion terms (synonymy and antonymy) and this task relied on the speaker’s intuitive

164

Chapter Eleven: What Can Self-Organizing Maps Reveal about the Structure of Emotion Concepts?

knowledge about the similarities and dissimilarities of the concepts. The second task addressed the qualities of emotional experience. The task forced the participants to explicate some facets of their subjective and most likely largely implicit and sub-conceptual emotional experience. We used the following scales: • pleasantness vs. unpleasantness of the emotional state; • strong vs. weak emotion; • long in duration vs. short in duration; • increasing vs. decreasing action readiness; • felt in the mind vs. felt in the body; • depends mostly on others vs. depends mostly on oneself; • preceding vs. following an event. The scales task was inspired by Osgood’s method of semantic differentials (Osgood, Suci and Tannenbaum 1975), consisting of three degrees on each side and a neutral zero value in the middle. The participants were instructed to mark their first opinion about a given concept on one side of the scale, indicating the degree of relevance of a specific feature. In the case of semantic irrelevance, a zero value was suggested, and in the case of ambivalent relevance an additional mark for secondary opinion was available.

Results Figure 2 presents the structure of Estonian emotion concepts according to the results of our first task. This task accessed the emotion concepts through the most abstract and symbolic level of representation of emotion knowledge (see Figure 1). The task addressed the concepts through the semantic interrelations (synonymy and antonymy) of emotion terms. The results of the second task (Figure 3) characterize how the conceptual organization of emotions emerged from sub-conceptual and experiential level of knowledge. The concepts were addressed through their relation to the individual perceptions and memories of episodes of emotional experience as measured on the scales with opposing values. As a result, two maps with seemingly divergent structures were obtained (see Figures 2-3). Still, there was a clear distinction (a dark area of nodes) visible in both of the maps between positive vs. negative emotion concepts. In addition, a third cluster of alertness-related states (concern, anxiety, excitement, fear) appears on the SOM of the first task (Figure 2). In the results of the second task (Figure 3) there is a recognizable dimensional layout. The vertical dimension represents the

Toomas Kirt and Ene Vainik

165

opposition of positive vs. negative concepts and on the horizontal axis the situation preceding alertness-related states (concern, anxiety, excitement, fear) are situated on the right edge as opposed to bigger part of the concepts on the left edge that were assessed as mostly following the event. 'rage' 'anger' 'sadness' 'oppression' 'pity' 'disappointment' 'envy' 'contempt'

'concern' 'anxiety'

'excitement' 'fear'

'desire' 'passion'

'shame' 'guilt' 'surprise' 'enthusiasm' 'fun' 'pride' 'happiness'

'love' 'pleasure' 'joy'

Figure 2. Results of the task addressing semantic interrelations of concepts

According to this analysis our hypothesis of structural isomorphism occurred as not clearly supported. Simple visual comparison of the two maps was insufficient to discover a coherent structure of emotion concepts in Estonian. A remaining question was the degree of precision by which the conceptual structure itself can be described. Should we rely more on the data gained from symbolic or sub-conceptual level of representation?

166

Chapter Eleven: What Can Self-Organizing Maps Reveal about the Structure of Emotion Concepts?

'pleasure' 'fun' 'love'

'enthusiasm' 'happiness' 'joy'

'passion' 'excitement' 'desire'

'surprise' 'pride' 'anxiety' 'pity' 'concern'

'rage' 'envy' 'anger'

'fear' 'guilt' 'disappointment' 'contempt'

'sadness' 'shame' 'oppression'

Figure 3. Results of the second task addressing qualities of emotions as measured on the scales

Beyond the SOM The maps of the two tasks were not identical and therefore a question about the coherent structure of the emotion concepts remained unclear. To find out whether there is a common, shared structure we used a new methodology to systematically measure the similarity between two SOMs (Kirt, Vainik and Võhandu 2007). The methodology is based on the idea that the SOM projects close units of the input space into nearby map units and therefore the local neighbourhood relations should remain quite similar. To find the common structure, firstly the matrix of all neighbourhood relations present on two maps was formed. The following criteria were used to define neighbours on the SOMs: (i) concepts were collocated in the same node; (ii) concepts were located no further than one node apart.

Toomas Kirt and Ene Vainik

167

Based on this data of similarity a new meta-level analysis was performed and the isomorphic sub-graph of our two maps was visualized to see the commonly shared information between the two maps (Figure 4).

Figure 4. The maximum isomorphic subgraph

The structures found on both maps (Figure 4) reveal two main bundles of concepts – the superordinate categories of positive and negative emotions – respectively. Both classes show up tendencies of inherent radial organization as well as presence of some peripheral members. In addition, interrelated pairs of concepts (pride–surprise and passion–desire) belonging to neither of the main subcategories appeared. The alertness related states (concern, anxiety, excitement, fear) are not part of the isomorphic sub-graph due to the polymorphism of their sub-conceptual and symbolic representations. To measure the similarity between the neighbourhood matrixes we decided to use the Simple Matching Coefficient (SMC) (Tan, Steinbach and Kumar 2005). The SMC is a commonly used coefficient that rates positive and negative matches equally and can be used if positive and negative values have equal weight. The SMC coefficient can be used if the

168

Chapter Eleven: What Can Self-Organizing Maps Reveal about the Structure of Emotion Concepts?

samples are exclusive like in the case of lexical study. There are two exclusive groups of data – positive and negative emotion concepts – that had weak mutual neighbourhood relations. In the case the neighbourhood range is set on 1 the value of the SMC is 0.84, which indicates rather strong similarity between the two neighbourhood matrixes.

Discussion and conclusion This chapter presented a survey into the structure of the Estonian emotion concepts. In order to find out if there is a coherent structure of emotion concepts two lexical tasks were carried out and the data was visualized by the self-organizing map (SOM) as a method that preserves the topology of input data and reveals the inherent hidden structure of data. The two selforganizing maps did not give a direct visual indication that the conceptual structure could be coherent if accessed through alternative levels of representation (symbolic and sub-conceptual, respectively). Even though the SOMs were not visually one-to-one comparable, the local neighbourhood relations remained mostly the same. This gave an indication of the similarities in the input spaces. In our analysis a new meta-level approach was used to reveal the structural similarities of the SOMs of two tasks which allowed us to discover the intermediate level of conceptual structure, possibly. After performing a systematic comparison of neighbourhood relations we could identify the coherent patterns of neighbourhood relations inherent in the data. The kernels of both positive and negative subcategories of emotion concepts could be clearly identified. This finding is in accordance with the structure of emotions found in the most of the studies carried out in the field of psychology (e.g., Larsen and Diener 1992; Russell 1980; Watson and Tellegen 1985). One possibility is that the structural elements that were found to be shared between our two SOMs (Figure 4), have the greatest chance of belonging to the conceptual level of representation, which is assumed to be on the intermediate position in the hierarchy of abstractness (Figure 1) (cf. Gärdenfors 2000). The details of the local neighbourhood relations which were not shared between our two SOMs, as well as the cluster-like versus dimension-like occurrence of situation related alertness, could be interpreted as belonging only to the symbolic and sub-conceptual levels accordingly. In this chapter we were able to demonstrate a method how to detect isomorphic structure on the SOM visualizations of differently accessed lexical data. We have made an attempt to visualise the conceptual level of representations, which is not directly accessible according to the theory

Toomas Kirt and Ene Vainik

169

that we adopt, i.e. Conceptual Spaces. In other theories, however, e.g. that of Mandler (2004) concepts are accessible to consciousness. We have used some indirect measures of both symbolic and sub-conceptual levels of representations and relied on the intuitions of the Estonian laypersons. Although the gained meta-level structure is an approximation to the original data, it could be said to represent the shared understanding and knowledge of the emotions carried in the minds of the participants of the survey. Our results can, thus, be interpreted as evidence that the multilevel model of representations of Gärdenfors (2000) can be used to represent the structure of emotion concepts. The three levels of symbolic, conceptual and sub-conceptual representation of emotions are compatible also with the levels proposed by other students of emotions (Niedentahl, Setterlund and Jones 1994). The relevance of the conceptual realm of emotions has gained considerable support in recent theories which claim that emotions can be understood as states that are both affective and conceptual at the same time (Feldman Barrett 2006).

Acknowledgements This research was partially supported by the Grant no. 7149 of the ESF: “The dynamic aspects of conceptualization of the emotions,” by the targetfinanced theme No. 0322709s06 of the Estonian Ministry of Education and Research and by the National Programme for Estonian Language Technology. Our grateful thanks belong to the editors of this volume whose comments and suggestions have helped us to improve our original manuscript.

References Barsalou, L.W. 2003. Situated simulation in the human conceptual system. Language and Cognitive Process 18(5/6): 513-62. —. 2008. Grounded cognition. Annual Review of Psychology 59: 617-645. Cacioppo, J.T. and Berntson, G.G. 1999. The affect system: Architecture and operating characteristics. Current Directions in Psychological Science 8: 133-137. Feldman Barrett, L. 2006. Solving the emotion paradox: Categorization and the experience of emotion. Personality and Social Psychology Review 10(1): 20-46.

170

Chapter Eleven: What Can Self-Organizing Maps Reveal about the Structure of Emotion Concepts?

—. 2004. Feelings or words? Understanding the content in self-report ratings of experienced emotion. Journal of Personality and Social Psychology 87(2): 266-281. Gallese, V. and Lakoff, G. 2005. The brain's concepts: The role of the sensory-motor system in conceptual knowledge. Cognitive Neuropsychology 22(3/4): 455-479. Gärdenfors, P. 2000. Conceptual spaces. The geometry of thought. London: The MIT Press. Harnad, S. 1990. The symbol grounding problem. Physica D 42: 335-346. Kirt, T., Vainik, E. and Võhandu, L. 2007. A method for comparing selforganizing maps: case studies of banking and linguistic data. In Y. Ioannidis, B. Novikov and B. Rachev (Eds.), Proceedings of eleventh East-European conference on advances in databases and information systems, 107-115. Varna, Bulgaria: Technical University of Varna. Kohonen, T. 2000. Self-organizing maps (3rd ed.). Berlin: Springer. Larsen, R.J. and Diener, E. 1992. Problems and promises with the circumplex model of emotion. Review of Personality and Social Psychology 13: 25-59. Mandler, J. 2004. The foundations of mind: Origins of conceptual thought. Oxford: Oxford University Press. Murphy, G.L. 2002. The big book of concepts. London: The MIT Press. Niedenthal, P.M., Setterlund, M.B. and Jones, D.E. 1994. Emotional organization of perceptual memory. In P.M. Niedenthal and S. Kitayama (Eds.), The hearts eye. Emotional influences in perception and attention, 87-113. New York: Academic Press. Ortony, A., Clore, G.L. and Collins, A. 1988. The cognitive structure of emotions. Cambridge, UK: Cambridge University Press. Osgood, C. E., Suci, G. J., & Tannenbaum, P. H. 1975. The measurement of meaning. Urbana and Chicago: University of Illinois Press. Prinz, J.J. 2005. Passionate Thoughts: The Emotional Embodiment of Moral Concepts. In D. Pecher and R.A. Zwaan (Eds.), Grounding cognition: The role of perception and action in memory, language and thinking, 93-114. New York: Cambridge University Press. Russell, J.A. 1980. A circumplex model of affect. Journal of Personality and Social Psychology 39: 1161-1178. Tan, P.-N., Steinbach, M. and Kumar, V. 2005. Introduction to data mining. Boston: Addison Wesley. Watson, D. and Tellegen, A. 1985. Toward a consensual structure of mood. Psychological Bulletin 98: 219-235.

A THREE-DIMENSIONAL FORMAT FOR MODELLING LEXICAL ORGANIZATION MICHAEL FORTESCUE

Abstract. Existing models of the organization of the mental lexicon vary considerably as to how closely they are intended to reflect cognitive reality. The present proposal takes processing seriously and is inspired by Burnod (1990), a milestone attempt to bring AI modelling closer to what is actually known of the workings of the brain. Elements from more recent neurolinguistic approaches that emphasize the distributed nature of lexical representation are also integrated into the model. The templatic formalism is tailored to mesh both with a procedural approach to language and with the basic architecture of cortical columns organized into extended “multimodule” networks according to the symmetry of cortical connectivity as described by Burnod. It generates a number of testable hypotheses, given modern neuroimaging techniques. The chapter is intended as an overall introduction to the formalism.

Introduction Linguistic models of the lexicon are generally conceived of in terms of representations of meaning-form couplings and their combinatorial potential. None of them – with the partial exception of Lamb (1998) í is directly geared to what is known of specific processes in the brain that utilise the mental lexicon “on line”, nor to the physical distribution of lexical information in the cortex. Lamb does discuss localisation (op. cit.: 352ff), but admits that his own proposal for placing “lexis” in one particular area (around the angular gyrus) is misleading, since it is probably widely distributed (as both Deacon 1997 and Pulvermüller 2002 assume). The present proposal, inspired by Burnod’s (1990) “adaptive neural network”, takes both processing and localisation in the brain seriously. Burnod’s neurological model is premised on the “sixdimensional” symmetry of cortical connectivity, the six dimensions being respectively: left-right (between hemispheres), associative-frontal (posterior-

172

Chapter Twelve: A Three-Dimensional Format for Modelling Lexical Organization

anterior), spatial-molecular (outer cortical surface-limbic system), visualauditory, parietal-temporal, and primary-secondary (sensorimotorassociative). The aim of my “templatic” approach (to be spelled out in much greater detail in a forthcoming book) is to provide a bridge between neuroscience and theoretical linguistics in the area of the lexicon. In meeting this aim, the model, by virtue of its hybrid nature, faces the inevitable risk of not satisfying practitioners of either discipline. I am not alone among cognitively orientated linguists, however, in believing that the divide will eventually disappear, although there is still a long way to go. It is not premature, it seems to me, to try and get a grip í even in a tentative way í on the intermediate territory remaining to be explored and explained. Though much of the neurological detail is necessarily speculative (controversy still abounds as regards interpreting the relevant neuroimaging evidence), the graphic framework that I propose will, I hope, at least help clarify what kind of overall modelling we need to undertake in order to generate testable hypotheses in this area. My model can be said to combine a classical AI approach, involving the manipulation of symbols by and between discrete modules (e.g. a dictionary, a grammar and a parser), with the more recent parallel distributed processing approach of connectionists in terms of neural networks whose sub-symbolic connections (or rather their “weights”) determine their input-output relations. On my model (as on Burnod’s) the connectionist perspective is reflected in the notion of “call trees” instantiated in the “tuned” synaptic connectivity of neural columns, while the modular, symbol-based perspective is reflected in the deterministic “wiring” between symmetrically related modules of the cortex and in the compositionality of word meanings. Ultimately, both neurolinguistic and computational adequacy are aimed at. The templatic formalism is meant to reflect the basic architecture and symmetry of cortical columns organised into neural networks. Cortical columns are “adaptive automata” that generate a sequence of actions adjusted to external conditions in order to attain a goal, an equilibrium state, via a series of sub-goals (Burnod 1990: 64f.). They may exist in various discrete states of activation (Burnod op. cit.: 91), which I call “eigenstates”. It is the intermediate “modular” level of lexical organization that I address, a level where “call trees” point from mediatory word columns towards sensory and other “affordances”. These correspond approximately to the features tested for by Miller and Johnson-Laird’s (1976) “decision tables”. Note that when I talk of individual “word columns”, it should be understood that this is for practical reasons: in

Michael Fortescue

173

many if not all cases I shall actually be referring to aggregates of more than one neural column, corresponding to Burnod’s notion of the “multimodule” (op. cit.: 129). These are interconnected assemblies of cortical columns that are not necessarily contiguous one to another but share a common “goal” (for example the production or recognition of a single word). They are coordinated in such a way (by both long and shortdistance connections) that their common “goal” can be accessed from a number of different starting points within the network. This property they share with Pulvermüller’s (2002: 156ff) general “functional web”. The model does not assume that all lexical information apart from phonological and morphological “form” is gathered in unitary “lemmas” in the manner of Levelt (1989: 187f.).1 It sees this information as distributed more radically across a number of specific cortical regions, as determined by the functioning of sets of cortical columns within an overall, symmetrically organised network – though by no means as radically as in connectionist models. The network, as conceived, is anchored in association cortex but is subserved by sensory-specific subregions and traversed by two major concentric circuits within the left hemisphere, a relatively fast inner “word/sentence” one (of mediatory linguistic form), and a slower outer “image/relational” one (of semantic content), as in Burnod’s model. To this I add a still faster innermost phonological circuit, directly linking Broca’s and Wernicke’s areas. These circuits can be compared with Deacon’s (1997: 291) notion of concentric “tiers” operating at different speeds. The essential innovation in the present approach is a distinction between cortically more posterior (sensory) and more anterior (functional) sub-parts of lexical entries, associated in specific ways via diverging/converging routes across the dominant hemisphere, a more ventral “nominal” one and a more dorsal “verbal” one, analogous to the well known “what” and “where” routes of visual processing. Démonet, Thierry and Cardebat (2003: 76ff) present and discuss the relevant neuroimaging data. Arbib et al. (1998: 243) prefer the term “how” to “where” here, since it is a matter of “how to interact with an object” 1

Levelt distinguishes a “lemma lexicon” from a “word form lexicon”. Whereas the former combines semantic and syntactic information about words, the latter combines morphological and phonological aspects of words. While the present model differs on this point, in other respects Levelt’s conception of the “lemma” fits well with it. It extends Morton’s (1969) notion of the “logogen”, with which it shares the property of being a parallel lexical accessing devise which collects “evidence” from various sources until a threshold is reached when “enough” has been gathered to trigger activation of the word’s form.

174

Chapter Twelve: A Three-Dimensional Format for Modelling Lexical Organization

(location being only one aspect of this). Henceforth I shall refer to the dorsal “where/how” route.2 As regards the posterior (parietal) end of the two visual routes, Burnod describes them as forming a crossing “combination matrix” which allows for continuous adjustment for the position of an object focused on (Burnod op.cit.: 229f.). Call tree networks here “anticipate” the transformation of images produced by movement. I suggest that a “combination matrix” also lies at the convergence of the two linguistic routes in inferior left frontal cortex. Accordingly, I divide lexical features into sensory and functional affordances (or associations).3 The latter are taken to be instantiated in higher (supramodal) cortical columns hierarchically above and at a considerable physical distance from the basic word columns mediating the former, sensory affordances, where local call trees are initiated through posterior cortex in search of sensory “goals” of increasing specificity. The goal might, for example, be the construction or recognition of an image. Call trees are “adaptive functions that can guide a sensorimotor system towards a goal” (Burnod 1990:123ff). They function by successive exploratory approximations as they search for matching patterns in other columns. It is now widely accepted that the principal sensory affordances for basic nouns and verbs (both “pointed towards” from mediatory columns in association areas) are located separately in the cortex, anchored respectively in temporal and – primarily – pre-motor areas, as neuroimaging studies corroborate (cf. Pulvermüller 2002: 45). In a lexical 2

See Hurford (2003) – and the peer review comments to it – for differing views on the relevance of the two routes to language. For readers worried about this assumed overlap between the visual and the linguistic systems or ”modules” I should point out that I am simply adopting a position common to all forms of cognitive linguistics, namely that the various cognitive sub-systems of the brain overlap, with some processes and structures being shared between them, others being specific to only one system. In the present case this is a matter of the correlation of the linguistic and visual systems. This close correlation is what allows for efficient communication and ”translation” back and forth between the two systems in cooperative tasks (which includes most forms of rational thinking). Clearly the inner ”phonological circuit”, on the other hand, has little to do with the visual system (except in reading). 3 Note that others, such as MacWhinney (1999), limit the use of the word “affordance” mainly to sensory associations (it was originally used by Gibson in the sense “parameters for motor interaction signaled by sensory cues” í cf. Arbib et al. 1998: 169). I have extended this usage since I follow Burnod in seeing the same underlying mechanism – that of the “call tree” – behind both kinds of affordance, although they are anchored in different cortical regions.

Michael Fortescue

175

decision task using a PET scan technique Perani et al. (1999), for example, found that left dorsolateral frontal cortex (i.e. Broca’s area) was activated by verbs but not nouns (a finding corroborated by aphasic evidence). The same overall area is activated during the perception of actions, especially those involving the hands (op. cit.: 2341f.). Other areas were also activated by verbs, in particular in mid-temporal and parietal parts of the left hemisphere. The mid-temporal activity may reflect the sensory affordances of typical object noun arguments (i.e. of entity types typically performing the activities concerned), just as nouns may in turn have associations with specific actions or states expressed by verbs. The networks associated with verbs are clearly quite complex, although with a common denominator. The sensory (or rather “sensorimotor”) affordances of verbs are taken on the model to be organised along orthogonal motor-spatial and (secondarily) auditory-visual call tree axes. The motor-spatial axis links the (pre-)motor areas of the cortex with the corresponding (and interconnected) inferior parietal region that Burnod sees as involved in the sensory affordances of verbs of transferral and other telic verbs containing a resultant state component (cf. Burnod 1990: 277f). The axes defining the sensory affordances of nouns again include the auditory-visual one, but now it is orthogonal to the spatial (parietal)temporal pole axis. The latter pole constitutes a higher level convergence zone rather than a primary sensory region, and is related to the prelinguistic object-classifying function of the inferior and middle temporal lobes. It appears to form a gradient from the more general (“natural kinds”) to the more specific (“individuals”) in a posterior to anterior direction (Damasio and Damasio 1992: 70). Let us call this the “object synthesis” dimension, one essential to the “binding” of separately analysed sensory features of objects during perception. “Natural” semantic fields (organised centripetally by degrees of proximity/similarity on both orthogonal axes simultaneously) fall out from this view of sensory affordances. The functional affordances of nouns are organized on the model’s templates into crossing paradigmatic (inherent property) and syntagmatic (combinatorial) axes. The former contains grammar-relevant semantic and referential features like animacy and countability, while the latter contains associations to syntactic complements or modifiers, etc., with which the noun typically combines. These features initiate call trees, like those associated with sensory affordances, but now on a higher, abstract grammatical plane. It is less clear than in the case of sensory affordances where exactly the functional affordances of nouns might be cortically

176

Chapter Twelve: A Three-Dimensional Format for Modelling Lexical Organization

located, but the frontal and anterior temporal lobes are known to be connected via the uncinate fasciculus and this may be crucial in the semantic and syntactic integration of nouns (or of “terms” containing them) with verbs (or their predicate frames) during the integration of predications (cf. Blank et al. 2002: 1839). Similarly for the functional affordances of verbs: the paradigmatic axis contains temporally extended event-type features (“aktionsart”), while the syntagmatic one specifies particular argument types required for the verb to form predications. The syntagmatic axis is assumed to be oriented orthogonally to those of nouns when they meet in the frontal “grammar” area in such a way that nouns (or, rather, the NPs they head) fill the argument slots of verb frames, thus facilitating their integration into predications without having to stipulate abstract word-class “labels” adhering to them. The syntagmatic affordances of nouns may in fact be little more than “pointers” into the grammar templates formed around the predicate frames of verbs. The functional affordances of verbs (which also embody “event structure” information) are assumed to be located somewhere in those regions of the dominant frontal lobe that lie anterior to Broca’s area. In fact, they may constitute the lower level of “grammar cortex” where generalized grammar templates are taken to be instantiated. They are thus seen as lying further along the same “motor” dimension on which the mediatory columns of verbs are located. There are of course further differences according to part of speech and sub-type of the noun or verb involved (thus prepositions appear to be primarily anchored in the parietal region and adjectives of shape, texture and colour in secondary occipital visual cortex), but the basic principles of symmetrical connectivity and the selective integration of bidirectional, orthogonally intersecting activation flows apply to all areas and all word types. This includes closed-class function words and pronouns that display anterior (functional) affordances only and may thus be diagrammed on the upper halves of word templates alone. At a still higher functional level, many verbs – and some nouns, especially those for artefacts of all sorts – are further associated with specific “scenarios” (or “scripts”) anchored in frontal cortex. Such scenarios relate argument and event types in higher level conceptual aggregates containing elements of causality and intentionality (both individual and institutional). These cognitive schemata can be termed “macro-functional” í as opposed to language-specific, grammatically relevant “micro-functional” affordances í and can be taken to extend also into right-hemisphere regions that contain essentially non-linguistic (though perhaps linguistically “moulded”) context types. Joanette et al.

Michael Fortescue

177

(1990) discuss the relevant literature, mainly from aphasia studies, indicating that the right hemisphere is intimately involved not only in “script” knowledge, but also in general discourse/contextual cohesion. When terms (activated via the temporal lobe) are combined with predicate frames in frontal cortex, there must be consonance between the parts combined at all levels (sensory, micro-functional and macro-functional, semantic as well as syntactic). The functional “qualia” (to use the term of Pustejovsky 1995) associated with a given predicate frame mesh with those of nominal terms integrated as appropriate arguments. The abstract “tags” employed on the model for characterizing these qualia of entity and event types are limited to those that can be relationally defined within the coding system of the grammar itself. A major organising principle of the model states that the micro-functional affordances of words must be mappable from sensory and/or macro-functional ones (through common abstracting processes). In the following section I shall illustrate how the model works by setting up schematic “word templates” for some typical words belonging to the major word classes of English. Thereafter I shall give two brief examples of how word columns interact in terms of “derivational” processes and linkages via macro-functional scenarios.

Some sample word templates The simplest starting point is to consider a concrete, countable noun, such as ‘parrot’. I shall assign to it the lexical “shape” represented on Figure 1. The format employed is schematic and not meant to be taken too literally as corresponding to the organisation of individual cortical columns; the fact that it nevertheless suggests their layered structure is not fortuitous, however. Figure 1 – corresponding to a single mediatory column í is to be understood in the following manner. The vertical line represents the columnar structure (a cluster of cortical cells forming a functional unit) and the medial arrow traversing it indicates phonological input (via the thalamus) from Wernicke’s area, where linguistic input through either the visual or the auditory channel is analysed. The lower half of the template (constituting the linguistic “sign” as such) associates the phonological form with the call trees corresponding to the word’s sensory affordances. Since this is a common noun referring to a “natural kind” the column will be situated somewhere in the (heteromodal) association areas of the medial or inferior temporal lobe, approximately in a position defined by the shortest distances to the relevant sensory areas

178

Chapter Twelve: A Three-Dimensional Format for Modelling Lexical Organization

habitually involved in activating the word or, in reverse, in being activated by it. Here Lamb’s “Proximity hypothesis” (op. cit.: 218) can be invoked. The exact lateral position along the wide swathe of “territory” defined (a cortical “band”) is a function of the relative strength of associated call trees along the orthogonal axis. In the case of ‘parrot’ this will involve both the auditory-visual axis and the orthogonal spatial-object synthesis axes, as indicated on the crossing lines at the bottom of the template. Activation of the word will send call trees searching for a visual image of a parrot, back through successive stages of analysis in partly redundant parallel pathways for colour, shape, distinctive parts, etc., towards the primary visual area of the occipital cortex. This “call” is rarely carried out to the point of producing a fully-specified image: it can stop at any point along the path as soon as a good enough match with top-down expectations is made (e.g. just some vague bird-like shape moving in some characteristic manner). At the same time as this principal call, others will go out to search for auditory images (e.g. a parrot’s characteristic shriek) or kinaesthetic ones set in space – the “feel” of a flying bird’s flapping wings, for example. Expanding activation in all four directions (relatively slow since this takes place via local cortical links) will result in more and more specific delineation of the object type ‘parrot’.

[párԥt]

auditory object synthesis

N syntag.: _V (fly) paradig.: animate

spatial visual (wings, colour, etc.)

Figure 1. A representation of ‘parrot’

The arrow pointing towards the object synthesis pole on templates for nouns indicates the more persistent activation produced by such words along this axis í namely towards unified multimodal images, i.e. towards the most individualized level of imagery possible. This is crucial in the specification of potentially referential “terms” (NPs). More contingent, “encyclopaedic” knowledge about parrots can be accessed along this axis

Michael Fortescue

179

(or its right hemisphere “echo”), but such knowledge (especially if structured propositionally) is not necessarily limited to proximity to the word’s sensory affordances, it may presumably be associated with higher level functional affordances, i.e. be in the form of “macro-functional” scenarios and frames based on book knowledge rather than on personal experience with the creature. The word ‘parrot’, along with other specific bird species nouns, can furthermore be assumed to be located (literally) within a natural semantic field of ‘birds’ (cf. Pulvermüller 2002: 88-90). This is defined by the call trees initiated by all these related words, namely towards a summation of all those properties (along both axes) shared by them at an early stage of the relevant call tree searches. The semantic field is hinted at by the small line crossing the object-synthesis dimension, but in fact could better be indicated by a square joining up all four sensory dimensions. In the upper half of the template, the short parallel lines indicate that the link to the functional affordances of the word (above) traverses a relatively long distance compared with that to the sensory ones. What this is meant to suggest is a contrast between local call tree axes (sensory affordances) and long-distance connections through myelinated fibre tracts to the functional affordances in correlated frontal cortex. The latter connections presumably pass from the axons of the pyramidal cells of mediatory word columns to the dendritic trees of cortical columns located at a considerable distance from them. The template can thus be interpreted as actually linking two widely separated columns. Further macrofunctional levels (implying still other, hierarchically superior columns) may be superimposed on a single template above a word’s microfunctional affordances. The micro-functional affordances, organised on orthogonal paradigmatic and syntagmatic axes within inferior frontal cortex, provide the trajectories for activated call trees that radiate outwards in all directions, searching for their functional “goals”. In the case of the word ‘parrot’, these include features relevant to the grammar templates, such as ‘noun’ (as marked by the bolded ‘N’), also – paradigmatically – ‘animate being’, and – syntagmatically – the common occurrence of the word before activity verbs like ‘fly’ (the italics indicate an actual phonological word). All the features relevant for the correct grammatical functioning of the word can be indicated in this way by short crossing lines. If one now compares these noun templates with one for an activity verb such as ‘fly’, as on the left of Figure 2, it will be seen that the orientation of the phonological input to the column has been swivelled through 90 degrees. This is meant to indicate that the activation it has

180

Chapter Twelve: A Three-Dimensional Format for Modelling Lexical Organization

caused is here being traced along the more dorsal (parietal) “where/how” route, rather than the more ventral (temporal) one associated principally with nouns; it further reflects the “U-bend” of the supramarginal gyrus, where the two routes branch out from (approximately) Wernicke’s area. Although the auditory-visual axis remains the same (still parallel with the flow of phonological input), the orthogonal axis traversed by the dorsal route is the sensorimotor one linking the spatial (parietal) and the premotor (frontal) regions of the cortex. The bolding indicates the primary importance of this axis for verbs (motion). On the syntagmatic functional axis a suitable NP to fill the single argument slot of the verb’s predicate frame is indicated – here an animate ‘agent’ noun, by default an exponent of the category of ‘bird’. Note that the position of features indicated along the syntagmatic axis is arbitrary for both nouns and verbs (positioned right or left for convenience on the templates), whereas the position of a feature along the paradigmatic axis of verbs with time-extended meaning is determined by its relative position along the “arrow of time” (as mapable from the adjacent “story” circuit). paradig.: motion syntag.: NP_ ag.(bird, etc.)

mot. vis. (bird)

V paradig.: past [flái]

[flu:] [flái]

(aud.)

spat.: air

Figure 2. A representation of ‘fly’

The broken vertical line links the sensory affordance of something winged like a flying bird with a suitable argument type (e.g. noun ‘bird’) at the functional level – the word ‘bird’ indicated on the latter is not itself strictly necessary here (note again the convention of putting actual words in italics). This represents the basis for the compatibility of ‘bird’ (or its hyponyms) with the selection restrictions required by ‘fly’: it is difficult to imagine ‘flying’ without simultaneously imagining something (typically a bird) doing it. Motion and change is central to action verbs, but can also enter a noun’s affordances in the form of typical visual actions associated with

Michael Fortescue

181

them, just as there are typical argument types associated with verbs. A similar vertical line could also have been added to map the feature ‘motion’ onto the motion dimension of the sensory affordances. The form of ‘fly’ discussed so far has been its infinitive or “citation” form. What of inflected forms like past tense ‘flew’? The relation between this and ‘fly’ is indicated on the partial template to the right above, which could be integrated with the one on the left – it is assumed that the word column as a whole may be activated by either of these allomorphs as phonological input (and the sensory affordances are of course shared). The node-like blobs have been added to make the connections between the two forms clearer. The past tense form is actually only “summoned” at a macro-functional level, namely by the higher-order requirements of the “story circuit”. Since the form of the past tense of ‘fly’ is irregular, it cannot simply be handled by a general grammar template “derivation” that adds ‘-ed’ to the lexical stem (as needed for regular verbs). One must assume that the form ‘flew’ is anchored in the micro-functional affordances of the ‘fly’ word column itself, ready for higher level activation, although it “maps back” to paradigmatic sets of forms in Wernicke’s area, where such formally related items may be gathered redundantly. The mediatory column, in other words, has two “eigenstates”, past and non-past, according to the grammatical demands of the given discourse context. An irregular past tense ‘went’ form must also be marked on the template for ‘go’, but since there is no formal overlap at all between the two phonological forms they will presumably not be adjacent in Wernicke’s area í ‘went’ will have its own representation there, only indirectly linked to the citation form via the mediatory word column for ‘go’. paradig.: dimension

Adj standard

syntag.: _N [wáid] (aud.) (obj.)

spat. vis.

Figure 3. A representation of ‘wide’

182

Chapter Twelve: A Three-Dimensional Format for Modelling Lexical Organization

Next, let us consider adjectives. On Figure 3 is portrayed the simple adjective ‘wide’. According to the Proximity hypothesis, the sensory affordances of this word should be placed somewhere in the parietal association cortex along the “what” route, as for nouns. The spatial dimension is central to its meaning, in particular its somato-sensory extreme (which includes kinaesthetic and tactile affordances), since the spatial perception of size is, as a default, relative to the perceiver’s own bodily experience (though also the visual dimension is involved). The opposite dimension on that axis, that of object synthesis, is less relevant here, hence the lack of the arrow in that direction (there is no degree of “individuation” involved). Note that the “standard” marked on the paradigmatic axis for ‘wide’ (‘wide for a í’) is determined by the particular nominal involved (hence the dotted lines forming a right-angle). This is part and parcel of the general mutual adjustment of meaning between lexical items in phrases or compounds. The syntagmatic affordance given for ‘wide’ reflects its primary function of modifying nominal heads. The orientation vis-à-vis head nouns is given by the short line crossing the syntagmatic axis (modifiers í both adjectival and adverbial í are orthogonal to the syntagmatic axes of nouns and verbs).

A “derived” word An analysis of the word ‘frightened’, a participial form derived from a verb, will allow us to fill out some dimensions of the sensory and functional affordances of words that have not been dealt with up to this point. As regards sensory affordances, it illustrates a level of call tree activity representing one of Burnod’s axes of symmetry not yet discussed, namely the limbic (or “molecular”) one of emotional reaction. The ‘fear’ meaning element of ‘frightened’ is added on Figure 4 as an extension “downwards” of the central axis of the cortical column for the word (cf. Pulvermüller 2002: 91 on the involvement of the amygdalae with such “word tails”). Although this axis is especially important for certain adjectives and verbs of emotion, it is potentially there to be exploited by other types of words too. Still more important for the overall model of the neurological embodiment of word meaning is the extension “upwards” from the (micro) functional pole of the word’s template to the higher level where the adjective is related to the verb from which it is derived, namely ‘frighten’, to its right. In Figure 4 this is indicated by the horizontal line joining the two columns (presumably a relatively short-distance link). Specifically, it joins the event type associated with the adjective (a state) with that of the

Michael Fortescue

183

verb (a causal action), via a causal link that projects the relationship between the two arguments of the verb (the causer and the causee) “upwards”, associating the latter with the experiencer argument of the adjective. While the “causee” must be animate (an “experiencer”), note that the “causer” is only prototypically an “agent”: NP1 could be represented by an inanimate or higher order entity, depending on broader semantic and situational compatibility. The slanting broken lines on both columns indicate this projective mapping í the “cumulative” or “abstracting” nature of higher levels will in general be symbolized in this fashion, with the two short parallel lines again indicating a shift in hierarchical level, as with the divide between sensory and micro-functional affordances. Compare the (micro-)functional affordances of the two words here. The adjective is labelled as Past Participle (a functional sub-variety of adjective) and its paradigmatic axis indicates that it refers to an emotional state – this of course maps onto its sensory affordances, but is relevant at the functional level for the various constructions it can enter, since it syntagmatically requires an experiencer nominal head. The functional affordances of the verb are those typical for causal action words, linking a causal event with a resultant state (note the direction of the arrow). What the large arrow connecting the two words then actually indicates is not the causal event as such but the derivation of an adjective suitable to describe the resultant state of the undergoer of the type of event expressed by the verb, a purely linguistic relationship. This “derivational” process (let us call resultant state (D1) Adj (PaP)

V _NP1: exp .-er _N: exp.-er paradig.: emot. state

e1: cause e2: result NP2_: ag

[fráitԥn]

[fráitԥnd] mot. vis.

aud. spat. limbic: fear

mot.

aud

vis.

spat. limbic: fear

Figure 4. A representation of ‘frightened’ and ‘frighten’

184

Chapter Twelve: A Three-Dimensional Format for Modelling Lexical Organization

it “D1”) can be generalized as a frontal cortex “grammar” template applicable to the class of (transitive) verbs undergoing it. The orientation of the orthogonal lines defining the relevant affordances, both within and between the columnar structures, are important. Note the short lines crossing respectively the syntagmatic and paradigmatic axes of both the verb’s and the adjective’s functional affordances – as elsewhere with orthogonally oriented features of affordances, they form a right angle when extended and joined up. Thus the lines indicating ‘emotional state’ and ‘experiencer’ on the adjective’s affordances have been extended and joined by dotted lines to associate that paradigmatic function with that syntagmatic argument, just as ‘result’ and ‘experiencer’ can be joined on the corresponding affordances of the verb. Observe that although the derived form is an adjective, it has inherited the (verbal) sensory affordances (and their orientation) from the verb that it is derived from. It is the micro-functional syntagmatic axis that is crucial for its integration into sentences, and here it is indicated – correctly í as orthogonal to its head noun host. The “derivational” process here is of course historically established and does not preclude the two words (and their columnar networks) existing independently. It is nevertheless potentially productive and can be triggered by a sentence-level context above the level of the individual word. Derivational processes of the type we have just seen feed back into the system of grammar templates and may enter into the recursive morphosyntactic accretion of whole sentences. It should be clear by now that by “derivation” I mean all manner of lexical extensions, syntactic as well as morphological, above the level of the individual word stem.4 This includes generalized grammar templates and constructions, also the extraction of event structures from nominals in certain “qualia unification” processes in Pustojevsky’s theory, a matter I shall return briefly to below. “Derivations” represent the “glue” that holds the lexico-grammar together as a whole.

4

My approach to “derivation” is germane to Jackendoff’s recent “l-rule” approach (Jackendoff 2002: 180ff), in which there is essentially no distinction between lexicon and grammar, with phrasal constructions taking a middle ground as regard the preponderance of variables as opposed to specified lexical items. I do not, however, accept his clear-cut distinction between lexical items (stored individually in long term memory) and grammatical “words” (generated on-line from rules). “Grammar cortex” on my model can be envisaged as a network of interrelated constructions ranging hierarchically from individual predicate frames to completely general clausal structures.

Michael Fortescue

185

Scenarios and qualia The following pair of templates, for ‘rifle’ and ‘shoot’, introduces another important aspect of lexical meaning not yet addressed, namely the “telic qualia” of artefact words, in this case the function of shooting and killing game associated with rifles. Such a function can be represented as on Figure 5, in the form of a higher-order “scenario” contained within a macro-functional circle.5 It should be clear from the amount and kind of information contained in the circle that the full scenario is unlikely to be instantiated in a single cortical column. ‘shooting for game’ action: aim fire (intention: kill) instr.: weapon ag.: hunter pat.: game

V

N

action

weapon

V_ (shoot)

_NP2: pat. NP1_: ag.

(with_) NP3: instr. (weapon) (telic) [šu:t]

[ráifԥl] aud.

spat. (parts)

obj. vis.

mot.: (aim) (fire) vis.

aud. (loud report) spat.

Figure 5. A representation of ‘rifle’ and ‘shoot’

“Scenarios” in general probably represent quite complex aggregates or networks of columns, though the essential features within them (those that 5

Various terms have been used within Artificial Intelligence, Sociolinguistics and Cognitive Linguistics for these world knowledge representations. Usage ranges from the frame semantics of individual predicates to detailed action scenarios, e.g. for the sequence of actions one typically carries out in a restaurant í my own usage is closer to the latter (compare also Fillmore and Atkins (1992) for a general “commercial transaction” frame).

186

Chapter Twelve: A Three-Dimensional Format for Modelling Lexical Organization

map onto the micro-functional affordances of individual word columns) may be abstracted by mediatory right hemisphere columns from openended arrays of sensory associations, similar to the way in which mediatory word-columns function in the left hemisphere. The scenario (let us call it ‘shooting for game’) is mapped via the slanting line onto the micro-affordances of the word ‘rifle’ as shown í it is also mapped onto those of the verb ‘shoot’ and ‘kill’ and other associated words. The feature ‘weapon’ there is a sub-type of ‘instrument’, note. The words within the circle should be regarded as arbitrary labels of perceptual categories/events. Observe that the scenario has both temporal and spatial aspects. There is a potential problem with the sensory affordances of the noun, since on the one hand it clearly lies on the “what” route, as a visually identifiable physical object (it could be recognized even by someone not familiar with its exact function), and on the other hand it is an instrument or tool which is handled in a specific manner for a specific purpose – which suggests that the “where/how” route (involving the motor-spatial axis of physical manipulation) is also crucial, although the latter is more closely associated with verbal actions and spatial relations. Pulvermüller (2002: 58) presents evidence of strong premotor activity associated with “tool” words, as opposed to strong occipital activity with easily visualized animal words.6 The solution I have adopted here (partially based on the existing neuroimaging data, but still awaiting unambiguous confirmation) is to utilize the macro-functional scenario already postulated, namely the word’s indirect links to ‘shoot’ (itself presumably situated along the motor-spatial axis). On the template the affordances of the verb and of the noun match and are linked, suggesting that the affordances of a word like ‘rifle’ are actually distributed along both routes, in so far as its function as a weapon is projected via the scenario into the ‘instrument’ slot of the verb’s functional affordances. As with all the templates presented in this chapter, only essential detail has been added – scenarios also contain a wealth of secondary associations (even to salient token occasions). Note that the different phases of shooting map indirectly via the scenario to the ‘motion’ affordances of the verb: the motion dimension is inherently time-extended, in the order of sub-events given. It should be emphasized that only a portion of the total meaning potential of the verb ‘shoot’ is indicated here – in particular, the 6

There is in general still a good deal of controversy as regards the interpretation of neuroimaging evidence for the localization of different semantic categories – see Martin and Caramazza (2003) for a summary of this research going back to the early 1980s.

Michael Fortescue

187

combination with the word ‘rifle’ inhibits those aspects of its meaning that are associated with killing another human being by shooting them. This is an example of the general process that Talmy (2000: 2, 253) calls ”resection”, whereby an inherent element of a verb’s basic meaning is removed by a suitable ”framing event”. The feature ‘telic’, note, is optional here and is defeasible by, for example, derivational extension with ‘at’.

Conclusions If the model is to be taken seriously as reflecting a neurological basis, it clearly needs to be in conformity with the on-going results of neuroimaging studies. In fact it highlights some specific areas where further investigations of this kind need to be undertaken. The hypotheses the model generates include the basic premise of the disassociation of the functional (i.e. grammatical) and sensory affordances of all “content” words. This should show up in studies devised to compare neurological responses to words sharing the same syntagmatic behaviour (but quite different sensory affordances) with words sharing similar sensory affordances (near synonyms) that differ in syntagmatic behaviour. Also more specific predictions as regards activity along the hypothesized linguistic “what” and “where/how” routes could be tested with relatively simple (and well known) methods involving prompts for appropriate words presented to subjects during scanning (cf. Démonet, Thierry and Cardebat (2003), Perani et al. (1999), and Martin and Caramazza (2003) for a selection of neuroimaging results attained so far). What I have attempted to demonstrate in this chapter, in a very sketchy way, is how the distributed nature of lexis need not be a stumbling block for the modelling of the mental lexicon. Taking up Lamb’s challenge, I have tried to be more specific in this respect by focusing on what is known of the localisation of the separate kinds of meaning associated with individual words, setting them within an overall framework provided by Burnod’s model, where “call trees” in search of their “goals” play across a symmetrically inter-connected neural substrate of “cortical columns”. The basic unit on which my linguistic approach and Burnod’s neurological one converge is the lexical “multi-module”. The templates of the model should thus be understood as each reflecting several interconnected cortical columns, with the component elements widely distributed but located in specifiable cortical areas, namely phonological forms in or close to Wernicke’s area, sensorimotor affordances in specific areas of posterior and pre-motor cortex, micro-functional affordances in inferior frontal

188

Chapter Twelve: A Three-Dimensional Format for Modelling Lexical Organization

cortex anterior to Broca’s area (all in the dominant hemisphere), and macro-functional affordances in (pre-)frontal cortex of both hemispheres. The model focuses especially on the micro-functional component since this is what meshes directly with grammatical processes and can be abstracted away from the open-ended multiplicity of associations at other levels. This separation should, in theory, make it easier to isolate aspects of the organisation of the mental lexicon for correlating with the findings of neuroimaging studies and for setting up new ones.

References Arbib, M., Érdi, P., and Szentágothai, J. 1998. Neural organisation. Cambridge, Mass. and London: MIT Press. Blank, S.C., Scott, S.K., Murphy, K., Warburton, G. and Wise, R.J.S. 2002. Speech production: Wernicke, Broca and beyond. Brain 125(8): 1829-1838. Burnod, Y. 1990. An adaptive neural network. The cerebral cortex. London: Prentice Hall, Paris: Masson. Damasio, A. and Damasio, H. 1992. Brain and language. Scientific American, September 1992: 63-71. Deacon, T. 1997. The symbolic species. London: Penguin Books. Démonet, J.-F., Thierry, G. and Cardebat, D. 2003. Renewal of the neurophysiology of language: functional neuroimaging. Physiological Review 85: 49-95. Fillmore, C. and Atkins, B. 1992. Towards a frame-based lexicon: The semantics of RISK and its neighbours. In A. Lehrer and E. Kittay (Eds.) Frames, fields, and contrasts, 75-102. Hillsdale, N.J.: Erlbaum. Hurford, J. 2003. The neural basis of predicate-argument structure. Behavioural and Brain Sciences 26: 261-316. Jackendoff, R. 2002. Foundations of language. Oxford: Oxford University Press. Joanette, Y., Goulet, P. and Hannequin, D. 1990. Right hemisphere and verbal communication. New York, Berlin and Heidelberg: SpringerVerlag. Lamb, S. 1998. Pathways of the brain. Amsterdam/Philadelphia: John Benjamins. Levelt, W.J.M. 1989. Speaking: From intention to articulation. Cambridge, Mass. and London: The MIT Press. MacWhinney, B. 1999. The emergence of language from embodiment. In B. MacWhinney (Ed.), The emergence of language, 213-256. New Jersey: Lawrence Erlbaum.

Michael Fortescue

189

Martin, A. and Caramazza, A. 2003. Neuropsychological and neuroimaging perspectives on conceptual knowledge: an introduction. Cognitive neuropsychology 20(3/4/5/6): 195-212. Miller, G.A. and Johnson-Laird. P.N. 1976. Language and perception. Cambridge, Mass.: Harvard University Press. Morton, J. 1969. The interaction of information in word recognition. Psychological Review 76: 165-178. Perani, D., Capa, S.F., Schnur, T., Tettamanti, M., Collina, S., Rosa, M.M., and Fazio, F. 1999. The neural correlates of verb and noun processing. A PET study. Brain 122: 2337-2344. Pulvermüller, F. 2002. The neuroscience of language. Cambridge: Cambridge University Press. Pustejovsky, J. 1995. The generative lexicon. Cambridge, Mass.: The MIT Press. Talmy, L. 2000. Towards a cognitive semantics. Cambridge, Mass. and London, England: MIT Press.

III. METAPHOR

MORE THAN A METAPHOR: METAPHORIC STRATEGIES IN DYNAMIC PROCESSES OF MEANING CONSTRUCTION CLAUDIO VANDI AND RICCARDO FUSAROLI1

Abstract. This chapter aims at presenting a conception of metaphors as the result of dynamic processes of meaning construction. Through the analysis of Conceptual Metaphor Theory, Blending Theory and the metaphor “that man is a caiman”, we highlight the need for a dynamic perspective on embodied cognition, where sedimented conceptual structures are locally deployed as resources and not as determining types. Considering cultural, discursive, lexical and co-textual factors, it is possible to describe the metaphor as a local diagram of semantic relations that can be manipulated, as a part of wider interpretative strategies. Through this analysis, a conception of schemata and conceptual structures as attractors sedimented in the course of practices is proposed.

Introduction An ever growing number of studies persuasively demonstrates that everyday cognition is continuous and dynamic, not a punctual transformation of stable states, but a process of partially overlapping fuzzy grey areas that are drawn over time (Thelen and Smith 1994; Port and van Gelder 1995; Clark 1997; Spivey 2006). The aim of this chapter is to situate metaphorical occurrences in this unfolding of cognitive processes. Assuming a particular kind of Cognitive Semantics approach – in which “meaning is cognition” (Evans and Green 2006) and cognition is ecological and deeply shaped by socio-cultural practices – we will focus on linguistic data and not on the experiences of empirically tested subjects. This perspective does not need to neglect discursive frames and contexts. 1 While the article has been thoroughly thought and discussed together, Claudio Vandi has materially written “Between embodiment and situatedness” and “Deep into the analysis” except “Which blends” and “Metaphors and diagrammatic reasoning”; Riccardo Fusaroli: “Introduction”, “Cognitive linguistics and metaphors” and “Conclusions”.

194

Chapter Thirteen: Metaphoric Strategies in Dynamic Processes of Meaning Construction

Since cognition is constituted by continuous overlappings, the construction of the meaning related to a specific linguistic expression will emerge in a wider field of expectations. This kind of analysis aims at tracking the conditions through which the relation between text and context emerges, the way a more schematic conceptual structure is constitutively situated and the reason why a configuration of meaning should be conceived as a moment in a trajectory rather than as a stable state. In this sense, the first part of the chapter is dedicated to the review of two of the main approaches to metaphors in Cognitive Linguistics (Conceptual Metaphor Theory, from now on CMT, and Blending Theory, from now on BT), highlighting the unsolved tension which they create between the constitutive dynamicity of meaning construction and the existence of stable and embodied conceptual structures. The second part of the chapter aims at addressing this tension, by sketching a way in which to account for the processual nature of meaning construction as unfolding in social arenas, mixing textual analysis with theoretical reflections. We believe that the real challenge for the scientist lies more in the investigation of the process than in the investigation of the state (Vygotsky 1978); that it is the trajectory that constitutes the meaning of the single moment and not vice versa.

Cognitive linguistics and metaphors The study of metaphors goes far back in time and – contrary to what is often reconstructed – metaphors were not always treated as rhetorical ornaments. The importance of analogy for constructing knowledge was acknowledged and investigated by Aristotle (cf. Eco 2005) Kant, Cassirer, Blumemberg, Weinrich (cf. Jäkel 1999), and with his usual emphasis Peirce stated that metaphors are “the only logical operation which introduce any new idea” (Peirce 1931-1958: 5.171). However, it is only with the cognitive turn in the study of metaphors – to which Cognitive Linguistics has greatly contributed (Lakoff and Johnson 1980; Ortony 1993; for an introduction, see Evans and Green 2006) – that the central role they play in processes of thought has entered mainstream research. Abstract concepts are re-defined as structured through the repetition – in CMT – or the construction – in BT – of patterns. The role of metaphors is to make possible the unfolding of meaning through acts of conflation, and through the connection of different conceptual domains thus make possible the understanding of something new through something already known.

Claudio Vandi and Riccardo Fusaroli

195

Conceptual Metaphor Theory CMT emerges mainly from the works of Lakoff and Johnson (Lakoff and Johnson 1980, 1999; Lakoff 1987; Johnson 1987). Its aim is to uncover the conceptual structure that underlies language use. In this model, imageschemata are specific and recurring action paths formed along time by embodied experience. They act as regularities to orient future experiences and are open to reconfiguration. Cognition also works through the projection of these bodily-grounded patterns onto uncharted conceptual domains, giving reasons for claiming that abstract thought too has a bodily basis. Metaphorical mapping is the cognitive mechanism performing the projection, and metaphors are the sedimented results of repeated projections. New metaphors have the power to create a new reality. This can begin to happen when we start to comprehend our experience in terms of a metaphor, and it becomes a deeper reality when we begin to act in terms of it (Lakoff and Johnson 1980: 146).

If CMT tends to emphasise the uni-directionality of this metaphoric projection, from concrete domains to more abstract ones, it has to be stressed that these concrete patterns, these image-schemata, are not that simple and mono-dimensional (cf. Hampe 2005). Image-schemata are related to sensorimotor patterns, but these patterns emerge in the blending of culture, language, history and bodily mechanisms (Johnson 1987). Image schemata are relatively abstract conceptual representations that arise directly from our everyday interaction with and observation of the world around us (Evans and Green 2006: 176).

This is a world that is deeply shaped by social and cultural practices. However, in CMT, this complexity is problematic. A first stage of the conceptual formation consists of a complete conflation of two basic experiential domains (e.g. MORE IS UP, where quantity and verticality are conflated through the “scale” schema) due to their experiential correlations. These “primary metaphors” emerge “automatically and unconsciously” as we experience various concrete and “subjective” events together. Soon, however, there is a second stage of development, “differentiation”, in which we recognize these conflated domains as different, thus separating them conceptually. Conceptual metaphors – emerging in this phase – are systems of more conventional conceptual mappings held in long-term memory but still grounded in embodied experience (e.g. TIME IS MONEY, in which the more abstract domain of

196

Chapter Thirteen: Metaphoric Strategies in Dynamic Processes of Meaning Construction

time is structured by an economic domain through a set of more basic image schemata related to UNITY/MULTIPLICITY, OBJECT, BALANCE, etc.). The conflation is thus replaced by a more structured projection claimed to be an invariant. In a related way, the analytical process performed by CMT (cf. also Lakoff and Turner 1989) mainly consists of the individuation of types under which discursive metaphorical occurrences are subsumed without focusing on their local differences. What we are looking at here, is what Kant called a determinative judgement, the application of a rule (Kant 1781-1787). But Kant himself described different kinds of judgements, the reflective and aesthetic ones (further developed by Peirce through the concept of abduction, Peirce 1931-1958). In these judgements, or cognitive processes, we do not have a deterministic rule to apply. We have different regularities, different habits, Peirce would say, that are applicable. Cognition is like bricolage, using the different resources at hand and playing with them. To account for this dynamic functioning, a semantic analysis should show how cognitive resources are employed in the development of linguistic phenomena, not neglecting the specificities of the single metaphors and their relation to the context.

Blending theory BT has been developed in the works of Fauconnier and Turner (2002) as derived from the Mental Spaces Theory of Fauconnier (1985).2 The first crucial step that separates BT from CMT is the focus on mental spaces rather than directly on conceptual domains. In CMT conceptual metaphors are defined as sedimented projections from one domain of established knowledge to another. In BT the focus is on partial and temporary representational structures that speakers construct when thinking or talking. Mental spaces are thus certainly structured by domains, but they are of a more unstable and local nature. A metaphor can be described by a specific kind of blending between two or more of these mental spaces. The “blending” – made possible by a partially common core structure –

2

Mental Spaces have been criticised for focusing only on the “inside of the skull” conception of cognition. In this chapter, we have in mind Brandt and Brandt’s (2005) version of the theory, where mental spaces are constituted through interactions with the environment and pertinences established in a more distributed Semiotic Space.

Claudio Vandi and Riccardo Fusaroli

197

generates a new and richer structure, which cannot be reduced to the sum of its parts.3 Thus, metaphorical mappings are not resolved in the repetition of preestablished patterns of embodied meaning but in more general mental operations which actively construct patterns and “spaces” of meaning. The strong focus on embodiment that CMT built seems to be at least partially lost here. Blending operations can certainly be grounded in neural processes, but they seem to be free from embodied image-schemata in their creation of patterns, even though they have as a goal the achievement of blends – not better specified – at a “human scale” (Fauconnier and Turner 2002: 312).

Critical evaluation Through CMT it is possible to appreciate the embodied nature of our conceptual patterns, and the way these metaphorical structures and projections imbue the way we think and talk. Through BT we are able to observe the emergence of new meanings and patterns, to focus on local occurrences (thus describing not only linguistic metaphors determined by conceptual ones, but metaphors in general in their putting into play linguistic and conceptual elements). Unfortunately in BT, the strong aspect of embodiment that was constructed in CMT is lost. To overcome this unsolved tension, the reflection on metaphors needs to take some additional steps: x x x

to acknowledge the situatedness of metaphors in their environment (culture, genre, situation, discourse and practice); to acknowledge the dynamics between language and cognition, which cannot be reduced to a determinative judgement; to acknowledge the dynamic nature of conceptual and pre-conceptual structures which cannot possibly be reduced to rigid types.

Thus embodiment can be preserved and the implied conception of cognition in cognitive semantics can be re-attuned to other developments in the cognitive sciences (ecological psychology, social cognition, distributed cognition, dynamic systems approaches, etc.) 3

In particular, a blend is considered metaphoric when: (1) a single element in the blend corresponds to an element in each of the input spaces; (2) certain very salient aspects of the input domains are prohibited from entering the blend; (3) some salient structure in the blended space is prevented from floating back to the inputs; (4) there are degrees of metaphoricity (cf. Grady, Oakley and Coulson 1999).

198

Chapter Thirteen: Metaphoric Strategies in Dynamic Processes of Meaning Construction

Between embodiment and situatedness A political metaphor In 2004, the most important Italian newspaper, at least in terms of sales, and readership, La Repubblica, published the following metaphor: “that man is a caiman” (in the original version: “quell'uomo è un caimano”). Analysing this as an independent token of language would leave us puzzled. Even dwelling on the established conceptual metaphor PEOPLE ARE ANIMALS, which traits of ‘caiman’ are projected onto ‘man’? Or, employing Fauconnier and Turner, which common structure can be found and which non-common traits can be blended, giving rise to new meaning? It is hard to say. The analysis needs to account for the great relevance of context in creating expectations through which a common structure can be re-constructed, and the emergence of new meaning can happen. In order to do so we need to critically analyse the work of two authors that, while studying metaphors, focus on the issue of cultural variation (Kövecses 2002, 2005) and on the relevance of genre (Steen 1994).

The cultural element A fundamental study of the cultural influence in metaphors is Metaphor and culture by Zoltán Kövecses (2005). Kövecses attempts a systematic investigation of Chinese, Hungarian and English, in order to show how differences exist both in the way the source domain is used to conceptualize the same target domain (e.g. both English and Chinese use UP and LIGHT to conceptualize happiness, but only Chinese has HAPPINESS IS FLOWERS IN THE HEART) and in the way the same source is used to conceptualise different targets domains (e.g. the BUILDING source, that is used in English to speak of relationships, but not in Tunisian Arabic). These cross-cultural variations are often used as an argument against the claim that it is possible to sketch an at least potentially limited list of conceptual and metaphorical universals that shape human cognition, universals that are grounded in our embodied experience and extended to abstract domains via metaphorical mappings (Lakoff and Johnson 1980; Lakoff 1987; Gibbs 1998). Kövecses (2002, 2005) defines two main causes usually responsible for cross-cultural variations: the Broader Cultural Context from which a metaphor arises and the Natural Physical Context constituting its source domains.

Claudio Vandi and Riccardo Fusaroli

199

However, cultural idiosyncrasies could be easily integrated at a higher level, given the very abstract structure of primary metaphors (Grady 1997; Grady, Oakley and Coulson 1999) on which conceptual metaphors are grounded. A better approach, according to Lakoff and Johnson (2002), is to define “specific level differences” according to how a specific metaphor is instantiated by a specific culture in a specific context, leaving to anthropologists the task of investigating the way the general schema is “filled out” differently between cultures (Kövecses 2005).

The discursive element This takes us to the second sense in which the expression “variation in metaphors” can be interpreted, namely in the sense of intra-cultural variations. This means focusing on the discursive nature of metaphors and analyzing how the same conceptual metaphor can produce different mappings when instantiated in different contexts. A common hypothesis is that in conceptual metaphors certain schematic elements get mapped from the source domain onto the target domain without changing their basic structure. Olaf Jäkel’s investigation of religious discourse in Western culture (Jäkel 2002) maps out the variations of LIFE IS A JOURNEY in this specific context, despite the highly stable set of mappings that this metaphor usually shows (Lakoff 1990, 1993). While a great number of linguistic occurrences can be found – for instance “following God’s path” – we can note some differences in the source domain of JOURNEY. While this metaphor usually furnishes the possibility of defining multiple paths through which one can reach one’s goal, an analysis of many passages of the Old Testament shows some interesting variations: 1.

You must follow exactly the path that the Lord your God has commanded you, [...]

2.

My foot has held fast to his steps; I have kept his way and have not turned aside. (examples taken from Jäkel 2002)

These sentences clearly show that multiple paths are not given, i.e. they are not inferential possibilities of LIFE IS A JOURNEY in this kind of religious discourse. Analogous variations can be found in many discursive genres. In business discourses, “to take a shortcut to reach one's own goal” is quite a widespread and generally positive metaphor, while in a religious text it would not be so and the few occurrences are generally used to describe the negative habits of sinners and people who are weak in their faith.

Chapter Thirteen: Metaphoric Strategies in Dynamic Processes of Meaning Construction

200

Defining context The conception of conceptual metaphor as a rigid type to be deterministically applied to local occurrences is now more clearly denied in favour of a more dynamic structure constitutively varied in different contexts. This dynamic conception is perfectly in line with certain experimental developments in cognitive poetics. Conceptual structures – prototypes in the quoted passage – emerge as “temporary constructions in working memory constructed on the spot from generic and episodic information in long-term memory, rather than as stable structures stored in long-term memory” (Gibbs 2003:32). Taking into account the previous observations and further elaborating with usually less considered factors, we can sketch a non-exhaustive list of elements to be taken into account in our analysis in order to define how a metaphor emerges: 1.

Socio-historical constraints: previous instantiations of the same metaphor, context of reference;

2.

Discursive context: the genre; the situation;

3.

More local expectations, like the sedimented style of the author, the specific subgenre, etc.

4.

The co-text: the sentences that surround the analysed metaphor.

5.

Lexical choices.

Deep into the analysis Style of the text and sentence processing To analyse “that man is a caiman” we have to sketch the field of expectations, or habits, that the style of the text gives rise to (Kövecses 2005). The metaphor appeared in a political commentary by Franco Cordero in La Repubblica, a leftist newspaper. Thus the expectations in its interpretation are of a limited amount of complexity, since a newspaper is not usually subjected to in-depth reading (Steen 1994). The co-text makes it easy to understand that the referent of ‘man’ is Silvio Berlusconi, a man well known to the Italian audience and towards whom both the newspaper and the author of the article have always displayed a strong critical stance, something which is shared by their audience. We could sketch the sedimented knowledge of Berlusconi in this

Claudio Vandi and Riccardo Fusaroli

201

context as follows: Berlusconi is a specific individual, owner (among other things) of three national television channels, the most influential Italian advertising company and a very successful football team. At the time in which the article was published Berlusconi was the Italian Prime Minister, economically and politically involved in many Italian and international industries, and undergoing a good deal of trials for corruption, financial crimes and never clarified connections with well-known criminal organisations. The sedimented knowledge to which ‘a caiman’ gives access is more complex to analyse. The caiman is a South American relative to the crocodile, looking slightly more slender than its African counterpart. There is also a less defined awareness of the fact that it lives in rivers, of its thick skin, and of its voracity. These contextual expectations are part of the meaning of the metaphor (Steen 1994; Kövecses 2005) and influence the way in which “entrenched associations in long-term memory” are recruited in the on-line processes (Grady et al. 1999) so that the potentialities of each term are already constrained by the context and meaning construction is conceived as distributed.

Which blends? It can easily be said that “that man is a caiman” is a token of the more general PEOPLE ARE ANIMALS conceptual metaphor and that the common mapping in these cases consists of an attribution to the person concerned of some moral quality related to traditional representations of the animal, or behavioural traits recontextualised, and so on. The problem is that in Italian culture, there is no traditional representation of the caiman. A second powerful element in this metaphor is the potential creation of a sharp “image metaphor”, an image-based metaphor that is rich in imagistic details (Lakoff and Turner 1989; Kövecses, 2002). There is in Italian culture quite a common imagery of crocodiles, to which, for lack of a better understanding, caimans are assimilated. But which images are to be used? The image of “crocodile tears”, of an immobile crocodile sleeping on the bank of a river, or of its powerful jaws at work? The only way in which the blending can be oriented without needing to “posit specific mechanisms that override parts of source-to-target-domain mappings in primitive metaphors” (Gibbs 2006: 118) is by sticking to the embodied basis of metaphorical thought. But to articulate the mappings that are triggered by the metaphor, to reconstruct a common structure that allows the idealised subject to construct new meaning, we also have to rely on the contextual expectations previously highlighted. The prejudiced

202

Chapter Thirteen: Metaphoric Strategies in Dynamic Processes of Meaning Construction

audience makes negative traits highly relevant. The strong presence of Berlusconi in the media, his active invading of any possible field, makes the sleeping image of the caiman less relevant. This activity reveals instead a deep commonality with the image of the caiman’s jaws, a commonality which emerges from this unfolding and tentative comparison. The common structure which allows for blending is that of a common force dynamics (Talmy 2000), that is seen through a dynamic interaction of the two mental spaces. Just like the jaws of the caiman, conceived as an agonist, are able to overcome any resistance in what they are chewing (the antagonist), Berlusconi is able to destroy any boundary and to break any resistance to his invasions.

The lexical element In our description of the entailments that follow from the use of ‘caiman’ as a source space, we referred to the ‘crocodile’ as an animal to which the caiman is usually associated in Italian culture. But we have not yet considered the lexical aspects of this metaphor. Could the analysis made up to now have worked equally well for “that man is a crocodile”? The term ‘crocodile’ would have been simpler to comprehend since the caiman is quite a marginal figure in the Italian zoological encyclopaedia. The use of ‘caiman’ implies a less structured and vaguer mental space. Idiomatic expressions like “to cry like a crocodile” (referring to people that cry for the negative outcomes of actions for which they are nevertheless responsible and which they would probably repeat) would constrain the space of interpretation and make the emergence of a more adequate blend in which voracity is the emergent meaning, more difficult. With a stronger source space (with the more established conceptual structures related to ‘crocodile’), the metaphor would have been subjected to a more directional mapping, while with a weak source the blend is due to negative aspects of both the source and the target, thus producing a more negative portrayal of the target (Berlusconi). The cultural specificities of one source space are thus fundamental if we want to explain how metaphors are used to conceptualize experience inside a particular culture. Concrete elements used as sources are always inherently culturally embodied: “recognizing that what is cognitive (and embodied) is inherently cultural should be a fundamental part of how we do our work as cognitive psychologists, linguists, and anthropologists” (Gibbs 1999: 156).

Claudio Vandi and Riccardo Fusaroli

203

Journalistic metaphors One possible critique to this analysis could be that we have not taken into account evidence from Steen (1994) concerning the differences between reading journalistic metaphor and literary metaphor. Steen, after having set up many observations and experiments on subjects reading both literary and journalistic texts, states that when dealing with literary texts readers more often do the following: 1.

Process the focus verbalizations.

of

metaphors

in

thinking-out-loud

2.

Build contexts in terms of author intentions.

3.

Identify metaphors explicitly.

4.

Express their judgement.

5.

Tend to re-conceptualise metaphors at later stages of the reading process.

While these findings are highly relevant, their quantitative perspective has to be integrated into a more local and qualitative approach. The fact that we are dealing with a newspaper is not per se enough to state anything about the way readers will interpret our metaphor. When analysing “that man is a caiman”, we are dealing with a particular journalistic text. Franco Cordero, the author, is renowned for his complex articles where Latin quotations, puns and highly specialised juridical lexicon are common. This has led to the creation of a very specialised audience that quickly adopts and experiments with his metaphors. Moreover, La Repubblica has a tradition of constructing strong identities for its best journalists and editorialists, leading to a situation similar to the second point highlighted by Steen on the specificities of literary metaphors: the attention to (reconstructed) authorial intentions.

Metaphors and diagrammatic reasoning The model of meaning construction underlying metaphorical occurrences that we are sketching is a tentative procedure in which sedimented structures are employed as resources, and re-deployed, in order to make new meaning emerge. Conceptual structures are thus conceived not as rigid types, but as more dynamic and evolving configurations, something that it is possible to manipulate in social practices. In particular the process described can be conceptualised as a diagrammatic reasoning: “By

204

Chapter Thirteen: Metaphoric Strategies in Dynamic Processes of Meaning Construction

diagrammatic reasoning, I mean reasoning which constructs a diagram according to a precept expressed in general terms, performs experiments upon this diagram, notes their results, assures itself that similar experiments performed upon any diagram constructed according to the same precept would have the same results, and expresses this in general terms” (Peirce 1976 Volume 4: 47-48). A diagram is a particular configuration of more general relations that can be manipulated, in order to make new inferences possible. Through the semantic processes highlighted we have actually constructed particular configurations of semantic traits that can be manipulated. As already reported, Peirce stressed the idea that metaphors were the only way in which a singularity could generate a regularity, in which a new concept could emerge. From the semantic diagram we constructed, many consequences, many new interpretative tentative habits could be drawn. Any new action performed by Berlusconi could be conceptualised as expressions of his caiman-like voracity, of his need to invade every public field, to increase public charges crushing any boundary and resistance. The whole of Berlusconi’s past and future career is re-read in this way. Not even the (poor) caiman is left untouched. It is now hard to say caiman in Italian without stressing its voracity, which was not so salient before; or without summoning the figure of Berlusconi. An entire movie, based on this metaphor, namely Il Caimano by Nanni Moretti, later on definitely stabilised this process. The diagrammatic nature of this process escapes any rigid schematicity, bringing to the foreground the dynamic nature of meaning in general and of metaphors in particular. Indeed, new interpretative habits and new contexts deeply reshape the metaphor. Let us make a thought experiment. Let us imagine a right wing newspaper (generously funded by Berlusconi) publishing the text “of course that man is a caiman: he needs a thick skin to cope with all the insults that the communists throw at him”. Here we can find a reverted FORCE DYNAMICS schema in which Berlusconi as an agonist has to resist the pressure of the antagonist (the political rivals). Another possibility could be “of course that man is a caiman, the turbulent waves of Italian politics require no less to be tamed. Thus Berlusconi is exactly the man we need”. We leave to the reader the pleasure of finding the trajectory through culture and conceptual structure that crosses this manipulation of the metaphor. .

Claudio Vandi and Riccardo Fusaroli

205

Conclusions: On variations and attractors Adopting a dynamic (though qualitative) approach to cognition, we have sketched a theoretical and analytical path through which metaphors can articulate the known and the unknown, regularities and singularities. Linguistic metaphors trigger processes of meaning construction that are diagrammatic and not determined by a strict type-token logic. Different semantic traits are available to be attempted and stabilised in a field of expectations. Their stability is also defined according to their conceivable consequences on the following cognitive processes and on the influence they will have on further discourses. In this model, schemata and established conceptual structures like conceptual metaphors do not lose their importance. They are simply pushed further away from any possible conceptualisation as rigid types determining local tokens. They are structures culturally and ontogenetically emerging from local variations and they keep this dynamics. Thus they can be conceived as attractors (Thelen and Smith 1994), i.e. patterns of recurring experience that help a system organise itself through the memory of the previous states. We are not alone in stating this. Gibbs has already proposed that schemata should be conceived as attractors which profile depends “on the overall state of the organism involved in some activity and past basins of attractions created within the system” (Gibbs 2006:115). Our contribution is thus to have sketched a (very tentative) model of the way these attractors mediate between situatedness and generality, and the way in which they are re-deployed in processes of meaning construction. Processes which can be defined as diagrammatic and which are constitutively situated. Language emerges as an arena for thought rather than as a place where thought is expressed. Language is a symbolic artefact and environment through which thought emerges and is engaged in its human specificities (Clark 2006), through which conceptual structures are put into play and potentially remodulated. To use and to understand conceptual structures, but also local metaphorical occurrences, involves at least an implicit understanding of their conceivable consequences, and of the way in which they can be put to work in the unfolding cognitive process, in the dynamic field of expectations in which meaning construction processes take place. And since language use is prominently social and public, “conceptual” here does not mean that metaphors are inner and subjective facts, but on the contrary that they are among the strongest tools social communities have to share and guide their experience. Metaphors can be used to enact (Varela et al. 1991; Gibbs 2006) the embodied, concrete experience in our

206

Chapter Thirteen: Metaphoric Strategies in Dynamic Processes of Meaning Construction

mental activity; as Gibbs (1999) states it by quoting Kirmayer (1992): “metaphors are tools for working with experience”. It is indeed high time to […] move metaphor out of our heads and put it into the embodied and public world […] (This move) does not make metaphor any less cognitive than if we had long lists of metaphors nicely encoded in our heads. All this move attempts to do is acknowledge the culturally embodied nature of what is cognitive and to suggest that there is much less of a difference between what is cognitive and what is cultural than perhaps many of us have been traditionally led to believe. (Gibbs 1999:162).

References Brandt, P. Aa. and Brandt, L. 2005. Making sense of a Blend. Annual Review of Cognitive Linguistics 3(1): 216-249. Clark, A. 1997. Being there. Cambridge Mass.: MIT Press. —. 2006. Material symbols. Philosophical Psychology 19(3): 1–17. Eco, U. 2005. Metafora e Semiotica Interpretativa. In A.M. Lorusso (Ed.) Metafora e Conoscenza. 257-290. Milano: Bompiani. Evans, V. and Green, M. 2006. Cognitive Linguistics. An introduction. Edinburgh: Endinburgh Press. Fauconnier, G. 1985. Mental spaces: Aspects of meaning construction in natural language. Cambridge, MA: MIT Press. Fauconnier, G. and Turner, M. 2002. The way we think. New York: Basic Books. Gibbs, R. 1998. Cognitive science meets metaphor and metaphysics. Minds and Machines 8: 433–436. —. 1999. Taking metaphor out of our heads and putting it into the cultural world. In R. Gibbs and G. Steen (Eds.) Metaphor in Cognitive Linguistics, 145-166. Amsterdam: John Benjamins. —. 2003. Prototypes in dynamic meaning construal. In J. Gavins and G. Steen. Cognitive poetics in practice, 27-40. London and New York: Routledge. —. 2006. Embodiment and cognitive science. Cambridge: Cambridge University Press. Gibbs, R. and Steen, G. (Eds.) 1999. Metaphor in Cognitive Linguistics. Amsterdam: John Benjamins. Grady, J. 1997. Theories are building revisited, a typology of motivation for conceptual metaphor. Cognitive Linguistics 8: 267–290.

Claudio Vandi and Riccardo Fusaroli

207

Grady, J., Oakley, T. and Coulson, S. 1999. Metaphor and blending. In Gibbs, R. and G. Steen (Eds.). Metaphor in Cognitive Linguistics. 101124. Amsterdam: John Benjamins. Hampe, B. 2005. When down is not bad, and up not good enough: A usage-based sssessment of the plus–minus parameter in image-schema theory. Cognitive Linguistics 16(1): 81-112. Jäkel, O. 1999. Kant, Blumenberg, Weinrich: Some forgotten contributions to the cognitive theory of metaphor. In R. Gibbs and G. Steen (Eds.) Metaphor in Cognitive Linguistics, 9-28. Amsterdam: John Benjamins. —. 2002. Hypotheses revisited: The cognitive theory of metaphor applied to religious texts. Online: http://www.metaphorik.de/02/jaekel.htm Johnson, M. 1987. The body in the mind. Chicago: University of Chicago Press. Kant, I. 1781-87. Kritik der reiner Vernunft, in Kants gesammelte Schriften, III-IV, Berlin-Leipzig 1903-04. Kirmayer, L. 1992. The body’s insistence of meaning: Metaphor as presentation and representation in illness experience. Medical Anthropology Quarterly 6: 323-346. Kövecses, Z. 2002. Metaphor: A practical introduction New York: Oxford University Press. —. 2005. Metaphor and culture. Cambridge: Cambridge University Press. Lakoff, G. 1987. Women, fire and dangerous things. Chicago: Chicago University Press. —. 1990. The Invariance Hypothesis: Is abstract reason based on image schemas? Cognitive Linguistics 1: 39-71. —. 1993. The contemporary theory of metaphor. In A. Ortony (Ed.) 1993, 202-251. Lakoff, G. and Johnson, M. 1980. Metaphors we live by. Chicago: University of Chicago Press. —. 1999. Philosophy in the flesh. Chicago: University of Chicago Press. —. 2002. Why cognitive linguistics requires embodied realism. Cognitive Linguistics 13(3): 245–263. Lakoff, G. and Turner, M. 1989. More than cool reason. Chicago: University of Chicago Press. Ortony, A. (Ed.) 1993. Metaphor and thought (2nd edition). Cambridge: Cambridge University Press. Peirce, C.P. 1931-1958. Collected papers of Charles Sanders Peirce, voll. I – VI edited by C. Hartshorne and P. Weiss, 1931-1935, voll. VII– VIII edited by A.W. Burks, 1958. Cambridge, Mass.: Belknap Press.

208

Chapter Thirteen: Metaphoric Strategies in Dynamic Processes of Meaning Construction

—. 1976. The new elements of mathematics by Charles S.Peirce. Vol. I-IV (ed. by Carolyn Eisele). The Hague-Paris/Atlantic Highlands, NJ: Mouton/Humanities Press. Port, R. and van Gelder, T. 1995. Mind as motion. Cambridge, MA: MIT Press. Spivey, M. 2006. The continuity of mind. New York: Oxford University Press. Steen, G. 1994. Understanding metaphors in literature. London & New York: Longman. Talmy, L. 2000. Towards a cognitive semantics. Cambridge (MA): MIT Press. Thelen, E. and Smith, L. 1994. A dynamic approach to the development of cognition and action. Cambridge MA: MIT Press. Varela F., Thompson, E. and Rosch, E. 1991. The embodied mind. Cambridge MA: MIT Press. Vygotsky, L.S. 1978. Mind in society. Cambridge: Harvard University Press.

GOING TOWARDS THE UNKNOWN: EXPRESSIONS FOR DYING IN EUROPEAN LANGUAGES ANNA VOGEL

Abstract. The chapter investigates expressions for dying in five European languages with an onomasiological approach. The data was collected from questionnaires, dictionaries, web resources and literature, both fiction and non-fiction. The results indicate that expressions for dying show great similarities across the languages. Expressions involve verbs with high agentivity. A location (where one will be after death) is often specified. Expressions often make use of trivial tasks, such as “put down the receiver”. Further, death is associated with lack of breath and sleep. Here, the study supports earlier research. Finally, a seemingly paradoxical way of relating death to both cold and heat is explained by discussing specific conceptualizations, separating the body from the location of the body.

Introduction The present chapter aims at studying patterns of metaphorical expressions for ‘dying’ in a few European languages. Thus the perspective is onomasiological: the concept of ‘dying’ and its linguistic realisations are in focus. According to Berendt, Maeda and Tanita (in press), expressions for ‘dying’ in Japanese and English involve the following conceptual metaphors: DEATH IS DEPARTURE, DEATH IS THE END OF A JOURNEY, DEATH IS SEPARATION and DEATH IS SLEEP. François (in press) shows that the ‘act of breathing’ is strongly tied to ‘life’ in many languages. The present study takes a theoretical vantage point in the metaphor theory as expressed by Lakoff (1993). According to Lakoff, a metaphor should be understood as a conceptual phenomenon, involving mappings across domains. The data include Swedish, Norwegian, Danish, German and French expressions for ‘dying’. For all languages except Swedish, the data were collected by questionnaires that were filled in by linguists who had the pertinent language as their first language. One linguist from each language

210

Chapter Fourteen: Expressions for Dying in European Languages

participated. The questionnaire was designed for the study. For Swedish, the data were extracted from questionnaires filled in by students. The number of student was 25. Here, too, the questionnaire was designed for the study. Data were also collected from dictionaries (the main contemporary dictionary from each language), web resources and from other fiction and non-fiction texts. All data were collected in 2007. The results are related to agentivity, location and degree of triviality. Further, the results are discussed in relation to metaphor theory where the domains of temperature, (lack of) light, breathing and sleeping are mapped onto the domain of dying.

What do we do and where do we go? We know very little about dying. To die is a transition from one state (life) to another (death) and of all the transitions we are subject to, this is the one that we know the least about. Three points can be made concerning this transition. First, dying is most of all a passive transition (suicides omitted). Second, we do not know what “happens” after the transition, although religion and myths provide us with various alternatives such as coming to paradise or hell. If something that we call “the soul” is taken to another place, we do not know anything about this place. Third, although every human being must die, death is not something trivial. For oneself and for friends and family, death can be, and often is, associated with anxiety and despair. Therefore, it is interesting that the results from the study show that expressions for ‘dying’ ignore exactly these three points.

High agentivity Expressions for ‘dying’ often show high agentivity, as shown in (1). In general, ‘dying’ is a transition during which the dying person is subject to a process he or she cannot influence. The transition may very well involve activities such as a struggle, but the main point here is that the transition from life to death often has a passive course in terms of intentionality and voluntariness. The Swedish verb gå ‘go, walk’ in (1) is intentional and involves self-propelled motion. Another example is given in (2). The German verb abtreten ‘step down’ is, like gå ‘go, walk’, intentional and includes physical motion. (1)

när deras pappa gick when their father go.PAST ‘when their father passed away’

bort away

Anna Vogel

(2)

von der from ART.DEF.DAT ‘to get off the stage’

211

Bühne abtreten stage step down

The reason why expressions show high agentivity despite the low agentivity of ‘dying’ is probably that human beings tend to avoid mentioning threatening concepts by their real names. Death is typically such a threatening concept, as is sex. The denial of the involuntariness of ‘dying’ helps us to approach the transition. In general, taboo concepts tend to generate a high number of synonyms or equivalent expressions (Halliday 1978: 165; Carter 1987: 93), which is confirmed in the present study.

Location is specified Expressions for ‘dying’ often include a specified location, as in (3). The Danish noun jagtmarker ‘hunting-grounds’ indicates a certain place that will be reached after the transition. In (4), an example from the French data is given. (3)

drage pull

til to

de ART.DEF

evige eternal

jagtmarker hunting-ground+PLUR ‘go to the happy hunting-grounds’ (4)

aller au go to.ART.DEF ‘go to heaven’

ciel sky

Both (3) and (4) point out a location in a religious context. Note that both (3) and (4) also include verbs with high agentivity. Other expressions involve the physical place where the corpse is buried, as in (5). (5)

men but

då then

ligger lie+PRES

under under ‘but then you are six feet under’

du you

sex six

fot foot.PLUR

212

Chapter Fourteen: Expressions for Dying in European Languages

The Swedish example (5) is collected from pop lyrics. The expression also occurred in the data collected from students.1 When comparing expressions such as (3)–(4) and (5), it turns out that the type where the specified location has a religious context is the most common in the data. This can be spelled out even more directly, as in the Norwegian example shown in (6). (6)

gå hjem go home ‘go to God’

til to

Gud God

In (6), God is mentioned directly, but expressions such as Swedish chefen ‘the boss’ or änglarna ‘the angels’ also occur in the data. In those cases, God is mentioned indirectly, or is implied. Then, the denial of the involuntariness of ‘dying’ helps us to approach death, and the denial of the ignorance of what happens after death gives us courage to talk about it.

Dying is something trivial A good deal of the expressions for ‘dying’ involve a trivial act; something which is performed in our everyday life. The Norwegian example (7) shows such an expression. (7)

legge på røret put down receiver+DEF ‘put down the reciever’

The act of hanging up a phone call is trivial. Another example, this time from Danish, is given in (8). (8)

stille træskoene put away clog+PLUR+DEF ‘put away the clogs’

Clogs are (or have been) traditional footwear, and they are put away every night before going to bed. Here, there is also a connection to sleep, which will be discussed later. Again, something non-threatening/trivial is used for something threatening (death). These trivial expressions may be 1

The expression is clearly influenced by English. The distance six feet indicates the depth of traditional Christian graves although today, this dimension varies from cemetery to cemetery.

Anna Vogel

213

contrasted to example (3), (4) and (6), where a religious context is evoked. It could be discussed whether example (7) and (8) represent a secularized attitude towards death. Further studies would be needed in order to investigate if and how societies a few hundred years back in time, societies, which were probably more religious than the Swedish and Danish societies are today, used expressions like (7) and (8).

Conceptual metaphors for ‘dying’ Berendt, Maeda and Tanita (in press) found that in Japanese, as well as in English, death is associated to sleep. The results of the present study support their research. In the next sections, the metaphor DEATH IS SLEEP will be discussed along with the metaphors LIFE IS BREATHING, DEATH IS DARKNESS and DEATH IS COLD.

Death is sleep Expressions involving sleeping are common in the data. In German, the expression is entschlafen ‘fall asleep, pass away’, in Swedish, it is somna in ‘fall asleep, pass away’. In general, these expressions are not used for ‘sleeping’ anymore. In German, the verb einschlafen ‘fall asleep’ is preferred while in Swedish, somna ‘fall alseep’ without the particle is preferred. Sleep is interesting, since it is trivial (something we do every night), and since its inactivity resembles death. Further, sleep has an indirect connection to darkness, since sleep is mostly performed during the dark hours.

Life is breathing As shown by François (in press), ‘breathing’ is closely connected to life. As a consequence, to stop breathing is connected to death. The data in the present study include expressions such as the French expirer ‘exhale’ and the Danish udånde ‘exhale’. The results support François’s study. Like putting down the receiver, as in (7), exhaling is something we do everyday, but in spite of this, the expressions involving breathing are separate from the everyday task mentioned so far. To exhale marks the stop of a breathing cycle (inhaling, exhaling and so forth), and the (simplified) idea is that when the person has exhaled for the last time, the person is dead. Thus, the connection between the source domain (breathing) and the target domain (dying) includes a strong logical part

214

Chapter Fourteen: Expressions for Dying in European Languages

(cause–impact), which is weaker for expressions such as (7), where the connection is to finish something.

Death is darkness The results further show that death is associated with darkness and that life is associated with light. The French data include s’éteindre ‘extinguish’ for ‘dying’. The Swedish expression given in (9) relates to a torch. (9)

vända ner facklan turn down torch+DEF ‘extinguish the torch’

Examples of life being associated with light can be collected from the Bible. In (10) a sample in Swedish from John (1:4) is given. (10)

Orden word+DEF.PLUR livet life+DEF

var be.PAST

var be.PAST

liv, life

och and

människornas people+DEF+PLUR+GEN

ljus. light ‘In him was life, and that life was the light of men.’

Death is cold Death is associated with cold, as in Swedish kallna ‘get cold’. In the Nordic countries, where Swedish, Norwegian and Danish are spoken, the connection between cold and death may be due to the existence of seasons. During the warm seasons, such as spring and summer, plants may grow, and this is also the time for animals to get off-spring. Thus, warmth is associated to fertility and life. The connection between cold and death also has to do with the fact that a dead body loses its heat (37 degrees Celsius) after death. This is true in climates where the air temperature is below 37 degrees Celsius, which is certainly the case in the areas where the pertinent languages are spoken. There is, however, a competing metaphor, which can be expressed as Death is heat. This metaphor is connected to another one, expressed as Life is shade. These metaphors have their origin in the Biblical context. They will be discussed in the next section.

Anna Vogel

215

Death is heat – Life is shade According to Christian belief, all people will not go to heaven, but some will go to hell. Hell is a very hot place where sinners suffer, see (11) which is taken from the Bible (Mark 9:44). (11)

i

helvetet,

i

den

eld

in

hell+DEF

in

ART.DEF

fire

som aldrig slocknar which never become extinct+PRES ‘into hell, to the unquenchable fire’ The climate of the Middle East, where the Bible was written/taken down, may help us to understand why heat is associated with death. The region contains abundant desert landscapes. These are not fertile; the temperature and lack of water prevent the growing process. In such a place, shade means relief and possibilities of cultivating. In (12), also taken from the Bible (Ps. 57:2), it is clear that shade is associated with safety, possibly with life. (12)

i in

dina your

vingars wing+PLUR+GEN

skugga shade

tar jag min tillflykt take I my refuge ‘in the shadow of your wings I take shelter’ Since shade is a type of darkness, it turns out that both life and death can be associated with darkness. At the same time, death can be associated both with heat and with cold. This can be explained by revealing that cold is associated to the body itself (a cold body = a dead body), while heat is associated to a place where the body is located (a hot place = either a place that may kill you or a place where you come after death). Thus, two separate specific conceptualisations, body and location, occur.

216

Chapter Fourteen: Expressions for Dying in European Languages

Conclusions The results indicate that European languages show similarities concerning how death is described. First, although ‘dying’ involves low or no agentivity, verbs showing high agentivity are used in order to express the transition of ‘dying’. Second, even if we do not know anything about what “happens” after death and if the “soul” arrives at some location, expressions specifying locations are common. Third, ‘dying’ is not trivial, but this is denied in expressions using everyday actions to describe ‘dying’. It is suggested that the reason for denial of these three major points is that such a denial helps us to approach death, in the same way as euphemisms are used for threatening concepts (Halliday 1978; Carter 1987). Further, results supporting François (in press) and Berendt, Maeda and Tanita (in press) have been presented. In accordance with François (in press), life is connected to breathing, and as a consequence, death is connected to exhaling (a “last” exhalation is implied in such expressions, although this is a simplification of the breathing process towards the end). In accordance with Berendt, Maeda and Tanita (in press), death is associated with sleep. Finally, the relation between temperature/light and life/death can be highlighted. At first, the results seem contradictory, since both cold and heat may be associated with death, and both light and shade may be associated with life. However, this can be explained by pointing out that while cold is associated to the body itself, heat is associated to location.

Acknowledgements The author wishes to express her gratitude to Annika Johansson and Carita Lundmark for valuable comments on earlier versions.

References Berendt, E., Maeda, A. and Tanita, K. In press. Various papers under the theme The Right to Kill; the Right to Die (preliminary title), Intercultural Communication Studies vol. XVI. Carter, R. 1987. Vocabulary: Applied Linguistics Perspective. London: Allen and Unwin. François, A. 2008. Semantic maps and the typology of colexification: Intertwining polysemous networks across languages. In M. Vanhove,

Anna Vogel

217

(Ed.) From polysemy to semantic change: Towards a typology of lexical semantic associations, 163–215. Amsterdam: John Benjamins. Halliday. M.A.K. 1978. Language as social semiotics. Baltimore, Md.: University Park Press. Lakoff, G. 1993. The contemporary theory of metaphor. In A. Ortony, (Ed.), Metaphor and thought, 202–251. Cambridge: Cambridge University Press.

THE SWEDISH ADJECTIVE VARM (‘WARM’): A COMPARISON BETWEEN LEXICAL MEANINGS AND ASSOCIATIONS ANNIKA BERGSTRÖM

Abstract. In this chapter, variants of lexical meanings of the Swedish adjective varm (‘warm’) are compared to associations to the same word. It is postulated that linguistic structures and structures of (immediate) thought might differ in some respects, and the comparison is made as a way of investigating if this assumption is correct. The 24 variants of lexical meanings were found in corpora (newspapers and novels). Out of these, 14 are physical, eight are metaphorical, and two variants are physical and metaphorical at the same time. Nine of the physical variants, two of the metaphorical, and one of the physical and metaphorical variants, did have corresponding associations. The associations were made by 17 informants. The variants that were found in written language as well as in the associations have to do with everyday phenomena of which warmth is an inherent, or at least important, quality (like the human body, the sun, fire, ovens, climate, nature phenomena, food and drink, clothes etc.), with feelings of affection which are metaphorically linked to the human body, and with common synaesthetic impressions (like “warm” colours). One important reason why concepts are mentally, as well as linguistically, related seems to be that they have a certain quality in common. The variants that were only found in written language have to do with physical warmth as a result of different processes, with physical measurement, and with conventional metaphorical expressions. There are also associations that do not correspond to the lexical meanings: paraphrasing associations, judgements, and (individual) chains of associations. According to the results, linguistic structures and structures of (immediate) thought do differ in some ways, although they are both, when it comes to a certain domain, like temperature, governed by the same central concepts.

Annika Bergström

219

Introduction In Cognitive Semantics, language and thought about the world we live in are believed to shape each other (e.g. Lakoff and Johnson 1980). This general idea is certainly plausible, but it is also possible that there may be at least some differences between linguistic structures and structures of thought. In order to examine these differences, it is necessary to execute different kinds of studies, using different methods. The need to compare introspective findings (the original method of Lakoff and Johnson) to the results of other kinds of studies, by other methods, is also recognised by, for example, Talmy (2000). In this chapter, I present a comparison between (variants of) lexical meanings of the Swedish adjective varm (‘warm’) and associations to the same word. Based on the results of this comparison, I argue that linguistic structures and structures of thought do differ in some respects, although, at the conceptual level, they are both governed by what is fundamental to human beings in their everyday experience.

The domain of temperature The domain of temperature is essential to human beings. It has to do with life and death (the warmth of the living body, the coldness of the dead body), as well as with fundamental conditions of human life (climate). In Nordic countries like Sweden, the changing of seasons is an important factor in shaping the way people live, physically as well as socially. Accordingly, Lakoff and Johnson (1980: 57) include warmth and cold in “some of the central concepts in terms of which our bodies function”, together with up – down, light – dark, etc., and Lakoff (1987: 271) considers hot and cold to be basic-level properties, along with tall, short, hard, soft, and so on. Temperature is also mentioned by Langacker as a “basic domain”, a concept which he defines as “cognitively irreducible representational spaces or fields of conceptual potential” (2002: 4). Other such “basic domains” are, for example, time and colour. The domain of temperature is also rich in metaphors, which makes it an interesting area of study within the theory of Cognitive Semantics. Examples of metaphors including temperature are mentioned by Lakoff (1987), Lakoff and Turner (1989) and Kövecses (1995), among others. Grady (1997) lists what he calls “primary metaphors” (metaphors which are grounded in our bodily experience, and which constitute fundamental connections in our perception of the world surrounding us). Among these are three metaphors linking together bodily temperature sensations with

220

Chapter Fifteen: The Swedish Adjective varm (‘warm’)

emotions, namely INTENSITY OF ACTIVITY IS HEAT, AFFECTION IS WARMTH, and INTENSITY OF EMOTION IS HEAT. Melnick (2000) reasons in a similar way, but from a more psychoanalytic perspective. According to Melnick, conceptual metaphors consist of constellations of interrelated terms, like the example “COLD/HARD: solid, uncomfortable, unsympathizing, reliable, well-defined, comprehensible in detail, sharp (sometimes blunt), precise, difficult”. (Melnick 2000: 23) Temperature (swe. varm (‘warm’) – kall (‘cold’), eng. hot – cold) is often mentioned as an example of antonyms (e.g. Croft and Cruse 2004; Jones et al. 2008). It should be pointed out that varm (‘warm’) in Swedish normally correspond to hot in English (Swe. ‘het’). Koptjevskaja-Tamm and Rakhilina (2003) compare temperature adjectives in Swedish and Russian, showing that the relationship between adjectives that seem to correspond to each other in dictionaries, is often more complicated in reality. In my forthcoming thesis (Bergström 2010) three kinds of conceptual relations within the domain of temperature in Swedish are analysed: the relations between the most important adjectives: varm (‘warm’), ljum/ljummen (‘lukewarm’ or ‘tepid’), het (‘hot’), kall (‘cold’), kylig (‘chilly’), sval (‘cool’), mild (‘mild’), the relations between these adjectives and other words and expressions relating to temperature, and the relations between temperature in language and associations to temperature. The linguistic expressions relating to temperature are studied in corpora, in dictionaries, and by close reading of texts, such as novels. Three different kinds of association tests have been conducted, investigating, among other things, the influence of linguistic form, first language, and culture.

The model for analysis In Bergström (2010) variants of lexical meanings of the temperature adjectives are collected and analysed according to a particular model of analysis which includes five dimensions: (1) concrete (or physical) – abstract (or metaphorical), (2) domain, (3) way of experiencing the temperature, (4) the character of the experience (namely grade, expectations, and value), and (5) structure (the relationship between the temperature and the thing etc. that has got the temperature, as well as the process of temperature change, if any). The variants of lexical meanings are defined according to the five dimensions of analysis. A variant of a lexical meaning is to be regarded as a practical, analytic tool, useful in the

Annika Bergström

221

comparison between the adjectives. The exact lexical status of each variant is not analysed, although it is noted whether the variant is central or peripheral to the understanding of the adjective. The point of departure is the experience of temperature by human beings. This experience often differs from technical, or scientific, temperature. For example, a winter’s day might be perceived and described as warm, compared to a “normal” winter’s day, although the temperature, measured in degrees, is still quite low. The observation that the experience of temperature may differ from technical temperature is confirmed by Knez and Thorsson (2006). In a questionnaire, Japanese people estimated the weather as warmer than Swedish people did, although the thermal conditions were the same. An important assumption for my work is that language is both shaped by as well as reflects, and influences, the common experiences and values of the people within the language community.

The Swedish adjective varm (‘warm’): variants of lexical meanings and associations The variants of lexical meanings of varm (‘warm’) have been collected from two corpora: a Swedish newspaper (Göteborgs-Posten 2004, abb. GP04, comprising about 20 million words), and Swedish novels (called Bonniersromaner II, abb. romii, comprising about 3 million words). The adjective varm occurs, in utrum (varm), 518 times in GP04 and 289 times in romii. In neutrum (varmt) the adjective occurs 632 times in GP04, and 276 times in romii. Other grammatical forms have been studied as well, but no extra variants of meanings have been found this way. The associations to varm were made by 17 informants (university students in Swedish language and native speakers of Swedish) in the autumn of 2004. The informants were given the words varm (‘warm’) and kall (‘cold’), and they were told to simply write down whatever they came to think of when they saw the two words. There were no restrictions as for the number of associations. In Table 1 and 2 the variants of the lexical meanings of varm are shown, together with the corresponding associations (in the cases where they had any). Examples of lexical expressions are also given, together with the meaning variants. According to this analysis varm has 24 meaning variants. The total amount of variants for all of the Swedish temperature adjectives (varm, het, kall, and so on) is 33. (This is the reason why letters are missing in the tables; meaning variants of other adjectives fill in the gaps.)

222

Chapter Fifteen: The Swedish Adjective varm (‘warm’)

It should be said that the translations into English in Table 1 and 2 (of the lexical expressions and the associations) are literal. Varm in Swedish is translated into warm in English, although idiomatically varm should probably in several cases be translated into hot in English. The same is true of fixed expressions in Swedish, like gå varm. Since the aim of the chapter is to analyse the Swedish adjective varm, literal translations have been preferred to more idiomatic ones. Table 1. Concrete (physical) lexical variants with corresponding concrete (physical) associations to varm (‘warm’). The number of informants who have made the association is given within parenthesis, if the number exceeds one. Concrete lexical meaning variants of varm (‘warm’) a) about being physically warm in the whole body or in bodyparts, ex. varm kropp (‘warm body’), hand (‘hand’), hud (‘skin’) b) about products of the human body, ex. varm urin (‘warm urine’), varmt blod (‘warm blood’) c) about becoming warm by physical activity, ex. springa sig varm (‘to run until one gets warm’) d) about the sun (and fire), ex. solen är varm (‘the sun is warm’)

Corresponding associations svettig (‘sweaty’) (2), handflata (‘palm’), katt (‘cat’), kropp (‘body’), mage (‘stomach’), min hund (‘my dog’), min panna (‘my forehead’), svettas (‘sweating’) blod (‘blood’)

-

sol (‘sun’) (6), eld (‘fire’) (4), glöd (‘embers’), solen (‘the sun’)

e) about day and night, seasons etc., temperatures outdoors and indoors, ex. varm dag (‘warm day’), lägenhet (‘apartment’)

sommar (‘summer’) (7), bastu (‘sauna’), solig (‘sunny’), sommaren (‘the summer’), inomhus (‘indoors’), strand (‘beach’)

g) about nature phenomena, ex. varm vind (‘warm wind’), ånga (‘steam’)

ånga (‘steam’)

h) about water (not drink), ex. varmt bad (‘warm bath’), vatten (‘water’)

bad (‘bath’)

Concrete lexical meaning variants of varm (‘warm’) i) about surfaces and materials, ex. varm asfalt (‘warm asphalt’), vägg (‘wall’)

Corresponding associations

j) about objects that produce warmth, ex. varm ugn (‘warm oven’)

element (‘radiator’), grill (‘grill’), kakelugn (‘tiled stove’), spis (‘stove’), ugn (‘oven’)

k) about objects that have been heated, varm tallrik (‘warm plate’), stekpanna (‘frying pan’) l) about food and drink, ex. varm mat (‘warm food’), varmt kaffe (‘warm coffee’)

-

m) about objects that have been heated as a result of (mechanical) use, ex. gå varm (ex. kullager, motor) (‘go warm’ (ex. ‘ball bearing’, ‘motor’) n) about clothes and similar objects, ex. varm filt (‘warm blanket’), tröja (‘sweater’) o) about warmth that has been physically measured (degrees), ex. 40 grader varmt (‘40 degrees of warmth’)

-

te (‘tea’) (3), kaffe (‘coffee’), macka (‘sandwich’), mat (‘food’), glass (‘icecream’) -

filt (‘blanket’) (2), mössa (‘cap’), socka (‘sock’), tröja (‘sweater’), täcke (‘quilt’), ullvantar (‘wool mittens’) -

Annika Bergström

223

Table 2. Abstract (metaphorical) variants, together with corresponding abstract (metaphorical) associations to varm (‘warm’). The number of informants who have made the association is given within parenthesis. Abstract lexical meaning variants of varm (‘warm’)

Corresponding associations

Abstract lexical meaning variants of varm (‘warm’)

Corresponding associations

p) about physical signs and physical experiences of emotional reactions (concrete and abstract), ex. varm rodnad (‘warm blush’), bli varm i hela kroppen (‘get warm in the whole body’)

-

y) about commitment of people into someone, or something, ex. varm anhängare (‘warm supporter’), varmt rekommendera (‘warmly recommend’)

-

q) about bodily contact in connected to emotion (concrete and abstract), ex. varm kram (‘warm hug’), varmt handslag (‘warm handshake’)

kram (‘hug’)

ä) about positive states, relations etc. within a certain environment, a community, in a country or between countries, etc., ex. varm miljö (‘warm environment’), varmt samhälle (‘warm community’)

-

s) about qualities, emotions, relationships, and states within or between people, ex. varm person (‘warm person’), stämning (‘atmossphere’), kärlek (‘love’)

kärlek (‘love’) (3), kärleksfull (‘loving’) (2), gemenskap (‘feeling of solidarity’), glädje (‘joy’), känsla (‘feeling’), samhörighet (‘solidarity’), vänskap (‘friendship’)

Į) about frequent (nonbodily, non-mechanical) use, ex. gå varm (ex. låt, telefon, dator) (‘go warm’) (ex. ‘song’, ‘telephone’, ‘computer’)

-

w) about looks and other physical expressions that might be interpreted as signs of positive feelings etc., ex. varm blick (‘warm look’), röst (‘voice’), leende (‘smile’)

-

ȕ) about positive feelings in a verbal account or other (artistic) expressions, ex. varm berättelse (‘warm story’), film (‘film’)

-

x) about abstract signs of appreciation, ex. varmt beröm (‘warm praise’), tack (‘thanks’)

-

Ȗ) about synestethic impressions, ex. varm färg (’warm colour’), ton (‘musical tone’), varmt ljus (‘warm light’)

ljus (‘light’) (2), färger (‘colours’), färgglad (‘colourful’), rött (‘red’)

As shown in Table 1 and 2, there are two kinds of associations: syntagmatic associations, e.g. handflata (‘palm’), and paradigmatic associations, e.g. svettig (‘sweaty’). The syntagmatic associations are far more common than the paradigmatic ones. Informants also made other kinds of associations: paraphrasing associations, like het (‘hot’), and

224

Chapter Fifteen: The Swedish Adjective varm (‘warm’)

värme (‘warmth’), so called judgements, like behagligt (‘comfortable’), and brännande (‘burning’), as well as associations to previous associations, which are no longer connected to warmth, like badkar (‘bath tub’), and hungrig (‘hungry’).

Discussion By taking a quick look at the results in Table 1 and 2, one can easily notice that most of the associations are concrete, or physical. There are 14 concrete meaning variants of varm (‘warm’). Nine of these have corresponding associations. Out of the eight purely abstract, or metaphorical, variants of varm, only two have corresponding associations. There are also two variants (p and q), which are at the same time concrete and abstract. One of these (q) has one corresponding association. The meaning variants of varm which exist in written language as well as in the associations have to do with the human body (a, b), (common) sources of warmth, like the sun, fire (d), and artefacts like radiators and ovens (j), climate, seasons etc. (e), nature phenomena (g), water (h), food and drink (l), clothes, blankets etc. (h), physical expressions of feelings, feelings, human (emotional) qualities (q, s), and synaesthetic impressions, like colours (Ȗ). In other words, these variants have to do with everyday objects etc. of which warmth is an inherent, or at least important quality, with feelings of affection that are metaphorically linked to the human body (e.g. AFFECTION IS WARMTH), and with synaestethic impressions, like colours of “warm objects” (red, for example, is the colour of blood). The variants that were only found in written language, on the other hand, have to do with people or objects getting warm by physical movement or frequent use (c, m), surfaces, materials, and objects which have been heated (i, k), measuring physical warmth in degrees (o), physical signs etc. of emotions (p), metaphorical interpretations of certain physical expressions (w), abstract, non-physical, signs of appreciation (x), commitment (y), metaphorical descriptions of states and relations within a country, city, or working community (ä), frequent (non-bodily, nonmechanical) use (Į), and positive feelings in artistic expressions (ȕ). In other words, these meaning variants have to do with things which are not warm in themselves, physical measurement, and conventional, metaphorical linguistic expressions which are agreed upon in the (Swedish) language community. It is interesting that two of Grady’s postulated “primary metaphors” that have to do with temperature (INTENSITY OF ACTIVITY IS HEAT and INTENSITY OF EMOTION IS HEAT) were only found in written language in this study. The

Annika Bergström

225

AFFECTION IS WARMTH-metaphor, on the other hand, seems essential to explain the connection between the physical and the metaphorical, in written language as well as in the associations. The AFFECTION IS WARMTH-metaphor builds on the fact that warmth is an inherent quality of the human body, while in the other two metaphors warmth (or heat) is created by some external physical or mental activity. The fact that concepts have a certain quality in common seems to be a very important reason why they are mentally, as well as linguistically, related. This quality should be inherent, or at least important, to the concept as it is; it should not be the result of external processes of some kind. According to the dictionaries (Hellquist, Svenska Akademiens ordbok, Söderwall), early uses of varm (Old Swedish varmber, varmr) had to do with the warmth of blood (slaughter of animals), warm (female) milk, embers, burning, to seethe, to boil, and with the warmth of summer – in other words, with concepts of which warmth is an inherent quality. This is interesting, although one always has to be careful not to draw sweeping conclusions from etymology. The meaning variants of lexical meanings to which no associations were made have to do with warmth as a result of processes of different kinds, and with conventional, metaphorical expressions. These metaphorical expressions may originally have been motivated by the human body, climate, culture, or other factors, but the connections to temperature are less obvious than in the case of especially the AFFECTION IS WARMTH-metaphor. When it comes to these more conventional metaphors, there might be other ways of describing the phenomena than by temperature; alternative metaphors might be found in other languages, in other cultures. As for the associations, they seem to be governed by (common) experiences of the world (which also govern language), by personal experience (chains of associations, sometimes judgements), and by language itself (paraphrasing associations). Associations made out of personal experience or purely linguistic associations (the paraphrasing ones) do not correspond to variants of lexical meanings. Of course, there is much to be said about word association tests as a method. The associations of the informants are influenced by the way the words are presented to them, by the instructions given, by the external circumstances of the test situation, as well as by their personal experiences. In this particular study, the number of informants was also quite low (only 17). Possibly corresponding associations to more variants of lexical meanings of varm would have turned up if the number of informants had been higher. Nevertheless, word association tests may

226

Chapter Fifteen: The Swedish Adjective varm (‘warm’)

serve as a valuable complementary method to lexical studies. They seem to point out what is essential to people, in this case what is central about warmth, in both language and mind. Naturally, one might object that all linguistic structures are, in some ways, also structures of thought, since they are created by the human mind, and since they are accessible to thought in language use. Nevertheless, the meaning variants that were only found in written language do not seem to govern thought in the same way as the variants that do have corresponding associations; they represent more conventional linguistic expressions. Perhaps the correct way of interpreting the associations is to say that they represent immediate thought. When comparing the (variants of) lexical meanings of varm to the associations to the same word, there does indeed, beside the similarities, seem to be differences between linguistic structures and structures of (immediate) thought.

Conclusions One important reason why concepts are mentally, as well as linguistically, related seems to be that they have a certain quality in common. In this respect, the (variants of) lexical meanings of varm (‘warm’) and the associations to the same word, are very much alike. The human body, climate, the sun, fire, everyday objects like ovens, food and drink, clothes, affections, and synaesthetic impressions of colours, are all central concepts when it comes to warmth, in both language and mind. But the linguistic structures and the structures of (immediate) thought, as represented by the associations, do also differ in some ways. Physical warmth as a result of different kinds of processes, as well as most of the (conventional) metaphorical expressions, may only be found in written language. On the other hand, there are associations (chains of associations, judgements, and paraphrasing associations), which do not correspond to lexical meanings.

References Bergström, A. 2010. (forthcoming) Temperatur i språk och tanke. University of Gothenburg, Sweden. Bonniersromaner II. 1980-81. (corpus). Online http://spraakbanken.gu.se Croft, W. and Cruse, A.D. 2004. Cognitive Linguistics. Cambridge University Press.

Annika Bergström

227

Grady, J. 1997. Foundations of meaning: Primary metaphors and primary scenes. University of California, Berkeley. Göteborgs-Posten 2004. (corpus). Online http://spraakbanken.gu.se Hellquist, E. 1922 (1957). Svensk etymologisk ordbok. Tredje upplagan. Lund: Gleerups förlag. Jones, S., Paradis, C., Lynne Murphy, M., and Willners, C. 2008. Googling for ‘opposites’: a web-based study of antonym canonicity. In: Corpora Vol. 2 (2): 129-154. Knez, I. and Thorsson S. 2006. Influences of culture and environmental attitude on thermal, emotional and perceptual evaluations of a public square. Int J Biometeorol. 50(5): 258-268. Koptjevskaja-Tamm, M. and Rakhilina, E.V. 2003. “Some like it hot”: on semantics of temperature adjectives in Russian and Swedish, Online http://www.ling.su.se/staff/tamm Kövecses, Z. 1995. Anger: Its language, conceptualization, and physiology in the light of cross-cultural evidence. In J. Taylor and R. MacLaury (Eds.), Language and the cognitive construal of the world, 181-196. Berlin & New York: Mouton de Gruyter. Lakoff, G. 1987. Women, fire, and dangerous things: What categories reveal about the mind. Chicago: University of Chicago Press. Lakoff, G. and Johnson, M. 1980. Metaphors we live by. Chicago and London: University of Chicago Press. Lakoff, G. and Turner, M. 1989. More than cool reason: A field guide to poetic metaphor. Chicago: University of Chicago Press. Langacker, R.W. 2002. (2nd ed.) Concept, image, and symbol: the cognitive basis of grammar. Berlin & New York: Mouton de Gruyter. Melnick, B.A. 2000: Cold hard world/warm soft mommy: The unconscious logic of metaphor. The Annual of Psychoanalysis 28: 225244. Svenska Akademiens ordbok (SAOB) Online http://spraakdata.gu.se Söderwall, K.F. Ordbok öfver svenska medeltidsspråket Online http://spraakdata.gu.se Talmy, L. 2000: Toward a cognitive semantics. Volume II: Typology and process in concept structuring. Cambridge, Mass.: MIT Press.

DOMAIN KNOWLEDGE IN CHILDREN’S UNDERSTANDING OF METAPHORS: DO SOURCE DOMAINS HAVE ANY IMPACT? AIVARS GLAZNIEKS

Abstract. According to theories from developmental psychology, there is one prominent predictor for understanding metaphorical expressions: Conceptual knowledge of both the source domain and the target domain. In conceptual metaphor theory, it is claimed that in metaphor a given source domain serves to explain a given target domain (Lakoff and Johnson 1980). According to this assumption, knowledge of the source domain is crucial for understanding a metaphorical expression. This chapter presents a study of German-speaking preschoolers around the age of five and pupils around the ages of eight and ten. In two comprehension tasks, participants were tested with several metaphorical expressions denoting the emotions fear and anger. The study focused on whether participants understand metaphorical expressions in a domainspecific way. The results show that the understanding of metaphorical expressions develops in a domain-specific way only with respect to target domains (FEAR vs. ANGER). By contrast, domain-specific development of the understanding of metaphorical expressions cannot be confirmed with respect to source domains (e.g. in the case of FEAR: COLDNESS vs. DEFECATION vs. WEAKNESS). Against the predictions of conceptual metaphor theory, it seems that source domains do not have any impact on the understanding of metaphorical expressions in language acquisition, while target domains do.

Introduction According to Winner (1988), children begin to understand their first metaphorical expressions around the age of four.1 Although still limited, 1

Children produce figurative utterances even around the age of two during symbolic play. Since in so-called symbolic play metaphors the direction of the mapping is from the target to the source, I do not consider this kind of metaphorical expressions to be metaphorical at all (cf. Burger 2006; Gibbs 1994).

Aivars Glaznieks

229

children’s ability to understand metaphorical expressions2 consistently increases between the ages four and fourteen. Comprehension of metaphorical expressions is influenced by various factors such as the development of so-called theory of mind (cf. Sodian and Thoermer 2005), as well as knowledge of the conceptual domains connected in a given metaphorical expression, i.e. domain-knowledge (Keil 1986). Young children are thus able to understand metaphorical expressions like the car was thirsty because the source domain of human nutrition is very familiar to them. In contrast, a metaphorical expression such as he was a smooth person is not very likely to be understood because neither the domain of surface texture nor the domain of personality traits are common concepts in young children. In addition, Keil found an “all-or-none” effect: children who understand one metaphorical expression that connects certain domains will also understand other metaphorical expressions with the same domain features. In conceptual metaphor theory, it is assumed that knowledge of the source domain influences the way metaphorical expressions are understood, or as Lakoff and Johnson put it: “the essence of metaphor is understanding and experiencing one kind of thing in terms of another” (Lakoff and Johnson 1980: 5). Some studies in psycholinguistics have shown that listeners activate their tacit knowledge of metaphors when interpreting figurative idioms such as spill the beans (e.g. Gibbs and O’Brien 1990). Gibbs and O’Brien’s study shows evidence of both the existence of conceptual metaphors and of the participant’s awareness of source domains. However, the question of the impact of source domain knowledge on children’s developing metaphorical competence has not yet been investigated in Cognitive Linguistics. This chapter deals with the impact of source domains on the understanding of metaphorical expressions in German-speaking preschool and elementary-school children, who were tested with several metaphorical expressions denoting the emotions of fear and anger in German.

Metaphors in the Domains of Fear and Anger According to the conceptual metaphor theory, emotions are conceptualized metaphorically (cf. Lakoff and Johnson 1980; Lakoff 2003). Various 2

To avoid any terminological confusion I will adopt Lakoff’s definition of metaphor as “a cross-domain mapping in the conceptual system” while metaphorical expression “refers to a linguistic expression […] that is the surface realization of such a cross-domain mapping” (Lakoff [1993] 2006: 186).

Chapter Sixteen: Domain Knowledge in Children’s Understanding of Metaphors

230

metaphorical expressions denoting fear and anger in German can be attributed to metaphors. In the domain of fear, for instance, Dobrovol‘skij (1997) found the following metaphorical mappings: x x x

(eine Gänsehaut bekommen “to get goose bumps” ‘to get the jitters’) FEAR IS DEFECATION (sich in die Hose machen “to make in one’s pants” ‘to be scared shitless’) FEAR IS WEAKNESS (weiche Knie bekommen “to get soft knees” ‘to have one’s legs turn to jelly’).3 FEAR IS COLDNESS

The domain of anger in German consists of metaphors such as: x ANGER IS A LIQUID IN A CONTAINER (jemandem kocht das Blut in den Adern “someone’s blood is boiling in her/his veins” ‘to boil with rage’) x ANGER IS INSANITY (wild werden “to become wild” ‘to be enraged’) x ANGER IS SICKNESS (Schaum vor dem Mund bekommen “to get foam at the mouth” ‘to fret and fume’) (cf. Lakoff and Kövecses (1983) for metaphors in American English).4 As pointed out in the introduction, according to the domain knowledge account, knowledge of the source and target domains should have a crucial impact on the understanding of metaphorical expressions. From this assumption we can predict that children will be able to understand metaphorical expressions only when they have knowledge about the domains that are mapped onto one another. According to conceptual metaphor theory, an especially strong influence of the source domain of a metaphorical mapping can be expected. Knowledge of target domains, in contrast, should be less important. On this basis, the following two hypotheses can be formulated concerning children’s understanding of metaphorical expressions in the domains of fear and anger. x 3

H1: Development of the understanding of metaphorical expressions in the target domains of fear and anger proceeds separately.

Dobrovol’skij (1997) states four main models of figurativeness in the domain of fear in German. For methodological reasons, I have left aside the model FEAR IS A PREDATOR. 4 Unfortunately, there is no systematic investigation of the metaphors in the domain of anger in German; for studies of Spanish and French anger metaphors see e.g. Herrera Burstein (1997) and Weber (1993), respectively.

Aivars Glaznieks

x

231

H2: Development of the understanding of metaphorical expressions within the target domains of fear and anger proceeds along with development of knowledge of the source domains.

According to H1, different results in understanding metaphorical expressions denoting fear and anger are to be expected. H2 is a reformulation of the “all-or-none” effect mentioned earlier: If participants understand one metaphorical expression, they will also understand other metaphorical expressions that share both target and source domains and are thus instantiations of the same metaphor (Keil 1986).

Method Participants were drawn from three groups: 24 preschool children (5;1 to 6;0, average 5;5), 30 second-graders (7;7 to 8;11, average 8;1) and 30 fourth-graders (9;10 to 11;9, average 10;7). Each group included an equal numbers of boys and girls. All participants came from a fairly homogeneous population from a small town west of Munich and were monolingual native speakers of German. All participants were tested with a set of metaphorical expressions that in German conventionally designate the emotions fear and anger. For the sake of comparability, participants were matched on a non-verbal IQ test (Raven-Matrices).

Design In Test 1, participants had to sort 50 cards with 21 different metaphorical and non-metaphorical terms for each of the emotions anger and fear. Additionally, 8 terms from the semantic field of joy were used as distracters. Participants were first asked to stack the cards according to similarity. Cards with unknown terms were to be put aside. Afterwards, they were asked to name the piles they had created. In Test 2, 14 short stories were presented within a multiple-choice task. At the end of each story, the protagonist found herself in a situation evocative of a certain emotion (again: fear, anger, and joy). All stories concluded with a question regarding the character’s feeling. Four possible answers were presented for each story: (a) a metaphorical answer denoting the intended emotion, (b) a metaphorical answer denoting a different emotion, (c) a metaphorical answer denoting something other than an emotion and (d) the option of rejecting all answers. The order of the answers (a) – (c) was randomised.

232

Chapter Sixteen: Domain Knowledge in Children’s Understanding of Metaphors

Except for one item, all metaphorical items tested in Test 2 were also used in Test 1.5 A total number of six items was used for each emotional domain of fear and anger. Two items of each domain are instances of the same metaphor. The following three pairs of metaphorical expressions were used from the domain of fear: sich in die Hose machen (“to make in one’s pants” ‘to be scared shitless’) and die Hosen voll haben (“to have one’s pants full” ‘to be scared shitless’) are instances for the metaphor FEAR IS DEFECATION, eine Gänsehaut bekommen (“to get goose bumps” ‘to get the jitters’) and es läuft einem eiskalt den Buckel herunter (“someone’s back runs cold” ‘someone’s blood runs cold’) for FEAR IS COLDNESS, and weiche Knie bekommen (“to get soft knees” ‘to have one’s legs turn to jelly’) and kaum noch atmen können (“hardly be able to breathe” ‘one’s breath caught’) for FEAR IS WEAKNESS. In the domain of anger the three pairs of metaphorical items were as follows: an die Decke gehen (“to go up to the ceiling” ‘to hit the roof’) and jemandem kocht das Blut in den Adern (“someone’s blood is boiling in her/his veins” ‘to boil with rage’) as instances for the metaphor ANGER IS A LIQUID IN A CONTAINER, Gift und Galle speien (“to spit poison an bile” ‘to breath fire and brimstone’) and Schaum vor dem Mund bekommen (“to get foam at the mouth” ‘to fret and fume’) for ANGER IS SICKNESS, wild werden (“to become wild” ‘to be enraged’) and auf 180 sein (“to be on 180” ‘to be furious’) for ANGER IS INSANITY. The main difference between Tests 1 and 2 was the way the metaphorical expressions were presented. Whereas in Test 1 participants had to sort all items in a context-free situation, metaphorical expressions were presented within a linguistic context in Test 2. The data were analysed quantitatively in a binary-coded evaluation process. Every item was evaluated either as correctly or incorrectly understood by the participants. In Test 1, an item was interpreted as correct only if it was sorted onto the appropriate pile, representing, respectively, the domains of fear and anger. In Test 2, only selections of (a), metaphorical expressions denoting the intended emotion, were evaluated as correct. Answers (b)-(d) were rated as incorrect.

5

The item kaum noch atmen können (“hardly be able to breath” ‘one’s breath caught’) was replaced by Blut und Wasser schwitzen (“to sweat blood and water” ‘to be in a cold sweat’). Since both metaphorical expressions belong to the same metaphor (FEAR IS WEAKNESS), the difference should not have any impact on the results.

Aivars Glaznieks

233

Results As Figure 1 shows, the understanding of metaphorical expressions improves with increasing age. Few differences between the participants’ performance in Test 1 and 2 could be seen, suggesting that linguistic context did not facilitate the understanding of the metaphorical expressions in this particular case. Figure 1 also shows a difference in the understanding of metaphorical expressions in the domains of fear and anger in the 8-year-old participants: As can be seen, fear was understood much better than anger. The 5-year-olds and 10-year-olds did not show this difference. The younger participants hardly understood any metaphorical expressions at all, whereas 70-80% of the older participants did so fairly well. This result indicates that the understanding of metaphorical expressions is domain-specific with respect to target domains, and that it develops separately in each target domain. H1 can thus be regarded as supported. In a second step, the participants’ results of the two items relating to the same metaphor were compared. Since they belong to the same metaphor, they feature the same source domain. Expecting an “all-ornone” effect, the results of the two items should be almost equal, since the participants should either understand both items or none. Figures 2 and 3 show that almost all results are divergent within the six pairs of metaphorical expressions. This finding is consistent in all age groups and in both target domains. The data indicates that the development of the understanding of metaphorical expressions within the target domains of fear and anger does not proceed along the assumed source domains. Moreover, the data suggest that source domains did not have an impact on the understanding of metaphorical expressions at all. If they had such an impact, differences in understanding between the tested metaphorical expressions would have followed the divisions between the source domains. Due to the fact that the arrangement of the results is rather arbitrary, H2 can be regarded as falsified.

Chapter Sixteen: Domain Knowledge in Children’s Understanding of Metaphors

percentage of participants

234

1 0,9 0,8 0,7 0,6 0,5

Anger (Test 1) Anger (Test 2) Fear (Test 1)

0,4 0,3 0,2

Fear (Test 2)

0,1 0 5-year-olds 8-year-olds 10-year-olds

Figure 1. Correct understanding of metaphorical expressions, divided by target domains (Test 1 and Test 2)

percentage of participants

1,00

HEAT An die Decke gehen HEAT Jemandem kocht das Blut in den Adern

0,80 0,60

SICKNESS Gift und Galle speien SICKNESS Schaum vor dem Mund bekommen INSANITY Wild werden

0,40 0,20 0,00 5-yearolds

8-year- 10-yearolds olds

INSANITY Auf 180 sein

Figure 2a. Correct understanding of metaphorical expressions within the target domain of anger based on source domains (Test 1)

percentage of participants

Aivars Glaznieks

235

HEAT An die Decke gehen

1,00 0,80

HEAT Jemandem kocht das Blut in den Adern

0,60 0,40

SICKNESS Gift und Galle speien

0,20

SICKNESS Schaum vor dem Mund bekommen

0,00

INSANITY Wild werden

5-yearolds

8-year- 10-yearolds olds

INSANITY Auf 180 sein

Figure 2b. Correct understanding of metaphorical expressions within the target domain of anger based on source domains (Test 2)

percentage of participants

1,00

DEFECATION Sich in die Hose machen

0,80

DEFECATION Die Hosen voll haben

0,60

COLD Eine Gänsehaut bekommen

0,40 0,20 0,00 5-yearolds

8-year- 10-yearolds olds

COLD Es läuft einem eiskalt den Buckel herunter WEAKNESS Weiche Knie bekommen

Figure 3a. Correct understanding of metaphorical expressions within the target domain of fear based on source domains (Test 1)

Chapter Sixteen: Domain Knowledge in Children’s Understanding of Metaphors

236

percentage of participants

1,00 0,80 0,60 0,40 0,20 0,00 5-yearolds

8-yearolds

10-yearolds

DEFECATION Sich in die Hose machen DEFECATION Die Hosen voll haben COLD Eine Gänsehaut bekommen COLD Es läuft einem eiskalt den Buckel herunter WEAKNESS Weiche Knie bekommen WEAKNESS Kaum noch atmen können

Figure 3b. Correct understanding of metaphorical expressions within the target domain of fear based on source domains (Test 2)

Discussion and conclusions The results of the study showed that children’s understanding of metaphorical expressions is not necessarily constrained by their knowledge of source domains. This means that children do not (always) understand metaphorical expressions by “projecting” from the source domains, on the basis of pre-existing metaphorical conceptual structures as predicted by conceptual metaphor theory (Lakoff 2003). Explaining these findings must take into account the fact that the tested metaphorical expressions are common and conventionalized metaphorical expressions in German. In this respect, they are comparable to idioms. Although an idiom’s meaning might be transparent to adult speakers, children usually learn complex idioms as a chunk. They do not consider the meanings of its constituents to process the meaning of the entire expression (Häcki Buhofer 1997). Unlike adults, they do not activate existing source domains when confronted with metaphorical expressions (Gibbs and O’Brien 1990) and they do not derive meaning from a tacit metaphorical knowledge about emotions. Rather, they seem to understand metaphorical expressions in a “top-down” manner, i.e., as one

Aivars Glaznieks

237

part of the whole linguistic and non-linguistic context in which metaphorical expressions appear in verbal interaction. To test this explanation, a further factor must be considered: The frequency of the tested items in the children’s linguistic environment, i.e. in mother-child communication and in the language of teachers, educators, siblings and peers. Highly frequent metaphorical expressions might be familiar to children and thus understandable as “chunks”. To understand infrequent metaphorical expressions, in contrast, children would need contextual information to discover their meanings, while the two tests in the study probably did not provide enough contextual information to permit the children to figure out the meaning of an unknown metaphorical expression. To obtain a complete picture of children’s metaphorical competence, a comparison of the comprehension data with production data in the domains of fear and anger is to be conducted. The way children speak about emotions is bound to be no less revealing than their comprehension.

References Burger, H. 2006. Phraseologie im Spracherwerb – Evidenz aus der Spontansprache. In A. Häcki Buhofer and H. Burger (Eds.), Phraseology in Motion 1: Methode und Kritik. Akten der internationalen Tagung zur Phraseologie (Basel 2004), 339-356. Bultmannsweiler: Schneider Verlag Hohengehren. Dobrovols’kij, D. 1997. Idiome im mentalen Lexikon. Ziele und Methoden der kognitivbasierten Phraseologieforschung. Trier: Wissenschaftlicher Verlag. Gentner, D. 1988. Metaphor as structure mapping: the relational shift. Child Development 59: 47-59. Gibbs, R. W. 1994. The poetics of mind. Figurative thought, language, and understanding. Cambridge: Cambridge University Press. Gibbs, R.W., and O’Brien, J.E. 1990. Idioms and mental imagery: The metaphorical motivation for idiomatic meaning. Cognition 36: 35-68. Glaznieks, A. (forthcoming): Why focus on target domains. In S. Handl and H.-J. Schmid (Eds.) Windows to the mind. Metaphor, metonymy, and conceptual blending. Berlin/New York: Walter de Gruyter. Häcki Buhofer, A. 1997. Phraseologismen im Spracherwerb. In R. Wimmer and Berens, F.-J. (Eds.), Wortbildung und Phraseologie, 209232. Tübingen: Gunter Narr Verlag. Herrera Burstein, M. 1997. Sprachliches Erfassen von Emotionen im Spanischen. Dissertation University of Freiburg.

238

Chapter Sixteen: Domain Knowledge in Children’s Understanding of Metaphors

Keil, F.C. 1986. Conceptual domains and the acquisition of metaphor. Cognitive Development 1: 73-96. Müller, C. 2007. A dynamic view of metaphor, gesture and thought. In S. D. Duncan, J. Cassell and E. Levy (Eds.), Gesture and the dynamic dimension of language. Essays in honor of David McNeill, 109-116. Amsterdam/Philadelphia: John Benjamins. Lakoff, G. and Johnson, M. 1980. Metaphors we live by. Chicago: University of Chicago Press. Lakoff, G. and Kövecses, Z. 1983. The cognitive model of anger inherent in American English. (Berkeley Cognitive Science Report No. 10) Trier. Lakoff, G. 2006. The contemporary theory of metaphor. In D. Geeraerts (Ed.), Cognitive Linguistics: Basic Readings, 185-238. Berlin/New York: Mouton de Gruyter. (Reprint of The contemporary theory of metaphor. In A. Ortony (Ed.), Metaphor and Thought, 202-251. Cambridge: Cambridge University Press. 1993. Sodian, B. and Thoermer, C. 2005. Theory of mind. In W. Schneider, and B. Sodian (Eds.), Enzyklopädie der Psychologie: Entwicklungspsychologie, Band 2: Kognitive Entwicklung, 495-608. Göttingen: Hogrefe. Weber, F.E. 1993. Denken in Metaphern. Kognitive Semantik und französische Gefühlsmetaphorik. Frankfurt/Main: Peter Lang. Winner, E. 1988. The points of words. Children’s understanding of metaphor and irony. Cambridge: Harvard University Press.

IV. GRAMMAR

LEXICAL AND CONSTRUCTIONAL ORGANIZATION OF ARGUMENT STRUCTURE: A CONTRASTIVE ANALYSIS JOHAN PEDERSEN

Abstract. In this chapter I outline some principles for a contrastive analysis of basic clausal expressions in English and Spanish. They are formulated within a general framework of Cognitive Linguistics. The proposed principles for crosslinguistic variation are a challenge to the principles of parametric variation in syntax as envisioned in Chomsky (1981) and defended by Snyder (2001). From a theory-internal point of view, they lead to a reinterpretation of Talmy’s descriptive typology of macro-events (Talmy 1991, 2000). Particularly, they solve a serious flaw in Talmy’s typology due to his exclusive focus on lexicalization patterns. Contrastive analysis may provide insight into differing ways of organizing grammatical information. Construction grammar (CXG) suggests that clausal core information is organized by integrating at least two construction types: A) schematic constructions, B) lexical constructions (e.g. Croft 2001; Fillmore 1988; Goldberg 1995, 2006). In addition, clausal expressions are, according to some CXG-frameworks (e.g. Croft 2001), supposed to be built on language-specific construction types. I hypothesize that languages may differ according to the level of constructional specificity at which the core information is organized. English (and presumably other Germanic languages to some extent) tends to organize principal clausal information in schematic argument structure constructions, leaving secondary information for lexical (verbal) specification. Spanish (and presumably other Romance languages to some extent) seems to organize principal clausal information lexically in verbal argument structure constructions, leaving secondary information for schematically organized specification.

Introduction Snyder (2001) presents evidence that crosslinguistic variation in syntax is a domain in which general, explanatory principles are operative. He provides converging evidence from child language acquisition and comparative syntax for the existence of a syntactic compounding parameter in the classical sense of Chomsky (1981). This parameter determines the availability in some languages (e.g. Germanic languages) of a range of syntactic complex-predicate constructions, e.g. verb-particle constructions, and

242

Chapter Seventeen: Lexical and Constructional Organization of Argument Structure

unavailability in other languages (e.g. Romance languages). In this chapter, I will identify similar typological differences. But I will also show that there is in principle no reason why Snyder’s findings should be taken as evidence for a generative, parametric understanding of crosslinguistic variation in syntax. I will demonstrate that a construction grammar based contrastive analysis of basic clausal expressions may reveal similarly differing ways of organizing grammatical information. Such a framework may provide the same kind of principles for crosslinguistic variation in syntax that have been put forward by Snyder, though with a different theoretical perspective. In addition, it may account for a restricted availability of complex event expressions in Spanish that has not been accounted for in Snyder’s proposal. Within the general framework of cognitive linguistics, the theoretical implications lead to a reinterpretation of Talmy’s descriptive typology of macro-events (Talmy 1991, 2000). Construction grammar (CXG) suggests that the clausal information is organized in constructions at different levels of specificity (e.g. Croft 2001, Fillmore 1988, Goldberg 1995). There are, from this point of view, at least two basic devices for organizing the core information of the clause: A) Lexical constructions B) Schematic constructions According to some CXG-frameworks (e.g. Croft 2001), clausal expressions are supposed to be built on language specific construction types. Hence, assuming that clausal information is organized in language-specific constructions, it is here hypothesized that languages may differ according to the level of constructional specificity in which the information is organized. In typical usage it appears as if organizing principles are predictable from the properties of the clausal constituents. For instance, in (1) we cannot see whether the basic transfer meaning: ‘X caused Y to receive Z’ is organized exclusively in a lexical argument (valence) structure, centred in a trivalent verb with rich semantic content, or whether it is encoded holistically as a schematic pairing of form and meaning: [SUBJ, V, OBJ1, OBJ2] / ‘X caused Y to receive Z’, and specified lexically by the verb. The confusion that a prototypical ditransitive may cause is due to the verb being trivalent and the transfer-meaning of the clause being perfectly compatible with the basic meaning of the verb: ‘to give (something to someone)’. It therefore seems as if a schematic level is not needed to express the transfer-meaning here since the verb appears to be the organizing device. However, we still do not know whether there is a division of la-

Johan Pedersen

243

bour between a schematic form-meaning pairing, taking care of the basic transfer meaning, and a verbal form-meaning pairing that specifies the basic transfer meaning. It could be objected that if this were the case, the organization of grammar reflected in (1) would be characterized by informational overlap and redundancy. One could respond that even if redundancy may seem inappropriate, there is no reason to believe that simplicity or nonredundancy are psychologically valid characteristics of grammatical knowledge and representation (see e.g. Croft 2001; Goldberg 2006). Atypical combinations of words and expressions provide knowledge about linguistic diversity and variation. It may, however, also serve as a valuable source of data for the linguist to understand how basic clausal information is organized in the grammar, since clausal organizing principles are not necessarily transparent in usage. Atypical usage may thus offer clues on grammatical organization, which are not directly assignable to the constituents. When the basic clausal meaning cannot be derived from the meaning of the verb, this may be an indication that this information is organized at another, more schematic level in grammar, independently from the encoding of the verb, as in (2) from Goldberg (1995). (1) (2)

Peter gave her a present. Peter baked her a cake.

The chapter is structured as follows. Firstly, I will present a short introduction and discussion on the term construction, as it is used in the construction grammar framework, and its application in this chapter. Secondly, a tentative hypothesis for a typology of constructional specificity will be formulated for English and Spanish. Thereafter, I will analyze expressions in English and Spanish of volitional transfer (the ditransitive) and caused motion. Specifically, I will show that a contrastive perspective offers insight into principles for clausal organization, and important implications for previous typological research will be briefly outlined. Finally, conclusions will be drawn.

Constructions at different levels of specificity The term construction has always existed in the history of linguistics. However, prior to the formulation of the construction grammar framework, it was only loosely defined without significant theoretical interest. It was understood simply as a complex linguistic unit, often with reference to the clausal level of analysis. Indeed, it often retains this traditional meaning. In construction grammar frameworks (e.g. Croft 2001; Fillmore

244

Chapter Seventeen: Lexical and Constructional Organization of Argument Structure

1988; Goldberg 1995, 2006; Langacker 1987/91) emphasis is specifically placed on the theoretical importance of the term construction. In a synchronic perspective, constructions are non-derived pairings of form and meaning. Constructions exist in the grammar at different levels of specificity; see e.g. Goldberg (1995, 2006). Hence, a construction grammar emphasizes both abstract linguistic patterns and more substantial types and instances of usage. Constructions link formal properties (morphological, syntactic and phonological) to meaning (mostly semantic, but also discourse-functional and pragmatic) in a conventionalized relation. Grammars consist of many different construction types. Some examples are provided in Figure 1. Construction type

Form

Meaning

Spanish lexeme Spanish morpheme Spanish se Spanish idiom English ditransitive English rule

[casa] [-s] [se-V-a-OBJ] [más vale tarde que nunca] [SUBJ-V-OBJ1-OBJ2] [SUBJ-VP-AND-Ø-VP]

‘house’ ‘plural’ ‘impersonal meaning’ ‘better late than never’ ‘X causes Y to receive Z’ ‘conjunction reduction’

Figure 1. Different construction types

It is important to recognize that most expressions referred to here as specific constructions, contain different construction types in their internal structure. This is, e.g., the case in (3). (3)

¿Qué le hizo Pedro a la hija de Fernando? what DAT.3SG do.PST.3SG Pedro to the daughter of Fernando ’What did Peter do to Fernando’s daughter?’

Expression (3) involves the integration of a number of different constructions: x [Someone, do, to someone, what] – constructions x [Interrogative] – construction x [Double indirect object] – construction x [VP] – construction x [NP] – construction x [Lexeme] – constructions x [Inflection] – constructions (e.g. mode/aspect/tense)

Johan Pedersen

245

From this perspective, grammar consists of many different types of constructions, integrated into clausal expressions. Should we then understand every single form-meaning pair in grammar as a construction? Are there elements of coded meaning in grammar that are not constructions? The answer is yes, though it is crucial to emphasize that the term construction is meant to be very general and comprehensive. How do we then identify constructions? Constructions are interrelated and have properties inherited from other constructions, while also retaining an autonomous status in the grammar as carriers of grammatical information. Constructions are thus defined synchronically as non-derived pairings of form and meaning. Consequently, a subject is a form-meaning pair in grammar, but it is not a construction given that the subject can be derived directly from argument structure constructions, e.g. the transitive or the ditransitive construction, in which the subject plays an integrated role. The subject does not have features that cannot be derived from the form-meaning pairings of the argument structure constructions. Users continually (re)analyze and (re)categorize linguistic input in different construction types. The identification of constructions is therefore basically an empirical matter of measuring user’s categorization of linguistic input into different construction types (see e.g. Croft 2001; Tomasello 2000, 2003). Whether or not some element of form-meaning pairing is a construction, is a relevant empirical question, but it has limited theoretical interest. More interesting, from the point of view of construction grammar, are the following questions: x x

What types of construction organize the clausal core information? How is the core information organized in different languages, and by means of which construction types?

As mentioned in the introduction, two kinds of construction at different levels of specificity seem to play a privileged role in clausal organization: 1) The lexical (verbal) constructions, in which the information is organized by the verb in a verbal argument structure construction. 2) Schematic argument structure constructions.

A tentative typology of constructional specificity One important insight provided by construction grammar is that English tends to organize the basic clausal information in schematic argument structure constructions, and that this information is specified in lexical constructions, particularly by the verb (Goldberg 1995). Below I provide

Chapter Seventeen: Lexical and Constructional Organization of Argument Structure

246

evidence that Spanish expressions are not organized in the same way. Tentatively, I formulate the following hypotheses: x

x

English tends to organize basic clausal information in nonderived schematic argument structure constructions. The basic clausal information is complemented and specified lexically, primarily by the verb. Spanish tends to organize basic clausal information lexically by means of a verbal argument (valence) structure construction. The basic clausal information may be complemented by an schematically organized, and non-verbally derived, constructional specification.

Contrasting English and Spanish constructions In the next subsections, the hypotheses outlined in the previous section will be verified and exemplified. We will examine some frequent expressions in English and Spanish from a contrastive point of view, with respect to the constructional specificity and strategy of clausal organization. I will discuss how two principal construction types seem to be used differently in English and Spanish as devices for organizing core information. In particular, it will be demonstrated that the analysis has important implications for previous typological research.

The ditransitive construction The central sense of the ditransitive construction involves transfer between a volitional agent and a willing recipient (Goldberg 1995), as in (4). Spanish has parallel constructions, as in (5). (4) (5)

She made Peter a cake. Le DAT. 3SG

hizo una tarta a Pedro make.PST.3SG a cake for Pedro

In typical expressions of the ditransitive, as in (4) and (5), there are no indications that show whether the transfer-meaning is encoded in a lexical (verbal) construction on the basis of valence relations, or in a schematic argument structure construction, in which the formal pattern of SUBJ, V, OBJ1, OBJ2 codes the transfer-meaning, or by means of both devices. Atypical examples, such as (6), indicate, however, that the transfer-

Johan Pedersen

247

meaning is encoded in a ditransitive argument structure construction, and that the specification of the activity is encoded by the verb. (6)

She baked Peter a cake. (Goldberg 1995)

Schematic ditransitive construction: [SUBJ,V,OBJ1,OBJ2] / ‘X causes Y to receive Z’ Verbal construction: [baked] / ‘specification of the verbal process’ The main reason is that the transfer-meaning cannot plausibly be part of the lexical meaning of bake. If the central information, the transfermeaning, were organized by the verb, the lexical meaning of bake should include a special ditransitive variant, which is not plausible. The ditransitive argument structure is a frequent meaning pattern in Spanish, as exemplified in (5), though it does not in general combine with verbs that do not predict the characteristic transfer-meaning per se (Martínez Vázquez 2003), as in (6). Nevertheless, some examples like (8), in which the verb does not predict the ditransitive meaning, are perfectly acceptable in Spanish. (7) (8)

She cooked him a joint of meat. María le cocin-ó un asado. María DAT.3SG cook-PST.3SG a joint of meat

However, the use of the dative in the Spanish version in (8) does not necessarily imply a volitional agent and a willing recipient, which is the central meaning of the ditransitive construction in English. Example (8) could be uttered in at least the following contexts: (8a) (8b)

(8c)

María le cocinó un asado para dárselo. ‘María cooked him a joint of meat with the intention of giving it to him.’ María le cocinó un asado antes de irse para que no tuviese que hacerlo él. ‘María cooked a joint of meat before she left, so that he wouldn’t have to do it.’ María le cocinó un asado para demonstrarle que sabía cocinar. ‘María cooked a joint of meat to show him that she knew how to cook.’

248

Chapter Seventeen: Lexical and Constructional Organization of Argument Structure

Therefore, Spanish expressions with a dative object in which the verb does not predict the ditransitive meaning, as exemplified in (8), do exist. But they are not evidence for the existence of non-derived schematic ditransitive constructions in Spanish that are comparable with the English ditransive. They are rather evidence that Spanish expressions of ditransitive meaning, as in (5), are basically organized lexically by verbal constructions, and that the schematic ditransitive construction, characteristic in English, does not exist in Spanish as a non-derived element in the grammar. What (5) and (8) do suggest is that Spanish has a very commonly used non-derived dative-construction, which is not equivalent to the English ditransitive construction. I will tentatively hypothesize that it is a morphological construction with the abstract meaning of pointing out a beneficiary. The dative construction combines with a verbally based ditransitive argument structure as in (5), depending on a number of syntactic and pragmatic rules. In fact it combines with almost all types of argument structure, as the transitive structure in (8), or the intransitive structure in (9). (9) (9’)

¿Cómo le fue? How DAT.3SG go.PST.3.SG ‘How did it turn out for him?’ ¿Cómo fue? ‘How did it turn out?’

The possibility of combining the dative construction with almost all kinds of argument structure indicates its non-derived status as a productive grammatical construction. In examples like (8) and (9), the dative construction may be left out without changing the basic meaning of the clause, as shown in (9’). This is an indication that the dative construction is not involved in the encoding of the central meaning of the clause. It rather provides supportive clausal information by pointing out, schematically, a beneficiary. Spanish versions of the ditransitive meaning thus seem to support the hypothesized characteristics of Spanish as opposed to English: The basic information in Spanish is verbally organized, and tends to be specified by abstract schematic constructions.

Johan Pedersen

249

The caused motion construction The central meaning expressed in (10) is caused motion. (10)

He kicked the ball into the box.

This meaning appears to be provided by the verb since the semantic frame of kick implies ‘someone who kicks’, ‘something to kick at’ and ‘somewhere to kick it’, which is the basic meaning of the clause. When the semantics of the verb does not imply a sense of caused motion that is similar to the basic meaning of the clause, the existence in the grammar of a nonderived schematic caused motion construction is indicated (Goldberg 1995). (11)

He sneezed the napkin off the table.

Core information: [SUBJ,V,OBJ, OBL] / ‘X caused Y to move Z’ Secondary information: [(SUBJ) V ] / ‘(A) sneezed’ We are now disposed also to analyze (10) as an expression of caused motion that combines a schematic caused motion construction with a verbally organized specification. (10)

He kicked the ball into the box.

Core information: [SUBJ,V,OBJ, OBL] / ‘X caused Y to move Z’ Secondary information: [(SUBJ)V(OBJ)(OBL)] / ‘(A) kick (B) (C)’ The encoding in (10) is redundant in the sense that the two subconstructions (schematic + lexical construction) involve the same basic argument structure. But as already argued in the introduction, grammatical coding is very often redundant to some extent. In Spanish, we may perfectly well express the same meaning of caused motion, see (12), though it seems to be impossible when the verb does not predict the caused motion pattern per se, as in (13). (12)

Met-ió la pelota en la caja de una patada. place-PST.3SG the ball in the box with a kick ‘He kicked the ball into the box.’

Core information: [(SUBJ)metió (OBJ)(OBL)] / ‘X caused Y to move Z’. Secondary information: [ADV form] / ‘causal specification’.

250

(13)

Chapter Seventeen: Lexical and Constructional Organization of Argument Structure

*Estornud-ó la servietta de la mesa. sneeze-PST.3SG the napkin from the table ‘He sneezed the napkin off the table.’

Example (12) may then be analyzed as an expression of caused motion whose principal clausal information is organized lexically by the verb. The causal specification may be organized by a non-derived schematic adverbial construction (de una patada). This analysis provides further arguments for the idea, already supported by the contrastive analysis of the ditransitive construction, that in Spanish the principal clausal information is organized lexically by the verb, as opposed to English, in which it is organized in more complex pairs of form and meaning (schematic constructions). The analysis has been carried out within the construction grammar framework, and has therefore important implications for how we should interpret Snyder’s findings on cross-linguistic variation in syntax (Snyder 2001), as outlined in the introduction. His findings may not, as he claims, be due to the existence of a syntactic compounding parameter in the sense of Chomsky (1981). Snyder’s data may rather reflect a crosslinguistic difference of constructional specificity in the sense of the analysis suggested in this chapter. In particular, the present analysis has demonstrated that expressions of complex events, comparable to English counterparts, such as the double object dative, are not excluded in Spanish. They have, though, to be licensed and organized by the verb. The partial availability of complex event expressions in Spanish cannot be accounted for by Snyder’s global parametric variation.

A construction-based typology of macro-events A construction-based analysis of expressions of caused motion, cf. the previous section, has also interesting implications for Talmy’s descriptive typology of macro-events (Talmy 1985, 1991, 2000), and for its theoretical status. Talmy found that in expressions of macro-events - the most important sub-domain is the motion event – some languages, e.g. Germanic languages, tend to lexicalize the main event in a satellite (so-called “satellite framing”), and the co-event by the verb. Other languages, e.g. Romance languages, tend to lexicalize the main event by the verb (socalled “verb framing”), and may express the co-event outside the verb, typically by adding an adverbial. Talmy’s typology shows, very convincingly, the predominance in some languages, as opposed to other languages, of certain patterns of form-meaning mapping in expressions of

Johan Pedersen

251

macro-events. In Pedersen (2009), however, I have argued that the typology suffers from being formulated exclusively in terms of lexicalization patterns. Its status as a universal typology requires a framework that goes beyond having the lexeme as the basic unit of the typology. It should include both the lexical level and more schematic constructional levels of analysis. It would thus profit, I argue, in terms of explanatory power and predictability, from being fitted into a construction grammar framework, in which constructions of different degree of specificity are the basic constituents. I argue that despite the considerable variation that can be observed, Germanic languages tend to map the main information of expressions of macro-events onto a complex schematic construction and the secondary information onto a lexical (verbal) construction. Romance languages tend to map the main information onto the verb, i.e. a lexical construction, while the secondary information may be mapped onto a schematic construction. These regularities represent a special case of the more general principles outlined in the present chapter, which is concerned with the organization of clausal information in general, not only with the construction of macro-events. According to some linguists, one of the main shortcomings of Talmy’s typology is that some languages do not seem to fit in his binary typology (e.g. Slobin and Hoiting 1994, Slobin 2004 and Zlatev and Yangklang 2004). In fact, from many studies, most of them concerned with the study of the motion event, a picture has begun to emerge of substantial deviations from the Talmian typology in almost all languages (e.g. Aske 1989; Berman and Slobin 1994; Gennari et al. 2002; IbarretxeAntuñano 2004a, 2004b; Pedersen 2009; Slobin and Hoiting 1994; Slobin 1996, 1997, 2000, 2004; Zlatev and Yangklang 2004). See also Talmy (2005), and Beavers et al. (2008) for overviews. In Pedersen (2009), I suggest a specific strategy for research into macro-events. Research projects should develop a typology of constructions, not of languages. This is in line with a more general trend in typological research away from typologizing languages as a whole, to typologizing particular situation types expressed in a language (see e.g. Croft et al. 2008). This strategy has important implications for how we should interpret the data that do not fit the patterns suggested by Talmy. Pedersen (2009) provides the examples reproduced here as (14)-16), in which the macro-event is state change. The English and Danish versions may show the typical “Germanic” pattern, as demonstrated in (14), in which the main information (MI = state change), according to Talmy, is expressed by the satellite out/ud. However, a “Romance” type is available as well. In Danish, it is very common to construe the main information (the state change)

Chapter Seventeen: Lexical and Constructional Organization of Argument Structure

252

by the verb: jeg slukkede (‘I put out’) stearinlyset, as in (15). In the Spanish version in (14), the main information (MI) is expressed by the verb apagué, whereas the secondary information, the causal factor, may be expressed by an adverbial construction. An alternative would be, as in (16), to express only the secondary information (SI), by means of the verb: soplé (‘I blew’) la vela, leaving the main information for inferential interpretation. (14)

I

blew

out. MI Jeg pustede stearinlyset ud. MI Apag-ué la vela de un soplido. put out-PST.1SG the candle by a blow MI

(15)

Jeg I

(16)

Sopl-é blow-PST.1SG SI

the candle

slukke-de stearinlyset. put out-PST the candle MI la vela. the candle

(Talmy 1991) (Danish) (Spanish)

(Danish)

(Spanish)

Examples (14)-(16) demonstrate what has been pointed out by many scholars, that even though there are important typological differences between e.g. Germanic and Romance languages in expressions of macroevents, there is no simple clear-cut distinction. In Pedersen (2009), the typological patterns that characterize expressions of macro-events are interpreted as an information structure phenomenon. Macro-event constructions, i.e. constructions of the main information (MIC) and the supportive information (SIC), are the basic constituents of the typology. It is hypothesized that pairs of MIC/SIC are distilled out of usage due to the user’s constant generalizations from usage. They are procedural devices for organizing the clausal information. Some organizational procedures (types of MIC/SIC) are thus more entrenched in the grammar of some languages than in others. This is the essence of the proposed typological principles. A construction-based typology of macro-events, as the one developed in Pedersen (2009), is not opposed to a typology of lexicalization. Nevertheless, it has a more general scope, and it places the invaluable observa-

Johan Pedersen

253

tions made by Talmy in a different theoretical perspective. Furthermore, it provides a concrete solution to a serious flaw in Talmy’s typology. In (17) it is counterintuitive that the satellite lexeme (out), whose basic meaning is: ‘leaving a container’ (Rudzka-Ostyn 2003), should determine the main event (‘X caused Y to move Z’). (17) (18)

Peter pushed him out of the restaurant. Pedro lo ech-ó del restaurante Pedro ACC.3SG.MASC throw-PST.3SG out from the restaurant a empujones. by pushing ‘Peter pushed him out of the restaurant.’

In a construction-based framework, the characteristic feature of the English version in (17) is that the main clausal information is determined by a schematic caused motion construction: [SUBJ V OBJ OBL] / ‘X caused Y to move Z’. This information is specified by a lexical construction (pushed). The Spanish version in (18), opposed to the English version, organizes the main clausal information in a lexical argument structure construction centred in the verb (echó), and specifies this information in a schematically organized adverbial construction. This analysis indicates that the cross-linguistic difference that may be observed in expressions of caused motion like example (17)-(18) should not be interpreted as differing lexical mappings of directed motion, as hitherto claimed, but rather as manifestations of differing constructional organization of clausal information in English and Spanish. The existence of basic typological differences in expressions of macro-events, as the ones originally observed by Talmy, cannot be denied, no matter how they are described and understood. In the present chapter, I have discussed the possibility that such regularities and differences may reflect general typological principles that govern the organization of clausal information. The essence of the principles of parametric variation (Snyder 2001) is that some languages have a compounding parameter in the Chomskian sense, and other languages do not. The compounding parameter enables, for instance, English to organize expressions of complex events such as resultatives, or double object datives. It is a problematic implication of Snyder’s proposal that languages like Spanish that lack the compounding parameter are supposed not to permit expressions of complex events that are comparable to the English counterparts. The present framework, however, accounts for the fact that very similar versions of, e.g., English resultatives, or double object datives, do exist in Spanish. We

254

Chapter Seventeen: Lexical and Constructional Organization of Argument Structure

have seen that the condition for Spanish, as opposed to the condition for English, is that the argument structure has to be lexically licensed, and organized, by the verb. We may, therefore, tentatively state that the present proposal represents a challenge to the parametric principles of crosslinguistic variation proposed by Snyder (2001).

Conclusion and perspectives Construction grammar frameworks suggest that two principal construction types determine how information is organized in simple clauses: Schematic constructions and Lexical constructions. I have argued, on the basis of contrastive data, that English and Spanish, and presumably other West European languages, are different regarding how these organizing devices are used, and specifically at which level of constructional specificity the basic clausal information is organized. It has been demonstrated that English tends to organize basic clausal information in schematic argument structure constructions, leaving secondary information for lexical (verbal) specification. Spanish seems to organize basic clausal information lexically in a verbal argument structure construction, leaving secondary information for schematically organized specification. This framework provides a reinterpretation of Talmy’s descriptive typology of macro-events that solves a serious flaw in the typology due to its exclusive focus on lexicalization patterns. Acquisition data indicate that Spanish learners find it particularly difficult to comprehend and acquire English expressions in which the basic schematic meaning cannot be derived directly from the lexical meaning (Martínez Vázquez 2003). It is therefore also likely that research into L2acquisition of argument structure will turn out to be supportive for the hypothesis formulated in this chapter. The hypothesis represents a challenge to the Chomskian claim (1981), defended by Snyder (2001), that the difference between English and Spanish is due to a parameter setting of cross-linguistic variation in syntax.

References Aske, J. 1989. Path predicates in English and Spanish: A closer look. In K. Hall, M. Meacham, and R. Shapiro (Eds.), Proceedings of the Fifteenth Annual Meeting of the Berkeley Linguistics Society, 1–14. Berkeley: Berkeley Linguistics Society. Beavers, J., Levin, B. and Shiao Wei, T. 2008. The typology of motion expressions revisited. Draft.

Johan Pedersen

255

http://www.stanford.edu/~bclevin/motion-typology-working.pdf Berman, R.A. and Slobin, D.I. 1994. Relating events in narrative: A crosslinguistic developmental study. Hillsdale, NJ: Erlbaum. Chomsky, N. 1981. Lectures on government and binding. Dordrecht: Foris. Croft, W. 2001. Radical construction grammar. Oxford: Oxford University Press. Croft et al. 2008. Revising Talmy's typological classification of complex events. Ms. http://www.unm.edu/~wcroft/WACpubs.html Fillmore, C. J. 1988. The mechanisms of Construction Grammar. In S. Axmaker, A. Jassier and H. Singmaster (Eds.), Proceedings of the Fourteenth Annual Meeting of the Berkeley Linguistic Society, 35-55. Gennari, S. P., Sloman, S.A., Malt, B.C. and Fitch, W.T. 2002. Motion Events in language and cognition. Cognition 83: 49-79. Goldberg, A. E. 1995. Constructions. A construction grammar approach to argument structure. University of Chicago Press, Chicago. —. 2006. Constructions at work. The nature of generalization in language. Oxford: Oxford University Press. Ibarretxe-Antuñano, I. 2004a. Motion events in Basque narratives. In S. Strömqvist and L. Verhoeven (Eds.), Relating events in narrative: Typological and contextual perspectives, 89-111. New Jersey: Lawrence Erlbaum. —. 2004b. Dicotomías frente a continuos en la lexicalización de los eventos de movimiento [Dichotomies vs. Continua in the lexicalization of motion events]. Revista Española de Lingüística 34.2, 481-510. Langacker, R.W. 1987/91. Foundations of Cognitive Grammar, Vol. I + II. Stanford, CA: Stanford University Press. Martínez Vázquez, M. (ed.). 2003. Gramática de Construcciones. Contrastes entre el inglés y el español. Universidad de Huelva. Pedersen, J. 2009. In press. The Construction of Macro-events. A typological perspective. In C. Butler and J. M. Arista (Eds.), Deconstructing constructions, 25-62. Amsterdam/Philadelphia: John Benjamins. —. 2008. Organización léxica y esquemática – una perspectiva tipológica. Papers from the VIII congress on general linguistics 2008. Madrid. Rudzka-Ostyn, B. 2003. Word power: phrasal verbs and compounds. A cognitive approach. Berlin: Mouton de Gruyter. Slobin, D. I. 1996. Two ways to travel: verbs of motion in English and Spanish. In M. Shibatani and S. A. Thompson (Eds.), Grammatical constructions: Their form and meaning, 195-219. Oxford: Oxford University Press.

256

Chapter Seventeen: Lexical and Constructional Organization of Argument Structure

—. 1997. Mind, code and text. In J. Bybee, J. Haiman and S. Thompson (Eds.), Essays on Language Function and Language Type, 437-68. Amsterdam: John Benjamins. —. 2000. Verbalized events: a dynamic approach to linguistic relativity and determinism. In S. Niemeier and R. Dirven (Eds.), Evidence for linguistic relativity, 107-138. Amsterdam: John Benjamins. —. 2004. The many ways to search for a frog: linguistic typology and the expression of motion events. In S. Strömqvist and L. Verhoeven (Eds.), Relating events in narrative: Typological and contextual perspectives, 219-257. New Jersey: Lawrence Erlbaum. Slobin, D. I. and Hoiting, N. 1994. Reference to movement in spoken and signed languages: Typological considerations. In S. Gahl, A. Dolbey and C. Johnson (Eds.), Proceedings of the Twentieth Annual Meeting of the Berkeley Linguistic Society, 487–505. Berkeley: Berkeley Linguistics Society. Snyder, W. 2001. On the nature of syntactic variation: Evidence from complex predicates and complex word-formation. Language 77(2): 324-342. Talmy, L. 1985. Lexicalization patterns: Semantic structure in lexical forms. In T. Shopen (ed.), Language typology and syntactic description (vol. 3): Grammatical categories and the lexicon, 57-149. Cambridge: Cambridge University Press. —. 1991. Path to realization: a typology of event conflation. In L. A. Sutton, C. Johnson and R. Shields (Eds.), Proceedings of the Seventeenth Annual Berkeley Linguistics Society, 480-519. —. 2000. Toward a cognitive semantics. Vol.1 and 2. MA Cambridge: MIT Press. Tomasello, M. 2000. First steps toward a usage-based theory of language acquisition. Cognitive Linguistics 10(1): 61-82. —. 2003. Constructing a language. A Usage-based theory of language acquisition. Cambridge, MA/London: Harvard University Press. Talmy, L. 2005. Written interview on my work conducted by Iraide Ibarretxe: Part 1. Annual Review of Cognitive Linguistics, vol. 3, 325-347. Amsterdam: John Benjamins. Zlatev, J. and Yangklang, P. 2004. A third way to travel: the place of Thai in motion event typology. In S. Strömqvist and L. Verhoeven (Eds.), Relating events in narrative: Typological and contextual perspectives 159-190. New Jersey: Lawrence Erlbaum.

IMPAIRED ATTENTION TO PHONOLOGICAL INPUT, DYSFUNCTIONAL LEXICAL “FREQUENCY COUNTERS” AND IMPAIRED GRAMMATICAL MORPHOLOGY: A POSSIBLE CAUSAL CHAIN? JUSSI NIEMI, ALEXANDRE NIKOLAEV AND KENNETH HUGDAHL

Abstract. Psycholinguistic studies of Finnish morphology show that in this language – excluding the very high frequent end of the continuum – inflected nouns are morphologically parsed into their components during perception. The present experimental study discusses the ontogenetic path of processing of these types of lexical items by Specific/Familial Language Impairment (SLI/FLI) speakers and non-impaired controls. FLI speakers have been previously shown to display reduced neural activation in temporal and frontal lobe areas critical for speech processing and phonological awareness (Hugdahl et al. 2004). Moreover, our previous studies have shown that they, unlike controls, are not able to “switch ears” in a forced left-ear condition in a dichotic task using CV syllables (Niemi et al. 2003). Finally, our most recent analyses of this population show that they are, unlike controls, immune to various lexical frequency indices (lexical, lemma, hapax legomenon frequencies). Thus, expressed through a functional metaphor, their lexical-level “frequency counters” – wherever/whatever these may be – do not function normally. All in all, we hypothesize that these three abnormal areas of FLI information processing are causally related, most probably with the attentional deficit as the (most) underlying cause.

Introduction Finnish nouns may be claimed to have as many as 82 (e.g., Nykysuomen sanakirja ['Dictionary of Present-day Finnish'], 1973, 2401 pp.) or 49 inflectional classes (e.g., Suomen kielen perussanakirja ['Basic Dictionary

258

Chapter Eighteen: Impaired Attention to Phonological Input

of Finnish', 1990], 2008 pp.). Allowing for drastic morphophonological and phonological abstractions the number of these paradigms can be reduced to half a dozen (see, e.g., Karlsson 1982). Whatever the number of the paradigms in the descriptions of autonomous morphology, it is a pretheoretical fact that in Finnish multi-member phonological strings (partially) determine the various surface inflectional patterns (for mathematical modeling of Finnish verb inflection, see the analogy model of Skousen 1989; for an optimality-theoretical view of non-categorical noun inflection in this language, see Anttila 1997). Another feature that makes the present language a highly complex one inflectionally is that speakers often have to choose among both stem and suffix allomorphs to arrive at the correct output (Table 1).1 As for nouns, they (and other noun-like lexical classes like adjectives and numerals) have 15 productive cases, many with different plural allomorphs. Table 1. A partial number and case matrix of a subset of Finnish noun paradigms. Nom.sg.

Part.sg

Gen.sg

Part.pl

'bottle'

pullo

pullo+a

pullo+n

pullo+ja

Ill.pl pullo+ihin

'horse'

hevonen

hevos+ta

hevos+en

hevos+ia

hevos+iin

'bullet'

kuula

kuula+a

kuula+n

kuul+ia

kuul+iin

'face'

naama

naama+a

naama+n

naamo+ja

naamo+ihin

'mine'

kaivos

kaivos+ta

kaivokse+n

kaivoks+ia

kaivoks+iin

'boot'

saapas

saapas+ta

saappaa+n

saappa+ita

saappa+isiin

'product'

tuote

tuote+tta

tuottee+n

tuotte+ita

tuotte+isiin

'teddy'

nalle

nalle+a

nalle+n

nalle+ja

nalle+ihin nim+iin

'name'

nimi

nime+ä

nime+n

nim+iä

'color'

väri

väri+ä

väri+n

väre+jä

väre+ihin

'fire'

tuli

tul+ta

tule+n

tul+ia

tul+iin

In addition to competing stem and affixal allomorphy, the assumed processing load in Finnish noun inflection is enhanced still as we observe that (i) each base (non-derived, non-compound) noun can have as many as 150 paradigmatic forms, and with possessives and clitics as many as 2000 forms, like saappa+ita+mme+ko+han, which is parsed as 'boot'+partitive pl.+possessive 'our'+question clitic+pragmatic clitic (Karlsson and Koskenniemi 1985). (ii) Moreover, the lexical frequency of nouns with 1

For the importance of stem allomorphs for processing in Finnish, see Niemi, Laine and Tuominen (1994) and Järvikivi and Niemi (2002a, 2002b).

Jussi Niemi, Alexandre Nikolaev and Kenneth Hugdahl

259

more than one stem form is high, since only one fifth (viz., 21%) of the noun lexemes exhibit one stem form (see Table 1, for relevant lexical frequency data see table in Karlsson 1982, p. 201). (iii) In running texts, the ratio of non-zero inflectional morphs to whole words is as high as 0.62 to 1 (calculations based on data from Pajunen and Palomäki 1984). (iv) Moreover, according to the corpus analyses performed by Vannest et al. (2002) as few as 2.6 per cent of Finnish noun lexemes in running text are monomorphemic nouns in the nominative singular form (no. of nouns: 1,022,944), and (v) in the same newspaper corpus over 95% of the morphologically complex words do not occur more than once in a million words (Järvikivi, Bertram and Niemi 2006). The most modest psycholinguistic hypothesis that can be constructed on the basis of the structural facts and frequency of use data ventilated above is that a speaker of Finnish – whether a child or an adult – can rarely expect to encounter the nominative singular as the first exemplar of a novel noun. Furthermore, even if s/he were to encounter this form of a particular lexeme, it would very often be of ambiguous cue value as to the morphological paradigm membership of the word. Similarly, due to frequent affixal homonymy across inflectional paradigms encountering one or two inflected forms may not be of any help either. Rather, in some instances, in order to correctly fill in the slots of the morphological matrix the speaker may have to experience several forms of the lexeme, most probably instances of the set of the so-called focal forms of Field Morphology (Paunonen 1976, see also Table 1).

Method Stimuli In order to test the cue validity of word shape analogies (schemas) in noun inflection, sets of disyllabic pseudowords were constructed on the basis of the phonological bodies of real lexemes (body 'phonological word minus C onset'). Based mostly on recurrent and frequent error patterns reported in studies of early stages of the acquisition of Finnish morphology (e.g., Nesser 1981; Niemi and Niemi 1985, 1987; Räisänen 1975; Toivainen 1980) as well as on a study involving 10-year-old SLI children and their age-matched peers (Niemi 1999) a stimulus set of 180 pseudoword items was prepared for the wug procedure (Berko 1958) in which the carrier frame was (see, Niemi, Järvikivi and Heikkinen 1999 for details):

260

Chapter Eighteen: Impaired Attention to Phonological Input

Tässä on yksi PSEUDOWORD: NOM.SG [given]. lit. 'Here is one X' Tuolla on paljon PSEUDOWORD: PARTITIVE PLURAL [expected] lit. 'There are a-lot-of X's' (i.e. 'there we have a lot of X's') The pseudoword paradigms are either phonologically (rule-driven) or lexically conditioned. In the rule-driven categories it is the phonological structure of the item that unequivocally determines the shape of its partitive plural. The item types are as follows. A. The phonologically conditioned set is composed of four inflectional categories: (i) In the vullo class the expected partitive plural is the agglutinative vullo+ja and (ii) in the tevonen type the expected response would be tevos+ia. The inclusion of the nen forms was motivated in order to test the productivity of the highly idiosyncratic nen : se alternation. This paradigm is formally unique and a morphological island in view of, e.g., Field Morphology. Two sets of a-final items were created: (iii) the so-called vuula type, in which the labial initial syllable vowel dictates that the second syllable /a/ is to be deleted in, e.g., the partitive plural (vuula : vuul+ia), and (iv) the vaama type, in which the final /a/ is raised to /o/ (vaama : vaamo+ja) (see 'bullet' vs. 'face' in Table 1). B. The lexically conditioned set is made up of representatives of five categories, of which two are s-final: (i) Words in one group are expected to change the stem-final /s/ into kse in, e.g., partitive plural (e.g., naivos : naivoks+ia, see 'mine' in Table 1), and (ii) those in the other to drop the /s/ (e.g., vaapas : vaappa+ita). The former type is the more productive of the two, as it incorporates novel lexemes (e.g., slang words, lexical loans, clippings, acronym-based words) more effectively than does its competitor. Moreover, it also attracts words from the competing paradigm(s) in L1 and L2 acquisition of Finnish (for the former, see Niemi and Niemi 1987, for the latter, see Nikolaev 2002). In addition, half of the s : kse items were built on the analogy of lexemes that show retrograde gradation.2 The retrograde gradation items are crucial if we wish to test whether speakers really operate on lexical connections when inflecting these pseudowords. 2

Typically, forms subject to consonant gradation in Finnish carry the strong form in the nominative singular, as in laƾka (nom.sg.) : laƾƾa+n (gen.sg.) 'thread, yarn' (orthographically and , respectively). However, there are fossilized forms that have the strong form in the oblique forms, as kaƾƾas (nom.sg.) : kaƾkaa+n (gen.sg.) 'fabric, cloth' (see also (iii), below).

Jussi Niemi, Alexandre Nikolaev and Kenneth Hugdahl

261

(iii) The present e-final test stimuli are all based on the fossilized lexical stock with -itA as the expected partitive. Like the s-final nouns, this e-final stock may carry retrograde gradation as in tuote (nom.sg.): tuottee+n (gen.sg.) 'product'. Thus, as in the case of s-final nouns, half of the present e stimuli are based on word shapes with retrograde gradation. Finally, two competing i-final categories, viz., (iv) simi and (v) käri types (see Table 1 for 'name' and 'color') are included in the test. The analogically expected partitive plurals for these types would be sim+iä and käre+jä, respectively. As expected, the more isomorphic käri : käre+jä is the productive one of these two.

Subjects and procedures The experimental questionnaire with its 180 pseudowords was administered to the 10- and the 14-year-old groups and to four FLI speakers. The group-tested 10- and 14-year-old subjects (N = 16 in both groups) gave written responses to the questionnaire, while the four FLI speakers and the 7-year-old group (N = 66) were individually tested and they gave the responses in oral form. The responses of the FLI group were transcribed on site by a second interviewer, while one of them presented the list orally. The responses of the 7-year-old group were audio taped and subsequently transcribed by a phonetically trained Linguistics PhD student. For practical reasons (fatigue, attentional factors), only the first 60 items of the 180-item questionnaire were presented to the 7-year-old children, and the number of item types per category for them were: vullo 6, tevonen 7, vuula 6, vaama 7, naivos 7, vaapas 6, kuote 7, simi 7, and käri 7. The 98 non-impaired speakers were tested in their schools, the FLI speakers in their homes. The FLI speakers come from an extended family with Specific Language Impairment attested for members of three generations (5 impaired speakers out of the 22 that we have tested). The affected family members have been shown to exhibit abnormal response patterns to linguistic stimuli, both in lateralization tasks with dichotic listening using CV syllables (Niemi et al. 2003), and in fMRI scans testing their attention to vowel, syllable, pseudoword and real word stimuli (Hugdahl et al. 2003) as well as in neural activity (fMRI scans, see Hugdahl et al. 2004). Due to her inability to internalize the gist of the wug task, Grandmother’s responses are excluded from the present analyses. Thus, the FLI data are derived from Mothers 1961 and 1963 and Sons 1985 and 1991. (See Figure 1 for the pedigree of the FLI extended family.)

262

Chapter Eighteen: Impaired Attention to Phonological Input

Figure 1. The pedigree of the members of the FLI extended family tested for the linguistic study.

Results Among the normal speakers, the numbers of overall correct responses increase drastically with age, as the youngest group obtains a 49.3 per cent correctness rate, the 10-year-olds a 60.5 per cent correct, and the teenagers respond correctly in four out of five instances (79.7 per cent) (Table 2). There is, however, no consistent age effect among the four FLI speakers, although Mother 1963 scores higher correctness percentage than her Son (56.1 % and 43.5 %, respectively), since for the other mother and son pair the pattern is opposite (Mother 1961: 52.0 %, and her Son 1985: 60.2 %). When we compare the phonologically and lexically conditioned categories, it becomes evident that – not unexpectedly – the former categories are mastered better than the lexically conditioned ones (see Table 2). There is also a clear ontogenetic path of development in the mastery of these forms in the non-impaired groups.3 3

In the present context we will not discuss the results paradigm by paradigm. For this purpose the interested reader may analyze the response data in Table 3 in detail, and s/he will observe clear effects of (i) type of morphological operation (here: degree of agglutination – fusion), (ii) overall productivity, and (iii) paradigm competition (like s-paradigms)/paradigm uniqueness (-nen paradigm). The third effect could be aptly called that of local paradigm density.

Jussi Niemi, Alexandre Nikolaev and Kenneth Hugdahl

263

Table 2. Number of correct responses (with total N's in second lines) in phonologically conditioned and lexically conditioned morphological types. 7-year-olds Phonologically conditioned Lexically conditioned Overall Correct

10-yearolds

14-yearolds

FLI speakers

69.1

80.0

88.7

65.2

%

1443

1280

1280

319

N

36.0

44.3

72.0

43.7

%

2142

1536

1507

364

N

49.3

60.5

79.7

53.0

%

1769/3585

1704/2816

2220/2787

362/683

N

Table 3. Total number of correct responses (with N's in second lines) per morphological type as well as the lexical frequencies (and logarithmical lexical frequencies) of morphological types used in the present experiment. 7-year-olds

10-yearolds

14-yearolds

FLI speakers

Lexical frequency (logarithmical)

93.1

96.3

94.7

90.1

1667 (7.42)

288

320

320

80

Tevonen

83.9

91.6

90.9

66.3

336

320

320

80

Vuula

48.7

57.2

83.4

36.3

378

320

320

80

59.6

75.0

85.6

67.5

441

320

320

80

Simi

38.3

29.8

54.5

48.5

441

336

336

80

Käri

56.0

69.1

84.9

48.5

441

304

304

68

47.2

50.7

97.6

37.5

441

288

288

72

20.9

38.9

53.3

60.3

378

288

272

68

15.6

35.0

71.0

26.3

441

320

307

76

Vullo

Vaama

Naivos Jangas Kuote

3287 (8.1) 2092 (7.65) 1090 (6.99) 108 (4.68) 4442 (8.4) 2628 (7.87) 440 (6.09) 828 (6.72)

264

Chapter Eighteen: Impaired Attention to Phonological Input

The effect of frequency – the most pervasive extra-grammatical factor in processing of lexical and sub-lexical structures – is seen in the present results through the fact that various indices of frequency correlate with the relative mastery of the present morphological categories among the nonimpaired groups (see Figure 2).

Figure 2. Correct responses per paradigm (%, y-axes) of 7, 10, and 14-year-old non-impaired children and FLI speakers against (a) logarithmical lexical frequency, (b) logarithmical hapax legomenon frequency and (c) median of logarithmical lemma frequency of paradigms (x-axes). Each circle represents an inflectional paradigm.

By calculating the linear regression values based on the lexical frequency data (a number of members of the inflectional type, Figure 2 (a)) and the correctness percentage scores in Table 3, we arrive at the following figures: for the youngest non-impaired group p-value=0.16,

Jussi Niemi, Alexandre Nikolaev and Kenneth Hugdahl

265

adjusted R square (henceforth R2) = 0.16 and correlation coefficient (henceforth r) = 0.51; for the 10-year-old group p-value=0.05, R2=0.36 and r=0.66; and for the oldest non-impaired group p-value=0.004, R2=0.68 and r=0.85. Lexical frequency is a strong predictor for productivity of the inflectional type and it does not appear to correlate with the correctness scores among the FLI speakers, since their correlation coefficient is about zero (viz., 0.06) and p-value=0.86 (R2= -0.14). Plotting correct responses of non-impaired children against hapax legomenon frequency (Figure 2 (b)), which is one of the productivity indices of paradigms (see, e.g., Baayen 1994), shows the same tendency to increase r and R2 with age (7-year-old group: r=0.37, R2=0.01, pvalue=0.32; 10-year-old group: r=0.54, R2=0.19, p-value=0.14; 14-yearold group: r=0.80, R2=0.59, p-value=0.009). Instead, the FLI speakers do not reveal any influence of the productivity level of paradigms: r=-0.08, R2=0.01, p-value=0.84. Lemma frequency of the inflectional type is not a strong predictor of productivity. However, it correlates with lexical frequency and this correlation is negative, when we use the median as an average of lemma frequency.4 The lemma frequencies for each word were counted from the CSC Language Bank corpora (which includes 131.4 million word tokens; http://www.csc.fi/english). As we can see from Figure 2 (c), there is the same tendency toward increasing r and R2 with age in non-impaired children and there is also a lack of correlation between given components in the FLI speakers (7-year-old group: r=-0.19, R2=-0.1, p-value=0.63; 10year-old group: r=-0.37, R2=0.01, p-value=0.32; 14-year-old group: r=0.68, R2=0.38, p-value=0.05, FLI speakers: r=0.25, R2=-0.07, pvalue=0.52).

Discussion On the basis of the experiments described in this chapter it is apparent that the mastery of the intricate inflectional patterns of Finnish has not 4

High lexical frequency and the specific distribution are qualities of an inflectional type. In the specific distribution the words with low lemma frequency predominate over the words with high lemma frequency (the novel words choose productive inflectional types and their lemma frequency, frequency of use, is mainly low). Nonproductive inflectional types, on the contrary, have low lexical frequency and these words are in frequent use for the most part (if a word of the nonproductive type has low lemma frequency, it is at risk to transfer to the productive type or to disappear). Therefore, lexical frequency of an inflectional type and the median of its lemma frequency are inversely proportional.

266

Chapter Eighteen: Impaired Attention to Phonological Input

stabilized at the beginning of the school-age (age 7) and not even later (age 14). During the last few decades similar observations regarding the slow emergence of grammatical categories/processes have been occasionally observed (in fact at least from the late 1960s onwards, see, e.g., C. Chomsky 1969, Harris 1978, Nippold, ed. 1988). Recently, using the wug paradigm, S. Niemi has found converging evidence for the slow stabilization of Swedish morphology, also with schema-analogical links guiding the plural inflection in that language (see Niemi 2002; for late stages of non-impaired acquisition of Finnish syntax, see Niemi and Hägg 1999; Pyykkönen, Niemi and Järvikivi 2003). When we look at the nine morphological categories separately (Table 3), it would at first appear that agglutination would be a major explanation for a relatively high number of correctness rates in the present data, since the vullo type (with its partitive plural vullo-ja) obtains close to the ceiling hit rates even among the youngest non-impaired group and the FLI speakers, both of which produce 9 out of 10 instances correctly. That about 30 to 40 % of the partitive plurals that can carry segments with retrograde gradation in fact did so in the present study is a definite proof that the three subject groups, including the FLI speakers, resort to lexical schema analogies in these experiments. Retrograde gradation is not a productive morphological operation in the s-nouns and thus there are no mere morphological grounds to inflect, e.g., the pseudoword stimulus jangas in partitive plural as janka+ita. Rather, in order for them to have arrived in such an end-result, the speakers must have referred to the lexical stock that carries this type of association, like kangas : kankaita 'fabric, cloth'. The claim for the use of the lexical link is further corroborated, if we observe that, in addition to this unproductive paradigm, retrograde gradation operates also in environments that are nearly unproductive, as in the case of e-final nouns, where the productive type is the fully agglutinative nalle : nalle+ja paradigm (vs. tuote : tuotte+ita). However, although non-impaired Finnish children at age 7 do not manifest full adult morphology, even they are better than the present two adult and two adolescent speakers with Familial Language Impairment in some respects. Non-impaired children as young as age 7 have, for instance, internalized the constraint that states that /a/ final disyllabic nouns with a labial first syllable vowel structure (like vuula) are to lose their final /a/ in some inflectional environments (e.g., vuul+ia). This is not the case with the present FLI speakers, who prefer vowel alternation over deletion.5 Cognitively the alternating pattern is the more natural alternative, 5

Note that terms like alternation and deletion are here used as mnemonics only, not as terms implying any spatiotemporal morphological process. In fact, the basic

Jussi Niemi, Alexandre Nikolaev and Kenneth Hugdahl

267

since it better retains the meaning-to-form isomorphy than does the deletion type, in which the affix is phonologically assimilated with the stem syllable (vuu.lia), while the alternating type retains the disyllabicity of the stem and adds a syllable cum affix to it (vaamo+ja, syllabified as vaa.mo.ja) (see Dressler 1985 for a semiotic view of morphology; for similar observations of non-impaired early acquisition patterns of Finnish morphology, see Niemi and Niemi 1987). Furthermore, in the i-final nouns the productive, morphologically uniform and isomorphic käri : käre+jä type gains more ground over simi : sim+iä in the non-impaired populations with increasing subject age. This ontogenetic development is taken to suggest that enhanced experience with Finnish lexicon and morphology adds to the schema-analogical capacity of non-impaired speakers. Similarly, the non-isomorphic and unproductive s-deletion type vaapas : vaap(p)a+ita and the unproductive retrograde derivation type kuote : kuot(t)e+ita attract more members in function of increasing age. Thus, the more instances of a morphological paradigm – even of an unproductive one – the speakers encounter, the stronger will the schematic associations among the analogical neighbors become. This strongly implies that intraparadigmatic, schematic associations are strengthened by frequency of use – a notion sine qua non in most psycholinguistic research, and especially in lexical studies. A major finding in the present context is also that this morpho-lexical strengthening extends well into the latter half of the first chronological decade, if not further. In both i-final types the FLI group behaves in a random fashion, thus indicating that when they have encountered a word of the present types, they have not tagged it with its morphological category, or they cannot construct schema-analogical links in this part of the Finnish lexicon. The reason for this incapacity may be found among the following factors: The käri : käre+jä type, albeit isomorphic and productive, may easily be subsumed by the non-alternating type(s) with -iA as the partitive plural marker. This in fact has taken place diachronically in some dialects of Finnish (although not in the two areas of the present speakers). In these dialects the partitive plural of 'color' is thus vär+iä, not väre+jä. Moreover, in the analysis of the noun paradigms as categorized in Nykysuomen sanakirja [Dictionary of Present-day Finnish] the käri : käri+n (gen.sg.) : käre+jä type does form a uniform morphological category in the sea of i-final nouns, where as many as 21 categories take their share of the remaining disyllabic i-final nouns (Karlsson 1982). This latter fact makes the set of Finnish disyllabic i-final nouns into an area of claim of the first author and his associates is that Finnish inflectional stem allomorphs behave mostly as units, see Introduction.

268

Chapter Eighteen: Impaired Attention to Phonological Input

intricate morphological associations. Local paradigm density is here used to refer to areas of relatively high amount of competing paradigms. Thus, if paradigm confusion is ever to be predicted for the word types of the present study, it is the käri and simi types with their overlapping paradigms (and schema analogies) that would be the best candidates for incorrect interparadigmatic associations. As for modeling the language of SLI/FLI/GLI speakers on the basis of the present results, it is almost needless to note that any explanatory claim based on memorization of forms (see, e.g., Gopnik and Crago 1991) cannot account for this set of findings in a satisfactory manner. Studying a language with each noun having circa 2 000 and each verb circa 20 000 forms, excluding derivation and compounding, a memorization hypothesis would not have been here the first a priori working hypothesis anyway. And since the speakers had never encountered this set of pseudowords, it is hard to claim that memorizing these items would account for the relatively good morphological performance of the FLI speakers. Similarly, any variant of the surface hypothesis for morphology (e.g., Leonard 1989) has to be enriched at least by high local paradigm density to account for the random use of the two i-final paradigms by the FLI group. More in detail, the surface hypothesis would predict that out of competing affixes the one favored by the FLI would be phonologically the most transparent (e.g., syllabic vs. asyllabic, phonologically more vs. less salient). Thus, forms like simi+ä (si.miä) should not receive such high FLI scores as they do here. As a general statement, most of the present FLI data – perhaps with the exclusion of the i-final paradigms only – show that we are not dealing with any lack (poverty) of morphological competence (or processing) (for similar views on agrammatic aphasia, see Menn and Obler 1990). In most stimulus types the FLI speakers produce morphological forms, although being often schema-analogically improper, that are understandable in terms of morphological productivity and meaning-to-form isomorphy, notions that account for the bulk of non-normative products of, e.g., early stages in language acquisition and in certain aphasic states. What could be a plausible causal mechanism underlying the aberrant morphological performance of the FLI subjects? Our tentative answer is the following: These same subjects have been previously shown to have impaired attention in a dichotic task, and their fMRI scans exhibit atypical neural activity in the cortical areas associated with attentional functions. Moreover, we now observed that they are “immune” to various frequency measures that play a marked role in non-impaired performance. Taking all these three observations together, we hypothesize that the attentional

Jussi Niemi, Alexandre Nikolaev and Kenneth Hugdahl

269

dysfunction is the more basic one of the three, and that it – in a manner that is now left completely open – causes the non-use of the lexical “frequency counters”, which in its turn would account for the atypical morphological performance of the FLI subjects.

Acknowledgments The study has been partially supported by an Academy of Finland grant (project: Genetic Language Impairment in Finnish, 1998-2000, principal researcher: JN) and by a NOS-H grant (project: Functional Magnetic Resonance (fMRI) and Behavioral Studies of Auditory, Phonological and Semantic Processing, 2001-2001, principal researcher: JN), and it is also part of the Canadian Social Sciences and Humanities Research Council MCRI study on the Mental Lexicon (1997-2006), Principal Investigator: prof. Gary Libben (Alberta). We also wish to thank Maarit Silvén and her associates at the University of Turku for collecting the 7-year-old group data for the present study and the staff at the University of Joensuu Teacher Training School for allowing us to collect the 10-year-old group data. Thanks go also to Janne Heikkinen and Juhani Järvikivi for their assistance in the construction of the test stimuli and data collection as well as to Mikko Takkunen for his help in the data analysis.

References Anttila, A. 1997. Variation in Finnish phonology and morphology. PhD Dissertation, Stanford University. Baayen, R.H. 1994. Productivity in language production. Language and Cognitive Processes 9: 447-469. Berko, J. 1958. The child's learning of English morphology. Word 14: 150-177. Bybee, J. 1985. Morphology: A study of the relation between meaning and form. Amsterdam: Benjamins. Chomsky, C. 1969. The acquisition of syntax in children 5 to 10. Cambridge, MA: MIT Press. Dasinger, L. 1997. Issues in the acquisition of Estonian, Finnish, and Hungarian: A crosslinguistic comparison. In D. Slobin (Ed.), The crosslinguistic study of language acquisition, Vol. 4, 1-86. Mahwah, NJ: Lawrence Erlbaum Associates. Dressler, W. 1985. Morphonology. Ann Arbor: Karoma. Gopnik, M. and Crago, M. 1991. Familial aggregation of a developmental language disorder. Cognition 39: 1-50.

270

Chapter Eighteen: Impaired Attention to Phonological Input

Heikkinen, J., Järvikivi, J. and Niemi, J. 2000. NP agreement and government in Familial Language Impairment: Evidence from Finnish. Poster presented at the Second International Conference on the Mental Lexicon, Montreal, October 2000. Hugdahl, K., Thomsen, T., Ersland, L., Rimol, L. and Niemi, J. 2003. The effects of attention on speech perception: An fMRI study. Brain and Language 85: 37-48. Hugdahl, K., Gundersen, H., Bakke, C., Thomsen, T., Rimol, L., Ersland, L. and Niemi, J. 2004. fMRI brain activation in a Finnish family with Specific Language Impairment compared with a normal control group. Journal of Speech, Language, and Hearing Research 47. Hyönä, J., Laine, M. and Niemi, J. 1995. Effects of a word's morphological complexity on readers' eye fixation patterns. In J.M. Findlay, R.W. Kentridge and R. Walker (Eds.), Eye movement research: Mechanisms, processes and applications, 445-452. Elsevier. Järvikivi, J., Bertram, R. and Niemi, J. 2006. Affixal salience and the processing of derivational morphology: The role of suffixal allomorphy. Language and Cognitive Processes 21: 394-431. Järvikivi, J. and Niemi, J. 2002a. Form-based representation in the mental lexicon: Priming with bound stem allomorphs in Finnish. Brain and Language 81: 412-423. Järvikivi, J. and Niemi, J. 2002b. Stem allomorphs in on-line processing. In R. Rapp (Ed.), Linguistics on the way into the Third Millennium, Part 2, 47-58. Frankfurt am Main: Peter Lang. Karlsson, F. 1982. Suomen kielen äänne- ja muotorakenne. Porvoo: WSOY. Karlsson, F. and Koskenniemi, K. 1985. A process model of morphology and lexicon. Folia Linguistica 29: 207-231. Laine, M., Niemi, J., Koivuselkä-Sallinen, P. and Hyönä, J. 1995. Morphological processing of polymorphemic nouns in a highly inflecting language. Cognitive Neuropsychology 12: 477-502. Leonard, L. 1989. Language learnability and specific language impairment in children. Applied Psycholinguistics 10: 179-202. Menn, L. and Obler, L. 1990. Cross-language data and theories of agrammatism. In L. Menn and L. Obler (Eds.), Agrammatic aphasia: A cross-language narrative sourcebook, Vol. 2, 1367-1389. Amsterdam: Benjamins. Nesser, A. 1981. Finska barns språkutveckling: fonologi och morfologi. FUSKIS/FIDUS, 2. Finsk-ugriska institutionen. University of Uppsala. Niemi, J. 1999. Production of grammatical number in Specific Language Impairment. Brain and Language 68: 262-267.

Jussi Niemi, Alexandre Nikolaev and Kenneth Hugdahl

271

Niemi, J., Gundersen, H., Leppäsaari, T. and Hugdahl, K. 2003. Speech lateralization in a Finnish family with specific language impairment. Journal of Clinical and Experimental Neuropsychology 25: 457-464. Niemi, J. & Hägg, M. 1999. Syntax at late stages of acquisition: Experiments with normal and SLI children. In B. Maassen & P. Groenen (Eds.), Pathologies of Speech and Language: Advances in Clinical Phonetics and Linguistics. London: Whurr. Niemi, J., Järvikivi, J. and Heikkinen, J. 1999. Suomen substantiivien monikkotaivutuksen epäsanatesti [Pseudoword plural inflection task of Finnish nouns]. Availability: Linguistics, University of Joensuu. Niemi, J., Laine, M. and Tuominen, J. 1994. Cognitive morphology in Finnish: Foundations of a new model. Language and Cognitive Processes 9: 423-446. Niemi, J. and Niemi, S. 1987. Acquisition of inflectional marking: A case study of Finnish. Nordic Journal of Linguistics 10: 59-89. Niemi, S. 2002. Schemas and competing paradigms in Swedish plural formation. Brain and Language 81: 464-472. Niemi, S. and Niemi, J. 2002. Emergent nature of morphological paradigms: Plural inflection in Swedish and Finnish. In R. Pajusalu and T. Hennoste (Eds.), Tähendusepüüdja/Catcher of the Meaning. A festschrift to Haldur Õim on the occasion of his 60th birthday. Publications from the Department of General Linguistics 3: 285-296. University of Tartu. Nikolaev, A. 2002. Eräiden suomen taivutustyyppien produktiivisuudesta. Puhe ja kieli 22: 113-124. Nippold, M. 1988. (Ed.). Later language development: Ages nine through nineteen. Boston, MA: Little, Brown and Co. Pajunen, A. and Palomäki, U. 1984. Frequency analysis of spoken and written discourse in Finnish, Vol. 1. Publication 30, Research Center for Domestic Languages. Helsinki: Valtion Painatuskeskus. Penttilä, A. 1963. Suomen kielioppi. Porvoo: WSOY. Pyykkönen, P., Niemi, J. and Järvikivi, J. 2003. Sentence structure, temporal order and linearity: Slow emergence of adult-like syntactic performance in Finnish. SKY Journal of Linguistics 16: 113 – 138. Skousen, R. 1989. Analogical modeling of language. Dordrecht: Kluwer. Toivainen, J. 1980. Inflectional affixes used by Finnish-speaking children aged 1-3 years. Helsinki: Suomalaisen Kirjallisuuden Seura. Vannest, J., Bertram, R., Järvikivi, J. and Niemi, J. 2002. Counterintuitive cross-linguistic differences: More morphological computation in English than in Finnish. Journal of Psycholinguistic Research 31: 83-106.

WHAT CAN CHILD LANGUAGE TELL US ABOUT PREPOSITIONS? ALIYAH MORGENSTERN AND MARTINE SEKALI

Abstract. Based on the spatial value of children’s first prepositions in English, parallels have been drawn between the acquisition of prepositions by children and the grammaticalization of prepositions in diachrony (Tyler and Evans 2003). Indeed, studies show that semantically charged prepositions are used by children several months before more functional (or more “grammaticalized”) ones. Yet, other studies stress the fact that ontogeny does not parallel phylogeny (Slobin 2004): the factors determining the acquisition of prepositions would be linguistic rather than cognitive, and linked to language use and frequency of input (Rice 1999). In order to tackle these questions, we conducted a contrastive corpus-based study of French/English acquisition of prepositions. Our findings support the claim that some aspects of children’s discourse are influenced by the particular structure of the language the child is acquiring (Talmy 2003; Hickman and Robert 2006), together with other parameters such as discursive organization and context. Ultimately, “grammatical words” can express and organize social interaction, and are acquired by children thanks to the mediation of adults.

Introduction Prepositions represent a problematic category for theories of syntax. Recent syntactic theory suggests a classification of prepositions according to either lexical or functional features: lexical prepositions contribute semantic content while functional prepositions merely assign case (see Cadiot 1997 for a review). Yet the same preposition can exhibit both features, and its classification will therefore depend on its use (the same preposition to in English can thus be used as a lexical spatial preposition: I’m going to London, or as a functional preposition assigning dative case: Give it to him!).

Aliyah Morgenstern and Martine Sekali

273

Language acquisition, when studied in spontaneous dialogical contexts, may give new insight on prepositions as a grammatical category. Which prepositions are used and acquired by children first? How are they used? In what order do they appear and why? Do French and English children use the same first sets of prepositions? In order to tackle these questions, we analysed the emergence of prepositions in spontaneous verbal interactions between children of age 1;08 to 2;04 and their parents, and have tried to determine the function of their first uses in context. We can observe that prepositions are used by children as soon as they elaborate two word utterances, at the end of the second year. The literature on the acquisition of English insists on the fact that prepositions appear quite early in children’s language: they are part of the first twenty items learnt by English speaking children according to Brown (1973), and are primarily spatial localizers. These results have enabled researchers to compare the development of language in acquisition and the history of languages (Traugot 1992; Lakoff 1987; Ziegeler 1997), since prepositions in the diachrony of the English language were spatial before being functional, which fits perfectly into the grammaticalization process underlined by cognitive linguists. But if this order of emergence in acquisition is true of English, we may wonder whether this is a general process, which is not language dependent. First prepositions in English are semantically rich (they are called “coloured” prepositions as opposed to the “colourless” prepositions, Spang-Hanssen 1963), but are these features specific to English, in which so-called prepositions can be used as adverbials in isolation as in down or up? Is it the case in French? We were interested in testing these claims and comparing first uses in English to first uses in French. We first conducted a quantitative analysis of the emergence of prepositions in two English and two French-speaking children, based on the syntactic opposition between lexical and functional features. We then focused on two French children for a qualitative analysis of their use of prepositions in context.

Quantitative analyses In order to check the claims found in the literature on English data, we ran a morpho-syntactic analyser through Peter’s data (taken from the CHILDES database collected by Lois Bloom, http://childes.psy.cmu.edy). It must be said that the analyser extracts morphemes according to their form and not their function, which, unfortunately, amalgamates prepositions and particles. The results for Peter can be viewed in table 1 showing the percentages of Peter’s first prepositions according to his age.

274

Chapter Nineteen: What can Child Language Tell us about Prepositions?

We notice that the morphemes related to spatial orientation up, on and especially in are quite frequent.

Table 1. Percentage of prepositions per age for Peter (CHILDES database)

The detail of the distribution of Peter’s prepositions according to functional and spatial uses can be observed in Figure 1, showing a vast majority of prepositions related to space.

Figure 1. Peter’s functional and spatial prepositions

Aliyah Morgenstern and Martine Sekali

275

The same analyses conducted on another child, Will, recorded by Katherine Demuth’s team at Brown University, shown in Figure 2 yield similar results, the proportion of “spatial” prepositions being significant as opposed to the proportion of “functional” prepositions.

Figure 2. Will’s functional and spatial prepositions

These results confirm previous research on English data. In order to compare these results to French data, we conducted the same analyses on Léonard and Madeleine, two French children living in Paris. Léonard’s data was collected by the first author before the beginning of the current research and is complete.1 The transcription is aligned with the video thanks to CLAN provided by the CHILDES project supervised by Brian McWhinney (1995).

1 Léonard’s data is a longitudinal follow-up from 1;08 to 3;03. For the purpose of this study, we analyzed his use of prepositions up to 2;04.

276

Chapter Nineteen: What can Child Language Tell us about Prepositions?

Figure 3. Leonard’s functional and spatial prepositions

Leonard’s data, shown in Figure 3, yield significantly different results, with a greater number of functional rather than spatial prepositions. In addition to that, the ‘spatial’ prepositions found mostly correspond to Leonard’s use of dans/(‘in’) in set expressions such as “dans le bain” (‘in the bath’) and in some utterances where the form dans stands for the adverb dedans, which shows us that quantitative analysis on forms counted as prepositions by the morpho-syntactic analyser have their limits. The same analysis was made on Madeleine’s data (Figure 4), which is part of the new French data we are currently recording and transcribing in a large research program financed by the French National Research Agency (Projet Léonard, 2005-2008). Madeleine was filmed by the second author, and we found her to be quite a precocious little girl, her linguistic development being very quick. At 2;02, she already had a large variety of prepositions in her repertoire. The distribution of functional and spatial prepositions in Madeleine’s data is quite similar to Léonard’s, but Madeleine has more types and more tokens at a younger age.

Aliyah Morgenstern and Martine Sekali

277

Figure 4. Madeleine’s functional and spatial prepositions

The results for the four children are quite interesting to compare: the English children confirm previous research and show a predominantly spatial use of first prepositions. But the French children tend to use mostly functional prepositions in their first usage of the category. These findings support the idea that what was presented as a general trend in child language might in fact be specific to English and should be reconsidered as far as other languages are concerned. Zlatev (2003) conducted a similar study of the acquisition of i (‘in) and på (‘on’) in two Swedish children, and found similarly to our results for the French data that functional uses predominate. Furthermore, a finer analysis of the data leads us to add several remarks. - The English-speaking children use prepositions such as in, on, or up very early, but in a “verb-like” manner and often in isolation. Interestingly enough, where the English-speaking children say up! or down! the Frenchspeaking children use verbs such as monter! (‘go up’) or descendre! (‘go down’). - The French-speaking children first use prepositions such as pour, à and de around 1;10 only (François 1977 ; Morgenstern and Sekali 1997). These mostly correspond to the possessive case in English. Where the French child says les jambes de maman, or à moi la poupée! the Englishspeaking child will say the corresponding Mummy’s legs and My doll! Yet Peter’s data shows the presence of possessive adjectives from the very first recordings (my turn) and the genitive morpheme appears as soon as the age of 1;10 (Patsy’s pencil). This remark tends to argue against the

278

Chapter Nineteen: What can Child Language Tell us about Prepositions?

temptation to transfer observations made on English data on the acquisition of French. - Around the age of 2;04, the four children use a larger variety of prepositions, semantically charged as well as more functional ones. We find pour, à, de, dans, sur, avec, en in French. In English at the same age, the variety of prepositions used is even more important, with at, by, for, from, in, of, off, on, out, over, to, under, up, and with. The quantitative analysis of prepositions in our data thus shows significant differences between English and French. Yet it is not sufficient to reveal the type of function that the children assign to this emergent category. Investigating early functions of morphemes can only be done through a qualitative analysis of utterances in context. The qualitative analysis of the French children’s first uses of the “functional” preposition pour in context will now show interesting recurrent patterns, suggesting that the acquisition of grammatical paradigms could be made through a series of transitory subsystems.

Qualitative analyses The in-depth examination of Léonard and Madeleine’s data required considering parallel and interacting linguistic processes in their development. In order to study the emergence of prepositions in the framework of the child’s development, we established an overview of salient features, first syntactic features, and then grammatical markers in parallel with the use of prepositions for each month from 1;08 to 2;04. This overview shows that, at 1;08, Leonard starts using two-word utterances and has his first tries at nominal determination. Pour is the first preposition used by Leonard in our data. It is first associated with a nominal element (pour papa / ‘for daddy’, pour moi / ‘for me’) and is then very quickly used with a verbal element (pour sauter / ‘to jump’, pour dessiner / ‘to draw’). In this context, “pour” is still called a preposition in French grammar, but could be considered as a complementizer introducing a new clause. The following example illustrates how Leonard uses the preposition pour in order to disambiguate a previous utterance. (1)

Father: Les belles saucisses! ‘What beautiful sausages!’ Léonard: Donne! ‘Give!’ Mother: J’te l’ai donnée! ‘I gave it to you’

Aliyah Morgenstern and Martine Sekali

279

Léonard: Pour papa. ‘For Daddy.’ In the situation, Leonard’s parents do not understand who is to be the “recipient” of the sausage; thanks to the use of pour in pour papa, Léonard clarifies the target and thereby makes the argument of the verb ‘give’ (donne) explicit. The preposition is thus used in the context of a misunderstanding with an argumentative value. The same argumentative feature appears in Léonard’s use of pour with infinitives. At the age of 2;02, Léonard is naked on his bed: (2)

Léonard: Maman, où est le petit lit? ‘Mummy, where’s the little bed?’ Mother: Le petit lit? ‘The little bed?’ Léonard: Pour sauter. ‘To jump’ ( = ‘For jumping’)

In this exchange, the “preposition” pour helps Léonard add an argument in order to elaborate his request of a bed. Here again, he is giving his reason for that request when faced with his mother’s incomprehension, explaining that he wants the bed because he wants to jump on it (and use it as a trampoline). Strikingly enough, during this period, whenever Léonard uses [pour+ verb], he is adding arguments in order to justify, explain, clarify the intentions he has previously expressed, but which his co-speaker did not understand: he wants his towel TO hide himself (“pour cacher”), he used chalks in school TO draw (“pour dessiner”). This argumentative function of the preposition pour goes along with its appearance as an emergent category. The French preposition à also figures among Léonard’s early prepositions, but is used with another function, as exemplified below. Léonard (2;02) is in his bath. Arianne, a friend of his parents comes into the bathroom to say goodbye. As soon as he sees her, he splashes and screams: (3)

Léonard: Pas beau. ‘Not you!’ Mother: Non, non Léonard arrête! ‘No Léonard stop that!’ Léonard: C’est à moi le bain! ‘That’s my bath!’

280

Chapter Nineteen: What can Child Language Tell us about Prepositions?

Mother: Mais oui, c’est à toi, ne t’inquiète pas. Ariane elle va pas aller dans ton bain. ‘But of course it’s yours, don’t worry. Ariane is not going into your bath.’ (A little later) Arianne: Et qu’est-ce que je vois sur tes fesses? Je vois quelque chose qui va ravir ‘And what do I see on this bottom? Wow! I see something that …’ Mother: C’est un grain de beauté. ‘It’s a beauty mark.’ Arianne: …qui va ravir les filles un peu plus tard Léonard. ‘…that girls are going to love Léonard…’ Léonard: C’est à moi les fesses! ‘That’s my bottom!’ Ariane: Oui les fesses sont à toi et le grain de beauté est à toi et ça va beaucoup plaire plus tard, beaucoup. ‘Yes the bottom is yours, and the beauty mark is yours, and girls are going to love that, really love it.’ In this dialogue Léonard reacts to Ariane’s intrusions as if she were taking away a beloved possession: not the objects themselves (the bath, the beauty mark), but his freedom to use them as he wishes. The bath is his kingdom, he asserts his freedom to refuse Ariane’s admission to it; he also forbids Ariane to make comments on HIS birth mark and sees her remarks as a violation of his privacy. In the utterances with à moi, the preposition à acts as a polemic re-enforcement of the property of objects described as being in the child’s possession. In all the examples we find in our data, Léonard uses à to assert himself as a sort of stage director, an orchestra conductor. The same uses can be found with à toi meaning ‘your’ or ‘your turn’ according to the context. In fact, the data shows that Léonard uses à and pour in complementary distribution. The following exchange is particularly interesting as to the alternating uses of à and pour: Léonard (2;01) has given Aliyah an imaginary piece of candy. His mother offers him a real piece of candy, and a situation of conflict ensues as to who is to be its recipient.

Aliyah Morgenstern and Martine Sekali

(4)

281

Mother: Il m’en reste plus qu’un! Un seul bonbon c’est pour toi. ‘I only have one left! Only one candy, it’s for you.’ Aliyah: Euh, c’est pour moi le bonbon. C’est pour moi! ‘Hey, the candy is for me, for me! ‘ Léonard: à moi. 'It’s mine’ Aliyah: Ah! 'Oh!’ Léonard: C’est pour moi. 'It’s for me.’ Aliyah: T’es sûr? 'You’re sure?’ Léonard: C’est moi Léonard, c’est à moi! ‘I’m Léonard, it’s mine.’ Léonard: C’est pour moi. ‘It’s for me.’

In this dialogue, the mother intends to give the candy to Léonard (‘it’s FOR you’), but Aliyah wants it too (‘the candy is for me’). Léonard makes an attempt at solving the conflict by alternating the use of pour and à in his discourse. He first uses à (à moi), thus redefining the candy as being his own (this property has been assigned to the candy by his mother and is set as immutable). The functional preposition à here links the candy to its TRUE possessor, Léonard. However, pour is also used immediately afterwards (pour moi), this time in order to solve the conflict raised by Aliyah over the recipient of the candy. The preposition pour is used with an argumentative value so as to redefine the target. After Aliyah’s incredulous reaction: ‘you’re sure?’ Léonard is forced to reassert his own identity (Léonard) in order to reassess the property of the candy (of it being HIS) in order for this legitimate recipient of the candy to be reestablished. The two prepositions therefore have different bearings: à puts the focus on the object, with a polemic redefinition of one of its inherent properties, while pour focuses on the recipient of the object. The syntactic level of analysis proves insufficient to understand the emergence of prepositions at this age; interpersonal relations as well as pragmatic speech-acts must be taken into account in the analysis of spontaneous children’s discourse (as opposed to electronic data). The pragmatic explanatory function of pour in situations of conflict or misunderstanding, is confirmed in Madeleine’s data.

282

Chapter Nineteen: What can Child Language Tell us about Prepositions?

At the age of 1;11, Madeleine uses the preposition pour several times during the same session but in an unconventional way. While she is reading a book with her mother, she is moved by the story of a mother crying because her children have nothing to eat, and wants to give the mother her bunny to console her. Picking up her bunny, she says: (5)

Le doudou pour Madeleine. ‘The bunny for Madeleine.’

Our first reaction to Madeleine’s use of the preposition pour here is that she is using it in the wrong way, or making a mistake, since her use of it seems to be the exact opposite of the function an adult would give to the preposition pour. In adult speech, pour before a noun indicates the destination, or target of an object, in a similar way to the English preposition for. Le doudou pour Madeleine would then mean ‘the bunny for Madeleine.’ Yet in the context of use, it is absolutely obvious that she does not mean that the bunny is ‘for’ her, but that she wants to give her bunny to the mother in the story so as to comfort her. Rather than considering the child’s use of the preposition as a mistake, especially since this particular use is recurrent, it is more interesting to find out what function she ascribes to it, in what may be seen as an “emergent category” (Clark 2001) in her language. Quite striking too is the fact that that particular session (1;11) showed a new surge of possessive determiners: mon/ma, son, sa (‘my’, ‘her’/his’) so that she had the linguistic ability to use the possessive form mon doudou if she had wanted to point at herself as the possessor of the bunny (and she does use it later on in the situation): (6)

Ne veux mon doudou. ‘I want my bunny.’

In fact, it seems that Madeleine’s use of the preposition pour is meant not to designate herself as the recipient or the possessor of the bunny, but rather to point to herself (in the 3rd person) as the agent of an intended action to console the mother, so that her utterance could be paraphrased as: ‘the bunny for Madeleine to console the mother.’ The “preposition” pour is given a particular function by Madeleine when she first uses it: the category of prepositions as it emerges in her speech is used not so much to assign case, but to assign argument structure to an implicit verb, and to explain and justify her intentions. The category thus emerges with a pragmatic and argumentative function in the dialogue

Aliyah Morgenstern and Martine Sekali

283

with her mother: the preposition pour directs the addressee to the agent rather than to the destination of her actions, so as to justify them. The destination of the bunny is then made obvious by a gesture: she puts the bunny on the picture of the mother and says Tiens! (‘Here!’). The same use of the preposition pour occurs a second time in the same session, when Madeleine is eating her lunch. Martine had just been telling her how to use her fork to pick the ham and put it in her mouth, and she suddenly points to the inside of her mouth and says: (7)

La bouche pour Madeleine. ‘The mouth for Madeleine.’

In the context, Madeleine does not mean that the mouth is ‘for’ her: she does not consider her mouth as a separate object given to her, nor does she indicate herself as the possessor of the mouth. The preposition pour here again associates the mouth with its use, so that Madeleine’s utterance can be interpreted as ‘the mouth for Madeleine to eat.’ This recurrent function – which Madeleine, at this age (1.11), assigns to this emergent prepositional marker used before a noun – is thus quite close to Léonard’s use of pour before an infinitive verb, except that Madeleine emphasizes herself as the agent of the process. The two children are acquiring the grammatical form “pour” both to introduce nouns (and they use it as a preposition) and to introduce new predications as in this last example in which “pour” could be considered as a nacent complementizer.

Conclusions On the syntactic level of analysis, prepositions are classified as being either semantically coloured with a primary lexical and spatial value, or semantically weak and marking syntactic function within the prepositional phrase. In this opposition, quantitative observations of early uses of prepositions show that the French and English children studied did not behave in the same way, with a clear priority of spatial first prepositions for the English children while the French children used mainly functional ones first. But a qualitative corpus-based analysis of the actual use of these emerging prepositions in context shows that this purely syntactic opposition is not entirely adequate and that another level of analysis would be more appropriate. In our corpus of French children, the category of prepositions emerges with a pragmatic rather than syntactic function: first

284

Chapter Nineteen: What can Child Language Tell us about Prepositions?

prepositions are used to mark a relation between speakers, objects and the situation of utterance and not just to link parts of speech or phrases within the utterances. In the acquisition of French, this particular grammatical category is first organised as a pragmatic paradigm, with prepositions used as tools for speech acts performed to justify action or to disambiguate intentions or interpersonal positioning. A closer look at English corpora shows that this spatial predominance should perhaps also be reconsidered. It appears that in quantitative morpho-syntactic analysis, the items that are gathered around the term “preposition” are free preposition-like morphemes that do not always have a prepositional syntactic function in child speech. The first items appearing in English: up and down, which Tomasello (1987) considers as “verb like”, could well belong to the category of verbal particles and correspond to truncated phrasal verbs (‘pick me up’, ‘put me down’) used as requests for the addressee’s actions in context. Those particles do not so much mark a spatial relation between objects or people but rather a request for action. In that case, particles would appear before prepositions (and not the opposite). There again, it is interesting to note that syntactically, all these free preposition-like morphemes form a kind of morphological “supercategory” when they emerge, and that the items within this category will be subcategorized as particles and prepositions (and complementizers) according to a functional and pragmatic discriminating feature. First uses by children are therefore interesting to analyse, and show that parallels between language acquisition and language histories may not be so accurate either (Slobin 2004). A better parallel to draw would perhaps be the one between first uses in language acquisition and the core, basic or schematic operations to be found behind those grammatical markers.

Acknowledgments We express our gratitude to the editors of this volume for helpful comments on a previous version of this text.

References Brown, R. 1973. A first language, Cambridge, MA: Harvard University Press.

Aliyah Morgenstern and Martine Sekali

285

Cadiot, P., 1997. Les paramètres de la notion de préposition incolore. Faits de langues – La preposition: une catégorie accessoire? 127-134. Paris: Ophrys. Clark, E. 2001. Emergent categories in first language acquisition. In M. Bowerman and S.C. Levinson (Eds.), Language acquisition and conceptual development, 379-405. Cambridge: Cambridge University Press,. Groussier, M.-L. 1997. Prépositions et primarité du spatial : de l’expression de relations dans l’espace à l’expression de relations nonspatiales, Faits de langues – La préposition : une catégorie accessoire ? 221-234. Paris: PUF. Jackendoff, R. 1996. The architecture of the linguistic-spatial interface. In P. Bloom, M.A. Peterson, L. Nadel and M.F. Garrett (Eds.), Language and space, PAGES. Cambridge MA: MIT Press. Jespersen, O. 1921/1964. Language, its nature, development and origin, New-York: Norton. Hickmann, M. 2003. Children’s discourse. Person, Space and Time across Languages. Cambridge: Cambridge University Press. MacWhinney, B. (1995). The CHILDES project: Tools for analyzing talk. Mahwah, NJ: Lawrence Erlbaum Associates. Morgenstern, A. and Sekali, M. 1997. L’acquisition des premières prépositions chez un enfant francophone, Faits de langues – La préposition : une catégorie accessoire?, 201-210. Paris: PUF. Morgenstern, A. and Sekali, M. 1999. Processus de clarification de la référence dans le dialogue adulte-enfant, in Travaux linguistiques du CERLICO : La référence – 2- Statut et processus, 313-333. Rennes: Presses Universitaires de Rennes. Morgenstern A., 2006, Un JE en construction. Ontogenèse de l’autodésignation chez l’enfant. Paris: Bibliothèque de Faits de langues, Ophrys. Slobin, D. 1997. The origins of grammaticizable notions : Beyond the origin of the individual mind. In D. I. Slobin (Ed.), The crosslinguistic study of language acquisition. Vol 5: Expanding the Contexts, 265323. Mahwah, NJ: Lawrence Erlbaum Associates. —. 2004. From ontogenesis to phylogenesis: What can child language tell us about Language evolution?, In J. Langer, S. T. Parker, and C. Milbrath (Eds.), Biology and Knowledge revisited. From neurogenesis to psychogenesis, 255-285. Mahwah, NJ: Lawrence Erlbaum Associates. Spang-Hanssen, E., 1963. Les prepositions incolores du français moderne, Copenhaguen: G.E.C. Gards Forlag.

286

Chapter Nineteen: What can Child Language Tell us about Prepositions?

Talmy, L., 1983. How language structures space, In H. Pick and L. Acredolo (eds), Spatial orientation: theory, research, and application, 225-282. New-York: Plenum. Tomasello, M. 1987. Learning to use prepositions: a case study. 79-98. Journal of Child Language. —. 1992. First verbs: A case study of early grammatical development. Cambridge: Cambridge University Press. Traugott, E.C. 1982. From propositional to textual and expressive meanings: Some semantic-pragmatic aspects of grammaticalization. In Lehmann, W. and Y. Malkiel (Eds.), Perspectives on Historical Linguistics. 245-271. Amsterdam: Benjamins. Tyler, A. and Vyvyan E. 2003. The semantics of English prepositions: Spatial scenes, embodied meaning and cognition. Cambridge: Cambridge University Press. Ziegeler, D. 1997. Retention in ontogenetic and diachronic grammaticalization, 207-241. Cognitive Linguistics. Zlatev, J. 2003. Beyond cognitive determination: Interactionism in the acquisition of spatial semantics. In J. Leather & J. van Dam (eds.) Ecology of language acquisition, 83-107. Amsterdam: Kluwer Academic Publishers.

ANIMACY AND CANONICAL WORD ORDER: EVIDENCE FROM PROCESSING OF ANAPHORA ESPEN A. LARSEN AND CHRISTER JOHANSSON

Abstract. Previous studies have shown that canonical word order is processed faster than non-canonical word order (Menn 2000; Kaiser and Trueswell 2004). Lyn Frazier’s Garden Path theory, described in the introduction to (Clifton, Frazier and Rayner 1994) states that only one syntactic possibility is tried at a time, and the first option is the simplest structure, in accordance with the Minimal Attachment Principle. In this chapter we ask if semantic factors interact with word order (Larsen 2007). Our hypothesis was that word order is selected from the available grammatical options, and new information is generally introduced early when the functional role of the first constituent can be easily guessed, thus avoiding reparsing by semantic cohesiveness, rather than by syntactic branching. Specifically we tested effects in an implicit anaphora resolution task, where there are always two antecedents: one is human the other is an inanimate object, an animal or another human. We evaluated for effects on both reaction time and accuracy, and the results showed a significant effect of animacy (and/or expectancies from new or given information) on human sentence processing.

Introduction This reaction time experiment investigates the role of semantic cohesion in the Human Sentence Processing Mechanism (HSPM; see Clifton, Frazier and Rayner 1994). Do we use available semantic information early, or do we solely rely on syntactic properties such as branching? Garden path sentences, such as “The horse raced past the barn fell” would suggest that we use mainly syntactic information. However, processing may be affected by fine semantic distinctions. For example the sentence “The car raced past the barn crashed” creates less of a garden path effect; volition, control, animacy and collocation might explain part of why this is so. We used the computer program E-Prime to create a reaction time experiment to measure response times and accuracy using a decision task. We designed small formalized three-sentence episodes, where the first

288

Chapter Twenty: Animacy and Canonical Word Order

sentence was a clarification question (such as “What did the lady eat?”). The second sentence, an answer to the question, came in two word order variants. The final sentence contained a pronoun, and would make a statement that was either congruent or incongruent with the context sentence. The first sentence sets up unambiguous functional roles in the second sentence. The second sentence contains the word order alternation that we investigate. The third sentence provides a task. We also used the third sentence to test for preferences in the “selectional restrictions” (in the context of our sentences; e.g. if the hunter shoots, the deer may die and the hunter may miss) of the verb. Some more examples “She was satisfied” (human subject in previous eat context), or “It was sour” (inanimate, fruit object in previous eat context). This experiment was designed to use a repeated measures variance analysis. We are interested in reaction time differences between two word orders SVO (Subject-Verb-Object) and OVS (Object-Verb-Subject) related to the “animacy” of the objects. The subject was always a human being, so we did not exhaust all possible alternatives, but sentences where for example people are eaten by apples, would be absurd from the start.

Design The experimental stimulus followed the pattern of Table 1 (see below), which gives one example of human subject vs. inanimate object. The first sentence introduces a question, and makes Per the subject. The second sentence presents an answer with either SVO or OVS word order. The third sentence presents a state, or a result, that either follows from the previous two sentences or not. Presentation order was randomized, so that the participants would not be trained in one sequence set. We made sure that the paired sentences would appear in different blocks. There was a pause between the blocks, and each block starts with some implicit training examples. We did not tell the participants that the first few items were for training. The training session had examples from each of the eight factor sequences, in a randomized order. We had more variation in the training samples, as those exemplars were not paired.

Espen A. Larsen and Christer Johansson Table 1. Stimulus for human subject and non-animate object # Clarify Prime Related 1

Expected answer

2

Hva spiste Per? What ate Per? ‘What did Per eat?’ Hva spiste Per?

Per spiste eplet. Per ate apple.def. ‘Per ate the apple.’ Per spiste eplet.

3

Hva spiste Per?

Per spiste eplet.

4

Hva spiste Per?

Per spiste eplet.

Hva spiste Per?

Han var sur.

Nei

5

Det ble mett. It became full.

289

Han ble mett. He became full. Det var surt. It was sour. Han var sur. He was sour.

Nei. No unconventional Ja. Yes conventional Ja Nei

6

Hva spiste Per?

Eplet spiste Per. Apple.def ate Per. ‘The apple ate Per.’ Eplet spiste Per.

Det var surt.

Ja

7

Hva spiste Per?

Eplet spiste Per.

Han ble mett.

Ja

8

Hva spiste Per?

Eplet spiste Per.

Det ble mett.

Nei

The training list consisted of 24 samples, and the trial list of 120 samples. Each sequence was composed of simple sentences, to reduce working memory effects during processing. SOA (Stimulus Onset Asynchrony, i.e. the time with a blank screen before the next stimulus) was set to 50 ms, allowing processing of stimuli, while giving the subjects limited time to think about the test. The presentation order of sequences was randomized, so that the participants would not be able to predict the word order of the context sentence. The three different animacy conditions would be randomly distributed across the experiment. Participants may identify three different groups, but no participant gave feedback that they had done so in debriefing. The design of the sequence set controls for potential differences related to word frequency, word length, and lexical semantics. The structure of the sets enabled us to always compare sentences with identical content; the only difference in the stimulus was the word order (SVO or OVS). This presents us with an opportunity to search for potential effects related to word order and animacy, and any effect we find will be related to this word order difference in the prime/context sentence. However, there could be an effect from the context sentence that affects pronoun resolution in the test sentence. All material in the sequences we compared was the same, even the expected answer from the participant. Every sequence pair

290

Chapter Twenty: Animacy and Canonical Word Order

we compared had the same expected answer, as to whether the test/related sentence presented a natural consequence, or not, based on what had been presented prior to the decision being made. For example, we compare the reaction time on “It died” on the two context sentences SVO “The hunter shot the deer” and OVS “The deer shot the hunter”. Both sentences mean the same in Norwegian, which is a “V-2 language”, as we set up the functional roles with the clarification question. Likewise, for “He missed”, and the two non-consistent sentences “He died” and “It missed”.

Procedure The participants were placed in a chair in front of a computer screen in a soundproof room. First, the experiment instructions were given orally. When the participant started the program, instructions were repeated on the screen, and he/she confirmed having read instructions before starting the actual test. This gave an opportunity to ask clarifying questions, if anything was unclear, before starting the experiment. The program presented one whole sentence at a time. One sequence consists of three stimuli sentences. The clarify-sentence remained on the screen for 1000 ms (milliseconds, i.e. 1 second), then a blank screen for 50ms, followed by the related-sentence, which was presented for another 1000 ms, a blank screen for 50ms, and finally the related-sentence was presented for 1700 ms. A preliminary pilot test had shown that the participants could process the information and make a decision in less than this time. If the subject hesitated too long, there would be a time-out and no recorded reaction time for that sequence; however most subjects had no time-outs. The participant pressed a button for either natural or unnatural continuation, given the sentences of each sequence. The responses were given through an E-Prime Deluxe Response Box.

Results There were 36 participants that took part in the experiment. From these, 27 completed in accordance with task criteria. Reasons for excluding participants were if they mostly used one option, or they pressed the wrong buttons. The participants that fulfilled the task criteria had no missing values, although some could have nearly 50% unexpected answers in one or two of the animacy conditions. They had at least 20 out of 40 expected (“correct”) answers in each category. See Table 2 for the actual accuracies.

Espen A. Larsen and Christer Johansson

291

Outlier values were removed from the data set, by removing all reaction times (RTs) outside 2.5 standard deviations. The outliers were probably not random; the analysis with outliers gave a stronger result in the same direction. The statistical analysis was performed for the dependent variables reaction time (RT) and accuracy (ACC). Reaction times are for the remaining pairs after outliers have been removed for both individual values and for paired differences. The analysis of reaction times only includes the items that were given with the expected answer. We found a strong effect of animacy for the difference in reaction time between SVO and OVS; F(2, 1067)=4.560, p