language as cognitive adaptation: from language evolution to language change

Language as Cognitive Adaptation

Language as Cognitive Adaptation

From Language Evolution to Language Change

Gábor Győri

Pécs

2005

Lingua Franca Csoport

Published by Lingua Franca Csoport, Pécs, Hungary [email protected] www.xanga.com/lingua_franca © Gábor Győri, 2005 Cover art: Remember the Owl?, copperplate engraving by P., 2004 Cover design by Márton Győri ISBN 963 642 049 1 Printed by Agora, Pécs Supervisor: Gyula Riger

For P.

This page is intentionally left blank.

Table of contents

Preface 11

Introduction: Language, cognition and adaptation 13

The evolutionary function of language 13

Explanations of language evolution 14

Structure and function in language evolution 16

The evolution of linguistic capacity 17

The cognitive functioning of language 23

Adaptive change in language 24

The adaptive character of semantic change 26

The organization of the book 29

Part I: Cognitive Adaptation and Language Evolution

1. Studying the prehistory of language 35

1.1 The origin of language and languages 35

1.2 On the emergence of symbol systems 38

1.3 Linguistic capacity and linguistic system:

issues of coevolution 42

2. The emergence of language and the problem of

evolutionary continuity 47

2.1 Structure, capacity and behavior in language evolution 47

2.2 Language as an evolutionary innovation and

the issues of continuity 49

2.3 Continuity and discontinuity in evolution 55

2.4 Genetic aspects of language evolution 57

3. Natural selection, adaptation and the evolution of

language 65

3.1 Some problems of language evolution 65

3.2 The adaptive value of language 68

8 Table of contents

3.3 Problems with the origin of human cognition 80

3.4 Symbolic cognition as the adaptive advantage of language 83

4. The emergence of referential behavior 87

4.1 Reference in biological perspective 87

4.2 Comparing animal communication and human language 89

4.3 Artificial language in apes as an indicator of

cognitive capacities 92

4.4 Biological and cognitive bases of linguistic reference 98

Part II: Semantic Change as Cognitive Adaptation Process

5. The cognitive function of language 109

5.1 What is cognition? 109

5.2 Language as a tool for individual and social cognition 112

5.3 Adapting language to cognition 121

6. Cognitive aspects of semantic change 130

6.1 Toward a cognitive explanation of semantic change 130

6.2 General issues of semantic change 135

6.3 Cognitive processes and the actuation of semantic change 140

6.4 Cognitive factors guiding semantic innovation 145

6.5 Semantic change and semantic knowledge 156

6.6 The cognitive basis of polysemy and its emergence

in the lexicon 158

6.7 Semantic change as adaptation process 168

7. Universal tendencies and linguistic relativity

in semantic change 172

7.1 How regular is semantic change? 172

7.2 The cognitive foundations of universal tendencies

in semantic change 177

7.3 A comparative cognitive analysis of emotion term etyma 182

7.4 Universals of semantic change versus linguistic relativity 195

8. The role of image schemata in semantic change 199

8.1 A note on the relevance of universal tendencies in

semantic change for semantic reconstruction 199

Table of contents 9

8.2 The relevance of image schemata in

the search for universals 202

8.3 The conceptualization of basic oppositions as

revealed through semantic change 206

8.4 Lexicalization processes based on image schemata 213

Conclusion: Language as cognitive adaptation 219

The evolution and adaptive value of symbolic cognition 219

Cognitive adaptation in language 223

References 228

Index 249

Preface The present book is the outcome of research I have been doing in two distinct areas, connected seemingly only by the fact that they are both concerned with language. These areas are the cognitive-biological aspects of language evolution and the cognitive aspects of semantic change. Within the first area I have mainly dealt with the question of how the biologically separate functions of communica-tion and cognition have become combined in the capacity for language and what evolutionary biological mechanisms were responsible for this development. In connection with semantic change I have mainly studied the cognitive processes behind speakers’ synchronic linguistic behavior when they modify conventional meanings and the diachronic linguistic process in which such new meanings become conventionalized.

The point of interest that has always connected these two directions of research for me is the general topic of the cognitive function of language. Based on this unifying theme, the book provides a synthesis of the results that have grown out of my work in these two research fields. These results have led to the realization that language evolution and language change can both be explained as adaptation processes driven by the cognitive function of language. The book is built to a large extent on papers that have been published previously, but here the results of that work have been woven together and organized organically around the topic of language as cognitive adaptation. Thus, the ultimate ground for the synthesis is the claim that language evolved as an adaptation to cognitive challenges in the social interaction with the environment, and consequently language change is an adaptation process driven by the function of language to serve as an adaptive and flexible tool for social cognition.

Here I would like to take the opportunity and express my appreciation to all of those who have helped me in shaping the ideas forwarded in this book. I am deeply indebted to Vilmos Csányi, who inspired me exactly two decades ago to view language from a much

12 Preface

broader perspective than customary within linguistics, and who has shaped my thoughts and views on cognition, evolution and other biological issues in numerous friendly and professional discussions of these topics. I also extend my gratitude to Duane Quiatt for his help and the support of my work and the friendly conversations with valuable suggestions in questions of primatology, evolution and cognition. I would especially like to thank Irén Hegedűs, who has discussed many issues of historical linguistics, language change and etymology with me and who has been an immense help in dealing with the etymological data. My special thanks are due to Zoltán Kövecses, from whose expertise on cognitive semantics and metaphor I have greatly benefited through his most useful suggestions concerning my work, and who also read the second part of this book in the form of a dissertation and provided me with helpful suggestions for improvement. I am grateful to Csaba Pléh for the encouragement I received from him on having read a slightly different version of the whole book in the form of two separate dissertations.

Several other friends and colleagues in these two diverging fields have also read parts of this work, mostly as independent pieces, and offered their advice, suggestions and comments. I would like to extend my thanks to all of them for the help I have received. Among them are Raimo Anttila, Bernard Bichakjian, Eugene H. Casad, A. Charles Catania, René Dirven, Adam Kendon, Edith Moravcsik, Gary B. Palmer, György Szépe, and the late Jan Wind. I would also like to thank András Kertész and Péter Pelyvás, who read an earlier version of the second part of this book, also in the form of a dissertation. My work has significantly benefited from their comments. Needless to say, none of the persons mentioned here are responsible for any errors and erroneous ideas the reader might find in the book, especially since I have not always heeded the advice offered and have stuck to my own views in the combination of evolutionary biology and linguistics.

Introduction: Language, cognition and

adaptation The evolutionary function of language The most conspicuous function of language is the communication of conceptual structures that have been coordinated through speaker-hearer interaction and have thus become conventionalized in a speech community. Therefore, language must have evolved as simultaneously serving cognitive and communicative functions in the social interaction of hominids, and it should be considered the product of both biological and social evolution. In animals, communication has an exclusively social function, which is to manipulate the behavior of others for the benefit of the individual (Dawkins and Krebs 1978; Wiley 1983; Catania 1994: 38). Cognition, on the other hand, is the biological capacity of organisms to construct and operate an internal model of the environment for the sake of appropriate orientation and survival. In proportion to the stable and the changeable components of the environment in which an organism evolved, this internal model is based partly on genetically determined knowledge of the environment and of the necessary fixed behavior therein and partly on individual experience supporting a larger flexibility of behavior as a response to more unpredictable fluctuations of the environment (Bonner 1980: 138; Csányi 1988; Plotkin 1994: 149). In this sense cognition is to a large extent a subjective process in an individual organism. However, it can be made social to the extent to which individually acquired knowledge can be made collective within a group, population, species, society, culture, etc. The sharing of such knowledge is based on information transfer, which can be achieved through different forms and mechanisms of communication. Humans are the species that possesses the most powerful device for sharing knowledge. Our capacity to exchange knowledge by far to the largest extent as compared to other animals is due to the joint cognitive-

14 LANGUAGE AS COGNITIVE ADAPTATION

communicative function of language (Tomasello 2003). Thus, language is the most powerful instrument of social

cognition and must therefore be a cognitive adaptation in the evolutionary biological sense (Jerison 1988). This means that language emerged and evolved because the cognitive capacity manifest in language had an adaptive value due to some kind of selection pressure in hominid evolution. Since the adaptive function of language is the facilitation of cognition, primarily social cognition, specific cognitive adaptive processes must keep language a functioning cognitive-communicative system in a changing natural and sociocultural environment. The processes that keep language up to date with variations of the environment are manifest in language change, mainly its semantic aspects. This book will take the reader from a cognitive adaptationist account of language evolution to the cognitive adaptation processes underlying language change (primarily semantic change), connecting up the two by sketching a unified view of language as cognitive adaptation. Explanations of language evolution In the neo-Darwinian view of evolution the emergence of particular biological traits in organisms is explained through natural selection and adaptation. For a long time such an explanation has been assumed also for the evolution of language. Since the accounts of language origins lay in general within the framework of the theory of natural selection and adaptation, the issue was not so much what evolutionary biological mechanisms were responsible for having shaped the human capacity for language, but rather what the selection pressures were to which evolution reacted by developing language in hominids. The major issues in these discussions were on the one hand the behavioral changes possibly connected to language evolution, like food sharing, hunting and tool use (e.g. Isaac 1976, 1981; Davidson and Noble 1993; Toth and Schick 1993), and on the other hand, the anatomical and physiological changes accompanying language evolution, like those related to encephalization and vocal tract development (e.g. Lieberman 1984; Gibson 1988). Naturally, also the relationship between language evolution and the origin of the human mind and consciousness have been discussed extensively

Introduction 15

and the probable effects of interactions between biological and cultural evolution on language evolution have been examined (Donald 1991).

Recently, however, the issue of evolutionary biological mechanisms has come more and more to the foreground in explaining the emergence of language, and explanations in terms of neutral evolution, preadaptation and exaptation have been offered among others as alternatives to natural selection and adaptation (e.g. Bierwisch 1992; Ragir 2001). At the heart of the controversy about the possible evolutionary mechanisms responsible for language evolution are three basic dichotomies of evolutionary biology: evolutionary continuity vs. discontinuity, gradualism vs. punctuationism, and adaptation vs. exaptation (e.g. Lenneberg 1967; Pinker and Bloom 1990). This shift in approach and interest originates partly in a linguistic problem. The Chomskyan theory of language postulates the existence of an innate universal grammar and claims that the structure of language as determined by this innate grammar is totally arbitrary because there is no biological or logical necessity why it should be as it is (Chomsky 1988). Thus, the Chomskyan point of view leaves no room for a causal, not to mention functional, explanation of the structural features of language.

It is no wonder that a linguistic theory rooted in philosophical rationalism and Cartesianism is inconsistent with the neo-Darwinian view of natural selection and adaptation. But since a serious biological claim is at the heart of these linguistic ideas, namely that universal grammar is genetically determined, a growing pressure has arisen to find an evolutionary explanation for the emergence of such an innate grammar. As a result it is claimed that if no causal-biological explanation can be given to this structure, then it is unlikely that language is an adaptation and the product of natural selection (Bickerton 1990, 1995). This opinion has necessarily lead to a serious controversy in the evolutionary account of language, and the debate over the origin of language structure soon made it inevitable to consider other evolutionary mechanisms.


Structure and function in language evolution There appears to be a necessary correlation between the form of the evolutionary explanation given for language and the fact whether language is approached from a structuralist or a functionalist point of view. The structuralism inherent in the Chomskyan theory makes it inevitably skeptical toward any adaptationist explanation. The structuralist approach is bound to seek an evolutionary explanation of language structure in itself, which, not surprisingly, will lead to a denial of adaptation. The reason for this is that structure in itself, that is, without any consideration of function, can by definition not be adaptive. The exclusion of any functional consideration will necessarily eliminate the possibility of any adaptationist explanation, since the term adaptation implies suitability for some function. However, this latter fact also shows that the functionalist endeavor to prove the adaptiveness of linguistic function is equally tautological because looking at function alone must ultimately lead to an adaptationist explanation. It should be realized that structure and function can only be considered conjointly because a given function is always the function of some definite structure that exerts a specific constraint on the possible functions and a given structure (often even one which is originally a so-called ‘spandrel’ [see Chapter 2]) is most of the time employed in order to fulfill some specific function.

The problem in the case of language arises from the fact that it is a rather complex and difficult task to bring the specific details of language structure in exact correspondence with the functions that we attribute to language. The intricacy of these correspondences caused by various indirect, joint, multiple, etc., effects makes it extremely difficult to examine to what extent particulars of structure can be explained in functional terms and vice versa. In spite of this it should be clear that structural and functional explanations do not only not exclude each other but cannot even be separated because neither structure nor function can be totally accounted for without the other. Only a combined explanation can be valid in which neither is subordinate to the other because structure always constrains function and function always exerts some influence on structure.

In line with the differences between a structuralist and a functionalist approach different evolutionary mechanisms and

Introduction 17

processes have been suggested recently in order to explain how linguistic capacities evolved. The most basic issue in this respect is that of evolutionary continuity vs. discontinuity, that is, whether language evolved in a straight line from other capacities or not. This is usually brought into correspondence with the opposition between gradualism and punctuationism. A continuous straight line evolution is often supposed to have involved gradual change whereas discontinuity is supposed to have arisen through punctuated evolution. Furthermore, it is generally believed that the mechanism underlying a continuous and gradual evolution was adaptation through natural selection. On the other hand, the sudden appearance of a trait (in this case language) and thus the discontinuity with earlier traits can be explained through mutation or the exaptation of an already given trait for a new function. This may manifest itself in punctuated equilibrium in the course of evolution.

However, the correlation between the above processes and mechanisms is not so unambiguous. It is only in the event of mutation that we have a real case of discontinuity. Exaptation does not cause discontinuity in the strict sense, since it is an old structure that is exapted for a new function, and it causes punctuated change as opposed to a gradual one only in the sense that it is a faster process than what is normally regarded as adaptation (cf. Dawkins 1986). Thus, it would be too hasty to say that on the one hand we may have continuity between traits, which is the result of gradual change through adaptation, while on the other punctuated equilibrium causes discontinuity through exaptation. The evolution of linguistic capacity In the case of language the controversy outlined in the above paragraph in a general form concerns the issue of the evolutionary connection between linguistic skills on the one hand and communicative capacities and general cognitive mechanisms on the other. The basic question is whether linguistic phenomena can be derived from such capacities and mechanisms, that is, whether the evolution of language can be explained as a cognitive adaptation for the sharing of knowledge structures or whether language emerged as an autonomous cognitive faculty. Though the cognitive reality of the


latter has always been a controversial point, its existence is widely assumed among linguists simply because it plays an axiomatic role in the theoretical construct of generative linguistics, which for various reasons has dominated the linguistic scene for a long time.

There are many details of language origins that are in one way or other connected to the controversy of the existence or non-existence of an autonomous language faculty. One of the main issues is the problem of natural selection and adaptation. Although from a biological point of view language, being a very complex trait, intuitively appears to be adaptive and the result of natural selection, the assumption of an autonomous language faculty generates a common bias against any selectionist view. Inherent in the claim that language is an autonomous cognitive faculty is the assumption that it is an evolutionary innovation in the hominid lineage and stands totally apart from other communicative systems and general cognitive capacities. Thus, such a claim excludes the possibility of an evolutionary continuity between language and the communicative systems and general cognitive capacities of the common ancestors which we share with our closest primate relatives and it allows no explanation of language emergence as a result of gradual evolution through natural selection from these systems and capacities.

There are two types of questions which need to be addressed when talking about language as a product of evolution. On the one hand our aim should be to provide a plausible description of how this capacity emerged, that is, to give an account of the successive steps of language evolution by listing the anatomical, behavioral, social etc. changes that had occurred during this process. On the other hand such an account will not be complete without an attempt to explain the evolution of language, that is, without trying to answer the question why language appeared in the course of human evolution. This endeavor requires the explication of the factors that started this development and the ones that determined its course.

Beside the double function of language (cognition and communication) mentioned at the beginning of this introduction – which undoubtedly must be considered as one of the major factors – language also exhibits a kind of double nature, which is especially important to consider when studying language origins. This double nature consists in the fact that language comes about only through a simultaneous existence of a symbol system and a biological capacity

Introduction 19

for handling this system. On the one hand, an existing symbol system presupposes some kind of specific capacity in the organism that has created such a system, while on the other, a symbolic or linguistic capacity alone, based on some kind of biological substrate, does not in itself constitute language. Thus, it appears to be a crucial issue of language origins in what way the emergence of language as a biological capacity on the one hand and as a symbol system on the other could have constituted parallel and complimentary processes.

However, since a reconstruction of the emergence of a symbol system is practically impossible, our primary concern should obviously lie with the biological capacity when studying the emergence of language within the framework of human evolution. Logically, such a capacity appears to be a prerequisite for the construction of a symbol system and so it has been argued that a full-blown capacity must have evolved before any kind of symbol system could have come into existence (Bierwisch 1992). But it is by far not clear whether evolution can latently produce a complex capacity like the one underlying language or whether its emergence is a result of some kind of coevolutionary process between a capacity and a symbol system in which a continuous interaction between the two could have provided the driving force of development (Győri 1992).

The evolutionary connections between communication and cognition and between biological capacity and symbol system in the emergence of language are among the most prominent issues when trying to answer the question where language comes from. Maybe the most crucial problem to be solved in connection with them is that of evolutionary continuity versus discontinuity, that is, whether the emergence of the human capacity for language can be related to any capacities that can be found in animals, especially our closest relatives, the great apes (Győri 1995). Of course all the physical characteristics that are involved in the capacity for language can without problems be derived from structures found in animals, first and foremost primates. This is the case with the anatomy of the human vocal tract and the anatomy of the human brain. Though the specific functional characteristics of certain brain areas involved in language might be innovations in the hominid lineage, even hemispheric dominance, which for a long time was considered a unique feature of the human brain, has been proven to exist both anatomically and functionally in other mammals and maybe even


other vertebrates as well. Thus, it is certain that even this human characteristic had precursors in our ancestors.

While nobody doubts the continuity of physical traits that underlie language between humans and their primate ancestors, the issue of continuity at the behavioral and mental levels is rather controversial. If we move from physical characteristics to cognitive capacities, a consensus over common characteristics and common origins between humans and other animals can hardly be upheld. The issues in this case are whether language is descended from the communicative behavior of our primate ancestors and whether there is continuity between the cognitive capacity manifest in language and other cognitive mechanisms that appeared earlier in evolution. Of course it is not the origin of human general cognitive capacities that is controversial but the origin of capacities connected to human rationality and language. Contrary to the obviousness of the continuity of physical traits and even some of the general mental characteristics, the continuity of the cognitive capacities involved in reason and language seems to be a highly controversial issue. For many linguists and philosophers in the rationalist tradition it is an obvious case that language cannot be linked to general cognitive abilities (Piatelli-Palmarini 1989). On the other hand, several linguists working within the functionalist and cognitive paradigms accept the view that the linguistic capacity must have evolved from and is based on capacities that emerged earlier in evolution, especially spatial perceptual mechanisms, like the visual information processing system (Givón 1998: 46; Lakoff 1990: 73).

The controversy stems from the fact that language skills appear to be qualitatively so different from other types of communicative behavior and mental capacities that for many it seems unlikely that these latter could have served as forerunners for the evolution of the former. Also, the lack of rudimentary language skills in non-human primates is often seen as supporting this line of reasoning. However, it must be emphasized here that what we should be looking for are not rudimentary skills but indeed forerunners. Such forerunners are not yet language but are specific capacities that in their original function and form have nothing to do with language since they evolved earlier. Only later did they develop into or gave rise to linguistic capacities and thus in retrospect should by definition be looked upon as forerunners. Claiming that evolutionary continuity

Introduction 21

requires the existence of rudiments of a given trait before a speciation event takes place is a serious confusion of rudimentary traits with real forerunners. Bridging the gap between forerunners and the trait in question on the one hand and bridging the gap between rudimentary forms and full-blown ones on the other should be two distinct issues in evolutionary biology.

An explanation of the development of rudimentary linguistic capacities into full-blown language is an issue that must be treated within the framework of hominid evolution. But before that we should come to grips with the issue which traits served as precursors for linguistic capacities. Which were the traits that transformed into initial rudimentary linguistic capacities after (or, depending on how we date the beginnings of language, simultaneously with) the branching off of the hominid lineage? This is what the issue of continuity is about. From this it also follows that even if language skills turn out to be uniquely human, a search for forerunners of language in our non-human ancestors, or for homologues in our closest primate relatives, is a necessary and perfectly logical enterprise.

Another common misunderstanding arises from a widespread but false belief about evolutionary continuity and discontinuity. In discussions about the origins of human and animal traits, especially in connection with language and cognition, it is often assumed that a capacity which is unique to humans cannot be continuous with traits that appeared earlier in evolution simply because of the qualitative difference that exists between them. However, the existence of unique forms and traits that constitute qualitative differences in a cross-section view of evolution does by no means contradict evolutionary continuity. Evolutionary continuity is even expected to produce qualitative differences, since it is obvious that many of the unique forms and traits that constitute the different species of organisms are results of continuous evolution. As a prime example consider the evolution of limbs in the phylum of Vertebrata, which has lead to a differentiation between fins, legs, wings and arms. These are unique in their forms and even more so in their functions. Thus, a difference in kind between successive traits does not automatically entail discontinuous evolution.

It is necessary to have a clear understanding of the continuity issue when undertaking the task of bridging the gap between the


communicative behavior and cognitive capacities of non-human primates and human language. Above I distinguished between linguistic capacity and symbol system as two sides of the notion of language. Looking at the system side, it appears to be obvious that our knowledge about the structure of primate communication systems and that of present-day natural languages does not help us bridge the gap between the two. We simply do not have the methodology for reconstructing a transitional system or a primordial form of language. But in spite of the impossibility to describe such a rudimentary structure, there is at least a theoretical need to explain the emergence of language as a symbol system. Plausible scenarios in the form of logically consistent hypotheses could help us understand where the human symbolic capacity (as manifest in language) comes from. Such an explanation should minimally provide answers to questions like how such a complex system arose, how it acquired its unique combinatorial properties and hierarchical structure, and whether it acquired its complexity by (gradual) development from simpler forms.

When inquiring into the origin of the biological capacities for language we also have to start out from what we find today in humans and other primates. The real hard data for our starting point are provided by the kind of language capacities that we attribute to ourselves and the communicative and cognitive capacities of other primates as well as the biological substrates of these. Based on this, we can then proceed and make inferences about earlier stages of human linguistic capacities. In making these inferences, we are of course also guided by a given conception of an earliest stage, and our task is to bridge the gap between that stage and the capacities characteristic of modern Homo sapiens. Our conception derives, among others, very indirectly from data gathered from the fossil record and other archeological material on the one hand and considering our closest primate relatives as a kind of model for our common ancestor species on the other. It is on this basis that we can entertain certain assumptions about the cognitive and communicative skills of our ancestors at this starting stage. Based on our knowledge of the comparative neural and vocal tract anatomy and physiology of primates, on our knowledge of evolution in general and of the fossil record, we can make inferences about the forerunners of these capacities in our hominoid ancestors, about their possible primordial

Introduction 23

forms in early and later hominids and about the possible course of their evolution, and maybe also about the evolutionary factors that might have triggered and then influenced such a development. The cognitive functioning of language If language evolved as a cognitive adaptation for the sharing of knowledge structures, then this must reveal itself in the cognitive functioning of language. The general function of cognition in both human and non-human individuals is the acquisition of adaptive knowledge about their environment. This involves the construction and operation of an internal model of this environment for the sake of optimal interaction with, and hence survival in it. However, the environment is made up of an array of conditions, which vary from enduring to less stable and even totally transient ones. As a result, such a model must be able to deal with changes in these conditions that might affect an organism’s interaction and survival, so that it can adjust its behavior accordingly. Therefore, the adaptiveness of an internal model is proportionate to the degree in which its complexity and flexibility match the corresponding features of the environment.

Human cognition is unique only with regard to the fact that its general mechanisms are supplemented by a special device: language. Language is a tool not only for individual cognition, but due to its symbolic nature language enormously enhances the possibilities for social cognition. Even though, the cognitive function of language derives from the general function of cognition, but complemented with social characteristics. Thus, this function is to provide a socially shared cognitive model of the environment and to serve as an adaptive cognitive-communicative tool for the members of a particular speech community in their interaction.

A particular language – as a cognitive model of cultural validity in a human community – will serve its function only if it is properly adapted to the particular natural and sociocultural environment of its speakers. From this it follows that language must also be flexible enough to accommodate any change of cultural relevance in this environment and – given the human cultural and intellectual complexity – also in the perspectives and attitudes the community


collectively takes on it. Thus, language must incorporate a mechanism which can optimally handle its adaptation to these changes.

Also, as already stated earlier, within functionalist and cognitive linguistics it is commonly accepted that the linguistic capacity can not be completely separated from general cognitive mechanisms, since it must have evolved from such capacities, especially the spatial perceptual mechanisms, like the visual information processing system. Thus, it should be not surprising that the linguistic mechanism which handles adaptation to environmental changes involves underlying general cognitive mechanisms. This is most obvious in the categorization function of language. The continuous adaptation of language to model the changing conditions of and social attitudes to the particular environment in which it is used happens through semantic-lexical change. This is the historical linguistic process in which words acquire new meanings in the course of historical time. However, such changes may sometimes also occur within the lifetime of one generation.

Thus, language is only seemingly a stable and unchanging system of communication. Viewed from a historical perspective, language is always in motion and changes are continuously going on in all parts of the system. Compared to other parts, the lexicon of a language appears to be the most unstable in this respect. Every language, at any given point in its historical development, codes a relatively well-defined and finite system of culturally significant conceptual categories (cf. Tomasello 2002: 180-181). This culturally valid category system is inherited across generations via language. Thus, Anderson’s (1988: 93) claim, made from the perspective of cognitive psychology, that language stabilizes concept structure against fragmentation appears to be valid in this special historical sense, too. Adaptive change in language It must be added, however, that this well-definedness and finiteness of linguistically coded cultural categories is only theoretically true, since the category system coded in the lexicon of a language can never be captured in a completely motionless state. New expressions

Introduction 25

(words) continuously emerge in the lexicon and existing expressions tend to acquire new meanings giving rise through this to new conceptual categories that are shared by a whole speech community and have thus cultural validity. According to Rosch (1978: 28), as already indicated above, the formation of particular categories in a culture is the result of the operation of two specific psychological principles of categorization – cognitive economy and perceived world structure. If these principles are to be psychologically real, they must in practice be operating in the minds of language users. Thus, Rosch seems to be reasoning along the lines of cognitive psychology (like Anderson, see above) when she claims that the formation of culturally valid categories happens through their coding by the language of the given culture. In fact, one of the major linguistic mechanisms for the temporary coding of newly arising conceptual categories is (context-bound) semantic extension in actual language use. However, for a conceptual category to reach complete cultural validity, the original (ad hoc) semantic extension (which codes the category) must move on from context dependent usage and interpretation to semantic change in order to become a conventional expression in the language system (cf. Paul 1920: 84). Therefore, semantic change appears to be one of the most important linguistic mechanisms for the coding of conceptual categories the validity of which reaches cultural dimensions.

Thus, Rosch’s psychological principles of categorization cannot be operating directly in category formation at the cultural level, i.e., in semantic change. Cognitive processes go on in the minds of individuals, but semantic change is a historical linguistic process taking place over relatively long periods of time in the language system, i.e., in the semantic structure of a particular language as such. Therefore, Rosch’s principles can have only a theoretical explanatory value for cultural category formation, and hence semantic change, which is based on their role in actual language use, or in other words, the concrete coding behavior of speakers. This theoretical explanatory value, however, is of great importance in an elucidation of the role of semantic change in the cognitive functioning of language. Anttila (1989: 153) has stressed that language change – including changes in meaning – is the inevitable result of the use of language in a speech community, and Winters (1992: 508-509) has pointed out that it is synchronic linguistic


activity that is ultimately responsible for diachronic changes. In accordance with these insights, Croft (2000: 8) states that language change is the result of two distinct mechanisms operating at different levels: the “mechanisms for innovation” and the “mechanisms for propagation.” The mechanisms for innovation operate in synchronic linguistic activity, i.e., in the actual language use of speakers, while the mechanisms for propagation are responsible for the spread of any innovative usage in the speech community and ultimately for a particular change to occur in the language system.

The connection between the above two mechanisms does not only involve the traditional linguistic, pragmatic and sociocultural aspects mentioned at the beginning of this introduction. Although the alteration of usage in synchronic linguistic activity, invoked as the source of any diachronic change, is certainly a pragmatic factor, actual usage of language is based on a definite cognitive activity in the interlocutors’ minds. The pragmatic explanation of semantic change as the result of the conventionalization of context dependent modification of usage must therefore be supplemented by investigating the general cognitive mechanisms that a change in usage relies on. Thus, Rosch’s psychological principles must actually operate in specific cognitive processes underlying innovative usage because cognitive factors provide the ultimate material underpinning for the pragmatic factors. Because of this, a comprehensive explanation of semantic change cannot do without taking into account the cognitive aspects of innovative usage. The adaptive character of semantic change Language change can primarily be explained in adaptive functional terms. Geeraerts (1983a, 1997: 102-108) has identified two basic principles of novel usage governing the change in speaking habits due to communicative needs: expressivity and efficiency. These principles complement rather than compete with each other, since “expressivity is always the primary cause of change, whereas efficiency involves the choice of the linguistic means realizing the expressive intention” (Geeraerts 1999: 105). The context dependent temporary semantic modifications of conventional expressions to which the interlocutors resort for the sake of immediate

Introduction 27

expressiveness in their communicative interactions are based on various cognitive mechanisms in the form of novel metaphor, metonymy, or meaning extension/restriction. Thus, changes in meaning ultimately originate at the cognitive level because the purpose of modifying the conventional meaning is to adapt the language to temporary cognitive-communicative needs. These needs arise in cases where conventional expressions are judged unsuitable for conveying an idea at hand because no available expression seems to match the speaker’s momentary conceptualization of some aspect of reality. Such an innovative but context-dependent spontaneous usage of an expression will temporarily become coded in the language in the form of a non-conventional expression (cf. Tomasello 2002: 168). Such expressions are occasional wordings with a figurative meaning which would normally require quotation marks in writing to indicate their unusualness and to draw the reader’s attention to the fact that the conventional meaning has been slightly altered. Although most of these fade away shortly, some will spread and find their way into the system of the language.

The cognitive-communicative aspect of change has a strong sociocultural background. Although the source of actuation of semantic change is the cognitive activity underlying the synchronic linguistic behavior of the individual speaker, the answer to the question why change happens at all can be found by looking into the sociocultural aspects of change. No change in meaning happens in a vacuum but always in a sociocultural context, and it reflects the changes – and may even contribute to them – in the sociocultural environment (Anttila 1989: 152). In this respect the real causes for any change in meaning derive from extralinguistic factors. When changes occur in the sociocultural setting of the speakers, the linguistic category system reflecting this setting must get adapted to the new circumstances. This is required by the categorizing function of language. From the semantic innovations – in the form of novel expressions – to which speakers resort in order to cope with the arising cognitive-communicative challenges, only the ones that prove to be of cultural relevance will spread and contribute to the alteration of the semantic structure of the language.

The strong connection between the cognitive and sociocultural aspects of semantic change has an adaptive functional reason. The speakers’ realistic cognition of their natural, social, cultural, etc.


environments – and reflecting upon them in language in appropriate ways – serves their social and cultural success to a great extent. Even meanings that appear to be about external reality per se are not separable from the speakers’ sociocultural awareness. A basic premise of cognitive linguistics is that linguistic communication is not about the world as such but about the way we see it. On this basis, a cognitive-functional explanation of language change has to be a sociocultural one at the same time, since communication, as a function of language, is a sociocultural activity.

However, language change has not only been explained as induced by function but also by fashion to a large extent. But Vennemann (1993) points out that there is no polar opposition between the two: they overlap because fashion change is functionally motivated. I think that this functional motivation of fashion changes can very well be explained on the basis of some of the communicative maxims Keller (1985) has identified as playing a crucial role in language change. The semantic modification (innovative usage) of conventional expressions elicited by fashion may for instance be due to maxims like “Speak in such a way that people pay attention to you” or “Speak amusingly, wittily, charmingly, etc.,” whereas the maxim “Speak like the others of the group you feel you belong to” may facilitate the spread of such innovations. It is quite evident that the types of linguistic behavior referred to in these maxims are elicited not only by fashion but at the same time by their sociocultural – or even adaptive ethological – functions, like greater popularity in the group, better integration in the group, higher position in the hierarchy of the group, etc.

The cognitive function of language is manifest in semantic change at two interconnected levels. The first level is the level of innovative usage and its cognitive underpinnings, in which semantic change originates. The second level is the level at which semantic change shapes the conceptual structure of a language. Many authors have pointed out – from linguistic, sociocultural, and cognitive perspectives – that the interconnection of these two levels is manifest in the fact that processes at the first level (innovative usage) provide the input to the modifications at the second level (semantic changes) (e.g. Antilla 1989; Croft 2000; Geeraerts 1997; Keller 1989; Milroy 1993; Sweetser 1990; Traugott 1999a; Winters 1992; etc.). However, the reverse side of this interconnection has not

Introduction 29

been dealt with, although Anttila (1989) alludes to this issue several times. This consists in the fact that the output of the second level furnishes the material on which the processes of the first level operate, which, in my opinion, is equally important – especially with regard to the cognitive functioning of language. The adaptive character of semantic change can be explained in terms of the cognitive significance of the circularity of this interconnection.

I will explicate my views on the adaptive character of semantic change within the framework of cognitive semantics. At the basis of my approach is the axiom that meaning relies on human understanding, i.e., meaning originates in the human interpretation of reality based on a non-objectivist mental elaboration of our physical experience (Johnson 1987). In the cognitive semantics framework human understanding – or making sense of the world – involves conceptual mappings from familiar domains of experience to unfamiliar or less well-understood domains. I claim that this cognitive activity lies at the basis of the adaptive character of semantic change and is manifest in the change at the two levels mentioned above. At the first level, in innovative usage speakers rely on the analogical character of the human mind for the sake of efficient communication. In their choice of a conventional expression for semantic modification they utilize familiar knowledge in the form of conceptual mappings, which provide the basis for the interpretation of occasion-bound meanings, i.e., a kind of meaning construction (cf. Fauconnier 1994, 1997). The organization of the book The book is divided into two parts according to the role of cognitive adaptation in language evolution and language change respectively. Part I deals with cognitive adaptation as an evolutionary biological process leading to the emergence of language as a social cognitive-communicative device. Part II discusses how the social cognitive-communicative function of language is continuously maintained by a linguistic-cognitive adaptation process in language change.

After the Introduction, Chapter 1 investigates problems in the study of linguistic prehistory. Its major aim is to differentiate a phase of prelinguistic evolution in the emergence of hominids from the


full-blown linguistic stage reached by anatomically modern Homo sapiens and to characterize the transitional period between the two. The first phase is that of cognitive adaptation resulting in the emergence of language, while the second is that of cognitive adaptation flexibly shaping language to function appropriately as a cognitive-communicative device.

Chapter 2 deals with the continuity problem in the areas of physical traits, cognitive capacities and communicative behavior, and looks at the possible rate and curve of language evolution from these perspectives. Its major claim is that even if human evolution, and within that the evolution of language, might not have been completely gradual but punctuated at times, this does not speak against language as an adaptive trait. Punctuated evolution can be the result of enhanced selection pressure and evolutionary continuity must exist in all the three areas mentioned above.

In Chapter 3 I take a look at the adaptive value of language and claim that language is a product of natural selection. Language must have emerged as a cognitive adaptation in the social interaction of evolving hominids for sharing and representing individually acquired knowledge and thus creating a common mental model for the benefit of the whole group.

Chapter 4 describes the referential function of language from an evolutionary perspective, relating it to communicative and cognitive capacities found in other primates. I view reference as the major function of language in which the joint communicative and cognitive character of language is manifest.

Part II starts with Chapter 5, which provides a transition from adaptation processes of an evolutionary biological nature, having lead to the emergence of language as cognitive adaptation, to adaptation processes taking place in language, maintaining it as a functioning cognitive-communicative device. As an outcome of the discussion in Part I, here I sketch a functional theory of cognition, which holds that cognition is a biologically based process by which an individual acquires knowledge of the environment for the sake of adaptive orientation in it. In this chapter I will explicate the relationship of language and cognition on the basis of this view and will look into the special role language plays in cognition for human beings. I will also investigate how language – being a social instrument – contributes to social cognition. These considerations

Introduction 31

will serve as the general foundation for the subsequent discussion of semantic change as a cognitive adaptation process in language.

Chapter 6 will focus on the interaction of the two adaptive levels of semantic change: (i) the cognitive processes in innovative usage adapting the language to unconventional situations and perspectives through ad hoc semantic modification and (ii) the mechanism of the spreading of innovations as an adaptive process of social cognition in which semantic change shapes the category structure of language in order to adapt it to changing environmental conditions.

Chapters 7 and 8 will deal with the universal and relativistic phenomena in cultural category formation, which phenomena are predicted by a functional theory of cognition. In Chapter 7 I will provide an explanation for universal linguistic mechanisms of semantic change in terms of specific cognitive processes employed in understanding and categorizing the world. I will also discuss the phenomenon of universal conceptual avenues of semantic change, which is more puzzling than the universality of the mechanisms, and will examine various cognitive factors that may be responsible for inducing universal conceptualizations. However, the analysis will also show that the universal conceptualizations attested through various universal tendencies in semantic change are balanced by relativity effects. The explanation of how and why this phenomenon occurs can be found in the cognitive function of semantic change.

Chapter 8 will deal with the resemblances in the conceptual content of semantic change from a more specific point of view. I will look at different levels of specificity, i.e., levels exhibiting a resemblance in more or less general aspects of the content of the conceptualizations, and analyze in more detail the level at which resemblances in content (i.e., conceptual universals) most naturally occur: the level at which metaphorical and metonymical projections from image schemata operate and thus influence particular semantic extensions.

In the Conclusion, by summarizing the main ideas of the book, I will give a concise outline of a unified view of language as a cognitive adaptation and offer an integrated account of how language has evolved as an adaptive social cognitive device and how it functions as such.

Part I:

Cognitive Adaptation

and

Language Evolution

Chapter 1

Studying the prehistory of language 1.1 The origin of language and languages Probably one of the best mottoes for studying the prehistory of language is a claim made by Charles Hockett more than three decades ago:

Thus, just as surely as all living organisms in the world today trace their ancestry back all the way to the beginnings of life itself, so also all the thousands of languages of the ethnographic present trace back, with no break in continuity, to the earliest language-like behavior of our incipiently human ancestors. To push the question to the extreme and ask where the first linguistic forms came from is simply to ask about the beginnings of language. (Hockett 1973: 120)

The main merit of this claim is its emphasis on continuity, and by this carrying the conviction that language is a product of evolution. However, the above quote also presents certain problems. Hockett speaks about linguistic forms that go back to the earliest language-like behavior. While this claim may hold in itself and serve as a theoretical basis for any sound study of linguistic prehistory, its practical implications may have consequences that create special difficulties for research. These difficulties arise from the fact that language (as a complex phenomenon) can minimally be divided into a biological-mental capacity and a symbol system, which interact in linguistic behavior (see Figure 1). I will call this state of affairs the double nature of language (Győri 1995a). Although the study of language in prehistory is naturally supposed to comprise both aspects, most of the time it appears useful to keep these two areas of study apart. Thus, when talking about linguistic prehistory, there are


two main issues to be dealt with: the evolution of the human biological capacity for language on the one hand and the origin and emergence of natural languages on the other. Throughout this chapter I will be concerned with how to accommodate these two aspects of language in a study of origins through showing how they relate to each other within a more complex framework, but I will also point to theoretical and methodological reasons for distinguishing them.

LANGUAGE

biological endowments mental capacities

| |

speech organs abstraction capacities

brain lateralization combinatorial capacities

speech center

symbolic system

|

symbol inventory

combinatorial rules

linguistic behavior

(the manifestation of biologically based mental capacities in the usage of symbols and combinatorial rules)

Figure 1: The double nature of language

Generally it can be stated that a historical investigation is the proper procedure when trying to find out about the origin of the object of study. For instance, studying the history of life entails also raising the question of the origin of life. Pursuing a similar objective

Studying language in prehistory 37

in the case of language, we will soon find out that the distinction between the biological capacity for language and language as a symbol system is crucial for our purposes. If we try to trace them back simultaneously it will turn out that our methods of inquiry will specify two different time depths for these two aspects, corresponding to a linguistic and prelinguistic period (see Figure 2). Our possibilities of tracing the history of language as a symbol system confine us to a period that is too recent for an evolutionary study of language as a biological capacity. At the same time, it appears to be rather controversial whether the difference in time depths follows necessarily from the nature of the matter, or is purely a consequence of our methodological competence. I will take up this issue later.

Figure 2: Hybrid schematic family tree of human evolution and the emergence of languages


If we approach the complex question of linguistic evolution with all the above in mind, two basic approaches lend themselves for practical research. Trying to track down the history of language as a symbol system, we can take the languages of today and start going back in time using comparative reconstruction, which in the course of the more than 150 years of historical linguistics has developed itself into a sophisticated method.

In the case of inquiring into the history of language as a biological capacity, too, we can only start out from what we know about the present state of this capacity. Based on our knowledge of the comparative anatomy and physiology of primates, our knowledge of evolution in general and the fossil record we can make inferences about the forerunners of these capacities in our hominoid ancestors, their possible primordial forms in early and later hominids and the possible course of their evolution and maybe also about the factors that might have triggered and then influenced such a development.

The question arises if the results of these two directions of inquiry converge upon each other, i.e., if these two approaches meet some-where in time and connect up the prelinguistic period with the linguistic one. In other words, can we hope to furnish a complex and coherent explanation (if only a theoretical one) for the transition period in which the biological evolution of the capacity for language reaches a state that can account for the emergence of a structurally full-blown linguistic symbol system? Can our investigations of biological evolution be smoothly continued and their results utilized by historical linguistics to explain the evolution and divergence of the different languages? Based on a solid commitment to continuity of the sort emphasized by Hockett, I hold the conviction that at least in theory there must be meeting points of these two areas. However, there are many problems involved in attempting to answer the above questions. 1.2 On the emergence of symbol systems While the biological capacity for language is ‘real’ in the sense that it is not an abstraction and can be studied as such, language as a symbol system is nothing more than an abstraction from particular symbol systems. Thus, the object of investigation in practical


research can only be the different instantiations of language as a symbol system, i.e., real languages. If we set out to find out about the history of particular languages until we get down to the point of their actual emergence, there are at least two difficulties involved. The first difficulty concerns the applicability of the comparative method beyond a certain time depth. Standard historical linguistics stops at a time depth to which the ages of the so-called reconstructed protolanguages correspond, i.e., at about 7-11.000 years before present. This is the time interval in which the ancestor languages of the established linguistic macrofamilies have been dated.

In my opinion, the only sound reason for stopping at this time depth is a methodological one. We have to see clearly that this boundary is an artificial one that historical linguistics has set for itself. This self-imposed limit is not a result of methodological considerations in the first place (they only came later) but simply a tradition, since this is the time depth that was reached relatively easily and early in the course of the emergence of historical linguistics as a discipline (mainly in the study of the Indo-European languages). The methodological doubts advanced later make the objection that multiple level reconstruction always gets more and more uncertain. Though this is of course true in general, there is nothing specific that compels us to say that a two-level reconstruction can still be trusted while a three-level one cannot or that a three-level one can and a four-level one cannot. Who is to say where the crucial dividing line is? Denying that with accurate and meticulous work multiple level reconstruction can be made more and more precise, refined and sophisticated is unnecessarily hampering the progress of science. Any scepticism would be valid only if it could be shown that any next level of reconstruction by necessity has to involve a qualitative jump in the methodology and not only a quantitative one.

Another argument against multiple level reconstruction is not simply based on the uncertainty involved but claims that the methodology as such is not applicable at such time depths. Here again we have to do with the problem of qualitative/quantitative difference, since the methodology can only cease to be applicable if the material we want to apply it to is qualitatively different from the material it was worked out for. However, within a period of the same structural properties also the same principles of change and


development must have been at work. Because of this, the methodology worked out and applicable for more recent stages of the evolution of languages must be able to capture any change within the whole structurally uniform period. The argument for the non-applicability of the comparative methodology to these time depths is only valid if it can be proven that the languages of those times were structurally significantly different from today's ones. In other words, the methodology should theoretically be applicable to the time depth (point in time) of the emergence of structurally full-blown languages. The linguistic criterion for full-blown language (as a symbol system) is that it be structurally comparable to today's languages.

How are we to assess the time depth in which we can still speak of structurally full-blown languages? It seems to be logical to postulate that the actual emergence of structurally full-blown languages (symbol systems) must have coincided with the emergence of the capacity to speak a structurally full-blown language. Since, by definition, it is only anatomically modern man that possesses the biological prerequisites for speech such as a modern vocal tract and a modern brain with the appropriate neural organization, the time depth of the existence of structurally full-blown languages must have coincided with the emergence of Homo sapiens sapiens (anatomically modern Homo sapiens). Thus, the limits of historical linguistic reconstruction (admittedly taken to the extremes) are theoretically relative to the time depth of the emergence of Homo sapiens sapiens unless we have firm reasons to believe that there has been a considerable evolution in the structural properties of the languages extant within the period that is marked by the existence of anatomically modern man.

Because of this, when pondering the problem of a full-blown language, we have to ask ourselves if there has been any evolution after its emergence? It is undeniable that a kind of social adaptation is constantly going on in a linguistic system, which can be described as evolution. One such process is the continual adjustment of the system to communicative and cognitive needs through the emergence of culturally coded categories (Győri 1994, 1996). Another one, according to Bichakjian (1988), is a continuous evolution involving structural changes in a well definable direction from later to ever earlier acquired structures (in an ontogenetic


sense). This modification, however, does not proceed from less developed structures to more developed ones and because of this does not affect the issue of the applicability of the comparative method.

Theoretically, also the following question might have some bearing on the properties of symbol systems: Has there been any evolution since the emergence of modern man that involves the capacity for using such a system? In principle, any evolution in capacity might have engendered structurally important changes in the system. The best answer we can give to this question is probably that even if the human capacity for language has kept on evolving after it had reached the stage that enables the use of a linguistic symbol system, the time span is so short that the changes that might have occurred since then cannot be measured in evolutionary steps, i.e., the structural properties of the symbol system could not be affected by it.

Let us return now to the problem of the applicability of the comparative method and of the plausibility of the resulting reconstructions at extreme time depths. Though it is still quite unorthodox in historical linguistics, there are attempts to go beyond the generally accepted protolinguistic stages of the established language families. This is called deep reconstruction or paleo-linguistics. According to such approaches, reliable linguistic reconstructions can be made up to the time depth of about 15-20 thousand years before present. These limits are still rather recent compared to the time of the emergence of modern man and though it might be theoretically possible to expand them (as I tried to show above), any practical realization of this is very difficult. While such a limitation might not be of direct relevance to a theory of language evolution, another problem involved in this issue is much more so. This has to do with the assumption of (an) earliest language(s) of modern man. It should be reasonable to claim that any ultimate time depth of such (a) language(s) can only have a virtual status. There must have been a transitional period from a primordial linguistic stage to a full-blown language with a fuzzy boundary between the two. The beginnings of (a) structurally full-blown language(s) are definable only in principle because such an estimate can only be given relative to the time and circumstances of the emergence of anatomically modern Homo sapiens. Thus, the estimation of the time


depth of the earliest human language(s) hinges on the results of other disciplines, such as archeology and paleoanthropology.

At this point inevitably also the question of monogenesis arises. When raising the issue of monogenesis, we should clearly keep apart the problem of monogenesis of language and that of monogenesis of languages. This distinction is once again based on the contrast between the evolution of the biological capacity for language and that of language as a symbol system. The framework for addressing the first problem is hominid evolution, and the question to be answered is whether the capacity for language emerged in only one lineage or not. On the other hand, the question of the monogenesis of languages clearly has to be addressed in connection with the problem of the monocentric or polycentric emergence of anatomically modern Homo sapiens. Practically, this latter issue is whether the number of the principle protolanguages (of the established language families) can be reduced essentially to one ancestral human language, which could be considered the first language of anatomically modern Homo sapiens. 1.3 Linguistic capacity and linguistic system: issues of coevolution Parallel to the general distinction between language as a biological capacity and language as a symbol system, the tracing of language in prehistory thus entails the distinction of the question of language origins and that of the origin of languages. Although at present these seem to be two totally different questions, it is not excluded that we might find meeting points between these two issues. It is mainly the tools that are at our disposal that compel us to keep these two aspects apart and relate them to different periods of prehistory. The only means that we have for investigating past forms of language as a symbol system is historical linguistic reconstruction. As we have seen, this will take us at best (i.e., in theory only) back to the period of the emergence of Homo sapiens sapiens. However, it would also be of theoretical interest if we could give an account of the symbol system before the full-blown stage. Here again we encounter a problem that is due to the double nature of language. While we can apply the same biological knowledge to investigate all stages of the


evolution of the biological capacity, no primordial form of a symbol system can be subjected to a linguistic analysis. The methods of linguistic analysis apply only to aspects of a full-blown symbol system, e.g. combinatorial properties, hierarchical structure, and structurally dependent mechanisms of internal processes. At best we could subject it to some kind of semiotic analysis like any other sign system, but it is exactly because of this state of affairs that any attempt at the reconstruction of such a primordial language is futile.

It might be also in place here to clarify a terminological confusion in the case of the term ‘protolanguage.’ Though the term indeed suggests some kind of primordial linguistic system, it has long been reserved by historical linguistics for a reconstructed common ancestor of a group of languages. In their glossary Jeffers and Lehiste (1979: 183) explain protolanguage as the “[t]erm used to refer to the earliest form of a language established by means of the comparative method of reconstruction.” It is of course common that different disciplines use the same term with totally different meanings and there is no danger of confusion (e.g. the term ‘morphology’ in linguistics and biology respectively). In the case of ‘protolanguage’ however, the possibility of misunderstanding is real, since it is used in both of its senses to denote stages in the prehistory of language. It is crucial to distinguish clearly between primordial forms of a symbol system and earlier stages within the full-blown linguistic period. In the case of the latter, the consensus in linguistic theory is that the languages of that period (including both past and presently extant languages) cannot be arranged on a scale of being less or more developed. Despite the fact that all languages change constantly, the processes in them are not processes of improvement (nor of deterioration). Thus, a protolanguage in the primary historical linguistic sense of the word does not denote a structurally inferior linguistic system.

Since language exhibits the kind of double nature we are concerned with here, a theory of the emergence of language has to account for it. Thus, in spite of the impossibility to reconstruct a primordial form of language, there is at least a theoretical need to explain the emergence of language as a symbol system beside the biological capacity. The questions that such an explanation should minimally attempt to answer are the following: How did such a complex system arise? How did it acquire its combinatorial


properties and hierarchical structure? Did it acquire its complexity by gradual development from simpler forms or as a result of some sudden evolutionary jump? Though these questions concern language as a symbols system, they are definitely issues of language origins, even if it appears largely impossible to address them in effect. However, there is one question that must be addressed at least at the theoretical level. This is whether the emergence of language as a biological capacity and that of language as a symbol system constituted parallel and complimentary processes or not. Thus, at least in theory the capacity issue and the symbol system issue do not automatically and naturally pertain to two different time depths (and different questions) of language evolution.

When inquiring into the origin of the biological capacities for language we also have to start out from the linguistic state of today. The real hard data for our starting point are provided by the kind of language capacities that we attribute to ourselves. Based on this, we can then proceed and make inferences about earlier stages of these capacities. In making these inferences, we are of course also guided by a given conception of an earliest stage and our task is to bridge the gap between that stage and the capacities characteristic of anatomically modern Homo sapiens. Our conception derives, among others, very indirectly from data gathered from the fossil record on the one hand and considering our closest relative, the chimpanzee, as a kind of model of our common ancestor species on the other. It is on this basis that we can entertain certain assumptions about the cognitive and communicative skills of our ancestors at this starting stage. However, especially among linguists, psychologists, philoso-phers and cognitive scientists the view is common, that such skills could not have served as forerunners for the evolution of the human linguistic capacity since they are incompatible with many (if not all) properties of the latter. As it will be shown in Chapter 2, a difference in kind between successive traits does not automatically entail discontinuous evolution and, what is more, evolutionary continuity is even expected to produce qualitative differences.

It is very often not realized that not only our assumptions of an initial stage of language evolution necessarily contain hypothetical elements, but also many of the features of our present linguistic capacity are assumed ones. What I particularly have in mind are assumptions that derive from inferences that we make about the


biological capacity based on our analyses of language as a symbol system. While on the whole it is a valid procedure to draw such inferences, we should bear in mind that many issues in the linguistic theorizing of our days that might have a connection to questions of language origins are far from being settled. But this may only be the lesser wrong. Far more serious problems are posed by the philosophical commitments (often verging on dualism) inherent in many theories about language. Unfortunately, there are several explanations that are pure assumptions standing simply by virtue of their coherence, but widespread enough to appear as undisputed cases. What I have in mind are especially various Cartesian and rationalistic views in generative linguistics and cognitive science (e.g. Chomsky 1985; Fodor 1987; Katz 1990).

The general and most important issue in an account of language origins is whether the capacity for language can be explained on the foundations of general cognitive mechanisms or we have to invoke the idea (and thus look for the emergence) of an autonomous language faculty. Though the cognitive reality of the latter has always been a controversial point, its existence is widely assumed among linguists simply because it plays an axiomatic role in the theoretical construction of generative linguistics, which for various reasons (but certainly not those of ultimate truth) dominates the linguistic scene. There are many details of language origins that are in one way or another connected to the controversy of the existence or non-existence of an autonomous language faculty. One issue, however, that is of overall importance for the language origins problem is adaptation. Although from a biological point of view, language, being a very complex trait, intuitively appears to be adaptive and the result of natural selection, the assumption of an autonomous language faculty and an arbitrary grammatical structure, on the other hand, generates a common bias against any selectionist view (e.g. Piattelli-Palmarini 1989).

As opposed to this, Pinker and Bloom (1990) have tried to counter these ideas and account for the selective advantage of language based on the adaptive value of its structure. They, however, leave function almost completely out of the picture, which, in my opinion is a cardinal point in an adaptationist explanation. The next chapter will deal with this issue in more detail and examine how structure and function are mutually dependent on each other in the


course of natural selection and adaptation to a new niche.

Chapter 2

The emergence of language and the problem

of evolutionary continuity 2.1 Structure, capacity and behavior in language evolution One of the most prominent questions in connection with language evolution is whether the human capacity for language can be related to any capacities found in animals, especially our closest relatives, the great apes. Concerning the anatomical structures involved in language, evolutionary continuity is obvious, but in the case of cognitive capacities and behavior it is rather controversial.

Thus, the physical characteristics involved in the capacity for language can without problems be derived from structures found in animals, first and foremost primates. This is the case with the anatomy of the human vocal tract (e.g. Lieberman 1984) and the anatomy of the human brain (e.g. Deacon 1988). Though the specific functional characteristics of particular brain areas involved in language might be innovations in the hominid lineage, even hemispheric dominance, which for a long time was considered a unique feature of the human brain, has been proven to exist both anatomically and functionally in primates, even various mammals and maybe even other vertebrates (Bradshaw and Rogers 1993, Hamilton 1990; Hamilton and Vermeire 1991; Walker 1980) Thus, even lateralization, usually considered crucial to our linguistic capacity, appears to have had precursors in our ancestors (Holloway and de la Coste-lareymondie 1982; Heilbroner and Holloway 1988; MacNeilage 1993).

When we move from physical structure to cognitive capacities that underlie language, consensus on commonalities and common origins is harder to reach. For many linguists, philosophers and psychologists it is an obvious case that our linguistic capacity is not


based on general cognitive capacities (e.g. Chomsky 1988; Fodor 1994; Jackendoff 1992). Even the biologists Maynard Smith and Szathmáry (1995: 281) argue that “the ability to learn to talk is independent of general intelligence.” However, they acknowledge that “one aspect of language competence, namely the ability to form and manipulate mental representations, did evolve because of selection for general mental ability” (Maynard Smith and Szathmáry 1995: 291).

On the other hand, there are linguists for whom it seems to be clear that the linguistic capacity has evolved from and is thus based on cognitive capacities that we share with other animals, most probably the sense of space, both visual and kinesthetic. Thus, for instance, Givón (1998: 46) claims “that a big chunk of the neurology that nowadays support[s] human language is but an evolutionary outgrowth of the visual information processing system.” In the same vein Lakoff (1990: 73) argues that

… certain mechanisms for the perception of spatial relations that appear to be present in lower animals are used by human beings in abstract reasoning – that aspect of human beings that has traditionally been taken as separating man from the lower animals. But biology has shown us that man is not a completely separate life form; rather human beings use many biological capacities present in animals that evolved earlier. From the evolutionary point of view, of course, it would not be surprising if human reason were to use and build on mechanisms for representing spatial relations that are present in lower animals. Indeed, the idea that abstract reason makes use of spatial perceptual mechanisms present in lower animals makes much more sense than the idea that reasoning came in all at once with man as a totally separate new cognitive faculty. The idea that abstract reason also evolved just makes more biological sense.

We should keep in mind too that language is not just a matter of anatomical structures and mental capacities, but also a matter of behavior, which must also emerge in an evolutionary sequence (Catania 2001). Thus, the issue of continuity also arises with respect to the question whether human linguistic behavior is the

Emergence of language and evolutionary continuity 49

evolutionary continuation of the communicative behavior of animals. Arguably, it is in social behavioral features that human language and animal communication have most in common – at least, and especially, insofar as both are behaviors designed for “getting another organism to do something” (Catania 1994: 38). Catania claims that this is also the primary function of language, namely transferring information in order to influence the behavior of other organisms, and all other functions of human language can be derived from this (cf. also Catania 1991).

The debate about the issue of continuity with respect to language origins originally arouse in connection with the question whether the difference between human language and other types of animal communication systems is of a quantitative or qualitative nature (cf. Lenneberg 1967: Chapter 7). Traditionally, biologically centered approaches to language were mostly inclined to accept an evolutionary continuity between language and the communication systems of other animals and, based on this, some even postulated a difference in degree only (e.g. Gallup et al. 1977; Hill 1978; Savage-Rumbaugh et al. 1980). However, theoretical considerations of language usually claimed that language was qualitatively different from any type of animal communication, which was commonly taken as a proof of evolutionary discontinuity (i.e., the lack of any traits in ancestor species that could have served as evolutionary antecedents) (e.g. Chomsky 1979; Premack 1985; Sebeok 1991; Corballis 1991; Burling 1993). These points of view imply a correlation between quantitative difference and evolutionary continuity on the one hand and qualitative difference and evolutionary discontinuity on the other. Recently these two positions have even incorporated claims of gradual and punctuated evolution of language respectively, suggesting a further correlation between these notions. In this chapter I will look into the problems arising from these suggested correlations. 2.2 Language as an evolutionary innovation and the issue of continuity If we look at language in all its complexity, it is unquestionable that it is an evolutionary innovation in every sense of the term in the


human lineage (cf. Nitecki 1990: 5). Language is a trait not found in other animals, since there are crucial qualitative differences between it and other systems of animal communication. The most obvious qualitative differences is revealed through a comparative structural analysis of the two. Let it suffice here to mention the well-known fact that in contrast to the communication systems of other animals language is an open (i.e., creative) system due to two basic features. One is duality of patterning (or structure), which means that the vast number of meaningful units of language (words, phrases, etc.) is made up of a very limited number of meaningless units, which are the human speech sounds (Hockett and Altman 1968). The other has been called productivity by Chomsky (1968). This is the special characteristic of the system which enables us to form an indefinitely large number of sentences from a finite number of elements (words) and the rules of their combination due to the recursiveness of these rules.

From a functional point of view language also appears to be not just more, but different. It is not only an instrument of communication but simultaneously also one of cognition. This means that linguistic signs do not only function as communicative signals but at the same time also as symbolic mental representations of reality in which knowledge is stored. As for the origin of the coupling of communication and cognition, originally two separate biological functions in organisms, Armstrong et al. (1995: 28) claim that “gesture is a critical link running through the evolution of perception, conceptualization, and language.” Most probably, the adaptive value of language resides in this joint function (Jerison 1988; Győri 2001; cf. also Maynard Smith and Szathmáry 1995: 284, 286). Symbolic mental representations enhance the effectiveness of the mental model that the brain creates and operates by freeing it considerably from the limitations of direct experience. This means that much of the knowledge contained in these representations is about aspects of the world which have never been subject to our individual experience, either because they are not of a perceptual nature or otherwise not directly experienced (Győri 1995, 1997). This joint communicative and cognitive function also makes language an instrument of social cognition, which may be its major evolutionary significance. With the help of language we can gain knowledge through the experience of other individuals because


language makes the parallel connection of modeling brains possible and thus creates a mental supermodel shared by a whole community (Csányi 1992). The result of this is that humans are capable of exchanging information to a degree to which no other species comes even close. The manifest structural difference between human language and animal communication systems obviously stands in service of this representational function.

As mentioned in the introduction, for many linguists, philosophers and psychologists these structural and functional characteristics of language entail an evolutionary discontinuity between language and the communication systems of primates. More than three decades ago even the biologist Eric Lenneberg made the following prominent claim in support of his discontinuity theory of language origins:

There is no evidence that language comes about by a gradual accretion of skills. If this were so, we should be able to see all but a few of such skills in our closest relatives; further relatives should show a few less, and so on down the line of evolution. Nothing like that seems to be the case. (Lenneberg 1967: 233)

This quote represents a common misunderstanding in connection with evolutionary development. It is often believed that for continuity to hold in the case of a certain trait rudimentary forms of the trait in question are required in related species. However, there is a crucial difference between rudimentary forms and what should normally be called forerunners.

A general and basic example to demonstrate this issue could be the case of organic macromolecules, which were necessary forerunners of life, but definitely not rudimentary forms of life, since they are not living organisms. Rudimentary forms of life, on the other hand, are for instance procaryotes. A target trait, which we set out to explain by exploring its evolutionary origins, will by definition always be qualitatively different from a forerunner and only quantitatively different from a rudimentary form. In the above example the fact that organic macromolecules were necessary forerunners of life also justifies the claim for a qualitative difference. On the other hand, it seems to be fair to say that as for


the trait which we call ‘life’ there is only a quantitative difference between procaryotes, as rudimentary forms of life, and say, mammals, although in other respects (i.e., concerning other traits) there are of course innumerable and enormous qualitative differences. Because of this, the verification of a special property in an organism as a species specific trait cannot mean the refutation of the existence of any forerunners of this trait in an ancestor species and reflexes of it in other (i.e., sibling) species.

In the case of language we should consider as forerunners (by definition) those conditions and traits possessed by our direct primate ancestors, and probably also to be found in today’s apes, which were shaped into language by certain evolutionary processes. Such could be various forms of behavior, certain communicative skills and certainly the cognitive capacities referred to by Givón (1998) and Lakoff (1990) (see above). These non-linguistic capacities found today in other animals, apes and maybe even more distant relatives, may have served as a starting point for a “gradual accretion of skills.” Dawkins (1986: Chapter 9) has pointed out that gradualism should not be understood as a process of regularly occurring steps of equal size. Rather, gradualism in a proper sense means the lack of disruptive gaps in the development without any implication to the constancy of speed. Thus, language evolution could have proceeded in an accelerated and uneven fashion in the hominid lineage and still be continuous with earlier traits. The differences in kind could have been engendered through changes to these previously existing traits (with other functions) (for a more detailed explication of this claim see further below). Consequently, the lack of even the most rudimentary language skills in apes does in principle not allow an automatic conclusion to evolutionary discontinuity and thus the refutation of the possibility that certain features or elements of primate communication and cognition are forerunners of language, i.e., that such traits provided the material for the onset of language evolution.

As opposed to forerunners, rudimentary forms are forms that have already developed past the traits that can be considered the origins of a new trait. Thus, if it turned out to be the case that some rudimentary form of language is possessed by non-human primates, it would simply mean that language had started to evolve some time earlier than the evolutionary branching point between humans and


chimpanzees. In this case the origin of language would have to be looked for in even earlier periods of hominoid evolution and we would still have to continue our search for those traits that language can be derived from. Because of this, it cannot be the case that for continuity to hold, rudimentary forms of language must be found all the way down the line of ancestor species and also in sibling species. The relevant question to be asked about language evolution is thus whether certain elements of non-human primate communication and cognition are forerunners of the human linguistic capacity, or whether these traits indeed have nothing to do with one another. Could it have been the case that certain evolutionary mechanisms had shaped elements of these communicative and cognitive capacities first into rudimentary forms of language and later into full-blown linguistic capacity? The evolutionary mechanisms that come to mind are of course natural selection and adaptation. It should be noted here, however, that language could just as well have arisen in a new niche or adaptive zone as a result of some kind of preadaptation. Though Mayr (1988: 409) talks about evolutionary discontinuity in such a case, obviously even this does not mean the lack of any antecedents. Also, mutation is often mentioned as the evolutionary event responsible for the origin of language, thus apparently causing a discontinuous evolutionary process. I will return to this quandary further below.

It also appears to be obvious from the above that there must be a transitional developmental period between forerunners and rudimentary forms. While several decades ago Lenneberg (1967: 234) was convinced that “the antecedents of the human propensity for language ... are not in the least obvious” in animals, new research seems to have discovered certain skills in related species which lie at the basis of our linguistic capacity. These could be placed somewhere along the scale between forerunners and rudimentary forms. The auditory capacity for phonetic/phonemic distinction in various primates has been known for a long time (e.g. Morse 1976). More recently Seyfarth and Cheney (1992) have claimed that vervet monkey alarm calls have a referential/semantic function. And newly Újhelyi (1996) and Újhelyi and Buk (2001) have reported the discovery of a rudimentary syntactic capacity in various primates and claimed that the syntax of various calls is often communicatively relevant. Thus, if we take these findings at face


value, it should not appear impossible that language emerged as the accretion of such communicative skills coupled with a growth in cognitive capacity. What the rate and curve of such an accretion may have looked like is another matter and will be dealt with later on.

The interpretation of the term “closest relative,” which Lenneberg uses in the quotation above, also has some bearing on the continuity-discontinuity issue. This notion is not without problems and in applying it it is necessary to realize its relativistic nature. For instance the difference between the rates of the evolution of macromolecules and anatomical or behavioral features is responsible for the fact that chimpanzees and humans are very close relatives genetically, but when we take behavior as a measure, they are quite distant ones (King and Wilson 1975). Maynard Smith and Szathmáry (1995: 292) point out that “[t]he intellectual arms race took place within the species itself” and this most probably served as the basis for the unusually accelerated evolution of Homo sapiens at the behavioral level. Because of this it is not enough to consider only closest living relatives but also related paleospecies must be considered. Thus, the evolution of the species within the genera Australopithecus and Homo (habilis, erectus, and sapiens) can in theory provide the time for an “accretion of [language] skills,” which need not have been a gradual process, but may have been accelerated at times and slowed down or in stasis at others. Although Mayr (1988: 409) clearly differentiates evolutionary innovations from adaptations as involving large vs. small changes, he admits that “[t]he end stage of ... an adaptive development ... can amount to a veritable innovation, particularly when the intermediate stages become extinct.”

Thus, the seeming missing links in the case of language should not suggest to us the impossibility of any continuous evolution. Elsewhere even Lenneberg (1976: 200) himself defends the view that every new trait must have its origin in previously existing ones. The fact that chimpanzees do not possess rudimentary forms of language should not mean that the rudiments of such skills, which we normally postulate for hominid paleospecies, could not have evolved from traits that existed in our hominoid ancestors, and which thus were forerunners of language, even if this involved developing new functions. Evolutionary history abounds in cases where qualitative differences can be traced back to common ancestor


traits. This seems to be rather the rule than the exception. Thus, any qualitative difference between animal (or more precisely primate) communication and human language should, at least in theory, be compatible with a continuity theory of language and the human mind in spite of the problems and difficulties such an approach involves (cf. Donald 1991: 24). 2.3 Continuity and discontinuity in evolution The point I would like to illustrate in the following is that a species specific or, in other words, unique trait, like for instance language, is not necessarily an independent development in the sense of being the result of a discontinuous evolutionary process. As I will show below, it is by far not impossible for related species to display species specific and hence qualitatively different characteristics that have developed from common ancestor traits. In fact, it is quite obvious that the majority of species specific traits are homologous with qualitatively different traits in related species, i.e., they are results of evolutionary changes to some kind of ancestor trait shared with other species. It is important to emphasize this commonplace because claims about the uniqueness and evolutionary novelty of human language have most of the time implied that language is a trait without any evolutionary precursors outside the hominid lineage.

However, in evolutionary theory there is no justification for the claim that any parallelism exists between the type of difference and the type of evolutionary process that engendered it. The main reason why this kind of correlation does not hold is that the issue of continuity or discontinuity and the issue of difference in degree or kind relate to different aspects of evolution. When looking at the continuity/discontinuity problem, we investigate sequences of changes that have occurred in the course of the evolutionary process. On the other hand, when trying to decide on differences in degree or kind, we examine cross-sections of the evolutionary process. We compare various stages of evolution and evaluate the difference we find. An examination of this kind may involve either successive stages in one lineage or contemporaneous stages in different but related lineages.

One of the most obvious and prominent examples of a case of


homology is the evolution of the locomotive organs in vertebrates. The most illustrative for our purposes is a consideration of the origin of forelimbs in amphibians, reptiles, birds, and mammals. Nobody would question the fact that they evolved from fish fins, i.e., that forelegs in various quadrupeds, wings in birds, and hands (or arms) in primates developed from the same ancestor trait (Benton 1990: 227-234). However, in spite of being homologous developments, it appears to be fully justified to consider them as qualitatively different. It would seem absurd to claim that they only differ in degree because they have the same evolutionary origin. The reason for attributing to them a difference in kind can of course not be simply intuitive. In general this is a matter of the structure these particular organs have and the function they serve. Most of the time it should be more elucidating to make a judgment on the basis of the function, which reflects the adaptedness to particular environments (cf. Mayr 1988: 128), while the evolved variation in the structure may also be a result of morphogenetic processes (e.g. Goodwin 1996: 157).

In the case of human and other mammalian forelimbs the functional difference is rather obvious. Human arms and hands, though homologous to the forelimbs of other mammals, do not serve locomotion any longer but evolved to manipulate objects and handle tools. Even the hands of non-human primates are more specialized for locomotor function than prehension and object handling (Jouffroy 1993). In contrast, fins, wings and forelegs all serve locomotion. Furthermore, a functional, and hence qualitative difference between these latter can also be claimed on certain grounds. Our considerations of a possible functional difference will have to involve more subtle detail and we will have to break down the concept of locomotion into different types like swimming, flying and various kinds of terrestrial quadrupedal locomotion. Thus, functional characteristics seem to be quite suitable for distinguishing quantitative and qualitative differences.

It is of course obvious that physical traits are far easier to track down than cognitive-mental or behavioral ones. In the case of the evolution of the human arm and hand the comparative anatomy and physiology of primates provides unequivocal clues for a reconstruction of evolutionary changes (Jouffroy 1993). However, despite the huge qualitative differences between the forelimbs of


humans, other mammals and further other vertebrates, they unquestionably have a common origin. The important general point to note is that there is no necessary correlation between a qualitative difference and discontinuous evolution and thus a difference in kind can be no evidence against evolutionary continuity.

The discussion above about limbs and language concentrated on the phenotypical level because it is in phenotypical features, physical or behavioral, where quantitative and qualitative differences are best manifested. However, we must not forget, as Dobzhansky, Boesiger and Sperlich (1980: 131) warned quite some time ago, that at this level even closely related species form discontinuous groups because most of the time no continuum of the smallest variants exists between them. Obviously, evolutionary continuity will not only hold when some kind of continuum of quantitative differences can be found between related species, since each different species is constituted by its own unique form and traits. A proper understanding of continuity requires the realization that no real intermediate forms are needed between related species or in the evolutionary record. 2.4 Genetic aspects of language evolution For a more subtle perspective on the continuity-discontinuity issue we also need to look at the genetic basis of evolution, since at the molecular level we might find a more complex picture of the matter. Actually, the genotypic differences that we find between individuals are by nature of a qualitative kind. Such differences provide the material for microevolutionary changes, which are constituted by changes in the gene frequency of a population. In this way even very small evolutionary steps mean qualitative (i.e., discontinuous) change at the genotypical level. Thus, at the molecular level any change is by nature a change in kind, even the smallest one, though at the phenotypical level we may still perceive a continuous process of change in degree. However, whether we interpret the process of evolution to be continuous or discontinuous will depend on the size of the phenotypical differences thus arising, which in turn are the function of the size of the underlying genetic changes.

Phenotypical features are always a combined expression of


genes and environmental factors, but the heritable portion of the features has a genetic basis. Thus, new phenotypical features usually arise through changes at the molecular level, such as mutation and the recombination of genes. Of the two it is mutation that may have a larger effect on the phenotype. Because of this the question of mutation takes a prominent place in arguments for the discontinuity between language and primate communication systems (e.g. Chomsky 1988: 170, 183; Bickerton 1995: 83). Since it is rather obvious that much of the human linguistic capacity is genetically determined, mutation, as a major source of evolutionary change, must surely have played a major role in the evolution of language. Because of this, by claiming that language arose through mutation nothing has been said that is not self-evident. The real issue is rather the rate and size of the genetic changes that played a role in the emergence of language. Could language have arisen as an entirely novel trait through some kind of large beneficial mutation or do we have to do with a relatively slow accretion and growth as a result of small genetic changes to already existing traits?

As a matter of fact, mutation is an inevitable mechanism in the evolutionary process because no evolutionary change could take place without it. The genetic variability to be found in a population at a certain point in time is always the result of prior mutations and genetic recombinations, and natural selection acts on this variability and may even fix certain features in the population. However, these mutations relate most of the time to existing traits in the sense that they modify these traits instead of creating completely new ones. Given the necessary time, so much genetic change may be involved that the accumulation of quantitative differences will inevitably be perceived as, or in fact yield, a qualitative difference (cf. Mayr 1988: 408-409). Such an accretion and growth need not be a gradual process in the sense of a constant speed development. It can take place at an accelerated rate in a relatively short time in evolutionary terms and even lead to macroevolutionary changes (cf. Mayr 1988: 459). Although this may look as a jump from a distant perspective, a qualitative difference at the phenotypical level may in this way be the result of continuous evolutionary change. Continuity means nothing more than that certain traits in earlier species were modified by mutation and recombination to yield with time a functionally and thus qualitatively different trait in later species.


The reason why the idea of mutation is invoked in accounts of language evolution is not simply the wish to explain the qualitative difference between primate communication and human language through evolutionary discontinuity. A crucial assumption in this respect is that only a full-blown linguistic capacity can be functional (Chomsky 1988: 170, 183; Bierwisch 1992; Bickerton 1995: 70-75). The alleged problem is the same as in the famous case of the evolution of the eye, i.e., whether the eye could have evolved step by step when theoretically all interlocking parts must fall in place simultaneously in order for the eye to function properly (cf. Dawkins 1986: 80). If a trait is not functional, i.e., it is not of any special use or advantage in the given environment, natural selection cannot act on it and so it cannot evolve through adaptation (Piatelli-Palmarini 1989). Thus, the idea of discontinuity and especially the assumption that only a full-blown linguistic capacity could have a selective advantage implies a sudden emergence of language.

In the following I will first examine how legitimate it is to claim that an evolutionary trait is fully developed and then take a look at the possible evolutionary scenarios of a sudden emergence of language and assess their probabilities. The crucial question to be asked about the assumption that only fully developed language can be functional is on what grounds an evolutionary trait can be judged to be fully developed. There are at least three ways in which a trait can be judged fully developed, but all of them seem to be more or less problematic. First of all we could base our judgment on some kind of essentialist assumption, which invokes the dubious metaphysical idea of a true language. Obviously, this is not a legitimate scientific approach. Second, we could take the structure of natural languages spoken today as the measure of fully developed language structure and the capacity to handle it as the measure of full-blown linguistic capacity. However, there is quite a bit of anthropocentric bias in this approach. This will become clear when we take a look at the third, evolutionary biological aspect of the issue.

Pondering a less and a more developed level of a trait in evolutionary terms is a relativistic enterprise and thus legitimate only in retrospect, i.e., when comparing two evolutionary stages of a trait. Otherwise every organism can be said to have 100 percent of a particular trait in the sense of this particular trait being perfectly


functional in the organism’s niche, since 100 percent of a trait can mean nothing more than the extant adaptedness of it to the particular environment of the organism possessing it. Claiming that a trait is fully developed out of its evolutionary context implies that there is a perfect state toward which every organism and their traits evolve. This kind of teleological thinking is totally impermissible in evolutionary biology.

In the case of a trait like language it is especially hard to see why there cannot be several intermediates between no language and fully developed language (cf. Maynard Smith and Szathmáry 1995: 308). From a linguistic point of view the dichotomy is obviously between non-hierarchical structures and hierarchical ones, between which a transitory state is apparently difficult to imagine. But whereas one element can in fact not form a hierarchy (though neither even a linear sequence), two elements can already form either a simple linear sequence or a hierarchy, or even both at the same time, especially if it is recognized that a linear sequence is the simplest possible hierarchy. From here on the avenue for the evolution toward more complex structures is opened up. Less complex structures will simply allow less information to be transferred and processed. This way even a selective pressure on the capacity of handling more complex structures could be accounted for if niche changes made this necessary. Furthermore, “[t]he need to speed up discourse processing” also must also have played an important role in the appearance of grammar in language evolution (Givón 1998: 92).

Because of this, looking at the issue from the perspective of the communicative and cognitive effectiveness of language appears to make more sense. Through simpler structures in a language-like system less communicative and cognitive power can be achieved. It must be specially emphasized here that the term language-like does in no way invoke the idea of a subsequent true language structure. In itself such a system could also be called language and it is only in retrospect, compared to the present state of language evolution, that we can use the term language-like. Such a language-like system, or any other trait for that matter, does not and cannot show any signs of inferiority per se at any particular state of evolution, which is actually equivalent to the perspective of a particular organism interacting with its environment. In other words, every trait is fully developed from the particular organism’s point of view, irrespective


of what turns evolution might take in the future. The issue whether the emergence of language was a sudden

event or a slow process in human evolution cannot be separated from the question whether it was a micro- or macroevolutionary event (cf. Armstrong et al. 1995: 205-206). Several authors (e.g. Bickerton 1995: 69) have found support for their hypothesis for the abrupt appearance of language on the evolutionary scene in the claim forwarded by Eldredge and Gould (1972) that speciation proceeds by abrupt changes, i.e., macroevolution consists of a series of “punctuated equilibria” (Gould 1980). Mayr (1988: 415) claims that our ancestry may have been characterized by such punctuated evolution because “[n]o transition from Australopithecus africanus … to Homo habilis has so far been found, nor from Homo habilis to H. erectus.” Thus, while the ancestor species did not go through much evolutionary change, the new species must have originated in isolated founder populations. Rightmire (1981) examined fossil skull and tooth dimensions of Homo erectus and came to the conclusion that the species indeed represents a period of stasis in human evolution, implying a punctuated curve of the evolution of the successive hominid paleospecies. On the other hand Wolpoff (1984) has found that the time span of Homo erectus is characterized by significant changes in a number of cranial, mandibular and dental features, suggesting a much smoother transition to Homo sapiens than a punctuated one.

It has also been suggested that the idea of punctuationism entails macromutations and ‘hopeful monsters.’ It is of course conceivable that a certain trait arises entirely through mutation, in which case a claim for continuity may be difficult to maintain, since under such circumstances the given trait can in no way be related to another more archaic trait. But obviously, such a mutation must be a beneficial one, since the major way for genetic changes to spread in populations is when their phenotypic effects confer a selective advantage on the affected individual(s) (Elena et al. 1996). The question is whether language is a trait that could have arisen in this way. To answer this question first we have to bear in mind that such a huge genetic change in an individual, if conceivable at all, must be an extremely rare occasion, to say the least. However, any claim of a sudden emergence of a ‘linguistic hopeful monster’ would also have to account for the spread of such a macromutation in a population in


order to lead to evolutionary changes. For such an explanation it must be borne in mind that language, as a behavioral trait, is a social phenomenon and an inevitable product of complex social interaction in human evolution (Csányi 2000; cf. also Armstrong et al. 1995: 199). Because of this several individuals must bear the same trait in order for it to be of any advantage to any of its possessors. This is also why the mechanism of group selection has been justly invoked in accounts of language evolution (Csányi 2001; Mayr 1988: 79-80). If group selectional mechanisms played a role in language evolution, the possibility that it was a huge beneficial mutation which spread is the more unlikely.

The main problem, however, with the idea of a beneficial macromutation is with its genetic plausibility. When trying to look into the genetic bases of language evolution, we have to start out from the fact that language is a complex polygenic trait. This is rather straightforwardly suggested by the huge phenotypical complexity of the linguistic capacity, which is manifest both in vocal tract and neural anatomy and physiology on the one hand and behavior and cognitive skills on the other. Purely on this basis language appears to be too complex a trait to have emerged full-blown. Maynard Smith and Szathmáry (1995: 309) also argue that changes in anatomy, in motor control, in sound perception, and in grammatical competence, all required for the emergence of language, could not have been instantaneous. The sudden emergence of a trait with such a complex polygenic character would have required a series of concorded mutations occurring at all the loci which are involved in language. Also, a huge macromutation of this kind with totally coordinated gene functions is simply statistically rather improbable. Mayr (1988: 483) points out that “seeming evolutionary saltations … can be explained without invoking systemic mutations and other mechanisms in conflict with molecular genetics.”

Thus, it is not necessarily only through some kind of macromutation that a trait can suddenly emerge. Some evolutionary biologists have argued against the overall role of natural selection in evolution, since in their view the rate of evolution can only be explained by a rather high percentage of neutral mutations (e.g. Kimura 1983). These mutations can spread and become fixed in peripheral founder populations through genetic drift and lead to


evolutionary changes (cf. Mayr 1988: 368). In this way a trait can latently emerge and become functional suddenly if niche changes make it necessary. Bierwisch (1992) has based his hypothesis of the latent emergence of a full-blown linguistic capacity on this theory, claiming that the linguistic capacity could not have been selected for till it was really functional. However, it seems to be questionable whether a complex capacity like language could have been assembled from independent neutral mutations through accident, i.e., without any selection (for a detailed critique of Bierwisch see Győri 1992).

There are other possibilities than the spread of a beneficial macromutation or genetic drift for the emergence of language to have been a relatively sudden event. Theoretically, the origins of language could have involved spandrels that emanated as side effects of other structures. In this case it is even possible that these are real innovations in the sense of having appeared in the hominid lineage without antecedents in prehominid ancestors. However, any legitimate claim of a trait being a spandrel must also include an account of the other structure that the trait in question is supposed to be a spandrel of. In what other way could the idea that a trait begun as a spandrel be justified? Maynard Smith and Szathmáry (1995: 290) point out that even if language originated as a spandrel, it must be a modified one simply because of its complexity. In this case the modifying force must have been natural selection, which makes a stepwise transition plausible. Though Maynard Smith and Szathmáry (1995: 309) acknowledge that this transition could have been a rapid process, they firmly reject the idea of an instantaneous emergence of language.

A very likely, if not a necessary way for language to have arisen is from preadapted earlier traits. Givón (1998) has shown rather convincingly that the neural substrate responsible for the visual information processing system in prehominid ancestors must have served as the evolutionary forerunners of the neural substrate responsible for the various components of our linguistic capacity. Maynard Smith and Szathmáry (1995: 290) also remark that the evolution of most complex structures involve preadaptations, i.e., structures which had served other functions earlier. The reason why the majority of evolutionary developments involve preadaptations is that the probability of new genetically complex traits coming into


existence out of nothing through mutation is extremely low. A further reason for preadaptation playing a major role in evolutionary changes is that

[n]o organism can invade a new adaptive zone unless it has a minimum of structural, physiological, and behavioral attributes that preadapt it to succeed in this shift ... [because] … [n]atural selection makes use of whatever genic or phenotypic material is available in order to answer a newly arising need. (Mayr 1988: 410)

No doubt, the emergence of language must have been facilitated by such an ecological shift. When this shift took place in human evolution is another question. For instance, Quiatt (2001: 25) concludes on the basis of genetic and cultural evidence that “glottogenesis, however and whenever it began, gained speed and was completed with the emergence and dispersal out of Africa of anatomically modern Homo sapiens.” If such an ecological shift is a rapid one, it will obviously result in punctuated evolution. However, this is not incompatible with the idea of continuity because a more fine-grained look should reveal nothing more than an unusually accelerated change and not an ex nihil emergence.

Chapter 3

Natural selection, adaptation and the

evolution of language 3.1 Some problems of language evolution The Darwinian concept of evolution implies a process of change in which organisms adapt to their environment. If such an approach is valid for language, recognized by many as a complex adaptive trait (e.g. Jerison 1988; Donald 1991; Plotkin 1994), then language evolution must reflect the interaction of humans with their environment. Any change is inherently characterized by the constraints put on it by the nature of its material and the environment it takes place in. Because of this Medin and Wattenmaker (1987) claim that biological-cognitive constraints become embodied in organisms as a result of their interaction with the environment during evolution. This chapter will look at how such constraints influence language evolution and what the adaptive value of language is. If language indeed emerged in humans as a cognitive adaptation, then the constraints should depend on the role language plays in this interaction and should be reflected in language structure and function.

There are two main issues in the debate on the evolution of language. One of these is the question if emergence happened by gradual accretion or abrupt appearance of skills and the other is if language has any adaptive value. The previous chapter investigated the first problem in detail and came to the conclusion that a continuity theory of language does not automatically imply a step by step emergence of the human linguistic capacity at a constant rate, nor does it contradict the claim for a relatively sudden appearance of language on the evolutionary scene. All it entails is that the ancestors of hominids must have possessed traits (most probably


retained by our closest living relatives) that gave rise to the capacity for language. Below I will discuss some of the evolutionary biological aspects of the emergence of human language. The need for a systematic treatment of these questions arises from the fact that recently some controversy has surrounded a neo-Darwinian view of language evolution. Interestingly, much of the scepticism stems from arguments within linguistic theory. At the heart of this controversy is the issue of the possible evolutionary mechanisms responsible for the emergence of the structure of language as we find it today.

For a long time discussions of the origin of language were purely of a philosophical-linguistic nature. However, as evolutionary thinking gained ground in biology a growing interest has arisen in the connection between the origin of language and human evolution. Even general discussions of human evolution have often included treatment of the emergence of language as it obviously has been a major factor in becoming human. The major issues in these discussions were on the one hand the behavioral changes possibly connected to language evolution, like food sharing and hunting (Isaac 1976; Montague 1976) or tool use (e.g. Hewes 1973; Frost 1980; Toth and Schick 1993; Davidson and Noble 1993), and on the other hand, the anatomical and physiological changes accompanying language evolution, like those related to encephalization (e.g. Jerison 1988) and vocal tract development (e.g. Lieberman 1984). Naturally, also the relationship between language evolution and the origin of the human mind and consciousness have been discussed extensively (e.g. Donald 1991) and the probable effects of interactions between biological and cultural evolution on language evolution have been examined (Csányi 2001; Quiatt 2001). In most of these approaches, often involving scenario like explanations, it has been assumed, tacitly or not, that a neo-Darwinian view of language evolution was in place. Thus, as the accounts lay in general within the framework of the theory of natural selection and adaptation, the issue was not what evolutionary biological mechanisms were responsible for having shaped the human capacity for language, but rather, natural selection taken for granted, what the selection pressures were to which evolving hominids reacted by developing language.

Recently, however, the problem of evolutionary mechanisms has come more to the foreground in discussions of evolutionary biological aspects of language origins. It has been repeatedly

Natural selection, adaptation and evolution of language 67

questioned whether natural selection and adaptation could have been responsible for the emergence of language. This scepticism arose through a growing pressure to bring the so-called innateness hypothesis in linguistic theory in line with an evolutionary explanation. For a long time this did not seem to be at issue. The Chomskyan theory of language, in the mainstream of linguistic theorizing for the past four decades, postulated the existence of a genetically innate universal grammar, which is responsible for the particular grammatical structures found across the different languages of the world. Thanks to Chomsky, rationalist thinking prevailed in linguistics and the idea of innateness was mainly a philosophical creed debated only due to some criticism on the grounds of empiricism (e.g. Lakoff 1987; Johnson 1987). Although it has always been obvious that the idea of innateness should coincide with the facts of genetics, linguistics did not have much contribution to make in this respect and thus innateness was supposed to serve as an ultimate explanation for the rules of grammar that linguist were to uncover. On the other hand, the science of genetics is still not in a position to offer a detailed explanation for the capacity of language. Thus for biologist, not being concerned directly with structures of grammar, simply the idea of genetic innateness in the case of a complex capacity like language has never seemed such a crucial issue, since as Jan Wind (1988: 140) put it, “to biologists the statement [that ‘language has a genetic basis’] sounds somewhat trivial.”

However, the reason for the newly emerged interest in the evolutionary foundations of innate grammar seems to be fairly clear. With the development of generative grammar the claim for innateness did not remain at the general level and more and more specific grammatical structures were claimed to have a direct genetic basis (Chomsky 2000a, 2000b). In spite of the seeming biological orientation of these claims, this meant that no further explanation was required as to the origin of the ‘language organ’ as the seat of these structures, a claim found “baffling” by the evolutionary biologists Maynard Smith and Szathmáry (1995: 190). Since Chomsky (1988: 46) found “no logical reason” for language structure to be as it is, the idea of arbitrary and autonomous language structure made a search for the evolutionary origin of language structures not only unnecessary but also futile, and natural selection


and adaptation no longer seemed necessary or even evident explanations (Piattelli-Palmarini 1989).

However, even among generative linguists there was a dissatisfaction with a ‘solution’ that deprived linguistics of the chance of providing a causal explanation for the structural features of language. Also, the question of evolutionary mechanisms seems to have been such a sensitive issue that even some linguists were unwilling to abandon the neo-Darwinian view, which has proved so useful in explanations for the evolution of complex organs and capacities. Several linguists attempted to give a neo-Darwinian explanation for the essentially arbitrary structure of universal grammar and to account for an innate universal grammar as an adaptation that has been selected for in human evolution (Pinker and Bloom 1990; Newmeyer 1991). Hurford (1991) tried to show the adaptive advantage of the language faculty and Corballis (1991) tried to explain generativity as the adaptive feature of the human brain. It has to be noted here that the autonomousness and arbitrariness of grammatical structure is by far not generally accepted among linguists (e.g. Haiman 1983), especially in the relatively recent cognitive linguistics tradition (Langacker 1987, 1991). 3.2 The adaptive value of language The basic question whether language is an adaptation that has been selected for or whether language has come about through some other mechanism cannot be totally separated from the problem of gradual versus sudden emergence discussed in the previous chapter. Thus, the issue is whether any particular capacities of our (animal) ancestors could have been selected for under certain environmental conditions so that with time they evolved into what we today call language or whether we have to consider language a true innovation in the sense of some kind of a sudden or latent emergence. We also have to bear in mind that in talking about what is usually called the ‘selective advantage of language’ some kind of rudimentary form of language is already presumed. Obviously, this cannot be taken as the starting point of language evolution, and first the essential question must be asked: Where does any rudimentary form of language come


from, i.e., what could the evolutionary precursors of language have been if any?

For an explanation of the emergence of language in terms of adaptation and natural selection one should be able to give a characterization of the niche that provided the evolutionary pressure on language evolution. It is too often stated simply that there was some kind of evolutionary pressure, without making an attempt to specify what this pressure could have consisted in. A trait cannot be advantageous just by itself, only under very specific conditions. Thus, better communicative and cognitive capabilities could have emerged only under circumstances that favored such capabilities. There is an ever ongoing interaction between a trait and the niche it exists in and this niche will be responsible for the adaptive value of that trait. The adaptiveness of the trait (i.e., its functional value) will necessarily have to be measured against the given environment.

One of the doubtless selective advantages of language is more efficient communication. Theoretically, more efficient communication in itself would be advantageous for all organisms under all circumstances. But as explained above, any trait’s advantage is relative to the environment in which the organism possesses it, or more precisely, to the interaction (interactive relationship) existing between organism and environment. It is not the environment per se in which a trait is advantageous but the behavior of the organism in the given environment. As Mayr (1988: 408) remarks, “[i]n animals, almost invariably, a change in behavior is the crucial force initiating evolutionary innovations” and this sets up a selection pressure. Since all organisms exhibit communicative behavior in their environment, which means that all environment necessitates communicative behavior, especially because conspecifics form a part of the environment, the statement that more efficient communication would be advantageous under all circumstances has no explanatory force.

When searching for the factors which provided a selection pressure for the emergence of language, it is not enough to say that speech/language conferred some advantage, like for instance enhanced abilities for cooperative behavior. It definitely did, though there are some disadvantages involved as well, like less effective breathing, swallowing, etc., which must have been outweighed by the advantages of speech. A common argument against an


adaptationist explanation is that other species would also have benefited from speech. This is posing the problem the wrong way because this is not what is meant by selective advantage in evolution. By the nature of the matter evolution always takes a minimalist approach as concerns natural selection. Selection can only occur on the basis of the given possibilities in the phenotype, and what is more important, selection occurs only in the interest of enhanced fitness or as a solution for survival, often only as one possible avenue in the range of the phenotypical traits exposed to selection.

Evolutionary changes are more likely to occur when the environment changes than when it is stable. And most of the time environmental changes are for the worst, at least from the point of view that a species is usually more or less well adapted to a certain environment, i.e., to the niche it lives in. Thus it is not correct to say that hominids developed language because it was so advantageous for them, but they most probably developed it under environmental pressure where already given capacities could be exploited to overcome certain disadvantageous environmental changes or even maybe the threat of extinction. It is important to emphasize already given capacities to avoid claims that might suggest that a certain capacity was developed because it was a necessary condition for survival. There has to be some material given for evolution to select from in the form of preadaptation. These given capacities might have been exploited (1) to a much lesser degree, (2) for different purposes (3) or simply have been ‘spandrels’ for other capacities, (4) or even some kind of a combination of the three previously mentioned points.

Thus, selection pressure is ‘real’ pressure. It is always those populations that evolve which are forced into the harsher conditions and thus most affected by environmental changes. This is so since an organism is always more or less well adapted to the environment it lives in so no change seems to be necessary. This is also the reason for punctuated equilibria, that is why evolutionary change is not going on continuously, but periods of stasis are followed by relatively fast evolutionary changes. Of course it is possible that some kind of mutation confers some further advantage on an individual under unchanging circumstances, since there hardly is any absolutely total or perfect adaptation. But the point is that whenever


we talk about selective advantage we must not forget to specify that advantage in terms of environmental conditions and niche changes, putting a severe pressure on the survival chances of a population. Because of this gradualism can never mean that evolution is slowly pacing towards greater perfection or greater complexity. If it were so, there would be no explanation why organisms also stagnate, i.e., there are organisms ‘stuck’ at all kinds of evolutionary levels. We simple mean that in the case of such complex traits like language or the eye there simply have to be intermediate stages that conferred some advantage under certain pressing environmental conditions. Although it could also have been the case that there were intermediate stages that were tolerated in the process of so-called neutral evolution. Thus, if these traits did not function until complete, a series of preadaptations must have lead up to them which is a gradual process in the proper sense of consisting of at least structurally if not functionally consecutive intermediate stages, which are favored by natural selection, i.e., exapted (cf. Gould and Vrba 1982), when the time comes. In this sense there is of course preadaptation involved in the emergence of language. The traits that served as precursors of language could well have been preadapted because they proved to be advantageous under certain unfavorable environmental conditions or niche changes. However, language is such a complex trait that the whole capacity for language could hardly have occurred as a preadaptation.

Looked at it this way, there is nothing strange in the fact that natural selection might occasionally result in punctuated equilibria, especially when earlier spandrels or structures with previously different functions come to be selected for due to certain niche changes. Thus, the question of the rate and curve of language evolution is only indirectly related to the continuity-discontinuity debate. While certain features of the primate communication system, cognitive capacity and social behavior must have served as sources, most probably as preadaptations, for language evolution to build upon, the rate and curve of the changes must have depended on the extent of the selection pressures to which evolving hominids, at whatever stage of their evolution, responded by developing language, in which the formerly separate capacities of communication and cognition became conjoined for the sake of a more efficient social cognitive interaction.


Theoretically, the abrupt emergence of some capacity is of course also possible. However, in the case of language the idea of the abrupt appearance of skills should lead to a paradox, which stems from the fact that language is not just a faculty but also a system of signs and rules. Whereas language as a biological faculty can in theory emerge suddenly or latently as a spandrel or preadaptation, this can hardly be said about language as a system, which in reality is always one particular code of communication, however primordial, that has to be learnt. It must be clear that the language faculty in itself could not have conferred any selective advantage on its possessor(s) without the knowledge of an acquired symbol system. A further, and probably even more serious problem is that language (at least as a system of signs and rules) is a social phenomenon, which means that in order for the language faculty to carry selective advantage several individuals had to have it coupled with a shared knowledge of a system of signs and rules. But an abrupt appearance of skills does not leave any room for the evolution of a symbol system.

The idea that language did not emerge through natural selection, makes some of the above problems disappear, but leaves the emergence of a code that has to be learned unexplained. Bierwisch (1992) also assumes that a full-blown language faculty had to precede the use of any symbol system and claims the functions of language to be irrelevant to its evolution. In order to come round the problem of the spreading of the language faculty trait, he rejects the idea that the language faculty could have provided a selective advantage. He also claims the functions of language to be irrelevant to its evolution as in his view all functional explanations of the origin of language are Lamarckistic and teleological. In his opinion functional explanations take the necessity of communication to be the driving force of the evolution of the language faculty. I do not think that asking ‘why’ type of questions in evolution is unjustified, as the term necessity is not used in a teleological sense in evolutionary biology. By selective advantage we mean that from the genetic variability of a population certain genotypes have a greater chance of survival and reproduction. These are the genotypes whose phenotypes are better adapted to the environment, i.e., those that display characteristics that are advantageous in the given environment. We do not assume that certain characteristics


developed for a given purpose, because the mutations underlying genetic variability are not goal directed. Hailman (1976: 15) points out that ethologists speaking about the function of a kind of behavior do not assume that this behavior evolved because of its selective advantage. The appearance of rudimentary traces of certain behavioral patterns or of other properties is accidental due to genetic change through mutations, recombination etc. Selective advantage simply means that individuals exercising such behavior or showing such properties in a changing niche may have a greater chance of survival than those not equipped with that behavioral pattern or property. Because of this they may have more offspring and their genes become more frequent in the population while the genes of the others become rarer (cf. Slater 1985: Chapter 7). To explain the emergence of language with its selective advantage is not a teleological explanation, but a teleonomic one, which is simply the explanation of adaptive processes with natural selection (cf. Lorenz 1981: Chapter I/1). If we speak about the necessary evolution of a function, we do not mean that certain functions evolved because they were necessary for the adaptation to the given environmental circumstances, but that individuals displaying certain characteristics necessarily had a greater chance of survival and reproduction. By this their genes contributed to an ever greater degree to the gene pool of the population, i.e., the characteristics in question occurred more and more often in the population.

It must also be realized that at the basis of the debate about adaptive value of and natural selection for language lies the ‘ideological’ controversy between structuralism and functionalism. I do not think that function and structure can be rightfully separated for other than methodological reasons. If we look at structure in itself and refuse to take function into account (or even deny the existence of function), it is no wonder that the issue of adaptiveness cannot even be raised. Though the question of adaptation is about function, it should not be overlooked that it is always some kind of structure that carries some kind of function. In fact it is the function that a structure has that is of adaptive value, and never the structure in itself, since evolution can create different structures for the same function. In other words, a structure can be adaptive only to the extent of the function it carries. The real issue is of course whether structure is always functional, i.e., whether a given structure


possesses the capacity to function in a certain way or not. The above considerations show that we have to constrain what

is meant by selective advantage, i.e., how it occurs. According to Mayr (1988), evolution is opportunistic: it takes the traits it finds given and tries to make the best of them when environmental changes require it. Evolution is also accidental: it reacts to the immediate needs of the organism, which are the challenges of the environment. There is no design for the future or any kind of teleology. As Dawkins (1986) says, evolution is a tinkerer. But at the same time it is trivial to claim that the evolution of certain structural properties is the result of chance. Evolution is always forced to build onto given material and is thus constrained in its structural possibilities.

This accidentalness in evolution is the manifestation that there is not the slightest teleology involved and thus the term function must also be void of any teleological interpretation. Moreover, different structural make-ups can be exploited for the same function (though often with varying degrees of efficiency, but this is exactly what natural selection is about).For example the capacity for vision evolved in structurally totally different ways in insects and mammals and thus it may seem that the structures of the insect and mammalian eyes are arbitrary formations, still we cannot say that they have no function and thus no adaptive value. A further example is locomotion which is clearly an adaptive capacity but has many structural realizations among animals. I do not want to claim that structure can never be the result of chance, and in this sense even arbitrary, as long as we realize that it is always dependent on the structural possibilities evolution finds given. Because of this I do not think that the selective advantage of language has to be or even can be accounted for by the adaptive value of its structure alone. It is impossible to have a selectionist approach to language evolution built on a purely structuralist basis and to disregard any possible functions of language in an explanation. However, as Pinker and Bloom (1990) have stressed, it is not necessary for every individual part in a structure to have a function in order for the whole to have its own function. Because of this, arbitrary phenomena of language structure cannot be cited as arguments against the adaptive value of language as a whole. Only the functioning totality of a trait can be the measure of its adaptiveness. This is why generative grammar is


so much at odds with any sound evolutionary biological explanation of language.

A rejection of any kind of functional explanation of language thus naturally leads to the rejection of its emergence through natural selection, because any selective advantages of language can work only through some kind of function. There is a prejudice against functional approaches also because function is often equated with purpose and thus any functional explanation is considered teleological (Bierwisch 1992). However, function in an evolutionary sense is nothing more than the use to which a certain structure is or can be put, or simply the position a structure occupies in a superstructure, without any implication of purposefulness in its emergence (Plotkin 1994). It is only in this sense that we can speak of language as having communicative and cognitive functions and it is indeed mistaken to say that language evolved for some purpose. Though Bierwisch recognizes the social advantage of language as a symbol system, he rules out the possibility of the simultaneous emergence of the faculty and the symbol system side of language, because he is convinced of the abrupt emergence of the language faculty as an ‘essential totality.’ Sebeok (1991) also denies the social advantage of language in the period of its emergence, though he recognizes the cognitive-modeling function of language as having a primary selective advantage compared to its social-communicative function. Sebeok correctly realizes that language is qualitatively different from animal communication systems because of its being a cognitive-modeling system. However, there seems to be a contradiction between Sebeok’s claim that the modeling function in itself could have conferred selective advantage and his assumption that a very high degree of subjectivity and individuality was at work in the emergence of language. It has to be stressed here again that also the cognitive modeling function of language has to be based on a specific system of socially learned signs and rules (even if the language faculty is supposed to be equipped with an innate universal grammar). Thus, the communicative function of language could not have been secondary and have appeared long after its cognitive-modeling function (cf. Sebeok 1991).

Other species share their cognitive structures mostly on a biological basis and to some degree on shared experience. However, the emergence of shared cognitive structures proceeds in a


completely novel way in humans, namely by language. That we can communicate our individual cognitive structures by a system of symbols does not mean simply that we activate shared cognitive structures in other individuals. Animal communication also rests on shared cognitive structures (Menzel and Johnson 1976). But human language is more than that. It enables us to develop the same cognitive structures without shared experience and in spite of the fact that our concepts are not genetically determined. I think that this function of human language is its primary unique feature, as it is totally different from the function of animal communication systems or even ape language. The structural uniqueness of human language, either syntactic or semantic, results from this different function. Because of this we have to contradict Chomsky in his opinion:

If we hope to understand human language and the psychological capacities on which it rests, we must first ask what it is, not how or for what purpose it is used. (Chomsky 1968: 2)

However, I think that in the case of many things one has to know the purpose of their usage in order to understand what they really are. For instance one can never really understand what a chair is without knowing that it is used for sitting on it. It is true that language is not a human artifact like a chair. But Chomsky himself says that language was something like a bodily organ. How is one to understand what bodily organs, like the hands, eyes, the heart etc. really are without knowing the purpose they serve? This is just the point with language and because of this I would put the problem differently. If we want to understand what language really is, we must ask what function it serves in humans.

Unfortunately, it has quite commonly been assumed in comparative studies that he function of human language is communication (e.g. Gallup et al. 1977). Also most text books on general linguistics discuss language as a communication system. Hence the belief that human language is communication at a higher level and with more efficiency. If this were really the only addition that human language were capable of, then there would indeed only be a difference in degree between it and animal communication systems in spite of the complex syntactic and semantic structures


found in natural languages. However, there seem to be quite serious structural differences. Human language is not just a means of communication as animal signaling systems and not even one of purposeful and conscious communication (Savage-Rumbaugh et al. 1985). Recently it has been found that not even the communication of higher primates is totally subject to the limbic system, but is partly also governed by neocortical processes, and can thus be voluntarily controlled by the animal to a certain degree (Steklis 1985). Though communication is always a cognitive process, as Menzel and Johnson (1976) have pointed out, it is not by chance that it has always been felt that true language somehow affects the mind in a fundamental way. One of the commonplaces about the effects of language on mind has been, especially since Vygotsky (1962), that language makes abstraction possible. But studies in animal psychology have shown that animals are also capable of abstraction (Herrnstein 1985; Pepperberg 1987). The claim that languages increases abstraction is of course a natural reaction to these findings. It also seems quite plausible, but Premack (1985: 273) has found that names for objects do not add to the abstractness of concepts. Names, however, greatly aid the retrieval of the concepts they are attached to (Premack 1985: 246). This leads Premack to say that it is not clear how words can increase abstractness. He is right in claiming that naming a concept does not add to its abstractness, but the case is that words are not just names for concepts in the sense of functioning as attached labels, though many cognitive psychologists hold this view. If they were, language would indeed only be a means of communicating ideas existing independently in non-linguistic forms. If word learning in human children simply were the association of sound sequences to concepts, then the achievements of apes demonstrated in the ape language projects would count as the learning of a very rudimentary language, because associative learning is just what apes can do. It has been demonstrated that they possess concepts and also have the ability to map signs (or names) onto these concepts (e.g. Savage-Rumbaugh, Rumbaugh, Smith and Lawson 1980). In fact this has also been demonstrated for other species, like the African grey parrot (Pepperberg 1987).

Thus, it is obvious that humans were not the first organisms in which the ability of the cognitive modeling of the environment appeared. Herrnstein (1984) attributes such an important selective


advantage to the ability of categorizing that he considers it characteristic for all living organisms. But with the emergence of language mental modeling could be done in a much more efficient way than ever before. Language enhanced and qualitatively changed the faculty of the human brain for modeling the environment. According to Oakley (1985) human language takes the highest place in the evolution of cognition, and with this language did not only enable humans to construct much more precise and detailed mental models of their environment, but according to Csányi (1992) it also made the parallel connection of the modeling brains possible, which resulted in the construction of a supermodel that was not based on the experience of a single individual any more, but was a product of a group of individuals being continuously and mutually in touch. The function of the brain of any animal (including humans) is to construct and operate such models for the sake of adaptation to and survival in the environment. But for the construction of this model non-human animals have to rely on their individual experiences, because no animal’s communication is powerful enough to be able to exchange parts of their models to any significant degree (cf. Csányi 1992). Though the cognitive-modeling ability in organisms existed already prior to language, the appearance of language can be explained by its vast social advantage, which in terms of group selection also meant a selective advantage. The social advantage of language does not simply mean the role of information transfer that communication plays in animals, but it means the advantage that the communication of mental models can have for a group.

Via this supermodel members of a population could share each other’s individual experience to a degree unmatched in other animals. Such a supermodel means that all members of a speech community can benefit from the knowledge of others, even from that of previous generations, without direct individual experience. This unique capacity can be described as symbolic cognition (Győri 2001). Symbols of language carry information about reality and can substitute such experience. It is quite evident that the larger part of our knowledge that we possess in a linguistic form is not of empirical character and many symbolic structures do not even qualify as representing something that can be physically experienced.

Linguistic communication made it possible for the individuals


of a group to have similar knowledge of their environment without physically sharing all the experience. It also enabled the group to construct more and more adequate models of reality, since it secured the control of the adequacy of these models by making the comparison and correction of individual models based on subjective experience possible. The evolutionary role of linguistic communication was thus the diminishing of the subjectivity of individual models, which vastly increased the possibility of effective cooperation. Neither the increased communicative potential nor the increased modeling capacity in themselves can account for the emergence of language. Without being linked to symbols that bear a cognitive-representational function in the mind communication can increase in volume and scope but not in efficiency. The same goes for cognition. Its efficiency cannot be changed qualitatively unless it is freed from being locked up in the mind of the individual. That is why language could only have emerged as the ability to manipulate cognitive symbols that can be used both externally in communication and internally in cognitive representation simultaneously. Thus, if language evolved with the function of making the cooperative interaction of humans with their environment more effective, then there have to be corresponding biological and cognitive constraints embodied in the faculty of language.

Because of this we have to look for the beginning of the evolutionary process of language emergence in the times when the separate biological abilities of communication and cognitive modeling were joined in our ancestors as a result of selection pressure, i.e., when in the interest of survival it became advantageous for the individuals in a group to bring their own cognitive models of their environment based on individual experience to the knowledge of other members of the group. The interest of survival is very important to emphasize as it has often been mentioned as a criticism that language would also be advantageous for other species, still they did not develop a similar system. In the case of humans they must have got into an evolutionary niche where given certain preconditions in the form of a high level of communicative and cognitive abilities, and refined and elaborate social structure the embarking on the road towards language offered itself as an evolutionary solution for survival. It


seems that language is the product of a selection process for cognitive adaptation in the social interaction of evolving hominids for sharing and representing individually acquired knowledge (cf. Steele 1989). 3.3 Problems with the origin of human cognition According to Ulric Neisser (1976: 1), “[c]ognition is the activity of knowing: the acquisition, organization and use of knowledge.” I believe that almost everybody using the term ‘cognition’ would more or less agree on such a general definition. However, within this frame the term ‘cognition’ is used in two basically distinct senses in the literature and these two senses imply two different interpretations of what cognition is.

One of the senses is mostly used in the so-called cognitive sciences inspired by the philosophical traditions of rationalism and Cartesianism. In this sense cognition is the specific human mental capacity of true reason, which is the only way of acquiring real knowledge about the world. This view presupposes a metaphysical world of objective truth. In other words, it is assumed that the physical world is organized and structured according to some transcendent principle that exists independently of the human mind. Cognition, then, has nothing to do with the physiological perception of external reality and processing that information in the mind, but is the working of human rationality in order to find out about this ‘higher’ truth. More specifically, this mentalistic idea considers cognition as a process of formulating this objective truth about the world in propositional form. Since propositional thinking is dependent on the human capacity for creative language use, cognition is proclaimed to be the sole privilege of humans. According to this view cognition is independent of the physical make up and biology of any organism, since there is only one true way of knowing reality (Fodor 1987). This does not necessarily mean that cognition has nothing to do with the biology of humans. Today’s advocates of rationalism and Cartesianism are of course not dualists in the same sense as Plato and Descartes were. They are explicit about the fact that humans are endowed with a unique biological capacity for rationality and creativity, which is manifest in our “combinatorial disposition” (kombinatorische Disposition)


(Bierwisch 1992: 36) or our ability to handle “discrete infinity” (Chomsky 1988: 169), i.e., manipulate systems of discrete elements in an infinitely creative fashion. It is supposed that a property of this kind forms the basis of the human capacity for generative grammar, which in turn enables us to form propositions about objectively existing truths in the world. However, this biological disposition is said to be without any evolutionary precursors. Another characteristic of such a view is also that cognition, though based on a biological disposition, is not seen as a biological process of the organism in the sense that it pertains to biological interaction with the environment, but some kind of a superior mental activity found only in humans. Chomsky (1988) denies any biological necessity for the structure of human language.

Though the above view is widespread in cognitive psychology, not everybody within the discipline shares it. Neisser (1976: 1) does not intend his definition, mentioned above, to be valid for humans only, but he extends it to other organisms as well. Thus, in the second sense, used in the life sciences and in the more biologically oriented approaches to cognitive psychology, cognition means knowing the world in a way that is permitted and at the same time required by the biological functions of an organism.

Both views of cognition agree on the fact that humans are capable of cognition on the basis of some biological endowment because humans have an organ, namely the brain, that serves as the physiological basis of cognition. But this seeming agreement conceals very different implications. The main controversy between the two conceptions is about the function of cognition. The rationalist assumption is that cognition is a higher order mental function of the human brain, practically equal to human reason. To say that cognition is a biological function, on the other hand, implies the view that the human brain is just one kind of adaptation for cognition and is phylogenetically related to other kinds. The brain as an organ is the product of evolution, namely with the function of building a model of the environment for facilitating the organism’s interaction with it, i.e., to process and store incoming information and coordinate the organism’s behavior accordingly (Csányi 1992). A dishomology corresponding to the difference between linguistic and non-linguistic communication does not exist at the level of neural organization (Deacon 1988) because the human brain shares its structure, even the columnar organization of the neocortex, with other mammals


(Szentágothai 1978; Rockel et al. 1980). Since this proves that it must in all structural respects be a product of evolution, it is difficult to see why this should not be so functionally, at least in the respect of being a model building adaptation. Thus, language can only be said to be a completely new function of the human brain in the sense that it is a new mode of the brain for building a model of the environment. Phylogenetic relatedness in a functional respect means that any function must have other functions as prerequisites and forerunners. Language has as its prerequisites and forerunners the model building mechanisms that preceded it in cognitive evolution, and we should not forget that perception itself, as an active structuring process, is already a way of modeling reality.

Furthermore, not only a morphological lateralization can be detected in the brains of non-human primates, but also a functional one (see above Chapter 2). It is clear that at the anatomic-physiological level the difference between human and non-human brains is only quantitative, but this difference manifests itself in a much more complex neuronal interconnectivity of the human brain, which is accompanied by the appearance of new functional centers (Gibson 1988). We can conclude from this that the quantitative difference found at the anatomic-physiological level results in a qualitative difference at the cognitive and behavioral level. This much more complex neuronal interconnectivity is also a genetic endowment, which probably serves as a basis for linguistic ability among others. Combinatorial disposition or discrete infinity, which I understand as the ability to handle open systems, may well be the result of this much more complex neuronal interconnectivity and there is no need to postulate a genetically determined special and independent language faculty that is inexplicable in terms of evolutionary biology.

The mentalistic approach denies this on the basis that real knowledge can only exist in propositional form, i.e., it can be logically true or false, and no other animal than man can acquire and process information about the world in this way. It is of course possible to maintain a terminological difference between the propositional form of human information processing and the ‘inferior’ ways of information processing in non-human organisms by reserving the term ‘cognition’ for the former, but in this way we may not be faithful to the reality of evolution. There is nothing that compels us to define


cognition solely as the mental manipulation of propositional structures but the already mentioned philosophical tradition that preserves the knowledge (or rather the ignorance) prior to modern insight into evolution and the animal mind, claiming that subhuman organisms cannot possess knowledge, or if they do, it has no evolutionary relationship whatsoever to human type of knowledge. However, the ultimate problem with this latter approach is that as a result of its dualistic and metaphysical attitude it denies the importance of interaction between organism and reality in the explanation of cognition. Because of this real human cognition is supposed to be based on rationality, which is the alleged capacity of bringing internal knowledge structures into correspondence with external reality without requiring any kind of biological interaction with the environment. 3.4 Symbolic cognition as the adaptive advantage of language Here again the basic issue is that of mental continuity or discontinuity. The common belief about evolutionary continuity and discontinuity in discussions about human and animal traits especially in connection with language and cognition is that a unique capacity cannot be continuous with traits that appeared earlier in evolution simply because of the qualitative difference that exists between them (e.g., Corballis 1991). However, the existence of unique forms and traits that constitute qualitative differences in a cross-section view of evolution does by no means contradict evolutionary continuity (Győri 1995, 1999; see also previous chapter). Each different species is constituted by its unique form and traits, but there are not many (e.g. Gould 1980) who would deny the fact that continuous evolution plays an important role in speciation in general. But even Gould’s idea of punctuated equilibria does not postulate a total ex nihil emergence. Thus, it is impossible that the human mind should have absolutely no evolutionary forerunners in animal traits. In spite of the problems and difficulties involved in a continuity theory, it is the only sound approach to take (Donald 1991: 24).

Unless we want to hang on to the metaphysical idea that there is some objectively given truth about the world that humans are destined to find out, it seems to be permissible to speak about cognition in a


general form, simply because the acquisition, organization and application of knowledge/information is a biological function that is characteristic of all living organisms in one form or another. Even though cognition can take different forms, we can speak about the evolution of cognition as such, because cognition is not only characteristic of all living organisms, but it is an indispensable condition of life, since this is the fundamental basis of an organism’s contact with its environment. And if organisms can evolve into one another giving rise to new species and thus to qualitative differences, why should their capacities not do the same.

The realization that cognition is a biological function of living organisms in general calls for an investigation of the issue of what place human cognitive capacities take in the evolution of cognition. Even if we put the issue of evolutionary continuity and discontinuity aside, human cognition must be a product of evolution in one way or another. This makes it important for a correct conception of human cognition to emphasize its biological character, and with this the fact that not even human cognition can be a purely rational activity but must at some level be based on an interaction with reality. This is the only way the human capacity for knowledge, including propositional type of knowledge, can be accounted for in a scientific way (cf. Plotkin 1994).

In the following I am going to examine the evolutionary differences in cognitive functioning that characterize different living organisms. All organisms exhibit the cognitive functioning of picking up information from the environment, processing that information, and changing their behavior in accordance with that information in order to enhance their average probability of survival in a given system (Csányi 1989: 205). Evolution has produced two basic levels of this acquisition, organization and application of knowledge/information in organisms. For the first organisms to appear in the course of evolution it was genetically determined, i.e., hard-wired, what kind of information they could acquire, how they could organize this information and how they could apply it. In the case of this type of information processing, when performance is based on gene expression, it is customary to speak about a genetic memory (Csányi 1988) or genetic knowledge (Plotkin 1994). This is the information that has become genetically coded in the genom of the individuals of a species during evolution. This genetic level of information processing


was later supplemented by a completely new type of cognitive processing when later on in the course of evolution organisms evolved that were equipped with a nervous system. “The nervous system is the special organ of adaptation that evolved to adjust the organism to rapid changes” in the environment (Csányi 1988: 299). An organism with a nervous system became capable of a more flexible behavior because individual experience began to play a much greater role in the modification of its behavior. Thus, the possibility for the acquisition of individual knowledge became greatly enhanced in a direct proportion to the complexity of the nervous system. When the neural mode of cognition began to supplement and interact with genetic cognition, the adaptive flexibility of the organism became the function of the proportion of the two types of cognitive activities in an organism, because this determined the amount and type of experience that could be processed.

Csányi (1988) has described how these two cognitive mechanisms (genetic and neural memory) contribute to behavioral adaptation in animals. Human beings are no exceptions to the fact that their behavior can be described as the outcome of genetic and neural cognitive mechanisms, although they probably rely most heavily on neural memory among all animals. In spite of this, the nature of human cognitive systems appears to be qualitative different from that found in other animals. Just as there is a qualitative difference in the way of the acquisition of knowledge in purely genetically established cognitive systems and in experience elaborated ones based on a nervous system, humans demonstrate yet another evolutionary innovation in knowledge acquisition. Organisms with an experience-elaborated cognitive system acquire knowledge almost exclusively through sensory experience gathered individually. The qualitative difference of human cognitive activities is characterized by the fact that humans can acquire knowledge without direct physical experience. This type of cognition without experience is a result of human linguistic capacities, which is a sophisticated extension of the neural mode of cognition. Symbols of language carry information about reality and symbolic communication is capable of substituting direct experience. Humans gather by far the larger part of their knowledge from symbols in this way (see Figure 3). However, as we shall see later, even human symbolic knowledge is rooted in our interactive behavior with our environment.


Figure 3: Schematic diagram of the amount of various types of knowledge in organisms of different complexity

It is quite evident that the larger part of our knowledge that we

possess in a linguistic form is not of empirical character and many symbolic structures do not even qualify as representing something that can be physically experienced. But even in the case of symbolic structures that represent parts of perceivable reality, the type of knowledge contained in these structures differs in qualitative ways, because of the use of symbols in our cognitive processes, from the simple neural type of cognitive processing of physical experience. On this basis, we can distinguish a third type of cognition that has only evolved in humans and human beings can thus rightfully be said to have symbolic cognitive systems. This does not mean, however, that earlier evolved types of cognitive activity are not found in humans. They are, though to a very large extent, only supplemented by the symbolic one and interact with it in basic ways. We must not forget that symbolic structures are after all based on neural mechanisms. But this does not apply to the physiological level alone. As we shall see later, symbolic cognition is ultimately grounded in more basic genetic and experience-determined cognitive structures.

Chapter 4

The emergence of referential behavior 4.1 Reference in a biological perspective Reference, the cognitive capacity of referring symbols via mental representations to phenomena in the world, is considered to be at the very basis of linguistic behavior. Obviously, the high efficiency of linguistic information processing is due to this capacity. A biological-cognitive approach implies that reference is viewed as a cognitive capacity having emerged as a result of biological evolution and thus characteristic of certain biological species. Such an approach to cognition suggests an evolutionary continuity between non-human and human types of information processing, linguistic information processing representing the highest level of cognitive activities among organisms. As it has been shown in Chapter 2, evolutionary continuity between non-human communicative and cognitive behavior and human linguistic capacity is very well consistent with their difference in kind. This qualitative difference is manifest in the combined and indivisible communicative and cognitive function of language.

Working in an evolutionary framework means approaching language as a product of biological evolution. Thus, we can view language as an evolutionary innovation of combining communication and cognition indivisibly in one symbol system. In other words, with the emergence of language not only symbolic communication, but also symbolic cognition was achieved. However, it has been suggested that other species also exhibit at least rudiments of such a referential capacity (e.g. Menzel and Johnson 1976; Green and Marler 1979; Seyfarth, Cheney and Marler 1980b).

For a long time phenomena like mind and cognitive processes, e.g. decision making and categorization, were thought to be linked


exclusively to being human, or maybe directly to the capacity for language. With the decline of behaviorism, subdisciplines like animal psychology, cognitive ethology and neuroethology began to spread the view, based on experimental evidence, that cognition should be considered a common biological property of animals, including both humans and non-humans. The nervous system, and especially the brain in higher animals, is an evolutionary adaptation for organizing, storing and processing incoming information for the sake of orientation in the environment. This is done by constructing models of the environment and cognition is just this ability to construct such models (Csányi 1989, 1992). It has been shown that animals, especially birds and mammals, are capable of categorizing objects in their environment (Cheney 1984; Herrnstein 1985; Terrace 1985; Pepperberg 1987). Herrnstein (1984) considers the ability to categorize so fundamental among animals that he attributes adaptive value to it. Griffin (1978) has suggested for cognitive ethological studies that the communication systems of animals be viewed as a window on their cognition, just as human verbal communication, i.e., linguistic behavior, gives us insight on human cognition.

Earlier not much thought was given to the difference between animal communication and human language, since the difference, though not formulated explicitly, seemed to be obvious. But now, given the fact of at least in a rudimentary way, similar cognitive processes in animals and humans, it becomes inevitable to ask what makes their communicative behavior different. One answer to this question has long been that the uniqueness of human language lies in its referential function. The signs of language refer to things and events and other phenomena in the world, whereas the signs used by animals in their communication do not. This is also the source of the distinction between animal signals and human symbols. There are two sources for comparison of human language with animal communication. One is animal communication in its natural environment and the other is the ape language teaching experiments. In both cases the question is whether there is any trace of reference or referential ability in the communicative behavior of animals.

The emergence of referential behavior 89

4.2 Comparing animal communication and human language For a long time it was almost an axiom in the study of animal communication that animal signals are merely expressions or symptoms of the internal states of the animals. This idea presupposed that animal signals were ‘hard-wired,’ i.e., genetically determined. Today we know that especially species higher on the evolutionary ladder acquire parts of their signal repertoire through learning after birth. This has been demonstrated especially in certain species of birds, along with a cerebral asymmetry of vocal control (Marler 1975; Nottebohm 1977), and monkeys (Seyfarth, Cheney and Marler 1980a). But the idea of animal communication as expression of internal states still has its impact, and Passingham (1982) sees this as the main difference to human language, because the signs of language refer to phenomena in the external world and thus have referents. The reason why internal states are not considered referents of animal signals is that the traditional view says that these signals are not controlled voluntarily.

However, new data indicate that especially in the case of non-human primates, communication does not completely rely on the limbic system, but partly also on the neocortex, which suggests that these animals are capable, at least to a certain degree, of the voluntary control of their communicative behavior (Steklis 1985). The existence of acquired signs and voluntary control of signal emission really suggests a certain kind of referential function of signals. Green and Marler (1979) even go a step further and claim that occasionally and only to a certain degree even the linking of signals to new referents can be found in animal communication systems, which phenomenon is said to be one of the basic criteria of openness, the fundamental characteristic of human language. Green (1975: 312) has found “a degree of openness in the vocal communication system [of Japanese macaques], which allows the spread of a new signal pattern within a single generation, a similar time span to the occurrence of the new phenomenon with which it is associated.” Menzel and Johnson (1976) take a broad view of reference and consider symbolicity fundamental to and existent in all kinds of communication and defend the view of the referential function of animal signals. They write: “The ability of nonverbal animals to ‘tell’ each other the nature and location of objects and


events is limited only by the richness of a signaler’s behavioral organization and a receiver’s knowledge of the signaler and the common environment in which they are operating” (Menzel and Johnson 1976: 139-140)

A third alternative of interpretation of the communicative behavior of animals has been presented by Dawkins and Krebs (1978). It differs from both the symptomatic (expression of internal states) and the referential view in the sense that Dawkins and Krebs deny the possibility of any informational interpretation of animal communication. In their view the emission of animal signals is manipulative behavior. The only means of communication where Dawkins and Krebs (1978) allow for a flow of information is human language, where the end result is that the receiver learns something which he did not know before, from the sender. On the other hand, they consider animal communication a manipulative form of behavior, where signal transmission is a means for exploiting the physiology and anatomy of another individual, i.e., signaling serves the purpose of achieving certain goals through influencing and modifying the behavior of other animals. Because of this, in the case of animal communication, they speak about actor and reactor instead of sender and receiver of signals.

I think that all three interpretations of animal communication have some truth in them. Information cannot be excluded from animal communication on the basis that the sender is actually not sending the information. If we think of the information content of a signal as a message, the signal transmitter’s intention to inform is indeed doubtful. But there is no need for an active transmitter in order for us to be able to speak about information, because it is the function of perception in general to pick up information from the environment. In my view, all states and motions of the environment that bear some relevance to an organism's orientation and that the organism is capable of perceiving are information for the organism. The informational status of events actually depends on their role in the orientational behavior of a given organism. In a broad sense, all perception is information, because everything that is perceived is processed in the nervous system of the organism. In a narrow sense, only the orientationally and adaptively relevant perceptions count as information. On such a view it is true for all perception, not only for linguistic communication, that the receiver (or perceiver) learns


something which he did not know before. Because of this, language must be much more than just an information transmitting device.

In a certain sense there is also reference in all kinds of signal transmission. Reference in its broadest sense can be considered a relationship between a signal and a phenomenon it is related to, on the basis that a receiver of the signal can relate it to this certain phenomenon. Even if an alarm call is not emitted for the sake of warning conspecifics but as a symptom of the internal state of fear, the receiver relates it to certain events, actions or other phenomena, be it in the form of reacting with a preprogrammed behavioral pattern. But as we move higher on the evolutionary ladder, the function of signals as elicitors of such preprogrammed patterns is not so unambiguous. It has been shown that vervet monkeys learn their alarm calls from their conspecifics and relate the different calls to different predators in the process of learning these calls (Seyfarth, Cheney and Marler 1980b). During this process of sign learning the monkeys also learn how to behave in connection with the approach of different predators. This of course does not mean that the signal is a symbol of the predator. However, the referential relationship between a certain call and the receiver’s knowledge concerning a certain type of predator is evident from the receiver’s performance of a specific action when hearing a specific call related to a specific predator. Here the emission of the different signals cannot be related simply to an internal state of fear, since the same state of fear must be present in the animal whatever predator may approach. The type of signal the sending animal emits depends entirely on outer circumstances and the receiving animal reacts in accordance with these circumstances.

Thus, I think that after all it is safe to say that animal communication is a process of transmitting and receiving information. Of course it is probable that most of the time the units of information called signals relate to internal states of the transmitter. However, this should not mean that a referential interpretation is always incorrect. Internal states do not come out of nothing, as no organism can be separated from and viewed apart from its environment. All internal states of organisms are causally linked to states and motions of the environment and thus any signal expressing an internal state is indirectly related also to external phenomena. For the receiver any such signal refers to something,


either indirectly to some external phenomenon (as in the case of vervet monkey alarm calls), or directly to the internal state of the other individual, as he interprets the signal by reacting to it in some way.

Claiming the above does not mean either that there is no place for manipulation in the interpretation of animal communication. All animals react to the signals of their conspecifics and their reaction can be viewed as a result of influence from other individuals. At this point it turns out that the three interpretations of animal communication are actually three ways of looking at it: (1) The informational aspect looks at the transmitter as either actively or passively, i.e., voluntarily or involuntarily, modifying the environment (e.g. in the form of causing particles of the air to vibrate, even if this is only a symptom of internal states) and at the receiver as perceiving this modification (maybe in fact as a signal) and interpreting it for the purposes of orientation. (2) The referential aspect looks at the receiver of the signal and at his mode of interpretation as to what action to take upon perception of the signal. Taking a broad view, we can say that the way to interpret a signal can also be ‘hard-wired’ in the genom of the animal. However, young vervet monkeys watch their elders, first in order to learn how to behave in the case of the different signals, which already implies a narrower sense of the term interpretation. (3) The manipulative aspect looks at the result of the signal transmission, which is always the elicitation of some kind of behavior in the receiver, or indeed the reactor, acting upon some environmental influence, which in this case is the signaling behavior of some other individual, i.e., the actor. It must be clear that the three aspects complement one another in a global approach to animal communication instead of representing alternatives. 4.3 Artificial language in apes as an indicator of cognitive capacities Another area where we can explore what animals are capable of in connection with referential sign usage is experimental sign instruction. Such instruction has been conducted in an array of experiments usually referred to as ‘ape language teaching


experiments’ or ‘ape language projects.’ These experiments have shown that apes are capable of doing something that has been hitherto thought of as one of the bases of linguistic capacity. They can associate signs (though mainly visual and not acoustic ones) with objects in the external world. It seems to be trivial to remark that apes, just like humans, actually do not associate the signs with the objects themselves, but with mental categories, otherwise they could not use the signs in referring to different instances of the same class. On the surface this achievement seems to be similar to that of young children learning their mother tongue, in which case we find it only natural that they should be able to use e.g. the word apple for all apples after having learnt it in the presence of one instance of an apple.

However, the big achievement of the apes here is not that they can learn the category apple, because as we have seen, this capacity is surely present in primates and even necessary for survival in their natural habitats. What seems to be of greater interest is the pure fact of the associative learning of the signs and their appropriate usage, because as Premack (1985: 226) says, “there is no evidence that concepts, previously unknown to the animal, were introduced by language.” Thus, the real question seems to be whether associative labeling in apes is equivalent to representational naming in children (Savage-Rumbaugh, Rumbaugh and Boysen 1980). But what is the difference between the two? Associative labeling is attaching an arbitrary label to a preexistent category while leaving the structure of the category intact. It is clear that apes are capable of this. It is also clear that language acquisition is more than this. Representational naming as sign acquisition probably parallels concept formation, thus creating concepts; in the case of already preexistent nonverbal categories, it affects their structure. Strangely enough, Chomsky (1988) considers word learning in human children an associative process, but in his view the concepts labeled are unique to the human species as they are part of our genetic endowment. Anyway, if associative learning is part of our linguistic ability, then the apes’ achievement in this respect falls short of ours only in degree. But whereas the ‘words’ of language trained apes are in fact labels for concepts, this does not exactly seem to be the case with words in a human language. They seem to have quite a different function from just labeling concepts. Language acquisition is not


just a learning of labels for innately given concepts as Chomsky (1988) thinks. It is the construction of a conceptual system, which means that there is a level of representation of concepts (see below) where concepts are in fact introduced with language learning.

The issue of fundamental concern is indeed whether associative labeling in apes is equivalent to representational naming in children (Savage-Rumbaugh, Rumbaugh and Boysen 1980). But the ability of cognitive representation is not confined to humans. Animals also represent their environment mentally and they do so without language (Terrace 1985). What Savage-Rumbaugh et al. are concerned with is not simply represent-ational thought, but symbolic representation. However, non-linguistic representational thought characteristic of all higher organisms, also humans, provides only the ability to represent the world to the self. That is why Savage-Rumbaugh et al. consider true symbolic representation to be interindividual representation. But here the old problem crops up again: Interindividual representation implies that the function of language is communication.

It is hard to see why associative labeling could not serve interindividual representation. It serves the purpose very well in artificial languages. But this is exactly the difference between an artificial language and a natural language (or human languages). Although humans also have mental representations that have nothing to do with language, it is a characteristic feature of human thought that it is mainly linguistic, or in other words, it is mostly symbolic. Humans use the symbols of language even for individual representation. Thus, symbolic representation is just as much individual as interindividual. And this is just the point. There is surely no individual symbolic representation in animal minds. Whereas animal signs have only communicative and no cognitive function, words in a human language have both. The words of humans have a completely different status in the mind from the ‘words’ of language trained apes.

Premack draws the following conclusion from his data obtained from experiments with apes:

... the word is not the source of the abstractness. On the contrary, the preverbal concept is the source of the abstractness. The word must inevitably contaminate that


abstractness, for the word cannot be associated with all possible exemplars – only with specific ones. (Premack 1985: 274)

This is probably the case with ape ‘words,’ because associative labeling does not affect the concept. With only a label attached the concept will keep its structure and representational level. It will remain at the basic level of representation (cf. Rosch 1978) and based on gestalt perception, motor activity and behavioral function. These factors are probably the ones that also apes make use of in mental categorization. It is probably not by chance either that first they also learn basic level categories, and it may even be the case that the very first step in word learning is some kind of associative labeling. But human words very soon change their status from labels to representations of the concepts, and by this free the concepts from perception-boundedness. In this way new and even perception free, i.e., purely logical or imaginary, categorizations become possible by metaphorical extension, which is just the characteristic human way of conceptualizing our environment (Johnson 1987; Lakoff 1987). The ape’s ability to attach labels to concepts is of course an enormous achievement, but language hardly serves as a vehicle apple of thought and means of cognition for the ape. Or does the ape, after learning the name of the concept, think of the concept by the symbol? Probably not, because as Premack (1985: 255) puts it : “language does not enter the ape’s mind.” Because thinking of the concept by the symbol is just what humans do and here exactly lies the difference between the name for the concept and the word. Words are not signs for mental representations, but mental representations themselves. One would not have to operate with words in one’s own thought if words were only signs for mental representations, because mental representations in themselves are just as well available to the self.

Naming a concept does not make its representation symbolic. The representation of the concept remains to be based on the modes of perception, i.e., the concept of a class of phenomena is made up of the knowledge of the typical shape, color, smell, taste, etc., but also motion, usage, relations, etc. of the phenomena. This must be the reason for the fact that labeling a concept does not add to its abstractness. The label simply does not influence the representation


of the concept and is probably only used when the concept has to be externalized, i.e., communicated. I agree with Premack when he assumes that the fact that an animal can use the sign for OPEN to situations different from the one in which it learnt the OPEN sign, does by no means mean that it was the sign that enabled the animal to do so and it did not have access to this level of abstraction before acquiring the name (Premack 1985: 273). The concept of OPENNESS based on the perception of e.g. an open door is abstract enough to apply to open drawers, open books or even open mouths (all Premack’s examples), but probably not abstract enough to apply to an open shop (with its door closed), not to mention such notions as open mind, open heart, etc. The levels of abstraction in these latter examples can only be reached with the help of words, probably because symbolic representation diminishes perception-boundedness, and thus words after all can increase the abstractness of concepts, whereas simple labels for them cannot, labels are only translations of concepts for the purpose of communication and not inherent to the thought process of the animal.

Unfortunately, very little is known about the nature of words and their status in the human mind, but there surely exists a difference between words and simple labels for concepts. It is probably this difference that also enables us to have knowledge without experience. With the help of words we can have representations of concepts of phenomena that have never been object to our perception (e.g. solar system), or that do not exist (e.g. ghost). We can make non-natural categorization, i.e., categorize according to features that are not accessible to perception (e.g. scientist). This implies that also words that have underlying concepts the features of which are open to perception have a cognitive status different from names or labels. Even though the perceptual features of the concept CHAIR (a piece of furniture) do not seem to allow it, one can chair a meeting while standing, or in an informal situation, sitting on the table. Despite the seeming contradiction, the step of abstraction could freely be made at one time in the history of the English language. The extended meaning is by far not arbitrary and senseless, but rests on features that are only represented in the linguistic concept/category CHAIR, but not in the preverbal concept CHAIR, which is based on gestalt perception, motor activity and behavioral function. The features utilized in the metaphorical


extension cannot be added to the preverbal concept simply by labeling it. This is what I would call the real productivity of language: not the inclusion of different phenomena into the same concept at the same level of abstractness, but the possibility of ever increasing abstractness, i.e., creating new concepts from old ones.

Speaking of ape language projects, another aspect of language must also be mentioned. It has been emphasized previously that language is a two-faced phenomenon. While it is a biological capacity on the one hand, it is a system of structures on the other. However, the capacity side and the system side cannot be separated since it would be nonsensical to say that someone possesses language without knowing any signs, rules etc. of a particular language. This is because the only way we can draw conclusions about the linguistic capacity, is from the way of handling the system. Though it is the biological capacity that is supposed to be unique to humans, any claims for the uniqueness of human language have to be concerned with its structural side. That is why in the ape language projects the apes’ linguistic capacity is measured in terms of their production of certain structural features of human language. It has been found quite soon after starting the ape language projects that apes can learn a limited vocabulary of signs in the visual channel. They can learn to use arbitrary signs to refer to external stimuli and they can also understand such signs when used in this way. In a certain sense this resembles human linguistic communication. However, beside the referential usage of signs language has another important property. Language does not only use referential rules that tell you which signs refer to which stimuli, but also syntactic rules that tell you how to combine these signs to refer to complex phenomena.

Thus, it has often been claimed that grammar is a condition for human language. Terrace et al. (1979), after analyzing the data gathered from their project Nim and by other ape language projects, have found no evidence of an ape’s ability to use grammar. They have denied the existence of grammatical competence in apes on the grounds that apes seem to lack any syntactic ability, since their use of ‘word order’ is inconsequent. However, instead of pure word order special emphasis should be laid on the teaching of syntactic rules in accordance with semantic relations, because most syntactic rules affect sentence meaning in some way. Then we really could


learn something about the representation of particular semantic relations in the ape, such as actor, action, object, recipient etc., which Terrace et al. miss. Premack has carried out experiments of this kind with Sarah and concluded that such representations are not totally alien to the ape (Premack 1985: 279). Experiments prove that chimpanzees (Pan troglodytes) are capable of expressing spatial relations with the use of correctly ordered prepositional structures (Fouts and Couch 1976).

More recently bonobos (Pan paniscus) have been used in ape language experiments. Savage-Rumbaugh et al. (1993) have shown that they have an even greater affinity for grammatical relations. Their subject Kanzi acquired rudimentary language skills comparable to those of a two-year-old human child by the age of eight and was able to understand novel sequences of words. In tests he even scored better on certain aspects of syntax, especially with respect to the comprehension of word recursion. 4.4 Biological and cognitive bases of linguistic reference The above does not mean that conceptual categories cannot be represented prior to the learning of some kind of language or without language. If apes are capable of mentally representing categories without language, humans must be, too. But this type of representation occurs at a level different from the level of representation of linguistic categories and is phylogenetically older and thus an inheritance from prehuman times. At this level even the existence of innate ‘concepts’ is a fact. Neurobiologists and neuro-ethologists speak of a genetic memory, which is phylogenetically much older than neural memory, and most animal species recognize e.g. their predators with the help of genetically given conceptual categories (Csányi 1988). Such genetically given conceptual categories are also present in humans. The best studied example of this is the universality of the human perception of color categories as different qualities, whereas color on the physical level can only be defined by quantitative differences on a spectrum (Bornstein 1987). While color categories are an example for genetically given categories, probably most categories in humans are learnt by experience. The reason why we can speak of prelinguistic


categorization is that even without language we would perceive our environment in categories, whether genetically given or learnt, and represent it in a categorical manner, just as other primates do. But at this level of representation we speak of sensory representation as it depends directly on perception.

Harnad (1987) distinguishes three types of representations: iconic, categorical and symbolic. The sensory representations I have described above are in Harnad’s terms categorical representations, which function as category-specific feature detectors, and iconic representations, which are the analog sensory projections of objects. The term iconic is not to be understood as relating to visual images only. Iconic representations are direct reflections of concrete objects and/or their characteristics in the brain in any of the perceptual modalities. These two types of representations are of course characteristic of both human and non-human brains. Neither humans nor animals could survive without perceiving similarities between individual entities according to their shape, smell, taste, color, etc., and motion and even function. It is thus clear that categorical representation must be present also in animal brains. Some famous experimental evidence for this was supplied by Rensch (1968). The chimpanzee Julia was trained to remove a screw with a piece of T-shaped iron having a sharp edge and serving as a screwdriver. When later on she was presented with unfamiliar tools, among them externally similar ones to the T-shaped iron but without a sharp edge, and real screwdrivers, she chose one of the screwdrivers without hesitation. She was capable of recognizing that particular object as belonging to the category of screwdrivers even though most of its external features were different from the object she had learnt to use for removing screws. This also shows that the similarities that the categorization is based on cannot be just of any kind. They must be functional. The adaptive value of categorization (Herrnstein 1984) means that similarities are recognized according to their relevance to the animal's orientation in its environment.

It is clear that categorical representations are anchored in iconic representations, because it is the perception of individual objects that the category-specific feature detector works on (Harnad 1987). Harnad (1990) claims that symbolic representations are grounded in these two types of perception-based representations. These concern us most because such representations underlie language and are thus


the basis for its referential function. Coming back now to our original problem, we can formulate our question in this way: Do apes in the language teaching experiments have a symbolic representation after acquiring a sign? If we were to answer this question from the traditional point of view in cognitive psychology, which advocated the idea that word acquisition is nothing more than associating a sound sequence with a concept, the answer should be in the affirmative. But is symbolic representation really reached simply by attaching a label to iconic and categorical representations, which primate brains are also capable of? This would mean that language has absolutely no effect on our conceptual system. This view has now been overcome by the new cognitive perspective on language (Lakoff and Johnson 1980; Lakoff 1987; Johnson 1987; Langacker 1987), and it is clear that language helps us construct categories at a new level of conceptual categorization.

At this level symbolic representations make it possible for us to learn by description (Harnad 1987). While the acquisition of categories at the level of categorical representation requires direct sensory experience, at the level of symbolic representation categories can be constructed by symbol composition alone (i.e., a description system) (Harnad 1990). At this point one aspect of the true referential function of language becomes obvious. While reference in associative labeling is the simple and static denotation of a category by any arbitrary sign, real linguistic reference creatively and in a flexible way facilitates cognition and concept formation. This is most obvious in metaphorical processes in language where one category as source domain can stand for another category, the target domain, on the basis of some kind of analogy.

In the case of such a claim it is important to define terms like metaphor and analogy. A metaphor is always based on the recognition of an analogy, which is a relationship between two entities. This relationship, however, is not entirely objective. It is a subjective evaluation in the sense that it appears at the level of perception as the recognition of any kind of functional similarity between phenomena (Holyoak 1984). Thus, it is basic to all kinds of categorization. Analogy turns into a metaphor in linguistic categorization, where the symbolic representation of a conceptual category is used to denote another conceptual category by highlighting particular features of it.


Until now I have spoken of levels of representation and said that at the level of symbolic representation we can construct new concepts without direct sensory experience, though indirectly relying on the sensory levels of representation. This grounding of symbols on the psychophysical level (Harnad 1990) is reflected in the embodiment of language (Lakoff 1987; Johnson 1987). But when speaking about new conceptual categories, we have to specify the level of these categorizations. It is hard not to recognize a connection between the levels of representation and the levels of categorization. Since iconic and categorical representations are dependent on sensory experience, it seems to be obvious that the phenomena they represent are basic objects. According to Rosch (1978: 29) basic objects are “information-rich bundles of perceptual and functional attributes [in the perceived world] that form natural discontinuities,” and “basic cuts in categorization are made at these discontinuities.” The categories formed at these cuts are the basic level categories, which are characterized by incorporating mainly perceptual features. Because of this, basic level categories can be represented also without language in apes and humans alike, at the level of categorical representation. The difference or similarity in the precision, scope and way of representation of these categories in apes and humans, e.g. whether they are prototypical or not, is another matter. But these are the categories that apes, too, can attach labels to, and thus refer to. Humans, on the other hand, retaining their categorical representations of basic objects, also have symbolic representations of these basic level categories. When a basic level category is raised to the level of symbolic representation with the help of a word, the word is not just a label, but a container of encyclopedic knowledge about the category (cf. Haiman 1980). A symbolic representation, i.e., a word (or linguistic sign), not only refers to perceptual features of a category as, a categorical representation does, but to non-perceptual ones as well.

Basic objects can be represented at three levels. We can have an iconic representation of an individual object, but at the same time we also see that it is the instance of some category and thus have a categorical representation of it. All this is dependent on our perception of the basic object, either directly or by recollection of its features from memory. How we represent it symbolically in the form of a linguistic sign has been described above. From the way basic


level categories are represented symbolically, another highly relevant characteristic of language follows. Since symbolic representations can contain non-perceptual features of categories, language must be able to refer to categorizations that are not in concordance with natural discontinuities. Thus with the help of language we can construct categories other than basic level ones, i.e., different super- and subordinate categories, whose features are non-perceptual. The categorization of FURNITURE draws upon some culture-based knowledge of housing, while the symbolic representation of KITCHEN-CHAIR is anchored in the knowledge that some chairs, not necessarily perceptually different from other chairs, are used in kitchens. Non-perceptual features are those that imply other knowledge than can be gained through sensory (visual, olfactory etc.) information or perception of motion. With the help of words we can represent non-perceptual information about basic level categories as well as construct super- and subordinate categories often dependent on knowledge of cultural traditions, etc., and not only on perception. Symbolic representation is capable of even more than that. Through it we can also construct and represent categories that we have invented or imagined (e.g. CULTURE, SOCIETY or MARTIAN). This can be done because language as a symbolically represented cognitive model (see below) frees cognition from being directly bound to perception.

Because they lack this linguistic kind of symbolic capacity in all probability, it seems to be impossible that apes will ever be able to use labels to refer to such categories. The more so, because it is highly probable that such categories can only be represented symbolically, i.e., they can only exist in symbolic representations, which means that they are indivisibly linked to a symbol, a linguistic sign. What is, however, very intriguing is that the apes in the experiments mentioned above could use signs for non-basic level categories. Interestingly, some superordinate categories were also mastered by apes, though much less easily. Premack’s (1985) ape Sarah mastered signs for categories like FRUIT, CANDY and BREADSTUFFS, and the two chimps trained by Savage-Rumbaugh (1986), Sherman and Austin, had no trouble in labeling tools and fruit. At the present stage of our knowledge we can explain this in two different ways. One way would be to say that knowledge of function (the basis for the categorizations mentioned) can occur at


the level of categorical representations. This is not impossible, as in fact Rosch (1978) claims that functional attributes can contribute to natural discontinuities and that functional classification can also lead to basic level categories. In this case Premack would be right and the achievement of the apes thought due to the fact that such categories are present in their minds and just need to be labeled. This explanation would attribute the apes’ performance to pure associative labeling.

The problem with the above explanation is that the categories in question are in fact not basic level categories. This is not to claim that functional attributes are not used in basic level categorization, but in the cases mentioned it is superordinate categories that were differentiated by functional attributes. This fact and the claim of associative labeling would invite the claim that (at least some) superordinate categories can map natural discontinuities and thus can be represented at the categorical level. But what is the functional difference between fruit and, say, vegetables? The other explanation would solve this problem by acknowledging that the apes constructed these categories with the help of signs, but we are still far from proving this. However, the question that has puzzled researchers for a long time now is what such signs mean to the apes themselves. More precisely, the question is whether apes really ‘know’ that signs are signs, i.e., labels for things or for categories of things used to communicate intentions and knowledge concerning these things, i.e., to refer, or whether they just ‘play the game’ in order to earn rewards. There is no reason to doubt Savage-Rumbaugh (1986) when she says that the latter is hardly the case. Her research really seems to present evidence that Sherman and Austin understand the referential meaning of signs and produce them intentionally to communicate messages.

But what is the referential function of words for humans? It is surely not just to communicate messages. If it were, language would be nothing more than the mere translation of our mental models of the world into a modality that is accessible to other individuals. Human language has a very basic cognitive function as well. It is our mental model, because its building blocks, grammatical rules and linguistic signs, are the ‘material’ out of which we create this model. Thus, the referential function of language is not only to refer to phenomena and categories of phenomena in our communication with


other individuals. The individual also relies on language to build and operate his or her own cognitive model of the world. As mentioned earlier, animals also have cognitive models of their environment. But because of language, human cognitive models are constructed in a different way. Csányi’s (1989, 1992) explanation on how animal and human cognitive models differ is very insightful in this respect: The function of the brain of any animal (including humans) is to construct and operate such models for the sake of adaptation to and survival in the environment. But for the construction of this model non-human animals have to rely on their individual experiences, because no animal’s communication is powerful enough to be able to exchange parts of their models to any significant degree. But language makes possible the construction of a ‘supermodel’ that can be shared by all members of a speech community, whereby they benefit from the knowledge of others, even of previous generations, without direct experience. Such a possibility would of course be trivial if we could do so merely because we have agreed on different labels for different concepts.

But words are not labels, they are categorizing descriptions. In a study on naming, Carroll (1985) has shown that almost all naming in human communities happens in a descriptive categorizing manner. And this is actually what linguistic signs do: they refer to phenomena by describing them and categorizing them. Csányi’s supermodel is not a model that is shared by everyone because they have the same genetic endowment and the same experiences, the way that animals share their cognitive structures and are able to communicate about them. But with language new concepts can be planted in the minds of other individuals, thus substituting individual experience. This is far from meaning that there are objective categories that are pointed out by one individual to another. It means that a concept or some cognitive structure constructed actively and subjectively by one individual on the basis of experience can replace experience in other individuals. It is the symbolic representation of concepts in words that makes it possible for humans to store even non-experienced concepts via reference to their symbols. Thus, with words we not only refer to our concepts in communicating with others; we use them for reference to our own concepts also for ourselves when operating our cognitive model, or in other words, when we think, be it in a propositional or metaphorical manner.


Let me conclude this chapter by touching very briefly on the age-old debate about whether we think in or with language or not, because what has been said above also gives us some insight into that problem. The answer that seems to emerge is both yes and no, because human thinking has a double nature, using both phylogenetically older prelinguistic and newer symbolic ways of thinking. Thinking is operating one’s cognitive model and we always do this at the appropriate level of representations. We can manipulate all kinds of iconic or categorical representations in all the perceptual modalities and conceive of different actions with them, such as when solving a jigsaw puzzle, finding the most appropriate moment for crossing a busy road without being driven over, etc., of course without language by the nature of the thing. And we can also manipulate symbolic representations, for which we rely on language most of the time, as most of our symbols are linguistic signs. But we have some other symbols as well, like mathematical ones, which we also manipulate at the level of symbolic representations, just like linguistic signs, though without language, since the corresponding concepts are stored in non-linguistic symbols. (It is another matter that most non-mathematicians probably resort to the linguistic signs for mathematical concepts in their thinking.) The issue is that certain concepts must be linked to symbols, be they linguistic, mathematical, etc., in the mind. However, the real double nature of human thinking is that the operation of one’s cognitive model of the world probably goes on simultaneously most of the time at all representational levels, both at the perceptual and at the symbolic ones.

Part II:

Semantic Change

as

Cognitive Adaptation Process

Chapter 5

The cognitive function of language 5.1 What is cognition? The fact that language is a cognitive adaptation in humans, and as such a product of cognitive evolution, has far-reaching consequences for the cognitive functioning of language and the relationship between language and cognition. For an adequate understanding of this functioning and this relationship it is indispensable to be clear about what cognition in general is and what its general functions are. The view presented so far has been clearly biologically oriented – similarly to the one taken by Plotkin (1994) – for it is my conviction that such an approach has more explanatory value for the way language operates than one inspired by the philosophical traditions of rationalism. Thus, cognition should not be understood here as “the convergence of our ideas and the truth about the world” (Chomsky 1988: 158), but rather in the sense of Neisser’s (1976: 1) already quoted definition – not intended to apply to human beings alone – according to which “[c]ognition is the activity of knowing: the acquisition, organization and use of knowledge.” Cognition should be viewed as a biological adaptation, since the organization and application of information gathered from the environment is in general the fundamental basis of any organism’s contact and interaction with the environment it inhabits (cf. Plotkin 1994: 4).

As already mentioned in Chapter 3, and repeated here for convenience’ sake, this kind of cognitive functioning has two basic biological levels corresponding to the relative stability of the environmental conditions with which the organism interacts (cf. Csányi 1988). For the sake of dealing with particularly stable conditions of the environment a genetically programmed cognitive functioning is the most efficient solution. In other words, it is hard-wired in the organism what kind of information can be acquired, how


it can be organized and how it can be applied. This has been called genetic memory (Csányi 1988) or genetic knowledge (Plotkin 1994: 141). In many less complex organisms this level of cognitive functioning alone controls all of the organism’s behavior. In more complex organisms much of the interaction with the environment is controlled by a nervous system (or a brain) in addition to genetic cognition. The nervous system is a cognitive adaptation to more rapidly changing environmental conditions because it enables the organism to accumulate knowledge obtained in its lifetime through individual experience and to store and process this information in order to coordinate its behavior in a flexible way (in proportion to the complexity of its nervous system) (Bonner 1980: 137; Csányi 1988: 299).

Thus, cognition is primarily a biological function for constructing and operating an internal model of the environment (cf. Csányi 1992) by picking up relevant information, processing that information, and changing one’s behavior in accordance with that information. Cognition has an adaptive role because all this functioning has one aim: to enhance the organism’s average probability of survival in its environment by adjusting its behavior to expected situations (Csányi 1989: 205). Consequently, this aim will determine what environmental information counts as relevant in an organism’s interactive behavior. Cognition starts with perception, which is not a purely objective amassing of information about the world, but is basically selective (cf. Langacker 1987: 101) and involves an interpretation of reality in terms of the perceiver’s biology and in accordance with its interactive functioning in its environment. The cognitive processes based on the information supplied by perception involve an even more complex interpretation of the outside world, since they create a dynamic mental model of reality for the beholder. In other words, the function of cognition is knowing the world in a way that is required for an organism’s adaptive interaction with its environment. The cognitive mechanisms of any organism have been adapted to this interaction and permit therefore a species-specific perception of the environment and processing of incoming information. Rosch formulates this idea very clearly:

[T]he perceived world ... [is] ... not a metaphysical world without a knower. What kinds of attributes can be perceived

The cognitive function of language 111

are ... species-specific. ... What attributes will be perceived ... is undoubtedly determined by many factors having to do with the functional needs of the knower interacting with the physical and social environment. (Rosch 1978: 29)

Therefore, cognition is of a relativistic nature: On the one hand, the same environment requires different functional interactions, thus different ‘views’ of it, in different species, and on the other, the same environment may require different interactions on different occasions of the same individual, depending on a multitude of various internal and external factors. Even a chimpanzee is capable of perceiving, i.e., conceptualizing or ‘interpreting,’ a piece of stone as a tool on one occasion, and a weapon or a toy on others, depending on its role in a given interactive situation with its environment, although a piece of stone is none of these without the cognizer’s mental operations.

Cognition is thus not merely knowing reality, but knowing reality in a way that facilitates an organism’s optimal adaptation to reality, i.e., to its environment. Because of this, components of reality must be cognized according to the role that they play in the interaction between organism and environment. According to Rosch’s (1978: 29) “cognitive economy” principle of categorization, stimuli are considered similar as long as their differences are irrelevant to behavioral purposes. Thus, the perceived similarity of stimuli is relative to their role in an organism’s behavior. Similarity does obviously not reside objectively in the entities themselves, but emerges as a kind of functional analogy only in their subjective cognizing by an organism (Holyoak 1984: 204). Thus, cognition is in a sense a subjective process and the species specific cognitive processes create an Umwelt for the given organism (cf. Uexküll 1982).

Therefore, a functional approach to cognition should not exclude subjectivity, but view it as the basis for an organism’s adaptive behavior. Of course, subjectivity must be held within specific limits, which means that the correspondence between objective reality and its subjective cognition will be regulated by the adaptive value of the organism’s ‘view’ of reality. In other words, there has to be a feedback between adaptive value and subjectivity. The subjectivity of the internal model of reality operated by an organism cannot extend so far as to endanger the organism’s


survival, and so a construal of reality that hinders appropriate adaptation to the environment rather than facilitate it is not likely to gain validity. For different types of organisms, but also for the same type of organism (and even for different individuals) under different environmental conditions, different construals of phenomena may become valid according to their adaptive value. In the ideal case subjectivity goes just so far that the conditions of reality are utilized to an optimal degree in terms of the organism’s biology for a functional adaptation.

The notion of environment includes not only the natural and material environment but, in proportion with the complexity of the behavioral organization of a species, also their social and cultural environment. The human environment includes socially and culturally determined components to an exceptionally high degree and is thus a very complex phenomenon. Consequently, in the case of human beings the adaptive function of cognition does not relate to survival in the strict biological sense. Due to the extreme complexity of human behavior (as compared to other species), human cognition largely pertains to functional behavior and appropriate orientation in our sociocultural environment. In correspondence with this aspect of cognitive functioning, human cognition, in addition to the above mentioned genetic and neural levels of cognition, incorporates a third level, the symbolic one. This level, which is mainly manifest in language, is a cognitive adaptation to the complex human sociocultural environment (Győri 2001; Jerison 1988). The biological character of human knowledge and the fact that it is a “special kind of adaptation” (Plotkin 1994: 117) calls into question the claim that cognition is a purely rational activity and suggests that it must be based on an adaptive interaction with the human environment. This is the only way the human capacity for knowledge, including propositional type of knowledge, can be accounted for in a scientific way (cf. Plotkin 1994: 2). 5.2 Language as a tool for individual and social cognition As we have seen, the internal model operated in the cognitive process is partly based on genetically determined knowledge of the environment and of the necessary behavior therein and partly on


individual experience. In this sense cognition is to a large extent a subjective process in an individual. However, it can be made social to the extent to which individually acquired knowledge can be made collective within a group of individuals. Quiatt and Reynolds (1993: 141) define social cognition as “[t]he application of intelligence to the review of social information and the exploitation and management of social relationships toward attainment of short- and long-term goals.” Thus, different species participate in social cognition to the extent that they rely on social interaction for their survival. This is matched by the complexity of the different forms and mechanisms of communication through which the necessary sharing of information is achieved.

Humans are the species that possess the most powerful device for sharing knowledge, or to put it the other way round, the power of the human brain is largely due to its capacity for language. Due to the evolutionary innovation of combining the interindividual function of communication and the individual function of cognition in one system, a capacity emerged which made it possible to manipulate symbols, which can be used both externally in communication and internally in mental representation simultaneously (Győri 1999). But this does not mean that humans simply have communicative labels for their mental representations. The nature of the mental representations behind words is totally different from the iconic and categorical nature of non-symbolic representations (Győri 1995: 118-121; Harnad 1987: 554; Tomasello 2002: 133). Linguistic communication is a cognitive activity; we communicate about the contents of our minds: about mental representations, mental states, beliefs, etc. Thus, due to the function of “language as an instrument for organizing, processing and conveying [emphasis added] information” (Geeraerts 1997: 6), humans are capable of exchanging knowledge among themselves to an unparalleled extent. No other species possesses a system of communication that is capable of transmitting mental content to such a considerable degree.

G.B. Palmer (1996: 53) has stressed the evolutionary interdependence between the capacity for social cognition and language. An effective communicative system of a symbolic kind will enhance the power of a mental model of reality by lending it a social character. As a result of the capacity for language human


mental models do not remain confined to knowledge gained from direct and personal experience, and individuals will be able to partake of and benefit from the experience of others in extreme proportions (cf. Plotkin 1994: 10). By facilitating the representation and distribution of individually acquired knowledge, language creates a culturally shared mental model of reality for the advantage of the whole community. Such a model of reality is more powerful and less subjective than any individual model because the adequacy of the model can be controlled by the comparison of individual models. This is one of the most conspicuous functions of language: it is used for communicating conceptual structures that have been coordinated through speaker-hearer interaction and thus conventionalized in a speech community (cf. Clark 1996). Individuals sharing a particular language will also be able to share the same model of reality, which is qualitatively superior to any individual (i.e., private) model in range, accuracy, flexibility, etc.

Thus, the innovation in this new symbolic mode of the human brain for building a model of the environment – as opposed to genetic and neural memory – is that it is not directly linked to perception. Human cognition is largely characterized by an indirect way of acquiring knowledge because the information about reality stored in linguistic symbols is capable of substituting direct perceptual experience. By far the larger part of human knowledge about the world comes in a symbolic form and is of this indirect kind because the symbolic level of cognition far outweighs genetic and neural memory in importance in the cognitive processes involved in human behavioral interaction with the environment (cf. Figure 3 on page 86). Furthermore, many symbolic structures do not even qualify as representing ‘things’ that can be physically experienced. However, when talking about acquisition of knowledge without direct experience the following question immediately arises: How are humans capable of acquiring knowledge purely on symbolic grounds and how can real knowledge, i.e., the appropriate connection between knowledge and reality be secured in this way? In the case of non-symbolic types of cognition there is a direct (i.e., physical/material) connection between the knowledge stored by an organism and the reality the knowledge is about because genetic and neural cognition are grounded in genetic coding and perception respectively.

The existence of non-empirical knowledge in human beings is


obvious, but its explanation has posed a problem for philosophers for ages. Chomsky (1988: 3) refers to it as “Plato’s problem.” Plato’s solution to the problem was that human knowledge is contained in a priori ideas. Various rationalist views of language and cognition (Chomsky 1988; Fodor 1994; Katz 1990) are akin to this belief, though in a more modern way, since they postulate some kind of genetic endowment for these a priori ideas.

In traditional semantic theory an expression is considered to be meaningful if it is linguistically (analytically) true, i.e., not contradictory within a complex semantic system, because language is akin to a mathematical system in which meaningfulness can be equated with consistency in the system. Thus, the meanings of the symbols can be directly inferred from their relations to each other and are otherwise independent of the understanding of the knower. Cognition ‘happens’ because symbols objectively fit the entities of reality, and the relationship between them necessarily reflects the metaphysical relation that is supposed to exist between phenomena (e.g. Katz 1990). On these grounds, cognition is the manipulation of symbols: one either gets true or false propositions, i.e., descriptions that fit reality or not.

It is true that symbols carry knowledge that can be defined with other symbols (i.e., through the knowledge contained in them). An ideally complete dictionary of a language defines any symbol in that language with other symbols from that language. It is easy to see, however, that such an approach to the meanings of symbols will lead to circularity, since the number of symbols in a language is finite. (Infinity, on the other hand, would lead to the problem of infinite regress.) Going through the process of defining a symbol with other symbols, and then the symbols of the definition with yet other symbols, one would get back to the symbol one started with. The definition of the initial symbol would in the end contain itself.

In spite of all this, symbols undoubtedly are mental constructions whose knowledge content can be given through other symbols. Among others, this is how the mechanism of symbolic cognition can substitute direct perceptual experience and create abstract conceptual structures in the form of a socially shared mental model in order to cope with the complexity of the human sociocultural environment. However, the rationalist view of cognition has to postulate a priori ideas in order to explain understanding, otherwise pure symbol manipulation will run


into the so-called “symbol grounding problem” (Harnad 1990), i.e., when symbols meaningless in themselves are supposed to get their meanings only via other meaningless symbols.

Thus, Katz (1976) claims that the unique feature of language, effebility – which means “the full intertranslatability of natural languages” and that “every language can express every sentence sense” – can be accounted for by the “assumption that our concepts come from our genes” and not “from experience.” To Quine’s rhetorical question “... who would undertake to translate ‘Neutrinos lack mass’ into a jungle language?”, Katz remarks:

For rationalists, cases of failure to translate theoretical sentences represent only a temporary inability of the speakers, based on their lack of knowledge of the relevant sciences, to make the proper combination of primitive semantic concepts to form the appropriate proposition. That is, the failure represents a temporary vocabulary gap (rather than a deficiency of the language) which makes it necessary to resort to paraphrase, creation of technical vocabulary, metaphorical extension, and so on, to make translation possible in practice, as well as in principle. (Katz 1976: 40-41)

There seems to be nothing wrong with Katz’s view that the emergence of any “higher” concept is to be accounted for by “the proper combination of … concepts,” since conceptual combination is a real psychological process which creates new conceptual categories (Hampton 1997: 155). But the symbol grounding problem is not solved by postulating a priori ideas in the form of genetically determined primitive semantic concepts. For this reason Harnad (1990) claims that the type of knowledge contained in complex symbolic structures is ultimately grounded in more basic genetic (not in the rationalist sense) and experience determined cognitive structures, i.e., in pure perceptual information and the simple neural type of cognitive processing of physical experience.

Harnad (1987, 1990) differentiates between iconic and categorical representations for the types of knowledge that an organism can acquire through direct interaction with its physical world. These are the knowledge representations that non-symbolic types of cognition can give rise to. Thus, direct experience in any of the perceptual


modalities, creates either of two types of representations in the brain. These are iconic and categorical representations. Iconic representations are “analogs of the proximal sensory projections of distal objects and events,” and categorical representations are “learned and innate feature detectors that pick out the invariant features of object and event categories from their sensory projections” (Harnad 1990: 335). In accordance with this, symbols are of two kinds. Elementary symbols are the ones that are grounded in these iconic and categorical representations, while higher order symbols are the ones created through composition of elementary symbols and will thus become indirectly grounded because they inherit the grounding from the elementary symbols. Thus, elementary symbols provide the basis for the acquisition of knowledge without experience when composed into higher order symbols whose underlying representations are not sensory representations anymore, but symbolic representations.

There is another line of evidence which suggests that certain symbols are more fundamentally linked to sensory experience than others. Rosch (1978) has shown that there is a level of categorization of reality – the basic level –at which we make categorizations on the basis of natural discontinuities found in nature. In other words, we distinguish the entities that show maximal category resemblance with each other and minimal with others based on motor movements, gestalt perception and behavioral functions connected to them. Categorization above this level has no direct empirical ground, and even categorization below the basic level is not necessarily grounded in perceptual qualities but often involves knowledge other than the types basic level categorization is based on, among others also some type of sociocultural knowledge. This is why children first learn and name basic level categories. It is here, in the process of word acquisition in linguistic ontogeny where the emergence of elementary symbols is the most obvious. The child first learns the symbols for the categories that it gets into physical contact with and can most easily distinguish on perceptual grounds. Thus, the meanings of these words are indeed empirically grounded because they are connected to categorical representations based on perception.

In spite of emphasizing the perceptual grounding of symbolic cognition, Harnad’s approach is still close to the objectivist symbol manipulation view of language as an algorithmic device. His example for symbol composition is the following: If we know the elementary


symbols for STRIPE and HORSE (i.e., we have sensory representations of them), we can form a symbolic representation of a ZEBRA without ever having seen one. This is so because “the grounded names ... [are] strung together into propositions” (Harnad 1990: 343) and thus the symbolic representations “consist of symbol strings describing [new] category membership relations” (Harnad 1990: 335): ZEBRA = HORSE & STRIPES (An X is a Y that is Z). (Harnad draws our attention to the fact that the Chinese orthographic sign for zebra is composed of the ones for STRIPES and HORSE.) The higher order non-elementary symbols do not completely lose their empirical grounding in this way. They remain indirectly grounded, because through the composition they inherit the grounding of the elementary symbols.

Though in this way it is possible to account for our understanding of meanings and our symbolic knowledge without having to postulate a priori ideas, there is still one problem. If we think of symbol composition as a propositional act, we presume that we have to do with classical categories, which are made up of necessary and sufficient features (Smith and Medin 1981: 23), and which by definition also appear in the newly formed category. However, the inheritance of the grounding does not necessarily bring about the inheritance of the set of attributes of the composing symbols. Though language indeed creates new meanings on the basis of already existing ones, this is not done in the form of propositions most of the time. Although conceptual combination may sometimes be compositional, in most of the cases it will yield “emergent properties” in the combined concept which do not derive from either of the combining concepts (Hampton 1997: 147). The reason for this is that the inputs to a conceptual blend are rarely mental representations of classical categories, but rather mental spaces construed by speakers on the basis of encyclopedic knowledge that have been evoked through the current linguistic expressions (Coulson and Oakley 2000: 176). Thus, in natural languages symbol composition is mostly not a propositional act because the newly emerging category is usually not deduced according to the laws of logic from the composing ones. Thus, we may immediately understand that a ‘striped horse’ is a ‘zebra,’ even though as a proposition this is not true; the category HORSE does not include zebras. The fact that we still understand the ‘equation’ makes a strong case for the role of the human capacity for analogy and metaphoric


imagination in conceptual representation. It is indeed true that our symbolic knowledge often takes the form

of propositions that describe reality in a true or false fashion. It is a fact that there are so-called truths about the world, and it is the task of the natural sciences to deal with these aspects of reality. In this sense, there may be only one scientifically correct way of cognizing reality, but the cognitive function of natural human language is not to describe these. It is another matter that – since human beings do science – language is also used for the purpose of describing scientific truths. However, the general cognitive task of language is to provide an adaptive conceptual model of the world. Thus, to take a classic example, the expert knowledge that a tomato is technically a fruit because it is the seed-bearing part of the plant on which it grows may be scientifically correct (Smith and Medin 1981: 29), but it is much more adaptive for everyday human behavior to conceptualize it as a vegetable on the basis of the role it has in our culturally determined eating habits (Wierzbicka 1984: 328). That is to say, in our adaptive interaction with the environment we can do without the former knowledge but not without the latter.

Thus – strange as it may seem – in a strictly propositional sense the statement A ZEBRA IS A HORSE THAT HAS STRIPES gives us false knowledge about reality. Nevertheless, this false proposition carries useful information about the world and in spite of its falsity we consider it true, though not in a propositional but in a figurative sense. Having heard the statement A ZEBRA IS A HORSE THAT HAS STRIPES, we will easily recognize a zebra. We will even have other important information about zebras at our hands, information that we can infer from our knowledge about horses. Thus, a figurative interpretation of a symbol string may help us acquire knowledge about reality – even though it does not match so-called ‘objective truth’ – in the sense that it will facilitate cognition for the sake of adaptive orientation in the environment.

Actually, our proper understanding of the statement A ZEBRA IS A HORSE THAT HAS STRIPES relies much more on conceptual blending than on an algorithmic conceptual combination. The two domains of knowledge, HORSE and STRIPES, can be characterized in terms of mental spaces (Fauconnier 1994), which provide the elements and relational structure as inputs necessary for the construction of an emergent new meaning in the form of a conceptual blend (Coulson


2001: 164). The blend does not emerge in a compositional fashion but the elements and structure that the newly formed mental space (i.e., the emergent meaning) will inherit from the input spaces is influenced among others by contextual information and background knowledge (Coulson 2001: 116-117).

A blend will become a metaphor when the new mental space emerges as an analogical structure to the blended spaces. Thus, though A ZEBRA IS A HORSE THAT HAS STRIPES is actually a ‘miscategorization’ in the propositional sense, as a metaphor it is a relevant conceptual tool in cognizing the environment. The miscategorization is in a way intentional in the sense that its purpose is to yield a meaningful novel conceptual construal of reality by extending or stretching the category. Lakoff and Johnson (1980), Johnson (1987) and Lakoff (1987, 1990) have worked out a theory that explains the way humans make sense of the world through metaphorical projections from a source domain to a target domain. The topological mappings across these conceptual domains rely on ontological and epistemic correspondences and facilitate by this our comprehension of unfamiliar or less well-understood domains of experience in terms of established familiar knowledge. This is the reason why even scientific explanations cannot do without making use of metaphorical descriptions (Beck 1978).

Let us return now to the grounding problem. Several authors – in line with the above course of reasoning about elaborating on concrete knowledge in order to understand more abstract realms of experience – take a more ‘embodied’ view on the ‘grounding’ issue and consider human symbolic knowledge as rooted in our direct bodily interaction with our environment based on our sense of space, both visual and kinesthetic. Givón (1998: 46), for instance, argues “that a big chunk of the neurology that nowadays support human language is but an evolutionary outgrowth of the visual information processing system.” Lakoff (1990: 73) claims that spatial perceptual mechanisms lie at the bottom of human rationality. Langacker’s (1987) Cognitive Grammar is also founded on the conviction that our knowledge of spatial relations forms the basis of linguistic structures. Johnson (1987) has given a detailed explanation of how linguistic meaning emerges through the metaphorical projections of image schemata, which arise in the mind from bodily experience through interactive behavior with our environment.


In sum, it is true that language has the power of facilitating so-called objective knowledge about reality in the form of propositions. But language is also the medium that promotes figurative thinking (Tomasello 2002: 17), which enables us to construe our environment in an adaptively optimal way. Thus, figurative and propositional modes of cognition are equally important and necessary for an adequate cognizing of reality. This speaks for language as an adaptation in order to facilitate an increased flexibility in cognitive activity for the sake of a more effective adaptation to the environment. 5.3 Adapting language to cognition As I have claimed above, the basic cognitive function of language is that it serves as a culturally shared model of reality on which every individual in a community can rely for the construction and operation of their own mental models of the environment. We have also seen that the power of this model derives from the fact that the basis of the knowledge shared through it is neither some common genetic endowment nor the same experience, but its symbolic nature. This symbolic model – with the help of the components (grammatical rules and linguistic signs) constituting it – can be operated in various ways for processing information about the environment. Even new cognitive structures can be constructed actively and subjectively by any one individual and then mediated to other individuals in order to substitute direct experience for them or to provide them with abstract conceptual constructions for understanding various relations between phenomena of reality.

In order for this social cognitive process to function correctly, language – as a social instrument for cognizing the environment – must always suit the cognitive needs of a speech community. This means that it must be able to encode all the necessary information about reality and model it in a way that facilitates optimal accommodation to a given environment. In other words, language has to be such that it adaptively serves the acquisition, organization and application of knowledge for interaction with the environment, just those things that make up the function of cognition in general (cf. Neisser 1976: 1).

Since human cognition is characterized by its strong reliance on


symbolic structures in the form of language, language must inherently be designed to serve cognition. However, the way language is structured is obviously not influenced by reality itself in some objective fashion, but this influence must come indirectly through our interpretation of reality. The reason for this – as already described – is that the general biological function of cognition is knowing one’s ‘world’ for the purpose of interacting with it in an optimally functional way (Varela, Thompson and Rosch 1993: 205). Even though human cognition employs the symbolic power of language in the form of a sociocultural cognitive model, its function is in line with the general biological function of cognition – though in a much more complex manner. Human cognition is not a rational process of revealing some metaphysical truth about the world and language is not a rational symbol system for this end (Plotkin 1994: 15). Language functions as a flexible device for cognition as it provides a means to adaptively model, both socially and individually, the given environment and to accommodate any changes of sociocultural relevance that might occur in it. Therefore, in order to be a functional social model of reality, it is crucial that language be continuously adapted to cognition in the proper way. As Anttila (1989: 179) says, “[l]anguage serves the sociocultural ends and its task is thus to keep itself in an enduring state, to keep functioning, adapting itself to new environments.”

Thus, the basis for a adequate cognitive functioning of language is not its correct reflection of reality, but the reflection of our adaptive interpretation of it. In other words, our cognitive processes will necessarily tailor language to the needs of cognition: the way we see the world and think about it in non-symbolic ways clearly affects the form of language (cf. Clark 1996: 342). But this is not just a general effect. The influence of cognition on the shape of language has a specific adaptive purpose. A particular language must be adapted to the particular physical, social, cultural, historical, etc. environment which it is to model and in which it is to be used. When cognition shapes linguistic structure to its needs (though naturally within the boundaries of the general structural properties of natural language), these different environments will exert their effect on the various languages. The social validity of these structures is achieved in the process of conventionalization through the sanctioning by a speech community in speaker-hearer interaction (Langacker 1987:


65-66, 156). This is of course not to deny that due to the complexity of design, language will necessarily possess also ultimately arbitrary structural features, i.e., ones without any functional relevance, and which are derived effects of other structures or effects of general structural constraints. Such phenomena will inevitably also leave their mark on the way language is.

Geeraerts (1997: 8) talks in this respect about “[t]he perspectival nature of linguistic meaning [, which] implies that the world is not objectively reflected in language.” Comparing the semantic structures of different languages, it becomes immediately obvious that different languages impose different categorizations on the world. This obviously results from the way different languages adapt to their environments – in line with the general function of cognition and the cognitive function of language (Tomasello 2002: 127). An adequate orientation in a given sociocultural environment requires a specific category system and appropriate construals of particular phenomena. Thus, for instance, different peoples and cultures often construe the same phenomena of reality in different ways because their different environments demand different ways of adapting to them. Because of this, linguistic categorization very often reflects a very intricate and complex social and cultural environment. This can be seen among others in the case of various classifiers in many aboriginal languages (e.g. Lakoff 1987: Chapter 6; Palmer 1996: 126-141; Palmer and Woodman 2000). For instance, from the ten noun classes found in the Australian aboriginal language Nangikúrungurr and marked with separate prefixes, one contains only names of weapons, and another exclusively names of spears (Wierzbicka 1984: 314). This is probably due to the fact that weapons (and among them spears especially) play a special role in the lifestyle of these people. Semantically transparent expressions and the etymologies of many so-called literal expressions reveal a great deal about this process as they show how reality can be construed in alternate ways to facilitate this adaptation. It is quite apparent for instance that the English word glass, meaning ‘a vessel made of glass for drinking’ derives its name from the material it is made of. Further on, OE glæs derives its form and meaning from the Common Germanic stem *glaza-, as found in *glazam meaning ‘amber’ (Onions 1966: 400; Drosdowski et al. 1963: 224). The shift in meaning could be explained on the basis that glass (the substance)


is similar to amber with respect to being translucent and shining (and was even more similar before the development of more advanced technology in glass-making). Furthermore, as Drosdowski et al. (1963: 224) claim, the similarity to amber is also supported by the fact that the Germanic people got acquainted with glass through the Romans, and it was not only the transparency of the two substances that the meaning transfer was based on, but also the fact that the Romans used glass as jewelry in the form of beads just as the Germanic people used amber. This form with its attached meaning comes in turn from the Proto-Indo-European (PIE) root *ghel- meaning ‘to shine, glitter’ (Watkins 1985: 21).

Furthermore, the environment is never a stable metaphysical reality, but a changing one. Because of this, any changes in the environment that are relevant at the level of a speech community call for an adaptation of language to these changes. Thus, when cognizing reality, our conceptual system continuously exhibits an interplay between stability and flexibility in order to fit stable conditions, but at the same time also to be able to adapt to novel ones (Medin and Barsalou 1987: 468). Language must always reflect this motion in order to function as an efficient cognitive device. Efficiency means here that language must provide an interpretation of the world that proves to be adaptive in the given physical, social, cultural, etc. environment by best facilitating cognition. This is done by supplying ready-made knowledge about the environment the language users live in, but only relative to the stability of conditions over time (Palmer 1996: 52). This interpretation in the form of ready-made knowledge is largely manifest in a cultural system of categories, i.e., a certain common repertoire of conceptual categories stored in the minds of the individuals of a speech community. Since language is a device for the categorization of experience (Geeraerts 1997: 7-8, 20), it is obviously a lexical item that makes a conceptual category most easily accessible at the social level (cf. Rosch 1978: 28).

Content words clearly name categories, but the fact that language is a system of categories is apparent not only in the case of content words. Functional elements (e.g. articles, prepositions, suffixes, etc.) also categorize reality, as they are very general categorizations of relations between non-linguistic phenomena. Many prepositions, for instance, are linguistic instantiations of


various image schemata, i.e., they categorize recurring patterns in our experience, like in and out in the case of the CONTAINER schema, up and down in the case of the VERTICALITY schema, or from and to in the case of the SOURCE-PATH-GOAL schema (Johnson 1987: 30ff.; Lakoff 1987: 271ff.).

As Anderson (1988: 93) points out, the categorizing function of language reveals itself especially in the fact that it stabilizes conceptual structure against fragmentation. Phenomena of reality must be designated not only for the sake of discourse, but also for the sake of thinking about them, since fixed conceptualizations and stabilized conceptual structures are essential for economical and effective thought.

In a study by Carroll (1985) subjects were asked to make up names for various things, either unfamiliar or only lacking a conventional name. It was observed that the names generated tended to describe and categorize because they referred in some degree to properties of the name’s referent. When the subjects were asked to rate the names they produced according to quality, the names that were easy to learn and remember (i.e., descriptive, natural etc.) and easy to use (i.e., distinctive, brief, etc.) were rated as “good names” (Carroll 1985: 5). As the criteria for easy remembering and easy usage indicate, names are the better the more unambiguously they highlight a category. Coding in language evidently facilitates the activation of the appropriate cognitive routines and thus contributes to a category reaching a degree of entrenchment through which it achieves unit status (Langacker 1987: 100).

Since the cognitive function of language requires the socially shared category system to be an adaptive interpretation of reality, there must be good reasons why meanings of a language specify the categories they do and not others (cf. Clark 1996: 340). At every historical stage in the development of a language its lexicon defines a system of conceptual categories which provides ready-made knowledge about reality and thus serves the cultural inheritance of experience and knowledge across generations (Rosch 1978: 28; Wierzbicka 1984: 314). The conceptual categories stored in a linguistic form provide the ‘building blocks’ of a speech community’s social model of the environment, which constitutes an essential part of the culture of the community (cf. Tomasello 2002: 180-181). As Rosch (1978: 27) has claimed, the specific categories


of the human mind that get coded in any particular language are not the “arbitrary product of historical accident or of whimsy” but the product of functional principles of categorization, and working with those categories should be the most efficient way to deal with the environment. Consequently, the two basic psychological principles, “cognitive economy” and “perceived world structure” (Rosch 1978: 28-29), also influence what conceptual categories will be socially adaptive and will as a result achieve cultural significance to become coded in a language. Thus, the process of cultural category formation is functional in nature since it is based on a speech community’s social cognitive adaptation to situations its members might encounter in their environment and which they have to handle by thinking, reasoning and communicating about them.

The above implies that linguistic forms will necessarily influence the speakers’ perception and thinking about the world. If we did not perceive the categories supplied to us by our language, what sense would language make? It just would not be functional in its given environment. Such a situation would be a self-contradiction in view of the mentioned adaptiveness and the function of the two basic principles of categorization. Of course, due to changes in the environment or novel aspects we might adopt about it, there might be lapses between the categories offered by language and our perception of the world. Relativity does not mean that our perceptions must always match our linguistic categories and that categories other than those provided by language cannot be perceived. Instead, language will flexibly create and maintain an optimal amount of synchrony in this respect in line with its cognitive function. This also implies that linguistic determinism is not a stronger version of relativity, but in fact its opposite. Determinism would be totally inadaptive and harmful for the cognition of an environment which has the dual character of being relatively stable and changeable at the same time. Linguistic relativity, on the other hand, is a reflection of a cognitive system adapted to such an environment and follows from the function of cognition. Relativity is exactly the right evolutionary solution for the required stability of a cognitive system coupled with the necessary flexibility, or as Hays (2000) put it: “… linguistic relativity … [is] … an evolutionary survival trait.” These considerations can also provide the foundation for what Lucy (1997: 291) has called the “domain-centered-


approach” to the study of linguistic relativity. This approach seeks an explanation to the question why different languages encode experienced reality in different ways. In view of the above, an explanation along the following lines appears to be rather plausible: The influence of language on cognition is functional, which means that the structure of language, semantic and grammatical, is largely organized in a way that will make this influence useful.

The differences in the semantic structure of different languages are thus due to the formation of the mentioned culturally adaptive categories, which happens in the process of lexicalization, i.e., through semantic and lexical changes as well as compounding and derivations in the course of the history of a language. It follows from the cognitive functioning of language that it is the speakers’ conceptualizations of reality that engender the actuation of these processes. These new meanings are thus products of historical categorization processes, i.e., they are fossilized conceptualizations of previous generations which have gained cultural validity. They have outlived the speakers of the times of their emergence and later on they impose a given categorization of the world on future generations. But just as these linguistically coded categories are results of previous conceptualizations on the level of a whole culture, they also provide an ever-ready source for the operation of similar cognitive processes in the future. Or in Hopper’s (1990: 159) plain description of this historical linguistic state of affairs, “words come from other words.” When in this fashion newly introduced expressions with no established conventional meanings in a language are employed for the sake of adapting the language to changing communicative and cognitive needs, mutual intelligibility between speaker and hearer is a basic requirement (Palmer 1972: 309), and shared knowledge must serve as an inevitable basis (Fritz 1998: 21). For this purpose the interlocutors must coordinate their expectations of each other’s intentions on the basis of all those various commonalities that actually constitute their culture (Clark 1996: 325).

The best possible basis for mutual intelligibility is the analogical character of human mental processing: it is a basic characteristic of human thought that all new phenomena are mentally grasped via an analogy to already familiar cognitive structures (e.g. Heit 1997; Gentner and Markman 1997; Holyoak and Thagard


1997). As Anttila (1989: 141) puts it, “man has an innate capacity for analogy [...] and [...] language is only part of this.” We utilize familiar knowledge through analogical thinking when we categorize, make inferences and create and learn new abstractions. Analogy is crucial in making sense of the world by recognizing similarities, i.e., by noticing that certain new experiences are similar to old ones in specific ways because “[a]nalogy [...] is structured similarity with functional import” (Holyoak 1984: 204). Holyoak and Thagard (1997: 36) have identified three constraints in analogical reasoning. First, the analogy rests on perception of direct similarity. Second, structural parallels are sought for. And third, the analogy has a certain purpose, i.e., it is guided by what the reasoner intends to achieve by it. All these constraints are determined by the reasoner’s existent knowledge.

The basis of cognitive semantics is akin to these insights in cognitive psychology, as Langacker’s formulation testifies:

Our mental experience is coherent by virtue of the structure we impose on it. A pivotal aspect of this structuring capacity is the interpretation of novel experience with reference to previous experience, … . (Langacker 1987: 105)

Johnson (1987: 174) has also stressed the importance of familiar information in making sense of new experience, and Lakoff (1987: 346) has pointed out that motivation – in the sense of relatively easy cognitive processing due to certain clues providing mental support, like iconicity (cf. Anttila 1989: 152) – is crucial to our understanding, learning and storing of new information. The prototypical character of our conceptual structure reflects exactly this analogical way of thinking. In the semantic structure of language this prototypical character, through its flexibility and dynamism, ensures both the adaptability of meaning to new experience and the intelligibility of meaning extension (i.e., novel usages) by restricting their range and direction (Geeraerts 1997: 113-114). It is also this analogical character of human thinking that gains expression in figurative language. It has been concluded that our minds understand and interpret the world around us with the help of metaphorical and metonymical processes, image schematic projections, and idealized cognitive models (Lakoff and Johnson 1980; Lakoff 1987; Johnson


1987). The analysis of historical semantic data also shows that a huge

part of our symbol system is metaphorical and metonymical in nature (e.g. Dirven 1985; Hopper 1990; Sweetser 1990). A cognitive examination of such data reveals specific details about how and why semantic change is actuated. In the next chapter I will undertake such an investigation and look at the cognitive background of speakers’ linguistic behavior when initiating metaphorical, metonymical and other indirect references. I will show that such usage is actually the manifestation of the cognitive function of language inasmuch as it is the result of flexible adaptive linguistic behavior in the effort to effectively cope with the communicative and cognitive challenges posed by variations and transitions of the environment.

Chapter 6

Cognitive aspects of semantic change 6.1 Toward a cognitive explanation of semantic change The basic claim I am about to advance in this chapter is that semantic change relies on general principles of human cognition. In spite of the apparent plausibility of such a claim its elucidation appears to be imperative for the following reason. As much as change of meaning should be considered a semantic issue, traditional semantic theory has never been capable of convincingly explaining semantic change within its own domain and has thus mostly opted for its exclusion from its subject matter. However, as Fritz (1998: 8) points out, a sound explanation should be considered one of the major tests for the validity of any semantic theory. Lyle Campbell (1998: 254) laments this situation in the following way: “… we would be in a better position to understand semantic change if we could base our understanding of change in meaning on a solid theory of semantics.” As Geeraerts (1997) has recently shown, prototype-theoretical semantics can righteously aspire to be such an underlying semantic theory for the explanation of semantic change. However, for a long time modern semantic theory has been characterized by objectivism (cf. Johnson 1987: 173). In general, objectivist approaches to semantics exclude the possibility that the subjective mind and general cognitive mechanisms could be involved in meaning. Meaning is considered as some kind of objective correspondence between linguistic symbols and parts of reality either in the form of reference or truth conditions, or else as some kind of objectively given conceptual structure. An explanation of semantic change is difficult to provide in these terms because the idea of the objectivity of meaning entails its invariability. The relatively recent trend of non-objectivist cognitive semantics remedies this situation to a great deal and the analysis offered in this

Cognitive aspects of semantic change 131

chapter will make much use of its tenets. Historical linguistics has of course never been short of an

explanation for semantic change, but it is very much at odds with the mentioned traditional views of semantics. Explanations within the framework of historical linguistics rely on notions like analogy, metaphor, metonymy, etc., all mental processes that are based on general cognitive mechanisms. Such mechanisms are banned from the framework of traditional linguistic theory on the basis of the conviction that language is an autonomous mental faculty. Figurative language use, which often lies at the heart of semantic change, appears to be non-rule-governed and thus non-formalizable. As a result, it is not considered to be an organic part of linguistic functioning and has often been relegated to the field of pragmatics.

The pragmatic aspect of semantic change is of course highly relevant in any serious attempt at an explanation. The conventionalization of novel expressions is a sociocultural process, and the alteration in the usage of an expression across a whole speech community is thus the cumulative result of the communicative activity of its members. Because of this, several authors have emphasized the social character of language change (e.g. Milroy 1993) and stressed the point that no explanation of semantic change can be correct without assigning a central role to the speakers of a language (e.g. Joseph 1992), since “a language that is not used does not change” (Anttila 1989: 153). Thus, semantic change can only arise in speaker-hearer interaction and is thus always the result of context dependent alteration of usage. According to Traugott’s (1999a) Invited Inferencing Theory of Semantic Change, the speaker does the work of innovation but invites the hearer to infer his/her motivations behind the altered usage from the context. There are two interacting and inseparable levels in the communicative process. On the linguistic level the speaker faces the question which expressions have the appropriate semantic structure for conveying the ideas which he/she wishes to communicate. On the pragmatic level the language user decides which the best expression is to use for his/her communicative purposes. When none of the conventional semantic structures are deemed appropriate, the speaker resorts to the modification of these structures, most of the time unconsciously (cf. Geeraerts 1997: 107). Thus, with regard to the communicative process, ways of novel


132

usage, like metonymy, metaphor, etc., are pragmatic devices for altering usage within a conventionalized system. With time this may lead to changes in conversational practices and in the rules of usage and it is these changes that will ultimately find their way into the system of the language. This is the point when metonymy, metaphor, etc., become semantic processes because “a Lexeme L may gain semantic properties from the context in which [it is] typically used” (Traugott 1999a: 94). The new generation will receive this altered system as input during language acquisition and this is how the change is completed (cf. McMahon 1994: 43, and Nerlich 1989: 175).

Although the pragmatic aspects of semantic change are fundamental, the speaker’s innovative usage of an expression and the hearer’s making sense of an altered usage cannot be explained satisfactorily in pragmatic terms alone. The pragmatic behavior manifest in the reliance on the interaction between context and conventional meaning must be rooted in some specific mental activity, since the way an expression is used in a speech community is determined by the cognitive/conceptual structure behind that expression in the minds of the speakers. The modification of the conventional usage is thus a cognitive problem: by what cognitive mechanisms can the given semantic structures be modified in the appropriate way to serve the speaker’s communicative purposes, i.e., satisfy the principle of efficient communication. In this respect metonymy, metaphor, etc., are cognitive devices for sharing perspectives and conceptualizations. But these are still not the basic cognitive operations behind altered usage (and potential semantic change in the long run) because first of all the speaker must make a choice (intuitively and unconsciously) as to how to conceptualize a given phenomenon, i.e., which semantic structures to modify and what cognitive devices to apply. In other words, the question is what factors influence or maybe even determine how a speaker will conceptualize particular phenomena. However, this aspect of altered usage is usually not considered in a depth that would be essential for a proper explanation of semantic change. Therefore it is necessary to examine what cognitive factors govern context dependent innovation of usage and its comprehension, and what constraints these factors place on the ways usage can be altered by the speaker and then made sense of by the hearer.


Blank (1997) makes a start in this direction of research. He distinguishes the preceding associative process from the innovation itself, which is a linguistic process (Blank 1997: 149). He also identifies four associative relations serving as the basis for innovation and three associative principles, which may underlie these relations (Blank 1997: 155-156). The former are associations between designata, between sign contents, between whole signs, and between sign expressions, while the latter are similarity, contrast and contiguity, though not all the principles are applicable to all the relations. In my analysis I will go one step further and try to find out what governs the detection of similarity, contrast and contiguity and on what cognitive grounds speakers judge various instances of these as appropriate bases for innovative usage.

In historical linguistics there is a tradition for accepting reference to cognitive processes and to human cognition in general as a basis for an adequate explanation of change of meaning (Campbell 1998: 267). As Anttila (1989: 133) puts it: “semantic change shows the mental or psychological factors of change in their purest form.” Already Paul (1920: 75) spoke of semantic change as originating in a discrepancy in the conceptualization underlying the conventional and occasion-bound meanings of a word. Most historical linguists have appreciated an elucidation of individual changes in terms of metaphor, metonymy and other figurative language use and viewed cognitive mechanisms like analogy and association, and psychological factors in general, as playing an important role in the emergence of new meanings (see Anttila 1989: 141; 1992; Campbell 1998: 269).

It has to be mentioned though that there are attested cases of semantic change in which only pure chance and no cognitive processes seem to play a role. McMahon (1994: 175) mentions “the movement of the sense of flaunt towards that of flout” as an ongoing change in English. It is indeed hard to see what cognitive motivation could underlie this semantic shift, and it may be a simple case of misinterpretation, probably caused by the similarity of the signifiers (cf. Blank 1997: 40). As another famous example the etymology of the word bead could be mentioned. One explanation claims that in this case erroneous learning occurred between generations. Children may have misinterpreted the Old English (OE) word (ge)bed ‘prayer’ as meaning ‘the little spheres on the rosary’ when hearing the phrase


134

to count one’s prayers (McMahon 1994: 177; cf. also Anttila 1989: 137). On the other hand, Campbell (1998: 258) interprets this change in meaning as a “metaphorical extension from the ‘prayer,’ which was kept track of by the rosary bead, to the rosary bead itself.” I think that the two views do not contradict each other, since erroneous learning through misinterpretation is undoubtedly a cognitive act that should involve some standard cognitive mechanism, though wrongly applied.

Another rather common case of semantic change in which one might suspect the lack of cognitive motivation is when there is a change in the referent of a word. After all, it indeed seems to be true here that speakers’ cognitive processes do not have anything to do with the change. As an example of referent change Anttila (1989: 137) mentions English pen, which used to mean ‘feather’ and was an appropriate term “[w]hen quills were used for writing with ink.” As a consequence of change in material culture, pen today means ‘writing tool.’ However, a moment’s reflection will make it obvious that this development does not only depend on the change in referent, but requires the speakers’ willingness to subsume the new device under the same category on the basis of almost total functional sameness. In fact, the change in material culture will trigger a most probably automatic metaphorical transfer, which undoubtedly requires cognitive motivation. Though changes lacking any cognitive motivation probably do occur, it is not easy to assess their number, but in view of the above they should be relatively rare.

A demonstration of the fact that various cognitive factors guide semantic innovation will also cast doubt on the objectivist stance described above and strengthen the recent view in semantics that meaning is based on the human understanding of the world and consists of knowledge structures that are open-ended and encyclopedic in nature (Langacker 1987, 1990). If this turns out to be a valid property of meaning, then it will also become obvious that change is a necessary and natural characteristic of meaning, or, in Nerlich and Clarke’s (1988: 73) even stronger formulation: “the nature of meaning is change.” This type of cognitive semantics also holds that semantic structure is shaped by the characteristics of human cognitive abilities (Langacker 1987), as opposed to the mentalist-rationalist view according to which it matches reality in an objective way independent of how language users happen to see the


world (e.g. Katz 1990). Lakoff (1987), Johnson (1987), and others have shown how metaphorical and metonymical processes, image schematic projections, idealized cognitive models (ICMs), etc., help us understand and interpret the world around us. It is no wonder then that this leaves its mark on language and will fossilize as diachronic semantic phenomena. In line with these views several recent works on semantic change take a cognitive approach for granted (e.g. Geeraerts 1985; Lichtenberk 1991; Traugott 1985; Sweetser 1990).

The cognitive motivation for the innovative use of expressions will first lead to individual linguistic action, which will actuate semantic change only if such actions of language users eventually get summated. Both Anttila (1989: 408) and Keller (1985, 1989) speak of language change as the collective consequence of an “invisible hand process.” Such a process occurs when individuals perform certain actions intentionally but not with the consequences in mind that will nonetheless be caused in the end. For instance a beaten path across a lawn will ensue through individuals intentionally crossing the lawn but not with the purpose of creating such a path. The path originates as a non-intended collective consequence of the intentional individual actions. Language change is supposed to be a similar process. Within a formulation of the invisible hand explanation Keller (1985: 222) includes “the depiction of personal motives, intentions, goals, convictions (and so on) which form the basis of individual actions.” I think that an analysis of these motives, etc. does not provide a full explanation for the individual actions, i.e., the innovative usage of conventional expressions, but for a full understanding of semantic change its cognitive background must also be uncovered. 6.2 General issues of semantic change Our knowledge about semantic change derives from the study of word histories (cf. also Job 1982). Etymology provides the material for the study of semantic change and it is wrong to think that it is simply the pursuit of the original meanings of words. It is easy to show that there can neither historically nor logically be a meaning of a word that is its ultimate origin. The etymology of the word glass, already mentioned in Chapter 5, may serve as a good illustration:


136

English glass ‘a vessel made of glass for drinking,’ < Old English glæs ‘the material glass’ < Common Germanic *glazam ‘amber’ (*glaza-) < Proto-Indo-European (PIE) *ghel- ‘to shine, glitter’ (Onions 1966: 400; Drosdowski et al. 1963: 224; Watkins 1985: 21). In a historical sense any of these meanings could be thought of as original to the subsequent one(s), but at the same time any meaning is also a derived one from an even previous one. Theoretically, this must also hold for a meaning at the PIE stage, since it would be quite absurd to claim that morphemes were created ex nihilo at that stage (cf. Hopper 1990). When we speak of the PIE root as the ultimate origin of the above words, it has to be borne in mind that we are constrained by the time depth which is set by the framework of the Indo-European language family and which is permitted by the methodology of historical comparative linguistics.

From a logical point of view there seems to be no reason either for deriving one sense from the other and not the other way round. E.g. the word orange, originally designating an object (a certain kind of fruit), can now also stand for an attribute of that object, its color, while we find the reverse logic in the case of the word bear, which earlier had the meaning ‘brown’ (PIE *bher- ‘bright, brown’ > Eng. bear, Watkins 1985: 7). The claim about this bi-directionality may be controversial due to the fact that a CONCRETE > ABSTRACT development appears to be the rule, but as Campbell (1998: 273) notes, semantic changes like narrowing “often involve change toward more concreteness.” Admittedly, the COLOR > CONCRETE OBJECT development in the latter example is probably due to a taboo strategy. However, there are examples of this kind from present-day language use where this is not the case, e.g. Hungarian fekete ‘black’, also meaning ‘black coffee, espresso.’ Ellipsis, an important factor in semantic change (cf. Blank 1997: 302), plays a role in this development, but it must also have done so in the case of bear (viz. ‘the brown one’).

Let us now turn to more theoretical issues of semantic change. The task of historical linguistics is in general to investigate the following: (1) How a particular language has changed in the course of time; this involves the description of changes from one stage to another in the history of that language. (2) The generalization of changes, i.e., what general mechanisms (or laws, or principles) of change there are. (3) The explanation of change, i.e., what its causes


and reasons are, how it starts and spreads, etc. Of these three issues, (2) and (3) are of considerable theoretical

interest for historical linguistics and also for our purposes here. However, (2) has always appeared problematic because semantic change seems to be far less generalizable than any other type of language change. In contrast to historical changes at other levels of linguistic description, semantic change has proven to be an area where regularities in the form of systematic changes are difficult to find (cf. Anttila 1989: 147). The generalizability of sound change for instance means the possibility of finding regularities and establishing laws, i.e., describing certain systematic changes in the history of languages, whereas semantic change is essentially sporadic (Hock and Joseph 1996: 244). Changes in the area of morphology and syntax also exhibit systematic effects on which generalizations can be based and described as major events in the history of a language. In the case of semantic structure we are looking in vain for such generalizations of change because it is not comparable to other levels of linguistic structure on the same terms. The expectation of such regularity in semantic change is probably due to the influence of the misconceived checklist approach to meanings, falsely imported from phonology.

Although semantic changes can be described in terms of a few well-defined types of possible processes, like metaphor, metonymy, etc. (König and Siemund 1999: 237), sound laws capture processes of regular and systematic changes (of course with again regular and systematic exceptions) through a whole language system, something unthinkable for semantic changes. That the system characteristics are missing from semantic change is due to the fact that it involves an open-ended set of linguistic elements, namely lexical items, whereas other types of change (at other levels of linguistic structure) involve closed system items (i.e., limited sets of elements) (cf. McMahon 1994: 185). In the case of lexical fields systematic changes may occasionally occur (Hock and Joseph 1996: 245). Anttila (1989: 146-147) (cited also in McMahon 1994: 186) provides an interesting example of such a regular shift from Latin legal terminology.

Another very important reason is that meanings refer to mental content and are thus not characterizable purely through aspects of linguistic structure. This is why knowledge of the sociocultural


138

history of the speakers of a language is very often indispensable for discovering etymologies, but even so it will only provide explanations for individual cases (cf. Anttila 1989: 137, and Campbell 1998: 267). It is especially true in the case of semantic change, as Keller (1985: 234) has pointed out, that a proper understanding of language change requires that we recognize language as an “object of sociocultural evolution.” Only this approach combined with the necessary cognitive analysis, to be proposed on the following pages, will help us gain insight into the essence of semantic change, i.e., (3) above.

The generalizability of linguistic change also means the discovery of certain well-established mechanisms of change. As for this kind of generalizations, semantic change is not in a worse position than other types of change. Several mechanisms and directions of semantic change have been identified and various causes of semantic change have also been distinguished. Thus, e.g. McMahon (1994: 178-184) gives the following classification, based on traditional work done on semantic change. She distinguishes changes (1) according to range of meaning: (i) extension, (ii) restriction; (2) according to the attitude of the speakers and hearers: (i) pejoration, (ii) amelioration; (3) according to whether the change results from (i) similarity of meaning (metaphor) or (ii) contiguity of meaning (metonymy); (4) according to causes of the change: (i) external (historical, social, cultural, technological, etc.), (ii) internal (linguistic), (iii) psychological (exaggeration and emphasis, expressiveness and creativity, euphemism and taboo).

Other basic and general works on historical linguistics list the same types of semantic change, but often differ in their groupings. Thus, Hock (1986) lists metaphor and other figurative language use as the basis for change of meaning, while broadening and narrowing of meaning and melioration and pejoration are considered results of semantic change. Anttila (1989: 148) also suggests that semantic change can be classified according to range and evaluation and treats figures of speech as mechanisms of the change. A very comprehensive classification with an extensive list of examples can be found in Blank (1997) and Campbell (1998: Chapter 10).

Traugott (1985) claims that semantic change also has its regularities that allow for significant generalizations. König and Traugott (1988) identify a special type of semantic change in


grammaticalization: interpretive augmentation, which is based on maxims of cooperative interactions. There have also been attempts at generalizing semantic change in other ways. Job (1982) proposes a typology of semantic change on the basis of a fixed set of categories which at the same time describe and explain the change. Geeraerts (1983a) classifies semantic change on the basis of functional factors that are connected with the communicative purposes of language.

Although semantic and lexical change are generally distinguished from each other, I will not keep them strictly apart in the discussion below because the cognitive processes are essentially the same behind both historical developments (cf. also Blank 1997: 112, and Geeraerts 1997: 84). By definition, semantic change occurs when the meaning of a word (or morpheme) changes over time, while lexical change refers to the emergence of new items in the lexicon of a language, which often involves morphological processes. A distinction between semantic and lexical change may be a useful one if we look at changes in the language system as such. However, in many cases no clear distinction can be made between the two processes and there is also an obvious connection between the two processes. On the one hand, in a number of semantic changes the older meaning is also retained, since a state of polysemy appears to be a prerequisite for a subsequent change, which may give rise to cognate sets. In such instances the original form with the new meaning will function as a new item in the lexicon of the language (cf. Anttila 1989: 134-135, and Zgusta 1990: 390). Hock and Joseph (1996: 297, 313) refer to this process as coinage through semantic extension or change. The question whether such a new item will be recognized as polysemy or treated as homonymy will be discussed in more detail in Section 6.6.

Wilkins (1996: 267) considers semantic change a subtype of lexical change on the grounds that a lexical item is made up of its form, meaning, and combinatorial properties, and a change in any one of these constitutes lexical change. On the other hand, when a new word emerges through morphological processes (mostly derivation and compounding), in an overwhelming majority of the cases semantic change will also be involved (cf. Geeraerts 1997: 95). Darkened compounds provide good examples. Eng. window, although a borrowing from Old Norse (ON vindauga), is originally a coinage from Proto-Germanic *windaz ‘wind’ and Proto-Germanic


140

*augon ‘eye’ (Palmer 1978: 344; Onions 1966: 1008; Watkins 1985: 45, 73). However, even at the time when window was still transparent morphologically, the meaning of this new word could not have been purely compositional but must have involved some kind of semantic extension. This obvious connection between semantic and lexical change derives from the underlying cognitive processes (cf. Blank 1997: 112, and Geeraerts 1997: 84). Langacker (1987: 156) has pointed out that valence relations in compounding cannot be a simple function of meaning compositions but must involve encyclopedic knowledge as underlying the meaning representations of the parts. Thus, the understanding of such a new meaning on the basis of the original meanings involves associative processes rather than simple algorithmic operations. In the case of metaphor these associative processes comprise a blending of conceptual domains from which the emergent meaning emanates (Coulson 2001: 165).

With the above generalities in mind I will now turn to some general issues regarding the connections between semantic change and cognitive processes. 6.3 Cognitive processes and the actuation of semantic change A connection between cognition and semantic change becomes obvious when looking at the function of Rosch’s (1978: 28) principles of human categorization. She claims that these principles guide the formation of categories in a culture. Although Rosch does not talk about semantic change or lexicalization, and linguists have rarely described semantic change as cultural category formation, semantic change appears to be the linguistic manifestation of cultural category formation for the following reason. Meaning extension and restriction obviously parallel category extension and restriction, and metaphor and metonymy are important cognitive devices for creating categories in an attempt to make sense of the external world (Lakoff and Johnson 1980; Johnson 1987; Lakoff 1987). Warren (1999: 218, 220) also acknowledges the close connection between forming meanings and forming categories, but she claims that words containing evaluative-attributive features of meaning (non-criterial ones) do not categorize. In her opinion only


words containing defining (i.e., criterial) features categorize because only the latter fix reference, whereas the former, like good, bad, beauty, folly, wisdom, justice, right, wrong, etc. (all Warren’s examples), do not. Arguing for a ubiquitous connection between forming meanings and forming categories, it could be remarked, on the one hand, that these have always been treated as categories (objective or subjective in the philosophical sense) of fundamental interest in the history of philosophy and not simply evaluations (although even evaluation judgments do categorize). On the other hand, more of interest to our line of reasoning, etymologies reveal categorization processes also in these cases. By way of example consider the semantic development of PIE *weid- ‘to see’ into German wissen ‘to know’ and Eng. wise (Drosdowski et al. 1963: 769; Watkins 1985: 74), i.e., wise are those who have seen (much).

It should be made clear at the beginning that the alteration in the usage of an expression across a speech community as a cumulative result of the communicative activity of its members can only indirectly be characterized as a cognitive phenomenon, since the phase where cognitive mechanisms and principles of categorization operate in the process of semantic change is obviously its actuation in the communicative activity of individual speakers. This activity (or linguistic behavior) is governed by communicative motivations, which must go hand in hand with cognitive ones because the goal is to pass on mental contents effectively. The effort to convey ideas in an appropriate and effective way may lead to the production of occasion-bound meanings in the language. In general, such new variants of meaning arise when speakers feel the need to refer to parts of reality for which no conventional name exists in the language or wish to express new ideas or views or emotional stances about their environment and deem the available conventional expressions unsatisfactory for that purpose (cf. Anttila 1989: 139, and Zgusta 1990: 390). Blank (1997: 146-147) describes this situation as speakers possessing a “potential designatum” to which they need to find an adequate sign. In this way it is the individual speaker who actively (though not necessarily consciously) induces the change in an attempt to make his mental model of the world accessible to the hearer. It should be remarked here, however, that there may be cases of change that will not be explicable as a result of communicative needs but involve misinterpretation of meaning (see


142

Section 6.1). However, these also seem to require an explanation in terms of specific, though different cognitive mechanisms.

The communicative motives for inducing changes in meaning are quite obvious (see e.g. Blank 1997: Chapter VI). Historical linguistics has offered plausible causal and teleological explanations in the form of intralinguistic and extralinguistic factors that are responsible for the speakers’ linguistic behavior when altering their usage of language and thereby actuating change of meaning (see e.g. Algeo 1990). In the case of external causes of change (e.g. historical, sociocultural, etc.) the communicative reasons for the introduction of new meanings are self-evident, but intralinguistic causes also have essentially communication-related roots. Changes related to the elimination of homonymic clash, differentiation of synonyms, or chain shifts within lexical fields all stand in the service of making the language more efficient for communication. Even causes that have been described as psychological, such as “boredom” with a word (e.g. in the case of slang), allow the detection of communicative motives. Keller’s (1985: 233) communicative maxim “Speak in such a way that people pay attention to you” may definitely induce such changes. Communicative efficiency motivates even such processes of semantic change as grammaticalization, development into discourse markers, etc. (Traugott 1990; Traugott and König 1991; König and Traugott 1988). According to Anttila (1989: 181) “the driving force [behind language change] is the mental striving to adapt language for communication with least effort, that is, the psychological motive and the necessity of fulfilling the functions of speech.” And the function of speech is the continuous adjustment of individual mental models across a community.

In the following I will explore the special characteristics of cognition that provide the motivational basis for the production and comprehension of semantic innovations and thus influence the way semantic change is designed. These are the characteristics which underlie the sharing of individual mental models in communicative interaction in order to produce culturally valid models of reality.

When producing occasion-bound meanings as new variants due to specific communicative motivations, intelligibility by communica-tion partners is a necessary requirement (cf. Palmer 1972: 309), and shared knowledge must serve as an inevitable basis (Fritz 1998: 21).


Faltz has proposed a hypothesis for semantic change which shows that easy inference of one meaning from another, as facilitated by the context, plays a crucial role in the change:

If expression X is interpreted as A at a certain time, and if for (many? all?) contexts k which are typical for X, it can be easily inferred from A(k) that B, then a possible semantic change is for X to come to mean B at a later time. (Faltz 1989: 318)

For the preservation of intelligibility such easy inference is necessary; after all, the speaker specifically invites the hearer to make the inference (see Traugott 1999a). The speaker’s ultimate goal is to convey his/her ideas and be properly understood and he/she is thus obliged to facilitate the hearer’s inferencing process. This will inevitably put specific constraints on the ways communicative ends may be accomplished. Since inference and also understanding, which is the ultimate goal of communication, are cognitive acts, these constraints will be shaped by cognitive factors.

Since the crucial context for semantic change is speaker-hearer interaction, Keller (1985: 232) has rightfully emphasized the role of the maxims of communication for language change. However, Keller does not take the cognitive dimension of language into consideration. It is not the whole truth that “[t]he form in which language exists is the ability of each individual to communicate with others” (Keller 1985: 233). Language exists also as the ability to possess knowledge about the world, i.e., to create and operate a mental model of reality, which, by its nature, can be shared by a whole speech community. Acting according to the maxims is only effective if the speaker’s and hearer’s minds share the same cognitive structures and if these structures can also be activated in parallel. Keller (1985: 234) claims that language changes because the maxims of communication create new variants that are then selected for. However, this cannot be the whole story. The maxims of communication must be based on the cognitive function of language in the sense that communication is about speakers and hearers exchanging parts of their models of reality. Proper communication can only arise when the right cognitive structures are activated in both hearer and speaker. When speaking about speakers


144

and hearers acting according to the maxims of communication, it is not enough to add that communication is a cognitive phenomenon, because cognition, and cognitive processes, are not action in the same way as communication is. This means that the intentional levels involved are different. The importance of this is that semantic change seems to involve even less intentionality than Keller admits, since at the basis of semantic change there are individual cognitive processes with a very low level of intentionality (if any at all). Linguistic communication is a conscious and intentional activity whereas cognition is not. We are not aware of our cognitive processes but we are of the fact whether we are using language or not. Perceiving some kind of phenomenon already involves categorization and “deciding” whether an object is e.g. a chair (or a “glittering thing” or a “brown one,” for that matter, cf. above) or not is an automatic activity of the mind. Furthermore, even if using a metaphor may be a conscious action, processes like associations and the perception of analogies that lie at its basis are not.

Thus, when the maxims of communication operate, i.e., speakers act according to them, underlying cognitive processes operate in the speakers’ minds along with the maxims. When we communicate, we are only seemingly concerned with the world directly; in fact it is our knowledge of the world that we communicate about. Nerlich (1989: 175) assumes three types of knowledge organized in a hierarchical system of layers. The semantic knowledge base holds the conventional meanings, the meta-semantic knowledge base (or expert system) consists of the procedures for linguistic innovation, such as metaphor and metonymy, while the world-knowledge base includes referential frames, schemata of concepts and the like. Both the semantic and world-knowledge bases provide input for the meta-semantic expert system, which through universal cognitive procedures helps the speaker “[m]ake as much (new) sense as possible in [… a given …] situation” and helps the hearer “[m]ake as much sense as possible from [… the given …] utterance” (Nerlich 1989: 180). Thus, when the speaker uses an occasion-bound meaning, he extends the meaning of a word – in the sense of deriving a new but related meaning from a basic one, and not in that of increasing the range of its denotata – on a motivational basis, while the hearer, in turn, must be able to realize that a word is used in a different sense from the


conventional one and use the same motivational basis for recovering this new sense and for storing it for recognition on subsequent occasions. In Langacker’s terms, the hearer must at least partially sanction the speaker’s usage. One of the most efficient ways of creating and recovering the motivational basis of unconventional meanings is to exploit the analogical capacities of the human mind (see Section 5.3), mostly in the form of metaphorical meaning extensions. Analogies will provide an economical but effective motivational basis for the speaker’s own representational process and for guiding the hearer’s understanding, i.e., for serving communicative ends and at the same time fulfilling the requirement of intelligibility. Based on work by several other researchers, Johnson and Henley (1992) have shown how efficient we are in making sense of even randomly generated analogies with the help of a relatively small number of relational concepts, which are experientially based.

In the following I will undertake an examination of the specific cognitive factors that are responsible for the particular avenues semantic change takes and how they operate in the actuation of the change. 6.4 Cognitive factors guiding semantic innovation In this section I will discuss four cognitive factors, which guide speakers’ semantic innovations. These factors operate in categorizat-ion processes, which appear to play an important role in such innovations. I will therefore start by examining the parallel aspects of semantic change (or lexicalization) and categorization in more detail in order to gain a deeper insight into the motivational basis of meaning transfer.

As we have seen in Section 5.3, reorganizing familiar knowledge structures is a very efficient way to meet emergent referring and representing needs. To judge by the evidence of etymologies, these communicative and cognitive needs will be met at the linguistic level by occasion-bound changes in meaning. When new ideas are expressed in terms of familiar information through non-conventional language use, both the new knowledge is conveyed and the desired intelligibility is achieved. This economical procedure


146

is a crucial factor in linguistic (and probably all other types of) communication; however, an optimal degree of explicitness must also be attained. As Carroll (1985) has shown, the naming of features of a referent is an expedient method for keeping communication both economical and efficient at the same time. This can be accomplished either through metaphorical or some other kind of figurative language use often coupled with some morphological operation on lexemes, like derivational affixation or compounding. This procedure, the coding of new concepts in the lexicon, is often referred to as lexicalization and obviously amounts to the formation of new conceptual categories on the cognitive level, since a new meaning will express a new category. This is the way semantic change can be considered a cultural category formation process (see Section 5.3).

However, at first glance not all semantic change appears to be analyzable as a categorization process. An obvious exception is constituted by the above-mentioned cases of the misinterpretation of meanings leading to semantic change in the course of time (see Section 6.1). Here no new meaning seems to be created, though a closer look might reveal some kind of miscategorization, which is not an uncommon cognitive phenomenon. Much more intriguing in this respect is the question of grammaticalization, which indisputably entails change on the semantic level as well (Traugott 1990). Heine et al. (1991) have pointed out the important role of cognition even in grammaticalization. In particular, they have shown that the “primary function [of grammaticalization is] conceptualization by expressing one thing in terms of another” (p. 150) and that “the relation between input and output [... of grammaticalization processes ...] is metaphorical in nature” (p. 151). Furthermore, Heine et al. have also proposed to describe the process of grammaticalization “in terms of a few basic categories which can be linearly arranged in the following way: PERSON > OBJECT > PROCESS > SPACE > TIME > QUALITY” (p. 157) and called the relationship between them “categorial metaphors, like SPACE IS AN OBJECT or TIME IS SPACE” (p. 157). Thus, even grammaticalization as a type of semantic change can be interpreted as conceptual categorization.

For example, Traugott and König (1991) have shown that the semantic change manifest in the development of causals,


concessives, etc. is rooted in the conventionalization of conversational implicatures. However, as much as implicature is a pragmatic phenomenon, it necessarily involves cognitive processing. Since the cognitive function of language is the categorization of experience (e.g. Geeraerts 1997, 7-8), it should not be surprising that categorization can even be detected at the basis of semantic change involved in grammaticalization. Traugott and König (1991) have shown how the conjunction while developed a concessive meaning beside its temporal meaning from the Old English period on. This change was due to the fact that on the basis of temporal overlap the inference could be drawn “that the conditions specified in the subordinate clause serve not only as the temporal frame of reference for those in the main clause, but also as the grounds of the situation” Traugott and König (1991: 201). Traugott and König (1991: 201-202) also remark that concessive “while is still awkward for some speakers in contexts expressing anteriority of one event to another [, which] shows that this conjunction has not entirely lost its original meaning of temporal overlap.” Obviously, this temporal overlap is a prominent feature of our concept of concession and it can be used to conceptualize and refer to the latter, as also exemplified by Hungarian ugyanakkor ‘at the same time’ and míg ‘while’ and German während ‘while’, all having a concessive sense as well. However, other such relations can also be inferred from temporal overlap. It is an everyday cognitive phenomenon to think that when two events coincide, they are causally related because a salient feature of cause seems to be temporal coincidence. In German this kind of inferencing yielded a causal sense in the development of the connective weil (cognate of while). In the sentence Ich bleibe heute zu Hause, weil ich krank bin [I’m staying at home today because I’m sick] the relation between temporal coincidence and cause is still transparent. Here, as part of the grammaticalization process, the conceptualization of cause in terms of temporal coincidence and the reference to one by the other can be witnessed. These examples show that categorization is at work even in grammaticalization and that this process of semantic change also has an onomasiological side.

Efficient reference to and representation of a given phenomenon are most easily achieved if they are based on actually recognizable properties. Thus, a suitable expression applied to this end will


148

characterize that phenomenon either from perceptual or functional or spatial or temporal or other relational aspects, i.e., by naming such a property (or properties) and creating thereby a new conceptual category of the form SOMETHING EXHIBITING PROPERTY X (cf. Carroll 1985). The choice of properties to be employed for reference and representation will be influenced by the speakers’ interaction with their environment, which may range from the mere perception of real world attributes (if this is at all possible without functional considerations) to the recognition of certain attributes as functionally relevant for a particular kind of behavior. But the choice of properties must also be governed by some kind of joint salience for both speaker and hearer based on a common ground (cf. Croft 2000: 93, 100). Herein lies the relevance of our capacity for the alternate construal of scenes (Langacker 1987: 138). By being able to conceptualize situations in alternate ways we can adaptively interact with our environment with maximal (or at least optimal) efficiency and language plays an important role in this (cf. Palmer 1996: 52).

Medin and Wattenmaker (1987: 25) have termed the search for constraints that may become embodied in organisms due to their interaction with their environment cognitive archeology. We are doing the same in a very specific sense when we look at particular meaning changes or instances of lexicalization by investigating etymologies. We will not simply find information about how particular phenomena were referred to originally, but we will also be able to infer details concerning the particular conceptual construals of phenomena that have with time become conventionalized and coded in the language. In other words, we will learn about how the results of human interaction with the environment have become embodied in the language in the form of fossilized conceptualizations. E.g. the construal inherent in the etymology of German Wand ‘wall’ (< PIE *wendh- ‘to turn, wind, weave’) (Watkins 1985: 76; Drosdowski et al. 1963: 753, 767) gives us a hint about such an interaction, namely how walls were erected at one time: they were woven from branches. A semantic change like this one involves a change from one word class to another. This may be due to derivation, which may have played a role in the change (see also above in this section) and then became obscured by sound change. It may also be the result of ellipsis, which often plays a


major role in semantic change (cf. above Eng. bear < PIE *bher- ‘bright, brown,’ i.e., originally ‘the brown one’). Thus, German Wand could originally have meant ‘the woven thing.’ Anyway, the associations facilitated by the conventional meaning of the original term will serve as the motivational basis on the grounds of which a conceptual connection can be made with the new referent. The term so employed may undergo semantic change due to frequent usage in this association.

At this point the question may arise why speakers choose one conventional expression over another in order to express a particular new idea with it. Having the whole system of language in mind, we can also ask why semantic change takes certain avenues and not others. After all, several conventional expressions may present themselves for reference to the same phenomenon and the same expression can be used to refer to different phenomena. In other words, the same phenomenon can be categorized in different ways, other things being equal, while a certain feature can be characteristic of phenomena falling normally under different categories. Kövecses (2000) has shown that not only the same target domain can be characterized by different source domains, but also how the same source domain can be mapped onto different target domains, of course with a different focus. There appears to be nothing ultimately compelling about referring to something in one particular way. For instance, Eng. glass ‘a vessel for drinking’ derives its name from the material it is made of, while Hungarian üveg ‘the material glass’ has come to denote metonymically a different object made of this material, namely bottle. Theoretically, other metonymical expressions could have done the job just as well. Kövecses and Radden (1998) have made a detailed analysis of the specific cognitive and communicative principles that govern the selection of vehicles in novel metonymic expressions. Their results show that three determinants of conceptual organization, human experience, perceptual selectivity, and cultural preference, constrain the cognitive principles, while the communicative principles originate in the need for clarity and relevance.

The fact that the choice of an expression for reference and the judgment of its appropriateness is within the individual speaker’s scope of decision is obviously one manifestation of “[t]he perspectival nature of linguistic meaning [, which] implies that the


150

world is not objectively reflected in language” (Geeraerts 1997: 8). Thus, subjectivity and subjectification must naturally also play a role in semantic change. While Langacker’s (1999) notion of subjectification is cognate with the line of reasoning followed here only in the relatively wide sense that subjective construal or altered perspective can also play a role in the construction of occasional meanings, Traugott’s historical-functional perspective is more applicable. According to Traugott,

[s]ubjectification involves speakers recruiting forms with appropriate meanings to externalize their subjective point of view. This is an activity that draws on cognitive principles but takes place in the context of communication and rhetorical strategizing. (Traugott 1999b: 189)

Thus, subjectification is an important force in language change due to “the attempt on the speaker’s part to increase the informativeness to the interlocutor of what is being said, i.e., a cognitive-communicative motivation” (Traugott 1995: 49).

The linguistic processes of semantic and lexical change are based on a reorganization of categories at the conceptual level with the help of such cognitive mechanisms as metaphor, metonymy, extension, restriction, etc. These mechanisms of human cognition are all used to create new conceptual categories or modify old ones and the way this is done is reflected in the linguistic processes in which they are manifest. Because of this, essential parallels can be detected between linguistic category coding (lexicalization) and conceptual category formation. At the linguistic level categorization is always explicit to a certain degree. The explicitness manifests itself in marking certain features (though usually not more than one or two) of the category via a coding expression while others stay implicit. E.g., as the etymology of Eng. gold (< PIE *ghel- ‘to shine, glitter’) (Watkins 1985: 21) shows, GLITTERING (or SHINING) is the explicit feature that served as the basis for the coding. Though only part of the complete conceptual construal is revealed in this way, this parsimonious solution is probably due to a reflection of cognitive economy, one of the principles of category formation proposed by Rosch (1978). As “all that glitters is not gold,” i.e., gold is not the only glittering thing in the world, there are clearly also


other features participating in the categorization, though only implicitly.

In many cases the high correlational structure of attributes probably yields such a high level of feature integration in the category that when activating one feature, the totality of the connecting features is also activated (cf. Rosch 1978: 29, and Langacker 1987: 385). This must be the explanation for the fact that a coding expression will still function appropriately when naming only one or sometimes two features. But this leads to a new question, probably one of the most important ones, in connection with characterizing the cognitive processes involved in semantic change. We must identify the factors that influence or maybe even determine the choice of a feature, or features, that will be elevated to the function of coding expression and by this being assigned the activating role. Below I will take a look at four such factors: 1. Cue-validity; 2. Cognitive economy (already mentioned above); 3. Perceived world structure; 4. Conjunctivity. These appear to be very much in line with Kövecses and Radden’s (2000) cognitive and communicative principles mentioned above. As the most central factor I would like to suggest the cue-validity of the features inherent in a category. Rosch defines cue-validity as

a probabilistic concept; the validity of a given cue x as a predictor of a given category y (the conditional probability y/x) increases as the frequency with which cue x is associated with category y increases and decreases as the frequency with which cue x is associated with categories other than y increases. (Rosch 1978: 30)

The reason why a speaker’s choice of a novel expression for a phenomenon that needs to be communicated about but lacks a conventional expression in the language will be influenced by the cue-validity of the features of the phenomenon appears to be the following. It is obvious that a speaker will try to make the reference as precise as possible, essentially in accordance with the maxim “Speak in such a way that the other person understands you” (Keller 1985: 233). Pointing out a salient characteristic of that phenomenon should prove to be a very good strategy for this purpose. A feature with a high cue-validity will more likely call to mind the category in


152

question than a low cue-validity one and so it will be more suitable for reference because the hearer’s attention will be more easily directed to the referent. Furthermore, the optimality of such a feature (or features) for mentally representing the category of the referent for the speaker herself is self-evident.

But it should be borne in mind that even here our analogical way of thinking presides: the new phenomenon is perceived as bearing a likeness to other things characterized by the feature in question. Salience is always a function of existing knowledge: the features we discern as salient must be ones that we already have knowledge of as separate categories (cf. Geeraerts 1997: 44). As Rosch (1979: 29) says, “[o]ne influence on how attributes will be defined by humans is clearly the category system already existent in the culture at a given time.” Any linguistic innovation is therefore accomplished with the help of already conventional devices in the language (cf. Croft 2000: 104), which naturally puts specific constraints on such innovations, like in what linguistic domains they can be used and what subjective functions they can perform (Traugott 1995: 46). Examples are abundant. Thus, a hat is a thing that shelters (Eng. hat < PIE *kadh- ‘to shelter, cover’) (Watkins 1985: 26), a thumb is a thing (or one of the digits of the hand) that is swollen (Eng. thumb < PIE *teu- ‘to swell’) (Watkins 1985: 71), a hawk is a “thing” that grasps (Eng. hawk < PIE *kap- ‘to grasp’) (Watkins 1985: 27), etc. In the case of such solutions to reference problems, Croft (2000: 108) talks about the creation of salience because several terms might “successfully establish reference … and the properties chosen thereby become more salient” (Croft 2000: 109) (cf. also Fritz 1998: 21).

The next two factors playing a role in the speaker’s selection of the appropriate feature(s) for reference and representation are ‘cognitive economy’ and ‘perceived world structure,’ which Rosch (1978) mentions as the two most important general and basic principles of human categorization. These are closely connected to cue-validity, since using a feature with high cue-validity will also (by definition) satisfy the principle of cognitive economy because the processing of the category on the basis of such a feature will require the least effort. The role of the principle of perceived world structure may be even more obvious than that of cognitive economy. Since speaker and hearer share the same perceptual organization at


the physiological level, perceived world structure has a powerful influence on which features of particular phenomena we experience as salient and these will naturally be the ones that cue a given category most unambiguously. It may happen that perceived world structure is over-ruled by functional considerations, i.e., when knowledge of the particular function of an object dominates over the information picked up by the sense organs due to the characteristic interaction with a given phenomenon. E.g. the etymology of Eng. rain (< PIE *reg- ‘wet, moist’) (Watkins 1985: 54) reveals a perceptual feature as salient on which the conceptualization was based, while a functionally salient feature was used for that purpose in the case of Eng. hut (< PIE *(s)keu- ‘to cover’) (Watkins 1985 60).

The fourth factor influencing the choice of referential features (or of any referential expression), which is also largely responsible for the above cases, has been identified by Brown (1979). He claims that the creative usage we find in naming behavior is influenced by conjunctivity. The cognitive links based on conjunctive relationships can be determined logically, without appeal to empirical conditions. Conjunctive relationships are psychologically more salient on logical grounds and have thus possible cognitive priority. According to Brown (1979: 259) transitive relationships are prominent examples of conjunctivity: “[T]wo entities are conceptualized as bearing an immediate transitive relationship to one another” e.g. in “kind of” and “part of” relationships, which often underlie meaning restriction/extension as well as metaphor/metonymy and are thus utilized in naming behavior. It should not be too difficult to detect an affinity to cue-validity here either. A feature capable of cueing a category through its salience is usually one that connects the category to other categories exactly through “kind of” and “part of” relationships. Thus, as we could see in the already mentioned example of thumb, the feature SWOLLEN, which serves to identify and categorize it, also relates it to a larger category as a “kind of” finger (or thing), namely a swollen one. Similarly, a “part of” relationship can be detected in the name for Hungarian farkas ‘wolf,’ which can be analyzed into farok ‘tail’ and the derivative suffix -as forming adjectives from nouns and meaning approximately ‘having [something], with [something].’ Thus, the etymology of the Hungarian word for wolf is quite transparent: ‘[the


154

one] having/with a tail.’ Conjunctivity seems to be an even more general constraint in naming behavior than Brown claims, since it occurs in forms other than merely “kind of” and “part of” relationships. E.g. temporal/causal conjunctivity appears to be responsible for the development of PIE *weid- ‘to see’ into German wissen ‘to know’ and Eng. wise (Drosdowski et al. 1963: 769; Watkins 1985: 74), since knowledge occurs in conjunction with seeing, or in other words, knowledge is caused by seeing (i.e., knowledge comes about through perception). Several other such associative relations (e.g. provenience, contingency, function, spatiality) are listed by Casagrande and Hale (cited in Palmer 1996: 93), which can guide naming behavior, as clearly illustrated by Palmer’s examples.

For simplicity’s sake I have talked about features in the above explanations. However, in many cases of semantic change it is not really a particular feature of a category that is involved. Changes like meaning extension and restriction are not examples for highlighting salient features. Here the relevant feature will usually stay implicit and it will be the “kind of” relationship that will be highlighted directly, as in the development from OE fugol ‘bird’ to Eng. fowl ‘a kind of bird, namely a domestic one.’ Such a development is obviously the result of frequent reference to domesticated birds with the superordinate category on the basis of the conjunctive “kind of” relationship mentioned above. Nonetheless, it should be obvious that the necessary motivational basis is supplied just the same and the factors of choice listed above are valid in these cases as well. E.g. providing the information that a robin is a bird or that a bird can be for instance a robin will do the same cueing work as individual features. Broader and narrower categories are capable of cueing each other on the basis of overlaps within the same perceived world structures, thus making the processing of the necessary category cognitively economical. In many cases of metaphor and metonymy the bases for the derivations of meaning are not some kind of features either but again some conjunctive relationship between two phenomena. The perceived world structure of phenomena will naturally generate the recognition of various types of contiguity opening up avenues for metonymy (cf. Nerlich 1989: 179). As for metaphor, e.g. Hungarian fiók ‘drawer (in any type of furniture)’ is cognate with the words fiú ‘boy’ and fióka ‘young (of birds),


nestling’ (with a diminutive suffix) and thus designates its referent metaphorically as the young of a piece of furniture. The above mentioned case of Eng. window could be cited as another example here for metaphorical categorization. Given the appropriate associative routes, the perceived world structure of one phenomenon will be able to cue that of another in cognitively economical ways because, as Lakoff (1990) claims, during the creation of a metaphor the cognitive topology of the source domain will be preserved in the target domain when the structure of the former is mapped onto the latter. The mapping relies on the consistency between the two domains or their parts. According to Coulson (2001: 165), however, the conceptual domains serving as inputs to a blending process need not be consistent in order to give rise to the emergent metaphorical meaning.

In the case of the WINDOW metaphor the two input domains to the blend are EYE as source domain and WINDOW as target domain. Actually, the real target domain is OPENING IN THE WALL OF A HOUSE, since apparently no conventional referring expression existed for that experiential domain at the time of creation of the metaphor. The two domains are connected by a generic space through sharing basic structural features: they are openings in the side of closed (hollow) objects. The blended conceptual space inherits these features and the mapping between eye in a head/skull and opening in a wall occurs. In this case the composition process leaves out some structural and functional features: the opening is not blocked as it is in the case of a real eye (though if rather the SKULL than the HEAD matrix is involved the situation might be different), and seeing is not projected into the blend from the source domain. In completing and elaborating the new frame in the blend the function of a real window – letting in the light – will be part of the constructed meaning not found in either domains. However, one could argue that not only structural analogy played a role in the emergence of this new meaning but also the experience that one can look out of the house through this particular opening. Anyway, though it might seem that the structural features of the EYE domain are directly transferred over to the OPENING domain, the new meaning is not the target domain itself but the blended domain, since a window is special type of opening in a wall with a special function.


156

The above mentioned factors (cue-validity, cognitive economy, perceived world structure and conjunctivity) are all parts of the motivational basis mentioned earlier because they greatly facilitate the reliance on familiar knowledge when new categories are processed. Even an individual feature that is selected to refer to a new category will have to exist as a familiar category in its own right in the speaker’s knowledge if cueing is to be accomplished with it. Furthermore, only familiar categories can surface as features of other categories in our perception of world structure and familiarity with them will naturally contribute to processing ease and thus be cognitively economical. In addition, conjunctivity of any type appears to be the simplest and most basic activator of an associative act. These are the factors that ensure that the new information provided by the speaker in the form of a new category be properly motivated so that an adequate conceptual support is available to the hearer for the construction of the same category. 6.5 Semantic change and semantic knowledge When creating and interpreting a novel expression, i.e., an occasion-bound meaning, both speaker and hearer must rely not only on cognitive aspects for a motivational basis but also on the productive rules of the grammar (Anttila 1989: 138). Therefore, the notion of usage should not only imply the extra-linguistic (i.e., pragmatic) activity of correct contextual usage but also a much more basic intra-linguistic process. It should also involve the speakers’ linguistic knowledge for the simple reason that any usage as an activity by an individual speaker presupposes the knowledge of how to implement this usage. The linguistic knowledge of how to produce correct expressions must entail the knowledge of how to use elements and rules of the language. An internal grammar can therefore be interpreted as containing instructions about the usage of linguistic elements and rules in order to form grammatical expressions for communication. In other words, the speaker must be able to anticipate the acceptance of an innovative usage and judge its likelihood of being sanctioned by the grammar (Langacker 1987: 65-66, 157).

In order to have an adequate understanding of language change,


we have to be aware of these two interrelated aspects of language use, i.e., pragmatic usage and grammatical usage. Semantic change is mostly triggered by extra-linguistic circumstances, which set the cognitive processes in operation which then produce a variation in contextual usage. If the given context-based usage persists, it can affect the structure of the language in the long run because the particular contextual usage may infiltrate the grammar under enduring conditions. In other words, the new usage will become “more and more deeply entrenched through continued repetition” (Langacker 1987: 100). This is how change occurs in the language system, i.e., in the speakers’ cumulative knowledge of correct grammatical usage: contextual knowledge will become grammatical (i.e., linguistic) knowledge with time (cf. Traugott 1990). This process has been called “conventionalization” (Langacker 1987: 65-66).

When looking at the process of change from this perspective, we will find some inherent circularity in the process (cf. Nerlich 1989: 175), which is manifest in the following. The apparent change in the system will be ultimately one in the knowledge of the speakers, but it is also the structure of this knowledge that is responsible for the form of further changes. On the one hand, the major input to the cognitive processes operating in the modification of meanings is semantic knowledge, i.e., the culturally valid and thus linguistically coded conceptual category system, or in Nerlich’s (1989: 179) terms, the contents of a semantic knowledge base. On the other hand, the nature of semantic knowledge, i.e., the way meaning is represented in the mind, will influence the way meanings change. Since it is the nature of meaning to change, semantic knowledge must be made up of malleable structures (cf. Geeraerts 1999: 95; Johnson 1987: 30), and general cognitive processes (which underlie the mechanisms of semantic change) must obviously have access to semantic knowledge or even be more directly incorporated in it (cf. Langacker 1987: Chapter 3). There must of course be certain constraints involved in this regard, probably reflected in the structural limits of semantic change.

Thus, semantic knowledge must be flexible enough to be able to accommodate changing circumstances and conditions, first of all for short-term purposes in everyday communication. The semantic knowledge of an individual speaker must be adaptable through


158

specific cognitive processes when they are induced by certain communicative requirements in order to result in an efficient communicative behavior, i.e., an innovative, though appropriate usage of the language. The semantic knowledge of an individual hearer, on the other hand, must be adaptable so as to sanction the new and hence deviant usage at the time of its occurrence at least to a certain degree as not totally meaningless because otherwise it would be discarded as uninterpretable. Thus, it is not only cognitive processes that play a role in the actuation of change but also the flexible character of semantic knowledge. An occurring deviant usage may not only be sanctioned by the semantic knowledge of the hearer, but it is the same knowledge that produces it in a speaker, who therefore must unconsciously be convinced of its acceptability for communicative purposes. Fritz (1998: 67) points out that a further step in the conventionalization process is when the new usage is not only considered collectively suitable but also as “correct.” In Hermann Paul’s (1920: 84) onetime explanation, the criterion for this correctness, i.e., for change of meaning to have occurred, is when speakers of a language can already interpret the earlier occasion-bound meaning of a word without resort to the original usual meaning, i.e., when the formerly occasion-bound meaning ceases to be identifiable as a derived meaning. In terms of cognitive grammar, this is the state when the new meaning has achieved unit status, i.e., it has become “sufficiently well entrenched that it is easily evoked as an integrated whole” (Langacker 1987: 100). 6.6 The cognitive basis of polysemy and its emergence in the lexicon The discussion in this chapter of the cognitive aspects of how semantic changes are induced has been characterized so far by an onomasiological perspective. It is now time to supplement that approach by examining the semasiological side of these processes, since the naming of new concepts inevitably goes hand in hand with the polysemization of lexical items. After having looked in the previous section briefly at the way semantic change is incorporated in the grammar, i.e., at the speakers’ knowledge of their language, I


will now take a look at how change is incorporated in the language system. I will offer a unified synchronic and diachronic explanation for the emergence of polysemy as an over-arcing process of language adaptation from strictly context-bound usage to complete semantic change, through which the connection between the cognitive and linguistic aspects of this ubiquitous phenomenon in the lexicon of a language will become the most obvious.

When using familiar expressions for referring to new concepts or to novel aspects of old ones, and the usage gets conventionalized, these customary expressions will become polysemous. The cognitive background of this is clearly the analogical workings of the human mind. Furthermore, the prototypical character of meaning structures determines the ways changes might happen at all. Geeraerts (1983b, 1997: 23 and passim, 1999) has identified four characteristics of prototypicality which in specific ways influence semasiological change. Thus, due to the fact that not every member has equal status in a category, changes in the referential range of a category will primarily affect central cases. Furthermore, since there are more and less salient readings of a prototypical semantic structure, changes will necessarily affect the radial set structure of a category. Due to the fuzziness of prototypical categories various transient readings of a lexical item may emerge over time. And finally due the use of non-necessary attributes in prototypical categorization changes in word meaning often result from the encyclopedic knowledge about members of the category. In general it can be said that it is the prototypical nature of semantic structure that is responsible for the polysemic character of lexical items (Geeraerts 1997: 123-124).

The onomasiological approach to semantic change predicts on cognitive grounds that new meanings primarily develop as the result of a (re-) describing or (re-) naming process, which reflects an emergent common conceptualization within a shared cognitive model in a speech community. With reference to what has been said above on the cultural formation of categories and on the actuation of semantic change, we can construct a general cultural scenario for this linguistic process from an onomasiological point of view. Changes in the natural or sociocultural environment of a speech community can

(i) create new entities, or


160

(ii) make hitherto unknown ones known, or (iii) make already known ones be seen in new lights.

If these entities are in any way important at the level of a whole culture, then they will be designated in some way for purposes of communication about them. Etymological examples reflecting the above processes are not difficult to find: thus for (i) we have beside the already mentioned forms of Eng. hut < PIE *(s)keu- ‘to cover’ and hat < PIE *kadh- ‘to shelter, cover,’ and Germ. Wand ‘wall’ < PIE *wendh- ‘to turn, wind, weave’, e.g. Eng. wheel < PIE *kwel- ‘to revolve, move around;’ as an example for (ii) we could mention Eng. glass and gold < PIE *ghel- ‘to shine, glitter,’ and hawk < PIE *kap- ‘to grasp’ (see above); whereas for (iii) Eng. cloud < PIE *gel- ‘to form into a ball’ and Germ. Wolke ‘cloud’ < PIE *welg- ‘wet’ could be mentioned as examples, since Proto-Indo-Europeans already had a word for ‘cloud’ – *nebh- – actually surviving in Germ. Nebel ‘mist, fog.’ It has to be remarked that many etymologies, though they are clearly manifestations of conceptualizations at the cultural level, cannot be unambiguously related to any of these processes. E.g. we can infer from Eng. rain < PIE *reg- ‘wet, moist’ that rain was conceptualized as something wet at some point, but it would be absurd to claim that this points to a time when rain became known. It may have been reconceptualized, but no expression that would reveal some earlier conceptualization survives (as it is the case with Germ. Nebel).

In the following I would like to merge the cognitive and linguistic aspects of change by investigating how polysemy networks originate and function in a language. The reason why an approach from polysemy will naturally link up the cognitive and linguistic levels of semantic change is the following. On the one hand, polysemy appears to be a consequence of the flexibility of human thinking and its properties follow from the structure of human cognition (Deane 1988), just in the same way as I have tried to show for semantic change. On the other, semantic change, and with it the alteration of the semantic structure of language, is obviously rooted in the emergence of polysemy relationships in the lexicon.

If we want to model the process of semantic change as it occurs in the history of a language, we cannot do this without considering the fact that polysemy is ubiquitous among lexical items (Hock and


Joseph 1996: 218). It is the rule rather than the exception. Polysemy relations may even emerge among derivational (and maybe also inflectional) elements due to specific cognitive motivations (Győri et al. 1998). Because of this, McMahon (1994: 176) even considers the existence of polysemy a major condition for semantic change. But there are also other aspects of the relationship between polysemy and semantic change. There is a basic congruence between the two phenomena in the sense that both constitute a relationship between meanings in which one (or more) are derived from another. If a lexical item undergoes semantic change, polysemy might form the first step in the process, with both the basic and the derived meaning existing in parallel (cf. Campbell 1998: 268). It appears to be a matter of the time that has elapsed since the point of the divergence of meanings whether a cognate relationship (if not obscured by sound change) is considered semantic change or polysemy (cf. Lee 1990). Thus, e.g. the words hide ‘skin’ and hide ‘conceal,’ both going back to PIE *(s)keu- ‘to cover’ (Watkins 1985: 60), are not conceived as polysemous anymore, though they clearly were at one time. On the other hand, the two rather distant meanings of the highly polysemous word run, ‘to go steadily by springing steps so that both feet leave the ground for an instant in each step’ and ‘to carry on, to manage’ are still considered related.

The claim that polysemy is not just a condition for semantic change (McMahon 1994: 176), but already the first phase of the change (cf. Blank 1997: 407) is supported by a positive correlation between the age of words and the extent to which they are polysemous. Lee (1990) found that older words (i.e., earlier attested ones) were more polysemous than recent ones and that polysemy evolves diachronically often through metaphorization. This means that the emergence of polysemy itself involves a process of change in which a word evolves a new meaning (cf. Algeo 1990: 403). Furthermore, Lee (1990) also found that frequently used words are more likely to develop polysemy. This may be the case because of the fact that “[a] frequently used word wears out its expressivity and novelty and is—in the long run—absorbed into the stock of the words of normal usage” (Nerlich and Clarke 1988: 78), i.e., it is ready to change its meaning (cf. Clarke and Nerlich 1991).

Surprisingly, Lee’s (1990) investigations also revealed that words with concrete meanings are not more polysemous than words


162

with abstract meanings, though this would have been expected on the grounds that the predictable direction of meaning derivations is considered to be from concrete to abstract. On the basis of this directionality one would expect concrete words to be used metaphorically more often than abstract ones and thus have more derived meanings in general. However, this finding in itself does not contradict our general idea about the direction of change and can probably be explained through the relative frequency of cases in which change goes from concrete to concrete and abstract to abstract, which phenomenon can readily be accommodated within a wider and more general concrete to abstract tendency.

Sweetser (1990: 1) lists pragmatic ambiguity, polysemy and semantic change as cases of “multiple form-to-function mapping.” I think that ambiguity, polysemy and changed meaning can also be considered as three successive stages in one and the same diachronic process, where one stage eventually leads to the next. This appears to be essentially the same as Blank’s (1997: 119) process model of semantic change, which proceeds through the steps of idiosyncratic innovation by a speaker, the usualization of this innovation as a rule of discourse, and finally lexicalization as a rule in the language. Fritz (1998: 65, 67) describes this process as forms of usage dynamics and also observes three corresponding stages: routinization, standardization and conventionalization.

Pragmatic ambiguity occurs when for immediate expressive purposes the speaker alters the meaning of a word usually supported by a given context, like in the case of a novel metaphor or metonymy. For instance many etymologically attested metonymical names for things, like “the glittering thing” for gold, “the covering thing” for hat, “the woven thing” for wall, etc. (see above), must have surfaced first in this way. If the usage of a word persists in a context in which it was formerly regarded as ad hoc, the pragmatic ambiguity will start to wear off and then cease with time while the word will emerge as a polysemous one. The word will become semantically ambiguous. This development already constitutes a change in the meaning of the word. It is not by chance that in most polysemy relations it can be determined which member of the group bears the more basic, and also chronologically earlier meaning and which are the ones that were derived from it. The word paper, for instance, has already reached the stage of polysemy, but it is still felt


that its concrete meaning ‘substance manufactured from wood fiber, etc. in the form of sheets used for writing, printing, etc.’ is the one from which the meanings ‘newspaper’ and ‘scientific article,’ among others, have been metonymically derived, probably through ad hoc usages for immediate expressive purposes. This process may even occur in the case of bound morphological elements where any further semantic development of the polysemy relation will ultimately result in functional change and constitute a case of regrammaticalization (Győri et al. 1998).

After a polysemy relation has emerged, it indeed serves as the condition for further change (Campbell 1998: 269). In the following I will try show how subsequent meaning changes may be completed based on the structure of polysemy within the lexicon. Langacker (1990: 266) has proposed a network model for representing the semantic structure of a polysemous lexical item. The nodes of the network define the different related senses and the relationships among these are symbolized by different arrows. There is a node that is the global prototype for the category as a whole, i.e., some kind of basic meaning, while other nodes represent its extensions (cf. also Campbell 1998: 269). As the network gets more elaborate, local prototypes may also develop. An elaborate network may be structured into sub-networks and the more peripheral nodes may be linked to the central node (the global prototype) via a node which is central to a subpart of the total network. Such a node is considered a local prototype. E.g. not all meanings of run are derived directly from the global prototype meaning ‘to go steadily by springing steps so that both feet leave the ground for an instant in each step’ but some are derived from already extended ones (Langacker 1990: 266).

Since the first step toward semantic change is the emergence of polysemy in a lexical item (i.e., polysemic split), this network model should also be suitable to simulate semantic change. The question is then what happens within the network when polysemy moves on to semantic change. Obviously, one extension from the global prototype, i.e., one of the senses of the polysemous lexical item in question, will get gradually detached from the original network. In terms of the underlying cognitive processes this should happen when a given entity which has been conceptualized formerly as a derivation from a prototypical center of a category is recognized as a


164

central member of a new category. This gradually detaching node is most likely to be one of the local prototypes, since they are the ones that have already moved furthest away conceptually from the global prototype and have acquired a relatively high degree of independence. Semantic change occurs in the polysemy network when one of the nodes (probably a local prototype) gets detached by being raised to the status of a new global prototype in a network of its own. I would term this process prototypicalization and claim that a local prototype constitutes an instance which has already started out on this course.

As an example consider the development from PIE *(s)keu- ‘to cover’ to Common Germanic *huson ‘covering for the legs,’ and then to Eng. hose (Watkins 1985: 60). In this case we may assume a category COVER with a prototype structure and claim that the global node of the network must have been the sense ‘a spreading over something.’ From this the garment hose (cf. German Hose ‘pants’) must have been conceptualized by extension. This type of special cover (i.e., leg covering) must have been culturally so salient that within the category network of COVER this particular extension acquired the status of a local prototype, i.e., a relatively independent sense. As further historical development testifies, the local prototype node representing the sense of ‘leg dressing’ must have become so salient that it got detached from the original network and became prototypicalized in its own right. In the status of a new global prototype it then gave rise to new extensions, like e.g. the metaphorical sense ‘flexible tube,’ which has nothing to do any more with the category COVER.

Considering the whole cycle of semantic change, prototypical-ization appears to be complementary to the process which Geeraerts (1997: 123) has termed polysemization. The semasiological basis of the process of polysemization are the four consequences of the prototypical nature of meaning structure for semantic change (Geeraerts 1997: 23), while its cognitive basis is “[t]he multiple actualizability of a prototypical concept into variously deviant nuances [... due to its ...] inherently flexible, dynamic structure” (Geeraerts 1997: 114). As shown above, such nuances, under the appropriate circumstances, may develop individual meanings, i.e., prototypicalize. These new individual meanings, due to their prototypical character, will then start out on the road of polysemi-


zation. Since polysemy – due to the way human cognition works

(Deane 1988) – is the natural semantic state of words into which all words will get with time (Lee 1990), any new prototypical center will eventually also start to build its own network. What remains to be explained is what the factors are that determine whether proto-typicalization will run its whole cycle and detach a node from its original network, or if a node will remain a local prototype at best. Geeraerts (1983b) has suggested that prototypical categorization may be the cognitive basis of the fluidity of meanings to allow for change. Instances of the category may lie at different distances from the prototype center depending on their degree of typicality and it is this prototypical organization of our conceptual categories which, through its dynamic nature, allows for this kind of flexible and adaptive categorizing behavior (Geeraerts 1997: 112 and passim, 1999: 98-99). E.g. the fact that German Hahn ‘rooster’ derives from PIE *kan- ‘to sing’ (Drosdowski et al. 1963: 244) suggests that a rooster used to be referred to and thus categorized as ‘something singing’ or as ‘the one that sings.’ Even though it is not a typical singing creature, this must have appeared as a cognitively plausible categorization and hence a good solution to the reference problem. The basis for such a solution is obviously the fact that “the prototype enables you more easily to understand peripheral instances of use that are novel to you” (Geeraerts 1997: 110). Thus, when a new instance is categorized, it may be recognized first as belonging to an established category, but due to its low degree of prototypicality it will be far away from the prototype center, just like the case of German Hahn exemplifies. This distance will determine the strength of its attachment to the category, i.e., how strongly it will be integrated. If it is too loosely attached to the category, it may get easily detached and recognized as a category in its own right, i.e., differentiated. Such a development seems to have taken place in the history of Hahn, since a rooster is far from being a typical singing creature. Seiler (1985: 117) has called this cognitive process “differentiating integration.” A new prototype, around which a concept can be organized, should easily be created in this differentiation process, since the observation of a single exemplar can already constitute an initial prototypical center for the emerging category (Zimmerman 1979: 64). These appear to be the underlying


166

cognitive processes that are responsible for the dynamism of a polysemy network and thus determine the behavior of the nodes representing the individual meaning instances of the polysemy relation.

Differentiating integration appears to be akin to analogy, making it easy for the brain to grasp any new knowledge in familiar terms. Polysemy and semantic change (i.e., semantic split) do not simply occur as a result of changed usage, but they constitute a cognitively motivated “derivation” of meaning founded in the semantic knowledge of the speakers (cf. Lakoff 1987: 345, and Sweetser 1990: 9). Deriving one meaning from another is a cognitively real process and the ensuing polysemy network must have an actual knowledge basis, i.e., the existence of polysemy “depends on the psychological reality or awareness of the speaker” (Anttila 1989: 181). This awareness of the congruity of meanings within the polysemy network may also facilitate the understanding and use of the related meanings. However, after some time speakers may not recognize the fact that certain meanings are related and the results of the original meaning derivations will not be mentally represented as polysemy. This is when their semantic relationship becomes obscured in the course of time and the new meaning becomes established in its own right. Their mental representations may therefore be said to have become homonymous. Therefore, we can talk of semantic change when the speakers’ awareness of polysemy has ceased. On these grounds the word pupil does not constitute a case of polysemy anymore but two distinct cases of semantic change from an earlier sense of ‘small person’ resulting in the senses ‘young student’ and ‘the black round aperture in the iris of the eye.’ This also implies that the difference between polysemy and semantic change is a matter of the speakers’ awareness, i.e., of their mental representation of semantic structure. Thus, semantic change may also be manifest in the existence of identical forms, and sound change is not a necessary requirement. This is of course not to say that a distinction between conceptual polysemy and “system” polysemy is not useful. The recognition of the latter is definitely justified and profitable in a diachronic investigation of the language system because it describes the linguist’s knowledge that two identical forms with different (and seemingly unrelated) meanings are historically related as opposed to the case when such forms are


not (cf. Algeo 1990: 403). The explanation for this phenomenon is that the connections in

the network are not rigid but flexible enough to make both the expansion of the network and its splitting in the form of semantic changes possible. Thus, a polysemy network does not represent a state but a process between the derivation of meaning and an eventual detaching of a node from the network in the course of time. It is this process character of the network that can explain how knowledge of meaning structures may by and by cease to incorporate semantic relatedness in the face of the lack of phonological distinction, as in the case of pupil.

The process of semantic change, as implied by the above analysis of the emergence of polysemy, can be basically twofold. The two types of the process can best be explained with an example from evolutionary theory. A species can evolve either by anagenesis or cladogenesis. Anagenetic evolution is when a species as a whole goes through genetic changes and evolves into a different new species because of environmental changes in the niche of the species. Cladogenetic evolution, on the other hand, is a bifurcation process. In this case a population gets (usually geographically) isolated from the rest of the species and evolves into a separate species as a result of different environmental influences, while the original species also goes on existing in parallel. Just in the same way there can be anagenetic and cladogenetic change in meaning. Anagenetic change is when a word with meaning x changes its meaning into meaning y (in the course of time), while meaning x ceases to exist. The process is as follows. Word w with the usual meaning x develops the occasional meaning x’. With time, this latter one develops into a new usual meaning y, while the old usual meaning x disappears completely, i.e., word w now has the meaning y. Of course word w may go through certain sound changes in the meantime. An example is OE wrítan ‘scratch, carve’ > Modern Eng. write. On the other hand, cladogenetic change of meaning is when a word with meaning x gives rise to a meaning y but also retains meaning x. The process again starts with a word w developing the occasional meaning x’ from its usual meaning x. However, while the occasional meaning x’ develops into a new usual meaning y, the old usual meaning is also retained. In this case there is usually also a bifurcation process in the sound form, which is a basic cause for the


168

etymology of a word to become obscure. An example is Eng. hut < PIE *(s)keu- ‘cover, conceal,’ while we still have Eng. hide < PIE *(s)keu- ‘cover, conceal.’ In both kinds of change the occasional meaning x’ can only be understood with respect to the usual meaning x. However, after a certain period of time speakers no longer have to invoke the usual meaning x, but understand the occasional meaning x’ on its own, and as a rule, they do not even know that meaning x’ is related to meaning x. This is the (theoretical) point when Paul’s (1920: 75) criterion for the emergence of a new usual meaning has been reached and when it makes more sense to speak about meaning y rather than meaning x’.

For simplicity’s and clarity’s sake this description is only a schematic one. Changes in meaning are usually much more complex processes and cannot be treated without considering sound change as well. Also, only one occasional meaning of a lexical item is reckoned with, although it is more usual to find bunches of cognates. Morphological processes are not included either in this schematic description, although the emergence of new meanings, especially in the case of coding new categories, usually goes together with some kind of word formation process: derivation or compounding. As examples for these we could mention Eng. belief and Germ. Glaube ‘belief,’ which are both derived from PIE *leubh- ‘to care, desire, love’ with a prefix attached, and also the obscured compound Eng. window, the etymology of which has already been discussed above. 6.7 Semantic change as adaptation process Anttila (1989: 179) remarks that change in language is a necessary phenomenon in order to adapt the language to new circumstances so that it remains a functional communicative and cognitive system. In this regard semantic change is especially important with reference to the conceptual category system coded in language. Therefore, semantic change can be regarded as part of the general adaptation process in language (see primarily Section 5.3), ensuing from the meaning extensions in innovative usage (as described in the previous sections), which can be considered a cognitive solution to what Geeraerts (1997: 102ff.) has described as the problem of efficient communication, i.e., the emergence of new referring and


representing needs. The reason for such problems to arise can be described – in accordance with Geeraerts’ (1983a; 1997: 102-108) ‘expressivity’ and ‘efficiency’ principles – as follows. As I have argued in Chapter 5, the principal cognitive function of language is the categorization of experience, through which language serves as a conventionalized symbolic model of the natural and sociocultural environment of a speech community. This model is usually well adapted to this environment and facilitates the proper exchange of beliefs, ideas, knowledge, etc. about it by providing a particular interpretation of the world in the form of different categorizations. But reality, and particularly our interpretation of it, does not remain stable through time. Changes which appear to be relevant at the level of a speech community demand language to adapt to these changes. This strongly suggests that semantic change – as the mechanism of this adaptation process – is responsible for the emergence of the different semantic structures (or category systems) of different languages.

However, the claim that language becomes adapted to the world around us does not refer to the results of historical change alone. Adaptation is going on continuously, since whenever we use language, we use it in such a way that it best describes the world for the purposes of our communicating about it. Thus, the adaptation happens spontaneously and unconsciously through the communicative interaction between speakers and hearers (Anttila 1989: 408; Keller 1994). This sociocultural function serves as the ground of the process of language change, and so there are two levels at which language is adapted to our physical or social reality. First of all, we can speak of a short-term (or synchronic) adaptation for communicative and cognitive purposes during everyday usage, i.e., when Nerlich’s (1989: 179) meta-semantic expert system is consulted for linguistic innovations due to various environmental challenges. This is the phase when speakers adhere to Keller’s (1985) maxims in their individual communicative actions. This is nicely exemplified by “semantic polygenesis,” the phenomenon when similar “transient meanings […] spring into existence at [… various …] moment[s] in the history of a word” (Geeraerts 1997: 64). Second, in proportion to the stability of the circumstances that have triggered the linguistic innovations and to the need for communicating about them (and also representing them mentally),


170

the adapted language use will be repeated over and over and may get fixed, in which case long-term (diachronic) adaptation, i.e., change, will occur. This means that the process of speakers adapting their language repeatedly to the same circumstances in the same way will lead to the conventionalization of the new ways of reference and representation. The expressions that are the most likely to get fixed should be the ones that “evoke and reinforce adaptive imagery [… which …] guides or promotes adaptive behaviors” (Palmer 1996: 52).

The conventionalization of novel expressions is a sociocultural process that is based on selection from a pool of linguistic variation (i.e., variability) (cf. Fritz 1998: 73, Keller 1985: 234, and McMahon 1994: 225). According to Croft’s Theory of Utterance Selection, variation comes about through altered replication of linguistic forms as “a result of speakers adjusting the mapping from language structure to external function …, that is, meaning in context” (Croft 2000: 8). When speakers select such non-conventional variants, they gradually establish a convention through the use of these variants in appropriate contexts (Croft 2000: 7, 30). In the case of semantic change then, the ultimate source of variation is the speakers’ novel usage of their language governed by the cognitive factors discussed above. This communicative behavior is triggered by various “phenomena of culture [… which …] elicit various responses to nomination, for example, metaphor, metonymy, or other figures of speech, and, as a result, synchronic variation increases. This variation is the basis of semantic change […]” (Anttila 1989: 153). Though language does not change in a predetermined direction, on the above grounds it is undeniable that language is inherently a goal-directed system (Anttila 1989: 194). This non-predetermined but still goal-directed character of language change is described by Keller in the following way (cf. also Croft 2000: 31):

[...] whereas, in nature, the variations evolve according to chance, with regard to communicating we create variation already in anticipation of the selection to be expected. (Keller 1985: 235)

Selection is thus not simply the choice from a given variability but


more like creating a solution. Heine et al. (1991: 150) have described grammaticalization “as the result of a process which has problem-solving as its main goal.” Problem-solving appears to be involved in other adaptive processes of language as well. This is especially obvious in semantic change in connection with communicative and cognitive challenges. When particular variants created in response to such challenges come under a lasting selective pressure in the form of communicative needs of wide-ranging sociocultural validity, change will occur in the language system. Contrary to biological evolutionary changes, linguistic changes are often teleological processes, but as we have seen, their explanation can “aspire to statistical generalizations and probabilistic predictions at most” (Geeraerts 1997: 151).

Though Geeraerts’ ‘pessimistic’ remark is definitely valid, certain universal tendencies can nonetheless be detected in semantic change. These come in the form of universal mechanisms of change due to universal laws in linguistic and cognitive processes on the one hand, and similar or even universal conceptualizations of the world on the other. These phenomena will be the subject of Chapter 7.

Chapter 7

Universal tendencies and linguistic relativity

in semantic change 7.1 How regular is semantic change? When discussing general issues of semantic change in Section 6.2, I also looked briefly at the question of regularity, since one important area of theoretical study within the field of historical linguistics is to provide a generalization of the changes that occur in the history of languages at the various levels of linguistic analysis. Here I will go into more detail concerning the generalizability of semantic change because the description of the general mechanisms (or laws, or principles) that operate in language change constitutes the first step in the search for universals of change. I have already mentioned that historical changes at the levels of phonology, morphology and syntax have been found to exhibit regular and systematic effects on which generalizations can be based and from which laws can be established. In many cases these even constitute major events in the history of a language, affecting the whole language system. As I will show below, such effects cannot occur in semantic change due to structural constraints. Semantic change has been considered to be basically sporadic, devoid of regularities in the form of systematic changes (Anttila 1989: 147; Hock and Joseph 1996: 244). Thus, the investigation of the generalizability in this area of language change has always been a problematic issue (McMahon 1994: 175). In spite of this, certain types of regularities have been found also in the case of semantic change (e.g. Traugott 1985), but the question to what extent they are comparable to the regularity and systematicity of the changes found at other levels of linguistic analysis has remained rather controversial.

Most of the generalizing work on semantic change has been

Universal tendencies and linguistic relativity 173

concerned with classifications of the changes according to various mechanisms, results, attitudes, causes, etc. (e.g. Algeo 1990; Blank 1997; Campbell 1998: 256-266; Hock 1991: 284-305; McMahon 1994: 178-184). The classification of the changes appears to be crucial with respect to finding universals, since this should provide the basis for the search. Unfortunately, some confusion seems to surround these classifications because they often seem to suggest that certain changes can be classified by causes while others by mechanisms and still others by range, results or attitude, etc. However, these aspects do not exclude each other; rather, a single change can and should be described from several of these aspects in order to give a complete and precise characterization of the change in question. Also, broadening and narrowing of meaning are classified as mechanisms by some authors, as results by others, while still others classify them as changes according to range. Most of the time it is not made clear that there is a strong systematic overlap between these various aspects of classification. It should be obvious that any change is set into motion by some kind of external or internal (or even psychological) cause and the change is then accomplished by some kind of mechanism like metaphor or metonymy. Broadening and narrowing are often classified as results of the change, and this is true as far as the comparison of the old and new meanings is concerned. However, it should be realized that the broadening and narrowing of a certain meaning are at the same time ways of changing that given meaning. Thus, just like metaphor and metonymy, broadening and narrowing should also be considered mechanisms of meaning change which have been triggered by some cause. On the other hand, metaphor and metonymy are also results in a special sense, since the ensuing new item in the lexicon will be a metaphor or metonymy, at least until its origin is completely obscured. Finally, certain changes, triggered by some cause and accomplished by some kind of mechanism, can also be characterized by the speakers’ attitude inherent in the once novel usage, as in the change from Old English cniht ‘attendant, servant’ to Eng. knight.

From the above aspects mechanisms appear to be generally applicable in the sense that every change can be identified as one in a limited set of well-defined mechanisms. While this fact provides considerable proof of at least one type of regularity in semantic change, it also means that the general mechanisms that have been


demonstrated for semantic change are so obvious that their universality is almost self-evident. However, there are at least two other levels of specificity at which generalizations about semantic change can be made and are useful to make (cf. Geeraerts 1994: 3800). These relate to the content rather than the form of the changes. In the following I will examine the extent to which semantic change is generalizable in order to explore the feasibility of establishing universals which may be linked to the form of change. Then I will turn to the question of potential universals pertaining to the content of semantic changes.

One of the reasons why semantic change has been found mostly irregular as compared to changes at other levels of linguistic analysis is that there is an essential structural difference between these levels and semantics. This difference is manifest in the fact that in the case of semantics the change operates on an open-ended set of linguistic elements, namely lexical items, while changes at the levels of phonology, morphology and syntax concern closed system items (i.e., restricted sets of elements) (cf. McMahon 1994: 185). For this reason, a terminological difference between generalizability and regularity should be made. Semantic change is generalizable rather than regular because various established general aspects – the mechanisms mentioned above – can be applied in the characterization of any single change. However, semantic change is not regular in the sense that the change of the meaning of one lexeme will have specific definable effects on the whole semantic system of a language, i.e., it lacks predictability, whereas in the case of sound change most changes will affect the complete phonological system. Such effects will alter the constellation of the phonological inventory, especially in the case of sound shifts which totally recast the distribution of phonemes in the system. The reason for this is the existence of a phonological space, which physically and physiologically limits the possibilities for the changes. On the other hand, the semantic structure of a language does not have the same system characteristics as does the phonological structure. In the case of lexemes and their meanings the status of individual elements of the system does not formally and systematically determine the status of other elements, i.e., they do not condition each others properties and position in the system, as is the case with phonemes. As a matter of fact, in the rare cases when semantic space exhibits (relative)


system characteristics, i.e., where parts of the semantic structure appear to be closed or at least semi-closed, like in the cases of homonymy, synonymy and lexical fields, the changes will be conditioned accordingly. These special cases of systematic changes of meaning are the elimination of homonymic clash (Hock 1991: 297-298), the differentiation of synonyms (Berndt 1989: 98-102), and chain shifts within lexical fields (Hock and Joseph 1996: 245ff.; cf. also Anttila 1989: 146-147; McMahon 1994: 186). However, they represent only a relatively small minority of meaning changes in languages. The reason for this is that apart from these cases of systematic changes, semantic change provides solutions for problems of efficiency in communication and mental representation, while phonological, morphological and syntactic changes solve structural problems in the system.

Another reason for semantic change to be different from other types of linguistic change pertains exactly to this last point, i.e., the function of semantic change. Since, as we have seen, meaning can be characterized only partly through aspects of linguistic structure, this reason is even more important to be taken into consideration in a search for universals. Research in cognitive semantics has shown that semantic knowledge is by far not autonomous. Meanings are based on encyclopedic knowledge with specific constraints (Langacker 1987: 153), and represent socially shared and culturally valid conceptualizations. In other words, semantic structure is conventionalized conceptual structure (Langacker 1987: 99). Armed with this theory, cognitively oriented historical semantics has made considerable progress in the theoretical account of meaning change (e.g. Geeraerts 1997), as opposed to traditional logic-based semantics, which has failed to give any explanation of such change. However, as much as a semantic theory founded on the open-ended nature of meaning can give a solid account for why and how semantic change happens, the fact still remains that most semantic changes require individual explanations based on our knowledge of the sociocultural history of the speakers of a language (cf. Anttila 1989: 137, and Campbell 1998: 267). And this is no wonder in view of the fact that meanings represent parts of a speech community’s conventionalized mental model of their natural and sociocultural environment.

Despite the difficulties, semantic change is not completely


ungeneralizable. König and Siemund (1999: 237) claim that recent studies have seriously questioned the irregularity of semantic change because “all semantic changes are instances of a very limited set of possible processes, such as metaphor, metonymy, ellipsis, narrowing, broadening, etc.” However, while ellipsis is a linguistic device promoting economy of expression, metaphor, metonymy, meaning restriction and extension (in fact category restriction and extension) are the basic cognitive mechanisms that yield novel conceptualizations of the world, and on which therefore semantic extension is based. In fact, these are the four well-established mechanisms of semantic change that almost all investigated languages make use of (cf. Traugott 1985). When speakers perform these cognitive operations on entrenched meanings for the sake of enhancing communicative efficiency, the linguistic manifestations of these operations may get conventionalized and new concepts of cultural relevance may get established. When this happens, semantic change has taken place.

Another area of research into the generalizability of semantic change has been concerned with the direction of the changes. It has been claimed that this direction follows general principles. Thus, Traugott (1990) points out three tendencies of change in which later meanings increasingly reflect the way speakers subjectively view the world. Wilkins (1996), discussing the semantic domain which he calls “parts of a person,” argues that synecdochic change is unidirectional because normally “a term referring to a visible part … [will] … come to refer to the visible whole of which it is an intermediate, and a spatially and/or functionally integral part” (Wilkins (1996: 275). A more general tendency, which embraces the above two, has also been established: meaning changes usually follow a concrete to abstract development (e.g. Hock 1991: 290, and Sweetser 1990: 18). However, Campbell (1998: 273) draws attention to the fact that the unidirectionality principle does not always hold because semantic restrictions “often involve change toward more concreteness,” as in the case of Eng. fowl ‘domestic bird’ < Old English fugol ‘bird,’ or Eng. deer < Old English deor ‘animal.’

When studying the generalizability of semantic change, one of the most controversial questions is whether, in addition to describing general mechanisms and general directions of change, generalizations concerning the content of meaning changes can also


be made. It appears to be self-evident that this aspect of semantic change is the most culture-dependent, and the influence of the specific sociocultural environment of the speech community is an important factor in the change. However, in spite of the culture-dependent character of the content side of semantic change, certain general tendencies have been found here as well. These general tendencies reveal themselves in similar conceptualizations across languages, and are thus the prime candidates for universals of semantic change. Especially onomasiological change may reflect certain universal directions of human thought. When looking at whole lexical fields, we may find that the various mechanisms of change may lead to similar conceptual avenues in referring to particular phenomena (cf. Anttila 1989: 147). Campbell (1998: 270-272) provides a good overview of the general tendencies that have been observed in certain kinds of changes, e.g. MAN > HUSBAND, WOMAN > WIFE, or BODY > PERSON. Haser (2000) has looked at a huge amount of data of semantic change in various lexical fields in a large number of genetically unrelated languages and compiled a long list of the similar trends of development between source and target concepts, for instance HEAR > OBEY, SEE > BEWARE, GRASP > UNDERSTAND, etc. 7.2 The cognitive foundations of universal tendencies in semantic change The task of research on universals is not only to describe what natural tendencies of change there are, but also to account for their occurrence. In my view, the causes behind the generalities found in the linguistic manifestations of semantic change are of a cognitive nature. In other words, they do not derive from the structural properties of linguistic systems, but from the way the human mind operates in perceiving and understanding the world. Crosslinguistic universal tendencies in semantic change originate therefore in the universal cognitive processes of the human mind in an effort to solve the problem of efficient communication. As described in Chapter 6, when facing communicative challenges, speakers adopt a particular conceptualization of the phenomena they wish to communicate about. The communication of new ideas, views, attitudes, etc. about


the world raises a cognitive problem, since the speaker’s mental model must be made accessible to the hearer in order to ensure mutual intelligibility. We have seen that due to the analogical nature of the human mind this is most evidently done through the exploitation of familiar knowledge (Holyoak and Thagard 1997) and the application of the fundamental cognitive mechanisms serving the exploitation of such knowledge. These mechanisms are category extension and restriction, and especially metaphor and metonymy, by which the human mind makes sense of the world in general (Dirven 1993; Johnson 1987: xx, 100; Lakoff and Johnson 1980: 36; Lakoff 1987: 77) through imposing new structure on the familiar knowledge (Langacker 1987: 105) residing basically in the already-coded category system of the language.

Thus, the universality of the four general linguistic mechanisms of semantic change (metaphor, metonymy, semantic broadening and semantic narrowing) seems to be rather self-evident considering the fact that they originate in universal human cognitive mechanisms. I would call these ‘universals of form’ due to the fact that these mechanisms pertain to the mode of conceptualizations of experience inherent in the innovative usage of conventional expressions. These universal mechanisms constitute the lowest level of specificity at which universal tendencies of semantic change can be discovered crosslinguistically (cf. Geerearts 1994: 3800).

In contrast to the general mechanisms of human cognition that shape our conceptualizations and inevitably lead to universal ways in the modification of meanings, the content of these conceptualizations may be influenced to a considerable degree by the cultural context. As we have seen in Section 5.3, the effect of the environment on semantic structure is largely filtered through the speakers’ cognitive systems. The question is whether this filtering process shows any tendencies that might lead to universal conceptualizations in semantic change. Thus, a much more challenging undertaking appears to be the search for universals that relate to the content of the conceptualizations on which the innovative usage of conventional expressions is based. Such universals, which could be called ‘content universals’ of semantic change, should be due to cognitive influences which reduce or even cancel the effects of the cultural context and bring about similar conceptualizations in various languages under different cultural


conditions. One such obvious influence is the mechanisms of change

themselves (cf. Anttila 1989: 148). They may lead to universal conceptualizations because familiar knowledge can be utilized only by way of the cognitive processes underlying these mechanisms. Thus, the reason why these mechanisms may universally induce certain specific conceptualizations is the generality of particular types of knowledge, which count as familiar, independent of the cultural context. The four special cognitive factors (identified in Section 6.4) governing the selection of the most suitable expressions for semantic modification will in addition influence the salience of specific features utilized in category formation. This should be the most obvious in the case of metonymy, which probably requires the least cognitive effort due to the very often explicit contiguity of the referents of the source and the target domains of the semantic extension. The cognitive underpinning of this seems to derive once again from Rosch’s (1978) two basic principles of categorization: perceived world structure and cognitive economy. The fact that “the perceived world is not an unstructured total set of equiprobable co-occurring attributes”(Rosch 1978: 29) will influence the way humans universally perceive things and conceptualize them across different cultures, and so it will also narrow down the range of possible lexicalizations. The cognitive salience of contiguity derives from the perception of this close correlational structure of the world, which appears to be left unaffected by the cultural context.

Conceptualizing certain phenomena in terms of these universally perceived correlations will also satisfy the principle of cognitive economy by providing “maximum information with the least cognitive effort” (Rosch 1978: 28). Cognitive economy will thus engender universal tendencies in the content of conceptualizations since the forming of categories will be the more economical cognitively the more salient features are utilized. However, it is partly responsible also for universals of form because the four basic cognitive mechanisms obviously best fulfill its requirements.

Conjunctivity, the third factor governing semantic change, will influence the contents of conceptualizations in the actuation of semantic change because conjunctive relationships are psychologically highly salient. These cognitive links are likely to be


so deeply entrenched universally that their utilization should have considerable priority in conceptualization and in the ensuing lexicalization process.

The remaining factor, cue-validity, will engender universal tendencies in the actuation of changes jointly with the other factors. By virtue of their cognitive status, perceived world structure and conjunctivity are the most evident cognitive grounds shared by interlocutors striving for mutual intelligibility. Thus, features deriving from perceived world structure and conjunctive relationships are likely to have the highest cue validity, i.e., cue their respective categories most efficiently. Since the principle of perceived world structure may universally influence the way humans perceive phenomena and conjunctivity can be determined on logical grounds, the salience of features induced by them, and the cue validity of these, is likely to be independent of cultural context, giving rise to universal conceptualizations.

The above processes serve only as the basis of the more specific similarities in the content of universal conceptualizations. The universally perceived world structure and conjunctivity relations are only starting points for more abstract levels of conceptualizations. They are likely to influence the topological structure of conceptual domains that may eventually serve as the source domains of metaphorical and metonymical extensions, and this structure is further preserved in mappings onto a target domain (cf. Lakoff 1990). Since ‘perceived world structure’ also influences our taxonomical view of the world, category extension and restriction, as manifest in semantic broadening and narrowing, may also yield universal conceptualizations by way of a similar chain.

In general, content universals of semantic change display a higher level of specificity than the universal mechanisms of change. However, these universals exhibit differences in specificity determined by the fact that the human conceptualizing capacity itself – on which they depend – may function at various levels of specificity (Lakoff 1987: 281). In other words, universals of content are expected to appear at various degrees of conceptual abstraction. Of very low conceptual specificity is the universal tendency of the unidirectional development from concrete to abstract in semantic extensions. This tendency is not simply a linguistic phenomenon but derives from a fundamental characteristic of human cognition.


Harnad (1990) has shown that all symbolic representations must be empirically grounded. Our direct perceptual experience will produce two kinds of mental representations of the concrete world, iconic ones, which are mental analogues of concrete objects and events, and categorical ones generated by innate and learned feature detectors. Symbolic representations acquire their grounding indirectly by being composed of these directly grounded ones. Categorical representations appear to be akin to basic level categories, which are categories of concrete objects at an intermediate level of our conceptual hierarchy in being “the most abstract categories for which an image could be reasonably representative of the class as a whole” (Rosch 1978: 34). These categories have principal psychological salience in human cognition because they function as cognitive reference points: objects are first recognized at this level and they are also the first ones to be learned by children (Rosch 1978: 35).

The notion of embodiment in cognitive semantics is also based on the claim that meaning and abstract reason in general originate in concrete and direct perceptual experience and bodily interaction with our environment. Johnson (1987) claims that it is through this behavioral activity that we recognize recurrent patterns which generate pre-conceptual mental structures called image schemata. These are gestalt structures whose internal organization makes our experience of the world meaningful by lending it “regularity, coherence, and comprehensibility” (Johnson 1987: 62). Furthermore, the abstract domains of our experience are understood via metaphorical projections from these image schematic gestalts.

As we have seen, the universal tendency of the concrete to abstract development in semantic change originates in what is probably one the most fundamental principles of human cognition, the general disposition of the mind to grasp new experience in terms of familiar experience. As a result of this analogical disposition cognition usually proceeds from concrete to abstract, i.e., we understand abstract domains in terms of concrete ones. Correspondingly, the universal tendency of semantic change to proceed in this direction is of very low specificity. However, the concrete to abstract trajectory of human thought is manifest in the functioning of image schemata only in a very general way. The various image schemata also function in much more specific ways


corresponding to the particular recurrent patterns of bodily experience, which further constrains the generality of the content of conceptualizations in concordance with the gestalt structures of the various schemata (cf. Johnson 1987: 112ff.). Since the various image schemata are expected to arise in all cultures due to the basic ways of human interaction with the environment, the specific conceptualizations emanating from them may still lead to universal semantic extensions, exhibiting a further step in conceptual specificity. Due to the particular relevance image schemata play in universal tendencies of semantic change, this issue will be treated separately in Chapter 8. 7.3 A comparative cognitive analysis of emotion term etyma In this section I will examine the lexicalization processes of basic emotions terms in different languages and look for universal tendencies of onomasiological change. I will try to assess both commonalities and differences in their conceptualizations using etymological data from three Indo-European languages (English, German and Russian) and a Uralic language (Hungarian). Emotion concepts lend themselves rather obviously to such an analysis for the following reasons. Basic emotions are universal psychological phenomena that seem to have universal physiological correlates across different human populations. Basic emotions also seem to be culturally salient and socially important enough for every speech community to conceptualize and code these conceptualizations in the language for general communicative availability.

In the following I will provide a list of basic emotion terms. I collected the most common expressions for the emotion pairs HAPPINESS vs. SADNESS and LOVE vs. HATE, as well as for ANGER and ANXIETY in the languages mentioned. On the basis of this list I will compare how these emotions were conceptualized diachronically. The reason why some familiar terms are not listed is that the dictionaries used list them as being of unknown (or unclear) origin, and thus these expressions could not be of interest here. This was most often the case with the Hungarian expressions and because of this the Hungarian material is the shortest. The list is not exhaustive in another respect either. Semantically transparent terms


were considered only exceptionally when they were (among) the most commonly used ones, since they are best treated in synchronic studies. Uncertainties in origin or in reconstructed meaning have been marked with a question mark in brackets. In most of the cases the etymologies speak for themselves, i.e., it is not too difficult to guess the motivation for the assumed conceptualization. I suggest that these semantic developments can be conceived of as metaphorical or metonymic projections with the etyma as source domains and the particular emotions as target domains. Thus, after every item a short explanation of the correspondences within these domains follows. Kövecses (1998) has found in his synchronic analysis that most of the emotions seem to be understood metonymically in terms of their causes or effects (behavioral responses). My historical analyses have yielded similar results.

A few words should be said about the principle of grouping the words under the various headings. Buck (1949) provides an extremely helpful and valuable list of Indo-European synonyms, among them emotion terms. The reason why I have not always exactly followed his groupings is that I found it rather difficult to establish exact semantic correspondences between emotion expressions across the various languages (see also Wiezbicka 1995). Because of this I have chosen to group the expressions under more general headings according to the fields of the basic emotions. For this same reason I have not given English equivalents as the meanings of the expressions but only marked the non-English words for noun (n), verb (v) or adjective (a). (1) HAPPINESS (i) English happy < Proto-Indo-European (PIE) *kob- ‘to suit, fit, succeed’

(Watkins 1985): Cause for emotion: SITUATION WHEN THINGS FIT / WHEN ONE SUCCEEDS FOR HAPPINESS (happiness is conceptualized as a situation when things fit / when one succeeds). This interpretation can be refined knowing that happy is a derivative of hap, which already in the Middle English period had acquired the meaning ‘good fortune’ (Onions 1966).


Thus, we have a case of metonymic chain: SITUATION WHEN THINGS FIT FOR GOOD FORTUNE > GOOD FORTUNE FOR HAPPINESS.

glad < Old English glæd ‘cheerful; shining’ < PIE *ghel- ‘to shine’ (Watkins 1985): Behavioral (psychological) response and/or effect for emotion: SHINING FOR HAPPINESS (happiness is conceptualized as shining).

cheer < PIE *ker- ‘horn, head’ (Watkins 1985): (Object used in the) behavioral response for emotion: OBJECT USED TO EXPRESS HAPPINESS FOR HAPPINESS. The motivation in the semantic development from this PIE root can apparently be explained by the fact that horns were used as musical instruments to express happiness.

merry < PIE *mregh-u- ‘short’ (Watkins 1985): Effect for emotion: SHORT TIME / TIME PASSING QUICKLY FOR HAPPINESS. Note here the semantic similarity with the German noun Kurzweile ‘state of having fun, passing the time happily,’ literally ‘short while,’ the antonym of which is Langweile ‘boredom,’ literally ‘long while’ (cf. Buck 1949). A similar semantic shift can also be found in Hungarian mulat ‘to enjoy oneself,’ meaning originally ‘to make (time) pass.’

joy < PIE *gau- ‘to rejoice; to have religious fear or awe,’ via borrowing from Old French (Watkins 1985, Onions 1966): Cause for emotion: HAVING RELIGIOUS FEAR OR AWE FOR HAPPINESS (happiness is conceptualized as having religious fear or awe), and/or the original meaning has been retained.

pleased < PIE *plak- ‘to be flat’ > suffixed form *plak-e- ‘to be calm (as of the flat see),’ via borrowing from Old French (Watkins 1985; Onions 1966): Behavioral (psychological) response and/or effect for emotion: BEING CALM FOR HAPPINESS (happiness is conceptualized as the state of being calm).

(ii) German Glück (n) < Middle High German gelücke ‘(good) fate, chance’

(Drosdowski 1963; Kluge 1975): Cause for emotion: (GOOD) FATE FOR HAPPINESS (happiness is conceptualized as having good fate). This word is of unknown origin and attested only since the 12th century, though cognate with English luck,


Swedish lykka, Dutch geluk, etc. froh (a), Freude (n) < PIE *preu- ‘to hop’ (Drosdowski 1963; Kluge

1975): Behavioral response for emotion: HOPPING FOR HAPPINESS (happiness is conceptualized as when someone hops around).

heiter (a) < PIE *kai- ‘to shine, light’ (Drosdowski 1963; Kluge 1975): Behavioral (psychological) response and/or effect for emotion: SHINING FOR HAPPINESS (happiness is conceptualized as shining).

(iii) Russian rad (a) < PIE *ar[ ]-/re[ ]- ‘to fit, suit’ (?) (Vasmer 1950-1958):

Cause for emotion: SITUATION WHEN THINGS FIT FOR HAPPINESS (happiness is conceptualized as a situation when things fit). The origin of this word is uncertain. Vasmer suggests that rad may be cognate with Russian radet’ ‘to look after’ and thus derivable from the above PIE root. On the other hand, Mann (1984-87) derives rad from PIE *rad- ‘glad’ (?). In this case the original meaning has been retained. However, the validity of this root is uncertain, since Mann reconstructs it purely on the basis of this form and Old English rot ‘cheerful, noble, excellent’ (which has no reflex in Modern English).

scast’e (n) < Proto-Slavic *s cest je: *s < PIE *su- ‘good’ + *cest ‘part’ (Vasmer 1950-1958): Cause for emotion: GOOD PART/FATE FOR HAPPINESS (happiness is conceptualized as having good part/fate). This explanation for this semantic development is made plausible by the fact that Russian cast’ also has the somewhat archaic meaning ‘fate’ (just like English lot means both ‘portion, part’ and ‘fate’).

(iv) Hungarian öröm (n), örül (v) < Proto-Finno-Ugric (PFU) *irw3 ‘to be happy’

(Rédei 1986-1991): The original meaning has been retained. (2) SADNESS (i) English sad < PIE *sa- ‘to satisfy’ (Watkins 1985): Behavioral


(psychological) response and/or effect for emotion: SATISFACTION FOR SADNESS. The basis for this surprising semantic development and metonymical extension could be the general observation that satisfaction induces quietness and one is quiet when sad. Thus, the basic metonymy could be QUIETNESS FOR SADNESS (sadness is conceptualized as quietness). Though this interpretation may be questionable, Buck (1949) also suggests a semantic development from Old English sæt ‘sated’ through Middle English sad ‘steadfast, firm, serious, grave’ to Modern English sad. However, it may also be speculated that we have to do with a metaphor here: (the quietness accompanying) SADNESS IS (similar to) (the quietness accompanying) SATISFACTION

sorrow < PIE *swergh- ‘to worry, to be sick’ (Watkins 1985): Cause for emotion: WORRYING FOR SADNESS (sadness is conceptualized as worrying).

grave, grief < PIE *gwere- ‘heavy’ (Watkins 1985): Behavioral (psychological) response and/or effect for emotion: HEAVY(NESS) FOR SADNESS (sadness is conceptualized as something heavy).

(ii) German traurig (a) < PIE *dhreu- ‘to fall, flow, drip, droop’ (Drosdowski

1963; Kluge 1975): Behavioral (psychological) response and/or effect for emotion: BEING (OR HAVING ONE’S HEAD) DROOPED FOR SADNESS (sadness is conceptualized as when one / one’s head is drooped).

betrübt (a) < PIE *dher- ‘to make muddy’ (Drosdowski 1963; Kluge 1975): Behavioral (physiological) response and/or effect for emotion: HAVING TURBID BODY HUMORS FOR SADNESS (sadness is conceptualized as having turbid body humors). This word is a derivative of trüb ‘turbid’ and the shift in meaning is most probably based on the folk theory of the four body humors.

(iii) Russian gor’e (n) < PIE *gwher- ‘to heat, warm,’ (Vasmer 1950-1958):

Behavioral (psychological) response and/or effect for emotion: BURNING FOR SADNESS (sadness is conceptualized as burning). This word is cognate with Russian goret’ ‘to burn.’


grust’ (n) < PIE *gwere- ‘heavy’ (Vasmer 1950-1958): See grave, grief above.

(iv) Hungarian szomorú (a) < PFU *som3 ‘hunger, thirst’ (Rédei 1986-1991): Cause

for emotion: BEING HUNGRY/THIRSTY/UNSATISFIED FOR SADNESS (sadness is conceptualized as when one is hungry/thirsty/unsatisfied).

(3) LOVE (i) English love < PIE *leubh- ‘to care, desire, love’ (Watkins 1985): Behavioral

response for emotion: CARING / DESIRING FOR LOVE (love is conceptualized as caring/desiring). Even more compelling is the idea that CARE and DESIRE should be viewed as inherent concepts in the cognitive model of LOVE and that in this instance this emotion was conceptualized through them (cf. Kövecses 1991). However, it may also be the case that the original meaning has been retained.

like < Old English lician ‘to please’ < Common Germanic *lik- ‘body, form; like, same’ (Watkins 1985): Cause for emotion: THE SITUATION WHEN THINGS ARE ALIKE FOR LOVE (love is conceptualized as the situation when things are alike). One is pleased by experiencing appropriate/alike forms.

(ii) German Liebe (n) < PIE *leubh- ‘to care, desire, love’ (Drosdowski 1963;

Kluge 1975): See love above. mögen (v) < PIE *magh- ‘to be able, to have power’ (Drosdowski

1963; Kluge 1975): Cause for emotion: (NO) ABILITY/POWER FOR (NO) LOVE ((no) love is conceptualized as (no) ability/power). This semantic development constitutes a special case. Today’s meaning of mögen ‘to like’ developed in Middle High German in negative environments and the meaning ‘not be able’ first induced the connotation ‘not like’ and then the complete semantic shift.


(iii) Russian ljubov’ (n) < PIE *leubh- ‘to care, desire, love’ (Vasmer 1950-

1958): See love above. (iv) Hungarian szeret (v) < PFU *ser3 ‘row, order’ (?) (Rédei 1986-1991): Cause

for emotion: THE SITUATION WHEN THERE IS ORDER FOR LOVE (love is conceptualized as the situation when there is order).

(4) HATE (i) English hate < Old English hete ‘hate, envy’ < PIE *kad- ‘sorrow, hatred’

(Watkins 1985): The original meaning has been retained. (ii) German Hass (n) < PIE *kad- ‘sorrow, hatred’ (Drosdowski 1963; Kluge

1975): See hate above. (iii) Russian nenavidet’ (v) < Proto-Slavic *navedeti ‘to look at with pleasure’

with prefixed negative particle (Vasmer 1950-1958): Cause for emotion: UNPLEASANT VIEW FOR HATE. Also behavioral response for emotion: NOT LOOKING WITH PLEASURE FOR HATE. Hate is conceptualized as looking at something unpleasant.

(iv) Hungarian No data from Hungarian could be considered here because the two

basic terms utál and gyűlöl, both meaning ‘to hate,’ are of unknown origins.

(5) ANGER (i) English angry < PIE *angh- ‘painful, tight’ (Watkins 1985): Behavioral


(physiological) response and/or effect for emotion: TIGHTNESS / FEELING OF CONTRACTION IN THE BODY FOR ANGER. Also cause for emotion: FEELING PAIN OR TIGHTNESS FOR ANGER. Anger is conceptualized as some tightness/contrac-tion/pain in the body.

(ii) German Zorn (n) < PIE *der- ‘to split, peel, flay’ (?) (Drosdowski 1963;

Kluge 1975): Behavioral (psychological) response and/or effect for emotion: SPLITTING (UP) FOR ANGER (anger is conceptualized as splitting up).

böse (a) < Old High German bosi ‘bad, worthless’ < PIE *b(h)eu- ‘to swell’ (Drosdowski 1963; Kluge 1975): Behavioral response for emotion: ILL WILL FOR ANGER (anger is conceptualized as ill will). This word also carries the meanings ‘bad’ and ‘ill willed.’ The development from PIE may be based on the metonymy SWELLING (ON THE BODY) FOR BAD.

(iii) Russian serdit’sa (v) < PIE *kerd- ‘heart’ (Vasmer 1950-1958): Seat of

emotion for emotion: HEART FOR ANGER (WHERE IT IS SEATED) (when one is angry one feels one’s heart contracting/pounding etc.). This word is a derivative of serdtse ‘heart.’

jadovityj (a) < PIE *ed- ‘to eat, bite’ (Vasmer 1950-1958): Body humor for emotion: POISON (IN THE BODY) FOR ANGER (anger is conceptualized as poison in the body). This word is a derivative of jad ‘poison,’ which figuratively also means ‘anger, fury.’

(iv) Hungarian mérges (a) < PFU *mirkk3 ‘poison’ (?) (Rédei 1986-1991): Body

humor for emotion: POISON (IN THE BODY) FOR ANGER (anger is conceptualized as poison in the body). This word is a derivative of méreg ‘poison,’ which figuratively also means ‘anger, fury.’

harag (n) < PFU *kur3 ‘anger,’ cognate with PFU *kura- ‘to peel, flay’ (Rédei 1986-1991): Behavioral (psychological) response: SPLITTING (UP) FOR ANGER (anger is conceptualized as


splitting up). (6) ANXIETY (i) English anxious < PIE *angh- ‘painful, tight’ (via Latin) (Watkins 1985):

Behavioral (physiological) response and/or effect for emotion: TIGHTNESS / FEELING OF CONTRACTION IN THE BODY FOR ANXIETY (anxiety is conceptualized as tightness / feeling of contraction in the body).

fear < PIE *per- ‘to try, risk’ (Watkins 1985): Cause for emotion: RISKING FOR FEAR (fear is when you risk something).

(ii) German Angst (n) < PIE *angh- ‘painful, tight’ (Drosdowski 1963; Kluge

1975): See anxious above. Schreck (n) < PIE (s)ker- ‘to leap, jump’ (Drosdowski 1963; Kluge

1975): Behavioral response for emotion: JUMPING FOR ANXIETY/FRIGHT (anxiety/fright is conceptualized as jumping for fear).

(iii) Russian bojat’sa (v) < PIE *bhoi- ‘to be afraid, tremble’ (Vasmer 1950-

1958): Behavioral (physiological) response and/or effect for emotion: TREMBLING FOR ANXIETY (anxiety is conceptualized as the situation when one trembles).

strach (n) < PIE *storg- ‘stiff’ (Vasmer 1950-1958): Behavioral (physiological) response and/or effect for emotion: BEING/BECOMING STIFF FOR ANXIETY (anxiety is conceptualized as the situation when one is/becomes stiff/motionless).

(iv) Hungarian fél (v) < Proto-Uralic *pele- ‘to fear’ (Rédei 1986-1991): The

original meaning has been retained.


First I will look at corresponding conceptualizations in the above list. The list shows how the same emotion was conceptualized in the different languages as attested in the lexicalization processes, and some noteworthy diachronic semantic correspondences can be observed. A certain degree of universality may be inferred from the cases in which different languages conceptualized the same emotion in a similar or the same way. There are also cases in which the same source domain served to conceptualize different emotions. In the following explanations I have added some examples from other languages as well. Corresponding conceptualizations of the same emotion Conceptualizations of HAPPINESS: In three cases the same or similar conceptualizations seem to underlie the lexicalization process of HAPPINESS in spite of the lexemes not being cognates. English glad and German heiter ‘joyous’ seem to be based on the metonymy SHINING FOR HAPPINESS. English happy, German Glück (n) and Russian scast’e (n) were conceptualized through the metonymy GOOD FATE/FORTUNE FOR HAPPINESS. Both English happy and Russian rad ‘happy’ (apart from the uncertainties in the etymology of the latter) are grounded in the conceptualization HAPPINESS IS WHEN THINGS FIT. Conceptualizations of SADNESS: We seem to have contradictory conceptualizations of the same emotion in the case of English sad and Hungarian szomorú ‘sad,’ since they imply satisfaction and lack of satisfaction respectively. However, we can hypothesize that English sad is based on a characteristic that this emotion shares with the feeling of satisfaction, namely calmness. Conceptualizations of LOVE: The diachronic semantic developments yielding English like and


Hungarian szeret ‘to like, love’ seem to find their origin in man’s attraction toward systematicity. Conceptualizations of ANGER: An astonishing correspondence is provided by German Zorn ‘anger’ and Hungarian harag ‘anger’ since they belong to two different language families. However, the similarity in conceptualization may not be so difficult to explain. The metaphor ANGER IS EXPLOSION (OF THE BODY) seems to be very common in many languages today (cf. Kövecses 1995a). It is also worth noting that the body humor metaphor functions both in the case of Russian jadovityj ‘angry’ and Hungarian mérges ‘angry.’ Both are adjective formations to corresponding nouns meaning ‘poison’ and figuratively also ‘anger, fury.’ Though not mentioned in the list, another interesting correspondence exists between Russian serdit’sa (v) ‘to be angry’ and Hittite kartimiia ‘be furious.’ Both derive from PIE *kerd- ‘heart’ and are derivatives of the corresponding nouns serdtse ‘heart’ and kard- ‘heart’ respectively. However, the semantic developments based on the metonymic conceptualization SEAT OF ANGER FOR ANGER are independent. Conceptualizations of ANXIETY: The conceptualization ANXIETY IS TREMBLING, which can be observed in the development to Russian bojat’sa ‘to be afraid,’ is very common in other languages as well. Though in a semantically transparent form, it can also be found in Hungarian reszket and retteg ‘to be afraid, tremble,’ in German zittern ‘to be afraid, tremble’ and in the English idiom tremble with fear. Corresponding conceptualizations of different emotions In the list above there are also several cases in which one source domain led to the conceptualization of different emotions. In the following I will list these source domains with their various target domains.


TIGHTNESS / FEELING OF CONTRACTION IN THE BODY as source domain: ANGER: English anger ANXIETY: English anxious (via Latin); German Angst ‘fear’

SADNESS: Old Norse angr ‘sorrow, grief’ (Watkins 1985). (Note hat this adds SADNESS IS TIGHTNESS / FEELING OF CONTRACTION IN THE BODY as behavioral (physiological) response and/or effect for emotion to the list of conceptualizations of SADNESS).

CALMNESS as source domain: HAPPINESS: English pleased SADNESS: English sad POISON as source domain:

ANGER: Hungarian mérges ‘angry, furious;’ Russian jadovityj ‘angry, furious’

SADNESS: Croatian jad ‘sorrow, grief;’ Serbian jad ‘sorrow, grief.’ (Note that this adds SADNESS IS POISON to the list of conceptualizations of SADNESS.)

HOPPING as source domain:

HAPPINESS: German froh ‘happy, glad,’ Freude ‘happiness, gladness;’ Middle High German scherzen ‘to leap with joy’

ANXIETY: German Schreck ‘fright’ Let us now turn to some cognitive implications of the above data. As we have seen, the terms in the same emotion domain even within closely related languages do mostly not belong to the same cognate set. Thus, the one-time conceptualization of a particular emotion as revealed by the etymology of its linguistic expressions is very often different from branch to branch within a language family, or even from language to language within one branch. This implies that the conceptualizations of emotions are as much subject to changes as that of any other culture-dependent phenomenon. However, Key (1988) argues that the direction of semantic change is predictable in many domains (e.g. kinship, corporeal, perception, etc.) because the


findings from different languages reflect universal semantic relationships. The material surveyed here very strongly suggests that this is also the case in the domain of emotions. If we look closely, the different etymologies show that different languages lexicalized different parts of a complex conceptualization of that emotion. These seem to combine to one cognitive model of the particular emotion, similarly to what has been found in synchrony within one language (cf. Lakoff and Kövecses 1983; Kövecses 1991). Kövecses (1995b) has also shown that metaphors constitute rather than just reflect a given cultural model. Thus, it can be hypothesized that any new conceptualization can only be made within the framework of that cultural model of the particular emotion. The conceptualizations, however, as the material shows, are not emotion-specific. The etyma functioning as source domains in the historical conceptualization processes have yielded many other metaphorical and metonymic concepts beside that of the particular emotion. This diachronic finding is in line with Kövecses’s (1998) synchronic claim that the source domains of emotion metaphors are not specific to that emotion.

The fact that we have a wide variety of non-specific source domains for emotion metaphors and metonymies probably accords with the rather high abundance of figurative emotion expressions in present day language use. In the languages surveyed these basic emotions can be expressed by several other semantically more or less transparent emotion expressions which are so commonly used that speakers do not really feel or consider them figurative any longer. Apparently emotions have always invited the human mind to metaphorize about them. Thus, in spite of the fact that basic emotions are universal human psychological states with their specific physiological correlates and that the cognitive mechanisms and factors discussed earlier in this section often lead to universal tendencies in their conceptualizations across different cultures, a considerable number of the conceptualizations are different and show culture-dependence, or even the freedom and creativity of human thought, though within general constraints. In the next section I will briefly discuss factors that work against universal tendencies.


7.4 Universals of semantic change versus linguistic relativity The earlier discussion reveals that content universals of semantic change derive mainly from the cognitive factors that govern the choice of conceptualizations when making sense of the world. However, as I will try to show below, the cognitive factors discussed – primarily cue-validity – are not only a source for universal conceptualizations but at the same time a source of culture-dependent ones as well. Cognitive processes may lead to universal conceptualizations in the case of universally salient conditions, but these may also be dominated by culture-dependent factors, due to which the conceptualizations may show relativistic tendencies.

The reason for this duality can be found in the primary task of cognition: to give a functional interpretation of the world. As a result, cue-validity – the dominating factor in the choice of the content of conceptualizations – cannot have a deterministic cognitive effect. The features universally deriving from perceived world structure and conjunctive relationships only have high cue-validity other things being equal. Under specific cultural circumstances these will most probably be overridden by culture-specific aspects. A culturally determined shift in accent concerning the salience of features might result in other features exhibiting high cue-validity, which means that cue validity might occasionally stand under strong cultural influence. This will engender culture-specific conceptualizations as bases of lexical coding in the language. Furthermore, not only culture-specific aspects but also the subjectiveness of the human psyche may override universal perspectives, as the so-called psychological causes of semantic change like exaggeration, emphasis, expressiveness, creativity, euphemism, or taboo testify.

Apart from the above cognitive influences, there is another very important source of cultural relativity in conceptualization working against universal conceptual avenues in semantic change. One of my main points in Chapter 6 was that the actuation of semantic change originates in the individual speaker’s efforts to meet changing communicative and cognitive needs (i.e., to comply with communicative and cognitive challenges) by modifying the meanings of conventional expressions. The basic aim of this adaptation of language to new circumstances in the form of novel


expressions is to make the speaker’s mental model of the world accessible to the hearer when no conventional expression is deemed suitable for that purpose. The most evident cognitive basis of this innovative language use is the exploitation of familiar knowledge, which can provide the necessary motivational support for both the production and the comprehension of occasion-bound meanings. Reliance on familiar knowledge largely enhances the efficiency and economicalness of reference and representation.

The relevant familiar knowledge that can be exploited for ensuring mutual intelligibility obviously resides in conventional expressions and in the connotations (or encyclopedic information) attached to them by the speech community. Thus, linguistically coded categories will provide the input to cognitive mechanisms of metaphorical and metonymical transfer when new experience or new perspectives are conceptualized. Consequently, semantic change does not only show how cognition influences what categories will be created in language. It also shows how the linguistically established categories influence further categorizations. As Rosch (1978: 29) says, “[o]ne influence on how attributes [in category formation] will be defined by humans is clearly the category system already existent in the culture at a given time.” Thus, while the linguistically coded categories are results of previous conceptualizations on the level of a whole culture, they also provide an ever-ready source for the operation of basic cognitive processes that guide the altered usage of conventional expressions for the sake of efficient communication and cognition. This will necessarily constrain the choice of applicable and modifiable expressions, biasing by this the way new experience is described, and thus ultimately influencing the content of new conceptualizations of reality.

As the above considerations make it clear, it is not in the origin of meanings (i.e., the etymology of words) that linguistic relativity shows itself, since the origins get obscured with time anyway, and linguistic relativity is an effect from actual language use on speakers’ cognitive processes. In other words, any influence from etymology on the cognitive processes of speakers is blocked by the fact that speakers’ semantic knowledge contains no information on how particular meanings emerged in their native language. What is interesting and telling in semantic change from the point of view of linguistic relativity is not the results of individual semantic changes,


but the actuation of such changes because they involve the speakers’ semantic knowledge and cognitive processes. The validity of this view hinges on the insight that semantic change is not just creating a label for a conceptual category but creating the category itself, although Rhodes (2000) argues that semantic change does not reflect change in conceptual structure. However, if semantic structure is conceptual structure (Langacker (1987: 99), i.e., the structure of meanings – although being conventionalized cultural creations – mirrors cognitive structures, then the different semantic structuring found in different languages must have relativity effects.

There are at least two basic cognitive facts that suggest that the semantic structure of a language will influence the way we see the world. Many cognitive psychologists argue that our expectations about the world are biased by our previous knowledge and the concepts we have (Heit 1997: 8; Neisser 1976). According to Smith and Medin (1981: 8) “concepts have a categorization function … [and] … are essentially pattern recognition devices.” Therefore, if meanings are conceptual structures, then they must also have a knowledge storing function, which means that they will influence our expectations. Language – as a cognitive device – will provide us with mental categories for ordering our experience into, or in other words, we are likely to experience the world in terms of the categories supplied to us by the meanings of our language. This is exactly in line with the cognitive function of language: any language must provide the proper material for a symbolic model of the environment in which it is to be used. The second fact, actually supporting the first one, is the perceptual phenomenon of categorical perception. This is when “[f]or certain perceptual categories, within-category differences look much smaller than between-category differences even when they are of the same size physically” (Harnad 1987: 535). Harnad (1987: 546) suggests that learned labels with their underlying categories might influence discriminability and thus engender a Whorfian effect. Medin and Barsalou (1987: 470) also propose that the effect of categorical perception could also operate at levels of cognition higher than perception, e.g. semantically defined generic-knowledge categories. In the case of linguistic categories this should mean that speakers perceive the world in terms of the categories supplied to them by their native language and will resort to these categories as starting


points of new conceptualizations when modifying conventional meanings.

Chapter 8

The role of image schemata in semantic

change 8.1 A note on the relevance of universal tendencies in semantic change for semantic reconstruction The assumed ad hoc character of semantic change has not only posed a theoretical problem in the search for generalities and universals (and thus in establishing a theory of semantic change), but also a practical one in the field of semantic reconstruction (cf. e.g. Sweetser 1990: 26). The postulation of possible cognate forms has two basic requirements. The first one is the existence of regular sound correspondences between the forms, which is by far the more compelling evidence for the reconstruction of a common etymon. However, the theoretical primacy of this requirement derives from our established knowledge of the regularity of sound change as opposed to the lack of regularity in semantic change, on which a postulation of semantic correspondences could be based. Thus, the second requirement is to find some semantic relationship between the assumed cognates and on this basis we have to be able to supply a plausible explanation for the semantic development from some earlier underlying meaning. For the sake of achieving a rigor similar to phonological reconstruction in the reconstruction of meanings, the best available method for semantic reconstruction appears to be a feature analysis of the assumed related meanings, from which we then process some kind of a lowest common denominator. The bundle of features yielded in this way will then be posited as the original meaning. However, as Sweetser (1990: 24) has clearly shown, these meanings do not seem to be realistic at all, since such a procedure yields a protovocabulary full of abstract meanings, which contradicts our knowledge of semantic change running from concrete to abstract in the


vast majority of the cases. The problem with the above method of semantic reconstruction

does not lie in its imprecision but in the view it maintains about the nature of meaning. Langacker (1987: 157) has pointed out that semantic extension – among other everyday semantic phenomena – cannot be handled by an autonomous feature-based approach but only in an encyclopedic view of meaning. In Sweetser’s (1990: 24) opinion, a cognitive theory of meaning cannot subscribe to the idea that the basic mechanisms of semantic change can be reduced to loss, addition and recombination of semantic features. In any case, no semantic reconstruction can be initiated unless we can find some semantic relationship between our tentatively cognate forms and supply a plausible explanation for the given development. An encyclopedic approach to meaning incorporating the view of the prototypical nature of conceptual categories is especially suited to explain semantic extension and change because it is based on natural cognitive capacities which underlie meanings.

In phonological change, for instance, naturalness of a reconstructed phonemic inventory and naturalness of the processes by which it is possible to derive the attested phonemic inventories from the protosystem is an important governing principle. That is, one of the tests of the validity of particular reconstructions has been mostly the existence of typological parallels across languages (see Fox 1995: 253f. on problems of the applicability of typological considerations in phonological and syntactic reconstruction). Joseph and Karnitis (1999) state that research on the change in components of grammar other than semantics has always benefited from work on naturalness constraints, whereas the study of semantic change and the search for cognates have at best “the traditional methodology of looking for parallels to get a handle on the wide range of semantic extension” (Joseph and Karnitis 1999: 152).

Wilkins (1996) also points out that the uncertainties and controversies in semantic reconstruction derive mainly from our lack of knowledge of natural semantic shifts. He quite correctly claims that by identifying natural tendencies in semantic change reconstruction could be made more precise, and distinguishes five main types of natural tendencies in the naming of body parts (Wilkins 1996: 273-274), summarized below:

The role of image schemata in semantic change 201

i. VISIBLE PERSON-PART VISIBLE WHOLE (e.g. ‘navel’ ‘belly’ ‘trunk’ ‘body’ ‘person’);

ii. PERSON-PART SPATIALLY CONTIGUOUS PERSON-PART (e.g. ‘belly’ ‘chest’; ‘skull’ ‘brain’);

iii. PARTS OF UPPER BODY / PARTS OF LOWER BODY [where the waist provides a midline] (e.g. ‘elbow’ ‘knee’; ‘uvula’ ‘clitoris’; ‘anus’ ‘mouth’);

iv. ANIMAL-PART PERSON-PART (e.g. ‘snout’ ‘nose’; ‘beak’ ‘face’)

v. ACTION INVOLVING THE USE OF A PERSON-PART PERSON-PART (e.g. ‘walk’ ‘leg’; ‘hold’ ‘hand’).

All of these appear to be properly grounded in universal cognitive procedures, which should be a governing principle in explanations of any natural tendency and thus of universals of change. Thus, tendencies i., ii., and v. are based on metonymic transfer, probably the most elemental of the cognitive mechanisms, requiring the least processing effort due to its explicitness based on the perceptual salience of the contiguity involved. In these cases cultural influence is also completely excluded and thus the cognitive/semantic link appears to be absolutely natural, making them perfect candidates for universals. The naturalness and hence universality of tendencies iii. and iv. is also quite plausible, since they are based on metaphorical transfer, one of the basic cognitive mechanisms, which involves structured mappings from a source domain to a target domain. In the case of tendency iv. these domains are universal domains of human experience not only with regard to perception, but also with regard to culture due to the role animals play in human culture in general.

In the following I would like to show that universal/natural tendencies of semantic change may not only originate in metaphorical/metonymical transfers from concrete domains of experience. Although universals most obviously derive from universal cognitive mechanisms of the mind operating on domains of universal human experience, in many cases even more fundamental experiential structures seem to serve as the basis for the operation of these mechanisms. Thus, I think that the actual conceptualizations found in many of the semantic changes, especially in the case of cognate groups, are more likely to be triggered by underlying image schemata – from which the semantic extensions develop in a parallel


fashion – than by particular concrete domains. 8.2 The relevance of image schemata in the search for universals In this and the previous chapter my argumentation centers around the conviction that if semantic extension in general is rooted in cognitive processes of the human mind, then universals of change should derive from what is universal about human conceptualization. The emergence of image schemata in our everyday bodily interaction with our environment and their employment in making sense of new and more abstract experience is such a pervasive and universal aspect of human mental life (Gibbs and Colston 1995) that when looking for sources of universal tendencies of change concerning conceptualization content, they should be considered prime candidates. Haser (2000: 187) has also called attention to the image schematic character of semantic extensions and implied that the observed universal tendencies in lexical developments might derive from the exploitation of similar source domains for similar target domains, though she remarks that “[a] more elaborate description of the underlying pattern … [than offered by image schemata] … seems necessary” for a more accurate explication of crosslinguistically similar extensions (Haser 2000: 187).

Given the pervasiveness of the image schematic structuring of human experience in general, this appears to be a valid claim. After all, the function of image schemata is to structure more concrete domains of experience. In other words, the more specific the content of universal human cognitive processes, for instance the particular conceptualizations on the basis of image schematic projections, the more specific are the semantic changes found crosslinguistically, and the “more elaborate description of the underlying pattern” is needed. The levels of specificity of these universal semantic extensions will be determined by what level of specificity universal conceptualizations can reach. Considering the rather general structuring function of image schemata and the pervasiveness of experiential domains that are image schematic in themselves (Clausner and Croft 1999: 21), we could say that image schematic structuring alone seems to account for a relatively low level of specificity in universals of semantic change. However, their


projections into more concrete domains will yield higher levels of specificity, though still universal, since particular concrete domains of human experience invite particular image schematic structuring universally.

I also completely agree with Haser’s claim that not all semantic extensions originate in “a more or less fixed inventory of image schemata” but rather “[a] comparative analysis of similar metaphorization processes may help discover relevant ‘structures’ triggering the extensions in the first place” (Haser 2000: 186). Among others, this is one of the insights which emerges from the etymological analysis below. But in all fairness to Johnson, it must be mentioned that he himself remarks that “[i]f one understands ‘schema’ more loosely than … [he does] …, it might be possible to extend … [the] … list … [of schemata] … at length” (Johnson 1987: 126).

An account of similar semantic developments across different languages on the grounds of particular underlying image schemata may not only be helpful in the explication of universal tendencies of semantic change but may also have considerable practical significance. As I will show below, the discovery of such structures may be helpful in verifying particular reconstructions of meaning. In the following I would like to examine some similar semantic developments across different languages. I suggest that they can be accounted for on the grounds of one particular image schema. Furthermore, I would also like to demonstrate by this that the explication of these similar semantic developments in terms of an underlying image schema may be relevant for developing a more accurate view of the historical developments in question, notably, it may yield new insights in the search for possible cognates (cf. Joseph and Karnitis 1999: 154). For my analysis I have chosen one of Wilkins’ examples to his third natural tendency, the semantic development between the senses ‘elbow’ and ‘knee.’ Since the term for ‘chin/jaw’ is cognate with the one for ‘knee’ in many Indo-European languages and strengthens my case for the particular image schema underlying the various semantic extensions found in this cognate group, I included it in the analysis as well.

Table 1 below shows the distribution of two groups of words with the senses ‘knee’ and ‘chin/jaw’ in Indo-European languages.


Table 1: The diachronic interconnection of ‘knee’ and ‘chin/jaw’

‘KNEE’ ‘CHIN/JAW’

Old English

cneo cin(n) ‘chin’, cinban ‘jawbone’

Old High German

kniu, kneo kinni ‘chin’, chinne ‘jaws’

Greek gónu genus ‘jaw, cheek’

Latin genu gena ‘cheek’

Welsh glin (< *glu-nes <*gnu-nes)

gen ‘jaw, chin’ (< ? Latin)

Hittite genu- <gi-[e]nu->

Tokharian A, kanwem, B, kenine A, çanwem ‘jaws’ (dual)

Armenian cunr cnawt ‘jaw’

Sanskrit janu- hanu- ‘jaw’ (h < *j)

Avestan žnu- zanauua ‘both jaws’

Lithuanian žándas ‘jaw’

Proto-Indo-European

*gónu-/*gnu- *genu-

The two groups of words have an ancient etymological relationship and must be derivatives from the same source, i.e., ablauting forms of one and the same ancient Indo-European root with different root extensions. From a conceptual point of view, there does not seem to be a primacy of any of the senses and the semantic extensions are probably based on the same underlying perceptual pattern. In this cognate group the resemblance in shape, presumably the notion of ‘angle,’ is the prominent characteristic feature associated with both of the senses ‘knee’ and ‘chin/jaw,’ which provides the semantic connection between the names of these body parts (cf. Buck 1949: 221). This supposition is reinforced by the fact that in some languages we can find cognate words that still have the meaning ‘angle,’ e.g. Greek gónia ‘angle.’ Pokorny gives the reconstructed meaning for PIE *genu-/*gonu-/*gnu- as ‘knee, corner, angle’ and ‘jaw(bone)’ (Pokorny 1959: 380). Furthermore, the meanings ‘jaw’,


‘chin’ and ‘cheek’ can interchange. Under these circumstances it is possible to find the common concept of similarity in shape, which was formulated by Buck as ‘something projecting’ or a ‘hook’ (Buck 1949: 224).

Interestingly, parallel semantic developments can be observed in the Uralic language family. In Uralic languages two etyma have been reconstructed for the notion of ‘knee’: Proto-Uralic *polwe (Rédei 1986-1991: 393) and Proto-Uralic (or Proto-Finno-Ugric) *sänc3 (the reconstruction of *sänc3 is perhaps not really as deep as the PU stage, since the inclusion of the Samoyedic data is questionable) (Rédei 1986-1991: 471). In the case of PU *polwe ‘knee’ the individual languages have not only the semantic component of the body part but also the meaning ‘curve, bend’, e.g. Finnish polvi ‘knee; extremity; curve, bend’. The question, of course, is which meaning component is primary and diachronically earlier. If we suppose that the notion of ‘bend’ is a frequently occurring underlying concept in the names of the body parts - as in the case of Indo-European ‘knee,’ then the polysemy found in Finnish seems to maintain an archaic state.

The other etymon for ‘knee’,*sänc3 could perhaps be derivationally connected to the etymon *sine ‘bend, curve’ (Rédei 1986-1991: 480) if they can be proven to be morphologically complex. Besides these two etyma, Rédei also lists PFU *pic3 (püc3) ‘bend(ing) of a body part (e.g. of the knee, elbow)’ from which we have reflexes both for ‘elbow’ and for ‘knee’: Vogul pisi/päs ‘elbow’ and Votyak pîd’es ‘knee’ (Rédei 1986-1991: 376). This is a most illustrative example of the linguistic representation of the underlying conceptual relationship.

The perceptual pattern of ‘curve/bend’ has served as a cognitive basis for the semantic extension also in the case of terms denoting other body parts with this shape. Thus, we can adduce evidence from several languages for the meaning ‘elbow’ deriving from the notion of ‘bend,’ e.g.: English elbow < Old English elnboga < Proto-Germanic *alino-bugon ‘bend of the forearm.’ The Uralic Etymological Dictionary also suggests the comparison of Proto-Finno-Ugric *kinä (künä) ‘elbow’ with Proto-Indo-European *genu- ‘knee.’ An interesting example is the above mentioned PFU *pic3 (püc3) ‘bend(ing) of a body part (e.g. that of the knee or the elbow),’ from which both Vogul pisi/päs ‘elbow’ and Votyak pîd’es ‘knee’


derive (Rédei 1986-1991: 376). These historical semantic data suggest the postulation of a

BEND/CURVE image schema, from which metaphorical and metonymical projections give rise to various senses. Of course, a proper verification of the existence of a BEND/CURVE schema needs a more detailed analysis and the collection of a large amount of evidence also from synchronic linguistic data in a similarly precise fashion as, for instance, Cienki (1998) has done for STRAIGHT. However, the above semantic extensions appear to have their cognitive foundations in an experiential schema emerging from a relatively well delineated, reoccurring perceptual and kinesthetic pattern, which figures in many types of bodily interaction with our environment. In the historical semantic data above this pre-conceptual BEND/CURVE schema seems to lead first to a more elaborate structuring of conceptual space, from which then the lexicalized concepts derive at the linguistic level. 8.3 The conceptualization of basic oppositions as revealed through semantic change In some languages (and across language families) there are certain lexical correspondences that are treated as cognates, although their semantic contents manifest conceptual oppositions. In this section I will look at such data in some Indo-European and Finno-Ugric languages and raise the assumption that they reveal important facets of human cognition. I will try to deduce some cognitive implications of this diachronic phenomenon and examine what cognitive processes might lead to these linguistic developments.

The phenomenon of a lexeme acquiring an opposite meaning and at the same time retaining the original one is absolutely not uncommon. E.g. Latin dare ‘give,’ Proto-Slavic *da- ‘give,’ and Hittite da- ‘take’ are cognates deriving from PIE *do- ‘give.’ In this case we can witness a semantic development from one pole of an opposition to the other. My objective here is not to offer an explanation for this type of semantic change, but rather I will suggest that another type of process could also have led to the emergence of cognate sets involving opposites, especially in the case of basic oppositions, i.e., oppositions in basic perceptual domains.


Thus, I will limit my analysis to basic oppositions and examine how their cognitively special status may have influenced their semantic development. Consider the examples in Tables 2-5 below:

Table 2: Semantic changes from PIE *bhel- ‘to shine, flash, burn’

BLACK WHITE

Proto-Germanic *blak- ‘black’ >

Old English bæl ‘flame, funeral pile’

Old Norse bal ‘fire’ Old English blæc ‘black’ Icelandic blakkr ‘dark’ Swedish bläck ‘ink’ Old High German blah ‘ink’ Old Swedish Blakker ‘black,

dark’'

Proto-Germanic *blaik- ‘white’ >

Old English blæc-an ‘to bleach’

Old Norse bleikr ‘shining, white’

Swedish blek ‘pale, fallow’ German bleich ‘pale’ Old Swedish Blakker ‘pale,

fallow’

Table 3: Semantic changes from PIE *kel- ‘burning, warm’

HOT/WARM COLD

Lithuanian šiltas ‘warm’ Lithuanian šáltas ‘cold’

Latvian silts ‘warm’ Latvian salts ‘cold’

Old Church Slavic slana ‘hoarfrost’

Proto-Germanic *hlewaz > German lau ‘tepid’ Old Norse hlýr ‘warm’ Old English hleo(w) ‘cover (from cold)’


Table 4: Semantic changes from PIE *upo ‘under, up from under, over’

UP UNDER

Proto-Germanic *upp- ‘up, in’ > Old English up-, upp- ‘upward’ Gothic iup ‘upward’ --------------------------------------------- Proto-Germanic *ufana ‘on, above’ > Gothic ufar ‘over’

----------------------------- Gothic uf ‘under’

Table 5: Semantic changes from PFU *pal'a icecrust, frost; to freeze’

BURN (HOT) FREEZE (COLD)

Finnish pala- ‘burn, flame’ Finnish palele- ‘feel cold, freeze’

Lapp buolle- ‘burn’ Lapp buolâš- ‘frost, frosty’

Mordvin palo- ‘burn’ Mordvin palo- ‘freeze’

Estonian pala- ‘burn’

Hungarian fagy ‘frost; freeze’

Since it appears to be difficult at first glance to find a plausible explanation for the above semantic developments, theoretically we could postulate that the protoform already had two meanings and what we have to do with are in fact “binary homophones” or “homophonic antinomies” (York 1993: 238). However, an explanation of these data as cases of homonymy does not seem to be totally convincing because of two reasons. On the one hand, there is a relatively large number of cases exhibiting this type of semantic correspondence (the examples in Tables 2-5 provide only a selection), and on the other, this kind of rather disturbing homonymic clash would probably have resulted in the lexical change of one of the lexemes of the antonym pair. It is well known that


lexical change is a general solution for such cases in the development of languages (Palmer 1978: 331; Hock 1986: 298). Thus, if we had to do with homonymic clash in these instances, it would be strange why lexical change did not occur in their case.

The possibility of polysemy as an explanation for the above data is an option as well. After all, these lexemes may always have been polysemous. However, other data of polysemy (and semantic change) imply that this is not possible because polysemy always implies some prior (diachronic) bifurcation process (see Section 6.5). Based on this, it could be said that at the given stage of language development (Proto-Indo-European and Proto-Finno-Ugric in the examples) the bifurcation process had already taken place and the lexical items in question are polysemous after all. This attempt at a plausible solution brings up another problem. In the case of polysemy there is always one meaning that can be considered basic conceptually, i.e., all other meanings must be explicable as derived meanings. Such an explanation is a possibility only to the extent to which e.g. the concept WHITE can be assumed to derive from BLACK or vice versa. A plausible way out could be that the basic meaning was already extinct in Proto-Indo-European times and we are facing two parallel extensions from it. The strange thing is, however, that these parallel developments are symmetrical in the sense that they are exact opposites, i.e., the two opposing poles of a conceptual domain.

Having outlined the problems manifest in the above etymologies and considered them from the point of view of semantic change in some detail, I will now move on to investigate the kind of conceptualization that might underlie the semantic phenomenon in question. My starting point is the fact that any phenomenon can only be conceptualized on the basis of its relationship to other phenomena. The importance of this for semantic structure is that the semantic pole of an expression – a predicate – must always be characterized relative to a domain (Langacker 1987: 147). When we say that a predicate is characterized relative to a domain, we mean that a domain used in the characterization is always one level less abstract than the domain that the predicate itself defines. This holds for most predicates (Langacker 1987: 150). Thus, if a predicate is characterized by a basic domain, the predicate itself already pertains to an abstract domain that is one level above the basic level. The


predicates treated here all pertain to basic domains directly, since the oppositions dealt with are all perceptual oppositions, i.e., our knowledge about them derives from direct bodily interaction with our environment (see Table 6 below). As Langacker writes:

By definition, basic domains occupy the lowest level in hierarchies of conceptual complexity: they furnish the primitive representational space necessary for the emergence of any specific conception. (Langacker 1987: 149)

Table 6: Oppositions and their basic domains

OPPOSITION PERCEPTUAL MODALITY

SENSORY DOMAIN

black : white visual color

hot : cold haptic temperature

up : down visual, kinesthetic space

For my analysis this means that the predicates in question (viz. the semantic poles of the etyma in question) can only be characterized by the domains to which they themselves belong, since these domains are not reducible any more to more fundamental ones. Due to the fact that such a situation exhibits a special case, a special mechanism for characterizing such predicates should be assumed. This requires a more detailed examination of how knowledge about such domains is exactly derived.

In our direct bodily interaction with the environment, i.e., in our physical experience gained through our sensory organs, we detect recurrent perceptual patterns (e.g. visual, kinesthetic, etc.). This gives rise to preconceptual configurations in our minds, which are then used to organize our experience. These preconceptual configurations are what Johnson (1987: 29) and Lakoff (1987: 278) call image schemata. We perceive our experience as organized because such image schemata structure the domains of our experience. Since relying on previous experience is a fundamental property of the mechanism of structuring all experience (e.g.


Langacker 1987: 105), image schemata also function in imposing structure on more abstract domains of experience. It is by projecting them onto such more abstract domains that we make these meaningful. Since the domains considered here are all physical ones, they must be directly structured by image schemata. The image schemata that structure them are not projected on them from other domains because these are the domains in which these image schemata originally exist. In other words, we must assume that basic domains are structured by their own image schemata.

The next step I will take in my analysis is to examine how image schemata structure domains, especially basic ones. As Johnson (1987: 41) has shown, they are capable of structuring domains because they themselves have an internal structure that they impose on the given domain. Thus, e.g. the UP-DOWN image schema, which arises from visual and kinesthetic perception, imposes its internal structure directly on the domain in which it exists (or on another domain it is projected to). This is why we conceive of space as extending upwards and downwards among others.

The internal structure of image schemata is basic level structure in the sense that it is analyzable but not decomposable, i.e., its elements are inseparable from each other. This is so because they pertain to basic domains, which are not characterizable relative to other domains, but only to themselves. Johnson defines the internal structure of image schemata as experiential gestalts. Image schemata have internal gestalt structure as they are coherent unified wholes within our experience:

Any given schema can, of course, be analyzed and broken down simply because it has parts. But any such reduction will destroy the integrity of the gestalt, that is, will destroy the meaningful unity that makes it the particular gestalt that it is. (Johnson 1987: 44)

It is important to bear in mind that image schemata are characteriz-able as irreducible gestalts. Most gestalts can be broken down physically, which destroys their integrity, but what we get are new gestalts. In the case of image schemata, however, we have to do with basic gestalts, which can only be broken down theoretically, but not physically. What should be understood then by irreducible gestalts?


For any phenomenon to have a structure, at least two contrasted elements are needed. Since every image schema possesses two such elements (see below for more explanation), it is obvious that they have an internal structure. However, it must also be obvious that these structures are irreducible basic gestalts. There is no OUT without IN, or DOWN without UP, etc., and vice versa. In other words, we cannot conceptualize one element of the structure without the other. Pertaining to the basicness of an image schema, there is also a seeming circularity in its conceptualization (or preconceptualization). Let me illustrate this with the example of the CONTAINMENT schema. We cannot have a sense of IN and OUT unless we have a sense of CONTAINMENT, and we cannot have a sense of CONTAINMENT unless we have a sense of IN and OUT. But this circularity should not disturb us; it is a result of irreducibility, i.e., a true sign of basicness.

This irreducible structure is not only characteristic of image schemata. In general it can be stated that we cannot perceive something as something unless there is something else to contrast it to in the given perceptual domain, i.e., across one perceptual modality. As Langacker (1987: 101) has pointed out, comparison is one of the most fundamental cognitive abilities relevant for semantic structure. Anything perceived will gain significance in our cognitive processing only in as much as it functions as a target that is compared to a standard. Because of this, for the perceived element of an opposition to be able to stand out as a figure the other element necessarily has to serve as the ground (cf. Langacker 1987: 120). Thus, it is the nature of oppositions, just like that of image schemata, that they can only be perceived as gestalts, i.e., as unified wholes made up of two opposing poles of a domain.

Above I assumed that the basic internal structure of image schemata always involves two contrasted elements. Therefore, I also think that image schemata naturally have an internal polarity. All the image schemata mentioned in Johnson (1987: 126) and Lakoff (1987: 272-275) conform to this assumption. Consider the following image schemata as examples. The LINK schema involves an entity A as separate from an entity B to which it is connected, though of course the relationship is symmetrical. The PATH schema has to involve by definition a SOURCE and a GOAL as separate points in space making up the basic structure of a PATH. Even schemata like


LINEAR ORDER or FORCE conform to this “polar” structure. At least two separate entities are necessary to make up the most primitive linear order and in the case of force there always has to be one entity exercising force on another separate entity, since “there is no schema for force that does not involve interaction” (Johnson 1987: 43). Everything we perceive is perceived on the grounds that it contrasts with something else, i.e., we compare it to other experience or conceive of it as a figure that stands out against a contrasting ground (Langacker 1987: 121). Because of this I suggest that the natural way for basic perceptual oppositions to appear preconceptually is in the form of image schemata (and maybe there is also some ground for the claim that image schemata themselves are preconceptual forms of perceptual oppositions). The basis for such a claim lies in the observation that it is exactly perceptual oppositions that structure our sensations and make them meaningful, i.e., we make sense of our perceptions in terms of oppositions. This can well be observed in fundamental cognitive mechanisms as comparison or figure/ground alignment, where a target always stands in opposition to a standard, or a figure to a ground, respectively. These oppositions basically include the perception of features vs. non-perception of features on the one hand, and the perception of different degrees along a scale on the other. 8.4 Lexicalization processes based on image schemata Let us now turn to the language material in the light of the above theoretical considerations. On the basis of these I claim that just as it is impossible to conceptualize one pole of a conceptual opposition without conceptualizing the other, or conceptualize only one part of an image schema, it is impossible to have a lexical gap at one pole of an opposition. This means that either the domain in question is not coded in any way in a language or both poles have to be coded. York (1993: 238) points out that “[t]hese homophonic antinomies are not a universal feature of the proto-language, but one which occurs often enough to be indicative of the possible IE tendency toward polarized perception ... .” In my view, this phenomenon should not be called homophony because it is exactly “polarized perception” that requires both poles to be coded in parallel at any historical stage in a


language. This presumes a simultaneous coding process, which can only be the case if the coding of the two poles has one common origin. This means that the two forms are cognates and therefore polysemic antinomies rather than homophonic ones. This type of simultaneous coding, i.e., deriving the lexemes for the two poles from one common source, is not simply a convenient linguistic solution, but as we have seen, also has psychological reality. At least at the conceptual (or preconceptual) level the simultaneous recognition of the two poles is compelling and may even be preceded by a holistic phase of knowledge of the phenomenon in question. This would speak for the universality of this phenomenon, which of course might have already been obscured in many cases by the Proto-Indo-European stage.

On this basis I will try to give a new interpretation of the semantic developments within the cognate groups of the conceptual oppositions I dealt with above. We have seen that a conceptual opposition appears preconceptually in an image schema. Because of this I would suggest that the etyma in question are cases in which a complete image schema became lexicalized, i.e., coded in the language. I propose that the notion “archilexeme” can be usefully applied in the historical sense to this linguistic phenomenon. An archilexeme, as used in lexical field studies, is a lexeme which neutralizes two poles of a semantic dimension that are distinct in lexemes with otherwise identical semantic contents (cf. Kastovsky 1988: 197, Mettinger 1988: 151). Cruse (1986: 255) defines neutrali-zation as the non-appearance of a semantic contrast. Thus, by way of example, the lexeme child is an archilexeme in the sense that the gender difference readily apparent in its two hyponyms boy and girl is obliterated in it (Mettinger 1988: 151).

By utilizing the notion archilexeme in a historical framework, we can consider the etyma in question to be archilexemes in the sense that the semantic contrast apparent in their later developments can still be found in a neutralized state in the Proto-Indo-European etyma themselves. In cognitive linguistic terms we can say that the image schema structures were still coded (lexicalized) in a homogeneous way at the protolinguistic stage. I can be assumed that these archilexemes were at best only internally differentiated for poles, and expressed a preconceptual idea of ‘verticality,’ ‘temperature,’ and ‘darkness-brightness.’ However, the internal


dynamism of the underlying schemata (cf. Johnson 1987: 29), due to their polarity, caused their splitting up into their poles, and a semantic development into opposite directions began. This splitting up took place of course only at the lexical level, because as we have seen, an image schema is not decomposable at the conceptual level. Thus, the etyma gave rise to parallel but contrasting lexical extensions in their own rights.

This kind of semantic development can plausibly be postulated for basic oppositions, since the conceptual inseparability of the two poles makes it likely that the development was not from one pole to the other, but it was a parallel one starting out from a common etymon denoting the complete notion of the opposition as such with only internally differentiated poles (a holistic phase). In the case of non-basic oppositions, however, especially if they are non-binary, i.e., the two poles are conceptually separable, a semantic conversion of one pole into the other seems to be possible. An example for this is the already mentioned case of PIE *do- ‘give,’ which retained its meaning in Latin (dare ‘give’) and in the Slavic languages (e.g. Russian dat’ ‘give’), but changed its meaning to its opposite in Hittite (da- ‘take’). The semantic reconstruction and thus the postulated semantic development are plausible because the two notions can be maintained independently: taking does not necessarily imply giving by the other party and vice versa. In spite of this, it would be mistaken to assume that such an opposition has no image schema structure. The opposition GIVE vs. TAKE is non-basic in the sense that several more basic domains form its matrix. Minimally it can be reduced to the basic oppositions contained in the LINK and SOURCE-PATH-GOAL schemata.

Thus, the semantic developments of PIE *bhel- ‘to shine, flash, burn’ into its derivatives with these separate meanings imply the metonymy BURNING FOR BRIGHTNESS/DARKNESS. The metonymical projection from the source domain BURNING to the target domain BRIGHTNESS/DARKNESS can be explicated in the following way: Brightness/darkness is caused by burning in the sense that when something burns there is brightness but at the same time the thing that burns will become dark (i.e., burned). Because of this brightness/darkness is conceptualized as the state(s) caused by burning. Among the derivatives of PIE *bhel- terms denoting some shade of color are more frequent than terms denoting ‘burn.’ This


fact together with the psychological reality of simultaneous conceptualization (or preconceptualization) of oppositions seems to indicate that the metonymical projection might have run in the other direction as well. Since both domains (source and target) are physical ones, this is not in the least inconceivable. A reversal of source and target domains in this case may be even more realistic on the grounds of the compelling simultaneity of the coding of the two poles. This would speak for the initial coding of a coherent image schema as a holistic phenomenon. On this view, the metonymy BRIGHTNESS/DARKNESS FOR BURNING could be postulated as the basis of the extensions. Beside the above reasons for such a reversal, there is an even more fundamental motive. On the basis of Langacker’s (1987: 149) distinction between basic and abstract domains, only the domain of BRIGHTNESS/DARKNESS can be considered to be a basic one because it is grounded directly in sensory experience. Furthermore, due to its gestalt structure, it can even be postulated as a basic image schema. Even though both domains (i.e., both BRIGHTNESS/DARKNESS and BURNING) pertain to physical reality, only the conception of BRIGHTNESS/DARKNESS relies directly on our visual sensation of light and is irreducible in the sense that it cannot be explicated in terms of more fundamental domains. BURNING on the other hand is clearly an abstract domain, since it is readily characterizable in terms of, i.e., reducible to, basic domains like the sensation of light and temperature. It is on this basis that BRIGHTNESS/DARKNESS is more likely to function as source domain for the projection (and semantic extension) to the target domain BURNING. Since BURNING is saliently characterized by BRIGHTNESS/DARK-NESS, BURNING is best conceptualized as the case when there is BRIGHTNESS/DARKNESS (i.e., light plus burned substance as the outcome).

In the case of PIE *kel- ‘burning, warm’ and PFU *pal’a ‘icecrust, frost; to freeze’ the reconstructed semantic contents of the original etyma serve as an even less adequate starting point for a plausible explanation of semantic development, since they are reduced to only one pole of the opposition. In view of the above considerations, and of the common knowledge that the touching of objects both with very high and very low temperatures (burning and freezing) cause similar physical sensations, semantic developments


comparable to the case of PIE *bhel- could be postulated. Such developments could be based on the metonymies BURNING FOR HOT/COLD and FREEZING FOR HOT/COLD, with an underlying HOT/COLD basic image schema. The semantic reconstruction of PIE *upo ‘under, up from under, over’ from its branching derivatives seems to be more in line with the present approach to the lexicalization of perceptual oppositions, and should be explicable in terms of the UP/DOWN schema.

A brief remark should be added here concerning the relevance this analysis (especially of the semantic development of PIE *bhel-) might have on the theory of the evolution of basic color terms. Berlin and Kay (1969) consider the stage in the evolution of color terms when only macro-white and macro-black are coded in a language to be stage one. On the basis of what has been said above, a stage prior to stage one could be postulated, i.e., to the ‘macro-white macro-black’ stage, which I would like to call stage zero, i.e., the stage of the coding of the undifferentiated image schema. I think that such a stage zero could be postulated for all the investigated oppositions.

Of course, one could question the reality of stage zero and wonder what the real meaning of the archilexeme might have been. In other words, can we postulate a realistic meaning, i.e., a realistic archisememe? Obviously, a totally neutralized form would have been communicatively inefficient and thus unthinkable; it could serve no communicative purpose. However, as I have claimed, these archilexemes are lexicalized image schemata and because of this they should reflect the nature of image schemata. An image schema is a gestalt, a unified whole, but has internal structure given by its polarity. Thus, the archilexeme cannot be anything but a gestalt itself that displays internal polarity. The Lithuanian antonym pair šiltas ‘warm’ šáltas ‘cold’ may serve as a model for illustrating this idea. In this word pair the consonant structure can be considered to give the gestalt (i.e., the perceptual dimension) and the vowels together with suprasegmental features (circumflex vs. acute accent) provide the internal polarity. Thus, in Proto-Indo-European the expression of the internal polarity could have been realized by different ablaut grades, or if Proto-Indo-European indeed had politonous accent (which may also be surviving in Lithuanian), then it could even have been realized by this, i.e., rising, falling, or rising


falling accent, or as a third alternative, simultaneously by both ablaut grades and accent differences. Again Lithuanian may provide a good example: it has an accent that can be semantically distinctive. Many linguists consider Lithuanian to be archaic in this respect, and this may be a relic feature surviving from Proto-Indo-European times (cf. Szemerényi 1972: 137).

Conclusion: Language as cognitive

adaptation

The evolution and adaptive value of symbolic cognition Language provides a means of symbolic representation, in which an exclusively interindividual function, viz. communication, and a primarily individual function, viz. cognition, have been inseparably tied together. Thus, human language unites two kinds of biological capacities that are distinct in non-human animals: communication and cognition. Although human cognition is definitely continuous with cognition in non-human animals in the evolutionary biological sense, this fact adds a qualitative difference to human cognition because, as a consequence of this joint function of language, human cognition can operate also in the symbolic mode. This unique capacity, i.e., symbolic cognition, is a third phase in the evolution of cognitive capacities, building on the mechanisms of genetic and neural cognition, which appeared earlier in evolution. The adaptive advantage of symbolic cognition is twofold. On the one hand, such cognition vastly enhances the possibility to acquire knowledge about reality without direct experience based on perceptual mechanisms of the individual, and on the other hand, it provides the possibility and mechanism to construe reality in alternate ways, which facilitates a wider and more flexible perspective of the environment.

Thus, humans are the species that possess the most powerful device for sharing knowledge. With the help of language we are capable of exchanging knowledge by far to the largest extent as compared to other animals (Csányi 1992). However, human cognition differs from the cognition of primates only by being supplemented by a symbolic capacity, which enables the cultural dissemination of knowledge with the help of language, and by this the formation of a common perspective of their natural and cultural environment for a whole community (Tomasello 2002). This fact suggests that language must have evolved as simultaneously serving


cognitive and communicative functions in the social interaction of hominids. Therefore, language should be considered the product of both biological and social evolution.

The capacity for symbol usage can only be based on an extraordinarily developed nervous system, and nervous systems are evolutionary adaptations for the capacity of processing individual experience with the environment and effecting a more flexible behavior as a result. Consequently, it is highly unlikely that language could be discontinuous with any antecedent biological, cognitive and behavioral trait (Maynard Smith and Szathmáry 1995), and thus elements of the primate communication system, cognitive capacity and social behavior must have served as evolutionary forerunners of language in the form of preadaptations. The changes must have depended on the extent of the selection pressures to which evolving hominids, at some stage in their evolution, responded by developing language, in which the formerly separate capacities of communication and cognition became conjoined for the sake of a more efficient social cognitive interaction.

Hence, language should be seen as the product of a selective process for cognitive adaptation in the social interaction of evolving hominids for sharing and representing individually acquired knowledge (cf. Catania 2001; Steele 1989). Through language all members of a speech community can benefit from the knowledge of others, even from that of previous generations, without direct individual experience. In this sense language is indeed an innovation of the hominid lineage, even though this can not mean that it is a development totally independent of animal communication and cognition. Even in the symbolic mode the human acquisition of new knowledge starts with the activation of existing cognitive structures, as is characteristic of cognition in general (Neisser 1976). But language enables us to construct new cognitive structures from existing ones without the need for direct empirical contact with the parts of reality that the new cognitive structures represent. There appear to be two basic questions to be asked in connection with this kind of cognitive functioning of language: (i) how symbols are connected to reality and (ii) how they mediate knowledge about reality.

As already mentioned above, a characteristic feature of human cognition is that even phenomena that fall outside perceptual and

Conclusion 221

individual experience can become its objects. This is possible because in the course of human evolution the capacity for symbolic cognition emerged to supplement the capacity for processing genetically transmitted and experience based knowledge. Mental representations based on symbols can store knowledge that is not the result of individual experience. This fact has puzzled philosophers for two millennia, and in the rationalist tradition the way symbols work has been explained in the following manner (Chomsky 1988). Symbols get their meaning via their relation to each other based on the fact that they stand in a one to one relationship to reality and thus their relation to each other mirrors the relation between entities of the world in an objective way. Symbols are indeed special in the sense that they exhibit the property of inclusion of each other’s meanings and can thus be defined relative to each other, though this might not exclusively be the case. This characteristic of symbols also allows for the formation of symbols that are not directly experientially grounded and makes a symbol system possible that is only partially and rather indirectly linked to the physical world. However, an explanation to the effect that symbolic cognition serves the understanding of reality through the manipulation of abstract symbols remains unsatisfactory because it results in a vicious circle of an unresolvable chain of symbols (Harnad 1990).

If symbols do not get their capacity to objectively mirror the world – their meaning, in other words – purely on the basis of their structural relation to each other or some kind of a priori structural knowledge, then there has to be some deeper empirical and functional connection between symbols and reality. It should also be noted here that not even the autonomousness and arbitrariness of grammatical structure is generally accepted among linguists (Haiman 1983, Langacker 1987). If language is a cognitive adaptation in the evolutionary biological sense, as it obviously is, then it can only be the function that a structure has that is of adaptive value, and not the structure in itself. This is why an account of the selective advantage of language in terms of the adaptive value of its structure alone, which Pinker and Bloom (1990) have tried, is doomed to failure at the onset. Structure in itself is not adaptive, and evolution can create different structures even for the same function. Structure can even be said to be a result of chance as long as we realize that evolutionary solutions are always dependent on the possibilities of already given structures


(Dawkins 1986, Mayr 1988). However, the fact that symbolic cognition is not about the

objective representation of reality does not mean an imperfect and inaccurate cognition of reality, but rather a much more flexible and multilateral adaptation to it. Any cognitive adaptation to the environment and the orientation in it is founded in the acquisition of knowledge about it, and the acquisition of this knowledge and its application for interaction with the environment is a biologically determined ability in every organism, including humans (Plotkin 1994). Therefore, at the very basis of symbolic cognition appear to be other human cognitive capacities that can be derived from general cognitive capacities both cognitively and in terms of evolutionary biology. Such appears to be our capacity for analogy, where one mental representation imposes its structure on another, i.e., directs our attention to another phenomenon than the one it represents by highlighting particular features of it (Holyoak and Thagard 1997). Analogy turns into a metaphor in linguistic categorization when a conceptual category is used to denote another conceptual category on the basis of this mechanism. In this way our abstract symbol system is empirically grounded as a result of metaphorical projections of sensory experience (e.g. Johnson 1987; Lakoff 1987; Givón 1998; Tomosello 2002).

The question how symbolic cognition adapts language to the environment has only been treated very briefly and in very general terms by only a few scholars. Such an inquiry entails the following questions: how the category system of a language reflects the environment of its speakers; how and why the category system of a given language has developed; and what factors influence its change and modification as a response to environmental fluctuations. Answers to these questions can only be given in the light of the place symbolic (linguistic) cognition takes in human cognition in general and in the more extensive scope of cognition in non-human organisms, or in other words, in terms of the way the general features of biological cognition influence or determine the characteristics of linguistic cognition.

The general biological function of cognition is the creation and operation of an adaptive functional internal model for an organism about its environment for the sake of the regulation and facilitation of orientation and survival in it. Thus, my basic claim is that the

Conclusion 223

evolutionary biological function of the capacity for language as cognitive adaptation carries over to the way language as a symbol system functions cognitively, i.e., how any particular language adapts to the environment in which it is used by a speech community to serve as a functional cognitive model. This model is flexibly and adaptively shaped in the communicative interaction of speakers and hearers so that it can functionally represent their particular natural and sociocultural environment in order to regulate the common interaction of the interlocutors with this environment and thus to ensure adaptive orientation in it. Cognitive adaptation in language According to Rosch (1978), specific principles of categorization govern the formation of a category system representing the environment in an adaptive and functional way in every organism, and in the case of humans these principles are manifest in the cultural formation of linguistic categories. Since this category formation can by the nature of the matter only be a language specific process, the language differences across cultures obviously reflect – from a functional and cognitive point of view – the adaptation of particular languages to the different environments in which they are spoken. More precisely, a given environment is reflected in the linguistic categories, i.e., in the conceptual structures forming their meanings, of the language spoken by a particular speech community. The adaptation process in which the category system of a language is formed and modified is the historical linguistic mechanisms of semantic-lexical change. This mechanism adjusts the semantic structure of a language to function appropriately as a culturally shared system of conceptual categories in the natural and sociocultural environment of a speech community. Herein lies the real adaptive significance of semantic-lexical change with regard to the cognitive functioning of language.

Thus, it can be concluded that the semantic structure of a particular language mirrors the way human cognition has filtered the different specific and universal influences from the natural, social and cultural environment that have affected speakers of that language. The universal tendencies and relativity effects that can


both be observed in semantic change follow directly from this functionality. Both are due to the embodied nature of meaning, which originates in the evolutionary function of cognition as the biological mechanism for the coordination of adaptive interaction with the environment and from which also the cognitive function of language derives. Universally occurring patterns in human perception and bodily interaction with the environment, coupled with universal human cognitive mechanisms, will naturally lead to similar ways of making sense of reality and thus to universal tendencies in particular semantic extensions. In other words, the observable similarities in more or less general aspects of various conceptualizations across languages (as manifest in their compatible semantic structures) derive mainly from the way definite universal cognitive factors affect the choice of the underlying familiar knowledge structures to be elaborated on in particular conceptualizations. Such familiar knowledge structures may originate either in culture-specific or universal human experience, or even genetic endowment like our species-specific physiological capacity for color vision leading to our conceptual knowledge of colors. The universal cognitive factors influence or even determine the cognitive salience of definite aspects of these knowledge structures. However, certain culture-specific factors may alter this salience and thus override the influence of the universal factors when the natural or sociocultural environment requires an alternate functional interpretation – or construal – of the world when this proves more adaptive. In addition to culture-specificity, language specificity is not a negligible factor either. Conceptualizations are also influenced by the category system available in linguistic form. The way new experience or new perspectives are conceptualized will also be constrained by the semantic knowledge of the speakers of that particular language.

The nature of language as cognitive adaptation reveals itself in the form of two different adaptation processes in language at two interconnected levels, the cumulative effect of which results in semantic change. At the first level the process of ad hoc innovative usage in the everyday linguistic activity of speakers adjusts the language temporarily to novel conditions occurring in the form of immediate cognitive-representing and communicative-referring needs. At the basis of this occasion-bound usage of conventional

Conclusion 225

expressions is the cognitive process of adaptive modification of conceptual structures. This happens through the flexible, adaptive and functional application of the categories (meanings) of the language in order for the speaker to see and make seen the phenomena of the environment from the cognitive perspective which he deems most suitable for his purposes. The adaptive value of the innovations of individual speakers is determined by certain linguistic and sociocultural factors, which then influence the survival of the modified and novel meanings at the group level.

Thus, the adaptive innovation in the usage of expressions feeds the second level, which is the historical process of their spreading and conventionalization. At this level language as a cognitive-communicative system becomes adapted to the conditions which have originally triggered the innovative usages at the first level. If these conditions persist and become culturally salient, the original innovations resulting from the adaptive linguistic behavior of individual speakers are preserved to become established in the system of the language. This is the process of cultural category formation, i.e., the linguistic coding of conceptual categories necessary for communicating about and mentally representing any changes in the specific environment of the speech community or in the perspectives collectively taken on it.

The fact that the processes of change are based on selection mechanisms at both levels provides evidence for the Darwinian nature of these adaptations. The mechanisms of selection operate simultaneously in the form of cognitive, cultural and linguistic constraints. In everyday linguistic activity selection occurs when the speaker makes an (unconscious) choice from the variability of conceptual structures – available in the form of conventional expressions – as to the most potent one for semantic modification in order to succeed in the given communicative circumstance. Selection is affected here by specific cognitive factors as to its adaptiveness and also by structural limitations in the form of the linguistic constraint of the contingent sanctioning of the new usage by the grammar. In addition to these concrete constrains on selection, cultural aspects of acceptability, like considerations of political correctness, politeness, traditions, group customs, fashion, etc., also play a significant role.

Selection processes are at work also in the spreading of


innovations within the speech community and across generations, which is the actual process of semantic change in the diachronic linguistic sense: as a result of the spreading of new usage conventionalization will take place and the semantic structure of the language will become modified. During the diffusion of novel usage selection occurs from the variability of available innovations through an (unconscious) preference by the speech community, which preference is in fact the manifestation of an adaptive linguistic behavior. The innovations that come to be selected are the ones that prove to be the most appropriate functional and adaptive conceptualizations of given phenomena as determined by the persistence of the conditions triggering the new usage and by its subsequent acceptance by the members of the speech community. The selection is affected here by the same cognitive, linguistic and cultural factors as in the case of the selection mechanisms working in ad hoc innovations of individual speakers. The reason for this is on the one hand that acceptance of an innovation by the speech community actually means the (unconscious) positive evaluation of the new usage by other individual speakers, and on the other that the innovation happens in (unconscious) anticipation of this acceptance.

Thus, the semantic-lexical changes going on in a language, when taken together, form a continuous cognitive adaptation process in the language system as a social cognitive model. This historical process is the result of adaptive linguistic behavior at the level of synchronic language use, which is manifest in the perspectival application of categories, or meanings. This is possible because the cognitive background of meanings is formed by conventionalized but malleable conceptual structures. Thus, linguistic categories are representations adapted to the given environment of the language users, and continuously adapt to the occurring cognitive needs in the course of the speakers’ interaction with their environment.

Consequently, the cognitive function of language cannot be separated from the general biological function of cognition, which is the flexible adaptive and functional representation of the given environment for the sake of guiding the interaction with that environment, including the adaptation of these representations to those environmental changes that have a bearing on the appropriate behavior in it. The task of linguistic cognition is therefore the formation of culturally shared mental categories which represent the

Conclusion 227

human environment in an adaptive and functional way. From this it follows that this function is manifest only in the particular category systems of individual languages as well as in the flexible application and continuous modification of these category systems. Thus, language is not an absolute model conveying some kind of metaphysical truth about reality, but every individual language is the relative (i.e., adaptive and functional) social-cognitive model of the immediate environment of its speakers.

References Algeo, John 1990, Semantic change. In: Edgar C. Polomé (ed.),

Research guide on language change. Berlin and New York: Mouton de Gruyter, 399-408.

Anderson, John A. 1988, Concept formation in neural networks: Implications for evolution of cognitive functions. Human Evolution 3: 81-97.

Anttila, Raimo 1989, Historical and comparative linguistics. (2nd revised edition) Amsterdam and Philadelphia: John Benjamins.

Anttila, Raimo 1992, Field theory of meaning and semantic change. In: Günter Kellermann and Michael D. Morrissey (eds.), Diachrony within synchrony. Language history and cognition. Frankfurt am Main: Peter Lang, 23-83.

Armstrong, David F., William C. Stokoe and Sherman E. Wilcox 1995, Gesture and the nature of language. Cambridge: Cambridge University Press.

Beck, Brenda E.F. 1978, The metaphor as a mediator between semantic and analogic modes of thought. Current Anthropology 19(1): 83-94.

Benton, M.J. 1990, Vertebrate paleontology. London: Chapman and Hall.

Berlin, Brent and Paul Kay 1969, Basic color terms. Their universality and evolution. Berkeley and Los Angeles: University of California Press.

Berndt, Rolf 1989, A history of the English language. Leipzig: Verlag Enzyklopädie.

Bichakjian, Bernard H. 1988, Evolution in language. Ann Arbor, MI: Karoma.

Bickerton, Derek 1990, Language and species. Chicago: The University of Chicago Press.

Bickerton, Derek 1995, Language and human behavior. Seattle: University of Washington Press.

Bierwisch, Manfred 1992, Probleme der biologischen Erklärung natürlicher Sprache. In: Peter Suchsland (ed.), Biologische und

References 229

soziale Grundlagen der Sprache. (Linguistische Arbeiten 280: 7-45). Tübingen: Max Niemeyer.

Blank, Andreas 1997, Prinzipien des lexikalischen Bedeutungswandels am Beispiel der romanischen Sprachen. Tübingen: Max Niemeyer.

Blank, Andreas and Peter Koch (eds.) 1999, Historical semantics and cognition. Berlin and New York: Mouton de Gruyter.

Bonner, John T. 1980, The evolution of culture in animals. Princeton, NJ: Princeton University Press.

Bornstein, Marc H. 1987, Perceptual categories in vision and audition. In: Stevan Harnad (ed.), Categorical perception. The groundwork of cognition. Cambridge: Cambridge University Press, 287-300.

Bradshaw, John L. and Lesley J. Rogers 1993, The evolution of lateral asymmetries, language, tool use, and intellect. San Diego, etc.: Academic Press.

Brown, Cecil H. 1979, A theory of lexical change: With examples from folk biology, human anatomical partonomy and other domains. Anthropological Linguistics 21: 257-276.

Buck, Carl Darling 1949, A dictionary of selected synonyms in the principal Indo-European languages. Chicago and London: The University of Chicago Press.

Burling, R. 1993, Primate calls, human language, and non-verbal communication. Current Anthropology 34: 25-53.

Campbell, Lyle 1998, Historical linguistics. An introduction. Edinburgh: Edinburgh University Press.

Carroll, John 1985, What’s in a name? An essay in the psychology of reference. New York: Freeman.

Catania, A. Charles 1991, The phylogeny and ontogeny of language function. In: N.A. Krasnegor, D.M. Rumbaugh, R.L. Schiefelbusch and M. Studdert-Kennedy (eds.), Biological and behavioral determinants of language development. Hillsdale, NJ: Lawrence Erlbaum, 263-285.

Catania, A. Charles 1994, The natural and artificial selection of verbal behavior. In: S.C. Hayes, M. Sato and K. Ono (eds.), Behavior analysis of language and cognition. Reno, NV: Context Press, 31-49.

Catania, A. Charles 2001, Three varieties of selection and their implications for language origins. In: Gábor Győri (ed.),

230 References

Language evolution. Biological, linguistic and philosophical perspectives. Frankfurt am Main: Peter Lang. 55-71.

Cheney, Dorothy L. 1984, Category formation in vervet monkeys. In: Vernon Reynolds and Rom Harré (eds.), The meaning of primate signals. Cambridge: Cambridge University Press, 58-76.

Chomsky, Noam 1968, Language and mind. New York: Harcourt, Brace and Jovanovich.

Chomsky, Noam 1985, Knowledge of language. Its nature, origin and use. New York: Praeger.

Chomsky, Noam 1988, Language and problems of knowledge. The Managua lectures. Cambridge, MA: The MIT Press.

Chomsky, Noam 2000a, New horizons in the study of language and mind. Cambridge: Cambridge University Press.

Chomsky, Noam 2000b, The architecture of language. New Delhi, etc.: Oxford University Press.

Cienki, Alan 1998, STRAIGHT: An image schema and its metaphorical extension. Cognitive Linguistics 9: 107-149.

Clark, Herbert H. 1996, Communities, commonalities, and communication. In: John J. Gumperz and Stephen C. Levinson (eds.), Rethinking linguistic relativity. Cambridge: Cambridge University Press, 324-355.

Clarke, David D. and Nerlich, Brigitte 1991, Word-waves: A computational model of lexical semantic change. Language and Communication 11: 227-238.

Clausner, Timothy C. and William Croft 1999, Domains and image schemas. Cognitive Linguisitcs 10: 1-31.

Corballis, Michael C. 1991, The lopsided ape. Evolution of the generative mind. New York and Oxford: Oxford University Press.

Coulson, Seana 2001, Semantic leaps. Frame-shifting and conceptual blending in meaning construction. Cambridge: Cambridge University Press.

Coulson, Seana and Todd Oakley 2000, Blending basics. Cognitive Linguistics 11: 175-196.

Croft, William 2000, Explaining language change. An evolutionary approach. Harlow: Longman.

Cruse, D. Alan 1986, Lexical semantics. Cambridge: Cambridge University Press.

Csányi, Vilmos 1988, Contribution of the genetical and neural memory to animal intelligence. In: Harry J. Jerison and Irene

References 231

Jerison (eds.), Intelligence and evolutionary biology. Berlin: Springer-Verlag, 299-318.

Csányi, Vilmos 1989, Evolutionary systems and society. A general theory. Durham, NC: Duke University Press.

Csányi, Vilmos 1992, The brain’s models. In: Thomas A. Sebeok and Jean Umiker-Sebeok (eds.), The semiotic web. Berlin: Mouton de Gruyter, 27-43.

Csányi, Vilmos 2000, The “human behavior complex” and the compulsion of communication: key factors of human evolution. Semiotica 128(3/4): 45-60.

Csányi, Vilmos 2001, An ethological reconstruction of the emergence of culture and language during human evolution. In: Gábor Győri (ed.), Language evolution. Biological, linguistic and philosophical perspectives. Frankfurt am Main: Peter Lang, 43-53.

Davidson, Ian and W. Noble 1993, Tools and language in human evolution. In: Kathleen R. Gibson and Tim Ingold (eds.), Tools, language and cognition in human evolution. Cambridge: Cambridge University Press, 363-388.

Dawkins, Richard 1986, The blind watchmaker. New York and London: Norton.

Dawkins, Richard and J.R. Krebs 1978, Animal signals: Information or manipulation? In: J.R. Krebs and N.B. Davies (eds.), Behavioral ecology: An evolutionary approach. Oxford: Blackwell, 282-309.

Deacon, Terrence W. 1988, Human brain evolution: I. Evolution of language circuits. In: Harry J. Jerison and Irene Jerison (eds.), Intelligence and evolutionary biology. Berlin: Springer, 363-381.

Deane, Paul D. 1988, Polysemy and cognition. Lingua 75: 325-361. Dirven, René 1985, Metaphor as a basic means for extending the

lexicon. In: Wolf Paprotté and René Dirven (eds.), The ubiquity of metaphor. Amsterdam: John Benjamins, 85-119.

Dirven, René 1993, Metonymy and metaphor: Different mental strategies of conceptualisation. Leuvense Bijdragen 82: 1-25.

Dobzhansky, Theodosius, Ernst Boesiger and Diether Sperlich 1980, Beiträge zur Evolutionstheorie. Jena: Gustav Fischer.

Donald, Merlin 1991, Origins of the modern mind. Three stages in the evolution of culture and cognition. Cambridge, MA: Harvard University Press.

232 References

Drosdowski, Günther et al. 1963, Das Herkunftswörterbuch. Die Etymologie der deutschen Sprache. Mannheim: Bibliogra-phisches Institut.

Eldredge, Neil and Stephen J. Gould 1972, Punctuated equilibria: An alternative to phyletic gradualism. In: T.J. Schopf (ed.), Models in paleobiology. San Francisco: Freeman, 82-115.

Elena, S.F., V.S. Cooper and R.E. Lenski, 1996, Punctuated evolution caused by selection of rare beneficial mutations. Science 272: 1802-1804.

Faltz, Leonard M. 1989, A role for inference in meaning change. Studies in Language 13: 317-331.

Fauconnier, Gilles 1994, Mental spaces. Aspects of meaning construction in natural language. Cambridge and New York: Cambridge University Press.

Fauconnier, Gilles 1997, Mappings in thought and language. Cambridge: Cambridge University Press.

Fodor, Jerry 1987, Psychosemantics. The problem of meaning in the philosophy of mind. Cambridge, MA: The MIT Press.

Fodor, Jerry 1994, The elm and the expert. Mentalese and semantics. Cambridge, MA: The MIT Press.

Fouts, Roger S. and Joseph B. Couch 1976, Cultural evolution of learned language in chimpanzees. In: Martin E. Hahn and Edward C. Simmel (eds.), Communicative behavior and evolution. New York: Academic Press, 141-161.

Fox, Anthony 1995, Linguistic reconstruction. An introduction to theory and method. Oxford: Oxford University Press.

Fritz, Gerd 1998, Historische Semantik. Weimar: J.B. Metzler. Frost, Thomas F. 1980, Tool behavior and the origin of laterality.

Journal of Human Evolution 9: 447-459. Gallup, G.G. Jr., J.L. Boren, G.J. Gagliandi and L.B. Wallman 1977,

A minor for the mind of human, or will the chimpanzee create an identity crisis for Homo sapiens? Journal of Human Evolution 6(3): 303-313.

Geeraerts, Dirk 1983a, Prototype theory and diachronic semantics: A case study. Indogermanische Forschungen 88: 1-32.

Geeraerts, Dirk 1983b, Reclassifying semantic change. Quaderni di Semantica 4: 217-240.

Geeraerts, Dirk 1985, Cognitive restrictions on the structure of semantic change. In: Jacek Fisiak (ed.), Historical semantics and

References 233

historical word-formation. The Hague: Mouton, 127-153. Geeraerts, Dirk 1988, Cognitive grammar and the history of lexical

semantics. In: Brygida Rudzka-Ostyn (ed.), Topics in cognitive linguistics. Amsterdam: John Benjamins, 647-677.

Geeraerts, Dirk 1994, Semantic change (Laws of). In: R.E. Asher (ed.), The encyclopedia of language and linguistics. Vol. 7. Oxford, New York, etc.: Pergamon Press, 3800.

Geeraerts, Dirk 1997, Diachronic prototype semantics. A contribution to historical lexicology. Oxford: Clarendon Press.

Geeraerts, Dirk 1999, Diachronic prototype semantics: A digest. In: Andreas Blank and Peter Koch (eds.), Historical semantics and cognition. Berlin and New York: Mouton de Gruyter, 91-107.

Gentner, Dedre and A.B. Markman 1997, Structure mapping in analogy and similarity. American Psychologist 52: 45-56.

Gibbs, Raymond W. Jr. and Herbert L. Colston 1995, The cognitive psychological reality of image schemas and their transformations. Cognitive Linguistics 6: 347-378.

Gibson, Kathleen R. 1988, Brain size and the evolution of language. In: Marge E. Landsberg (ed.), The genesis of language. A different judgement of evidence. Berlin: Mouton de Gruyter, 149-172.

Givón, Tom 1998, On the co-evolution of language, mind and brain. Evolution of Communication 2: 45-116.

Goodwin, Brian 1996, How the leopard changed its spots. The evolution of complexity. New York: Touchstone. (1

st edition

1994) Gould, Stephen J. 1980, Is a new and general theory of evolution

emerging? Paleobiology 6: 119-130. Gould, Stephen J. and E.S. Vrba 1982, Exaptation – a missing link in

the science of form. Paleobiology 8: 4-15. Green, Steven 1975, Dialects in Japanese monkeys: Vocal learning

and cultural transmission of locale-specific vocal behavior? Zeitschrift für Tierpsychologie 38: 304-314.

Green, Steven and Peter Marler 1979, The analysis of animal communication. In: P. Marler and J.G. Vandenbergh (eds.), Handbook of behavioral neurobiology. Vol. 3. New York: Plenum Press, 73-158.

Griffin, Donald R. 1978, Prospects for a cognitive ethology. Behavioral and Brain Sciences 1: 527-538.

Győri, Gábor 1992, Über Spezifik und Entstehung der

234 References

Sprachfähigkeit. In: Peter Suchsland (Hrsg.), Biologische und soziale Grundlagen der Sprache. (Linguistische Arbeiten 280: 47-53). Tübingen: Max Niemeyer.

Győri, Gábor 1994, Historical motivation in the linguistic sign and its cognitive origin. In: Jan Wind, Abraham Jonker, Robin Allott and Leonard Rolfe (eds.), Studies in Language Origins. Vol. 3. Amsterdam: John Benjamins, 221-234.

Győri, Gábor 1995a. The evolution of language and languages: The problem of origins. LOS Forum. Journal of the Language Origins Society. No. 21: 17-26.

Győri, Gábor 1995b, Animal communication and human language: Searching for their evolutionary relationship. In: Stanislaw Puppel (ed.), The biology of language. Amsterdam: John Benjamins, 99-126.

Győri, Gábor 1996, Historical aspects of categorization. In: Eugene H. Casad (ed.), Cognitive linguistics in the Redwoods. The expansion of a new paradigm in linguistics. Berlin: Mouton de Gruyter, 175-206.

Győri, Gábor 1997, Cognitive archaeology: A look at evolution outside and inside language. In: Roger Blench and Matthew Spriggs (eds.), Archaeology and language. I. Theoretical and methodological orientation. London and New York: Routledge, 43-52.

Győri, Gábor 1999, Evolutionary continuity and the rate and curve of language evolution. Evolution of Communication 3: 207-226.

Győri, Gábor 2001, Symbolic cognition: Its evolution and adaptive impact. In: Gábor Győri (ed.), Language evolution. Biological, linguistic and philosophical perspectives. Frankfurt am Main: Peter Lang, 113-129.

Győri, Gábor, Irén Hegedűs and Attila Dobó 1998, From plural to abstract: A cognitive-naturalist approach to diachronic morphology. In: József Andor, Béla Hollósy, Tibor Lackó and Péter Pelyvás (eds.), The diversity of linguistic description. Studies in honour of Béla Korponay. Debrecen: KLTE, 69-88.

Haiman, John 1983, Iconic and economic motivation. Language 59: 781-819.

Haiman, John 1980, Dictionaries and encyclopedias. Lingua 50: 329-357.

Hamilton, C. 1990, Hemispheric specialization in monkeys. In: C.B.

References 235

Trevarthen (ed.), Brain circuit and functions of the mind. Festschrift for R.W. Sperry. Cambridge, MA: Cambridge University Press, 181-195.

Hamilton, C.R. and B.A. Vermeire 1991, Functional lateralization in monkeys. In: F.L. Kitterle (ed.), Cerebral laterality. Theory and research. Hillsdale, NJ: Lawrence Erlbaum, 19-34.

Hampton, James 1997, Conceptual combination. In: Koen Lamberts and David Shanks (eds.), Knowledge, concepts and categories. Cambridge, MA: The MIT Press, 133-159.

Harnad, Stevan 1987, Category induction and representation. In: Stevan Harnad (ed.), Categorical perception. The groundwork of cognition. New York: Cambridge University Press, 535-565.

Harnad, Stevan 1990, The symbol grounding problem. Physica D 42: 335-346.

Haser, Veronica 2000, Metaphor in semantic change. In: Antonio Barcelona (ed.), Metaphor and metonymy at the crossroads. A cognitive perspective. Berlin and New York: Mouton de Gruyter, 171-194.

Hays, Paul R. 2000, From the Jurassic dark: Linguistic relativity as evolutionary necessity. In: Martin Pütz and Marjolijn Verspoor (eds.), Explorations in Linguistic Relativity. Amsterdam and Philadelphia: John Benjamins, 159-172.

Heilbroner, P.L. and Ralph L. Holloway 1988, Anatomical brain asymmetries in new world and old world monkeys: Stages of temporal lobe development in primate evolution. American Journal of Physical Anthropology 76: 39-48.

Heine, Bernd, Ulrike Claudi and Friederike Hünnemeyer 1991, From cognition to grammar - Evidence from African languages. In: Elizabeth C. Traugott and Bernd Heine (eds.), Approaches to grammaticalization. Amsterdam and Philadelphia: John Benjamins, 149-187.

Heit, Evan 1997, Knowledge and concept learning. In: Koen Lamberts and David Shanks (eds.), Knowledge, concepts and categories. Cambridge, MA: The MIT Press, 7-41.

Herrnstein, Robert J. 1984, Objects, categories, and discriminative stimuli. In: Herbert C. Roitblat, Thomas G. Bever and Herbert S. Terrace (eds.), Animal cognition. Hillsdale, NJ: Erlbaum, 233-261.

Herrnstein, Robert J. 1985, Riddles of natural categorization. In:

236 References

Lawrence Weiskrantz (ed.), Animal intelligence. Oxford: Clarendon Press, 129-144.

Hewes, Gordon W. 1973, An explicit formulation of the relationship between tool-using, tool-making, and the emergence of language. Visible Language 7: 101-127.

Hill, Jane H. 1978, Apes and language. Annual Review of Anthropology 7: 89-112.

Hock, Hans Heinrich 1986, Principles of historical linguistics. Berlin and New York: Mouton de Gruyter.

Hock, Hans Heinrich 1991, Principles of historical linguistics. Berlin and New York: Mouton de Gruyter.

Hock, Hans Heinrich and Brian D. Joseph 1996, Language history, language change, and language relationship. An introduction to historical and comparative linguistics. Berlin and New York: Mouton de Gruyter.

Hockett, Charles F. 1973, Man’s place in nature. New York: McGraw-Hill.

Hockett, Charles F. and Stuart A. Altman 1968, A note on design features. In: Thomas A. Sebeok (ed.), Animal communication. Bloomington, IN: Indiana University Press, 61-72.

Holloway, Ralph L. and M.C. de la Coste-lareymondie 1982, Brain endocast asymmetry in pongoids and hominoids: Some preliminary findings on the paleontology of cerebral dominance. American Journal of Physical Anthropology 58: 101-110.

Holyoak, Keith J. 1984, Analogical thinking and human intelligence. In: Robert J. Sternberg (ed.), Advances in the psychology of human intelligence. Vol. 2. Hillsdale, NJ: Lawrence Erlbaum Associates, 199-230.

Holyoak, Keith J. and Paul Thagard 1997, The analogical mind. American Psychologist 52: 35-44.

Hopper, Paul J. 1990, Where do words come from? In: William Croft, Keith Denning and Suzanne Kemmer (eds.), Studies in typology and diachrony. Amsterdam and Philadelphia: John Benjamins, 151-160.

Isaac, Glyn L. 1976, Stages of cultural elaboration in the pleistocene: possible archeological indicators of the development of language capabilities. In: Stevan R. Harnad, Horst D. Steklis and Jane Lancaster (eds.), Origins and evolution of language and speech. (Annals of the New York Academy of Sciences, 280: 275-

References 237

287). Isaac, Glyn L. 1981, Emergence of human behaviour patterns.

Archaeological tests of alternative models of early hominid behaviour: excavation and experiments. Philosophical Transactions of the Royal Society, London B 292: 177-188.

Jackendoff, Ray 1992, Languages of the mind. Essays on mental representation. Cambridge, MA: Cambridge University Press.

Jeffers, Robert J. and Ilse Lehiste 1979, Principles and methods for historical linguistics. Cambridge, MA and London, England: The MIT Press.

Jerison, Harry J. 1988, Evolutionary neurobiology and the origin of language as a cognitive adaptation. In: Marge E. Landsberg (ed.), The genesis of language. A different judgement of evidence. Berlin: Mouton de Gruyter, 3-9.

Job, D.M. 1982, Semantic change and etymologies. In: Anders Ahlqvist (ed.), Papers from the 5th International Conference on Historical Linguistics. Amsterdam and Philadelphia: John Benjamins, 163-170.

Johnson, Mark 1987, The body in the mind. The bodily basis of meaning, reason and imagination. Chicago and London: The University of Chicago Press.

Johnson, Michael G. and Tracy B. Henley 1992, Finding meaning in random analogies. Metaphor and Symbolic Activity 7: 55-75.

Joseph, Brian D. 1992, Diachronic explanation: Putting speakers back into the picture. In: G.W. Davis and G.K. Iverson (eds.), Explanation in historical linguistics. Amsterdam and Philadelphia: John Benjamins, 123-144.

Joseph, Brian D. and Cathrine S. Karnitis. 1999, Evaluating semantic shifts: The case of Indo-European *(s)meuk- and Indo-Iranian *muc-. OSU Working Papers in Linguistics 52: 151-158.

Jouffroy, Francoise K. 1993, Primate hands and the human hand: the tool of tools. In: A. Berthelet and J. Chavaillon (eds.), The use of tools by human and non-human primates. Oxford: Clarendon Press, 6-33.

Kastovsky, Dieter 1988, Structural semantics or prototype semantics? The evidence of word formation. In: Werner Hüllen and Rainer Schulze (eds.), Understanding the lexicon. Meaning, sense and world knowledge in lexical semantics. (Linguistische Arbeiten 210: 190-203). Tübingen: Max Niemeyer.

238 References

Katz, Jerrold J. 1976, A hypothesis about the uniqueness of human language. In: Stevan R. Harnad, Horst D. Steklis and Jane Lancester (eds.), Origins and evolution of language and speech. (Annals of the New York Academy of Sciences, 280: 33-41).

Katz, Jerrold J. 1990, The metaphysics of meaning. Cambridge, MA: The MIT Press.

Keller, Rudi 1985, Towards a theory of linguistic change. In: Thomas T. Ballmer (ed.), Linguistic dynamics: Discourses, procedures and evolution. Berlin: Walter de Gruyter, 211-237.

Keller, Rudi 1989, Invisible-hand theory and language evolution. Lingua 77: 113-127.

Keller, Rudi 1994, On language change. The invisible hand in language. London and New York: Routledge.

Key, Mary R. 1979, Opposites - once again. Interfaces. Linguistics and Psychoanalysis Newsletter. (University of Delaware, Newark, Delaware) No. 11: 2-5.

Key, Mary R. 1985, An approach to semantics through comparative linguistics. In: U. Piper and G. Stickel (eds.), Studia linguistica diachronica et synchronica. Berlin: Mouton de Gruyter, 467-475.

Key, Mary R. 1988, Semantic space as seen in comparative linguistics. In: The Linguistic Society of Korea (ed.), Linguistics in the morning calm, 2. Hanshin Publishing Co.: Seoul, 369-373.

Key, Mary R. 1990, An example of cognitive-semantic processes in language change. In: B.G. Gurón and P. Levy (eds.), Memorial volume for Jorge Suárez. México, D.F.: El Colegio de México, 315-322.

Kimura, M. 1983, The neutral theory of molecular evolution. Cambridge: Cambridge University Press.

King, M.C. and A.C. Wilson 1975, Evolution at two levels in humans and chimpanzees. Science 188: 107-116.

Kluge, Friedrich 1975, Etymologisches Wörterbuch der deutschen Sprache. Berlin: Walter de Gruyter, (21st edition).

König, Ekkehard and Elizabeth C. Traugott 1988, Pragmatic strengthening and semantic change: The conventionalizing of conversational implicature. In: Werner Hüllen and Rainer Schulze (eds.), Understanding the lexicon. Meaning, sense and world knowledge in lexical semantics. Tübingen: Max Niemeyer, 110-124.

König, Ekkehard and Peter Siemund 1999, Intensifiers as targets and

References 239

sources of semantic change. In: Andreas Blank and Peter Koch (eds.), Historical semantics and cognition. Berlin and New York: Mouton de Gruyter, 237-257.

Kövecses, Zoltán 1991, Happiness. A definitional effort. Metaphor and Symbolic Activity 6/1: 29-46.

Kövecses, Zoltán 1995a, The “container” metaphor of anger in English, Chinese, Japanese and Hungarian. In: Zdravko Radman (ed.), From a metaphorical point of view: A multidisciplinary approach to the cognitive content of metaphor. Berlin: Walter de Gruyter, 117-145.

Kövecses, Zoltán 1995b, Metaphor and the folk understanding of anger. In: James A. Russell, Jos‚-Miguel Fernandez-Dols, Antony S.R. Manstead and J.C. Wellenkamp (eds.), Everyday conceptions of emotion. An introduction to the psychology, anthropology and linguistics of emotion. Dordrecht: Kluwer Academic Publishers, 49-71.

Kövecses, Zoltán 1998, Are there any emotion metaphors? In: Angeliki Athanasiadou and Elzbieta Tabakowska (eds.), Speaking of emotions. Conceptualisation and expression. (Cognitive Linguistics Research 10) Berlin and New York: Mouton de Gruyter, 127-151.

Kövecses, Zoltán 2000, The scope of metaphor. In: Antonio Barcelona (ed.), Metaphor and metonymy at the crossroads. A cognitive perspective. Berlin and New York: Mouton de Gruyter, 79-92.

Kövecses, Zoltán and Günter Radden 1998, Metonymy: Developing a cognitive linguistic view. Cognitive Linguistics 9: 37-77.

Lakoff, George 1987, Women, fire and dangerous things. What categories reveal about the mind. Chicago and London: The University of Chicago Press.

Lakoff, George 1990, The Invariance Hypothesis. Is abstract reason based on image-schemas? Cognitive Linguistics 1: 39-74.

Lakoff, George and Mark Johnson 1980, Metaphors we live by. Chicago and London: The University of Chicago Press.

Langacker, Ronald W. 1987, Foundations of cognitive grammar. Volume 1: Theoretical prerequisites. Stanford, CA: Stanford University Press.

Langacker, Ronald W. 1990, Concept, image, and symbol. The cognitive basis of grammar. Berlin and New York: Mouton de

240 References

Gruyter. Langacker, Ronald W. 1991, Foundations of cognitive grammar.

Volume 2: Descriptive applications. Stanford, CA: Stanford University Press.

Langacker, Ronald W. 1999, Losing control: grammaticization, subjectification, and transparency. In: Andreas Blank and Peter Koch (eds.), Historical semantics and cognition. Berlin and New York: Mouton de Gruyter, 237-257.

Lee, Christopher J. 1990, Some hypotheses concerning the evolution of polysemous words. Journal of Psycholinguistic Research 19: 211-219.

Lehmann, Winfred P. 1993, Theoretical bases of Indo-European linguistics. London and New York: Routledge.

Lenneberg, Eric 1967, Biological foundations of language. New York: Wiley and Sons.

Lenneberg, Eric 1976, Problems in the comparative study of language. In: R.B. Masterson, W. Hodos and H.J. Jerison (eds.), Evolution, brain and behavior. Persistent problems. Hillsdale, NJ: Lawrence Erlbaum, 199-213.

Lichtenberk, Frantisek 1991, Semantic change and heterosemy in grammaticalization. Language 67: 475-509.

Lieberman, Philip 1984, The biology and evolution of language. Cambridge, MA: Harvard University Press.

Lipka, Leonhart 1990, Metaphor and metonymy as productive processes on the level of the lexicon. In: Werner Bahner, Joachim Schildt and Dieter Viehweger (eds.), Proceedings of the XIVth International Congress of Linguists. Vol. II. Berlin: Akademie Verlag, 1207-1210.

Lorenz, Konrad 1978, Vergleichende Verhaltensforschung. Grundlagen der Ethologie. Wien: Springer Verlag.

Lucy, John A. 1997, Linguistic relativity. Annual Reviews in Anthropology 26: 291-312.

MacNeilage, Peter F. 1993, Implications of primate functional asymmetries for the evolution of cerebral hemispheric specializations. In: Jeanette P. Ward and William D. Hopkins (eds.), Primate laterality. Current behavioral evidence of primate asymmetries. New York, Berlin, etc.: Springer-Verlag, 319-341.

Mann, Stuart E. 1984-1987, An Indo-European comparative dictionary. Hamburg: Helmut Buske Verlag.

References 241

Marler, Peter 1975, On the origin of speech from animal sounds. In: James F. Kavanagh and James E. Cutting (eds.), The role of speech in language. Cambridge, MA: The MIT Press, 11-37.

Maynard Smith, John and Eörs Szathmáry 1995, The major transitions in evolution. Oxford, New York, etc.: W.H. Freeman.

Mayr, Ernst 1988, Toward a new philosophy of biology. Observations of an evolutionist. Cambridge, MA and London, England: Harvard University Press.

McMahon, April M.S. 1994, Understanding language change. Cambridge: Cambridge University Press.

Medin, Douglas L. and Lawrence W. Barsalou 1987, Categorization processes and categorical perception. In: Stevan Harnad (ed.), Categorical perception. The groundwork of cognition. New York: Cambridge University Press, 455-490.

Medin, Douglas L. and William D. Wattenmaker 1987, Category cohesiveness, theories, and cognitive archeology. In: Ulric Neisser (ed.), Concepts and conceptual development. Ecological and intellectual factors in categorization. Cambridge: Cambridge University Press, 25-62.

Menzel, Emil W. and Marcia K. Johnson 1976, Communication and cognitive organization in humans and other animals. In: Stevan R. Harnad, Horst D. Steklis and Jane Lancaster (eds.), Origins and evolution of language and speech. (Annals of the New York Academy of Sciences, 280: 131-142).

Mettinger, Arthur 1988, Pay Caesar what is due to Caesar ...: Semantic features vindicated. In: Werner Hüllen and Rainer Schulze (eds.), Understanding the lexicon. Meaning, sense and world knowledge in lexical semantics. (Linguistische Arbeiten 210: 148-156). Tübingen: Max Niemeyer.

Milroy, James 1993, On the social origins of language change. In: Charles Jones (ed.), Historical linguistics. Problems and perspectives. London: Longman, 215-236.

Montague, Ashley 1976, Toolmaking, hunting and the origin of language. In: Stevan R. Harnad, Horst D. Steklis and Jane Lancaster (eds.), Origins and evolution of language and speech. (Annals of the New York Academy of Sciences, 280: 266-274).

Morse, A.P. 1976, Speech perception in the human infant and rhesus monkey. In: Stevan R. Harnad, Horst D. Steklis and Jane Lancaster (eds.), Origins and evolution of language and speech. (Annals of

242 References

the New York Academy of Sciences, 280: 694-707). Neisser, Ulric 1976, Cognition and reality. Principles and

implications of cognitive psychology. San Francisco: Freeman. Nerlich, Brigitte 1989, Elements for an integral theory of language

change. Journal of Literary Semantics 18: 163-186. Nerlich, Brigitte and David D. Clarke 1988, A dynamic model of

semantic change. Journal of Literary Semantics 17: 73-90. Newmeyer, Frederick 1991, Functional explanations in linguistics

and the origins of language. Language and Communication 11: 3-28.

Nitecki, Matthew H. 1990, The plurality of evolutionary innovations. In: Matthew H. Nitecki (ed.), Evolutionary innovations. Chicago and London: The University of Chicago Press, 3-18.

Noble, William and Ian Davidson 1996, Human evolution, language and mind. A psychological and archeological inquiry. Cambridge: Cambridge University Press.

Nottebohm, Fernando 1977, Neural asymmetries in the vocal control of the canary. In: S.R. Harnad, R.W. Doty, L. Goldstein, J. Jaynes and G. Krauthamer (eds.), Lateralization in the nervous system. New York: Academic Press, 23-44.

Oakley, David A. 1985, Cognition and imagery in animals. In: David A. Oakley (ed.), Brain and mind. London: Methuen, 99-131.

Onions, C.T. 1966, The Oxford dictionary of English etymology. Oxford: Clarendon Press.

Palmer, Gary B. 1996, Toward a theory of cultural linguistics. Austin: University of Texas Press.

Palmer, Gary .B. and Claudia Woodman 2000, Ontological classifiers as polycentric categories, as seen in Shona class 3 nouns. In: Martin Pütz and Marjolijn Verspoor (eds.), Explorations in linguistic relativity. Amsterdam and Philadelphia: John Benjamins, 225-249.

Palmer, Leonard R. 1978, Descriptive and comparative linguistics. (2nd edition) London: Faber and Faber.

Passingham, Richard E. 1982, The human primate. Oxford: Freeman.

Paul, Herman 1920, Prinzipien der Sprachgeschichte. (5th edition) Halle: Max Niemeyer.

Pepperberg, Irene M. 1987, Acquisition of the same/different

References 243

concept by an African Grey parrot (Psittacus erithacus): Learning with respect to categories of color, shape, and material. Animal Learning and Behavior 15: 423-432.

Piatelli-Palmarini, Massimo 1989, Evolution, selection and cognition: From “learning” to parameter setting in biology and the study of language. Cognition 31: 1-44.

Pinker, Steven and Paul Bloom 1990, Natural language and natural selection. Behavioral and Brain Sciences 13: 707-784.

Plotkin, Henry 1994, Darwin machines and the nature of knowledge. Cambridge, MA: Harvard University Press.

Pokorny, Julius 1959-1969, Indogermanisches etymologisches Wörterbuch. Bern: Francke.

Premack, David 1985, “Gavagai!” or the future history of the animal language controversy. Cognition 19: 207-296.

Quiatt, Duane 2001, Language and evolution in the Middle Stone Age. In: Gábor Győri (ed.), Language evolution. Biological, linguistic and philosophical perspectives. Frankfurt am Main: Peter Lang, 25-42.

Quiatt, Duane and Vernon Reynolds 1993, Primate behaviour. Information, social knowledge, and the evolution of culture. Cambridge: Cambridge University Press.

Ragir, Sonia 2001, Toward an understanding of the relationship between bipedal walking, encephalization, and language origins. In: Gábor Győri (ed.), Language evolution. Biological, linguistic and philosophical perspectives. Frankfurt am Main: Peter Lang, 73-99.

Rédei, Károly (ed.) 1986-1991, Uralisches etymologisches Wörterbuch. Band I-III. Budapest: Akadémiai Kiadó.

Rensch, Bernhard 1968, Manipulierfähigkeit und Komplikation von Handlungsketten bei Menschenaffen. In: Bernhard Rensch (Hrsg.), Handgebrauch und Verständigung bei Affen und Frühmenschen. Bern: Huber, 103-126.

Rightmire, G. Philip 1981, Patterns in the evolution of Homo erectus. Paleobiology 7: 241-246.

Rockel, A.J., R.W. Hiorns and T.P.S. Powel 1980, The basic uniformity in the structure of the neocortex. Brain 103: 221-244.

Rosch, Eleonor 1978, Principles of categorization. In: Eleonor Rosch and Barbara B. Lloyd (eds.), Cognition and categorization. Hillsdale, NJ: Lawrence Erlbaum Associates, 27-48.

244 References

Savage-Rumbaugh, E. Sue 1986, Ape language. From conditioned response to symbol. New York: Columbia University Press.

Savage-Rumbaugh, E. Sue, J. Murphy, R.A. Sevcik, K.E. Brakke, S.L. Williams and D.M. Rumbaugh 1993, Language comprehension in ape and child. (Monographs of the Society for Research in Child Development. Serial No. 233. Vol. 58, Nos. 3-4.)

Savage-Rumbaugh, E. Sue, Duane M. Rumbaugh and Sarah Boysen 1980, Do apes use language? American Scientist 68: 49-61.

Savage-Rumbaugh, E. Sue, D.M. Rumbaugh, S.T. Smith and J. Lawson 1980, Reference – the linguistic essential. Science 210: 922-925.

Savage-Rumbaugh, E. Sue, R.A. Sevick, D.M. Rumbaugh and E. Rubert 1985, The capacity of animals to acquire language: Do species differences have anything to say to us? In: Lawrence Weiskrantz (ed.), Animal intelligence. Clarendon Press: Oxford, 177-185.

Sebeok, Thomas A. 1991, In what sense is language a “primary modeling system”? In: Thomas A. Sebeok, A sign is just a sign. Bloomington: Indiana University Press, 49-58.

Seiler, Thomas B. 1985, Sind Begriffe Aggregate von Komponenten oder idiosynkratische Minitheorien? In: Thomas B. Seiler and Wolfgang Wannenmacher (eds.), Begriffs- und Wortbedeutungs-entwicklung. Berlin: Springer, 105-131.

Seyfarth, Robert M. and Dorothy L. Cheney 1992, Meaning and mind in monkeys. Scientific American 267(6): 122-128.

Seyfarth, Robert M., Dorothy L. Cheney and Peter Marler 1980a, Monkey responses to three different alarm calls: Evidence for semantic communication and predator classification. Science 210: 801-803.

Seyfarth, Robert M., Dorothy L. Cheney and Peter Marler 1980b, Vervet monkey alarm calls: Semantic communication in a free ranging primate. Animal Behaviour 28: 1070-1094.

Slater, Peter J.B. 1985, An introduction to ethology. Cambridge: Cambridge University Press.

Smith, Edward E. and Douglas L. Medin 1981, Categories and concepts. Cambridge, MA: Harvard University Press.

Steele, James 1989, Hominid evolution and primate social cognition. Journal of Human Evolution 18: 421-432.

References 245

Steklis, Horst D. 1985, Primate communication, comparative neurology, and the origin of language re-examined. Journal of Human Evolution 14: 157-173.

Sweetser, Eve 1990, From etymology to pragmatics. Metaphorical and cultural aspects of semantic structure. Cambridge: Cambridge University Press.

Szemerényi, Oswald 1972, Comparative linguistics. In: Thomas A. Sebeok (ed.), Current trends in linguistics. Vol. 9/1. The Hague: Mouton, 119-195.

Szemerényi, Oswald 1990, Einführung in die vergleichende Sprachwissenschaft. (4th edition) Darmstadt: Wissenschaftliche Buchgesellschaft.

Szentágothai, János 1978, The neuron network of the cerebral cortex: A functional interpretation. Proceedings of the Royal Society of London, Series B 201: 219-248.

Taylor, John R. 1989, Linguistic categorization. Prototypes in linguistic theory. Oxford: Clarendon Press.

Terrace, Herbert S. 1985, Animal cognition: Thinking without language. In: Lawrence Weiskrantz (ed.), Animal intelligence. Oxford: Clarendon Press, 113-128.

Terrace, Herbert S., L.A. Pettito, R.J. Sanders and T.G. Bever 1979, Can an ape create a sentence? Science 206: 891-902.

Tomasello, Michael 2002, Gondolkodás és kultúra. Budapest: Osiris kiadó. (Hungarian translation of The cultural origins of human cognition. Cambridge, MA and London, England: Harvard University Press. 1999.)

Tomasello, Michael 2003, The key is social cognition. In: Dedre Gentner and Susan Goldin-Meadow (eds.), Language in mind. Advances in the study of language and thought. Cambridge, MA: The MIT Press, 46-57.

Toth, Nicholas and Kathy Schick 1993, Early stone industries and inferences regarding language and cognition. In: Kathleen R. Gibson and Tim Ingold (eds.), Tools, language and cognition in human evolution. Cambridge: Cambridge University Press, 346-362.

Traugott, Elizabeth C. 1985, On regularity in semantic change. Journal of Literary Semantics 14: 155-173.

Traugott, Elizabeth C. 1990, From less to more situated in language: The unidirectionality of semantic change. In: Sylvia Adamson,

246 References

Vivien Law, Nigel Vincent and Susan Wright (eds.), Papers from the 5th International Conference on English Historical Linguistics, Cambridge, 6-9 April, 1987. Amsterdam and Philadelphia: John Benjamins, 497-517.

Traugott, Elizabeth C. 1995, Subjectification in grammaticalisation. In: Dieter Stein and Susan Wright (eds.), Subjectivity and subjectification. Linguistic perspectives. Cambridge: Cambridge University Press, 31-54.

Traugott, Elizabeth C. 1999a, The role of pragmatics in semantic change. In: Jeff Verschueren (ed.), Pragmatics in 1998: Selected papers from the 6

th International Pragmatics Conference. Vol. II.

Antwerp: International Pragmatics Association, 93-102. Traugott, Elizabeth C. 1999b, The rhetoric of counter-expectation in

semantic change: a study in subjectification. In: Andreas Blank and Peter Koch (eds.), Historical semantics and cognition. Berlin and New York: Mouton de Gruyter, 237-257.

Traugott, Elizabeth C. and Ekkehard König 1991, The semantics-pragmatics of grammaticalization revisited. In: Elizabeth C. Traugott and Bernd Heine (eds.), Approaches to grammaticalization. Amsterdam and Philadelphia: John Benjamins, 189-218.

Uexküll, Jakob von 1982, The theory of meaning. Semiotica, Vol. 42. (Special edition, ed. by Thure von Uexküll)

Újhelyi, Mária 1996, Is there any intermediate stage between animal communication and language? Journal of Theoretical Biology 180: 71-76.

Újhelyi, Mária and Pál Buk 2001, On the origins of syntactically structured vocalizations in primates. In: Gábor Győri (ed.), Language evolution. Biological, linguistic and philosophical perspectives. Frankfurt am Main: Peter Lang, 101-111.

Varela, Francisco J., Evan Thompson and Eleanor Rosch 1993, The embodied mind. Cognitive science and human experience. Cambridge, MA: The MIT Press. (First paperback edition.)

Vasmer, Max 1986, Etimologicheskij slovar’ russkogo jazyka. Moskva: Progress. [1950-1958] [Translated from: Russisches etymologisches Wörterbuch. Heidelberg.]

Vennemann, Theo 1993, Language change as language improvement. In: Charles Jones (ed.), Historical linguistics. Problems and perspectives. London: Longman, 319-344.

References 247

Vygotsky, Lev S. 1962, Thought and language. Cambridge, MA: The MIT Press. (original publication: Moscow, 1934)

Wahrig, Gerhard 1978, Deutsches Wörterbuch. Gütersloh: Bertelsmann Lexicon-Verlag.

Walker, S.F. 1980, Lateralization of function in the vertebrate brain. British Journal of Psychology 71: 329-367.

Warren, Beatrice 1999, Laws of thought, knowledge and lexical change. In: Andreas Blank and Peter Koch (eds.), Historical semantics and cognition. Berlin and New York: Mouton de Gruyter, 215-234.

Watkins, Calvert 1985, The American Heritage dictionary of Indo-European roots. Boston: Houghton Mifflin.

Wierzbicka, Anna 1984, Apples are not a ‘kind of fruit’ - the semantics of human categorization. American Ethnologist II: 313-328.

Wierzbicka, Anna 1995, Everyday conceptions of emotion: A semantic perspective. In: James A. Russell, Jos‚-Miguel Fernandez-Dols, Antony S.R. Manstead and J.C. Wellenkamp (eds.), Everyday conceptions of emotion. An introduction to the psychology, anthropology and linguistics of emotion. Dordrecht: Kluwer Academic Publishers, 17-47.

Wiley, Haven 1983, The evolution of communication: Information and manipulation. In: Tim Halliday and P.J.B. Slater (eds.), Communication. Oxford: Blackwell, 156-189.

Wilkins, David P. 1996, Natural tendencies of semantic change and the search for cognates. In: Mark Durie and Malcolm Ross (eds.), The comparative method reviewed. Regularity and irregularity in language change. New York and Oxford: Oxford University Press, 264-304.

Wind, Jan 1988, Language evolution and paedomorphosis. In: Marge E. Landsberg (ed.), The genesis of language: A different judgement of evidence. Berlin: Mouton de Gruyter, 137-144.

Winters, Margaret E. 1989, Diachronic prototype theory: On the evolution of the French subjunctive. Linguistics 27(4): 703-730.

Winters, Margaret E. 1992, Diachrony within synchrony: The challenge of cognitive grammar. In: Martin Pütz (ed.), Thirty years of linguistic evolution. Studies in honour of René Dirven on the occasion of his sixtieth birthday. Amsterdam and Philadelphia: John Benjamins, 503-512.

248 References

Wolpoff, Milford H. 1984, Evolution in Homo erectus: the question of stasis. Paleobiology 10: 389-406.

York, Michael 1993, Toward a Proto-Indo-European vocabulary of the sacred. Word 44: 235-254.

Zgusta, Ladislav 1990, Onomasiological change: Sachen-change reflected by Wörter. In: Edgar C. Polomé (ed.), Research guide on language change. Berlin and New York: Mouton de Gruyter, 389-398.

Zimmerman, Barry J. 1979, Concepts and classification. In: Grover J. Whitehurst and Barry J. Zimmerman (eds.), The functions of language and cognition. New York: Academic Press, 57-81.

Index Abstract domain, 181, 209,

211, 216 adaptation, 13-18, 23-24, 29-

31, 40, 45-46, 53-54, 59, 63, 65-73, 78, 80-81, 84-85, 88, 104, 109-112, 121, 123-124, 126, 159, 168-169, 195, 219-226

adaptationism, 14, 16, 45 adaptedness, 56, 60 adaptive, 14, 16, 18, 23, 26-

31, 45, 50, 53-54, 64-65, 68-69, 73-74, 83, 85, 88, 90, 99-112, 119, 121-122, 124-127, 129, 148, 165, 170-171, 219, 221-227

adaptiveness, 16, 23, 69, 73-74

analogical thinking, 29, 100, 111, 118, 120, 127-128, 131, 133, 144-145, 152, 155, 159, 166, 178, 181, 222

animal communication, 49-51, 75-76, 88-92, 220

apes, 19, 47, 52 ape language projects, 76-77,

88, 92-95, 97-103 archilexeme, 214, 217 archisememe, 217 associative labeling, 93-95,

100, 103 Australopithecus, 54, 61

Basic domain, 209-211, 215-216

basic level category, 95, 101-103, 117, 181

behavior, 13-14, 18, 23, 25, 27-28, 30, 35-36, 47-48, 52, 54, 56-57, 62, 64, 66, 69, 73, 81-82, 84-85, 87-92, 95-96, 110-112, 114, 117, 119-120, 148, 153, 158, 170, 181, 183-190, 193, 220, 226; communicative, 13, 20, 22, 30, 49, 69, 87-90, 158, 170; cooperative 69; linguistic, 25, 27-28, 35-36, 48, 87-88, 129, 132, 141-142, 225-226; naming, 153-154; social, 49, 71, 220

Categorical perception, 197 categorization, 25, 87, 95-96,

99-103, 111, 117, 120, 123-124, 126-127, 140-141, 144-147, 150, 152, 155, 159, 165, 169, 179, 196-197, 222-223

chimpanzee, 44, 52, 54, 98-99, 111

coevolution, 19, 42 cognition, 13-14, 18-19, 21,

23, 27, 30-31, 50, 52-53, 71, 78-88, 95, 100, 102, 109, 110-117, 119, 121-

250 Index

124, 126, 130, 133, 140, 142, 144-146, 148, 150, 154, 160, 165-166, 178, 181, 195-197, 206, 214, 219-220, 222-223, 226-227; function of, 23, 81, 110, 112-113, 121-123, 126, 222, 224, 226; neural, 85, 98, 112, 114, 116, 219; social, 14, 23, 31, 50, 112-113; symbolic, 78, 83, 85-87, 112, 114-117, 121-122, 219-222

cognitive, function of language, 23, 28, 87, 109, 121, 123, 125, 129, 143, 147, 169, 197, 224, 226-227; functioning, 23, 25, 29, 84, 109, 112, 122, 127, 220, 223; motivation, 133-135, 160; structure, 75-76, 86, 104, 116, 121, 127, 143, 197, 220

complexity, in evolution, 18-19, 22-23, 43, 49, 62-63, 68, 71, 85-86, 110, 112; in language structure, 60, 76, 113, 115-116, 122-123, 210

concept, 76-77, 93-97, 100-101, 104-105, 116, 118, 144, 146, 158-159, 164-165, 197, 205

conceptual blend, 118-120, 140, 155; category, 25, 98, 100-101, 116, 124-126, 146, 150, 222-223; combination, 116,

118-119; mapping, 29, 120, 155, 162, 170, 180, 201; structure, 125, 175, 197; universal, 31

conceptualization, 31, 95, 119, 125, 127, 132, 147-148, 160, 178-180, 182-183, 187, 191-196, 198, 202, 209, 212-213, 216, 224

construal, 111, 120, 123, 148, 150, 224

context, speech, 25-27, 120, 131-132, 143, 147, 150, 156-157, 159, 162, 170, 178-180

continuity in evolution, 15-21, 30, 35, 38, 40, 44, 47-49, 51-55, 57-59, 61, 64-65, 71, 83-84, 87, 219-220

conventionalization, 13, 26, 114, 122, 131-132, 146, 148, 157-159, 162, 169-170, 175-176, 197, 225-226

cue-validity, 151-153, 155, 180, 195

culture, 13-14, 23, 25-28, 102, 112, 115, 117, 122-123, 125, 127, 131, 134, 137, 140, 142, 152, 159-160, 169-171, 175-177, 179, 182, 193-196, 201, 219, 223-225

culture-dependence, 177, 193-195

Diachrony, 25-26, 135, 158,

161-162, 166, 170, 182, 191-192, 194, 204-206,

Index 251

209, 226 Embodiment, 65, 79, 101,

120, 148, 181, 224 emotion, 141, 182-194 encyclopedic

knowledge, 101, 118, 134, 140, 159, 175, 196; view of meaning, 200

environment, 13-14, 23-24, 27, 30-31, 56-57, 59-60, 65, 68-70, 72, 74, 77, 79, 81, 83-85, 88, 90-92, 94-95, 99, 104, 109, 110-112, 114-115, 119-126, 129, 141, 148, 159, 167, 169, 176-178, 181-182, 188, 197, 202, 206, 210, 219-220, 222-223, 225-227

essentialism, 59 etymology, 133, 135, 148,

150, 153, 162, 167-168, 182, 191, 193, 196, 203-204

exaptation, 15, 17, 71 Figurative language, 128,

133, 138, 146 functionalism, 16, 20, 24, 73 General cognitive

mechanism, 17, 24, 26, 45, 130-131

genetic, 57-58, 61-62, 67, 72-73, 82, 116, 167, 177, 224; memory, 84-85, 98, 110, 114

genotype, 57, 72 gestalt, 95-96, 117, 181-182,

211-212, 216-217 gradualism in evolution, 15,

17-18, 22, 30, 43, 49, 51-52, 54, 58, 65, 68, 71

grammar, 15, 60, 67-68, 74-75, 81, 97, 156-158, 200, 225

grammaticalization, 139, 142, 146-147, 163, 171

grounding, 99, 101, 114-118, 120, 181, 216, 221-222

Hominid, 13-14, 18-19, 21,

23, 30, 38, 42, 47, 52, 54-55, 61, 63, 65-66, 70-71, 80, 220

Homo erectus, 54, 61 Homo habilis, 54, 61 Homo sapiens, 22, 30, 40-42,

44, 54, 61, 64 homonymy, 139, 142, 166,

175, 208 human evolution, 18-19, 30,

37, 61-62, 64, 66, 68, 221 Iconicity, 128 idealized cognitive model,

128, 135, 187, 194 image schema, 31, 120, 124,

128, 135, 181-182, 199, 201-203, 206, 210-217

inferencing, 143, 147 Invited Inferencing Theory of

Semantic Change, 131 innateness, 15, 67-68, 75, 94,

98 innovation,

evolutionary 18-19, 47, 49, 54, 63, 68-69, 85, 87, 113-

252 Index

114, 220; in language use, 26-29, 31, 131-135, 142, 144-145, 152, 156, 158, 162, 168-169, 178, 196, 224-226

intelligibility of innovative language use, 127-128, 142-143, 145, 178, 180, 196

intentionality, 144 invisible hand theory, 135 Knowledge structure, 17, 23,

83, 134, 145, 224 Language,

emergence of, 15, 18-19, 22, 29-30, 37, 43, 47, 58-59, 61-64, 66-67, 69, 71, 73, 75, 78-79, 87; evolution of, 14-15, 17-18, 29-30, 40-41,44, 47, 49, 52-53, 57-60, 65-66, 68-69, 71, 74; double nature of, 18, 35-36, 42; faculty, 18, 45, 68, 72, 75, 82; family, 39, 41-42, 136, 192-193, 205-206; full-blown, 21, 40-41, 72; function of, 13-14, 18, 23-24, 27-30, 49, 75, 87, 94, 100, 103, 109, 121, 123, 125, 129, 143, 147, 169, 197, 219, 224, 226; structure of, 15-16, 31, 59-60, 65-67, 74, 127, 170; use, 25-29, 31, 36, 80, 124-125, 128-129, 131-133, 135, 141-142, 145, 149, 153, 156-159, 161-162,

166, 168-170, 173, 178, 196, 224-226

language-like behavior, 35; system, 60

lexicalization, 127, 140, 145-146, 148, 150, 162, 179-180, 182, 191, 217

linguistic relativity, 31, 126, 172, 195-197, 223

Macroevolution, 58, 61 macromutation, 61-63 mental space, 118-120 mentalism, 80, 82, 134 metaphor, 27, 31, 95, 97, 100,

104, 116, 118, 120, 128-129, 131-135, 137-138, 140, 144, 146, 150, 153-155, 161-162, 164, 170, 173, 176, 178, 180-181, 183, 186, 192, 194, 196, 201, 203, 206, 222

metaphorical extension, 95, 97, 116, 134

metaphysical truth, 122, 227 metonymy, 27, 31, 128-129,

131-133, 135, 137-138, 140, 144, 149-150, 153-154, 162, 170, 173, 176, 178-180, 183, 186, 189, 191, 196, 201, 206, 215

microevolution, 57, 61 model,

internal, 13, 23, 110-112, 222; cognitive, 23-24, 75, 77-79, 81-82, 88, 102-105, 114-115, 119, 121-122, 159, 223, 227; mental, 30, 50, 78, 103, 110, 113-115,

Index 253

121, 141-143, 176, 178, 196; social, 125, 159, 226-227; super-, 51, 78, 104; symbolic, 169, 197

monogenesis of languages, 42 mutation, 17, 53, 58, 70, 72;

macro-, 61-63 Naming, 77, 93-94, 104, 146-

147, 151, 153, 158-159, 200

natural selection, 14-18, 30, 45-46, 53, 58-59, 62-64, 66-75

neuronal interconnectivity, 82 neutral evolution, 15, 71 Objectivism, 117, 130, 134;

non-, 29, 130 onomasiology, 147, 158-159,

182 Paleolinguistics, 41 paleospecies, 54, 61 perception, 48, 50, 62, 80, 82,

90, 92, 95-96, 98-102, 110, 114, 117, 126, 128, 144, 148, 154, 156, 179, 194, 197, 201, 211, 213-214, 224

phenotype, 57-58, 61-62, 64, 70, 72

polysemy, 139, 158-167, 205, 209, 214

preadaptation, 15, 53, 63-64, 70-72, 220

preconceptual, 210, 212-214, 216

prehistory of language , 29,

35, 42-43 prelinguistic, 30, 37-38, 98,

105 primate, 18-22, 30, 38, 47, 51-

54, 56, 58-59, 71, 77, 82, 89, 93, 99-100, 219-220

protolanguage, 39, 41-43, 213-214

prototypicality, 101, 128, 130, 159, 163-165, 200

punctuated equilibrium 17, 61, 70-71, 83

punctuated evolution, 17, 30, 49, 61, 64, 70-71, 83

punctuationism, 15, 17, 61 Qualitative difference of

evolutionary traits, 21, 44, 49, 50-51, 54, 56-59, 82-85, 87, 219

quantitative difference of evolutionary traits, 39, 49, 51, 57-58, 82, 98

Rationalism, 15, 20, 45, 67,

80-81, 109, 115-116, 134, 221

recombination, of genes, 58, 73; of semantic features, 200

reconstruction, linguistic, 19, 22, 38-43; semantic, 183, 199-200, 203-205, 215-217

recursive rule, 50 reference, 30, 53, 87-92, 97-

98, 100, 103-104, 125, 128, 130, 134, 141, 144, 146-149, 151-154, 159, 165,

254 Index

170, 179, 196 relativity of cognition, 111 representation,

categorical, 99-101, 103, 105, 113, 116-117, 181; iconic, 99-101, 105, 113, 116-117, 181; mental, 48, 50, 87, 94-95, 113, 118, 166, 175, 181, 221-222; symbolic, 50, 94-96, 99-102, 104-105, 113, 117-118, 181, 219

representational naming, 93-94

Saltation in evolution, 62 sanctioning in language, 122,

145, 156, 158, 225 selection pressure, 14, 30, 66,

69-71, 79, 220 selectionism, 18, 45, 74 selective advantage, 45, 59,

61, 68-70, 72, 74-75, 78, 221

semantic change, 14, 24-29, 31, 127, 129-143, 145-146, 148-151, 154, 157-164, 166-168, 170-181, 193, 195-196, 199-203, 206, 208-209, 223-224, 226; generalizability of, 137-138, 172, 174, 176; pragmatic factors in, 26, 131-132, 146, 156, 162; regularity in, 137-138, 172, 174, 176, 199; unidirectionality of, 176, 180; universals of, 31, 171-182, 195, 199, 201-203

semantic extension, 25-26, 28-29, 31, 139-140, 176, 179-180, 182, 200, 202-206, 216, 224-225

semantic knowledge base, 144, 157

semantic structure, 25, 27, 76, 123, 127-128, 131-132, 134, 137, 159-160, 163, 166, 169, 174-175, 178, 197, 209, 212-223, 226

semasiology, 158-159, 164 spandrel, 16, 63, 70-72 speciation, 21, 61, 83 speech community, 13, 23, 25,

78, 104, 114, 121-122, 124-125, 131-132, 141, 143, 159, 169, 175, 177, 182, 196, 220, 223, 225-226

structuralism, 16, 73-74 symbol manipulation, 48, 79,

81-82, 105, 113, 115, 117, 221

symbolic structure, 78, 85-86, 114, 116, 121

Teleological explanation

of evolutionary change, 60, 72, 74-75; of semantic change 142, 171

Universal grammar, 15, 67-

68, 75 Utterance Selection,

Theory of, 170

Variability, 58, 72, 130, 170, 225-226

Other publications by Lingua Franca Csoport: Learning lessons: Innovations in teacher education and assessment. Edited by Marianne Nikolov and József Horváth. 1996. A nyelvtanulás folyamata és mérése [The process and assessment of language learning]. Edited by Péterné Deák and Györgyi Máté. 1997. Advanced writing in English as a foreign language: A corpus-based study of processes and products. By József Horváth. 2001. Színes eszmék nem alszanak: Szépe György 70. születésnapjára [Colorful ideas never sleep: On the 70th birthday of György Szépe]. Edited by József Andor, Tibor Szűcs and István Terts. 2001. A nyelv nevelő szerepe: A XI. Magyar Alkalmazott Nyelvészeti Kongresszus előadásainak válogatott gyűjteménye [The educational role of language: Selected papers from the 11th Hungarian Congress in Applied Linguistics]. Edited by Ágota Fóris, Eszter Kárpáti and Tibor Szűcs. 2002. Studies in English theoretical and applied linguistics. Edited by József Andor, József Horváth and Marianne Nikolov. 2003. Nostratic Centennial Conference: the Pécs Papers. Edited by Irén Hegedűs and Paul Sidwell. 2004. Lingua Franca Csoport on the web: www.xanga.com/lingua_franca

language as cognitive adaptation: from language evolution to language change

Documents