The Origins of Grammar: An Anthropological Perspective

The Origins

of

Grammar

CO22552.indb 1

23/04/2010 14:13

Also available from Continuum Epistemology: Key Concepts in Philosophy, Christopher Norris Science: Key Concepts in Philosophy, Steven French

CO22552.indb 2

23/04/2010 14:13

The Origins

of

Grammar An Anthropological Perspective

Martin Edwardes

CO22552.indb 3

23/04/2010 14:13

Continuum International Publishing Group The Tower Building 80 Maiden Lane 11 York Road Suite 704 London New York SE1 7NX NY 10038 www.continuumbooks.com © Martin Edwardes 2010 All rights reserved. No part of this publication may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopying, recording, or any information storage or retrieval system, without prior permission in writing from the publishers. Martin Edwardes has asserted his right under the Copyright, Designs and Patents Act, 1988, to be identified as Author of this work. British Library Cataloguing-in-Publication Data A catalogue record for this book is available from the British Library. ISBN: 9  78-1-4411-1438-9 (hardcover) 978-1-4411-7098-9 (paperback) Library of Congress Cataloging-in-Publication Data Typeset by Fakenham Photosetting Ltd, Fakenham, Norfolk Printed and bound in India by Replika Press Pvt Ltd

CO22552.indb 4

23/04/2010 14:13

For Philip: my strongest critic, my fiercest defender.

CO22552.indb 5

23/04/2010 14:13

This page intentionally left blank

Contents

 Acknowledgements 1 Why All the Fuss? The Problem of Brainpower Two Legs, Two Hands Making Tools Hunting and Culture Language: the final frontier? The Genetic Problem of Language What Is Language for? Mapping the Journey

2 The Story So Far Language Language Language Language Language Language Language

xi 1 2 3 4 6 8 10 12 13

15

Is Tool Use Is Play Is a Signal of Fitness Is Gestural Is Cognition Is Social Construction Just Is

16 17 18 21 22 24 25

3 The Heavy Hand of Generative Linguistics

27

Linguistic Structure Extending Structure Principles and Parameters Small Is Beautiful Generative Origins Is Generative Grammar an Inimical Environment for Language Origins?

4 Other Views on Language A System of Functions Systemic Functional Grammar Other Views on Functional Grammar Grammar without Tiers? Linear Grammars Functionalism and the Origins of Grammar

CO22552.indb 7

28 29 30 32 33 36

38 38 39 42 44 45 47

vii

23/04/2010 14:13

viii

Contents

5 It’s All in the Mind A Short History of Cognitive Linguistics The Nature of Cognitive Linguistics Embodiment The Modularity Debate The Nature of Cognitive Grammar Cognitive Linguistics and Language Origins

6 Being Human Physical Differences Manual Dexterity Is Social Dexterity? Working Together The Problem of Culture The Costs of Reproduction Beating the Cheats Making Models

7 The Weirdness of Self Planning and Modelling Human Social Models The Self and Language Selfishness and Self-awareness Four Selves Awareness of Self

8 How Did We Come to Be Human? Altruistic Punishment Metaphor in Cognition THE GROUP IS AN ENTITY Where Does the GROUP Come from? Altruistic Punishment as an Engine of Socialization THE GROUP IS AN ENTITY: building social structures THE GROUP IS AN ENTITY: an ancient metaphor? What Happened, and When?

49 49 51 54 55 57 58

61 62 63 64 66 67 69 70

72 73 75 75 76 79 80

82 83 84 85 86 88 89 90 91

9 How Did We Come to Use Grammar?

93

Grammaticalization Grammaticalization and Language Origins Overture and Beginners, Please Not Required at the Origin of Grammar Becoming Complex From Non-grammar to Grammar

94 96 97 100 101 102

CO22552.indb 8

23/04/2010 14:13

ix

Contents

10 What Nonhumans Tell Us about Being Human Animals and Grammar Primate, Know Thyself Multiple Intelligences Accommodating Others Empathy Not about Language?

11 What Young Humans Tell Us about Being Human Children Children Children Children

and and and and

Language Origins Co-operation Selfhood Language

12 What Time Tells Us about Being Human Getting Tense Doing Other Things with Time Adding Depth Time, Uncertainty and Fiction Becoming Time-aware Three Time Points, Three Voices? Time and Being Human

13 The Evolution of Grammar Basic Communication Social Modelling Uttering Language What Language Did Next Becoming Myself Are There Grammar Universals? And Finally . . .

104 105 106 107 108 110 111

115 116 118 121 123

125 126 128 129 130 131 132 133

135 136 137 138 140 141 143 144

 References

147

 Index

169

CO22552.indb 9

23/04/2010 14:13

This page intentionally left blank

CO22552.indb 10

23/04/2010 14:13

Acknowledgements

without the input and help of many people. First, I must thank my PhD supervisors, Chris Knight and Camilla Power, for keeping me on track and gently steering my stranger ideas into more conventional channels. Particularly I would like to thank Tom Dickins, who stepped in as a third supervisor when my anthropology took a distinctly psychological direction; he helped me to realise that anthropology goes on inside the individual as well as around them. Thanks also go to Jim Hurford, my external examiner, who, with Tom, has provided continuing support for my rather odd post-doctoral trajectory. They have provided me with opportunities to meet leading figures in the field, and to improve my ideas into theories. The people at the City of London Academy Southwark have also been very helpful, allowing me the time to remain involved with the Academic world and still take a full role at the school. Particularly I must thank Martyn Coles, the Principal, for his support while preparing this book. Thanks go to Don Barnes, who kindly sent me a paper I couldn’t get hold of by other means, and to the many others who have directed me towards useful articles and books. I would also like to thank the people who kindly read and commented on the drafts for this book, Philip Rescorla and Stephen Colmer. Their suggestions have improved the readability of the text considerably. Any remaining issues with the writing and argumentation of the book are, of course, down to me.

This book would not have been possible



CO22552.indb 11

xi

23/04/2010 14:13

This page intentionally left blank

CO22552.indb 12

23/04/2010 14:13

1 Why All the Fuss? This question has exercised human minds since we first noticed our humanity, and it has often been conflated with the more complex but less fundamental question, what is special about the specialities that make us human? This second question is about perspectives that can be taken on skills that seem to define our species: do we view them from the inside looking out, using them as benchmarks against which to judge other species; or do we see them from the outside looking in, as instances in a range of possible natural specializations? These two, largely exclusive, positions are what makes the second question less fundamental than the first: any answer tells us more about the position taken by the respondent than about the question posed. Both questions need to be addressed, however, if we are to understand both our specialness and our continuity with the rest of nature. This chapter will review the first question, what is so special about being human, in terms of particular physical and cognitive attributes. It will, however, use a particular meaning of “special”: the attributes of humans that are unusual in nature. The word “special”, therefore, is not intended to create the impression that humans are somehow isolated from the rest of nature. In terms of the second question, our human specialities will mainly be viewed from the outside looking in, taking the position that humans are atypical in having language and grammar rather than extraordinary. The intention is not to identify differences that isolate humans from the rest of nature but to show that our species, like every other, has speci-al capacities that can be viewed as species identifiers. In particular, this chapter is concerned with those human attributes that are implicated in language grammar. This would seem to be a very small part of the question of being human; but the intention of this book is to show that grammar is not a self-contained, stand-alone language engine (as some linguists believe), it is emergent from a series of cognitive, social and communicative capacities which have their own, non-linguistic functions. Individually, these capacities may seem unrelated to language or grammar; but collectively they create both a need for communicative complexity and a structure to accommodate that complexity. An important feature of human language grammar is that it seems to be unrepeated in nature; it does, therefore, seem to be a candidate for what

What is so special about being human?



CO22552.indb 1

1

23/04/2010 14:13

2

th e orig in s of g r am mar

differentiates us as a species. As an emergent capacity, however, it cannot be the whole story: the capacities from which it emerges must be more deeply implicated in our humanness than grammar itself. This chapter therefore looks at some of the attributes which have been named as decisive in the evolution of our species, starting with that often-cited but ill-defined quality: brainpower.

The Problem of Brainpower Many animals have brains. Certainly all vertebrates have a nervous system with a primary control node located at one end of a spinal column. All of these brains are involved in doing much the same thing: processing sensations from various parts of the body, producing holistic interpretations of the actual world outside the body, and generating whole-body reactions to those sensations and interpretations. In this approach, brains are viewed as processing devices, converting inputs to outputs through the application of rules. This, however, is a very general description, individual brain types vary enormously in their construction, the sensations they attend to, and the rules they use to generate bodily reactions. The value of the brain, though, is not in its processing capacity but in how well its operations serve the organism of which it is a part. Bigger is not necessarily better in brainpower. Nonetheless, looking at brains from inside our species, what seems to be important is our capacity for intelligence—and the fact that intelligence allows us to interact with our environment in ways that cause our environment to change drastically. We are not the only form of life to constructively interact with its environment; indeed, it can be argued that life itself is just an arrangement of chemicals that can utilize its local environment to make copies of itself. We are also not the only form of life to change our environment to make it more favourable to us. Laland et al. (1999) have called this adjustment of the local environment by a species, “niche construction”; and they have shown that it is a feature of life forms from worms upwards. The effect that human intelligence has on the environment, both locally and globally, is, however, orders of magnitude greater than that of any other species. Indeed, it currently seems to be orders of magnitude greater than all other species aggregated together. This ability not just to react to the environment but to appropriate it in support of our own species is a direct product of our cognitive sophistication; and we therefore see cognitive sophistication, or intelligence, as a marker of relative fitness, both within our own species and between species. Unfortunately, bare intelligence does not really indicate anything useful in a Darwinian sense. The evolutionary significance of something is not in its quantity but in the advantages it gives, and the abstract quality of intelligence by itself seems to be a rather poor gauge of evolutionary success; only when it is socially applied (Gavrilets & Vose, 2006; Nettle & Pollet, 2008) does intelligence become a useful marker of human reproductive success.

CO22552.indb 2

23/04/2010 14:13

Why All the Fuss?

3

The physical indicators of the brain itself are even less reliable. If we measure brain power simply by brain size then elephants have us far outclassed (Shoshani et al., 2006). If we rely on the surface complexity of the brain itself then whales and dolphins are ahead of us (Marino et al., 2007). If we choose brain/body ratio (the size of the brain compared to the size of the body) as Leslie Aiello and Peter Wheeler (1995) propose, then our extinct cousins, the Neanderthals (Homo neanderthalis), put us in the shade (Stringer & Gamble, 1993); nonetheless, they are extinct and we are still here. If we look at the absolute number of neurons in the cortex then humans, at 11.5 billion, come out just ahead of, but not significantly ahead of, whales and elephants (Roth & Dicke, 2005). All of these measures are, however, very rough-and-ready reckoners of brain power, equivalent to counting the value of notes and coins in circulation to calculate the wealth of a nation. It is not the amount of cash available but how it is used that makes wealth. If physical measures of brains do not explain the difference between us and other animals, perhaps it is the “computing power” of our brains that sets us apart. Abstract measures of brainpower, such as quickness and versatility of thought, things that can be measured with IQ tests, do allow us to make a comparison of intelligence between humans. Mostly, though, the tests cannot be applied to nonhumans; they contain elements which are human-centric and of no relevance to other animals. This means that our “IQ ruler” cannot be applied across species, and it cannot therefore be a useful measure of evolutionary intelligence. The value of IQ tests in measuring even human intelligence has been questioned by some cognitive scientists (Gould, 1981). If we turn to individual cognitive mechanisms as measures of intelligence then we are on stronger ground. Identifying these cognitive mechanisms relies, however, on finding out what they do: we have to measure mental difference by describing physical differences. This approach explains differences in species lifestyles, but does not help with our peculiarly human mental world. Nonetheless, there are several physical attributes that we think of as distinctly human, and which do appear to be implicated in cognitive processes that help to define us as a species.

Two Legs, Two Hands Bipedalism was an early candidate for human uniqueness, and it has important implications for cognition. One particular early theory, however, illustrates the difficulty of working from physical difference through mental control structures to evolutionary need. Gordon Hewes (1961) proposed that walking on two legs instead of four freed the hands to do other things, specifically to transport large amounts of food from where it had been found or killed to a home base. This was an attractive idea because it implied high levels of co-operation, and it explained the origins of human manipulation as well as bipedalism: a drive towards

CO22552.indb 3

23/04/2010 14:13

4

th e orig in s of g r am mar

manipulation, motivated by the need to carry things, preceded and opened the way for tool use. However, current evidence of australopithecine scavenging (and, possibly, hunting) techniques, although still scant, indicates that, despite being bipedal, they went to the food and ate it where it was; they didn’t waste calories carrying it around the landscape (Conklin-Brittain et al., 2007). Bipedalism by itself is problematic as the agency which defined us as human, because it is not exclusive to humans; other species use bipedal locomotion because it raises the head and thus increases visual range, or because it is a more energy-efficient mode of travel (Day, 1986). These remain valid reasons why early hominins were able to walk on two legs, and they require no specifically human explanation. Bipedalism is therefore likely to be implicated in the process by which we became human—freeing the hands from the job of walking was a significant event—and it clearly created a niche around which several human cognitive functions seem to cluster; but it is unlikely to be the sole cause of difference between humans and other animals. What about tool use? If bipedalism does not explain our humanness, perhaps one of its products, using our hands to manipulate tools, is what differentiates us from other animals. This idea is certainly plausible, humanity is clearly the most efficient tool-using lifeform on the planet. Tools extend our physical capacities, enhancing strength, reach, manipulation and the senses. Modern humans use tools for a large number of tasks; we even use them where they serve no practical purpose, such as cutlery for eating. For decades, tool use was considered the most likely differentiator between humans and other animals. However, studies of chimpanzees (Pan troglodytes) in the wild have shown that they indulge in various tool-using activities, such as poking sticks into termite mounds and eating the termites that swarm onto the stick (Sanz et al., 2004). Other studies show that they know how to crack hard-to-open nuts between a stone hammer and a stone or wood anvil (Boesch-Achermann & Boesch, 1993)—a skill, incidentally, which is definitely not innate and which requires practice and learning. Two potential differentiators were thus ruled out by this one study: both tool use and in-life learning of tool-related skills are clearly not exclusively human. Bipedalism and tool use have now been identified in a wide range of animals, and they are no longer seen as solely human capacities. Nonetheless, both have been enshrined in our evolutionary taxonomy: the first early human fossil to be discovered, in 1891 at Trinil on the island of Java by Eugene Dubois, is now called Homo erectus, “upright human”; and the most ancient fossil type identified as a member of the Homo genus belongs to the species Homo habilis, “skilful human”. Despite the names, however, both were upright walkers and both used stone tools.

Making Tools If tool use is not our sovereign domain, perhaps it is tool making that distinguishes us from other animals. Picking up a handy rock and hitting a handy nut

CO22552.indb 4

23/04/2010 14:13

Why All the Fuss?

5

on a handy tree root, or picking up a stick and pushing it into a hole, do not require the tools to be made, only used. Unfortunately, once again, nonhumans have disproved the hypothesis. At first, there was evidence only of tool refinement: rather than just pick up any stick, chimpanzees break off suitable twigs and strip off the leaves before using them to fish for termites (Goodall, 1988, pp. 34–36). Evidence of other tool creation began to accumulate, however. Among other examples, chimpanzees were found to make brush-like tools by chewing twigs, which they then used to fish in bee nests for honey (Brewer & McGrew, 1990); chimpanzees have also been observed sharpening thin branches with their teeth and then using the branch as a spear to hunt for bushbabies in tree hollows (Pruetz & Bertolani, 2007); and there seems to be evidence that a beaver cut and moved a willow branch log to use as a prop, enabling it to work at a greater height (Barnes, 2005). Perhaps the most impressive examples of nonhuman tool making have been demonstrated by the New Caledonian crow (Corvus moneduloides). These birds have been observed in various tool-creation activities, both in captivity and in the wild. In the wild they use their beaks to cut pandanus leaves into various types of insect spears, some to deal with grubs in the ground, some to extract them from trees (Hunt & Gray, 2004). In laboratory conditions they have been observed bending metal wires into hooks to lift a container of food out of an otherwiseinaccessible pipe (Weir et al., 2002). The New Caledonian crows show that the capacity to plan and manufacture tools is not exclusively human, nor even exclusively primate. An argument has been made that the manufacture of stone tools is cognitively different from other types of tool manufacture, and it is this that sets us apart from other animals (Davidson & McGrew, 2005); so does lithic technology have implications for other forms of cognition? To make any tool you must have an idea of the shape and size of the finished tool, and you must have a knowledge of how to alter the base object into the final object. This planned conversion from raw material to finished object is a complex cognitive task, involving planning, visualization, intention, and the ability to constantly triangulate from the current state of the unfinished tool to the desired outcome; it is a cognitive task that has clear implications for general cognition. In contrast, the material used to make the tool requires cognitive capacities in the tool maker which are specific to the material, and which are not readily transferable to other cognitive fields. Knowing how to bash rocks together effectively is a remarkable skill, but can it really help in solving other lifestyle problems? Stone tool manufacture does seem to be a uniquely human feature, but it is not clear that this ability leads to a general and sophisticated tool technology involving a wide range of materials, or that it leads to any generalized cognitive capacities. Indeed, the hominid capacity to make stone tools, when viewed in terms of technology used, seems to have remained unchanged for about a million years (Ambrose, 2001)—not what you would expect from a cognitive capacity with useful transferability to other areas.

CO22552.indb 5

23/04/2010 14:13

6

th e orig in s of g r am mar

Nonetheless, there remains something very specifically human about our tool manufacture: human tools exceed those made by other animals in complexity of form, complexity of purpose and complexity of manufacture (Byrne, 2004). We make tools with multiple parts, often out of multiple materials; we make tools that have multiple functions; and we make tools to make other tools, sometimes involving multiple levels of manufacture to create the final product. This capacity to plan a final product by breaking its production down into discrete steps does seem to be peculiarly human; but it also seems to be something we became able to do over quite a considerable period of time (Henshilwood & Marean, 2003). This does not, therefore, look like a single genetic event creating a new primary capacity, it seems more like something emerging slowly from other cognitive capacities. It is possible that tool making emerged from the same group of primary cognitive components as did language, but it is unlikely to have followed the same trajectory: tool making is an essentially solitary activity that relies at each stage on the genius of one brain; language is a social activity that relies on negotiation between brains, and should therefore be developmentally slower. In practice, language seems to innovate frequently and change rapidly, while innovation in tool creation was, until recently, a very slow process. Human tool making is certainly something special in evolutionary terms, but its role as a driver of the development of language and grammar is tenuous.

Hunting and Culture Several other candidates have been proposed as behaviours peculiar to humans, and therefore as possible features that define us as human. Most of them, however, (such as religiosity) require humans to already be in a universe of symbols created by language and human culture: they are products of, rather than causes of, being human. There is one non-symbolic behaviour, however, that does seem to have an individually human signature to it, although it is definitely not an exclusively human activity: the fact that we hunt. It is not hunting itself that sets us apart, many carnivores hunt. It is also not the fact that we hunt in groups: several species rely on pack hunting, and they have different sharing strategies to ensure that co-operation in the hunt is worthwhile for the individual. There is, nonetheless, something unusual in the way humans are able to co-operate in a hunt, and in the way we share the resulting food throughout our community. Today’s human hunter-gatherers seem to have a common separation of specializations based on gender, in that men hunt and women do not (Balme & Bowdler, 2006). There are exceptions to this (Noss & Hewlett, 2001), but it does represent the situation for most of the remaining hunter-gatherer cultures. Why this gender specialization should be so common is not immediately obvious, it is not a separation of roles based on capabilities. The hunter-gatherer exceptions show that women are fully capable of organizing communal hunts to regular schedules; and modern Western women, like

CO22552.indb 6

23/04/2010 14:13

Why All the Fuss?

7

Sarah Palin, can also be avid hunters. Mythology, too, is replete with goddesses who hunt: Astarte, Artemis, Diana and Freya (Indo European); Yama-No-Kami (Japanese); Cihuacoatl (Aztec);and Achimi (Kabyle), to name a few. One feature of human subsistence hunting that does seem to be universal is sharing: the habit of hunting to primarily benefit others rather than the self seems to occur in all human hunter-gatherer groups. Sharing surpluses is not uncommon in nature, it generates a web of reciprocal debts that allows an individual to rely on others when personal foraging has been less successful (Wilkinson, 1984); but giving up the whole of your foraged portion of food every time, as many human hunters do, is rare. This surrendering of food has to be a ritualized act, there must be an expectation that there will be some compensation for the food; but it is expressed both ostensibly and culturally as an act of altruistic modesty. This peculiarly human dislocation of the relationship between hunting and food is an important expression of species difference; and, as we shall see, has significance for language. The fact that the product of hunting is subject to ritualized distribution implies that there is a mechanism to enforce this ritual—a system of moral and coercive processes that can be summed up by the word culture. When we use the word culture, however, we need to be careful about what we mean. One definition would be that it consists of the non-genetic capacities of individuals which are transmissible between individuals, and it therefore relies on mechanisms which allow the sharing of information. This definition of culture can be divided into three key components: the capacity for innovation; the capacity for transmission; and the capacity for personal learning. This, though, is not a definition that limits it to humans: apes (Lycett et al., 2007), macaques (Flack et al., 2006) and other monkeys (de Waal, 2004), as well as dolphins (Rendell & Whitehead (2001), have all been observed producing innovative actions that subsequently propagate through the community, a few individuals at a time. This definition of culture has, however, been questioned: there remains an important difference between the transmission of directly advantageous technical innovations, like nut cracking, and the adoption of belief systems that have currency only in terms of group identity (Premack & Hauser, 2001). Human culture is not just quantitatively different from other cultures, it has a depth and breadth unrepeated in nature. Our culture is not just a group repository for knowledge which has not been encoded at the genetic level; it is not just a mechanism for social compromise which allows us to live in groups; it is not just a mechanism for differentiating group members from outsiders; and it is not just a costly signal of willingness to co-operate. It is, of course, all of those things, but it is also arbitrary. The knowledge we exchange, the compromises we make, the groups we form, our flags of membership, and the costs we pay to signal co-operation, are very often based around what, in Darwinian terms, can be classed as pointless activities. It seems to be solidarity for solidarity’s sake. If we are looking for the sources of language, therefore, the human cultural

CO22552.indb 7

23/04/2010 14:13

8

th e orig in s of g r am mar

environment produced by co-operative hunting should provide a fertile ground for the search. Language is symbolic, allowing arbitrary relationships between sounds and meanings; human culture is arbitrary, allowing symbolic relationships between individuals, such as nationality, to be treated as real relationships. Language permits metaphor, allowing apparently incompatible meanings to be compared and combined to generate arbitrary hybrid meanings; human culture allows very different personal contributions to be compared and evaluated against arbitrary hybrid measures of worth, such as money. Language is segmented, differentiated and hierarchical, allowing specialized components to be combined according to subjective rules into shareable meta-meanings, which can then combine into further meta-meanings; human culture allows individuals operating in specialist roles to work together within subgroups of a social group, creating meaningful and stable economic systems within the group itself. These are important cognitive similarities between language and human culture that point towards a common source. While cognitive complexity is clearly necessary to enable linguistic cognition, and bipedalism leading to tool use and tool manufacture all indicate cognitive complexity, it is hunting that provides the first signpost for exploring the important differences between humans and other animals. Our co-operative hunting and food sharing strategies opened the way for a complex co-operative culture which required a communication system—language—capable of describing the complexities of that culture.

Language: the final frontier? If human culture is quantitatively and qualitatively different from the culture of other animals then we have an important reference point for species difference; and if language is a reflection of the human cultural system then the idiosyncrasies of language, when compared to other forms of signalling, should give us important clues to being human. The multipurpose and flexible communication system that we call language occupies a signalling niche that, on current evidence, other animals do not appear to have exploited. True, other species do have effective signalling systems with unique features; but, when considered holistically as systems, they are unlike human language in significant ways. When we look at the features of language in detail, however, exclusivity begins to fray a little. Several features once believed to be exclusive to language have been identified in other signalling systems. For instance, the ability to refer to non-immediate locations has been demonstrated by honey bees (Apis mellifera) (von Frisch, 1973); the ability to use reduced-cost indexical signals has been demonstrated in chimpanzee play (Flack et al., 2004); diana monkeys (Cercopithecus diana) are able to use segmented signals in which meanings are changed depending on which segments are used (Zuberbühler, 2000); one particular grey parrot (Psittacus erithacus) has been able to successfully

CO22552.indb 8

23/04/2010 14:13

Why All the Fuss?

9

understand both the naming of things and the identification of attributes of things (Pepperberg, 1999); and the capacity to identify novel names for novel items has been demonstrated by a collie dog in Germany (Kaminski et al., 2004). Possession of the necessary capacities to use language appears to be a matter of degree rather than an absolute difference. Yet there remains something very different between what humans can do with language and what animals can do with their signalling systems. Attention has therefore recently turned to the complexity of language, and the device that makes that complexity possible: language grammar. Grammar would appear to represent the pinnacle of humanness. It is itself complex, which means that it greatly increases the complexity of human signalling over other signalling systems; and it appears to be missing, at least in its complex version, from nonhuman signalling in the wild. Grammar provides a schematic which allows objects not just to be named but to be mapped into a cognitive landscape of interrelationships; and it allows these mappings to be shared between minds by converting that landscape into a one-dimensional stream of sound. This cognitive landscape is not actuality itself but a model of the actual world: it allows for the representation of not just actual things and events, but of imagined objects and fictional events; and language, as the map of this cognitive landscape, allows these fabrications to be exchanged between minds. This poses a problem for Darwinian signalling theory. Signals are only valuable to the receiver if they are, in some way, a reliable indicator of a significant event in the actual world; so how can a signalling system which relies on fictional representations become endemic in a species? What advantages accrue to individual members of a species from being able, essentially, to lie to each other, and from being able to find value in the lies of others? In Darwinian evolutionary accounting, there has to be a countervailing advantage for the individual in being part of this world of fiction; the individuaI’s own reproductive fitness must somehow be enhanced by co-operating in what Nietzsche (1976 [1874]) saw as the big lie of language. How could this be? Why are hearers of language willingly deceived, when susceptibility to deception is usually an indicator of evolutionary inefficiency? And how are speakers of language able to generate untrue messages when duplicity in a signalling system should create distrust in the value of the signals, and thus lead to the disruption of that system (Zahavi, 2003)? What makes co-operative deception work in the case of language? We can uncontroversially claim that language is an essential element in human co-operative culture; and if our co-operative culture is itself of sufficient advantage to the individual then that may be all we need to explain how language can be both deceptive and effective. This, however, only pushes the problem into a different sphere without explaining it: we co-operate in the deceptions of language because of our co-operative culture, but what is the fitness advantage an individual gets from being part of a co-operative culture? The problem of co-operation is vital to an understanding of language

CO22552.indb 9

23/04/2010 14:13

10

th e orig in s of g r am mar

grammar and its origins, and it is therefore central to this book. The Darwinian paradox that co-operation creates must, however, be explored using Darwinian tools: individual fitness, reproductive fitness, kin selection, reciprocal altruism and costly signalling. These tools are all extractions from the basic scientific facts of evolution: that all life consists of a series of remarkable molecules which have the ability to copy themselves; and that those molecules all access the same limited pool of raw chemicals from which to build copies of themselves. To understand co-operation as an emergent feature of this highly competitive, selfish-gene universe (Dawkins, 1989), we have to remember that it must have a genetic explanation. Individuals co-operate because they are closely related (Hamilton’s Kin Selection, 1964), because they can expect co-operative behaviour in return (Trivers’ Reciprocal Altruism, 1971) or because co-operation makes them attractive to potential mates (the Zahavis’ Handicap Principle, 1997). Our explanations of the co-operative behaviour behind socialization, culture and language should therefore be consistent with evolutionary fitness.

The Genetic Problem of Language If we consider language as a genetic solution to a fitness need—an effective strategy at the phenotypic level—then we encounter the problem of what we mean by language. Traditionally we see language as a single system for communication or cognition, but it actually incorporates a number of subsystems which, while being part of language, also seem to function outside of language. In the traditional model, language consists of semantics (the meaning behind an utterance, what the utterance is being generated to do) and grammar (the rules the utterance must follow to ensure that the meaning understood by the receiver matches the meaning intended by the sender). Grammar is further divided into morphology (how words are internally constructed), and syntax (how words combine into utterances). Between semantics and grammar there are lexis (the words themselves, which have both morphological and semantic content) and phonology (the sounds and gestures used to represent the words). Leaving phonology to one side for the moment, this model of language does not appear to be a system which could evolve incrementally. A language utterance has to be able to mean something, and that meaning comes out of both the parts that make up the utterance and the way in which they are combined: semantics requires lexis and grammar, lexis requires grammar and semantics, and grammar requires lexis and semantics. How could this Mexican standoff have evolved? In theory, semantics can exist as a cognitive mechanism without needing to be expressed through language; but what would it have evolved to do, and how would it work without access to tags to carry semantic content—words, or a cognitive precursor to words? Similarly, grammar can exist cognitively as a separate engine; but what would its purpose be in evolutionary terms, and how could it work without conceptual

CO22552.indb 10

23/04/2010 14:13

Why All the Fuss?

11

segmentation—words, or their cognitive precursors? Semantics and grammar have no cognitive validity without lexis, implying that any incremental evolution of language must start with lexis, or a discrete, combinatorial form of cognition. What use, however, are discrete units of cognition unless they mean something? And if the units need to be combined, then rules are needed to differentiate useful combinations from useless. Lexis cannot evolve without semantics and grammar. There are two ways out of this paradox. The first is to assume that all of the elements necessary for language—lexis, grammar and semantics—evolved in a single massive genetic change—a macromutation (Chomsky, 2006, pp. 176–184). The advantage of this approach is that it removes the need to explain the paradox incrementally: the problem of which came first is avoided because lexis, grammar and semantics all appeared together. The disadvantage is that successful macromutations seem to be extremely uncommon. The majority of individual genetic mutations that create changes at the phenotypic level are disadvantageous to the phenotype; only the occasional change is fit enough to become established in a population. The chance of a cluster of advantageous mutations all occurring together (a genotypic macromutation) is highly improbable. A second type of macromutation, where a small change at the gene level produces large changes at the individual level (a phenotypic or homeotic macromutation), is more common, but still rare. In phenotypic macromutations the change at the gene level is small, but it has a cascade effect on protein levels which can significantly affect the body form at the level of the phenotype. Phenotypic macromutations that successfully produce large-scale changes tend, however, to be even rarer than successful genotypic events: the process of producing “hopeful monsters’ much more commonly creates unhopeful monsters. Very few cases of successful phenotypic macromutation have been identified, and those are mostly in the plant kingdom (Theissen, 2006). Macromutations do occur: the change from 48 chromosomes in apes to 46 in humans may be an example of a genotypic macromutation; and the appearance of different skin shades in many mammals is classifiable as phenotypic macromutation. However, genotypic macromutations tend to be deep coding changes which have very subtle, or no apparent, effects on the phenotype; and phenotypic macromutations tend to have obvious but cosmetic effects on the phenotype. Neither of these matches what we would need from a macromutation to generate language. We cannot dismiss as impossible that a macromutation caused language; but we should explore all other options fully before considering it as probable. The second way out of the lexis-grammar-semantics paradox is an incremental approach: the road to language started with a little bit of semantic content, a tiny lexis and very simple grammar, and this change met a specific fitness need in a limited area of cognition (Jackendoff, 2002). There is still a single genetic change which simultaneously produced semantics, grammar and lexis, but it is a response to a small and limited need, not to the pervasive and disparate needs that language meets. Once, however, there are rudimentary forms within

CO22552.indb 11

23/04/2010 14:13

12

th e orig in s of g r am mar

cognition of all the paradoxical aspects of language then, even though language itself does not exist, the hard part is over. There remain only two, simpler questions: which increments occurred to which aspects of language to change it from a specialized cognitive device to a generalized communication system; and in what order did they occur? The “only” in the previous sentence is not intended to imply that these problems are trivial; the incremental changes, and their order, are crucial. It is, however, easier to explain a series of small changes in evolutionary terms than a single large one. This does mean that the timescale over which language developed has to be viewed as hundreds of thousands rather than thousands of years—there has to be enough time to allow the many small changes to accumulate. However, as the changes proposed are cognitive there is no need for a series of proto-languages to precede full language (although the model does not preclude this). The phonological problem, how the cognitive structures became expressed as signals, can have a separate solution to the lexis-grammar- semantics problem, and will not be examined further here.

What Is Language for? If we wish to discover why language evolved, a good starting point is to look at what we use it for today. It is unlikely that, when it began, language was as versatile as it is now; but somewhere in the current wide range of uses we may find clues to the original purpose. On the broadest level, language can be viewed today as occupying two particular niches: cognition and socialization. The cognitive uses for language can be summarized as problem-solving and planning; but not all cognition needs language, as a simple thought experiment shows. Imagine an object, any object; rotate the object so that you can see the other side; now turn the object upsidedown. While this visual experiment was activated by words, the transformational cognition itself required no language, and there is no reason to believe that this mental visualization is a product of language. In fact, if we consider what we think about through an average day, very little of it has linguistic content. True, we often seem to be conducting an internal monologue, but this is more about what is happening rather than causing things to happen. The internal monologue does, however, represent a continuing linguistic reaction to what we are thinking. Language may not provide the majority of our cognition, but the thinking about thinking—our metacognition—is dominated by language. It is this metacognition that we humans use when we analyse a problem or speculatively plan. We can also consider language in terms of what we use it for socially. Some of these uses have implications for cognition; but it is in socialization, and not in cognition, that we can see them as solutions to evolutionary needs. The social uses of language can be divided into four types: phatics, direction, negotiation and information sharing.

CO22552.indb 12

23/04/2010 14:13

Why All the Fuss?

13

The first two social uses of language are relatively simple. Phatics is the production of sounds and gestures to acknowledge our relationships with others. It does not require complex constructions and can rely on verbal gestures with simple meanings; it is therefore closely related to non-verbal grooming in the rest of nature (Dunbar, 1992). The second social use, direction, is a product of a willingness to engage in joint enterprises. It requires a level of co-operation between individuals, but it does not require complex constructs—simple imperatives, possibly accompanied by phatics, suffice. Negotiation, the third social use of language, requires a more complex signal: as well as the action of asking-for (or demanding) that direction necessitates, there is the quid pro quo action of offering. Negotiation therefore involves the communication of two different actions involving a range of objects; and there is also the need to cognitively link pairs of action-object constructs to form an exchange. Negotiation, however, does not require the full power of language: it can be carried out using simple one-argument forms (verb-plus-noun), with restricted temporality (present tense only) and limited denotation (physical pointing rather than labelling with words). The final type of social language use, information sharing, requires the full complexity of language. Information sharing gives the receiver access to the experience and viewpoint of the sender, so it involves events which are no longer current and which occurred elsewhere. To share information the sender and receiver must each model a “virtual stage” on which the sender’s experience can be re-enacted. This requires a shared metacognition about the information being shared, and a meta-signalling system to create the shared metacognition. Information sharing involves not just signalling about the information, it also requires signalling about the information context; the signalling system must have enough complexity to signal multiple temporal relationships, to share absent events, to describe the sender and receiver of the signal as third parties within the signal, and to co-ordinate a series of events into a story. Information sharing is where language finds its full communicative purpose. So if we want a plausible evolutionary explanation for grammar in language, we should concentrate our search on the social process of information sharing. The question of why we need grammar is tied to the questions of how a social structure evolved requiring the exchange of complex information, and what that social structure was.

Mapping the Journey This book considers the origins of grammar from an anthropological perspective. Anthropology, however, covers a wide area, and it can be seen as including a range of other disciplines. As the study of becoming and being human, anthropology includes both sociology, the study of humans in groups, and psychology, the study of human minds; and language, the lubricant for our peculiarly human

CO22552.indb 13

23/04/2010 14:13

14

th e orig in s of g r am mar

groups and the engine of our peculiarly human cognition, is therefore of primary interest to anthropology. This journey to the origins of grammar will therefore attempt to map an academically wide area, but attention will remain focused on the key question: how does being human relate to having grammar? The book will follow a traditional approach: setting out the issues, suggesting a solution, and reviewing the solution in terms of available evidence. Chapter 2 looks at current work on the origins of language, and some of the routes that have been proposed to get from prehuman signalling to human language. Of particular interest is the significance that these routes have for the origins of grammar. Chapters 3 to 5 look at current linguistic theories, showing that linguistics is a science still at the comparative theory stage; the considerable disagreement over what is involved in having and using language means that linguistic theories of language origins remain hotly disputed. In chapter 6 we look particularly at the anthropological perspective: what is the peculiar nature of human co-operation, and what does this mean for grammar as a cognitive and communicative device? The psychological perspective is presented in chapter 7. Being able to recognize and work with models of ourselves and others is essential for human co-operation; this chapter examines how this modelling works—and why it is, in evolutionary terms, weird. Chapters 8 and 9 then bring together the themes discussed in the previous chapters to propose a story of how we became human. They concentrate particularly on the modelling of self and others as a generator of both co-operation and language grammar. With chapters 10 to 12 we move on to consider some of the evidence: what nonhumans, human children, and our understanding of time tell us about being humans with language and grammar. Self- and other-modelling is shown in these chapters to be a determining principle for much language grammar. Finally, in chapter 13 the itinerary of the book is reviewed, identifying the major milestones and attempting to establish the universals of grammar. In this chapter we have explored many different topics around language: brains, bipedalism, tool-making, culture, co-operation and genes. This has not been an attempt to comprehensively cover the field of human origins, but to raise some issues that have significance for a study of grammar genesis. As we shall see through the rest of this book, however, the new ways of thinking that came with grammar are heavily implicated in our species uniqueness.

CO22552.indb 14

23/04/2010 14:13

2 The Story So Far The origin of language is a toxic subject; at least, that was the impression

given by the scientific community for over 100 years. In 1866 the Linguistics Society of Paris banned any debate on the genesis of language (Aitchison, 1996, p. 5) after a series of papers that owed more to theology than to scientific method. Language origins became a subject to be avoided, even less respectable than investigating paranormal phenomena. This situation did not really change until 1996, when the first of a series of conferences on language evolution was held in Edinburgh (published proceedings Hurford et al., 1998). This series, which has become known as Evolang, is held regularly every two years, and it is now one of several conference series that have an interest in the subject. Many academic conferences across a range of disciplines now accept papers on language origins, and the search to discover how language came about is beginning to give us some viable answers, backed by a solid theoretical framework. However, the intellectual climate which allowed the language origins debate to be reopened can be traced back even earlier, to Noam Chomsky’s publication of Syntactic Structures (1957). This revolutionized the way linguistics was done—a remarkable feat for such a short book. Before Syntactic Structures the attention of linguists was on what made languages different from each other (e.g. Bloomfield, 1933). The approach was largely structural, viewing languages as coding structures existing mainly between minds, and to which individual minds “signed up”. Chomsky’s generativist approach, in contrast, looked at language inside minds. Despite the title of his book, he did not ask the structural question, how do individual minds come to language? Instead he asked a question about the form of language, what features do human minds have in common that make language possible? As we shall see in chapter 3, some of the proposed answers to this question have not survived critical analysis; but the fact that Chomsky asked this question—and so established a new way of looking at language—made a critical difference to the way linguistics is studied. When the generativist view of linguistics, that language is an internal cognitive mechanism, encountered the Darwinian approach to evolution, the question of language origins once again became a respectable subject for study.

CO22552.indb 15

15

23/04/2010 14:13

16

th e orig in s of g r am mar

Whatever physical characteristics we have must be part of the genetic inheritance from our forebears; and these inherited features survived because they made our forebears more successful at reproduction than other individuals without them. Language is a near-universal characteristic of humans, so it is likely that it is a product of selection for reproductive fitness among our ancestors—it is either itself a successful strategy or a useful adjunct to one (Dawkins, 1996, ch. 6). This chapter looks at some of the promising avenues now being investigated in language origins research. It is not intended to be a comprehensive summary of language origin theories—there are too many of them to be covered in the few pages available here. Instead, the main approaches to the functions of language will be reviewed, looking in each case at the particular feature of language being explained. Of particular interest are the significances these approaches to language have for the origins of grammar.

Language Is Tool Use That language involves the use of tools is, on some level, clearly right: language is a process of assembling components, such as words, into metaconstructs, such as sentences. Just as human tool use allows us to construct and assemble components into increasingly complex tools and objects, language has a constructional process in which the product at one level becomes the tool or component at the next level. Language is not just like tool use, in this definition it is tool use (Stout & Chaminade, 2009). This approach has important implications for the evolution of grammar. If the cognitive process of tool production became, at some stage and in some way, applicable to general cognition and communication then the genesis of grammar is unremarkable: the cognitive processes that allow us to construct complex tools are the same ones that allow us to construct complex language. Seeing language as tool use, therefore, does seem to provide all the answers for the origins of grammar that we need. The ability to analyse a problem into a range of different components, the ability to subdivide a problem into separately soluble segments, and the ability to solve the main problem by integrating all the individual solutions of the problem segments; these are what we need both for complex tool use and for constructing language utterances. There is, however, a flaw in this model: the archaeological evidence indicates that complex language probably preceded complex tool-making. We can place the appearance of complex tools at about 40,000 years ago, because there is a clear change in the archaeological tool record at about that time (Gravina et al., 2005). Before this, the tool set was carefully knapped stone, but essentially limited to cutters and scrapers; after, it is a mixture of bone and stone punches, drills, saws, even needles, as well as a larger range of specialized cutters and scrapers. Wooden artefacts have not survived, and they may have provided some kind of continuity of complexity between the two tool sets; but this is only hopeful speculation.

CO22552.indb 16

23/04/2010 14:13

The Story So Far

17

On the evidence there seems to have been a technological event about 40,000 years ago that changed both the quality and range of tools made; and this new technology seems to have propagated swiftly through most of the human groups on the planet. This speed of propagation indicates several things. First, it is too fast for it to have been an evolutionary event, it is much more likely to have been a teachingand-learning event: the speed of transfer between individuals reflects a contagious rather than an inherited pattern. Second, social mechanisms that enabled the widespread transmission of knowledge must already have been in existence for fast transmission to take place. Third, the communication system available must have been complex enough to both negotiate the exchange of knowledge and to describe the technological complexity involved in manufacturing and using the new tools—in short, it would have required a pre-existing language-like grammatical complexity (Szathmáry and Számadó, 2008). The Upper Palaeolithic Revolution of 40,000 years ago is a significant event in our species history, but it is unlikely to have been the time when language began. Language is tool use, yes; but it does not seem to be the case that tool use generated language.

Language Is Play Language as play is another proposition that, in some ways, is obviously correct: language involves the exchange of tokens that are themselves largely valueless (they are just sounds), but which have value in terms of a permissive social environment of exchange (human culture) (Tomasello et al., 2005). Play shares this nature with language. When we play we are taking part in a model of reality which has its own rules, and in which some of the usual imperatives of existence have been suspended. Play signals are reduced versions of earnest actions: a play bite is a reduced bite, a play fight should not result in real injury, a playful submission does not indicate actual submission. In play, roles can be reversed because they have no significance outside of the game (Bateson, 1985). Actions in play simultaneously represent reality and deny it: the play bite stands for a real bite, but it also indicates that there is no intention to really bite. This corresponds to the representational value of words: a word like lion both represents and denies the real world: I can talk about an actual lion in a zoo, or an imagined lion which only has existence in my mind; I can extend the word lion to cover not only gregarious cats in Africa but also solitary cats in America, which are clearly a different species; I can describe First World War soldiers as lions led by donkeys; and I can use the term to refer to characters in The Wizard of Oz and The Lion, the Witch and the Wardrobe, although they are fictional and share only limited characteristics with the actual lion in the zoo. Language uses the same conditional representations as play and, like play, relies on the receiver being willing to accept the arbitrary representations offered by the sender (Knight, 1998).

CO22552.indb 17

23/04/2010 14:13

18

th e orig in s of g r am mar

In most animals, play is an activity of subadults. Adults do not play because the games are about activities they have to take deadly seriously. Play can be seen as a way of practicing adult activities in relative safety before puberty plunges the individual into a world where survival, eating and reproduction are uncompromisable goals. In contrast to this, adult humans indulge in many different types of play, which indicates that the imperatives of adulthood have somehow been neutralized (Cook, 2000, ch. 4). One type of play specific to humans is language itself—our willingness to co-operate together in a signalling environment which permits speculation, opinion and fantasy seems to be unique in nature. So language can be classed as a form of play. Unfortunately, though, this does not seem to have any implications for the origins of grammar. The key feature of play is its arbitrariness—rules are conditional and adaptable. The only basic rule of play is that there should be agreement between the players; every other rule can be negotiated and renegotiated. This allows play to take a range of different forms, from highly regimented professional sports to freeform “mucking about” (Caillois, 1958, ch. 6). Language grammar does not have this level of flexibility, there are some features that are uncompromisable. At a minimum, language grammar requires a differentiation in a message between action and object roles (nouns and verbs); and it seems also to require a structure which allows at least two objects to be related together through an action, one as the cause of the action and the other as the recipient (the subject and object roles of nouns) (Allen & Saidel, 1998). As we will see in chapters 3, 4 and 5, these are only some of the ways in which language grammar follows systematic rules, and therefore differs from play. There is another important way in which the grammar of language differs from play. Play is a levelling event, making everyone equal despite differences; grammar, on the other hand, relies on the ability to create hierarchies. Noun phrases can take the place of nouns, and can contain their own noun phrases; verb forms can contain other verbs to create nested meanings, such as I began to try to find an answer. Grammar requires hierarchy, which in turn creates the need for a structuring of rules; hierarchy is not a basic requirement of play. Language as an activity is certainly a form of play: it is an interaction between individuals which requires those individuals to occupy a shared imaginary space where meaning is negotiated rather than reliant on actuality. Language as a rule-set, however, does not share the arbitrariness that play possesses. It is the need to communicate particular types of message that imposes structure—or grammar—on language. These communicative needs do not correspondingly exist in play.

Language Is a Signal of Fitness A problem that perplexed Darwin was the peacock’s tail: how could something so elaborate have evolved just to make the peacock look good to the peahen?

CO22552.indb 18

23/04/2010 14:13

The Story So Far

19

Why would females favour such a non-functional signal of fitness (Darwin, 1859, pp. 87–90)? The answer was finally provided over 130 years later: the peacock’s tail is a cost to the male and therefore reliably indicates to the female that this male can pay the cost and still thrive (Zahavi & Zahavi, 1997). Natural selection will favour females who choose males with elaborate tails because their offspring will tend to be fitter, and the tendency towards elaborate tails will therefore be selected for both by female choice and by the necessary fitness of the male tail-carrier. Could language be a similar badge of fitness? Were talkative partners favoured because the capacity and willingness to talk represented an arbitrary cost demonstrating reproductive value? It is certainly true that language represents a high cost in many ways. Cognitively, language requires the capacity to remember both rules of language and components (words), the capacity to quickly formulate acceptable utterances to express intended meaning, the capacity to control the mechanisms of speech articulation in rapid and complex movements, and the capacity to interpret a stream of sound into a stream of meanings. Being able to meet these cognitive costs is significant even today: talking effectively is a skill that is valued in careers which involve leadership, entertainment or teaching. However, being a good talker is not the only route to social (and presumably, therefore, reproductive) success. The traditional markers of fitness—physical skill, motoric dexterity, body symmetry, creativity—are valued as highly or more highly than speech skills, as the incomes of some sports people, musicians and entrepreneurs demonstrate. Talking, unlike the peacock’s tail, does not by itself seem to be a certain route to reproductive success. If language really were a costly sign of fitness then we would expect it to be a one-way signal: it would be unequally distributed between the sexes, and it would predominate in the sex that has the lowest reproductive cost. The peahen does not need an ornate tail because she pays all the costs of incubating and protecting the young—effectively, she has control over the process of reproduction. In contrast, the male only contributes sperm, which is plentiful and cheap. The fitness of the female is of minor importance to the male—impregnating one female does not limit his opportunities to impregnate others. The fitness of the male is, however, of paramount importance to the female. If the wrong male impregnates her then the best she can do is to abandon the eggs or offspring; but by then she has already made a costly commitment of time and resources to reproduction, and it is a cost she cannot abandon lightly (Campbell, 1999). It is in the female’s interest to discriminate fitness before copulation, and therefore in the male’s interest to demonstrate it to gain access to reproduction; and that is why male peacocks have magnificent trains, and females do not. This separation of roles is not something we see in human language in any marked way. While it is true that there are subtle statistical differences in the way men and women use language (Eckert & McConnell-Ginet, 1998), there is no massive separation of language capacities by gender. Both male and female

CO22552.indb 19

23/04/2010 14:13

20

th e orig in s of g r am mar

humans seem able to use the full range of language, and there are no forms that are clear to one gender and opaque to the other. If we look at the meanings being shared in language rather than the complexity of sharing itself, then we are in even more difficult territory. By sharing meanings the sender is making information available, information that is valuable to the receiver because it enhances their fitness; but where is the cost to the sender in making the signal? Giving information to a sexual rival is costly because anything that enhances their fitness relatively decreases the sender’s fitness; but to be a costly signal it would have to be done ostentatiously in the presence of prospective sexual partners. Sharing meaning is costly to the sender, but it only makes sense as a costly signal if the receiver is of the same sex and there are members of the opposite sex to see the cost being paid. This is a very specific way of signalling which explains only a fraction of what humans do with language: we share information, often privately in pairs, or in groups containing only individuals of our own sex; and, most importantly, we share not just valuable true meanings but apparently valueless fictions. The fictional nature of many language utterances poses a problem for utilitarian approaches to language origins, such as costly signalling. Why would a receiver be happy to be told lies? This will be looked at more fully in chapter 9, because any theory of language origins should be able to explain this apparent fitness anomaly. Humans like being told stories, and good story-tellers are socially valued, as J. K. Rowling can attest. It seems likely that story-telling was one of the first uses to which language was put, and it may even have been one of the primary causes of language. Certainly, nowadays, the entertainment industries together form a major component of the world economy. What does costly signalling imply about the nature of grammar? If complexity of form is an indicator of cost then producing complex utterances should be an indicator of fitness. We would expect humans, especially males, to use complex forms whenever possible; complexity should be valued and rewarded with reproductive opportunities, as happens with songbirds. We would expect talking competitions to be commonplace, and we would expect the ability to produce complex utterances to be positively correlated with breeding success. This does not seem to be the case. Instead, we seem to value clarity over complexity, and prefer simpler utterances over complex ones (Denton, 2006). Using inappropriate complexity of form is seen as an unattractive trait, and unavoidable complexity is viewed as a necessary evil. An important feature of language is that we use it to work with people rather than against them—we value co-operative dialogue over competitive argument. If there is cost to the sender in a language utterance it seems to lie in the meanings offered—the information being giving away—rather than the complexity of form (Buzing et al., 2005). It is true that there is a physical cost in being able to use grammar to produce utterances containing complex meanings, but it is the complexity of meanings to be expressed that drives the need for grammatical

CO22552.indb 20

23/04/2010 14:13

The Story So Far

21

complexity—the one-dimensional stream of sound has to be capable of expressing ideas which are often multi-dimensional. In fact, expressing complex meanings in simple forms is what we consider most co-operative. If language by itself is seen as a costly signal of reproductive fitness then complexity is an indicator of the sender’s cognitive costs, and duration of utterance is an indicator of physical cost; and these are what we would value in a speaker. This, however, is birdsong, not language. Instead, there seems to be another costly signal, co-operation, for which language and grammar are necessary tools; and it is the complexity of signalling required to facilitate human co-operation that drives the complexity of grammar (Kirby, 1998).

Language Is Gestural Language as gesture is another view of language that is clearly correct: language is the conversion of meanings into a series of orofacial gestures which produce sounds, and these sounds are interpreted by the receiver’s brain back into meanings. It is also true that language is mode-independent: it can work through any channel where the sender is able to produce motoric gestures to express meanings, and the receiver is able to perceive those motoric gestures and interpret them back into meanings (Corballis, 2003, ch. 9). Your reading of this page is evidence of this: you are interpreting meanings from ink marks on paper, produced in the first instance by my fingers pressing keys to produce contrast effects on a screen. Language can be produced as sounds, marks, a series of dots and dashes, waved flags—and, most importantly, hand and arm gestures. Modern deaf community signed languages are examples of language by other means; and we should be in no doubt here, they are real languages. A signer can express all the utterances that a speaker can produce, they convey the same meanings and use similar complexities of form. In terms of grammar, sign languages have constructions to express time relationships and non-present and nonexistent events, to describe, and to link utterances together into metaconstructs. They are segmented (they use word forms); they are differentiated (they use different types of words, such as nouns, verbs and prepositions1); they are hierarchical (they have adjectives dependent upon nouns, adverbs dependent upon verbs, and noun phrases which contain noun phrases); and they are rulebound (the same basic form is used for most utterances, and order is significant) (Kyle & Woll, 1985). Gesture, however, allows sign languages to have real place-marking. There is no “over-thereness” about the spoken words over there, but the sign language equivalent does indicate relatively where there is. Up is up and down is down, in front and behind are where you would expect. Sign language is able to put events 1.  Strictly speaking, a preposition always occurs before its noun phrase, and some languages have postpositions, which occur after the noun phrase, instead. The correct linguistic term for both types is adposition, but preposition will be used here as it is more familiar.

CO22552.indb 21

23/04/2010 14:13

22

th e orig in s of g r am mar

onto a virtual stage in front of the receiver that spoken language cannot easily replicate—and this is why we use a lot of gesture when we are speech-talking (Hanks, 2005). This link between actuality and utterance has caused some commentators to see signed language as a possible halfway house between non-symbolic prehuman signalling and symbolic language. Ape vocalizations are heavily constrained and are used mainly to express emotion, they do not seem to be under volitional control. In contrast, the dextrousness of apes is notable, and they appear to have quite proficient control over their gestures (Arbib, 2005). Could it be that language first appeared in the gestural mode, and only later converted to the vocal mode? This idea has been explored (Iverson & Goldin Meadow, 1998; Corballis, 2002; Arbib, 2005, among others), and several possible developmental processes have been suggested, with varying timescales. Essentially, though, the proposal is that volitional gesture in australopithecines developed slowly to cover a range of different signalling needs, such as hunting—gestural communication remains an important part of many hunting cultures today (Lewis, 2009). At some stage the vocal channel became subject to greater conscious control—possibly because, like birds, singing became a signal of fitness (Mithen, 2005, ch. 9)—and it became possible to generate meaning-rich vocal gestures as well as manual gestures. At that stage the preferred channel for exchanging meanings became the vocal channel, and gesture became an important, but secondary, accessory (Steklis & Harnad, 1976). What does this model imply for grammar? Unfortunately, not much: if both gesture and vocalization can use grammar of equal complexity, the channel used gives us no indication of when or how the complexities arose. The mode of the signal is transparent when we look for grammar origins—and, to a certain extent, it is unimportant. The cognitive changes that allowed vocalization to become volitional and controllable do not really have any implications for the cognitive changes that created the basis for grammar. Language is currently partially gestural and may well have evolved from a completely gestural signalling system; but language grammar is modality-independent, and its development is unlikely to have been affected by changes in the signalling channel.

Language Is Cognition It is uncontroversial to say that language is a product of thought and is about thought: it is the sharing of ideas that distinguishes language as a signalling system in nature (Kendon, 1991). There are, however, two adjunct questions about language and cognition that are less clear-cut: is language the cause of human cognition or a product of it; and is language-thought a product of a specific module in the mind which only humans have? Since the 1970s there has been an enterprise in linguistics, referred to as

CO22552.indb 22

23/04/2010 14:13

The Story So Far

23

cognitive linguistics, which has been looking at the relationship between thought and language. It views language production as the outcome of cognitive meaningmappings: language provides a framework to convert thought into an arbitrary communication system, and back into thought. The enterprise has been described as an archipelago of ideas rather than a peninsula of knowledge (Geeraerts, 2006, p. 1) but it is beginning to identify strong links between cognitive functions that have no direct linguistic content, such as vision, and the language forms we use. To give a simple example, body knowledge about up and down, a product of the external force of gravity, is reflected in our use of the terms: up is skywards even if you are standing on your head. In contrast, body knowledge about left and right is view-specific, if you turn around their external orientation changes. For that reason the terms are often accompanied by possessives (my left, your right) or by gesture to indicate direction; the terms, like the view, are negotiable. This negotiability allows body knowledge terms to be extended to other, metaphorical, purposes: the political left and right; and the cultural significance of terms like dextrous, sinister, adroit, gauche and cack-handed. This seems to imply that, semantically at least, language is both the product of general cognition and a driver for it: some meanings are projected onto thought (and thus onto language) by external imperatives; some are generated by thought and projected onto language and external actuality; and some are generated by language use and moderate thought and cultural reality (Kövecses, 2002, ch. 3). Linguistics has at least two views of language and its relationship with cognition. The position taken by cognitive linguistics, that language is a product of general cognition, contrasts with that of generative linguistics, that language is a product of a language-specific module in the brain. This has considerable implications for the origins of grammar: the cognitive linguist looks for grammar in general cognition and attempts to relate grammar origins to the need to communicate particular cognitive constructs; the generative linguist has an interest only in where the language module, or the key component that defines language, originates. Cognitive linguistics therefore offers a naturalistic view of language and cognition which is open to a Darwinian explanation of grammar origins (Palmer, 2006); generative linguistics offers a modular view in which language can be isolated from general cognition, and which requires a non-gradual explanation of grammar origins (Chomsky, 2002, pp. 84–91). The cognitive linguistics view of grammar origins is discussed in detail in Chapter 5, but the importance of the approach needs a short preliminary description. A comprehensive review of this area has been produced by Johansson (2005), and he gives a largely uncontroversial analysis of current cognitive linguistic thinking. He sees syntax as consisting of four levels of elaboration, which apply to individual development and may apply to language evolution, too (ibid, pp. 230–235): first comes structure, allowing simple two-word forms; then there is hierarchy, allowing phrase structure; then comes recursion, allowing potentially infinite forms to be generated; and finally comes flexibility, which

CO22552.indb 23

23/04/2010 14:13

24

th e orig in s of g r am mar

allows the same thought to be expressed in different ways, and which can be apprehended before or after recursion. The drivers for linguistic forms are twofold: social scripts, which allow the communication of interpersonal relationships; and image schemas, which allow the description of events. This description of grammar gives both an explanation of the way grammar evolved and possible reasons why that evolutionary process began and continued. What it does not do is directly address the question of fitness: why is having a grammar in language a “good thing” and what advantage does it give to those possessing it? This question is less important to a cognitive linguist because it is part of the evolutionary conundrum of language as a whole: why should senders be willing to give away valuable information; and why should receivers be happy to accept the information given when they have no way of detecting whether that information is fact, opinion or lies? Compared to this paradox at the origin of language, the origin of grammar would seem to be simple: once the language paradox has been overcome then the need to share increasingly complex mental constructs inevitably creates the need for more complex language structures to make sharing possible. The problem of grammar origins is not how complexity arose, but why those particular complexities arose (Cronin, 2005). This cognitive view of language as a co-operative activity inevitably defines it as a social phenomenon. Language exists because humans co-operate in stable social groups, and the way they co-operate drives what they need to communicate. This brings us to the next description of language, as a way of making human groups work.

Language Is Social Construction Language as a social lubricant describes much of what we do with language, and is clearly correct on many levels. We use language to work together in groups to achieve common ends; we use language to build social institutions; we use language to pass on ideas, allowing our ideas to live on for generations after our death, thus creating cultural continuity; and we use language to negotiate, to include and exclude people, to entertain . . . essentially, to socialize. It is difficult to imagine a simple human society without language, and impossible to imagine complex civilization without it. Human societies have all the features we find in grammar. They are segmented, consisting of individuals who are nonetheless capable of working together to produce solutions where individuals cannot. They are differentiated, with individuals taking particular complementary roles to solve problems; some of the roles are ad hoc (“you push, I’ll pull”), but others become formalized and begin to define individuals both to others and to themselves (“I’m a potter”). Finally, societies are hierarchical, with some individuals deferring to others in ritualized ways which sometimes appear not to promote the deferring individuaI’s reproductive fitness.

CO22552.indb 24

23/04/2010 14:13

The Story So Far

25

The social aspect of language has been explored in some detail. A branch of linguistics, pragmatics, is devoted to it (e.g. Thomas, 1995); a major grammatical theory, systemic functionalism, is based around the fact that language is exchange (Halliday, 1994, ch. 4); and, in terms of language origins, Searle (1999), Dessalles (2007), Worden (1998), Locke (1998), Dunbar (2004) and many other researchers have emphasized the importance of social construction as a source of language. Social construction has considerable implications for grammar. If language is the information conduit for a social group then it must be capable of expressing the complexities of that social group—it has to allow group members to tell other members about the group. At the very least it should be able to signal relationships between others, and between objects and others, as events. It must also be capable of placing those relationships both in space and in time, because the events being communicated are not necessarily current. The role of grammar in the communication of social constructs has important ramifications for the origins of grammar, and these will be explored in more detail in chapters 6 and 7.

Language Just Is The discussion in this chapter demonstrates that language seems to be central to the panoply of effects that define us as human. We can see language as a tool-like device that enables us to do things we could not otherwise do: it allows us to formulate a problem as a segmented series of sub-problems, and it allows us to recruit others to assist in solving problems that are beyond an individual. We can see language as playful, allowing us to communicate ideas in an environment where the pressure of reproductive fitness has somehow been switched off, or considerably dampened. We can see language as a signal of fitness, allowing us to directly show our cognitive fitness and to indirectly demonstrate that we can give away our knowledge, compete with our conspecifics, and still win. Language can be seen as a gestural act, no different from other gestures in that it is both a way to do things and a way to communicate things. We can look at language as a process of cognition which allows us to generate complex models of relationships in the world around us, and to then manipulate those models to test different outcomes. And we can look at the communicative nature of language to analyse what it is that language allows us to do in groups. Language is all these things, it does all these things; and a theory of language origins that does not address all of these issues is likely to miss something important. In relation to grammar, however, the problem is much simpler: grammar probably emerged from a human capacity which involved segmentation, differentiation and hierarchy, such as the cognitive architecture needed to negotiate human social systems; and the cognitive mechanisms which allow human social structures to exist are our internal models of reality. How we see the world affects how we relate to it, and the co-operative way we relate to the world is what permits a level of social co-ordination unusual in nature. Yet our capacity

CO22552.indb 25

23/04/2010 14:13

26

th e orig in s of g r am mar

for co-operation does not mean that there is a simple link between language and socialization; instead, there are cognitive functions which lay behind them both, and which drive the forms of both. Before we were able to co-operate we must have had the cognitive systems to permit co-operation, so it is likely that there was a continuous and intertwined development of cognition for co-operation, co-operation itself, and the grammar forms required to communicate that co-operation. This very rough developmental guide will be reviewed in more detail in chapters 8 and 9.

CO22552.indb 26

23/04/2010 14:13

3 The Heavy Hand of Generative Linguistics can currently be divided into two main eras, pre-Chomskyan and Chomskyan: the importance of Chomsky to the science is that great. Before Syntactic Structures was published in 1957, linguistics was largely an anthropological enterprise: linguists went to other parts of the globe, lived with the locals, and learned their cultures and their languages as a single package (Boas, 1938). Social theories of language were common, such as the view that an apparent gap in the available functions of a language indicated a similar gap in culture (Sapir, 1921), and even a gap in cognition (Whorf, 1956). The approach to language was mostly behavioural—language was seen as culturally learned, with children being trained into language—and the idea of language as an inherent human capacity based on universal genetic attributes was largely unconsidered. In the 1920s and 1930s Bloomfield had adopted a mathematical approach to linguistics (Tomalin, 2004), and the Prague School (notably Jakobson, publ. 1987) had developed a structural approach based on de Saussure’s earlier work (publ. 1972), but the general approach to language remained mainly anthropological. When Chomsky’s generative grammar programme appeared, it altered the whole way linguistics was done. Generative grammar views language as more than learned, it has an innate, genetic nature with its own rules and forms. What makes human languages similar is not that they do similar jobs, they are similar because they are all generated by the same language engine, and they do similar jobs because those are the only jobs that languages can do. After 1957, linguistics and anthropology largely parted company: culture no longer had significance for the basics of language, it could only impose a light dusting of difference onto a solid core of sameness. This chapter will give a short overview of the development of generative grammar since 1957, and will hopefully provide an insight into a truly exciting half-century in linguistics.

The history of linguistics



CO22552.indb 27

27

23/04/2010 14:13

28

th e orig in s of g r am mar

Linguistic Structure In Syntactic Structures (1957), Chomsky reviewed the grammar models available at the time, showing them to be inflexible, incomplete, and unable to correctly parse complex grammatical utterances. He proposed the development of a new model, which he labelled Transformational Grammar. This was to be a representation that encompassed the grammars of all possible human languages, past, present and future—effectively, a Universal Grammar. The transformational rules in this grammar would make it generative and not just descriptive: it would explain existing utterances, but it would also predict all possible grammatical forms in any language from a limited set of utterances offered to it. This was an essentially computational approach, offering a grammar engine which would work just as well on a computer as in a human brain. However, while Chomsky provided examples of how the transformational rules could be arrived at, he did not provide a complete, working grammar. Syntactic Structures was a promissory note for a solution still to come. In Aspects of the Theory of Syntax, Chomsky (1965) elaborated what has become known as the Standard Theory. This proposed two structural levels in language, deep and surface. The deep structure consisted of the underlying “rules” which define the nature of a language, and which rely on innate language capacities common to all humans—the Universal Grammar. The surface structure consisted of language-specific rules which therefore have to be learned. In Standard Theory, a set of transformational rules interpret between deep structure forms and surface structure utterances, and a simple process then converts between surface structure and the phonological signal. The signal itself is therefore just the tip of the generative iceberg of language. Chomsky identified three resources as generators of deep structure: the lexicon (words themselves), phrase structure rules (how words are used together), and the encoded semantic values (what the words mean). These resources are not language-specific in Standard Theory, they are part of the universal resources available to all humans. Although Chomsky thought that Standard Theory should be able to effectively describe Universal Grammar, he left it to others to identify the mechanisms of that engine. Without a detailed definition of the theory, however, different people found different solutions. Fillmore (1971) proposed Case Grammar, in which deep structure was based around the effects of the verb on other sentence components; Relational Grammar (Postal, 1968, among others) was almost the mirror-image, being concerned with the noun hierarchical relations of subject, object and indirect object; and George Lakoff’s Generative Semantics (1971) concentrated on the semantic content of deep structure. Very soon after the publication of Aspects, it became clear that there were problems with the Standard Theory (Chomsky, 1977, pp. 151–152); in particular, semantic interpretation seemed to be more variable than should be the case for a deep structure resource. The role of semantic interpretation therefore had to

CO22552.indb 28

23/04/2010 14:13

The Heavy Hand of Generative Linguistics

29

be changed, to allow it to moderate surface structure as well as deep structure. This, in turn, meant that surface structure no longer had a simple one-to-one relationship with the phonological interpretation; so another set of rules was needed to convert between surface structure and both the phonological and the semantic interpretations. However, there still remained a problem with this amended model. Lexicon had been divorced from phonological and semantic interpretations; but then, what is a mental lexicon if not a cross-referencing of meanings with sounds? Standard Theory did not provide a satisfactory answer to this. The only feature that could occupy the role of lexicon in this model was the word class (noun, verb, etc.), but Chomsky placed this firmly in the phrase structure rules. Of course, it is possible to reduce the importance of the lexicon as a moderator of deep structure, as was done for the semantic interpretation; but this leaves deep structure even further impoverished.

Extending Structure The Standard Theory was revised, mainly by Jackendoff (1972), and became known as the Extended Standard Theory (EST). Jackendoff placed the semantic interpretation parallel to the transformation between deep and surface structures, allowing access to it from both structures and from the transformation rules themselves. The semantic interpretation became a resource available to other processes rather than a process itself. This model proved robust enough to allow productive work to be done for over a decade, and the theory was expanded in several complementary directions. It seemed that a stable base for generative linguistics had been found. One important approach in the EST was X-bar theory. This stated that language phrases consist of an X-value (the phrase-defining word) and an optional specifier. So the phrase simply happy man consists of a noun phrase (X-value man, specifier simply happy), and an adjectival phrase (simply happy), which consists of an X-value (happy) and a specifier (simply). This two-part relationship occurs at all levels of an utterance, creating a binary tree structure with meanings combining at each level to eventually create sentential meaning; and it occurs in a similar way in every language, making it a language universal. This binary tree structure was an important contribution to the understanding of language generation and is still a major tool of linguistic analysis today. Chomsky, however, looked for a greater emphasis on language universals in the model, and during the middle 1970s he was working on what became known as the Revised Extended Standard Theory (REST). This theory emphasized the Move function in grammatical construction, an effect that is usually illustrated by the English interrogative construct. When phrasing a question, such as Who did you see?, we appear to take the statement form you saw X, move the object to the front of the construct and insert a neutral verb (such as do or be). This sort

CO22552.indb 29

23/04/2010 14:13

30

th e orig in s of g r am mar

of movement occurs commonly to indicate particular grammatical forms, and can be seen in English in the passive form (you saw the man → the man was seen by you) and the noun phrase (you saw the man → the man you saw [was happy]) as well as the interrogative. Movement is important in generative linguistics because it emphasizes that the cognitive construction of the action between two objects, one as actor (or doer) and the other as patient (or done-to), is independent of the surface utterance; and there has to be transformation by movement to allow the deep structure thought to become the surface structure utterance. There are, however, issues with this, not least the question of the function that movement serves. If the deep structure of thought has a natural order, why is this subverted to the unnatural order of the utterance? For instance, the interrogative can be, and often is, not subject to movement, with you saw X becoming you saw whom? The subject interrogative form is also not subject to movement (X saw you → who saw you?), implying that movement applies to only a subset of interrogatives. So what function does the moved interrogative serve that is not met by the unmoved interrogative? There are also problems where constructs which appear to have been move-transformed do not seem to work in their unmoved state (Aren’t I right? ← * I aren’t right1), a problem known as the Move Paradox. The Move function requires the concept of the linguistic “trace”, an element of the structure that is not expressed. This phonologically zero element marks the pre-move position of a moved element, so it represents the reflection of deep structure into surface structure. Chomsky took the view that the zeroing of traces was a function of the phonological rules. So, for example, the surface structure of who did you see t? is rendered to the phonological form who did you see? The surface structure interrogative form comes from a transitive base form (you saw X), so the trace is a marker of transitivity for the verb see. The trace, like Move, is important to generative linguistics because it emphasizes underlying cognitive structures, although at a different level of utterance construction.

Principles and Parameters By 1982 Chomsky was pursuing the zero element with vigour, and developing the theoretical structure that became known as Principles and Parameters (P&P). In this theory there are certain features of deep structure which are cognitively present in the final utterance even if they are physically absent. These features are passed through the transformations of the Extended Standard Theory without excessive change because they are “pre-parameterized” in the process of language-learning. The Universal Grammar available at birth permits several different language structures; but the first language learned switches off some of those options, which reduces the range of permissible language forms and creates 1.  In linguistics, an asterisk is used to indicate an ungrammatical construction, a question mark is used to indicate disputed or uncertain grammaticality.

CO22552.indb 30

23/04/2010 14:13

The Heavy Hand of Generative Linguistics

31

a standard template for utterances. If any element of the standard template is not actually present in a final utterance, it can be interpolated as a trace. At this stage, Chomsky saw Universal Grammar as a potential in a newborn human’s mind which could generate one of a large but finite set of real languages (Chomsky, 1986, pp. 38–39). P&P introduced new terminology to the Extended Standard Theory. The initial state of language, identifiable with Universal Grammar, became S0; and the attainable state, identifiable with an individuaI’s grammar, became SL. There was also I-language, language internal to an individuaI’s mind, which was contrasted with E-language, social or external language. This reflected Humboldt’s (1836, pp. 74–76) earlier distinction between Language as a capacity and Languages as communication. Vital to P&P were the principles of Government and Binding. Government was really an extension of X-bar theory, setting out the rules by which X- values governed their specifiers (Cook & Newson, 1996, p. 51); it reflected the fact that languages seemed to show preferences for specifier-first or X-value-first constructs. So if a language created sentences by putting the governing verb before the governed object, it was also likely to put the governing noun before the governed adjective. However, examples of non-compliance with this rule, such as English (verb before object but noun after adjective), weakened it as a clear universal. It is now mainly used to illustrate the relationships between verb inflection and subject, between verb and object and between preposition and indirect object. So in the sentence John saw a house with chimneys, see governs John as a past tense form, see governs a house with chimneys as a verb-object form, and with governs chimneys as a prepositional form. Binding is the semantic process by which lexical items refer back to previously defined items, sometimes at some distance. Pronouns are the most obvious example, but adjectives like same (as in the same idea) or adverbs like again (it’s happened again) can also have binding properties (Chomsky, 1988, p. 52). The search for a simple universal system of grammar was becoming complicated. Chomsky himself described the P&P model as consisting of X-bar theory, Government theory, Binding theory, q-theory (Theta theory), Case theory, Bounding theory and Control theory (Chomsky, 1982, p. 6). q-theory states that all components required to formally define a sentence are expressed in the deep structure even if they are missing in surface form. Case theory is concerned with the assignment of case where it is needed (for agreement or tense), but as an abstract in deep structure. Bounding theory covers the conditions of separation operating on items subject to binding. Finally, control theory is concerned with the rules governing phonologically zero elements, both the old t and the new PRO, used to explain languages like Spanish that drop subject pronouns. Yet, despite this large theoretical framework and what was becoming decades of investigation by a legion of linguists, the exact definition of Universal Grammar remained more promissory than real.

CO22552.indb 31

23/04/2010 14:13

32

th e orig in s of g r am mar

Small Is Beautiful In an effort to place Universal Grammar back at the heart of generative linguistics, Chomsky went back to first principles, and in 1995 he produced The Minimalist Program. In this book Chomsky reiterates the P&P approach of the REST as providing vital tools for linguistic analysis, but in chapter 4 he drops a somewhat enigmatic bombshell: A linguistic expression of L is at least a pair (p, l) meeting this condition [capable of Full Interpretation]—and under minimalist assumptions, at most such a pair, meaning that there are no levels of linguistic structure apart from the two interface levels PF and LF [Phonetic Form and Logical Form]; specifically, no levels of D-Structure or S-Structure. (Chomsky, 1995, p. 219.)

Chomsky restated this radical position in 2000 (p. 10); suddenly, differentiation between deep and surface structures is swept away. This has not endeared him to many in the American linguistics community, who have spent a large part of their lives identifying the components of the two structures and mapping the transformational rules between the two. With each new incarnation of Generative Grammar there have been theorists who have continued working on the old model, although most have eventually moved on to the new one. However, in the case of the Minimalist Program, Chomsky has not carried many adherents with him. As Newmeyer says: If I were to write this book several years from now, I would opt for the MP [Minimalist Program]. However, at the present time, I find the concrete claims of the MP so vague and the total set of mechanisms that it requires (where I have been able to understand them) so unminimalist that I see no reason to encumber the exposition with my interpretation of how the phenomenon in question might be dealt with within that approach. It is also worth pointing out that even leading developers of the MP typically appeal to strictly GB [Government and Binding] principles in presentations to general audiences of linguists. (Newmeyer, 2000, pp. 12–13.)

The Minimalist approach to grammar has, however, inspired a new generation of linguists, often with exciting results. For instance, Hornstein (2001) has proposed that the seven theories of REST (X-bar, q, Case, Binding, Bounding, Control and Government theories) can all be reduced to a single form, that of movement. In doing so he argues against Chomsky’s (1995, pp. 297–312) theory of attraction, or attract/move, which replaced the old doctrine of merge and move. It has to be said that Hornstein’s theory satisfies the minimalist requirement of being simpler than Chomsky’s, and may be a sign of even greater simplification possible in the generative model. One of the more exciting parts of the initial Syntactic Structures theory was the promise of a mechanism which could predict any utterance in any

CO22552.indb 32

23/04/2010 14:13

The Heavy Hand of Generative Linguistics

33

language—a Universal Grammar. With the Minimalist program, Chomsky views Universal Grammar as not necessarily existent in the minds of adults. It is active in the minds of children, and it guides acquisition of their first language; but, once all the grammatical switches have been set to give them their first language, the Universal Grammar then effectively ceases to exist in their brains. Like a dress pattern it can be cut to a particular size; but, once cut, it cannot be restored to its “universal size” state. The Universal pattern is destroyed in the process of producing the usable, specific pattern. One feature that Chomsky emphasizes in the Minimalist Program is the recursive nature of language: Human language is based on an elementary property that also seems to be biologically isolated: the property of discrete infinity, which is exhibited in its purest form by the natural numbers 1, 2, 3, . . . (Chomsky, 2000, p. 3.)

When he uses the term “infinity” Chomsky is referring to the mathematical concept of open sets. This view of infinity is comprehensible only abstractly, by reference to cases: no matter how large a number, you can always add 1 to make another number; and no matter how long a sentence, you can always add to it to make a longer sentence. In a trivial way, it is true to say that language constructs can be strung together to make an infinite sentence. It is also true that the range of possible sentences in a language is larger than the number of sentences that will ever be uttered, but this is also trivial. Language is dialogic, it is not about what can be done but what needs to be done—and what is done—to communicate; and infinity is neither needed nor produced in actual language. Chomsky sees our possession of the property of “discrete infinity” as key to understanding our capacity for recursion. Recursion involves reusing language forms at new levels of construction: the noun phrase stands in place of a noun and contains a noun; but the noun in the noun phrase can itself be a noun phrase, and contain its own noun or noun phrase. This hierarchical reuse requires sophisticated mental modelling, and it can reasonably be viewed as a human-only characteristic; Chomsky’s recent approach has therefore been to emphasize the importance of recursion as the thing that made language possible (Hauser et al., 2002). The assumptions behind this approach have been questioned, however (Pinker & Jackendoff, 2005), and, as we shall see in chapter 9, the role of recursion as the source of grammar is far from settled.

Generative Origins One of the major problems for generative grammar theories is the genesis of language. It is not an issue that Chomsky really tackles, and he has consistently maintained the view that language cannot have pre-language states: it either exists as a complete construct or it does not exist. This has led to the belief that Chomsky supports a sudden and catastrophic appearance of language in humans.

CO22552.indb 33

23/04/2010 14:13

34

th e orig in s of g r am mar

However, while “sudden appearance” best typifies the generativist approach to language genesis, Chomsky himself displays only a cursory curiosity about the subject. By assuming sudden appearance he is able to treat Universal Grammar as an ideal structure, because it is the only game in town. There is no reason to look for variation in this universal engine, by definition there is no significant variation to be found (Chomsky, 2006, pp. 106–113). The generativists who do take an interest in language genesis have to deal with this Chomskyan assumption of catastrophic evolution. This is problematic in evolutionary terms, because catastrophe is usually just that: highly disadvantageous to the phenotype. Three different approaches to this paradox are looked at here. They are far from the only solutions offered, but they do give a flavour of the problem that faces generativists: how can a complex, innate system like language come into being as an integrated system in a Darwinian universe? The first approach is to formulate language origins as a continuous evolutionary event, with a smooth and steady transition from general communication to language throughout human species development. Each small increment in language capacity is selected for by the (arguable) fact that language in any form is an advantageous thing to have (Pinker, 1994, p. 333). A second approach proposes instead a small number of large steps between pre-lingual communication and full language. Bickerton (1998) sees only one intermediate signalling stage, which he refers to as protolanguage. He provides evidence from child language and aphasic language for this single stage, and also from the development of unstructured pidgin languages, which occur spontaneously where two different linguistic communities interact (Bickerton, 1990). Pinker, however, considers the gap between protolanguage and full language to be too great for a reasonable evolutionary explanation, and refers to Bickerton’s approach as “reminiscent of hurricanes assembling jetliners” (Pinker, 1994, p. 366). If, though, Universal Grammar exists in the form suggested by generativism then Bickerton’s position is more tenable than Pinker’s. for generativists, Universal Grammar is an “organ” (Chomsky, 2000, p. 4) which is present in humans and absent from all other animals. It is species-invariant, and it does not assist language—it permits it. Most importantly, it is indivisible: the rules do not relate to specific language constructs or forms, they determine the whole structure of language (Bickerton, 2000). There is no partway-house with Universal Grammar: it is either present, and full language can be generated; or it is absent and full language is impossible. As Bickerton says, “a common code is an all or nothing thing—you either have one or you don’t” (Calvin & Bickerton, 2000, p. 96). Universal Grammar is a principle that both Bickerton and Pinker appear to support, but their different approaches show that it is not fully compatible with both evolutionary theory and protolanguage. The change from protolanguage to full language remains catastrophic in Bickerton’s model (Bickerton, 1998), and the rules of Universal Grammar remain one system, indivisible. Bickerton acknowledges no social factors in the genesis of

CO22552.indb 34

23/04/2010 14:13

The Heavy Hand of Generative Linguistics

35

protolanguage, instead he sees it as a product of interactions between individuals concerning the environment: it contains only information about foraging and other environmental matters. The truthfulness of these environmental utterances is provable by fairly immediate testing, so deception is not an issue in this model (Bickerton, 2002). However, the model assumes the pre-existence of a level of socialization and co-operation which makes the sharing of information worthwhile for both sender and receiver. Bickerton has overcome the problem of language genesis by placing his speakers in a very language-friendly environment, but he does not explain how that language-friendly environment came about. While Bickerton’s protolanguage remains problematic, it should not be dismissed out of hand. It is indeed likely that full language was preceded by a functional communication system which had aspects of segmentation, differentiation and hierarchy; and it is the nature of this system that Wray (2000, 2002a, 2002b) examines. For Wray, the problem of protolanguage is one of continuity: it has to form an intermediate state between the holistic communication of primates, where the whole signal represents a single idea, and the analytical language used by humans; yet it seems from Bickerton’s theories to have continuity with neither. Wray solves this problem by showing that modern languages contain a holistic element, in some ways similar to primate signals. This is evident in those difficult little words like yes, no and thanks, and even in apparently segmented utterances. Idioms like how are you? can be analysed equally well as segmented or holistic constructs, and the speech outcomes of these exclusive analyses are indistinguishable. Wray describes use of holistic utterances as “performance without competence”, and shows that they are more common than we think. The segmentation of utterances into words is sometimes more illusory than real. The origin of grammar is not, in Wray’s view, a question of how separate meaning-units were built up into utterances, but of how holistic utterances became arbitrarily divided into separate meaning-units. Wray proposes two possible routes from holistic utterance to segmented language. The first is a slow increase in the number, range and use of analytic constructs throughout the history of protolanguage; the second is a slow evolution in cognition of the features which allowed analytic language to emerge, but with the actual emergence into speech being a single event. As the utterance of analytical language itself imposes limitations on the analytical structures possible, Wray favours the second solution, which allows the cognitive enablers of language to evolve unrestrained by the communicative limits of language itself. There remain issues with Wray’s analysis, not least her reasons why grammar appeared at all. She also does not address why language is so clearly overengineered for communicational purposes, or why it can so easily be used to produce lies. But her theory does offer a middle path between Chomsky’s and Bickerton’s unlikely catastrophic events and Pinker’s optimistic incremental evolution.

CO22552.indb 35

23/04/2010 14:13

36

th e orig in s of g r am mar

Is Generative Grammar an Inimical Environment for Language Origins? By placing Universal Grammar and language parameters at the genetic level, generativists are able to state that language rules are universal. However, they remain unable to describe the genetic process that created the human universal grammar engine, and so far they have been unable to describe definitively the structure of the grammar engine itself. This may well be due to insufficient time to study the area: the science may just be too young to provide all the answers (Baker, 2001, ch. 7). However, the generative view of language is based essentially on the idea that the universal parameters of language are small in number and powerful. Universal Grammar must therefore have a finite, and probably highly compact, structure. It is not a moving target, and it does not involve elusive components. True, it has to be studied through instances of actual language, which can only show Universal Grammar “through a glass, darkly”; but the number of actual language utterances is enormous, much greater than any single linguist could analyse in a lifetime—if universal structure remains mysterious, it is not for want of data. Continuing difficulties in discovering the universals of grammar and how they work must raise questions about their nature and even about their existence. At the heart of generativist linguistics is the view that language is optimized for cognition and not for communication, but in its cognitive role it is a near perfect system (Chomsky, 2002, pp. 105–109). Kinsella has questioned this approach, particularly in relation to the Minimalist Program (2006 as Parker, 2009). She has shown that some of the constraints on which the principles of Universal Grammar have so far been built are less simple and less universal than is necessary in a near perfect system. If the constraints are to be seen as universal then they must be seen as compromises and not near-perfect; and if they are to be seen as near-perfect then they cannot be seen as truly universal. There are contradictions at the heart of the Minimalist Program which need to be addressed if it is to be compatible with a Darwinian approach to language evolution. A major problem for generativist linguistics is that language changes: somehow the rules that transform thought into utterance in one generation or group are subverted and redefined by other groups and generations. This is a process known in linguistics as grammaticalization (Hopper & Traugott, 1993). According to generativist theory, changes to one part of the system should create cascade effects in other parts of the system, so that the parameterization of the new form of the language is consistent with the dictates of Universal Grammar. Yet the historical evidence seems to show no such heavy parameterization; instead we see in many languages a richness of valid forms, as each grammaticalization adds to, rather than replaces, acceptable grammatical forms. For instance, at this point we diverged has a similar meaning to we diverged at this point, and this is the point at which we diverged, and our divergence

CO22552.indb 36

23/04/2010 14:13

The Heavy Hand of Generative Linguistics

37

occurred at this point, and even we diverged here. All of these forms express the same idea and they have a semantic relationship which cannot be explained in grammatical terms. While other linguistic theories seek to explain this semantic relationship, Principles & Parameters and Minimalist theory are largely silent. Indeed, the traditional generativist interpretation is that the grammatical differences mean that they are different constructs, and cannot be co-analysed. Yet it would seem reasonable that a theory of linguistics should be able to address this; language is largely about meanings, whether being manipulated in a single mind or exchanged between minds, and it would appear reasonable for semantics to be central to any linguistic theory. Chomsky (2002, pp. 110–111) disagrees with this view, but he recognizes that he is ploughing a lonely furrow. Another question that generativism sidesteps is whether full grammar is always needed for language utterances. Words themselves encapsulate meaning, and they can work without a grammatical overlay. Indeed, some of our commonest utterances, like yes and no, live in a strange grammatical limbo; and an utterance like You! Here! Now! demonstrates that messages can be clear yet have no need for grammar. Idioms also have to be treated with care, so constructs like that’s something I wot not of (I have no knowledge of that) cannot be analysed in a standard generativist way. Local language variations also create problems: for instance I’m gonna go see he’s there (I am going to go and see if he is there) can be viewed either as aberrant grammar from someone who knows correct English grammar, or correct grammar in itself—or even as someone using the minimal amount of grammar needed to convey meaning. It could be argued that all of these difficult utterances show that surface structure must have a deep structure to be understood, but that sounds like the argument for phlogiston in early physics: things that burn must contain something to make them burn. Deep structure could be the phlogiston of linguistics, a position not incompatible with Chomsky’s Minimalist Program. However, we must also recognize that there do seem to be some true universals in human languages: nouns and verbs are one (however we name them), and the Subject-Verb-Object construction is another (although not necessarily in that order). These generalized language features point toward some form of universal basic structure behind language. There is also evidence from child studies that some bootstrapping device or natural sensitivity to language is already present in a child’s mind from birth. Generative Grammar therefore does have something to offer the search for the origins of grammar, and this book will not totally abandon generativism as a tool.

CO22552.indb 37

23/04/2010 14:13

4 Other Views on Language was to give linguistics the same scientific basis as the other natural sciences (Chomsky, 2002, pp. 56–60). There has long been a view among some scientists that their own discipline is doing “real science” about real things, and other disciplines are merely playing with data of questionable validity that do not produce credible, applicable knowledge. The divide between the natural, or “hard”, sciences and the social sciences is particularly hard-fought, and even today some physicists, chemists and biologists refuse to accept economics, anthropology, sociology and psychology as real sciences. If science is about the external physicality of what you work with then this is a valid viewpoint; if it is about how you approach your subject matter then it is not. Chomsky’s desire to make linguistics into a full science was a commendable aim, and necessary if linguistics was to progress beyond mere description. The generative approach led to new levels of meta-analysis: no longer was linguistics just about languages and their apparent differences, but about language as a human capacity. However, the emphasis by generativism on language as computation inevitably produces a computational theory of structures, forms and rules. Other approaches to linguistics have produced different outcomes. This chapter looks at the contribution of Systemic Functional linguistics, and some other functionalist grammar theories, to the origins of grammar debate; and it also reviews the importance of linear approaches to grammar. One of Chomsky’s aims

A System of Functions Language has purpose, it exists to carry out a range of functions: it enhances cognitive processing in a single brain; it provides a reliable conduit of information between people; it enables negotiation to agreed meaning between individuals; it expands the processing power of a single brain by co-opting brains around it; and, in the case of writing, it co-opts non-linguistic resources for offline storage and communication. These functions are not individually isolated, they are interrelated; they are parts of a system which makes language, in Darwinian terms, a fit strategy. Systemic Functional (SF) linguistics is a programme which aims to identify the full range of language functions and the system they occupy. 38

CO22552.indb 38

23/04/2010 14:13

Other Views on Language

39

The theories of SF linguistics have developed out of Halliday’s work in the 1960s (e.g. 1969). However, while the theoretical base of generative linguistics has changed several times over the years, leading to the abandonment of old theories and the development of new ones, the theoretical base of SF linguistics has grown incrementally. The seminal text for the programme remains An Introduction to Functional Grammar, first published in 1985 (Halliday) and now in its third edition (Halliday & Matthiessen, 2004). The ideology of SF grammar can be summarized as “semantic (concerned with meaning) and functional (concerned with how the language is used)” (Bloor & Bloor, 1995, p. 2). This description has correspondences with a rogue generativist theory, Lakoff’s (1971) Generative Semantics (GS); but the GS emphasis is on semantic structures, which are universal and deep and which produce meaning. GS is concerned with “nonsyntactic semantic regularities” (Ziff, 1960, p. 42) in language constructs, which produce meaning independently of the social and grammatical contexts in which they are produced. The SF view of meaning is related to the early semantic theories of Sapir and Whorf. Sapir was especially concerned with the effect of culture on language; he believed that culture had such a large effect that it actually created different structures in different languages (Sapir, 1921, p. 119). Whorf also took this view, saying “linguistics is essentially the quest for MEANING” (Whorf, 1956, p. 73). In many ways Whorf’s views were more extreme than Sapir’s: he took the view that culture affected not just language structure but the thought processes behind structure. Basically, culture affected language by changing how a person thought. In the SF view, language is more integrated into cognition than Sapir and Whorf suggest: language for SF is a set of cognitive systems which offer the speaker/writer choice in the expression of meaning. For instance, the constructs give me a drink, please, please may I have a drink, and drink, please are all ways in which a drink can be requested. On a semantic level their meanings are similar, but on a structural level they are very different. What the speaker/ writer is doing is not transforming internal language into external language, but expressing internal meaning in a convenient and clear manner while suppressing superfluous meanings. This is a major difference between generative and SF analyses: generative grammar starts from the principle that language is mostly a product of subconscious processes with only a thin overlay of choice; SF grammar sees language as mostly about choice. Thompson (1996, p. 6) says that transformational-generative grammar “does not reflect how the users themselves view language. They respond above all to the meanings that are expressed and the ways in which those meanings are expressed”.

Systemic Functional Grammar For systemic functionalists, the transfer and encoding of meaning occurs at the clause level; it is a product of both words (lexis) and grammar, and it is therefore

CO22552.indb 39

23/04/2010 14:13

40

th e orig in s of g r am mar

described as occurring on a lexicogrammar continuum. In some cases individual words carry the meaning, sometimes it is the idiomatic combination of words, and sometimes the meaning is a product of a specific combination of words (Morley, 2000, p. 21). For instance, the word cold implies a temperature which is less than ideal, but when it is in the construct this isn’t a cold fish there are several semantic effects at work. First, the term isn’t implies that the actual temperature is not less than ideal and may be greater than ideal. Second, there is an ambiguity in the construct which depends on what is viewed as the comparator: is a cold fish being compared with an uncold fish, or is it being compared with something completely different, such as a beefsteak? Is there a semantic relationship between cold and fish which renders them a single entity for this meaning construct? Or is cold fish being used in its idiomatic meaning of an unemotional person? This example shows how the context of the utterance affects its meaning and therefore its analysis. SF linguistic analyses are based around the modes of meaning that are conveyed in language. These modes are grouped into three metafunctions, Textual, Interpersonal and Ideational. The Textual metafunction describes the clause as a message, and is concerned with the words and forms chosen to convey meaning. The Interpersonal metafunction describes the clause as an exchange, and is concerned with the social transfer of information between minds. The Ideational metafunction describes the clause as a representation, and is concerned with the conversion of thought into utterance and vice versa (Halliday, 1994, p. 34). Each metafunction carries a thread of meaning in an utterance. They are not mutually exclusive, an utterance carries all three metafunctions simultaneously. The Ideational metafunction is subdivided into two further metafunctions: the Experiential, producing representation inside the clause; and the Logical, producing representation between clauses. The Logical metafunction is what makes SF analysis so powerful: the other metafunctions are all concerned with analysis of the clause, but the Logical metafunction is an analysis of discourse— how clauses work together. The logical metafunction is perhaps Halliday’s greatest gift to the linguistics community; his work on the other three metafunctions of the clause had been prefigured by Harman (1968, p. 68.): Theories of meaning may attempt to do any of three different things. One theory might attempt to explain what it is for a thought to be the thought that so-and-so, etc. Another might attempt to explain what it takes to communicate certain information. A third might offer an account of speech acts. As theories of language, the first would offer an account of the use of language in thinking; the second, an account of the use of language in communication; the third, an account of the use of language in certain institutions, rituals, or practices of a group of speakers.

Harman’s analysis gives a close, although not complete, correspondence to the Experiential, Interpersonal and Textual metafunctions.

CO22552.indb 40

23/04/2010 14:13

Other Views on Language

41

We can analyse the ways of asking for a drink to show how SF grammar works. The textual metafunction works in terms of theme (the key item in the utterance, usually the first word or phrase) and rheme (the rest of the utterance, supporting the theme). In the construct give me a drink, please, the theme is giving, establishing a task-related intention to the utterance. Compare this to please may I have a drink, which requests the receiver to satisfy the sender’s need rather than just perform an act. Drink, please emphasizes the object itself rather than the action or the sender’s needs. Interpersonally, the first construct establishes the receiver in the role of servant to the sender, the second places them more equally as working in a joint enterprise, and the last largely discounts the role of the receiver, emphasizing outcome over personal roles. Experientially, all three constructs are concerned with an actor (you), a process (supplying a drink) and a circumstance (satisfying my thirst), and all should result in the same actual outcome. The experiential context of the utterance, however, is all-important: using the wrong construct in some contexts may result in the drink being poured over my head rather than placed in my hand. Halliday’s multi-mode analysis may appear unwieldy, but it has proven productive. It brought back into linguistics the study of stress, intonation, tone and rhythm (prosodics and melodics), which are largely ignored by generativists; it placed emphasis back onto the discourse rather than the sentence; and it allowed issues of idiosyncratic language use to be addressed. For instance, the Logical metafunction allows SF grammar to identify a difference between parataxis and hypotaxis in discourse. A paratactical discourse is one where the clausal elements are linked at the same level by simple connection (e.g. I went upstairs and got my hat, and then I went to the market); a hypotactical discourse consists of clausal elements linked hierarchically (e.g. I went upstairs to get my hat so that I could go to the market). Tannen (1994, chs 3–5) has shown that this difference in discourse is identifiable in cross-cultural misunderstandings, and even in cross-gender misunderstandings. To say that men are hypotactic while women are paratactic is an exaggeration, but men tend to use and be more comfortable with hypotaxis than parataxis, and women vice versa. This is a discovery that could not have been made in a strict generativist tradition. One problem for grammaticians of all kinds is that of cognitive dissonance, where the message understood by the receiver does not match the intentions of the sender. How do these misunderstandings arise, and what part of the signalling process fails? This is not a case of the sender wishing to deceive the receiver, or of the signal being misspoken or misheard, this is a case of the communicative meaning engine breaking down. If language is a formal structure whereby internal representations are converted to external signals, then the only way dissonance can occur is if there are differences between the internal languages of sender and receiver. If, however, language is seen as a system of functions then it is possible for dissonance to occur in the cultural assumptions of the sender or receiver. Functional conflicts between the different metafunctions of the utterance

CO22552.indb 41

23/04/2010 14:13

42

th e orig in s of g r am mar

can mean that the message given is not what the sender intended to convey; or the different contexts that the sender and receiver bring to the utterance can create different interpretations. SF grammar gives us the mechanisms to analyse the causes and outcomes of these dissonances; generative grammar largely does not.

Other Views on Functional Grammar Systemic Functional grammar is marginally younger than TransformationalGenerative grammar, but it is part of as long-standing tradition of functional approaches to language, dating back to the Prague Linguistic Circle. This remarkable group of researchers established the basics of phonology, signal theory, semantics and semiotics, and they were the first researchers to consider linguistics as a rule-based science. The principles set down by the Prague Circle have inspired several interrelated research programmes, and they in turn have produced their own theory structures. While SF grammar is arguably the main theoretical base in functional linguistics, it is not the only one. One of the other approaches, initiated by Hervey in the late 1970s, is Axiomatic Functionalism, or AF (Hervey, 1979; Mulder & Hervey, 1980). This attempts to reconcile functional theories with the axiomatic approach of the generativists. The intention is to allow functionalism to use many of the same tools as generativism, such as hierarchical tree structures. Like generativism, AF is rule-based and looks for regularity and commonalities in utterances. Also like generativism, it is introspective, using judgements of what makes a well-formed construct to identify rules. Unlike generativism, however, it attempts to describe rule systems not just for syntax, not just for language, but for the entire semiotic process. It is an ambitious project. An AF analysis of the sentence the quick brown fox jumped over the lazy dog would work as follows. First, the nominal [fox] has dependencies of determiner [the] and adjective [quick] and [brown]; similarly the nominal [dog] has dependencies of determiner [the] and adjective [lazy]; the verb [jump] has dependencies of past tense [-ed] and preposition [over]; and it also has nominal dependencies of [fox] and [dog], thus forming both the sentence and the semiotic syntagm of the sentence. Like a generative analysis this is essentially hierarchical, with syntagms at lower levels combining to create higher syntagms; but where a generativist analysis is concerned with syntactic form, the AF analysis is interested in how meaning is generated. Problems for the AF approach come mainly from the large size of the project and the small size of the research team. Where generativism and SF grammar can pour teams of postgraduates over any theoretical glitches that may arise, Axiomatic Functionalism has fewer human resources and therefore longer solution times. The large size of the project and the small size of the team conspire to make progress in Axiomatic Linguistics slow. Functional Discourse Grammar (FDG) (Hengeveld & Mackenzie, 2008) is

CO22552.indb 42

23/04/2010 14:13

Other Views on Language

43

another approach which has potential but so far few resources. It is a relatively recent research project, but it has already established a dedicated following. The principle idea is that the metafunctions of SF grammar work together in a strict hierarchy to produce meaning. The interpersonal level represents the speaker’s intentions, and this governs the representational level where meaning is introduced. This in turn governs the structural level, where grammar occurs; and the structural level governs the phonological level where sound or writing happens. This hierarchy of language production differs significantly from that of the generativists in that grammar is a product of, and governed by, intention and meaning; meaning structure is more important than rule structure in the production of language. FDG, as the name suggests, is discourse-bound and not word- or rule-bound. This change of emphasis produces an analysis very different from both generative models and SF grammar: pragmatics, the study of the contexts and assumptions in language, comes first; then semantics, then grammar, then phonology. Analysis is a top-down process: the reasons for language production and the meanings involved must be interpreted before the grammar structures and the sign patterns used can be understood. Once again, what stands in the way of FDG is lack of academic popularity. However, despite being new it is already offering insights that other functional theories do not. It is an approach that we are likely to hear more from in coming decades. Danish Functional Linguistics (DFL) is yet another methodology in the functional linguistics tradition. It is largely based on the work of Hjelmslev (Harder, 1996), who was instrumental in bringing together the Copenhagen Linguistic Circle in the 1930s, a group inspired by the earlier Prague Linguistic Circle. As it is characterized by inclusion rather than dogma, the theoretical basis of DFL remains somewhat disparate. Hjelmslev’s emphasis on phonological form has been supplemented by semantic and pragmatic threads, and there have been some interesting multiple language analyses produced by DFL researchers. Although it does not have the same consistent theoretical base as the other traditions discussed here, it nonetheless provides a non-judgemental haven where good research can and does happen. It provides an exemplary research environment which is a model for other academic institutions. The last methodology discussed here is Functional/Typological Linguistics (FTL) (Noonan, 1999). This is a largely North American enterprise, although it is now being taken up by other linguists around the world. It has strong links to Cognitive Linguistics (see chapter 5) and identifies itself as being a non-structuralist theory. This places it in direct opposition to generativism, but also in opposition to the structural aspects of other functionalist grammars. In this respect it has some similarities to Integrationist Linguistics, which does not acknowledge a need for any systems in linguistic analysis (Toolan, 1996). FTL is a theory of linguistic processes. In the traditional Chomskyan

CO22552.indb 43

23/04/2010 14:13

44

th e orig in s of g r am mar

differentiation between linguistic competence (the capacity to produce language) and performance (the actual language produced), FTL must be viewed as performance-driven. Language is not an adjunct to knowledge, an engine to transfer information between minds; it is itself a form of knowledge, and indistinguishable in its working from other forms of knowledge. This view of language has inevitable effects on the approach to grammar. Grammar for FTL cannot be an extant thing which dictates the production of language, it has to be an emergent property of the use of language. There are universals in language grammar, but they are products of universal needs to share particular types of knowledge. The common cognitive mapping of the knowledge being shared imposes its own regularities on the sharing process, creating the illusion that the regularities are within language itself and not emergent from general cognition. According to FTL, this illusion has allowed linguists who are seeking rule systems in language to find them, even though they are not actually present. FTL therefore offers an important reminder that all language theories rely on basic assumptions, and these should not be allowed to go unchallenged. Popularity is not always a guarantee of accuracy.

Grammar without Tiers? The functional linguistic models discussed here by no means exhaust the range of theories available; many more exist (e.g. Lockwood’s Stratificational-Cognitive Linguistics, 2002). However, the five traditions described do demonstrate an important fact about the range of theories available: functional linguistics is a broad church. Yet there are some things that functional linguistic models do share in common. First, language is about meaning, and this has to be the primary driver when looking at language rules. Second, language is about signalling, so the relationship between sender and receiver has to be part of the language model. Third, language is about signs, so the rules of language must account for the ways that signs are manipulated in the production and comprehension of language utterances. So why are functional models “grammar without tiers”? There would seem to be as much differentiation of levels in functional linguistics as in generativist analyses. Generativism describes utterances in terms of order and dependencies, and uses tree structures to illustrate the hierarchical nature of those dependencies; functionalism describes utterances in terms of different threads of meaning which are being simultaneously expressed through the single utterance—and in the case of FDG, the threads have a hierarchical relationship to each other. If there are separate threads and hierarchy then there are tiers of meaning in functionalist analyses to match the generativist tiers of structure. There is one important difference, however: because of the inherent binary hierarchy of generative analysis, utterances must be deconstructed two-dimensionally; yet speech is, by nature, a one-dimensional vector of

CO22552.indb 44

23/04/2010 14:13

Other Views on Language

45

sound—and writing, even though expressed on a two-dimensional surface, retains this one-dimensionality. Although Functionalism does identify language as a multi-vectored object, it analyses the vectors separately as one-dimensional processes—essentially, grammar, or grammars, without tiers. The one-dimensionality of language utterances has, however, inspired several attempts to represent the final state in the production process. Linear grammars attempt to describe how multidimensional information is formulated as a single signal stream, and they therefore recognize the important unidimensional reality of language production in a way that other grammars can ignore. For that reason, a review of grammars would be incomplete without some examples of linear grammar.

Linear Grammars Generative grammar is a hierarchical description of nested constructs, definitely not linear. Functionalist grammars analyse utterances as a series of levels, or metafunctions; so, although each metafunction is essentially linear, they together form a multi-dimensional description of the utterance. In contrast, a linear grammar looks at the relationship between neighbouring segments of an utterance; and the relationship between non-contiguous linear segments is analysed only as a secondary step. This means that analysis can differ between languages: for instance johan hat marie im zug gesehen produces a different linear solution to john saw mary on the train. Linear grammars tend to be less didactic than other grammars about the need for universal or standard templates, and they do not set out to solve the entire problem of language in a single model (Croft, 2001, p. 202). Instead, they address some particular aspects and attempt to explain those. KathoI’s (2000) approach builds a linear adjunct to generativist binary hierarchy. He analyses sentences in both generativist tree structure form and in linear functionalist domains of meaning, and shows that a hybrid approach informs and enriches the analysis. His model drills down through the tree structure to the semantic units, and then applies a linear analysis to those units. It also applies a linear analysis at the clause level, reflecting the SF division of ideation into the experiential and logical metafunctions. Sentential coherence is particularly important in this analysis. Kathol looks at the utility of his hybrid approach in relation to Germanic languages: English, German, Swedish, Yiddish, Dutch, Icelandic, Danish and Norwegian. He finds that the linear analyses of each language have similar features, but they are not identical: linear grammar is language-specific. Kathol shows that apparent similarities in tree structure analyses of different languages can rely on a strictness of word order that actual usage does not have. The nature of the language, whether it is inflecting (incorporating grammatical markers like tense and case into words) and agglutinative (incorporating several meaning-units

CO22552.indb 45

23/04/2010 14:13

46

th e orig in s of g r am mar

in one word) like German or mainly isolating like English, affects the basic assumptions made in a tree structure analysis. English, as a rather extreme isolating language, may not be the best place to start when looking for universal language features. Bod (1998) uses a comparison of English and Dutch grammars to argue for a model he calls Stochastic Context-Free Grammar. He is concerned only with constructs which have a chance of occurring in real language usage; and he uses generative transformations to analyse non-terminal verb phrase strings into non-terminal plus terminal verb phrase strings. These transformations change the theoretical sentence, S, into a produced form, R. For Bod, the grammar of a language is probabilistic: some constructs are common and usually acceptable, some are less common but still usually acceptable, and some are uncommon and usually unacceptable. Using what he calls Data Oriented Parsing (DOP) techniques, Bod creates an analysis model at three levels: DOP1, where the common, acceptable generative transformations occur; DOP2, which uses the structure of known constructs to predict the structure of new constructs encountered; and DOP3, which uses probability analysis to predict meaning from structure. Bod’s approach is therefore semantic as well as syntactic, and it is the semantic interpretation that is done linearly. Although he relies on tree structures for DOP1 and DOP2, he finds it easier to analyse segments of meaning at DOP3 in a linear way. Bod recognizes two problems with computational analyses: the rule structure used by one individual may not be the same as that used by another; and the rule structure used in one context may not be the same as that used in another, even by the same individual. These differences can occur without apparent breakdown in communication between individuals; as long as utterances are within the receiver’s tolerances of comprehension, the detail of the uttered form need not be significant. Language grammar is therefore, at the level of analysis used by Bod, a probabilistic entity, and does not necessarily operate in the same way in every user. There may be no truly universal grammar in the way generativists propose (Bod, 1998, p. 145). Perhaps the most effective linear grammar analysis is given by Hudson (1998), in a system he calls Word Grammar. This sees language constructs as a series of heads controlling dependents. For instance, adjectives are dependents of nouns, and nouns therefore act as heads in noun phrases. Nouns also act as dependents in verb and prepositional constructs, creating an apparent hierarchy in the sequential word-bound constructs. However, this is actually a single dimension of suspended expectations, it is not a two-dimensional cognitive hierarchy being expressed in a one-dimensional medium. For instance, when we hear the partial construct, this is a matter of . . ., our expectation is for a noun, not a noun phrase as in the generativist model. The fact that we encounter an adjective, . . . great . . ., does not remove the expectation of a noun, merely suspends it: the adjective is a dependent of a noun head so it is

CO22552.indb 46

23/04/2010 14:13

Other Views on Language

47

acceptable in this position. If we had encountered an adverb, . . . really . . ., then our expectation would then be for an adjective, further suspending our noun expectation. Grammaticality is produced not by completion of a formal grammatical tree, but by fulfillment of our expectation about what should happen next. So, in the construct above, we see the construct as ungrammatical or incomplete until our expectation of a noun has been met: this is a matter of really great interest. Word Grammar deals with the single dimension of speech as a single dimension of meaning; it is not a two- (or multi-) dimensional grammar being squeezed into a single dimension. Currently, most of the work on Word Grammar is in English, but there have been some favourable outcomes when it has been applied to other languages (e.g. Creider, 2000). However, as it is case-driven and explanatory rather than predictive, there is no search for universal principles of language in Word Grammar. From the models reviewed here, it is clear that linear grammars tend to be data-driven: theory must emerge from the evidence, not vice versa. They all seem to accept that detailed grammatical structure is likely to be ad hoc and not subject to universal significance, and universal features of language are likely to be found outside the specifically linguistic grammatical structure as well as within it. Linear grammars remind us that sometimes the obvious structure is all the structure we need.

Functionalism and the Origins of Grammar Functionalist grammars are providing answers where generativism is silent. Partly, this is because speculation and discovery are encouraged by the inclusive nature of functionalism: it is open-ended, there are no heresies. Compare this to generativism, where each new level of theory has generated a range of paratheories, many of which have been condemned as non-canonical, or forced into apostasy by a change in the main theory (Harris, 1993). This has meant that many of the debatable issues in generativism have still not been fully debated. For instance, while phrase structure grammar is a good description of language structure it is a poor description of language production. If production followed the mappings produced by transformational generative models then we would expect the hierarchically lowest parts of the construct to be evaluated first, so that they can be slotted into the higher levels. This seems to be counterintuitive in view of what actually happens. For instance, if we take the sentence, This is the dog that chased the cat that caught the rat that ate the malt that lay in the house that Jack built, we would have to reverse-engineer it: we must evaluate the nature of the house (that Jack built) before we can evaluate the malt (that lay in the house that Jack built), and so on. This begs the question why we say it in the order we do; why not start with the part we need first? Halliday’s theme and rheme structure in the Textual metafunction gives a better explanation of how this construct works: each theme except the first—the cat, the rat, the malt, the house—is also

CO22552.indb 47

23/04/2010 14:13

48

th e orig in s of g r am mar

the rheme of the previous part of the construct. The regular structure, using the logical linker that, creates an expectation of semantic nestedness. We are able to imagine concepts containing concepts in a top-down way more easily than grammatical constructs contained by grammatical constructs in a bottom-up structure. However, one major problem with functional grammars for this book is that they make few predictions concerning language innateness or language genesis. There is no direct philosophy in functional linguistics which requires an explanation of these matters, unlike the Universal Grammar of generativism. Considerable work has been done in describing childhood language acquisition in an SF model (Peters 1995; Craig, 1995; Gaylard, 1995; Torr, 1997, among others), and the origins of SF grammar itself lie in Halliday’s analysis of his own son NigeI’s early linguistic development (Halliday, 2003). Many other researchers of child language acquisition, while not officially working in a functionalist framework, adopt a pragmatic approach based around evidence (e.g. HirshPasek & Golinkoff, 1996; Chiat, 2000). They find that the open-ended theories of functionalism tend to fit better with their evidential approach than the more constrained theories of generativism. As theories of language origins also tend to be pragmatic and evidential (although the evidence itself is necessarily indirect), they seem to fit better with functionalism than generativism. There is, however, remarkably little literature that addresses language origins from a strictly functionalist perspective. One thin volume of four papers (Benson & Greaves, 2005) does give a functionalist analysis of the capacities of nonhumans to produce human language. In particular, it looks at Kanzi and the bonobo group at the Yerkes institute, interpreting their communication in terms of the interpersonal metafunction, the ideational metafunction, the syntactic element of the textual metafunction, and phonology. The papers show that the metafunctions of human language have precursors in nonhuman signalling; and, in the case of human-language-trained bonobos, can be recruited to provide an effective bridge between bonobo and human communication. Bonobos may remain incapable of debating the relative merits of different linguistic theories, but they do have the capacity to reliably receive information from us and transmit their own information back to us. The apparent lack of interest by functionalists in the origins of grammar is, however, more apparent than real. Many researchers who use functional analysis methodology for discourse will use a related methodology, cognitive linguistics, when working with language in more abstract terms. Not all functionalists are also cognitivists, and not all cognitivists are also functionalists; but, because the theories start from similar assumptions and reach similar conclusions, the overlap is wide. As we will see next, cognitive linguistics does offer a lot to linguists ­interested in language—and grammar—origins.

CO22552.indb 48

23/04/2010 14:13

5 It’s All in the Mind The major linguistic theories , generativism

and systemic functionalism, seem to have only a cursory interest in language origins. This is to be expected, linguistics is traditionally the study of current language use, and historical linguistics is concerned only with languages for which there is written evidence. Speech leaves no record, so the forms of speech before the advent of recording are somewhat conjectural; and before writing they are highly conjectural. It would seem that the early history of language is not directly discoverable through linguistics. One branch of linguistics does have a direct interest in the origins of language, however. Cognitive Linguistics is a more recent discipline than generativism and Systemic Functionalism, but it already has a dedicated following. More importantly, it has produced some effective theories of how language is used, and strong evidence for those theories in terms of actual language usage; and it has much to offer us in the study of language origins. Cognitive Linguistics should not be thought of as an offshoot of Cognitive Psychology. Although they share a common adjective and a common origin, their trajectories have been somewhat different. They both advocate the scientific method of checking theory against reality by experiment, and they both recognize the significance of inputs and outputs in the study of cognitive processes. The emphasis for Cognitive Psychology, however, is on the mind as a computational device (Eysenck & Keane, 1995, p. 1); the emphasis for Cognitive Linguistics is on the mind as a meaning device (Geeraerts, 2006, p. 3). This is not a trivial difference in treatment, it produces different models of cognition: Cognitive Psychology sees the mind’s processes as largely modular and reactive, they interpret inputs and produce outputs in a largely predictable way; language, as a module of cognition, is separately examinable. Cognitive Linguistics sees language in the mind as distributed, with a degree of idiosyncrasy and unpredictability between individual minds; language can therefore only be examined concurrently with other cognitive processes.

A Short History of Cognitive Linguistics Although the principles of cognitive modelling of language have been around since the 1970s, it was in the late 1980s that it began to be recognized as a

CO22552.indb 49

49

23/04/2010 14:13

50

th e orig in s of g r am mar

valid way of doing linguistics. Before the 1980s, work had been done by Chafe (1970), Fillmore (1976) and Lakoff & Johnson (1980) on semantic systems in language; and by Langacker (1987 & 1991) on syntax. These research initiatives were, however, being pursued in parallel with each other and alongside the functionalist approaches discussed in chapter 4. What united these initiatives was both commitment to the view that language is a feature of general cognition, and opposition to the view that language is a unique and isolated phenomenon. In the study of language and origins, therefore, Cognitive Linguistics and generative linguistics have very different starting points. In the late 1980s the theoretical basis of Cognitive Linguistics began to be codified. Fillmore was working with Lakoff on the foundations of Construction Grammar, and Langacker published the first volume of his two-volume work on Cognitive Grammar (1987). Lakoff also published his key work on metaphor and embodiment in language, Women, Fire and Dangerous Things (1987). In 1989, the Cognitive Linguistics community began to coalesce, with the first conference at Duisberg in Germany, and the foundation of the International Cognitive Linguistic Association. The first community journal, Cognitive Linguistics, began publication in 1990. Cognitive Linguistics continued to grow throughout the 1990s, mainly due to two important developments. The first was the introduction of functional Magnetic Resonance Imaging (fMRI). With an image capture rate of up to 60 images a minute, fMRI allowed fast, simple and relatively cheap imaging of the human brain. It suddenly became possible to test traditional theories of linguistic cognition against what was happening in actual brains, and the results were not always clear-cut vindications of tradition. The first theory to be brought into question was the idea that language occurs exclusively in one part of the brain: the scans showed large parts of the brain involved in linguistic processing, and in both hemispheres. While there did seem to be some statistical specialization of brain areas, the specialist areas did not always map closely between individuals (Thulborn et al., 1999). Language cognition seemed to be less isolated from general cognition, and less universally organized, than generativism had predicted. This new evidence led to a development in generative linguistics of some significance to Cognitive Linguistics: Chomsky’s Minimalist Program. This, as we saw in chapter 3, was an attempt to redefine the universal component of language and make it a conceptual universal of the mind rather than a physical universal of the brain. Minimalism and its effect on generativism have already been covered, but its effect on other linguistic traditions should not be underestimated. It represented a significant retrenchment for generativism from the view that language is largely hard-wired and physically modular; and it therefore also vindicated some contra-generativist positions. Cognitive Linguistics has continued to gain adherents in the new millennium. Tomasello’s work on animal signalling (e.g. Tomasello & Call, 1997) and human child communication (e.g. Tomasello, 2003a) has been significant, as has the

CO22552.indb 50

23/04/2010 14:13

It’s All in the Mind

51

work of Steels on robotic communication (e.g. Steels, 1998). Croft (2001) has continued the codification of Cognitive Linguistic syntax, and Evans and Green (2006) have unified the different research initiatives in Cognitive Linguistics into a single volume. Cognitive Linguistics has now become mainstream and academically respectable.

The Nature of Cognitive Linguistics There are many threads to the Cognitive Linguistics initiative, each representing a particular cognitive approach to a particular aspect of language. In terms of the traditional divisions—semantics, pragmatics, syntax and phonology—Cognitive Linguistics has a presence in every camp. This means that the solutions that Cognitive Linguists propose can be problem-specific rather than universal. While Cognitive Linguistics is attempting to solve all of the questions that language generates, it is not trying to produce a single macro-theory to explain everything. Cognitive Linguistics has nonetheless uncovered some important relationships, both within language and between language and other cognition. For instance, Palmer (2006) shows how the cognitive models of language acquisition and usage have informed and enriched anthropological models of the acquisition of culture (and vice versa); and Evans (2005) has shown how a cognitive approach to the language we use to express time can illustrate how we think about and understand time. In keeping with the Cognitive Linguistics enterprise, language is seen as only part of the problem to be solved; and the solutions found are therefore likely to explain more than just how language works. The Cognitive Linguistics initiative can be characterized by two parallel aims. The first is that any description of language should be based on what we know about the mind and the brain, and not reliant on unevidenced speculative theory. The second aim is that linguistic evidence should be analysed in terms of what it means generally for cognition, and not just in terms of what it means for language (Evans et al., 2007). These two aims give the Cognitive Linguistics enterprise its own distinctive trajectory. Language features of particular interest to cognitivists are metaphor and metonymy. Metaphor is the capacity to identify systemic features from one model of reality and apply them to another, thus creating a new understanding of the second reality. For instance, the concept of interpersonal interactions can be applied to larger groups; so the idea that nations can have relationships with each other—such as “special friendships” and “understandings”—emerges naturally from the idea that groups are entities as well as being composed of entities. Metonymy is a specific form of metaphor, allowing objects to be arbitrarily associated with labels that are only indirectly referential. For instance, when I use the idiom cakehole to refer to the mouth I am relying on shared context, recognition of usage and common culture in my hearer to ensure faithful transmission of my intended meaning (Lakoff & Johnson, 1980).

CO22552.indb 51

23/04/2010 14:13

52

th e orig in s of g r am mar

Metaphor and metonymy are products of arbitrary associations between intention and utterance in language, implying that there are two cognitive mechanisms at work. The first is translational, between the concept and the utterance; the second is relational, between the intention and the concept. The intention to identify a mouth is related to its role as a conduit for eating; and the role of eating is related to the substance eaten, giving us the metonymous cakehole for mouth. The phrase shut your cakehole, however, refers to the mouth in its role as conduit for speech, making this particular usage metaphorical as well as metonymous. Language uses metaphor extensively, it is not just a poetic trope. For instance, in the last sentence, the word language is metonymous, I am using it as if it is an entity; but language as an entity is a cognitive construct, it has no actuality. There are certain physical relationships in the actual world that are so pervasive that they seem to provide a rich source of metaphor—Lakoff and Johnson refer to these as experiential metaphors. For instance, the metaphorical equivalence that MORE IS UP is a simple outcome of the observation that as piles get bigger they also get higher. The governor of this metaphor is the universal human experience of gravity. There is a concomitant, but weaker equivalence that MORE IS BETTER, which can be merged with MORE IS UP to give UP IS BETTER. This is represented in cognitively generated relationships such as heaven and hell; but it isn’t universal, MORE IS UP can take precedence. For instance a rise in crime or a growing fear are counter-examples of UP IS BETTER. Metaphor also has an effect on how we view grammar. Verbs, for example, can be treated as items with a functional commonality because of the relationships they have with other words in a sentence. So the verb phrases in John is a man, John saw a man, John greeted the man, and John moved away from the man can be viewed as sharing a common role, despite the fact that the relationship they establish between John and the man is, in each case, very different. But can we truly treat the reflexive nature of is in the same way we treat the interactive nature of greeted? This is an issue of interest to cognitive linguists. This metaphorical nature of word classes is something we often fail to recognize. For instance, the adjectives in happy teacher and drama teacher would appear to be in the same class; yet we can say the teacher is happy but not *the teacher is drama, and we can say teacher of drama but not *teacher of happy. Do we class drama as a descriptive noun? But then why is a drama teacher also a teacher of drama, but a University teacher is not a *teacher of University? This is only really explicable if the phrase forms are seen as rough applications and exaptations of rules created for other purposes. Metaphorical grammar does not propose idealized forms from which individual languages draw their particular actual forms, it describes mechanisms by which common communicative needs can be met in convenient ways (Dąbrowska, 2004, ch. 8). Although Cognitive Linguistics is concerned with language in brains, it does have a strong interpersonal aspect to it. If language is about human cognition, then the fact that humans devote a large part of their cognition to social

CO22552.indb 52

23/04/2010 14:13

It’s All in the Mind

53

interaction has to be accommodated in a description of language. Cognitive Linguistics recognizes that some meanings are the product of social definitions. For instance, Westerners see a timeline with the future ahead of them and the past behind them, but the Aymara of South America put the past in front (what is visible) and the future behind (what cannot be seen) (Núñez & Sweetser, 2006). This difference in cognitive mapping indicates that moving forwards into the future is not a universal cognitive representation of time, a social fact that is of vital significance to a cognitivist linguist, but of less significance to a generativist or functionalist. Cognitive Linguistics is also interested in grammaticalization, the process by which the grammar of a language changes over time (Hopper & Traugott, 1993). For the cognitivist, changes in a grammar are traceable to changes in word usage and meaning. For instance the use of the auxiliary will to indicate the future tense seems to come from its Middle English meaning of intend to; and the other future tense form, be going to, is an extension of the spatial meaning of go to the temporal dimension. Without accommodating the cognitive and semantic origins of grammatical forms it is difficult to explain the mechanisms by which language changes, and therefore why it changes. Cognitive Linguistics makes extensive reference to frames and schemata, an approach to metacognition originated in psychology (Eysenck, 1993, pp. 33–38). In a frames and schemata analysis, every utterance is comprehensible only in terms of its context (the frame), which includes the intentions of the sender, the interpretations of the receiver and their pre-existing shared knowledge. The items of knowledge shared in the utterance provide the schema, and these can be innate body knowledge, innate or acquired environmental knowledge, and acquired social or cultural knowledge. So when I use the phrase “I’m left-handed” I’m not just informing you about one of my hands, I’m giving you information about my hand preference for writing, the possibility of me having mental and physical traits statistically linked to left-handedness, my likely social stance with regard to handedness, my social conformity . . . a lot can be read into that short phrase. The frame may indicate solidarity (I’m in a minority, too), differentiation (I’m in a minority and you aren’t), capacity (I have appropriate dexterity to do a particular task), incapacity (all left-handers are naturally awkward), identity (this is one of my key traits) . . . a phrase has only trivial meaning in isolation from its context. Frames and schemata emphasize the communicative nature of language, and they subordinate structure to the needs of meaning. Cognitive Linguistics can therefore be seen as a broad church, in terms of theory and adherents. Unlike other schools of linguistics, it includes many scientists whose interests are not primarily linguistic. It is able to encompass any topic which has elements of language and cognition—but it remains, at base, a cognitive theory about language rather than a linguistic theory about cognition.

CO22552.indb 53

23/04/2010 14:13

54

th e orig in s of g r am mar

Embodiment Embodiment, an important aspect of Cognitive Linguistics, states that the brain is primarily a device for regulating the body: it controls hormonal flows and therefore emotion; it controls physical action through the nervous system; it has the capacity to model emotions and physical actions—the capacity to think; and, in the case of humans at least, it has the capacity to model the thinking process itself—to think about thinking, or metacognize. The brain performs most of its activity autonomously, using internal models of the body and the environment to act unconsciously or subconsciously; but it also appears to have a system for recursively modelling itself into models of the environment, creating an awareness of the self as both model and modeller, a process which will be examined further in chapter 7. For Cognitive Linguistics, cognition happens in the body and is all about the body. Any capacity we have for abstract thinking is an abstraction from our physical relationship with the world—our cognition is grounded in actuality (Johnson, 1987). This would seem to be an obvious fact, but many theories about cognition see it as an adjunct to the physical processes of the phenotype. Even Empiricism, which takes the view that all cognition is experience-based, identifies cognition itself as somehow different from the things thought about. This led, in Bishop Berkeley’s case, to the strange idea that the external world only has actuality if it is being thought about (Berkeley, 1710). Fortunately for Berkeley’s theory, God was available to do all the thinking to keep everything in existence. This mind-body dichotomy is insidious: while recognizing that we are clearly holistic objects, all parts of which are governed by the same DNA program, we nonetheless see ourselves as riding around in our heads, controlling our internal machine in the same way we control external machines. Cognitive Linguistics views this dichotomy as spurious: thinking and thinking-about-thinking are of the same nature and use the same systems. The fact that they feel very different from inside the thinking does not actually make them different. Cognition, at its most basic, is about controlling the body to improve its survival and reproduction. We have the cognitive processes we do because they improved the survival of our forebears and enhanced their reproductive opportunities. Cognition is a product of very physical, body-related processes, so we should expect it to be mainly about our physical selves—or embodied. We have already seen examples of this in the discussion of metaphor: the application of a personal experience of gravity to measure abstract quantities; and the positioning of the self in an abstract description of time passing. In fact, whenever we are faced with non-physical abstractions we rely on physical metaphors to allow us to share meaning. We have and exchange thoughts, we produce ideas, we seek agreement: it is all about what we would do if the abstractions were physical objects in a physical landscape (Gibbs & Perlman, 2006). Embodiment is key to understanding Cognitive Linguistics: it is the doctrine

CO22552.indb 54

23/04/2010 14:13

It’s All in the Mind

55

that separates it from other philosophies of language. If you wish to study language as a cognitive entity then there is no value in looking at language in isolation to identify rules and regularities; behind language (and containing it) is cognition, and behind (and containing) cognition is embodiment. This is where all searches for knowledge about language have to begin.

The Modularity Debate One issue that divides linguists—and cognitive scientists of all kinds—is whether the brain is structured as discrete modules, each with its own specialism, or organized in a more ad hoc way. Do particular aspects of cognition have their own brain areas or do they gravitate towards areas on a functional basis? Few people nowadays believe in the fully free-form brain of John Locke’s tabula rasa—and those that do have to explain away a growing tide of evidence for localized and specialized thinking in the brain (Pinker, 2002). Brain scanning studies have shown clearly that, for the large majority of people, a significant proportion of language processing occurs in the left hemisphere of the brain, in the regions known as Broca’s area, Wernicke’s area and the arcuate fasciculus (Beeman, 2005). The terms “large majority” and “significant proportion” do, however, point to the statistical nature of this evidence. At birth, the available communication systems seem to be generally distributed across the brain; but there then seems to be a concentration into the left hemisphere by age 5 in 95% of people. Of the others, most have language in both hemispheres, and a tiny minority have it in the right hemisphere only (Carter, 1998, pp. 46–47). There are indications that left-side lateralization is higher in right-handers than in left-, and higher in males than in females. This seems to make language look less like an innate and independent brain module and more like a functionally formed subsystem of general cognition. Large-scale brain modularity was first suggested by Fodor (1983). He predicted that the already-established modularity of physical control systems in the brain, such as for movement, vision and hearing, would be found to be part of a general cognitive modularity covering things like language, memory, selfhood and Theory of Mind. There was already evidence for this in relation to language: In the latter half of the 19th century, Broca and Wernicke had identified a correlation between particular language difficulties and lesions in the brain areas that came to be named after them. Other language problems were correlated with the angular gyrus and the arcuate fasciculus, implicating these brain areas in language production and comprehension. At the statistical level, Fodor’s weak modularity has proved to be correct: selfhood, theory of mind, and planning have all been correlated with different parts of the cortex. There do indeed seem to be certain functions that largely congregate in the same places in different human brains. The cortex does seem to

CO22552.indb 55

23/04/2010 14:13

56

th e orig in s of g r am mar

be less generalized and more modular than a blank slate approach can explain, but Fodor is careful to emphasize that plasticity of brain development is also significant. A stronger view of modularity has been taken by evolutionary psychologists. Cosmides and Tooby (1994) describe the brain as like a Swiss Army knife, consisting of a range of specialized components “each of which is well designed for solving a different problem”. In this model, the structure of the brain has been dictated by natural selection of individuals who are best able to meet the fitness challenges they encounter: everything in the brain is there for a reason, and we need to view the brain in terms of the problems it is able to solve. It is a commendable aim to ensure that cognition is explained in terms of evolution, but the strongly modular approach has problems, especially in relation to language. The first is exaptation: the purpose that a particular brain area is put to today may not be what it evolved for. For instance, dyslexia is a recognized brain-related issue, with the size of the left planum temporale area implicated in its severity; but reading and writing have only been widespread human activities for the past 300 years at most, the brain area cannot have evolved in the hope that one day humans would have widespread literacy. Another problem is stroke recovery: damage to the official language areas can severely diminish language capacity in the short term, and often in the long-term, too. Yet recovery is not rare, and some brain “rewiring” seems to occur without stimulation (Mosch et al., 2005). However, perhaps the most telling argument against Evolutionary Psychology’s strong modularity in the brain is its own claim that everything is modular. Humans, like all animals, come in a very wide variety of cognitive models. Natural variations which appear to create notable differences in brain function include handedness, sexuality, manipulative skill and imaging skill. Some people are good at cognitive skills like maths, others not. If all of these capacities are hard-wired into the brain then we would expect to see quite large, predictable variations between brains—it should be possible to identify an athlete’s brain, say, when compared to a painter’s brain. We cannot do this, and it seems unlikely that we will ever be able to do so. The argument for strong modularity remains, at best, unproven. The Cognitive Linguistics position on modularity is that, if it exists, it is weak and functional rather that strong and structural. Language modules would need to be leaky, and highly porous to external influence—because language is a subset of, and highly dependent on, general cognition. Language cannot be studied as an entity by itself, it is a region of cognition that shades into other regions, and which cannot be properly understood without including the “shadow lands”. Language did not evolve as a generative solution to a communicative need, it is an application of general cognition which has been exapted for communication, a way of exchanging ideas which proved to be a fit strategy. This strategy then became subject to the same evolutionary pressures that every strategy is subject to,

CO22552.indb 56

23/04/2010 14:13

It’s All in the Mind

57

and began to evolve under its own selection pressures. Any apparent modularity is a product of this post hoc evolution, there could not have been a separation of language from general cognition without losing important functionality.

The Nature of Cognitive Grammar Cognitive grammars have a distinct philosophy which sets them apart from other grammars: they are concerned with grammar as a tool of communication and not grammar as an organizing principle of language. While it certainly does organize language, this is just a secondary property of grammar; the primary function is to make the production and comprehension of meanings work. As with functional linguistics, there are several different models of grammar in Cognitive Linguistics; and, like functional linguistics, one model is becoming the tool of choice for linguistic analysis. Croft’s Radical Construction Grammar (RCG) (2001) is a direct descendant of Fillmore’s Construction Grammar, but it also incorporates features of Langacker’s Cognitive Grammar. It provides a consistent grammatical model within a Cognitive Linguistic framework; and for that reason it will be used here to illustrate the key features of Cognitive Linguistics grammars. Croft does not just present RCG as an alternative to generativist Minimalism, he places it in direct opposition to generativist theories. Far from seeing universal syntactic structures, Croft sees only diversity: grammar comes from the choices made in the production of an utterance, it is ad hoc and subordinate to the needs of meaning. Apparent regularities in the system are the product of structural metaphor and common framing: a structure which is useful to explain one circumstance is extended to cover circumstances with similar descriptive needs; and descriptions of less common events are shoehorned into pre-existing frames used to describe more common events. RCG therefore indicates that grammars do not need universal rules to work, regularities are sufficient to create shared meanings between sender and receiver. In this there is a clear closeness between Systemic Functional and cognitive ideologies, and a common distance from generativism. Croft differentiates between a component approach and a construction approach, a difference which affects the relationship between the syntactic and semantic representations of a sentence. In the component approach, separate syntactic and semantic components are linked by a lexicon; in the construction approach, the semantic and syntactic features are embedded in the lexicon. Where the component approach is deconstructive, attempting to deal with meaning, grammar and lexis separately, the construction approach is constructive, attempting to model them simultaneously. Croft shows how the component approach arbitrarily attaches words to word classes, even where they demonstrate features of two or more classes. Croft suggests that an open-ended approach, creating new word classes as needed, gives a better explanation of

CO22552.indb 57

23/04/2010 14:13

58

th e orig in s of g r am mar

what is really happening. He maintains that grammatical universals need not rely on universal “atomic” primitives, and instead proposes universal primitives in the interrelationships of language (Croft, 2001, pp. 47–61). This means that the proposed universals have a complex nature and not a simple one, and the atomic features of the complex primitives do not need to be the same in every language. To use an analogy, it is possible to see a car as a “primitive” of transport, it is a device for transferring people from one place to another; but the components of two different types of car need not be similar (one could have a petrol engine, one a diesel). The components themselves are only indirectly involved in the process being addressed, namely transport: a diesel or petrol engine by itself gets nobody anywhere, so it is useless to try to describe the process of transport in terms of engines. The idea that primitives can be complex is liberating in terms of syntactic analysis. The linguist is no longer searching for the basic particles of language and ignoring the functions being performed; instead, the function becomes paramount, and language production can be seen as an outcome of its functions and not its components. If one language appears not to have an adjectival word class then this has no universal significance, unless the language is unable to express adjectival qualities in any way. Similarly, if a language (like English) can express adverbial qualities in several ways (apply again, reapply, apply once more) this also has no universal significance, and the different methods can be attributed to new word classes if this proves useful. RCG enables Croft to take syntax out of the straitjacket of words and consider it as a conceptual space onto which individual language syntaxes can be mapped. To look for universals between languages is unproductive because each language occupies only a part of syntactic space. Instead, RCG attempts to define the limits of this syntactic space, and the exclusivities (where having one feature automatically prevents a language from having the other). He admits that this task is enormous and that he has only begun to scratch the surface, but an RCG approach does seem to be both more productive and more revealing than a Minimalist approach.

Cognitive Linguistics and Language Origins Of all the branches of linguistics, Cognitive Linguistics is the one with the greatest interest in language origins. For generativists, language is a discrete entity which is generated by a single evolutionary event; intermediate states are unlikely because language only works in its complete state. Although some generativists do propose part-way stages, such as protolanguage, those stages cannot have hierarchy or recursion (which would make them full language), and they can have only limited segmentation and differentiation (e.g. Bickerton, 1990). For functionalists, language is just one case on a continuum of communication systems, and there is nothing in the way language works that is not present in some way in

CO22552.indb 58

23/04/2010 14:13

It’s All in the Mind

59

pre-linguistic animals. There is interpersonal significance in primate grooming—it is given to specific individuals to establish social relationships with them; there is ideation in primate warnings—they are about specific threats and seem to signify the same things to sender and receiver; and there is textuality in diana monkey (Cercopithecus diana) modal warnings—if a warning call is preceded by a “boom” vocalization then it moderates the meaning from “threat” to “probably threat”, and the receiver alters their behaviour accordingly (Zuberbühler, 2000). In contrast, a cognitive approach to linguistics emphasizes the evolutionary nature of language. If there are apparent discontinuities in the evolution of language it is because the evolution story is not just about language; and if there are traceable precursors of language in other animals it does not alter the fact that human language needs its own evolutionary explanation. There are many researchers working on grammar origins within a cognitive framework. For instance, Johansson (2005, pp. 236–240) emphasizes the social structure of other primates as a model for grammatical structure in language. Baboon societies in particular have structure and hierarchy, and a hierarchy of individuals within a hierarchy of family groups—a limited form of recursion. Hurford (2007) describes how primate cognition seems to involve a system for generating nested meanings which uses segmentation, differentiation and hierarchy. He shows how this forms a grammatical stepping-stone between cognition and communicated human language. Heine and Kuteva (2007) use a reverse grammaticalization process to discover what early language forms were like. They see language as developing from one-word noun phrases to noun-verb constructs, and then on to two-argument, three-argument and then increasingly complex forms. Complexity was introduced incrementally to language rather than being fully present at the start. Tomasello (2003a) and his team have conducted extensive primate and child investigations to show that there are certain aspects of Theory of Mind that apes do not seem able to understand (such as pointing and attention), but which children appear to grasp from a very early age. Tomasello’s work indicates that Theory of Mind is a fertile area in which to look for differences between humans and other primates. Brighton, Kirby and Smith (2005) use computer modelling to show how language is itself part of the learnability process, and adapts to the prevailing culture to become easier to learn. These are just a few examples of how researchers using cognitive models of language are advancing our knowledge of how language came to be. Within linguistics, the cognitive approach is currently offering the most complete analysis of language origins—and, indeed, of the genesis of grammar. If we are to properly understand how grammar works then it is necessary to address why grammar exists: what functions does it perform that cannot be performed in any other way? What features of being human are specifically implicated in its existence? What structures does it use, and where did they come from? These are questions which associate general cognition with the origins of grammar, and which therefore should be amenable to a cognitive approach. There is also

CO22552.indb 59

23/04/2010 14:13

60

th e orig in s of g r am mar

an important change of emphasis between the cognitive approach to linguistics and other analytical methods. For the cognitivist the question is not “how does grammatical language define us as humans?”, it is “how did becoming human generate a need for grammatical language?” This question will be explored in more detail next.

CO22552.indb 60

23/04/2010 14:13

6 Being Human Humans (Homo sapiens) are members of a single species;

and every species is comprised of individuals who are physically close copies of each other, and who differ significantly from individuals in other species. This, to a large extent, is what the word species signifies. However, individuals within a species are not exactly the same—which is why genetic recombination makes offspring different from parents. To belong to the same species, the biological definition requires individuals to be similar enough to successfully interbreed, a testable measure of similarity. Nature, however, does not always subscribe to this classificatory system. For instance, the gull species who live around the North Pole have a complex genetic relationship: the European herring gull (Larus argentatus) interbreeds readily with its neighbour, the American herring gull (L. smithsonianus), and this in turn interbreeds with its other neighbour, the East Siberian herring gull (L. vegae). This interbreeds with its subspecies, Birula’s gull (L. v. birulae), and this interbreeds with Heuglin’s gull (L. heuglini). Finally, Heuglin’s gull interbreeds with the lesser black-backed gull (L. fuscus). Yet, despite both living in Europe, European herring gulls do not interbreed with lesser black-backed gulls (Liebers et al., 2004). Another example is the European white-headed duck (Oxyura leucocephala). This has become an endangered species, partly because of the introduction of the ruddy duck (O. jamaicensis) from North America. Not only are the two species able to interbreed, white-headed female ducks seem to mate preferentially with ruddy duck males (Henderson & Robertson, 2007). This is creating a major conservation issue, with ruddy ducks being subject to extermination programmes in Spain and Britain. The speciation of the Larus gulls is disputed, with between two and eight species being recognized by different authorities; and the Oxyura ducks, while being recognized as different species, are distressingly cross-fertile. So where does subspecies end and new species begin? At the genetic level, species as a differentiator can all but disappear. For example, the genetic difference between humans and chimpanzees (Pan troglodytes) is a fraction of our DNA (Marks, 2002), while the genetic variation within a single species of yeast (Saccharomyces paradoxus) is greater than that

CO22552.indb 61

61

23/04/2010 14:13

62

th e orig in s of g r am mar

between chimpanzees and humans (Liti et al., 2009). Because of this low level of difference between species and high level within a species, it is often difficult to differentiate between species using DNA alone; which is why the sequencing of the human genome, while a magnificent project, told us little about being human. It is the physical and behavioural differences produced by the genotype within the phenotype that explain the uniqueness of the human species.

Physical Differences So what are the attributes that make the human species different? If we are interested in the external physical features that differentiate us from our closest surviving co-species, the chimpanzee, then we must start with bipedalism. This increased our height, allowing us to see more; it enabled us (eventually) to increase our speed and endurance, allowing us to run faster and further; and it freed our hands from the task of walking, allowing them to become more flexible and accurate—more dextrous. Manual dexterity, an outcome of bipedalism, is another adaptation that allows many other new capacities to develop. It may be a mere side-effect of bipedalism, but it has its own developmental trajectory, enabling tool use, tool manufacture, accurate throwing, and a whole range of other useful skills. The original primary function for manual dexterity is, for our purposes, unimportant: what is important is the fact that it permits a skill set that is peculiarly human, and which our close genetic relatives can only palely imitate. Manual dexterity is a product of cognitive control over the muscles of the hand and arm, and it requires increases in brain capacity to accommodate this control. The part of the brain involved in motor control of the hand is known as the knob, and in most human brains it is located about 2cm above the ears, which places it close to the usual language areas. Like the language areas, there is asymmetry between left and right halves of the brain; but, in the case of the knob, this is related to handedness. The knob is unsurprisingly bigger in humans than in other primates, but it is easily identifiable in all primates and many monkeys (Hopkins & Pilcher, 2001). This asymmetry and proximity to language areas has raised the possibility of language being a by-product of dexterity (Frey, 2008), especially if prehuman signalling went through a manual gestural stage (Corballis, 2002). Indeed, many parallels can be drawn between dexterity and language. Manual activity requires the manipulation of discrete objects, just as language requires the syntactic ordering of discrete morphemes. Manual activity requires recognition of target objects and of action relationships between them, while language requires noun-verb constructs. Manual activity requires recognition that target objects can be products of action relationships, and of the capacity to plan production; language requires recognition that meanings at the sentence and discourse level are constructed from relationships between morphemes, not just by the morphemes themselves.

CO22552.indb 62

23/04/2010 14:13

Being Human

63

Yet the need for a manual gestural stage in language origins remains unproved, and evidence of continuities between chimp and human vocalizations continues to accumulate. It used to be claimed that chimpanzees have only limited volitional control over their vocal signals while humans have highly expressive vocal control, but this is now disputed (Slocombe & Zuberbühler, 2007). Ape vocalizations are controlled by the same face and voice brain areas in the left hemisphere as are human vocalizations (Reynolds et al., 2008); they seem to be subject to learning and local cultural significance, like human languages (Marshall et al., 1999); and, in bonobos (Pan paniscus) at least, they are used multimodally with gesture (Pollick & de Waal, 2007). While gesture remains an expressive mode of signalling in both apes and humans, there is no need to posit an abandonment and re-adoption of vocalization in the history of language genesis. It is much more likely that both vocalization and gesture are subservient to a cognitive need to communicate, and each is pressed into service as appropriate. After all, the modern phenomenon of writing is just another example of speech by gesture, and one that has not had time to develop genetically. Seeing writing as an exaptation of syntactic communicative gesture explains its explosive appearance simply and elegantly.

Manual Dexterity Is Social Dexterity? Manual dexterity is about more than just tools and signalling, however, it has an important effect on socialization and co-operation. First, and probably least, it enhances grooming: The finer finger control allows a range of new physical interactions. While chimpanzees tickle each other (Johnson, 2003), they do not massage each other, and caressing is not really part of their repertoire. This enhanced ability to “be nice” to each other is likely to have, in a small way, encouraged interpersonal co-operation in early Homo species. Second, manual dexterity creates the possibility of complex manufacture: humans have the ability to tie things together, or bind things to other things. A lot has been said about lithic technology, but stone-bashing is essentially about taking bits away to leave the final object; tying is about making separate objects work together as a new object, which often creates a dramatic technological improvement—for instance, a spear is about ten times more effective than a sharp stone by itself (Massey, 2002). This manufacture by binding may have had an effect on the way social groups were viewed: just as stones and sticks can be tied together to make a spear, so individuals can be tied together to make hunting parties. The capacity to make stones and sticks “co-operate” is extensible to the co-operation of individuals within tribes—a metaphorical exaptation from one domain to another. Metaphors, unfortunately, don’t leave a trace in the archaeological record, so this is unlikely to be properly testable. Another social function made possible by manual dexterity is shared manufacture. Humans have the capacity to work together to achieve a common

CO22552.indb 63

23/04/2010 14:13

64

th e orig in s of g r am mar

goal that could not be achieved by individual effort. This is a capacity they share with other primates (Wolkenten et al., 2007); and, like other primates, this capacity is governed by the ability to detect and deal with cheats (Fehr & Gächter, 2002). However, the enterprises humans can undertake together are unlike any others in the animal kingdom. Part of the reason for this is that manual dexterity makes co-operation reliable: the other party to the venture can be relied upon to be accurate in their supporting activity. For instance, the simple act of holding a post while someone else hits it can only work if the holder knows reliably that the hitter is going to hit the post and not their hand. Shared manufacture needs not just co-operative intent to work, there has to be the capacity to co-operate effectively (Gibson, 2002). All of these social outcomes of manual dexterity are likely to be involved in the enhanced co-operation that typifies us as a species. They are not, however, the likely proximate causes of that co-operation, or of the signalling system needed to sustain that co-operation. Being able to do something at the individual level does not automatically select for that capacity at the species level, the capacity has to give a reproductive advantage to the individual using it: it has to be an evolutionarily effective capacity.

Working Together One very obvious difference between us and other primates is the size of our social groups and the level of co-operation within them (Dunbar, 1993). While it is true that all apes live in social hierarchies, there are nonetheless important differences in the number and types of social relationships available to humans compared to other apes, and important differences in the ways an individual can gain status. Chimpanzee society is based around tribes of about fifty individuals, and it is divided primarily by gender: the males, who are mostly related, form close alliances of two or three individuals, and looser alliances of five or six individuals. The females are unrelated and more solitary, but they do have a social hierarchy based mainly on age and personality. Females seldom challenge their place in the hierarchy, waiting instead for death to move them up the social scale. Chimpanzees are patriarchal (ruled by the males) and mostly patrilocal (the males stay in their birth territory, the females move to new territories) (Goodall, 1990, ch. 2). The hierarchy of males is therefore a more earnest matter than that of females. Bonobos (Pan paniscus), although very closely related to chimpanzees, have a different social model. It is actually different in only one detail—it is matriarchal rather than patriarchal—but this difference has a major effect on the way bonobo society is organized. For bonobos, the unrelated females are able to suppress the aggressive excesses of the males by joint intervention. This solidarity of females against isolated males places the females, and their agenda, at the centre of

CO22552.indb 64

23/04/2010 14:13

Being Human

65

bonobo society (de Waal & Lanting, 1997, ch. 4). Chimpanzee society is based around the reproductive imperatives of the males, so the emphasis is on sex, with child-rearing subsidiary. Bonobo society is based around the reproductive imperatives of the females, so the emphasis is on child-rearing, with fertile sex (and males) as subsidiary. Surprisingly, one side-effect of this is that recreational sex is more common among bonobos: the reduction in the importance of reproductive sex seems to be accompanied by a more permissive attitude to sex in general. Where chimpanzees are only interested in sex when females are fertile, bonobos use heterosexual and homosexual sex as a form of social grooming and bonding. Sex is used to defuse conflicts and to establish alliances—and, sometimes, just as a greeting. Although human sexuality has, in many ways, greater similarity to that of bonobos than chimpanzees, we should nonetheless remember that the social structures of bonobos are closer to those of chimpanzees than those of humans. Like chimps, bonobos live in groups of about fifty, are patrilocal, subsist on a similar diet, have similar lifespans, and have similar body shape. Humans live in complex groups which can be numbered in millions, depending on the hierarchical scale being considered; we are patriarchal, matriarchal and egalitarian; patrilocal, matrilocal and fully nonlocal; our diet is omnivorous and highly variable depending on locality and social grouping; our lifespan is on average twenty years longer than chimpanzees and bonobos, and is marked, in the case of females, by an extended period of infertility in later years; and our bodies— bipedal, more gracile, largely hairless—are very different from our genetically close primate cousins. Human social structure also differs from that of chimps and bonobos in fundamental ways related to our different psychologies. Chimps and bonobos exist in an individualistic, Darwinian environment where genetic survival and personal promotion limit the range of social interaction; humans live in societies where an overlay of culture has mitigated the pursuit of genetic survival. We still operate under natural selection, and we have many characteristics that can be directly associated with genetic imperatives. Sexual selection, interpersonal competition, recognizing social rank, the existence of schadenfreude (joy at the misfortune of others), all are indicators of natural selection at work in humans. Yet we also have a set of group strategies for solving conflicts, addressing inequities and maintaining social cohesion; and we mostly seem to be happy adhering to the codes that these group strategies, or cultures, impose on us. We should not, however, see culture as something outside of evolution. We have evolved to co-operate in arbitrary ways because that co-operation is somehow advantageous to each individual. Although human culture appears to contradict Spencer’s survival of the fittest maxim (Spencer, 1898, p. 444), it exists because individuals who co-operated were better at getting their genes into future generations than selfish individuals. We co-operate because we have genes that make us co-operative; and we have co-operative genes because, in the case of

CO22552.indb 65

23/04/2010 14:13

66

th e orig in s of g r am mar

humans, they are reproductively more successful than selfish ones. This poses a huge dilemma for any theory of human origins—social, linguistic or grammatical: how can co-operative mechanisms overcome the natural tendency to succeed through individual selfishness?

The Problem of Culture Co-operation in nature is problematic because of the freerider effect (Boyd et al., 2003). Any co-operative behaviour works by providing advantages to all co-operating parties: everyone puts in a little and gains a lot. Unfortunately, this model is open to exploitation by cheats, those who do less than their fair share but still profit from the enterprise. These individuals should prosper because they are taking a larger net gain than the co-operators: cheats are both increasing the burden of investment for others and reducing their gain. If cheats prosper sufficiently then, over time, they replace the co-operators and co-operation collapses. The co-operators need a strategy to protect co-operation and disadvantage freeriding; and one possible solution is altruistic punishment (Egas & Riedl, 2005). If the freeriders can be excluded from the co-operating group, or punished for not doing their share, then co-operation can prosper despite cheats. Unfortunately, the mechanisms for detecting and dealing with cheats do not come cost-free, and this cost allows for new cheating mechanisms: individuals who co-operate in the original enterprise but do not punish freeriders will prosper more than individuals who pay both the cost of co-operation and the cost of altruistic punishment; so altruistic punishment will be selected against and, when it becomes too low, freeriders will go unpunished. Once again, the whole mechanism collapses (Fehr & Renninger, 2004). There is, however, a final mechanism which can make the rest of the system work. Individuals have different capacities, and therefore the cost they pay to co-operate, even though largely equal in absolute terms between individuals, is variable relative to the individuaI’s capacity. Those individuals who have a capacity to act which exceeds what they need to co-operate are clearly fitter individuals, and should therefore be selectively chosen by mates—if the mates can identify them. Altruistic punishment gives these individuals the opportunity to signal their fitness: not only can I co-operate but I am also fit enough to punish those who do not co-operate. Selective mating with these fit individuals will then spread the punisher’s genes through the population and allow co-operation with altruistic punishment to become the default individual disposition for individuals. Fowler (2005) shows that this will happen as long as the individual gains from co-operation are greater than from non-co-operation, and the individual cost of punishment is less than the cost paid by punished cheats. The problem of human co-operative culture therefore comes down to this: what makes co-operation more valuable than non-co-operation; what makes cheating relatively costly; and what makes punishing freeriders relatively cheap?

CO22552.indb 66

23/04/2010 14:13

Being Human

67

To answer these questions we need to look more specifically at the nature of human reproduction: what is costly (and cheap) in the way we produce the next generation?

The Costs of Reproduction Human reproduction is based around developmental principles that do not obtain in other apes: our infants are born in a considerably more helpless state, have a longer period of dependence, a longer period of childhood before sexual maturity, and a peculiar limbo state of adolescence. For this increased child dependency to have evolved, there must have been increased nurturing already available; but where could this have come from? The major costs of raising offspring—pregnancy and nurturing—fall on the female in most species, so one way to reduce those costs is to get males involved in nurturing. For instance, if the male provides food to the female then she has to do less foraging for herself. In a Darwinian universe, though, males are unlikely to commit to an individual female unless they feel the female is part of their genetic future—that is, her offspring are also the male’s. If a male is certain that a child is his then he should be willing to support it to ensure his genes get into the future; and males who support their own offspring will tend to be more successful reproductively because their offspring have an extra source of nurture. There are two ways to involve males in nurturing: paternity certainty and paternity uncertainty. With paternity certainty, the female ensures the male knows it is his child by truthfully signalling fertility and only mating with that one male while fertile. This is the strategy of gorillas (Gorilla gorilla), but in their case it does not lead to increased nurturing. Instead, males take on harems, which spreads the nurturing of a single male over several females and offspring—gorilla male nurturing is therefore limited to play and protection, there is very little provisioning. With the alternative strategy, paternity uncertainty, the female may not indicate her fertility (humans do not); or, if fertility is signalled, the females mate with as many males as possible to confuse the males about paternity. If every male has the possibility of being the father of all the children then it makes sense for all the males to provide at least some level of support for all the children. This is the strategy of chimpanzees (Pan troglodytes), but it does not lead to better provisioning for the females or children—precisely the opposite: because the commitment by each male to each child is statistical rather than absolute, children are a relatively cheap commodity. Bonobos (Pan paniscus), like chimps, use paternity uncertainty; but with the important difference that a female coalition is in charge. This means that bonobos are able to call on a second type of nurturing: female-to-female. Among bonobos there is greater female co-operation, both in food-sharing and in alliance against male aggression. There is also greater tolerance of the offspring of other females, although evidence of direct nurturing remains uncertain.

CO22552.indb 67

23/04/2010 14:13

68

th e orig in s of g r am mar

So how do humans co-operate to meet the costs of reproduction? Some anthropologists look to the Western monogamous pair-bond of one male and one female as a model for human evolution and origins. From this model they argue that a human female engages the close support of a single male by offering him absolute certainty that her children are also his (Deacon, 1997). This would be a reasonable model if the evidence supported it, but in so many ways it doesn’t (Clutton-Brock & Isvaran, 2006). Human females do not give paternity certainty at conception because their ovulation is concealed. They do not give certainty at birth because only 10% of babies are delivered on the due date—counting from conception doesn’t work. Some human cultures design paternity uncertainty into their societies by encouraging multiple matings through the partible paternity myth that it takes many fathers and many contributions of sperm to make one child (Beckerman & Valentine, 2002). Even in the pair-bonded West, estimates of the proportion of children being raised by men who are unknowingly not their biological fathers ranges from 9% to 30% according to one study (Baker & Bellis, 1995). If human males rely on paternity certainty to ensure their nurturing effort is not being wasted on another man’s children then they clearly need better verification mechanisms. If humans do not have paternity certainty, what is it that enables the high level of reproductive co-operation in our species? Female-female co-operation is one solution, and this is available to human mothers from both their own offspring and their own mothers. Human children remain subadult and therefore not involved in their own breeding programme for a few years after their foraging dependency on their parents has ended, which gives them time to support their mothers in the raising of their siblings (Sear & Mace, 2008); and human females cease to be fertile at about the time that their own children are giving birth, allowing them to help their daughters raise their grandchildren. Bonobos and chimpanzees also have a pre-fertile pubescence from approximately age seven, with first conception being at about age thirteen; but their patrilocality means that females move to new tribes away from their mothers at about age seven. They also have a post-fertile stage, but it is rare for a female to survive long enough to reach it. Humans would therefore appear to have greater opportunities to exploit their kin for child-rearing than chimpanzees. Opie (2004), however, shows that these forms of female-female co-operation have limited provisioning value and can only explain a part of human reproductive success—male co-operation must have been involved, too. It seems clear that, in order to raise our human, big-brained, helpless babies there has to be a high level of co-operative support available to the mother—more than can be provided by a simple pair-bond. How this need for co-operation and the co-operation itself co-evolved will be revisited in chapter 8. For now, we must leave this issue with the knowledge that human offspring require a high level of reproductive co-operation, and that this is likely to involve both females and males, and both the young and the old. It truly takes a village to raise a human child.

CO22552.indb 68

23/04/2010 14:13

Being Human

69

Beating the Cheats If human reproduction requires a high level of co-operation then it is clearly open to cheating, especially by males: a male can philander, getting females pregnant but not following through with childcare; he can freeride, relying on the efforts of others to help his children survive; or he can attempt to dominate, forcing others to support his reproductive effort at the expense of their own. All of these cheating strategies are used by human males, so it is not sufficient to treat co-operation as a genetic given: there must be mechanisms at work to keep most males co-operative by identifying cheats and punishing them. The first anti-cheating mechanism humans use is memory: we remember when someone has offended us and remove our trust from them. Of course, in order for this to work we have to have a default state of trust existing—we cannot withdraw what we have not already given. Evolution has therefore developed a series of hierarchical trust mechanisms in humans: we tend to trust strangers only enough to not ignore them; as we build up a history in memory of their reliability we trust them more and expand the joint enterprises we undertake with them; eventually we are willing to work together on enterprises where failure is highly costly to us. Essentially, we use memory to build a personal knowledge of an individuaI’s reputation. On top of this personal knowledge of the reputation of others, humans tap into the knowledge that others have about reputations: we share our knowledge of the relationships that exist between others, and that exist between others and ourselves (Dunbar, 1996). This sharing of information, though, has its own costs—and it therefore generates new ways of cheating: an individual can listen and not talk, increasing their own knowledge without revealing what they know; or, even worse, they can lie, creating false impressions of the reliability of others. How do we deal with this without adopting the nuclear option of distrusting everything we hear, thus rendering knowledge-sharing pointless? The first thing we can do, and indeed appear to do, is interpret no utterance, no information given, in isolation. We treat most of the social information we receive as neither truth nor falsehood but as expressing a position: Beth says Sue is friendly, Joe says she is not; what does this tell me about relationships between Sue, Joe and Beth? And, taking account of my relationships with Beth and Joe, what does this tell me of Sue’s likely reaction to me? Every item of social information represents a two- or three-argument form: it is about the relationship between two people, and it is given by one of the two people or by a third person; and every item is tagged with the reputations of the people, which are in turn built up from previous encounters with them or from previous social information. In this treatment, honesty is not an issue. Even downright deception, when detected, gives me useful information about the deceiver and their relationship with me. It is never worthwhile not listening. So what makes it worthwhile to talk, to give away valuable social knowledge

CO22552.indb 69

23/04/2010 14:13

70

th e orig in s of g r am mar

about others? One possibility is that sharing social information is a costly signal, a sign that I am able to give away this information without compromising my fitness. The purpose of this costly information-sharing, though, is to advertise my fitness, so it should be loud and public and it should be made preferentially in the presence of the opposite sex; this is not a good description of how we share social knowledge through language. Another possibility is that the giving of social knowledge indirectly benefits my genes by giving genetic advantages to my relatives. Once again, this does not reflect the way we use language: we talk to anyone, not just our kin. A third possibility is that social knowledge exchange is reciprocal, I give you information today in exchange for information you have already given to me or will give to me. This seems more like the way we use language, but even this does not capture the full garrulousness of humans: gossiping about others is not just a reciprocal duty, it can be a positive joy. Humans get pleasure out of talking and listening, which indicates that it is likely to have been to our evolutionary advantage to do so: pleasure is Nature’s way of telling you you’re doing something which has advantaged your ancestors over several generations, and which is likely to similarly advantage you. Getting pleasure from an activity makes the individual more likely to pursue it; and, if the activity advantages the individuaI’s genome, both the activity and the pleasure gained from doing it will prosper. So the pleasure we take in language, accompanied by its near-universality among humans and the relative success of the human species, together indicate that it is likely to have advantaged us as individuals. A major advantage of social information is that it lets us co-ordinate our efforts, creating co-operative enterprises that can achieve more than we can as individuals. Could this, by itself, be sufficient reason for language to have evolved? Could physical co-operation and communicative co-ordination have co-evolved, each feeding off the other in a ratchet effect? We have certainly seen this process at work recently: the physical co-operation in the shared enterprises of design and manufacture produced computers, which generated new ways of exchanging data, which produced the Web, which generated new ways of exchanging data, which produced the Internet, which generated new ways of exchanging data, which produced the blogosphere . . . each time the new way of co-operating enhances knowledge-sharing, which then creates a new way of co-operating; and, even despite the viruses, worms, trojans and other malware, co-operation continues to grow through enhanced e-communication. The ratchet effect does seem to have worked effectively in this case.

Making Models One important feature of sharing social knowledge is that it largely involves information from another time, and references people who are usually not present. It requires, therefore, that both sender and receiver are able to tokenize

CO22552.indb 70

23/04/2010 14:13

Being Human

71

their representations of others and manipulate them cognitively—essentially, they must be able to make mental models of other individuals, events and relationships. Humans are supreme model-makers: we are able to create entire worlds within our heads, and share them with each other. Not only can we model things that are no longer there, we can model things which are not yet there, and things which never will be there. Indeed, it can be argued that the whole of human culture is based upon a counter-reality where representation—the model—is more significant than the actuality (Knight, 1999). For instance, in the West we tend to count value in units of money, a fictitious and notoriously variable measure; but the comparison of relative values that money allows has a significant effect on our views of the world and of other people. The capacity to make models is not just part of our social understanding, the same mechanism seems to be involved in the human ability to make tools: the envisioning of an end product and the steps necessary to get there from the starting products are a form of modelling. Modelling should, however, be differentiated from planning. For example, if we look at chimpanzees foraging then we can see that they start from a simple state of being hungry. Hunger is a sensation, endemic in nature and unremarkable; if the chimp decides to move in a particular direction because fruit is likely to be there then that is planning; but imagining a juicy pear in your mouth is modelling. Planning works because it embodies a purpose, and there is no need for the end result of that purpose to be cognitively modelled: hunger is satisfied by foraging based on planning without the need for the outcome, eating, to be modelled. Chimpanzees forage, using their local knowledge to maximize their foraging effort, and so meet their generalized needs; but humans shop, modelling a future in which their wants have been satisfied in specific ways by specific items. For humans, what matters is not just the satisfaction of sensations, it is the conversion of models into reality. We are interested not just in having but in owning. Making models seems, therefore, to be a form of cognition at which humans are particularly adept (Sloman, 2008). It is involved in complex tool-making; in the mechanisms of human culture; and in the social calculus which determines, and allows us to navigate, the complexities of our social groups. Additionally, it uses the same system that we find in language grammar, of placing objects in relationships with each other. A mental model of a social event requires segmentation, to identify the actors in a model; it requires differentiation, to identify which actors are doing and which being done unto; and it requires hierarchy, both of time—to identify what happens first—and of space—to identify what happens closest. The mechanisms involved in mental modelling of social events are the same mechanisms we are looking for to explain grammar.

CO22552.indb 71

23/04/2010 14:13

7 The Weirdness of Self for clues to the origins of grammar, and similarities between the cognitive mechanisms of socialization and grammar were discussed. These similarities become explicable if, despite Chomsky’s view that language evolved for thinking (Chomsky, 2000, pp. 27–29), language actually evolved as a social mechanism to enable us to share cognitive social models. The chapter also explored the difference between planning means to achieve something and modelling the ends to be achieved, and suggested that extensive and complex model-making seems to be a peculiarly human preoccupation. Modern human socialization relies on models of the actual world that are subscribed to and enforced by the majority in a social group; they allow us to live in a reality (money, manners, ritual, conventions) which is an abstraction of actuality (food, sex, survival), and they allow us to create shared virtualities (abstract ideas, art, nations) which can themselves become realities if accepted by enough people as being so. Popper (1967, ch. 4) describes these as World 1 (actuality), World 2 (virtuality) and World 3 (reality); and he identified World 3, reality, as key to understanding human knowledge. Certainly, there is no other animal on this planet which has demonstrated the capacity to live in, and share, these three worlds simultaneously. To understand where cognitive social modelling comes from, we must look at the social structures we have to encode and manipulate in order to navigate the human social landscape. At base there is a need to recognize individuals, and to tag those individuals with emotional flags generated by past and present knowledge. This object-action differentiation maps to the noun-verb one-argument form in grammar. It also reliably allows for appropriate social reactions to be expressed— submission, dominance, irritability, tolerance—and therefore for alliances and other relationships to be built. This is a very basic capacity, though, and not limited to humans. Hurford (2003) shows that the predicate-argument (objectaction) distinction is neurologically based, and is a capacity available to apes and monkeys as well as humans. As well as the one-argument form, social calculus requires the two-argument form, linking two objects together with an item of knowledge which involves The last chapter looked at human socialization

72

CO22552.indb 72

23/04/2010 14:13

The Weirdness of Self

73

them both. For instance, understanding the simple concept that Beth likes Colin requires a mind able to bring two conceptualized individuals together and establish a relationship of liking between them. This is a more complex form of cognition than the basic one-argument form, and is probably not as widespread; but it does appear that chimpanzees at least can use two-argument social calculus, once again making it non-exclusive to humans (Tomasello & Call, 1997, ch. 10). This chapter looks at four interrelated types of cognitive modelling that humans use to understand and exploit their relationships with others and the relationships between others. First, there is modelling of the intentions of others: understanding others not just as objects to be manipulated but as beings with agendas. Second, there is modelling of events in the past and future, which allows items to be compared and collated outside of time, and which therefore allows reputations to be assembled and allocated to individuals. This involves not just remembering the feelings produced by events and projecting those feelings onto current situations, it requires the capacity to remember an event in terms of what happened. It also requires a disengaged view of the event, seeing the self inside the event as a third party. The third type of modelling is modelling of what-if: we can analyse not just future probabilities and past certainties, but events that are unlikely to, or did not, happen; and, most importantly, we can differentiate between what-ifs and actual events. And the fourth, and most significant, type pf cognitive modelling is modelling of our own self. The question of selfhood has long been recognized as vital for an understanding what makes us human. Even before Darwin’s theory of evolution by descent, von Humboldt (1836, p. 4) stated that “. . . the appearance of a greater individuality in individuals and in peoples, practically inexplicable by any derivation, interferes suddenly and without warning with the course more obviously determined by cause and effect.” Even earlier, Descartes (1641, pp. 79–86), proposed a duality of the physical automaton self, and the spiritual controlling self which was of a different substance; selfhood, in this theory, was a product of spiritual, not physical, cognition. In terms of embodiment, Lakoff and Johnson (1999, ch. 13) reverse Descartes’ dichotomy. They draw a distinction between the subject, which is the physical entity and largely unconscious of itself, and the self or selves, the cognitive constructions through which a person evaluates their physicality. It is the physical subject that is the essential person in their description, not the cognitive constructions. The modelling of self and others described here will use the Lakoff and Johnson approach, so Descartes’ famous dictum should be read as I (the physical me) think, therefore I (the modelled me) am.

Planning and Modelling The difference drawn in this book between planning and modelling is crucial in understanding the distinct nature of human cognition, so it will once again

CO22552.indb 73

23/04/2010 14:13

74

th e orig in s of g r am mar

be outlined here. Planning is devising a means which will make a particular end achievable, but it does not require the end itself to be envisaged. For instance, deciding to go to bed at the end of a hard day does not need us to imagine ourselves asleep, our bodies resting, our brains re-organizing. The sensation of tiredness is sufficient to put us through a series of pre-ordered activities which should end with us in bed, unconscious. In contrast, deciding to buy a bed does require us to imagine ourselves asleep; and, when we test beds, it involves the comparison of an idealized unconscious state with several consciously assessed examples. We cannot know ourselves asleep, but we can model ourselves asleep, and our plan (to buy a bed) can include that cognitively modelled endpoint. Planning does not require a modelled endpoint: we do not have to imagine ourselves asleep to go to bed. Other species can certainly plan. In particular, chimpanzees and bonobos are, like us, Machiavellian thinkers. They share with us the understanding that the actions of others can be influenced, that others are useful or dangerous, and that others have relationships with each other. To express this as a calculus, we have emotive mental constructs of our relationships with others which range from fear through to affection; and emotive mental constructs for all other relationships in our group. These mental constructs of relationships are used to plan behaviour. For instance, Colin fears Alf but has a good relationship with Beth, and Beth has a fair relationship with Alf. If Colin stays close to Beth, Alf is unlikely to attack him because Beth is more likely to support Colin than Alf, and Alf therefore risks unfavourable odds. If this were a description of chimpanzee behaviour then this planning could be described as a product of emotional tagging. Being alone with Alf is stressful, so Colin will avoid it. Being alone with Beth is pleasant, so Colin will seek it. Being in the presence of both is less stressful than being in the presence of Alf alone, so if Alf is getting too close then Colin will gravitate towards Beth. There is planning in this cognition, but there need be no conscious modelling involved in it. Colin does not need to imagine the outcome of being alone with Alf—in fact, the stress created by imagining an event can be similar to the stress created by the actual event, so why imagine it? Can other animals make cognitive models? There are some studies of primates in the wild which seem to show that they are making models of outcomes and planning towards them; but most of these examples can be explained equally well by operant conditioning—pursuing activities that result in reward and avoiding activities that don’t—and they do not necessarily need interpersonal modelling or even theory of mind (Byrne, 1995, ch. 9). However, there remains a small core of isolated examples from a range of primate species that do look like the employment of means to achieve envisioned ends (Tomasello & Call, 1997, ch. 7). There is also now extensive evidence that, when exposed to a human environment and trained, chimpanzees (Premack & Premack, 1983), bonobos (Segerdahl et al., 2006) and gorillas (Tanner et al., 2006) can also produce behaviours that are

CO22552.indb 74

23/04/2010 14:13

The Weirdness of Self

75

easiest explained as involving modelling of outcomes. Yet, although we cannot say that nonhumans do not model, we can point to the relative socio-cultural achievements of humans and other primates as an indicator of how different human modelling is in magnitude and effect.

Human Social Models If we look at the Alf-Beth-Colin example as a human relationship then we have to see it in terms of the ability to make models—because we know that this is what we, as humans, do. Colin is able to model the relationship between Alf, Beth and himself in terms of the separate relationships between Alf and Beth, Alf and Colin, and Beth and Colin. Two types of knowledge are needed: knowledge of how you (the immediate other) react to me, and knowledge of how others react to each other. These both involve second order intentionality, but of very different types: the first models your thoughts about my thoughts, and the second models their thoughts about their thoughts (Dunbar, 2004, pp. 45–51). In the first, your thoughts and intentions directly affect me; in the second, they only indirectly affect me through my own intentions towards others. As humans, we are each aware of our own and others’ individuality. However, I am aware not just of myself, and not just that you can model me, I am aware that others can have models of the relationships between me and you; and, most importantly, that the viewpoint of those others can be adopted by both me and you. I must simultaneously have an awareness of my own self-image, the image of me maintained by you and others, and the images of the relationship between me and you held from a possible fourth-person viewpoint (the person watching me analysing the relationship between you and me). Homo sapiens may be the only species that does this.

The Self and Language When reviewing the issue of self in language the first question that needs to be addressed is simply, what is the self? Or, to put it in more immediate terms, what is me? This is not the same question as the ubiquitous “who am I?”, which aims to identify an intimate self, mostly through introspection. “What is me?” attempts to describe the self as an externalized model—the viewpoint is not that of the interested self but of a disinterested fourth-person. Jaynes (1990, pp. 59–65) describes the modelled self as the metaphor “me”, in contrast to the analog “I”. Greenfield answers the question “what is me?” by reference to an always-present self and an awareness of that self through linguistic self-reference: . . . language gives us a symbol for something that normally does not make inroads into our senses, simply because it is always there: one’s self. As soon as we have a simple word for ourselves then we can inter-relate the self in

CO22552.indb 75

23/04/2010 14:13

76

th e orig in s of g r am mar

context. We can become self-conscious. This self-consciousness, combined with the ability to escape from the here and now, is surely what really distinguishes us from almost all other animals, as well as seemingly inhuman human infants. (Greenfield, 2000, p. 169.)

In Greenfield’s formulation the self in context (the metaphor “me”) is a conscious reflection, or model, of the true physical self (the analog “I”). We are able to describe our modelled self because it is a direct product of our conscious cognition; but we can only imperfectly describe our true self—others can see and describe our true selves more accurately than we can. Tomasello (1999, pp. 70–77) sees this process the other way around: it is our increasing self awareness in childhood that creates our knowledge of the capacities of others; but this implies that our knowledge of self should be more accurate than our knowledge of others, and this does not seem to be the case. Dunning et al. (2004) show that self-judgements on intelligence have a low correlation with real intelligence, completion of our tasks is always behind our estimated completion, and we are incurable optimists when it comes to our own health. Baumeister et al. (2005) show that our selfesteem does not match well with the esteem given by others, and it also does not correlate well with academic achievement. Benjamin Franklin said “there are three things extremely hard: steel, a diamond, and to know one’s self”. This would appear to be a reasonable view in light of the evidence. It seems likely, therefore, that we understand others better than we understand ourself, and most knowledge of ourself comes not from introspection but from modelling the minds of others and their intentions towards us. Self awareness seems to be better served by comparison than by introspection.

Selfishness and Self-awareness Cognition is costly: the human brain uses a massive 20% of our energy intake, a cost which can only be justified by the value of what it does. Any cognitive process must be viewed in terms of its considerable cost, and the existence of a cognitive capacity in a species indicates that there is a clear fitness advantage in paying the cost. The capacities to understand self and others must therefore be products of the fitness advantages they bring to the individual; but in considering the advantages we must remain aware of the high costs they also entail. Self-awareness is a conscious cognitive act for humans, but it is not the same as the self interest, or selfishness, that underpins the evolutionary process (Dawkins, 1989, ch. 5). Genetic selfishness is not concerned with activities at the phenotypic level but at the genotypic level. At the phenotypic level, genetic activities are expressed as autonomic responses to environmental stimuli. Genes do not choose to be selfish, they are selfish because only selfish genes survive: genetically inspired actions that favour the phenotypic self and its reproduction lead directly to genotypic survival, actions that favour other selves do not. There may be indirect fitness gains in favouring others, but they have to be advantageous

CO22552.indb 76

23/04/2010 14:13

The Weirdness of Self

77

enough to outweigh the direct path of non-co-operation before such altruism should appear as a species trait. There is no intentionality in Dawkins’ gene model of selfishness, any more than there is intentionality in a tornado: both are natural phenomena and controlled by fully explicable rules external to the phenomenon. Genetic selfishness is different from sense of self, which in turn is different from self awareness. Genetic selfishness is a default state which does not require consciousness. It ensures survival, and requires no more knowledge of the self than that the world is divided into self and not-self. What is inside the line, the self, is the ends; the rest, the other, is just means. With a binary model such as this only one of the pair of items needs to be identifiable to the self. The self would seem to be the easiest for the self to know, but it is also the least useful. The self is that part of the Universe that is already under control, so it is trivial; much more important is that part of the Universe that has to be manipulated and negotiated. A feature of genetic selfishness is therefore likely to be a lack of comprehension of the self. Sense of other gives immediate advantages, it allows an organism to subvert the survival of those others to its own purposes; sense of self gives no such immediate advantages. Yet, although sense of self does not give direct advantage to an individual, it does give indirect advantage: it allows an organism to exercise choice between strategies. In any situation there is usually more than one viable strategy of advantage to the organism, and the ability to choose effectively between them maximizes the advantage of each strategy. Adopting a single strategy for a situation relies on there being no organism with effective choices at the other end of the strategy; so as soon as a single strategy approach is met by a variable response (which an effective choice allows) then it ceases to work as effectively, and it can even become counter-productive. However, in order to make choices an organism must have a rudimentary understanding that there is a self to make the choices. This understanding does not have to be a conscious act, it need be no more than a recognition at the genetic level that the other half of the binary division into self and non-self exists. Where there is an understanding that there are choices, however, it can become advantageous to model those choices onto other organisms. If I come from a lineage that has been successful because it is able to make choices, then my immediate same-species rivals are also likely to come from that lineage. The ability to anticipate their choices is the next logical step in gaining a reproductive edge. It becomes advantageous to develop other awareness, a knowledge that others have choices that can affect my choices, and the ability to anticipate those choices. The model of the choices available to other organisms (and their possible responses) can be no greater than the knowledge I have of my own choices, so increases in the number of my choices lead to more sophisticated models of others. There are now two different types of knowledge, however: my choices are unconscious, they are selected by emotional bias and dictated by my feelings;

CO22552.indb 77

23/04/2010 14:13

78

th e orig in s of g r am mar

in contrast, my analysis of your choices has to be, on some level, a conscious act—I have to be aware of your choices in order to cognitively evaluate them. In analysing your choices I am trying to evaluate your intentions: which choice will you make? But I am not evaluating my own intentions—they emerge autonomically from my feelings about your intentions and do not need conscious analysis. With other-awareness, you become a modelled entity in my consciousness; but I remain unmodelled and not part of my conscious cognition. My model of you includes the fact that you have intentions; so I see you, to use Tomasello’s (1999) definition, as a mental agent. My model does not, however, include conscious empathy for you: not having a model of me, I cannot put myself “in your shoes”. Other awareness allows me to generate an increasingly sophisticated model of your intentions, but it provides only a limited toolbox to deal with those modelled intentions. It provides Machiavellian intelligence, but not the empathic social co-operation of human culture. With humans and language we have a new type of self to be recognized. Tomasello and Call (1997, pp. 337–338) describe this as self as social agent; Pinker (1997, pp. 134–136) labels it self-knowledge; and Bruner (1986, ch. 4) calls it the transactional self. Here, this final stage will be labelled self awareness. Somehow humans are able to extrapolate from making Machiavellian models of others to making models of ourselves, which allows us to conceptualize ourselves as if we are looking in from the outside. The picture we have of ourselves is often inaccurate, but the ability to generate it at all is an evolutionary conundrum: how have we become able to take a disinterested viewpoint of ourselves? Self awareness is a process that is implicated in what we do with language: essentially, it allows simple modelling of the intentions of others to become recursive models of the intentions of the self and others. The self becomes identified with the other to such a degree that their roles in cognition and language are interchangeable. I can see myself simultaneously in three ways: as the person generating an action, as the person to which the action is done, and as part of the context in which the action happens—I can occupy the first, second and third person roles in an utterance. I can also see you in any of the three roles, and both of us can be replaced in a construct by third parties: yesterday’s “you” becomes today’s “they”, which I can tell another “you” about. The traditional three persons of grammar do not, however, tell the whole story: all of this self-modelling within an utterance has to be done from outside the utterance—a fourth-person viewpoint which can model the self in all the other roles. The multiple viewpoints of the self permitted by self-modelling are prevalent throughout language. It is even possible to identify the roles in specific English usages. For instance, the apparently interchangeable reflexive forms, I hate me and I hate myself, seem to have different roles in the identification of the nature of self (Edwardes, 2003), with me representing Jaynes’ analog “I” and myself representing the metaphor “me”; and idioms like “if I were you” emphasise the role-taking that self-modelling permits.

CO22552.indb 78

23/04/2010 14:13

The Weirdness of Self

79

Four Selves Self awareness would seem to be a function of socialization. Malik argues that self awareness is intimately tied to language and social living—unless we have the knowledge that others have intentionality we can never have knowledge of our own intentionality: “The existence of a community of beings possessing language allows us to make sense of our inner world, and hence to become self-conscious. At the same time, I am only conscious of myself insofar as I am a member of such a community” (Malik, 2000, p. 220). This socialization dimension means that two further capacities become possible with self awareness. The first of these is anticipation, or a second-guessing between intentions: I know your options, and I know my options, so I should choose the one that gives me the best result in response to your best choice. But then you know my options and you know your options, so you may choose the option that gives you the best result in anticipation of my best response to your best choice, so I should choose the best response to that option. . . . There is a recursion between your intentions and mine within both of our minds, and this recursion is one of the capacities that enable language, as the Hauser, Chomsky and Fitch (2002) model predicts—although it is a product of pre-existing social modelling and not a feature limited to language as Hauser et al. suggest. The other capacity that self awareness makes possible is speculation on the intentions of others to each other, without the need for direct reference to the self’s own intentions. Other individuals are modelled not just as animate agents with linking relationships between them, but as mental agents with their own intentions. Modelling the intentions of others in not done to identify strategies which are directly useful to me, but simply to identify what is going on. It is this capacity that enables and informs the insatiable and disinterested curiosity of humans (Stewart & Cohen, 1997, pp. 163–164). So out of self awareness come these three further types of cognition: Reflexion, the analog “I” is separable from, and interchangeable with, the metaphor “me”; Anticipation, there is a recursion between the intentions of me and others; and Speculation, the intentions of others are modelled in language into both the recipient of the action and the instigator—roles become interchangeable. The four stages of mental modelling (sense of other, sense of self, other awareness and self awareness) take us from the reactive state of genetic replicators to the human ability to anticipate the thoughts of others. For Paul Bloom the last stage, self-awareness, is an important feature not just of being human but of language learning itself: children do not learn words by a process of association, they learn them by inference from the intended meaning of others. This means that children, when they begin to utter their first associative words, already have sufficient modelling ability to understand that the word-sign is a negotiation between them and other people. They also understand enough about intentionality to know that the meaning of a word-sign is in the intention of the

CO22552.indb 79

23/04/2010 14:13

80

th e orig in s of g r am mar

speaker (sender), and it is the role of the listener (receiver) to try to apprehend that meaning (Bloom, 2002, ch. 3). Children are able to attain a Theory of Mind because they are born with a theory of theory. They seem to understand implicitly the process of thesis-antithesis-synthesis which is the heart of human scientific method. They apprehend the world, make models of it, check those models against new realities as they arise and modify their models appropriately. Gopnik et al. (1999, pp. 155–162) call this “the scientist as child”, comparing the childhood modelling which builds adult competence with the human ability to continue modelling into adult life. Humans continue to play in the “mental sandpit” of modelling throughout their lives. Self modelling raises the issue of temporality: humans have an image of themselves as continuous with their past selves and future selves, but they are also able to see those past and future selves as if they were other people. This is something that non-linguistic animals are unlikely to be able to do: it is probable that they have a sense of the continuity of the self inasmuch as survival is its own testament to continuity, but it is a trivial sense which only serves to inform the current self. The trick of seeing time as episodic through the eyes of past and future selves requires a mechanism for identifying those past and future selves. Nonhumans can see time as passing before the self, but they are unlikely to be able to see the self as passing through time. To be human means being able to use second and higher orders of intentionality; it means being able to see others as mental agents, with a cognitive life as rich as the self’s; it means being aware of my own mental life as a metacognitive event—being able to think about my thinking; it means being aware that there is a me to be thought about, to be planned for, and to have unrealistic expectations about; and it means having the ability to create a model world inside my head which is as significant to me as the actual world outside my head. All of these features rely on a cognitive ability to make models of myself. This is a very unusual talent, and problematic in Darwinian terms: to make models of myself I have to step backward from myself: I have to try to view the “real” me from a fourth-person viewpoint. This means I have to be disinterested about myself, to try to see myself as others see me; and this is a skill that we are far from practiced at. Our self models are almost invariably wrong in material ways: we overestimate ourselves and delude ourselves about our abilities. There is one person in the Universe that we need to be totally honest about and, in large part, we cannot do it.

Awareness of Self Knowing that I am me, having a meta-awareness of my own existence, is not a trivial item of knowledge. The natures of the I doing the knowing and the me being known have exercised philosophers over thousands of years, and an early

CO22552.indb 80

23/04/2010 14:13

The Weirdness of Self

81

comment on meta-awareness was provided by Aristotle (350bce, bk 9, ch. 9): “being conscious that we are perceiving or thinking is being conscious of our own existence”. The ability to adopt the view of a self outside of time leads naturally to a belief in an eternal selfhood which is the observer, and maybe controller, of the self within time. The multiplicity of modelled selves permitted by speculation or what-if leads naturally to an uncertainty about the nature of the “real self”, and allows for the possibility that different modelled selves represent different ideal targets for the self to aim to become. The view of the self as second or third person creates the possibility that, contrary to our basic evolutionary instincts, other people are as valuable—or more valuable—than me. The interpretation of others’ models of me as being as valid as my own models of me adds to the number of targets to aim for, but also adds targets that are not necessarily in my evolutionary interests to pursue. If sufficient people challenge my own models of me then I am liable to come to the conclusion that the best me is the one that conforms to the majority stereotyping of me—and if enough people offer me a negative image of myself then I am more likely to adopt my own negative image of myself (Granberg et al., 2001). So being meta-aware of ourselves may be an empowering intellectually, but it can also be disturbingly unDarwinian in its effects on our personal survival. Yet within this acquiescence to the wills of others lies an important clue to human co-operation. To maintain a healthy self-image we must ensure that the images of ourselves that others are giving to us are positive; and the way we do that is to ensure that they think well of us. Our self-image is reliant on the reputation we build up in the minds of others, and doing things for others becomes part of the way we preserve our reputation and therefore our positive self-image. Altruistic punishment ensures that sociopathy—the ability to retain a positive self-image without reference to the opinions of others—is kept in check; but it remains a deep, genetic capacity in all of us, and ensures that our self-image is usually not fully reflective of our reputation. The constant interplay between what we know by introspection and what we know through received knowledge leaves humans in a constant state of doubt, not just about the external world but about our internal model of the world. It is this uncertainty in our models of our selves that makes possible the co-operation of human society and that makes necessary the social exchange of language. As Voltaire expressed it, “Doubt is not a pleasant condition, but certainty is absurd”. Self-doubt may be stressful, but for humans living in human groups, it is less stressful than unmitigated self-interest.

CO22552.indb 81

23/04/2010 14:13

8 How Did We Come to Be Human? (Stringer & Andrews, 2005, pp. 114–117) and 4.5mya (Byrne, 2000), somewhere in Africa, a species that was the precursor of the chimpanzees, bonobos and humans began to differentiate into two populations. One population took to, or stayed in, the forest and the other stayed in, or took to, the more open savannah plains. The first population, the Pan family, remained in Africa and developed into the modernday chimpanzee and bonobo; the other population was more nomadic and therefore more subject to allopatric (or geographic) speciation, in which isolated populations go their own evolutionary way (Jolly, 2009). The range of species and subspecies in this hominin population is therefore much greater than in Pan, and includes at least four genera: Ardipithecus, Australopithecus, Paranthropus and Homo. Eventually a branch of this family tree developed into modern-day Homo sapiens, which is the only remaining species of the hominin family. The physical story of how we came to be human remains incomplete, but we do know some things. We know that bipedality was an early adaptation in our ancestors, and dates to at least 6mya (Richmond & Jungers, 2008); and that early hominin brain size did not really differ from that of the Pan family, remaining at about 450cc until the genus Homo embarked on a threefold increase, which started at about 2.5mya (Dunbar, 1993). We know that tool use is a signature of our lineage, and that there were several technological breakthroughs: the simple stone choppers of the Oldowan, from about 2.4mya; the teardrop “ax-heads” of the Acheulean, from about 1.4mya; the flake technology of the Mousterian, from about 200kya (thousand years ago); the Upper Paleolithic technological revolution of about 40kya; and the Neolithic domestication revolution of about 10kya (Johanson & Edgar, 1996, pp. 250–261). In terms of art, There are indications of body adornment from about 80kya (Henshilwood et al., 2004), production of sculpted objects from 32kya (Conard, 2003), and depictive art (cave paintings) from about 30kya (Balter, 2009). Yet, despite all of these capacities being limited to the human lineage, they do not address what is particularly weird about humans. Foremost, we are a highly

Sometime between 6.5mya (million years ago)

82

CO22552.indb 82

23/04/2010 14:13

How Did We Come to Be Human?

83

socialized species, and that has implications for what and how we share: we share enterprises, co-operating in long-term undertakings on the understanding of our future right to a share of the gain; we share our social models, generating socially agreed reputations about others and about ourselves; and we have language to make all this possible (Key & Aiello, 1999). Our socialization is contractual, based around reciprocal altruism—an act of generosity received today creates the expectation and obligation for an act of generosity to be returned (Taylor & Day, 2004); and it is backed up by a system of altruistic punishment—we have property-related concepts such as theft and cheating, and we take revenge against freeriders (Chiappe et al., 2004). This altruistic punishment is itself a social act which occurs at the group level—we have rules and laws and we recognize ownership. It is not individuals punishing other individuals, it is the collective (or abstraction) of society punishing the individual. Because human individuals are mostly reliant upon approval by others to prosper, we seem to have developed a genetic predisposition for submission to the group: in human evolution, those who co-operated in social groups have largely outbred the loners (van Vugt & Schaller, 2008). This, in turn, brought us the advantages of specialization: I do not have to make everything I need myself because I can rely on the social structures of reciprocal altruism and altruistic punishment to exchange my labour for yours. This exchange may initially have been a continuous cycle of reciprocal giving, relying on individual reputations (Mauss, 1950); but, as the circle of giving widened, the reputation became symbolically stored into agreed units of exchange, or money. This leaves one further weirdness in which humans indulge, a trait we explored in the last chapter: we make models of ourselves. This chapter explores the way in which our objective self-imaging relates to our substantial capacity for altruistic punishment, and to our capacity to reify the group as an entity above ourselves. Both these traits pose problems for an evolutionary explanation of human origins, and for different reasons; but they both seem to have played an important part in the process by which we became human.

Altruistic Punishment Altruistic punishment is unusual in nature, but not unknown; and, where used, it is often a response to inaccurate or deceptive signalling (Tibbetts & Dale, 2004). In many cases, the honesty of a signal is maintained by its cost to the sender: the threshold at which the receiver accepts the signal is so high that cheating by the sender is too expensive (Zahavi & Zahavi, 1997). There is no need for altruistic punishment in this type of costly signalling. Costly signalling is useful where the receiver must make judgements based solely on the signal value (such as mating signals), but it is less useful for kin selected or reciprocally altruistic signals. In these cases, a “contract” outside of the signal means the signal can be reduced in cost without affecting value

CO22552.indb 83

23/04/2010 14:13

84

th e orig in s of g r am mar

(Bergstrom & Lachmann, 1997); and this is what we see in warning and social signals between group members, especially relatives. These cheap signals are open to cheating, however, so the receiver must be able to evaluate the signal/signaller combination rather than just the signal itself. Humans are exceptional in terms of signalling honesty. We use cheap signalling despite living in groups with limited relatedness; and we have complex societies dedicated to enforce signal honesty (Knight, 2008). Not all our signals have to be honest, we also value the shared fantasy of story-telling; but where signal/referent veracity is expected by the receiver and not delivered by the sender, we have systems of punishment available at the group level to use against the cheat (Gintis, 2008). As these punishments are initiated by the group, individuals must be pooling their resources to punish not just those who have offended against them personally, but those who have offended unrelated others. This group-driven form of altruistic punishment is difficult to explain as an evolutionary outcome: how does supporting unrelated others enhance my personal reproductive fitness? Yet it is what humans do; we seem to treat groups as if they are entities, and our membership of the group as an extension of our kin or, possibly, our own selfhood. There seems to be a cognitive representation, or metaphor, which allows us to treat the group in the same way that we treat individuals.

Metaphor in Cognition Metaphor, as a communicative device, does seem to be an exclusively human preoccupation. As a cognitive device, however, it is more common. For instance, mimicry relies on a mechanism through which inferences from one sensory event influence another (Wickler, 1965). Recognition of black-and-yellow striping, commonly occurring in nature, as being a sign of danger saves a lot of painful experimentation; and the discovery that local lakes contain crocodiles can, by extension, create a useful aversion to stretches of open water. These metaphors rely on personal experience, but they can also become encoded at the genetic level in a straightforward way: individuals for whom the metaphor is innate do better than those who need to learn it through experience. Metaphors of this type therefore existed as cognitive extrapolations long before humans first appeared on the planet. The communicative use of metaphor was a later evolutionary development than its cognitive use. Between minds, metaphor is a source of obfuscation and confusion (Searle, 1993); but inside a single mind it has practical superordinating, subordinating and co-ordinating functions. As a superordinating mechanism it represents the cognitive shortcut that the instance, x, is part of a related group of instances, y, so the features of x must predict the features of y (this elephant has big ears, so all elephants have big ears); as a subordinating mechanism it represents both the shortcut that the features of y must predict the features of x

CO22552.indb 84

23/04/2010 14:13

How Did We Come to Be Human?

85

(elephants are dangerous, so this elephant must be dangerous), and that a shared nature of x and y predicts that x is y (elephants have trunks, this has a trunk, so this is an elephant); and as a co-ordinating mechanism it represents that x shares some features with an unrelated object z, so it is likely to share other features, too (wasps are black-and-yellow and sting, this is black-and-yellow so it stings). All of these metaphor models are concerned with what works and not with what is logically true; their strength is that they provide working shortcuts that are correct in enough cases to give a fitness advantage to the animal that uses them. Metaphor is essentially a self-deceptive mechanism, a shortcut that allows immediate action based upon incomplete information. Sometimes, lying to ourselves is a valuable strategy, allowing us to react even when we don’t have all the necessary evidence to do so. True, over-reaction does incur a cost for the individual, but the cost of under-reaction may be much higher: better to spook at branches that look like snakes than to ignore the snake that looks like a branch.

THE GROUP IS AN ENTITY The way we are able to treat groups as if they are entities seems to be central to understanding how we became human. This understanding can be expressed, in the notation of Lakoff & Johnson (1980), as THE GROUP IS AN ENTITY, the key metaphor of human socialization. THE GROUP IS AN ENTITY is a co-ordinating metaphor: the GROUP shares some features with the unrelated single being, the ENTITY, such that one can stand for the other in cognitive modelling. The metaphor also has superordinating features, in that the individuals who make up a group are each entities themselves; so the attributes of the individuals predict the attributes of the group. The ENTITY concept is likely to be common in nature: cognitive segmentation of the universe is vital for many basic activities, such as identifying sexual partners, food and predators, so it will have been an early evolving cognitive function. In the notation of chapter 7, it can be equated with sense of other. It is also likely to have elaborated over evolutionary time, making it possible to classify entities in various ways on a range of interrelated scales; for instance, moving versus stationary, living versus nonliving; friend versus foe; in-group versus out-group. These four classifications have been deliberately selected to illustrate increasing sophistication of entity identification. Moving versus stationary allows the differentiation of things of interest (moving) from less important aspects of the environment (stationary); important things (sexual partners, predators, and sometimes food) tend to be mobile. However, important things also tend to be living, and to move differently to unimportant things; so it becomes useful to distinguish between different ways of moving. Living versus nonliving becomes a difference worth identifying (Corning, 2000). These two levels, moving versus stationary and living versus nonliving, are

CO22552.indb 85

23/04/2010 14:13

86

th e orig in s of g r am mar

basic, and most animals are able to make these distinctions. The next level, friend versus foe, is more sophisticated, and relies on being able to identify objects not just by rough class but as individuals. Social animals need to be able to classify other individuals in their environment as friendly or antagonistic; at least, it is important to do this for those living objects that have a long-term and material effect on the animal. Identification of an individual, however, requires the capacity to identify the characteristics that make them individual—the individual is recognized by a holistic bundle of sensory inputs (Tibbetts et al., 2008). So friend or foe, when it includes identification of individuals, requires a new level of cognitive complexity, corresponding to other-awareness in the notation of chapter 7. Nonetheless, most warm-blooded animals (mammals and birds) seem to be capable of identification at the individual level. This is certainly the case for social mammals, where the imperatives of feeding and breeding require the maintenance of a complex set of relationships with other individuals, as well as the cognitive overhead of individual identification. The Machiavellian intelligence of the apes (Byrne, 2000) adds a new level to this other-awareness: there is a clear fitness advantage in not just identifying individuals and attributing characteristics to them, but in manipulating those individuals to the advantage of the manipulator. The last cognitive classification of entities, in-group versus out-group, or Us versus Them, is the most problematic; not least because, as humans, we see it where it is unlikely to actually occur. For instance, in eusocial insects we can see behaviours that seem to be describable as group recognition, consensual voting and even cultural conformity. It is, however, very unlikely that these behaviours are actually products of conscious cognition. Instead, they seem to be emergent properties from convergent individual genetic imperatives (Bourke & Franks, 1995, pp. 56–66). For true classification of in-group versus out-group there has to be cognitive recognition by the classifier of the group itself (de Waal, 2006, pp. 52–58); but the evolution of this group recognition is not easy to explain in terms of Darwinian fitness.

Where Does the GROUP Come from? The recognition of the concept GROUP would seem to be uncomplicated: things occur in multiples, and can be handled either as individual items or as a single mass—just as a bunch of grapes is both a group of individual grapes and a single bunch. This concept of GROUP is, however, external to and does not involve the individual. If we consider GROUP in terms of in-group versus out-group then we have a very different and much more complex concept to explain. First, the concept of in-group requires a knowledge of the self as an entity: there is an object, group, and another object, self, and the group contains the self. However, the self-object is, in a vitally intimate way, also me. It is me both as a cognitive model made by me of myself, and as the person who is generating

CO22552.indb 86

23/04/2010 14:13

How Did We Come to Be Human?

87

the cognitive model. Basically, I am viewing myself in a way that allows me to be dispassionate about myself—to view myself as if I were an external third person (Jordan, 2003). This is an evolutionary conundrum: where is the fitness in being dispassionate about yourself when all others around you are passionately defending their own corner? An objective viewpoint of the self would seem to offer a ticket to genetic oblivion; and yet, somehow, this capacity has become a defining genetic feature of our species, to the point where failure to exhibit sufficient skill in it is seen as pathological—the autism spectrum (Jordan, 1998) and sociopathy (Pitchford, 2001) being obvious examples of this pathology. What mechanisms allowed self-objectivity, albeit not fully accurate self-objectivity, to become part of our species-defining genetic make-up? Self-objectivity does help to explain how we objectify the group, though: if the self is seen as a third party, it is possible to treat it in the same way as any other third party—it can be seen as part of a group. The objective grape is part of a bunch, the objective me is part of a tribe. There is no fitness implication in this unless the tribe is seen not just as a product of the individuals that make it up, but also as a superordinate of those individuals. If the tribe is reified and anthropomorphized as being itself an individual, then the imperatives of the individual, me, are subordinate to the imperatives of the entity, the group. Or, to put it another way, my actions become subject to group sanctions, not just at the group level but in a self-censoring way, too. I begin to see myself as no more important than any other member of the tribe, and must be willing (and cognitively able) to subordinate my needs to those of others. The subordination of self in the emergent metaphor, THE GROUP IS AN ENTITY, is a startlingly un-Darwinian thing to be willing and able to do. There must, therefore, be a countervailing fitness that makes self-effacement (or morality) a good strategy for the individual. Group living does convey all kinds of advantages and is widespread throughout nature; but the particular socialization that makes self-sacrifice a viable strategy is rare. The eusociality of ants, bees, wasps and termites is an obvious example: it works by reducing the reproductive capacity of individuals, sometimes to zero, thus making them reliant upon the breeding success of close relatives to get their genes into the next generation. This is not the case in humans: we retain our own full reproductive capacities, yet somehow we have generated a pseudo-eusocial group structure that allows us to work together on complex projects. This co-operative capacity appears to have become established at the genetic level: we are happier in groups than by ourselves, we are happier working with others than working alone, and we are happier conforming to group norms than rebelling (Pressman et al., 2005). We cannot put this down just to our niceness, our nastiness, our willingness to altruistically punish, must also play a productive role.

CO22552.indb 87

23/04/2010 14:13

88

th e orig in s of g r am mar

Altruistic Punishment as an Engine of Socialization Altruistic punishment is another conundrum in evolution. What is the fitness advantage in punishing others? To mete out punishment we have to give up time and resources, and we also have to be willing to accept any costs caused by the punished individual fighting back (Fowler, 2005). Obviously, punishment works best when a group of individuals gang up on the “sinner”, but this raises the mutant problem. All evolutionary change must start with a single mutant; but if a single mutant altruistic punisher arises in a population they will be paying all the punishment costs when others pay nothing, so they are compromising their own fitness and advantaging the fitness of others—how do their genes successfully come to dominate in a population, and thus create team punishment? (Fehr & Fischbacher (2005.) Nonetheless, altruistic punishment has been observed in several species, and the more socialized the species, the higher the level of punishment. It may be that altruistic punishment is not an engine of socialization but a product of it. If an individual needs group membership in order to thrive then the mere withdrawal of that membership may be sufficient to punish the individual. For instance, if grooming is necessary for good health, and individuals in a group groom only individuals they like, then all that is needed is a mechanism whereby individuals view social sinners unfavourably. The sinners get less grooming, and getting their genes into the next generation becomes that little bit harder. A simple mechanism of Vigilant Sharing (Erdal & Whiten, 1994)—each individual being aware of their share in terms of the shares of others—accompanied by low-cost social ostracism can be sufficient to create an environment where stronger altruistic punishment can appear. This Vigilant Sharing social environment does not favour alpha individuals, whether male or female. Any individual who attempts to seize more than their share of the resources is likely to suffer a reduction of fitness caused by their ostracism, reducing their capacity to act in an alpha role. This Vigilant Sharing can then lead on to Boehm’s (1999) Reverse Dominance, where alphas are suppressed by active group punishment, and where being modest about altruism becomes a costly signal of fitness: I am so fit that I don’t even need to advertise my generosity. Modesty is perhaps the most unusual form of altruistic punishment: the individual is punishing (or censoring) themself in order to prove their fitness as a member of the social group. They are making a costly signal by showing themself able not just to put the needs of the group above their own needs, but to do so in a way that values the group needs as greater than their own—they have so much spare capacity for the group because their own needs are a trivial draw on their resources (Barclay & Willer, 2007). The altruistic punishment inflicted on social cheats is also likely to be inflicted on signalling cheats: deceptive signals will not just be ignored, they will

CO22552.indb 88

23/04/2010 14:13

How Did We Come to Be Human?

89

be positively discouraged. The high level of socialization allowed by Reverse Dominance means that signal honesty becomes particularly significant, and deception has the capacity to create significant losses for the group. In eusocial insect species such as the hymenoptera and isoptera (Queller, 1994), and in mole rats (Heterocephalus glaber and Cryptomys damarensis) (Burland et al., 2002), deceptive signalling is rare and usually punished. There is, however, usually a single dominant fertile female per group in these species and therefore high relatedness between group members; and relatedness is itself a powerful Darwinian incentive to keep signals honest. The power of altruistic punishment can be harnessed to enhance socialization only in very limited circumstances, and only where a level of socialization is already present. It cannot, by itself, generate socialization in an unsocial species, and it is not a fit evolutionary strategy in those circumstances. Once harnessed, however, altruistic punishment can move socialization to levels of co-operation that create some very un-Darwinian effects in the individual, such as self-sacrifice.

THE GROUP IS AN ENTITY: building social structures The metaphor, THE GROUP IS AN ENTITY, is both a product of social structure and a source of it. The cognitive concept of GROUP is only relevant if it is part of the individuaI’s daily experience. Yet being part of a group need not be cognitively recognized by the individual in order for the group to exist. For instance, it is not necessary for eusocial insects to have any concept of nestmates or nest in order to work together in what appears to be a highly organized way (Hölldobler & Wilson, 2009, ch. 3). Insect eusociality can even produce the illusion of hierarchy and central organization—partly because, as humans, we metaphorically extend our own pseudo-eusocial models to explain full eusociality. The group can thus emerge as an entity from individual co-operations, and does not need to be cognitively recognized by its participants; but what can bridge the gap between the unrecognized emergent group and the recognized group that we humans clearly have? The first step is probably the maintenance of individual accounts of co-operation, to identify those individuals who are regularly co-operating with me and those who are not. This becomes a necessary mechanism in any species with a level of cognitive control over whether they co-operate or not: cognitive control creates the possibility of freeriding by others, so co-operative accounting becomes necessary for me to identify those freeriders (Gardner & West, 2004). The next stage of co-operative accounting is to identify not just how individuals co-operate with me, but how they co-operate with each other; and this opens up the possibility of forming alliances based on mutual interests. These alliances can be ad hoc, changing as the web of co-operation itself changes; so individuals with the necessary intelligence to negotiate this web of social commitments will tend to do better than less Machiavellian individuals. There is some

CO22552.indb 89

23/04/2010 14:13

90

th e orig in s of g r am mar

group awareness in this kind of cognitive modelling, but it is merely a representation of the boundaries of the group: it divides the world into those individuals who have a place inside my social modelling and those who don’t. There is no representation of the group as an entity of itself. When THE GROUP IS AN ENTITY does become a cognitive concept, however, it creates a new world of possibilities. The group no longer needs to reflect physical reality, it can become a concept purely of the imagination—it becomes a Popperian World 3 entity given reality by consensus. The identification of in-group and out-group may appear logical and reasonable in these cases, but it is often based upon arbitrary mutuality. Why do individuals who are good at pushing bits of wood across a tessellated board according to arbitrary conventions need to group together in chess clubs? What fitness advantages do they get? Seeing the group as an entity allows a society to consist of not just one group but many sub-groups, and membership of one sub-group does not automatically preclude membership of another. Treating the group as an entity can make conformism to the arbitrary rules of a single group a fit strategy, but it can also make membership of multiple groups a fit strategy. The arbitrary meta-rules of culture determine the nature of the groups available for individuals to join, and the arbitrary rules within the group determine which individuals will join them. The appearance of THE GROUP IS AN ENTITY as a cognitive concept, therefore, creates the possibility of a very different kind of social structure. Individuals who can manipulate the concept will have an important advantage over those who cannot, being able to create alliances in new ways; and they will also be more skilled at negotiating the new and more complex social web that the concept creates.

THE GROUP IS AN ENTITY: an ancient metaphor? The capacity to use metaphor in general cognition is unlikely to be a purely human trait; it offers useful cognitive short-cuts that provide tangible advantages even for simple life-forms. Some metaphors must, therefore, have their origins in pre-human cognition. For instance, the classic metaphor MORE IS UP is a simple product of gravity: as a pile grows it also rises. It therefore becomes a fit cognitive short-cut to identify tall things as more and short things as less. The metaphor, THE GROUP IS AN ENTITY, does not have a natural relationship to the actual world, however. Indeed, it seems highly counter-intuitive in a Darwinian Universe. This alone argues for it being a very recent occurrence, and probably limited to the human lineage. Subordination of the self to the aims of the group is explicable in eusocial animals, but not in species with individual breeding rights, no matter how socialized they may be. There are small hints that other primates may have some type of group recognition, and the friend versus foe metaphor (which they very likely do have) is a good basis on which to build recognition of the group. Goodall (1990, ch. 10) has even reported that

CO22552.indb 90

23/04/2010 14:13

How Did We Come to Be Human?

91

chimpanzees (Pan troglodytes) go to war—although it is a very different concept of war from the organized conflicts that even hunter-gatherer humans are able to mount against each other; and it is very unlikely that any chimp would give cognitive houseroom to the idea, dulce et decorum est pro patria mori. The origins of the metaphor, THE GROUP IS AN ENTITY, must lie in the complexities of human socialization—it is an essentially social expression. Machiavellian socialization (I get more advantages in the group than out) must have been supplemented by a level of co-operative altruistic punishment by members of the group to suppress unsocial activities. This is likely to have been a product of Vigilant Sharing—every individual ensuring that their personal share was enough, and individually punishing the greedy (and, indeed, punishing those that did not themselves punish the greedy). When the greedy are punished for their actions, whether simultaneously by the group or serially by individuals, they are likely to suffer in terms of breeding success. This punishment of the greedy therefore enhances Vigilant Sharing as a fit strategy. Vigilant Sharing in turn leads to Reverse Dominance, where alphas are suppressed by group action—alphas are alpha because they are greedy, and greed, in this social model, is not good for the individual. The whole model however, can be maintained with a simple friend versus foe identification system, and a willingness to change alliances to disfavour individuals seeking alpha status. Reverse hierarchy does not by itself require in-group versus out-group identification. It seems, therefore, that the metaphor THE GROUP IS AN ENTITY is the product of a late evolutionary event in our journey to being human. Before it could appear we must have already been a highly socialized species with a high level of co-operation. Without the metaphor, though, the complexities of our modern social structures would be impossible.

What Happened, and When? We cannot put a reliable timescale on the processes by which we became recognisably human, but we can give an order to the events described here. We start with social individuals co-operating in activities like hunting and group defence, but not yet sharing social models. Communication is certainly needed to co-ordinate these activities, but it requires only the sharing of current events and objects, what Tomasello (2008, ch. 3) describes as requesting. There is no need for syntactic grammar at this point because all that needs to be expressed are simple, one-argument, action-object combinations; and “you, there” has in this circumstance the same meaning as “there, you”. At some stage it became a fit strategy for humans to exchange social models with each other. Why this happened is not explored here, but a radical change in social structure is likely to have been the proximate cause. Somehow reputations became the driver of social cohesion, and reliability in signalling became

CO22552.indb 91

23/04/2010 14:13

92

th e orig in s of g r am mar

incorporated into reputations. Knight (1991) describes one mechanism that could have generated this change. Whether or not he is correct in all details, he does identify the key issues: male provisioning must somehow have been appropriated by females to feed increasingly helpless babies; and ritual is likely to have acted both as a conduit for expression of the concept of group and as a badge of belonging. The exchange of social models would have required a communication system capable of expressing relationships between individuals—two-argument, or object-action-object, combinations. If the action between the two objects expresses a two-way relationship there is no need for syntax: if Alf likes Beth and Beth likes Alf then the two forms are interchangeable in terms of describing the relationship. If, however, the relationship is one-sided then Alf likes Beth is no longer equivalent to Beth likes Alf—indeed, one of the forms becomes deceptive. In this circumstance, syntax, or identification of roles in the utterance, becomes crucial for understanding the meaning behind the utterance. If we add reputation into these utterances then it becomes necessary for the receiver to cognitively associate an utterance with the person making it: every A-relationship-B utterance must be tagged with by-C. Mental modelling of a three-dimensional calculus of social relationships, and of a hierarchy of utterance within utterance, becomes necessary to capture this. Adding gossip (Dunbar, 1996) to this mix means that these tagged utterances, A-relationship-B-by-C, need communicative mechanisms that permit them to be shared. As we will see in chapter 9, this is where the structural requirement for grammar really begins, and the need to negotiate to a common meaning becomes paramount.

CO22552.indb 92

23/04/2010 14:13

9 How Did We Come to Use Grammar? the reputations of others to make our highly co-operative social structures work; and we need the system of language to enable us to share our social models, and therefore our views of the reputations of others. Behind this is a recognition that the viewpoint of others is different from that of the self, and that the viewpoints of self and other are exchangeable. Relevant (although not necessarily accurate) information is the currency of exchange in this system (Dessalles, 2000), and supply of non-relevant information must be discouraged by punishment. The guarantee of relevance, which is also the mechanism for punishing non-relevance, is the reputation being shared between others, underpinned by the capacity to share social information via language. The grammar mechanisms within language are therefore likely to be the ones that make sharing of reputation models easier. These mechanisms include: object-action differentiation; the one-, two- and three-argument forms; hierarchies of meaning within an utterance; descriptors (adjectives and adverbs); and the identification of the three voices involved in a signalling event (speaker, hearer and other, or first, second and third person). In this model of language origins there are two possible routes to language grammar: the individual mechanisms could each have their own genesis, or the whole structure could arise as a single event. If we look for a single origin for the whole structure of grammar then we have a problem which is difficult in several dimensions. The first question we need to answer is, what fitness pressure required the grammar engine to spring fully-formed into the minds of humans? The second question is, what evolutionary event made the grammar engine a necessity in communication? The third is, what underlying single principle produced all the complexities we see in language grammar? And the fourth is, how does a single language grammar engine produce so much variation between individual languages? If language grammar is the product of a single cognitive engine then any research effort directed at finding that engine should at least attempt to answer these questions.

So far, we have seen that humans need to know



CO22552.indb 93

93

23/04/2010 14:13

94

th e orig in s of g r am mar

The canon of generativism, which espouses this approach, has had mixed success in providing answers to all of the questions, and solutions offered to answer one of the questions have often proved difficult to extend as general rules covering the other questions. In contrast, an incremental evolutionary approach to language origins, separately identifying the key features that differentiate our cognition from that of other animals, seems to implicate our social relationships rather than our language structure as the source of our cognitive complexity (Tomasello, 2003a, pp. 282–283). This indicates an indirect and complex, rather than a direct and simple, evolutionary source for language grammar. In terms of non-generativist approaches, the interpersonal nature of linguistic communication seems to intrude at every level. A semiotic approach, studying the forms of language in terms of the tasks they have to do, seems to require the specifics of human socialization to be addressed (Hurford, 2007, ch. 7); an evidenced grammatical approach, looking for commonality in the different grammatical forms of different languages, continues to produce a disturbingly large number of “ungrammatical” forms of actual language use which, nonetheless, are signals acceptable to sender and receiver (Sampson, 2005, ch. 5); and an anthropological approach, explaining why and how signalling complexity became a necessary driver for human culture, remains largely unexamined by generativism (Steels, 2003). (The phonological approach, explaining how speech complexity arose and stabilized in the human lineage, poses its own problems for generativist analysis; but that is not an issue for this book). This does not mean that generativism will never achieve its grail-quest for a single, elegant solution to the whole of language; but, in the face of the difficulties already encountered in this undertaking, this book will abide by Albert Einstein’s dictum: “If you are out to describe the truth, leave elegance to the tailor.”

Grammaticalization If language did not emerge fully-formed from the mouths of early humans there must have been a process by which it came into being. This is likely to have involved both novel grammatical developments and realignments of existing forms to accommodate new complexity, a process known as grammaticalization. No language is a fully fixed and stable structure, there is constant change within it. Some of these changes can lead to the creation of new dialects or even new languages; and this is a process we can see at work today. For instance, one small change is the appearance of the term innit in colloquial English: originally a contraction of isn’t it, itself a reanalysis of is it not, it was originally used as a pragmatic marker to share an external event or experience (it’s hot, innit or this is the way, innit). It has now evolved into a general marker for sharing any information (I’m going home, innit or you’re mad, innit). The normal rules of agreement, that the subject of the main sentence dictates the form of the subordinate subject and verb (I’m happy, aren’t I or he’s going home, isn’t

CO22552.indb 94

23/04/2010 14:13

How Did We Come to Use Grammar?

95

he) have disappeared, and innit has become a lexical unit in its own right. This particular grammaticalization is part of an ongoing process: the analysable isn’t it has its own peculiarity in that the uncontracted form, *is not it, is not seen as an acceptable English utterance. Grammaticalization seems to follow rules. These rules are not fixed, but they do have statistical relevance. Hopper & Traugott (1993, ch. 3) show that language change happens pragmatically (in terms of social language use) and semantically (in terms of meaning), as well as in terms of morphology (language structure) and phonology (sound or gesture representations); and it is produced by two mechanisms: reinterpretation of language forms within single brains over a lifetime, and the misinterpretation of forms between brains, such as between generations. Reanalysis is one such change mechanism: although the form itself remains unchanged, the underlying interpretation of the form changes. An example of this in English is the noun takeaway: originally used to refer to cooked food not eaten at point of purchase it is, by reanalysis, now used to refer to cooked food delivered to the home which involves no taking away. One aspect of grammaticalization that is of particular interest here is directionality: do grammatical changes occur in one direction only, or can they oscillate between forms? Hopper & Traugott (1993, ch. 5) are neutral on this, detecting both processes at work. But Dixon (1997, pp. 41–43) sees change as cyclical. He describes grammaticalization in terms of typological change: isolating languages, in which each word is a single meaning-unit, tend to become agglutinating languages, in which words contain multiple but separable meanings; agglutinating languages tend to become fusional languages, in which single syllables can perform multiple meaning functions; and fusional languages tend to become isolating. The approach to directionality dictates the approach to language origins. In multi-directional grammaticalization there is no reason to see language complexity as increasing or decreasing: movement occurs equally easily in any direction between systems, and there is no natural tendency towards any form. As there is no mechanism for complexity to grow, it is likely that language started complex and remained so. In contrast, one-directional grammaticalization does allow language complexity to develop, but it has a “slope” problem: if there is a natural movement towards a particular typology then we would expect to see all languages tending towards the same form, and this does not accord with the data. Cyclical grammaticalization is a one-directional system that does not have the “slope” problem; but, like multi-directional systems, there can be no preferred typology if the cycle is to be continuous. An alternative approach to complexity is to see it not as a feature of grammaticalization but as a cultural imposition on language: language needs to become complex as culture becomes complex. In this model the multi-directional and cyclical models can acquire complexity with each grammaticalization; so the process of grammaticalization is not just drift, it is driven by the need to communicate cultural complexity. The typology of a language is not itself a

CO22552.indb 95

23/04/2010 14:13

96

th e orig in s of g r am mar

marker of complexity, complexity can grow (or diminish) with each move to a new typology—multi-directional systems can be steplike, and cyclical systems can be spiral.

Grammaticalization and Language Origins So what do the proponents of grammaticalization have to say about language origins? For Dixon (1997, pp. 63–66) the cyclical nature of typological change means that there is little scope for language enhancement; he sees language as making a sudden appearance, proceeding from nothing to high complexity within a few generations, and then becoming largely fixed in terms of potential. Deutscher (2005, ch. 7) takes the view that we cannot know the origins of language before the two-word stage. This was the point at which symbolic usage took off, and any description of language origins cannot usefully go back before this point. Pre-symbolic communication cannot be called language, so we cannot know the origins of language except through language itself. In contrast, Heine & Kuteva (2002) see language complexity as growing slowly and steadily through a series of layers: nouns; verbs; adjectives and adverbs; demonstratives, prepositions, aspects, and negation; extensions of form to produce complex constructs; reduction of meaning to convert meaning words into marker words; and the reduction of markers to morphemes. In this description the earliest grammar consisted of nouns only, simple labelling of objects and people. This layer of nouns was then supplemented by actiondescriptors, or verbs; then came qualifiers for nouns and verbs; and then, in a series of steps, all the other potentials of language were realised. For Allen & Seidenberg (1999) words must have been present before grammar began. Grammaticality is a statistical relationship which emerges from connections made between words when they are used; it emerges from common non-conscious agreements in the actual usage of words—a negotiation towards grammar. Hopper & Traugott (1993, pp. 33–38) do not speculate on language origins, but they do make an important point about early languages: we cannot assume anything in early languages that is not evidenced in at least one language known about today. If early languages had grammatical features that disappeared before languages were first recorded, we have no way of finding out how those grammatical features worked. Nettle (1999, ch. 2), however, does offer us a way out of this dilemma. Language change seems to follow the same rules as genetic change: when a single language population becomes split into isolated groups, variation occurs; contact between different languages can produce consolidation, where the two separate languages become more alike, or differentiation, where they move further apart, depending on the nature of the contact. This is not a neutral process driven only by language, change in these cases is a product of social contact. Early languages would have had fewer social contacts with other groups, and so would have undergone less change. We cannot know for certain

CO22552.indb 96

23/04/2010 14:13

How Did We Come to Use Grammar?

97

that the grammar features described today are a definitive set; but we can expect modern levels of change to be much faster, and therefore the current variation to be greater, than that of early languages. The chance of a particular grammatical feature being present in the narrow variation of early languages and not in today’s wide variation is likely to be small. One proponent of grammaticalization who looks in detail at language origins is Carstairs-McCarthy (1999, ch. 5). He takes the two-argument form as the basic syntactic structure, and compares the binary generative analysis of this form with the traditional analysis of the syllable. A syllable consists of an onset, the sound that announces the syllable, and a rhyme which completes the syllable; so the syllable dog has an onset of d and a rhyme of og. The rhyme is also divisible, consisting of the nucleus (o) and the coda (g). A generative analysis of a simple two-argument sentence such as Alf likes Beth produces a similar tree structure to that of the syllable: a noun phrase (Alf ) links to a verb phrase (likes Beth); the verb phrase is then divisible into a verb (likes) and a noun phrase (Beth). It is certainly true that the syllable and the basic two-argument sentence form have comparable structures which can be described as [A[uB]]; but what does this signify? While the collocation of the two forms is indicative of the human capacity to identify analogies, there is no clear evidence that the form of the syllable and the form of the sentence are actually coherently linked.

Overture and Beginners, Please If grammaticalization does not by itself provide a route back to the origins of grammar, what needs to be added to the model to make it work? We should first make an important distinction between cognitive grammar, the algorithms we use to make our mental models, and communicative grammar, the algorithms we use to share our mental models. Both forms of grammar are about organizing objects and actions into meaningful relationships—establishing the actions, or relations, which exist between different objects; but they are also different in an important way. Cognitive grammar is internal to the individual mind, so idiosyncrasies are significant only if they advantage or disadvantage the individual; in contrast, any idiosyncrasies in communicative grammar must be capable of interpretation by other minds, which imposes limits on expressive difference. We can state uncontroversially that cognitive grammar preceded communicative grammar. In the model proposed here, the forms of cognition required by social modelling calculus must prefigure the forms required to exchange the social models—and, indeed, communicative grammar is, at the points of production and reception, just another type of cognitive grammar. It is unlikely, however, that all of our communicative grammar is necessarily prefigured in non-communicative cognitive grammar; and this raises the question, which forms would have been prerequisites for the appearance of language, and which developments occurred after language began?

CO22552.indb 97

23/04/2010 14:13

98

th e orig in s of g r am mar

The consensus from grammaticalization studies is that nouns, verbs and at least the one-argument (action-object) form would have been available. If we introduce the exchange of social models, though, we need at least the two-argument form (the object-action-object form) to express the relationship between two individuals. This is also the simplest basic structure requiring syntax: if the relationship between the two objects is bi-directional then order is unimportant; but if it is one-directional then a problem arises about which object is the actor or instigator, and which is the patient or recipient—Alf hit Beth has different implications in social calculus to Beth hit Alf. So, while the origin of language may be uncertain, the fact that syntax has a minimal requirement means that the origins of grammar are much clearer. We cannot say whether the one-argument form was being communicated before the two-argument form made syntax a necessity, we can only say that communication of the two-argument form is the minimum requirement for syntax. That is, however, a very important thing to be able to say. We also cannot say when it became necessary to communicate two-argument forms, but we can say that sharing mental models of social relationships requires the capacities to express and comprehend two-argument forms. This makes it very likely that the two argument form was prefigured in cognition before it was used communicatively. The important first step of language grammar was not the creation of a new cognitive engine but the reuse of an existing one. What other features of grammar are likely to be prefigured in general cognition? There clearly has to be the capacity to label objects, and this needs to be present in cognitive social modelling before the need arises to reliably share the models and the labels. The objects being modelled must be effectively tagged, so that the modelling mind can use the same object in different models and know that it is the same object. Inside cognition the label can be holistic (partly recognition of physical attributes, partly emotional responses, partly association with past events, and so on); and, most importantly, it can be idiosyncratic—my tag for the object need not correspond to your tag for the same object. When the models become shareable between minds there has to be a negotiation to a common tagging, and tags will need to become less holistic and more neutral representations. Simple nominal labels would seem to work best in this circumstance (and it can be argued, although it is outside the scope of this book, that the vocal channel is superior to the gestural channel for this purpose). The noun, or the name at least, is therefore necessary as a prefigured cognitive item for syntactic language: nominalization is a necessary part of cognitive social modelling, and therefore a necessary part of sharing those models, although there still has to be negotiation between minds to establish common meaning. The relationship between objects is also necessarily prefigured in cognition: if [A[uB]], the one-directional two-argument form, is the basic syntactic structure then, as well as nouns (A and B), there is a need for linkage between nouns, which is the role of verbs (u). These verb-linkers represent an important cognitive trick

CO22552.indb 98

23/04/2010 14:13

How Did We Come to Use Grammar?

99

that Machiavellian thought makes possible. If my model of Alf is accompanied by fear, and my model of Beth is accompanied by fear, how do I generate a model of Alf and Beth together that is not dominated by fear? If I am to model the relationship between Alf and Beth as one of alliance I have to be able to do this. Somehow I need to be able to see Beth through Alf’s eyes and Alf through Beth’s eyes; or, in terms of my own cognition, I need to see the relationship between Alf and Beth on a different level to my relationships with each of them. So the trick of adopting the viewpoint of another requires a method for isolating the relationships between others from my own relationships with those others. The cognitive linkage that is represented by verbs needs to have a level of disinterest to be effective. The final communicative capacity that needs to be prefigured in general cognition is the ability to negotiate to common tags for objects. To make this work, there has to be at least one metaphor operating in the minds of the negotiators: THE NAME IS THE OBJECT. Tagging involves creating a shorthand label which stands in cognition for the object itself, such that any of the natures of the object itself can be evoked by the tag. With pre-existent language this process seems obvious: words represent objects in an arbitrary but agreed way, so the use of a word evokes a similar set of impressions in speakers of the same language community, even though their emotional reactions to those impressions may differ markedly. There is a “thingness” to words even when they are abstract and physically unbounded. This capacity to represent objects with tags must have been prefigured in cognition before language appeared. The dislocation needed to treat Alf both with fear and with disinterest implies a representation of Alf which is abstracted away from personal reaction. All representations of Alf are associated with a single “Alfness” which is completely arbitrary; and this arbitrary tag stands for the whole knowledge of the Alf-object in ways that associations evoked by viewpoints cannot. The tag, or THE NAME, represents THE OBJECT faithfully and intimately because it is dislocated from any single viewpoint of Alf. It is easy to see how this disinterested, dislocated tagging could have become useful in disinterested, dislocated negotiation to common meanings. We can thus see that for the appearance of grammar in communication we need the ability to segment utterances into separable meaning-units, and for these meaning units to be interchangeable within a communicable structure. We also need the capacity to use different meaning types in the structure—at minimum there has to be differentiation between objects, or nouns, and relationships, or verbs. Finally, we need hierarchy, to allow different parts of the structure to govern other parts; in the one-directional two-argument form we require a mechanism able to describe one object acting upon another. What does not seem to be present at the origin of grammar, contrary to the prediction of the HauserChomsky-Fitch (2002) model, is recursion.

CO22552.indb 99

23/04/2010 14:13

100

th e orig in s of g r am mar

Not Required at the Origin of Grammar The model proposed so far gives us nouns, verbs, the one-directional two-argument form, disinterested modelling, and a mechanism for negotiating to common meaning. All of these come out of the need to share models of social relationships within a group. It should not be assumed that all of these capacities were marshalled in general cognition and released together as language, that is not the claim being made here. Instead it is being argued that these mechanisms must have been available for communication when the first syntactic utterances, exchanging social models, were made. So what aspects of modern language were notable by their absence? The lack of linguistic complexity in early grammar is unremarkable: there is no need to look for adverbs, adjectives, multi-argument forms, subordination, iteration or recursion at the point of origin, these can all be seen as later developments. There is, however, an aspect that we tend to think of as basic to language, but which is not part of the origins picture painted so far. Selfhood, and the use of voices (me, you, they) would seem to be a minimal requirement for linguistic communication: when exchanging social models, the ability to label the speaker and listener and those who are neither would appear to be basic. Yet there is no function in nonlinguistic cognition for a mechanism allowing recognition of the performers in a speech act; without language all the internally generated social models are expressible as object-relationship-object, and all of the objects are it. There is no need for the unlanguaged mind to recognize the special object you because there is no external dialogue requiring differentiation of you from it—there is no “other” voice requiring differentiation from the internal voice. Indeed, the fact that there are no other voices means that there is no need to recognize the internal voice, me, either: there is only one voice, so there is no point in viewing it as “my” voice—or anything other than “the” voice. The selfhood needed by the unlanguaged mind has to recognize the edge between self and other, but it does not need to produce models of the self in the same way that social calculus requires models of others. The languaged mind, in contrast, needs to recognize both the special nature of the you receiving my models and the me providing them, and the you giving me models and the me receiving them. Being aware of you means being aware that some of the third parties in the models I’m offering are actually second parties—I am offering my receiver information about themself; and it means being aware that the models being offered to me are being offered intentionally. The intentions of you become significant, both in terms of what I say and what I hear. Being aware of me has even more far-reaching effects. The first is that my voice is no longer the only voice, it is one among many, although it remains the most important. The second is that others are offering me their models of myself, which means that, to properly interpret their models, I have to have the capacity to make my own models of myself—to treat myself as a third party like all the

CO22552.indb 100

23/04/2010 14:13

How Did We Come to Use Grammar?

101

other third parties in my cognitive modelling. However, the feature of social modelling that makes it so powerful is the capacity to treat the entities modelled with dispassionate disinterest; and treating the self objectively would seem to be contrary to all good evolutionary sense: where is the advantage of treating myself with disinterest when all around me are passionately self-interested? Self-modelling would seem to be an unfortunate side-effect of sharing social models, yet it has somehow become the very mark of being human; how could this have happened? In a social environment driven by reputation, sharing models creates an environment where being honest, informative and co-operative enhances social standing and therefore fitness; and self-disinterest then allows me to model the effect that my generosity in sharing information has on others’ image of me. Self-modelling is therefore an outcome of sharing social models, and only has value in that environment: selfhood comes out of language, it does not originate it. Vigilant Sharing, Reverse Dominance and altruistic punishment generated the environment in which sharing of social models becomes a viable fitness strategy, and the sharing of social models made self-modelling and disinterested self-effacement possible; but it was the subordination of the self to the group, made possible by self-effacement, that permitted full language to develop.

Becoming Complex When the circumstance arises that makes exchange of social models a fit strategy, their exchange in the form of one-directional two-argument forms creates its own evolutionary pressures. Minds that are able to easily exchange models will obviously do better than those that find it more difficult, so the species will move in the direction of enhanced signalling. Odling-Smee & Laland (2009) call this niche-building: language creates its own evolutionary fitness constraints, affecting the signalling environment by its very existence. For instance, minds that can tag received models with the identity of the sender will do better than minds that cannot differentiate received models by source. Knowing the individual creating the model adds information about the reliability of the model being offered, and it can generate understanding of the motivation of the sender offering the signal. This means that the offering of the model becomes valuable whether the model offered is accurate or not. Gossip becomes valuable not just for what is said but for who says it, and why they are saying it. This capacity to attach a sender-label to a shared model can be represented by the form [C[A[uB]]]—or, in English, Alf-relationship-Beth-by-Colin. This structure has the potential to become recursive (Doris said that Colin said that Alf likes Beth), but only if it, too, can enter the communication structure. Clearly it did, but what evolutionary pressures allowed this to happen? The benefits for the receiver seem obvious: iterated labelling gives enhanced knowledge of the reliability of the message, it creates greater reliability in personal modelling of social relationships, and it makes the receiver aware of the sender’s sources of

CO22552.indb 101

23/04/2010 14:13

102

th e orig in s of g r am mar

knowledge, among others. The motivation for the sender to provide this information is, however, more difficult to explain. Obviously, there is a quid pro quo effect here, you tell me yours and I’ll tell you mine, but what really makes giving away this extra information valuable to the sender is reputation. If information is socially valuable, and I become known as a reliable source of social information, my status increases. There is therefore a fitness pressure towards the sharing of reliable information. If, however, I offer inaccurate third-party models as my own it is my reputation that suffers; so a mechanism which allows me to be the person offering the information, but which distances me from the accuracy of the information, will be advantageous to me. Other pressures exist to create complexity in language. Language can become a vehicle for signalling the capacity to produce language itself, and utterance can be used as a costly signal if it can be made with sufficient demonstrated skill. Although costly signalling is not the cause of complexity, complexity in language can become usable as a demonstration of language capacity—which means that we would expect to see a higher proportion of demonstrative complex language coming from males as a costly signal of fitness. There do, indeed, seem to be aspects of complex demonstrative language that are statistically the domain of males (Tannen, 1994; Cameron, 1998a), but, as Cameron (1998b) points out, they are far from being solely male preserves; every human seems to have the capacity to use demonstrative language, although males are marginally more likely to actually to do so. Complexity to indicate fitness is not currently a significant feature of language. Complexity will also come from the need to differentiate the in-group from the out-group (Nettle, 1999). If language is a measure of group membership then complexity, and the capacity to handle complexity, are badges of group membership. The most celebrated example of this is the slaughter of the Ephraimites by the Gileadites at the fords of the River Jordan (Book of Judges, ch. 12). The Ephraimite language lacked a /sh/ sound, so they could be differentiated from the Gileadites by being forced to say the word shibboleth. There is some evidence that the word Scheveningen was used in a similar way by the Dutch in 1940 to identify German infiltrators. These pressures towards complexity rely, however, on a pre-existing basic language system: a willingness to communicate, something to communicate about, and a system to formalize communication. They exapt the pre-existing language system for the good, old-fashioned, Darwinian purposes of getting mates and disposing of rivals. Just because language is something different in Darwinian terms doesn’t mean that it cannot be used as a Darwinian device; in fact, its value as a sign of fitness indicates that it inevitably will be.

From Non-grammar to Grammar There seems to be a stepwise arms race between cognition and model sharing. The personal one-argument knowledge that I have relationships with others

CO22552.indb 102

23/04/2010 14:13

How Did We Come to Use Grammar?

103

(relationship-B or [uB]) becomes detectable by others through my actions, making their possession of a two-argument calculus (A-relationship-B or [A[uB]]) a fit strategy. The sharing of these two-argument forms through syntactic communication then makes the possession of a three-argument calculus (A-relationship-B-by-C or [C[A[uB]]]) fit—tagging messages with their source allows the receiver to construct a reputation of the sender. Because of the nature of the signalling environment, however, it also becomes a fit strategy to share these three-argument forms; and this, in turn, generates the need for a limited iterative capacity in modelling. As this iteration can become recursive, it creates the illusion that a generalized recursive engine is at work; but, in practice, the iteration has a limited range. Dunbar (2004, ch. 3) shows that three or four levels of third person allocation, a five- or six-argument form, is the limit for most people. This is therefore not the infinite recursion engine of Hauser-ChomskyFitch (2002), it is a limited capacity which emerges from the fitness of sharing three-argument social calculus (interpersonal relationship models which have been tagged with their sources). The language structure used to express this capacity is indeed theoretically infinite, but this is only a secondary effect of a structure designed to work at three or four levels of iteration. Of course, if you define language as the product of recursion then everything described here is pre-language or protolanguage; if, on the other hand, you define language as symbolic manipulation then the recursion described here comes late in the development process. The point at which language begins is dependent on the definition of language, and the definition of language used here (grammatical utterances) produces its own point of origin (the first grammatical utterance). This definition is no better, but no worse, than the others, and it does allow a concrete description of grammar origins, if not language origins, to be given. This chapter has offered a developmental path from pre-grammar to complex grammar. It has, inevitably, only touched the surface of what may have happened. It has left big questions unanswered: what was the event that made exchange of social models a fit strategy? How does a language propagate through a community? How does negotiation to common meaning happen? It has not addressed language complexity except in the abstract, and has talked about only nouns and verbs. It has reviewed multi-argument structures, but only in linear form; and it has said little about the complex sentences in which the theory is being described. It has, however, provided a possible route from non-grammar to grammatical human language which indicates that grammar does not need to be viewed as a strange isolate from the rest of nature.

CO22552.indb 103

23/04/2010 14:13

10 What Nonhumans Tell Us about Being Human (Fouts & Mills, 1997); bonobos who can understand and use syntactic structure (Savage-Rumbaugh & Lewin, 1994); a grey parrot who can recognize number symbols, do simple arithmetic, and understand adjectival properties (Pepperberg, 1999); dolphins who can parse multi-argument sentences (Herman & Uyeyama (1999): all of these represent experiments carried out by humans to try to prove the non-uniqueness (or uniqueness) of human language. The question behind all of these experiments is: can animals learn human language, thus proving that it cannot be a solely human trait; or are all the efforts merely evidence of the irreconcilable difference between animal signalling and human language? While it is true to say that Homo sapiens is just another type of animal it has, like every animal, certain characteristics which make it the animal it is. By studying only those characteristics it is possible to conclude that language is something special in nature (which it seems to be), and that it has no antecedents in nonhumans (which seems unlikely). Every species is suited to its environmental niche. Indeed, that is a definition of how evolution works: a species grows towards its niche through the reproductive success of the individuals most fitted for that niche and the penalizing of the less fit. The appearance of language in any species must therefore be the product of an environmental feature that makes language a fit strategy for that particular combination of species and niche. If language really is a product of human socialization then its appearance in humans and its non-appearance in other animals are both unremarkable outcomes. Human socialization is, by definition, species-specific; and its products, such as language, are therefore also likely to be so. However, behind the issue of uniqueness there is another question which is just as significant for our understanding of language origins: why should animals want to learn human language, of what use can it be to them? What is remarkable about language use by nonhumans, therefore, is that it happens at all. Somehow these nonhumans are able to use our signalling system, however incompletely, to convey information to us. Some, but not many, humans are able to do this

Chimpanzees who can sign

104

CO22552.indb 104

23/04/2010 14:13

What Nonhumans Tell Us about Being Human

105

in reverse, producing sounds which are recognisable to nonhumans as information carriers; and it may be that, in our hunting past, this capacity was more widespread and important (Lewis, 2009). However, the usual reason why human hunters use animal signals is to get the animaI’s attention or distract them—the hunters use the signals deceptively for their own ends. As scientists, what mostly interests us in our experiments involving animals using human language is veracity. What we expect from the animals in these experiments is therefore different from our expectations of how humans use language. When conversing with humans, we expect and allow for the deceptions of metaphor, obliquity, hyperbole and fiction as part of everyday discourse; we treat these as normal language use and signs of communicative success. In our animal experiments (although not usually in our social linguistic interactions with the same animals) we treat these linguistic deceptions as communicative failures. Is this a necessary consequence of using scientific method to assess the experiments, or does it represent an unspoken prejudice against the idea that animals have minds similar to ours? (Savage-Rumbaugh & Lewin, 1994, pp. 256–258).

Animals and Grammar The central thesis of this book is that the capacity in humans for language grammar is a product of the need to exchange social models. The cognition behind this, the capacity to model social groups as a series of interlinked bipartisan relationships, is probably not exclusive to humans, and is likely to be a necessary cognitive function for any species able to manipulate others in a Machiavellian way. The expression of social modelling within signalling, however, is rare—as far as we can tell, only one species does it (Donald, 2001). There is no evidence for the exchange of social models by nonhumans, either in natural signalling or in animals who have been trained in human language use. We have detected no cases in the wild of animals telling other animals about social relationships; and our language experiments with nonhumans have, so far, not involved testing of social modelling by asking the equivalent of “what do you think of Beth?”, or “what does Alf think of Beth?” As the capacity to signal social models has never been tested we cannot know how good other animals are at exchanging them; but, from what we know about their social structures, it seems likely that chimps and bonobos in the wild would be able to understand the cognitive modelling behind the first question (expressing the relationship of another with the unmodelled self; Tomasello & Call, 1997, pp. 338–341) and individuals acculturated to human society would probably understand the cognitive modelling behind the second question, too (the relationship of another to a second other; Premack & Premack, 1983, ch. 3). It is likely that our closest primate relatives are able to use cognitive social modelling involving both relationship-B and A-relationship-B forms; yet they do not have

CO22552.indb 105

23/04/2010 14:13

106

th e orig in s of g r am mar

communicative systems to enable them to exchange their models. Why should this be? The simple answer is that the very capacity to model social relationships indicates a type of cognition that precludes the sharing of those models. If the ability to model relationships in a group is common, it implies a fitness advantage for individuals able to do so; and, if that social environment is also Machiavellian, the ability to manipulate others through their friendships and rivalries seems to be an obvious likely fitness advantage. By modelling the relationships of others, I can tailor my approach to those others, allowing me to build a coherent set of alliances and to correctly establish my position in the social pecking order. This modelled social information is valuable to the modeller because it enhances their knowledge base and reduces costly confrontation; giving away this valuable information reduces its exclusive value and therefore reduces its fitness advantage for the giver. In a Darwinian universe, it shouldn’t happen—and, indeed, in every species universe except that of humans, it doesn’t happen. If we are to study the origins of grammar in nonhumans, therefore, we need to look for social modelling in cognition, and not in communication. Fortunately, a lot of work has been done in this area, and a clear picture of the social modelling capacities of other primates—and some other animals—is beginning to emerge.

Primate, Know Thyself The temple of the Oracle at Delphi is supposed to have had the words know thyself above the entrance. However, in the tricksy language with which the Oracle made all of her pronouncements, this has several meanings: know that you have existence, know that you are a person, know the person that is you, know yourself as others see you, and know your place, among others. When we are looking for signs of self-awareness in nonhumans we should be aware of these different ways of being self-aware, and of the particular needs that different species have for selfand other-awareness. As Stephen Budiansky (1998, p. 162) says: There is a certain flavor of anthropocentric bias in the very hunt for selfawareness in other animals, a hint that conscious self-awareness is the best thing evolution has yet to produce—and we want to know how animals stack up against this standard of ultimate perfection. Yet it is no insult to animals that they might do what they do without self-awareness as we understand it; nor is it a particular compliment to animals to see how closely they share our peculiar cognitive abilities.

When we are looking for the ways in which nonhumans display—or don’t display—self- and other-awareness, therefore, we need to be clear about exactly what we are testing for and measuring. One of the earliest tests for recognition of self was Gallup’s mirror test (1970). He showed that some, but far from all, chimps are able to recognize

CO22552.indb 106

23/04/2010 14:13

What Nonhumans Tell Us about Being Human

107

the image in a mirror as being themself, whereas monkeys cannot do this. The experiment involved first acculturating the chimps to a reflective surface. The chimpanzee was then anaesthetized and a water-based mark put on their face in a position they could not see directly. On seeing their reflection and noticing the mark, the chimpanzee touched the mark on their body rather than on the mirror, using their image to direct their investigation. Monkeys either ignored the mirror image or treated it as another monkey. This recognition of the image in the mirror as equating to the physical self seemed to Gallup to indicate a clear cognitive difference between what chimps and humans can do and other animals cannot. The situation, however, has now become less certain. Bottle-nosed dophins (Tursiops truncatus) pass the mirror test (Reiss & Marino, 2001); one Asian elephant, Happy, has also passed (Plotnik et al., 2006); and even some monkeys (Cebus apella), excluded from self-recognition by Gallup’s tests, have proved to be mirror-aware, at least in terms of other objects in the room (de Waal et al., 2005). Pigs have also proved very capable at recognizing mirrors as reflectors of the real world (Broom et al., 2009). Perhaps most intriguing is evidence for mirror recognition in magpies, which seem to be able to pass the mark test with the same reliability as chimpanzees (Prior et al., 2008). The capacity to recognize the image in the mirror as a representation of the physical self (what Povinelli (2000, pp. 328–337) calls the kinaesthetic self) seems to be quite widespread in nature. If, though, physical self-recognition is not limited to the Pan/Homo family, it cannot be a useful indicator of human speciation.

Multiple Intelligences In the early 1980’s, as a counter to the growing tyranny of IQ as a single measure of intelligence, Gardner (1983) proposed the theory of multiple intelligences. It is not possible to judge a human as evolutionarily more or less fit based on the standard IQ test because it tests only one fitness-relevant capacity, what Gardner called logical-mathematical intelligence. He added to this five other fitnessrelevant capacities which can be described, in the wider sense, as intelligences: linguistic, musical, spatial, bodily-kinaesthetic and personal intelligence (or selfawareness). Each intelligence in this suite of measurable human capacities marks a potential difference between us and other animals; and they therefore all have the possibility of being involved in our genetic speciation. Yet all of these intelligences represent internalized knowledge about the self. The obvious and most significant difference between us and other animals, our level of socialization, seems to be missing from the list. As early as 1927, Thorndike had used the term social intelligence to describe the fact that some academically gifted students were nonetheless failing the social side of university life. The Universities were, wrongly, equating logical-mathematical intelligence with success in life. For this reason, Goleman (1995) added a seventh item, emotional intelligence, to Gardner’s list.

CO22552.indb 107

23/04/2010 14:13

108

th e orig in s of g r am mar

At base, every intelligence relies on curiosity. Intelligences need the capacities to discriminate and to retain knowledge, but there also has to be a capacity to acquire new knowledge; curiosity is the engine that powers that acquisition. In terms of social intelligence, curiosity about the cognition of others is represented by an interest in their states and relationships, and it works on three levels: curiosity about the physical states of others, their strengths, weaknesses and habits (kinaesthetic knowledge); curiosity about the mental states of others, what they are thinking and how this can be used to advantage (Machiavellian knowledge); and curiosity about emotional states of others, what they are feeling (empathic knowledge). As well as kinaesthetic and Machiavellian intelligence about others, capacities well documented in other primates (e.g. Arbib, 2005; Whiten & Byrne, 1988), humans also have empathy with those others. This empathy is, however, difficult to explain in evolutionary terms: to empathize we need to be able to “feel the pain” of others, and where is the advantage to the individual in taking on the problems of others? Surely the individual who concentrates on their own problems will do better than the individual who adds the problems of others to their own? Clearly this evolutionary difficulty has been solved, because humans do not have the same societies as chimps; but how did it happen? And is empathy a possible key indicator of what makes us human?

Accommodating Others There is certainly a difference between the tolerance of others among humans and among apes, which seems to create a greater willingness in humans to make the effort to understand others. Hrdy (2009) describes some experiments using chimps and macaques raised by human carers after being rejected by their birth mothers. These infants initially displayed a level of imitation similar to human babies. However, where humans continue to imitate throughout their childhood, and possibly their life, the chimps ceased to do so after about eleven weeks, and macaques after a mere seven days. The willingness to study the physical postures of others indicates an interest in how those others work, both physically and cognitively. The early suppression of curiosity in this area among chimps and macaques indicates an important difference in socialization between these species and humans. Fossey (1983, pp. 70–71, pp. 218–219) describes the use of infanticide by male gorillas to remove a rivaI’s offspring and bring females back into oestrus, and the apparent absence of mourning by the mothers for their lost offspring. It makes sense in a male-dominated gorilla group for a mother to discontinue investment in offspring that cannot earn her the protection of the dominant silverback, and for a silverback to encourage this transfer of a female’s investment to ensure maximum reproduction during his reign. Empathy in this circumstance is not evolutionarily fit. Cheney & Seyfarth (1990, pp. 235–236) state that the few descriptions

CO22552.indb 108

23/04/2010 14:13

What Nonhumans Tell Us about Being Human

109

of empathy in the actions of monkeys are likely to have other explanations. Monkeys just do not seem to exhibit compassionate traits such as care for the elderly, the sick, the bereaved or the defeated. While they do care for infants, it seems to be largely innate caring, and they do not adjust their caring regimes if their offspring becomes sick or distressed. When grooming others, monkeys tend to treat wounds as points of interest and probe them without consideration for the wounded animal. There seems to be no compassion in the actions of monkeys, and any social benefits of co-operation seem to be products of innate mechanisms rather than conscious cognition. This does not mean that empathy is missing from other primates, just that it has a different quality. Warneken & Tomasello (2006) conducted tests on young chimpanzees and human infants to assess their willingness to help others. The tests were on an increasing range of difficulty to assess the capacities of the subjects to mentally model the needs of the person needing help. Both children and young chimps proved willing to help, but their actual support, and their understanding of the need for assistance, differed markedly. The human infants helped in more circumstances, provided help that better modelled the needs of the person needing assistance, and helped regardless of whether they knew the person needing assistance. Chimps were less likely to help, less effective at helping, and tended to help only people they knew. On the basis of these tests, chimpanzee empathy seems to be more constrained than human empathy. Yet there are some examples of nonhuman empathy which are eerily humanlike. De Waal (1997, pp. 154–160) discusses a strange proto-symbolic activity of bonobos which indicates a respect for the emotional states of others. Bonobos build night nests for sleeping; but they sometimes also build day nests, and the purpose of these nests seems to be mainly to establish personal space—a purpose which is respected by other bonobos. The nests are built to provide private feeding spaces or just to deter others from approaching; even close allies will not invade the sanctity of the day nest, and offspring will beg at the edge of the nest for their mother’s permission to enter. It has even been recorded that one male successfully used nest-building to deter an aggressive opponent. Clearly there must be, on some cognitive level, recognition of, and respect for, the Garboesque message of day nest-building: “I want to be alone”. De Waal (1996, pp. 148–150) also describes an experiment involving capuchin monkeys, in which monkeys in twin cages were selectively fed and allowed to share the food with another known capuchin. The unfed monkey had no direct access to the food, and relied on the fed capuchin to give them a share. The experimenters found that the capuchins shared readily if they had a pre-existing good relationship, but refused to share with their enemies or unknown monkeys. In the wild this would translate to assisting kin: capuchins mostly stay in small, closely related kin groups, so familiar monkeys are also kin. There does seem to be recognition here of the needs of others, and a willingness, albeit limited, to provide for those needs.

CO22552.indb 109

23/04/2010 14:13

110

th e orig in s of g r am mar

Empathy Empathy is not just about understanding of the needs of others, there has to be an attempt to meet those needs. This differs from Theory of Mind, which can operate in a Machiavellian environment in which the needs of others are exploited, rather than accommodated. Premack & Premack (1983, ch. 3) showed that chimpanzees are able to model the probable future actions of others based on their previous actions. However, chimps do not seem able to model another individual modelling the beliefs of a third individual—something which humans do regularly, with sentences like “Alf thinks Beth is unhappy”. Dunbar (2004, ch. 3) shows that, in terms of Theory of Mind, chimpanzees are able to impute motivation to others, but not to impute motivations about motivations. Most humans can work at about five levels of motivation (which Dunbar describes as “A believes that B thinks that C wants D to suppose that E imagines . . .”, ibid, p. 48). This is a clear difference in capacity for mental modelling between humans and other primates. However, Tomasello (2008, pp. 342–245) shows that this advanced Theory of Mind is not sufficient to get us to language by itself, it is the sharing of our models of the intentions of others that is vital. It is not our cognitive sophistication that demands language but our co-operative sophistication; and empathy is the engine that powers our co-operation. It does seem that there is a species-hierarchy of empathy in primates, with monkeys displaying little, chimpanzees more, bonobos even more, and humans most. Empathy is clearly a trait that has been subject to evolutionary pressure in the human lineage. It is not an all-or-nothing difference between humans and other primates, however, and many individual cases of behaviours that could be classed as empathic have been recorded in nonhumans. This is especially true of primates who have learned to communicate in human language, which may be because their enhanced humanlike communicative behaviour allows us to see their other humanlike qualities; or it may be that their exposure to human culture has created the possibility for all the humanlike qualities they have to be expressed. For instance, Kanzi, the bonobo taught to communicate with humans via a special keyboard, has been reported as not just being friendly with the other primates in the Yerkes Primate Centre, but also requesting visits, and taking gifts for the apes he visits (Savage-Rumbaugh & Lewin, 1994, p. 155). Nonetheless, even with language-competent nonhumans like Kanzi, the distance between human empathy and that of other animals remains significant. So how did humans move so far along the scale of empathy and co-operation, ending up in a very different species-niche to our closest relatives, the chimpanzees and bonobos? This is a major problem that any evolutionary theory of human origins has to address: empathy leads to greater co-operation, and greater co-operation leads to the advantages of specialization, enhanced communication and social tolerance; but co-operation also leads to very effective cheating by individuals who use empathic knowledge for Machiavellian ends. How did we

CO22552.indb 110

23/04/2010 14:13

What Nonhumans Tell Us about Being Human

111

humans get past this road block and become the highly co-operative species we are? Sober and Wilson (1999, ch. 4) offer one solution: evolution does not move a species towards a set of species-benefitting goals but towards behaviours that benefit the individual. A sharing behaviour can offer personal benefits if it acts to smooth the stressful peaks and troughs of existence for an individual—allowing them to receive support in bad times in exchange for giving it in good times. The stress has to be sufficiently extreme for the individual, such that the costs of not receiving support in bad times are significantly higher than the cost of giving it in good; and the stress has to affect individuals rather than the whole community. In this circumstance, failure to establish a network of support exchanges will disadvantage the individual. Cheating is counter-productive because of the exchange mechanism: taking and not giving will only work for a short while, very quickly the sources of support for the cheat will dry up. In this rather specific environment the evolutionary pressure is towards co-operation, and the development of enhanced empathy to anticipate the needs of others. There is a second evolutionary pressure towards enhanced empathy: If a behaviour benefits the group of which the individual is a member as well as the individual then it may be secondarily beneficial for the individual. Empathy acts to reduce antagonistic relationships in a group, allowing larger, more concentrated, and more co-operative groups with less stressful interrelationships. Constraints on individuals to conform to the group “morality” (essentially, to obey the golden rule, “do as you would be done by”) then becomes a matter of positive pressure. Instead of just withdrawing support from cheats, the possibility of group-driven altruistic punishment of individuals not exhibiting empathic behaviour arises (Singer et al., 2006). Empathy can work to create a tyranny of the commons as well as individual co-operation.

Not about Language? This chapter has not directly addressed the question that is usually treated as paramount in language origins studies: can nonhumans use human language? This is not a trivial question but, in terms of the origins of grammar, it is not necessarily helpful. To use language a nonhuman has to understand the nature of the sign as an arbitrary representation which has meaning only through the collusion of others; it has to be able to articulate—not necessarily with vocal speech, but with some signalling system that can allow segmented signs to be used in a coherent way; it has to be able to understand the hierarchy of meaning that allows separate signs with individual meanings to be brought together to create sign combinations which have new meanings; and it has to understand the concept of combinatorial rule systems—which are themselves arbitrary. Only this last is of real interest in the origins of grammar, and could be demonstrated through puzzle-solving rather than language use. The fact that language use has

CO22552.indb 111

23/04/2010 14:13

112

th e orig in s of g r am mar

been achieved by nonhumans is, however, remarkable in its own right and should not be ignored. Aitchison (1998, p. 35) discusses two early failures to teach chimps to speak English. Gua, trained by Winthrop and Luella Kellogg in the 1930s, never uttered a single word, although she appeared to understand a vocabulary of about 70 words. Viki, trained by Keith and Kathy Hayes in the late 1940s, managed to produce approximations to the words papa, mama, cup and up, but that proved to be the sum total of her spoken vocabulary—although, once again, she seemed to understand many more words. The reason for these failures was simply mechanical: chimpanzees do not have the laryngeal and orofacial control mechanisms to handle human speech, so their inability to produce it says nothing about their capacity to handle human language. Fouts (Fouts with Mills, 1997, ch. 2) describes a different technique of communication between humans and chimps, using American Sign Language. Allen and Beatrix Gardner taught Ameslan to a female chimpanzee called Washoe during the late 1960s. After one year she had a vocabulary of about 25 signs, according to stringent rules used to judge when she had learned a new word. She had also begun to combine them in a similar way to children at the two-word stage of language. Nonetheless, many linguists refused to accept that anything language-like was going on with Washoe (ibid, ch. 5). Fouts himself worked with another chimp, Lucy. She gained a wide and productive vocabulary, combining signs to describe new items, especially food. For instance, watermelon became CANDY-DRINK, while radish became CRY-HURT-FOOD. He also gave Washoe an adoptive baby chimp called Loulis to see whether Washoe could teach a young chimp to sign. To ensure that Loulis did not learn sign from the human helpers, all but ten signs were replaced by English—Washoe seemed to have no difficulty understanding that a change of communication channel did not affect meaning. The experiment proved successful, and Loulis learned his first 55 signs from Washoe (ibid, ch. 10). Another Ameslan experiment was conducted by Patterson, involving gorillas rather than chimps. Because of the limited motor control in gorilla hands, Patterson refers to the gorillas’ signing as GSL (Gorilla Sign Language) rather than Ameslan. Patterson’s claims for Koko, her star pupil, are impressive (Tanner et al., 2006), but so different from those for Washoe that they represent either a major difference between chimpanzees and gorillas, or a major difference in what is defined as a sign. For this reason, Patterson’s approach has been somewhat discounted by other scientists (Savage-Rumbaugh & Lewin, 1994, p. 148) as not sufficiently scientifically rigorous. These positive studies are contrasted with Terrace’s (Terrace et al., 1979) work with the chimpanzee, Nim Chimpsky. Terrace concluded that Nim was only copying the signs of his tutors; it was a simple copy-for-reward behavioural response, there was no evidence of original signing. Terrace took this to mean that not only could Nim not sign linguistically, neither could any other ape.

CO22552.indb 112

23/04/2010 14:13

What Nonhumans Tell Us about Being Human

113

However, Terrace’s position is criticized by Fouts, who placed Nim in a group of Ameslan-signing chimps in Oklahoma after Terrace had completed his experiment. At Oklahoma, Nim showed a dramatic increase in spontaneous signing as he integrated with the existing community of signing chimps (Fouts with Mills, 1997, p. 278). Savage-Rumbaugh (1999) has also taken issue with Terrace’s conclusions. Savage-Rumbaugh (1994) did her own research with a group of bonobos, although her first test subjects were two chimpanzees, Sherman and Austin. She used a keyboard of arbitrary symbols to communicate with her subjects, but her first bonobo subject, Matata, proved to be disappointing: Matata’s understanding of the symbolic nature of the task remained suspect, and inferior to that of Sherman and Austin. However, while Matata was training she was also mothering Kanzi, and he spontaneously picked up the meaning of several symbols without direct tutoring. By 17 months, Kanzi was producing novel combinatorial signs that he had not seen his tutors produce. The number of signs available to Kanzi has dramatically increased over time, and his keyboard now consists of over 300 symbols, compared to the under 100 of Sherman and Austin, the chimpanzees trained in the same programme. Kanzi has also supplemented the symbols on his keyboard with vocalized and gestural signs. In another human-chimp communication study, Premack & Premack (1983) used a magnetic board to which their subjects could attach metal objects of different colours and shapes. The study involved five chimpanzees, Sarah, Gussie, Elizabeth, Peony and Walnut. The chimps, with varying success, learned concepts such as on, arranging cards accurately to differentiate between GREEN ON RED and RED ON GREEN. They also showed understanding of the concepts same and different. All of these examples, however, serve to illustrate the empathy problem rather than the language problem. All of these animals were taught language in a pre-existing human co-operative environment in which the language game is a fit individual strategy. Many kinds of animals who work with humans seem able to acquire language comprehension skills, which is why there have been successful language-related experiments involving non-primates. Alex the grey parrot (Pepperberg, 1999), dolphins such as Akeakamai and Phoenix (Herman & Uyeyama, 1999) and Rico the collie dog (Kaminski et al., 2004) all show that co-operative communication is within the capacities of a wide range of species—as long as there is a co-operative human being at the other end of the communicative act. The question raised by these studies appears to be not “can nonhumans learn human language?”, but “can nonhumans learn human co-operation?” The studies of nonhumans using human language do tell us some important things about grammar origins, however: first, that several aspects of language grammar are prefigured in nonhuman cognition, and available for use in

CO22552.indb 113

23/04/2010 14:13

114

th e orig in s of g r am mar

communication; second, that they are only used for communication after careful teaching within a human environment; and third that, even then, nonhumans remain unable to use human language to the same level as a human five-yearold. Evolutionarily, there are both old and new aspects to human language and grammar, and both nonhumans and young humans must be studied to fully understand the origins of grammar.

CO22552.indb 114

23/04/2010 14:13

11 What Young Humans Tell Us about Being Human Anecdotal evidence about child development is ubiquitous, every parent has their stories; and scientific evidence is also common, which means that most debate on the subject has been pragmatic and data-driven. This does not mean, however, that all key theories in this area are completely experiential. For instance, Chomsky’s (1980) view, that some grammar must be innate because a child has insufficient correct input to learn it all (the Poverty of Stimulus argument), remains theoretical. It has been challenged (Sampson, 1997, pp. 38–45; MacNeilage, 1998), and this key concept of Generative Grammar remains far from proved. Nonetheless, most approaches to child development have been pragmatic. Some features of language, such as phonetic control, are innate and best explained in genetic terms; other features, such as lexis, are more easily explained in terms of socialization and culture; and features such as grammar need to be explained in terms of both genes and acculturation. There is something intrinsic about the syntax of the two-argument form, and probably about the three-argument form; whereas, for example, the English use of determiners, absent in many other languages, is clearly learned. Childhood is marked by a series of milestones—first tooth, first step, first word—so it seems to make sense to analyse it as a series of phases. Piaget (1923), was one of the first to formally do this, identifying four stages. First is the sensorimotor stage, from birth to about age two, in which children experience the world by sensing movement. The preoperational stage, from ages two to seven, is involved with the acquisition of motor skills, and it is in this stage that most language acquisition occurs. This is followed by the concrete operational stage, running from ages seven to eleven, in which children begin to think logically about concrete events. The final stage is the formal operational stage: after age eleven children become adolescents and begin to develop abstract reasoning (Smith et al., 2003). Piaget’s model proposes a fixed schedule for language acquisition, but he overlays the four main stages with sub-stages and autonomous events. For

CO22552.indb 115

115

23/04/2010 14:13

116

th e orig in s of g r am mar

instance, he puts the emergence of consciousness, a key event, between stages at about nine months (Piaget, 1947, pp. 113–114). At the other end of development, adolescence may not be the final stage: adolescent brains are still unable to fully model social reality (Sabbagh, 2006), which affects their use of language, and indicates at least one extra milestone not included in Piaget’s model. The details of Piaget’s theories have fallen out of favour with some linguists because of the generativist view that much human development is innate rather than learned (Parisi & Schlesinger, 2002). However, his basic concept, that child language acquisition occurs in a series of identifiable stages, has become canonical. Vygotsky (1934, pp. 80–83) proposed a different model, in which child development is driven by learning. This learning is seldom the outcome of deliberate teaching, it is part of the process of socialization which we, as humans, are innately prepared for. Where Piaget takes the view that thought precedes and produces language, the situation for Vygotsky is more complex. Forms of pre-thought and pre-language exist before age two, when they merge to produce verbal thought, a hallmark of being human. Vygotsky sees child development as being a continual process: different threads of cognition develop continuously and simultaneously, driven both by physical maturation and by learning—each enhances the other. For Vygotsky, new learning relies on pre-existent learning; and each thing learned opens up the possibility for other things to be learned. He describes this scope for, and limit on, new learning as the Zone of Proximal Development (Vygotsky, 1935). Piaget’s model emphasizes the stages of development and provides an approximate timetable for events to happen; but it does not adequately explain the transitions between the stages. Vygotsky’s model emphasizes the process of continuous development, but it sees child development as an individual response to specific learning circumstances, and does not really explain why almost all human children seem to pass through the same stages of development at approximately the same times. The full story of human development is therefore likely to come from a synthesis of these two ideas.

Children and Language Origins If language involves both a programmed acquisition schedule and a system for ad hoc learning, these capacities must have evolved out of something, or have already existed. So what does language acquisition by human children tell us about how the species got language? One answer is given by Deacon (1997). He sees language as a learned response moderated by genetic limitations. Although now an elegant and integrated system, language was initially disjointed and complex, a series of responses to different signalling needs. Effectively, the first languages had no single grammar, only circumstance-specific sets of rules; but these rules were refined and integrated over many generations until we reached our modern, open-ended languages.

CO22552.indb 116

23/04/2010 14:13

What Young Humans Tell Us about Being Human

117

This process, Deacon believes, was not a collective conscious attempt to make better language, instead it was the outcome of an accumulation of errors made by children over generations. These errors simplified and generalized language by making it more “child-friendly”. Deacon’s theory reverses the normally accepted train of events: language does not start simple and get complex, it starts complex and gets simple. Language is not a unitary thing that suddenly sprang into being (Chomsky, 1988, pp. 183–184), sprang into being in steps (Bickerton, 1990, p. 128) or slowly developed as a single entity (Pinker, 1994, p. 366); nor is it a thing that appeared out of a single signalling need and then expanded into other areas (Mithen, 1996, p. 213); instead, it is the merging of separated, already-existing signalling systems, each with their own complexities. Language has no primogenitor, it is a by-product of socialization. Deacon does not, however, address how or why complexity appeared in the pre-language signalling systems. Deacon’s theory does, in many ways, match the current evidence. In English, Children over-generalize linguistic constructions—for example, standard case endings like plural –s and past perfect –ed. From 1 to 3 year-olds we can hear constructs like “where those dogs goed?” and “childrens”. There even seems to be a stage where already-learned correct irregular past tenses are reinterpreted when the general rule is grasped. So a child will change from saying “ate” to “eated” and then back to “ate” (Plunkett, 1995). More interestingly, in view of Deacon’s theory, some childhood forms do not revert. Thus the past perfect form of burn used to be burnt, an irregular form left over from the Germanic roots of English. Nowadays it is perfectly acceptable to use burned—in the recent past the overgeneralization of –ed applied to burn has changed the acceptable form. Another example would be the plural forms of formula: the anglicized formulas is steadily replacing the latinate formulae. Both of these are examples of English moving from a complex set of rules to a simpler set. Deacon does not advocate a Universal Grammar, although he does not dismiss universality from his theory. For Deacon, language universals emerge from the interaction between linguistic possibility and phenotypic limitations (Deacon, 1997, p. 116). They are not the accidental side effects of cognitive evolution that Gould and Lewontin’s (1979) spandrel theory proposes, they are ability barriers that limit the possibilities available to language, as Nettle (1999, pp. 5–11) proposes. Deacon sees the major difference between our minds and those of apes as the ability to think symbolically with ease. We do not need to change mental gears to enter a symbolic thought-universe, we are born into it, and it is only with difficulty that we can change mental gears to leave it. Deacon attributes this ability to think symbolically to the fact that our prefrontal cortex is greatly expanded compared to other primates (Schoenemann et al., 2005). For Deacon, the speciation event that generated language was the appearance of symbolic thought. This, however, requires symbolic thinking to be excluded from the capacities of all other animals. It is true that, in their own environments,

CO22552.indb 117

23/04/2010 14:13

118

th e orig in s of g r am mar

no other animal has been shown to use symbolic representation in their signalling; but it cannot be excluded as a cognitive capacity. When we look at animals trained in human language we cannot explain their behaviour easily without recourse to the conclusion that some kind of symbolic representation is going on. It may be that symbolic representation is more common in animal cognition than we believe, and it is the absence of symbols from their signalling that needs to be explained. And if that is the case then it is not symbolic capacity that differentiates us from the other animals but whatever needed that symbolic capacity to become shareable. The social structure Deacon believes generated the first language is long-term pair-bonding, which he refers to as marriage. He is careful to state that the marriage he is referring to is not the Western monogamous model, it can include multiple serial partnerships; but he takes the view that these partnerships have to be monogamous and stable over long periods, and that “two males almost never have simultaneous sexual access to the same reproducing female” (Deacon, 1997, p. 385). This just does not match the data available from pre-urban cultures, (e.g. Bergstrom, 1994; Beckerman & Valentine, 2002; Goldstein et al., 2002) where stable pair-bonding is neither the economic nor the moral basis of societies. Deacon recognizes the importance of meat to hominin groups, and that co-operative meat-sharing between males and females is advantageous; but the model he proposes to explain co-operation (males give meat in return for sexual fidelity from the females) is both unenforceable at the individual level, and isolating at the gender level: males out hunting have no way of enforcing fidelity in their wives if both husband and wife are part of a social group, so they would have to sequester their wives away from the group—which means that their children would have no social group in which to produce consensuses of new language forms. However, the problems with Deacon’s proposed transmission mechanism should not be overstated. The heart of his theory, that language is in a constant state of change because of infidelities in transgenerational transmission, is clearly true. The way human children acquire language is not just a reflection of the way humans got to language, it is the way humans got to language. Grammaticalization began when the first segmented two-argument [A[uB]] utterances were made, and it continues to happen today (Hopper & Traugott, 1993).

Children and Co-operation Humans co-operate, and co-operation is at the heart of language (Grice, 1989, pp. 26–31). But are we born co-operative? Looking at the behaviour of infants it would seem to be so, and there are good evolutionary reasons to believe it is so. Co-operation seems to be a trait that defines a human being as fit in evolutionary terms (Taylor & Day, 2004); which means that, statistically, successful adult co-operators should raise more children than non-co-operators, thus emphasizing

CO22552.indb 118

23/04/2010 14:13

What Young Humans Tell Us about Being Human

119

co-operation as a genetic trait (Traulsen & Nowak, 2006). Co-operation remains a strange trait to be emphasized genetically because it is open to exploitation by non-co-operators; but if altruistic punishments like social exclusion are not too costly for the perpetrators and are sufficiently debilitating for the non-cooperators then co-operation remains the fitter strategy. We are not, however, born fully co-operative. We have certain skills present at birth that promote co-operation, such as an awareness of and attraction to other humans, especially our primary carer (Healy, 1994, pp. 15–17); and we tend to treat most of our encounters with others, up to the age of two at least, as benign. There is a default assumption, which appears to be innate, that people are co-operating with me even if I am not yet able to understand how to co-operate with them (Gopnik et al., 1999, pp. 25–31). Nonetheless, most of our co-operative behaviours are acquired in childhood. Children do not play co-operatively until about age three: below that age they use parallel play, playing together in the same place, sometimes with the same objects, but not at the same game (Brewer, 2001, pp. 30–31). Three-year-old children, even though they are willing to share an imagined universe for play purposes, are co-operating because their desire to play their game must mean that everyone wants to play: there is only one intention in the universe. At age four children are aware that others have their own intentions, and may or may not want to play. This ability to view others as having intentions is often regarded as the start of Theory of Mind, and therefore the starting point for the use of language as true dialogue rather than vocalized thought (Foley & Thompson, 2003, pp. 25–29). Before this point the child does not have an internal dialogue of “inner speech”, only external “social speech” (Meares & Sullivan, 2004). If co-operation is both innate and a product of socialization, can we get any clues to how these two effects work by looking at cases where language has gone wrong? Two types of deprivation can be studied here: sociologically deprived children—those given a grossly abnormal childhood; and pathologically deprived children—those living at the extreme end of the autistic spectrum. Feral children offer a window onto sociological deprivation, but recorded cases of such extreme child neglect are rare. The best documented case is that of Genie, who was kept isolated by her father in a world without language and human company until age 13. Her subsequent exploitation by scientists and abandonment by Social Services meant that she never experienced anything like a normal life; and, as a test of nature versus nurture, Genie’s case is hopelessly compromised. She eventually disappeared into obscurity when she dropped off the radar of scientific novelty (Newton, 2002, ch. 7). Other feral children provide similar study problems: the extreme circumstances of their lives mean that it is hard to isolate causes and effects in the abnormality of their language. However, some general conclusions can be drawn. The first is that, in all cases, both socialization and language were abnormal, and neither developed to acceptable normality. The second is that the levels of

CO22552.indb 119

23/04/2010 14:13

120

th e orig in s of g r am mar

socialization and language finally achieved varied considerably, although in all cases high language achievement mapped to high socialization, and vice versa. Third, although the level of co-operation is also highly variable, where language and socialization are high then so is co-operation; but it is impossible to judge whether co-operation is the source or outcome of language. When looking at autistic children, nature and nurture are easier to disentangle. Most autistic children are exposed to childhoods similar to those of other children, it is their response to the care and support offered that is unusual. Autistic children are desocialized in predictable ways: they seem to have a faulty Theory of Mind, they are literal in their linguistic comprehension, and they are deficient in their linguistic production. They also have difficulties co-operating with others, whether in shared enterprises or at play, preferring stereotyped patterns of behaviour to experimentation (Smith et al., 2003, pp. 477–481). There is something clearly different about the autistic brain. Frith (1993) believes a single cognitive component is damaged in the brains of autistics, and identifies this component as “the ability to think about thoughts or to imagine another individuaI’s state of mind”. For Baron-Cohen (1995, chs 4–5) the problem is more complex: there are two stimulus detectors that are significant in development of Theory of Mind. The first is the Intentionality Detector, which maps the actions of others onto a mental representation of desires and goals; for instance, a grimace indicates a dislike of something. The second stimulus is the Eye Direction Detector, which maps the gaze of others as indicating desires and goals; for instance, the observed person is likely to be looking at what is making them grimace. These stimuli are both dyadic representations of the observed agent and their goal, but they are combined in the Shared Attention Module to give a triadic relationship between the agent, the self and the goal. This Shared Attention Module sets the stage for Theory of Mind, which makes us fully human. For Baron-Cohen, autism is an impairment of the Shared Attention Module. Autistic children interpret the actions of others in a mechanistic way, and are able to identify where the attention of others is directed; but they cannot understand that there is a reason why others have goals, so there is no triadic relationship possible between self, other and goal. Theory of Mind is compromised before it begins. It is now becoming clear that autism is identifiable in terms of brain function. When performing tasks that involve modelling the minds of others, non-autistic and autistic brains have very different patterns of activation: the prefrontal cortex is heavily used by non-autistics but remains inactive in autistic brains (Carter, 1998, pp. 141–143). This area of the brain is larger in humans than other animals, and is associated with distinctly human cognitive faculties: planning, imagination, selfhood, other awareness, working memory, and spacetime cognition (Greenfield, 2000, pp. 144–153). Through the study of autism we can thus see both the significance of co-operation for humans, and what part of the brain is involved. As the prefrontal

CO22552.indb 120

23/04/2010 14:13

What Young Humans Tell Us about Being Human

121

cortex in archaic Homo sapiens was almost as well-developed as in modern humans (Aiello & Dean, 1990, ch. 10), we can say with some confidence that our speciation event did not involve a dramatic change in this area of the brain. But we can also see that, through the prefrontal cortex, co-operation is intimately tied to modelling of self and others. Do we co-operate because of our modelling, or do we model to enhance our co-operation? The answer is probably that each enhances the other: modelling allowed our ancestors to anticipate and accommodate the intentions of others, which then enhanced our co-operation; and co-operation gave us better understanding of the intentions of others, enhancing our modelling.

Children and Selfhood The link between language acquisition and selfhood has a long pedigree. For Wilson (1937, pp. 143–146), the emergence of consciousness of self was the defining event in becoming human. Piaget (1923, pp. 39–43) describes young children as linguistically egocentric: their utterances mostly involve their internal monologue rather than social dialogue, so their inadequate model of selfhood makes for a suboptimal use of language. For Vygotsky (1934, pp. 217–235), the emergence of consciousness is gradual, a series of emergences of “consciousness of”. Egocentric speech is the first step in the internalization of language, and is replaced at about age five by inner speech. Only when this has been achieved does self-consciousness begin. We know from our own experience that knowledge of self is not present from birth: few if any humans remember their first year of life, and memories before age four are usually disjointed and isolated from the life-memories we use to define our self. In the first six months, babies are unaware that they are individuals and seem to treat other people and objects as physical extensions of themselves (Brewer, 2001, pp. 17–23). Between six months and two years, the infant is acculturated by their care-givers, what Kaye (1982, p. 205) describes as a development “from an organism to a person”. After age two there is an identifiable self being asserted, and this often creates a period of carer-child conflict referred to as “the terrible twos” (Brewer, 2001, pp. 218–222). Gopnik et al. (1999, ch. 2) provide a comprehensive description of the developing child’s knowledge of self and others. In the first few months the baby is building on innate knowledge. It knows about human faces and voices, and it is busy learning to identify its caregivers—the baby is learning how to identify others. It also begins to learn about expressions, and how they relate to the caregiver’s behaviour. By age one, infants begin to see others as agents— other people attend to, and have an effect on, the world. The child learns about pointing and eye direction, and will look at the objects being attended to by others. Additionally, they are learning that the world can be affected remotely if they can recruit the muscle-power of others.

CO22552.indb 121

23/04/2010 14:13

122

th e orig in s of g r am mar

By eighteen months the child is usually aware that the usefulness of others can be variable—sometimes they will help, other times not. The child is also beginning to understand that others may not know everything the child knows, so giving as many clues as possible will facilitate satisfaction of the child’s wishes. It is probably not coincidental that children begin to use indexical word labelling at this age. Empathy for others begins at about age two, and it is the conflict in the child’s mind between getting their own way and pleasing others that leads to the tantrums. At age three the child is beginning to take control of their emotions. They are also fending for themselves in important ways, such as feeding, washing and dressing: given the raw materials, they will have a go at finishing the job. At this age they are also learning about deception. They are aware that others have beliefs different from their own, and that these beliefs can be manipulated— although their attempts to deceive are usually hopelessly ineffective. By age four, children are recognized to have a complete Theory of Mind: they can make effective guesses about what others are thinking. They also learn two important lessons at this stage: successfully deceiving others who are aware that they can be deceived is a costly process in terms of cognition, involving deception in multiple modalities (Wray, 2002a, pp. 128–129); and other humans are more than willing to altruistically punish those they discover deceiving them (Fowler, 2005). In Gopnik et al. (1999) we see a steady increase of awareness, in lockstep with socialization and co-operative behaviour. The young baby has no need for a concept of self or other, but the ability to co-opt the muscle-power of others—and the willingness of parents to have their muscle-power co-opted—soon creates a situation where modelling others becomes advantageous. As the child grows it learns that others are not just objects to satisfy demands, they are agents who may or may not assist in particular circumstances. This leads on to the concept of manipulating others, the Machiavellian intelligence of apes. For humans raised in a highly co-operative linguistic culture, however, the modelling of the child by others also becomes obvious to the child. The simple sentence “let’s go to the park” requires the child to understand that the adult has a model of the child in a different place and time to now. Comprehending that others are making models of you allows you to make a model of yourself. This self-modelling is not a direct representation of the self, though: your model of you is actually your model of their model of you (Gopnik et al., 1999, p. 47). From the modelling of others, the modelling of the self as a first-person agent becomes possible. We begin to build models of what others think about us, and of what we think about ourselves: we define and build our own individual personalities. Co-operation, socialization and culture drive us towards language, language drives us towards self awareness, and self awareness drives us towards further co-operation. Our innate need to co-operate is the engine that powers language acquisition and Theory of Mind; but it is the co-operative culture of altruistic punishment that creates the environment in which these systems can flourish.

CO22552.indb 122

23/04/2010 14:13

What Young Humans Tell Us about Being Human

123

Children and Language Did species grammar genesis follow a roughly similar path to childhood grammar acquisition? The mental structures of grammar in the first ever language are likely to be the same mental structures that a child uses today, so it should be possible to identify the same stages of syntax acquisition in language origins and in child language. Additionally, the child’s mental modelling development should be mappable to the same checkpoints as grammar. These checkpoints are not analogue—we would not expect to encounter two-and-a-half word utterances— so a Piagetian staged approach should work. Several attempts have been made at producing a staged model of language acquisition. Brown (1970) proposed four stages: pre-language up to about six months; one-word utterances up to about 18 months; two-word utterances to about three years; and full language starting at about age three. Hirsh-Pasek & Golinkoff (1996) produce a slightly different agenda: up to about nine months the child is associating sounds with things; up to about two years they are matching words to things and learning about attention; up to about three years they are dealing with two-argument forms; and from then they are dealing with full language. Halliday (2004, ch. 14) identified three phases of language development: language learning, which is dominant up to about two years; learning through language, which starts at about two years of age; and learning about language, which starts at about age four. For Halliday, language is not a skill fully achieved in the early years, it continues to be acquired throughout life. The phase shifts of language given by all the models above are highly variable, and nobody would back a species-specific calendrical schedule for language learning (Bates et al., 1995). However there is agreement that there are several stages in language acquisition; and, more importantly, there seem to be detectable phase changes between the stages—as if a new set of rules is being learned and applied, sometimes replacing the old set and sometimes supplementing it. So is there a schedule that will cover language acquisition at both the individual and species levels? Let us start with a six-stage model of how children move from sounds to language. • The pre-language stage: a sound is a sound. It can attract attention but there is little differentiation between sounds. • The phonetic stage: some sounds appear to elicit better responses from carers than others. These are the language sounds that the carers associate with “being human”, but the child is not aware of this (Berg, 1972, p. 7). • The word stage: sounds have meanings. Objects can be requested or named with sounds, and personal wants can be better met by making the sound associated with the effect desired. Sounds, however, are indexical at this stage and not symbolic. • The one-argument stage: words can be combined to produce more accurate

CO22552.indb 123

23/04/2010 14:13

124

th e orig in s of g r am mar

requests and enhanced outcomes. Some of the sound combinations are recognized by the child to be segmented words which can be used in a range of circumstances (Tomasello, 2003a, pp. 139–140). There is also a growing recognition that different word combinations have different effects: kiss teddy means that mummy is to kiss teddy; teddy kiss means that teddy is to kiss mummy (Gopnik et al., 1999, p. 117). This seems to be the highest stage reached by animals taught human language in a human cultural environment (Tomasello, 2003b). • The two-argument stage: for every action there is something active and something being acted on. Mummy kiss teddy is different from teddy kiss mummy; and while juice in cup is one outcome, juice in bottle is another, although they both involve getting juice. • The full language stage: the child’s syntactic knowledge is largely complete. While complex language forms still need to be learned, no new syntax is needed to understand them—although there will sometimes be individual rule exceptions to be learned. In terms of grammar genesis, this model largely maps how we would have got to language as a species. The first two stages are clearly prehuman, and the word stage seems to be available to all primates. The one-argument form seems to be cognitively available to apes, but is only communicated when they use human language. Apes may also be able to use two-argument forms in cognition, which would certainly enhance their social intelligence; but examples of communicative use, even among apes using human language, are rare and disputed. The final stage of full language does seem to represent both a cognitive and communicative difference between humans and other animals. Humans have a predisposition towards language. This has evolved over the generations of human selection within a linguistic culture, and it is likely to be at least partially encoded at the genetic level: humans born predisposed to language are likely to have had greater reproductive success than those born with less sensitivity in that area. Language, and the forms of co-operation enabling it and enabled by it, have been fit strategies for our species for generations, and we appear to have adapted to them both. Language, however, remains largely learned, an incremental building of personal knowledge; and the stages by which a child learns its language inform us about the way humans became languageusing animals.

CO22552.indb 124

23/04/2010 14:13

12 What Time Tells Us about Being Human Time is ubiquitous ,

it is not a product of human cognition or evolutionary imperatives. Like gravity and light, it is a key component of the universe, defining everything in relation to itself. Time, in cosmological terms, is just another dimension, similar to the three dimensions that define space; but the nature of time means that it is perceived by humans in very different ways to those other dimensions (Reichenbach, 1956 [1927], pp. 109–113). The first way we perceive time is as a journey. Every movement in the three dimensions of space automatically means movement in time; but, unlike space, every movement in time occurs in only one direction and at only one rate. There is some theoretical evidence that movement is possible from present to past, or to the future at an accelerated rate (Hawking, 1996, ch. 10), but this time travel is not part of our everyday human experience. Instead, we have the impression that we are all heading inexorably into the future at the same rate: the journey of time is both communal and inescapable. The second effect of time upon human perception is that it is indirect: we do not see, hear, or feel time but we are nonetheless aware of its passing. Unlike the other dimensions, we cannot know time directly; we can only know it in terms of metaphors, usually of the other three dimensions. We can view it from inside time as a line ahead and behind us, with either the self as fixed and time flowing around us, or with time as a fixed path and the self moving along it. We can also view it from outside time as a horizontal line from left to right in front of us, with us or the line moving; or we can view the line as vertical instead. We even view time as cyclical, moving regularly through a repeating cycle—a representation used in rotary clocks and watches, creating a key Western metaphor of time, the cyclical day; and we can see time as helical, going through cycles that nonetheless move forward and do not repeat exactly. The third human perception of time is its directionality. We move through time (or time moves through us) in one direction only, from a remembered, recorded past to an unknown future. Past and future are, as directions, very different experiences from the directions of 3-D space: we see where we are going



CO22552.indb 125

125

23/04/2010 14:13

126

th e orig in s of g r am mar

in space because it is in front of us, and we can turn and see where we have come from equally well. Time does not have this bidirectionality; we can see only where we have come from, we cannot see where we are going. This is not a trivial effect, it can create very different cultural views of time, as we saw in chapter 5: for the Aymara, the past is what they see, so it must be in front of them; the future cannot be seen, so must be behind them (Núñez & Sweetser, 2006). This directionality of the body as a metaphor for time reverses our Western experience of the future being ahead and the past behind; a different cultural perspective produces different analyses of the same metaphor of time. All of these representations are metaphors of time as space, allowing us to create clocks, calendars, schedules and timelines. It is this metaphor that allows us to forecast, to model the future; and to remember, to model the past. It is a conceptualization that relies not just on the metaphor TIME IS SPACE (Evans, 2007) but on the capacity to model the self as an entity in the past and future, and to see the current self as a continuity of these modelled selves. Self-modelling must be part of cognition before effective modelling of time becomes possible.

Getting Tense Our perceptions of time dictate the way we model self into time, and govern the mechanisms we use in language to express temporal relationships. It is, however, possible to have a sense of time without self-modelling: the unmodelled self is fully capable of envisioning events in the past and future. A sense of time can even be generated by a simple punishment and reward system using memory, forecast and the uniformity of reality. For instance, wasps will sting in the future because they have stung in the past, so avoid wasps. There is no need for a modelled self being stung to make this planning effective. What happens, though, when the capacity to model the self into the future or past appears? The simple temporal model, consisting of two components (the unmodelled self fixed in the present and the memory or forecast of the event in the past or future), becomes extendable. If a modelled self can be projected into the past or future then it can, like the unmodelled self in the present, have memories and make forecasts. For instance, the self modelled into the future can have memories of events that are in its past but still in the future for the unmodelled self in the present—and, therefore, actually unknown to the unmodelled self. This ability to refer to events in the future as if they are in the past or future of a future time, and the ability to refer to events in the past as if they are in the future or past of a past time, are described by Reichenbach (Areces & Blackburn, 2005) as an interaction between point of speech (S, always the present), the point of event (E, when the action of the verb actually takes place), and the point of reference (R, which corresponds to the location of the modelled self). This is a powerful model which encapsulates the vast majority of what languages do with tense.

CO22552.indb 126

23/04/2010 14:13

127

What Time Tells Us about Being Human

Using Reichenbach’s three points, seven natural tenses emerge: the present, in which the event, the modelled self and the unmodelled self are telescoped into one time point; the past, requiring memory of an event that has happened; the future, requiring forecasting of an event going to happen; the past of the past, requiring a memory as a feature of a self modelled into the past; the future of the past, requiring a forecast as a feature of a self modelled into the past; the past of the future, requiring a memory as a feature of a self modelled into the future; and the future of the future, requiring a forecast as a feature of a self modelled into the future. These seven tenses are represented in different languages in different ways. For instance, in English they are formed from a mixture of inflections and auxiliaries: I do, I did, I will do, I had done, I was going to do, I will have done, and I will be going to do. Other languages use different methods, but expression of these seven tenses seems to be possible in every human language; the capacity to use them is certainly seen as a marker of successful language use (Paradis & Crago, 2000). Compare this to bonobos using human language, where simple past and future tenses are within the capacities of some, but not all, of the apes; and where tenses using a point of reference outside the present seem to be beyond them (Savage-Rumbaugh et al., 2005). The seven tenses do not exhaust what we are able to do with Reichenbach’s three points. Proximity allows us to use the modelled self at the point of reference as an assumed present. We can thus use expressions like Daddy’s taking us to the zoo tomorrow: the present tense indicates that the event is current, but the word tomorrow tells us that it is actually going to happen in the future. Similarly, reportage often uses the present tense to create immediacy in a story about past events (for instance, It’s the eve of Waterloo; Napoleon is in his tent . . .). Proximity adds two more tenses to the original seven, as Figure 12.1 shows. Tense

Point of Speech (S) Unmodelled self

Point of Reference (R) Modelled self

Simple Present

Point of Event (E)

Present

Simple Past

Present

Past

Simple Future

Present

Future

Past of Past

Present

Past

Past of (R)

Future of Past

Present

Past

Future of (R)

Past of Future

Present

Future

Past of (R)

Future of Future

Present

Future

Proximate Past

Present

Past

Proximate Future

Present

Future

Future of (R)

Figure 12.1—The Nine Tenses

CO22552.indb 127

23/04/2010 14:13

128

th e orig in s of g r am mar

All of these tenses can be viewed as using Reichenbach’s three points, but with movement of the modelled self to produce the simple tenses and the proximate tenses. Simple past and future merge the point of reference with the point of speech in the present; proximate past and future merge the point of reference with the point of event in the past or future; and present tense merges all three points in the present (Hornstein, 1990, ch. 1). Reichenbach’s three point analysis thus gives us an effective way to describe the key tenses used in languages. Of course, it is possible to extend this system by adding a point of reference to the point of reference, producing four-term constructs such as I will have been going to do; but these constructs are not easy to understand, and tend to introduce effects that are not strictly tense-related. For instance, the difference between I have been going to do and I had been going to do is that the first indicates an intention continuing into the present, while the second indicates a former intention now abandoned. There is no true four-point analysis in I had been going to do, other temporal effects are at work.

Doing Other Things with Time Reichenbach’s three-point system encapsulates the way time is expressed in language through modelling of self into past and future; but it can also be used to illustrate a series of other linguistic temporal effects. The first of these is continuity. The point of event can represent a single complete event (I wrote a letter); a single ongoing event (I am writing a letter); or one of a series of events (I am writing some letters). It can even refer to a series of events of which none are happening at the point of event (I write a letter every week). Continuity therefore adds width to the point of event, allowing it to express duration as well as point in time. A second linguistic temporal effect is imminence, which is about the distances between Reichenbach’s points. A point of event can be close in time to a point of reference or more distant; and a point of reference can be close to or further from the point of speech in the present. The nine tenses given above dictate the temporal ordering of point of event and point of reference, but imminence determines the distance between them. Although imminence is in reality highly variable, in many languages only near and far are recognized. In some East African languages, there are two past tenses to indicate imminent and non-imminent events (Lee, 1992, p. 9), and this is partially the case in English. For instance, in the sentences I wrote a letter and I have written a letter, the point of reference is the same (the present) and the point of event is also the same (the past); but the point of event of the second sentence is closer to the present than the point of event of the first. Imminence can occur in the future, too. In the sentences I will write a letter and I am going to write a letter, the point of reference and point of event are both the same (the present and the future respectively); but, once again, the point

CO22552.indb 128

23/04/2010 14:13

What Time Tells Us about Being Human

129

of event of the second sentence has greater imminence. In this case we can create even greater imminence with I am about to write a letter, indicating that, in English, imminence is not just a binary dichotomy of near and far. In English, imminence is often expressed with relative adverbials, like soon and just. It can also be indicated by absolute adverbials, like tomorrow and last week, or with prepositional phrases, like by tomorrow or before next week. The relative adverbials tend to affect the distance between point of reference and point of event, while the absolute terms tend to affect the distance between point of reference and the present. Thus, in tomorrow, I will have almost finished it, almost indicates that the point of reference of tomorrow is close to the point of event of finishing, while tomorrow fixes the distance between the point of reference and the present as one day. Almost has a second role, converting the event from completed at the point of event to incomplete. It therefore also has an effect on the continuity of the construct, showing that temporal effects cannot always be isolated linguistically.

Adding Depth Temporality is not just limited to individual events, it is also involved in defining order between events. With language, each signalled event is no longer isolated, it can be connected to other events in the past or future. This process, here called connectivity, is not an expression of temporality within a single language construct, it creates temporal connections between constructs; and it therefore strongly corresponds with the Systemic Functional Logical metafunction. Connectivity, in terms of temporality, is the feature that facilitates the never-ending discourse of language. Temporal connectivity can identify events as contemporary or sequential. For instance, in he looked and listened, he looked before listening and he looked after listening, the connective determines the order of the events. Connectives can also place identities into a time series: in he ate the plum, then the peach and finally the banana, the event, eating, is being applied to a series of objects in turn. While some connectors, like after and before, explicitly create the temporal relationship between events, this is not true for all connectors. For instance, in he jumped on his horse and rode into the sunset we see and as linking two events serially: both events are in the past, but the first has to happen before the second can occur. In comparison, in he sat on his horse and stared at the sunset the two actions are probably contemporary. To convert the actions to a series we would use and then or just then. Our knowledge of context is, therefore, at least as important as the words uttered in determining temporality. Connectives allow events to be placed into a structured temporal relationship, a capacity at the heart of human story-telling. This is no small side-effect of language, it is central to it. Every time we make models we are telling ourselves a story, extrapolating existing circumstances through a net of possibilities to reach

CO22552.indb 129

23/04/2010 14:13

130

th e orig in s of g r am mar

a conclusion; and if our first story does not end as we wish we can model others until we get the result we want. Continuity works at the point of event, determining the duration of the event; and imminence works between point of event, point of reference and point of speech, determining the distances between the points. This gives a rich single dimension for linguistic expression of time. Connectivity adds another dimension to temporal space which allows individual events, each with their own timeline, to be linked together. Connectivity is perhaps the most important dimension in terms of narrative: without the ability to link constructs together logically and semantically, dialogue becomes an exchange of unrelated facts and narrative becomes impossible. Connectivity links single utterances together into the continuous interpersonal narrative that language has become, and it is therefore a clear differentiating feature between human language and other signalling.

Time, Uncertainty and Fiction There is a third dimension of temporality, which is concerned with how language deals with the certainty or uncertainty of events. Conditionality allows events to be placed onto a vector of probability, which works with the other two vectors of continuity and time itself. In English, conditionality is mainly expressed through adverbials, it has only limited expression through auxiliary verbs. For instance, I may have done and I may do are permissible English forms, but *I may had done and *I may will do are not. With adverbials the range of temporal expression is wider: perhaps I had done, I will possibly do, I have likely done, I was probably going to do, maybe I will have done, hopefully I will be going to do . . . These all add uncertainty onto pre-existing verb constructs. Because our experience of past and future time is non-symmetrical, the effect of conditionality in the past and future is somewhat different. Events in the future of the point of speech already have uncertainty in that the future, by its nature, is unknown, and adding conditionality only increases the uncertainty. Events in the past, in contrast, have greater certainty, and adding conditionality can convert certainty into uncertainty. This is why conditionality in the future tends to be about volition, establishing personal control over an undetermined future, while in the past it is about review—and often regret. Auxiliary conditional verbs (may, could, should, etc) also reflect the asymmetry between past and future, and the replacement of will with may illustrates this particularly well. I may have done does not express the same temporality as I will have done: while will expresses a point of reference in the future, may causes the point of reference to merge into the present. It seems as if this form of conditionality moves the point of reference through the vector of probability instead of through the time vector, which indicates that seeing conditionality as a separate dimension of temporality is a productive metaphor. What does the vector of conditionality give us? In terms of the future, it allows

CO22552.indb 130

23/04/2010 14:13

What Time Tells Us about Being Human

131

us to plan, to choose between a range of alternatives; in this form, therefore, it probably existed before Homo sapiens. If we look at tool-making as an indicator of planning then it is indeed ancient. Experiments with New Caledonian crows (Corvus moneduloides) have shown them capable of planning the retrieval of difficult-to-access food, by making the tool necessary to achieve access and then using it appropriately. In the laboratory the crows worked with unfamiliar materials and an unnatural environment, but they were still able to bend a metal strip into a hook and use the hook to lift a pot of food out of an otherwiseinaccessible hole (Weir et al., 2002). Attributing to the crow the capacity to plan gives the simplest and most likely explanation for this behaviour. Other experiments with chimpanzees have shown them capable of working together in tasks that require planning for co-operative activity (Melis et al., 2006), although they do seem to be better at planning for competition than for co-operation (Hare & Tomasello, 2004). The full power of conditionality, however, only becomes available with the capacity to model the self into past and future. If the unmodelled self in the present can model what-ifs in the future then the models of the self projected into the past and future can also model what-ifs; so the self modelled into the past can model conditionality into a future which is still the unmodelled self’s past. From this modelling into probability space comes all our fiction—and, indeed, a lot of our history. We can model from known facts to possibilities and, if enough facts point in the same direction, we can develop a consensus view of what has probably happened. The old Soviet adage that “the future is certain, it is the past we cannot predict” is, for historians, too real to be funny. Nonetheless, the power of fiction, unleashed by conditionality merged with self modelling, has been a powerful and defining feature of being human. We are a story-telling animal (Niles, 1999), probably the only one; and, if our storytelling is indeed unique, it is an important difference between us and other animals.

Becoming Time-aware We are not born with a full model of temporality: children under the age of about four seem to divide the world into “now” and “not-now”. Not-now has a considerably reduced value compared to now, so events in the not-now are less valuable than now events. In a televised experiment, children of various ages were asked to choose between having a single piece of chocolate now or a bar of chocolate in 10 minutes: the under-fours universally opted for the small piece now, while the over-fours universally opted to wait (Brewer, 2001, pp. 179–183). In the terminology used here, it is reasonable to view the under-fours as unable to project a modelled self into the future to accept the whole bar, all they have is the unmodelled self in the present to accept the single piece. The way children come to understand temporality and use it in language is, however, progressive rather than instantaneous, there are several stages in the

CO22552.indb 131

23/04/2010 14:13

132

th e orig in s of g r am mar

mastery of tense forms. Weist (1986) describes four stages: for the first 18 months the child talks only about events in the present; from 18 months to 36 months they refer to past and future using simple tenses (-ed morpheme and auxiliary will); from 36 to 48 months they use temporality involving adverbials, but only where they agree with the simple tense (e.g. We went to the park yesterday); and from 48 months they begin to use complex tenses involving a separate point of reference. This is a somewhat simplified view of tense acquisition—and, of course, nobody expects all children to achieve these milestones at exactly the same time: Shirai & Miyata (2006) show that, as well as individual differences, the language being learned affects the timetable of the learning process. This tense acquisition schedule does show, however, that modelling of self, needed to use the point of reference effectively, is not something immediately available to the infant. Acquisition of temporality seems to involve a range of processes: tense seems to be acquired to one timetable, aspect (equating to continuity and imminence) to another (Valian, 2006). There are also links between aspect and the nature of the action, with the perfective (non-continuous) past tense being used by young children for verbs that change states, such as stop, hit, give; and the imperfective being used for activity verbs, such as playing, doing and making. There also seem to be issues with the direction of time, with future tense being harder for young children to understand than past tense; and they also seem better able to grasp near future than far future. Additionally, young children tend to use the gonna future tense only for their own plans, with other future events being indicated by will (Tomasello, 2003a, pp. 217–224). Clark (2003, pp. 258–261) shows that younger children seem to have a default rule that events mentioned first occurred first; so sentences like he leapt after looking and before he leapt he looked tend to be misinterpreted at 36 months, although by 54 months the correct rule is usually being applied. The process by which children acquire tense and aspect seems, therefore, to be more complex than just a matter of learning the rules; it relies on how the action of the verb is viewed, the perceived difference between past and future, reasonable but sometimes faulty syntactic hypothesizing, and other factors. Learning temporality in language seems to be tied to understanding meaning and intention; but it is also heavily reliant on the capacity to model events into the past and the future, and on the flexibility that modelling the self gives in terms of recursive modelling.

Three Time Points, Three Voices? Do Reichenbach’s three points correspond to the three voices in language? These voices are the self (me, the sender), the directly addressable non-self (you, the receiver), and the non-self that is not directly addressable (them, the referenced objects). Bloomfield (1933, p. 224) refers to these as “speaker, hearer and third person”, and we traditionally refer to them as the three persons, or the three

CO22552.indb 132

23/04/2010 14:13

What Time Tells Us about Being Human

133

voices. The three voices express something fundamental to all languages, possibly something fundamental to language as a communicative device (Benveniste, 1970, pp. 217–221): the need to explicitly reference the sender and receiver at each end of the signal, and to separately identify the subject matter of the signal itself. Every human language so far discovered and described has the capacity to indicate these roles with a special class of words which can collectively be described as pronouns. Different languages use different ranges of pronouns, but they all come down to the three voices of the sender, the receiver and the referent. Pronouns usually have singular and plural forms, although the range and nature of plurals in each voice also varies from language to language. If we look at these three voices in terms of Reichenbach’s points we can see that there is some correspondence. The sender is always at the point of speech, in the present; and the referent is always at the point of event. The receiver, however, is not so easy to allocate: the receiver is in the present with the sender for spoken signals (although in the future for writing), and they are being invited by the sender to model the points of reference and event indicated by the utterance. The receiver and the point of reference do, however, share one feature in the mind of the sender: they are both models with the capacity to make their own models. The sender invites the receiver to join them in modelling a self at the point of reference, thus permitting that modelled self to model the point of event. It is this collusion in modelling at the point of reference that associates the receiver with the point of reference. The correspondence between Reichebach’s points and the three voices is therefore approximate, but it does demonstrate that self-modelling could be behind both processes. It is probably no coincidence that there are three voices (I, you and they), three objects to a signal (sender, receiver and referent), and three points in Reichenbach’s tense model (points of speech, reference and event). At some level there are correspondences between the three components in each of these models.

Time and Being Human We have seen that temporality in language is a complex of functions for describing point of reference, point of event, continuity, imminence, connectivity and conditionality. In English it is served by inflection (-ed, -ing) and auxiliary verbs (was, have, will); and by time-related words (after, yesterday, tonight, eventually, and so on). This array of functions is able to express all the complexity needed for human temporal expression. Although temporal expression varies between languages, behind those variations are universal constraints based around the basic tools of language—the noun-verb distinction, the three-argument “subject/ verb/object/indirect object” form, and the modelling of self and other. Because the range of tools is limited, the range of rules that can be derived from them is also limited—but the rules actually derived do not themselves need to be universal. Time and its expression in language, temporality, are key to understanding

CO22552.indb 133

23/04/2010 14:13

134

th e orig in s of g r am mar

what makes us human. Because we are able to make models of ourselves, we are able to overcome the problem of viewing everything from the present. By modelling ourselves as future or past entities we create the twin possibilities of planning and reviewing our actions; and by modelling ourself as a model of a model, we can review our plans before they are enacted, or plan them after they have occurred. Our models are not tied to the present in the way that our unmodelled selves are. Self-modelling gives us a type of recursion, the capacity to model a self modelling a self. We can project our first level models into our past and future, and then project the second level model into the past or future of the first-level model. When it appeared, the sharing of our models of self and others is likely to have enhanced our already-high level of co-operation; and this, in turn, enabled us to plan co-operatively, and gave us access to what Donald (2001, pp. 269–271) describes as the Plan-Execute-Review mimetic cycle; and, because of the enhanced co-operation, we were able to use the Plan-Execute-Review cycle as a shareable experience rather than just an internal, personal one. With self-modelling we became able not just to share our plans but to work together in novel ways, exploiting new levels of socialization and co-operation. Language is clearly a product of, and enabler for, high-level co-operation; and it is this extreme co-operation which creates the willingness to give and receive truthful messages. Robert Burns was almost right when he said in his poem, To a Louse: “O wad some Pow’r the giftie gie us, to see oursels as others see us!” But it is the ability to see ourselves as we see others that is the reason why we have the full complexity of language. For Chomsky, the capacity for recursion is the key difference between human and other minds (Hauser, Chomsky & Fitch, 2002). However, as Dickins (2003) points out, recursion cannot stand alone, it must operate within a system. It requires exchangeable tokens or symbols, a process of exchange which transmits meaning as well as the symbols, and a structure in which form can recur without meaning recurring. If the theory proposed in this book is correct, then recursion is only an emergent iterative property of the modelled selves, a property which is exemplified in grammatical temporality: the Russian dolls of self modelled within self form a potentially infinite (although in practice profoundly constrained) recursion, and the orders of intentionality they create form the basic structure for recursion throughout language (Dunbar, 2004, pp. 47–69). Recursion is not, therefore, the source of all things linguistic; instead, it is a product of self modelling, which is in turn a product of the sharing of multiple-argument models through language.

CO22552.indb 134

23/04/2010 14:13

13 The Evolution of Grammar and other as a route to language grammar, but it has taken the unusual approach that modelling of self is an outcome, not a driver, for modelling others. While unusual, this approach is not completely unsupported: Nichols & Stich (2003) show that Theory of Mind does not require self-awareness, and that other-awareness and self-awareness may be the product of two distinct processes; and Carruthers (2009) shows that it is likely in evolutionary terms that mindreading (required for modelling others) preceded metacognition (the capacity to think about thinking, required for modelling the self). This does not mean that there is no sense of self without metacognition, knowledge that the world is divided into self and non-self has its own fitness advantages and is therefore likely to be selected for by evolution. In this division, however, it is more useful to recognize the non-self rather than the self: autonomic control ensures that the self part of the universe usually operates optimally for the self; it is the uncontrolled rest of the universe that needs active, and sometimes conscious, intervention. This means that the ability to model others, despite having a large cognitive cost, can become a useful capacity because it allows more accurate anticipation of the actions of those others—it provides effective objective information. The ability to model the self provides subjective information, which does not require the same type of accuracy: it is often better (and, therefore, evolutionarily fitter) to have an inaccurate self image than an accurate one (Epley & Whitchurch, 2008). If we view the key features of language grammar as modelling of others (which is likely to be common in nature) and modelling of self (which seems to be rare), are we able to use this to generate a description of how language may have developed? The answer is a definite maybe: we can certainly plot a series of grammatical enhancements in the development of language, placing them into a dependent time series; but we cannot yet put a timescale on when these enhancements occurred, nor how long the gap was between each enhancement. Nonetheless, communication of the modelling of self and others does seem to be implicated in a range of grammatical forms, and its development does tell a coherent tale about how grammar came about. This book has been about modelling of self



CO22552.indb 135

135

23/04/2010 14:13

136

th e orig in s of g r am mar

Basic Communication The first event we need to consider on our path to grammar is the nature of the pre-existing signalling system. It is not unreasonable to presume that any signalling capacity currently demonstrated by other primates was within the capacity of Australopithecines, four million years ago; and that is a presumption that does indeed give us a wide range of signalling possibilities. It includes the imperative mode of address: informing conspecifics to behave in particular ways by vocalizing rather than by physical intervention is widespread among primates. Those who achieve alpha status often use the standard ordering form of the imperative, but the negative imperative of submission is even more common. There are also the alarm calls of many primate species which, as well as containing reference to the object being warned about (at least for the sender), act as instructions to the receiver to carry out a course of action (Cheney & Seyfarth, 1990, ch. 5). The prevalence of courtship, territorial, threat and alarm calls mean that instruction to specific action is a fundamental feature of many signalling systems (Bradbury & Vehrencamp, 1998, ch. 18). Australopithecines are also likely to have had the capacity for segmented signals—signals containing more than one meaning-unit. Male diana monkeys (Cercopithecus diana) have different warning calls to indicate threats from leopards and eagles; but they also seem able to convey information about direction and distance (and, therefore, threat level) by modulating the start of the call (Zuberbühler, 2000). This represents a level of segmentation in that modulation of part of the call alters the intensity of the call. Similarly, apes taught to use human language seem to have no problem with the segmentation of a construction into units representing different actions and objects, and there is even some indication of syntactic rules in Kanzi’s utterances (Savage-Rumbaugh & Lewin, 1994). While signal segmentation may be a latent ability in wild chimps and bonobos, the fact that it can be accessed through acculturation to humans indicates that it must be there to be accessed, and that it probably serves a pre-existing non-communicative cognitive role. Another important capacity that chimpanzees trained in language can demonstrate is differentiation between types of segments in utterances. They understand the basic object-action distinction, that some symbols represent things and others represent actions done to things; they understand comparators like same and different, colours, and descriptors like sliced; they correctly interpret prepositions like on, under, above, below; and they display some comprehension of conditionals, such as “if this happens then that happens” (Premack & Premack, 1983, ch. 8). In language terms we would say that they have the grammatical concept of word classes, but we should be wary of assigning conscious knowledge to what they do. The word classes may be used appropriately merely because of collocation, the words around them promote their correct use and suppress their incorrect use; but even this implies that some kind of rule system has been internalized.

CO22552.indb 136

23/04/2010 14:13

The Evolution of Grammar

137

Grooming also should not be underestimated as a signal (Dunbar, 1996). It is a highly socializing activity, being part of the important process of forming and maintaining alliances; and it is clearly pleasurable for both parties. It is often a mutual activity, too, making it both interpersonal and interactive. It is an activity that places individuals in an extended dialogue that is language-like in many ways, and it is an important precedent for the extended exchanges of language— as can be attested by anyone who has experienced the warm feeling of calming pleasure that a good conversation brings. The imperative form, segmentation, differentiation and grooming give a rich environment in which hominin signalling could develop; and we can assume that, even if the australopithecines had only very limited vocal production, the range of their vocalizations was potentially quite sophisticated. Of course, we can never know whether these capacities became actual communicative realities; but we can say that a simple grammar was within the competence of Australopithecus and Homo.

Social Modelling The appearance in cognition of relationship modelling, in the form of A-relationship-B, is another capacity with significance for language. This modelling is a useful tool for keeping track of interactions in a medium-sized group, although it does rely on the pre-existence in cognition of representational tokens, and segmentation of thought. Relationship modelling is particularly important in Machiavellian social environments, where individuals can enhance their personal fitness through a series of alliances. Each primary relationship with another individual becomes a secondary relationship with their allies; and, in every social species, picking the right relationships is a vital part of survival and successful reproduction. So A-relationship-B social modelling can become cognitively important for a species living in moderately sized groups with high levels of social interaction. There are two ways of cognitively encoding these three-part models: each of the A-relationship-B sets required to model the social network of the group can be separately encoded as unsegmented tokens; or a basic token can be allocated to each individual and to each possible relationship between individuals, and segmented metatokens can then be generated to represent the A-relationship-B constructs. Clearly there is a cognitive overhead to the segmented approach, so what are the likely advantages that offset this? The simpler relationship-B model needed to record relations between the unmodelled self and others works better using unsegmented tokens: simple emotional reactions triggered by the presence of the individual save cognitive load in terms of both segmentation and modelling capacity. This means that there is something very different in cognitively modelling my relationships with others and modelling relationships between others: my relationships with others

CO22552.indb 137

23/04/2010 14:13

138

th e orig in s of g r am mar

are identified directly with those others; each relationship is not just a representation of the emotion in the relationship, it is the emotion. In A-relationship-B modelling, the relationship represents an emotional state which is abstract, and which is unrelated to my own emotional state. Attending to the emotion in A-relationship-B forms should not actually elicit the emotion—it is not, after all, my emotion but that of A. My reaction to my model can be visceral, and very different from the reaction of A to B within my model; as Erica Jong says, “jealousy is all the fun you think they had”. The segmented metatokens of the A-relationship-B model also permit social calculus. In unsegmented modelling, A-relationship-B, B-relationship-A and B-relationship-C can be compared only as holistic units; the fact that B is common to all three forms is incidental. Segmented modelling, on the other hand, allows the modeller to compute A’s likely relationship to C, based on the information already known about A, B and C. The more that is known about relationships between individuals, the more accurate are the predictions that can be made about unknown relationships. Knowledge is power in a Machiavellian social environment, so acquiring and accumulating social knowledge makes an individual, in evolutionary terms, fitter for their environment. Although this type of calculus has its own cognitive overhead, the advantage it gives in terms of social interaction—and even social manipulation—seems to more than offset the cost.

Uttering Language If the theory proposed here is correct, at some stage the ability to communicate the two-argument A-relationship-B social model became advantageous. What particular set of circumstances would have allowed this to come about? We are unlikely to ever get a full answer to this question, but we can identify some key features. Humans must have been living together in large groups where social knowledge would be particularly valuable; and, because these groups probably exceeded 100 individuals each (Dunbar, 1993), the level of kinship was probably low. There would have been a high level of co-operation in hunting and in most other activities, and this would have been facilitated by a low level of interpersonal conflict and a high level of altruistic punishment. Fehr & Gächter (2002) show that humans have a natural aversion to freeloading and we punish it even at cost to ourselves; so altruistic punishment does seem to be part of human nature. Early humans are likely to have had a level of individual specialization, with individuals more adept at, say, tool-making being able to exchange their product for the produce of others, such as hunters. There would thus be a primitive exchange economy enabled by the altruistic punishment of economic freeloaders and renegades (Henrich & Boyd, 2008). Of course, it would not be exchange in a form we would immediately recognize, involving barter, compromise, agreement and finally delivery, it is more likely that it followed the principles of gifting

CO22552.indb 138

23/04/2010 14:13

The Evolution of Grammar

139

described by Mauss (1950): a network built up over time of obligations created and fulfilled. Each gift contains within it the suggestion of “your turn next”, which modern economics avoids. To see this system as primitive communism or primitive capitalism is to lose its true nature; it wasn’t a system of our time but one that worked for early humans. Erdal & Whiten (1994) offer one explanation of how this exchange culture could have come about. Their Vigilant Sharing model proposes that early humans had a culture of sharing and a sense of fairness about entitlement. These counterdominant tendencies acted to suppress greed and to provide a collective (although not necessarily co-operative) basis for altruistic punishment: not sharing fairly when everyone else is doing so takes advantage of, and therefore offends, everyone else. If the renegade is punished by every individual in the group then this amounts to punishment by the group. Boehm (1999) offers another explanation for the genesis of the exchange culture: collective action against controlling individuals that he refers to as Reverse Dominance. In this model, individuals work together to suppress alphas and their behaviours, creating a cultural environment where individual modesty and group achievement are valued. Once again, altruistic punishment keeps renegades in line, but this is done by co-operative group action rather than a series of individual actions. Vigilant Sharing and Reverse Dominance are not mutually exclusive, and it is likely that both played their part in early human culture. In terms of communication, the A-relationship-B model would, in this story, be the first grammatical form to be uttered. Initially, getting this message-form across would probably have been laborious, involving a mixture of gesture and vocalization; it would have required the receiver to be brought to the understanding that the sender was attempting to express the cognition model as a signal; and it would have required a high level of monitoring by both parties to ensure the message intended was the message received. It would have been neither easy nor efficient; but, once the trick had been learned, subsequent utterances would have been quicker and simpler—it would have been in the interests of both parties to work toward reducing the cost of the signal by emphasizing salience over detail. Grammatical language would have initially been difficult, but convergence of the interests of sender and receiver in ensuring faithful transmission would have created a powerful pressure toward simplicity. So the first two-argument A-relationship-B form to be uttered would have been produced in an environment of social equality, with individual specialization, exchange of goods, and within a large group. Social knowledge would certainly be valuable in this environment, but the large group size means that each individual must either spend a lot of time gathering the knowledge they need to fully map the social group, or they must settle for a partial map. If a person appeared in this environment who was able to honestly share their own social map with others, then they would enhance the social knowledge of others in the group and thus create better group consensus and cohesion.

CO22552.indb 139

23/04/2010 14:13

140

th e orig in s of g r am mar

Would this sharing enhance the fitness of the sharer, though? It would certainly make them popular, an ally worth cultivating; and the giving of knowledge would, in a gift-driven environment, create the need for reciprocity. Also, while sharing social knowledge is relatively cheap for the sharer it is quite valuable for the receiver; so the sharer should show a “profit”, creating valuable obligations for the receiver with little personal sender effort—in Darwinian terms, gaining fitness. Whether this sharing of two-argument forms was a genetic change or a cognitive innovation, it would have enhanced the fitness of the sharer; and it would, therefore, have spread through the group where it began, and then beyond, either by reproductive success or by behavioural adoption. Reputation keeps this social model-sharing honest (Fitch & Hauser, 2003); the sharing only works if what is shared is useful to the receiver. Giving out false information will lead, at least, to the discounting of that sender’s signals and a reduction in gifts exchanged for it; so any short-term advantage in lying will be offset by long-term losses. Basically, lying reduces fitness, being honest improves it.

What Language Did Next The two-argument form did more than bring grammar into signalling, it also introduced narrative description of noncurrent events. The A-relationship-B message is, essentially, “telling about” rather than instructing, and it augments the role of signalling in ways vital for full language. With telling-about, the relationship in the message is not immediately verifiable by the receiver, so it relies on a pre-existing level of trust between sender and receiver: the receiver accepts the signal not because it is inherently trustworthy but because the sender is trustworthy. There has to be a level of interpersonal interaction in the signalling process which, in turn, introduces Halliday’s systemic-functional metafunctions (Halliday & Matthiessen, 2008): the signal is no longer just a product of the text (the textual metafunction), it involves the sender and receiver as people (the interpersonal metafunction), and the ideas surrounding and behind the production of the utterance (the ideational metafunction). Telling-about turns a signalling environment into a communication system. The introduction of narrative into signalling also changes the modelling environment. As well as having cognitive mechanisms to handle two-argument forms, it now becomes advantageous to be able to model three-argument forms. This allows A-relationship-B models received from others to be tagged with the identity of the sender, making hierarchical [A-relationship-B]-by-C models. There is clearly an extra cognitive overhead involved here, and it is unlikely that this capacity emerged as soon as A-relationship-B messages began to be uttered; but being able to allocate utterances to their author allowed the relative trustworthiness of the author to be factored into the received two-argument forms. It is at this point that the receiver’s social modelling becomes iterative, with the modelling of the individuals in the message (A and B) being contained

CO22552.indb 140

23/04/2010 14:13

The Evolution of Grammar

141

within a model of the sender (C). It is not, however, recursive at this stage: while the modelled sender contains models of others, these models of others are not themselves model-makers. Once, however, the [A-relationship-B]-by-C model becomes utterable as the three-argument form, language begins to escalate. Once again it is to the receiver’s advantage to be able to tag the received model with the identity of the sender, creating nested hierarchies of [[A-relationship-B]-by-C]-by-D. These contain true modelling recursion: the receiver’s model contains D modelling C modelling A and B, so there is an iteration of a model-maker (D) modelling a model-maker (C) within the model. This recursive iteration is theoretically infinite but in practice quite limited, with possibly only three or four levels of recursion achievable by the average human. This, however, still represents a capacity not demonstrated by any other primate (Dunbar, 2004, ch. 3). The telling-about of the A-relationship-B model is also likely to have initiated various new grammatical tools. The ability to differentiate between people known to both sender and receiver can be done with simple tagging, or naming; but the establishment of those tags would have required (and still does require) careful negotiation towards common meaning. The individuals in the models being exchanged are often not available for direct reference, such as pointing; and, even where they are, discretion may dictate a more indirect form of reference. Telling-about therefore benefits from ways of describing both the individuals being talked about and their relationships, introducing grammatical roles such as adjectives, adverbs, determiners and negators. These descriptors, which probably developed slowly over many generations, vastly expanded the range of narratives possible. They are not part of the story being told here, but they are a significant outcome of exchanging A-relationship-B models.

Becoming Myself As has already been discussed in chapter 7, a significant outcome of exchanging A-relationship-B models is the realization that some of the models being exchanged involve me and my relationships with others. There is the potential for me to see in the modelling by others of others that sometimes they are modelling me as another. Once I am aware that others are modelling me I can use my own cognitive capacity for modelling others to model myself. While the capacity to model myself may be innate (inasmuch as the capacity to model others is innate), it is not necessarily an ability that comes fully-formed at birth. Chapter 11 showed that a young child has only basic models of others and themself, and it takes years before the child can fully take part in adult discourse. In chapter 12 we saw that self-modelling is a vital capacity in our understanding and communication of temporality. Our use of tense in language signalling allows us to place events onto a time line, and out into the what-if realms of conditionality. In terms of intentionality, Reichenbach’s three points

CO22552.indb 141

23/04/2010 14:13

142

th e orig in s of g r am mar

can be summarized as I (now) believe that I (in the future or past or conditional universe) can conceive of the event (in the future or past or conditional universe). The capacity to model myself is an emergent feature of my ability to model others and of others sharing their models of others; and the capacity to model myself modelling is an emergent feature of my ability to model myself and of seeing others as intentional beings. It is the capacity to model myself modelling which is at the heart of temporality in language. This iterative modelling is also behind the roles of person or voice in language: the special relationship between the sender (the first person, me) and the receiver (the second person, you) in talking about others (the third persons, them) is explicitly recognized in language, indicating that there is meta-awareness of the signalling process itself. Language utterances are tailored by the sender to match what they believe are the receiver’s needs: we use different registers to talk to different people in different circumstances, a capacity that has not been identified in other animal signalling. In fact, most human communication is dictated by the sender’s model of the likely receivers of the signal; and impairment in this ability to make models of the needs of others seems to be implicated in the condition of autism (Frith, 2003, ch. 5). It seems that the capacity to model self and others is not just a product of other ways of being human, it is closely implicated in the definition of humanity itself. One strange, apparently anti-Darwinian, outcome of our capacity to model others as equal to ourself is that the self is often suppressed as the logical target of all intention. The unmodelled self is no longer the centre of the universe around which everything else revolves, the universe has its own existence separate to the self and does not rely on the self. Suppression of self occurs elsewhere in nature only among eusocial animals, where coercive control of individual reproduction means that the only way an individual can get its genes into the future is by supporting its fertile relatives. In eusocial insects, the queen produces all the offspring for the nest, whether fertile breeders or infertile workers (Bourke & Franks, 1995). It is fitter in this environment for the individual worker to support the reproductive effort of the queen than to pursue a pointless personal agenda. Humans, however, are not fully eusocial, we each retain our individual ability to reproduce. This should impose on us the standard imperative of nature: everyone else is either a potential mate or a potential rival, and should be treated as such. It seems, though, that we had already evolved a level of social co-operation before self-modelling appeared: our offspring are feeble and require extensive nurturing; our capacities as individuals are limited, but specialism and co-ordinated effort have a multiplier effect; we are able to share skills by teaching and learning; and we are able to remember and react to the reputations of others, judging their present actions by their past reliability. It is likely that all these markers of socialization would have been present in early Homo, creating an environment where the self is better served by co-operation than by selfishly

CO22552.indb 142

23/04/2010 14:13

The Evolution of Grammar

143

pursuing personal ends. For co-operation to work it has only to be a slightly better strategy than individualism; over time, small differences in fitness become powerful trends for the species. Into this environment the first social model signaller appeared and began their own species trend, a process which led eventually to language. This trend relies on the capacity for an increasingly complex signalling grammar; but that grammar is not itself something novel, it is based on pre-existing cognitive mechanisms which were needed to navigate the social environment in which the individuals found themselves. In this respect it has some of the features of the likely origins of eusociality: co-operation leads to enhanced fitness, but also opens the way for higher levels of co-operation. Eventually, a social structure appears that is so complex and so co-operative that it is difficult to explain using only the reference points of Darwinian selfishness and the existent sociality. That difficulty is, however, a problem of our explanatory capacity, not of evolutionary mechanics.

Are There Grammar Universals? Generativist linguists believe that grammar has a single set of universal principles which are innate in all humans. All languages therefore rely on, and contain, mechanisms that are generated by a language organ which provides a unified structure to human language. This language organ is a product of a specific, language-related genetic mutation which has occurred only in the human lineage (Chomsky, 2002, p. 64). There certainly seem to be universal features underlying human languages, but are they the product of a single engine dedicated to language, or are they emergent from other cognitive systems devised only indirectly for communication? The argument set out here indicates that language is more of a Heath-Robinson affair than a well-tuned dedicated engine. The first grammatical utterances had to communicate A-relationship-B models; but the act of generating them produced unexpected outcomes which, when communicated, created more outcomes . . . and so on. Yet the nature of the messages to be communicated does impose constraints on the form of the messages. The A-relationship-B model requires a three-component message; it requires recognition of objects and actions as two different classes; it requires tagging of individuals with labels; and it requires negotiation towards a common meaning in terms of what the components of the message refer to. Each grammatical innovation relies on cognitive processes that evolved for other, non-communicative purposes, and which therefore have their own genetic bases. Thus, while the language organ remains improbable, universals can be imposed on language by the cognitive processes that language uses. We would expect to see in every human language the one-, two- and three-argument forms; the objectaction distinction; similar expression of temporality; iteration and recursion; Halliday’s metafunctions; and the capacity to indicate non-present items. Indeed,

CO22552.indb 143

23/04/2010 14:13

144

th e orig in s of g r am mar

the differences between languages should, at this level of comparison, seem quite trivial. Based on the theory of grammar origins proposed here, we can also suggest a solution to the ongoing debate about language complexity: did language start simple and become complex (Burling, 2005, ch. 9), or did it start complex and become simple (Wray, 2002a)? The answer seems to be both: in terms of communicable grammatical forms it started simple with only the A-relationship-B form, and out of this a range of more complex grammatical structures developed. Getting the message across, however, would initially have been very laborious, with each meaning having to be negotiated; there were no pre-existing givens shared by sender and receiver. As language spread through the community, and each utterance joined the already-existing discourse, these givens would have begun to appear as consensuses established in meaning and form. So instead of a directional change from complexity to simplicity, we should see instead an exchange of complexity in utterance for complexity in form: as the simple forms became easier to share they created room for more complex forms to be shared. Language started complex and remains complex, but in different ways.

And Finally . . . This book has hopefully given an overview of the development of language grammar out of the cognitive mechanisms of social modelling. In a highly socialized species such as ours, the capacity to map social interactions is valuable, so it is likely to be selected for in terms of evolution. Being able to plug into the social maps of others would therefore be equally valuable and, if a way to do it could be found, it would probably become fixed genetically in the species. The sharing of social maps is not the whole story of grammar; but it is a foot in the door for other social communication, and it is the richness of this extended social signalling that generates the need for complex grammar. At base, though, grammar is a device for sharing social models: a product of sociality, a measure of sociality, and a social exchange. This book has left some subjects unaddressed. For instance, how did the sharing of social models become a general sharing of subject-verb-object forms that contain no direct social significance (such as the couch had a blue cover)? The problem of honesty has also been somewhat cursorily reviewed, with an appeal to the necessity for honesty as evidence for the honesty of the first-ever speaker. The honesty problem certainly does need to be examined further, but the wider issues of morality it raises are too large to tackle here. There are several other aspects of grammar origins that need to be addressed in greater detail than has been given here; but they are also tasks for another day. The problem of the origins of grammar should not be underestimated: it is possible that we will never have a satisfactory answer to the question of where language grammar came from, and that the Linguistic Society of Paris was

CO22552.indb 144

23/04/2010 14:13

The Evolution of Grammar

145

correct in 1866 to ban speculation on language origins (Aitchison, 1996, p. 5). Yet speculation is at the heart of scientific endeavour—and an important part of being human, too. While this book has been a speculation on the origins of grammar, it has tried to stay within the scientific usage of the term “speculation”: the proposal of a case which is testable against evidence. This book is unlikely to have provided the definitive answer to the origins of language grammar, but it has hopefully told a consistent and effective story which explains more than it obscures. Whether or not the theory proposed here contributes to an explanation of grammar origins, the search for the genesis of language must continue to be pursued with vigour: how we came to rely on grammatical language is crucial to our understanding of how we became human.

CO22552.indb 145

23/04/2010 14:13

This page intentionally left blank

CO22552.indb 146

23/04/2010 14:13

References

Aiello, L. & Dean, C. (1990). An Introduction to Human Evolutionary Anatomy. London, UK: Academic Press Ltd. Aiello, L. & Wheeler, P. (1995). “The Expensive Tissue Hypothesis: the brain and the digestive system in human and primate evolution”. In Current Anthropology vol. 36 no. 2 April, 1995. Aitchison, J. (1996). The Seeds of Speech: language origin and evolution. Cambridge, UK: Cambridge University Press. Aitchison, J. (1998). The Articulate Mammal: an introduction to psycholinguistics (4th edition). London, UK: Routledge. Allen, C. & Saidel, E. (1998). “The evolution of reference”. In D. D. Cummins and C. Allen (eds), The Evolution of Mind. Oxford, UK: Oxford University Press. Allen, J. & Seidenberg, M. S. (1999). “The Emergence of Grammaticality in Connectionist networks”. In B. MacWhinney (ed), The Emergence of Language. Mahwah, NJ, USA: Lawrence Erlbaum Associates. Stanley H. Ambrose (2001). “Paleolithic Technology and Human Evolution”. Science 291, 2 March 2001: pp. 1748–1753. Arbib, M. A. (2005). “From monkey-like action recognition to human language: An evolutionary framework for neurolinguistics”. Behavioral and Brain Sciences 28, pp. 105–167. Areces, C. & Blackburn, P. (2005). “Reichenbach, Prior and Montague: a semantic get-together”. In Articles in Honor of Dov Gabbay’s 60th Birthday. London, UK: Kings College Press. Aristotle (2004 [350 BCE]). Nichomachean Ethics. London, UK: Penguin. Baker, M. C. (2001). The Atoms of Language: The mind’s hidden rules of grammar. New York, USA: Basicbooks. Baker, R. R. & Bellis, M. A. (1995). Human sperm competition: Copulation, masturbation, and infidelity. Chapman & Hall: London, UK. Balme, J. & Bowdler, S. (2006). “Spear and Digging Stick: The origin of gender and its implications for the colonisation of new continents”. Journal of Social Archaeology 6: pp. 379–401. Balter, M. (2009). “On the Origin of Art and Symbolism”. Science 323: pp. 709–711.



CO22552.indb 147

147

23/04/2010 14:13

148

References

Barclay, P. & Willer, R. (2007). “Partner choice creates competitive altruism in humans”. Proceedings of the Royal Society B 274: pp. 749–753. Barnes, D. M. (2005). “Possible Tool Use by Beavers, Castor canadensis, in a Northern Ontario Watershed”. The Canadian Field-Naturalist 119–3: pp. 441–443. Baron-Cohen, S. (1995). Mindblindness: an essay on autism and Theory of Mind. Cambridge, USA: MIT Press. Bates, E., Dale, P. S. & Thal, D. (1995). “Individual Differences and their Implications for Theories of Language Development”. In P. Fletcher & B. MacWhinney (eds), The Handbook of Child Language. Oxford, UK: Blackwell Publishers. Bateson, G. (1985). “A Theory of Play and Fantasy”. In R. E. Innis (ed), Semiotics: an introductory anthology. Bloomington, USA: Indiana University Press. Baumeister, R. F., Campbell, J. D., Krueger, J. I. & Vohs, K. D. (2005). “Exploding the Self-esteem Myth”. Scientific American Mind 16–4: pp. 50-57. Beckerman, S. & Valentine, P. (2002). “Introduction: The concept of partible paternity among Native South Americans”. In S. Beckerman & P. Valentine (eds), Cultures of Multiple Fathers: The theory and practice of partible paternity in Lowland South America. Gainesville, FL, USA: University Press of Florida. Benson, J. D. & Greaves, W. S. (eds) (2005). Functional Dimensions of Ape-Human Discourse. London, UK: Equinox Publishing Ltd. Benveniste, É. (1970). “The Nature of Pronouns”. In Problems in General Linguistics. University of Miami Press: Miami, USA. Berg, L. (1972). Look at Kids. London, UK: Penguin Books. Bergstrom, C. T. & Lachmann, M. (1997). “Signalling among relatives I. Is costly signalling too costly?” Philosophical Transactions of the Royal Society of London B, 352: pp. 609–617. Bergstrom, T. C. (1994). “On the Economics of Polygyny”. eScholarship Repository, University of California. (http://repositories.cdlib.org/ucsbecon/ bergstrom/1994A) Berkeley, G. (1910 [1710]). A Treatise Concerning the Principles of Human Knowledge. (Facsimile reprint). Chicago, USA: Open Court Publishing. (Available via Scribd.) Bickerton, D. (1990). Language and Species. Chicago, Ill., USA: University of Chicago Press. Bickerton, D. (1998). “Catastrophic Evolution: the case for a single step from protolanguage to full human language”. In J. R. Hurford, M. StuddertKennedy & C. Knight (eds), Approaches to the Evolution of Language: social and cognitive bases. Cambridge, UK: Cambridge University Press. Bickerton, D. (2000). “How Protolanguage Became Language”. C. Knight, M. Studdert-Kennedy & J. R. Hurford (eds), The Evolutionary Emergence of

CO22552.indb 148

23/04/2010 14:13

References

149

Language: social function and the origins of linguistic form. Cambridge, UK: Cambridge University Press. Bickerton, D. (2002). “Foraging Versus Social Intelligence in the Evolution of Protolanguage”. In A. Wray (ed), The Transition to Language. Oxford, UK: Oxford University Press. Bloom, P. (2002). How Children Learn the Meanings of Words. Cambridge, USA: MIT Press. Bloomfield, L. (1933). Language. New York, USA: Holt, Rinehart & Winston. Bloor, T. & Bloor, M. (1995). The Functional Analysis of English: a Hallidayan approach. London, UK: Arnold. Boas, F. (1938). “Introduction”. In F. Boas (ed), General Anthropology. Boston, Mass, USA: D C Heath and Company. Bod, R. (1998). Beyond Grammar: an experience-based theory of language. Stanford, Cal, USA: CSLI Publications. Boehm, C. (1999). Hierarchy in the Forest: the evolution of egalitarian behaviour. Cambridge, Mass, USA: Harvard University Press. Boesch-Achermann, H. & Boesch, C. (1993). “Tool Use in Wild Chimpanzees: New Light from Dark Forests”. Current Directions in Psychological Science, 2–1: pp. 18–21. Bourke, A. F. G. & Franks, N. R. (1995). Social Evolution in Ants. Princeton, NJ, USA: Princeton University Press. Boyd, R., Gintis, H., Bowles, S. & Richerson, P. J. (2003). “The evolution of altruistic punishment”. Proceedings of the National Academy of Science, 100–6: pp. 3531–3535. Bradbury, J. W. & Vehrencamp, S. L. (1998). Principles of Animal Communication. Sunderland, USA: Sinauer Associates inc. Brewer, S. M. & McGrew, W. C. (1990). “Chimpanzee use of a tool set to get honey”. Folia Primatologica 54: pp. 100–104. Brewer, S. (2001). A Child’s World: a unique insight into how children think. London, UK: Headline Book Publishing. Brighton, H., Kirby, S. & Smith, K. (2005). “Cultural Selection for Learnability: three principles underlying the view that language adapts to be learnable”. In M. Tallerman (ed), Language Origins: perspectives on evolution. Oxford, UK: Oxford University Press. Broom, D. M., Senaa, H. & Moynihan, K. L. (2009). “Pigs learn what a mirror image represents and use it to obtain information”. Animal Behaviour, 78–5: pp. 1037–1041. Brown, R. (1970). “Three Processes in the Child’s Acquisition of Syntax”. Psycholinguistics: selected papers by Roger Brown. New York, USA: Macmillan Press. Bruner, J. (1986). Actual Minds, Possible Worlds. Cambridge, Mass., USA: Harvard University Press.

CO22552.indb 149

23/04/2010 14:13

150

References

Budiansky, S. (1998). If a Lion Could Talk: how animals think. London, UK: Weidenfeld & Nicolson. Burland, T. M., Bennett, N. C., Jarvis, J. U. M., & Faulkes, C. G. (2002). “Eusociality in African mole-rats: new insights from patterns of genetic related­ness in the Damaraland mole-rat (Cryptomys damarensis)”. Proceedings of the Royal Society of London B, 269–1495: pp. 1025–1030. Burling, R. (2005). The Talking Ape: how language evolved. Oxford, UK: Oxford University Press. Buzing, P. C., Eiben, A. E. & Schut, M. C. (2005). “Emerging Communication and Cooperation in Evolving Agent Societies”. Journal of Artificial Societies and Social Simulation, 8–1. Byrne, R. (1995). The Thinking Ape: Evolutionary Origins of Intelligence. Oxford, UK: Oxford University Press. Byrne, R. (2000). “Evolution of Primate Cognition”. Cognitive Science 24–3: pp. 543–570. Byrne, R. (2004). “The manual skills and cognition that lie behind hominid tool use”. In A. E. Russon & D. R. Begun (eds), The Evolution of Thought: Evolutionary Origins of Great Ape Intelligence. Cambridge, UK: Cambridge University Press. Caillois, R. (1958). Man, Play and Games. Chicago, Ill., USA: University of Illinois Press. Calvin, W. H. & Bickerton, D. (2000). Deus ex Machina. Cambridge, Mass, USA: MIT Press. Cameron, D. (1998a). “Performing Gender Identity: young men’s talk and the construction of heterosexual masculinity”. In J. Coates (ed), Language and Gender: a reader. Oxford, UK: Blackwell Publishers Ltd. Cameron, D. (1998b). “What Has Gender Got To Do With Sex?” In R. Harris & G. Wolf (eds) Integrational Linguistics: a first reader. Oxford, UK: Elsevier Science Ltd. Campbell, A. (1999). “Staying alive: Evolution, culture, and women’s intrasexual aggression”. Behavioral and Brain Sciences (1999) 22–02: pp. 203–252. Carruthers, P. (2009). “How we know our own minds: The relationship between mindreading and metacognition”. Behavioral and Brain Sciences, 32–02: pp. 121–138. Carstairs McCarthy, A. (1999). The Origins of Complex Language: an inquiry into the evolutionary beginnings of sentences, syllables and truth. Oxford, UK: Oxford University Press. Carter, R. (1998). Mapping the Mind. London, UK: Weidenfeld & Nicolson. Chafe, W. (1970). Meaning and the Structure of Language. Chicago, Ill., USA: University of Chicago Press. Cheney, D. L. & Seyfarth, R. M. (1990). How Monkeys See the World: inside the mind of another species. Chicago, Ill., USA: University of Chicago Press. Chiappe, D., Brown, A., Dow, B., Koontz, J., Rodriguez, M. & McCulloch, K.

CO22552.indb 150

23/04/2010 14:13

References

151

(2004). “Cheaters Are Looked At Longer and Remembered Better Than Cooperators in Social Exchange Situations”. Evolutionary Psychology, 2: pp. 108–120. Chiat, S. (2000). Understanding Children with Language Problems. Cambridge, UK: Cambridge University Press. Chomsky, N. (1957). Syntactic Structures. Berlin, Germany: Mouton de Gruyter. Chomsky, N. (1965). Aspects of the Theory of Syntax. Cambridge, Mass., USA: MIT Press. Chomsky, N. (1977). Language and Responsibility. In On Language: Chomsky’s classic works Language and Responsibility and Reflections on Language. New York, USA: The New Press. Chomsky, N. (1980). Rules and Representations. New York, USA: Columbia University Press. Chomsky, N. (1982). Some Concepts and Consequences of the Theory of Government and Binding. Cambridge, Mass., USA: MIT Press. Chomsky, N. (1986). Knowledge of Language—its nature, origin, and use. Westport, CT., USA: Preager Publishers. Chomsky, N. (1988). Language and Problems of Knowledge: the Managua lectures. Cambridge, Mass., USA: MIT Press. Chomsky, N. (1995). The Minimalist Program. Cambridge, Mass, USA: MIT Press. Chomsky, N. (2000). New Horizons in the Study of Language and Mind. Cambridge, UK: Cambridge University Press. Chomsky, N. (2002). On Nature and Language. Cambridge, UK: Cambridge University Press. Chomsky, N. (2006). Language and Mind (third edition). Cambridge, UK: Cambridge University Press. Clark, E. V. (2003). First Language Acquisition. Cambridge, UK: Cambridge University Press. Clutton-Brock, T. H. & Isvaran, K. (2006). “Paternity loss in contrasting mammalian societies”. Biology Letters 2: pp. 513–516. Conard, N. J. (2003). “Palaeolithic ivory sculptures from southwestern Germany and the origins of figurative art”. Nature, 426–6968: pp. 830–832. Conklin-Brittain, N. L., Wrangham, R. W. & Smith, C. C. (2007). “Relating Chimpanzee Diets to Potential Australopithecus Diets”. Paper at 14th International Congress of Anthropological and Ethnological Sciences, Williamsburg, Virginia. web.cast.uark.edu/local/icaes/conferences/wburg/ posters/nconklin/conklin.html (ref. June 2009) Cook, G. (2000). Language Play, Language Learning. Oxford, UK: Oxford University Press. Cook, V. J. and Newson, M. (1996). Chomsky’s Universal Grammar: an introduction (second edition). Oxford, UK: Blackwell Publishers Ltd. Corballis, M. C. (2003). From Hand to Mouth: the origins of language. Princeton, NJ, USA: Princeton University Press.

CO22552.indb 151

23/04/2010 14:13

152

References

Corning, P. A. (2000). “Biological Adaptation in Human Societies: a “basic needs” approach”. Journal of Bioeconomics 2: pp. 41–86. Cosmides, L. & Tooby, J. (1994). “Beyond intuition and instinct blindness: towards an evolutionary rigorous cognitive science”. Cognition, 50: pp. 41–77. Craig, H. K. (1995). “Pragmatic Impairments”. In P. Fletcher & B. MacWhinney (eds), The Handbook of Child Language. Oxford, UK: Blackwell Publishers. Creider, C. (2000). “Mixed Categories in Word Grammar: Swahili infinitival Nouns”. Linguistica Atlantica 21: pp. 53–68. Croft, W. (2001). Radical Construction Grammar: syntactic theory in typological perspective. Oxford, UK: Oxford University Press. Cronin, H. (2005). “Adaptation: ‘a Critique of some Current Evolutionary Thought’ ”. The Quarterly Review of Biology 80–1: pp. 19–26. Dąbrowska, E. (2004). Language, Mind and Brain: some psychological and neurological constraints on theories of grammar. Edinburgh, Scotland: Edinburgh University Press. Charles Darwin (1859). On the Origin of Species. London, UK: John Murray. Davidson, I. & McGrew, W. C. (2005). “Stone Tools and the Uniqueness of Human Culture”. Journal of the Royal Anthropological Institute, (NS) 11: pp. 793–817. Dawkins, R. (1989). The Selfish Gene. Oxford, UK: Oxford University Press. Dawkins, R. (1996). Climbing Mount Improbable. London, UK: Penguin Books. Day, M. H. (1986). “Bipedalism: pressures, origins, and modes”. In B. Wood, L. Martin & P. Andrews (eds), Major Topics in Primate and Human Evolution. New York, USA: Cambridge University Press. Deacon, T. (1997). The Symbolic Species: the co-evolution of language and the human brain. London, UK: Penguin Books. Denton, D. K. (2006). “Keep it Simple: business decisions don’t have to be complicated to be good”. Industrial Engineer, May 2006: pp. 30–35. Descartes, R. (1998 [1641]). “Meditations on First Philosophy: replies to second objections”. In D. M. Clarke (tr), René Descartes, Meditations and Other Metaphysical Writings. London, UK: Penguin Books. Dessalles, J. L. (2000). “Language and Hominid Politics”. In C. Knight, M. Studdert-Kennedy & J. R. Hurford (eds), The Evolutionary Emergence of Language: social function and the origins of linguistic form. Cambridge, UK: Cambridge University Press. Dessalles, J. L. (2007). Why We Talk: the evolutionary origins of language. Oxford, UK: Oxford University Press. Deutscher, G. (2005). The Unfolding of Language: the evolution of mankind’s greatest invention. London, UK: Arrow Books. Dickins, T. E. (2003). “General Symbol Machines: the first stage in the evolution of symbolic communication”. Evolutionary Psychology 1, pp. 192–209. Dixon, R. M. W. (1997). The Rise and Fall of Languages. Cambridge, UK: Cambridge University Press.

CO22552.indb 152

23/04/2010 14:13

References

153

Donald, M. (2001). A Mind So Rare: the evolution of human consciousness. London, UK: W W Norton & co. Dunbar, R. I. M. (1992). “Why gossip is good for you”. New Scientist 136–1848: p. 28. Dunbar, R. I. M. (1993). “Co-Evolution of Neocortex Size, Group Size and Language in Humans”. Behavioral and Brain Sciences 16–4: pp. 681–735. Dunbar, R. I. M. (1996). Grooming, Gossip and the Evolution of Language. London, UK: Faber & Faber Ltd. Dunbar, R. I. M. (2004). The Human Story: a new history of mankind’s evolution. London, UK: Faber & Faber Ltd. Dunning, D., Heath, C. & Suls, J. M. (2004). “Flawed Self-Assessment: Implications for Health, Education, and the Workplace”. Psychological Science in the Public Interest 5–3: pp. 69–106. Eckert, P. & McConnell-Ginet, S. (1998). “Communities of Practice: where language, gender, and power all live”. In Jennifer Coates (ed), Language and Gender: a reader. Oxford, UK: Blackwells Publishing. Edwardes, M. (2003). “I like Both Myself and Me”. Camling 2003: proceedings of the University of Cambridge first postgraduate conference in language research. Cambridge, UK: CILR. Egas, M. & Riedl, R. (2005). “The Economics of Altruistic Punishment and the Demise of Cooperation”. Forschungsinstitut zur Zukunft der Arbeit, discussion paper 1646, July 2005. Epley, N. & Whitchurch, E. (2008). “Mirror, Mirror on the Wall: Enhancement in Self-Recognition”. Personality and Social Psychology Bulletin, 34–9: pp. 1159–1170. Erdal, D. & Whiten, A. (1994). “On Human Egalitarianism: an evolutionary product of Machiavellian status escalation?” Current Anthropology, 35–2: pp. 175–183. Evans, V. (2005). The Structure of Time. Amsterdam, Netherlands: John Benjamins. Evans, V. & Green, M. (2006). Cognitive Linguistics: an introduction. Edinburgh, Scotland: Edinburgh University Press. Evans, V. (2007). “How We Conceptualise Time: language, meaning and temporal cognition”. In V. Evans, B. K. Bergen & J. Zinken (eds), The Cognitive Linguistics Reader. London, UK: Equinox Publishing. Evans, V., Bergen, B. K. & Zinken, J. (2007). “The Cognitive Linguistics Enterprise: An Overview”. In V. Evans, B. K. Bergen & J. Zinken (eds). The Cognitive Linguistics Reader. London, UK: Equinox Publishing. Eysenck, M. W. (1993). Principles of Cognitive Psychology. London, UK: Psychology Press. Eysenck, M. W. & Keane, M. T. (1995). Cognitive Psychology: a student’s handbook, 3rd Edition. London, UK: Psychology Press. Fauconnier, G. & Turner, M. (2008). “Rethinking Metaphor”. In R. Gibbs (ed), Cambridge Handbook of Metaphor and Thought. Cambridge, UK: Cambridge University Press.

CO22552.indb 153

23/04/2010 14:13

154

References

Fehr, E. & Gächter, S. (2002). “Altruistic punishment in humans”. Nature 415–6868: pp. 137–140. Fehr, E. & Renninger, S. V. (2004). “The Samaritan Paradox”. Scientific American Mind, November 2004: pp. 15–21. Fehr, E. & Fischbacher, U. (2005). “Human Altruism—Proximate Patterns and Evolutionary Origins”. Analyse & Kritik 27: pp. 6–47. Fillmore, C. (1971). Types of Lexical Information. Semantics: an interdisciplinary reader in Philosophy, Linguistics and Psychology. Cambridge University Press: Cambridge, UK. Fillmore, C. (1976). Frame semantics and the nature of language. Annals of the New York Academy of Sciences: Conference on the Origin and Development of Language and Speech, 280: pp. 20–32. Fitch, W. T. & Hauser, M. D. (2003). “Unpacking ‘Honesty’: Vertebrate Vocal Production and the Evolution of Acoustic Signals”. In A. Simmons, R. R. Fay, & A. N. Popper (eds), Acoustic Communication. New York, USA: Springer. Flack, J. C., Jeannotte, L. A. & de Waal, F. B. M. (2004). “Play Signaling and the Perception of Social Rules by Juvenile Chimpanzees (Pan Troglodytes)”. Journal of Comparative Psychology, 118–2: pp. 149–159. Flack, J. C., Girvan, M., de Waal, F. B. M. & Krakauer, D. C. (2006). “Policing stabilizes construction of social niches in primates”. Nature 439–7075: pp. 426–429. Fodor, J. (1983). The Modularity of Mind. Cambridge, Mass, USA: MIT Press. Foley, J. & Thompson, L. (2003). Language Learning: a lifelong process. London, UK: Arnold. Fossey, D. (1983). Gorillas in the Mist. London, UK: Phoenix. Fouts, R. with Mills, S. T. (1997). Next of Kin: my conversations with chimpanzees. New York, USA: Avon Books Inc. Fowler, J. H. (2005). “Altruistic Punishment and the Origin of Cooperation”. Proceedings of the National Academy of Science, 102–19: pp. 7047–7049. Frey, S. H. (2008). “Tool use, communicative gesture and cerebral asymmetries in the modern human brain”. Philosophical Transactions of the Royal Society B 363: pp. 1951–1957. von Frisch, K. (1973). “Decoding the Language of the Bee”. Nobel Lecture, December 12, 1973. Frith, U. (1999 [1993]). Autism. The Scientific American Book of the Brain. Guilford, USA: The Lyons Press. Frith, U. (2003). Autism: explaining the enigma (2nd edition). Oxford, UK: Blackwell Publishing. Gallup, G. G., (1970). “Chimpanzees: self-recognition”. Science, 167–3194: pp. 86–87. Gardner, A. & West, S. A. (2004). “Cooperation and Punishment, Especially in Humans”. The American Naturalist 164–6: pp. 753–764.

CO22552.indb 154

23/04/2010 14:13

References

155

Gardner, H. (1983). Frames of Mind: the theory of Multiple Intelligences. New York, USA: Basic Books. Gavrilets, S. & Vose, A. (2006). “The dynamics of Machiavellian intelligence”. Proceedings of the National Academy of Science, 103–45: pp. 16823–16828. Gaylard, H. L. (1995). “Phrase Structure in a Computational Model of Child Language Acquisition”. PhD thesis, School of Computer Science, University of Birmingham. Geeraerts, D. (2006). “Introduction: a rough guide to Cognitive Linguistics”. D. Geeraerts (ed), Cognitive Linguistics: basic readings. Berlin, Germany: Mouton de Gruyter. Gibbs, R. W. & Perlman, M. (2006). “The Contested Impact of Cognitive Linguistic Research on the Psycholinguistics of Metaphor Understanding”. In G. Kristiansen, M. Achard, R. Dirven & F. J. Ruiz de Mendoza Ibáñez (eds), Cognitive Linguistics: current applications and future perspectives. Berlin, Germany: Mouton de Gruyter. Gibson, K. R. (2002). “Evolution of Human Intelligence: The Roles of Brain Size and Mental Construction”. Brain, Behavior and Evolution 59: pp. 10–20. Gintis, H. (2008). “Punishment and Cooperation”. Science 319–5868: pp. 1345–1346. Goldstein, M. C., Jiao, B., Beall, C. M. & Tsering, P. (2002). “Fertility and Family Planning in Rural Tibet”. The China Journal 47: pp. 19–39. Goleman, D. (1995). Emotional Intelligence: why it can matter more than IQ. London, UK: Bloomsbury. Goodall, J. (1988). In the Shadow of Man. London, UK: Weidenfeld & Nicolson. Goodall, J. (1990). Through a Window: thirty years with the chimpanzees of Gombe. London, UK: Phoenix. Gopnik, A., Meltzoff, A. & Kuhl, P. (1999). How Babies Think. London, UK: Weidenfeld & Nicolson. Gould, S. J. & Lewontin, R. C. (1997 [1979]). “The Spandrels of San Marco and the Panglossian Paradigm: a critique of the adaptationist programme”. In M. Ridley (ed), Evolution. Oxford, UK: Oxford University Press. Gould, S. J. (1981). The Mismeasure of Man. London, UK: Penguin. Granberg, Å., Armelius, K. & Armelius, B-A. (2001). “Self-Image Pattern and Treatment Outcome in Severely Disturbed Psychiatric Patients” Umeå Psychology Reports 1: pp. 1–13. Gravina, B., Mellars, P. & Ramsey, C. B. (2005). “Radiocarbon dating of interstratified Neanderthal and early modern human occupations at the Chatelperronian type-site”. Nature 438–7064: pp. 51–56. Greenfield, S. (2000). Brain Story: unlocking our inner world of emotions, memories, ideas and desires. London, UK: BBC Worldwide. Grice, P. (1989). Studies in the Way of Words. Harvard, Mass., USA: Harvard University Press.

CO22552.indb 155

23/04/2010 14:13

156

References

Halliday, M. A. K. (2003 [1969]). “A Brief Sketch of Systemic Grammar”. In J. J. Webster (ed), On Language and Linguistics. London, UK: Continuum. Halliday, M. A. K. (1985). An Introduction to Functional Grammar. London, UK: Arnold. Halliday, M. A. K. (1994). An Introduction to Functional Grammar, 2nd edition. London, UK: Edward Arnold. Halliday, M. A. K. (2003). The Language of Early Childhood. London, UK: Continuum. Halliday, M. A. K. & Matthiessen, C. M. I. M. (2004). An Introduction to Functional Grammar, third edition. London, UK: Arnold. Halliday, M. A. K. (2004). “Three Aspects of Children’s Language Development: learning language, learning through language, learning about language”. The Language of Early Childhood. London, UK: Continuum. Hamilton, W. D. (1964). “The Genetical Evolution of Social Behaviour II”. The Journal of Theoretical Biology, 7: pp. 17–52. Hanks, W. F. (2005). “Explorations in the Deictic Field”. Current Anthropology 46–2: pp. 191–220. Harder, P. (1996). Functional Semantics: A Theory of Meaning, Structure and Tense in English. Berlin, Germany: Mouton de Gruyter. Hare, B. & Tomasello, M. (2004). “Chimpanzees are more skilful in competitive than in cooperative cognitive tasks”. Animal Behaviour, 68: pp. 571–581. Harman, G. H. (1971 [1968]). “Three Levels of Meaning”. In D. D. Steinberg & L. A. Jakobovits (eds), Semantics: an interdisciplinary reader in philosophy, linguistics and psychology. Cambridge, UK: Cambridge University Press. Harris, R. A. (1993). The Linguistics Wars. Oxford UK: Oxford University Press. Hauser, M. D., Chomsky, N. & Fitch, W. T. (2002). “The Faculty of Language: what is it, who has it, and how did it evolve?” Science 298–5598: pp. 1569–1579. Hawking, S. (1996). The Illustrated a Brief History of Time. London, UK: Bantam Press. Healy, J. M. (1994). Your Child’s Growing Mind: a practical guide to brain development and learning from birth to adolescence. New York, USA: Doubleday. Heine, B. & Kuteva, T. (2002). The Genesis of Grammar: a reconstruction. Oxford, UK: Oxford University Press. Henderson, I. & Robertson, P. (2007). “Control and Eradication of the North American Ruddy Duck in Europe”. In G. W. Witmer, W. C. Pitt & K. A. Fagerstone (eds), Managing Vertebrate Invasive Species: Proceedings of an International Symposium. Fort Collins, CO, USA: USDA/APHIS/WS, National Wildlife Research Center. Hengeveld, K. & Mackenzie, J. L. (2008). Functional Discourse Grammar: A Typologically-Based Theory of Language Structure. Oxford, UK: Oxford University Press. Henrich, J. & Boyd, R. (2008). “Division of Labor, Economic Specialization,

CO22552.indb 156

23/04/2010 14:13

References

157

and the Evolution of Social Stratification”. Current Anthropology 49–4: pp. 715–724. Henshilwood, C. S. & Marean, C. W. (2003). “The Origin of Modern Human Behavior: Critique of the Models and Their Test Implications”. Current Anthropology 44–5: pp. 627–651. Henshilwood, C. S., d’Errico, F., Vanhaeren, M., van Niekerk, K. & Jacobs, Z. (2004). “Middle Stone Age Shell Beads from South Africa”. Science 304–5669: p. 404. Herman, L. M. & Uyeyama, R. K (1999). “The dolphin’s grammatical competency: Comments on Kako (1999)”. Animal Learning & Behavior, 27–1: pp. 18–23. Hervey, S. G. J. (1979). Axiomatic Semantics: a theory of linguistic semantics. Edinburgh, UK: Scottish Academic Press. Hewes, G. W. (1961). “Food transport and the origin of hominid bipedalism”. American Anthropologist 63: pp. 687–710. Hirsh-Pasek, K. & Golinkoff, R. M. (1996). The Origins of Grammar: evidence from early language comprehension. Cambridge, Mass., USA: MIT Press. Hölldobler, B. & Wilson, E. O. (2009). The Super-Organism: the beauty, elegance, and strangeness of insect societies. London, UK: W W Norton & co. Hopkins, W. D. & Pilcher, D. L. (2001). “Neuroanatomical Localization of the Motor Hand Area With Magnetic Resonance Imaging: The Left Hemisphere Is Larger in Great Apes”. Behavioral Neuroscience 115–5: pp. 1159–1164. Hopper, P. J. & Traugott, E. C. (1993). Grammaticalization. Cambridge, UK: Cambridge University Press. Hornstein, N. (1990). As Time Goes By: tense and Universal Grammar. Cambridge: Mass, USA: MIT Press. Hornstein, N. (2001). Move! A Minimalist Theory of Construal. Oxford, UK: Blackwell Publishers Ltd. Hrdy, S. B. (2009). Mothers and Others: the evolutionary origins of mutual understanding. Cambridge, Mass., USA: Belknap Press. Hudson, R. (1998). English Grammar. London, UK: Routledge. von Humboldt, W. (1999 [1836]). On Language: on the diversity of human language construction and its influence on the mental development of the human species. Cambridge, UK: Cambridge University Press. Hunt, G. R. & Gray, R. D. (2004). “Direct observations of pandanus-tool manufacture and use by a New Caledonian crow (Corvus moneduloides)”. Animal Cognition 7: pp. 114–120. Hurford, J. R, Studdert-Kennedy, M. & Knight, C. (eds) (1998). Approaches to the Evolution of Language. Cambridge, UK: Cambridge University Press. Hurford, J. R. (2003). “The Neural Basis of Predicate-Argument Structure”. Behavioural and Brain Sciences 26–3: pp. 283–284. Hurford, J. R. (2007). The Origins of Meaning: language in the light of evolution. Oxford, UK: Oxford University Press.

CO22552.indb 157

23/04/2010 14:13

158

References

Iverson, J. M. & Goldin-Meadow, S. (1998). “Why People Gesture when They Speak”. Nature, 396–6708: p. 228. Jackendoff, R. (1972). Semantic Interpretation in Generative Grammar. Cambridge, Mass., USA: MIT Press. Jackendoff, R. (2002). Foundations of Language: brain, meaning, grammar, evolution. Oxford, UK: Oxford University Press. Jakobson, R. (1987). Language in Literature. Cambridge, Mass., USA: Harvard University Press. Jaynes, J. (1990). The Origin of Consciousness in the Breakdown of the Bicameral Mind. London, UK: Penguin. Johanson, D. & Edgar, B. (1996). From Lucy to Language. London, UK: Weidenfeld & Nicolson. Johansson, S. (2005). Origins of Language. Amsterdam, Netherlands: John Benjamins Publishing co. Johnson, M. (1987). The Body in the Mind: the bodily basis of meaning, imagination, and reason. Chicago, Ill., USA: University of Chicago Press. Johnson, S. (2003). “Emotions and the Brain”. Discover Magazine, April 2003. Jolly, C. J. (2009). “Fifty Years of Looking at Human Evolution: Backward, Forward, and Sideways”. Current Anthropology 50–2: pp. 187–199. Jordan, J. S. (2003). “Emergence of self and other in perception and action: An event-control approach”. Consciousness and Cognition 12: pp. 633–646. Jordan, R. (1998). “Is Autism a pathology? Reflections from theory, research and practice”. Psychobiology of Autism: Current research and practice. Sunderland Autism Research Unit, Durham conference papers. Jung-Beeman, M. (2005). “Bilateral brain processes for comprehending natural language”. Trends in Cognitive Sciences 9–11: pp. 512–518. Kaminski, J., Call, J. & Fischer, J. (2004). “Word Learning in a Domestic Dog: Evidence for ‘Fast Mapping’ ”. Science, 304–5677: pp. 1682–1683. Kathol, A. (2000). Linear Syntax. Oxford, UK: Oxford University Press. Kaye, K. (1982). The Mental and Social Life of Babies: how parents create persons. London, UK: Methuen & co. Kendon, A. (1991). “Some Considerations for a Theory of Language Origins”. Man 26–2: pp. 199–221. Key, C. A. & Aiello, L. C. (1999). “The Evolution of Social Organisation”. In R. Dunbar, C. Knight & C. Power (eds), The Evolution of Culture. Edinburgh, UK: Edinburgh University Press. Kinsella, A. R. (2009). Language Evolution and Syntactic Theory. Cambridge, UK: Cambridge University Press. Kirby, S. (1998). “Language evolution without natural selection: From vocabulary to syntax in a population of learners”. Edinburgh Occasional Papers in Linguistics April 3, 1998. Knight, C. (1991). Blood Relations: menstruation and the origins of culture. New Haven, Mass., USA: Yale University Press.

CO22552.indb 158

23/04/2010 14:13

References

159

Knight, C. (1998). “Ritual/speech coevolution: a solution to the problem of deception”. In J. R. Hurford, M. Studdert-Kennedy & C. Knight (eds), Approaches to the Evolution of Language. Cambridge, UK: Cambridge University Press. Knight, C. (1999). “Sex and Language as Pretend Play”. In R. Dunbar, C. Knight & C. Power (eds), The Evolution of Culture. Edinburgh, UK: Edinburgh University Press. Knight, C. (2008). “Language co-evolved with the rule of law”. Mind & Society 7–1: pp. 109–128. Kövecses, Z. (2002). Metaphor: a practical introduction. Oxford, UK: Oxford University Press. Kyle, J. G. & Woll, B. (1985).Sign Language: the study of deaf people and their language. Cambridge, UK: Cambridge University Press. Lakoff, G. (1971). “On Generative Semantics”. In D. D. Steinberg & L. A. Jacobovits (eds), Semantics: an interdisciplinary reader in Philosophy, Linguistics and Psychology. Cambridge, UK: Cambridge University Press. Lakoff, G. (1987). Women Fire and Dangerous Things: what categories reveal about the mind. Chicago, Ill., USA: University of Chicago Press. Lakoff, G. & Johnson, M. (1980). Metaphors We Live By. Chicago, USA: University of Chicago Press. Lakoff, G. & Johnson, M. (1999). Philosophy in the Flesh: the embodied mind and its challenge to Western thought. New York, USA: Basic Books. Laland, K. N., Odling-Smee, J. & Feldman, M. W. (1999). “Niche Construction, Biological Evolution and Cultural Change”. Behavioral and Brain Sciences 23–1: pp. 131–175. Langacker, R. W. (1987). Foundations of Cognitive Grammar: volume I, theoretical prerequisites. Stanford, Cal., USA: Stanford University Press. Langacker, R. W. (1991). Foundations of Cognitive Grammar; volume II, descriptive application. Stanford, Cal., USA: Stanford University Press. Lee, D. (1992). Competing Discourses: perspective and ideology in language. Harlow, UK: Longman. Lewis, J. (2009). “As Well as Words: Congo Pygmy hunting, mimicry and play”. In R. Botha & C. Knight (eds), The Cradle of Language. Oxford, UK: Oxford University Press. Liebers, D., de Knijff, P. & Helbig, A. J. (2004). “The herring gull complex is not a ring species”. Proceedings of the Royal Society of London B 271–1542: pp. 893–901. Liti, G., Carter, D. M., Moses, A. M., Warringer, J., Parts, L., James, S. A., Davey, R. P., Roberts, I. N., Burt, A., Koufopanou, V., Tsai, I. J., Bergman, C. M., Bensasson, D., O’Kelly, M. J. T., van Oudenaarden, A., Barton, D. B. H., Bailes, E., Nguyen, A. N., Jones, M., Quail, M. A., Goodhead, I., Sims, S., Smith, F., Blomberg, A., Durbin, R. & Louis, E. J. (2009). “Population genomics of domestic and wild yeasts”. Nature, 458–7236: pp. 337–341.

CO22552.indb 159

23/04/2010 14:13

160

References

Locke, J. L. (1998). “Social Sound-Making as a Precursor to Spoken Language”. In J. R. Hurford, M. Studdert-Kennedy & C. Knight (eds), Approaches to the Evolution of Language. Cambridge, UK: Cambridge University Press. Lockwood, D. G. (2002). “Highlighting in Stratificational-Cognitive Linguistics”. In W. Spruiell, M. Cummings, D. Lockwood & P. Fries (eds), Relations and Functions Within and Around Language. London, UK: Continuum. Losin, E. A. R., Russell, J. L., Freeman, H., Meguerditchian, A. & Hopkins, W. D. (2008). “Left Hemisphere Specialization for Oro-Facial Movements of Learned Vocal Signals by Captive Chimpanzees”. Public Library of Science One, 3–6: e2529. Lycett, S. J., Collard, M. & McGrew, W. C. (2007). “Phylogenetic analyses of behavior support existence of culture among wild chimpanzees”. Proceedings of the National Academy of Sciences, 104–45: pp. 17588–17592. MacNeilage, P. F. (1998). “Evolution of the mechanism of language output: comparative neurobiology of vocal and manual communication”. In J. R. Hurford, M. Studdert-Kennedy, C. Knight (eds), Approaches to the Evolution of Language. Cambridge, UK: Cambridge University Press. Massey, D. S. (2002). “A Brief History of Human Society: The Origin and Role of Emotion in Social Life”. American Sociological Review, 67–2: pp. 1–29. Malik, K. (2000). Man, Beast and Zombie: what science can and cannot tell us about human nature. London, UK: Phoenix. Marino, L., Connor, R. C., Fordyce, R. E., Herman, L. M., Hof, P. R., Lefebvre, L., Lusseau, D., McCowan, B., Nimchinsky, E. A., Pack, A. A., Rendell, L., Reidenberg, J. S., Reiss, D., Uhen, M. D., van der Gucht, E. & Whitehead, H. (2007). “Cetaceans Have Complex Brains for Complex Cognition”. Public Library of Science Biology, 5–5 e139: pp. 966–972. Marks, J. (2002). What it Means to be 98% Chimpanzee: apes, people, and their genes. Berkeley, USA: University of California Press. Marshall, A. J., Wrangham, R. W. & Arcadi, A. C. (1999). “Does learning affect the structure of vocalizations in chimpanzees?” Animal Behaviour, 58–4: pp. 825–830. Mauss, M. (1950). The Gift: the form and reason for exchange in archaic societies. London, UK: Routledge. Meares, R. & Sullivan, G. (2004). “Two Forms of Human Language”. In G. Williams & A. Lukin (eds), The Development of Language: functional perspectives on species and individuals. London, UK: Continuum. Melis, A.P., Hare, B. & Tomasello, M. (2006). “Engineering cooperation in chimpanzees: tolerance constraints on cooperation”. Animal Behaviour, 72: pp. 275–286. Mithen, S. (1996). The Prehistory of the Mind: a search for the origins of art, religion and science. London, UK: Phoenix. Mithen, S. (2005). The Singing Neanderthals: the origins of music, language, mind and body. London, UK: Weidenfeld & Nicolson.

CO22552.indb 160

23/04/2010 14:13

References

161

Mosch, S. C., Max, J. E. & Tranel, D. (2005). “A Matched Lesion Analysis of Childhood versus Adult-Onset Brain Injury Due to Unilateral Stroke: Another Perspective on Neural Plasticity and Recovery of Social Functioning”. Cognitive and Behavioral Neurology, 18–1: pp. 5–17. Morley, G. D. (2000). Syntax in Functional Grammar: an introduction to lexicogrammar in systemic linguistics. London, UK: Continuum. Mulder, J. W. F. & Hervey, S. G. J. (1980). The Strategy of Linguistics: papers on the theory and methodology of axiomatic functionalism. Edinburgh, UK: Scottish Academic Press. Nettle, D. (1999). Linguistic Diversity. Oxford, UK: Oxford University Press. Nettle, D. & Pollet, T. V. (2008). “Natural Selection on Male Wealth in Humans”. The American Naturalist, 172–5: pp. 658–666. Newmeyer, F. J. (2000). Language Form and Language Function. Cambridge, Mass, USA: MIT Press. Newton, M. (2002). Savage Girls and Wild Boys: a history of feral children. London, UK: Faber & Faber. Nichols, S. & Stich, S. (2003). “How to Read Your Own Mind: A Cognitive Theory of Self-Consciousness”. In Q. Smith & A. Jokic (eds), Consciousness: New Essays. Oxford, UK: Oxford University Press. Nietzsche, F. (1976 [1874]). “On Truth and Lie in an Extra-Moral Sense”. W. Kaufmann (tr), The Portable Nietzsche. New York, USA: Viking Press. Niles, J. D. (1999). Homo Narrans: The Poetics and Anthropology of Oral Literature. Philadelphia, Penn, USA: University of Pennsylvania Press. Noonan, M. (1999). “Non-structuralist syntax”. In M. Darnell, E. A. Moravcsik, M. Noonan, F. J. Newmeyer & K. M. Wheatley (eds), Functionalism and Formalism in Linguistics, Volume I: General papers. Amsterdam, Netherlands: John Benjamins. Noss, A. J. & Hewlett, B. S. (2001). “The Contexts of Female Hunting in Central Africa”. American Anthropologist 103–2: pp. 1024–1040. Núñez, R. E. & Sweetser, E. (2006). “With the Future Behind Them: Convergent Evidence From Aymara Language and Gesture in the Crosslinguistic Comparison of Spatial Construals of Time”. Cognitive Science 30: pp. 1–49. Odling-Smee, J. & Laland, K. N. (2009). “Cultural Niche Construction: evolution’s cradle of language”. In R. Botha & C. Knight (eds), The Prehistory of Language. Oxford, UK: Oxford University Press. Opie, K. (2004). Testing the Grandmothering Hypothesis: the provisioning of Homo erectus infants and juveniles. Dissertation submitted in partial fulfilment of the requirements for the degree of M.Sc. in Human Evolution and Behaviour (UCL) of the University of London. Palmer, G. B. (2006). “Energy Through Fusion at Last: synergies in cognitive anthropology and cognitive linguistics”. In G. Kristiansen, M. Achard, R. Dirven & F. J. Ruiz de Mendoza Ibáñez (eds), Cognitive Linguistics: current applications and future perspectives. Berlin, Germany: Mouton de Gruyter.

CO22552.indb 161

23/04/2010 14:13

162

References

Paradis, J. & Crago, M. (2000). “Tense and temporality: A comparison between children learning a second language and children with SLI”. Journal of Speech, Language and Hearing Research, 43–4: pp. 834–848. Parisi, D. & Schlesinger, M. (2002). “Artificial Life and Piaget”. Cognitive Development 17: pp. 1301–1321. Parker, A. R. (2006). Evolution as a Constraint on Theories of Syntax: The Case against Minimalism. PhD thesis, Edinburgh University, March 2006. Pepperberg, I. M. (1999). The Alex Studies: cognitive and communicative abilities of Grey Parrots. Cambridge, Mass., USA: Harvard University Press. Peters, A. M. (1995). “Strategies in the Acquisition of Syntax”. P. Fletcher & B. MacWhinney (eds), The Handbook of Child Language. Oxford, UK: Blackwell Publishers. Piaget, J. (1959 [1923]). The Language and Thought of the Child. London, UK: Routledge. Piaget, J. (1950 [1947]). The Psychology of Intelligence. London, UK: Routledge. Pinker, S. (1994). The Language Instinct. London, UK: Penguin. Pinker, S. (1997). How the Mind Works. London, UK: Penguin. Pinker, S. (2002). The Blank Slate: the modern denial of human nature. London, UK: Penguin. Pinker, S. & Jackendoff, R. (2005). “The faculty of language: what’s special about it?” Cognition 95: pp. 201–236. Pitchford, I. (2001). “The Origins of Violence: Is Psychopathy an Adaptation?” The Human Nature Review 1: pp. 28–36. Plotnik, J. M., de Waal, F. B. M. & Reiss, D. (2006). “Self-recognition in an Asian elephant”. Proceedings of the National Academy of Science, 103–45: pp. 17053–17057. Plunkett, K. (1995). “Connectionist Approaches to Language Acquisition”. In P. Fletcher & B. MacWhinney (eds), The Handbook of Child Language. Oxford, UK: Blackwell Publishers. Pollick, A. S. & de Waal, F. B. M. (2007). Ape gestures and language evolution. Proceedings of the National Academy of Science, 104–19: pp. 8184–8189. Popper, K. (1985 [1967]). “Knowledge: subjective versus objective”. In D. Miller (ed), Popper Selections. Princeton, NJ, USA: Princeton University Press. Postal, P. M. (1968). “Underlying and Superficial Linguistic Structure”. In R. C. Oldfield & J. C. Marshall (eds), Language. London, UK: Penguin. Povinelli, D. J. (2000). Folk Physics for Apes: the chimpanzee’s theory of how the World works. Oxford, UK: Oxford University Press. Premack, D. & Premack, A. J. (1983). The Mind of an Ape. London, UK: W W Norton & co. Premack, D. & Hauser, M. D. (2001). “A whale of a tale: Calling it culture doesn’t help”. Behavioral and Brain Sciences, 24–2: pp. 350–351. Pressman, S. D., Cohen, S., Miller, G. E., Barkin, A., Rabin, B. S. & Treanor, J. J. (2005). “Loneliness, Social Network Size, and Immune Response to

CO22552.indb 162

23/04/2010 14:13

References

163

Influenza Vaccination in College Freshmen”. Health Psychology, 24–3: pp. 297–306. Prior, H., Schwarz, A. & Güntürkün, O. (2008). “Mirror-Induced Behavior in the Magpie (Pica pica): Evidence of Self-Recognition”. Public Library of Science Biology, 6–8: pp. 1642–1650. Pruetz, J. D. & Bertolani, P. (2007). “Savanna Chimpanzees, Pan troglodytes verus, Hunt with Tools”. Current Biology 17: pp. 412–417. Queller, D. C. (1994). “Extended parental care and the origin of eusociality”. Proceedings of the Royal Society of London B 256: pp. 105–111. Reichenbach, H. (1956 [1927]). The Philosophy of Space & Time. New York, USA: Dover Publications inc. Reiss, D. & Marino, L. (2001). “Mirror self-recognition in the bottlenose dolphin: A case of cognitive convergence”. Proceedings of the National Academy of Science, 98–10; pp. 5937–5942. Rendell, L. & Whitehead, H. (2001). “Culture in whales and dolphins”. Behavioral and Brain Sciences, 24–2: pp. 309–382. Richmond, B. G. & Jungers, W. L. (2008). “Orrorin tugenensis Femoral Morphology and the Evolution of Hominin Bipedalism”. Science, 319: pp. 1662–1665. Roth, G. & Dicke, U. (2005). “Evolution of the brain and intelligence”. Trends in Cognitive Sciences, 9–5. Sabbagh, L. (2006). “The Teen Brain, Hard at Work—No, Really”. Scientific American Mind, 17–4: pp. 20–25. Sampson, G. (1997). Educating Eve: the “language instinct” debate. London, UK: Cassell. Sampson, G. (2005). The “Language Instinct” Debate (revised edition). London, UK: Continuum. Sanz, C., Morgan, D. & Gulick, S. (2004). “New Insights into Chimpanzees, Tools, and Termites from the Congo Basin”. American Naturalist, 164: pp. 567–581. Sapir, E. (1921). Language: an introduction to the study of speech. Orlando, Fla., USA: Harcourt Brace & co. de Saussure, F. (1972). Course in General Linguistics (tr. Roy Harris). London, UK: Duckworth. Savage-Rumbaugh, S. & Lewin, R. (1994). Kanzi: the ape at the brink of the human mind. New York, USA: John Wiley & Sons Inc. Savage-Rumbaugh, S. (1999). “Ape Language: between a rock and a hard place”. In B. King (ed), The Origins of Language: what non-human primates can tell us. Santa Fe, NM, USA: SAR Press. Savage-Rumbaugh, S, Fields, W., Segerdahl, P., & Rumbaugh, D. (2005). “Culture Prefigures Cognition in Pan/Homo Bonobos”. Theoria: An International Journal for Theory, History and Foundations of Science, 54: pp. 311–328. Schoenemann, P. T., Sheehan, M. J. & Glotzer, L. D. (2005). “Prefrontal White

CO22552.indb 163

23/04/2010 14:13

164

References

Matter Volume Is Disproportionately Larger in Humans than in Other Primates”. Nature Neuroscience, 8–2: pp. 242–252. Scott-Phillips, T. C. (2008). “On the Correct Application of Animal Signalling Theory to Human Communication”. In A. D. M. Smith, K. Smith & R. Ferrer-i-Cancho (eds), Proceedings of the 7th International Conference on the Evolution of Language. Singapore: World Scientific. Sear, R. & Mace, R. (2008). “Who keeps children alive? A review of the effects of kin on child survival”. Evolution and Human Behavior, 29–1: pp. 1–18. Searle, J. R. (1993). “Metaphor”. In A. Ortony (ed), Metaphor and Thought 2nd edition. Cambridge, UK: Cambridge University Press. Searle, J. R. (1999). Mind, Language and Society. London, UK: Weidenfeld & Nicolson. Segerdahl, P., Fields, W. & Savage-Rumbaugh, S. (2006). Kanzi’s Primal Language: the cultural initiation of primates into language. Basingstoke, UK: Palgrave Macmillan. Shirai, Y. & Miyata, S. (2006). “Does past tense marking indicate the acquisition of the concept of temporal displacement in children’s cognitive development?” First Language, 26–1: pp. 45–66. Shoshani, J., Kupsky, W. J. & Marchant, G. H. (2006). “Elephant Brain Part I: Gross morphology, functions, comparative anatomy, and evolution”. Brain Research Bulletin, 70: pp. 124–157. Singer, T., Seymour, B., O’Doherty, J. P., Stephan, K. E., Dolan, R. J. & Frith, C. D. (2006). “Empathic neural responses are modulated by the perceived fairness of others”. Nature, 439: pp. 466–469. Slocombe, K. E. & Zuberbühler, K. (2007). “Chimpanzees modify recruitment screams as a function of audience composition”. Proceedings of the National Academy of Science, 104–43: pp. 17228–17233. Sloman, A. (2008). “Putting the Pieces Together Again”. In R. Sun (ed), The Cambridge Handbook of Computational Psychology. Cambridge, UK: Cambridge University Press. Smith, P. K., Cowie, H. & Blades, M. (2003). Understanding Children’s Development (4th edition). Oxford, UK: Blackwell Publishing. Sober, E. & Wilson, D. S. (1999). Unto Others: the evolution and psychology of unselfish behavior. Cambridge, Mass, USA: Harvard University Press. Spencer, H. (1898). The Principles of Biology volume 1. New York, USA: D Appleton & Co. Steels, L. (1998). “Synthesising the origins of language and meaning using co-evolution, self-organisation and level formation”. In J. Hurford, C. Knight & M. Studdert-Kennedy (eds), Approaches to the Evolution of Language: social and cognitive bases. Cambridge, UK: Cambridge University Press. Steels, L. (2003). “Social Language Learning”. In M. Tokoro & L. Steels (eds), The Future of Learning. Amsterdam, Netherlands: IOS Press. Steklis, H. D. & Harnad, S. R. (1976). “From hand to mouth: Some critical stages

CO22552.indb 164

23/04/2010 14:13

References

165

in the evolution of language”. In S. Harnad, H. D. Steklis & J. Lancaster (eds), Origins and Evolution of Language and Speech. New York, USA: New York Academy of Sciences. Stewart, I. & Cohen, J. (1997). Figments of Reality: the evolution of the curious mind. Cambridge, UK: Cambridge University Press. Stout, D. & Chaminade, T. (2009). “Making Tools and Making Sense: Complex, Intentional Behaviour in Human Evolution”. Cambridge Archaeological Journal 19–1: pp. 85–96. Stringer, C. & Gamble, C. (1993). In Search of the Neanderthals. London, UK: Thames and Hudson. Stringer, C. & Andrews, P. (2005). The Complete World of Human Evolution. London, UK: Thames & Hudson. Szathmáry, E. & Számadó, S. (2008). “Language: a social history of words”. Nature, 456: pp. 40–41. Tannen, D. (1994). Gender and Discourse. Oxford, UK: Oxford University Press. Tanner, J. E., Patterson, F. G. & Byrne, R. W. (2006). “The development of spontaneous gestures in zoo-living gorillas and sign-taught gorillas: from action and location to object representation”. Journal of Developmental Processes 1: pp. 69–102. Taylor, P. D. & Day, T. (2004). “Cooperate with thy Neighbour?” Nature, 428: pp. 611–612. Terrace, H. S., Petitto, L. A., Sanders, R. J. & Bever, T. G. (1979). “Can an Ape Create a Sentence?” Science, 206: pp. 891–902 Theissen, G. (2006). “The proper place of hopeful monsters in evolutionary biology”. Theory in Biosciences, 124: pp. 349–369. Thomas, J. (1995). Meaning in Interaction: an introduction to pragmatics. Harlow, Essex, UK: Longman Ltd. Thompson, G. (1996). Introducing Functional Grammar. London, UK: Arnold. Thorndike, E. (1927). The Measurement of Intelligence. New York, USA: Columbia University. Thulborn, K. R., Carpenter, P. A. & Just, M. A. (1999). “Plasticity of LanguageRelated Brain Function During Recovery From Stroke”. Stroke, 30: pp. 749–754. Tibbetts, E. A. & Dale, J. (2004). “A socially enforced signal of quality in a paper wasp”. Nature, 432: pp. 218–222. Tibbetts, E. A., Sheehan, M. J. & Dale, J. (2008). “A testable definition of individual recognition”. Trends in Ecology and Evolution 23–7: p. 356. Tomalin, M. (2004). “Leonard Bloomfield: linguistics and mathematics”. Historiographia Linguistica xxxi: 1: pp. 105–136. Tomasello, M. & Call, J. (1997). Primate Cognition. Oxford, UK: Oxford University Press. Tomasello, M. (1999). The Cultural Origins of Human Cognition. Cambridge, USA: Harvard University Press.

CO22552.indb 165

23/04/2010 14:13

166

References

Tomasello, M. (2003a). Constructing a Language: a usage-based theory of language acquisition. Cambridge, Mass., USA: Harvard University Press. Tomasello, M. (2003b). “On the Different Origins of Symbols and Grammar”. M. H. Christiansen & S. Kirby (eds), Language Evolution. Oxford, UK: Oxford University Press. Tomasello, M., Carpenter, M., Call, J., Behne, T., & Moll, H. (2005). “Understanding and sharing intentions: The origins of cultural cognition”. Behavioral and Brain Sciences 28: pp. 675–735. Tomasello, M. (2008). Origins of Human Communication. Cambridge, Mass., USA: MIT Press. Toolan, M. (1996). Total Speech: an integrational linguistic approach to language. Durham, NCa., USA: Duke University Press. Torr, J. (1997). From Child Tongue to Mother Tongue: a case study of language development in the first two and a half years. Nottingham, UK: Monographs in Systemic Linguistics. Traulsen, A. and Nowak, M. A. (2006). “Evolution of cooperation by multilevel selection”. Proceedings of the National Academy of Science, 103–29: pp. 10952–10955. Trivers, R. (1971). “The Evolution of Reciprocal Altruism”. The Quarterly Review of Biology, 46: pp. 35–57. Valian, V. (2006). “Young Children’s Understanding of Present and Past Tense”. Language Learning and Development, 2–4: pp. 251–276. van Vugt, M. & Schaller, M. (2008). “Evolutionary Approaches to Group Dynamics: An Introduction”. Group Dynamics: Theory, Research, and Practice, 12–1: pp. 1–6. Vygotsky, L. (1986 [1934]). Thought and Language. Cambridge, Mass., USA: MIT Press. Vygotsky, L. (1978 [1935]). Interaction between Learning and Development. In Mind in Society: the development of higher psychological processes. Cambridge, Mass., USA: Harvard University Press. de Waal, F. B. M. (1996). Good Natured: the origins of right and wrong in humans and other animals. Cambridge, Mass., USA: Harvard University Press. de Waal, F. B. M. & Lanting, F. (1997). Bonobo: the forgotten Ape. Berkeley, Cal., USA: University of California Press. de Waal, F. B. M. (2004). “Peace Lessons from an Unlikely Source”. Public Library of Science: Biology, 2–4: pp. 434–436. de Waal, F. B. M., Dindo, M., Freeman, C. A. & Hall, M. J. (2005). “The monkey in the mirror: Hardly a stranger”. Proceedings of the National Academy of Science, 102–32: pp. 11140–11147. de Waal, F. B. M. (2006). Primates and Philosophers: how morality evolved. Princeton, NJ, USA: Princeton University Press. Warneken, F. & Tomasello, M. (2006). Altruistic Helping in Human Infants and Young Chimpanzees. Science, 311: pp. 1301–1303.

CO22552.indb 166

23/04/2010 14:13

References

167

Weir, A. A. S., Chappell, J. & Kacelnik, A. (2002). “Shaping of Hooks in New Caledonian Crows”. Science, 297: p. 981. Weist, R. M. (1986). “Tense and aspect: Temporal systems in child language”. In P. Fletcher & M. Garman (eds), Language acquisition: Studies in first language development. Cambridge, UK: Cambridge University Press. Whiten, A. & Byrne, R. W. (1988). “Tactical deception in primates”. Behavioral and Brain Sciences, 11: pp. 233–273. Whorf, B. L. (1956). Language, Thought and Reality. Cambridge, Mass., USA: MIT Press. Wickler, W. (1965). “Mimicry and the Evolution of Animal Communication”. Nature, 208: pp. 519–521. Wilkinson, G. S. (1984). “Reciprocal food sharing in the vampire bat”. Nature, 308: pp. 181–184. Wilson, R. A. (1937). The Miraculous Birth of Language. London, UK: British Publisher’s Guild. van Wolkenten, M., Brosnan, S. F. & de Waal, F. B. M. (2007). “Inequity responses of monkeys modified by effort”. Proceedings of the National Academy of Science, 104–47: pp. 18854–18859. Worden, R. P. (1998). “The Evolution of Language from Social Intelligence”. In J. R. Hurford, M. Studdert-Kennedy & C. Knight (eds), Approaches to the Evolution of Language. Cambridge, UK: Cambridge University Press. Wray, A. (2000). “Holistic Utterances in Protolanguage: the link from primates to humans”. In C. Knight, M. Studdert-Kennedy & J. R. Hurford (eds), The Evolutionary Emergence of Language: social function and the origins of linguistic form. Cambridge, UK: Cambridge University Press. Wray, A. (2002a). “Dual Processing in Protolanguage: performance without competence”. In A. Wray (ed), The Transition to Language. Oxford, UK: Oxford University Press. Wray, A. (2002b). Formulaic Language and the Lexicon. Cambridge, UK: Cambridge University Press. Zahavi, A. & Zahavi, A. (1997). The Handicap Principle: a missing piece of Darwin’s puzzle. Oxford, UK: Oxford University Press. Zahavi, A. (2003). “Indirect selection and individual selection in sociobiology: my personal views on theories of social behaviour”. Animal Behaviour, 65: pp. 859–863. Ziff, P. (1960). Semantic Analysis. New York, USA: Cornell University Press. Zuberbühler, K. (2000). “Referential labelling in Diana monkeys”. Animal Behaviour, 59: pp. 917–927.

CO22552.indb 167

23/04/2010 14:13

This page intentionally left blank

CO22552.indb 168

23/04/2010 14:13

INDEX

ability 2, 5, 8, 10, 16, 18, 20, 63–64, 71, 75–81, 99–100, 106, 117, 119–120, 122, 126, 130, 134–136, 138, 141–142 abstract 2, 31, 33, 48, 54, 72, 99, 103, 115, 138 action 13, 18, 30, 41, 54, 62, 72, 78, 79, 85, 91–93, 96, 98, 124, 126, 132, 136, 139, 143 actor 30, 41, 98 actuality 9, 18, 22–23, 52, 54, 71–72 adjective 31, 42, 46–47, 49 adulthood 18, 80, 118, 122, 141 adverb 47 Africa 17, 82 Aiello, L. 3, 83, 121 Aitchison, J. 15, 112, 145 Alex the grey parrot 113 allopatric speciation 82 altruism 10, 77, 83, 88 altruistic punishment 66, 83–84, 88–89, 91, 101, 111, 119, 122, 138–139 Ambrose, Stanley H. 5 America 17, 53, 61 Ameslan 112, 113 analog ‘I’ 75–76, 78–79 animal 50, 64, 72, 85–86, 104–105, 109, 118, 131, 142 anthropology 13–14, 27, 38, 51, 94 ape 112 Apis mellifera 8 Arbib, M. A. 22, 108 archaeology 16, 63

CO22552.indb 169

arcuate fasciculus 55 Ardipithecus 82 argument, one- 13, 72–73, 91, 98, 102, 123–124 argument, three- 59, 69, 93, 103, 115, 133, 140–141, 143 argument, two- 59, 72–73, 92, 97–101, 103, 115, 118, 123–124, 138–140 Aristotle 81 attention 14–15, 59, 105, 120, 123 Australopithecus 4, 82, 137 autism 87, 119, 120, 142 autonomic 76, 135 auxiliary 53, 130, 132, 133 awareness 54, 75–76, 78–81, 86, 90, 106–107, 119, 122, 135, 142 awareness, other 77, 79, 120 awareness, self 76–79, 122 axiomatic Functionalism 42 Aymara 53, 126 baboon 59 Baker, M. C. 36, 68 Balter, M. 82 Barnes, D. M. 5 Baron-Cohen, S. 120 Bates, E. et al. 123 Bateson, G. 17 Baumeister, R. F. et al. 76 beaver 5 bee 5 behaviour 6, 10, 59, 66, 74, 110–111, 118, 120–122, 131 169

23/04/2010 14:13

170

Benveniste, É. 133 Berg, L. 123 Berkeley, G. 54 Bickerton, D. 34–35, 58, 117 bipedalism 3–4, 8, 14, 62, 65, 82 birdsong 21 Bloom, P. 79–80 Bloomfield, L. 15, 27, 132 Bloor, T. & Bloor, M. 38 Boas, F. 27 Bod, R. 46 body 2–3, 11, 19, 23, 53–54, 65, 82, 107, 126 Boehm, C. 88, 139 Boesch-Achermann, H. & Boesch, C. 4 bone 16 bonobo 48, 64–65, 82, 110, 113 Bourke, A. F. G. & Franks, N. R. 86, 142 Bowdler 6 Boyd, R. et al. 66, 138 Bradbury, J. W. & Vehrencamp, S. L. 136 brain 2–3, 6, 21, 23, 28, 38, 50–51, 54–56, 62–63, 76, 82, 120–121 Brewer, S. 5, 119, 121, 131 Brighton, H. et al. 59 Broca’s area 55 Broom, D. M. et al. 107 Brown, R. 123 Bruner, J. 78 Budiansky, S. 106 Burland, T. M. et al. 89 Burling, R. 144 bushbaby 5 Buzing, P. C. et al. 20 Byrne, R. 6, 74, 82, 86, 108 Caillois, R. 18 Calvin, W. H. & Bickerton, D. 34 Cameron, D. 102 Campbell, A. 19

CO22552.indb 170

Index

capacity 1–2, 4–7, 9, 19, 25, 27–28, 31, 33–34, 38, 44, 48, 51, 53–54, 56, 62–64, 66, 71–73, 76, 79, 81–83, 86–90, 93, 97–103, 105–110, 112–113, 116–118, 126–127, 129, 131–137, 140–144 Carruthers, P. 135 Carstairs McCarthy, A. 97 Carter, R. 55, 120 case grammar 28 cat 47 catastrophic evolution 34 Cercopithecus diana 8, 59, 136 Chafe, W. 50 channel 21–22, 98, 112 cheat 84, 111 Cheney, D. L. & Seyfarth, R. M. 108, 136 Chiappe, D. et al. 83 Chiat, S. 48 children 34, 37, 48, 50, 59, 65, 67–68, 80, 115–117, 119, 121–124, 132, 141 chimpanzee 4–5, 8, 61–65, 67–68, 71, 73–74, 82, 91, 107, 109–110, 112–113, 131, 136 Chomsky, N. 11, 15, 23, 27–38, 50, 72, 79, 99, 103, 115, 117, 134, 143 circumstance 41, 57, 91–92, 98, 101, 108, 111, 116 civilization 24 Clark, E. V. 132 Clutton-Brock, T. H. & Isvaran, K. 68 cognition 3, 5, 8, 10–12, 14, 16, 22–23, 25–27, 35–36, 39, 44, 49–57, 59, 71, 73–74, 76, 78–79, 86, 90, 94, 97, 98, 99, 100, 102, 105–106, 108–109, 113, 116, 118, 120, 122, 124–126, 137, 139 Cognitive Dissonance 41 Cognitive Grammar 50, 57, 97

23/04/2010 14:13

Index

Cognitive Linguistics 23, 43, 48–54, 56, 57, 58 Cognitive Psychology 49 collie 9, 113 communication 1, 8, 10, 12–14, 16–18, 22–25, 31, 34–36, 38, 40–41, 46, 48, 50–53, 55–58, 63, 70, 84, 92–94, 96–103, 105–106, 110, 112–113, 124, 133, 135–137, 139–144 community 6–7, 15, 21, 32, 40, 50, 79, 99, 103, 111, 113, 144 competition 10, 20, 25, 65, 131 complexity 1, 3, 6, 8–9, 13, 16–17, 20–22, 24, 86, 94–96, 100, 102–103, 117, 133–134, 144 component 16, 20, 23, 50, 57, 120, 125, 143 computer 3, 28, 59 Conard, N. J. 82 conditional 17–18, 130, 142 Conklin-Brittain, N. L. et al. 4 connectivity 129, 133 consciousness 22, 74, 76–79, 81, 86, 96, 106, 109, 116–117, 121, 135–136 construct 16, 29, 33, 40–42, 46–48, 52, 78, 103, 129 Construction Grammar 50, 57 context 13, 40–41, 46, 51, 53, 76, 78, 129 continuity 1, 16, 24, 35, 80, 126, 128–130, 132, 133 Cook, G. 18 Cook, V. J. and Newson, M. 31 co-operation 3, 6–7, 9–10, 13–14, 18, 21, 24, 26, 35, 63–70, 77–78, 81, 83, 89, 91, 109–111, 113, 118–122, 124, 131, 134, 138, 142–143 Corballis, M. C. 21–22, 62 Corning, P. A. 85 cortex 3, 55, 117, 120–121

CO22552.indb 171

171

Corvus moneduloides 5, 131 Cosmides, L. & Tooby, J. 56 costly signalling 7, 10, 19–21, 66–67, 69–70, 76, 83, 88, 102, 106, 119, 122 Craig, H. K. 48 Creider, C. 47 Croft, W. 45, 51, 57–58 Cronin, H. 24 culture 6–10, 14, 17, 23–24, 27, 38, 41, 51, 53, 59, 63, 65–66, 71, 75, 78, 86, 90, 94–95, 110, 115, 122, 124, 126, 139 currency 7, 93 Dabrowska, E. 52 Danish Functional Linguistics, DFL 43 Darwin, Charles 18–19, 73 Darwinian Evolution 2, 7, 9–10, 15, 23, 34, 36, 38, 65, 67, 80, 86–87, 89–90, 102, 106, 140, 142–143 Data Oriented Parsing 46 Davidson, I. & McGrew, W. C. 5 Dawkins, R. 10, 16, 76–77 Day, M. H. 4, 83, 118 Deacon, T. 68, 116–118 deception 9, 20, 35, 41, 69, 83, 85, 88–89, 91–92, 122, 140 deep Structure 28–31, 37 Denton, D. K. 20 Descartes, R. 73 Dessalles, J. L. 25, 93 Deutscher, G. 96 dexterity 19, 23, 53, 62–64 dialogue 20, 100, 119, 121, 130, 137 diana monkey 8, 59, 136 Dickins, T. E. 134 differentiation 18, 25, 32, 35, 44, 53, 58–59, 71–72, 85, 93, 96, 99–100, 123, 136–137 Discrete Infinity 33 Dixon, R. M. W. 95–96 dolphin 3, 7, 104, 113

23/04/2010 14:13

172

Donald, M. 105, 134 Dubois, Eugene 4 Dunbar, R. I. M. 13, 25, 64, 69, 75, 82, 92, 103, 110, 134, 137–138, 141 Dunning, D. et al. 76 eating 4, 18, 52 Eckert, P. & McConnell-Ginet, S. 19 Edinburgh 15 Edwardes, M. 78 Egas, M. & Riedl, R. 66 elephant 84–85, 107 embodiment 50, 55, 73 emergence 1–2, 10, 35, 44, 86–87, 89, 116, 121, 134, 142–143 empathy 78, 108–111, 113 environment 2, 8, 17–18, 25, 35, 43, 54, 65, 74, 85–86, 88, 101, 103, 106, 110–111, 113–114, 122, 124, 131, 137–140, 142–143 Epley, N. & Whitchurch, E. 135 Erdal, D. & Whiten, A. 88, 139 eusociality 86–87, 89–90, 142–143 Evans, V. 51, 126 Evolang 15 evolution 2, 10, 11, 15–16, 23, 35–36, 56–57, 59, 65, 68, 73, 83, 86, 88, 104, 106, 111, 117, 135, 144 Evolutionary Psychology 56 Extended Standard Theory, EST 29–31 extinction 3 Eysenck, M. W. 49, 53 Fehr, E. & Fischbacher, U. 64, 66, 88, 138 female 19, 61, 67–68, 88–89, 108, 112, 118 fiction 9, 105, 131 Fillmore, C. 28, 50, 57 fish 5, 40 Fitch, W. T. 79, 99, 103, 134, 140

CO22552.indb 172

Index

fitness 2, 9–11, 16, 19, 20–22, 24–25, 56, 66, 70, 76, 84–88, 90, 93, 101–103, 106–107, 135, 137, 140, 143 Flack, J. C., et al. 7–8 Fodor, J. 55–56 Foley, J. & Thompson, L. 119 food 3–8, 67, 72, 85, 95, 109, 112, 131 foraging 7, 35, 67–68, 71 Fossey, D. 108 Fouts, R. with Mills, S. T. 104, 112–113 Fowler, J. H. 66, 88, 122 frame 53 freerider 66, 69 Frey, S. H. 62 Frisch, K. von 8 Frith, U. 120, 142 functional Discourse Grammar, FDG 42 functional linguistics 42–44, 48, 57 Functional Magnetic Resonance Imaging, fMRI 50 Functional/Typological Linguistics, FTL 43 Gallup, G. G. 106–107 game 17, 34, 113, 119 Gardner, A. & West, S. A. 89 Gardner, Allen & Beatrix 112 Gardner, H. 107 Gavrilets, S. & Vose, A. 2 Gaylard, H. L. 48 Geeraerts, D. 23, 49 gender 6, 19–20, 41, 64, 118 gene 10–11, 14, 65, 142 Generative Semantics 28, 38 generativism 23, 27–30, 32–34, 36–39, 42–44, 46–50, 56–57, 94, 97 Genesis 14–16, 33–35, 48, 59, 63, 93, 123–124, 139, 145

23/04/2010 14:13

Index

genetics 6–7, 10–11, 16, 27, 36, 61–62, 65, 67, 69, 70, 76–77, 79, 81, 83–84, 86–87, 96, 107, 115–116, 119, 124, 140, 143 genie 119 genotype 11, 62, 76 Germany 9, 50 gesture 21–23, 25, 62–63, 95, 98, 113, 139 Gibbs, R. W. & Perlman, M. 54 Gibson, K. R. 64 gift 40, 139–140 Gintis, H. 84 Goldstein, M. C. et al. 118 Goleman, D. 107 Goodall, J. 5, 64, 90 Gopnik, A. et al. 80, 119, 121–122, 124 gorilla 67, 108, 112 gossip 92 Gould, S. J. 3, 117 Government & Binding, GB 31–32 grammar 1–2, 6, 9–14, 16–18, 20–33, 35–39, 41- 48, 52–53, 57–60, 66, 71–72, 78, 91–100 102–103, 105–106, 111, 113–116, 123–124, 134–137, 139–141, 143–145 grammaticalization 36, 53, 59, 94–98 Granberg, Å. et al. 81 Gravina, B. et al. 16 gravity 23, 52, 54, 90, 125 Greenfield, S. 75–76, 120 grey parrot 8, 104, 113 Grice, P. 118 grooming 13, 59, 63, 65, 88, 109, 137 group 6–8, 25, 36, 40, 42–43, 48, 65–66, 72, 74, 83–92, 100–102, 106, 108, 111, 113, 118, 137, 139, 140 Halliday, M. A. K. 25, 38, 40–41, 47–48, 123, 140, 143

CO22552.indb 173

173

Hamilton, W. D. 10 Hammer and Anvil 4 hand 18, 21, 35, 41, 53, 62, 64, 103, 138 handedness 23, 53, 56, 62 Handicap Principle 10 Hanks, W. F. 22 Harder, P. 43 Hare, B. & Tomasello, M. 131 Harman, G. H. 40 Harris, R. A. 47 Hauser, M. D. 7, 33, 79, 99, 103, 134, 140 Hawking, S. 125 Healy, J. M. 119 Heine, B. & Kuteva, T. 59, 96 Henderson, I. & Robertson, P. 61 Hengeveld, K. & Mackenzie, J. L. 42 Henrich, J. & Boyd, R. 138 Henshilwood, C. S. 6, 82 Herman, L. M. & Uyeyama, R. K. 104, 113 herring gull 61 Hervey, S. G. J. 42 Hewes, G. W. 3 hierarchy 8, 18, 21, 23–25, 28, 33, 35, 42–46, 58–59, 64–65, 69, 71, 89, 91–92, 99, 110–111, 140 Hirsh-Pasek, K. & Golinkoff, R. M. 48, 123 Hjelmslev 43 Holism 2, 35, 54, 86, 98, 138 Hölldobler, B. & Wilson, E. O. 89, 111, 121 homeotic macromutation 11 hominid 5 hominin 82, 118, 137 Homo 3–4, 61, 63, 75, 82, 104, 107, 121, 131, 137, 142 Homo erectus 4 Homo habilis 4 Homo neanderthalis 3

23/04/2010 14:13

174

Homo sapiens 61, 75, 82, 104, 121, 131 honesty 69, 80, 83–84, 89, 101, 140, 144 Hopkins, W. D. & Pilcher, D. L. 62 Hopper, P. J. & Traugott, E. C. 36, 53, 95–96, 118 Hornstein, N. 32, 128 Hrdy, S. B. 108 Hudson, R. 46 human 1–9, 13–17, 19, 21–22, 24–25, 27–28, 31, 33–34, 36–38, 42, 48, 50, 52, 55–56, 59–60, 62–63, 65–76, 78–85, 90–91, 94, 97, 101–105, 107–114, 116, 118–121, 123–125, 127, 129–131, 133–134, 136, 138–139, 141–143, 145 Humboldt, W. von 31, 73 Hunt, G. R. & Gray, R. D. 5 hunter-gatherer 6–7, 91 hunting 4–8, 22, 63, 91, 105–106, 118, 138 Hurford, J. R. 15, 59, 72, 94 hypotaxis 41 imminence 128–130, 132–133 imperative 136–137, 142 index 122–123 individual 2–3, 6–7, 9–11, 15–16, 18, 23–25, 31, 40, 46, 49, 52, 58, 64–67, 69, 70, 76–77, 83, 85–91, 93, 97, 101, 108, 110–111, 113, 116, 118, 120, 122, 123–124, 129–130, 132, 137–139, 142 information 7, 12–13, 20, 24–25, 35, 38, 40, 44–45, 48, 53, 69, 70, 85, 93–94, 100–102, 104–106, 135–136, 138, 140 innateness 4, 27–28, 34, 48, 53, 55, 84, 109, 115–116, 119, 121–122, 141, 143 input 115

CO22552.indb 174

Index

integrationism 43 intelligence 2–3, 76, 78, 86, 89, 107–108, 122, 124 intention 1, 5, 17, 41–43, 52, 79, 119, 128, 132, 142 intentionality 75, 77, 79–80, 134, 141 internal monologue 12, 121 interpersonal 24, 43, 48, 51–52, 59, 63, 65, 74, 94, 103, 130, 137–138, 140 interrogative 29–30 IQ (Intelligence Quotient) 3, 107 Iverson, J. M. & Goldin-Meadow, S. 22 Jackendoff, R. 11, 29, 33 Jakobson, R. 27 Java 4 Jaynes, J. 75, 78 Johanson, D. & Edgar, B. 82 Johansson, S. 23, 59 Johnson, M. 50–54, 73, 85 Johnson, S. 63 joint enterprise 13, 41, 69 Jolly, C. J. 82 Jordan, J. S. 87 Jordan, R. 87 Jordan, River 102 Kaminski, J. et al. 9, 113 Kanzi the bonobo 48, 110, 113, 136 Kathol, A. 45 Kaye, K. 121 Kendon, A. 22 kin selection 10, 83 Kinsella, A. R. 36 Kirby, S. 21, 59 knapping 16 Knight, C. 17, 71, 84, 92 knowledge 5, 7, 17, 23, 25, 37, 38, 44, 53, 55, 59, 68–72, 75–81, 86, 99, 101–102, 106–108, 110, 121, 124, 129, 135–136, 138–140

23/04/2010 14:13

Index

Koko the gorilla 112 Kövecses, Z. 23 Kyle, J. G. & Woll, B. 21 labelling 13, 96, 98–101, 122 Lakoff, G. 28, 38, 50–52, 73, 85 Laland, K. N. et al. 2, 101 Langacker, R. W. 50, 57 learning 4, 7, 17, 30, 63, 79, 116, 121–123, 132, 142 Lee, D. 128 leg 3–4 Lewis, J. 22, 105 lexicogrammar 40 lexicon 10–12, 28–29, 39, 57, 115 Liebers, D. et al. 61 life 2, 4, 10, 80, 90, 107–108, 119, 121, 123 Linguistic Society of Paris 144 linguistics 14–15, 22–25, 27, 29, 30, 32, 36–44, 48–50, 53, 57–60 lion 17 lithic technology 5, 63 Liti, G. et al. 62 Locke, J. L. 25, 55 Lockwood, D. G. 44 Lycett, S. J. et al. 7 macaque 7, 108 Machiavellianism 74, 78, 86, 89, 91, 99, 105–106, 108, 110, 122, 137–138 MacNeilage, P. F. 115 macromutation 11 male 19, 67–69, 88, 92, 102, 108–109 Malik, K. 79 mammal 11,86 Marino, L. et al. 3, 107 Marks, J. 61 Marshall, A. J. et al. 63 Massey, D. S. 63 Mauss, M. 83, 139 meaning 1, 10, 18–20, 22–23, 29, 32,

CO22552.indb 175

175

35–47, 49, 51, 53–54, 57, 59, 79, 80, 91–93, 95–96, 98–100, 103, 111–113, 132, 134, 136, 141, 143–144 Meares, R. & Sullivan, G. 119 meat 118 Melis, A.P. et al. 131 melodics 41 membership 7, 84, 88, 90, 102 mental 3, 12, 24, 29, 33, 53, 71, 74, 78–80, 97–98, 108, 110, 117, 120, 123 merge 32, 116, 128, 130 message 18, 40–42, 101, 109, 139–140, 143–144 metacognition 12–13, 53, 135 metaconstruct 16, 21 metafunction 40–41, 45, 47–48, 129, 140 metafunction, experiential 41, 45, 52, 115 metafunction, ideational 48, 140 metafunction, interpersonal 24, 43, 48, 51–52, 59, 63, 65, 74, 94, 103, 130, 137–138, 140 metafunction, logical 40, 45, 48, 77, 90, 107, 142 metafunction, textual 41, 48, 140 metaphor 8, 50–52, 54, 57, 75–76, 78–79, 84–85, 87, 89–91, 99, 105, 125–126, 130 metaphor ‘me’ 75–76, 78–79 metonymy 51–52 mexican standoff 10 mimesis 134 mind 17, 22, 31, 37, 49–51, 54–55, 73–74, 84, 97–98, 100, 120, 122, 133 minimalism 32 Minimalist Program 32–33, 36–37, 50 mirror test 106–107 Mithen, S. 22, 117 modality 22

23/04/2010 14:13

176

modelling 9–10, 12–13, 16–17, 22, 28–29, 31–32, 34–35, 43–46, 48, 51, 54, 56–57, 59, 64, 66, 68, 71–72, 74–81, 86–87, 91, 93, 95, 97, 99, 100–102, 105–106, 109–110, 115–116, 118, 121–124, 126, 130–135, 137–143 modesty 7, 139 modularity 22–23, 49, 55–57 money 8, 71–72, 83 monkey 59, 107, 109 morality 7, 87, 111, 118, 144 Morley, G. D. 40 morphology 10, 95 Mosch, S. C. et al. 56 move 14, 29–30, 32, 64, 68, 71, 85, 89, 96, 101, 110–111, 123, 125 Mulder, J. W. F. & Hervey, S. G. J. 42 mutation 143 mythology 7 nature 1, 7, 13–14, 17–18, 20, 22, 25, 27–28, 33, 35–36, 44–45, 47, 52–55, 58–59, 66, 67, 71, 73, 78, 81, 83, 85, 87, 90, 94, 96, 100, 103–104, 107, 111, 113, 119–120, 125, 130, 132–133, 135–136, 138–139, 142–143 Neanderthal 3 negotiation 6, 12, 38, 79, 96, 98–99, 103, 141, 143 nervous system 2, 54 Nettle, D. 2, 96, 102, 117 neuron 3 New Caledonian crow 5, 131 Newmeyer, F. J. 32 Newton, M. 119 niche 2, 4, 8, 101, 104, 110 Nichols, S. & Stich, S. 135 Nietzsche, F. 9 Niles, J. D. 131 Nim Chimpsky the chimpanzee 112 nonhuman 5, 9, 48, 109, 111, 113

CO22552.indb 176

Index

Noonan, M. 43 Noss, A. J. & Hewlett, B. S. 6 noun 13, 18, 21, 28–31, 33, 46–47, 52, 59, 62, 72, 95, 97–98, 133 Núñez, R. E. & Sweetser, E. 53, 126 Odling-Smee, J. & Laland, K. N. 101 offering 13, 28, 43, 59, 68, 100–102 Opie, K. 68 oracle 106 organism 2, 77, 121 origin 15–16, 24, 35, 49, 93, 98–100, 103 orofacial 21, 112 output 2, 49 Palin, Sarah 7 Palmer, G. B. 23, 51 Pan paniscus 63–64, 67 Pan troglodytes 4, 61, 67, 91 Paradis, J. & Crago, M. 127 Paranthropus 82 parataxis 41 Paris 15 Parisi, D. & Schlesinger, M. 116 Parker, A. R. 36 paternity 67–68 patient 30, 98 peacock 18–19 Pepperberg, I. M. 9, 104, 113 person 39–40, 73, 75, 78, 80, 86, 92, 102, 106, 109, 120–121, 139, 142 first 122, 142 fourth 75, 78, 80 second 142 third 69, 78, 81, 87, 93, 103, 132, 142 Peters, A. M. 48 phatics 12–13 phenotype 10–11, 34, 54, 62, 76, 117 phlogiston 37

23/04/2010 14:13

177

Index

phonology 10, 12, 28–30, 42–43, 48, 51, 94–95 phrase 21, 23, 28–30, 33, 41, 46–47, 52–53, 97 physical 1, 3–4, 13, 16, 19, 20–21, 50, 52–55, 62–63, 70, 73, 76, 82, 90, 98, 107–108, 116, 121, 136 Piaget, J. 115–116, 121 Pinker, S. 33–35, 55, 78, 117 Pitchford, I. 87 planning 5, 12, 55, 71–74, 120, 126, 131, 134 play 8, 17–18, 67, 80, 87, 119–120 Plotnik, J. M. et al. 107 Plunkett, K. 117 Point of Event 127 Point of Reference 127 Point of Speech 127 pointing 8, 12–13, 32, 36–37, 55, 59, 75, 87, 91, 95–96, 100, 103, 119, 121, 126–133, 140–141 Pollick, A. S. & de Waal, F. B. M. 63 Popper, K. 72 possessive 23 Postal, P. M. 28 Poverty of Stimulus 115 Povinelli, D. J. 107 pragmatics 25, 43, 48, 51, 94, 115 Prague School 27 pre-grammar 103 prehuman 14, 22, 62, 124 Premack, D. 7, 74, 105, 110, 113, 136 preposition 21, 31, 42 Pressman, S. D. et al. 87 primate 5, 35, 59, 65, 74, 105, 136, 141 Principles and Parameters 30–32 Prior, H. et al. 107 process 4, 6, 11, 13, 16, 19, 24–25, 28–31, 33, 36, 41–45, 53–54, 58–59, 70, 76, 78–80, 83, 94–96, 99, 103, 116–117, 122, 129, 132, 134, 137, 140, 142–143

CO22552.indb 177

pronoun 31, 133 prosodics 41 protolanguage 12, 34–35, 58, 103 proximity 62 Pruetz, J. D. & Bertolani, P. 5 Psittacus erithacus 8 psychology 13–14, 38, 53 Queller, D. C. 89 Radical Construction Grammar 57 Ratchet effect 70 reality 17, 23, 25, 45, 49, 51, 71–72, 90, 116, 126, 128 receiver 9–10, 13, 17, 20–22, 35, 41–42, 44, 46, 53, 57, 59, 70, 80, 83–84, 92, 94, 100–101, 103, 132–133, 136, 139, 140–142, 144 recipient 18, 79, 98 reciprocity 7, 10, 70, 83, 140 recursion 23–24, 33, 58–59, 78–79, 99–101, 103, 132, 134, 141, 143 Reichenbach, H. 125–128, 132–133, 141 Reiss, D. & Marino, L. 107 Relational Grammar 28 relationship 7, 23, 29, 37, 40, 44–45, 52, 54, 57, 61, 69, 73–75, 90, 92, 96, 98–101, 103, 105, 109, 120, 129, 137–144 relevance 3, 89, 93, 95, 107 religiosity 6 Rendell, L. & Whitehead, H. 7 reproduction 2, 9–10, 16, 18–21, 24–25, 54, 64–65, 67–69, 76–77, 84, 87, 104, 108, 124, 137, 140, 142 reputation 69, 81, 83, 92–93, 101–103 Reverse Dominance 88–89, 91, 101, 139

23/04/2010 14:13

178

Revised Extended Standard Theory 29, 32 Richmond, B. G. & Jungers, W. L. 82 Rico the collie dog 113 ritual 7, 24, 72, 92 role 6, 25, 28–29, 33, 36, 41, 52, 78, 80, 87–88, 98, 129, 136, 140 Roth, G. & Dicke, U. 3 Rowling, J. K. 20 rule 18, 21, 31, 42–44, 46, 111, 117, 124, 132, 136 Sabbagh, L. 116 Sampson, G. 94, 115 Sanz, C. et al. 4 Sapir, E. 27, 38–39 Sarah the chimpanzee 7, 113 Saussure, F. de 27 Savage-Rumbaugh, S. 104–105, 110, 112–113, 127, 136 schema 53 Schoenemann, P. T. et al. 117 Sear, R. & Mace, R. 68 Searle, J. R. 25, 84 Segerdahl, P. et al. 74 segmentation 11, 25, 35, 58–59, 71, 85, 99, 136–137 self 1, 7, 14, 54, 73, 75–81, 83, 85–87, 89–90, 93, 100–101, 105–107, 120–122, 125–128, 131–135, 137, 141–142 self awareness 76–79, 122 Selfish Gene 76 selfishness 66, 76–77, 143 semantics 10–12, 28–29, 31, 37–40, 42–43, 45–46, 48, 50–51, 53, 57 sender 10, 13, 17, 20–21, 35, 41–42, 44, 53, 57, 59, 70, 80, 83–84, 94, 101–103, 132–133, 136, 139, 140–142, 144 sensation 71, 74 sense 2, 20, 67, 77, 79, 80, 85, 101, 107–108, 115, 126, 135, 139

CO22552.indb 178

Index

sentence 12, 28, 31, 33, 41, 42, 46–47, 52, 57, 62, 94, 97, 122, 128–129 sex 19–20, 65, 70, 72 sharing 6–8, 12–13, 20, 22, 24, 35, 44, 52, 67, 69, 70, 91, 93–94, 98, 101–103, 106, 110–111, 118, 134, 139–140, 142, 144 Sherman & Austin 113 Shirai, Y. & Miyata, S. 132 Shoshani, J. et al. 3 sign 19, 21, 32, 43, 70, 79, 84, 102, 104, 111–112 sign language 21, 112 signalling 7, 13, 19–22, 25, 28, 35, 41–42, 45, 66, 70, 83–84, 88–89, 101–102, 105, 133, 136–137, 139, 140, 142 Singer, T. et al. 111 skill 4–5, 19, 56, 62, 80, 87, 102, 123 Slocombe, K. E. & Zuberbühler, K. 63 Sloman, A. 71 Smith, P. K. et al. 115, 120 Sober, E. & Wilson, D. S. 111 socialization 10, 12, 26, 35, 63, 72, 79, 83, 85, 87–91, 94, 104, 107–108, 115–117, 119–120, 122, 134, 142, 144 sociology 13, 38 sociopathy 81, 87 songbird 20 sound 9, 19, 21, 43, 45, 95, 97, 102, 123–124 speaker 21, 39, 43, 80, 93, 100, 132, 144 spear 5, 63 species 1,–4, 6–9, 14, 17, 34, 61–64, 67–68, 70, 74–77, 82–83, 87–91, 101, 104–106, 108, 110–111, 113, 116, 123–124, 136–137, 143–144 speech 19, 22, 35, 40, 44, 47, 49, 52, 63, 94, 100, 111–112, 119, 121, 126, 128, 130, 133

23/04/2010 14:13

Index

Spencer, H. 65 spinal column 2 Standard Theory, ST 28–29 Steels, L. 51, 94 Steklis, H. D. & Harnad, S. R. 22 Stewart, I. & Cohen, J. 79 Stochastic Context-Free Grammar 46 stone 4–5, 16, 59, 63, 82 story 2, 13–14, 20, 59, 78, 82, 84, 115–116, 127, 129, 130–131, 139, 141, 144–145 Stout, D. & Chaminade, T. 16 Stratificational-Cognitive Linguistics 44 Stringer, C. 3, 82 structure 15, 27–29, 32, 39, 43, 56–57, 92 subadult 68 subject-verb-object 18, 28, 31, 37, 133, 144 surface structure 28–30, 32, 37 syllable 97 symbol 75 syntactic structures 15, 27–28, 32 syntax 10, 23, 42, 46, 48, 50–51, 57–58, 62–63, 91–92, 97–98, 100, 103–104, 115, 123–124, 132, 136 system 7–10, 12–13, 17, 22–23, 31, 34–36, 38, 41, 46, 54, 57, 59, 61, 64, 66, 71, 83, 91–93, 95, 102, 104, 111, 116, 126, 128, 134, 136, 139, 140 systemic functionalism 38–39, 42, 49, 57, 129 Szathmáry, E. & Számadó, S. 17 tabula rasa 55 Tannen, D. 41, 102 Tanner, J. E. et al. 74, 112 Taylor, P. D. & Day, T. 83, 118 teaching 17, 19, 112, 114, 116, 142

CO22552.indb 179

179

temporality 13, 53, 80, 126, 128–134, 141–143 tense 13, 31, 42, 45, 53, 126–128, 132–133, 141 termite 4 Terrace, H. S. et al. 112–113 Theissen, G. 11 theory of mind 55, 59, 80, 110, 119–120, 122, 135 Theta theory 31 Thomas, J. 25 Thompson, G. 39, 119 Thorndike, E. 107 thought 3, 12, 22–24, 28, 30, 36, 39–40, 49, 54, 80, 99, 116–117, 119, 137 Thulborn, K. R. et al. 50 Tibbetts, E. A. 83, 86 time 6–7, 12, 14, 16–17, 19, 21, 25, 28, 32, 36, 51, 53–54, 63, 66, 68, 70–71, 73, 79–81, 85, 88, 113, 120, 122, 125–130, 132–133, 135, 139, 141, 143 Tomalin, M. 27 Tomasello, M. 17, 50, 59, 73–74, 76, 78, 91, 94, 105, 109, 110, 124, 131–132 tool 4–6, 8, 14, 16–17, 25, 29, 37, 57, 62, 71, 82, 131, 137, 138 Toolan, M. 43 Torr, J. 48 trace 30–31, 63 trait 20, 77, 83, 90, 104, 110, 118–119 transformation 12, 28–30, 32, 39, 47 transitive 30 Traulsen, A. and Nowak, M. A. 119 Trinil island 4 Trivers, R. 10 universal grammar, UG 28, 30–34, 36, 48, 117 universals 7, 16, 27–29, 31, 33–34,

23/04/2010 14:13

180

36–38, 44–47, 50–53, 57–58, 133, 143 Upper Palaeolithic Revolution 17 utterance 10, 20–22, 29–32, 35–37, 40–42, 44–45, 52–53, 57, 69, 78, 92–93, 95, 102–103, 133, 140, 144 Valian, V. 132 value 2–3, 9, 17, 19–21, 29, 31, 55, 68, 71, 76, 83–84, 101–102, 106, 131 verb 13, 18, 28–31, 42, 46, 52, 59, 62, 72, 94, 97–98, 126, 130, 132–133 vertebrate 2 viewpoint 13, 38, 75, 78, 80, 87, 93, 99 vigilant sharing 88, 91, 101, 139 Viki the chimpanzee 112 vocalization 22, 59, 63, 98, 111, 137, 139 voice 63, 100, 133, 142 volition 130 Vugt, M. van & Schaller, M. 83 Vygotsky, L. 116, 121 Waal, F. B. M. de 7, 63, 65, 86, 107, 109 walking 4 war 91 Warneken, F. & Tomasello, M. 109

CO22552.indb 180

Index

Washoe the chimpanzee 112 wealth 3 Weir, A. A. S. et al. 5, 131 Weist, R. M. 132 Wernicke’s area 55 whale 3 Whiten, A. & Byrne, R. W. 108 Whorf, B. L. 27, 38–39 Wickler, W. 84 Wilkinson, G. S. 7 Wilson, R. A. 121 Wolkenten, M. van et al. 64 wood 4, 16, 90 word 1, 7, 17, 21, 23, 29, 40–41, 43, 45–46, 52–53, 57–59, 61, 75, 79, 95–96, 99, 102, 112, 115, 122–124, 127, 136 word grammar 46–47 Worden, R. P. 25 worth 8, 32, 85, 140 Wray, A. 35, 122, 144 X-bar theory 29, 31 Yerkes Institute 48, 110 Zahavi, A. 9, 19, 83 Ziff, P. 38 Zone of Proximal Development 116 zoo 17, 127 Zuberbühler, K. 8, 59, 63, 136

23/04/2010 14:13