New Essays on the Origin of Language [Reprint 2011 ed.] 9783110849080, 9783110170252

Table of contents :
Introduction: New perspectives on an old academic question
1. Biological aspects of the question
On the subcortical bases of the evolution of language
Origin of the human language faculty: the language amoeba hypothesis
2. The first language
The apparent paradox of language evolution: can Universal Grammar be explained by adaptive selection?
Elementary forms of linguistic organisation
From potential to realisation: an episode in the origin of language
Protothought had no logical names
The birth of rules
How language changed the genes: toward an explicit account of the evolution of language
3. Beyond biolinguistics
The narration “instinct”: signalling behaviour, communication, and the selective value of storytelling
Taxonomic controversies in the twentieth century
The origin of origins: a play in five acts, with a prologue im Himmel and an epilogue auf der Erde
References
Index

Citation preview

New Essays on the Origin of Language

W DE G

Trends in Linguistics Studies and Monographs 133

Editors

Werner Winter Walter Bisang

Mouton de Gruyter Berlin · New York

New Essays on the Origin of Language

Edited by

Jürgen Trabant Sean Ward

Mouton de Gruyter Berlin · New York

2001

Mouton de Gruyter (formerly Mouton, The Hague) is a Division of Walter de Gruyter GmbH & Co. KG, Berlin.

© Printed on acid-free paper which falls within the guidelines of the ANSI to ensure permanence and durability.

Library of Congress Cataloging-in-Publication

Data

New essays on the origin of language / edited by Jürgen Trabant, Sean Ward. p. cm. - (Trends in linguistics. Studies and monographs ; 133) Includes bibliographical references and index. ISBN 3-11-017025-6 (cloth : alk. paper) 1. Language and languages—Origin. I. Trabant, Jürgen. II. Ward, Sean, 1963III. Series. P116.N48 2001 401 - d c 2 1

2001034295

Die Deutsche Bibliothek — CI Ρ-Einheitsaufnahme New essays on the origin of language / ed. by Jürgen Trabant ; Sean Ward. - Berlin; New York : Mouton de Gruyter, 2001 (Trends in linguistics : Studies and monographs ; 133) ISBN 3-11-017025-6

© Copyright 2001 by Walter de Gruyter GmbH & Co. KG, 10785 Berlin All rights reserved, including those of translation into foreign languages. No part of this book may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopy, recording or any information storage and retrieval system, without permission in writing from the publisher. Printing & Binding: Hubert & Co, Göttingen. Cover design: Christopher Schneider, Berlin. Printed in Germany.

Contents

Introduction: New perspectives on an old academic question Jürgen Trabant

1

1. Biological aspects of the question On the Lieberman subcortical bases of the evolution of language Philip

21

Origin of the human language faculty: the language amoeba hypothesis Eörs Szathmäry 41 2. The first language The apparent paradox of language evolution: can Universal Grammar be explained by adaptive selection? Manfred Bierwisch 55 Elementary forms of linguistic organisation Wolfgang Klein

81

From potential to realisation: an episode in the origin of language Bernard Comrie

103

Protothought had no logical names James R. Hurford

119

The birth of rules Jean Aitchison

133

How language changed the genes: toward an explicit account of the evolution of language Daniel Dor and Eva Jablonka

149

vi

Contents

3. Beyond biolinguistics The narration "instinct": signalling behaviour, communication, and the selective value of storytelling Volker Heeschen

179

Taxonomic controversies in the twentieth century Merritt Ruhlen

197

The origin of origins: a play in five acts, with a prologue im Himmel and an epilogue auf der Erde Henri Meschonnic

215

References

229

Index

257

New perspectives on an old academic question Jürgen Trabant 1. The origin of language at the Berlin Academy

1.1.

The question of the origin of language was one of the eighteenth century's major philosophical problems. And it is once again, at the turn of this century, one of the most controversial scientific topics thanks to the confluence of linguistics, psychology, biology, and paleoanthropology. Since the New York Academy conference of 1975 (Harnad et al. 1976), numerous books on this subject have been published.1 The renewed interest is noteworthy because linguistics officially forbade the discussion of language origin in 1866 when the Soci0te de linguistique de Paris struck consideration of both universal language and language origin from its agenda. Linguistics was considered a historical science that should not speculate about language's prehistoric beginnings or its post-historic end. Instead, it was supposed to dedicate itself exclusively to the historically documented middle and to faits positifs. Since then, linguistics has largely adhered to the Paris Society's ban. Right up to our times, hardly any prominent linguist has written about language origin, with the characteristic exception of Hugo Schuchardt, who did so in 1919-21 at the Berlin Academy, the institution that has served as the discussion's traditional forum. But the verdict of linguistics did not, of course, affect the life sciences, which, from Charles Darwin onwards, have dramatically increased our knowledge about the natural history of humans and consequently about the natural aspects of language. Today, evolutionary biological and paleoanthropological insights and a more precise understanding of the genetic make-up of man and of the functioning of the brain combine with a linguistics which is no longer exclusively historically oriented. The question of language origin is back on the agenda of the language sciences because quite a few sciences have turned out to be language sciences which previously were not and because linguistics - at least an influential part of it - has itself become a natural science. The linguistic profession's prohibition against the origin problem is obsolete. The new linguistics integrates insights from the other

2

Jürgen Trabant

natural sciences into its own research agenda, thus generating new scenarios of the origin of language and new perspectives on its evolution, scenarios which preceding centuries could not have imagined. But even if the new discussion is clearly initiated by the natural sciences, its development shows a growing interest for and a growing participation of the cultural disciplines and culturally oriented linguists. Human beings belong to both realms, to nature and to culture, and these realms are interwoven in the most intricate ways in every field of human reality, and in language more than in any other. To discuss the origin of language is, hence, necessarily also to discuss how nature and culture come together. This is what the myth of Babel already tried to explain: how cultural differences originate within the universal and natural unity of an adamitic mankind.

1.2.

Ever since its foundation, the Berlin Academy has been the most prominent forum for the European debate about language origin. It is a topic which Gottfried Wilhelm Leibniz himself, the Academy's founder, launched in the Academy's first publication, the Miscellanea Berolinensia of 1710, with his article about the origins of peoples and languages: "Brevis designatio meditationum de originibus gentium ductis potissimum ex indicio linguarum". In it Leibniz considers a primordial human language, a lingua adamica, pieced together from onomatopoeic roots. When sensualist philosophy had made its way into the Berlin Academy - the French philosopher Maupertuis was the Academy's first president - the empiricists argued with their more orthodox Christian colleagues about the origin of language, a subject which had achieved major philosophical significance in the wake of Etienne Bonnot de Condillac's Essai sur I 'origine des connaissances humaines of 1746. The debate not only concerned language, but nothing less than humans' role in the cosmos. Do we, by inventing language, create ourselves or does the Creator give us the gift of language? This is what was at stake in 1769 when the Academy posed its famous prize question about language origin. Johann Gottfried Herder's winning essay, Abhandlung über den Ursprung der Sprache (1772), settled the dispute with a sort of sensualist exegesis of the Bible. Fifty years after Herder, Wilhelm von Humboldt, in his inaugural Academy lecture on the topic of "comparative language studies" (1820), referred to Leibniz's and to Herder's theories on language origin, but gave a transcendental-philosophical twist to the problem which simul-

New perspectives

on an old academic question

3

taneously brought about its scientific end. Because the beginning of language is outside the scope of empirical knowledge, we cannot say anything scientific about it. We do, however, know the "eternal" fountain of language because it is inherent in every word, at all times. Every speech act is the origin. Jakob Grimm's ideas (1851), initiated by Schelling, ignored Humboldt's insight and did not thus differ much from eighteenth-century speculations about the early history of language. Heymann Steinthal (1851), by contrast, reaffirmed the Humboldtean answer to the question that the Academy had raised anew. That is, he closed the debate by interpreting the origin as an eternal source and not as a temporal beginning. In the twentieth century, however, Schuchardt returned to the question of the temporal beginning of language in his essays presented at the Berlin Academy in 1919-21.

1.3.

Since the Berlin Academy is where the discussion on the origin of language was held in the past, the Academy decided, as part of its tricentenary celebrations, to bring the new discussion back to Berlin. It invited researchers who are currently involved in the debate to a colloquium at the Academy in December 1999. This volume presents the results of that discussion. Gordon Hewes was right when he predicted that the conference at the New York Academy of Sciences in 1975 might be a turning point, but was wrong when he continued: "or perhaps it will be another 200 years before any Academy of Sciences thinks the topic is worthy of such treatment" (Hewes 1976: 3). In 1997 the Californian Academy organised a conference on the topic (see Jablonski and Aiello 1998), and it only took 25 years for the debate to arrive in Berlin.

2. Herder on language origin Hans Aarsleff, the historian of the eighteenth-century debate on language origin at the Berlin Academy (see Aarsleff 1974), was correct to criticise the tendency of German linguistic historians of the past to underestimate the discussion's wider European context in the eighteenth century. Herder inserts himself in a European discourse: he discusses Condillac, Rousseau, and English theoreticians. Moreover, the Prussian Academy is an institution which is thoroughly integrated into the French tradition. Not only was

4

Jürgen Trabant

Maupertuis, a follower of Condillac, its president, but also King Frederick II of Prussia - Voltaire's friend and correspondent - preferred to philosophise in French. In the eighteenth century philosophy and science were international affairs. But this does not mean, as Aarsleff's critique of German historiography seems to imply, that Herder's response to the question posed by the Academy in 1769 could therefore not be completely novel. The famous question - in French, of course - was: "En supposant les hommes abandonnes ä leurs facultes naturelles, sont-ils en etat d'inventer le langage? Et par quels moyens parviendront-ils d'eux memes ä cette invention? On demande une hypothese qui explique la chose clairement et qui satisfasse ä toutes les difficultes" (Herder 1772: 138-139). Herder's answer is revolutionary because he inverts the traditional conception of language. His response to the origin question differs fundamentally from those of his predecessors. Condillac, whose Essai was the Berlin competition's point of departure, famously tells the first enlightenment story about the origin of language as part of the history of the Human Mind's ascent from perception to Reason. His tale is a kind of thought experiment in the manner of Psammetichus. He imagines two children living alone in a desert. First they develop the basic intellectual operations that they have in common with the other animals. Then one day the primordial linguistic scene takes place. The first child cannot reach an object he desires and cries out in distress. The second child observes him, perceives both the body movement and the cry, and - moved by pity comes to his assistance. This interaction is repeated, reversed, memorised, habitualised, and finally reproduced by free will. The gesture and the concomitant cry become the first arbitrary (that is, the first voluntarily created) sign. From here, the evolution of human semiosis proceeds. The gesture is transformed into dance, the cry into words. And from here as well, the ascent of the Human Mind toward Reason takes place. Jean-Jacques Rousseau sketches an alternative story. It was not published until after his death (Rousseau 1782) and was therefore unknown to Herder when he wrote his Abhandlung. Rousseau criticises Condillac for presupposing something that must be invented, namely pity. For Rousseau, the human being in its prehuman phase is a loner. Its social sense must first be created. Humans' developing an interest in other humans requires an evolutionary step. It is only at the moment of the apparition of love that language arises: as a love song (which later, in another situation, can then be transformed into the expression of pity).

New perspectives

on an old academic

question

5

Despite their poetic charm, these two stories are traditional in that they conceive of language mainly as a communicative device. They merely continue what common sense always knew and what, more than two millennia before them, Aristotle taught about language in De interpretatione: language is a device for designating "ideas" and for communicating those ideas to other people; language is sound (which differs from country to country); language is a communicative instrument. But Condillac's and Rousseau's origin stories do not reflect what their predecessor, John Locke, had already recognised, what European linguistic theory had gradually realised beginning in the Renaissance, and what Condillac himself in fact also knows, namely: that language is also - or perhaps primarily "thought", cognition. Rousseau is not interested in "thought" or the cognitive nature of language. For him, language is exclusively social and communicative (and passionate). As for Condillac, he tells a communicative story to explain the genesis of a cognitive device. This is precisely what Herder criticises. Language, human language (that is, language-thought), could never have arisen, he argues, from a communicative, passionate, and expressive act. Herder reverses the priorities, conceives of language as primarily cognitive, and tells an origin story that reflects language's cognitive character. The human being creates the first thought via an encounter with the world, namely with a sheep. And the human's representation of the sheep's bleating - this completely interior mental act - is language. Language is cognition. The sound of the voice and communication are added later in a second step after the initial and essentially cognitive event. Communication is something humans share with animals. This explains the Abhandlung's famous opening sentence: "Schon als Tier hat der Mensch Sprache [Already as an animal the human being possesses language]". Humans have animal language, externalised communication, by virtue of their being animals. Human language is essentially internalised language, cognition. Herder is the first Chomskyan in the history of linguistic theory (for more details, see Trabant 2000). After this first radical step, Herder's Abhandlung continues with reflections on the evolution of the first wild language, on the first social uses of language, and on the genesis of linguistic diversity. Man's encounter with the sheep - ovine bleating as the initial (acoustic) event that via the ear's mediation (acroamatically) creates the first thought-word - is the new story.2 It is easily recognisable as a version of a very old story: Adam's naming of the living creatures.

6

Jürgen Trabant

3. The debate and its traditions

3.1. The eighteenth century was of course not the first time that the question of language origin was posed. The question seems to be asked whenever humans ask who they are, which often means that they ask where they come from. The foundational texts of Western culture - the Bible and Ancient Greek philosophy - contain the founding myths about the origin and the first rational discussion of the origin of language. Like all myths, the Biblical story represents a first attempt to understand the world and its mysteries: Adam's naming of the living creatures, the Serpent's seductive conversation with Eve, and the Tower of Babel are "linguistic" stories that try to grasp the origin of language and words, the effects (and the dangers) of speech, and the origin of linguistic diversity. These stories have informed the conception of language for centuries. But they are still stories. In the West's first enlightenment, the Greeks critically examined the certainties of myth. Plato's question about whether language is "natural" or "established by man" is an expression of rational doubt. But religion and myth reigned again in the Christian period. Only when theology and religion loosened their grip on European intellectual life was it possible to take a critical view of religious myths. And it is exactly at such moments that the question of language origin arises.

3.2. Thus for instance around 1300, in a time and a place of nascent enlightenment, the most pious of the Italian poets, Dante, already calls the Biblical myths into question. Nunc quoque investigandum esse existimo cui hominum primum locutio data sit, et quid primitus locutus fuerit, et ad quern, et ubi, et quando, nec non et sub quo ydiomate primiloquium emanavit. [Now I think that we have to investigate to which humans speech was first given, what was said in the beginning as well as to whom, where, when, and, finally, in what language this first speech emanated] (Dante [1304] 1988: 14, my translation).

New perspectives

on an old academic question

7

Let me briefly sketch his answers: Who, quis hominurnt Dante is firm: it must have been Adam and not Eve, despite what we read in Genesis where it was clearly Eve (after the Serpent). What, quid] The first word was El, "God" (the Bible does not say anything about the first word). And since the first word was a whole utterance, it not only had a semantic meaning, but also an illocutionary function. It was an answer, responsio. To whom, ad quem! The addressee was of course God. When, quando? Adam uttered the first word immediately after having been given breath, spiritus, by God: immediately after his in-spiration. Adam's first word is thus his breath, his first ex-spiration in God's presence: Breath, human life, and language are the same. Dante has more to say about the pragmatic status of the first expiration. The first word is a response to God, but its function is not communication. After all, what can Adam tell God that God doesn't already know? What God wants is adoration. Glorifying God: that is what man and language are created for. Where, ubP. This question is not as important to Dante. It was wherever God gave breath to Adam (theologians disagreed about whether it was inside or outside the Garden of Eden). What language, sub quo ydiomate, did Adam speak? Dante is very Chomskyan. He posits a "certa forma locutionis a Deo cum anima prima concreata [a certain form of language which God had created together with the first soul]" (Dante [1304] 1988:24). Language is innate to the mind. But Dante's innate language seems to be more complete than Chomsky's. It is not only syntax plus mentalese. It has all three parts of grammar: lexicon, syntax, and phonology: vocabula, constructio, and prolatio. The first forma locutionis that God created within Adam's soul was of course a concrete language, namely Hebrew. Later, Hebrew was neither lost nor confused at Babel because the sons of Heber were not involved in the tower's construction, the source of language diversity. As we can see, Dante "corrects" and adds missing information to the Biblical narrative. For Dante, the Bible doubtless contains the truth about language origin, but it is still possible to ask questions. And questioning revealed truth is the first step of a free mind (which can still be a pious mind).

3.3.

But the crucial age for the question of language origin was of course the age of scientific inquiry usually referred to as the Enlightenment - and in which we are still living. Now, all truths have to appear before the Court of Reason, that is, of Science. This is why eighteenth-century philosophers construct

8

Jürgen Trabant

their alternative theories of language origin. Condillac proposes that language - a device for designating "ideas" - is created by the necessities of mutual practical interaction (pity). Rousseau replies that love must be invented first and that the first language must have been a love song, an expression of feeling. Herder rebuts both by stating that neither social interaction nor communication could have led to the creation of something that is essentially cognitive. These "theories" remain in the form of narratives and have strong intertextual links to the original Biblical stories. But unlike religious myths, they are not meant to be read as revelation. They are heuristic scenarios and have a completely different epistemic status. They are, as the Berlin Academy explicitly stated, "hypotheses".

3.4.

If now - more than two hundred years after Condillac, Rousseau, and Herder - there is again a passionate interest in the question of language origin, it seems to me it is because a critical reflection on myth is (again) at work. We are implicitly or explicitly creating alternatives to these stories, these "hypotheses", which we often view as rather naive myths. We tend to be proud of what we know today. But however smart we are, however sophisticated our knowledge may be, it seems to me that it would be fitting for us to show the same respect for these stories that the enlightened philosophers showed for the Biblical narratives. Enlightenment philosophers asked the right questions and came up with very intelligent answers on the basis of available knowledge. And this is precisely what modern theorists do. They use their knowledge about the functioning or the history of language, life, and society to generate conjectures about the prehistorical past. Manfred Bierwisch is aware of this methodological characteristic of origin research when he remarks that his reflections on the genesis of language are based on what he knows about the structural properties of normal linguistic expressions. Even if we today know more than eighteenth-century philosophers did, what we do is essentially the same. We construct hypotheses on the basis of our knowledge. Our hypotheses are less narrative than the eighteenthcentury variety. There are (unfortunately) no children in a desert, no lovers dancing around a well, and no bleating sheep. But our hypotheses still retain the form of "scenarios", a term which clearly designates a poetic creation, an imaginative construct. If, as Ernst Mayr says, "scenarios" are important heuristic instruments for modern research in the life sciences, it would be

New perspectives

on an old academic question

9

churlish to deny to the stories of the past the dignity of being scientific instruments (Mayr 1998: 295).

4. New essays on the origin of language

4.1.

The new discussion about the origin of language depends to a large extent on the success of the linguistic theory of Noam Chomsky and consequently on some of the central assumptions of Chomskyan linguistics: that language is an innate human faculty (and as such the result of an "exaptive" step in evolution), that language is mainly syntax, that language is a specific mental capacity independent of general intelligence, and that language has nothing (or very little) to do with communication or "speech". The debate on language origin largely represents a critical examination of these fundamental assumptions. Chomsky's universalistic biological conception of language has forced linguistics (which had previously been a cultural and social science) to examine the findings of the other life sciences of which it had become a part: evolutionary biology, neuroscience, genetics, primatology, paleoanthropology, and so forth. Linguistics is, all of a sudden, surrounded by biology. As a natural science, linguistics must attempt to tackle the problem of the origin of language, which is a crucial aspect of the natural history of man. Some of the most important steps in the evolution of the human being have to do with language (or speech): upright posture (which implies, as Leroi-Gourhan [1964-1965] says, the liberation of the "face" from prehension and the liberation of the hand from locomotion), enhanced brain size, brain lateralisation, the development of the vocal apparatus, and so forth. Becoming a natural science was thus a somewhat risky step for Chomskyan linguistics. Whatever the individual contributors to it might think, this volume as a whole suggests that the Chomskyan school's reductionist view of language is ultimately untenable. For this volume subjects its central assumptions to critical revision from an evolutionary perspective. "Language" is perhaps not only or primarily syntax: lexicon plays a crucial role in its development. "Language" is closer to animal behaviour than the rationalistic abyss would indicate. "Language" and its evolution cannot be separated from general intelligence, communication, or the bodily prerequisites of "speech". 3 Finally, "language" is likely to turn out to be the

10

Jürgen Trabant

result of highly complex adaptive and selective processes, not merely of a single exaptive leap.

4.2. The two papers of the first part of the book deal with the biological problem of selection for language. They show right from the beginning the profound divergence in the understanding of what "language" essentially is and thus represent two fundamentally different approaches to its origin. Philip Lieberman is known as a fierce critic of Chomsky's conception of language as an innate cognitive device and evolutionary "miracle" (see Lieberman 1984, 1991, and 1998). In his paper he adds further arguments to his scenario of a continuous adaptive selection for language as speech. Vocalisation is the centre of his concept of language which is therefore linked to an older part of the brain, namely the subcortical basal ganglia. This proposition roots language much deeper in the history of life than the "normal" cortical localisation. Eörs Szathmäry, by contrast, sides with Chomsky's conception of language as syntax and with Steven Pinker's and Derek Bickerton's account of language origin (see Smith and Szathmäry 1995 and Szathmäry 1996). For Szathmäry, the origin of language is thus primarily the origin of hierarchical syntax. He argues that recent advances in primatological, psycholinguistic, genetic, and neurological research do not make solving the origin problem any easier. Neural activity patterns point to a "language amoeba" in the human brain, a notion that in fact vastly complicates the possible evolutionary processes involved in creating this "simple" faculty.

4.3. The second part of the book contains papers that deal primarily - but not exclusively - with the formation of the first language. What evolutionary steps are necessary to form a fully fledged language? What is the relation between lexicon and syntax - is there an automatism of an innate universal grammar at work? What are the communicative constraints on these processes? Manfred Bierwisch starts with an account of the origins problem from a Chomskyan perspective: the language faculty is the result not of adaptive

New perspectives

on an old academic question

11

selection (of a cognitive or a communicative behaviour), but of exaptive evolution. In other words, it does not offer a selective advantage (see also Bierwisch 1994). On this genetic basis, first a lexicon is formed, after which the combination of these lexical items must develop. Wolfgang Klein shares these theoretical assumptions. He distinguishes between the language faculty, the language system, and communication or speech (communication for Bierwisch and Chomsky is only a pleasant bonus). Klein treats the language faculty as a given and devotes his essay to the evolution of the language system. He provides additional evidence for what the second step, the creation of syntax, might have looked like. Like Bickerton, who used the genesis of grammar in Creole languages as a proof for the innate language faculty, Klein imports observations from second language acquisition research. Like Creoles, the varieties of "natural" second language learners display universal structural features in whatever languages are learned. He calls these the Basic Variety. The Basic Variety's universal features are candidates for proving the underlying universal language faculty and might be traits of the first language. This is the place to mention Ray Jackendoff's paper, which was given at the conference, but which has already been published elsewhere (JackendofF 1999). This already very influential paper sketches something like a plausible "normal" evolutionary scenario on the basis of our current knowledge about language. Bernard Comrie asks what makes Dumbo start to fly? That is, what makes an innate disposition turn into concrete activity? What is it that "ignites" language, what makes syntax go? He discusses the problem by giving a critical account of the famous cases of language-impaired humans and of the creation of "new" languages which have been used as evidence for the existence of a language "instinct" independent of general intelligence: 4 Genie, Ildefonso, Creole grammar, and Nicaraguan sign language. His rather astonishing conclusion is that lexicon makes Dumbo fly; lexicon makes language go. The available evidence clearly shows that the crucial prerequisite for the creation of a fully fledged language is the input of a lexicon (at the right developmental moment) and that from then on grammar can be created anew within a speech-group of the right size. James Hurford approaches the same problem by inquiring into "protothought". He studies Dumbo's ears before Dumbo uses them to fly. But what he actually discovers is that syntax does not depend merely on representational protocognitive mental activities, but that "protosyntax" requires communicative activity where topic and comment are involved.

12

Jürgen Trabant

This conclusion is similar to Jean Aitchison's, whose review of the most important problems of the language evolution process culminates in a scenario for the creation of syntax through informational constraints. As we know from her seminal book on language origin (Aitchison 1996), Aitchison does not believe in the cognitive origin scenario, but conceives of language as a profoundly social event. She assumes a slow process of habitualisation. Words are created first, followed by a categorial shaping of the words. Finally, order is established according to informational weight (newsworthiness). The communicative aspects of the evolution of grammar sketched by Comrie, Hurford, and Aitchison are underscored by Daniel Dor and Eva Jablonka, who set out to demonstrate that grammar is not independent of meaning, that is, of a language-specific communicational device. Dor and Jablonka refuse to choose between communication and cognition. They attempt to show that language evolution is the evolution of conceptual structures with the aid of more complex communicative tools. Language, the conceptual system, and motor control evolve together in a sort of spiral. Bridging the gap between communication and cognition, the authors mediate as well between culture and genetics (see also Jablonka and Rechav 1996). What all the papers in this section have in common is an emphasis on syntax - the Chomskyan heritage - even if it is criticised. Yet even though lexicon plays a crucial role in some of the contributions, the creation of the lexicon itself is more presupposed than tackled directly. The emphasis on syntax as the centre of language is of course the profound difference to eighteenth-century approaches. But if, as Comrie shows, the growth of a large lexicon is the condition for the launch of syntax, and if Hurford's concern for unities of "protothought" is correct, then the creation of "first words" should attract further attention. Interestingly enough, Terrence Deacon, who is not a linguist and therefore unburdened by linguistic prejudices (and who was also at the Berlin conference), concentrated in his 1997 book on the creation of a "word" as the first step toward language. It is not unlikely that the invention of words will be the next focus of the language origin discussion (see Martin 1998).

4.4.

The third part of the book consists of three articles that in one way or another do not share the concerns of the biolinguistic mainstream. Volker Heeschen's horizon is ethnolinguistic, Merritt Ruhlen is a historical com-

New perspectives

on an old academic question

13

parative linguist, and Henri Meschonnic's approach is a very "culturalist" historical anthropology of language. Acknowledging that the construction of reality (naming, mapping the world) - cognition more broadly defined than in "cognitive" linguistics - is the first function of language and conceding that communication might only be a side effect of language, language use in small traditional societies sheds a completely new light on the priorities. Much of the direct speech that we think to be the essence of linguistic behaviour might just be a consequence of the social organisation of large modern societies. According to Heeschen, traditional societies often make only indirect use of language or use language for non-communicative purposes. But these aesthetic uses of language - play, embellishment, the poetic creation of alternative realities are not simply "parasitical" (Austin), but rather are vital necessities in such societies. Heeschen's reflections, based on his experiences in New Guinea, show how much our theoretical assumptions about language depend on our own linguistic experiences in modern Western societies, assumptions that might be subject to revision in the light of different cultural experiences. Merritt Ruhlen addresses some of the theoretical issues of his wellknown provocative reconstruction of the first language (Ruhlen 1994d). If it is true that modern humans all descended from ancestors who wandered out of Africa some 100,000 years ago, then why not assume that they spoke one language - Proto-Sapiens - and that traces of this Mother Tongue can still be found in modern languages? Traditional diachronic linguistics refrains from going further back than the well-established language families like Indo-European allow, that is, about 6,000 years. But since we now have information on all the world's languages, why not go further? Ruhlen's bold reconstructive efforts represent, by the way, a realisation of what Leibniz, the founder of historical comparative linguistics, started in his first academic text for the Berlin Academy, the "Brevis designatio". Leibniz was convinced that there had been one first language - lingua antiqua - and that extant languages contain its traces, which he attempted to find. The volume ends on a very sceptical note with the essay of Henri Meschonnic, the French linguist and poet. For him (a little bit like the Sociitä de linguistique de Paris), language is essentially historical. And so because the origin question is ahistorical, it cannot be asked - let alone answered - scientifically (see also Meschonnic 1996). Meschonnic's paper also points to some ideological questions that have traditionally been linked to the origin question. Whether or not we share Meschonnic's doubts, they lead us to the question of why the question is asked and why it is asked now,

14

Jürgen Trabant

and hence to a reflection of further motivations for the origin question outside its narrow scientific context.

4.5.

To understand our own motives, it might be useful to look back again at the eighteenth century. At stake in the language origin discussion was the very notion of Man, especially of how much was his own creation and how much was God-given or, as the philosophes preferred to say, given by Nature. The typical enlightened, optimistic answer was that human beings played the decisive role in the creation of language. Today's situation displays a number of parallels. The sciences - biology, genetics, the neurosciences, paleoanthropology - have so dramatically changed the place of the human being in the cosmos that the question has returned. How much is given by Nature, how much do we create by our own cultural activities? The modern answers seem to be more sceptical regarding human autocreation. Another issue raised by Meschonnic in his sceptical remarks is whether, beyond these scientific developments, there are other, more general historical conditions for the timeliness of the question, mainly whether there are political reasons for it. Perhaps. In the eighteenth century, Europe discovered the depth of historical and cultural differences between human beings and tried to hold things together by means of universalistic assumptions (like Reason, Human Nature, and Human Rights). Is it possible that after two centuries of emphasising differences - and of the sometimes deadly political consequences of this emphasis - universalistic assumptions are today politically important for mankind's survival?

5. Convergences and differences

5.1.

Even if the aim of the conference at the Berlin Academy was to discuss recent developments in the area of language origin, as a historian of ideas I could not resist the temptation, in this introduction, to make some connections to the past. I think that not only the old institution can learn something

New perspectives

on an old academic question

15

new, but also that the modern debate can learn from the past. For instance, it is fascinating - and rather enlightening - that, notwithstanding the differences, the opposition between Condillac and Herder is structurally still a primary point of contention in the current discussion: is communication or cognition the very essence of language and hence the nucleus of its origin? And when Merritt Ruhlen reconstructs the oldest language of mankind, the Ursprache, he is realising a research project proposed by Leibniz in 1710. On a larger scale, the discussion goes back to the founding myths of our civilisation. Science continues the incipient rationalisation of myth. Just as philosophy is - according to Whitehead's wonderful remark - footnotes to Plato, the origins debate is, of course, footnotes to the Bible: to Adam's naming of the living creatures, to Eve's seductive first speech act, and to the diversification of language at Babel. Dante's systematic approach to Genesis is already an example for how human intelligence struggles with myth. The Enlightenment wants to get rid of the Biblical myth, but all enlightened origins stories have very clear intertextual relationships to the Bible. Modern science as a whole continues this process of emancipation from myth. But it seems to me that here again science is not so far away from the storytelling of the past: in their scenarios, the authors of this volume use their creative imagination as good storytellers. And, as a whole, our volume reflects a traditional compositional scheme. Structurally, it follows Condillac's and Herder's essays (or even Dante's fourteenth-century treatise), which start by telling the story of the very first moment of "language", the Promethean fire, the sparking of the first "word". Then they ask what the first "wild" language might have looked like. For instance, what part of speech was first: verbs, nouns, or adjectives? And finally they turn their attention to the historical development of languages. Similarly, the first section of this book also addresses the genesis of the language faculty, the second part deals with the first language and its probable evolution, and the third section treats questions of a more cultural and historical nature.

5.2. Notwithstanding the connections and structural convergences with the past, the way of doing things at the Berlin Academy in the year 2000 is marked by three major differences to the ways of the eighteenth century.

16

Jürgen Trabant

First, instead of an essay contest, modern academies organise conferences. This has to do with the possibilities offered by modern means of transportation and with a difference in the style of dealing with scientific issues: we believe in discussion and debate as important means of advancing knowledge. This is connected to a second difference. The old Academy believed in the possibility of finding a definitive answer: "On demande une hypothese qui explique la chose clairement et qui satisfasse ä toutes les difficultes". Today we lack the eighteenth century's optimism. We gathered in Berlin certain that we would not solve all the difficulties. But we hoped to make at least some progress in identifying them. Third, we use a different language. In 1769 the question was asked in French, and the contestants could answer in French, German, or Latin. Herder answered in German, which, by the beginning of the nineteenth century, became the Academy's official language. The language of our conference was English. This is quite a remarkable change, especially at this Academy, because the other linguistic topic the Prussian Academy was famous for was its question about the universality of French in 1784. Antoine de Rivarol won the prize with his "Discours sur Γ universal ite de la langue frangaise". Rivarol was sure that the splendour of French culture and literature as well as the language's structural properties predestined French to be the universal language of the future: the Latin of modern times. Moreover, he was convinced that French had won the geopolitical struggle with English, the more so since the American colonies had just won their independence from England with French help. But we know today that this victory is exactly the reason why this book is written in English. These three differences between our debate and the eighteenth-century debate have resulted not in one prize-winning Abhandlung über den Ursprung der Sprache, but in a plurality of studies: New Essays on the Origin of Language.

Notes 1. See e.g. Aitchison 1996, Beaken 1996, Bickerton 1981 and 1990, Chomsky 1986, Danesi 1993, Deacon 1997, Gajdusek et al. 1994, Gessinger and von Rahden 1989, Hewes 1975 and 1977, Hurford et al. 1998, Jablonski and Aiello 1998, Lieberman 1984, 1991, and 1998, Lyons 1988, Maynard-Smith and Szathmäry 1995, Pinker 1994, Ruhlen 1994d, Trabant 1996, and Zimmer 1986.

New perspectives

on an old academic question

17

2. That the ear - and not the eye or the hand - is the sensory organ at the origin of human beings' cognitive evolution is another aspect of Herder's radical anthropological revision of classical European epistemology. See Trabant 1990. 3. "[W]e can go beyond the idealization, perhaps appropriate for 1965 but not for now, of the language faculty as a sui generis mental phenomenon, unrelated to general cognition" (Jackendoff 1999: 279). 4. See Pinker 1994. The most famous of these cases, the "Gopnik case", seems to have been dropped in the research debate because it does not offer reliable evidence.

1. Biological aspects of the question

On the subcortical bases of the evolution of language Philip

Lieberman

1. Introduction Although most studies on the evolution of the neural bases of human language no longer overtly accept the tenets of phrenology, they implicitly accept the proposition that particular regions of the brain constitute the "seats" of language, thinking, memory, and so on. And since the very beginnings of neuroscience the neural bases of human language and cognition have focussed on the neocortex. In the past forty years, linguists adhering to Noam Chomsky's theories have equated language with syntax, hypothetically specified by an innate, cortical organ, the "Universal Grammar". I shall attempt to shift the focus. I shall show that speech is perhaps the central "unique" aspect of human linguistic ability which also includes lexical and syntactic components, and that all of these elements are learned skills, based on a distributed neural system involving much of the brain, the "Functional Language System" (FLS). Neither the anatomy nor the physiology of the FLS can presently be specified with certainty. However, evidence from many behavioural and neurobiological studies suggests that the neural basis of human language is a distributed network that crucially involves subcortical structures, the basal ganglia. The evolution of functional basal ganglia can be traced back in time to animals similar to present-day frogs (Marin 1999) and their role in the acquisition of learned complex communicative behaviour appears to be similar in other mammalian species as well as in birds (Brainard and Doupe 2000). The evolution of human language, in this light, can be traced back hundreds of millions of years. However, the particular neural network that regulates human language, the FLS, appears to be a unique, "derived" human characteristic, differentiating humans from closely related species. Like other distributed neural systems that regulate complex behaviour, FLS architecture consists of circuits linking segregated populations of neurons in structures distributed throughout the brain, cortical and subcortical, including the traditional "language" areas (Broca's and Wernicke's areas) as well as other parts of the neocortex. The FLS rapidly integrates sensory information with stored knowledge; it is a dynamic system, enlisting additional neural resources in response to task difficulty. Regions of the frontal lobes of the human

22

Philip

Lieberman

neocortex, implicated in abstract reasoning and planning, and other cortical areas are recruited in the FLS as task difficulty increases. The FLS also provides direct access to the information coded in a word, that is, primary auditory, visual, pragmatic, and motor information. The mental operations carried out in the brain are not compartmentalised in fixed modules and the neural bases of human language are intertwined with other aspects of cognition, motor control, and emotion. No other living species possesses the neural capacity to command spoken language, which serves as a medium both for communication and thought. However, while the human FLS is unique, it appears to have evolved from neural structures and systems that regulate adaptive motor behaviour in other animals. In this light, the subcortical basal ganglia structures usually associated with motor control that are key elements of the FLS reflect its evolutionary history. Natural selection operated on neural mechanisms that yield adaptive - that is, cognitive - motor responses in other species. Given the evolution of the FLS, which as we shall see is reflected in its anatomy and physiology, there is no reason to believe that the basic operations of the human brain differ for motor control and language. Although the neural architecture that regulates motor control and syntax is part of our innate endowment, the details are learned. And the early stages of the evolution of the cortico-striatal neural circuits that regulate human language and thought may have been shaped by Natural Selection to meet the demands of upright bipedal locomotion, the first defining feature of hominid evolution.

2. Functional neural systems The traditional view of the brain bases of language derives from nineteenthcentury phrenology. Phrenological studies claimed that discrete parts of the brain, which could be discerned by examining a person's cranium, were the seats of various aspects of behaviour or character. Neophrenological theorists such as Fodor (1983) do not claim that a bump on your skull shows that you are virtuous, but phrenology lives on. The traditional Broca-Wernicke model of the neural bases of language is in essence phrenological. Paul Broca (1861) ascribed the word-finding difficulties and speech-production deficits of his patient to damage to a frontal region of the neocortex, "Broca's area". In 1874 "receptive" deficits involving comprehension were ascribed to damage to a posterior area of the cortex, "Wernicke's area". Lichtheim's 1885 model, which was adopted without change by Geschwind in 1970,

On the subcortical bases of the evolution of language

23

claimed that the neural basis of human language was a cortical system linking Wernicke's area with Broca's area. According to this model, Wernicke's area processes incoming speech signals; information is then transmitted via a cortical pathway to Broca's area which serves as the "expressive" language output device. The Broca-Wernicke model is taken by linguists such as Chomsky (1986) and Pinker (1994) to be a valid description of the neural architecture underlying human linguistic ability. According to Pinker, "Genuine language ... is seated in the cerebral cortex, primarily the left perisylvian region", identifying "the human language areas..., Wernicke's and Broca's areas and a band of fibres connecting the two" (Pinker 1994: 334, 350). Although the traditional model has the virtue of being simple, neurophysiology studies show that it is wrong. It is apparent that particular regions of the neocortex are specialised to process particular stimuli, visual or auditory, other regions participate in regulating motor control or emotion or holding information in short-term (working) memory, and so forth. But complex behaviours like reaching for an object or talking are regulated by neural circuits that constitute distributed networks that link activity in many different neuroanatomical structures. As Mesulam notes, "complex behaviour is mapped at the level of multifocal neural systems rather than specific anatomical sites, giving rise to brain-behaviour relationships that are both localised and distributed" (Mesulam 1990: 598). A given part of the brain typically supports many segregated neuronal populations that project to different parts of the brain forming circuits that regulate different aspects of behaviour. In other words, although specific operations are performed in particular neuroanatomical structures, overt behaviours are regulated by a network, "functional neural systems", that integrates activity in structures in many regions of the brain. Studies that relate brain activity to behaviour in humans and other species show that a class of functional neural systems exists which generate rapid responses to environmental challenges and opportunities. These neural systems integrate incoming sensory information with stored knowledge and modify or generate goal-directed motor activity to enhance biological fitness. The postulated human FLS regulates language, which sets human beings apart from other living species. The role of subcortical basal ganglia in the FLS is crucial. Basal ganglia structures form part of neural networks that regulate sequencing in seemingly unrelated activities like moving your fingers (Cunnington et al. 1995), talking (Lieberman et al. 1992; Pickett et al. 1998), comprehending distinctions in

24

Philip

Lieberman

meaning conveyed by syntax (Lieberman et al. 1990, 1992; Grossman et al. 1991, 1993; Natsopoulos et al. 1993), and solving cognitive problems (Lange et al. 1992; Pickett et al. 1998). Basal ganglia sequence the pattern generators governing motor activity as well as cognitive operations, by means of segregated neuronal populations in distributed neural networks that project to neuronal populations in other subcortical structures and cortical areas throughout the brain.

3. The FLS and basal ganglia: cortico-striatal circuits The evidence for basal ganglia activity in the FLS is discussed in Lieberman (2000) in some detail. Two sources of evidence are relevant. Since only human beings possess spoken language and complex cognitive behaviour, it is not possible to employ highly invasive techniques that might reveal the neural circuitry of the FLS or the "computations" that are effected in its component neuroanatomical structures. However, our physiology is manifestly similar to that of other species, so valid inferences concerning the human brain can be derived from the study of the brains of other species. Comparative neurophysiologic studies of other species have revealed many aspects of basal ganglia circuitry and function. In some instances, comparable data of human brains using advanced imaging techniques are available. Neurophysiologic studies clearly show that neural circuits link basal ganglia structures and the cerebellum to prefrontal cortical areas implicated in cognition as well as to cortical areas associated with motor control. Experiments-in-nature dating back to Broca's time constitute the second line of enquiry. Studies of the behavioural effects of brain damage resulting from trauma or disease provide indisputable evidence for the role of subcortical FLS structures. They demonstrate that subcortical structures are essential components of the FLS. Language usually recovers after humans suffer cortical damage, perhaps reflecting cortical plasticity (Elman et al. 1996). However, damage to subcortical circuits results in permanent language deficits (Stuss and Benson 1986). Speech, lexical access, the comprehension of meaning conveyed by sentences, and various aspects of "higher" cognition are regulated by parallel circuits that involve basal ganglia and other subcortical structures as well as other neocortical structures.

On the subcortical

bases of the evolution

of language

25

One of the major findings of clinical studies since the 1980s is that behavioural changes once attributed to frontal lobe cortical dysfunction can be observed in patients having lesions in subcortical basal ganglia. Cummings in his 1993 review article identifies five parallel basal ganglia circuits of the human brain: a motor circuit originating in the supplementary motor area, an oculomotor circuit with origins in the frontal eye fields, and three circuits originating in prefrontal cortex (dorsolateral prefrontal cortex, lateral orbital cortex and anterior cingulate cortex). The prototypical structure of all circuits is an origin in the frontal lobes, projection to striatal structures (caudate, putamen, and ventral striatum), connections from striatum to globus pallidus and substantia nigra, projections from these two structures to specific thalamic nuclei, and a final link back to the frontal lobe (Cummings 1993: 873).

4. Representative experimental data: the syntax of rat grooming Most linguists believe that the defining characteristic of human linguistic ability is syntax, which links words together into well-formed sentences that can convey an unbounded set of meanings. Studies of rodents show that they too make use of a "syntax", regulated in basal ganglia, to bind individual word-like submovements into well-formed grooming programs. The grooming movements of rats do not convey an unbounded set of meanings - or perhaps any meaning at all. However, experiments show that damage to the striatum of rats disrupts the integrity of the sequences of gestures that normally occur, but does not disrupt the individual gestures that would make up a grooming sequence. In other words, the "syntax" of rat grooming is regulated in the basal ganglia. Damage to other neural structures - prefrontal cortex, primary or secondary motor cortical areas, or cerebellum does not affect the grooming sequence. Electrophysiologic data (Aldridge et al. 1993) show that neuronal activity in basal ganglia regulates the sequence in which these individual movements (forelimb strokes, body licks) occur.

26

Philip Lieberman

4.1. Learned

behaviour

Rats raised in isolation from other rats execute the same grooming patterns as rats raised in a normal rat environment. Therefore, rodent grooming patterns appear to be coded by a genetically transmitted Universal Grooming Grammar analogous to the Chomskyan Universal Grammar that supposedly determines the syntax of all human languages. Although many linguists accept some form of Chomsky's theory, virtually all developmental studies suggest that human beings learn the particular syntax of the languages that they command by means of cognitive processes and neural mechanisms similar in manner and kind to those employed in learning to play a violin or tennis or to walk (reviews of these studies can be found in Elman et al. 1996; Lieberman 1991, 2000). Neurobiological studies of monkeys and birds reveal the role of basal ganglia in the acquisition of learned behaviour. The basal ganglia circuits that regulate learned motor tasks in monkeys clearly are shaped by associative processes. Kimura et al. (1993) studied the responses of striatal interneurons as monkeys learned a classic Pavlovian conditioned motor task - the monkeys heard a sound that preceded a task for which they were rewarded. The data showed that neuronal populations in basal ganglia "coded" the learned response. The independent studies of Graybiel and her colleagues (1994) confirm these results. Dopamine-sensitive striatal interneurons (neurons that connect to other neurons within the striatum) responded contingent on reward. The striatal architecture noted by Graybiel et al. (1994) is similar to computer-implemented models of distributed neural networks (Elman et al. 1996) that carry out both associative Hebbian learning and supervised learning. Other independent studies in which monkeys learned tasks when they were rewarded with fruit juice confirm the role of reward-based, "appetitive", activation of midbrain dopamine-sensitive neurons. Interruption of a basal ganglia-forebrain circuit in songbirds results in the birds no longer being able to learn new songs or modify previously learned songs (Brainard and Doupe 2000).

4.2. Human finger

sequencing

Experiments that indirectly monitor human basal-ganglia activity yield data consistent with these studies of other species. Human basal-ganglia circuits regulate sequential, self-paced, manual motor control tasks. Depleted production of the neurotransmitter dopamine degrades basal ganglia activity

On the subcortical

bases of the evolution

of language

27

in Parkinson's Disease (PD), largely sparing cortex (Jellinger 1990). Studies of PD consistently show abnormalities in motor sequencing (Harrington and Haaland 1991). Cunnington and his colleagues (1995) monitored the activity of the supplementary motor area of the cortex in both PD and normal subjects, by means of techniques that used external EEG electrodes to record movement-related potentials: electrical signals in the brain that are emitted before a movement. Subjects pushed buttons with their index fingers in various experimental conditions; signals were recorded from the supplementary motor area before and during each button-push. The buttonpushing data reveal basal ganglia activity similar to that noted by Aldridge and his colleagues for rats as well as studies of spatial sequencing in monkeys that make use of invasive techniques. The basal ganglia: activate the preparatory phase for the next submovement, thereby switching between components of a motor sequence. Since the basal ganglia and supplementary motor area are more involved in temporal rather than spatial aspects of serial movement, this internal cueing mechanism would coordinate the switch between motor components at the appropriate time, thus controlling the timing of submovement initiation (Cunnington et al. 1995: 948).

4.3. Stereotaxic

surgery

Studies of the results of surgery offer another source of data concerning basal ganglia function in human subjects. Stereotaxic surgical techniques, in use for many years, selectively destroy basal ganglia structures such as the internal and external segments of the globus pallidus or the targets of circuits from these structures in the thalamus. Thousands of operations were performed before Levadopa treatment was available to treat PD by offsetting the dopamine depletion that is its immediate cause. These operations often reduced the debilitating rigidity and tremor of PD patients. Marsden and Obeso (1994) review the outcomes of these surgical interventions and similar experimental lesions in monkeys. The seeming paradox is that surgery that destroys subcortical structures, known to regulate various aspects of motor control, has little effect on motor control though it reduces tremor and rigidity. The answer appears to be the distributed parallel nature of the basal ganglia system regulating motor control. Marsden and Obeso note that:

28

Philip Lieberman neurons in supplementary motor area, motor cortex, putamen and pallidum, all exhibit very similar firing characteristics in relation to movement. For example, populations of neurons in each of these regions appear to code for the direction of limb movement ... and may alter their discharge in preparation for the next movement.... Within each of these various motor areas, neuronal populations seem to be active more or less simultaneously, rather than sequentially. They appear to cooperate in an overall distributed system controlling the shape of movement (Marsden and Obeso 1994: 886).

Marsden and Obeso conclude that the basal ganglia have two different motor control functions in human beings. First, their normal routine activity may promote automatic execution of routine movement by facilitating the desired cortically driven movements and suppressing unwanted muscular activity. Secondly, they may be called into play to interrupt or alter such ongoing action in novel circumstances.... Most of the time they allow and help cortically determined movements to run smoothly. But on occasions, in special contexts, they respond to unusual circumstances to reorder the cortical control of movement (Marsden and Obeso 1994: 889).

But, as many studies show (see Lieberman 2000), the basal ganglia circuitry implicated in motor control does not radically differ from that implicated in cognition: [T]he role of the basal ganglia in controlling movement must give insight into their other functions, particularly if thought is mental movement without motion. Perhaps the basal ganglia are an elaborate machine, within the overall frontal lobe distributed system, that allows routine thought and action, but which responds to new circumstances to allows a change in direction of ideas and movement. Loss of basal ganglia contribution, such as in Parkinson's disease, thus would lead to inflexibility of mental and motor response... (Marsden and Obeso 1994: 893).

4.4.

Aphasia

These conclusions are supported by the studies of aphasia that have structured theories of mind and brain since Broca's time. Although the most apparent linguistic deficit of the syndrome named for Broca, "Broca's aphasia", is laboured, slow, slurred speech, other disruptions of normal behaviour can occur such as deficits in fine manual motor control and oral apraxia (Stuss and Benson 1986). Broca's aphasics often also have difficulty exe-

On the subcortical bases of the evolution of language

29

cuting either oral or manual motor sequences (Kimura 1993). "Higherlevel" linguistic and cognitive deficits also occur. The utterances produced by Broca's aphasics were traditionally described as "telegraphic". When telegrams were a means of communication, the sender paid by the word, and "unnecessary" words were eliminated. Hence the utterances of Englishspeaking aphasics who produce messages like "man eat fish" have the appearance of telegrams. Aphasie telegraphic utterances were thought to be a compensatory behaviour. Aphasie speakers presumably produced short utterances to minimise their speech production difficulties. However, it is evident that Broca's aphasics also have difficulty comprehending distinctions in meaning conveyed by moderately complex syntax. Although agrammatic aphasics are able to judge whether sentences are grammatical with high error rates, these comprehension deficits have been replicated in many independent studies (see Blumstein 1995). Cognitive deficits also occur; Goldstein noted the loss of the "abstract capacity, deficits in planning, and deriving abstract criteria", and "executive capacity generally associated with frontal lobe activity" (Goldstein 1948: 6). Acoustic analyses show that the characteristic speech production deficit of Broca's syndrome involves impaired sequencing of motor commands. It first is useful to note the aspects of speech production that are intact in Broca's aphasia. The production of the formant frequency patterns that specify vowels and consonants is unimpaired, though there is increased variability (Ryalls 1986). Formant frequency patterns are determined by the configuration of the supralaryngeal vocal tract - primarily tongue and lip activity. We can thus conclude that the control of these structures is unimpaired. The "encoding" or "melding" of formant frequency patterns (Liberman et al. 1967) that characterises the production of human speech is likewise preserved in Broca's aphasia.

5. Speech encoding A short digression on the nature and selective advantage of speech encoding is perhaps germane. Speech encoding is one of the keys to human linguistic ability. It allows us to communicate rapidly, transcending the limits of the human auditory system and the bounds of short-term auditory memory. Formant frequencies are determined solely by the shape and length of the supralaryngeal vocal tract. As a person moves his or her tongue, lips, velum and larynx position, the shape of the vocal tract gradually changes, and so

30

Philip Lieberman

do the formant frequencies. The result is an acoustic melding of the formant patterns that specify "individual" sounds or phonemes into syllable-sized units (Liberman et al. 1967). For example, it is impossible to produce the isolated sound [t] without also producing a vowel or "continuant" such as [ta] or [ts]. The changing formant frequency pattern specifies the initial consonant as well as the vowel: consonant and vowel are fused into a syllable. This process yields the high information transfer rate of human speech because the encoded syllables are transmitted at a rate that does not exceed the fusion frequency of the auditory system. Listeners then resolve the encoded syllables into the phonetic code. The encoding process results in a transmission rate of 20 to 30 phonetic units per second, which exceeds the fusion frequency of the human auditory system. If we were forced to communicate at the slow syllabic rate we generally would forget the beginning of a complex sentence before we heard its end. The encoding process involves two factors. Inertial left-to-right, "coarticulation" effects inherently encode the formant frequency pattern. For example, when producing the sound [t] of the syllable [ta], the tongue blade must initially be in contact with the palate (the roof of the mouth). The tongue cannot instantly move away from the palate to the lower position necessary for [a]. Consequently, the vocal tract shape gradually changes causing the formant frequency pattern to gradually change as the tongue moves from its syllable-initial position to the [a] position. A similar effect holds for the syllable [tu] except that the consonant "transition" moves into the formant frequency pattern of the vowel [u]. However, inertia cannot account for the "anticipatory" coarticulation typical of speech. Human speakers always plan ahead as they talk, anticipating sounds that will occur. Speakers, for example, "round" their lips (move their lips forward and towards each other) at the very start of the syllable [tu], anticipating the vowel [u]. They don't do this when they produce [ti] because the vowel [i] must be produced without lip-rounding. (It is easy to see this effect if you look into a mirror and say "tea" and "to".) The details of anticipatory planning vary from one language to another, and children learn to produce these encoded articulatory gestures in the first few years of life. Acoustic analyses of anticipatory coarticulation in aphasics show that aphasics do not differ markedly from normal controls (Katz 1988).

On the subcortical bases of the evolution of language

31

6. Voice-onset-time The defining speech production deficit of Broca's syndrome involves sequencing. Broca's aphasics lose control of the sequencing between larynx and supralaryngeal vocal tract activity. The acoustic cue that differentiates stop consonants such as [b] from [p] in the words bat and pat is voiceonset-time (VOT), the interval between the "burst" of sound that occurs when a speaker's lips open and the onset of periodic phonation produced by the larynx. The sequence between laryngeal phonation and the burst must be regulated to within 20 milliseconds. Broca's aphasics are unable to maintain motor sequencing control; their intended [b]s may be heard as [p]s, [t]s as [d]s, and so on. Control of duration is preserved in Broca's aphasics since the intrinsic duration of vowels is unimpaired.

7. Prefrontal activity and aphasia Although aphasia is by definition a "language" disorder, cognitive deficits occur. In fact, Goldstein, a leading figure in aphasia research, stressed loss of the "abstract" attitude. Goldstein (1948) noted the difficulties that aphasic patients had planning activities and strategies, shifting strategies, formulating abstract categories, and thinking symbolically. Subsequent research has found that these cognitive deficits are associated with impaired frontal lobe, particularly prefrontal, cortical activity (Stuss and Benson 1986). But frontal lobe cognitive deficits do not necessarily result from damage to frontal lobe structures alone. Studies employing Positron Emission Tomography (PET) and CT scans show that damage to elements of stitato-cortical circuits, to either prefrontal cortex, or subcortical structures supporting these circuits, can yield "frontal lobe" cognitive deficits. Metter et al. (1989) found that all of their Broca's patients had damage to the internal capsule and parts of the basal ganglia. PET scans showed that the circuit damage in these patients resulted in vastly reduced metabolic activity in the left prefrontal cortex and Broca's region.

32

Philip Lieberman

8. The subcortical locus of aphasia Marie (1926) claimed that subcortical lesions were implicated in the deficits of aphasia. This was reasonable, since the middle cerebral artery is the blood vessel most susceptible to damage. The central branches of this artery supply the putamen, caudate nucleus and globus pallidus, and one of the central branches of this artery is the thin-walled lenticulostriate artery which is exceedingly vulnerable to rupture. Modern brain imaging techniques confirm Marie's position. It has become apparent that permanent aphasia does not occur, absent subcortical damage. As Stuss and Benson note in their review of studies of aphasia, damage to "the Broca area alone or to its immediate surroundings ... is insufficient to produce the full syndrome of Broca's aphasia.... The full, permanent syndrome (big Broca) invariably indicates larger dominant hemisphere destruction ... deep into the insula and adjacent white matter and possibly including basal ganglia" (Stuss and Benson 1986: 161). In fact, contrary to the traditional theory, subcortical damage that leaves Broca's area intact can result in Broca-like speech production deficits (Alexander, Naeser, and Palumbo 1987; Mega and Alexander 1994). Damage to the internal capsule (the nerve fibres that connect neocortex to subcortical structures), the putamen, and the caudate nucleus can result in impaired speech and agrammatism similar to that of the classic aphasias in addition to other cognitive deficits. Alexander and his colleagues (1987) reviewed 19 cases of aphasia resulting from lesions in these subcortical structures. Language impairments occurred that ranged from fairly mild disorders in the patient's ability to recall words to "global aphasia" in which the patient produced very limited nonpropositional speech. In general, the severest language deficits occurred in patients who had suffered the most extensive subcortical brain damage. The locus for the brain damage traditionally associated with Wernicke's syndrome includes the posterior region of the left temporal gyrus (Wernicke's area), but often extends to the supramarginal and angular gyrus, again with damage to subcortical white matter below (Damasio 1991). Recent data again show that linguistic capability can be recovered after complete destruction of Wernicke's area (Lieberman 2000). As D'Esposito and Alexander conclude, it is apparent that a "purely cortical lesion - even a macroscopic one - can produce Broca's or Wernicke's aphasia has never been demonstrated" (D'Esposito and Alexander 1995: 41).

On the subcortical bases of the evolution of language

33

9. Neurodegenerative diseases Studies of the behavioural consequences of diseases such as PD and Progressive Supranuclear Palsy provide evidence for the role of basal ganglia in the FLS. These neurodegenerative diseases result in major damage to basal ganglia, mostly sparing the cortex until the late stages (Jellinger 1990). As Marsden and Obeso (1994) note, tremors, rigidity, and movement disorders occur. However, subcortical diseases also result in linguistic and cognitive deficits that in extreme form constitute a dementia (Albert, Feldman, and Willis 1974). Sentence comprehension deficits linked to syntax have been noted in independent studies of PD (Lieberman et al. 1990, 1992; Grossman et al. 1991, 1992; Natsopoulos et al. 1994). Illes et al. (1988) found that the sentences produced by PD subjects are generally short and have simplified syntax. Illes and her colleagues attributed these effects to the subjects compensating for speech production difficulties. However, it is clear that the syntactic deficits of PD are not simply a consequence of impaired speech. The data of Lieberman et al. (1990) showed sentence comprehension deficits in PD. The subjects in this experiment simply had to utter the number (one, two, or three) that identified a line drawing that best represented the meaning of the sentence that they heard. Nine out of 40 non-demented PD subjects showed comprehension deficits. The sentence comprehension test that Lieberman et al. (1990) used had been designed for hearing-impaired children. Therefore, the test's vocabulary can easily be comprehended by sixyear-old children, which argues against the subjects having any difficulties with vocabulary. Slight cognitive loss was associated with impaired sentence comprehension; the subjects who had sentence comprehension deficits showed no symptoms of dementia of Alzheimer's type, but cognitive decline was apparent to the neurologist who had observed them over a period of time.

10. Verbal working memory The sentence comprehension deficits of Broca's Aphasia and PD may reflect impairment of "verbal working memory". The concept of working memory derives from research on short-term memory. Short-term memory is usually thought of as a buffer in which information is briefly stored; working memory includes computation as well as storage. In a series of experiments

34

Philip Lieberman

that span thirty years, Baddeley and his colleagues (Gathercole and Baddeley 1993) showed that "verbal working memory" was implicated in both the storage of verbal material and the comprehension of sentences. Verbal working memory involves two components: an "articulatory loop", whereby subjects maintained speech sounds in working memory by subvocally rehearsing them using the brain mechanisms that regulate overt speech, and a "central executive" process. The central role that speech plays in the FLS is evident in phonetic "rehearsal", whereby words are subvocally maintained in working memory using the neuroanatomical structures that regulate speech production. Many experiments show that subjects have more difficulty recalling a series of longer words than shorter words, as might be predicted if the articulatory buffer had a finite capacity. When articulatory rehearsal is disrupted by having subjects vocalise extraneous interfering words - the numbers one, two, three - during the recall period, recall dramatically deteriorates. Verbal working memory appears to be an integral component - perhaps the key component - of the human functional language system, integrating speech perception, production, semantics, and syntax. The sentence comprehension test used in the Lieberman et al. (1990, 1992) studies included sentences having syntactic distinctions that are known to place different processing demands in verbal working memory in neurologically intact adult subjects: "centre-embedded" sentences such as The boy who was fat sat down; "right-branching" relative clause sentences such as I saw the boy who is fat; conjunctions like The boy swam and the girl rowed; and "simple" declarative sentences like I saw the boy. Some of the sentences, such as The apple was eaten by the boy, were semantically constrained, others semantically unconstrained (the boy was kissed by the girl.). Whereas control subjects make virtually no errors when they take the test battery used in Lieberman et al. (1990, 1992), the overall error rate was 30 percent for some PD subjects. The subjects' comprehension errors typically involved repeated errors on sentences that had particular syntactic constructions. Therefore, their syntax comprehension errors could not be attributed to general cognitive decline or attention deficits. The highest number of errors (40 percent) was made on "left-branching" sentences that departed from the canonical pattern of English having the form subjectverb-object. An example of a left-branching sentence is Because it was raining, the girl played in the house. Thirty percent errors occurred for right-branching sentences with final relative clauses like Mother picked up the baby who is crying. Twenty percent error rates also occurred on long conjoined

On the subcortical bases of the evolution of language

35

simple sentences like Mother cooked the food and the girl set the table. This again points to verbal working memory load being a factor in sentence comprehension; longer sentences place increased demands on verbal working memory. Similar sentence comprehension error rates for non-demented PD subjects have been found by Grossman et. al. (1991, 1993) and Natsopoulos et al. (1994) as well as by Pickett et al. (1998) for a subject having brain damage limited to basal ganglia. Grossman et al. (1991) also tested PD subjects' ability to copy unfamiliar sequential manual motor movements (a procedure analogous to that used by Kimura [1993] for Broca's aphasia). The PD subjects' manual sequencing and sentence comprehension deficits were correlated. The correlation is consistent with Broca's area playing a role in manual motor control (Lieberman 1984; Rizzolatti and Arbib 1998) and verbal working memory through circuits supported by basal ganglia (Marsden and Obeso 1994).

11. Sequencing deficits in speech, syntax, and cognition Similarities between the pattern of deficits in PD and Broca's aphasia were found by Lieberman et al. (1992). Acoustic analysis showed a breakdown in nine subjects' VOT control similar in nature to Broca's aphasia. The speech of the PD subjects was similar to that of Broca's aphasics in other ways; they produced appropriate formant frequency patterns and preserved the vowel length distinctions that signal voicing for stop consonants when they occur after vowels. The PD subjects who had VOT overlaps had significantly higher syntax error rates and longer response times on the test of sentence comprehension than the VOT non-overlap subjects; moreover, the number of VOT-timing errors and of syntax errors was highly correlated.

12. VOT deficits: sequencing or laryngeal control? Impaired laryngeal control, usually heard as hoarse, dysarthric speech and low amplitude speech (hypophonia), is a sign of PD. Therefore, it would be reasonable to suppose that impaired control of laryngeal muscles might be the root cause of the VOT deficits of PD. A study of Chinese-speaking PD subjects (Lieberman and Tseng 1994) showed that this was not the case.

36

Philip Lieberman

Chinese makes use of "phonemic tones", controlled variations in the fundamental frequency of phonation (FO) which is controlled by the larynx, to differentiate words. The syllable [ma] produced with a level FO contour in Mandarin Chinese, for example, signifies 'mother', whereas it signifies 'hemp' when produced with a rising FO contour. Twenty PD subjects read both isolated words and complete sentences in test sessions shortly after or before they took medication that increased dopamine levels. VOT overlap for PD subjects, significantly greater than that of age-matched normalspeaking controls for both the pre-medication and post-medication test sessions, decreased significantly after medication for half of the PD subjects. This showed that dopamine-sensitive subcortical circuits were implicated in sequencing the laryngeal and SVT motor commands that yield VOT distinctions. Acoustic analysis showed that the PD subjects were always able to generate the controlled FO patterns that specify Chinese phonemic tones. Since the FO patterns that specify these phonemic tones are generated by precise laryngeal manoeuvres, it is apparent that sequencing deficits are responsible for the observed VOT overlaps. In short, the VOT overlaps that can occur in PD appear to reflect impaired sequencing of the individual motor commands that constitute the motor programs that generate speech. This is not surprising in the light of studies of sequential motor activity in PD (Marsden and Obeso 1994; Cunnington et al. 1995), and the sequencing of "submovements" of rodent grooming noted by Aldridge et al. (1993). If the human neural circuitry regulating voluntary laryngeal activity during speech production is similar to that of monkeys, then the locus of VOT sequencing deficits may be the coordination of the activity of two independent neural circuits: a laryngeal circuit involving anterior cingulate gyrus, and independent circuits involving neocortical areas that regulate supralaryngeal vocal tract manoeuvres. Significantly, no neocortical areas appear to be implicated in the regulation of nonhuman primate vocalisations (Sutton and Jürgens 1988). Discussions of other experimental data consistent with basal ganglia regulating and shifting sequential motor and cognitive acts are noted in Lieberman (2000). Hypoxia on Mount Everest resulting from exposure to altitude results in VOT and sentence comprehension deficits (Lieberman et al. 1994, 1995). Pickett et al. (1998), in a study of a subject having brain damage limited to basal ganglia, found speech motor sequencing deficits as well as deficits involving sequencing in the comprehension of distinctions in meaning conveyed by syntax and in cognitive tasks. Vargha Khadem et al. (1998) documented speech and language deficits in the alleged "language

On the subcortical bases of the evolution of language

37

gene" family (misrepresented in Pinker 1994): subjects who have a genetically transmitted anomaly that results in bilateral reduction of caudate nucleus volume. Cognitive deficits similar to those occurring with frontal lobe damage can be traced to impaired basal ganglia activity in PD (Lange et al. 1992). Cerebellar damage also yields similar deficits (Pickett 1998). It is clear that subcortical structures are critical elements of the FLS that regulates human language and some aspects of cognition.

13. On the evolution of adaptive behaviour Despite somewhat strident claims to the contrary (Pinker 1994), languagetrained chimpanzees produce about 150 words using manual sign language or computer keyboards, roughly equivalent to the abilities of two year-old children. These chimpanzees also can understand spoken English words (Gardner and Gardner 1984; Gardner et al. 1989; Savage-Rumbaugh and Rumbaugh 1993). Indeed, other species comprehend spoken words. Pet dogs almost always comprehend a few words, and the circus dog "Fellow" understood at least 50 words (Warden and Warner 1928). Therefore, lexical ability itself, dissociated from speech production, is a primitive feature of language and undoubtedly existed in the ancestral species that was the common ancestor of human beings and apes as well as all archaic hominid lineages. Syntactic ability, which generally is taken by linguists (e.g., Pinker 1994; Calvin and Bickerton 2000) to be a unique human attribute, is present to a limited degree in language-trained chimpanzees. Analyses of the American Sign Language used by the Project Washoe chimpanzees show that they used two-sign combinations and some three-sign combinations. Savage-Rumbaugh et al. (1986) demonstrate that the six-year-old pygmy chimpanzee Kanzi comprehends simple sentences having the canonical form of English in which the subject always proceeds the object. Kanzi responded correctly about 75 percent of the time when he heard sentences like: Put the pine needles on the ball. And: Put the ball on the pine needles. Kanzi appropriately placed the pine needles on the ball or the ball on the pine needles in response to English-language commands, demonstrating a sensitivity to basic word order in English. The sole aspect of human linguistic ability that chimpanzees lack is speech. Despite attempts spanning the past 300 years, no one has been able to train a chimpanzee to talk. The supralaryngeal vocal tract anatomy of chimpanzees prevents them from producing vowels such as [i] and [u] (the

38

Philip

Lieberman

vowels of the words see and do). However, speech communication could take place without producing the vowel [i], albeit with increased error rates (see Lieberman 1984, 1991,2000). Acoustic analyses of chimpanzee vocalisations and computer modelling of the range of sounds that their vocal tracts can produce show that ape anatomy would permit them to speak producing bilabial and "alveolar-dental" consonants like [b], [p], [m], [t], [d], [s], and so on as well as all vowels save [i], [u] and [a] (Lieberman 1968, 1984). But chimpanzees cannot even freely permute the sounds that occur in their natural repertoire of calls. The factor limiting chimpanzee speech is the chimpanzee brain. Chimpanzees cannot produce vocalisations that are not bound to specific emotional states (Lieberman 1994a): "Chimpanzee vocalisations are closely bound to emotion. The production of a sound in the absence of the appropriate emotional state seems to be an almost impossible task for a chimpanzee"(Goodall 1986: 125). Human beings are able to freely permute the motor commands, the "submovements" that constitute the sounds of human speech to form words. The absence of productive speech and the limited cognitive and syntactic abilities of apes and other species may reflect their basal ganglia circuitry. It is quite possible that our human qualities, our ability to unbind articulatory submovements to generate a virtually unlimited number of words, regulate complex syntax, and abstract thought, derive from some unique aspects of human cortico-striatal circuitry. Further study using noninvasive techniques that will allow us to map the detailed striatal circuits of humans and apes will answer this question.

13.The antiquity of hominid speech The probable absence of a modern human vocal tract in Neanderthals and its almost certain absence in Australopithecines and Erectus hominids (Lieberman 1984), does not indicate the complete absence of speech. As the initial Lieberman and Crelin (1971) paper on Neanderthal speech capabilities carefully pointed out, Neanderthals undoubtedly possessed speech, albeit less efficient speech than modern humans. This conclusion follows from the logic of natural selection. The perceptual "decoding" of the formant frequency patterns that convey particular speech sounds must take account of the length of a speaker's vocal tract (longer vocal tracts produce lower formant frequencies for the "same" speech sound than shorter vocal tracts). The vowel [i] is optimal for this process (Nearey 1979). It is possible to perceive speech without [i], but its presence makes speech perception less

On the subcortical

bases of the evolution of language

39

susceptible to error. However, the human vocal tract increases the risk of choking to death when we swallow solid food (noted by Charles Darwin) and has other negative consequences (increased risk of death due to impacted teeth and less efficient chewing). Therefore, the restructuring of the hominid SVT to enhance speech perception would not have contributed to biological fitness unless speech and language were already present in the hominid species ancestral to modern Homo sapiens. There otherwise would have been no reason for the retention of the lower laryngeal position of the human SVT. In short, we can conclude that speech and some form of language (including syntactic ability, present in rudimentary form in living apes) must have already been present in Neanderthals and in the common ancestors of Neanderthals and modern humans, Homo erectus and perhaps Australopithecines. Indeed, it has become clear that some aspects of human speech are very primitive in an evolutionary sense. The neural processing that allows us to determine the length of the SVT of the person to whom we are listening occurs in many other species; studies of monkeys suggest that they also can judge the length of another monkey's vocal tract, which is a good index of a monkey's size, using a similar process (Fitch 1997). Early hominids surely must have possessed this ability. Nonhuman primate calls also can be differentiated through formant frequency patterns generated when monkeys or apes close or open their lips as they phonate (Lieberman 1968). The fundamental FO, which is determined by laryngeal muscles and alveolar (lung) air pressure, is one of the principal cues that signals the end of a sentence and major syntactic units. Most human languages make use of controlled variations of FO to produce "tones" that differentiate words. Since apes possess laryngeal anatomy that can generate FO contours, early hominids must have had this ability. The roots of speech communication may extend back to the earliest phases of hominid evolution.

14.Walking and basal ganglia About 5 million years ago a species lived that was the common ancestor of present day apes and humans. The fossil remains of early hominid species, like those of the 4.4 million-year-old Ardipithecus ramidus, resemble apes who could have walked upright (White et al. 1994). It is quite possible that upright bipedal locomotion may have been the preadaptive factor that selected for neural mechanisms that enhanced motor ability (Jesse Hochstadt,

40

Philip

Lieberman

personal communication). Human beings learn to walk. The "walkingreflex" that exists in new-born human infants is controlled by a quadripedal neural "pattern-generator" that reflects our hominoid ancestry. Heel strike, which marks efficient bipedal locomotion, takes toddlers several years to develop (Thelen 1984). The subcortical basal ganglia structures of the FLS regulate upright, bipedal locomotion. One of the primary signs of PD, in which basal ganglia circuits are degraded, is impaired locomotion. Thus, upright, bipedal locomotion may have been the initial selective force for the enhancement of the subcortical-cortical circuits that regulate sequencing of both motor and cognitive acts. Careful developmental studies relating the onset of walking to language, cognition, and brain development in young children will show whether this suggestion has any merit. Many topics relating to the subcortical brain and the evolution of language are discussed in Lieberman (2000). The dopamine depletion that characterises PD may also directly reduce verbal working memory span. Ongoing research suggests that this effect can occur independent of sequencing deficits. Studies of the neural bases of motor control also show that complex behaviours are generally learned rather than innate. Since the same subcortical neuroanatomical structures participate in both motor control and language, it is likely that syntax is learned rather than specified by a Universal Grammar. Neurophysiologic studies also show that algorithmic descriptions of motor behaviour are at best metaphors; the neural activity that governs motor control is parallel and distributed. Since there is no evidence that the physiology of the brain differs in any fundamental manner for motor control and language, we must conclude that algorithmic descriptions of language processing have the same status as they do for motor control. In short, the cumbersome and inadequate "rules of grammar " - innate Universal Grammar specifying the "rules" of syntax proposed by Chomsky and his disciples - are biologically implausible.

The origin of the human language faculty: the language amoeba hypothesis Eörs Szathmäry 1. Introduction The origin of the eye, once considered one of the hardest problems for evolutionary science, now looks almost trivial compared with the problem of the origin of human language. What makes the question of eye origin now seem to be so easy? There are at least four reasons. First, the anatomy and physiology of the eye are relatively well understood. Second, there are many eye forms that can serve as plausible evolutionarily intermediate stages (even the vertebrate eye has a close analogue in cephalopods). Third, although apparently related to an ancestral light-sensitive spot, eyes with various anatomical and physiological structures evolved about forty times. Fourth, the genetics of eye development is now better understood. Compare these with the corresponding list for language origin. First, our knowledge of the "language organ" is fairly limited. Second, there seems to be no extant intermediate between "protolanguage" (broadly understood as limited vocabulary without syntax) and human language. Third, even the much simpler protolanguage evolved only a few times (candidate species where protolanguage could be used in the wild include bottle-nosed dolphins, grey parrots, and chimps), and natural language evolved only once. Fourth, the genetics of human language is largely unknown. It may seem, then, almost impossible to say anything useful about the origin of natural language. Nevertheless, there are at least five sources of scientific input that can, I think, contribute to enhancing our understanding of this conundrum: • The spread of elements of vocabulary and language rules can be analysed using the tools of population biology (Nowak and Krakauer 1999, Nowak et al. 2000). • A better understanding of primate communication and social life greatly contributes to narrowing the evolutionary and intellectual gap between our ancestors and us. • The anatomy, physiology, and genetics of specific language impairment (SLI) expand our knowledge of language's biological foundations.

42

Eörs

Szathmäry

• New, noninvasive brain analysis methods (such as PET) reveal a surprisingly dynamic neural foundation for human language. • We are in the process of formulating a selectionist picture, connected to brain epigenesis, of what happens inside the brain during at least some forms of learning. In this essay I will elaborate on the last four points. I will argue that our increased understanding in these areas points toward a "language amoeba" in the human brain. The language amoeba is the neural activity pattern that essentially contributes to processing of linguistic information, especially syntax. It is a sort of dynamic manifestation of Chomsky's language organ. It finds its "habitat" in the developing human brain, whereas the brains of other primates apparently cannot sustain it. Why? I attempt to outline a tentative answer. Since the gap between other primates and us is fairly narrow, the genetic and correlated functional changes could not have been too numerous. What were the critical changes? I argue that an appropriate and rather widespread connectivity pattern in the immature human brain renders it a habitat for the emerging language amoeba. This change does not require too many altered (probably regulatory) genes, but there are great risks involved, which make this "major transition" truly difficult.

2. Primate communication and cognition: the narrowing gap Primates are especially relevant, since they are our closest evolutionary relatives. Bickerton (1992) correctly emphasises that the right starting point for understanding the origin of natural language is something like protolanguage. Our closest relatives seem to possess it in some form. They transmit signals that may well turn out to be Saussurean signs: there is an orderly and reversible relation in the brain that links object to concept to sign. Sensitivity to conspecific vocalisations, but not to other types of auditory input, seems to be lateralised (Ghazanfar and Hauser 1999). Apes can be taught some elements of human language, but their syntactic abilities are clearly limited. This is even true for "action grammar": the hierarchical structure of embedded actions. Chimps seem to be severely limited with recursive embedding (a crucial element of natural syntax): they usually cannot go beyond three levels. Remarkably, "linguistically" trained chimps do somewhat better (Greenfield 1991). The case of Kanzi demonstrates that childhood learning is better than adult learning and that understanding

The origin of the human language faculty

43

is easier than production. But this should not surprise a human too much. In the social context, apes seem to carry all the necessary Machiavellian weapons that we do. In an insightful paper, Worden (1995) deduced a "speed limit" for evolution that seems surprisingly uniform despite the fact that the relative contributions of component processes (mutation, recombination, selection, and drift) are allowed to vary. He argues that the number of genes specifically associated with the "language organ" must be rather limited, which prompts one to think that it was not the number of genes, but their identity (i.e., mainly regulatory effects) that was peculiar during the evolution of natural language. All this leads me to two conclusions. First, apes can do everything that we can do - expect for things that require language or neural processing abilities that are likely to have evolved through natural selection for language. Second, there are likely to be surprisingly few genes associated with language's neural substrate that are critically different from ape genes/alleles.

3. Specific language impairment (SLI): phenomenon and genetics There is no disagreement that SLI is real. What is contested is how closely it is limited to, or rooted in, a specific grammatical impairment. The famous Gopnik (1999) case has been very stimulating because of its characterisation as "feature-blind" dysphasia and its obvious genetic background (a single dominant allele). Whether other cognitive skills are also, or even primarily, affected has been debated ever since (Vargha-Kadem et al. 1998). Recently, more evidence involving other linguistic groups has accumulated (Dalalakis 1999, Rose and Royle 1999, Tomblin and Pandich 1999). A recent paper (Van der Lely et al. 1998), sadly without genetics, claims to demonstrate that grammatically limited SLI exists in "children", notwithstanding the fact that it only studies one child. Clearly, more genetics is needed. But the task is formidable owing to the obvious restrictions. I expect some crucial input to come from the genome projects. This input includes a list of genes and alleles that act as liability genes (Flint 1999) for language and that can cause SLI when mutated. Moreover, extension to a primate genome project could reveal genes and alleles that are closely and specifically associated with human language. It is worth remembering that the genetics of human cognitive tasks is a notoriously difficult problem. The clinical characterisations are usually not

44

Eörs

Szathmäry

sufficient as descriptions of phenotypes (Flint 1999). Nevertheless, a consensus seems to be emerging that the genes involved are so-called "liability genes" that, when present in the right allelic form, significantly enhance the probability of developing the respective cognitive skills. We can safely conclude the following: • Specific language impairment does exist, and some of its forms seem to be familial. There is disagreement as to whether it is restricted to grammar. The likely explanation behind these contrasting views is that the genetic factors in question distort smaller or larger parts of the brain. When language becomes localised, the morpho-physiological distortions involved partially overlap with the habitat of the language amoeba. It is easy to see that the relative size of the overlap with the outlying regions and the nature of the affected regions of the habitat will influence the degree to which the impairment is specific to language or, more specifically, to grammar (see Figure 1). • Since many of the genes are likely to influence the epigenesis of a certain brain area, incomplete penetration can cause more or less severe forms of impairment. a

b

c

grammar

impaired regions

Figure 1.

language amoeba

The origin of the human language faculty

45

4. Neural dynamics and localisation of language: noninvasive studies The analysis of neural activity during the performance of cognitive tasks has become a growth industry. The equipment's sensitivity has increased over the years. These methods are increasingly used to record brain activity during linguistic performance. It is now widely accepted that neural localisation of language can be plastic (Nobre and Plunkett 1997, Neville and Bavalier 1998, Musso et al. 1999). Studies of brain injury have revealed that damage to the left hemisphere before a critical period is not necessarily debilitating because the right hemisphere can take over the necessary functions (Müller et al. 1999). This does not contradict the finding that in most people Broca's area does seem specialised for syntax (Embick et al. 2000). It thus seems that the common left-hemisphere localisation of language is just the most likely outcome when there is no genetic or epigenetic disturbance. PET studies have revealed a truly shocking feature of language development: the localisation of linguistic processing shifts during normal ontogenesis. The outcome in "normal" people is also highly variable.

growth

transient redundancy

selective stabilisation

Figure 2.

46

Eörs Szathmäry

Analysis of a particularly interesting genetic syndrome called Williams syndrome also reveals surprisingly dynamic manifestations during ontogenesis. Whereas affected children seem to be bad at language and good with numbers, adults perform the other way round (Paterson et al. 1999). Needless to say, the classic characterisation of the disease has been based on adult performance. I draw the following conclusions from brain studies. First, language localisation is not fully genetically determined: even major injuries can be tolerated before a critical period. Second, language localisation to certain brain areas is a highly plastic process, both in its development and its end result. Third, a surprisingly large part of the brain can sustain language. True, there are recognised areas that seem to be most commonly associated with language, but these are by no means exclusive, either at the individual or the population level, or during either normal or impaired ontogenesis. Fourth, whereas a large part of the human brain can sustain language, no such region exists in apes.

5. Brain epigenesis: plasticity and selection It's time for a confession: on the whole, we don't understand how the brain works. Nevertheless, some crucial elements seem to be apparent. One is that the normal brain's development is enormously plastic, even though the power of genetic factors is obvious. One classic example is that in the same brain area of identical twins the two hemispheres of the same individual resemble each other more closely than the twin's corresponding hemispheres (Changeux 1983). Another insight is that a tremendous amount of variation and selection is going on during brain ontogenesis. This is doubtless a Darwinian process. As the psychologist William James recognised a long time ago, natural selection of heritable variation is the only known force that can lead to adaptations, which has led to its application to both brain ontogenesis and problem-solving. There are several studies that all regard the brain, in one way or another, as a "Darwin machine" (see Calvin and Bickerton 2000). I think Changeux's formulation (1983) is the most relevant to the language problem. According to this view, the functional micro-anatomy of the adult cortex results from the vast initial surplus stock of synapses and their selective elimination according to functional criteria (see Figure 2).

The origin of the human language faculty

47

The previous section indicated that a large part of the human brain can process linguistic information and perform syntactical operations. This means that there is no fixed macro-anatomical structure that is exclusively dedicated to language, but that the micro-anatomical structure must be appropriate, otherwise it could not sustain language. There is consequently some statistical connectivity feature of a large part of the human brain that suits it to linguistic processing. From the selectionist perspective (see Figure 2) there are three options: the initial variation in synaptic connectivity is novel; the means of selecting functional criteria are novel; or both. I think it is likely that both component processes are different in the relevant human brain areas. I would not dare to speculate on their relative importance. It is necessary to see this idea in close connection with Rapoport's observations (1999) on the bottom-up coevolution of brain and cognition. According to the bottom-up view, a genetic change of some neural structure is subjected to selection, and based on its performance it either spreads or does not. There is, however, a so-called top-down mechanism that could have contributed more significantly to the evolution of human cognitive skills, including - especially, I argue - language (see Figure 3). The topdown view has three crucial components. First, due to the plasticity of brain development, enhanced demands on a certain brain region lead to less synaptic pruning (a known mechanism). Second, less synaptic pruning leads to more elaborate (and more adaptive) performance. Third, natural selection will favour any genetic change contributing to the growth of that brain area. Heritable variation in brian neuronal and synaptic numbers and neuroplasticity among immature primates + N e w cognitively, behaviourally, culturally or socially stressfull environment

Activation o f brain regions by ideation and attention during language acquisition during maturation

Reduced synaptic pruning, increased network elaboration in language-related regions

U

Adult genotypes with more extended language-processing systems are most cognitively competitive, thus most fecund

Spreading o f their genes in the population leads to natural selection o f an enhanced language faculty

JJ

Figure 3.

Further increases in cognitive, social or behavioural pressures caused by evolved population

t ΐ ί ί

48

Ears

Szathmäry

Two additional observations are necessary. First (as Rapoport himself conceded), the top-down mechanism is a more detailed exposition of the late Alan Wilson's theory. According to Wilson, a larger brain, due to its more complex performance, alters the selective environment (in social animals it is composed largely of conspecifics), which selects for an even larger brain, and so on. Second, and perhaps more importantly, this mechanism is also a neat example of a Baldwin effect (or genetic assimilation), in which learning guides evolution. As Terry Deacon (1997) points out, applying the notion of genetic assimilation to language is trickier than is usually thought. The reason for this is that the performed behaviour must be sufficiently long lasting and uniform in the population. It is thus hard to imagine how, say, specific grammatical rules could have been genetically assimilated. Here, however, we are dealing with a different phenomenon: the genetic assimilation of a general processing mechanism that is performed via the connectivity of the underlying neural structures. My claim is that the network's ability to process syntactical information was the most important - and largely novel faculty selected for. The specific hypothesis is that linguistically competent areas of the human brain have a statistical connectivity pattern that renders them especially suitable for syntactical operations. This leads me to three conclusions. First, the origin of human language required genetic changes in the mechanism of epigenesis in large parts of the brain. Second, this change affected statistical connectivity patterns of the neural networks involved. Third, due to the selectionist plasticity of brain epigenesis, coevolution of language and the brain resulted in the genetic assimilation of syntactical processing ability as such. But if this is true, then why isn't language more common?

6. The origin of language: a difficult transition? Some major transitions in evolution (such as the origin of multicellular organisms or that of social animals) occurred a number of times, whereas others (the origin of the genetic code or of language) seem to have been unique events. One must be cautious with the word "unique", however. Because we lack the "true" phylogeny of all extinct and extant organisms, we can give it only an operational definition. If all the extant and fossil species that possess traits due to a particular transition share a last common ancestor after that transition, then the transition is said to be unique. It is obviously possible that there have been independent "trials", but we do not have comparative or fossil evidence for them. Two factors can lead to "truly" unique transition.

The origin of the human language faculty

49

1) The transition is variation-limited. This means that the set of requisite genetic alterations has a very low probability. "Constraints" operate here in a broad sense. 2) The transition is selection-limited. This means that there is something special in the selective environment that favours the fixation of variants that are otherwise not particularly rare. Both abiotic and biotic factors can contribute to this limitation. There are interesting sub-cases for both types of limitation. For 1), one can always enquire about the time-scale. "Not enough time" means that given a short evolutionary time horizon the requisite variations have a very low probability, though this could change with a widened horizon. An interesting sub-case of 2) is "pre-emption", meaning that the traits resulting from the transitions carry out selective overkill and sweep through the biota so quickly that further evolutionary trials are competitively suppressed. The genetic code could be a case in point. It is hard to assess at the moment why language is unique. Even the notenough-time case could apply, which would be amusing. But pre-emption, due to the subsequent cultural evolution that language has triggered, may render further trials very difficult. There is, however, another indication that language could be variation-limited in a deeper sense. The habitat of the language amoeba is a large, appropriately connected neural network. Most of the information processing within the network elaborates on information coming from other parts of the network. There is a special type of processing required, namely that of hierarchically embedded syntactic structures. I see the following difficulties: • Neural networks contain a large number of cycles. Syntactic structures of language are tree-like. It seems difficult to process large trees without getting into loops. • Overproduction of initial synapses or decreased pruning, both implied in the origin of language, may easily lead to "solipsist" network dynamics. This has two consequences. First, the network's activity is detached more than optimally from external sources of information. Second, exaggerated internal processing leads to too much "internal talking": linguistic processing for its own sake.

50

Eörs

Szathmäry

I think schizophrenia is a case in point. It may be, as some have contended, the "price for language" (Crow 2000, Maddox 1997). True, this syndrome affects a surprisingly large portion - up to 5 percent - of the human population. But this is exactly the type of pattern one would expect from an evolutionarily recent major transition. Insufficient adaptive fine-tuning causes the device to malfunction fairly frequently. Brain lateralisation is partly a means to decrease the spontaneous fission of the language amoeba, which could otherwise happen more easily. After all, a large habitat can sustain two amoebas. Moreover, there is evidence that schizophrenic people display reduced lateralisation (Gold and Weinberger 1995, Shapleske et al. 1999, Zaidel 1999, Niznikiewicz et al. 2000). The idea that a schizophrenic language amoeba is split into two (or more) functional offspring can be tested by, for example, PET and related methods. In practice, however, this might be hampered by the fact that the two entities' spatial separation may not be apparent enough to be observable.

7. Modelling One of the ways to test my ideas would be to model neural network architectures that are particularly suited to syntactic processing. Obviously, it cannot be a clumsy application of some current connectionist model (see Elman et al. 1996). There are three reasons that discourage me from applying such models. First, they tend to use some variant of the back-propagation algorithm, which is known to be grossly unrealistic (Crick 1998). Second, they cannot figure out abstract rules in the way people seem to do readily (Marcus 1998, Marcus et al. 1999). Third, they do not incorporate the implementation of neuronal selection. My suggestion for a more fruitful modelling of linguistic operations by neural network is straightforward: implement the selectionist picture (Figure 2) as an algorithm. The plasticity involved could more easily lead to the appropriate architecture. My recommendation contains an additional twist. It comes from the recognition that even the rules of epigenetic plasticity are themselves evolutionarily plastic; that is, they can be moulded by variation and selection. This evolutionary plasticity is probably also necessary for a successful research project. Very recently, Rolls and Stringer (2000) presented such a novel method for the design of neuronal networks. They specify a number of gene types influencing the development of neuronal networks. A small number

The origin of the human language faculty

51

of genes specify the generic properties (e.g., the number of neurons in the class and the firing threshold) of a certain neuron class. A larger number of genes influences synapse formation from neurons in other classes onto neurons in the given class. Genetically determined traits include whether the connection is excitatory or inhibitory, the nature of the learning rule at the synapse, and the initial synaptic strengths. A genetic algorithm was used to evolve networks performing a particular task, including competitive learning, pattern association, and auto-association. Such an approach would also be welcome in language studies.

8. Where to start? It is not obvious where to start evolving a linguistic neuronal network. This problem relates to the question of whether the connectivity pattern in the language amoeba's habitat is unprecedented among primates. Probably not. First, because it is a statistically determined property, a few such areas must appear by chance. Second, my guess is that apes use similar areas for social cognition, but that limited cortical size does not provide a sufficient habitat for the language amoeba. The idea that social cognition (including the mental simulation of actions in a partially hierarchical network) is a good prerequisite for language is not new (Calvin and Bickerton 2000). Deacon's (1997) observation that an increase in the size of the prefrontal cortex could have critically liberated parts of the brain from visceral tasks is another case in point.

2. The first language

The apparent paradox of language evolution: can Universal Grammar be explained by adaptive selection? Manfred Bierwisch 1. Three introductory remarks The evolution of natural language has again become the subject of lively debate. Like in previous discussions of this topic, an important distinction is often ignored: the origin of both a given language and of languages in general is not the same as the origin of the language capacity as a speciesspecific disposition. The corresponding developments must be clearly distinguished.

1.1. Language and language

capacity

Ferdinand de Saussure (1916) draws a crucial distinction between langue, parole, and langage. For Saussure, the faculte de langage is the speciesspecific capacity to acquire and use a natural language; it is a biological property of the human organism. Langue refers to natural languages like French, Hungarian, or Hebrew; it is a social institution that consists of the knowledge shared by the members of a given speech community. Parole refers to the intentional use of linguistic expressions and comprises the actual, psycho-physical processes determined by the knowledge of a language and by other situational conditions. Noam Chomsky (1965) employs the terms language capacity, linguistic knowledge or competence, and language use or performance to draw similar distinctions, though in a slightly different way. Whereas Saussure is primarily concerned with the socio-cultural aspects of language as an institution (in other words: with langue), Chomsky is essentially interested in the mental structure of individual speakers' knowledge. For Chomsky, language as a social institution depends on there being sufficient similarity between speakers' knowledge. For even if language as a social institution is not merely the linguistic knowledge shared by the members of a given speech community, individuals' knowledge remains an indispensable condition.

56

Manfred Bierwisch

We will see that the asymmetrical interdependence between Saussure's emphasis on language's social aspect and Chomsky's interest in language's mental structure will have intriguing consequences for the issues related to phylogenetic development. Both Saussure and Chomsky view the language faculty as an inherited biological condition and the use of language as a process carried out by individual, psycho-physical mechanisms operating under various conditions. To suggest how linguistics can contribute to the understanding of language phylogenesis, we need to make the above concepts slightly more specific. Following Chomsky (1986), this can be done along the following lines. The formal structure of the human language capacity that organises possible systems of linguistic knowledge is called Universal Grammar (UG). The system of knowledge that determines a possible natural language (L), the "internal language" L, is characterised by the Grammar (G) of L. The structure of a given utterance (u) determined by L is characterised by the structural description (SD) of u. UG, G, and SD are theoretical constructs corresponding to or describing mental structures that are schematically related in the following way: =====> UG > G phylogenesis ontogenesis (language origin) (language acquisition)

— > actual genesis (language use)

SD

The processes indicated by the arrows are, of course, fundamentally different. First, language use - primarily the production and comprehension of utterances - consists of short-term processes lasting seconds or less. They are relevant because they are the source of all primary linguistic data. One might furthermore argue, like Hubert Haider (1991), that UG's properties are essentially due to the processing routines by which humans deal with linguistic utterances. Second, the language acquisition process is part of individual human organisms' (presumably epigenetic) development. It is relevant because it depends (in addition to the actual experience a child is exposed to) on the language capacity - that is, UG. UG renders the process possible and thus shapes the structure of the resulting knowledge. Third, the phylogenetic process from which UG originates is subject to biological evolution. Yet the origin of UG should obviously not be confused with the origin of a primordial language. The former is part of biological evolution, whereas the latter is part of cognitive or social evolution. Nearly

The apparent paradox of language evolution

57

all speculations about language origin - including those of Plato, Leibniz, Herder, and Engels - either address the origin of language knowledge (usually: words) or simply confound the two issues.

1.2. Evolution Language capacity - and thus UG as its structural aspect - is a species-specific trait of homo sapiens. In other words, its genetic foundation must reside in the 1.5 percent that, according to a recent assessment, distinguishes the human genome from that of our closest relatives. Like any other genetically determined trait, language capacity is the result of evolution. Following Gerald Edelman (1987), evolution can be characterised by three conditions. Variation creates randomly alternative properties, the origin of which is not causally connected to the conditions governing subsequent processes of selection. Selection favours certain variants over others during encounters with an independently changing environment. Heredity distributes the favoured variants within a population via differential reproduction or amplification. Variation applies to the individual's genotype, selection applies to its phenotype, and heredity to the population the individual belongs to. The standard Darwinian view conceives of selection as adaptive in the sense that the variants survive via reproduction if they guarantee their bearer a benefit over possible competitors in survival and reproduction, thus spreading the selected property within the population. Selection thus guarantees adaptation according to environmental conditions. This notion of adaptive selection, typically referred to as "survival of the fittest", requires two significant amendments. First, random variants that do not provide a benefit might survive and be reproduced - without adaptive consequences - as long as they do not represent a manifest disadvantage. Stephen J. Gould (1982) has used the term exaptation to describe this extension of Darwinism. Exaptive properties can, however, become advantageous (or disadvantageous) if the environment subsequently undergoes relevant changes. In other words, between the origin of a variant and what might be called "delayed selection" other properties or conditions may change, thus assigning the variant a different role. Second, genotype variation might generate adaptive properties that are causally connected to concomitant, but nonadaptive properties. Known as "emergence", this phenomenon might be insignificant, like for example the colour of the iris as opposed to its shape. Of greater importance are, for

58

Manfred

Bierwisch

example, certain concomitants of increased brain size. Phenomena of emergence - combined with those of exaptation - are of particular relevance considering the genome's enormous complexity and the largely unknown means by which it determines the equally complex proteome, that is, the system of proteins that are responsible for the phenotype's morphology and behaviour. With respect to language, there is no reason to suppose researchers will one day identify a discrete group of genes that separately and completely determine nothing but a well-defined, adaptively selected brain structure supporting just the capacity described by UG.

1.3. Signals and thoughts Nevertheless, language capacity and the knowledge it gives rise to comprise a biologically determined mental system that organises a specific component of human behaviour. It interacts with other systems of the mind/brain like vision and hearing as well as locomotion and other motor activities. Even though there are obvious and relevant conditions shared by several of these systems - including linear organisation, hierarchical structure, and invariant patterns retrievable from memory - language capacity has its own, domainspecific properties. Roughly speaking, language capacity recruits two phylogenetically prior systems of mental organisation, thereby creating a systematic correspondence between their representations. The first is the system of articulation and perception (A-P) that underlies the production and recognition of invariant structures of external signals. The second is the system of conceptualisation and intention (C-I) that allows for conceptually organised and intentionally controlled representations of experience. Both A-P and C-I, which support what is usually termed the form and meaning of linguistic expressions, might in themselves be complex aggregates of mental organisation. A-P is normally instantiated by the systems that control vocal articulation and auditory perception. But recent research initiated by Edward Klima and Ursula Bellugi (1979) clearly demonstrates that sign language, based on the production and perception of visual signals, represents a fully adequate alternative. C-I must be construed as integrating the full range of systems involved in conceptual organisation, intentional control, and motivational instigation of experience and behaviour. Both systems are products of previous evolutionary stages. They might also exhibit new properties, since the language capacity exploits their possibilities.

The apparent paradox of language evolution

59

Schematically, these considerations can be abbreviated as follows: signal
A-P < v

> C-I < V

> environment

'

language capacity We can also refine things somewhat. The knowledge of L, characterised by G, determines the correspondence between A-P and C-I by means of the phonetic form interface (PF) and the semantic form interface (SF), where PF and SF represent those aspects of A-P and C-I on which the correspondence between the two domains relies. These refinements yield following schema: signal < = > A-P PF
SF C-I < = = > environment G It UG

Based on the possibilities provided by UG, G determines the relation given by pairs (π, σ), where π and σ belong to PF and SF, respectively. The crucial point is that this relation ranges over an unlimited set of pairs. G must therefore be a system for computing new pairs on demand. This has far-reaching consequences.

2. Properties of UG It is an empirical fact that normal human beings acquire the language spoken in their environment on the basis of varying and incomplete input. Yet external input only partially determines the structure and result of this epigenetic process. It must also rely on internal, biologically fixed conditions. UG specifies the structural conditions this internal disposition contributes to the structure and result of the language acquisition process. These internal conditions cannot be inspected directly. They therefore must be identified hypothetically by examining the structure of the observable result. Section 2.1 sketches UG's main traits.

60

Manfred Bierwisch

2.1. Lexical items and their combination As a necessary condition, a system of linguistic knowledge must organise a fairly large set of basic expressions or lexical items (roughly: words), each associating a form π (a representation in PF) with its meaning σ (a representation in SF). However, for pairs (π, σ) to be able to function as proper linguistic expressions, they must be categorised by grammatical features determining their possible combination into larger expressions. (1) below provides the most elementary illustration of this point. It combines the words daddy and jump in different ways, leading to different categories of expressions: (1) (a) daddy jumps (b) daddy's jump (c) jumping daddy In (la), jump is a verb, the final -s indicates the personal inflection, and the whole expression a clause. In (lb), jump is a noun, the -s attached to daddy indicates the possessive case, and the whole expression is a nominal phrase with jump as its nucleus. In (lc), the inflection -ing turns jump into an adjective modifying daddy, which in this case is the nucleus. The knowledge of lexical items thus comprises three types of information (π, γ, σ), where π and σ represent form and meaning, as already noted, and γ constitutes what might be called the "grammatical form" (GF). GF categorises the expression in question and determines its combinatorial properties. Hence, knowledge of English includes lexical items like (2), where /jump/ indicates a feature matrix of PF, JUMP an abstract characterisation of a particular type of motion, and [Verb] and [Noun] the relevant features of GF. (2) (a) (/jump/, [Verb], JUMP) (b) (/jump/, [Noun], JUMP) (2) (a) and (b) must presumably be treated as one lexical entry, allowing for [Verb] and [Noun] as alternative specifications of GF. These simplified examples indicate that UG must, among other things, provide the following conditions: (3) (a) Accessibility of primitive elements in terms of which the interface representations PF and SF and the categorisation GF can be specified and fixed in memory, (b) The general format of lexical data structures comprising (PF, GF, SF).

The apparent paradox of language evolution

61

Three things need to be emphasised. First, the general availability of specific types of primitive elements, their properties, and their organisational format in lexical items are neither obvious nor trivial. They are crucial determinants of the way in which linguistic expressions can be organised. Moreover, they specify the structure of representations supporting the mental computation of complex expressions. Second, GFs like Noun, Verb, Genitive, and so forth determine which lexical information goes beyond the mere association of form and meaning. It is the grammatical information g that determines the combinatorial possibilities of words and distinguishes them from labelled concepts. Finally, lexical items can be subject to inflection and derivation - the two branches of Morphology - , which is another of GF's facets. Morphological distinctions allow recurrent conditions in the SF-PF correspondence to be explicitly related to and fixed by features of GF. This provides guidelines for the combinatorial aspect of linguistic expressions, as even simple cases like (1) above indicate. The following constructions with the particle again demonstrate how intricate lexical information and its combinatorial aspects are. Their PF differs only by nuclear-stress placement (marked by capital letters). But the corresponding SF requires different conditions (informally added in parentheses): (4) (a) Mary had LEFT again, (presupposition: Mary was absent before) (b)Mary had left AGAIN, (presupposition: Mary had left once before) In other words, again adds the notion of repetition to both assertions about Mary having left. If, as in (4b), again is stressed, it repeats the event of Mary's leaving, whereas if, as in (4a), left is stressed, only the state of Mary's absence is repeated. In (5), the required lexical information is indicated in a rather provisional way: (5) (a) (/again /, [Particle], REPETITION OF P) (b) Content of Ρ is determined by position and stress. "REPETITION OF P" indicates that Ρ presupposes that Ρ was the case before (Stechow 1996 offers a detailed discussion of the complexities of again). I want to emphasise two things. First, knowledge of particles like again, also, or almost involves intricate conditions relating lexical information to the combinatorial requirements of syntactic, semantic, and even phonetic structure. Second, complexities of this sort are by no means

62

Manfred Bierwisch

exotic or marginal phenomena, but belong to the core of linguistic knowledge. Even though their specific details differ from language to language and are subject to language acquisition, UG provides the general possibility of their structure.

2.2. Hierarchy and

compositionality

The principles by which the combinatorial conditions of lexical items are realised are largely predetermined and subject to highly restricted variation in language acquisition. With respect to PF, combination essentially consists of sequential ordering. For example, John is followed by walks, which is followed by slowly in John walks slowly. Things are less trivial if prosodic consequences - stress and intonation - are considered, but these aspects are still related to the signal's linear organisation. The crucial step is due to SF which imposes an inherently nonlinear, hierarchical structure on the sequential combination of basic expressions. This hierarchy and its constituents create the GF, usually represented by tree structures or bracketings, as illustrated in (6). Here, constituents are categorised by features originating in the GF information of lexical items:

(6) (a)

( b ) [ s [ D Mary] [ v " [ v " [ v left] [D' [ D the] [ N book] ] [p' [ p on] [ D ' [ D the] [ N desk] ] ] ] ] ]

The category symbols (D, Ν, Ρ, V, C for Determiner, Noun, Preposition, Verb, and Clause and X' for complex categories with the head or nucleus X) are standard abbreviations. Their particular properties need not concern us here. (6a) is equivalent to the labelled bracketing (6b). The principle underlying this type of structure can be characterised as follows:

The apparent paradox of language evolution

63

(7) Two constituents X and Y combine into a complex constituent [ γ Χ Υ], where the category γ is determined by the category of the head. This formulation is somewhat simplified. In particular, it ignores the nontrivial selectional conditions lexical items impose on potential co-constituents, as illustrated above with respect to again (see Bierwisch 1997 for a detailed discussion). Example (6) illustrates what is traditionally called Constituency or Phrase Structure. It is an essential factor in the computation by which complex expressions derive their meaning. It gradually integrates the lexical items' SF into the SF of more complex constituents. The SF of the Determiner the therefore combines with the conceptual conditions of the Noun desk, selecting from the (situational) context a particular object that meets just these conditions. The next step turns the individual thus specified into the anchor point of the relation expressed by the Preposition on, which fixes the location of the act ascribed to Mary. More generally, the hierarchy described by Phrase Structure controls the Compositionality of the mental computation, by which a complex expression's SF derives from the SF of its constituents on the basis of their combination. This effect can be observed directly in cases of Phrase Structure differences. For instance, if we replace the verb leave by the verb buy, the natural structure would be (8). Here, on the desk combines with book, adding further conditions to the object in question rather than to Mary's act of buying: (8)

[ c Mary [ y bought [ D the [ N book [ p on the desk] ] ] ] ]

In many cases, two alternative structures that determine clearly different meanings . Thus in (9), in which only the relevant differences in constituency are indicated, we get (9a), in which Robert Redford is most likely a protagonist of the film and (9b), in which he is a participant of the discussion. (9) (a) they [discussed [the film with Robert Redford] ] (b) they [ [discussed the film] [with Robert Redford] ] In other words, the structural ambiguity of linguistic expressions can come from an alternative Phrase Structure being assigned to the same PF.

64

Manfred Bierwisch

2.3. Chains of positions A further characteristic property of natural languages seems to go beyond the minimal requirements for the mapping between PF and SF. A constituent's sequential position in PF may not correspond to the role it plays in SF. A characteristic case in point is the position of the finite verb in German, Dutch, and a number of other languages: (10) (a) daß das Konzert erst eine halbe Stunde später anfing, (b) das Konzert fing erst eine halbe Stunde später an. ((that) the concert didn't began until a half hour later) Because anfangen must be registered as a lexical item, the SF of which cannot be derived from that of an and fangen, the two parts of the Verb separated in (10b),fing and an, must be one constituent with respect to SF, as is overtly the case in the subordinate clause construction (10a). Hence, fing has two functions in (10b): an overt position in PF and a covert participation in the Verb's SF. The relevant structure resembles (11), in which fing indicates the covert position; the arrow connects it to its overt realisation: D

das Konzert] [ v [ v fing] [ [ A erst eine halbe Stunde später] [ v an [ v ftftg] ] ] ] ]

t

I

The principle that leads to chains of positions can be formulated as follows: (12) A constituent X can occupy more than one position in a structure Κ depending on the grammatical features of X and K, forming a chain (X n ... Xj) of positions, in which only X n is realised in PF. The features on which chain formation depends originate in the lexical items involved. The details are anything but trivial, but need not concern us here. As cases like (13) show, a complex expression may contain more than one chain, whereas (14) illustrates the fact that chains may be involved in the ambiguity of linguistic expressions:

The apparent paradox of language evolution

65

(13) (a) Was fängt Peter damit an? (What is Peter doing with that?) (b) [was [fängt [Peter [damit [was [an fängt] ] ] ] ] ] Τ

Τ

I

1

(14) (a) Mary had [plans [to leave] ] (b) Mary had [plans [to leave plans] t

I

In (14a), Mary intends to leave, whereas in (14b) Mary had plans to drop off. Typically, (14a) has nuclear stress on leave, (14b) on plans.

2.4. The innateness

ofUG

To sum up, UG must essentially provide the following conditions or principles: (15) (a) Primitive elements from which the interface representations PF and SF as well as the categorisation required in GF can be invented; see (3a). (b) The organisation of lexical items, including morphological indicators; see (3b). (c) Phrase Structure, which supports SF's Compositionality; see (7). (d) Chain formation, which assigns one constituent to different positions; see (12). The above description of UG's content is, of course, simplified. For my purposes here it is important to note that the conditions (15)(a) to (c) - with the possible exception of Morphology - are conceptually necessary for any system generating a correspondence between A-P and C-I that goes beyond a list of pairs (π, σ). By contrast, (15d) is an empirical fact about the human language capacity that does not seem to be logically necessary. In any event, only systems with the above four abbreviated structural principles can support the knowledge and use of a natural language. By the same token, these principles delimit the range of possible natural languages. This includes their geographical, historical, or social variation.

66

Manfred Bierwisch

What is the status of the principles constituting UG? We don't know how genetically determined aspects of complex behaviour are realised in the brain, let alone how the brain's relevant properties are determined by DNA structure. Nevertheless, overwhelming ethological evidence suggests that certain aspects of human behaviour have an inherited and genetically fixed basis. There is no reason to doubt that the language capacity is among them. There seem to be two logical possibilities regarding the more specific assumptions summarised in (15): (16)Nativist position: UG characterises domain-specific, genetically fixed principles that merely support the ability to acquire and use a natural language. (17) Empiricist position: UG consists of principles that emerge from the interaction of general mechanisms of association and combination with actual linguistic input. There are two versions of the nativist position. The maturational hypothesis (see Borer and Wexler 1987) assumes that components of UG become available according to the schedule of (early) ontogenetic development. Pinker's (1994) homogeneity hypothesis assumes that UG is in place from the very beginning. Although recent research into brain maturation seems to render the maturational hypothesis more plausible, both versions presuppose that UG is genetically fixed. The empiricist position (see Elman et al. 1996), on the other hand, assumes that the properties of linguistic knowledge emerge from general principles of cognitive organisation. This eliminates the need to stipulate innate conditions specifically supporting the language capacity. The problem with this position is that it either has to claim (contrary to fact) that nonhumans will acquire and use language if exposed to appropriate input or has to include the relevant disposition as a specific component of its general cognitive equipment and its genetic foundation. But this is simply another way of positing what is ultimately a biologically fixed basis for the conditions summarised in (15).

The apparent paradox of language evolution

67

3. The evolution of the language capacity Assuming that conditions like those in (15) must somehow be fixed in humans' genetic endowment, how could the language capacity have developed phylogenetically?

3.1. The paradox of adaptive

selection

The theory of adaptive evolution appears to provide a straightforward explanation. A random genetic variation determines a change in brain structure that supports the capacity to produce and comprehend signals that systematically represent complex cognitive structures. Individuals with the ability to communicate verbally have a selective advantage over competitors who lack it. Hence, adaptive selection favours the language capacity, though its genetic basis originated by chance. But this explanation contains a vicious circle. First, although it does not matter at this point whether (15) correctly characterises the basic traits of this heritage, it is crucial that the genetic basis distinctively accounts for the ability in question. Second, adaptive selection favours this ability if and only if its benefits can be exploited in actual behaviour. This requires at least a limited population to communicate with. The members of this population must therefore already have and use the capacity in question. In other words, explaining the language capacity by adaptive selection presupposes the property it attempts to explain. The theory of evolution has to cope with this problem for all cases of genetically determined social behaviour. Yet it is of particular intricacy in the case of linguistic communication because the selectional advantage presupposes not only the population whose members previously developed the capacity, but also a language - that is, a system of knowledge based on this capacity - without which the capacity would be of no adaptive value. The language capacity is an empirical fact, one that is controversial only with respect to its specificity and not to its biological foundation. So there must be ways to avoid the above paradox. In my opinion, there are at least three ways of addressing the issue. They are not mutually exclusive.

68

Manfred

Bierwisch

3.2. Avoiding the paradox The first option is to abandon the restriction imposed by adaptive selection and to assume that the language capacity's genetic foundation is due to exaptation (see section 1.2). More specifically, the language capacity and UG's principles sketched in (15) can be genetically fixed and inherited without immediate behavioural consequences. Their full potential will be realised only if appropriate conditions arise. Moreover, the genetic foundation might be a by-product of other (perhaps more general) changes like the relative growth of the brain or the modification of its architecture. This seems to be Chomsky's (1988) somewhat sceptical position regarding the adaptive explanation. His scepticism is supported by the fact that we know little about the details of genetic information, about the causal structure by which the brain is controlled by the genome, or about the way the brain effects the behaviour. All we do know is that the behaviour exhibits the specific properties illustrated above and that it is species-specific. Furthermore, we know that certain principles of linguistic behaviour (hierarchical structure or the fixing of information chunks in memory) are exploited in domains of behaviour that are not species-specific. Hierarchical structure occurs in various types of motor action, information chunks fixed in memory are crucial for visual perception, and so on. We do not, however, know what aspects of genetically fixed brain structure constitute language's species-specific domain, the domain that computes the systematic correspondence between conceptualisation and articulation. The second option is to retain adaptive selection and to emphasise the language capacity's cognitive - as opposed to its communicative - benefits. That is, the circularity is removed if the advantage of language does not depend on the behaviour of other members of the population. There is no need to assume that the capacity and its use are already in place, since the advantage only concerns the organism that exhibits the innovation. It consists of the increasing efficiency of cognition (including its far-reaching consequences) brought about by the access to conceptual structures via independently organised and memorised signals. Johann Gottfried Herder's essay on language origin (1772) stresses precisely this cognitive benefit. For Herder, Besonnenheit (reflection) is the crucial property that makes language possible. It not only enables humans to identify invariant characteristics, but also to associate them with reproducible signals. There are two reasons why Herder does not account for the language capacity. First, he is more concerned with naming objects than with addressing the combinatorial aspect.

The apparent paradox of language evolution

69

Second, he tries to explain the creation of words and takes for granted the capacity to name objects, namely "reflection". Moreover, the cognitive perspective on adaptive selection shares a significant problem with the communicative approach. In order to establish the relevant advantage, the language capacity must support the spontaneous acquisition and use of linguistic knowledge. Yet this presupposes that a linguistic system is available. True, the cognitive perspective can attribute the creation of such a system to an isolated individual, but this is an artificial assumption that requires justification. Evolutionary theory seems to offer the third and apparently most plausible option to avoid the paradox of adaptive selection. Evolution, after all, is the cumulative result of tiny steps. The vertebrate eye is the result of numerous minor changes. The same could be true of UG. There are at least two versions of this proposition. Pinker (1994) points out that UG combines a number of more or less self-contained subsystems or modules, like Morphology, Phonology, Phrase Structure, and so forth. These might well be the result of independent developmental steps, of which UG is the sum. Though this model is largely in line with generally accepted principles of evolution, it creates crucial difficulties if adaptive selection is assumed to be decisive with respect to the individual steps. In fact, Pinker runs into the same paradox I discussed above, creating even greater difficulties. Take, for instance, Morphology, the ability to systematically relate grammatical properties to (partial) conceptual interpretation. What is the adaptive advantage of such a capacity if there is, first, no grammatical system for it to improve and, second, no group of speakers whose behaviour is based on the same principles? This paradox applies to any component of UG one might single out for separate evolution. The problem is that though a marginal improvement in an organism's vision yields immediate selective advantages, there is no comparable benefit in acquiring a subcomponent of UG - unless we assume that it generates advantages in other behavioural systems. But then we would no longer be talking about the language capacity. Bickerton (1995) offers another version of the gradualist position. It differs in two respects. First, the gradual development consists of only a few stages. One of these is a so-called "protolanguage", from which the modern language capacity evolves. Protolanguage (comparable to the first stages of language acquisition or of pidginisation) is a restricted lexical system with a limited syntax for combining lexical items. Second, these stages are assumed to have adaptive value for cognition and communication.

70

Manfred

Bierwisch

3.3. Precursors? Did language emerge gradually from phylogenetic forerunners? To answer this question it is sometimes revealing to compare human languages with animal communication systems, including those - like birdsong - that are phylogenetically unrelated to human language. It is by now generally agreed that the language capacity is not a phylogenetic continuation of animal communication such as the gestural systems of nonhuman primates, though these are related to forms of nonlinguistic gestural and physiognomic communication. But these forms are all independent of the language capacity and the structure of UG. Moreover, no path leads from the ability to use a highly restricted repertoire of situationally dependent signals to the combinatorial capacity of human language. This assertion is unrelated to issues of modality, such as the question of whether, say, chimpanzees are better at visual or acoustic signals. That said, the language capacity clearly has a prelinguistic basis. I already noted that the language capacity can build on the systems abbreviated as A-P and C-I (or on their predecessors, assuming that these systems have undergone change). But I want to emphasise that they are prerequisites rather than early instantiations of the language capacity, particularly with respect to selectional conditions.

4. The origin of language From the Book of Genesis and Plato's speculations about whether the link between signifiers and signifieds is physei (natural) or thesei (conventional) to the speculations of Leibniz and Herder, the debate about language origin has focused on individual words. The origin of the language capacity is either taken for granted or confused with the origin of (the first) language. The combinatorial properties of human language have largely been ignored.

4.1. Stages of language

development

The phylogenesis of UG and the origin of a language L are manifestly different issues. But neither can be understood in isolation. This is the upshot of section 3. In particular, the different versions of adaptation as well as the radical position of strict exaptation must all deal with the origin of a first

The apparent paradox of language evolution

71

language, which eventually activated the language capacity. We lack direct evidence about both language origin and the evolution of the language capacity. Nevertheless, the relevant facts regarding language origin fall within the range of linguistic theorising, whereas this is hardly the case with phylogenetic issues regarding the language capacity. Following Bickerton (1995) and particularly Jackendoff (1999), in the remarks below I assume stages in the origin of language, stages that presumably interacted with the evolution of the underlying capacity. These stages do not correspond to modules of L - or of UG, as Pinker assumes - , but are closely related to the phenotype of language use. The scenario I describe below does not pretend to be a reconstruction of the actual developmental path, but represents a logical possibility. The necessary principles of UG must be related to properties of linguistic expressions that could be (or in fact are) realised, given the constitutive condition of mapping Articulation to Conceptualisation. This yields a conceptually necessary set of properties that can be arranged in developmental stages. Two indispensable capacities are to be identified in this respect, whose characteristic prerequisites and consequences will be articulated in due course: (18) Stimulus-free and situationally independent assignment of structured signals to conceptual representations (arbitrary sign formation). (19) Systematic, recursive combination of signs into structures supporting compositional interpretation (compositionality). Condition (18) roughly corresponds to Herder's notion of "reflection" as the source of language. It marks a decisive difference between human language and animal communication. In Bickerton's (1995) view, (18) and (19) represent two stages in the development of both the language capacity and of language. Condition (18) results in what Bickerton calls "protolanguage", which is transformed by the addition of (19) into modern human language. It might turn out, however, that arbitrary sign formation has the same condition of possibility as compositionality - in other words, that independence of stimulus control for basic expressions is tantamount to the capacity to combine them. I will leave the issue open, noting, though, that Jackendoff (1999) points out that modern languages exhibit fossilised cases of prelinguistic items like shh, hey, or ouch, which are dependent on specific situations and simultaneously devoid of syntactic status.

72

Manfred

Bierwisch

4.2. Developing a lexical system Whereas exclamations like hey or wow are only appropriate in relevant situations, words like jump or dog do not depend on a dog or a jumping action being present. The crucial point is that a structure in C-I becomes accessible by a pattern in P-A. As I mentioned above, the relation on which this accessibility rests is the major topic in the history of attempts to account for the origin of language. There are three basic proposals. First, Herder (1772) posits an act of naming caused by a salient feature of the thing to be identified. His example is a bleating sheep, which receives its name (internal bleating) by a sort of indexical relation. Second, Leibniz (1710) considers the act of naming to be based on the analogy between the shape of the name and the emotion induced by the perception of the object. His model postulates a synesthetic similarity between sound-pattern and object-sensation. The third proposal is that the links between words and meanings are conventional and arbitrary. All three semiotic types - indexical, iconic, and conventional signs - are involved in the mapping between A-P and C-I. Yet there can be little doubt that arbitrary signs comprise the overwhelming majority. More important than the origin of individual words are the general conditions and consequences of the "protolexicon": the forerunner of proper lexical systems. I will address two of them. First, the protolexicon does not have a strictly limited number of items. This distinguishes it from systems of primate calls and from all other prehuman sign systems (including the vocabulary of language-trained chimpanzees). This may or may not be a side effect of the capacity for stimulusfree sign use. The protolexicon's elasticity means that the system can expand incrementally. Its gradually increasing size and complexity may correspond to gradually increasing adaptive benefits. This avoids the theoretical pitfalls of adding up UG modules that lack independent useability. Moreover, once the protolexicon capacity is part of a population's genetic endowment, an expanding actual protolexicon can be transmitted as cultural rather than biological heritage. Second, an increasingly large set of items cannot be accommodated as global, unstructured chunks of information, but only on the basis of systematic organisation that relies on structural dimensions or features of representation. This is true for aspects of both domains linked by the lexical system: conceptual relations in C-I and articulatory conditions in A-P. This raises the question of whether primitive elements of PF and SF are geneti-

The apparent paradox of language evolution

73

cally fixed prerequisites of the linguistic structure. In Bierwisch (2000), I have argued that UG need not provide fixed repertoires, but only general conditions, on which epigenetic development then constructs the actual primes via triggering experience. In any case, the protolexicon integrates two systems that become systematically structured in their own right. This is not a post-hoc effect, but an initial condition that first enables the formation of stimulus-free pairs of form and meaning. The protolexicon's systematic nature seems to imply two other phenomena. The first is the initiation of Morphology. This involves formal features that are not (like phonetic features and semantic primes) directly based on A-P or C-I, but only have a status within the organisation of lexical items. Two stages can be identified. They build on each other logically, though not perforce developmentally. The examples below from English illustrate the two stages. Words like those in (20a) differ with respect to sex, but share the other conceptual conditions; no feature or segment in PF corresponds to the semantic distinction between male and female. By contrast, half the words in (20b) have the suffix -ess to mark the condition FEMALE; this might be said to reflect a distinction of SF in the makeup of PF. (20) (a) son : daughter; boy : girl; uncle : aunt; king : queen; husband: wife (b) actor : actress; prince : princess; duke : duchess; steward : stewardess This type of partial systematisation also shows up in other human sign systems like colour-coding in traffic signs and the use of subscripts or superscripts in formal languages. For example, the terms of (21a) are replaced in (21b) and (21c) by more explicit notation: (21) (a) x, y, z, u, ... (b) x, x', ", x"\ ... (c) x,, x 2 , x 3 , x 4 , ... This means that (20b) - like other arbitrary cases of derivation and inflection - relies on inherent properties of the capacity to organise nontrivial systems of pairs that relate form and meaning. The second stage emerges when in the correspondence /-es/ [FEMALE] the condition [FEMALE] (and the suffix related to it) acquire a formal, independent status. This status is directly accessible and supports the further systematisation of lexical elements. Once established, such a feature might be dissociated from the fixed interpretation in A-P and C-I. That is, [+ Feminine] can become a formal

74

Manfred

Bierwisch

feature of woman, girl, s/ze, and so forth as well as of items (like ship names) where the conceptual condition related to FEMALE does not apply. My proposals do not ascribe a proper morphology to the protolexicon. They merely employ present-day phenomena to indicate that elements of the sort on which morphological systems rely are available as a side effect of the structural conditions necessary for burgeoning protolexical systems. The second feature is presyntactic combination. The elements of the protolexicon allow this because they are stimulus-free and thus have an articulated structure. Due to PF's basic linearity, presyntactic combination can only amount to sequential juxtaposition. This might, however, lead to a structure in C-I that is more specific than that of the combined elements in isolation. The point is that we can imagine an elementary stage where elements of the protolexicon can be linked in the absence of specific combinatorial principles. Jackendoff notes fossilised residues of this in modern languages. For instance, compounds like steamboat, houseboat, and rowboat are formed by two elements linked conceptually according to our beliefs about the world, not according to syntactically based principles of compositionality. One could speculate about whether recurrent juxtaposition, if it corresponds to conceptual conditions taken up in SF (like in find apple, eat apple, take apple, and have apple), provides the inherent foundation for nascent compositionality, a foundation that is comparable to the extraction of grammatical features described above. Without straining plausibility, I will merely note that there are surprising presuppositions and consequences connected to the capacity to create structured lexical items that are not controlled by situational dependence.

4.3. Beyond

protolanguage

In a heavily debated report, Gopnik (1990) claims that certain grammatical deficits have genetic causes. Even if this is correct, we still don't know which aspects of the language capacity are genetically determined or how. The characteristics of Bickerton's "protolanguage" are thus speculative. This is of course equally true for the scenario I described above and for Jackendoff's (1999). But let's assume for a moment that the capacity to acquire and use protolexical knowledge is a precursor of the language capacity itself. The protocapacity can manifest itself in characteristic overt behaviour. Such behaviour consists of the use (and enables the accumulation)

The apparent paradox of language evolution

75

of shared knowledge in populations with a common heritage. But this is still different from the language capacity of recent homo sapiens. One could even argue that the big leap involves the change(s) that catapult the protolexicon capacity to UG. I will characterise three of its aspects. All are related to the central role of grammatical features, extending protolexical items (π, σ) into proper lexical items (π, γ, σ) discussed in section 2.1. One may, but need not, claim that this results from a single genetic condition emerging from a single phylogenetic change. The first aspect is the formation of Phrase Structure described in (6) and (7). It involves two conditions. First, it provides the hierarchy or bracketing that goes beyond the mere juxtaposition of words and supports specific relations that can be drawn upon by compositional semantics as in example (9). In fact, it is needed to account for simple differences like those between (22)(a) and (b). (22) (a) [Mary [asked Bill] ]

(b) [Bill [asked Mary] ]

The second aspect is the categorisation of the constituents that accounts for the different properties of, say, his sleep (nominal head) and he sleeps (verbal head). In fact, only the integration of constituency and categorisation formulated in (7) creates UG's basic combinatorial aspect. Categorisation, however, is the core effect of formal features. This means that categories like Verb, Noun, and Determiner are just bundles of formal features. Their role is to organise the correspondence between PF and SF (see Wunderlich 1996 for an overview). It need not concern us here whether the relevant properties of features like [Referential] and [Functional] can be explained in a manner similar to those for [Feminine], The point is that they determine a sort of second-order classification based on properties that do not derive directly from A-P or C-I, but from the correspondence to be established between these primary domains. The ability set up and use this second-order classification is a plausible candidate for the property that characterises the Big Leap. The second aspect comprises morphological features like [Feminine]. This may well be related to the role of formal features more generally and may assign an independent, central status to second-order classification. Indeed, morphological categories like Gender, Number, Case, and Aspect should not be considered as supplements made in modern languages, supplements that could possibly be avoided for the sake of greater simplicity (Klein and Perdue 1997 suggest precisely this on the basis of the "Basic

76

Manfred Bierwisch

Variety" of second-language learners). Morphological categories play a central role in organising the PF-SF correspondence. Let me sketch three interdependent aspects of this central role. First, it is well known that morphological features combine into rich inflectional and derivational systems. Even English, which has little inflectional morphology compared with, say, Hungarian or Georgian, displays feature combinations like (23). In these examples D stands for the category features of Determiners; Oblique and Plural stand for Case and Number, respectively: (23) (a) /him/ [D; - Feminine, + Oblique, - Plural] (b) /she/ [D; + Feminine, - Oblique, - Plural] (c) /them/ [D; + Oblique, + Plural] Inflectional systems can display remarkable intricacy, a subject beyond the scope of this paper. Second, the formal features are instrumental in establishing and regulating various types of grammatical relations imposed on the underlying Phrase Structure. One of them concerns the selectional constraints that relate lexical items to their complements, as illustrated in (24) and (25). The (semantically based) properties by which lexical items combine with complements are all specified by morphological features: (24) (a) he dreams something (b) he dreams of something (a1) *he sleeps something (b') *he sleeps of something (25) (a) *his dream something (b) his dream of something As shown by (24a) and (24b), the Verb dream requires a Subject and alternatively accepts an Object or a Prepositional Phrase. By contrast, the Verb sleep is restricted to the Subject. As demonstrated by (25), the Noun dream accepts a Prepositional Phrase, but not an Object. These conditions are based on conceptual content: both dream and sleep require an Agent, but dream has a Theme, whereas sleep does not. Nevertheless, formal features mediate their realisation. Another relation marked by formal features is the concord that pronouns require for binding to their antecedent, as illustrated in (26), where her can refer to Mary in (26a), but not in (26b):

The apparent paradox of language evolution

77

(26) (a) Mary wants him to help her/*herself (b) Mary wants to help her/herself The third aspect depends on formal features in yet another respect. Chains of positions as in (27) (see section 2.3. above) are an extension of Phrase Structure that requires one constituent to be available in two structural positions: (27) (a) I know, who he wants me to talk to. (b) [ c I know [ c who [ c he wants me [ v to talk [ P to who ] ] ] ] ] ΐ I A chain's positions are bound together by formal features. Above, [+Wh] comes with the Interrogative who, which simultaneously characterises the embedded indirect question clause. Chomsky (1995) supposes this relation to be the result of copying (or attracting) a feature from a source to a target position (the whole constituent that contains the feature being carried along). In any case, the connection between chain positions crucially depends on the availability of formal features. In addition, chain formation is ubiquitous in natural language and reconciles multiple requirements that cannot be met at the same structural position. Though it complicates the (apparently) simplest possible correspondence between PF and SF, it enables the discrepancies between conflicting conditions in Phrase Structure to be overcome. Finally, chain formation is a proper amendment to bare Phrase Structure. Like Phrase Structure and Morphology, it draws on formal features, but might result from a separate step in the evolution of UG. In this paper I have commented on the following aspects of UG via which the language capacity goes beyond the protolexical capacity and underlies stages in language origin: (28) (a) Free disposal of formal features, defining syntactic and morphological categories; (b) Phrase Structure, integrating general hierarchy formation with category assignment; (c) Morphological systems that make grammatical features accessible for selectional restrictions, agreement, concord (that is, relations that enrich Phrase Structure); (d) Formation of Chains of structural positions connected by formal features.

78

Manfred Bierwisch

One can emphasise aspects or components of the language capacity in slightly different ways. For instance, the important role played by Argument Structure and the linking of lexical heads to appropriate Complements are subsumed here under the selectional restrictions in (28c). One could easily claim a central position for this momentum. Delimiting components and subcomponents is thus a matter of argument. My remarks have of course been incomplete. I did not, for instance, say anything about the conditions that support the abstraction of elements and relations of SF from C-I. Without these, neither contrasts like behind the wall (location) and after the wall (time) nor the vast area of metaphors can be accounted for. Nevertheless, it is obvious that (28) (b) (c), and (d) are logically dependent on (28a). It addresses the centrality of formal features, which are - differing from phonetic and semantic primes - a kind of currency whose value is only fixed internally and due to the computational system that determines the correspondence between PF and SF. It is not obvious whether Categorisation in (28b) and Morphology (28c) are independent of each other. But it is clear that the complexity of Chain formation (28d) presupposes (28) (a) to (c). According to Borer and Wexler (1987), it is a prime candidate for later maturation into ontogenetic development. Perhaps it is also a relatively late addition phylogenetically.

5. Three concluding remarks First, attempting to sort out structural conditions like in (28) above by no means implies that components of G (or UG) correspond to separate neurophysiological counterparts. Kean (1992) points out that there is no direct and discrete representation of grammatical components in the brain. Moreover, this undermines speculations about simple relations between UG's properties and their genetic foundation. Yet if, as Kean suggests, a distinction can be made between brain systems for representation and those for re-representation of knowledge, then linguistic knowledge crucially depends on conditions of re-representation that mediate the primary representations of A-P and C-I. In this view, the availability of formal features and the options based on them would be among the language capacity's conditiones sine qua non because re-representation seems to be the most appropriate way of characterising the essence of formal features.

The apparent paradox of language evolution

79

Second, I have proposed how to avoid the paradox of language evolution by combining two distinct but interrelated problems, both of which have to be solved anyway: the evolution of the language capacity and the origin of linguistic knowledge. These frequently confounded issues must be clearly distinguished because they depend on fundamentally different conditions affecting the genetic heritage as well as possible knowledge based on it. But it seems that a plausible scenario emerges if they are construed to depend on each other in a non-vicious circle. The capacity to accumulate lexical items could gradually lead to a developmental stage where a random variation indeed leads to an improvement of the linguistic capacity, justifying the urgently desired selectional benefit. The stages in this type of scenario of course leave us with various unanswered questions. Suppose for a moment that Herder was basically right to suggest that "reflection" is the condition of possibility for the initial stage's stimulus-free naming. Is the combinatorial potential thus an implicit condition whose importance Herder failed to recognise? Or must we posit separate evolutionary steps that are not involved in building up comparatively complex protolexical items? In any event, it is clear that the accumulation and transmission of (proto) lexical knowledge relies on the social (communicative) dimension rather than on the strictly cognitive dimension in the development of the language capacity. Finally, the scenario's plausibility presupposes that the language capacity's evolution was gradual and that it relied on adaptive selection. This is by no means certain. In the end, perhaps Chomsky is right to have doubts about the adaptive explanation. In the terms of one of Murphy's Laws: for any complex problem there exists a solution that is simple, plausible, and wrong.

Elementary forms of linguistic organisation Wolfgang Klein What song the Sirens sang, or what name Achilles assumed when he hid himself among women, although puzzling questions, are not beyond all conjecture. Sir Thomas Browne

1. Introduction Studying the past is dull, dusty, and difficult and does not seem to provide a selective advantage. So why do we do it? The first and most obvious reason is simple curiosity. With the thrill of children exploring grandmother's attic, we dig among the material and immaterial remnants of past worlds. This is a respectable motive. After all, curiosity is at the origin of all research, of all systematic attempts to understand the world around us. The second reason is the old idea that the truth can be found in the past. It is perhaps not accidental that the study of language began with the quest for the origin of words - etymology. The idea was that things had "right" names. If you wanted find out the truth about something you had to uncover the real meaning of its name. Etymologies have since become less important. Men named "George" are no longer suspected of being peasants in disguise. Nevertheless, etymologies often seem to contain at least a grain of truth. According to Grimms' dictionary - one of the greatest scientific achievements of the Berlin Academy and of lexicographical research ever the German word Ehe ('marriage') is derived from the Germanic word aivs meaning 'eternity'. The English word ever has the same etymon. And aren't marriages supposed to last forever? The third reason is that we tend to believe that studying the past helps us to understand the present. Examining the present can tell us how things are, but not why they are as they are. This notion is not new. But it was only during the nineteenth century that it became a key concept of numerous scientific disciplines. In biology, the transition from static Linnean classification to evolutionary dynamism is a case in point. It was no less common in the scientific investigation of language. To Hermann Paul, whose 1880

82

Wolfgang

Klein

Principien der Sprachgeschichte marked the culmination of nineteenth-century linguistic thought, it was evident that there can be only one scientific approach to language: diachronic analysis. Studying the present may well uncover facts, but it cannot explain them. It is thus not truly scientific: Es ist eingewendet, daß es noch eine andere wissenschaftliche Betrachtung der Sprache gäbe, als die geschichtliche. Ich muss das in Abrede stellen. Was man für eine nichtgeschichtliche und doch wissenschaftliche Betrachtung der Sprache erklärt, ist im Grunde nichts als eine unvollkommen geschichtliche, unvollkommen teils durch Schuld des Betrachters, teils durch Schuld des Beobachtungsmaterials. Sobald man über das blosse Konstatieren von Einzelheiten hinausgeht, sobald man versucht, den Zusammenhang zu erfassen, die Erscheinungen zu begreifen, so betritt man auch den geschichtlichen Boden, wenn auch vielleicht ohne sich klar darüber zu sein. [It has been objected that there might be a scientific approach to language other than the historical one. I must refute this claim. What has been declared to be an ahistorical but nevertheless scientific approach to language is in actual fact nothing more than an imperfect historical approach - imperfect due partially to insufficiencies on the observer's part and partially to insufficiencies in the material under observation. As soon as we go beyond mere details, as soon as we attempt to grasp the relations among phenomena and to understand them, we enter the realm of history, albeit perhaps without being aware of it] (Paul 1886: 19-20, my translation).

A century later we are less convinced that a truly scientific analysis must be historical. There are explanatory factors beyond evolution, in biology as well as in linguistics. Still, the idea that to understand things we must examine how they came to be is deeply rooted in our thought. It is this idea that makes us feel that the study of how mankind came to language - or language to mankind - is more than a matter of mere curiosity. But there are two fundamental problems with this idea. First, the word "language" is used in many ways. It is consequently not clear what we mean when we talk about the origin of language. Second, there is something paradoxical about the idea that we should study the past in order to understand the present. After all, the past is past, and we have no access to it. What we have access to are its remains: bones, teeth, and petrified foot prints. And of course we have ourselves, products of past development. When we study the past to understand the present, what we are really doing is studying selected aspects of the present in the hope that this might help us to understand the past. If we want to know how language came about, there are only two paths to take. We can reconstruct it from what we have here and now as the

Elementary forms of linguistic organisation

83

result of past development. Or we can look at cases in which similar processes are occurring here and now. But are there such cases? The answer depends on what we mean by language and language origin.

2. Three notions of language and language origin Ever since Ferdinand de Saussure at the beginning of the last century, linguists distinguish among at least three types of linguistic facts. First, there are facts that are characteristic of the ability to create, learn, and use particular languages. Saussure called this ability the faculte de langage or simply langage. I will use the term "human language faculty". It is the faculty with which normal human beings are born; it belongs to our genetic endowment. This fact is beyond doubt, as is the fact that it must somehow be part of our nervous system and our body. But there are also a number of unresolved questions: 1. Is the language faculty specific to our species or is it also found in other animals like higher primates? And if so, why don't they typically use it? 2. Is it domain-specific? That is, is it just one of the many aspects of human memory and cognition in general, or is it a separate module in our brain? 3. How does it develop over one's life span? Is it fully developed at birth or does it develop over time? Does it deteriorate with age? And if so, when and at what rate? These are not easy questions. But they are sufficiently well defined to be investigated scientifically. Second, though the language faculty is part of our genetic endowment, we are not born with a language. The capacity as such is not enough. Children (and adults under appropriate circumstances) must learn a particular language like Russian, Urdu, or Kpelle - a langue, as Saussure said. A langue is a system of expressions with specific properties. Linguists disagree to some extent on what these properties are and how such a system should be analysed. But they concur on two points. First, a linguistic expression is a particular combination of a form (usually a sequence of sounds) and a meaning. Second, there are elementary expressions (words), and there are complex expressions formed by certain morphological and syntactical operations (phrases, sentences, and

84

Wolfgang Klein

texts). In short, every language has a lexicon (an inventory of elementary expressions) and a grammar (rules according to which words can be modified and put together to form extended expressions). There are also a number of questions concerning the properties of these systems: 1. What does the the most common word of the planet's most-spoken language - mean? 2. What is the function of inflectional morphology? 3. Why is possible to say the only book, but not an only book or three only books? 4. How does the meaning of the complex expression the only book result from the meaning of its parts? There are several thousands of these systems across the world. They differ considerably. But diversity does not mean that there are no shared characteristics. In fact, there may be properties - linguistic universals - that are found in all pairings of sounds and meanings. It makes sense to assume that these universals reflect properties of our innate language faculty. In other words, the properties of the human language faculty and the properties of specific languages - linguistic systems - must be related to each other. The third set of linguistic facts comprises what Saussure called parole, the actual communication between human beings in a given situation: gossiping, cursing, praying, holding lectures, describing a living room, and arguing about the origin of the language faculty. With some exceptions (like keeping a diary), linguistic communication involves more than one participant. There are a number of typical questions about parole: 1. Beyond the sound-meaning pair in question, which other components of the human mind - memory, reasoning, and so forth - play a role in communication? 2. What adaptive value does linguistic communication have for the individual and for his or her social group? 3. How does communication by sound-meaning coupling interact with other forms of human communication like facial expressions and gestures? 4. How does human communication differ from communication among bees, dolphins, or lobsters?

Elementary forms of linguistic organisation

85

These are also difficult questions. But they too can be investigated empirically, and much has been learned about them. The third type of linguistic phenomenon is the most complex. It has physical components like the acoustics of the room or the properties of the paper on which something is written. It has biological components like the participants' voice properties. It has social components like the personal relations among speakers. And it has cognitive components like the spatial knowledge needed to give directions or the planning capacity needed to construct a coherent text. What do we mean when we talk about language origin? Do we mean the origin of the human language faculty, the origin of the first linguistic system, or the origin of communication with the aid of such a system? It is important to remember that these are quite different notions of origin. The language faculty is part of our biological equipment. At some point in history, it was formed by genetic changes involving several parts of the human body. Some of the changes affected our nervous system where the central part of the language faculty is stored. Others affected peripheral organs like the larynx. We do not know when these changes occurred. And we do not know how and when they were synchronised into the capacity that is now standard human equipment. Regarding the central part of the language faculty, we do not know whether there was a single change or a whole series of them. Nor do we know whether the change or changes only produced the human language faculty or whether they are also responsible for other brain capacities. All the human language faculty needed was a language. But no such linguistic system was available, so the language faculty had to create one. This is the second meaning of language origin. The creation of the first sound-meaning coupling is very different from the formation of the human language faculty. We do not know how long the transition period was. There may have been hundreds of thousands of years between the origin of the language faculty and the origin of the first language. This raises a number of empirical and theoretical issues. First, was there one first language or did the language faculty independently create several first systems? The longer the time lag between the origin of the language faculty and its manifestation in a language, the likelier it was that several such systems were created independently. But we simply do not know. Structural equivalencies across languages are not evidence for a monogenetic origin, since they may simply reflect properties of the language faculty itself.

86

Wolfgang Klein

Second, what is the selective advantage of having a language faculty without a language? Such a major component of the brain consumes a lot of energy. And as long as it is not used - and it cannot be used if there is no language - it is anything but an advantage. A language faculty without a language would be a parasite unless it also served other functions. Alfred Wallace was the first to raise this issue for cognition in general. He proposed that the early growth of the human brain represented a moral rather than intellectual advantage. Charles Darwin did not share Wallace's rosy view, but he could not supply a satisfactory answer. In fact, he told Wallace that he was about to kill "your own and my child" (quoted in Desmond and Moore 1991: 642). And there is still no answer, at least with regard to language. Third, what did the first system (or systems) look like? We do not know. It seems unlikely that it was as complex as the first well-documented languages like Sanskrit. Maybe it was more like a pidgin. In order to qualify as a human language, it must have had a lexicon and a grammar. Is there is a "basic" lexicon or a "basic" grammar? How are they related to each other? Is grammar just a projection of the lexicon (as some linguists propose) or is it a completely different component? And is there any way to answer these questions? Fourth, how did it come into more than one head? The birth of the language faculty was a biological process, and it spread among individuals via a biological process. But the fact that the sound sequence /ho:mo:/ is associated with a particular meaning - and the fact that /ile ho:mo: pater me:us est/ is composed as it is and means what it means - is not biologically transmitted. We know it because we have learned it from others via experiential transmission: by ears and eyes rather than by genes. But this mode of transmitting linguistic knowledge cannot have applied to the first linguistic system. It was the joint creation of a social group that somehow agreed that a particular sound sequence is systematically coupled with a particular meaning and must be combined in a particular way with other sound sequences to form more complex expressions. How was this possible? Again, we do not know. But we do know that in contrast to the origin of the linguistic faculty, the origin of linguistic systems has a fundamentally social dimension. As soon as there was a first language it became possible to convert thoughts, feelings, and wishes into sound waves, to transmit them to others, and, in turn, to influence their thoughts, feelings, wishes, and behaviour. It is the use of a linguistic system that orients human beings in their environment differently from a monad in a world defined by the laws of pre-

Elementary forms of linguistic organisation

87

established harmony and from an ant in a world ruled by the rigid interactional principles of an ant colony. The verbal transmission of theoretical, practical, and situation-bound knowledge from one generation to the next sets the stage for the particular type of behaviour we consider human. It is the use of language that makes possible all higher forms of cognition as well as the characteristically human kind of interaction between members of the species. The crucial point here is that neither the mere existence of the biologically given human language faculty nor the mere existence of a first linguistic system suffices to achieve this - for the simple reason that linguistic communication involves more. It requires a complex interplay of various cognitive and social capacities. These include storing certain types of knowledge, appropriately selecting pieces of this knowledge, integrating expressions into an ongoing flow of information, and adapting to a social environment. This is the third meaning of language origin: the origin of (linguistic) communication. It presupposes the two other origins, but goes far beyond them (just as it goes far beyond other kinds of communication that do not require a linguistic system). It is only this type of language that has adaptive value. Linguists are primarily interested in the second notion of language: in the structure and functioning of linguistic systems. In fact, linguists cannot contribute much to answering questions about the origin of the language faculty, unless it is by pointing out which properties biologists and brain researchers ought to consider. The knowledge of a few signs does not require something as complex as the human language faculty. Dogs and cats have them or can learn them. So what is it that distinguishes human languages from other systems of communication?

3. Lexical repertoire and rules of composition Two notions seem uncontroversial among linguists. First, there must be a set of elementary expressions (lexemes). Second, there must be rules of composition that prescribe how complex expressions are formed from simpler ones. In other words, there must be lexicon and a grammar. This applies to all manifestations of the human language faculty, from elementary learner varieties (like the language of beginning second language learners or of pidgin speakers) to fully fledged languages like Latin or English.

88

Wolfgang Klein

3.1. The lexicon A lexeme (word) is a cluster of at least three types of features: semantic features that indicate an expression's lexical meaning (or "lexical content"); phonological features that describe an expression's phonological shape; and categorical features that characterise an expression's behaviour with respect to rules of composition. Let us consider the English lexeme horse. The visual shape on the paper is not the word, but rather it is one way of representing the word. The word itself is a cluster of features. They include the phonological information /h :s/, the semantic information "equine quadruped", and the categorical information "is a noun". Horse is a simple example, and things can get more complicated. In particular, other properties - graphematical features - may be linked to a lexeme. But they are not crucial (written language is a relatively late invention, and many, if not most, languages still lack it). There are many lexemes whose semantic properties are far more abstract than horse. We might consider the English morpheme the, the most frequent lexical item on earth. No linguist has managed to give a precise and clear characterisation of what it means. Bertrand Russell, for example, begged his readers not to reject his theory for its "apparently excessive complication" until they themselves had "attempted a theory ... on the subject of denotation" (Russell 1905: 493). In addition, linguists largely agree that in some specific cases semantic features can be entirely absent, as in the case of there in there is a slug in the salad. Similarly, phonological information can be absent like in the case of socalled "empty elements". What seems indispensable are categorical features: each lexical item must contain information about how it can be integrated into larger constructions. The first task the human language faculty must be able to perform is to create a lexicon. This means that it must be able to sort out these three types of features (semantic, phonologic, and categorical), to cluster them in some fashion, and to store them somewhere in the brain. There are two ways to achieve this: either by copying an existing repertoire or by creating new clusters. Today, with so many languages available, the first way predominates. It constitutes the lexical part of language acquisition and is an extremely complicated process, many aspects of which are still a complete mystery (see Clark 1993). Children or adult learners are not confronted with words, but with more or less continuous sound streams that are initially baffling. They must break such sound streams into smaller segments and associate these segments with semantic features and - what is more difficult -

Elementary forms of linguistic organisation

89

categorical features. This process, though little understood, is highly efficient. Children learn thousands of words within a few years. And although the capacity to learn new lexical items seems to deteriorate with age, it is hardly ever lost entirely. When our ancestors had the language faculty but no language, they had to invent lexemes. No one knows how they did it. But it was and still is a fertile field for speculation (see the survey in Kainz 1967: 267-334). Did the first lexemes express emotions like pain or fear? Did they imitate natural sounds? It is difficult to imagine how one might arrive at even a simple lexeme like tree. We simply have no historical evidence. But we can examine instances where the human language faculty still creates new lexical items. These cases are admittedly infrequent. Children sometimes create lexical items that are not based on existing words. Psycholinguists invent nonce words for experimental purposes. Pharmaceutical companies design names for new products. These names are often based on Latin or Greek roots, but are just as frequently new coinages. In German, the lexeme hungrig ('hungry') has an antonym that describes the state of having had enough food. The lexeme durstig ('thirsty') lacks such a counterpart. German has a lexical gap. We could of course fill it by a new lexeme, say schwock. (English even lacks an antonym for hungry.) What is characteristic of all observable cases of lexeme creation is that they begin with semantic features and associate them with phonological features. Was this also true for our earliest ancestors? It should be true unless we assume that the human language faculty itself has changed. But if the human language faculty has changed, then it is hopeless to say something about its origin. So if we want to understand the past, we must study the present. The creation of new lexical items is perhaps not the best example for this strategy. Children perfectly imitate the language around them. And when adults create a new item they do so in the context of a language or languages they already know. Nevertheless, I believe this is the only realistic approach. So far I have not addressed the way categorical features are added to a lexeme. It is a complete mystery. Most linguists believe that semantic and phonologic features can be absent from a lexeme. Categorical features, however, must be present. These features reflect the combinatorial side of a linguistic system: its grammar.

90

Wolfgang Klein

3.2. Rules of composition Grammatical rules are traditionally divided into morphological and syntactical rules depending on whether they operate within or go beyond individual words. There are a number of borderline cases (just as there are borderline cases between lexicon and grammar). The most salient example of morphological rules is inflection. The Western linguistic tradition has frequently equated inflection with inflectional morphology. This is because the languages that stood at the beginning of this research tradition - Greek and Latin - have rich inflectional systems. The first grammars of comparatively modern languages like English and German readily adopted this "morphology bias". The fact that they had less elaborate inflectional systems than classical languages was generally seen as a symptom of decay. This view is today obsolete. Inflectional morphology is a common but by no means indispensable aspect of languages. In other words, there is grammar beyond inflection. But as strange as the notion of grammatical decay may seem, it is a fact that to the extent to which we have historical records languages tend to reduce or to eliminate inflectional morphology rather than to expand it. There are some exceptions, such as the formation of future marking in Romance languages (aimerai from amare habeo). Although frequently referred to, these cases are rare and do not affect the overall picture. English, Dutch, and even German show very reduced inflectional systems when compared with their common Westgermanic origin, let alone when compared with older stages of Indo-European. As August Schleicher pointed out many years ago, English had is the modern equivalent to Germanic (more precisely Gothic) habaidedema. Chinese, the paradigmatic example of a language without inflection, is assumed by many scholars to have had suffixes at one time. But they are gone and have only left behind traces in the form of lexical tones. So, despite some exceptions, languages seem to develop away from morphology. This fact led Otto Jesperson to assume that in the beginning languages were much more complex than they are today: "we must imagine primitive language as consisting (chiefly at least) of very long words, full of difficult sounds, and sung rather than spoken" (Jesperson 1921: 421). His theory is based mainly on the development of languages for which we have written records. He sums up his argument with a general "law of development": "The evolution of language shows a progressive tendency from inseparable irregular conglomerations to freely and regularly combinable short elements" (Jesperson 1921: 429).

Elementary forms of linguistic organisation

91

This notion is counterintuitive. It is hard to imagine that our ancestors who designed the first linguistic system immediately created something as morphologically complex as Sanskrit. It seems more likely that the first systems had no complex morphology and only very elementary combinatorial rules. Again, the only way to find out is to examine situations in which the human language faculty does not copy an existing system but somehow creates elementary combinatorial rules. The distinction between morphological and syntactic rules is generally accepted in modern linguistics. There is also agreement that the latter are in a way more fundamental. After all, there are linguistic systems that keep lexemes intact, but there are no languages that do not allow the formation of expressions consisting of several separate lexemes. The next point is more controversial. I would like to make a rigid distinction between two types of rules of composition. There are rules that operate on lexical contents, and there are rules that serve to integrate an expression into a given context. I will call the former LC rules and the latter CI rules. LC rules serve to form complex lexical contents from simple ones. By doing so, they affect an item's semantic, categorical, and phonological features. For example: 1. The accusative singular of German nouns of paradigm class five is formed by attaching -n. This is a morphological rule based on categorical features. 2. A lexeme of the type "determiner" and a lexeme of the type "noun" form an expression of the type "noun phrase". This is a syntactic rule based on categorical information. 3. The constituent that expresses the agent comes first. This is a syntactic rule based on semantic features. Compositional rules can also - perhaps only - use the semantic information provided by lexemes rather than by categorical features. There are also compositional rules (such as French liaison) that only affect phonological information. All that matters is that they are stated in terms of the information provided by the lexemes involved. Let us turn now to the rules that integrate the utterance into the ongoing flow of contextual information. Such information may stem from preceding utterances, from the speech situation, or from the speaker's or listener's general world knowledge. Typical CI rules might be:

92

Wolfgang Klein 1. Focus constituents come last. 2. Lexemes that preserve information from the preceding sentence come first. 3. Lexemes that preserve information from the preceding sentence are de-accented. They also include rules regarding the communicative function, in particular a sentence's "illocutionary force" when it is uttered in a certain communicative context: 4. A question is marked by a final rise 5. An assertion is marked by having the finite component of the verb in second position 6. An imperative is marked by a bare stem in initial position.

Rules like the six above are not based on lexical information. After all, nothing in the meaning of the lexeme come says that it should be used as a question, an assertion, or an imperative, just as nothing in the lexical information of this lexeme tells us whether, in a given utterance, this information is new or retained from a preceding utterance. The distinction between LC rules and CI rules is not absolute. It does not preclude that a given language will cluster bits and pieces of both types together into a single complex rule. In fact, the apparent opacity of "fully fledged" languages is often due to such clustering, whereas the distinction is clearer in more elementary manifestations of the human language capacity. In the latter, we seem to have very simple rules such as "agent first" (an LC rule) or "focus last" (a CI rule). The problem is that under specific communicative circumstances the two types of rules come into conflict. Hence, when applied simultaneously, they do not allow the formation of a complex expression. Such cases call for additional devices. Indeed, this may be the driving force behind the development of linguistic systems that go beyond mankind's first languages. There is another reason why the distinction between LC rules and CI rules is important in the current context. It may be that these two types of compositional rules belong to different components of the human language faculty. In fact, they may have developed at different times. There was consequently not just a single period during which this faculty developed. We intuitively assume that the capacity to integrate meaningful stretches of sound into the ongoing flow of discourse precedes the capacity to build complex constructions based on the categorical information of lexemes. But we do not know for sure, and there is no easy way to find out. Again, all we can do is examine how this faculty currently functions when it goes about developing elementary linguistic systems.

Elementary forms of linguistic organisation

93

4. The Basic Variety If the human language faculty has undergone substantial changes since the days when it created the first linguistic systems, then there is little hope of finding out what it was like in those days or what these first systems looked like. If it has not undergone substantial changes, then we should examine how it currently operates when faced with a similar task. Such situations seem rare; the case which comes closest to them is language acquisition. We all are born with the language faculty, but not with a particular language. Hence, while mankind has many languages, each newborn has to go through a "latent language faculty" period in which he or she has the faculty but not the linguistic system. This case is substantially different, since today our innate faculty does not have to invent lexemes and rules of composition, but rather copy them from an existing language. This is also true for second language acquisition (SLA). But isn't it at least possible that the process of language acquisition displays aspects that are independent of the particular language being learned? No one seems to have made this claim for first language acquisition. As a rule, children can copy even the most irrelevant details of the language they are learning. True, there might be creative aspects to the acquisition process, but they are not obvious. The language-copying component of children's language faculty seems to outweigh - or render superfluous - the language-making component. This may be different for adult second-language learners, since they are less able or less willing to copy what they hear. The linguistic systems adult learners develop - their learner varieties lag considerably behind the target language they are trying to replicate. This may be because they have already mastered a language - their source language - or because of changes in their learning capacity. Teachers, linguists, and laymen tend to view these learner varieties as imperfect replications of the target. But to the extent that such systems also exhibit properties that are independent of the source and target languages, we must assume that they reflect creative processes of the underlying human language faculty. Is there such an overarching linguistic system whose structural properties are independent of a particular language? In a large cross-linguistic and longitudinal research project, we examined how 40 adult learners picked up the language of their social environment via everyday communication (Perdue 1993 contains a detailed description of this project, which involved about fifteen researchers across Europe). Their production was regularly recorded and analysed over about 30 months. This production and the way it evolved varied in many respects. But it also manifested a number

94

Wolfgang Klein

of striking similarities. One of the core findings was the existence of a special language form that we called the Basic Variety (Klein and Perdue 1997). It was developed and used by all learners, independent of source and target language. About one third of the learners fossilised at this level. This means that, minor variations aside, they only extended their lexical repertoire and learned to make more fluent use of the Basic Variety. But they did not complicate their utterances in other ways, particularly with respect to morphology or syntax. We believe the Basic Variety not only plays a particular role in the SLA process, but also that it represents a particularly natural and transparent interplay between function and form in human language. In a way, fully fledged natural languages are only elaborations of the Basic Variety. They add specific devices like inflectional morphology or focus constructions. They also add numerous decorative elements: pleasant to the ear, hard to learn, and faithfully transmitted from one generation to the next. But fully fledged languages essentially build on the same organisational principles.

4.1. The Basic Variety's lexicon There is no inflection in the Basic Variety, hence no morphological marking for case, number, gender, tense, aspect, or agreement. Typically, a single form corresponds to the stem, the infinitive, or the nominative in the target language. But it can also be a form that would be an inflected form in the target language. Sometimes a word appears in more than one form. Such variation does not seem to have a function; the learners simply try out phonological variants. The Basic Variety's lexicon varies in size and origin. Normally, it expands steadily during the acquisition process, though this increase varies considerably among learners. The main source is normally the target language. There are also numerous borrowings from the source language. Interestingly, the composition of the lexicon is remarkably constant among all learners. It essentially consists of a repertoire of noun-like and verb-like words as well as a few adjectives and adverbs. The pronoun system is extremely elementary. It includes minimal means to refer to speaker, listener, and a third person (functioning deictically and anaphorically). There are a few quantifiers, a word for negation, and a few prepositions with over-generalised lexical meanings. There are no subordinating conjunctions. In other words, the repertoire consists mainly of openclass items and a few closed-class items with lexical meaning. There are some determiners (in particular demonstratives), but hardly ever a determiner system. And there are no expletive elements like the English existential there.

Elementary forms of linguistic organisation

95

What can these findings teach us about the language-making side of the human language capacity? Not much. Most items are replications of something that already exists. What is telling, though, is the absence of some items that are typically found in fully fledged languages, notably semantically empty elements and closed-class items. Both facts are related to categorical properties, that is, to properties that link lexical items with rules of composition.

4.2. The Basic Variety's rules of composition How do speakers of the Basic Variety integrate their repertoire of lexemes into full utterances? The first and most salient point is the already noted absence of verb or noun inflection. There are no morphological rules (though there are some noun-noun compounds). But speakers of the Basic Variety are able to construct more complex utterances. The structure of these utterances is determined by the interaction of three types of constraints. First, there are absolute, "phrasal" constraints on the form and relative order of constituents (these are LC rules based on lexemes' categorical features). Second, there are "semantic" constraints relating to the case-role properties of arguments (these are LC rules based on lexemes' semantic features). Third, there are "pragmatic" constraints relating to the organisation of information in connected text (these are CI rules for the introduction and maintenance of reference as well as for topic-focus structure). The phrasal constraints observed in the Basic Variety allow three basic phrasal patterns with some subvariants (the subscripts correspond to differences in their possible internal structures): PHI a. PHlb. PHlc.

NPj - V N P j - V - NP 2 N P j - V - N P 2 - N P2

ADJECTIVE PH2.

NP, - COPULA - NP2 PREPOSITIONAL PHRASE V

PH3.

np2

COPULA

96

Wolfgang Klein

All patterns may be preceded or followed by an adverbial phrase (normally one of time or place) or by the conjunction and. The phrasal constraints impose narrow restrictions on possible sentence structures. But a pattern such as NP - V - NP does not mean that the first NP is the "subject" and the second NP is the "object". In fact, it is not easy to define these notions in the Basic Variety except by their apparent similarity to target or source language utterances. So which argument takes which position? We found that a semantic principle obtains based on the control asymmetry between referents of noun phrases. One can rank each argument of a verb by the greater or lesser degree of control that its referent exerts or intends to exert over the referents of the other argument(s). In the English sentence Clive sliced the salami, Clive ranks higher on the control hierarchy than the salami. The semantic constraint is: SEMI. The NP referent with the highest control comes first (controller first). Strength of control ranges from clear agent-patient relations (verbs like kick and push) to weak asymmetries (verbs like kiss and meet), to complete absence (as in copular constructions). Some verbs, notably verbs of saying and giving, take three arguments (four arguments are never observed in the Basic Variety). These verbs are regularly of the "telic" type; that is, their lexical meaning involves two distinct states (see Klein 1994: 79-97). It is crucial that the control relation between the various arguments is not the same in both states. In an utterance like George gave Eva a book there is a first state (the "source" state) in which George is "in control o f " the book and is active in bringing about a distinct state (the target state). In the target state Eva rather than George is "in control o f " the book. The control status of the NP that refers to the gift is low in both states. The principle "controller first" thus requires that this argument not come first. It does not prescribe, however, whether the controller of the source state or the controller of the target state comes first. "Controller first" must therefore be supplemented by an additional constraint defining the relative weight of source and target state in determining word order: SEM2. Controller of source state outweighs controller of target state.

Elementary forms of linguistic organisation

97

This principle also applies analogously to verbs of saying if we assume that the control of information changes in both states. There is one referent who controls the information in both states, and another referent who controls the information in the target state, but not in the source state. Thus, the speaker comes first, the hearer comes second, and what is said comes last. The two control constraints are not always operative, either because there is no asymmetry between the NP referents, or because the verb has only one argument. In these cases, the NP's position depends on how information is distributed across an utterance in context - that is, on pragmatic factors. The Basic Variety has two types of pragmatic constraints. They relate to information status - which information in the utterance is new and which is retained from the preceding utterance(s) - or to the topic-focus structure. These two factors must be kept distinct, although in practice they often appear together. The topic-focus structure reflects the fact that part of the utterance defines a set of alternatives (the topic) and selects the appropriate one (the focus). For example, the utterance Eva ate an apple can answer at least three different questions: (1) Who ate an apple? (2) What did Eva eat? (3) What did Eva do? In (1), the alternatives are the persons who could have eaten an apple (the topic) and the person specified by the NP Eva (the focus). In (2), the topic is the set of things that Eva could have eaten, and apple specifies one of them (the focus). In (3), the set of alternatives comprises all the events involving Eva that could have occurred on that occasion, and the verb phrase specifies the one selected from this set (the focus). (See Klein and von Stutterheim 1987 for more complex cases in which both factors - information introduction and retention as well as topic-focus structure - are combined in individual utterances and entire texts that constitute an answer to a quaestio: an explicit or implicit question.) Fully fledged languages can mark an expression as a focus or topic expression by specific devices that include intonation, clefting, and special particles. The Basic Variety mainly uses word order: PRAG. The focus expression comes last (focus last). If there is only one argument, then this argument has a semantic role. But there is no semantic role called "asymmetry", and so the controller constraints cannot apply. Hence, only PRAG and phrasal constraints interact. If the referent of the NP is topical, then pattern PHI is used; if it is in the

98

Wolfgang Klein

focus, then pattern PH3 is used. The same constraint stipulates the NP's position in copula constructions. Our model has the advantage of explaining word order without resorting to ill-defined notions like "subject" or "object". It also explains the "topic status" often associated with the notion of "subject". The other pragmatic factor that influences utterance structure is the "given-new distinction": is what an expression refers to retained from a preceding utterance or is it new? This distinction actually interacts with the topic-focus status. It does not, however, result in a simple word order rule like PRAG, but rather in different types of NPs. These, in turn, are restricted to certain positions indicated by the numbers in the phrasal rules PH 1 to PH3 noted above. Here we find some limited variation within the Basic Variety. In particular, we find some numerals and (rarely) a definiteness marker, usually a demonstrative. We indicate this in the following diagram by optional DET. As a rule, however, nouns are unadorned. This gives us the following main types: NP 2 NP t proper name proper name (determiner) noun (determiner) noun pronoun Zero (item without phonological features) The choice among these forms depends on whether a referent is introduced or retained and whether the referring expression is in topic or focus. The most general opposition is between use of a lexical noun (or proper name) and Zero (or a pronoun). The latter is used exclusively to maintain reference in the context of a controller moving from topic to topic in successive utterances. Maintaining semantic role and position (controller first) is thus not in itself sufficient to licence Zero where there are two potential controllers in the previous utterance (and is a further indication that "subject o f " is not a Basic Variety function). With names and lexical nouns, position is the sole indicator of the referent's topic/focus status. It follows from the observed distribution that reference maintenance in focus cannot be achieved by pronominal means. So there are clear constraints on how things can be expressed in the Basic Variety, and where, consequently, its speakers might run into problems. These problems are a major source of structural complexities.

Elementary forms of linguistic organisation

99

5. Conflicts The Basic Variety is a remarkably elegant and versatile system. Its structure seems to be independent of the learners' source languages (which in the study were Finnish, Punjabi, Italian, Turkish, Moroccan, Arabic, and Spanish) and target languages (which in the study were English, German, Swedish, Dutch, and French). It should thus reflect properties of the learners' underlying language faculty. It actually offers a number of advantages over fully fledged languages. It lacks irregular verbs and other inflectional nuisances. But problems arise when its neat principles come into conflict. The clearest case we noted was in describing a scene from a film. In order to control the learners' message content, we had them watch and describe scenes from Charlie Chaplin's Modern Times. In one of the scenes a girl is accused of stealing a loaf of bread. In the "German" version of the Basic Variety that is, in the version that primarily uses lexical material based on German - this can be easily described by: (4) Mädchen stehle Brot. There are two nominal arguments. The first is the controller, the second is focussed. These three rules taken together result in an utterance like (4). But the film's plot becomes more convoluted. The speaker now has to express that Charlie (not the girl) stole the bread. The speakers produced: (5a) Charlie stehle Brot. (5b) Brot stehle Charlie. In (5a) the speaker violates the pragmatic constraint PR1 because Charlie is focussed and so should be in final position. In (5b) the speaker violates SEMI because Charlie is the controller and so should be in first position. The Basic Variety breaks down in such cases. There are two ways to rectify the problem. The first consists of ranking the two principles: Semantic constraints outweigh pragmatic constraints. I suspect that native speakers of English have such a ranking principle. They would infallibly consider the first argument to be the controller. Sentence (5b) thus sounds bizarre to a native speakers of English, but much less so to native speakers of German, since in German the controller might easily be in final position. Hence, if there is ambiguity they tend to follow the opposite ranking. Nevertheless, one of the constraints is violated no matter which ranking is chosen. If we adopt the English strategy, it is not clear which argument is in focus. If we adopt the German strategy, it is not clear which element is the controller (though here it is unlikely that the bread is the controller).

100 Wolfgang Klein

The other way to solve the problem is to invent an additional device that allows the speaker to mark either what is in focus or what is the controller. Both the target and the source languages constrain the inventiveness of the Basic Variety. By contrast, the first homines sapientes, though in principle in the same situation, were without a paradigm. They had to invent something new. Natural languages seem to have either used supra-segmental means or to have created a specific segmental expression (a morpheme) to serve as the additional device. This morpheme may be free or attached to one of the relevant words. Supra-segmental devices are widely used for this purpose. But to the best of my knowledge they only mark an expression as focussed or non-focussed but never as agent, patient, or the like. They serve CI functions, not LC functions. The other choice, the formation of a specific morpheme, offers both options. It is possible to invent or adopt a morphological focus marker (or a non-focus marker), and it is possible to invent or adopt a morphological controller marker, a patient marker, and so on. In SLA, the first possibility is exemplified by some learners of French who use a particle [se] to mark an element in initial position as focussed. This particle is a precursor of the cleft construction c'est... que (see Klein and Perdue 1997: 330). The other possibility is tantamount to case marking, either by inflection or by some free morpheme. The controller can thus be marked by a special suffix, the non-controller by a different suffix, non-focus by still another suffix (thus indicating something like "topicness"), and so on. It may be that the relevant marking only occurs when at least two arguments are present (otherwise no confusion arises), but it is also possible that the case role is marked in all occurrences regardless of whether there is a second argument with which it can be confused. In the case of language acquisition, the learner is not free to choose among these various options and to build his or her own system. Eventually, the learner has to copy what the social environment does regardless of whether he or she really understands it. Adult learners may be somewhat reluctant to do this if they find it difficult and if they don't see the point. This may be one of the reasons why they often get stuck at a certain stage of proficiency. Children normally do not get stuck. This may be because they are better or more willing imitators of things they do not understand. Our ancestors, who first invented inflectional morphology (including case marking), were not influenced by an already existing system. But we have no direct evidence of what they did. All we have are the results of a long pro-

Elementary forms of linguistic

organisation

101

cess of transformation, elaboration, and reduction. A fully fledged language and its inflectional morphology resemble an old city in which many generations have left their traces. This explains many of the architectural oddities of modern languages. But it does not preclude them from having had quite systematic foundations.

6. Conclusion As Immanuel Kant said two centuries ago, it is the fate of the human mind to be haunted by questions it cannot answer. He did not mention the quest for the origin of language. But it seems like a good candidate. As William Dwight Whitney, the most eminent American linguist of the nineteenth century, put it in 1873: No theme in linguistic science is more often and more voluminously treated than this, and by scholars of every grade and tendency; nor any, it may be added, with less profitable result in proportion to the labour expended; the greater part of what is said and written upon it is mere windy talk, the assertion of subjective views which commend themselves to no mind save the one that produces them, and which are apt to be offered with a confidence, and defended with a tenacity, that are in inverse ratio to their acceptableness. this has given the whole question a bad repute among sober-minded philologists (quoted in Jespersen 1921: 412).

This is not encouraging. Still, I do not think, nor did Whitney, that the issue is necessarily beyond all reasonable research. But it is if no clear distinction is made between various notions of language and, as a consequence, various notions of language origin. First, there is the question of how the human language faculty came into existence. It must have been a complicated biological process involving changes in the brain and in some peripheral organs. We do not know when and how all of this occurred or how it came together to create this remarkable faculty. Once available, this faculty had to create the first linguistic system or systems. This is what nineteenth-century linguists usually understood by the origin of language. This also must have been a complex process involving the creation of a lexicon and of various types of compositional rules. It was not until such linguistic systems were available that communication by means of them was possible. This is the third type of language origin and is the sense in which, for example, Herder proposed that language had its origin in poetry.

102 Wolfgang Klein

The coming-into-existence of linguistic communication is probably the most problematic among the three notions of origin, since it involves the interaction of so many different parts of our cognition. The first point I tried to make in this paper was the need to keep these three notions distinct. The second point relates to the way in which we have access to these origins. All we can study is the present: the more or less elaborate relics of the past and those aspects of past processes that we still can observe. Today we rarely experience the birth of a new cognitive capacity comparable to the human language faculty. This seems to eliminate the second method. But we can experience new forms of communication, such as the introduction of written communication in oral speech communities or the rise of internet communication. These are interesting and important developments with considerable social consequences. But compared with the origin of linguistic communication they are only elaborations of an existing theme. From the linguist's point of view, the second notion of origin is the most interesting. How did and does the human language faculty create linguistic systems? Its typical task today is to replicate existing systems. It enables children and adults to learn languages. But there is evidence that its language-making ability is not lost, both with respect to the creation of lexemes and of rules of composition. And if we want to understand how it worked before, we should try to find out how it works now.

From potential to realisation: an episode in the origin of language Bernard Comrie 1. Introduction This is a speculative paper, dealing with one facet of the origin of human language. 1 While I make no attempt to hide its speculative nature, I believe that we do now have available to us empirical material that bears on the question I address and that can serve at least to narrow down the range of possible answers. But first, I wish to introduce Dumbo the flying elephant, as this will lead directly into the main question to be posed. It might not be immediately clear what Dumbo the flying elephant has to do with the origin of language. Those familiar with the story of Dumbo will recall that he was able to fly, but that for the earlier part of his life he was unaware that he had this ability and did not in fact utilise it. When placed in a situation where his life depended on being able to fly, he made the attempt and flew. This is, of course, fiction, but there is an analogy that carries over to one aspect of the origin of language. Let us suppose that a particular creature has the genetic ability to carry out a certain behaviour. What factors in the environment are necessary to bring the creature in question actually to manifest that behaviour? More specifically, I wish to address this question with respect to language. I will assume a point in time at which human beings had the necessary genetic ability to create language. I will remain neutral as to how detailed that genetic ability is, that is, as to the division of labor between nature and nurture in the development of any particular language in its speaker(s), though I will assume, probably uncontroversially, that some of the ability to speak a particular language is given genetically and that some is determined by the linguistic environment. But I am not here concerned with how human beings came to have whatever genetic endowment they have in order to make language acquisition possible. Given a human being with such genetically determined linguistic ability, what circumstances are necessary for this ability to be realized? There are many possible answers, and there remains much room for speculation about the correct answer or range of answers. Nonetheless, it seems that empirical

104 Bernard Comrie work, especially in recent years, may enable us at least to narrow down the range of possible answers. My original interest in this question arose somewhat indirectly, namely from an interest in whether all human languages share a common origin. A clear way of disproving the monogenetic thesis would be to find one that has a demonstrably different origin. This then feeds into the question of the circumstances under which such a language could arise. In the present article, I concentrate on this latter question, although the answer to this question will, of course, have implications for the probability that human languages have a common origin. For the lower the likelihood of a new language arising ex nihilo in a human, the greater the likelihood that all languages have a single origin, since the single-origin hypothesis would require only a single occurrence of what I would like to call the Dumbo Factor, that is, the recognition on the part of a member of the species that communication by language is possible and the practical realization of this means of communication. However, nothing in this paper will come even close to answering this question.

2. The range of scenarios At one extreme, we may consider ordinary child language acquisition. The child is a member of a particular speech community and grows up to speak the language of that community. While the child goes through acquisitional stages that are at times very different from that of the adult community, there is nonetheless remarkable convergence on the adult language goal. In particular, under normal conditions, children do not (with the possible exception of twin languages; see section 4.2) create their own languages as a long-term solution to the problem of their communication needs. (I deal briefly in section 4.1 with constructed languages, such as Esperanto.) An important additional factor is that successful language acquisition normally requires that the acquisition take place within a quite narrow developmental window. Indeed, one can go even further. Under these normal circumstances, it seems that normal children cannot help but acquire the language of their speech community; in other words, whatever role instruction and correction may play in fine-tuning the child's linguistic knowledge, the basic linguistic knowledge is independent of such explicit guidance. One might therefore wonder whether a child would not come up with a language - not of course

From potential

to realisation

105

that of any existing speech community - even in the absence of input. This was, for instance, the view of Johann Gottfried Herder, whose 1772 publication proposed that simply by virtue of being human, a human being will, even in isolation, come up with human language as a communication system. (It presumably also underlies experiments like that attributed to King Psammetichus of Egypt, who, as related by Herodotus, caused two children to be reared in isolation to see which language they would end up speaking; this presupposes that they would end up speaking a human language and goes further in also presupposing that there is some particular language - according to Herodotus, their first word was the Phrygian for 'bread' - that they would speak as a default.) One can add a related observation. Whatever the origin of human language, we know of no human community that lacks human language, even though there are human communities that lack most of the other abilities that distinguish humans from nonhumans. This applies equally to what is probably the most isolated known human group, namely the Tasmanians, who were isolated from the rest of mankind for over 10,000 years: from the flooding of the Bass Strait, which cut Tasmania off from the mainland of Australia, until the arrival of European explorers and in particular settlers at the beginning of the nineteenth century, at which time the Tasmanians numbered about 3,000 to 5,000. Although little is known of the languages of the Tasmanians, as a result of the genocide visited upon them by early settlers, it is clear that they did have language. We thus know of no human community that has abandoned language as its basic means of communication. 2 The kind of test cases that are particularly relevant to our question are thus those where children are exposed to restricted input, where "restricted input" can vary in different degrees from the kind of input to which normal children are exposed under normal circumstances. In what follows, I will be interested in the possible acquisition of a communication system that is comparable in complexity to what is found among normal speech communities. Bickerton (1990) introduces a concept of "protolanguage", exemplified for instance by pidgins, and perhaps by the communication systems acquired by apes in captivity, which can serve as rudimentary communication systems but which lack the complexity of normal human languages, in particular complex syntax. I am concerned specifically with the acquisition of language, not of protolanguage.

106 Bernard Comrie 3. Restricted input In this section, I will examine three sets of circumstances under which it is the case, or might be the case, that children are exposed to restricted input, and examine the consequences for language acquisition.

3.1. Feral children and related cases In the most literal sense, a feral child would be one who has been deprived of all human contact. Whether such a child would ever survive is doubtful, so actual cases come close to the literal sense rather than exactly matching it. Unfortunately, most supposed cases of feral children are woefully lacking in adequate documentation, certainly in the kind of documentation necessary for a scientific evaluation. But there is one reasonably well-documented case, that of the twins Kamala and Amala, described by the Reverend J.A.L. Singh in his diaries from 1920 to 1929 and published in Singh and Zingg (1942: 1-118). The two girls Kamala and Amala were found, apparently uncared for and certainly unable to speak any human language, in 1920 and taken into the care of the orphanage attached to Singh's mission station. Kamala's age was then estimated at eight years, i.e., well into the period at which a normal child under normal circumstances would have been acquiring language, and Amala's at one-and-a-half (Singh and Zingg 1942: 11), namely, a time when language acquisition might well not yet have manifested itself, even in a normal child under normal circumstances. The fact that two children are involved is potentially interesting, since they do constitute a potential minicommunity, in other words, one in which whatever social circumstances are necessary for language development might have been met. However, given Amala's probably young age, and the fact that she died the following year, it is doubtful whether this potential interest could actually have been expected to be realized. Kamala remained with Singh's mission until she died in 1929. Singh's diary is not primarily concerned with her linguistic development. As a clergyman he was much more interested in her moral development and to some extent in her physical behaviour, for instance the (partial) shift from quadrupedal to bipedal gait. Nonetheless, the earlier part of the diary does contain a certain amount of linguistic information. As the diary progresses, unfortunately, the linguistic information becomes increasingly

From potential to realisation

107

sparse and general, so that I do not feel that I have anything like a clear grasp of her linguistic abilities towards the time of her death. But from what Singh's account does say about her language, it seems that she acquired a pidgin-like competence in Bengali, the language of Singh's mission, without ever progressing beyond this. Moreover, there is no clear discussion of her general mental abilities, so the possibility cannot be excluded that language development might have been impaired by more general cognitive deficiencies. If Kamala did have a normal general mental level yet failed to progress beyond the protolanguage stage, then in one case where it is reasonably clear that a child had no or minimal linguistic input up to the age of about eight, that child did not subsequently succeed in acquiring a knowledge of language comparable to that of a normal child under normal circumstances. If the estimate of Kamala's age at about eight was correct, then this would suggest a rather narrow time window within which language acquisition must at least start. (However, in response to a question from a medical doctor, Singh adds in a footnote to the published version of the diary [Singh and Zingg 1942: 11] that the age estimates were based on guesswork, in which case poor nutrition or other factors could have led to an abnormally smaller body. The only medical evidence concerns tooth eruption, but even here the footnote says that no systematic record was kept.) A related case is that of Genie, described with scientific accuracy in Curtiss (1977). Genie (this is a pseudonym) was discovered at age 13 after having been kept imprisoned and isolated from exposure to language from about age one-and-a-half. At the time of her discovery, she was incapable of speech, thus providing even more direct evidence that absence of all input will probably lead to the absence of language development. However, in Genie's case one cannot exclude the possibility that general maltreatment might also have been a factor. But on examination, her general mental abilities turned out to be within the range of normal children. Genie did subsequently acquire the ability to use a form of English as a means of linguistic communication, but her abilities in English were clearly not those of one who has acquired English under normal conditions. While Curtiss (1977) provides extensive illustration of Genie's speech, there remain possibilities for different interpretations of just how much progress Genie made in acquiring English. Curtiss emphasises Genie's achievements, noting utterances that seem to evince quite complex syntax. Bickerton (1990: 115-118), however, interprets Genie's failures as indications that she has acquired protolanguage rather than language; for instance, he claims

108 Bernard Comrie that apparent instances of complex sentences (involving subordination) may rather be interpreted as fixed formulae (1990: 116-117). Although certain questions remain, the material discussed in this subsection suggests that in the absence of any input during the crucial time window, not only will a child not develop a human language spontaneously, but also the absence of language development during this critical period will mean that the child will not subsequently be able to advance beyond the stage of protolanguage.

3.2. Creoles The question of the origin of Creoles is perhaps one of the most controversial questions in current linguistics, and it is not my aim here to adjudicate among competing positions, but rather to assess their relevance for the general question posed in this article. Creoles are of particular interest because, at least to a considerable extent, they involve the creation of a new language, one that is substantially distinct from any of the languages that enter into its creation. One part of creole genesis seems to be reasonably clear, namely that the lexicon of a creole comes from one or more of the languages of the communities that are involved in the building of the community in which the creole comes to be spoken, hereafter referred to as the contributing languages. In the most typical (for social and historical reasons) instances of creole genesis, the lexicon comes primarily from the superstrate language, with contributions ranging from highly significant to quite marginal from the substrate language(s). Although making up a lexicon might seem a priori to be the easiest part of creating a new language, in practice it seems to be the option that is least often, if ever, resorted to. By lexicon, incidentally, I mean essentially the forms of lexical items and their core meanings. The precise range of a lexical item may be influenced by the semantics of a language other than the one that contributes the form, but the core meaning will coincide, as in the case of N d y u k a f u t u 'foot, leg', where the semantic range of the English-origin form is broader than that of English foot (Huttar and Huttar 1994: 609-610). However, given the nature of a plantation society, for the majority of members of the community access to the superstrate language is limited, particularly the kind of access that would lead to adoption of the grammar of the superstrate language. The controversy surrounds precisely the origin

From potential

to realisation

109

of Creole grammar. Some researchers claim that the grammar comes largely from the contributing languages, and Lefebvre (1998) argues at length that the grammar of Haitian Creole comes primarily from the substrate language Fongbe. If it is the case that the grammar of Creoles comes largely from the contributing languages, then Creole languages essentially cease to be of interest for our present enterprise, since neither in lexicon nor grammar do they illustrate creation anew of a language. (There is, of course, no reason why a group of languages should necessarily be of interest to the present enterprise. Indeed, the vast majority of the world's languages, the product of regular transmission from generation to generation, clearly are not. The comments in this paragraph are thus quite irrelevant to the evaluation of Lefebvre's relexification hypothesis of Creole genesis.) An alternative hypothesis suggests that the grammar of a creole does not come from any of the contributing languages, but is instead created by children whose input is the fluctuating grammar of adults; these adults are native speakers of a range of languages, who have developed a common lexicon, largely on the basis of the superstrate language, but who lack any consistent grammar and are in fact operating with what we have seen Bickerton call a protolanguage. Given that there is no systematic grammatical input, the children have to make up their own grammar, that is, to create this part of the language anew. On the basis of similarities among creole languages with different contributing languages, a point returned to by McWhorter (1998) and Bickerton (1984, 1999) argues not only that children create grammar anew, but that there is a specific path laid down genetically for them to follow. In this bioprogramme hypothesis there is a particular set of unmarked values for the major parameters along which languages can vary, and in the absence of positive evidence to the contrary children opt for the unmarked values. For my present purposes, it is actually irrelevant whether this specific bioprogramme hypothesis is correct, since the mere fact of creation of grammar anew, however it is done, would satisfy the requirement of part of a language being created anew and not on the basis of input. Moreover, there are accounts other than the bioprogramme hypothesis that are candidates in accounting for typological similarities across Creoles, for instance the more functional approach of Seuren and Wekker (1986). Given the controversy within creole studies - see Muysken and Smith (1990) for further contributions - we cannot use Creoles as a clear case of creation of a language or part of a language ex nihilo. At best we can say that if it is true that creole grammar has been created ex nihilo, then Creoles are

110 Bernard Comrie directly relevant and illustrate the rapid creation of a language on the basis of a given lexicon only. But this is a big if.

3.3. Deaf sign languages Perhaps because of the "muddy" - to use a term suggested to me by Pieter Muysken - nature of the Creole data, attention has shifted more recently to another kind of language where there may be more cogent evidence for creation anew, namely deaf sign languages. By "deaf sign language" I mean specifically a signed communication system that is the basic medium of communication of a community (which will typically consist largely of deaf members) and that is of a complexity comparable to that of spoken languages that are the basic medium of communication in hearing communities. The recent expansion of sign language studies has shown that such languages exist and, moreover, that they are not derivative of the spoken languages of the same or neighbouring communities, neither in their lexicon nor in their grammar. For an early but cogent demonstration of the extent to which American Sign Language (ASL) differs grammatically from English, reference may be made to Klima and Bellugi (1979). Although ASL is by far the best studied sign language, one disadvantage for our purposes is that the language was already in place when it began to be studied scientifically, so that we have no direct evidence of the early stages of its development. One point, however, that can be resolved concerning the lexicon (the actual shapes of the signs) is that at least many of them were initially iconic but rapidly lost their iconicity (Klima and Bellugi 1979: 67-83; as noted on page 67, this chapter was written by Nancy Frishberg). For instance, the sign for 'sweetheart' was originally made with the hands on the heart, iconic of the folk identification of the heart as seat of emotions connected with affection, but is now made in the centre of the chest (Klima and Bellugi 1979: 74-75). This is interesting in that, for the first time, we see a possible ultimate origin for the lexicon, although the extent to which this can be extended to the lexicons of spoken languages remains unclear to me. But it is equally interesting that even in cases of transparent iconic origin, there is a rapid diachronic shift to arbitrary signs so characteristic of human language. 3 Fortunately, another sign language has been studied scientifically from within at most a few years of its creation, namely Nicaraguan Sign Language (NSL) (Senghas 1995, Kegl et al. 1999; see also Stokoe 1995:

From potential

to realisation

111

334). NSL arose around 1980, when for the first time deaf children in Nicaragua were brought together in schools for the deaf. Scientific investigation began in 1986. NSL did have a precursor in the Lenguaje de Signos Nicaragüense,4 a protolanguage "home sign" in Bickerton's sense, that lacked many of the grammatical prerequisites of a fully fledged language. Senghas argues that in the brief period separating the creation of NSL from the time of her investigation, the grammar of NSL had expanded to include just these grammatical prerequisites. Thus, NSL in particular, and perhaps sign languages more generally, seem to provide evidence of the creation of grammar anew. One thing that is interesting from the viewpoint of the bioprogramme hypothesis mentioned in relation to Creoles in section 3.2 is that deaf sign languages seem to share typological traits that distinguish them from Creole languages. For instance, David Perlmutter informs me that all known deaf sign languages have the phenomenon of "verb agreement" (Klima and Bellugi 1979: 276-279, where it is referred to as "referential indexing"), whereby the sign for a verb involves either location at, motion from, or motion to the location(s) that have been assigned to its arguments. For instance, the sign for "give" will move from the location assigned to the giver to the location assigned to the recipient. There is nothing like this in Creoles, which are noted rather for their extreme paucity of inflectional morphology (McWhorter 1998: 792-793); nor indeed in spoken language generally, where there seems to be no analogue to location in deaf sign languages. This opens up even further possibilities for investigation of where the constituent features of newly emergent grammars come from. But the basic observation remains that, certainly in the case of deaf sign languages and perhaps in the case of Creoles, grammar is created anew. A further point that emerges from the work on NSL is the apparent need for a sufficiently large community of signers for a deaf sign language to take off, and the need for a continuous stream of new cohorts; in the case of NSL, a second cohort of children entering the system, some ten years after the first, modified the efforts of the first cohort in the direction of more stable encoding of certain semantic values. In the absence of such continuous input, as documented by Ragir (2000), deaf communities do not develop beyond the protolanguage stage. Finally in this section, we should cite one well documented case (Schaller 1991) of a deaf man (known in the literature as Ildefonso) who grew up away from contact with any deaf sign language or, apparently, even home sign. As an adult, though socialised in other respects, he was not only without a language but even had difficulty grasping

112 Bernard Comrie

the concept of arbitrary signs when confronted with them. Thus, whatever the cognitive prerequisites for language acquisition and language creation, there are clearly also social constraints.

4. Supplementary creation In this section I will examine cases in which there is already a community language, or at least access to a community language, but nonetheless individuals create a distinct language.

4.1. Artificial

languages

Artificial languages, such as Esperanto and Klingon, seem to bear primarily on social aspects of the question of language origin and language continuity. It is clear that there are artificial languages that have been deliberately created, in most cases probably closely following the lexicons and grammars of European languages, as with Esperanto. Other cases depart far from these norms, like Klingon, created for the Klingons in the Star Trek series and films and the subject of a substantial cult following. There are instances of adults learning such languages with high levels of success, most notably in the case of Esperanto, and, again especially in the case of Esperanto, of children being brought up with such languages as their native languages. Thus, it is clearly possible for humans who already speak one or more natural languages to create new languages in this way and for them to become native languages. And such languages might be clearly unrelated genetically to existing languages, even if the most successful cases of artificial languages have been quite close lexically and typologically to the languages spoken by their creators. In the overall history of human language, however, artificial languages have surely played a minor role, if indeed they have played any lasting role at all. Humans acquire a language in childhood within their speech community; social circumstances might lead them subsequently to learn another language, and perhaps even to rear their children bilingually. But all of this involves the transmission of already existing languages. People have no need to create artificial languages, and it is doubtful if many in premodern times have even sought to do so. So artificial languages, though clearly possible, are not an integral part of the overall scenario that is being constructed in this article.

From potential

to realisation

113

4.2. Twin languages There is, however, another set of circumstances under which a new language can arise, apparently unnecessarily. It has been noted that twins often develop a communication system between themselves that is not comprehensible to others, particularly to other members of their speech community, indeed of their family. The fullest study of twin languages (also called autonomous languages) is Bakker (1987), on which I rely heavily here. Although at first the lexicon of a twin language appears to be idiosyncratic, closer investigation has in each documented case revealed that the lexicon is heavily parasitic on the language(s) spoken in the twins' community. The forms of lexical items often undergo substantial phonetic transformation, usually in ways typical of child phonology more generally, but with the difference that these forms then become fossilised in use between the twins. Thus, even though the resulting forms are usually incomprehensible to outsiders, they are clearly derived from the language(s) of the twins' input, so that there is no question of creating a lexicon ex nihilo. Bakker's conclusion on examining the grammars of twin languages is that the grammars are not necessarily similarly derivative of the grammars of the input language(s). Indeed, as the subtitle of Bakker (1987) suggests, the author entertains the possibility, on the basis of grammatical similarities among twin languages, that they may represent a default setting of parameters along the lines of Bickerton's bioprogramme (see section 3.2). In any event, there is much more innovation in the grammar than in the lexicon of twin languages. And in the case of twin languages, of course, there are no other possible sources of the grammar as with the substrate languages of Creoles. Thus, the twin language phenomenon can be summarised as a lexicon that is largely given by the input and a grammar that, at least in some cases, seems to develop largely independently of the input.

5. Levels of input We may now return to our basic question, namely: what level of input is necessary for language to arise? In the few cases where we can be reasonably certain that a normal child has been exposed to no input, language has not developed. Now, when language first originated this must have been the scenario, that is, the Dumbo Factor must have come into play: a creature that had a certain ability but was unaware that it had this ability must have be-

114 Bernard Comrie

come aware that it had the ability. We know of no clear modern instances of this happening, but this is not in itself evidence, since most children have the possibility of acquiring language under normal conditions and do so. At best we can say that we have no direct evidence of the Dumbo Factor coming into play, but circumstances make it unlikely that we would encounter such evidence. In one sense, then, our basic question of how humans came to know that they could communicate by means of language remains unanswered. If a lexicon is provided, then it seems that, at least in the presence of a community of potential speakers, language will develop, and will develop rapidly. The best documented example here seems to be NSL, whose predecessor, the protolanguage Lenguaje de Signos Nicaragiiense, provided at least a rudimentary lexicon from which a fully fledged language could then develop. If Bickerton's bioprogramme hypothesis is correct, then the development of Creoles would be another example, although as indicated in section 3.2 the evidence is less clear here. Twin languages may be a further example. Thus, perhaps somewhat surprisingly, the main task in creating language seems to be providing the lexicon. Now, since protolanguage clearly has a lexicon, the early humans who had developed the ability to acquire human language but did not yet have a human language to acquire could in principle have simply taken off from whatever protolanguage they already knew and expanded it. If this scenario is correct, then while the provision of a lexicon is a task that does not in itself require the linguistic ability of humans, it is nonetheless a crucial catalyst for the realisation of that ability. Can a language arise simply by stimulus diffusion? This term, borrowed from anthropology, means that the mere recognition that some other individual or group has a particular ability would lead the observer to realise that ability. To return to our initial analogy, if Dumbo had observed other flying elephants, then he might have realised that he had this ability and have started to fly. The question is difficult to answer, given that the crucial step of provision of at least a basic lexicon seems to be a step in the development of language that does not in itself depend on fully fledged linguistic ability. Even NSL, which has been studied from so close to its origin, has a forerunner in the protolanguage Lenguaje de Signos Nicaragiiense, which provided the initial lexicon. Thus the question remains open whether a group of humans lacking language and a lexicon but observing another group of humans that have language would solely on this basis enter straight into language without going through the initial stage of developing at least a rudimentary lexicon. It may even be that the question is moot, with lexicon creation taking place anyway under stimulus diffusion.

From potential

to realisation

115

6. Analogies: writing and ape language It may be useful to consider an analogy to the conditions under which human language can arise by looking at the origin of writing systems. Of course, there are striking differences between the two. Under normal circumstances, language develops inevitably in the normal child, while writing is a recent invention in human history, is absent from a large number of speech communities, and is normally learned through explicit (and at times painful) teaching. There is thus no sense in which we would expect a human individual or community to come up with writing. The number of independent developments of writing systems is small. In all likelihood the development of writing in Meso-America that led to the Mayan hieroglyphic writing system was completely independent of developments in the Old World. And it is possible and in some cases probable that some Old World writing systems - Chinese, for instance - may represent ex nihilo developments (Boltz 1996: 189-190). Most writing systems can be traced back to at most a small number of ancestors, in the way that the English writing system can be traced back through Latin to Greek and beyond. However, such direct ancestry is not the only possibility. Writing systems can also arise by stimulus diffusion: "some visionary, aware simply of the existence of writing among nearby peoples..., sets out to devise his own system" (Daniels 1996: 579). Perhaps the clearest documented example is the Cherokee syllabary (Scancarelli 1996: 587). The inventor of the script, Sequoyah, was a monolingual Cherokee speaker. He observed that English speakers could use marks on paper to represent their language, and set about devising a writing system for his own language. While the actual symbols are largely based on Latin letters, their values in the Cherokee syllabary bear no relation to their values in English: a symbol like an uppercase W represents la, whereas one like an uppercase Ζ represents no. But even more strikingly, the linguistic basis of the two systems differs. The English system is, however imperfectly, an alphabetic system, with ideally one symbol per phoneme, while the Cherokee system is a syllabic system, with ideally one symbol per syllable. There is, for example, no consistent representation of the phoneme /o/: the symbol described for no bears no resemblance to the symbol for go (which looks like an uppercase A) or that for do (which looks like an uppercase V).

116 Bernard Comrie Thus, despite the considerable ontogenetic differences between language and writing systems, in the case of the latter we have extremely few instances of creation ex nihilo, rather more instances of stimulus diffusion, and a vast number of cases of direct ancestry. A different possible point of comparison would be the language-like communication systems that have, with some degree of success, been taught to nonhuman primates. What is particularly interesting here is that no nonhuman primate seems spontaneously to have come up with a substantial lexicon, although nonhuman primates clearly have the ability to acquire a reasonably substantial lexicon as a result of direct instruction and, in the case of Kanzi and other bonobos (formerly called pygmy chimpanzees), by imitation of others that had been taught this lexicon (Savage-Rumbaugh et al. 1986). Once again, provision of a lexicon appears as a prerequisite for linguistic development, here of the protolanguage (in Bickerton's sense) attainable by some nonhuman primates.

7. Conclusions On the basis of the empirical material5 and hypotheses relating to them as discussed in the preceding sections, it seems that we can go some way toward answering the question of the necessary input for successful language creation. Under normal conditions of language acquisition, essentially nothing new is created in the long run, and this is what has characterised the linguistic development of most humans throughout human history. At the opposite extreme, if no input is provided at the crucial age, it seems that language is not created anew, a conclusion that has interesting implications for the origin of human language in the first place. If a lexicon is provided, it seems that children can create a grammar anew, though apparently only in the presence of a sufficiently large community and only with continuous input from new cohorts who enhance the language-like nature of the communication system; the most striking evidence comes from deaf sign languages, while other possible sources of evidence might be Creoles and twin languages, although especially in the case of Creoles there are strongly competing hypotheses. Creation of a lexicon seems somewhat surprisingly to be something of a stumbling block: children who are not exposed to language at the relevant early age do not spontaneously create a lexicon. Creation of a lexicon is, however, possible, since deaf sign languages have lexicons that are in some cases of demonstrably recent origin. They may, however, have been created by stimulus diffusion.

From potential to realisation

117

Notes 1. This article originally appeared in Linguistics 38 (2000): 989-1004.1 am grateful to the editor and publisher of Linguistics for allowing this reprint. 2. One might exclude certain monastic orders with a vow of silence. But these are hardly normal communities, for instance in terms of the possibility of internal procreation. 3. Compare the way in which the transparent pictures of the earliest Chinese writing soon give way to conventionalised, noniconic transformations. Iconicity between form and meaning in the lexicon seems just not to be a strong desideratum of human language. 4. The language here referred to as Nicaraguan Sign Language is called Idioma de Signos Nicaragüense in Spanish. English unfortunately has no obvious way of distinguishing between Spanish idioma and lenguaje. 5. It may be worth noting that even as philosophical and apparently aprioristic a treatise as Herder (1772) was recognised by the author as being subject to revision in light of new empirical discoveries. In the first edition, Herder claims that apes are unable to speak human languages despite having the same anatomy of the vocal tract. In a revised edition, he notes that anatomical investigations carried out in Holland showed that his presupposition was incorrect: apes have a different anatomy of the vocal tract.

Protothought had no logical names James R. Hurford

1. Introduction The evolutionary history of any complex system, such as human cognition or the human language capacity, necessarily starts with something simpler. This is not to deny that evolution can sometimes simplify. But the dominant evolutionary trend is from simple to complex, and the original protosystems were undoubtedly simple. It is accepted as a working hypothesis that a precursor of the modern human capacity for complex syntactic language was a capacity for protolanguage (Bickerton 1990), a kind of communication system with no syntax. In protolanguage, although words may have been uttered in short sequences, there were no rules defining the wellformedness of strings, and therefore words in protolanguage could not be said to belong to separate syntactic classes, such as Noun or Verb. A view commonly encountered about the emergence of true syntactic language is that during the protolanguage period there existed a somewhat complex system of mental representation, capable of representing the structure of events with their participant agents and patients. In this view, the move to syntactic language was in large part an externalisation of this preexisting system of representation: "The mechanism [of syntax] was there all the time, but it was not being used for language" (Bickerton 1998: 350). Let us call such a system of mental representation "protothought". It is often assumed that the structure of this protothought was something like Predicate Calculus (but possibly without quantifiers).

2. Predicate logic, semantics, and psychology Assuming something like a modern Predicate Calculus form of mental representation for our prelinguistic ancestors actually attributes to them at least a fiveway distinction between types of atomic mental entities, as the following simple formulae illustrate. The first formula: CAME(john). The translation of this would be 'John came'. The second formula: 3x[TALL(x) & MAN(x) & CAME(x)]. The translation of this would be Ά tall man came'.

120 James R. Hurford

In such representations, one has to distinguish the following types of atomic elements, each of which functions in the representations and relates to external meanings in a distinctive way: • • • • •

predicates (CAME, TALL) individual constants (john) individual variables (x) connectives (&) quantifiers (3)

The proposals which derive the basics of modern syntactic structure from such prelinguistic mental representations are seldom fully explicit on points of detail, but typically seem to assume that individual constants are the evolutionary sources of Proper Nouns and perhaps also of Common Nouns, whereas the predicate constants are the sources of Verbs. One account which is completely explicit on this point is found in the computer models of Simon Kirby (Kirby 2000a and 2000b). Here it is clear that individual constants are the evolutionary source of what emerge as Proper Nouns, and predicate constants the source of Verbs. Such an apparatus for prelinguistic mental representation is implausibly rich. I will argue that a reduced apparatus is more plausible. Specifically, in my less complex format for mental representations, there are only predicates, individual variables (but no individual constants), just one connective (namely conjunction: &), and just one quantifier (the existential "3" implicit in all representations). The implication of this proposal is that formulae such as the first given above, CAME(john), were not available as prelinguistic mental representations, specifically because they contain individual constants. Why do I believe that individual constants cannot be justified as elements of mental representation pre-existing language? A brief look at the place of predicate logic in modern scholarship will help to set the scene. The originators of modern logical calculi did not intend them to be taken as psychological entities. Frege, for example, was averse to "psychologism", believing that individuals' private concepts are beyond investigation. The central tradition stemming from Frege interprets logical formulae by a direct relationship with a presumed objectively given world, thus bypassing any psychological stage which might be located in a user's mind or brain. In this view, meaning is a relation between (logical) language and the world; no mind intervenes or mediates in this: "According to Montague, the syntax, semantics and pragmatics of natural languages are branches of mathematics,

Protothought had no logical names

121

not of psychology" (Thomason 1974: 2). This approach to meaning has been characterised (and criticised) by George Lakoff as "objectivist semantics" (Lakoff 1988: 119-154). The designers of Predicate Calculus assumed, surely correctly, that the world contains individual objects like the sun, the moon, particular people, dogs, cats, tables, and chairs. These individual objects are the denotata of the individual constants in logical formulae, such as john. Since, in this tradition, no minds mediate the relation between language and the world, the function which assigns denotations to individual constants is merely stipulated; it is typically explained by the author of a text on logic with enough ordinary-language prose to make the idea of an inherent (or stipulated) link between the logical name and the object it denotes clear enough to the reader. Often, the author conjures up some fictitious little world with fictional characters in it, and can be confident that the reader will get the idea. Textbooks such as Cann (1993), Guttenplan (1997), and McCawley (1981) as well as more advanced accounts of meaning in logic assume that giving a real semantics to sentences involves mapping them to an assumed objective world, without involving any psychological entities such as concepts, thoughts, or ideas. When linguists began to build generative models of languages including their semantics, it seemed appropriate to assume that the classical and simple languages of logic, such as Predicate Calculus, could be taken in their entirety and used as mental representations of meaning, much as standard IPA phonetic transcriptions can be used to represent pronunciation. Generative grammar, as indeed most modern schools of linguistics, aims to model linguistic competence, that is, psychological representations of language in speakers' heads. But it cannot be safely assumed that a formalism, such as predicate logic, which was not designed to suit psychologistic purposes, will be appropriate for such purposes. A semanticist who accords psychological reality to semantic representations like logical formulae, faces not only the task of stating what in the world the components of these representations map onto, but also the task of accounting for how these mappings could come to be known by the language user. Theorists of language evolution who assume something like Predicate Calculus representations preceded language in the minds of early hominids also face the problem of accounting for the mappings between the mental entities (such as predicates and individual constants) and parts of the world. An account of how mental predicates (LITTLE, GREEN, MAN) come to be mapped onto properties of the world is, I claim, relatively unproblematic,

122 James R. Hurford

whereas an account of how mental individual constants (John, mary) come to be mapped onto particular individuals is impossibly problematic.

3. Perception and cognition of objects It is crucial to distinguish between perception and cognition. Mental representations of events and their participants are involved differently in these two psychological processes. Let us start with perception or, more precisely, "perception-as".

3.1. 3.1.1.

Perception Unfocussedperception

Some experiences come uninvited. A creature may notice a particular smell, or feel cold, or hear thunder. In prelinguistic hominids, these experiences would undoubtedly have caused different reactions and can be regarded as separate mental predicates whose denotations are the classes of stimuli giving rise to them. In hearing thunder or noticing a smell, the perceiver perceives only the smell or the thunder; no other object or person is perceived. Such perceptions can be represented schematically as isolated zero-place predicates: SMELL 31 , COLD, and THUNDER. These could be paraphrased in modern English as 'There is a certain smell (around here)', 'It's cold', and 'Ah, thunder'. Note the lack of any contentful grammatical subjects in these paraphrases. If the creature has no language, there is no question of these categories being connected with any words. I assume that nonlinguistic creatures have representations which involve mental categories and that they are able to judge whether or not perceived objects belong in these categories. Such mental categories are partitions of an innately given structured space of possible categories, acquired through experience. Activation of different categories, whether by immediate experience or by rumination, will stimulate different patterns of behaviour; this is what enables nonlinguistic creatures to behave systematically in relation to the world. Our imagined creature somehow carries categories like SMELL 31 , COLD, and THUNDER in its mind. Their existence in its mind has been distally triggered by its previous experiences. In the act of perception, these categories are reactivated.

Protothought

3.1.2. Attending to perceived

had no logical names

123

objects

We now bring particular objects into the discussion. For concreteness, let's say a creature looks, with some attentive deliberation, at a fresh green sycamore leaf. In this case, the focus of attention is the leaf itself, which is an object in the world. In attending to it and judging it to be a green leaf, the creature applies certain mental categories to it, which we can relatively safely represent as the mental predicates GREEN and LEAF. Let us assume that this much is possible, the creature's species having evolved to be able to recognise greenness and the class of leaves. Just as with thunder and coldness, the categories GREEN and LEAF are somehow carried in the creature's mind, having been planted there by its previous contacts with the world. In the act of perceiving a green leaf, these mental categories are brought to bear on an object of attention in the external world. It is natural, for illustrative purposes, to represent the principal elements in such an act of perception by a formula such as GREEN(x) & LEAF(x) where χ denotes the real-world object attended to, and the predicates correspond to the mental categories which the creature judges to apply to this object. (To the best of my knowledge, the first use of such formulae, suggested as prelinguistic representations of meaning, is in Batali 2000.) Such a formula, then, has a mixed, partly external objective, partly mental, semantics. This is appropriate because an episode of "perception-as" is an encounter between mental categories and real-wo rid objects. The χ in the formula is essentially deictic (or "indexical"). It is not the constant name of any particular object. It only stands, in this convenient expository representation of an act of perception, for whatever object happens to be the object of perception. If on another occasion the creature sees a different green leaf, the same representation as before, "GREEN(x) & LEAF(x)", would be appropriate, as this is another act of perceiving an object as a green leaf. If I may call this χ a variable, it is attention that associates the variable with a real perceived object in the world. The same mental predicate can be either a zero-place predicate or a oneplace predicate, depending on whether the stimulus comes from an object to which the perceiver is attending. Thus if the creature deliberately sniffed the leaf, its perception might also include SMELL 31 (x). There is a fuzzy borderline between attention to a specific object and awareness of ambient impressions. Attention involves selection of one source of stimulation over others which have an equally good chance of being selected or attended to. A creature might decide to have a good look at a hole

124 James R. Hurford

in a tree (as opposed to the tree's bark or an upper branch). A hole in a tree is something of a size that one can reasonably pay selective attention to. Compare this with a large cave, especially one in which the creature is sitting. As the cave is all around it, it is not clear whether to count a perceptual judgement about the cave as an unfocussed perception or the perception of an object attended to. That is, it is not clear whether this should be represented as DARK(x) & CAVE(x) & SMELLY(x) or as DARK & CAVE & SMELLY. (To keep things simple, I allow the connective "&" to conjoin more than two conjuncts. Because I will not be dealing with disjunction, this does not at all affect the argument.) It will pose no problem to the account proposed here if we allow that mental representations cannot be fully determined, or theorised about, in all cases. Either the creature represents its perceptions as simply ambient (the zero-place case) or more specifically about the place it is sitting in as opposed to some other place it has chosen not to attend to (the oneplace case).

3.2. Cognition: representing previously perceived objects The important aspect of cognition, for our purposes, is the memory or store of experiences which are no longer current. Consider how the nonlinguistic creature, while not actually perceiving a green leaf, stores in memory the fact that it did once see a green leaf. When objects are no longer attended to, the association of an external object with an attention variable is lost, since it is the very act of attention which brings about this association. In the transition from perception to cognition (that is, the process of putting into memory), then, the variable loses its former (extensional) meaning. The creature can remember that it once had a "GREEN(x) & LEAF(x)" experience, but there is now nothing pointing, as attention once did, to the object itself. The closest that Predicate Calculus can get to this is formulae with existential quantifiers, like 3x[GREEN(x) & LEAF(x)] and B[DARK(x) & DAMP(x) & SMELLY(x) & CAVE(x)]. The second formula here might be paraphrased as '(I remember) there was a dark damp smelly cave'. Memory of complex experiences (but not involving more than one object) is captured by conjunctions of such atomic proposition-predicates. Each conjunction or bundle of features is a separate entry in the mental database, a memory of a separate experienced event. Implicit in this account, and the central plank of the whole argument, is the theme that in prelinguistic cognition there are no components of mental

Protothought

had no logical names

125

representation with the logical status of individual constants, that is, elements, that, as the philosophers say, allow their possessors to re-identify particulars with absolute reliability.

3.2.1. There are no particulars in animal cognition The individuals most important to us are our near kin, mothers, and children. The following quotation demonstrates the prima facie attraction of the impression that animals distinguish such individuals, but simultaneously gives the game away: The speed with which recognition of individual parents can be acquired is illustrated by the "His Master's Voice" experiments performed by Stevenson et al. (1970) on young terns: these responded immediately to tape-recordings of their own parents (by cheeping a greeting, and walking towards the loudspeaker) but ignored other tern calls, even those recorded from other adult members of their own colony (Walker 1983: 215).

Obviously, the tern chicks in the experiment were not recognising their individual parents. They were being fooled into treating a loudspeaker as a parent tern. For the tern chick, anything which behaved sufficiently like its parent was recognised as its parent even if it wasn't. The tern chicks were responding to very finely grained properties of the auditory signal and apparently neglecting even the most obvious of visual properties discernible in the situation. In tern life, there usually aren't human experimenters playing tricks with loudspeakers, and so terns have evolved to discriminate between auditory cues just to the extent that they can identify their own parents with a high degree of reliability. Even terns presumably sometimes get it wrong: "[AJnimals respond in mechanical robot-like fashion to key stimuli. They can usually be 'tricked' into responding to crude dummies that resemble the true, natural stimulus situation only partially, or in superficial respects" (Krebs and Dawkins 1984: 384). A creature can appear to recognise a particular individual reliably if it gets sensory impressions which are finely grained enough to distinguish this individual from others with which it is likely to come into contact. To manage successfully in "recognising its mother", a creature need not have a mental distinction between individual and property. All that is needed is the ability to distinguish just the set of properties which in practice differentiates its mother from other candidates.

126 James R. Hurford

English dictionaries often distinguish between two senses of the word "recognise" as below: 1. If you recognise someone or something, you know who someone is or what someone is when you see them or hear them, because you have seen them or heard them before. EG She didn't recognise me at first... 2. If you recognise something ... you are able to identify it when you see it, for example because you have learned about it or had some previous experience of it. EG They are specially trained to recognise the symptoms of radiation sickness... There was a field planted with what he recognised as maize (Collins Cobuild English Dictionary).

This lexicographic distinction rests on the ontological distinction between individuals or particulars (I use the terms interchangeably) and universals. The central claim here is that prelinguistic creatures could not have had mental representations containing elements corresponding to particular individuals; they could only recognise things in sense 2 above, but not in sense 1.

3.2.2. Individual particulars,

universal

properties

Prelinguistic creatures could not have been able to re-identify particulars with absolute reliability because, in fact, neither can we. Given what we know about the world, it is, in the limit, impossible. This is not to deny that much of the time we are quite successful at re-identifying particulars, especially where it matters. But situations where re-identification is impossible are also common. For instance, every morning I pick up a bottle of milk from my doorstep. If I have a backlog of undrunk milk from previous days, I add today's bottle to the two or three that are already in the fridge in a particular spatial sequence so that I know which bottle is which. If someone disturbs my milk filing system, then I no longer know which is which because I cannot re-identify particulars (my dairy does not datestamp its products). Where it really matters to us to keep individuals apart, we either mark them somehow (as with my milk filing system) or we develop more refined perceptual categories so that we can remember the difference between individuals in terms of their slightly different properties. This is presumably why humans are so proficient at face recognition. But it is both impossible and unnecessary to refine one's perceptual categories to the point where every individual object is distinguishable from all others. One paperclip is

Protothought had no logical names

127

much the same as another, as are sheets of paper, grains of sand, first-year students, and so on. The argument above has concentrated on the difficulty with logical individual constants as components of mental representations. The impression may have been given that logical predicates are not problematic in the same way. Philosophers have emphasised to me that, in the limit, predicates are also subject to similar problems. Putnam's Twin Earth argument (Putnam 1975: 223247) establishes that there could be objectively (that is, from a God's-eye view) two different sorts of stuff that we label water. Our concept of water could fit both kinds of stuff, and the stuff that is chemically XYZ could, to all intents and purposes, serve exactly the same range of purposes as the stuff that is chemically H 2 0 to the degree that we wouldn't know the difference: "Extension is not determined by psychological state" (Putnam 1975: 222). There is, however, a great practical difference between an ability to reidentify particulars (such as the precise lion that was stalking me yesterday) and an ability to re-identify universals (such as lions in general), and hence a great practical difference between individual constants and predicates. Invariant factors in the environment in which humans have evolved have given rise to our perceptual space and the discriminability of regions within it. The species has evolved to see just the colours it can see, hear just the sounds it can hear, smell just the sounds it can smell, and so forth. Despite some individual variation, the same inputs from the world trigger broadly similar perceptions in everyone, across a wide range of possible inputs. More importantly, there are relatively few cases in which objectively disjoint stimuli trigger the same perception (objectively, that is, to the best of our current scientific knowledge). Putnam mentions such a case: "Although the Chinese do not recognise a difference, the term 'jade' applies to two minerals: jadeite and nephrite. Chemically, there is a marked difference. Jadeite is a combination of sodium and aluminium. Nephrite is made of calcium, magnesium, and iron. These two quite different microstructures produce the same unique textural qualities!" (Putnam 1975: 241). Examples such as jade pose no problem for a theory that prelinguistic creatures could have had mental predicates corresponding to various classes of stimuli. Their mental predicates would have had a perceptual and functional basis. If it looks like a duck, sounds like a duck, smells like a duck, feels like a duck, and tastes like a duck, it's a duck. Properties are procedurally prior to individuals. That is, we use an object's properties to try to determine which individual it is. We humans are often obsessed by this question of which individual is presenting itself to us. We can never be certain, but we get as close as we can by finer and finer

128 James R. Hurford

subclassifications of the properties we are observing. In life's ongoing calculation of "what is happening", the fundamental empirical premises are the properties which the world presents to us. The conceptual identities of individuals appear only in our conclusions. Central to the concept of an individual are the related ideas of uniqueness and spatio-temporal continuity. These are not part of the idea of a property. Many things are red, in different places, at the same time. But an individual cannot be in two places at once. Consequently, the evidence needed to assign the same property (a mental predicate) on different occasions is weaker than the evidence needed to conclude that something is the very same individual. To assign a property, in many cases all we need is the evidence of our senses: "No doubt about it, this is green". But without keeping a constant eye on an object or insulating it somehow, we can never be sure that it hasn't been replaced by a twin: a different individual.

3.3. Language hints at knowable identity Modern human languages contain devices which are interpreted as if their identification of particulars posed no problem. Proper names are the prime example. But in fact our ordinary usage with proper names is more rough and ready than the usage of logicians with individual constants. We all know many people called Jim, and usually the context of the speech situation clarifies which Jim we are talking about. Thus Jim is in fact only a shorthand for whichever of the people who have been regularly labelled Jim in the experience of the interlocutors is most salient in this context. Clearly, Jim in English does not pick out a single individual, though doubtless there are some rare names which do happen to pick out single individuals. But we would not wish to distinguish logically or grammatically between different proper names on the basis of their relative rarity of application. Markers of definiteness, like English the, are also basically shorthand for whichever of the things concerned is most salient in this context. Individual objects are less enduring over time than properties. The objects to which proper names are most commonly attached, namely people, only live for a few generations. But almost all of the properties which we see in an individual person can be seen, in different combinations, in other people, long before and long after the brief life span of the particular individual. Ars longa, vita brevis might be adapted to the case of properties and particulars, which is related to why, in philosophical discussions, properties

Protothought had no logical names

129

are often called universals. The relative impermanence of particulars, as opposed to properties, undermines the utility of constant expressions for particulars in the continuing language of a community. Place names are the most useful designators of enduring individual objects in modern languages, but our early ancestors would have had much less use for them. In any case, there is evidence that in modern languages all place names in fact derive historically from descriptions, that is, from expressions involving property terms, like the dark pool or the new castle. Probably, all person names are ultimately so derived as well. Linguistic devices suggest that humans are capable of mentally representing unique individuals using the mental equivalent of logical individual constants. We might even grant, for the sake of this argument, that modern humans do have this capacity. But it is only language that pushes us toward this conclusion. Without an insight into our language, outside observers of modern human behaviour could not tell that we are able to mentally represent particular individuals. We, in our turn, when we consider the minds of nonlinguistic creatures, must adopt the simplest hypothesis compatible with the observed facts. There is no need to hypothesise mental individual constants in animal minds. When animals respond systematically to impressively finely graded stimuli - even to the extent of reliably recognising what human observers believe to be unique individuals - there is no need to postulate a special category of mental representation like individual constants. Finely graded predicates or conjunctions of less finely graded predicates will do the job.

4. Social life without individual constants We are considering the emergence of language with modern characteristics from pre-existing cognitive capacities in prelinguistic creatures. I have maintained that the cognitive representations available to prelinguistic hominids for remembering the events of their lives are most unlikely to have contained the equivalent of logical individual constants. A more plausible form of prelinguistic mental representation would have had only predicates, individual variables, a conjunctive connective, and an implicit existential quantifier. The baby australopithecine or habiline would have had many stored memories of episodes involving its mother, many containing a form something like this: SMELL31(x) & SQUAWK(x) & WARM(x) & CUDDLY(x). (Here and henceforth, the existential quantifier is omitted from such re-

130 James R. Hurford

presentations. It is implicit. The variables, however, cannot be eliminated because they are needed for representing memories of events involving several different objects.) Our immediate prelinguistic ancestors certainly had complex social lives, as do modern chimpanzees. In modern human terms, they remembered who were their friends and enemies, who was whose child, who had recently fought with whom, and who had groomed whom. But the whos in this last sentence tend to impose our modern logical analysis on the animals' behaviour. It is possible, with finely graded predicates, to represent as accurately as is necessary any of the information that australopithecines or habilines would have stored about members of their social group. That this is generally true is shown by a transformation that one can apply to any firstorder logical formula containing individual constants, replacing the constants by corresponding (suitably fine-grained) predicates. For example, CAMEO'ohn) can be replaced by CAME(x) & JOHN(x) where JOHN is a predicate which is only satisfied by one individual. Of course, expressing it in terms of a predicate being "only satisfied by one individual" is again to express things from a modern, language-informed perspective. To a prelinguistic hominid, as to a tern chick, it was only important that the mental predicate be specific enough to be reliable in all the circumstances in which it was likely to be used. Two-place predications involving different individuals can be handled in this way, but with distinct variables. Thus Kanzi is Panbanisha's child could be represented as KANZI(x) & PANBANISHA(y) & CHILD(x,y). Clearly, representations without individual constants are sufficient to account for the Machiavellian intelligence of apes. It is assumed here that our prelinguistic ancestors were capable of forming associations between different concepts, allowing inferences between thoughts involving different predicates. Thus SMELL 14 (x) —> EDIBLE(x) might have been a mental rule useful in foraging. Similarly, "portmanteau" predicates could have been created, abbreviating longer conjunctions of more basic predicates. Thus the baby australopithecine we mentioned earlier could have had a rule such as [SMELL 3 1 (x) & SQUAWK 7 (x) & G A I T 1 2 ( X ) ] -> MOTHER(x).

5. So what? How does this argument relate to other issues in language evolution? The origin of syntax is a central problem. At the heart of syntactic structure is the

Protothought

had no logical names

131

Subject-Predicate dichotomy. So any claim about the origin of the SubjectPredicate dichotomy is central to theorising about the origins of modern human language. There are two tendencies in the search for explanations of features of language, whether these are evolutionary or otherwise. One tendency is to look in the direction of possibly innate, not necessarily communicatively functional, characteristics of the minds which gave rise to language. Β ickerton, in particular, has stressed the representational function of language, as opposed to its communicative function. Accounts of the rise of Subject-Predicate structure from the pre-existing-representation point of view assume a pre-existing dichotomy between two types of concepts, a predicate type, and a subject (or argument) type. Individual constants are the principal instances of concepts of the subject/argument type. In this view, pre-existing propositional representations of Subject-Predicate form get rather straightforwardly translated into linguistic clauses with corresponding structure. Such proposals push the question of the origin of Subject-Predicate structure back into prelinguistic representations, and that is considered far enough to absolve the proposers from further explanation. But as I have shown, it is unnecessary to attribute such pre-existing mental structure to prelinguistic creatures. They could have got along fine without it, and so they probably didn't have it. The opposing tendency in linguistic explanation is to appeal to communicative function. Communicative function cannot be a consideration relevant to prelinguistic (or precommunicative) mental structure. Pressure from communicative function can only arise once mental representations are externalised in public, conventionalised symbols. From this perspective, an alternative explanation for the origin of Subject-Predicate structure is communicatively functional Topic-Comment structure. It must be emphasised that Topic and Comment are not categories of classical (modern) logical systems. Topic and Comment are categories rooted in pragmatics, in the shared knowledge of speakers and hearers, and thus are not, and never have been, the concern of truth-conditionally oriented logicians. The argument presented in this paper has been largely negative. I have argued that our prelinguistic ancestors could not have represented the world in terms of formulae involving individual constants, where these individual constants were destined to become the evolutionary antecedents of modern referring expressions by a process of externalisation into public language.

132 James R. Hurford

This negative argument is a prelude to a positive argument that will describe a mechanism whereby essentially communicative pressures gave rise to the emergence of modern Subject-Predicate structure, based on prior representations in which the only constants were predicates. In brief, some predicates became selected as prototypically serving a Topic function, while others became selected as prototypically serving a Comment function. And this happened concurrently with the emergence from protolanguage of longer, more coherent strings of words. But that is another story.

Acknowledgements This work was supported by grant (Number R000237551) from the UK Economic and Social Research Council and by a Fellowship at the Collegium Budapest Institute for Advanced Study. I have benefited from helpful comments and advice from John Batali, Manfred Bierwisch, Peter Carruthers, Terry Deacon, Daniel Dor, Dan Nettle, Bruce MacLennan, Luc Steels, and Paul Taylor, but don't blame them for any faults in the paper.

The birth of rules Jean Aitchison

I. Chunks versus rules Language origin has finally become a topic of interest to mainstream linguists and others, and at long last, publications are beginning to proliferate (e.g., Aitchison 1996, Deacon 1997, Hurford et al. 1998, Jablonski and Aeillo 1998, Jackendoff 1999, Pinker and Bloom 1990, Trabant 1996). Now, as the twenty-first century begins, scholars are identifying some key questions. This paper will attempt to answer one of these: the chunks-versus-rules problem. Language is not a unitary phenomenon: parts of it are memorised chunks, and parts of it are rule-governed. These two linguistic types co-exist throughout language, as seen, for example, in irregular versus regular verbs (Pinker 1999). Yet the historical relationship between the two phenomena is unclear. Do the memorised chunks lead to rules? Or do they co-exist as independent types? This is the key question that will be explored in this paper. A parallel question used to be found in the child language literature. At one time, it was unclear whether youngsters' cries and babbles were separate from "real" language. This has now been settled: crying has been recognised as separate, prelinguistic behaviour. Babbling, on the other hand, becomes more sophisticated and develops into speech (Vihman 1996). This paper is divided into four main sections. First, it will summarise some background assumptions on language origin. Second, it will outline some presumed early stages in the human development of language, highlighting the importance of convergence (co-evolution). Third, it will discuss the possible origin of rule-governed behaviour. Fourth, it will consider what current-day language can reveal about these early rules.

134 Jean Aitchison 2. Background assumptions Three important background assumptions are outlined below, concerning innately guided behaviour, Baldwinian evolution, and the "general uniformity principle". These will form a backdrop to the remainder of the paper.

2.1. Innately guided behaviour For centuries, an age-old nature-nurture controversy has trundled on, swinging to and fro. Is language "natural", as when dogs naturally bark? Or is it "nurtured", as when dogs can be taught to beg? This debate has also been labelled a distinction between "hard-wired" and "soft-wired" behaviour, or between "instinct" and "learning". Yet the sharp divide between these two types is now recognised as unrealistic. Lenneberg (1967) pointed out that language is controlled by maturation: children acquire language "naturally" provided that they are properly "nurtured" - exposed to adequate linguistic data at the relevant time. Gould and Marler (1987) speak of "innate guidance": instinct guides humans to pay attention to particular features, whose finer points then have to be learned, much as bees primarily pay attention to the odour, but also shape and colour of flowers. These characteristics lead them to sources of nectar, which they visit with increasing certainty as they learn to identify particular flowers. Birdsong provides a further example (Gould and Marler 1987, Marler 1998). A number of species of birds have an outline knowledge of their own songs, but have to learn the finer details. So the nature-nurture controversy has largely faded away as the division between the two types of behaviour has blurred (Aitchison 1998).

2.2. Baldwinian

evolution

Around a century ago, James Mark Baldwin, an American psychologist, indicated how Darwin's theory of natural selection could be extended (Baldwin 1902). So-called Baldwinian evolution shows how new traits can emerge without the necessity of assuming (as did the discredited Frenchman Lamarck) that acquired responses to environmental challenges could be passed onto offspring directly. Behavioural flexibility and learning, Baldwin argued, can amplify and bias natural selection: a subgroup of a species can

The birth of rules

135

move into a niche slightly different from that of its ancestors. Species members who could via natural variation withstand the cold, for example, would survive better during a fierce winter. Future generations might then inherit, and gradually enhance, a genetic predisposition for enduring freezing weather. From the point of view of language, those who had, say, better memory skills, might be those with greater chances of survival. These survivors might pass on an ability to retain a large vocabulary to their offspring.

2.3. General uniformity principle The so-called "general uniformity principle" (Lass 1997: 28) that language in the past followed similar principles to language today, was clearly stated by William Dwight Whitney in the nineteenth century: "The nature and uses of speech ... cannot but have been essentially the same during all periods of its history... [T]here is no way in which the unknown past can be investigated, except by the careful study of its living present and recorded past, and the extension and application to remote conditions of laws and principles deduced by that study" (Whitney 1867: 34). The general uniformity principle ties in with the "uniform probabilities principle" (Lass 1997: 28), that the likelihood of any linguistic state of affairs has always been roughly the same as it is now. The practical upshot of these interlocking principles is that we can make deductions about the beginnings of language by looking at current-day language.

3. Early stages of language: humans versus other primates Humans, like all primates, are social animals. Primate social behaviour has at least three clear characteristics: strong family ties, active within-group interaction, and a well-defined ranking order. It is not surprising, therefore, that a primate should have developed a communication system which promotes these. Yet humans alone have developed language, though chimps show some linguistic precursors. The ways in which humans and chimps diverged are outlined below. The most noticeable differences are brain size, "theory of mind", voluntary vocalisation, and speech production abilities.

136 Jean

Aitchison

3.1. Brain size The enlarged brain size of humans may have arisen as a spin-off of the geographical location of protohumans. Humans are now generally thought to have emerged from Africa. One increasingly accepted view is that future humans lived in an area of Africa which underwent a severe drought, possibly east of the Great Rift Valley (Kortlandt 1968, Coppens 1994), though the exact location is still under discussion. This led to a cascade of further developments - especially meat eating - which alongside other factors promoted a big brain: the prefrontal area in particular is enlarged, compared with the brain of chimps (Deacon 1997). The enlarged brain involved premature births, by ape standards, and consequently neoteny (extended childhood). These immature infants were kept close to their parents, whose vocalisations they imitated during their lengthy infancy. Yet meat-eating cannot have been the only factor promoting a big brain: brain size, group size, and use of deception all intercorrelate strongly. The latter two, particularly deception, provided the impetus to developing advanced communication skills, especially a "theory of mind".

3.2. Theory of mind A so-called "theory of mind" was possibly crucial in language development. Signing chimps, it has been noted, restrict themselves mainly to asking for things they want, such as oranges, juice, tickles, and so on. They do not talk for the sake of talking. Humans have developed an ability to put themselves into another person's shoes. This ability is turning out to be multilayered and complex. Its origins may lie in the ability to deceive, since successful deceit requires an animal to see things from another's point of view: the nearer the primate species is to humans, the more efficient they appear to be at hoodwinking one another (Whiten and Byrne 1997). The theory of mind possibly led to symbolisation, which in turn led to the "naming insight", the realisation that a sequence of sounds can be a "name" for something - an ability which develops normally in infants, though which can be delayed in deaf humans (Schaller 1991).

The birth of rules

137

3.3. Voluntary vocalisation A less obvious linguistic forerunner is the ability to vocalise voluntarily, and equally importantly, to refrain from vocalising involuntarily. Many animals, including chimps, tend to vocalise when presented with an appropriate stimulus: chimps can under certain circumstances suppress vocalisations (Byrne 1994). But this is not an everyday occurrence. Gradually, vocalisation became normal and habitual. At some point, humans began to interact via vocalisation more than in any other way. This was a grooming replacement strategy, according to Dunbar (1996), when numbers in a group rose too high to make grooming feasible.

3.4. Speech production Chimps have similar auditory abilities to humans. The big problem arises with speech production. Apes do not have the firm tongue, L-shaped vocal tract, and lowered larynx which, arguably, allowed humans to produce clear sounds. Some of the differences are due to the flatter faces of humans: the higher quality animal-based diet also reduced chewing requirements, which led to a reduction of face and jaw size (Aiello 1998). The lowered larynx is possibly due to humans' upright posture. Yet an inability to make clear sounds could not have been the only factor holding other apes back: language could have emerged in another medium, notably sign, as some people propose happened prior to spoken language (e.g., Armstrong et al. 1995, Givon 1995).

3.5. Convergence

(co-evolution)

True language began, possibly, when some of the factors outlined above converged and amounted to more than the sum of the parts. As Givon (1995: 426) notes: "the socio-cultural, cognitive, communicative, behavioural and the neurological aspects of language probably evolved in parallel rather than serially. These profoundly interdependent, interactive changes thus coevolved". The big step forward could have arisen when a clear output converged with the naming insight. The convergence of these two abilities possibly led to a naming "explosion", a desire to name dozens of objects (Figure 1).

138 Jean Aitchison Stage 1

Habitual vocalisation Good hearing

Stage 2

Upright stance L-shaped vocal tract

Big brain Deception ability

Stage 3

Stable vowels

Theory of mind

Clear output

Naming insight

Stage 4

Naming explosion

Figure 1. Convergence of abilities leading to the "naming explosion".

4. The emergence of rule-governed behaviour A process of preferences to habits to rules (discussed further below) can be hypothesised as the basis for "real" language (Aitchison 1998). The date for the emergence of rule-governed language is still under discussion. The earliest suggested date is around 250,000 b.p., though this is most probably too early. The latest date put forward is around 50,000 b.p., though this is undoubtedly too late. Most researchers assume a date of around 100,000 b.p., though it seems increasingly likely that an earlier protolanguage existed for some time (Jackendoff 1999, Calvin and Bickerton 2000). This comprised maybe a handful of words for key needs - like come, look, and so on - with some habitual collocations.

4.1. Word combinations: inbuilt preferences The naming explosion presumably led to word combinations. Eventually too many words existed for them to be routinely uttered singly. In some cases, early words were possibly combined fairly randomly, though were always perhaps subject to preferences, of the type seen in ape signing. The chimp Nim, for example, combined words mostly in a random way. Yet love of

The birth of rules 139 eating led him always to place foods first in any sign sequence, as banana eat Nim, banana Nim eat, banana me eat, grape eat Nim, apple me eat. (Terrace 1979). Some human ordering preferences are so deeply inbuilt that they can be regarded as prelinguistic, such as a preference for placing "small on large". Humans prefer to say: The cat sat on the mat. It would be bizarre to say: The mat lay under the cat. Even if the mat was the topic of conversation, the order would likely be: On the mat, sat the cat (Landau and Jackendoff 1993). Other human preferences can be regarded as linguistic, since they do not inevitably result in the same order, as with a liking for placing animate subjects first: an order Horses eat hay or Horses hay eat would be more likely in the languages of the world than Eat hay horses or Hay eat horses, though these are not impossible. Another strong human preference is for action-patient closeness, though the order of action and patient is variable: an order Monkeys eat bananas is as likely as Monkeys bananas eat. Not all languages have a preferred word order, but in a sample of 402 languages from those that do, 87 percent combined the two linguistic preferences mentioned above, with subject-object-verb (SOV) accounting for 45 percent, and SVO for 42 percent (Tomlin 1986). At the origin of language, then, it seems likely that "animate first" and "action-object closeness" predominated.

4.2. Word classes Word order in language depends not just on habitual collocation, but on the existence of word classes. The earliest words were possibly nouns. Certainly late learners, such as deaf people who have been deprived of language, have discovered nouns before other types of word, as with the deaf Mexican Ildefonso (Schaller 1991). And the blind Helen Keller famously discovered the word for 'water' as her earliest vocabulary item (Keller 1903). Verbs possibly arose out of nouns via re-analysis, as in the pidgin Tok Pisin, where it is easy to see how a sequence mi singsing 'me song-anddance' could be interpreted either as Ί was at the song and dance festival'or Ί sang and danced'. This distinction is possibly based on a human ability to distinguish names from events. Adjectives were possibly a subdivision of verb: in some languages stative verbs and adjectives are indistinguishable, as in Miskito Coast Creole English: if yu wud sief yu wud ron.

140 JeanAitchison 'If want to be safe, you would run' (Holm 1988: 85). Here, the word class of sief is unclear. As these examples show, re-analysis is the mechanism by which numerous changes take place in languages today (Aitchison 2001). Such re-analysis would have been possible from the earliest combinations (in accordance with the uniformity principle).

5. Clues from current-day language Following the uniformity principle, clues about language origin can be culled from a variety of present-day sources, in particular cases where strong preferences are shown in behaviour that is formulaic, but not fully rule-governed. The distinction between an unanalysed formula and a rule-governed group of words is not necessarily clear-cut. A sequence of sounds may be just that, as with a child who learned to hammer on a door shouting obedide ( O p e n the door'), though who later learned to analyse his utterance (Peters 1983). And population groups may behave differently from one another. Gleitman and Gleitman (1979) found strong differences between educated and uneducated speakers in their ability to interpret a compound such as house-bird glass. In the sections below, three situations are discussed in which current day humans produce language which is quasi-formulaic, quasi-rule governed: "newsworthiness" order, sports commentating, and headlines.

5.1.

Newsworthiness

Only a portion of today's languages have a fixed word order. Some American-Indian languages have a superficially free word order, though in fact work according to a "newsworthiness" principle, bringing to the front some highlighted, newsworthy item (Mithun 1992), as with Cayuga. The (translated) answers to the (translated) questions show how this works: Q. 'Who are you going with?' A. 'SAM we're going with'. Q. 'What do you want to buy?' Α. Ά DRESS I'm looking for'. Q. 'How long were they there?'A. 'TWO WEEKS they were away'. Cayuga is not alone. Ngandi and Coos behave similarly. "Word order in these languages is thus based ... on the relative newsworthiness of the constituents to the discourse. An element may be newsworthy because it represents significant new information,

The birth of rules

141

because it introduces a new topic, or because it points out a significant contrast" (Mithun 1992: 39). In the earliest stages of language, therefore, as in these American-Indian languages, newsworthiness may have played a vital role.

5.2. Routines: auctioneers and sports

commentators

At the origin of language, humans may have had a less efficient short-term memory than humans today; they may have been less able to process speech fast, so some clues may be found in situations in which humans are pressured to speak fast. In such circumstances, they develop linguistic routines: "because it is necessary in some cases and desirable in others, to cut down the linguistic options for both speakers and hearers, speakers often resort to formulaic speech in routine contexts. Such psychological factors are essentially concerned with individual human memory and processing capacities" (Kuiper 1996: 92). Auctioneers have developed routines and formulae for the various stages of an auction. These are repetitious utterances, though with "slots" which change, notably, the item being auctioned, and the price bid. Each stage of the auction has its own formulae, as, for example, a first bid: How much for that? How much for the X? X dollars for it. X dollars I've got for it. X dollars for that. I've got X dollars for them (Kuiper 1996: 61). The closing stages also are predictable: Xgets 'em. X buys 'em. They go to Y. You 're the winner, Sir (Kuiper 1996: 71).

142 Jean Aitchison Horse racing has formulae for each section of the race, as with the start: There they go. They 're away and racing. They 're off and racing now. They're on their way (Kuiper 1996: 17). The finish also has formulae: They go to the post. X has won it. Xgot it won (Kuiper 1996: 18). These are slotted in around other formulae, such as "loop formulae", the arrival at the end of a lap of the course, and also around individual events, such as a particular horse falling: "the horse looks as though he is walking OK so he hasn't done any serious damage to himself and the driver is quite O K " (Kuiper 1996: 14). Children are known to cope with some aspects of language via routines (Gleason et al. 1984, Greif and Gleason 1980), as when American youngsters follow a "trick or treat" routine when they call round at neighbours' houses at Halloween (Gleason and Weintraub 1978). And successful second language learners acquire routines to help them on their way (Fillmore 1979). In short, routines which are partly memorised, and partly analysed, in that they have "slots" into which different items can be placed, appear to be an essential part of language under pressure of time or memory. Presumably, this has always been the case.

5.3. Newspaper

headlines

Newspaper headlines illustrate an intermediate stage between routines with occasional words slotted in, and true rule-governed behaviour. Noun sequences in headlines represent a headline style which became common in the 1960s in British English newspapers (Simon-Vandenbergen 1981). These noun sequences have their own incipient "rules", in that they have a high probability of appearing in a particular order, though this is not essential.

The birth of rules

143

Similar sequences are found in broadsheet and tabloid newspapers. Noun sequences involving the word murder are documented below for a six month period in two British broadsheet newspapers: The Times (T) and the Guardian (G); and two tabloids: the Sun (S) and the Daily Mirror (M) (Aitchison, Lewis, and Naylor 2000). These noun sequences sometimes formed the whole headline, at other times they were part of a longer one: Shotgun murder horror (M) Street murder inquiry reaches dead end (G) Within the 220 noun sequences involving murder, two-noun sequences predominated: 141 two-noun (64 percent): Turk jailed for wife murder (T) 73 three-noun (33 percent): Murder trial mistress tells of old flames (M) 6 four-noun (3 percent): WPC murder bid charge (S) At first sight, these noun sequences allowed a variety of possibilities, though murder was preferentially near the beginning of its sequence, rather than the end. On closer examination, the words which preceded and followed murder were not random. They followed a limited set of patterns, which were similar across all four newspapers analysed, and across two-, three-, and fournoun sequences. The noun immediately preceding murder within its sequence was most usually the victim (48 occurrences, 59 percent): Bell murder enquiry ends (T) Jury in child murder case was misled (G) Bride murder trial (S) Sex fiend wanted over barmaid murder (M) Place preceded murder in 16(19 percent) of the examples: University murder case opens (T) Street murder inquiry reaches dead end (G) Hospital murder man's prediction (S) Amazing case of the M50 murder and the vanishing moccasins (M)

144

JeanAitchison

Cause of death preceded murder in 10 cases (12 percent): Husband Barbecue Hammer Shotgun

guilty of acid-bath murder (T) smell that led to needless knife murder (G) murder victim cut wife out of will (S) murder horror (M)

The word immediately following murder was most commonly a legal and/or abstract term (139 examples, 79 percent), as bid, case, charge, enquiry, plot, quiz, rap, trial. For example: Murder on tv (T) Teenage girls on murder charge (G) Lorry man on 3-girl murder rap (S) Tears of murder trial girl (M) Of the remainder, 16 percent (28 examples) referred to some involved person, such as suspect or victim: Murder pair jailed for life (T) Murder victim looked lovely (G) Car murder hubby caged (S) Party's off as murder sisters get life (M) A number of three-noun sequences began with the noun murder, in which case the second word was a legal/and or abstract term, and the third usually some involved person: Murder trial judge praised (T) Murder case husband spent wedding night with accused (G) Murder quiz wife drugged (S) I lied to give murder case lover an alibi (M) Occasionally, however, another legal/abstract term was added: Murder case remand (G) Grieving dad on murder bid rap (S)

The birth of rules

145

The few four-noun sequences mostly involved a victim in front of the above three-noun sequences: Boy murder charge man in court (G) Carl murder quiz man freed by cops (S) Wpc murder bid charge (S) The murder headlines therefore showed a clear murder formula: A. Victim (most likely), or place, or cause. B. Word murder. C. Legal or abstract term. D. Person accused, or second legal term. Two-word sequences were mostly A + B, or Β + C; three-word sequences were mainly A + Β + C, or Β + C + D. The rare four-word sequences were (sometimes) A + Β + C + D. The main difference between newspapers was in vocabulary. The broadsheets preferred to use surnames, while the tabloids mainly used the first names of victims, as: Shaughnessy murder trial (T) Alison murder charge (S) In line with this formal versus informal trend, the broadsheets used relatively formal vocabulary to describe humans, such as mother, father, husband, child, friend: Murder case husband spent wedding night with accused (G) Jury in child murder case was misled (G) The tabloids, on the other hand, often referred to humans via short, informal vocabulary, such as mum, dad, hubby, tot, pah Car murder hubby caged (S) Tot is silent witness to mum 's murder (M)

146 Jean

Aitchison

Predictably, perhaps, the broadsheets used fairly formal legal or technical vocabulary like charge and inquiry: Driver faces triple murder charge (T) Murder enquiry arrests (G) The tabloids used fairly informal vocabulary like rap and quiz: Husband on murder Girls in murder quiz rap (M)(M) Overall, then, the similarities between newspapers were striking in that they all used the same murder formula, though the vocabulary slotted in to the formula differed: fairly formal in the broadsheets, shorter and more informal in the tabloids. These murder headlines illustrate how groups of words become melded into a fixed order, with the most newsworthy first, an extension of the process of grammaticalisation. Grammaticalisation is most typically the demotion of a full word to a grammatical morpheme (Hopper and Traugott 1993), though is also found with larger groups of words which co-occur and coalesce: "loose, paratactic, 'pragmatic' discourse structures develop - over time - into tight, 'grammaticalised' syntactic structures" (Givon 1979: 208). This coalescence is further explored in Tabor (1993) and Tabor and Traugott (1998). Compound nouns show some similar properties to the newspaper headlines in that co-occurrence leads to gradual coalescence and hence to a grammatical structure, as in dinner plate, finger nail, housewife, rescue worker, though the principles involved are not always clear-cut (Bauer 1998, Downing 1977, Ryder 1994).

6. Conclusion This paper has outlined a possible scenario for the birth of rules. It argued that human language diverged from the communication systems of other primates when a number of different factors converged: speech production ability converged with the naming insight to produce a large vocabulary. An ability to distinguish people and objects from events led to different word classes and a preference for particular orders. This led to the development of

The birth of rules

147

habits. Then habits became rules. It further argued that the development of preferences to habits, then habits to rules is observable today in the process of grammaticalisation. The question posed at the beginning of the paper was on the relationship between memorised chunks and rules. The conclusion is that memorised chunks may lead to rules, though do not inevitably do so. Rules arise when words are assigned to different types (word classes) and these types are assigned a typical order. This order is likely to have arisen partly out of natural preferences and partly out of "newsworthiness". Different types of activity may have given rise to different types of "newsworthiness".

How language changed the genes: toward an explicit account of the evolution of language Daniel Dor and Eva Jablonka

1. Introduction This paper represents an attempt to construct a programmatic framework for the evolution of human language. The theory we propose is interdisciplinary in its essence. It is based on what we believe are noteworthy recent advances in linguistics and evolutionary biology. The theory aims to resolve three fundamental paradoxes frequently encountered in the burgeoning literature on language evolution. These paradoxes leave many of the traditional conceptions of language evolution at a theoretical dead end and inhibit interdisciplinary cross-pollination among linguistics, evolutionary biology, anthropology, and the neurosciences.

1.1. The functional

paradox

Two radically opposite claims have traditionally been made in the linguistic literature concerning the functional nature of linguistic knowledge. As far as language evolution is concerned, both claims seem to lead to a theoretical impasse. Generative Grammar, by far the most influential linguistic theory of the last forty years, is famous for making the claim that the aspects of language that form the universal, innate basis of linguistic knowledge (Universal Grammar) are essentially structural, formal, and autonomous from notions of meaning. As such, they are explicitly non-functional. According to this view, what we can do with language - like communicate meanings to each other - cannot tell us anything significant about the linguistic system itself. This implies that a neo-Darwinian theory of language evolution is unattainable because such a mode of evolutionary explanation requires a satisfactory functional characterisation of the relevant evolved trait (here: the linguistic capacity) as a descriptive platform. This is why Noam Chomsky insists that attempting to say anything meaningful about language evolution is a waste of time. Different writers, like Pinker and Bloom (1990), Newmeyer (1998), Berwick (1998), and Jackendoff (1999),

150 Daniel Dor and Eva Jablonka

have tried to demonstrate that Generative Grammar does make sense, but we believe that Chomsky is right: from the evolutionary perspective, his innateness claim cannot be reconciled with his specific characterisation of language as a non-functional cognitive apparatus. The opposing view, according to which grammatical complexities are reducible to principles of general cognition, at first seems more tenable, since it implies a functional characterisation of linguistic knowledge. But on closer examination it does not do much better than the Generative view, for at least three complementary reasons. First, although the specific functional theories that attempt to reduce linguistic complexities to general cognitive principles (e.g., Givon 1995, Langacker 1987, and Wierzbicka 1988) propose preliminary approximations of the cognitive precursors of linguistic complexities, they do not usually provide rigorous explanations for specific grammatical facts. Therefore, they do not offer a real alternative to the Chomskyan characterisation of the object of evolution. Second, attempting to reduce language to general cognition is problematic. There is ample evidence - from language acquisition, language breakdown, and the formation of de novo languages (like the sign language developed by deaf Nicaraguan children) - that language is a unique and highly specialised cognitive capacity. Whereas Chomsky's theory captures the uniqueness of language at the expense of its functionality, cognitive theories attempt to salvage language's functional aspects at the expense of its uniqueness. Third, the cognitivist characterisation of language as a general-purpose communication tool is too broad. Asserting that we use language to communicate is similar to asserting that we use our vision to see. As Marr (1982) has convincingly argued, a functional characterisation of any cognitive system must be much more specific. A real functional characterisation of language should be both empirically viable and functionally specific.

1.2. The paradox of

domain-specificity

Linguists unanimously concur that linguistic knowledge is extremely domain-specific; that is, it manifests properties not found in other behavioural systems. The usual extension of this to the domain of language evolution goes as follows: if language is domain-specific (and if at least some of our linguistic ability is innate), then some of this domain-specificity should be reflected in brain structures. Ironically, the one clear assertion neuroscientists seem to agree on is that the brain is an organ of extreme plas-

How language changed the genes

151

ticity and generality (see Elman et al. 1996 and Deacon 1997), which means that the chances are slim of finding explicit representations of linguistic specificities innately encoded in brain tissue prior to acquisition. Characteristically, linguists either subscribe to domain-specificity or adopt the neurophysiological position of non-specificity and resort to a general learning theory to account for the acquisition of linguistic specificities from external input. This is paradoxical, since both linguistic specificity and brain plasticity seem to be irrefragable, as is the notion that language acquisition cannot be based solely on external input. So what we need is a theory of the relationship between language, genes, and the brain that a) reconciles language's domain-specificity with the brain's high level of plasticity and b) relates them to each other in non-contradictory ways, both in phylogeny and ontogeny.

1.3. The paradox of the dynamic and variable nature of language Most scholars who believe in linguistic innateness adhere to a static and universalistic conception of language. The generative theory of principles and parameters is the most famous such conception: even when some variability between languages is conceded, the theory encodes the variability in the genes. Children come to the world with a few parameters for each linguistic principle, and choose the right one for the language they encounter. This notion faces what we consider to be insurmountable difficulties. First, it cannot account for the fact that languages are dynamic entities that constantly change and evolve in their social contexts. Second, it cannot easily account for the considerable diversity among different languages. Third, it must posit a genetic mutation that enabled an individual in a hominid community to use the full range of future languages at a stage when no languages existed, and it must then assume that this property spread across the entire community. This is hardly a plausible evolutionary scenario. It is our goal in this paper to propose a framework that successfully resolves these paradoxes, explicates the theoretical prerequisites for an empirically viable theory of language evolution, and describes what we think are the fundamental properties of the evolutionary process. Our argument runs as follows: A. Our point of departure is the functional paradox. We will claim that a series of empirical results, accumulated in the last two decades in the field of linguistic semantics, strongly indicate that Chomsky's long-standing hypothesis of the autonomy of syntactic structures from meaning regulari-

152 Daniel Dor and Eva Jablonka

ties should be abandoned in favour of an explicit, semantically based, and empirically oriented theory of transparent meaning-form relations. B. Crucially, the specific semantic categories which turn out to determine syntactic regularities in languages are not reducible to general cognition or to conceptual structure. Instead, they seem to manifest uniquely linguistic properties. As such, they determine the expressive envelope of language. That is, they determine which meanings - and which meaning combinations - are expressible by means of natural language structures. C. This state of affairs allows for a novel characterisation of language as a cognitive capacity: neither a non-functional, formal system (the Generative characterisation), nor a functional, general-purpose communication system (the functionalist characterisation), but a unique and highly constrained communication system dedicated to the communication of a specific set of meanings. According to this characterisation, language is not just functional and unique. Its uniqueness is in the specificity of its function. In line with traditional cognitive psychology, we characterise language as a functional, unique, and transparent mapping system between the representational level of linguistic meaning (the set of constitutive semantic categories) and the representational level of linguistic form (the set of grammatical markers for these meanings). D. This characterisation of language enables us to reframe the evolutionary question as the question of the gradual expansion and sophistication of the linguistic mapping system; that is, the expansion and sophistication of the set of constitutive semantic notions, their interactions, and their modes of mapping onto the speech channel. Note that this is not the same as the question of the origin of language; that is, the question of the evolution of language's cognitive precursors, and the first "leap" from these to what we think of as natural language. We will concentrate on the question of evolution (as formulated above) and leave the question of origin for further research. E. The above reframing of the evolutionary question naturally calls for an answer in terms of the interaction between cultural and genetic evolution. Here, a distinction should be made between three closely related questions. How did language evolve? How did speakers (and linguistic ability) evolve? How did these two processes interact? F. According to our theory, cultural evolution played a major role in the evolution of language. The process of cultural-linguistic evolution consisted of the selection of, the social agreement on, and the cultural evolution of the constitutive semantic categories for communication, as well as the gradual

How language changed the genes 153 sophistication of the mapping system for these categories. In this long, gradual, and complex process, a social group gradually isolates certain aspects of its epistemology, sharpens and develops them, reaches social agreement about them, and develops sophisticated structural means for communicating about them within the community. Needless to say, this is a permanent process, which continues today. G. Throughout most of evolution, this process was made possible by homonids' great behavioural plasticity, which includes all aspects of linguistic behaviour: production, comprehension, acquisition, and transmission. H. Crucially, however, at various points in evolution, cultural evolution stretched speakers' behavioural plasticity to the point where differences in the ability to learn language became selectively important. At these points, genetic assimilation occurred - on all cognitive fronts. At every step, linguistic culture constituted the selective environment for genes that contributed to linguistic performance, acquisition, and transmission. The interaction between continuous, directional cultural evolution and partial genetic assimilation resulted in a consecutive set of evolutionary stages in which language's expressive envelope was expanded and sophisticated and in which speakers were selected on the basis of their linguistic performance. This process of cultural evolution and genetic assimilation gradually created what we think of as a linguistically biased cognition: a cognitive makeup which, without encoding linguistic specificities on a genetic basis, is still biased toward rapid learning of the linguistic mapping system. I. As we will claim, our approach successfully resolves the three abovementioned paradoxes, and has several additional, non-trivial consequences. It transforms the traditional dichotomy between innateness and learning into the question of how much, when, and what type of learning are necessary at each evolutionary stage. It renders the continuity-discontinuity debate obsolete by rejecting the idea that there is a relevant distinction between a syntactic and a presyntactic language. And it reconciles socially oriented as well as structurally oriented approaches to language evolution and incorporates significant insights from linguistic anthropology into the universalistic discourse on language evolution.

154 Daniel Dor and Eva Jablonka

2. On the nature of grammars Our point of departure is a theoretical reappraisal of Chomsky's long-standing hypothesis of the autonomy of syntactic structures from meaning considerations. As we have already indicated, recent empirical research on the interface between syntactic and semantic representations consistently demonstrates the dramatic extent to which syntactic phenomena are determined by semantic regularities. These studies strongly suggest that Chomsky's hypothesis should be abandoned in favour of an explicit, semantically based, and empirically oriented theory of transparent meaning-form relations. Before rehearsing one such demonstration, we need to make a few general observations about Chomsky's autonomy hypothesis. First, Chomsky's hypothesis is not a mere existence claim. It does not claim that natural languages display purely structural syntactic phenomena. This assertion would be self-evident because such structural phenomena one need only think of word order - obviously exist. The autonomy hypothesis goes much further. It claims that a significantly large set of core syntactic phenomena in natural languages cannot be theoretically correlated with a corresponding set of functions formulated in terms of meaning. The autonomy hypothesis takes the non-functionality of syntactic structures to be a necessary and fundamental property of human language, not a peripheral, accidental property of this or that construction in a specific language. From an evolutionary perspective this is crucial because according to Chomsky the set of non-functional syntactic generalities is innately given. Second, the autonomy hypothesis is ultimately irrefutable. It is a negative claim: the claim that a significantly large set of core structural phenomena is immune to explanation in terms of a meaning-based theory. Any attempt to empirically refute such a hypothesis should, in principle, provide for total coverage of the complete set of structural-syntactic phenomena of an entire language—and preferably all languages. It should demonstrate that every nonaccidental syntactic phenomenon can be correlated with a coherent functional theory of meaning. Because no linguistic theory, regardless of its ideological stamp, comes close to providing a fully explicit description of a single language (let alone of the universal parameters of language as such), the autonomy hypothesis is in no danger of refutation. This means that the fate of the autonomy hypothesis should be decided on the basis of empirical demonstration. We should take some complex, syntactic phenomena - which everybody agrees are significant and nonaccidental and which seem to be divorced from any considerations of meaning - and

How language changed the genes

155

show that they can be given an explicit meaning-based explanation. Then we should do it again. And again. To the extent that the analyses turn out to be empirically sound, they demonstrate that an explanation of the relevant structural phenomena in terms of meaning is not just possible, but necessary. To the extent that the phenomena constitute uncontroversial, core domains of linguistic knowledge, such meaning-based analyses gradually weaken the autonomy hypothesis as a default assumption. They strengthen the suspicion that the autonomy hypothesis reflects a stage in the history of the misunderstanding of the nature of meaning rather than a profound insight into the nature of language. In the last two decades, semanticists and lexical semanticists have generated empirical demonstrations of exactly this type (see Alsina, Bresnan, and Sells 1997; Butt and Geuder 1998; Dor 1996, 1999, and 2000a; Dowty 1979 and 1991; Frawley 1992; Goldberg 1995; Grimshaw 1990; Jackendoff 1983 and 1990; Levin 1993; Levin and Rappaport 1991 and 1995; Parsons 1990; Pustejovsky 1991 and 1995; and Van Valin 1997). Some scholars, ourselves included, believe that enough results have accumulated for us to entertain the idea that grammars are not autonomous from meaning. Below we briefly present one empirical demonstration regarding what is probably the most famous set of syntactic facts discussed in the Generative literature, namely island constraints. This cursory presentation is meant to provide a general sense of the proposed explanation and its functional nature (see Dor 2000a for details).

2.1. Island

constraints

Island constraints were discovered by John Ross (1967) and have since acquired a uniquely prominent status in syntactic theory. The different syntactic mechanisms postulated through the years to (partially) grasp these phenomena have been extremely complex, and island constraints have come to be known as the standard example of syntax's innately given complexity and of its autonomy from meaning. The basic facts are demonstrated in the following examples. Languages like English allow long-distance transformations of the type that we see in (1) below. Moving the wh-word from its "natural position" (after the verb meet) to sentence-initial position seems to be possible regardless of the distance between the original position and the final one:

156 Daniel Dor and Eva Jablonka

(1) a. b. c. d.

Who did John meet at the supermarket? Who did Bill say that John met at the supermarket? Who did Mary think that Bill said that John met at the supermarket? Who did George realise that Mary thought that Bill said that John met at the supermarket?

Crucially, however, some transformations are not allowed. For example, it seems that moving an NP out of some constructions is ungrammatical - at least in most cases. Below are two of the most famous constraints. First, it seems that NPs cannot be moved out of complex NPs: (2) The complex NP constraint: a. John kissed [np the girl who delivered [np the pizza]]. b. * What did John kiss the girl who delivered? Second, it seems that NPs cannot be moved out of subjects: (3) The subject constraint: a. [ s u b j Your obsession with [np Madonna]] annoyed your father. b. * Who did your obsession with annoy your father? Within the Generative framework, facts like those in (2) and (3) have traditionally been explained by structural means. Informally, the different versions of the theory have taken syntactic constituents to be structural barriers to extraction. Obviously, it is difficult to imagine the functional significance of such structural barriers. To the extent that constraints on extraction are taken to be innately given - and they certainly are within the Generative framework - then the question of their evolution becomes a prime example of the functional paradox. Dor (2000a), however, explains the relevant set of facts on a semantic basis. According to his theory, the grammatical extractions meet a semantic condition that the ungrammatical ones do not. The semantic condition has to do with the interaction between two semantic domains: event structure and epistemic licensing. Informally, when a speaker performs an assertive speech act - when she tells an interlocutor, for example, that John kissed the girl who delivered the pizza - the speaker performs two types of speech acts at once. She provides the interlocutor with the description of an event (like the event in which John kissed the girl who delivered the pizza) and she makes the epistemic

How language changed the genes

157

claim that this event took place in the world. The theory of event structure addresses the first part of the speaker's speech act. It specifies the semantic properties of events described by natural language sentences and the semantic properties of the participants in these events. Thus, informally, our speaker's sentence describes an event with the following representation: (4) There is an event e e culminates before now e is an event of the type kissing the agent of e is John the patient of e is the girl who delivered the pizza The agent and the patient in (4) are the thematic constituents of the described event (there are of course many other possible thematic constituents). The theory of epistemic licensing has to do with the second part of the speaker's speech act: her claim that the event actually took place. Making the factual claim that the event actually took place in the real world is just one of the speaker's many epistemic options. She may, for example, say any of the following sentences: (5) a. I believe that John kissed the girl who delivered the pizza. b. Bill told me that John kissed the girl who delivered the pizza. c. Mary saw John kissing the girl who delivered the pizza. In (5a), the speaker tells the interlocutor that she has some reason to believe that the event actually took place. In (5b), she tells the interlocutor that her epistemic claim is based on Bill's epistemic claim regarding the kissing event. In (5c), the speaker tells the interlocutor that Mary actually witnessed the event as it occurred and that this forms the basis for her epistemic claim. In all these cases (and in many others) the kissing event is epistemically licensed. The speaker specifies who is responsible for the claim that the event took place and the type and strength of the claim. Dor (2000a) proposes that the syntactic behavioural patterns we saw in examples (1) to (3) are best captured by the interaction of these two pieces of semantic theory. This is Dor's transformation rule: a transformation can apply to a syntactic constituent if and only if it refers to a thematic constituent of an epistemically licensed event. We have already seen that the thematic constituents of our kissing event are the agent {John) and the patient (the girl who delivered the pizza). The

158 Daniel Dor and Eva Jablonka

sentence is epistemically licensed by the speaker's factual claim, which means that a question transformation can apply to the syntactic constituents which refer to the agent and the patient of the event. (7) a. Who kissed the girl who delivered the pizza? b. Who did John kiss? Moreover, because the kissing event is also epistemically licensed in all the examples in (5), transformations can also apply to its thematic constituents: (8) a. Who did you believe kissed the girl who delivered the ρϊζζαΊ b. Who did Bill tell you that John kissed? c. Who did Mary see John kissing? Crucially, however, the pizza does not participate in the epistemically licensed event. It plays a thematic role in the delivering event, but not in the kissing event. The NP the pizza thus cannot be transformed: (9) * What did John kiss the girl who delivered? A similar type of explanation applies to the examples in (3): the participants in the annoying event are the experiencer (your father) and the source (your obsession with Madonna). Madonna herself is not a participant, and the question transformation in (3b) is thus ungrammatical. Despite its brevity, the above presentation is, we hope, specific enough to provide a sense of the theory's meaning-based nature. The relevant set of constraints on syntactic transformations in English is captured by a semantic constraint on the performance of the interrogative speech act. It seems that the interrogative speech act can only target the thematic constituents of an event that is epistemically licensed by the speaker. This is a functional theory in at least two complementary ways. First, it correlates the syntactic complexities with a well-defined set of semantic complexities, which we need to assume independently for the sake of semantic and pragmatic analysis. Second, the specific semantic categories that determine the relevant syntactic complexities are categories of linguistic communication. They are categories relating to the communication of information about events. When we tell each other about events in the world, this is exactly what we do. We provide descriptions of the events and their participants, and we evaluate the descriptions' epistemic validity. If island constraints can be ex-

How language changed the genes

159

plained on the basis of a set of semantic categories directly related to linguistic communication, then they are far from evidencing the autonomy of syntax from meaning. Indeed, they demonstrate the functional nature of grammatical structures. Does this refute the autonomy hypothesis? Strictly speaking, it does not. There remains a long list of syntactic phenomena that has not yet been explained in terms of meaning. Nevertheless, island constraints are often invoked as the prototypical example of the autonomy of syntax. And together with numerous other semantically based explanations of other major syntactic complexities, the above theory significantly weakens the autonomy hypothesis. Every advance in this field strips yet another syntactic phenomenon of its autonomous appearance. We think enough demonstrations have accumulated for us to reject Chomsky's hypothesis and to see where a transparent view of the syntax-semantics interface will take us.

3. On language as a unique mapping system Event structure and epistemic status are not the only two semantic domains that determine grammatical structures in natural languages. Logical categories like negation and conditionals determine some patterns of structural behaviour (such as negative polarity items); spatial categories (categories of spatial relations between physical entities) determine some behavioural patterns of verbs and prepositions; categories of time interact with event structure to determine aspects of the syntax and morphology of tense and aspect; classifier categories determine morphological patterns in many languages; and so on. Some pragmatic categories, such as topic and focus, seem to play a similar role, as does the inventory of speech acts. Some of these semantic and pragmatic categories seem to be universally relevant and to determine aspects of grammar in all languages. Others seem to be particular to a subset of languages. Where do the semantic (and pragmatic) categories that determine syntactic regularities in natural languages belong? What is their cognitive status? Some scholars, most notably Jackendoff (1983 and 1990), assign them to the conceptual domain and view them as conceptual categories. To us, the evidence suggests that these categories belong to a uniquely linguistic level of meaning representation, one closely related to, and yet crucially different from, conceptual structure (see Dor 2000b). Why? First, the semantic categories that play a constitutive role in determining syntactic generalisations in natural languages seem to belong to a small

160 Daniel Dor and Eva Jablonka

subset of all the semantic categories we use to think about and conceptualise the world. Even more significantly, these semantic categories cut across conceptual structures in ways that seem arbitrary from a conceptual viewpoint. Some semantic categories are grammatically relevant across many languages, whereas others are not. To take just one example, Frawley (1992) compares two important and robust distinctions of meaning: the distinction between natural and nominal kinds and the distinction between animate and inanimate objects. The former distinction relates to the difference between common nouns that denote inherently and those that denote compositionally. It is visible throughout the lexicon. Tiger, gold, hepatitis, heat, pain, and red are natural kinds. Car, wheel, coat, wedding, divorce, and president are nominal kinds. This distinction plays a central role in our cultural conceptualisation, but no language appears to mark the distinction by any sort of formal device. The distinction between animate and inanimate objects, on the other hand, is extremely relevant for linguistic structure: "the linguistic evidence shows that in every language there appears to be some grammatical reflex of the difference between animate and inanimate objects" (Frawley 1992: 9). Frawley concludes that the "fundamental question of philosophical semantics - what kinds of meaning are possible - contributes to the identification of a variety of potential meanings that language may encode. But only some of the results of an inquiry driven by this question are relevant. Not all possible meanings are grammaticalised; not all have empirical status" (Frawley 1992: 10). Second, constitutive semantic categories always determine the behavioural patterns of word classes and not of individual words. Verbs, for example, belong to verb classes. The members of each class denote different versions of the same event type. The members of the class of surface-contact verbs, for instance, denote different versions of the same event type: movement in contact with a surface. They include sweep, wipe, scrape, scratch, and scrub. The class of change-of-state verbs includes break, smash, crash, fracture, and shatter. They all denote different versions of an activity resulting in a change to the physical state of the patient (Levin 1993). As many researchers have demonstrated, all members of the same semantic class display the same pattern of syntactic behaviour. The meaning distinctions between the different members of the same verb class - between, say, sweep and wipe or between break and smash - are syntactically irrelevant. That is, there does not seem to be a syntactic generalisation that is sensitive to these meaning distinctions. It is as if all such verbs are grammatically identical. Conceptually, of course, we do distinguish between sweep and wipe, and

How language changed the genes

161

between break and smash. But as far as the structure of language is concerned, sweep and wipe are indistinguishable. Grammar only isolates their event type, which is thus a constitutive determinant of their structure. We therefore must adopt two lexical representations of the verbs' meanings: a fully detailed conceptual representation and a skeletal linguistic-semantic representation. Third, the semantic categories which turn out to determine the grammaticality of judgements do not constrain our ability to assign conceptual interpretations to ungrammatical sentences. We are fully capable of understanding what the non-grammatical transformations in (2) and (3) were supposed to mean. There is nothing inherently illogical or conceptually impossible about asking a question concerning the pizza the girl delivered or Madonna as the cause of your father's annoyance. To take another example, Rappaport-Hovav and Levin (1998) use the semantic properties of a large number of verb classes (including change-of-state verbs and surface-contact verbs) to account for intriguing phenomena like the contrast between the grammaticality of Terry swept the crumbs off the table and the ungrammatically of * Terry broke the vase off the table. The point is that we can readily imagine a situation where the vase was glued to the table and Terry removed it from the table by breaking it. The problem here is not with the ability to understand a certain event conceptually, but with the ability to describe it linguistically. This is a telling fact about the semantic categories of language, not about the conceptual categories in which we think. Finally, the semantic categories that determine syntactic generalisations seem to manifest discrete (or digital) properties, whereas conceptual categories seem to manifest continuous (or analogue) properties. Take, for example, the category of animacy. Making a conceptual decision about the animacy of some entity is not always easy. Conceptually, animacy is a continuous category. Some entities are prototypically animate, some are prototypically inanimate, and others are somewhere in between. But the linguistic category of animacy is not continuous. As far as grammar is concerned, an entity should be either animate or inanimate. This discreteness is manifested in the structural markers of animacy in languages. Languages that mark animacy in their morphology force an obligatory, discrete choice on speakers, a choice that speakers do not necessarily have to make when they conceptualise. All the above observations seem to point to the same conclusion. Linguistic-semantic categories comprise a constrained subset of all possible

162 Daniel Dor and Eva Jablonka

conceptual categories, a subset that is systematically highlighted, foregrounded, isolated, and digitalised for the purposes of linguistic communication. This set of linguistic-semantic categories determines language's expressive envelope. It determines which meaning combinations can be expressed in language, and which meaning combinations cannot. This conclusion echoes a major insight discussed by Aitchison (1996), Levinson (1997), and others: language is not the best tool to communicate all meanings. Some meanings are better communicated by means of visual imagery, music, body language and mime. Other meanings, especially narrative ones, are best communicated by language. They constitute its expressive envelope. All this allows for the construction of a novel characterisation of language as a unique and functional system: a transparent mapping system between the set of constitutive semantic notions (that determine its expressive envelope) and the set of linguistic markers used to express these meanings. The following is a schematic representation of this view of language: [conceptual representations]

[linguistic meaning] ο

[linguistic form] || [phonetic representations]

The above schematic representation characterises language as a transparent mapping system between the levels of linguistic meaning and linguistic form. The level of linguistic form, which we have not yet discussed, includes all the structural tools which are visible on the speech-channel and are used to mark linguistic meanings in natural languages. Besides phonology, these include morphological markers, linear order, adjacency, and so on. What is missing from this picture is abstract, autonomous, invisible syntax. As we have seen before, we have good reasons to assume that such syntactic representations are no longer needed. Moreover, the above schematic characterisation allows for both linguistic variability and universality - on all fronts. Different languages may, in principle, occupy the level of linguistic meaning with different semantic categories and different categorical combinations (this is the problem of linguistic relativity: see Gumperz and Levinson 1996). But some major categories - like event structure and animacy - seem to be universal. Different languages use different subsets of linguistic form to mark semantic categories. Some markers, however, are universal. Finally, different languages may map different semantic categories onto different markers; this is an essential property of the system. Thematic roles, for example, may be marked by linear order in some languages and by morphological case markers in others.

How language changed the genes

163

This novel characterisation of language immediately reframes the question of language evolution. It is now neither the question of the evolution of a formal, non-functional system, nor of the evolution of a general-purpose communication system. It is the question of the evolution of a specific communication system dedicated to the communication of a constrained set of meanings by means of sound concatenation. In cognitive terms, it is the question of the evolution of a mapping system: of the gradual expansion and sophistication of the representational levels of linguistic meaning and linguistic form, and their transparent mapping onto each other. The answer to this novel evolutionary question lies in the interaction between two different evolutionary processes: cultural evolution and genetic evolution.

4. Cultural evolution and genetic assimilation Genetic evolution involves a change in the nature and frequency of genes in a population. Similarly, cultural evolution involves a change in the nature and frequency of socially learned and transmitted behaviours in a population. Both cultural and genetic evolution clearly played important roles in the evolution of hominids. Early hominids must have had cultural traditions, which are the consequence of cultural evolution, and modern humans and early hominids are certainly genetically different. Traditions are ubiquitous in higher animals and encompass every aspect of their lives: modes of foraging, mate selection, avoiding predators, criteria for choosing a habitat, practices of parental care, and so on (Avital and Jablonka 2000). Traditions are particularly well studied in primates. Thirty nine different cultural traditions were recently described in seven populations of the common chimpanzee, and five of these traditions have something to do with communicative-social functions (Whiten et al. 1999). There is much more to be learned here, since researchers have only recently started to study animals' traditions systematically and comprehensively. That some form of culture and much cultural variation (some of which was associated with communication) existed in hominids, can be taken for granted. In hominids, cultural evolution is often cumulative and often leads to the gradual sophistication of a cultural practice. A good example is socially learned and transmitted improvements in tool-making techniques, which could have gradually led to the more elaborate fashioning of stone tools. That said, genetic evolution was certainly also involved in language evolution. After all, the linguistic differences between humans and chimpanzees

164 Daniel Dor and Eva Jablonka seem to be at least partially genetically based. The interesting question concerns the nature of this genetic difference, and the relationship between linguistic cultural evolution and the evolution of the genetic difference. Can the cultural evolution of languages be related to the genetic evolution of hominids? We claim that it can and that this interaction is particularly important for understanding language evolution. We suggest that the evolutionary process involved the co-evolution of genes and culture through a dynamic process of genetic assimilation. What was genetically assimilated was the increasing capacity to acquire language - a process that resulted in a cognition biased towards the acquisition of language.

4.1. Genetic assimilation and simple "instinct"

evolution

Genetic assimilation is the transition, through Darwinian selection, from an acquired (learned or induced) response to a more genetically fixed or "instinctive" response. Also known as the Baldwin effect (Simpson 1953), this process involves selection for the ability to respond rapidly and efficiently to the new stimulus. When individuals face a new environmental challenge, they usually first adapt to it by learning. If the selective pressure is ongoing (learning takes time and is costly), there will be selection for the best and fastest learners. This may culminate in a population of individuals who learn very quickly and even of individuals for whom a single exposure to the stimulus is sufficient to elicit the adaptive response. The learned response becomes an "instinct". Conrad H. Waddington, the British geneticist and embryologist who coined the term "genetic assimilation", showed experimentally how an acquired trait could be transformed, through Darwinian selection, into a trait that is partially or completely independent of environmental induction. He focused on induced physiological responses in fruit flies. Fruit flies normally have two wings. Waddington induced the development of four wings by treating fertilised fly eggs with ether. Not all treated flies developed four wings. But Waddington selected those that did, bred from them, treated their fertilised eggs with ether, and allowed them to become adults. He then selected again and repeated the whole procedure. In each generation he kept some eggs unexposed to ether and checked whether they developed into four-winged flies. In the first 20 generations, none did. But after 20 generations of systematic selection, a few flies with four wings started appearing in the selected line—even without the ether treatment. The

How language changed the genes

165

trait whose development was at first dependent on external induction by ether became genetically fixed and independent of the ether treatment (Waddington 1953). How did this work? The ability to develop four wings as a result of ether induction has a genetic basis. This genetic basis was exposed by the ether treatment and was then selected. By gradually selecting the gene combinations that produced an ether-induced, four-winged phenotype most effectively, a threshold was eventually crossed, and a particular combination of genes that enabled the development of four wings now appeared without the external inducement. R. F. Ewer and John B. S. Haldane used Waddington's analysis to explain the evolution of behavioural instincts (Ewer 1956, Haldane 1959). Haldane suggested that the innate, excited response of sheepdog puppies to the smell of sheep may have evolved through the genetic assimilation of an initially learned response. For hundreds of years, shepherds selected for dogs that performed their task effectively. Many sheepdog properties were selected for, including the ability to react excitedly to the smell of sheep. The combinations of genes that contributed to this excited response were gradually selected. A response that was initially learned by reward and punishment became an almost entirely automatic response. Another example, this time involving natural and not artificial selection, may be the innate avoidance response of spotted hyenas to the smell of lions, and the avoidance response of many small mammal and bird species to hissing, snake-like noises (Kruuk 1972, Edmunds 1974). Individuals that learn quickly and remember the sound or smell that should be avoided have a better chance of surviving, and the genetic constitution of these fast learners will be passed on to the next generation. After generations of selection for fast association between a certain sense impression and the danger - and for evasive action - the avoidance response will become "innate". Its expression will depend on very few or even a single exposure to the danger stimulus. It is important to note that although in these extreme cases a particular response may become independent of learning, it is in fact the ability to learn that was selected. Learning became increasingly efficient and rapid until it was ultimately internalised. In some cases, there is an additional factor: the learned response changes the individuals' environment. For example, assume that instead of merely learning to detect a predator's smell or sounds more quickly some organisms learn to hide by digging burrows. Here, there will be selection for learning to dig more effectively. However, because burrows amount to a new environment, individuals are now also selected for their ability to live in them.

166 Daniel Dor and Eva Jablonka

Those individuals that are both efficient diggers and efficient burrow-dwellers are positively selected. In this case, the pressure to avoid predators led to what is called niche construction. Organisms actually construct the environment in which they and their offspring are selected (Lewontin 1978, OdlingSmeeetal. 1996). The process of genetic assimilation, which involves the exposure of new genotypes to selection and often also the construction of the selective environment, explains how effectively blind genetic variations can simulate an acquired response within a brief period of evolutionary time. For the process to work we must assume that: a) populations have abundant genetic variation that is relevant to individuals' ability to respond to stimuli; b) different sets of genes become selectively relevant under new circumstances; c) phenotypically visible genetic variation can be recruited and organised into new adaptive genotypes via sexual reshuffling and selection; and d) selection for the adaptive genotypes (genotypes that enable more adaptive responses) is maintained for several generations. What we know of the nervous system and of the abundance of genetic variation in animals not only allows us to make these assumptions, but also suggests that such pro-cesses must have been common during evolutionary history.

4.2. Stretch-assimilate:

the sophistication of behaviour

Avital and Jablonka (2000) discuss an important consequence of genetic assimilation, which enables the lengthening of a sequence of learned behaviours by making a portion of it partially innate. Imagine, for example, a bird capable of reliably learning a sequence of four consecutive acts that culminate in a simple nest. Additional learning is difficult. Assume, for the sake of simplicity, that there is a constraint on the learning capacity of this species of bird. Improved learning ability is unlikely to evolve (perhaps because a large brain requires more energy, or there may be some developmental constraint on brain growth). However, if there is consistent selection for the efficient and reliable performance of the nest-building behavioural sequence, one of the steps becomes genetically assimilated: it becomes innate. The bird now needs to learn only three steps and will construct its simple nest more efficiently. Yet part of its unchanged learning capacity is now freed up. If selection for building good nests continues, the bird can now learn an additional nest-improving skill. For example, it may learn to use plant strips to secure the nest, thus enhancing the nest's stability in windy conditions.

How language changed the genes

167

There are now five consecutive acts, one of which is innate. If building nests rapidly and efficiently continues to be advantageous, another previously learned act can become assimilated and another newly learned one can be added, extending the behavioural sequence by a further step. It is thus possible to gradually lengthen the sequence of acts without changing the capacity to learn. Genetically assimilating previously learned behaviours frees the individual to learn additional acts without extending the limits set by its learning capacity. It is not necessary to assume that any of these acts is completely assimilated. It is sufficient to assume that the number of trials required for the effective performance of the behavioural act is significantly reduced. This is, in fact, the most likely effect of genetic assimilation (Hinton and Nowlan 1987, Behera and Nanjundiah 1995). Reduction in the amount of time and energy spent learning one activity allows for time and energy to be spent on another activity. The stretch-assimilate process may underlie the evolution of complex behavioural sequences that comprise both learned and partially or fully innate components. It could explain the evolution of complex behaviours such as nest building, bird singing - and human linguistic communication.

5. The linguistic spiral We are now in a position to propose an answer to our evolutionary question, the question of the gradual and directional evolution of the linguistic mapping system. Think about the process of language evolution as comprising an arbitrarily long number of stages, and concentrate on two early, consecutive stages: stage Ν and stage N + l . Assume that at stage N, a community of hominids shares the necessary precursors for linguistic communication: a certain minimal theory of the mind, a certain level of conceptualisation, a motivation for information sharing, a certain implicit understanding of social relations and hierarchies, and so on. Assume also that this community uses a preliminary, culturally transmitted system of linguistic or quasilinguistic communication: a system that maps a set of meanings onto a set of phonetic markers. The meanings and their markers need not be recognisable in present-day languages. They may include, for example, ritualised calls, social-emotional vocalisations, and so forth. Whatever stage N's specifics are, one thing is certain. The system's expressive envelope is much more limited than the expressive envelopes of our languages. But, more importantly, the system's expressive envelope is much more limited than the indi-

168 Daniel Dor and Eva Jablonka viduals' conceptual envelopes. Although their conceptual envelope is much narrower than ours, they still - like chimpanzees - can think and feel much more than they can say. Assume further that the individuals in this community use their quasilinguistic system comfortably and naturally and that their children comfortably acquire it. Finally, assume that this community has a particular genetic constitution that allows them to acquire and use the system - with the necessary amount of variability: some individuals are better at acquiring and using the system than others. Now assume that on the way to stage N + l at least two changes occur. First, the communication system developed, and its expressive envelope was expanded and became significantly more sophisticated. Second, the genetic constitution of the individuals in the community changed so that they comfortably acquired and used the more sophisticated system. Our explanation starts with cultural evolution. Assume that at different times throughout stage N, individuals or groups of individuals make linguistic innovations. The driving force behind such innovations must have been associated with a growing pressure for better communication within the group. It may have been related to increases in group size (Dunbar 1996), significant changes in ecological conditions, changes in tool usage, changes in the need for social cooperation, or changes in interactional patterns between different hominid populations. The range of linguistic innovations during the evolutionary process must have been wide, and most of them must have occurred repeatedly: new lexical items for specific referential meanings; new abstract markers for existing and novel conceptual distinctions; new epistemic markers and speech-act markers; new pragmatic conventions for linguistic communication; more sophisticated morphological and phonological structures; more sophisticated usage of linear order and adjacency to mark meanings; and so on. We will not speculate on the order of these innovations (but see Jackendoff 1999 for some interesting hypotheses). Let us assume that some of them occurred at stage N. Like any other type of cultural innovation (a new tool-making technique, say), these linguistic innovations may have been accidental or the result of conscious efforts by clever, inquisitive, or just lucky individuals who happened to be in the right social context at the right time. Many of these individuals were probably juveniles, who are particularly inclined to explore and innovate. In many cases the innovation - like the sign language developed by Nicaraguan children - may have been the result of group effort. The important point is that we do not need to invoke a genetic explanation for any of these cultural innovations. The cultural-linguistic innovations of stage Ν were within

How language changed the genes

169

the genetically based capacity of these talented - or serendipitous - linguistic innovators. Now, although only a small minority in any community is capable of real innovation, a much larger group of individuals is capable of learning to understand and use the innovation once it exists. This also does not require a genetic explanation. Research on chimpanzees and on human children has demonstrated the considerable plasticity of primate cognition. Because comprehension typically precedes production, and because social learning takes advantage of the system's considerable plasticity, the innovator has a good chance of being understood (for example, most of us can understand Newton's theories, though few of us have Newton's genius). This is especially true of the innovator's family members and close friends. Different individuals, however, will differ in their ability to understand and use the innovation. At least some of this variability will stem from variability in their genetic makeup. Some will grasp the innovation better than their peers, some will learn to use it themselves, some will manage to passively comprehend it, and others won't understand it at all. What happens to a cultural innovation once it is learned by a few members of the community? Its fate depends to a significant degree on its propagation and dissemination across the population. This is the real bottleneck. In the innovator's own generation, the propagation of the new linguistic tool may be unstable and uncertain. Many innovations, including some very adaptive ones, will probably disappear at this stage because their significance can sometimes be fully appreciated only when they are used by a significantly large and cohesive group of communicators (there is - up to a point at least - positive frequency-dependent selection). We may assume that innovations have a better chance of establishing themselves after the first learners transmit them to their offspring. Cultural evolution in primates and other animals (like the spread of food washing by Japanese macaques on Koshima island) demonstrates that children play an important role in establishing cultural traditions. We may also assume that for a long time after their invention, innovations undergo further correlated cultural evolution. They may be improved in all sorts of ways as well as become conventionalised and streamlined. The semantic categories, marked by the innovations, will go through a gradual process of differentiation, amplification, and sharpening. They will gradually acquire their discrete character. Moreover, categories will gradually dissociate themselves from emotional connotations, from specific prototypical contexts, and so on owing to their application to a wide variety of social contexts.

170 Daniel Dor and Eva Jablonka An innovation's chances of establishing itself do not, however, depend solely on its pattern of propagation. They are crucially dependent on its adaptive value as a tool of social communication. For a linguistic innovation to survive, its usage should benefit those who adopt it. In general, a linguistic innovation's adaptive value is a direct function of its information potential and an inverse function of its processing effort (Sperber and Wilson 1985). An innovation carries high information potential to the extent that it allows for the transfer of more information which is relevant to the community, to the extent that it adds relevant elements to the system's expressive envelope, and to the extent that it enables more precise production and interpretation. An innovation requires low processing effort by being relatively easy to acquire and use in contexts of social communication. The idea that the survival of a linguistic innovation depends on its adaptive value requires a few additional remarks. First, an innovation's information potential is not perforce related to the practical considerations usually discussed in the literature, such as cooperating efficiently in hunting or fighting as well as sharing information about the natural environment. Although these considerations are important, it seems to us that an innovation's information potential is also a social issue. It is related to the sharing of social information (social relations, social events, and social hierarchies), to the sharing of social narratives and myths, and to the construction of social epistemology (see Knight 1998 and Heeschen 2001). As we know from the recent literature on the relativity problem (see Gumperz and Levinson 1996), linguistic markers help determine the extent to which a community of speakers isolates and foregrounds some aspects of its environment, and thus establishes its epistemological perspective. This epistemological establishment, in turn, plays an important role in linguistically based social identity, which further strengthens the innovation's adaptive value. The categories we discussed in sections 2 and 3 - event structure, epistemic status, and animacy - seem to be especially relevant in this respect (see Dunbar 1996). Second, for a linguistic innovation to survive and propagate it must be adaptive for a sufficient amount of time and preferably in a wide variety of changing circumstances. This is especially true for categorical markers. New lexical items may come and go, but categorical markers that manage to survive probably remain adaptive throughout numerous social changes over a long time span. Third, some types of information - emotional messages, manual instructions - are effectively communicated non-linguistically through body language, facial expressions, mime, song, and dance. Linguistic innovations directed at these types of information may not sur-

How language changed the genes 171 vive (or may not be invented in the first place) because other means of communication render them unnecessary. Division of labour among different communication systems may thus play a significant role in the cultural evolution of language's highly constrained expressive envelope. Fourth, for linguistic innovations to survive they have to meet the conditions set by system constraints. These fall into at least two types. First, linguistic innovations have to comply with psychological constraints. Those innovations that correspond to pre-existing cognitive or developmental biases will probably be selected, since they are the easiest to learn, remember, and transmit (see Sperber 1996). Second, as the linguistic system evolves, it sets its own constraints on new innovations. They have to comply with the already established system. This means that, at least after a certain point in the evolution of language, the system itself dictates the direction of its own future evolution. Let us assume, on the basis of the above considerations, that some of the adaptive linguistic innovations of stage Ν manage to spread and establish themselves in the community. They endure because they are both dependent on, and constitutive of, the social structure and because social traditions are by their very nature self-perpetuating. This cultural change enhances the capacity of the members of the community to communicate. Yet it also raises the demands for social learning imposed on individuals in the community. They not only have to acquire the new innovations in order to participate in social communication, they also have to learn to look at the world in new ways, direct their attention to new aspects of reality, process and remember new types of information, and so forth. In short, the linguistic innovations that establish themselves in the community change the social niche, and the inhabitants of this new niche have to adapt to it. In adapting themselves to the niche, the individuals will probably be able to count on their mind's built-in residual plasticity. Individuals and cohesive groups of individuals that make better use of the innovations for efficient communication (for whatever cultural or social reason) will benefit. They will probably be reproductively more successful than others - and more likely to thrive. Gradually, however, the increasing cognitive demands set by the evolving linguistic niche will start to expose hidden genetic variation. Individuals will find the accumulating linguistic innovations more and more demanding. Residual plasticity will gradually be stretched. After a long period of consistent, directional cultural selection, genetic assimilation will occur. Some individuals will drop out of the race; others will survive. The population will not become genetically homogenous, but the frequencies of the gene com-

172 Daniel Dor and Eva Jablonka binations that contribute to easier language acquisition and use will increase via the continuous processes of genetic reshuffling and selection. Note that genotypes may contribute to acquisition and use in a wider variety of ways. Eventually, at stage N + l , we will find a community whose general genetic makeup enables individuals to use the system comfortably and children to acquire it comfortably. Now the whole process can start over again. Assimilation frees individuals to make further use of their cognitive plasticity. What might be genetically assimilated in our transition from stage Ν to stage N + l ? All the relevant aspects of general cognition were assimilated, to a certain degree at least, according to suggestions by Lieberman (1991), Donald (1991), Jablonka and Rechav (1996), Deacon (1997), and others. Individuals at stage N + l probably were more intelligent, had better memories, had better voluntary control of their sound production mechanisms, and were smarter social agents. We believe, however, that individuals at stage N + l had a cognitive constitution that was marginally more biased toward acquiring and using language than the cognitive constitution of individuals at stage N. In other words, after the long period of cultural evolution in which the community became increasingly dependent on linguistic communication and individuals' survival depended increasingly on their linguistic performance, the process of genetic assimilation must have targeted the cognitive capacities most useful for this specific type of behaviour. Some examples are the capacity to recognise discrete conceptual categories, to rapidly process the speech channel, to recognise linguistic-communicative intent, and to expand lexical memory. These are language-specific and must have been targeted by linguistically driven genetic assimilation. Moreover, the genetic assimilation of these capacities was, for two complementary reasons, probably partial rather than complete. First, genetic assimilation of specific linguistic behaviours cannot lead to a completely innate response because of the variability of the situations in which the linguistic behaviours are adaptive. This is particularly true in our case, where cultural evolution is an integral part of the process. The adaptive value of the linguistic behaviours has to track a constantly changing cultural environment. Second, partial assimilation makes learning easier and faster and reduces the selection pressure for additional assimilation. All this has straightforward linguistic implications. Very specific innovations, such as specific words or specific morphological markers, are not assimilated. They are too variable and context-dependent and change too rapidly throughout cultural evolution. But certain semantic categories can be partially assimi-

How language changed the genes

173

lated because they remain adaptive for a long time across many cultural environments. But even these categories will not be completely assimilated - they will not create a semantic "instinct" - because cultural change puts a high premium on epistemic flexibility. The process of linguistically based genetic assimilation may be related to the general evolution of human culture and human conceptualisation. As we have indicated, genetic assimilation also targets general intelligence. We know, after all, that there was no significant constraint on the evolution of the hominids' general intelligence. Hominid brains doubled in size in 2.5 million years. As the process of cultural and linguistic evolution constantly leads to an extension of the environment as perceived by the community, individuals are constantly faced with more information about the world. They can learn more about more aspects of the world because they can think and communicate more effectively. This creates a process of positive feedback. The more individuals learn about the world, the more they can communicate; and the more they communicate, the more they can learn. On the one hand, individuals and the whole community are now in a position to evolve their conceptual structures with the aid of a more complex communication tool: language. On the other hand, the evolution of conceptual structures, of general cognitive tools for learning, and of remembering aids the concomitant evolution of the linguistic system. The linguistic system thus spirals together with the conceptual system (and with the motor control system, which we have not discussed in this paper). This wider spiral also includes a wide variety of non-linguistic, culturally based evolutionary processes that interact with each other in complex ways. The process resulted both in the expansion of hominids' conceptual capacities and in the construction and expansion of their linguistic expressive envelope. This conception of the process renders theoretically unnecessary the traditional distinction between the syntactic nature of present-day languages and the supposedly presyntactic nature of so-called protolanguage (see Bickerton 1984 and 1990). Because we conceive of the gradual increase in grammatical complexity as a reflection of the gradual increase in the complexity of the expressive envelope, and because the capacity to acquire this complexity was gradually and partially assimilated on the basis of the system's cultural evolution, we conceive of the entire evolutionary process as a gradual and continuous one. The simple and crude structural properties of the mapping system in its first evolutionary stages were a reflection of the system's expressive envelope, just as the complex and sophisticated properties of present-day linguistic mapping systems reflect their elaborated

174 Daniel Dor and Eva Jablonka

expressive envelopes. There were no island constraints, for example, prior to the point at which the interaction between event structure, epistemic status, and the structure of speech acts became a constitutive part of linguistic semantics.

6. Conclusion The framework developed in this paper takes us a long way toward resolving the three constitutive paradoxes presented in the introduction. We started out by characterising language as a cognitive system which is both functional and unique: a transparent mapping system dedicated to the expression of a constrained subset of meanings by means of sound concatenation. As we claimed, recent advances in linguistic research support this conception in direct opposition to Chomsky's traditional hypothesis of the autonomy of syntax. We then characterised the evolution of this system - and the evolution of its social users - as the interaction between cultural and genetic evolution. We discussed the evolution of the linguistic system in cultural terms as the social process of innovation, production, comprehension, transmission, and propagation of linguistic conventions, in which a community isolates and foregrounds certain aspects of its epistemology and develops social agreement about the means of their expression. This process results in a functional (rather than formal) and highly constrained (rather than general-purpose) communication system because it is founded on a selected subset of semantic categories. At each stage of this long and continuous process, the system's expressive envelope expands, and the structural means of expression become more sophisticated. We then discussed the genetic evolution of this system's users in terms of partial genetic assimilation. Partial assimilation resulted in linguistically biased cognition, which enabled easier and more effective language acquisition and use. This conception allows for the resolution of the paradox of domain-specificity. Partial genetic assimilation does not copy linguistic specificities into brain structures, and it does not result in genes for linguistic specificities. Instead, it constructs a genetic make-up that supports the development of a cognition biased towards acquiring and using linguistic specificities (though a significant amount of learning remains mandatory). The question of innateness thus becomes the question of how much, when, and what type of learning are necessary at each stage of the evolutionary spiral. In line with the approach suggested by Elman et al. (1996), our model avoids the nature-nurture dichotomy.

How language changed the genes

175

Our view of the continuous interaction between cultural and genetic evolution is not only consistent with the dynamic nature of languages and with the attested variability among different languages, it actually considers these properties fundamental to the evolutionary process. Languages are constantly changing in their social contexts, and a certain degree of linguistic universality is accompanied by a certain degree of linguistic variability. This is exactly what one would expect as the result of this process. Finally, the framework developed in this paper reconciles the two major approaches to language evolution, one focusing on the evolution of language as a system of social communication, and the other focusing on the evolution of the structurally unique properties of language. As the exchange between Bickerton (1996) and Dunbar (1998) makes clear, scholars adhering to the two approaches have instituted an artificial division of labour. Socially oriented researchers have concentrated on the adaptive value of language as a communication system and largely ignored its formal properties. Structurally oriented scholars have focused on formal specificities and largely ignored social communication. According to our theory, language's formal properties are a reflection of meaning relations. These, in turn, have been selected throughout the evolution of language on the basis of their adaptive value in terms of social communication. The formal question and the social question are one and the same.

3. Beyond biolinguistics

The narration "instinct": signalling behaviour, communication, and the selective value of storytelling Volker Heeschen 1. Introduction: Human ethology, signalling behaviour, and narration The title of my essay echoes the title of Steven Pinker's well-known book. Pinker's "language instinct", however, alludes to language structures, whereas my title refers to language functions. Indeed, I will ignore language structures in this essay and concentrate instead on a single language function: the aesthetic use of speech. I will argue that this particular function correlates with - or perhaps even results from - the conditions underlying verbal behaviour in small societies. This claim is somewhat unconventional. For though books on so-called primitive cultures analyse artefacts like cave paintings and carvings, the verbal arts have largely been neglected. They are supposedly too culture-specific. Moreover, questions of evolution and origin were long considered passe in ethnology and linguistic anthropology. The last several years, however, have witnessed a burgeoning of interest in the biological origins of art, morals, virtue, and beauty (on biology and art see Aiken 1998, Cooke and Turner 1999, Dissanayake 1995, Eibl 1995, and Richter 1999; on the origins of moral behaviour see Ridley 1996 and de Waal 1996). In a similar vein, this essay combines research on present-day verbal art, the evolution of verbal behaviour, and the relationships between speakers' aesthetic repertoire and the behaviours enabled and enhanced by language. It proposes that distancing devices coevolved with the artistic uses of language. In other words, what Roman Jakobson (1960) dubbed the "poetic function" is not only responsible for language's aesthetic repertoire. It is also one of the driving forces behind language evolution. Speech is doubtless a new type of behaviour compared with other forms of communication. The ethnography of communication and linguistic anthropology investigate the functions of speech across different cultures. Yet there is no sub-discipline devoted to studying the evolution of speech. This is partly because speech (unlike, say, tool use) leaves no traces. We can only study present-day communication systems and verbal behaviour. Moreover, ethologists have tended to treat human beings like speechless animals (see Washburn 1978: 414). Yet "etholinguistics" (Eibl-Eibesfeldt 1997:

180 Volker Heeschen

744) and "linguistic ethology" (Sager 1995) can only be pursued succesfully by ethologists (like Irenäus Eibl-Eibesfeldt) who venture into ethnology and by linguists (like Sven Sager) whose research encompasses the nonverbal behaviour of humans and the communicative behaviour of nonhuman primates. Extending the research agenda in this fashion would lead ethologists to topics like the innateness of behaviour, functional equivalences between speech and phylogenetically older communication systems, and the selective value of speech as a behaviour. In other words, it would lead them to topics that have already been studied extensively by psycholinguists, philosophers, and anthropologists. If the ontogenesis of language acquisition has innate aspects, the urge to know the names of things is one of them. Both children and adults experience a feeling of gratification when they learn the name of an unfamiliar object or person. It is as if the mere act of naming enhances cognition. We also feel an urge to break silences during encounters. Human ethology has systematised these observations and proposed that language helps harmonise social life by establishing distance to emotions, by ritualising aggression, and by mitigating conflicts (Eibl-Eibesfeldt 1997: 744). It has also located similarities between humans and other primates: "The grooming behaviour of nonhuman primates participates in networks of social exchange that share many properties with networks of material object exchange in man" (Reynolds 1981: 199). Speech is related homologously or analogously to older communication systems. Yet it is also part of the cultural complex that includes the use of symbols in initiations, architecture, painting, carving, and object exchanges. What is the relationship between phylogenetically older communicative behaviour patterns and subsequent symbolic patterns? We have known since Charles Darwin's 1872 The Expression of Emotions in Man and Animals that humans, like other primates, express emotions physiognomically and gesturally. We also know that nonhuman primates are able to think without using names, to solve problems, and to learn by insight. Nonhuman primates regulate group behaviour via nonverbal signals (see Hauser 1996, Lock and Peters 1996, and Hurford et al. 1998). Humans, by contrast, no longer use only nonverbal signals to regulate behaviour or deal with objects. Instead, humans substitute verbal signs for nonverbal gestures and movements. Moreover, what was once instinctive behaviour is increasingly subject to learning and insight. More than half a century ago Konrad Lorenz (1935) observed that what our instincts tell us is not always sufficient: the gap must be filled by learning. Symbolic behaviour has apparently been decoupled from instinct.

The narration

"instinct"

181

The task of human ethology (and of linguistic anthropology) is twofold. First, it must articulate the path that leads from a code that regulates behaviour to a code that maps environmental and social data. Second, it must articulate a path that leads from thought-without-names to verbal behaviour. In this essay I will not address the common evolutionary phenomena that older functions are accomplished by new morphological means (like behaviour regulation via verbal codes) and that new functions utilise old morphological structures (like language employing the structures of vocal signalling behaviour). Instead, I will concentrate on new functions: representation, narration, and language's inherent aesthetic repertoire. Biologists like Harry Jerison (1976) and Terry Deacon (1998), who study the coevolution of brain size and symbolic behaviour, have only recently begun to follow in the footsteps of philosophers, linguists, and psycholinguists who have long stressed language's representational function, its distance from emotional states, and its independence from the time and place of the objects it refers to. Jerison asserts that speech and language developed as instruments to "map" new biotopes and to optimally exploit new or plentiful resources (Jerison 1976: 101). According to this view, language is primarily a system for referring to space (and to movement within this space) and to the names of places and living beings. That is, language is principally (and was initially) a system for processing information and for mapping the environment. Jerison maintains that we still need language for designing plans of action (plans for movement in space) and for telling stories. Stories "create mental images in ... listeners that might normally be produced only by the memory of events as recorded and integrated by the sensory and perceptual systems of the brain" (Jerison 1976: 101). For Jerison, communication was an evolutionary side effect Psycholinguistic, philosophical, and anthropological research on humans' narrative urge concurs with Jerison's distinction between language as a system for mapping reality and speech as a system for communicating. For example, children use language to refer to objects before they learn language's social uses (Freedle and Lewis 1977). Even very young children tell stories, but if you ask them questions or try to interact with them, they frequently remain silent (on the narration-interaction complex, see Heeschen 1988). Adults experience an urge to learn the names of unfamiliar people and objects. Philosophers like Arnold Gehlen (1971: 199) and Hans Blumenfeld (1996: 41) have also emphasised the importance of naming. Learning the names of objects gives them an aura of familiarity, as if we knew them from our own experience.

182 Volker Heeschen

The primary function of myths and nonsacred narratives is to establish order. Merely observing, like Jerison does, that language is necessary to tell stories obscures language's vital functionality. The ethnological literature has amply demonstrated that language's selectional advantages lie in its ability to simulate. Australian aboriginal myths describe travel routes and distant oases, information that could save the lives of parched travellers (Birdsell 1979). Myths report about land settlement and property rights, information that could help settle ownership disputes between rival tribes. And myths record past gift exchanges, information that could tell subsequent generations where to turn to for help in times of need (see Strathern 1971 and Wiessner and Tumu 1998). Language evolution amounts to a continual levelling out of the differences between language's original mapping function and its subsequent social uses. Language releases humans from primary functional cycles (like eating and copulating) and distances them from their emotions and urges (see Heeschen 1988: 201 and Eibl-Eibesfeldt 1997: 744). Yet language is equally suited to communicating desires, achieving interactional goals, and expressing consent or disapproval by means of digression, misdirection, and play. Hallpike (1985/1986) has demonstrated that although the task repertoires of traditional societies may be comparatively basic, their abundance of symbolic expressions renders such societies highly complex. A single code is not comprehensive enough to handle a wider variety of tasks (begging, for instance, can be accomplished by gestures, glances, and songs). This is why Western societies generalise their codes and use language to reduce ambiguity and to facilitate rapid decision making. They combine language's mapping function and its social uses in ways that differ from those of traditional societies. Deducing the origins of language from humans' present-day communicative functions and artistic practices is necessarily imprecise. Current communication systems may have developed homologously from early man's nonverbal and verbal behaviour, or they may have emerged analogously in the same environments and under the same functional pressures. Moreover, biological evolution becomes less irrefragable once humans enter the picture. Animals necessarily react in a given way in a given situation. Humans can choose. For instance, humans can respond to aggression by submitting silently, by retaliating immediately, by planning future revenge, by appealing to norms, and by using oratory to appease or ridicule the aggressor. Animal signals must be unequivocal and unmistakable. They result from ritualisation. For Wolfgang Wickler (1967), ritualisation is the process by which intended

The narration

"instinct"

183

motor activity (and/or motor activity from other functional cycles) assumes the character of a signal. For example, animals bare their teeth when they are about to bite another animal. Ritualisation is the process by which teeth baring itself becomes a minatory signal. Repetition, simplification, and overemphasis "semanticise" signals and render them unequivocal. But unlike signals, language is decoupled from functional cycles, is inherently equivocal, and only achieves full functionality - and acquires adaptive value- when it simulates a reality outside the interactional context. Human language does not guarantee reliability (Zahavi and Zahavi 1998: 371-373). No human society can forego reliable nonverbal signals. And verbal signals only become reliable when they are ritualised and reintroduced into functional cycles (like greeting formulas) or when they become cliches (see Eibl-Eibesfeldt 1988). Ritualisation results in culturally specific social uses of language and speech. The unreliability of language- an unusual feature in the evolution of communication - invites us to search for its adaptive value. Yet new codes have not completely replaced the older systems of signalling behaviour (see Hondrich 1999). In internet chat rooms, icons reintroduce nonverbal signalling behaviour. Visitors enter and exit chat rooms using formulas that are as ritualised as those of stone-age villages.

2. Speech characteristics in small communities A community's size and population density likely determine the speech characteristics of its members. There are no reliable figures for prehistoric population density. Most estimates are based on existing hunter-and-gatherer societies. Figures vary from 0.03 (Herbig 1986: 81) to 0.3 inhabitants per square kilometre prior to the agricultural age. At 0.1 inhabitant per square kilometre, present-day France would have a total population of 55,000 people (Herbig 1986: 81).' After 5,000 BC and during the Bronze Age, population density may have reached 3 or even 17 inhabitants per square kilometre (see Probst 1991 and 1996, L. and F. Cavalli-Sforza 1994, Renfrew 2000, and Birg 1993). Prehistoric bands likely comprised around 25 individuals, tribes somewhere between 175 and 475. A band of 25 people had a 50 percent chance of surviving for 177 years. They would have needed to cooperate and establish marriage relationships with other bands. Worldwide, the population of farming villages usually does not exceed 150 (see Dunbar 1996: 92-96). According to Forge (1972), 450 is the maximum size of an egalitarian com-

184 Volker Heeschen

munity; above 450, such communities become subject to fission. Social control becomes problematic in communities that exceed this number. In small communities, a softly spoken utterance is sufficient to reprimand an offender, whereas in larger communities such reprimands provoke arguments and protest. Robin Dunbar (1996) contends that the size of the neocortex correlates with the size of primate groups. That is, the larger the group, the more Macchiavellian intelligence is required to deal with the demands of social life. The optimal and maximum size of human primate communities is 150. How do the figures on group size compare with figures on the speakers of single languages? The 750 Papuan languages and the more than 250 Austronesian languages of New Guinea are spoken in an area that measures about 900,000 square kilometres. The average size of a language area is thus about 900 square kilometres - an area that can be traversed on foot (the inhabitable areas are actually somewhat smaller owing to swampland and mountains). This means that each language has an average of 3,000 speakers. Assuming that each language area comprises seven to ten hamlets (as in the valleys of the eastern mountains of West Papua) results in speech communities of 300 to 450 speakers (Heeschen 1992; see also Foley 1986 and Sankoff 1977). Australia and some areas of Africa display similar ratios between population size and language diversity. But considering that some languages in New Guinea are spoken by more than 50,000 people, that exogamous marriage rules and trading partnerships link speakers of different dialects or languages, and that endemic warfare separates speakers of the same language, the average number of speakers per language approaches 1,500 - the presumed size of tribes. Hence, fairly small groups and multilingualism seem to be part of the human condition. The task of this essay is to correlate the small size of ancient and present-day traditional communities and the number of speakers per language with the functions of speech, with verbal and nonverbal behaviour, and with communicative genres. The vast number of proper names for places, mountains, rivers, ecological niches, paths, borders, and settlements serve to map perceived space and the world beyond one's own experience. Myths and songs are sometimes nothing more than lists of names. Moreover, humans are highly oriented towards space. Myths are frequently accounts of wanderings and tales of territorial occupation. Up to half the vocabulary in conversations and texts implies a spatial reference. More than half of the nouns in languages of mountain-dwelling Papuans refer to plants and animals, all of which can be located. To define things is to delimit their location. Indeed, people enjoy

The narration

"instinct"

185

naming and describing imaginary spaces; giving accounts of origins and descents; and discovering kin relationships. Speech is the continual projection of a common social and spiritual territory. It replaces the territoriality of the animal kingdom. It also assigns significance to current events by connecting them to the past. The narration instinct can be thought of as a continuation of children's urge to name people and objects (Lorenz 1973). The naming urge is augmented by the urge to break silences, to give an account of current events, to discover the significance of past events, and to give order to vague plans. Speaking and narrating seem to be means of self-reward and self-gratification. In Western societies we learn to ask questions, make requests, and provide responses in verbal code. In small non-Western societies people communicate socially via nonverbal behaviour. They transmit knowledge via initiation ceremonies and settle conflicts via gift exchanges. My own research has shown that such societies distinguish more strictly between verbal and nonverbal codes. This accords with observations made in other small societies: "Precontact Fore [a mountain tribe in Papua New Guinea] interaction was based largely on common feeling, personal rapport, and familiarity. Subtle interactive behaviour, not questions and instructions, communicated needs, desires, and interests..." (Sorenson 1976: 15). And: "In most circumstances [Australian aborigines] do not attempt to constrain others to do their bidding in a direct, overt manner, and nobody is prepared to take orders from others..." (Kendon 1988: 445). Eipo (in West Papua, where I did research for a number of years; see Heeschen 1998a), Fore, and Australian aborigines use nonverbal codes to transmit information. They do not supplement it with verbal codes, which would draw too much attention to the interaction itself. In small societies, being too blunt is considered aggressive. But all societies transmit desires, needs, and questions verbally, as well. Speakers are confronted with the formidable task of making language - which primarily maps reality - suitable for social uses. They do this by digressing and making detours. They refer to objects outside the immediate interaction, to past events, to absent or never seen objects; they tell stories to appease anger and manage conflicts; and they chat to establish amicable relations prior to reaching decisions. Verbal signs refer to - and simulate - potential actions without initiating them. They divert attention from the immediate interaction, thereby rendering future interactions possible. In other words, verbal signs postpone action. I stress the importance of digressions' referential and representational character because Malinowski's term "phatic communion" and Dunbar's

186 Volker Heeschen

assertion that speaking is a homology of grooming suggest that the content of utterances is irrelevant. Phatic communion does not convey meaning. It is a type of speech in which "ties of union are created by a mere exchange of words" (Malinowski 1923: 315): "[S]peech is the intimate correlate of [humans' well-known tendency to congregate], for to a natural man, another man's silence is not a reassuring factor, but... something alarming and dangerous.... The breaking of silence ... is the first act to establish links of fellowship..." (Malinowski 1923: 314). Yet except for very short, highly ritualised formulas like greeting formulas, neither gossip, idle conversation, nor any utterance longer than a noun or verb phrase can do without language's representational function. All utterances rely on naming, on references to time and space - in short, on symbolic behaviour. Several characteristics of speech favour aesthetic narration. First, speech provides the quintessential means to multiply the steps required to reach a goal. Prolongation makes opposition less dangerous and refusal less likely. For example, when a new kindergartener joins an already existing play group, he or she rarely asks for permission to join the group, although explicit requests are rarely denied (Grammer 1985). The new child establishes contact by observing, imitating, and varying the other children's behaviour. One child wanting to join a play group stood at the window, pointed to a passing aeroplane, and exclaimed: "There's a plane!" The other children stopped playing and looked at the plane. This digression - which refers to an event outside the group - diverts attention, synchronises behaviour, establishes contact, and facilitates admittance into the group. The verbal act is an additional step. This example again illustrates the primacy language's representational function and humans' relatively late acquisition of social language skills. Second, speakers use misdirection to prolong interaction. Misdirection involves digressing as well as feigning a lack of interest in one's aims, needs, or desires. People talk about things outside their group. They adorn their words with song, oratory, and quotations. In small societies, blunt speech merging verbal and nonverbal behaviour during an interaction - would draw other people's attention to the speaker, filling him or her with shame. Such societies use formalised and ritualised forms of speech like songs and fairy tales to deal with important matters. Blunt words are considered offensive and invite resistance. In small societies, openly calling another person a thief or a miser can have only two outcomes. The accused either strikes his accuser or flees the village in order to avoid censorious glances and gossip. This is why everyday speech in small societies is filled with allusions, tropes, irony,

The narration

"instinct"

187

veiled speech, wordplay, oratory, and narration. I heard an Eipo woman refer to her own exhaustion by alluding to smoke rising from a nearby mountain and speculating that it came from a fire kindled by a long-suffering woman from the Fa valley exhausted from her daily tasks. Veiled speech and vague allusions are characteristic features of misdirection. Third, misdirection is augmented by information packaging. Instead of providing concise, precise information, speakers employ prolix utterances in order to maintain secrecy and withhold news. Speech in small communities seems to be less function-oriented and less suited to developing a generalised code for efficient information processing. Speakers are free to comment, tell stories, spin yarns, and play language games. In a society in which everyone knows practically everything there is to know, oratorical skills are used to reawaken interest in familiar topics. Furthermore, new information is too valuable to be revealed prematurely, and small societies handle it "thriftily" (Harrison 1986). Speech is not only a means of communicating, but also of stemming the flow of information and of differentiating oneself from other people. Men set themselves apart from women, initiation groups invent group-specific words, hunters develop their own argot, and speech communities even consciously invent or alter grammatical structures in order to be different from their neighbours and to prevent former allies from understanding them (see Camartin 1992: 39 and 48, Dixon 1997: 13, Heeschen 1998a: 95-102, and Laycock 1982). Fourth, speech fosters group cohesion and establishes bonds of trust and affection. It helps relieve social tensions (Malinowski 1923, Marshall 1961, and Eibl-Eibesfeldt 1997: 744) and is thought to have replaced grooming as a mechanism for social bonding (Dunbar 1996 and Foley 1997: 67-68). Narrative digressions harmonise social life, and misdirection and information packaging appease conflicting parties. Speech turns objects of primary interest - loved ones, enemies, or strangers - into objects of secondary interest. The speaker no longer sniifs, strikes, or stares, but instead sings a love song, tells about the enemy's origin and descent, or initiates a gift exchange. Speech establishes what the German sociologist Alois Hahn has dubbed Konsensfiktion·. fictional consensus (Hondrich 1999: 145). Speech erects a wall of fictions between group members' divergent interests, thereby fostering group cohesion. It can also continue sniffing, striking, and staring by other means. It transposes these actions into a code that is not suited to violence, but that nevertheless invites disagreement. Digression, misdirection, and fictional consensus make language suited to experimentation - and to lying (Dietzsch 1998: 67-86). Volker Sommer

188 Volker Heeschen

characterises lying as an exercise in mental tickling, mind-reading, and assessing (Sommer 1992: 167). Jean Aitchison describes lying as a valuable skill because it "involves displacement - reference to absent or non-existent events... Furthermore, narrating stories is deeply ingrained in all human cultures: most literature is based on the ability to make nonexistent events plausible" (Aitchison 1996: 21).

3. Aesthetic form, play, ritualisation Speech is essentially an organ for processing information. But once it is decoupled from regulating behaviour and from stimulus-response chains, it can be manipulated playfully and artistically as well as enriched Dissanayake would say: made special - by additional structures, tropes, and enigmatic forms. Such aesthetic features are inherent in our own everyday utterances and in the communicative genres of small societies (see Heeschen 1984; 1998a: 30-35). Speech seems to have a predisposition toward art. Are these aesthetic forms by-products of language evolution or do they represent additional adaptations? If narration is adaptive behaviour and if telling stories (as Jerison suggests) has selective value, then language and aesthetic forms probably coevolved. If, on the other hand, symbolic behaviour and representation became gradually (and exclusively) assigned to verbal codes, then aesthetic forms probably developed in later phases of language evolution (though what was initially a by-product could have had adaptive value later). There is no easy answer as to why the verbal code was isolated. Older and more recent communication systems - menacing gestures and verbal threats, for example - continue to supplement each other in face-to-face interaction. Yet isolation and detachment do occur. Information is increasingly transmitted on a single channel. Admonitory speeches, songs, veiled accusations, courting, and verbal duelling have gradually become detached from signals that warn of imminent actions. Myths are handed down as narratives and as drama, dance, pantomime, initiation ceremonies, and simple lists of names. Although the verbal code is universal, the degree of complementarity of the codes is highly culture-specific. Task specificity is crucial. When it comes to comforting, appeasing, teasing, criticising, establishing social bonds, or transmitting accumulated knowledge, speech is a conscious choice from a variety of options. The more or less conscious choice and the isolation of speech from other codes are prerequisites for the appearance of aesthetic forms of communication. These do not function in release-re-

The narration

"instinct"

189

sponse chains and do not form part of composite signals. Instead, aesthetic forms signal play, peaceful interaction, emotional detachment, experimentation, alternative worlds, and make-believe. They are, in short, vital to human beings. This means that smaller units of speech have selective value to the degree that they tend to develop into narratives. In addition to play, the driving forces behind language evolution are probably cooperation, conflict management, and prolonged socialisation (Heeschen 1988: 216-222). In traditional societies, cooperation and conflict management require misdirection, veiled speech, and information packaging. You don't say: "Let's plant a new garden." Instead, you chat, gossip, and narrate. The desired cooperation emerges from the reconfirmation of social bonding - not from a direct request. Elaborate speech fosters cooperation by putting participants in the right mood. And speech's ability to establish fictional consensus helps manage conflicts. Play is an important mediator for elaborate ways of speaking. It is found in children's word games and play acting as well as in secret languages, conscious language alteration, joking, ritualistic insults, and shouting matches of adults (see Goldman 1998, Duranti 1997, Foley 1997, Heeschen 1988 and 1998a; on play and the origins of poetry see Camartin 1992: 184-185). Creativity only emerges during play. Speech consists of a vast number of forms to express a small number needs. Creativity is the positive correlate of what Amotz and Avishag Zahavi call "unreliability" and what Dissanayake refers to as "unpredictability": "In play, novelty and unpredictability are actively sought, whereas in real life we do not usually like uncertainty" (Dissanayake 1995: 43). In the animal kingdom, play is supposed to prepare young animals for the conflicts and alliances of adulthood. When walking alone in the eastern mountains of West Papua, I was inevitably joined by a child or group of children. One of them would take me by the hand, accompany me for awhile, and perform melodious speeches on the virtues of giving. I was told I should give them pearls, peanuts, fish, and rice and allow them to visit me in my hut. Humans are perpetually threatened by hunger and must be capable of addressing strangers in order to obtain food. Perhaps aesthetic communication is a form of experimental play that prepares us for begging, making friends, and forging alliances in real life. Prolonged socialisation enables children to address "strangers", that is, members of their society who do not belong to their intimate "security circles" (see Lawrence 1984: 38-60). The Eipo and Yalenang say that children have reached adulthood when they are able to approach a stranger and beg for sweet potatoes. The step outside the security circle, band, or the

190 Volker Heeschen

speech community is the decisive element. This step leads to related but unfamiliar bands, to less familiar members of large villages, and to members of the language community. Within one's own family, band, or speech community, subtle nonverbal behaviour suffices. Outside this circle, speech is required for clarification, and nonverbal behaviour is reduced, controlled, and formalised. Why do children continue to learn language if one- or two-word sentences and nonverbal behaviour are sufficient for communicating within the security circle? As far as I know, only Jan Gleason has attempted to supply an answer: "[C]hildren have to learn to talk to their fathers and other strangers, and these people are not tuned to them in the warm, sensitive way their mothers are" (Gleason 1973: 293). I interpret Gleason's "father" and "mother" as placeholders for "language community" and "speech community". Ethnological reports on socialisation and childhood show that children are entirely absorbed with establishing social relationships, alliances, and friendships (see Beals 1962: 21-22). Ongoing language learning and honing the capacity to address oneself to strangers are doubtless aspects of prolonged socialisation. Future leaders are usually big talkers. There is a direct relationship between speech, play, and aesthetic forms and the selective value of narration and oratory. Moreover, the size of the speech and language communities correlates with the difference between groups displaying a systematic complementarity of verbal and nonverbal codes and groups in which speech becomes independent. The narration instinct mediates between these two spheres. But play has rules, creativity must be limited in real life, and speakers choose from many styles, genres, and settings. Creativity, play, and the narration instinct become socially useful via limitation and choice. Unreliable and unpredictable signals are rendered unambiguous. The equivocal signals of play and narration are formalised and ritualised. Ritualisation counteracts language evolution. Verbal material can be reduced to cliches, fragments of older or foreign languages, formulas, and names. For ethologists, ritualisation has its origins in signalling motions (like the baring of teeth I mentioned above) and acquires semanticity via repetition, exaggeration, and simplification. The new behaviour subsequently becomes an autonomous urge independent of the behaviour in which it originated (Lorenz 1978: 159). Ritualisation serves two functions. It curbs aggression and fosters social bonding (Lorenz 1978: 157). I believe that speech has developed into an independent urge. We cannot remain silent in the presence of others. We feel the urge to say where we come from and where we are

The narration

"instinct"

191

going. Narration, for example, curbs aggression and fosters bonding by means of digression. Yet what about situations I have labelled Bruchzonen (loosely translated: "danger zones") of social communication: minatory instants when people choose between flight and approach: in encounters with strangers, leave taking, courting, mourning, begging, sharing, admonishing, and managing conflicts? In these danger zones, speaking is no longer or never was - sufficient. Speech is too unpredictable for such situations. Instead, new and unequivocal composite signals gradually developed (or verbal behaviour never achieved complete independence from nonverbal behaviour) to deal with danger zones. For example, eyebrow movement and facial expressions accompany the verbal component of greeting scenes (Eibl-Eibesfeldt 1997: 332 and 633-637). Like play, many aesthetic forms and communicative genres have constraints. They become composite signals because the symbols of authorship, preferred style, setting, and audience contribute to their semanticity. Below are the major constraints and steps towards semanticity (see Dissanayake 1995: 42-49 and Heeschen 1984: 407). First, narration is by definition detached from urges, needs, and desires. Public speeches, sacred narratives, legends, and songs that call for cooperation, cohesion, bonding, or criticism further increase this emotional distance. Nonverbal communication signals are not used. The speaker masks his personality. Second, linguistic means are reduced. A series of proper names and verbal nouns suffices to create form-meaning relations. Conversely, everyday talk and noncommunicative narratives often employ extremely complicated syntax (see Heeschen 1998 a: 319-359). Third, the language of communicative genres refers to, and relies on, the meaningfulness of other semiotic systems (the ordering of objects in space, indexicality, and nonverbal signs). A place name, the name of a clan, and the act of keeping one's hands hidden sufficiently indicate an individual, property, and an unwillingness to share. Whereas narratives can make exaggerated use of tropes to a degree that baffles (or awes) the listener, communicative genres must utilise ironic and veiled statements unambiguously. Fourth, narration tends to use exclusively verbal codes. Some communicative genres re-introduce secondary codes. Myths are sung or use pantomime, public oratory turns into play acting (see Williams 1940 and Salmond 1976).

192 Völker Heeschen

Fifth, narrative and some communicative genres require an audience, whereas poetry, love songs, or speeches criticising prominent people do not. Here, the individual case becomes a general statement. Sixth, in all genres the participants choose just one of many possible options- namely speech - to address a problem, though the degree of manipulation remains distinctive. Narration pleases, oratory excites, myths indoctrinate, and songs appeal to the emotions. Seventh, the space is specified. For play acting, performed myths, dancing songs, and public speeches it is the centre of the village. Moreover, performers have narrowly defined roles. Songs can be sung anywhere, though frequently with one condition: the person who is mentioned or criticised in the song should not be present. What is true for play is true for all kinds of communicative performances: "Often special places are set aside for playing: a stadium, a gymnasium, a park, a recreation room, a ring or circle. There are special times, special clothes, a special mood for play think of holidays, festivals, vacations, weekends" (Dissanayake 1995: 43). Wherever people speak, they symbolically mark off an enclosed and consecrated space. Most genres select - and are defined by - places. Narration flourishes around campfires, greeting songs during encounters on the road, and songs wherever someone feels inclined to express his or her emotions (see Heeschen 1984). Eighth, aesthetic speech creates an atmosphere in which interest is directed away from the original events and toward the pleasure of discovering social meanings in enigmatic forms. This movement away from the object of primary interest and toward the object of secondary interest is precisely what characterises play. Aesthetic speech underscores the metamessage "this is play". This message is inherent in all kinds of speech, especially in every-day talk, gossip, and narration. But it is enhanced in communicative genres by reduction and formalisation so that these genres can serve to curb aggression and foster group cohesion. Ninth, whereas narration flourishes in the security circles, formalised genres address a wider public, namely "strangers". Experimentation and play make sense within groups of intimate partners. Communicating with "strangers", by contrast, calls for greater prudence, namely: ritualisation. What role does biology play? The universality of some arts (like bodypainting) and narrative themes (like patricide) provides only inconclusive evidence that art has a biological basis. It is the way humans process information that offers more convincing testimony. Symbols or cliches that evoke

The narration

"instinct"

193

release-response packages (symbols of territoriality or group cohesion) are universally pleasing. The ability to recognise order and patterns and to decode veiled symbols and messages are prerequisites for the sensory pleasure humans experience in dealing with form-meaning pairings (see Aiken 1998, Boas 1955: 13, Cooke and Turner 1999, and Eibl-Eibesfeldt 1988 and 1997: 899-953): "[H]uman information can be described as... kalogenetic..., a word coined from the Greek kalos (beauty, goodness, Tightness) and genesis (begetting, productive cause, origin, source). The human nervous system has a strong drive to construct affirmative, plausible, coherent, consistent, concise, and predictive ly powerful models of the world..." (Turner and Pöppel 1988:75). According to Dissanayake, who views art behaviourally (as opposed to, say, historically or sociologically), three indications suggest that art is an evolved behaviour: "The first is... that it 'feels good', and so people are positively inclined to do it. The second is that people spend a great deal of time and effort doing it. Frivolous pastimes that take energy and time from useful activity are not selected-for, particularly in large numbers of the population, which leads to the third criterion, universality" (Dissanayake 1995: 33). These three characteristics certainly apply to speech, narration, and most communicative genres. Though speech may be a frivolous pastime, I have already alluded to its survival value for children learning to address strangers. Art consists in making "special": embellishing, exaggerating, patterning, juxtaposing, shaping, and transforming (Dissanayake 1995: 38-83). Begging can also be made special by means of patterning, juxtaposing, and repetition. I believe that making "special" and digressions have survival value in societies where people must constantly step outside the security circle and address strangers. The good feeling engendered by speech results from the features it shares with play and ritualisation. These make it an autonomous urge (see Lorenz 1978: 159). Though Dissanayake admits that making "special" also has survival value in the context of "scenario-building", she mainly refers to activities associated with behaviour in the danger zones of human societies: "objects and activities that [were made special] were parts of ceremonies having to do with important transitions, such as birth, puberty, marriage, and death; finding food, securing abundance, ensuring fertility of women and of the earth; curing the sick; going to war or resolving conflict; and so forth" (Dissanayake 1995: 61). Stories give humans access to alternative worlds, to worlds beyond the horizon, and to behavioural models not sanctioned by familiar rules of sociability and knowledge structures. In traditional societies, borders and horizons are highly significant. The space belonging to an individual or a

194 Volker Heeschen

community is surrounded by symbols. Transgressing or transposing these borders is a symbolic act. Storytellers know how to transcend the ordinary. They know that beyond the horizon things might be radically different. Men might be women. And women might be men. Or, more fantastically, women might own pigs, educate the children, and be the guardians of festivals, manhood, fertility, and thunderstorms. Stories systematically explore alternative worlds (see Eibl 1995: 16). Creating alternative worlds and making them special via aesthetic forms and structures are biologically endowed needs or predispositions. All genres of poetry and narration depict individuals at work in order to make something special and in order to express personal emotions and opinions. Individuals suffer, individuals ruminate on unsolved problems, and individual narrators rapidly change and mix themes, styles, and genres (see Heeschen 1984 and 1998b; on genres see Luckmann 1988 and Foley 1997: 359-360). Papuan societies are highly individualistic. It seems inconceivable that members of these societies do not construct images of personhood and personal identity. Songs in New Guinea can be highly individualistic (see Feld 1982, Harrison 1986, Heeschen 1984, Strathern 1974, and Finnegan and Orbell 1985). However, like small societies' thrifty approach to new (and thus: valuable) information, images of personhood and personal identity must be handled carefully, toned down, and masked. Stressing the speaker's autonomy would be intolerable in the constant face-to-face communication typical of small communities. Numerous stylistic devices in every-day talk, songs, and speeches serve to mask the speaker's personality, and agenthood. Speech is a powerful tool for evoking alternative worlds. It is a way to say what should not be said and to enter social spheres beyond one's own security circle. Narration, ritualisation, self-expression, playful speech, and talking to strangers are all, I believe, part of humans' evolutionary history.

4. Conclusion: narration, hidden information, and veiled communication Over the course of evolution, language's mapping and construction-of-reality function has continuously been rendered suitable for social uses. Having become independent of other codes, the verbal code forms composite signals via ritualisation in the danger zones of social life. Misdirection, veiled speech, information packaging, and making "special" (that is, creating aesthetic forms) are correlates of unburdening, distancing, digression, and the unreliability of speech as communication. These capabilities and func-

The narration

"instinct"

195

tions simultaneously enable speakers to narrate and to create alternative worlds. They are learned during play and enable speakers to act prudently before a wider public and to use ritualised and formalised genres when approaching strangers. Nevertheless, narration and subtle nonverbal behaviour predominate communication in the security circles. The general aim of speech is to create fictional consensus. Finally, there is the individual's urge to express himself or herself, an urge that is less well understood. In small societies, self, ego, agenthood, and personal interests can only be expressed by aesthetic forms. Here, speech is less burdened with social functions, appeals, criticism, and self-expression must be masked using artistic means. Aesthetic forms of communication thrive in small societies because of the general rule of misdirection and the restrictions that apply to blunt language. Ritualisation has presumably transformed humans' urge to break silences, to name objects, places, and persons, to recount their own origin and descent, to create playful identities, and to tell stories. It refers to the biology of behaviour and to universal dispositions. Orality and literacy, genres, styles, narration, drama, and poetry are all subjects of extensive research. Yet we rarely ask ourselves why human beings narrate and why they trouble themselves about secrecy, style, making things "special", and creating beautiful things that are neither true nor easily understood. If the aesthetic function is assumed to be at work in all utterances - that is, in both every-day talk and elaborate communicative genres - one might even expect speech and particular language structures to coevolve with humans' drive to tell stories, to keep secrets, and to artistically narrate alternative and nonsensical worlds. In my opinion, the narration instinct is ready to be taken seriously as a concept. Early human societies were small and isolated. Knowledge was required to arrange meetings at certain times and places. Peaceful cooperation was an urgent necessity for bands of up to 25 individuals. Establishing marriage relationships and trading partnerships was vital in speech communities fewer than 450 people and in language communities fewer than 1,500 members. Means to approach strangers were needed. These factors suggest that mapping the real world, creating alternative worlds, making things "special", making detours, specifying the tasks of distinct codes, and learning the oratory required for communicating with strangers all had survival value. And it is at least possible that there was a correlation between narration and the selective values of misdirection, aesthetic forms, and reality simulation. Distinguishing between speech and signalling behaviour is an important part of the search for the diversity of language origins.

196 Volker Heeschen

Note 1. Herbig quotes Η. Μ. Wobst, "Boundary conditions for palaeolithic social systems: a simulation approach", American Antiquity 39.2: 147-148. See also Beaken 1996: 125-126, where the relationship between language, exogamous marriage rules, and long distance exchange is mentioned. Though settlement and spreading of homo sapiens, cultures and languages is now widely taken up by geneticists, linguists, and archaeologists, I have not found new figures and calculations, a remark that is also valid for population density. Thematisation of size, density, and language communities is practically nonexistant.

Taxonomic controversies in the twentieth century Merritt Ruhlen At the beginning of the twentieth century there were a number of hotly debated taxonomic controversies in linguistics, with scholars sharply divided into two camps. One camp argued that linguistic taxonomy had already progressed as far as it possibly could, and that attempts to find relatives of Indo-European were doomed to failure for the simple reason that evidence of genetic affinity disappears after 6,000 years - the presumed age of IndoEuropean - and thus no evidence could possibly still persist from earlier times, even if such evidence had once existed. Similar proposals to find relatives of Basque, Algonquian, and hundreds of other languages and language families were rejected for the same reasons. Obviously, in this context, any attempt to argue for monogenesis of all extant languages was met with disbelief and hostility. The second camp saw things quite differently. It argued that IndoEuropean's closest relatives were already quite obvious, as were the immediate relatives of Basque, Algonquian, and many other languages and language families considered "independent" by the first camp. And some members of the second camp - notably the Italian linguist Alfredo Trombetti - even dared to argue that existing evidence already quite strongly supported the idea of monogenesis. What is remarkable is that these same controversies remained even more hotly debated at the end of the twentieth century and, though the participants in the debate had of course changed, the controversies themselves had often remained remarkably similar in form and content. In this paper I will examine several of these controversies and argue that it is the ideas of the second group, particularly those of the much-maligned Trombetti, that are winning the debate. The linguistic evidence today is much richer than it was in Trombetti's day, and it confirms in virtually every respect Trombetti's daring hypotheses of a century ago.

1. The Sapir-Michelson debate I will begin with an obscure taxonomic dispute that occurred in the second decade of the twentieth century. In 1913 Edward Sapir announced a surprising discovery, namely, that two languages located side by side on the northern

198 Merritt

Ruhlen

California coast line - Wiyot and Yurok - were most closely related to the vast Algonquian family that extended from Montana to the eastern seaboard. In support of this relationship Sapir offered some 200 lexical and grammatical similarities. But the most potent piece of evidence was the virtual identity of the pronominal prefixes in the three groups.

Table I. Pronominal prefixes in Wiyot, Yurok, and Algonquian. my

your

his

someone's

Proto-Algonquian

*ne-

*ke-

*we-

*me-

Wiyot

du(?)-

khu(?)-

u(?)-

b-

Yurok

?ne-

k?e-

?we-/?u-

me-

The task of judging this evidence fell to the leading Algonquianist of the day, Truman Michelson, who recognised that "the importance of this discovery, if valid, can hardly be overestimated" (Michelson 1914: 362). Michelson's verdict was, however, entirely negative, and he dismissed all of Sapir's evidence as "fancied lexicographical similarities", misanalysis of morphological elements, accidental resemblances, and features in Wiyot and Yurok that were thoroughly un-Algonquian. His conclusion left no doubt that he considered Sapir's discovery without merit: "Enough has been said to show the utter folly of haphazard comparisons unless we have a thorough knowledge of the morphological structure of the languages concerned" (Michelson 1914: 367). As a consequence of Michelson's opposition the Algic hypothesis (Algonquian + Wiyot + Yurok) became one of the unresolved taxonomic controversies of the twentieth century. In the early 1950s, during his work on the classification of African languages, Joseph Greenberg examined a number of taxonomic controversies around the world, one of which was the Algic hypothesis. Upon examining the SapirMichelson debate and its attendant evidence, he concluded that this relationship was, in fact, "not very distant... and, indeed,... evident on inspection" (Greenberg 1953: 283). The real mystery was why anyone thought there was a mystery. During the 1950s the relationship became accepted and the decisive proof was often attributed by American Indianists to Mary Haas's work. Haas herself made no such claim. In fact, she seconded Greenberg's opinion and merely gave additional evidence for the relationship, implying, correctly, that

Taxonomic controversies

in the twentieth century

199

it was Sapir who had proved the relationship beyond a reasonable doubt (see Ruhlen 1994a for a more detailed discussion). There are, of course, just four possible explanations for linguistic similarities: common origin, borrowing, accident, and onomatopoeia. Which explanation best accounts for the similarity of the Algic pronominal prefixes? Clearly onomatopoeia can be ruled out since there is no intrinsic connection between any pronoun and any particular sound, despite ill-founded and unsubstantiated claims by Johanna Nichols (1992) and others. The possibility that these resemblances are accidental may also be easily ruled out. While the probability that such similarities could arise by accident is not zero, it is as close to zero as we need to get in historical linguistics. Borrowing too may be eliminated since the nearest Algonquian language (Blackfoot) lies over 600 miles to the east, on the other side of the Rocky Mountains, and there is no evidence that either Wiyot or Yurok was ever anywhere near an Algonquian language. A claim of borrowing, which has in fact never been made, would be little more than a deus ex machina. The only reasonable explanation for such pronominal similarities is common origin, as Sapir, but not Michelson, realised. In a letter to Sapir, Alfred Kroeber wrote "Michelson's review strikes me as puritanical: I have never had any doubt of the validity of your union of Wiyot and Yurok with Algonkin.... I hardly consider it worth while seriously to refute Michelson. His attitude speaks for itself as hypercritical and negative.... I regard the case in point so one-sided as to be already conclusively settled" (Golla 1984: 153). Less well known is that the Algic relationship was independently discovered by Trombetti, though later than Sapir. In sum, the Algic "controversy" was little more than one scholar's inability to see the obvious. Scholars with a broad knowledge of languages and an understanding of taxonomy - Sapir, Trombetti, Kroeber, Greenberg - realised immediately that the Algic pronominal similarities could only reasonably be explained by common origin. Narrow specialists like Michelson assumed - or perhaps hoped - that there could be some other explanation. It is quite clear that had Michelson congratulated Sapir on his brilliant discovery, as Kroeber did, there never would have been any controversy at all. But Michelson's position of power as the leading Algonquianist of the day allowed him to initiate a pseudo-controversy that lasted over half a century. The lesson to be learned from the Sapir-Michelson debate has been aptly stated by Greenberg in a recent article: "I believe that one lesson of the Sapir-Michelson controversy is that 'controversial' is not to be equated with 'doubtful'" (Greenberg 1997: 669).

2 0 0 Merritt

Ruhlen

2. The Amerind controversy The second taxonomic dispute I would like to examine is the Amerind controversy, initiated in 1987 by the publication of Greenberg's Language in the Americas, in which he argued that all New World languages belong to one of three groups, Eskimo-Aleut, Na-Dene, and Amerind. Of these, the first two had long been accepted as valid families; the controversy centred on Greenberg's claim that all other American Indian languages belonged to a single family, Amerind. This claim was at sharp variance with the prevailing opinion among American Indianists, who believed that what Greenberg called Amerind was really a group of over 200 unrelated families, or, to be more precise, 200 families among which there was no evidence of genetic affinity (Campbell 1997). Specialists concede that some of these families may be related, but the time depth is so great that any evidence of this fact would have long since disappeared. Once again pronouns are among the most convincing pieces of evidence, for Greenberg provided abundant examples of a specifically Amerind pronominal pattern: na Τ and ma 'you'. In fact precisely this pattern had been noted at the beginning of the past century by both Trombetti and Sapir. In a book published in 1905 Trombetti devoted an entire appendix to documenting this pattern in the Americas and concluded that, "from the most northern regions of the Americas the pronouns NI Ί ' and Μ 'thou' reach all the way to the southern tip of the New World, to Tierra del Fuego. Although this sketch is far from complete, due to the insufficient materials at our disposal, it is certainly sufficient to give an idea of the broad distribution of these most ancient and essential elements" (208). In a personal letter written in 1918, Sapir wrote: "Getting down to brass tacks, how in the Hell are you going to explain general American η- Ί ' except genetically? It's disturbing, I know, but (more) non-committal conservatism is only dodging, after all, isn't it? Great simplifications are in store for us" (quoted in Ruhlen 1994b: 87). Franz Boas, aware of the widespread n/m pattern, nonetheless did not explain it as due to common origin, but rather attributed it to "obscure psychological causes". Surprisingly, even though the n/m pattern in the Americas was recognised long ago, Greenberg's critics, such as Lyle Campbell, have claimed that this pattern is not particularly frequent in the Americas, no more frequent, according to Campbell (1994), than the m/t pattern that we will see characterises language families of northern Eurasia. I have shown elsewhere (Ruhlen 1995a) that Campbell's claim is false, but his assertion pinpoints

Taxonomic controversies

in the twentieth century

201

one of the major weaknesses of contemporary linguistics: the lack of a linguistic database that would quickly disprove Campbell's assertion. In addition to the distinctive Amerind pronominal pattern, there is a lexical item characterising the Amerind family that by itself virtually guarantees the validity of the family. Throughout North and South America there is a root t-na (- indicates an indeterminate vowel) with the general meaning of 'child, son, daughter'. A comparison of hundreds of such forms indicates that, in the original Amerind system, the first vowel in the root was correlated with the gender of the child. Thus, the root had three morphologically determined grades: masculine t'ina 'son, brother', feminine t 'una 'daughter, sister', and indeterminate t'ana 'child, sibling' (Ruhlen 1994c). No contemporary Amerind language preserves all three grades of this gender ablaut system in this particular root, but a number have retained two (e.g., Iranshe atina 'male relative', atuna 'female relative'; Tiquie ten 'son', ton 'daughter'). Elsewhere, however, a single language can preserve all three grades, for example in the Tucano numeral 'one': nik-e 'one (masculine)', nik-o 'one (feminine)', nik-a 'one (indeterminate)'. It is noteworthy that this numeral is the general Amerind numeral for 'one' (Ruhlen 1995b). Languages retaining only one grade of the root are far more abundant. Examples of the masculine grade include Molala pne: -t'in 'my older brother', Yurok tsin 'young man', Mohawk -?tsin 'male, boy', Proto-Tzeltal-Tzotzil *lih-ts'in 'younger brother', Cuicatec Idiino 'brother', and Yagua deenu 'male child'. Feminine examples are Central Sierra Miwok tu.ne- 'daughter', Salinan a-t'on 'younger sister', Tacana -tona 'younger sister', and Piokobye a-ton-kä 'younger sister'. Examples of the indeterminate grade are Nootka t'an 'a 'child', Coahuilteco t'an-pam 'child', Proto-Uto-Aztecan *tana 'son, daughter', Aymara tayna 'first-born child', and UrubuKaapor tal+in "child". There is no intrinsic connection between any vowel and any particular gender, and in fact in Afro-Asiatic i is feminine and u is masculine. The combination of this particular root (t-na 'child') with this particular gender ablaut system (i/u/a 'masculine/feminine/indeterminate') is a trait found only in Amerind languages. Its explanation can only be genetic. Significantly, the Amerind gender ablaut system, in conjunction with this root, can be reconstructed on the basis of just North American languages, or just South American languages. One might infer from this fact that the Amerind population must have passed through North America rapidly enough that the entire ablaut system reached South America intact. Had South America been populated by people with languages that retained only

2 0 2 Merritt

Kuhlen

two or just one grade of the root, the entire system could never have been put back together. This concords well with the archaeological record in the Americas, which appears to indicate just such a rapid migration by the first Americans throughout North and South America around 10,500 years ago (Klein 1999).

3. The isolation of Indo-European The third controversy that I would like to focus on is the supposed isolation of the Indo-European family. Throughout the twentieth century most historical linguists maintained that Indo-European had no known relatives. Nevertheless, the Indo-European family was the subject of two disputes in the twentieth century with regard to putative relatives. I will discuss them in reverse chronological order.

3.1. The discovery of Hittite It is largely forgotten today that the discovery of Hittite led initially to a dispute over whether Hittite (and later the other Anatolian languages) was, or was not, a member of the Indo-European family. At first scholars such as Antoine Meillet argued that Hittite was too different from the other IndoEuropean languages to be considered a member of the family: "The interpretation [of Hittite texts] is still very hypothetical, and the assertion that Hittite is Indo-European would seem to be very risky; it has been disputed by most of those who have examined the documents" (Meillet [1914] 1965: 99). Yet only a few years later the presence of certain diagnostic IndoEuropean traits led Meillet and others to conclude that Hittite was in fact an Indo-European language after all: "The Indo-European character of Hittite immediately struck the first interpreters" (Meillet [1922] 1964: 38). It is seldom realised today, when no one questions the Indo-European nature of the Anatolian languages, that Meillet's initial position was in fact correct. Hittite and the other Anatolian languages are not members of the Indo-European family as that family was understood at the time of its discovery. Part of the problem in this debate was that scholars focussed on two options: Hittite is Indo-European, or it is not Indo-European. What was overlooked was the third option: that Hittite is not Indo-European, but is related to Indo-European. It was Edgar Sturtevant who first pointed this out

Taxonomic controversies

in the twentieth century

203

in 1933 when he argued that Hittite was really a sister language to IndoEuropean, not a daughter language. According to Sturtevant, this higherlevel family, which he called Indo-Hittite, consisted of two branches, Hittite (Anatolian) and Indo-European, as that term had been understood before the discovery of Hittite. As Sturtevant pointed out, there were numerous traits shared by Indo-European languages which were absent in Hittite, one of which was the Indo-Hittite two-gender system, which developed in the IndoEuropean branch into a three-gender system. Paradoxically, Sturtevant's Indo-Hittite hypothesis was generally opposed by Indo-Europeanists during the past century (Warren Cowgill was a major exception), even though somewhat schizophrenically - they conceded its fundamental correctness: "A number of archaic features in morphology and phonology set Anatolian apart from the other branches, and indicate that it was the earliest to hive off. But Anatolian remains derivable from Proto-Indo-European; and periodic efforts to situate Anatolian as a sister language to Indo-European, with both deriving from a putative 'Indo-Hittite', have not found a following" (Watkins 1992: 209). What happened in the twentieth century was that taxonomy and genetic relationship were confused. Because Hittite had been convincingly shown to be related to Indo-European, it must therefore be a member of the family and therefore the term Indo-Hittite is not needed. But what really happened was that the term Indo-European was redefined to include languages that did not fit the original definition. This then raised the question of what to call the non-Anatolian branch of the redefined Indo-European, whose name had been usurped for the higher-level taxonomic unit. One finds in the literature that Indo-European is then referred to as "Early Indo-European" and the non-Anatolian branch is referred to as "Late Indo-European". But of course these two terms are simply different names for Indo-Hittite and IndoEuropean. In biology, where names of taxa are not allowed to be changed, this terminological confusion would not occur.

3.2. Nostratic The confusion of genetic relationship and taxonomy is even clearer in the case of Nostratic. The Nostratic hypothesis - that Indo-European is related to certain other families - was first advanced by Holger Pedersen in the first decade of the twentieth century. Indeed, the very definition of the family was based on relationship, not taxonomy, for Nostratic was defined as those

2 0 4 Merritt

Ruhlen

families that are related to Indo-European. It should be clear that a taxon cannot be defined by a relationship to some particular family. Of course one of the motivating factors for such a definition was that Indo-European was not supposed to be related to any other family, and the Nostratic hypothesis was an attempt to break this supposed barrier. One consequence of this definition was that Nostratic has included different families for different scholars. The Nostratic of Vladislav Illich-Svitych (1971-84), sometimes referred to as Classical Nostratic, included Indo-European (by definition), Afro-Asiatic, Kartvelian, Uralic, Altaic, and Dravidian. Aaron Dolgopolsky's (1984) version, developed at about the same time, also included ChukchiKamchatkan and Eskimo-Aleut, but omitted Dravidian. And both of these differ from the original conception of Pedersen, which contained Semitic, Finno-Ugric, Samoyed, Yukaghir, Altaic, and Eskimo-Aleut. Such differences are in part due to the improper definition of Nostratic. If, as I will argue below, all the world's language families are related, then Nostratic becomes identical with Proto-Sapiens. In recent years the accuracy of the Nostratic hypothesis has once again become the focus of discussion in historical linguistics (Renfrew and Nettle 1999), and one often hears the question "Is the Nostratic hypothesis correct?" It is crucial, however, to distinguish two different questions in this regard. First, does the Nostratic evidence prove that Indo-European is related to other families? Second, is Nostratic - in any of its definitions - a valid taxon, that is, a set of language families more closely related to one another than to any other family? In my (and Greenberg's) opinion the answer to the first question is yes, but the answer to the second question is no. Certainly the evidence offered by the Nostraticists shows overwhelmingly that IndoEuropean does have relatives, as was clearly recognised by Pedersen and Trombetti early in the twentieth century. However, none of the definitions of Nostratic are valid taxa. For example, in Illich-Svitych's version AfroAsiatic and Dravidian are included in the family, yet Chukchi-Kamchatkan and Eskimo-Aleut are not. But in fact these latter two families are clearly more closely related to Indo-European than is Afro-Asiatic or Dravidian (Greenberg 2000). Thus Illich-Svitych's version is not a valid taxon. In fact, it has recently been conceded by most Nostraticists that Afro-Asiatic is really a sister to Nostratic, not a daughter (Dolgopolsky, however, still includes Afro-Asiatic). It is for these reasons that Greenberg's version of Indo-European's closest relatives differs from the various versions of Nostratic. Greenberg's version, which he calls Eurasiatic, includes Indo-European, Uralic-Yukaghir, Altaic,

Taxonomic controversies in the twentieth century

205

Korean-Japanese-Ainu, Gilyak, Chukchi-Kamchatkan, and Eskimo-Aleut, as shown in Figure 1. Greenberg believes that these families do form a valid taxon. Several Nostraticists have recently come to conclusions similar to Greenberg's, and Allan Bomhard considers Eurasiatic one branch of a larger Nostratic family (Bomhard and Kerns 1994). The only differences in Sergei Starostin's view from that of Greenberg is that he would include Kartvelian in Eurasiatic, but not Ainu.

Atlantic Ocean

Μ

Indo-European

Altaic

1 ]

Japanese-Korean-Ainu

Ü1

Uralic

Gilvak

tSl

Chukchi-Kamchatkan

Figure I. The Eurasiatic language family. I will discuss Indo-European's relatives in terms of Greenberg's Eurasiatic family, though much of the evidence for Eurasiatic appeared first in the Nostratic literature (as Greenberg readily acknowledges). What is the evidence for Eurasiatic? One of the most salient pieces of evidence is the specific pronoun pattern mit Ί/you', already clearly recognised by Pedersen, Trombetti, and others at the beginning of this century. Trombetti remarked somewhat acerbically in 1905 that "it is clear that in and of itself the comparison of Finno-Ugric [Uralic] me T , te 'you' with Indo-European me- and te- [with the same meaning] is worth just as much as any comparison one might make between the corresponding pronominal forms in the IndoEuropean languages. The only difference is that the common origin of the Indo-European languages is accepted, while the connection between IndoEuropean and Finno-Ugric is denied" (Trombetti 1905: 44). Table 2 lists a few Eurasiatic cognates.

2 0 6 Merritt

Ruhlen

Table 2. Eurasiatic cognates. I, me, my

thou, thy

Indo-European

*me-

*tu

Uralic-Yukaghir

*-m

*te

Turkic

men

Mongolian

mini

Tungus

mini

Korean

-ma

who? k l-

what?

plural

'ma

*ke

*mi

*-k

*-t

*kim

*mi-

iki

-t

*ti

ken

*ma

ikire

*-t

-ti

*xa-

*ma

-ka

mai

Japanese

ml

Ainu

mak

Gilyak

two, dual

ti

-ka

-te

-ki

-ti

-gi

-t

Chukchi-Kamchatkan

-m

-t

k'e

mi-

-k

-ti

Eskimo-Aleut

-ma

-t

*kina

*mi

-k

-t

However, in addition to the evidence already advanced by the Nostraticists, Greenberg's book contains additional evidence that leaves no doubt whatsoever that Indo-European can hardly be considered an "isolate" in any sense. While the distinctive M/T pronominal pattern, by itself, constitutes compelling evidence, there is an additional complication in the first-person pronoun that is even more decisive. While first-person Μ is characteristic of every Indo-European language - and is abundantly attested in other branches of Eurasiatic - there is a peculiar subject form of this pronoun that has never been satisfactorily explained. Indo-European shows different roots for the subject and object forms of the first-person pronoun: English Ί , me', French je, me, Russian ya, menya, and so forth. The Proto-Indo-European reconstructions of these two forms are *eg(h)om Τ and *me 'me'. This particular suppletive alternation has always been considered a diagnostic trait of the Indo-European family. After all, the possibility that two unrelated languages would independently invent the same suppletive alternation is unlikely in the extreme. Greenberg shows, however, that this suppletive alternation, far from being an innovation of Indo-European, is in fact a trait that Proto-Indo-European inherited from an earlier Proto-Eurasiatic language.

Taxonomic controversies

in the twentieth century

207

The clearest evidence comes from Chukchi-Kamchatkan, at the other end of Eurasia, which shows a pronominal paradigm for the first- and second-person pronouns that is virtually identical with that of Proto-Indo-European: evam Ί ' , ma 'me'. Furthermore, Chukchi-Kamchatkan has extended the pattern to the second-person, erst 'thou', in which second-person -t replaces first-person -m. The extension of this pattern is also found in Uralic, for example, in the Hungarian object pronouns en-gem(-et) 'me' and te-ged{-et) 'thee'. Greenberg's hypothesis regarding the origin of this suppletive alternation is that Proto-Eurasiatic *egom Ί ' was originally a periphrastic form of the first-person pronoun used for emphasis. Just as in French c 'est moi 'it's me' can be used in place ofje for emphasis, a typologically similar development occurred in Proto-Eurasiatic, and the morphological analysis that Greenberg gives for *egom, *e-go-m 'this-is-me', is identical to that of French. The Kamassian language in Uralic supports exactly this analysis: i-gä-m Ί am' (= 'this-is-me'). Clearly neither onomatopoeia, accident, nor borrowing can be taken seriously as an explanation of these facts. Only common origin provides a plausible explanation for such pronominal similarities. The dozens of other grammatical items, and hundreds of shared lexical items pointed out by the Nostraticists and Greenberg, only confirm what can be surmised on the basis of just two pronouns.

4. A family of isolates: Dene-Caucasian The fourth controversy that I will discuss is the question of language isolates, of which Basque and Ket are two of the most famous. One of the more exciting developments in the past two decades has been the identification of a language family, now called Dene-Caucasian, that includes several of these supposed isolates. The six branches of this family - Basque, Caucasian, Burushaski, Ket (Yeniseian), Sino-Tibetan, and Na-Dene - are shown in Figure 2. The current conception of this family derives from the work of Starostin (1984), Sergei Nikolaev (1991), and John Bengtson (1991), though as usual there were precursors, and as usual one of the primary precursors was Trombetti, who devoted an entire monograph to the origins of the Basque language: "I connect Basque most closely with the Caucasian family. But this linguistic group is then most closely related to the Sino-Tibetan family" (Trombetti 1923: 6).

2 0 8 Μerritt

Ruhten

Figure 2. The Dene-Caucasian family.

Most of the recent work on Basque is the work of John Bengtson (1991) (the Nostraticists avoid languages that cannot be reconstructed). Table 3 shows six of the roots that characterise Dene-Caucasian. Note that Basque shares both the general Dene-Caucasian interrogatives, one in Ν and one in S. Note further that these interrogative pronouns are completely different from those of Eurasiatic, in which Κ, M, and Y form the basic interrogatives. Table 3. Dene-Caucasian cognates. Family

who?

what?

dry

day(light)

water

hungry

Basque

no-r

se-r

agor

egun

ur

gose

Caucasian

*na

*sa

*-GWVr-

*-GinV

*Tiwiri

*gasi

Burushaski

ana

be-SA-n

qaqar

goon

hur-

w

Sino-Tibetian

*naai

*su

*qar

*1 *k aai]

Yeniseian

*?an-

*sV-

*qVr

*ge?n

sa

*-Gaq

kuuq

Na-Dene

*xur *gas

Taxonomic controversies

in the twentieth century

209

5. Criticism What are we to make of the criticism levelled at Greenberg and the Nostraticists in recent years and the similar criticism aimed at Trombetti and Sapir at the beginning of the twentieth century? Space permits only a brief discussion of some of the most common criticisms.

5.7. The temporal limit of the comparative method: 6,000 years It has become an accepted dogma of contemporary historical linguistics that the Comparative Method is limited to roughly 6,000 years, before which time linguistic change has obliterated all evidence of genetic affinity. We have already seen many examples where this is clearly not true. The 6,000 year limit appears to be tied to the supposed age of the Indo-European family. In a sense it explains why Indo-European has no relatives, which, of course, we have already seen is incorrect. Furthermore, the Afro-Asiatic family must be over 10,000 years old since the language was pre-agricultural. There is really no empirical basis for the supposed 6,000 year limit on comparative linguistics; it is simply a self-imposed limitation of twentieth-century historical linguists.

5.2. Reconstruction and genetic affinity Probably the most common criticism of Greenberg's work is that he has not reconstructed anything and therefore he has not followed the real comparative method. It has become a dogma of contemporary historical linguistics that only reconstruction can prove genetic affinity; this stricture appears in virtually every historical linguistics textbook. According to Hock, for example, only reconstruction proves genetic affinity, and Indo-European, Uralic, Dravidian, Austronesian, Bantu, and Uto-Aztecan have all been proved by successful reconstructions (Hock 1986: 567). And yet all of these families were universally accepted as valid families before anyone even thought of trying to reconstruct the protolanguage. If reconstruction proved Indo-European, as Hock claims, then who proved it, and when? When I posed this question to two historical linguists at a meeting at the University of Cambridge in 1998 - Don Ringe and Roger Lass - they replied that it was neogrammarians like Brugmann and Delbrück who had proved Indo-European. This would come as

210

Merritt

Ruhlen

quite a shock to these scholars, who never imagined that they were proving Indo-European. Indo-European had long been accepted as a valid family by everyone, and reconstruction was never cited as "proof" of anything. This idea is entirely an innovation of twentieth-century scholars; I have found no trace of this bizarre notion in the nineteenth century. It is instructive to look at what the neogrammarians themselves had to say about the basis of genetic affinity, and it has nothing to do with reconstruction. Delbrück stated that "it was proved by Bopp and others that the so-called Indo-European languages were related. The proof was produced by juxtaposing words and forms of similar meaning. When one considers that in these languages the formation of the inflectional forms of the verb, noun, and pronoun agrees in essentials and likewise that an extraordinary number of inflected words agree in their lexical parts, the assumption of chance agreement must appear absurd" (Delbrück 1880: 121-122). Delbrück considered Indo-European to have been proved by the time of Bopp at the beginning of the nineteenth century, and the basis of this proof was the "juxtaposition of words and forms of similar meaning", a virtual paraphrase of Greenberg's methods. If one were to tell a biologist that one should not believe in mammals because no one has ever reconstructed Proto-Mammal and then explained how this animal evolved through all the intermediate mammals before arriving at the current array of 4,006 species, the biologist would laugh and think it's a joke. Yet if you tell a traditional historical linguist the same thing regarding, say, Amerind, he will nod solemnly and say "of course". The problem with modern historical linguistics is that the very nature of the comparative method has been misunderstood. Reconstruction has been confused with taxonomy. In reality the Comparative Method consists of two separate stages, Taxonomy and Historical Linguistics, and it is taxonomy that determines genetic affinity and necessarily precedes the normal concerns of historical linguistics. It should be obvious that one cannot reconstruct a protolanguage until one has somehow identified a group of related languages, and it is the first stage of the comparative method, taxonomy, that identifies groups of related languages, just as Delbrück explained. One finds, however, that in contemporary historical linguistics the reconstruction of protolanguages - with of course regular sound correspondences - is called the Comparative Method. Taxonomy has disappeared from modern historical linguistics, and one searches in vain in any modern historical linguistics textbook for any discussion of classification.

Taxonomic controversies

5.3. Internal

in the twentieth century

211

subgrouping

Another criticism that has been levelled at Greenberg's Amerind hypothesis is that he has not worked patiently backward through all the intermediate nodes, but has simply jumped to conclusions - Amerind - without even establishing the validity of the intermediate families. It should be obvious, however, that it is often far easier to discern more ancient groups than chronologically younger ones. For example, the Austronesian family, which extends from Madagascar to Easter Island and Hawaii and is thought to be about 6,000 years old, was recognised in the early eighteenth century, yet the internal subgrouping of the family remained largely unknown until recently. Whether or not it is possible to identify intermediate nodes is a function of the rate of expansion and divergence of a population and has nothing to do with the validity of the overall family. If a population spreads rapidly through unoccupied territory - as the Amerind population seems to have done in North and South America roughly 11,000 years ago - intermediate groupings will be more difficult to detect than the overall family precisely because there was not sufficient time for the defining innovations to develop in the intermediate groupings. But the presence of the n/m pronoun pattern and the ubiquitous tina/tana/tuna example (and hundreds of other elements) throughout Amerind languages, but not elsewhere, establishes the validity of Amerind, whether or not intermediate subgroupings can be worked out. Even for Indo-European the family as a whole is quite obvious, whereas the internal subgrouping - how the dozen or so branches actually split up - is poorly understood. Except for Indo-Iranian and Balto-Slavic (and the fundamental divergence of Anatolian) little is known of this internal structure.

6. Monogenesis of extant languages The final taxonomic controversy I will discuss is the question of the monogenesis of all extant languages. This topic is generally considered the most controversial of all, and most historical linguists regard even the possibility of ever proving monogenesis as intrinsically beyond the methods of comparative linguistics. There have, however, long been linguists who argued for monogenesis, and of course Trombetti is the best known. In fact, Trombetti's name is usually associated exclusively with the theory of monogenesis and, because of the stigma that has been attached to this idea, Trombetti's contributions to lower levels of taxonomy are generally unknown. Sapir, how-

212 Merritt Ruhlen

ever, had a different opinion of Trombetti's work, as revealed in a letter to Kroeber in 1924: "There is much excellent material and good sense in Trombetti in spite of his being a frenzied monogenist. I am not so sure that his standpoint is less sound than the usual 'conservative' one" (Golla 1984: 420). The question of monogenesis is an empirical one. Do ancient language families such as Niger-Kordofanian, Eurasiatic, Australian, and Amerind themselves share certain basic roots which would indicate a common origin? Figure 3 shows the world's major language families according to Greenberg. What would a comparison of these families show?

Minn tic Ocean Pacific Ocean Indian Ocean '

Khoisan Niger-Kordofanian Nilo-Saharan

Afro-Asiatic

?

Dravidian

III

Austric

*

Kartvelian

Η

Indo-Pacitic

i. ,

Eurasiatic

L;j

Australian

Dene-Caucasian

0

Amerind

5

Figure 3. Language families of the world. In the early 1990s John Bengtson and I (Bengtson and Ruhlen 1994) compared the roots that had been identified by specialists in 33 different families that included all the world's languages. We found that there were in fact a sizeable number of widespread roots and we argued, like Trombetti before us, that these similar roots could only be explained by common origin. One of the most widespread is pal 'two', examples of which are given in Table 4.

Taxonomic controversies Table 4. pal

in the twentieth

century

213

'two'.

Language-Family

Language

Form

Meaning

Niger-Congo

Nimbari

bala

'2'

Nilo-Saharan

Kunama

ba:re

'2'

Afro-Asiatic

Proto-Central Chadic

*-bwVr

'2'

Indo-European

Proto-Indo-European

*pol

'half'

Uralic

Proto-Uralic

*pälä

'half'

Dravidian

Proto-Dravidian

*pa:l

'part'

Austroasiatic

Proto-Austroasiatic

*?(m)bar

'2'

Indo-Pacific

Kede

-po:l

'2'

Australian

Proto-Australian

*pula

'2'

Miao-Yao

Proto-Miao-Yao

*(a)war

'2'

Austronesian

Proto-Austronesian

*ka(m)-bal

'twin'

Amerind

Wintun

pa:le-t

'2'

How can such similarities be best explained? We have, of course, the same four possible explanations. Onomatopoeia can immediately be eliminated because there is no intrinsic connection between the sounds pal and the meaning 'two'. Borrowing can also be eliminated because intercontinental borrowing - from Africa to Australia to the Americas - could not have taken place until only a few centuries ago. We are left with a choice between common origin and accidental resemblance. But is it plausible that so many large families would have independently chosen the sounds pal to represent the number 'two'. Of Greenberg's 12 large families,pal occurs in all but Khoisan, Kartvelian, and Dene-Caucasian, and in the Amerind family pal is found in 11 of the 13 Amerind branches. Even for a language with just seven consonants and three vowels, there are 147 possible consonantvowel-consonant (CVC) sequences, and pal is just one of them. For a language with 14 consonants and five vowels the possible CVC sequences increase to around 1,000. Clearly it is implausible that so many supposedly unrelated families should have independently chosen the same sequence of sounds to represent 'two'. The only reasonable explanation for these data is

214

Merritt

Ruhlen

common origin. As Sapir said: "It's disturbing, I know, but (more) noncommittal conservatism is only dodging, after all, isn't it? Great simplifications are in store for us" (quoted in Ruhlen 1991: 222).

7. Conclusion I would like to conclude by placing the linguistic evidence I have discussed here in the broader context of human evolution. Both archaeology (Klein 1999) and human genetics (Cavalli-Sforza, Menozzi, and Piazza 1994) indicate that all modern humans share a recent common origin in Africa. Anatomically modern people - who are identical with modern humans morphologically - first appear, in Africa, around 100,000 years ago. However, these people did not behave like modern humans; they behaved like Neanderthals, with both groups using essentially the same tool kit. Roughly 50,000 years ago there was a major (probably the major) transition in human evolution, as anatomically modern humans started, quite suddenly, to exhibit modern human behaviour. Artifacts that had changed little in hundreds of thousands of years, over vast geographical areas, suddenly became much more complex and began to change rapidly in both space and time. For the first time tools were made not just from stone, but also from bone, antler, tusks, shells, and other materials. It is at this time that the first clear indications of art appear, in both Australia and Europe. It was apparently also at this moment that these behaviourally modern people spread out of Africa, carrying with them the genes that attest to their recent African origin, and the language that has left traces, such as pal 'two', in contemporary languages down to the present day. Eventually these people spread throughout the world, replacing earlier hominids such as the Neanderthals and occupying for the first time Australia, Oceania, and the Americas. A number of scholars have maintained that this sudden and profound change in human behaviour could hardly have been accomplished without modern human language (Klein 1999). Indeed, the emergence of fully modern human language at this time is often seen as the underlying mechanism behind the swift change to behavioural modernity and the subsequent occupation of the entire world. If this is so, then the emergence of human language at this time is not only consistent with the Out-of-Africa hypothesis, it may help to explain it.

The origin of origins: a play in five acts, with a prologue im Himmel and an epilogue auf der Erde Henri Meschonnic Prologue The very origin of the question of origin is lost. It was already a problem for the Ancients, as evidenced by Herodotus's famous anecdote about King Psammetichus of Egypt, who isolated two children to find out what language they would naturally speak. Following Hebrew, Sanskrit (when it was discovered) played the happy role of Origin, that is, of the primeval unity of all European languages except Finnish and Basque. Which underscores the link between the question of origin and the question of unity. The origin question becomes a conflict between the unique and the manifold, where there remains at least the unity in the essence itself of language, which shifts, in turn, into the notion of a general or universal grammar. Theo-linguistics.

Act I. Unity versus Diversity, Nature versus History The play pits biology against history. The fascination with the question of language origin can be inverted into the question of the history and meaning of this fascination, the history of a displacement that enacts the very functioning of language, the vicious circle between nature and history. Just as thinking about the origin of poetry amounts to thinking about poetry. About the origin of Man. And so on. The origin is a displacement and a superlative: the sublime state of the question of Being. The problem with biology is that we are entirely biology. But we are also entirely history, entirely subjects, entirely art (and sense of art), entirely ethics, entirely politics. And so far there have been no concepts for thinking about the relationship between biology and history and the rest. Certainly cognitivism does not provide such concepts. For it tends to biologise and Darwinise the problem - thereby creating the illusion of solving it. Cognitivism is to the late twentieth and early twenty-first centuries what organicism was to the scientistic nineteenth century. If one day the gene for language is discovered, it will not change a thing about the problem of what

216

Henri Meschonnic

language is, how it works, what it implies what it does, and what it points to within the limits of various linguistic doctrines. The fascination with the question is continually renewed - despite the Paris Society of Linguistics' 1866 ban (see Trabant 1996). It even shows a return of what has been forced back into the unconscious - therefore the pleasure - and is thus interesting for an analysis of knowledge for two reasons. First, as evidence that neither nineteenth-century historicism (which relegated onomatopoeia to the "poultry-yard of language", according to the nineteenth-century Grand Dictionnaire Universel of Pierre Larousse) nor Saussure's theoretical linguistics was able to repress the question. But second, the fascination also demonstrates both the question's archaism and its novelty. Tzara said: "La pensee se fait dans la bouche" (see Tzara 1975). Are you looking for its origin? You have it in your mouth. Spit it out. What is at question in the origin question is the very idea of a theory of language. Seventeenth- and eighteenth-century enquiries into language's past or its varieties around the world were more a search for unity and identity than a study of diversity. In the early seventeenth century Claude Duret maintains that in the beginning Hebrew was the world's first and only language. Duret quotes Book 19 of St. Augustine's City of God: "linguarum diversitas hominem alienat ab homine... ita ut libentius homo sit cum cane suo quam cum homine alieno [the diversity of languages makes man a stranger for man... to the point that man has more joy being with his dog than with a stranger] (Duret 1972 [1613]: 8, my translation). For Court de Gebelin in his Monde Primitif things were simple: "Cette Origine est Divine". He rejects any gradual elaboration: "La Parole naquit avec 1'Homme" (Court de Gebelin 1775: 3: 66). "La Parole n'etant qu'une peinture, eile ne sauroit dependre de la convention" (Court de Gebelin 1775: 3: 69-70), and Hebrew was not the primordial language. There are three kinds of life in man, a vegetative one, an animal one, and the life of intelligence - "trois Ames" (Court de Gebelin 1775: 3: 97-98). For him, speech has its origin in intellectual life, which he would not separate from the "energie du geste" (Court de Gebelin 1775: 3: 103). The primordial tongue was language itself. By doing so he of course only suppressed the question, only to fall back immediately into another fiction: that of primitive Man. Monosyllables were the "berceau de la parole" (Court de Gebelin 1775: 3: 270). His analysis of the origin of speech rested on two bases: the organs of speech that are the same today as at the origin, and the idea that words had been formed analogously by choosing sounds similar to the objects and ideas they repre-

The origin of origins

217

sent. Which results in a generalised mimetics. Before laughing at it, one should remember that it is precisely the same notion that still informs the work of Roman Jakobson and Ivan Fonagy, whose theories are perhaps not so distant from that of President de Brasses, who uses the same examples: papa and mama (quoted in Court de Gebelin 1775: 3: 336). The origin? Nature!

Act II. Identity versus Purity There is a desire for origin. The point here is to show that it is necessarily linked to a quest for identity and purity. The search for the unknown of lost earlier times, for Paradise Lost, takes the form of a quest for identity as a multiple knot: it is a semiotic knot, since it is a question about the modes of making sense. Mallarme wrote: "toute äme est un noeud rythmique". It is a semantic knot, a knot of intelligibility. It is an ethical knot, for what is at stake is the question of the plurality of subjects. It is a political knot because it also includes the whole history of human cultures, conquests, and exterminations. The question of origin is thus the knot between the questions of identity, unity, and purity. From the origin, naturally. Strangers are naturally excluded. Nothing is more sensitive to dirt than purity. Naturally the desire is also a nostalgia. In the nineteenth century Arthur de Gobineau demonstrates how it is the question of lost purity and a sense of decline. Bopp and Schleicher share this idea (see Gobineau 1867: 9). In his 1867 Memoire sur diverses manifestations de la vie individuelle Gobineau writes: "Tant que les langages primitifs et typiques resterent isoles, l'organisation phonetique de chacun d'eux se conserva dans un etat d'originalite absolue. Aujourd'hui qu'il n'existe plus que des langues metissees, parce qu'il n'y a plus que des races melees, le desordre s'est mis dans tout le domaine... Ton apergoit les marques les plus convaincantes des mesalliances et de la corruption" (Gobineau 1867: 122). Even in Sanskrit and Greek, "la purete n'y est que relative" (Gobineau 1867: 122). "L'anglais, le frangais, l'allemand, l'arabe vulgaire montrent les traces evidentes de la decadence" (Gobineau 1867: 102). And: "Entre les langues pures, puis ä demi-pures, puis tout ä fait metissees qu'on observe actuellement, il y a bien des transitions d'ou l'affaiblissement a graduellement resulte" (Gobineau 1867: 136). If the origin of origin is a myth of unity and purity, everything in language that tends to recreate unity reinforces the illusion of originism.

218

Henri

Meschonnic

So, to my mind, does communicationism: the reduction of language to communication and of words to tools. This sort of toolism is quite common. And the monolinguistic communicationism of English, of which we are now the products, reenacts the fable of unity-identity. It goes on denying plurality-diversity-historicity. It thus still performs the theatralisation of the fiction of origin, the myth of Babel. It simultaneously abolishes the dispersion of Babel and repeatedly re-enacts Babel. The Gate of God (which is what bab-el means) as the Gate of Language. The origin question can therefore be compared with religion. It unites or is supposed to unite - but in fact separates. Lactantius defined the word religio by means of the verb religare ("to link"): that which links men to God and, via this link, links men together. It becomes evident that the question of origin, of the origin of language, is essentially a religious question, perhaps the quintessential religious question. It exterminates alterity and historicity. It is then at the very opposite of the question (and search) for historicity, which is the question of Saussure, when he says that whenever one looks for origin one finds the way language works: "Le probleme de l'origine du langage n'est pas un autre probleme que celui de ses transformations" (Godel 1969: 38). And: "II n'y a aucun moment ού la genese differe caracteristiquement de la vie du langage, et l'essentiel est d'avoir compris la vie. Inanite de la question" (Godel 1969: 38). Or: "l'origine de la langue est sans importance en regard de ce qui se passe continuellement" (Godel 1969: 67). By inverting origin into function, one may also show that the classical relationship between mother-language and the invention of thought gets reversed: natural languages are no longer mother-languages, it is the works of thought that are mother-works. It is not Hebrew that made the Bible, it is the Bible that made Hebrew, that is, that made it what it has become. And so on. Just as every poetics of thought transformed the natural language in which it intervened. The inversion of origin is crucial for thinking about the radical historicity of language, of languages, of speech, of works, of subjects, of societies, and of values, that is, in order to think atheologically, to think the atheologico-political, to think the specificity of ethics vis-ä-vis the religious. For before being the question of the religious (in the meaning of Lactantius), the question of origin is the question of the sacred and in a sense specific to language theory: the primitive union of words and things, of men and nature, the age of fables and tales (when beasts could talk), the

The origin of origins

219

mythical continuity that is the foundation for totemism and for analogous thought (on talking beasts see Rilke 1987). Which shows up again and again in any unitarian epistemology of social sciences and natural sciences: organicism in the nineteenth and so-called "stupid" century, and today in cognitivism, the new organicism.

Act III. Becoming versus Origin In his 1859 preface to Jacob Grimm's essay "On the Origin of Language", Ernest Renan writes that Grimm "est arrive par la philologie a la mythologie. Les fables et les mots ont ete pour lui inseparables, et il a cherche leur commune origine dans Γ esprit meme de la race qui les a crees, dans sa maniere d'imaginer et de sentir, dans ses instincts les plus antiques et les plus profonds" (Renan 1859: 1-2). He associated the "vivacite de ses [Grimm's] intuitions et le merveilleux sentiment qu'il a des choses populates, ä l'entente des questions d'origines" with his insight "dans l'intelligence du monde primitif" (Renan 1859: 2). But he opposed "le probleme essentiel de la philologie, qui est tout historique" to "l'attention du public vers les questions d'origines" (Renan 1859: 2). On the historical side, the most positive and surest research; on the side of origin, the general questions that touched "une solution vraiment philosophique des problemes de l'histoire de l'humanite" (Renan 1859: 3). Grimm starts his essay by recalling that language origin was the subject of a Berlin Academy prize question in 1770. Schelling proposed the question again and subsequently withdrew it. Grimm refers to Schelling's disappointment at Herder's winning essay from 1770. Grimm opposes the question of language origin to the progress in language studies generally, which had become a "science", by which he means comparativism: "the relationship of the languages to one another", the progress in "their simple morphology" (Grimm 1984: 1-2). Right from the start he compares linguistics with "natural history" (Grimm 1984: 2). And praises "the established dominion of the British in all parts of the world chiefly in India" (Grimm 1984: 3), since it had allowed the discovery of Sanskrit. For Grimm, Sanskrit is a "magnetic instrument" on the "linguistic ocean" (Grimm 1984: 3). It is the source of the interest in the "little suspected law of our own German language". Its study helps explain the "common movement and course of human language" and generates "perhaps to the most productive conclusions about its origin" (Grimm 1984: 3).

2 2 0 Henri

Meschonnic

But the objection comes immediately: "Does not the whole question fall in the realm of the impossible?" (Grimm 1984: 3). Language study, it seems, suffers by comparison to naturalists' study of the "secrets of natural life" (Grimm 1984: 3). This makes the difference "between creation and generation" a "power prevailing outside of the created being" (Grimm 1984: 4). The comparison seems to end with the double hypothesis of whether language was "created or uncreated" (Grimm 1984: 4). If it was created, "its first origin remains to our glances just as impenetrable as that of the first created animal or tree" (Grimm 1984: 4). A divine origin. If it was uncreated - formed, that is, "by the freedom of man himself" - "then it can be measured according to that law which indeed its history up to the oldest stem yields us. It may be stalked back over that empty abyss of millennia and in thought even cornered on the beach of its origin" (Grimm 1984: 4). Grimm attempts first to show that language could neither be "created" nor "revealed". In favour of divine creation, he evokes the diversity of "the species of languages": "do not the species of languages resemble the species of the plants, of animals, indeed of men themselves in almost endless varieties of their changing form?" (Grimm 1984: 4). But he opposes animals to man, animal cries to human language, and instinct to what keeps changing and "must be constantly learned" (Grimm 1984: 6). Surroundings and epoch. After arguing against the hypothesis of innateness as divine creation, Grimm gainsays the "idea of a revealed language", which presupposes "a state of paradisiacal innocence" followed by a "fall" (Grimm 1984: 9). Instead, Grimm invokes "our Holy Scripture": "only a long time after the fall of man does the confounding of tongues occur" (Grimm 1984: 9). And he rejects the Biblical narrative and its interpretation as "confusing"; he views the "endless multiplicity" of languages "as salutary and necessary" (Grimm 1984: 9). At this point, his inquiry addresses "a theological position" (Grimm 1984: 9). He asserts that the "miracle of the continuation of the world is fully equal to that of its creation" and that "the nature of man at the time of creation was not different from what it is today" (Grimm 1984: 9-11). He observes that, unlike the Bible's story of Babel, "neither Greek nor Indian antiquity has tried to ask and to answer the question about the origin and diversity of human tongues" (Grimm 1984: 12). The link between language origin and language diversity is clear. Grimm jettisons theology: "An innate language would have made men animals; a revealed one would have presumed gods of them" (Grimm 1984: 12). There remains only a single hypothesis: "it must be a human language acquired by

The origin of origins

221

us with complete freedom regarding origin and progress. It can be nothing else; it is our history, our inheritance" (Grimm 1984: 12). Grimm establishes an etymological link - via Sanskrit - between Mensch (human being) and the Indian King Manas, whose root is "man i.e. to think" (Grimm 1984: 12). From this Grimm deduces: "Man [Mensch] is not only called thus because he thinks, but is also man because he thinks, and he speaks because he thinks. This very close relationship between his ability to think and to speak, designates and guarantees to us the reason and origin of his language" (Grimm 1984: 12). He recalls that logos "means speech and reason" and adds that "men with the deepest thoughts - philosophers, poets, speakers - have also the greatest command of language"(Grimm 1984: 12). This has two consequences. First: "The power of language forms nations and holds them together" (Grimm 1984: 12). Second (and more significantly): "Without such a bond they would be scattered. The riches of thought in each nation are chiefly what solidifies its world empire" (Grimm 1984: 13). The power of thought. Not economic and political power. Contrary to appearances. What is striking, from my point of view, is that Grimm sees language as "a work" and diversity as a "precious acquisition" (Grimm 1984: 13). This rejects the notion of the inequality of languages: "This speaking, this thinking does not stand there separately for individual men, but, all languages are one common property come into history, and bind the world together" (Grimm 1984: 13). Grimm's sense of the radical historicity of language and languages even impels him to say that "the inventions of writing and printing ... confirm and complete the proof of its human origin" (Grimm 1984: 13). Here, he confounds two types of origin: an anthropological origin and a cultural origin. But when Grimm declares that he intends to turn to the main part of the question, the first notion he mentions - even if it is in order to brush it aside - is that of languages' "first formation, or to several formations only" (Grimm 1984: 14). At the origin of the question of origin, there are two conditions. The first is the opposition between unity and plurality. The second is the relationship between "great antiquity" and the anteriority that remains out of reach because it did not leave behind any traces (Grimm 1984: 15). Nineteenth-century historical linguistics was thus genealogical. It was a story of decline, which presupposes "a prior high point of greater form perfection" (Grimm 1984: 15). This high point consisted of case endings modelled on Sanskrit, a phase preceded by an invisible one, "a combination

2 2 2 Henri

Meschonnic

of analogous parts of words" (Grimm 1984: 15). This was based on the natural model of "leaf, blossom and ripening fruit" (Grimm 1984: 16). Grimm notes here that Chinese, which lacks case endings, remains stuck at that first period of formation. The ages of man provide the underlying model for Grimm's observations. Grimm turns his attention to historical phonetics, but not without granting symbolic values to the sounds of language: "Of the consonants / will designate the soft, r the rough" (Grimm 1984: 17). He combines this with an ideology of gender: "Obviously a feminine base must be ascribed to the vowels altogether; to the consonants a masculine one" (Grimm 1984: 17). He then inverts the idea of decadence into continual progress: "The conclusion is that human language, although considered only apparently and in an individual way in retrogression, has to be regarded when understood from a total viewpoint, as always in a state of progress and growth from its inner power" (Grimm 1984: 20). He adds immediately: "Our language is also our history" (Grimm 1984: 20). Childhood consists of "short, monosyllabic" words (Grimm 1984: 20). In the second period, "words have become longer and polysyllabic. Now masses of compounds are formed from the loose order... At this time we see language most highly suited to meter and poetry. For these beauty, harmony and exchange of form are essential, indispensable. The Indian and Greek poetry designate for us a peak reached at the right moment in immortal works later unattainable" (Grimm 1984: 21). But still, for Grimm, an "eternal, irresistible ascent" depends on "a still greater freedom of thought" (Grimm 1984: 21). While discussing the formation of modern European languages, Grimm notes that English, through its "plenitude of free mediants", has a power of expression "as it perhaps never yet was at the command of a human tongue", which has "resulted from a surprising marriage of the two noblest languages of later Europe, the Germanic and Roman" (Grimm 1984: 22). There is, it seems, an aesthetics of origin - a linguistic aesthetics - that passed from grammatical beauty (case endings) to a generalised aesthetics of language. Grimm attributes to English a universality which is incorrect on two counts. First, his lionisation of Shakespeare implies that English produced Shakespeare and ignores the problem of the relationship between a language and a literary work. Second, his definition of English as the union of two other languages presupposes an imaginary genetics of language properties: "Yes, the English language, may by all rights be called an universal language. Not in vain has it produced and brought forth the greatest and most superior poet of modern times in contrast to ancient classical poetry. Here I

The origin of origins

223

can only mean Shakespeare. Like the people themselves the English language seems chosen in the future to hold sway in a still higher degree at all ends of the earth. For in riches, reason and succinct combinations none of the presently living languages can stand beside it" (Grimm 1984: 22). Grimm considers the German language inferior. It is "torn to pieces as we Germans are fragmented. We would first have to shake off some of its imperfections before it could run boldly along in the race" (Grimm 1984: 22). But he harbours "hope", for "the beauty of human language blossoms not in the beginning but in its midpoint. It will only present its richest fruit one day in the future" (Grimm 1984: 22). It's all a bit confused and contradictory. A conflict, for which there is no concept, between the "warm hand of human freedom", the "incessant extension and folding of the wings, unceasing exchange", and the genius of language: "The quiet eye of the guardian spirit of language quickly closes up and heals all its wounds overnight. It keeps in order and preserves from confusion all its affairs, except that it has shown individual languages the highest favour and others less" (Grimm 1984: 22-23). Again the old idea of the inequality of languages, though without the concepts needed to know what Grimm means when he says: "One language is more beautiful and seems more productive than the other" (Grimm 1984: 24). For the genius of language and of languages not only is a non-concept, but simultaneously offers the illusion of thinking something and prevents thought - without one even knowing it.

Act IV. The origin every day That the origin is the way language works, which is the conception of Saussure, is shown by Robert Nicolai'. Under the rubric "origin" he examines the "evolution of languages" and "linguistic change" (Nicolai 1998: 157). He concludes: "What is at stake is more the understanding of the dynamics and constitution of languages, through the detailed study of their formation and the analysis of the identification of their dimensions useful for their description which all establish themselves on that anthropological ground where sociological, semiotic and cognitive relevances in the broadest meaning get connected" (Nicolai 1998: 178). A comparatist's work. By studying "creolisation or pidginisation", starting from what Trubetzkoy analysed as an "alliance of languages", the "phonological Sprachbund of

2 2 4 Henri

Meschonnic

languages" in the Balkans and a "cultural Sprachbund''' (Nicolai 1998: 167), Nicolai sets out to both "resume an old debate" and give it back its relevance (Nicolai 1998: 177). And by reappraising "the questioning on the origin of languages", in order to make it again a "full part of the general problem of the sciences of language from the moment when the theoretical tools apt to contribute bringing responses respond to relevances specific of those sciences themselves" (Nicolai 1998: 177-178), the question of origin is transformed. It is "no more ideological, nor philosophical" but, as he says himself, "considering to open up again the old questioning on the origin of language does not come back asking the same question!" (Nicolai 1998: 178).

Act V. The problem is to displace the problem The question of the origin of natural languages has a goal: to shed light on the question of the origin of language. And, for its part, this double question also has a goal: "to investigate human prehistory", as Merrit Ruhlen puts it (Ruhlen 1994d: 161). Ruhlen resumes the idea of a single origin for languages. He does this against the prevailing opinion of specialists, which he describes as a "hoax" (Ruhlen 1994d: 66). Though his critique of Indo-Europeanists' self-centeredness is probably correct (Ruhlen 1994d: 78-80). The preface to the French edition of Ruhlen's The Origin of Language (which I read prior to reading the English original) is written by a biologist who seems to draw from the common genetic origin of "present day humans" the "probable" character of a "mother language" (Ruhlen 1997: 6-7, my translation). And the translator observes that the linguistic use of the term "genetic", as in the locution "genetic classification of languages", is "derived by mere analogy of the meaning in biology", without saying that "the particular languages would be inscribed in the genes of their speakers" (Ruhlen 1997: 8, my translation). The question of origin is double. In any case it very quickly splits into two. In Ruhlen's analysis, it splits into the question of the origin of human language and the question of the origin of languages. Which, in turn, also splits into two: a "monogenesis", a "common source" for "presently extant languages" (Ruhlen 1994d: 3-4) and independent evolutions. Which amounts to the question of the relationship between languages. Which again splits into two: within a family of languages like Romance languages (where

The origin of origins

225

there is no problem) or between families (where there is a problem). And then the question splits again: where there is a written tradition and where there is none. And it is either a genealogy or a focus on resemblances. It is interesting to recall that it's primarily the differences that would have interested Humboldt. So the question will split again depending on the weight attached to resemblances or differences. Resemblances are possibly due to accidental convergence, borrowing, or a common origin. It becomes a question of method. That of the relationship between the notion of (genealogical) classification and the notion of a system. It is noteworthy that the comparisons largely remain on the level of "words" (Ruhlen 1994d: 34), "on the basis of a few words" (Ruhlen 1994d: 62), with a partial exception for "the pronominal pattern" (Ruhlen 1994d: 78) or the "sound symbolic" cases mama, papa, kaka ('older brother') (Ruhlen 1994d: 122-123). Another problem is the relationship between the motion of words and "the dispersal of modern humans" (Ruhlen 1994d: 19). In fact, the analysis shifts to the study of language families and families of families. Starting with William Jones's identification of the Indo-European family in 1786, the fundamental paradox of language typologies is that in order to historicise languages by separating them from mythology and from religion - that is, in order to make studying them into a science, a historical science, and historical phonetics - one locates them within natural history: one biologises them. Ruhlen refers to an analogy with "biological taxonomy" (Ruhlen 1994d: 129). Even if biology has itself had a paradoxically historical result, namely establishing that there are no "primitive peoples on earth" which parallels the linguistic assertion that there are "no primitive languages anywhere" and no "intrinsic connection between language and ethnicity" (Ruhlen 1994d: 148-149). Another epistemological problem is the one of proof. Ruhlen declares that linguistics escapes the notion of proof, "for 'proof' is a mathematical concept not applicable to empirical sciences like linguistics" (Ruhlen 1994d: 136). For him, the whole point is "that the modicum of data presented ... nonetheless strongly suggests that a common origin for all extant languages is the simplest, and most likely, explanation of the linguistic similarities observed" (Ruhlen 1994d: 136). And this likelihood transmogrifies into the ambiguous state of knowing: "One can see that all the world's families are

2 2 6 Henri

Meschonnic

related"(Ruhlen 1994d: 146). A knowledge mixed with hope: "But unfortunately there is as yet no understanding of the order in which these various groups branched off from the original linguistic trunk. Though one may hope that linguistic evidence may one day shed light on this difficult, but crucial, question, for the moment we must rely on evidence from other fields" (Ruhlen 1994d: 146). But there is at least one strong point - apart from the relevance of Greenberg's work on African and American Indian languages. It is Ruhlen's criticism of the general state of linguists, their "compartmentalized view of knowledge" and "collective myopia"(Ruhlen 1994d: 138). It is a criticism that is worth amplifying if one considers Humboldt's Sprachsinn and the necessity of thinking together the meaning of language, the theory of language, the theory of the art of language as well as ethics and politics in their interaction: their Wechselwirkung. And this very interaction can itself be seen as the origin of language as it works.

Epilogue What we did at the Berlin conference was in one sense a mimicry of origin. By all speaking the same language - English - we mimicked a unity that repeats the story of the tower of Babel. We enacted the nostalgia for a nonlanguage which might as well be mentalese. In the same way that French Spinoza specialists read Spinoza in French, in Italian when they are Italian, in English when they are native speakers of English, and so on. By doing so, they demonstrate that they have no problem with language or translation. One or two might consider isolated words in Latin and think that solves the problem. Only one has the courage to utter that because Spinoza wrote in Latin he had "no language at all". That is Yeshayahu Yovel in his Spinoza and other heretics. From this remark I am obliged to draw the conclusion that those lucky philosophers are the last speakers of the pre-Babelian, Adamic language that is the same for everyone and that grants direct access to thought - without having to pass through what one calls language. That is, logically, no language at all. My amazement is increased when I see that they go on writing, believing that they think. They are probably dreaming.

The origin of origins

227

If they're awake, they should remain silent. But they couldn't even do that, for silence is also part of language. I must seem the opposite of serious. On the contrary, I would call this kind of humour a superlative form of the serious, as is demonstrated by comedy which plays with serious matters. So does the comedy of thought. This is what we performed at the Berlin conference by enacting the question of the origin of language.

References Aarsleff, Hans 1982 From Locke to Saussure, 146-209. Minneapolis: University of Minnesota Press. Aiello, Leslie C. 1998 The foundations of human language. In: Nina G. Jablonski and Leslie C. Aiello (eds.), The Origin and Diversification of Language, 21-34. San Francisco: California Academy of Sciences. Aiken, Nancy 1998 The Biological Origins of Art. Westport, C.T.: Praeger. Aitchison, Jean 1996 The Seeds of Speech: Language Origin and Evolution. Cambridge: Cambridge University Press. 1998 On discontinuing the continuity-discontinuity debate. In: James R. Hurford, Michael Studdert-Kennedy, and Chris Knight (eds.), Approaches to the Evolution of Language: Social and Cognitive Bases, 17-29. Cambridge: Cambridge University Press. 2001 Language Change: Progress or Decay. 3rd edition. Cambridge: Cambridge University Press. Aitchison, Jean, Diana M. Lewis, and Bronwyn Naylor 2000 "Car murder hubby caged" and other murderous headlines. English Today 61, 16.1: 23-30. Albert, Μ. Α., R. G. Feldman, and A. L. Willis 1974 The "subcortical dementia" of progressive supranuclear palsy. Journal of Neurology, Neurosurgery, and Psychiatry 37: 121-130. Aldridge, J. W., K. C. Berridge, M. Herman, and L. Zimmer 1993 Neuronal coding of serial order: syntax of grooming in the neostratum. Psychological Science 4: 391-393. Alexander, M. P., M. A. Naeser, and C. L. Palumbo 1987 Correlations of subcortical CT lesion sites and aphasia profiles. Brain 110: 961-991. Alsina, Alex, Joan Bresnan, and Peter Sells (eds.) 1997 Complex Predicates. Stanford: CSLI Publications. Aristotle [1962] The Categories. On Interpretation. Prior Analytics. Edited by Harold P. Cook and Hugh Tredennick. London/Cambridge, M.A.: Heinemann/ Harvard University Press. Armstrong, David F., William C. Stokoe, and Sherman E. Wilcox. 1995 Gesture and the Nature of Language. Cambridge: Cambridge University Press.

230

References

Auroux, Sylvain 1989 La question de l'origine des langues: ordres et raisons du rejet institutionnel. In: Joachim Gessinger and Wolfert von Rahden (eds.), Theorien vom Ursprung der Sprache, 2: 122-150. Berlin/New York: de Gruyter. Avital, Eytan and Eva Jablonka 2001 Animal Traditions: Behavioural Inheritance in Evolution. Cambridge: Cambridge University Press. Bakker, Peter 1987 Autonomous Languages: Signed and Spoken Languages Created by Children in the Light of Bickerton's Bioprogram Hypothesis. (Publicaties van het Instituut voor Algemene Taalwetenschap 53.) Amsterdam: Universiteit van Amsterdam, Instituut voor AlgemeneTaalwetenschap. Baldwin, James Mark 1902 Development and Evolution. New York: Macmillan. Batali, John 2000 The negotiation and acquisition of recursive grammars as a result of competition among exemplars. In: E. J. Briscoe (ed.), Linguistic Evolution through Language Acquisition: Formal and Computational Models. Cambridge: Cambridge University Press. Bauer, Laurie 1998 When is a sequence of two nouns a compound in English? English Language and Linguistics 2: 65-86. Beaken, Mike 1996 The Making of Language. Edinburgh: Edinburgh University Press. Beals, Alan R. 1962 Gopalpur. A South Indian Village. New York: Holt, Rinehart, and Winston. Behera N. and V Nanjundiah 1995 An investigation into the role of genetic plasticity in evolution. Journal of Theoretical Biology 172: 225-234. Bellugi, Ursula, L. Lichtenberger, D. Mills, A. Galaburda, and J. R. Korenberg 1999 Bridging cognition, the brain and molecular genetics: evidence from Williams syndrome. Trends in Neurosciences 22: 197-207. Bengtson, John B. 1991 Notes on Sino-Caucasian. In: Vitaly Shevoroshkin (ed.), Dene-Sino Caucasian Languages, 67-129. Bochum: Brockmeyer. Bengtson, John B. and Merritt Ruhlen 1994 Global etymologies. In: Merritt Ruhlen (ed.), On the Origin of Languages: Studies in Linguistic Taxonomy, 277-336. Stanford: Stanford University Press. Berwick, Robert C. 1998 Language evolution and minimalist program: the origins of syntax. In: James R. Hurford, Michael Studdert-Kennedy, and Chris Knight (eds.), Approaches to the Evolution of Language: Social and Cognitive Bases, 320-340. Cambridge: Cambridge University Press.

References

231

Bickerton, Derek 1981 Roots of Language. Ann Arbor: Karoma. 1984 The language bioprogram hypothesis. Behavioral and Brain Sciences 7: 173-221. 1990 Language and Species. Chicago: University of Chicago Press. 1996 I chat, thereby I groom. Nature 380: 303. 1998 Catastrophic evolution: the case for a single step from protolanguage to full human language. In: James R. Hurford, Michael Studdert-Kennedy, and Chris Knight (eds.), Approaches to the Evolution of Language: Social and Cognitive Bases, 341-358. Cambridge: Cambridge University Press. 1999 How to acquire language without positive evidence: what acquisitionists can learn from Creoles. In: Michel DeGraff (ed.), Language Creation and Language Change, 49-74. Cambridge, M.A.: MIT Press. Bierwisch, Manfred 1994 Kommunizieren und Berechnen. Linguistik zwischen Biologie und Geisteswissenschaft. In: Berlin-Brandenburgische Akademie der Wissenschaften. Jahrbuch 1992/93: 187-215. Birdsell, Joseph B. 1970 Local group composition among Australian Aborigines: a critique from fieldwork conducted since 1930. Current Anthropology 11: 115-142. 1979 Ecological influences on Australian Aboriginal social organization. In: Irwin S. Bernstein and Euclid O. Smith (eds.), Primate Ecology and Human Origins: Ecological Influences and Social Organization, 117-151. New York: Garland STPM Press. Birg, Herwig 1993 Der überfüllte Planet. Lebenserwartung, generatives Verhalten und die Dynamik des Weltbevölkerungswachstums. In: Funkkolleg Der Mensch. Anthropologie Heute. Studieneinheit 27, Studienbrief 9. Deutsches Institut für Fernstudien an der Universität Tübingen. Blumenberg, Hans 1996 Arbeit am Mythos. Frankfurt am Main: Suhrkamp. Blumstein, Sheila Ε. 1995 The neurobiology of language. In: Joanne L. Miller and Peter D. Eimas (eds.), Speech, Language and Communication, 339-370. San Diego: Academic Press. Boas, Franz [1955] Primitive Art. New York: Dover Publications. First published 1927. Boltz, William G. 1996 East Asian writing systems. In: Peter T. Daniels and William Bright (eds.), The World's Writing Systems, 189-190. New York: Oxford University Press. Bomhard, Allan R. and John C. Kerns 1994 The Nostratic Macrofamily. A Study in Distant Linguistic Relationship. Berlin: Mouton de Gruyter.

232

References

Brainard, Μ. S. and A. J. Doupe 2000 Interruption of a basal ganglia-forebrain circuit prevents the plasticity of learned vocalizations. Nature 404: 762-766. Broca, P. 1861 Remarques sur le siege de la faculte de la parole articulee, suivies d'une observation d'aphemie (perte de parole). Bulletin de la Societe d 'Anatomie 36: 330-357. Bühler, Karl 1990 Theory of Language. The Representational Function of Language. Translated by Donald Fraser Goodwin. Amsterdam: Benjamins. Butt, Miriam and Wilhelm Geuder 1998 The Projection of Arguments: Lexical and Compositional Factors. Stanford: CSLI Publications. Byrne, Richard W. 1994 The evolution of intelligence. In: Peter J. B. Slater and Tim R. Halliday (eds.), Behaviour and Evolution. Cambridge: Cambridge University Press. Calvin, William H. and Derek Bickerton 2000 Lingua ex machina: Reconciling Darwin and Chomsky with the Human Brain. Cambridge, M.A.: MIT Press. Camartin, Iso 1992 Nichts als Worte? Ein Plädoyer für Kleinsprachen. Frankfurt am Main: Suhrkamp. Campbell, Lyle 1994 Putting pronouns in proper perspective in proposals of remote relationships among Native American languages. In: Margaret Langdon (ed.), Survey of California and other Indian Languages, Report 8,1-20. Berkeley: Survey of California and Other Indian Languages. 1997 American Indian Languages: The Historical Linguistics of Native America. New York: Oxford University Press. Candler, Douglas Keith 1993 Feral Children and Clever Animals: Reflections on Human Nature. New York: Oxford University Press. Cann, Ronnie 1993 Formal Semantics. Cambridge: Cambridge University Press. Cavalli-Sforza, Luca, Paolo Menozzi, and Alberto Piazza 1994 The History and Geography of Human Genes. Princeton: Princeton University Press. Cavalli-Sforza, Luca, and Francesco Cavalli-Sforza 1994 Verschieden und doch gleich. Ein Genetiker entzieht dem Rassismus die Grundlage. München: Droemer Knaur. Changeux, Jean-Pierre 1983 L'Homme Neronal. Paris: Fayard. Chomsky, Noam 1986 Knowledge of Language: Its nature, Origin and Use. New York: Praeger.

References

233

Clark, Eve V. 1993 The Lexicon in Acquisition. Cambridge: Cambridge University Press. Condillac, Etienne Bonnot de [ 1973] Essai sur I 'origine des connaissances humaines. Edited by Charles Porset. Auvers-sur Oise: Galiliee. First published in 1746. 1987 Philosophical Writings. 2 Volumes. Hillsdale/London: Erlbaum. Cooke, Brett and Frederick Turner (eds.) 1999 Biopoetics: Evolutionary Explorations in the Arts. Lexington: ICUS. Coppens, Yves 1994 East Side story: the origin of humankind. Scientific American 270.5: 62-79. Court de Gebelin, Antoine 1777 Monde primitif analyse et compare avec le monde moderne considέrέ dans son genie allegorique et dans les allegories auxquelles conduisit ce genie. Volume 3. Paris. Crick, Francis Henry Compton 1998 What Mad Pursuit: A Personal View of Scientific Discovery. New York: Basic Books. Crow, T. J. 2000 Schizophrenia as the price that Homo sapiens pays for language: a resolution of the central paradox in the origin of the species. Brain Research Reviews 31: 118-129. Cummings, J. L. 1993 Frontal-subcortical circuits and human behavior. Archives of Neurology 50: 873-880. Cunnington, R., R. Iansek, J. L. Bradshaw, and J. G. Philips 1995 Movement-related potentials in Parkinson's Disease: presence and predictability of temporal and spatial cues. Brain 118: 935-950. Curtiss, Susan 1977 Genie: A Psycholinguistic Study of a Modern-Day "Wild Child". New York: Academic Press. Dalalakis, J. E. 1999 Morphological representation in specific language impairment: evidence from Greek word formation. Folia Phoniatrica et Logopaedica 51: 20-35. Damasio, Hanna 1991 Neuroanatomical correlates of the aphasia. In: Martha Taylor Sarno (ed.), Acquired Aphasia. 2nd edition. New York: Academic Press. Danesi, Marcel 1993 Vico, Metaphor, and the Origin of Language. Bloomington/Indianapolis: Indiana University Press. Daniels, Peter T. 1996 The invention of writing. In: Peter T. Daniels and William Bright (eds.), The World's Writing Systems, 579-586. New York: Oxford University Press.

234

References

Dante Alighieri [1988] De vulgari eloquentia. Edited by Sergio Cecchin. Milan: TEA. Darwin, Charles 1872 The Expression of Emotions in Man and Animals. London: Murray. Deacon, Terry 1997 The Symbolic Species: The Co-evolution of Language and the Brain. New York: W.W. Norton. Delbrück, Bertold 1880 Einleitung in das Sprachstudium. Leipzig: Breitkopf und Härtel. Desmond, Adrian and James Moore 1991 Darwin. London: Michael Joseph. D'Esposito, M. and Μ. P. Alexander 1995 Subcortical aphasia: distinct profiles following left putaminal hemorrhage. Neurology 45: 38-41. Dietzsch, Steffen 1998 Kleine Kulturgeschichte der Lüge. Leipzig: Reclam. Dissanayake, Ellen 1995 Homo Aestheticus. Where Art Comes From and Why. Seattle/London: University of Washington Press. Dixon, Robert M. W. 1997 The Rise and Fall of Languages. Cambridge: Cambridge University Press. Dolgopolsky, Aaron 1984 On personal pronouns in the Nostratic languages. In: Otto Gschwantler, Käroly Redei, and Hermann Reichert (eds.), Linguistica et Philologia, 65-112. Vienna: Wilhelm Braumüller. Donald, Merlin 1991 Origins of the Modern Mind: Three Stages in the Evolution of Culture and Cognition. Cambridge, M.A.: Harvard University Press. Dor, Daniel 1996 Representations, attitudes and factivity evaluations: an epistemically based analysis of lexical selection. Ph.D. Dissertation, Stanford University. 1999 Towards a semantic account of that-deletion in English. Unpublished manuscript. 2000a Towards a semantic account of strong island constraints. Unpublished manuscript. 2000b From the autonomy of syntax to the autonomy of linguistic semantics: some notes on the correspondence between the transparency problem and the relationship problem. Pragmatics and Cognition (forthcoming). Downing, Pamela 1977 On the creation and use of English compound-nouns. Language 53: 810-841. Dowty, David R. 1979 Word Meaning and Montague Grammar. Dordrecht: Reidel. 1991 Thematic proto-roles and argument selection. Language 67: 547-619.

References

235

Dunbar, Robin I. M. 1996 Grooming, Gossip and the Evolution of Language. London/Boston: Faber and Faber. 1998 Theory of mind and the evolution of language. In: James R. Hurford, Michael Studdert-Kennedy, and Chris Knight (eds.), Approaches to the Evolution of Language: Social and Cognitive Bases, 92-110. Cambridge: Cambridge University Press. Duranti, Alessandro 1997 Linguistic Anthropology. Cambridge: Cambridge University Press. Duret, Claude [1972] Thresor de l'histoire des langues de cest univers. Geneva: Slatkine Reprints. First published in 1613. Edmunds, Malcolm 1974 Defense in Animals: A Survey of Anti-predator Defences. Burnt Mills: Longman. Eibl, Karl 1995 Die Entstehung der Poesie. Frankfurt am Main/Leipzig: Insel. Eibl-Eibesfeldt, Irenäus 1988 The biological foundations of aesthetics. In: Ingo Rentschier, Barbara Herzberger, and David Epstein (eds.), Beauty and the Brain. Biological Aspects of Aesthetics, 29-68. Basel: Birkhäuser. 1997 Die Biologie des menschlichen Verhaltens. Grundriß der Humanethologie. 3rd ed. Weyarn: Seehamer. Elman Jeffrey, E. A. Bates, Μ. Η. Johnson, A. Karmiloff-Smith, D. Parisi, and K. Plunkett 1996 Rethinking Innateness: A Connectionist Perspective on Development. Cambridge, M.A.: MIT Press.. Embick, D., A. Marantz, Y. Miyashita, W. O'Neil, and K. L. Sakai 2000 A syntactic specialization for Broca's area. Proceedings of the National Academy of Sciences of the USA 97: 6150-6154. Ewer, R. F. 1956 Imprinting in animal behaviour. Nature 177: 227-228 Feld, Steven 1982 Sound and Sentiment. Birds, Weeping, Poetics, and Song in Kaluli Expression. Philadelphia: University of Pennsylvania Press. Fillmore, Charles J., Daniel Kempler, and William S.-Y. Wang (eds.) 1979 Individual Differences in Language Ability and Language Behavior. New York: Academic Press. Fillmore, Lily Wong 1979 Individual differences in second language learning. In: Charles J. Fillmore, Daniel Kempler, and William S.-Y. Wang (eds.), Individual Differences in Language Ability and Language Behavior. New York: Academic Press. Finnegan, Ruth and Margaret Orbell (eds.) 1995 South Pacific Oral Traditions. Bloomington/Indianapolis: Indiana University Press.

236

References

Fitch, W. Τ. Ill 1997 Vocal tract length and formant frequency dispersion correlate with body size in macaque monkeys. Journal of the Acoustical Society of America 102: 1213-1222. Flint, J. 1999 The genetic basis of cognition. Brain 122: 2015-2031. Fodor, Jerry A. 1983 Modularity of Mind: An Essay on Faculty Psychology. Cambridge, M.A.: MIT Press. Foley, William A. 1986 The Papuan Languages of New Guinea. Cambridge: Cambridge University Press. Foley, William A. 1997 Anthropological Linguistics. An Introduction. Maiden, M.A./Oxford: Blackwell. Forge, Anthony 1972 Normative factors in the settlement size of Neolithic cultivators. In: P. J. Ucko, R. Tringham, and G. W. Dimbleby (eds.), Man, Settlement, and Urbanism, 363-376. London: Duckworth. Frawley, William 1992 Linguistic Semantics. Hillsdale, N.J.: Lawrence Erlbaum Associates. Freedle, Roy and Michael Lewis 1977 Prelinguistic conversations. In: Michael Lewis and L. A. Rosenblum (eds.), Interaction, Conversation, and the Development of Language, 157-185. New York: Academic Press. Gajdusek, D. Carleton, Guy M. McKhann, and Liana C. Bolis (eds.) 1994 Evolution and Neurology of Language. Amsterdam: Elsevier. Gardner, R. Allen and Beatrix T. Gardner 1984 A vocabulary test for chimpanzees (Pan troglodytes). Journal of Comparative Psychology A: 381-404. Gardner, R. Allen, Beatrix T. Gardner, and Thomas E. Van Cantfort 1989 Teaching Sign Language to Chimpanzees. Albany: State University of New York Press. Gathercole, Susan E. and Alan D. Baddeley 1993 Working Memory and Language. Hillside, N.J.: Lawrence Erlbaum Associates. Gehlen, Arnold 1971 Der Mensch. Seine Natur und seine Stellung in der Welt. Frankfurt am Main: Athenäum. Geschwind, Norman 1970 The organization of language and the brain. Science 170: 940-944. Gessinger, Joachim and Wolfert von Rahden (eds.) 1989 Theorien vom Ursprung der Sprache. 2 volumes. Berlin/New York: de Gruyter.

References

237

Ghazanfar, A. A. and M. D. Hauser 1999 The neuroethology of primate vocal communication: substrates for the evolution of speech. Trends in Cognitive Science 3: 377-383. Givon, Talmy 1979 On Understanding Grammar. New York: Academic Press. 1995 Functionalism and Grammar. Amsterdam/Philadelphia: John Benjamins. Gleason, Jean Berko 1973 Fathers and other strangers: men's speech to young children. In: Daniel P. Dato (ed.), Developmental Psycholinguistics. Theory and Applications, 289-297. Washington, D.C.: Georgetown University Press. Gleason, Jean Berko and Sandra Weintraub 1978 Input language and the acquisition of communicative competence. In: Keith Nelson (ed.), Children's language. Volume 1, 171-222. New York: Gardiner Press. Gleason, Jean Berko, Rivka Y. Perlmamm, and Esther B. Greif 1984 What's the magic word: learning language through politeness routines. Discourse Processes 7: 493-502. Gleitman, Harry and Lila Gleitman 1979 Language use and language judgement. In: Fillmore, Charles J., Daniel Kempler, and William S.-Y. Wang (eds.), Individual Differences in Language Ability and Language Behavior. New York: Academic Press. Gobineau, Arthur Comte de 1935 Memoire sur diverses manifestations de la vie individuelle. Edited by A. B. Duff. Paris: Desclee de Brouwer. Godel, Robert 1969 Les sources manuscrites du Cours de linguistique general. Geneva: Droz. Gold, J. M. and D. R. Weinberger 1995 Cognitive deficits and the neurobiology of schizophrenia. Current Opinion in Neurobiology 5: 225-230. Goldberg, Adele Ε. 1995 Constructions: A Construction Grammar Approach to Argument Structure. Chicago: University of Chicago Press. Goldman, Laurence R. 1998 Child's Play. Myth, Mimesis and Make-believe. Oxford/New York: Berg. Goldstein, Kurt 1948 Language and Language Disturbances: Aphasis Symptom Complexes and their Signifiance for Medicine and the Theory of Language. New York: Grune and Stratton. Golla, Victor (ed.) 1984 The Sapir-Kroeber Correspondence. Berkeley: Survey of California and Other Indian Languages. Goodall, Jane 1986 The Chimpanzees of Gombe: Patterns of Behavior. Cambridge, M.A.: Harvard University Press.

238

References

Gopnik, Myrna 1999 Familial language impairment: more English evidence. Folia Phoniatrica et Logopaedica 51: 5-19. Gould, James L. and Peter Marler 1987 Learning by instinct. Scientific American 256.1: 62-73. Grammer, Karl 1985 Verhaltensforschung am Menschen: Überlegungen zu den biologischen Grundlagen des "Umwegverhaltens". In: M. Svilar (ed.), Mensch und Tier, 273-318. Bern/Frankfurt am Main/New York: Lang. Graybiel, Ann Μ., T. Aosaki, A. W. Flaherty, and M. Kimura 1994 The basal ganglia and adaptive motor control. Science 265: 1826-1831. Greenberg, Joseph H. 1953 Historical linguistics and unwritten languages. In: Alfred L. Kroeber (ed.), Anthropology Today, 265-286. Chicago: University of Chicago Press. 1987 Language in the Americas. Stanford: Stanford University Press. 1997 Mary Haas, Algic, and the scientific consensus. Anthropological Linguistics 39: 668-672. 2000 Indo-European and Its Closest Relatives: The Eurasiatic Language Family. Stanford: Stanford University Press. Greenfield, P. M. 1991 Language, tools and brain: the ontogeny and phylogeny of hierarchically organized sequential behaviour. Behavioural and Brain Science 14: 531-595. Greif, Esther B. and Jean Berko Gleason 1980 Hi, thanks, and goodbye: more routine information. Language in Society 9: 159-166. Grimm, Jacob [ 1964] Über den Ursprung der Sprache. In: Kleinere Schriften, 1: 255-298. Berlin: Dümmler. First published in 1851. [1984] On the Origin of Language. Translated by Raymond A. Wiley. Leiden: E. J. Brill. 1859 De l'origine du langage. Translated by Fernand de Wegmann. Paris: Α. Franck. Grimshaw, Jane B. 1990 Argument Structure. Cambridge, M.A.: MIT Press. Grossman, M., S. Carvell, S. Gollomp, M. Stern, G. Vernon, and H. D. Hurtig 1991 Sentence comprehension and praxis deficits in Parkinson's Disease. Neurology 41: 1600-1628. Grossman, M., S. Carvell, S. Gollomp, M. Stern, B. Reivich, D. Morrison, A. Alavi, and H. L. Hurtig 1993 Cognitive and physiological substrates of impaired sentence processing in Parkinson's Disease. Journal of Cognitive Neuroscience 5: 480-498. Gumperz, John J. and Stephen C. Levinson (eds.) 1996 Rethinking Linguistic Relativity. Cambridge: Cambridge University Press.

References

239

Guttenplan, Samuel 1997 The Languages of Logic. 2nd edition. Oxford: Blackwell. Haldane, John B. S. 1959 Natural selection. In: P. R. Bell (ed.), Darwin's Biological Works, 101 -149. Cambridge: Cambridge University Press. Hallpike, Christopher Robert 1985 Social and biological evolution I. Darwinism and social evolution. Journal of Social and Biological Structures 8: 129-146. 1986 Social and biological evolution II. Some basic principles of social evolution. Journal of Social and Biological Structures 9: 5-31. Harnad, Stevan R., Horst D. Steklis, and Jan Lancaster (eds.) 1976 Origins and Evolution of Language and Speech. New York: The New York Academy of Sciences. Harrington, D. L. and J. Haaland 1991 Sequencing in Parkinson's Disease: abnormalities in programming and controlling movement. Brain 114: 99-115. Harrison, Simon 1986 Laments for foiled marriages: love-songs from a Sepik river village. Oceania 56: 275-293. Hauser, Marc D. 1996 The Evolution of Communication. Cambridge, M.A.: MIT Press. Heeschen, Volker 1984 Ästhetische Form und sprachliches Handeln. In: Inger Rosengren (ed.), Sprache und Pragmatik. Lunder Symposium 1984, 387-411. Stockholm: Almqvist and Wikseil. 1988 Humanethologische Aspekte der Sprachevolution. In: Joachim Gessinger and Wolfert von Rahden (eds.), Theorien vom Ursprung der Sprache, 2: 196-248. Berlin: de Gruyter. 1992 The position of the Mek languages of Irian Jaya among the Papuan languages; history, typology, and speech. Bijdragen tot de Taal-, Land- en Volkenkunde 148: 465-488. 1998a An Ethnographic Grammar of the Eipo Language Spoken in the Central Mountains of Irian Jaya (West New Guinea), Indonesia. Berlin: Reimer. 1998b New Guinea myths and fairy tales seen from the Irian Jaya mountains. In: Jelle Miedema, Cecilia Ode and Rien A.C. Dam (eds.), Perspectives on the Bird's Head of Irian Jaya, Indonesia. Proceedings of the Conference Leiden, 13-17 Octobober 1997, 291-312. Amsterdam: Editions Rodopi B.V 2001 The narration "instinct": signalling behaviour, communication, and the selective value of storytelling. In: Jürgen Trabant and Sean Ward (eds.), New Essays on the Origin of Language, 179-196. Berlin: Mouton de Gruyter. Herbig, Jost 1986 Am Anfang war das Wort. Die Evolution des Menschlichen. Munich: Deutscher Taschenbuch Verlag.

240

References

Herder, Johann Gottfried 1966 On the Origin of Language. Edited by John H. Moran and Alexander Gode. Chicago/London: University of Chicago Press [1978] Abhandlung über den Ursprung der Sprache. Edited by Wolfgang Proß. Munich: Hanser. First published in 1772. Hewes, Gordon W. 1975 Language Origins: A Bibliography. 2 volumes. The Hague: Mouton. 1976 Opening statement. In: Harnad, Stevan R., Horst D. Steklis, and Jan Lancaster (eds.), Origins and Evolution of Language and Speech, 3. New York: The New York Academy of Sciences. 1977 Language origin theories. In: Duane M. Rumbaugh (ed.), Language Learning by a Chimpanzee. The Lana Project, 3-53. New York: Academic Press. Hinton, Geoffrey E. and S. J. Nowlan 1987 How learning can guide evolution. Complex Systems 1: 495-502. Hock, Hans Henrich 1986 Principles of Historical Linguistics. Berlin: Mouton. Holm, John A. 1988 Pidgins and Creoles. Volume 1. Cambridge: Cambridge University Press. Hondrich, Karl Otto 1999 Das Jahrhundert der Elektronik und der Kommunikation. Mensch im Netz. Der Spiegel 18: 131-145. Hopper, Paul J. and Elizabeth Closs Traugott 1993 Grammaticalisation. Cambridge: Cambridge University Press. Humboldt, Wilhelm von [1907] Über das vergleichende Sprachstudium in Beziehung auf die verschiedenen Epochen der Sprachentwicklung. In: Gesammelte Schriften. Edited by Albrecht Leitzmann. Volume 4. Berlin: Behr. First published in 1820. 1997 Essays on Language. Edited by Theo Harden and D. Farelly. Frankfurt am Main: Lang. Hurford, James R. 1990 Beyond the roadblock in evolution studies. Behavioral and Brain Sciences 13: 736-7. Hurford, James R., Michael Studdert-Kennedy, and Chris Knight (eds.) 1998 Approaches to the Evolution of Language: Social and Cognitive Bases. Cambridge: Cambridge University Press. Huttar, George L. and Huttar, Mary L. 1994 Ndyuka. London: Routledge. Illes, J., E. J. Metter, W. R. Hanson, and S. Iritani 1988 Language production in Parkinson's Disease: acoustic and linguistic considerations. Brain and Language 33: 146-160. Illich-Svitych, Vladislav M. 1971-1984 Opyt sravnenija nostraticheskix jazykov [An Attempt at a Comparison of the Nostratic Languages]. 3 volumes. Moscow: Nauka.

References

241

Jablonka, Eva and G. Rechav 1996 The evolution of language in light of the evolution of literacy. In: Jürgen Trabant (ed.), Origins of Language, (Workshop Series No. 2.), 70-88. Budapest: Collegium Budapest. Jablonski, Nina G. and Leslie C. Aiello (eds.) 1998 The Origin and Diversification of Language. San Francisco: California Academy of Sciences. Jackendoff, Ray 1983 Semantics and Cognition. Cambridge, M.A.: MIT Press. 1990 Semantic Structures. Cambridge, M.A.: MIT Press. 1999 Possible stages in the evolution of the language capacity. Trends in Cognitive Sciences 3.7: 272-279. Jakobson, Roman 1960 Closing statement: linguistics and poetics. In: Thomas A. Sebeok (ed.), Style in Language, 350-77. Cambridge, M.A.: MIT Press. Jellinger, Kurt 1990 New developments in the pathology of Parkinson's Disease. In: Μ. B. Streifler, A. D. Korezyn, J. Melamed, and Μ. Β. H. Youdim (eds.), Advances in Neurology. Volume 53: Parkinson's Disease: Anatomy, Pathology and Therapy, 1-15. New York: Raven Press. Jerison, Harry J. 1976 Palaeoneurology and the evolution of mind. Scientific American 234.1: 90-101. Jespersen, Otto. 1968 Language. Its Nature, Development, and Origin. London: George and Unwin. First published in 1921. Kainz, Friedrich 1967 Psychologie der Sprache. Volume 1. Stuttgart: Enke. Katz, W. F. 1988 Anticipatory coarticulation in aphasia: acoustic and perceptual data. Brain and Language 35: 340-368. Kegl, Judy, Ann Senghas, and Mary Coppola 1999 Creation through contact: sign language emergence and sign language change in Nicaragua. In: Michel DeGraff (ed.), Language Creation and Language Change, 179-237. Cambridge M.A.: MIT Press. Keller, Helen 1903 The Story of My Life. New York: Doubleday. Kendon, Adam 1988 Sign Languages of Australia. Cultural, Semiotic, and Communicative Perspectives. Cambridge: Cambridge University Press. Kimura, Μ., T. Aosaki, and A. Graybiel 1993 Role of basal ganglia in the acquisition and initiation of learned movement. In: Noriichi Mano, Ikuma Hamada and Mahlon R. DeLong (eds), Role of the Cerrebellum and Basal Ganglia in Voluntary Movements. Amsterdam/New York: Excerpta Medica.

242

References

Kimura, D. 1993

Neuromotor Mechanisms in Human Communication. Oxford University Press.

Oxford/New York:

Kirby, Simon Learning, bottlenecks and the evolution of recursive syntax. In Briscoe, 2000a E.J. (ed), Linguistic Evolution through Language Acquisition: Formal and Computational Models. Cambridge: Cambridge University Press. Syntax without natural selection: how compositionality emerges from 2000b vocabulary in a population of learners. In: James R. Hurford, Michael Studdert-Kennedy, and Chris Knight (eds.), The Evolutionary Emergence of Language: Social Function and the Origins of Linguistic Form, 302-323. Cambridge: Cambridge University Press. Klein, Richard G. 1999 The Human Career. Human Biological and Cultural Origins. Chicago: University of Chicago Press. Klein,Wolfgang 1994 Time in Language. London. Routledge. Klein, Wolfgang and Christiane von Stutterheim 1987 Quaestio und referentielle Bewegung in Erzählungen. Linguistische Berichte 109: 163-183. Klein, Wolfgang and Clive Perdue 1996 The Basic Variety, or: Couldn't natural languages be much simpler. Second Language Research 13: 301-347. Klima, Edward and Ursula Bellugi 1979 The Signs of Language. Cambridge, M.A.: Harvard University Press. Knight, Chris 1998 Ritual/speech coevolution: a solution to the problem of deception. In: James R. Hurford, Michael Studdert-Kennedy, and Chris Knight (eds.), Approaches to the Evolution of Language: Social and Cognitive Bases, 68-91. Cambridge: Cambridge University Press. Koehler, Otto 1974 Das unbenannte Denken. In: Klaus Immelmann (ed.), Grzimeks Tierleben. Enzyklopädie des Tierreichs. Sonderband "Verhaltensforschung", 320-336. Zurich: Kindler. Kortlandt, Adriaan 1972 New perspectives on ape and human evolution. Stichting voor Psychobiologie. Krebs, John R. and Richard Dawkins 1984 Animal signals: mind-reading and manipulation. In: John R. Krebs and Ν. B. Davies (eds.), Behavioural Ecology: an Evolutionary Approach. Oxford: Blackwell Scientific Publications. Kruuk, H. 1972 The Spotted Hyena: A Study of Predation and Social Behavior. Chicago: University of Chicago Press.

References

243

Kuiper, Koenraad 1996 Smooth Talkers: The Linguistic Performance of Auctioneers and Sportscasters. Mahwah, N.J.: Lawrence Erlbaum. Lakoff, George 1988 Cognitive semantics. In: Umberto Eco, Marco Santambrogio, and Patrizia Violi (eds.), Meaning and Mental Representations, 119-154. Bloomington: Indiana University Press. Landau, Barbara and Ray Jackendoff 1993 What and where in spatial language and spatial cognition. Behavioral and Brain Sciences 16: 217-65. Langacker, Ronald W. 1987 Foundations of Cognitive Grammar: Theoretical Prerequisites. Volume 1. Stanford: Stanford University Press. Lange, Κ. W., Τ. W. Robbins, C. D. Marsden, M. James, A. M. Owen, and G. M. Paul 1992 L-Dopa withdrawal in Parkinson's Disease selectively impairs cognitive performance in tests sensitive to frontal lobe dysfunction. Psychopharmacology 107: 394-404. Lass, Roger 1997 Historical Linguistics and Language Change. Cambridge: Cambridge University Press. Lawrence, Peter 1984 The Garia. An Ethnography of a Traditional Cosmic System in Papua New Guinea. Manchester: Manchester University Press. Laycock, Donald C. 1982 Melanesian linguistic diversity: a Melanesian choice? In: R. J. May and Hank Nelson (eds.), Melanesia: Beyond Diversity. Volume 1, 33-38. Canberra: Australian National University. Lefebvre, Claire 1998 Creole Genesis and the Acquisition of Grammar: The Case of Haitian Creole. Cambridge: Cambridge University Press. Leibniz, Gottfried Wilhelm 1710 Brevis designatio meditationum de originibus gentium ductis potissimum ex indicio linguarum. In: Miscellanea Berolinensia, 1-16. Berlin: Papenii. [ 1966] Nouveaux essais sur I 'entendement humain. Edited by Jacques Brunschwig. Paris: Garnier-Flammarion. First published in 1765. Lenneberg, Eric H. 1967 Biological Foundations of Language. New York: Wiley. Leroi-Gourhan, Andre 1964-1965 Le geste et la parole. 2 volumes. Paris: Albin Michel. Levin, Beth 1993 English Verb Classes and Alternations: A Preliminary Investigation. Chicago: University of Chicago Press.

244

References

Levin, Beth and Malka Rappaport-Hovav 1991 Wiping the slate clean: a lexical semantic exploration. Cognition 41: 123-151. 1995 Unaccusativity: At the Syntax-Lexical Semantics Interface. Cambridge, M.A.: MIT Press. Levinson, S. C. 1997 From outer to inner space: linguistic categories and non-linguistic thinking. In: Jan Nuyts and Eric Pederson (eds.), Language and Conceptualization. Cambridge: Cambridge University Press. Lewontin, Richard C. 1978 Adaptation. Scientific American 239: 157-169. Liberman, A. M., F. S. Cooper, D. P. Shankweiler, and Michael Studdert-Kennedy 1967 Perception of the speech code. Psychological Review 74: 431 -461. Lichtheim, L. 1885 On aphasia. Brain 7: 433-484. Lieberman, Philip 1968 Primate vocalizations and human linguistic ability. Journal of the Acoustical Society of America 44: 1157-1164. 1984 The Biology and Evolution of Language. Cambridge, M.A.: Harvard University Press 1991 Uniquely Human: The Evolution of Speech, Thought and Selfless Behavior. Cambridge, M.A.: Harvard University Press. 1998 Eve Spoke. Human Language and Human Evolution. New York/London: Norton. 2000 Human Language and Our Reptilian Brain: The Subcortical Bases of Speech, Syntax, and Thought. (Perspectives in Cognitive Neuroscience.) Cambridge, M.A.: Harvard University Press. Lieberman, Philip and E. S. Crelin 1971 On the speech of Neanderthal man. Linguistic Inquiry 2: 203-222. Lieberman, Philip, J. Friedman, and L. S. Feldman 1990 Syntactic deficits in Parkinson's Disease. Journal of Nervous and Mental Disease 178: 360-365. Lieberman, Philip, Ε. T. Kako, J. Friedman, G. Tajchman, L. S. Feldman, and Ε. B. Jiminez 1992 Speech production, syntax comprehension, and cognitive deficits in Parkinson's Disease. Brain and Language 43: 169-189. Lieberman, Philip, B. G. Kanki, A. Protopapas, E. Reed, and J. W. Youngs 1994 Cognitive defects at altitude. Nature 372: 325. Lieberman, Philip and C-Y. Tseng 1994 Subcortical pathways essential for speech, language and cognition: Implications for hominid evolution. American Journal of Physical Anthropology (Supplement 16) 93:130. Lieberman, Philip, B. G. Kanki, and A. Protopapas 1995 Speech production and cognitive decrements on Mount Everest. Aviation, Space and Environmental Medicine 66: 857-864.

References

245

Lock, Andrew and Charles R. Peters (eds. 1996 Handbook of Human Symbolic Evolution. Oxford: Clarendon. Lorenz, Konrad 1935 Der Kumpan in der Umwelt des Vogels. Journal für Ornithologie 83. 1973 Die Rückseite des Spiegels. Versuch einer Naturgeschichte menschlichen Erkennens. Munich: Piper. 1978 Das Wirkungsgefüge der Natur und das Schicksal des Menschen. Gesammelte Arbeiten. Edited by Irenäus Eibl-Eibesfeldt. Munich: Piper. Luckmann, Thomas 1988 Kommunikative Gattungen im kommunikativen "Haushalt" einer Gesellschaft. In: Gisela Smolka-Koerdt, Peter M. Spangenberg, and Dagmar Tillmann-Bartylla (eds.), Der Ursprung der Literatur: Medien, Rollen, Kommunikationsituationen zwischen 1450 und 1650, 279-288. Munich: Fink. Lyons, John 1988 Origins of language. In: A. C. Fabian (ed.): Origins. The Darwin College Lectures, 141-166. Cambridge: Cambridge University Press. Maddox, J. 1997 The price of language? Nature 388: 424-425. Malinowski, Bronislaw 1923 The problem of meaning in primitive languages. In: C. K. Ogden and I. A. Richards, The Meaning of Meaning, 296-326. London: Routledge and Kegan Paul. Marcus, Gary F. 1998 Rethinking eliminative connectionism. Cognitive Psychology 37: 243-282. Marcus, Gary F., S. Vijayan, S., S. Bandi Rao, and P. M. Vishton 1999 Rule learning by seven-month-old infants. Science 283: 77-80. Marie, P. 1926 Traveaux et Memoires. Paris: Masson. Marler, Peter 1998 Animal communication and human language. In: Nina G. Jablonski and Leslie C. Aiello (eds.), The Origin and Diversification of Language, 1-19. San Francisco: California Academy of Sciences. Marr, David 1982 Vision: A Computational Investigation into the Human Representation and Processing of Visual Information. San Francisco: W. H. Freeman. Marsden, C. David and Jose A. Obeso 1994 The functions of the basal ganglia and the paradox of sterotaxic surgery in Parkinson's Disease. Brain 117: 877-897. Marshall, Lorna 1961 Sharing, taking, and giving. Relief of social tensions among the Kung Bushmen. Africa 31: 231 -249.

246

References

Martin, Alex 1998 Organization of semantic knowledge and the origin of words in the brain. In: Nina G. Jablonski and Leslie C. Aiello (eds.), The Origin and Diversification of Language, 69-88. San Francisco: California Academy of Sciences. Maynard Smith, John and Eörs Szathmäry, 1995 The Major Transitions in Evolution. Oxford/New York: Freeman. Mayr, Ernst 1998 Was ist eigentlich die Philosophie der Biologie? Β erlin-Brandenburgische Akademie der Wissenschaften. Berichte und Abhandlungen 5: 287-301. McCawley, James D. 1981 Everything that Linguists Have Always Wanted to Know about Logic, but Were Afraid to Ask. Oxford: Blackwell. McWhorter, John H. 1998 Identifying the Creole type: vindicating atypological class. Language 74:788-818. Meillet, Antoine [1964] Introduction ά I'etude comparative des langues indo-europeennes. Tuskaloosa, A.L.: University of Alabama Press. First published in 1922. [1965] Le probleme de la parentέ des langues. In: Antoine Meillet, Linguistique historique et linguistique genirale, 1: 76-101. Paris: Champion. First published in 1914. Meschonnic, Henri 1996 Si notre marche dans le langage est une chute depuis l'origine ou si on y va en trebuchant. In: Jürgen Trabant (ed.), Origins of Language, (Workshop Series No. 2.), 116-141. Budapest: Collegium Budapest. Mesulam, M. Marcel 1990 Large-scale neurocognitive networks and distributed processing for attention, language, and memory. Annals of Neurology 28: 597-613. Metter, E. Jeffrey, D. Kempler, C. Jackson, W. R. Hanson, J. C. Mazziotta, and Μ. E. Phelps 1989 Cerebral glucose metabolism in Wernicke's, Broca's, and conduction aphasia. Archives of Neurology 46: 27-34. Michelson, Truman 1914 Two alleged Algonquian languages of California. American Anthropologist 16: 361-67. Mithun, Marianne 1992 Is basic word order universal? In: Doris L. Payne (ed.), Pragmatics of Word Order Flexibility, 15-62. Amsterdam/Philadelphia: John Benjamins. Müller, R.-A., R. D. Rothermel, Μ. E. Behen, 0. Muzik, P. K. Chakraborty, and H.T. Chugani 1999 Language organization in patients with early and late left-hemisphere lesion: a PET study. Neuropsychologia 37: 545-557. Musso, M., C. Weiller, S. Kiebel, S. P. Müller, P. Bülau, and M. Rijntjes 1999 Training-induced brain plasticity in aphasia. Brain 122: 1781-1790.

References

247

Muysken, Pieter C. and Norval Smith (eds.) 1990 Substrata Versus Universals in Creole Genesis. Amsterdam: John Benjamins. Natsopoulos, D., G. Grouios, S. Bostantzopoulou, G. Mentenopoulos, Z. Katsarou, and J. Logothetis 1993 Algorithmic and heuristic strategies in comprehension of complement clauses by patients with Parkinson's Disease. Neuropsychologia 31: 951 964. Nearey, T. 1979 Phonetic Features for Vowels. Bloomington: Indiana University Linguistics Club. Neville, H. J. and D. Bavelier 1998 Neural organization and plasticity of language. Current Opinion in Neurobiology 8: 254-258. Newmeyer, Frederick J. 1998 On the supposed "counterfunctionality" of Universal Grammar: some evolutionary implications. In: James R. Hurford, Michael Studdert-Kennedy and Chris Knight (eds.), Approaches to the Evolution of Language: Social and Cognitive Bases, 305-319. Cambridge: Cambridge University Press. Nichols, Johanna 1992 Linguistic Diversity in Space and Time. Chicago: University of Chicago Press. Nicolai', Robert 1998 De l'origine des langues: relectures. Bulletin de la Societe de linguistique de Paris 93.1: 157-180. Nikolaev, Sergei 1991 Sino-Caucasian languages in America. In: Vitaly Shevoroshkin (ed.), Dene-Sino-Caucasian Languages, 42-66. Bochum: Brockmeyer. Niznikiewicz, Μ., R. Donnino, R. W. McCarley, P. G. Nestor, D. V. Iosifescu, B. O'Donnell, J. Levitt, and Μ. E. Shenton 2000 Abnormal angular gyrus asymmetry in schizophrenia. American Journal of Psychiatry. Nobre, A. C. and K. Plunkett 1997 The neural system of language: structure and development. Current Opinion in Neurobiology 7: 262-268. Nowak, Μ. A. and D. C. Krakauer 1999 The evolution of language. Proceedings of the National Academy of Sciences of the USA 96: 8028-8033. Nowak, Μ. Α., J. Β. Plotkin, and V. Α. Jansen 2000 The evolution of syntactic communication. Nature 404: 495-498. Odling-Smee, F. J., Laland, Κ. N. & Feldman, Μ. W. 1996 Niche construction. The American Naturalist 147: 641-8. Paul, Hermann 1886 Prinzipien der Sprachgeschichte. 2nd edition. Jena: Niemeyer.

248

References

Parsons, Terence 1990 Events in the Semantics of English: A Study in Subatomic Semantics. Cambridge, M.A.: MIT Press. Paterson, S. J., J. H. Brown, Μ. K. Gsödl, Μ. Η. Johnson, and A. Karmiloff-Smith 1999 Cognitive modularity and genetic disorders. Science 286: 2355-2357. Peters, Ann 1983 The Units of Language Acquisition. Cambridge: Cambridge University Press. Pickett, Emily R. 1998 Language and the cerbellum. Ph.D. Dissertation, Brown University. Pickett, E., R. Kuniholm, A. Protopapas, J. Friedman, and Philip Lieberman 1998 Selective speech motor, syntax and cognitive deficits associated with bilateral damage to the head of the caudate nucleus and the putamen. A single case study. Neuropsychologia 36: 173-188. Pinker, Steven 1994 The Language Instinct: How the Mind Creates Language. New York: William Morrow. 1999 Words and Rules: The Ingredients of Language. London: Weidenfeld and Nicolson. Pinker, Steven and P. Bloom 1990 Natural language and natural selection. Behavioural and Brain Sciences 13: 507-508. Probst, Ernst 1991 Deutschland in der Steinzeit. Bauern, Bronzegießer und Burgherrn zwischen Nordsee und Alpen. Munich: Bertelsmann. 1996 Deutschland in der Bronzezeit. Jäger, Fischer und Bauern zwischen Nordseeküste und Alpenraum. Munich: Bertelsmann. Pustejovsky, James 1991 The syntax of event structure. Cognition 41: 47-82. 1995 The Generative Lexicon. Cambridge, M.A.: MIT Press. Putnam, Hilary 1975 The meaning of "meaning". In: Hilary Putnam, Mind, Language and Reality: Philosophical Papers, 2: 215-271. Cambridge: Cambridge University Press. Ragir, Sonia 2000 Constraints on communities with indigenous sign languages: clues to the dynamics of language origins. Paper presented at the conference The Evolution of Language, Paris, April 3-6, 2000. Rapoport, Stanley I. 1999 How did the human brain evolve? A proposal based on new evidence from in vivo brain imaging during attention and ideation. Brain Research Bulletin 50: 149-165.

References

249

Rappaport-Hovav, Malka and Beth Levin 1998 Building verb meanings. In: Miriam Butt, and Wilhelm Geuder (eds.), The Projection of Arguments: Lexical and Compositional Factors. Stanford: CSLI Publications. Renfrew, Colin and Daniel Nettle 1999 Nostratic: Examining a Linguistic Macrofamily. Cambridge: McDonald Institute for Archaeological Research. Renfrew, Colin 2000 Die Indoeuropäer - aus archäologischer Sicht. Spektrum der Wisssenschaft. Dossier. Die Evolution der Sprachen, 1: 40-48. Reynolds, Peter C. 1981 On the Evolution of Human Behavior. The Argument from Animals to Man. Berkeley: University of California Press. Richter, Klaus 1999 Die Herkunft des Schönen. Grundzüge der evolutionären Ästhetik. Mainz: von Zabern. Ridley, Matt 1996 The Origins of Virtue. Harmondsworth: Penguin. Rilke, Rainer Maria [1987] Eine Begegnung. In: Rainer Maria Rilke, Sämtliche Werke. Volume 7. Frankfurt am Main: Insel. First published in 1907. Rivarol, Antoine de [1995] Discours sur l'universalite de la langue frangaise. In: Pierre Penisson, Academie de Berlin: De l 'universalite europeenne de la langue frangaise. 1784, 127-186. Paris: Fayard. Rizzolatti, G. and Μ. A. Arbib 1998 Language within our grasp. Trends in Neuroscience 21: 188-194. Rolls, Edmund T. S. M. Stringer 2000 On the design of neural networks in the brain by genetic evolution. Progress in Neurobiology. 61: 557-579. Rose, Y. and P. Royle 1999 Uninfected structure in familial language impairment: evidence from French. Folia Phoniatrica et Logopaedica 51: 70-90. Ross, J. R. 1967 Constraints on variables in syntax. Ph.D. Dissertation, Massachusetts Institute of Technology. Rousseau, Jean-Jacques [1981] Essai sur l'origine des langues. Edited by Charles Porset. Paris: Nizet. First published in 1781. 1966 On the Origin of Language. Edited by John H. Moran and Alexander Gode. Chicago/London: University of Chicago Press. Ruhlen, Merritt 1994a Is Algonquian Amerind? In: Merritt Ruhlen (ed.), On the Origin of Languages: Studies in Linguistic Taxonomy, 111-126. Stanford: Stanford University Press.

250

References

1994b

First- and second-person pronouns in the world's languages. In: Merritt Ruhlen (ed.), On the Origin of Languages: Studies in Linguistic Taxonomy, 252-260. Stanford: Stanford University Press. 1994c Amerind t'a?na 'child, sibling'. In: Merritt Ruhlen (ed.), On the Origin of Languages: Studies in Linguistic Taxonomy, 183-206. Stanford: Stanford University Press. 1994d The Origin of Language. Tracing the Evolution of the Mother Tongue. New York: Wiley. 1995a A note on Amerind pronouns. Mother Tongue 25: 60-61. 1995b Proto-Amerind numerals. Anthropological Science (Tokyo) 103: 209-225. 1997 L'origine des langues, Sur les traces de la langue mere. Paris: Belin. Russell, Bertrand 1905 On denoting. Mind 14: 479-493. Ryalls, J. 1986 An acoustic study of vowel production in aphasia. Brain and Language 29: 48-67. Ryder, Mary Ellen 1994 Ordered Chaos: The Interpretation of English Noun-Noun Compounds. Berkeley: University of California Press. Sager, Sven Frederik 1995 Verbales Verhalten. Eine semiotische Studie zur linguistischen Ethologie. Tübingen: Stauffenburg-Verlag. Salmond, Anne 1976 Hui: α Study of Maori Ceremonial Gathering. Wellington: A. H. and A. W. Reed. Sankoff, Gillian 1977 Multilingualism in Papua New Guinea. In: Stephen A. Wurm (ed.), New Guinea Area Languages and Language Study, Volume 3, Language, Culture, Society, and the Modern World, Fascicle 1, 265-307. Canberra: Australian National University. Sapir, Edward 1913 Wiyot and Yurok, Algonkian languages of California. American Anthropologist 15: 617-646. Savage-Rumbaugh, E. Sue, K. McDonald, R. A. Sevcik, W. D. Hopkins, and E. Rubert 1986 Spontaneous symbol acquisition and communicative use by pygmy chimpanzees (Pan paniscus). Journal of Experimental Psychology : General 115: 211-235. Savage-Rumbaugh, E. Sue and D. Rumbaugh 1993 The emergence of language. In: Kathleen R. Gibson and Tim Ingold (eds.), Tools, Language and Cognition in Human Evolution, 86-100. Cambridge: Cambridge University Press. Savage-Rumbaugh E. S. and R. Lewin 1994 Kanzi: The Ape at the Brink of the Human Mind. New York: John Wiley.

References

251

Scancarelli, Janine 1996 Cherokee writing. In: Peter T. Daniels and William Bright (eds.), The World's Writing Systems, 587-592. New York: Oxford UniversityPress. Schaller, Susan 1991 A Man without Words. Berkeley: University of California Press. Schlicht, Ε. 1979 The transition to labour management as a gestalt switch. Gestalt Theory 1: 54-67. Schuchardt, Hugo [1976] Sprachursprung. In: Hugo Schuchardt-Brevier. Ein Vademecum der allgemeinen Sprachwissenschaft, 254-310. Edited by Leo Spitzer. Darmstadt: WBG. First published in 1919-1921. Senghas, Ann 1995 Children's contribution to the birth of Nicaraguan Sign Language. Ph.D. dissertation, Massachusetts Institute of Technology. Seuren, Pieter and Herman Wekker 1986 Semantic transparency as a factor in creole genesis. In: Substrata Versus Universals in Creole Genesis, 57-71. Amsterdam: John Benjamins. Shapleske, J., S. L. Rossell, P. W. R. Woodruff, and A. S. David 1999 The planum temporale: a systematic, quantitative review of its structural, functional and clinical significance. Brain Research Reviews 29: 26-49. Simon-Vandenbergen, Anne-Marie 1981 The Grammar of the Headlines in The Times, 1870-1970. Brussels: Paleis der Academien, Koninklijke Academie voor Wetenschappen. Simpson, G. G. 1953 The Baldwin effect. Evolution 7: 110-117. Singh, Joseoph Amrito Lai and Robert M. Zingg [1966] Wolf-Children and Feral Man. Hamdon, C.N.: Archon Books. First published New York: Harper & Row, 1942. Sommer, Volker 1992 Lob der Lüge. Täuschung und Selbstbetrug bei Tier und Mensch. Munich: Beck. Sorenson, E. Richard 1976 The Edge of the Forest. Washington, D.C.: Smithsonian Institution Press. Sperber, Dan and Deirdre Wilson 1986 Relevance: Communication and Cognition. Oxford: Β. Blackwell. Sperber, Dan 1996 Explaining Culture: A Naturalistic Approach. Oxford: B. Blackwell. Starostin, Sergei A. 1984 Gipoteza ο geneticheskix svj/azjax sinotibetskix jazykov s enisejskimi i severnokavkazskimi jazykami [An hypothesis concerning the genetic affinity of the Sino-Tibetan languages with Yeniseian and North Caucasian languages]. In: Lingvisticheskaja rekonstruktsija i drevnejshaja istorija vostoka [Linguistic Reconstruction and Ancient History of the East] (Moscow) 4: 19-38.

252

References

Steinthal, Heymann 1851 Der Ursprung der Sprache im Zusammenhang mit den letzten Fragen des Wissens. Eine Darstellung der Ansicht Wilhelm von Humboldts verglichen mit denen Herders und Hamanns. Berlin: Dümmler. Stevenson, J., R. E. Hutchison, J. Hutchison, Β. L. R.Bertram, and W. H.Thorpe 1970 Individual recognition by auditory cues in the common tern (Sterna hirundo). Nature 226: 562-563. Stokoe, William 1995 Language: gene-created or handmade. Sign Language Studies 89: 331 -346. Strathern, Andrew 1971 The Rope of Moka: Big-Men and Ceremonial Exchange in Mount Hägen, New Guinea. Cambridge: Cambridge University Press. 1974 Melpa amb kenan. Courting Songs of the Melpa People. Port Moresby: Institute of Papua New Guinea Studies. Sturtevant, Edgar H. 1933 A Comparative Grammar of the Hittite Language. Baltimore: Linguistic Society of America. Stuss, Donald T. and D. Frank Benson 1986 The Frontal Lobes. New York: Raven Press. Sutton, D. and U. Jürgens 1988 Neural control of vocalization. In: Horst D. Steklis and J. Erwin (eds.), Comparative Primate Biology, 4: 625-647. New York: A. R. Liss. Szathmäry, Eörs 1996 On the origin of language: utility and enchantment. In: Jürgen Trabant (ed.), Origins of Language, (Workshop Series No. 2), 8-38. Budapest: Collegium Budapest. Tabor, Whitney 1993 The gradual development of degree modifier sort of and kind of: a corpus proximity model. Chicago Linguistic Society 24: 451-465. Tabor, Whitney and Elizabeth Closs Traugott 1998 Structural scope expansion and grammaticalization. In: Anna G. Ramat and Paul J. Hopper (eds.), The Limits of Grammaticalization. Amsterdam: John Benjamins. Terrace, Herbert S. 1979 Nim. New York: Knopf. Thelen, Esther 1984 Learning to walk: Ecological demands and phylogenetic constraints. Advances in Infancy Research 3: 213-250. Thomason, Richard (ed.) 1974 Introduction to Formal Philosophy: Selected Papers of Richard Montague. New Haven: Yale University Press. Tomblin, J.B. and J. Pandich 1999 Lessons from children with specific language impairment. Trends in Cognitive Science 3: 283-285.

References

253

Tomlin, Russell S. 1986 Basic Word Order: Functional Principles. London: Croom Helm. Trabant, Jürgen 1990 Herder's discovery of the ear. In: Kurt Mueller-Vollmer (ed.), Herder Today, 345-366. Berlin/New York: de Gruyter. 1996 Origins of Language (Workshop Series No. 2). Budapest: Collegium Budapest. 2000 Inner bleating, cognition and communication in the language origin discussion. In: Herder Jahrbuch/Herder Yearbook, 1-19. Stuttgart: Metzler. Trombetti, Alfredo 1905 L'unitä d'origine del linguaggio. Bologna: Luigi Beltrami. 1923 Le origini della lengua basca. Bologna: Luigi Beltrami. Turner, Frederick, and Ernst Pöppel 1988 Metered poetry, the brain, and time. In: Ingo Rentschier, Barbara Herzberger and David Epstein (eds.), Beauty and the Brain. Biological Aspects of Aesthetics, 71-90. Basel: Birkhäuser. Van der Lely, H. J. K., S. Rosen, and A. McClelland 1998 Evidence for a grammar-specific deficit in children. Current Biology 8: 1253-1258. Van Valin, R. D. Jr. and Randy J. LaPolla 1997 Syntax: Structure, Meaning and Function. Cambridge: Cambridge University Press. Vargha-Khadem, F., Κ. Ε. Watkins, C. J. Price, J. Ashbruner, K. J. Alcock, A. Connelly, R. S. J. Frackowiak, K. J. Friston, Μ. E. Pembrey, M. Mishkin, D. G. Gadian, and R. E. Passingham 1998 Neural basis of an inherited speech and language disorder. Proceedings of the National Academy of Sciences 95: 2695-12700. Vihman, Marilyn May 1996 Phonological Development: The Origins of Language in the Child. Oxford: B. Blackwell. Waal, Frans de 1996 Good Natured. The Origins of Right and Wrong in Humans and Other Animals. Cambridge, M.A.: Harvard University Press. Waddington, Conrad H. 1953 Genetic assimilation of an acquired character. Evolution 7: 118-126. Walker, Stephen 1983 Animal Thought. London: Routledge and Kegan Paul. Warden, Carl J. and Lucien H. Warner 1928 The sensory capacities and intelligence of dogs, with a report on the ability of the noted dog "Fellow" to respond to verbal stimuli. Quarterly Review of Biology 3: 1 -28. Washburn, S. Lancaster 1978 Human behavior and the behavior of other animals. American Psychologist 33:405-418.

254

References

Watkins, Calvert 1992 Indo-European languages. In: William Bright (ed.), International Encyclopedia of Linguistics, 2: 206-212. New York: Oxford University Press. Wernicke, Carl 1874 The aphasic symptom complex: a psychological study on a neurological basis. In: R. S. Cohen and Mark W. Wartofsky (eds.), Boston Studies in the Philosophy of Science, Volume 4. Boston: Reidel. White, T. D., G. Suwu, and B. Asfaw 1994 Australopithecus ramidus, a new species of early hominid from Aramis, Ethiopia. Nature 371: 306-312. Whiten, Andrew and Richard W. Byrne 1997 Machiavellian Intelligence 2: Extensions and Evaluations. Cambridge: Cambridge University Press. Whiten, Andrew, Jane Goodall, W. C. McGrew, T. Nishida, V Reynolds, Y. Sugiyama, C. E. G. Tutin, R. W. Wrangham, and C. Boesch 1999 Cultures in chimpanzees. Nature 399: 682-68. Whitney, William Dwight 1967 Language and the Study of Language: Twelve Lectures on the Principles of Linguistic Science. New York: Charles Scribner. Wickler, Wolfgang 1967 Vergleichende Verhaltensforschung und Phylogenetik. In: Gerhard Heberer (ed.), Die Evolution der Organismen. Ergebnisse und Probleme der Abstammungslehre, 420-508. Stuttgart: Gustav Fischer. Wierzbicka, A. 1988 The Semantics of Grammar. Amsterdam: John Benjamins. Wiessner, Polly, and Akii Tumu 1998 Historical Vines. Enga Networks of Exchange, Ritual, and Warfare in Papua New Guinea. With translations and assistance by Nitze Pupu. Illustrations by Akii Tumu. Washington/London: Smithsonian Institution Press. Williams, Francis E. 1940 Drama of Orokolo: The Social and the Ceremonial Life of the Elema. Oxford: Clarendon Press. Worden, R. P. 1995 A speed limit for evolution. Journal of theoretical Biology 176: 137-152. Yovel Yirmiyahu 1989 Spinoza and Other Heretics. Princeton, N.J., Princeton University Press. Zahavi, Amotz and Avishag Zahavi 1997 The Handicap Principle. A Missing Piece of Darwin s Puzzle. New York/ Oxford: Oxford University Press. Zaidel, Dahlia W. 1999 Regional differentiation of neuron morphology in human left and right hippocampus: comparing normal to schizophrenia. International Journal of Psychophysiology 34: 187-196.

References Zimmer, Dieter Ε. 1986 So kommt der Mensch zur Sprache. Uber Spracherwerb, hung, Sprache & Denken. Zurich: Haffmans Verlag.

255

Sprachentste-

Index Adaptive selection 67-69 Amerind controversy 200-202 Aphasia 28-29, 31-32 Artificial langauges 112 Autonomy (of syntax) hypothesis 154,

Feral children 106-108 Functional language system (FLS) 21-24,33,37 Greenberg, Joseph 198-200, 204-207,

159, 174-175 Baldwin effect (genetic assimilation) 47, 134-135, 164-166 linguistic innovations 167-174 Berlin Academy 1 -4, 15-16, 81, 219 Bickerton, Derek 10, 42, 69, 71, 74, 105,

209-211 Grimm, Jakob 3, 219-223 Grooming patterns 25-26, 180 Grimm, Jakob 3

109, 111, 114, 119 Brain

70-72, 78, 101, 105, 117,219 Hittite, discovery of 202-203 Hominid speech 38-39, Human language faculty 92, 95, 99, 102 origin of 83, 87-89 selective advantage of 85 Humboldt, Wilhelm von 2-3

plasticity 46-47 size 136 Broca's aphasia 28-29, 31, 35 Broca's area 21-24, 45

Herder, Johann Gottfried 2, 4-5, 15, 68,

Chain formation 77-78 Child language acquisition 104-105

Individuals, mental representations of

Chimpanzees, language abilities of 37-

126-129 Individual constants 129-130 Indo-European 197, 209-210 relation to Hittite 202-202 relation to Nostratic 203-207 Island constraints 155-159

3 8 , 4 1 , 4 2 , 72, 116, 137-138 Chomsky, Noam 9-12, 21, 26, 40, 55, 56, 68, 77, 79, 149-151, 154, 174 Compositionality 62 Compositional rules 90-92, 101 of Basic Variety 95-98 Condillac, Etienne Bonnot de 2-5, 8, 15 Constituency (Phase Structure) 63, 77 Convergence (co-evolution) 137 Creole languages 108-110 Dene-Caucasian 207-209 Domain-specificity 150-151

Jackendoff, Ray 11, 71, 74, 159 Language capacity 55-59 evolution of 67 Language organ 41-43, Leibniz, Gottfried Wilhelm 2, 13, 25, 70-72 Lexicon 88-89, 94, 114, 116 Lexical system 72

258

Index

Language, localisation of 45-46 Motor activity/control/sequencing 23, 26-29, 35-37, 39-40 Narration 191-192, 195 New York Academy of Sciences 1 Parkinson's Disease 27-28, 33-36, 40 Perception, mental representation of 122-123 Phase Structure (Constituency) 63, 77 Pinker, Steven 10, 23, 66, 69, 71 Playful speech 189-190 Predicate Calculus 119-121 Protolanguage41, 71, 74, 105, 109, 111, 119 Protolexicon 72-73, 77, 79 Ritualisation 190, 195 Rousseau, Jean-Jacques 4, 8 Sapir-Michelson debate 196-199 Saussure, Ferdinand de 55, 83-84, 216, 218

Schizophrenia 50 Second language acquisition (SLA) 93-94, 100-101 Semantic categories 159-163 Sequential position 64-65 Sign languages American Sign Language (ASL) 110 Nicaraguan Sign Language (NSL) 110-112, 114, 117 Societe de linguistique de Paris 1,3, 13, 216 Specific language impairment (SLI) 41, 43-44 Stimulus diffusion 114 Trombetti, Alfredo 197, 199,211-212 Twin languages 113 Universal Grammar 21, 26, 40, 55-60, 62, 68-70, 77-79, 149-150 innateness of 65-66 Voice-onset-time (VOT) 31, 35-36 Wernicke's area 21-23, 32 Working memory 33-34