Universal History of Linguistics: India, China, Arabia, Europe 9789027245526, 9027245525

This wide-ranging book presents the linguistic achievements of four major cultures to readers presumably conversant with

132 82 34MB

English Pages 379 [378] Year 1991

Report DMCA / Copyright

DOWNLOAD FILE

Polecaj historie

Universal History of Linguistics: India, China, Arabia, Europe
 9789027245526, 9027245525

Table of contents :
UNIVERSAL HISTORY OF LINGUISTICS: INDIA, CHINA, ARABIA, EUROPE
Editorial page
Title page
Copyright page
Preface
Table of contents
1. Introduction
2. India
2.1 The Relation of Linguistic Theory to Hinduism
2.2 Linguistics before Pānini
2.3 Pānini
2.3.1 General Remarks
2.3.2 The Main Peculiarities of Pānini's Grammar
A) Use of Conventional Labels and Determinatives
B) Brevity of Pânini's Style
C) Order of Rules
2.3.3 The Types of Rules Employed by Pānini
A) Metagrammatical Rules
i) Definitions
ii) Interpretation Rules
B) Grammatical Rules
i) Expression Rules
ii) Combination Rules
iii) Substitution Rules
2.3.4 On the 'Axiomatic' Character of Pānini's Grammar
2.3.5 Pānini's Descriptive Technique: How It Works
A) The Notion of'Käraka'
B) How to Describe a Simple Sentence Containing Simple Words: an Outline
C) How to Describe a (Semantically) Complex Sentence: an Outline
D) How to Describe a Complex Word: an Outline
E) How to Describe a Simple Word: a Detailed Example
2.4 Linguistics after Pānini
2.4.1 Kātyāyana and Patañjali
2.4.2Bhartrhari
2.5 Conclusion: Pānini and the Riddle of 'Progress'
Notes
3. China
3.1 Confucius
3.2 Mencius
3,3 The School of Names
3,4 The Taoist School
3.5 Mo Tzu and His School
3.6 Hsün Tzu
3.7 The Legalist School
3.8 The Aftermath
3.9 Conclusion: Pānini and Confucius
Notes
4. Arabia
4.1 The Cultural Context of Arab Linguistics
4.2 General Characteristics of the Arab Linguistic Tradition
4.3 Central Aspects of the Arab Syntactic Theory
4.4 Metatheoretical Implications
4.5 Sībawaihi's Conception of 'Discourse-Based' Linguistics
4.6 Residual Issues
4.7 Conclusion: Pānini, Confucius, and Sībawaihi
Notes
5. Europe
5.1 Linguistic Thinking in Greek Philosophy
5.1.1 Plato
5.1.2 Aristotle
5.1.3 The Stoics
5.1.4 The Relation between Mind and Reality in Greek Philosophy
5.2 The Alexandrian School
5.2.1 Dionysius Thrax
5.2.2 An Interlude: Varro
5.2.3 Apollonius Dyscolus
A) The Attitude towards the Data
B) Grammatical Explanation
C) Abstract Analysis
D) Deciding the Category-Membership
E) The Problem of Isomorphism
5.2.4 The Aftermath: Priscian
5.3 Linguistics in the Middle Ages
5.3.1 Precursors: Ammonius, Boethius, Abaelard
5.3.2 Philosophy of Linguistics
5.3.3 Isomorphism as the Basis for Universal Grammar
5.3.4 Grammatical Description
5.3.5 Logical Linguistics
5.3.6 The Demise of the Modistic Doctrine
5.4 Grammatical Theory 1500-1900
5.4.1 The 16th Century
5.4.2 The 17th Century
5.4.3 The 18th Century
5.4.4 The 19th Century
5.5 The 20th Century
5.5.1 Some Philosophical Background
5.5.2 Autonomous Linguistics
5.5.3 Causal or Functional Linguistics
A) A Taxonomy of Explanations
B) The Notion of Analogy
a) Extralinguistic Reality
b) Extralinguistic Reality and Language
c) Language
5.6 Conclusion: Unity in Variety
Notes
6. Implications for the Philosophy of Science
6.1 The Notion of Progress
6.1.1 Progress in the Natural Sciences
6.1.2 Progress in Linguistics
6.1.3 Scientific vs. Social Progress
6.2 Relativism vs. Universalism
6.3 Externalism vs. Internalism
Notes
7. Conclusion
References
Name Index

Citation preview

UNIVERSAL HISTORY OF LINGUISTICS

AMSTERDAM STUDIES IN THE THEORY AND HISTORY OF LINGUISTIC SCIENCE General Editor E. F. KONRAD KOERNER (University of Ottawa)

Series III - STUDIES IN THE HISTORY OF THE LANGUAGE SCIENCES

Advisory Editorial Board Sylvain Auroux (Paris); Ranko Bugarski (Belgrade) H. H. Christmann (Tübingen); Rudolf Engler (Bern) Hans-Josef Niederehe (Trier); R. H. Robins (London) Rosane Rocher (Philadelphia); Vivian Salmon (Oxford) Aldo Scaglione (New York); Kees Versteegh (Nijmegen)

Volume 65

Esa Itkonen Universal History of Linguistics

UNIVERSAL HISTORY OF LINGUISTICS INDIA, CHINA, ARABIA, EUROPE

ESA ITKONEN University of Turku

JOHN BENJAMINS PUBLISHING COMPANY AMSTERDAM/PHILADELPHIA 1991

Library of Congress Cataloging-in-Publication Data Itkonen, Esa. Universal history of linguistics : India, China, Arabia, Europe / Esa Itkonen. p. cm. -- (Amsterdam studies in the theory and history of linguistic science. Series III, Studies in the history of the language sciences, ISSN 0304-0720; v. 65) Includes bibliographical references and index. 1. Linguistics -- History. I. Title. II. Series. P61.I85 1991 410'.9--dc20 91-30574 ISBN 90 272 4552 5 (Eur.)/l-55619-360-2 (US) (alk. paper) CIP © Copyright 1991 - John Benjamins B.V. No part of this book may be reproduced in any form, by print, photoprint, microfilm, or any other means, without written permission from the publisher.

Preface

Methodology of linguistics was the topic of my two previous books Grammatical Theory and Metascience (henceforth to be abbreviated as GTM) and Causality in Linguistic Theory (to be abbreviated as CLT). Therefore I found it natural to choose history of linguistics as my next topic. When I started the research, I had an inkling that the results might ultimately agree with those of my previous books. However, this proved to be the case even more than I had expected. Accordingly, the three books might be said to constitute a trilogy dealing with the methodology-cumhistory of linguistics. Writing a 'universal history' of a scientific discipline is a venture whose viability may not be obvious to everybody. To some extent, this may be just a matter of terminology. My use of the expression 'universal history' is explained in Introduction. As for my use of sources, I know Sanskrit, Classical Greek, and Latin well enough to have read e.g. Pānini, Plato, and Varro in the original. Since I know only the merest rudiments of Arabic syntax, Chapter 4 is based on translations of primary sources and on secondary sources. Since I do not know Chinese at all, the same is true of Chapter 3. Without the existence of the series Studies in the History of the Language Sciences, the writing of this book would have been much more difficult.

Maison des Sciences de l'Homme, Paris June 1991 Esa Itkonen

Contents 1. Introduction

2. India 2.1 The Relation of Linguistic Theory to Hinduism 2.2 Linguistics before Pänini 2.3 Pänini 2.3.1 General Remarks 2.3.2 The Main Peculiarities of Pāmni's Grammar 2.3.3 The Types of Rules Employed by Pänini 2.3.4 On the 'Axiomatic' Character of Pānini's Grammar 2.3.5 Pāninïs Descriptive Technique: How It Works 2.4 Linguistics after Pänini 2.4.1 Kātyäyana and Patañjali 2.4.2 Bhartrhari 2.5 Conclusion: Pänini and the Riddle of 'Progress' Notes

3. China 3.1 Confucius 3.2 Mencius 3.3 The School of Names 3.4 The Taoist School 3.5 Mo Tzu and His School 3.6 HsünTzu 3.7 The Legalist School 3.8 The Aftermath 3.9 Conclusion: Pänini and Confucius Notes

5 5 10 12 12 14 23 38 44 70 70 79 83 85

89 89 95 97 102 108 113 115 116 119 122

UNIVERSAL HISTORY OF LINGUISTICS

VIII

4. Arabia 4.1 The Cultural Context of Arab Linguistics 4.2 General Characteristics of the Arab Linguistic Tradition 4.3 Central Aspects of the Arab Syntactic Theory 4.4 Metatheoretical Implications 4.5 Sïbawaihi's Conception of 'Discourse-Based' Linguistics 4.6 Residual Issues 4.7 Conclusion: Pānini, Confucius, and Sībawaihi Notes

5. Europe 5.1 Linguistic Thinking in Greek Philosophy 5.1.1 Plato 5.1.2 Aristotle 5.1.3 The Stoics 5.1.4 The Relation between Mind and Reality in Greek Philosophy 5.2 The Alexandrian School 5.2.1 Dionysius Thrax 5.2.2 An Interlude: Varro 5.2.3 Apollonius Dyscolus 5.2.4 The Aftermath: Priscian 5.3 Linguistics in the Middle Ages 5.3.1 Precursors: Ammonius, Boethius, Abaelard 5.3.2 Philosophy of Linguistics 5.3.3 Isomorphism as the Basis for Universal Grammar 5.3.4 Grammatical Description 5.3.5 Logical Linguistics 5.3.6 The Demise of the Modistic Doctrine 5.4 Grammatical Theory 1500-1900 5.4.1 The I6th Century 5.4.2 The 17th Century 5.4.3 The 18th Century

125 125 128 132 145 149 157 159 161

165 166 167 174 183 189 191 192 194 201 216 219 220 226 232 237 244 249 252 253 259 211

CONTENTS

5.4.4 The 19th Century 5.5 The 20th Century 5.5.1 Some Philosophical Background 5.5.2 Autonomous Linguistics 5.5.3 Causal or Functional Linguistics 5.6 Conclusion: Unity in Variety Notes

6. Implications for the Philosophy of Science 6.1 The Notion of Progress 6.1.1 Progress in the Natural Sciences 6.1.2 Progress in Linguistics 6.1.3 Scientific vs. Social Progress 6.2 Relativism vs. Universalism 6.3 Externalism vs. Internalism Notes

IX

276 292 292 297 309 320 322

325 325 325 328 336 338 344 346

7. Conclusion

347

References

349

Name Index

363

1. Introduction

In this book I shall pursue a twofold goal: on the one hand, I intend to present a 'universal history' of linguistics (in a sense to be explained immediately below); on the other, I intend to spell out its significance for the general philosophy of science. In both endeavours I have been motivated by the conviction that in my chosen area of research there are important facts which are either ignored or not known well enough. The linguistic traditions of India and Arabia as well as the past of the Western tradition are better than the average linguist has been led to believe; and the ('universal') history of linguistics is apt to illuminate the notion of 'scien­ tific progress' (and hence the notion of 'science' tout court) in a new and rather surprising way. Both components of the term 'universal history' need to be explained. What does 'universal' mean in the present context? I have not tried to account for each and every occurrence of linguistic thinking in the world. To achieve some general validity, I have concentrated, rather, on traditions which are both indigeneous and long-lasting. Auroux (1989:22) accepts the existence of only three such traditions, namely the Indian, the Chinese, and the Greek ones. For my part, I have added the Arab tradition to this list, for reasons to be discussed in 4.1. I have treated of my subject matter as extensively as is required by my notion of 'history'. Indeed, what does 'history' mean in the present context? That is, how should one write the history of a scientific discipline? There are two obvious alternatives: either just to tell what happened, and — if possible — why; or to make, in addition, value judgments concerning what happened. Having started actual research, I soon discovered that it was impossible consistently to maintain the value-free approach. The idea of a norm, i.e. what is good and what is bad, is at the very center of scientific activity, as is shown by the fact that scholars are constantly engaged in an attempt to show that their results are better than those of their rivals. This fact can be ignored, also in an historical context, only at the cost of distorting science beyond recognition. Therefore, the value-free approach had to be abandoned. (Incidentally, it is my considered opinion that standard large-scale works on the history of linguistics vacillate uneasily between these contradictory points of view, namely being value-free and not being value-free. That this is a

2

UNIVERSAL HISTORY OF LINGUISTICS

muddle, is easily overlooked because the corresponding practice is so wide-spread.) This also means that the history of science cannot be written like the history of wars. In science, you may 'lose' first and 'win' only much later. As a consequence, I gave up the idea of writing a 'comprehensive and balanced' history of linguistics. To be sure, I have retained the historical point of view, by presenting linguistic theories in the chronological order, and even pointing out causal relationships where this has seemed possible. But I have dwelt upon those thinkers whose views I value, and I have dealt with others more perfunctorily (except when, in my opinion, they have been wrongly extolled). All the time, I have tried to justify the value judgments that I have made; and it is a conceptual truth that (genuine) justification must make appeal to intersubjectively valid criteria. In this way, I have tried to combine value-ladenness and objectivity. From a strictly scientific point of view, the linguistic tradition of China is without much interest. Therefore I might seem to be acting against my own advice, when I nevertheless discuss it in this book. It must be clearly understood, however, that in emphasizing its 'archaic' character, I present it as a foil for the three other traditions. As such, the Chinese example possesses considerable interest from the universal (or global) point of view, because it shows that there is no inherent necessity for a developed culture to produce a first-rate linguistic tradition. Some people have let me understand that writing a 'universal history of linguistics' is a task which should not, because of its magnitude, be undertaken by one person. Above, I have given some of the reasons why I think they are wrong. An additional reason is, of course, the unity of perspective which is virtually always absent from a collection of articles, but is quite easy for a single author to achieve. In my re-evaluations of what have generally been taken as 'historical facts', I have been motivated not only by a wish to tell the truth, but also by a wish to do justice to personages who have been treated unfairly by the posterity. This kind of undertaking has sometimes been disparaged as being a chasse aux précurseurs. I cannot share this view. Of course, there is a right way and a wrong way to do everything. It is wrong to claim to have discovered an antecedent where none exists in fact. But it is just as wrong not to point out an antecedent where it does exist, even if it has passed unnoticed, and it is right to do the opposite. (This is just a consequence of the fact, mentioned above, that 'winning' and 'losing' mean differ­ ent things in science and in war.) Besides, on some issues it is a matter not of showing that past generations have been precursors, but of showing that today's generations are (knowingly or unknowingly) epigons. Thus, I often point out simi­ larities between the past and the present; but just as often I trust the reader to draw

INTRODUCTION

3

his own conclusions. (This is the case e.g. with my discussion of Edmund Husserl, Georg von der Gabelentz, and Hermann Paul.) In the preceding paragraph I used the word 'truth' deliberately. Many philosophers of history deny that there is any such thing as the objective truth about past events, with the consequence that historiographical work is made to appear as just 'story-telling'. None of this sounded very implausible to me until I actually started to write this book. Then it soon became clear that there is no real alternative to being guided by the notion of (objective, historical) truth; everything else is just empty posturing. More precisely, I came to see that there are truths, at least in my chosen field of study, whose validity is not limited by historical or cultural bound­ aries. Thus I rejected relativism and embraced universalism (cf. 6.2). As the striking similarity between ancient Indian linguistics and (part of) modern North American linguistics vividly illustrates, external circumstances have a negligible impact on progress in linguistics, i.e. on what makes one linguistic theory superior to another. This is why I think I have been justified to write my (universal) history of linguistics in the style of 'history of ideas' (cf. 6.3). In any historical and/or cross-cultural study there is the problem of descrip­ tive concepts: Is it just the case that one imports the concepts of one's own time and place into another time and/or place? I have no clear-cut answer to this query, nor has anyone else. Of course, I had to start with concepts that are familiar to me, but I chose such as are likely to have a general application. (If I found out I was wrong, I chose others.) The concepts had to be modified to some extent, often just a little, when they were applied to theories of another time and/or place. It is these modified concepts which I in turn imported into my own time and place, when I at the conclusion of my study 'came back', as it were.

2. India

Any universal history of linguistics must start with India, because linguistics has a longer and better tradition in India than anywhere else. As far as recorded history is concerned, Pānini's (. 400 B.C.) grammar is the origin of Indian linguistics. At the same time, it determines in a definitive way the nature of subsequent linguistics, which consists in elucidating and complementing Pānini's sūtras. This is a per­ vasive feature of Indian culture: sūtra literature, formulated as a group of aphor­ isms, is followed by bhäsya literature, or commentary. Pânini's most important commentators were Kātyāyana (. 250 B.C.) and Patañjali (. 150 B.C.). Kätyäyana's commentary consists of  4,300 brief remarks (= vārttikas) dealing with about 1,000 of Pânini's sütras. Patañjali's Mahäbhäsya (= 'The Great Commentary') treats, in a more expansive form, both Kätyäyana's värttikas and Pânini's sütras. As a consummate grammarian, Pänini concentrates on linguistic form, mainly morphophonology, whereas Kätyäyana and Patañjali are also interested in semantics and philosophy of language. Pänini, Kätyäyana, and Patañjali are the three 'seers' (= rsi, muni) of Indian linguistics. Between the 6th and the 18th centuries the 'scholastic' Pāninian tradition was continued above all by Bhartrhari (c. 500), Kaiyata (c. 1000), Bhattoji-dïksita (c. 1600), and Nāgeaa (d. 1755). Today this tradition is represented e.g. by J. D. Joshi who, together with J. A. F. Roodbergen, has in several volumes edited, and commented upon, an important portion of Mahäbhäsya.

2.1 The Relation of Linguistic Theory to Hinduism Indian thinking is characterized by a preoccupation with linguistic questions which is quite extraordinary, both in intensity and in the number of areas covered. It is natural to search for reasons to explain this fact; and to do so, it is natural to start by looking into the possible influence of religion on linguistic thinking. Hinduism is the prevailing religion in India. The oldest sacred scriptures of Hinduism are the four Vedas (. 1300-800 B.C.), the most literary of which is the Rg-veda,1 a collection of hymns whose topics range from the origin of the universe

6

UNIVERSAL HISTORY OF LINGUISTICS

and the praise of various gods to problems of sexual rejection. The importance of language is expressed graphically in the hymn 10.125, where language (= vαc), personified as a goddess, glorifies herself in the following terms: I gave birth to the father on the head of this world. My womb is in the waters, within the ocean. From there I spread out over all creatures and touch the very sky with the crown of my head. I am the one who blows like the wind, embracing all creatures. Beyond the sky, beyond this earth, so much have I become in my greatness. The idea seems to be that since language can express the holiness of the gods as well as the immensity of the universe, it too shares these same character­ istics. This interpretation seems to be confirmed by the fact that brahman, which simply means 'sacred word' in the Vedas, acquires in the Upanishads (c. 800-400 B.C.), the next set of sacred scriptures, the meaning of 'Absolute', 'Supreme Being', or 'Ground of the Universe', and thus becomes the central concept of Hinduism (cf. Zaehner 1962:46). We thus have the meaning shift 'sacred word' > 'sacred thing'. When Bhartrhari later expresses the view that the universe is ultimately of linguistic nature, and defines brahman more narrowly as śabdabrahman, i.e. 'word-brahman' (cf. p. 80), he in fact performs the reverse meaning shift 'sacred thing' > 'sacred word'. In both cases the overriding importance of language is emphasized. Originally, the motivation for linguistic study was a religious one: to maintain the intelligibility and the right pronunciation of the Vedas (cf. 2.2). Thus religion commanded linguistic analysis of hymns which themselves put a high premium on language. These two viewpoints must have reinforced each other. Gradually the link between linguistics and religion grew more tenuous. Pänini's sütras are much too concise and matter-of-fact to betray any religious feeling; and the Vedic material treated by him is scanty and remains extraneous to his grammar as a whole. Kätyäyana and Patañjali are a little more explicit on this issue. According to them, when the use of language is backed by grammatical knowledge, it produces the same religious rewards as the study of the Vedas. A Brahmin, i.e. a member of the priestly class, just has to study grammar. More precisely, it is his duty to study the Vedas with the six 'auxiliary sciences', of which grammar is the most important one (the others being phonetics, metrics, etymology, astronomy, and ceremonial). I think Biardeau (1964:42) is right in her view that although Patañjali expounds at length the religious merits accruing from the study of grammar, he is just repeating the traditional view, without much inner

INDIA

7

conviction (cf. p. 77). Bhartrhari, by contrast, was clearly a religious man (cf. 2.4.2). It is only in the Upanishads that Hinduism acquires its characteristic content: The history of the mankind is seen as an eternal wearisome cycle of births and rebirths ( = samsara)', it is ordained that one should strive after liberation ( = moksa) from samsara; moksa may be obtained by such means as meditation, with the result that one's self ( = αtman) merges with brahman. At this very high level of abstraction, there is no difference between Hinduism and Buddhism, apart from terminology. For instance, the Buddhist term for moksa is nirvana. Moksa is a transcendental phenomenon which lies far beyond the ordinary, visible world ( = nāma-r, lit. 'names-and-forms'). Therefore Upanishadic Hinduism constantly stresses the inability of language to express the ultimate religious truths (cf. Zimmer 1973:35-37). In this respect Buddhism goes, if possible, even farther. This kind of attitude is in clear contradiction with the Vedic reverence for the expressive power of language (cf. above). The notion of moksa with its attendant mystical pantheism is indeed quite alien to the Vedas. One cannot help using language even in religious teaching, and since logical language is unsuitable, the preferred means is metaphor. The most frequent type of metaphor, employed both in the Vedas and in the Upanishads, consists in an elaborate correspondence, or isomorphism, between microcosmos and macrocosmos: ...macrocosmic man is the prototype of microcosmic man, and his various parts correspond to the different organs of the human individual: between man and the world there is analogy of being, and since creation is a sacrifice on the macrocosmic scale, so is this sacrifice renewed on the microcosmic scale a creative act which ensures the continued orderly existence of the universe (Zaehner 1962:44). In moksa the two terms of the correspondence fall together, to the point of forming a full circle: I who am food, eat the eater of food! I have overcome the whole world! (p. 52) Because language is at most an unreliable medium, religious teaching is to a large extent imparted by non-verbal means: the pupil is supposed to regulate his behavior in accordance with the directions given by his teacher ( = guru), with the ultimate

8

UNIVERSAL HISTORY OF LINGUISTICS

intention of being able to imitate the latter's behavior and thinking in all the details. This type of education is motivated by the assumption that actions which one performs oneself can teach something that words heard from someone else cannot possibly do. Accordingly, the pupil—teacher relationship covers all aspects of the daily life. This agrees with the requirement, characteristic of the 'Oriental' attitude in general, that one's thoughts and actions should constitute a harmonious whole. Claims unsupported by one's own behavior have no value, however ingenious they may appear (cf. Zimmer 1973:56-62). This educational ideal is by no means restricted to the religious context only. In the teaching of grammar, too, it is by respectful imitation of the guru that one gradually acquires the tradition represented by the latter, even if the purely intellectual element is more prominent than in religious teaching. This may explain to some extent why in the Indian context criticism of one's predecessors tends to be rather muted (cf. Scharfstein 1978a:72). Of course, there have always been acri­ monious debates between different traditions. The Indian philosophy contains six orthodox systems, that is, systems which accept the validity either only of the Vedas or both of the Vedas and of the Upanishads. They have to struggle, accordingly, with the problem of 'revelation'. This is, in one sense, a linguistic problem. The most conservative system, Mīmαmsa, accepts no authority above the Vedas, for instance no God who would have issued or revealed them. As a consequence, Mīmαrnsa,while restricting transcendental considerations to a minimum, regards the Vedic texts, and the language employed therein, as eternal and immutable. The Nyäya system could be characterized as the most progressive one. It is concerned with logic, and continuously develops it in the direction of greater formalization. At the same time, Nyäya accepts the auth­ ority of one supreme God and regards the Vedas as issued by Him. Because issuing a text is an action situated in time, the Vedic language is non-eternal; in fact, it is not too different from ordinary language.2 Finally, a compromise between the two views is sought after by the Vedαntasystem, which is the philosophical expression of the most typical Hindu thought. As its name indicates, it represents the Upanishadic position (Veda-anta: 'end — or fulfillment — of Veda'). A brahman-like God issues the sacred scriptures at the beginning of each cosmic cycle; but they remain the same. So sacred language is both non-eternal and eternal (cf. Hiriyanna 1948:139, 101, and 169; Biderman 1978:139-146). The six systems, of which I have mentioned here three, constitute, pairwise, three superordinate systems. (For instance, Mīmαmsa and Vedαnta are the terms of one such pair.) Each of these three systems has its own fully articulated meta­ physics, with detailed answers given to all basic epistemological and ontological

INDIA

9

questions. The systems also deal with philosophy of language at a level not yet involved in the intricacies of 'revelation' (cf. Biardeau 1964: Chap. III). All this has an obvious relevance to my topic, but it would be difficult not to be over­ whelmed by the details. For my part, I have solved this problem in such a way that I shall let Bhartrhari's metaphysics represent that aspect of Indian philosophy which might be of interest to linguistic theory (cf. 2.4.2), with only a passing reference to divergent conceptions. In addition to the six orthodox systems, there are three non-orthodox ones. Two of these, viz. Buddhism and Jainism, have a religious core not too different from Hinduism. It is only the third one, variously called Lokäyata (= 'Doctrine of the Ordinary World') or Svabhuva-vûda (= 'Doctrine of the Nature in Itself) which represents the non-religious, materialistic point of view. It is extremely important to realize that it has been possible to develop also this kind of philosophy in India. So there has been no cultural determinism such that, in the post-Vedic age, every Indian had to consider the visible world either as just an illusion or at least as inferior to the invisible, transcendental world. Because religion has been intimately connected with the political power, and because the latter has to a large extent controlled the written transmission of knowledge, very few texts expounding Indian materialism have survived. For the most part, information about it must be inferred from what its opponents have written. Hiriyanna (1948:59) unwittingly bears witness to why, given the general intellectual climate of India, materialistic texts have all but disappeared: But even at its best the materialistic theory carries no conviction with it, since it tries to account for the higher principle of mind by the lower one of matter. And even here 'matter' should not be taken literally, because representatives of Lokäyata were willing to accept the notion of a (mortal) soul coexistent with the body and to regard social behavior as motivated by goals, viz. those of obtaining wealth and pleasure. An educated Indian had four basic goals in life, namely wealth and political power (= artha), pleasure (= kα), fulfillment of the socio-religious duties and rituals (= dharma), and liberation from the cycle of rebirths (= moksa). Zimmer (1973) is structured in such a way that the three first goals constitute the subject matter of the 'philosophy of the finite', whereas the fourth constitutes the subject matter of the 'philosophy of the infinite'. Now Lokäyata might be considered simply as a philosophical system which rejects moksa as mere superstition and

10

UNIVERSAL HISTORY OF LINGUISTICS

dispenses with dharma because of its intimate links with religion as well as with the caste system. In the same vein, Mīmαmsa concentrates on dharma and ignores moksa, whereas Vedαnta incorporates all four goals and constructs a hierarchy leading from material (= artha and kα) via ethical (= dharma) to transcendental (= moksa). — It may be added that to an outsider, dharma seems very much to exemplify an ethics of Brahminic self-indulgence. In the remainder of this chapter I shall disregard linguistic views which are justified by religious considerations, like the view that the Vedic language is eternal and immutable just because of the sanctitude of the Vedas. By contrast, I shall try to give a fair hearing to reasonable views which have been merely suggested by religious considerations. For instance, it might be argued that the considerations just mentioned induced Mīmαmsa philosophers to realize, quite correctly, that any given state of language can be viewed as an abstract, and in this sense 'timeless', entity.

2.2 Linguistics before Pānini Of the linguistic subdisciplines, phonetics and phonology were the first to develop. The Vedas were composed long before the advent of the writing system. They were recited aloud, and when the everyday language began to change in the direction of what was to be called 'Classical Sanskrit', it became more and more important to maintain the correct pronunciation of the Vedas. The concern with (correct) pronunciation quite naturally engendered an interest in how sounds are produced. The classification of sounds according to their place and manner of articulation goes back to the beginning of the post-Vedic time, i.e.  800-700 B.C. In addition to being pronounced correctly, the Vedas had to be understood. This required, among other things, an ability to establish the word-boundaries and thus to cancel the effects of external sandhi. The same treatment was extended to internal sandhi between the members of compound words and between word stems and certain endings. As a result, the concrete 'continuous recitation' (= samhitαpαtha) of the Vedas was analyzed as, or reduced to, an abstract 'word-for-word recitation' (= pada-pαtha) (cf. Scharfe 1977:81-82). The relation between the two levels can be illustrated with the following example, taken from Staal (1974:63): Padapätha: /a/urú/aprah/ámartyá/ni-vátah/deví/ut-vatah Samhitāpātha: órvaprā ámartyā niváto devyudvátah

INDIA

11

'The immortal goddess has pervaded the wide space, the depths and the heights' To speak only of the correspondence á/urú/aprαhvs. órvaprα, it is clear that every analysis of Vedic and Sanskrit must recognize just these same sandhi rules, i.e. 'a + u→ ' and ' u→v / _ V '. It is very important to realize that such results of general validity were achieved before the invention of any writing system, i.e. at a time when spoken language was analyzed, and the analysis was preserved, by means of spoken language. This should give pause to those who think that linguis­ tics as we know it today has been systematically distorted by some sort of 'written language bias' (cf. 2.3.1). It is quite natural to conceptualize the relation between the padapätha and the samhitäpätha in such a way that the former, being the more abstract of the two, changes into, or is replaced by the latter. From the orthodox Vedic point of view, however, such a conception engenders difficulties because the samhitäpätha is quite obviously temporally prior to the padapätha, and is moreover supposed to be the sacred word, whereas the padapätha is the result of human efforts. The phonetic-phonological analysis described above applies only to the Vedas, which constitute a closed corpus. This tradition has been preserved in socalled prαtiśàkhya manuals. These have generally been assumed to represent the original post-Vedic analysis, once it was written down after the invention of the alphabet. It is possible, however, that they belong to a later phase of the same tradition (cf. Scharfe 1977:127-128). This is not a question of great consequence, because the tradition was admittedly a very conservative one. To understand the Vedas, it was also required, of course, that one should know the meanings of the words whose formal identity was established by means of the padapätha analysis. The branch of linguistics which took care of this aspect was etymology (= nirukta). Our conception of the ancient Indian etymology is based on the Nirukta written by Yāska. Because Yāska's scholarly achievement stays far behind Pānini's, it has been assumed that he temporally preceded the latter. This is a non-sequitur, however. The example both of European and of Arab linguistics shows that diachronic linguistics may take a very long time to develop even after synchronic linguistics has reached quite a high degree of sophistication. Pānini mentions by name ten other grammarians. Because their work has disappeared without trace, it seems rather pointless to speculate about what it might have consisted in. So much is clear, in any case, that since grammar was one of the six auxiliary sciences of the Vedas (along with phonetics and etymology), there

12

UNIVERSAL HISTORY OF LINGUISTICS

must have been substantial achievements in grammatical analysis already before Pänini's time.

2.3 Pänini

2.3.1 General Remarks As its name indicates, Pänini's grammar AstädhyäyT contains eight chapters (= adhyäya); each chapter contains four quarter chapters (= pαda); each quarter chapter contains rules (= sūtra), the number of which ranges from 38 to 223. In all, the grammar contains a little less than 4,000 rules. This three-level structure has been invented by commentators. In accordance with it, each rule is customarily indicated with the aid of three numbers standing, respectively, for chapter, quarter chapter, and rule. Pänini investigated the spoken language (= bhäsä) of the educated class of his own time (cf. Cardona 1975:238, Scharfe 1977:106). As codified by him, this state of language, i.e. Classical Sanskrit, remained for some 2,000 years the principal, if not the sole, language of the Indian literary culture. (To be sure, the Indo-European tripartite accent disappeared from Sanskrit perhaps 200 years after Pänini.) Those few of Pänini's rules which deal with Vedic material are provided with the explicit indication chandasi (= 'in the Veda'). Thus, unlike the Prātiaākhyas, Pānini does not study a closed corpus: "[Pänini's grammar] does not concern itself with actual usage but only with possible correct usage" (Joshi 1968:x). For a relatively long period of time, i.e. at least for a couple of hundred years, Pänini's grammar was not written down at all. It existed only as memorized, and now and then as orally transmitted, by students of the grammatical science. Interestingly enough, it has always been the cornerstone of Brahmin education. A Chinese Buddhist who spent the years 671-695 in Sumatra and India, gives the following account of the contemporary curriculum in India (cf. Staal 1972:11-17). The teaching begins when the pupils are six year old. At the age of eight they memorize Pänini's grammar, which takes about eight months. Then during several years they are taught to understand AstädhyäyT at an elementary level; they also learn some appendices to it. At the age of fifteen they start to learn a more advanced commentary on Pänini, viz. Kαśikä-vrtti\ this takes about five years. After learning additionally logic and metaphysics, they spend a couple of

INDIA

13

years in places that we might call universities. Finally they go to the king's court and display their learning. Those who are successful receive grants of land and are appointed to the highest government posts. Meanwhile, the true scholars continue with the study of grammar. They learn first Patañjali's Mahäbhüsya and then Bhartrhari's commentary on it as well as his Vukyapadtya ('Investigation of Sentences and Words'); this takes some five additional years. It is only now that they may call themselves 'well-informed': "A person who has studied so far as this, is said to have mastered grammatical science." A more recent account has been given by a 20th century observer. He notes that memorization still continues to play an important role: There are various ways in which a child may begin in this [i.e. not-justreligious] method. A favorite way is to start him out with Sanskrit gram­ mar. ...He will be expected to memorize the rules and to learn to apply them. To watch children applying these rules is something utterly fascinat­ ing. ...At each step the child justifies what he has done by quoting the pertinent rule. ...Often there are conflicting rules or a choice of various methods of explanation. The constructions then are like the problems of geometry. They require not only memory but imagination and ingenuity. If a child can stand this training for the first year, he usually comes to enjoy it. ...The teaching of logic and philosophy proceeds in much the same way (Ingalls 1959:5). Those familiar only with the Western intellectual tradition cannot help being surprised at the fact that for such a long time there was no need to write down Pänini's grammar. For them, it is nearly impossible to imagine the existence of a thriving and sophisticated science without the concomitant existence of scientific texts.3 The nothing-but-oral mode of existence of Pänini's grammar may have much to do with the Bhrahminic habit of regarding knowledge as a secret which should be guarded as closely as possible (cf. Zimmer 1973:62-68, Scharfstein 1978a:86). Whatever the reasons for this aspect of Pänini's grammar, it refutes the standard Western concept of science, by showing that some of its basic ingredients are not necessary, but only contingent. The putative importance of writing can be highlighted in different ways. One way is to claim that science cannot exist without scientific texts. Another is to claim that a science like linguistics exhibits a 'written language bias'. This latter position has been argued at great length by Linell (1982). On the one hand, he maintains that linguists have always analyzed written, and not spoken sentences, with the consequence that units resulting from grammatical analysis, or segmen-

14

UNIVERSAL HISTORY OF LINGUISTICS

tation (i.e. sounds, syllables, words, phrases, clauses, sentences), are really characteristic of written language only. On the other hand, not just the language to be described (i.e. object language), but also the language which does the describing (i.e. metalanguage), is identical with written language, with the consequence that the theoretical-descriptive concepts too (e.g. the static hierarchical structure, or the 'deep vs. surface' distinction) are fashioned after the example of written language. Now Pänini's grammar refutes this line of thinking on both accounts. First, since Pänini gives neither spoken nor written examples, there is no reason to claim that he investigates only written language. In fact, it is questionable whether the Sanskrit alphabet already existed by Pänini's time.4 It certainly did not exist at the time when the first Prätiaäkhya-type analyses were made; and the results achieved by these are in basic agreement with Pānini's results (cf. 2.2). Secondly, since Pänini does not write his analysis at all, but speaks it, it hardly makes sense to claim that his (spoken) analysis is affected by writing. — Notice, incidentally, that Linell's claims are doubtful even when applied to modern Western grammarians. It is undeniable that they are in some sense operating with written sentences. But it does not follow at all that they are also investigating written sentences. It is more correct to say that they first imagine (or 'intuit') a correct spoken sentence, i.e. they imagine someone speaking this sentence; and it is only afterwards that they write this sentence down. That they do so, proves nothing, because (at least for Westerners) it is only normal to present one's scientific views in writing. An anthropologist writes down theritualshe has observed; but it does not follow that he is just investigating written sentences. The same applies to the grammarian, with only the difference that he has neither acted himself nor observed actions by others, but has imagined himself as either acting or observing.5 Linell (1982:31) is prudent enough to confine his claims to Western linguistics. But this does not help. Let us grant, for the sake of argument, that ever since the Stoics, Western linguists have done nothing but describe written sen­ tences. Although less well presented, the results they have achieved are identical with those achieved by Indian linguists investigating spoken sentences (cf. below). This shows that the opposition of speaking vs. writing is misconceived. What is at issue is just language, whether spoken or written.6

2.3.2 The Main Peculiarities of Pänini's Grammar Knowledge of general linguistics and of Sanskrit is not enough if one wishes to understand Pänini. In addition, the aid of secondary literature is absolutely necess-

INDIA

15

. It might be said that there is such a huge number of commentaries on Pānini's grammar, simply because it really needs to be commented upon. But why is this? What makes it so difficult? It seems advisable to answer this question, before tackling Pānini's grammar more directly. Personally speaking, the commentary which finally let me penetrate the secrets of Pānini's system was Junnarkar (1977 -1987). Without its help, I might have been forced to resign myself to William Jones' view that Pänini is "dark as the darkest oracle". Cardona (1988) too provides an excellent overall introduction to Pänini. Pineault (1989), though less extensive, also manages to place Panini's grammar in its larger historical context. Even for one who knows Sanskrit perfectly, Pānini's text is incomprehen­ sible, unless he has been schooled expressly in the Pacinian tradition. This general difficulty stems from three principal sources, which I present in a descending order of importance. First, an abundant use of conventional labels7 and 'determinatives' (i.e. auxiliary symbols) which have no meaning outside the Pāninian tradition. Secondly, an extreme brevity, evinced by the fact that rules often consist only of one or two words (which, to repeat, are for the most part incomprehensible to a non-Pāninian). Third, an apparent lack of systematicity in the order of the rules. (It may be added that the first point overlaps with the second insofar as conventional labels contain fewer sounds than ordinary words and thus abbreviate the rules in which they occur.) I shall now elucidate the three preceding points in more detail. In what follows, I shall often discuss the same set of Pānini's rules again and again from different points of view. In this way I hope to make these rules familiar to the reader. Since they are certainly among the most important rules of Pānini's gram­ mar, the reader will hopefully end up having a 'three-dimensional' picture of what the grammar is about. A) Use of Conventional Labels and Determinatives For clarity, theoretical-descriptive, non-Sanskrit elements will be indicated with capital letters. It should be remembered, however, that they were pronounced like ordinary sounds and words. I shall first give examples of conventional labels. Temporal and modal endings of the verb are collectively represented by the label L (3.4.77), which was pronounced and written as lah. This general category is further divided into ten subcategories, with labels like LAT (= present endings) or LAN (= imperfect endings). Absolute deletion of an affix is labelled LU (1.1.61), to be further sub­ divided, depending on the context, into LUK, ÉLU, and LUP. For instance, it is

16

UNIVERSAL HISTORY OF LINGUISTICS

stipulated for simplicity that all non-verbs have nominal endings; but, according to 2.4.82, these are deleted by LUK in connection with indeclinable words (defined in 1.1.37-41). LU is contrasted with lopa, which leaves the function of the deleted unit intact and lets it be represented by zero (1.1.62). Finally, that part of a word which begins with its last vowel is labelled 77 (1.1.64). The use of 77 will be illustrated later. Next I shall give examples of determinatives (called 'it' by Pānini, and 'anubandha' by the subsequent tradition). They are single sounds which are added to lexical or grammatical units and perform the most varied functions. They are literally 'auxiliary' because they must be deleted — by lopa\ — before the deri­ vation is over (1.3.9). That AstödhyäyT was indeed originally spoken, and not written, is conclusively demonstrated by the fact that only sounds can function as determinatives. Since there are hundreds of functions to be performed, but only a limited number of sounds, the result is a pervasive and multiple ambiguity, which strongly enhances the general difficulty of Pänini's text. It would have been much easier to use written diacritics, but apparently the written medium was either unavailable or undesirable. Later, of course, determinatives came to be represented by letters, but some of them — in particular, those originally indicated by accents — were lost in the process (cf. below). First of all, determinatives are employed for subclassification. For instance, adding Ñ to the verbal root (like in níÑ-) indicates that medial endings are used with the corresponding verb only if the action is for the benefit of the agent (1.2.72). By contrast, adding Ñ (like in dīÑ-) indicates that medial endings are necessary also when normal active meaning is to be expressed (1.3.12). Determinatives may also trigger various processes in their surroundings. Consider the following example: 1. trimśat + DaN 2. trimś + aN 3. traimś 4- a The functions performed by the determinatives can be directly read off from this derivation. D deletes the preceding vowel (and any intervening consonant), and disappears itself. A changes the basic degree of the stem vowel (= i, u, r, l) into its full form (= ai, au, αr, αl, called 'vrddhi'), and disappears itself. Determinatives are also used in the formation of 'condensation' labels (= pratyαhαra) which represent such sound classes (whether 'natural' or not) as occur

INDIA

17

in phonological rules. The pratyāhāra symbols are a central feature of Panini's descriptive technique. Let us take the vowels as an example: 1. 2. 3. 4.

a  e ai

i u 1 o au

N  N C

The symbol for a sound class is formed by taking the first intended sound and the determinative following the last intended sound (1.1.71). As a result, we get e.g. these symbols: 'aK' = {a, i, u, r, l}, or short and long vowels; 'eC' = {e, o, ai, au}, or 'diphthongs' (that is, e and  are always long, and often coincide with ai and au); 'aC' = all vowels. The pratyähära notation ignores quantity, nasality, and accent. (When these are taken into consideration, there are 144 vowel variants in all to be distinguished.) There are 34 semivowels and consonants. Their description necessitates 10 additional lines and as many determinatives. Pratyähära symbols are also used to express the categories of noun and verb. The first basic case ending is the nominative singular, or sU, and the last is the locative plural, or suP; thus, sUP = '(that which has) nominal endings'. The first basic verb ending is the active, indicative, present, third person singular, or tiP, and the last is the medial, indicative, imperfect, first person plural, or mahiW; thus, tiN = '(that which has) verbal endings'. As a first, rough approximation, 'word' (= pada) is defined thus: sUP-tiÑ-antam padam (1.4.14) ('A word is that which has either nominal or verbal endings'). It will be recalled that adverbs and particles too have sUP, which is then deleted by LUK. The rules 1.3.2-8 indicate which sounds may in which environments function as determinatives. As mentioned before, determinatives are multiply ambiguous. Often their functions are not stated directly, but must be inferred from how they are used. "Sanskrit is far and away the most elaborately structured language of the entire [Indo-European] family" (Baldi 1983:53-54). Pānini mentions practically every affix of Sanskrit, equipped for the most part with one or more determinat­ ives. The resulting grammatical entities are enumerated together with their uses by Katre (1981), in addition to the conventional labels and determinatives proper. It takes 260 closely printed pages to do all this. No wonder that AstādhyāyT could not be mastered in less than 15 years (cf. p. 12). In Pānini's theoretical-descriptive metalanguage e.g. the sound lah functions as i) a self-referential symbol representing the phoneme /1/ (e.g. 8.2.12): ii) a

18

UNIVERSAL HISTORY OF LINGUISTICS

conventional label representing the superordinate category 'temporal-modal ending' (3.4.77); iii) a determinative indicating that the preceding syllable has the high (udāttα) accent (6.1.193); iv) a determinative following the last sound of the sound table (thus aL = 'any sound'). The grammar of Pânini's metalanguage has been investigated by Scharfe (1971). In general, theoretical-descriptive units are inflected like ordinary nouns. In fact, there are almost no finite verbs in Pânini's rules. Thus the semantic complexity of the metalinguistic units as well as of the rules them­ selves is counterbalanced by the syntactic simplicity of Pânini's style. B) Brevity of Pânini's Style Conventional labels and pratyâhâra symbols, being shorter than the corresponding ordinary words, contribute to compressing the rules. Thus, a typical rule is as concise as this: iko yan aci (6.1.77) or, when the determinatives are indicated, iKo yaNaCi. Brevity is also achieved by exploiting the expressive resources of Sanskrit, such as the nominal sentence and the so-called bαhwvrtfH-construction (= possessive compound). The latter was already exemplified by the expression sUP-tiN-antam, which means, not 'sUP-tiN-end', but 'that which has a sUP-tiÑ-end'. Because of their formal simplicity, the meaning of rules must be unfolded by successive expansions. Consider the rule 1.1.46:

More explicitly: 'A unit having the -determinative precedes, and a unit having the K-determinative follows the unit under discussion.' Even more explicitly: {X, Ta} = aX, whereas {X, Ka} = Xa. Or consider the rule 1.1.64: aCo 'ntyādi 77. A literal translation would be something like this: 'The final beginning of a vowel (is called) '.' What it means is this: 'That part of a word which begins with its last vowel is called "IT.' Thus e.g. in the verb-forms anayata and anayetām, -a and -ām are, respectively, 77. To give one more example, consider the rule 1.1.55: anekāl śitsarvasya. Its more transparent form is an-eka-aL S-it sarvasya, literally 'Not-one-sound 5determinative instead of the whole'. The definitive meaning is 'A substitute

INDIA

19

consisting of more than one sound or having the Ś-determinative replaces the whole string of sounds occupying the position of a substituendum' (whereas, as indicated in 1.1.52, a substitute consisting of just one sound replaces only the last member of the string of sounds occupying the position of a substituendum). Thus (to put it anachronistically) XYZ → ab Sa} means that the left-hand side of the arrow is replaced by either one of the units on theright-handside, whereas XYZ→a means that Z is replaced by a. First and foremost, however, the brevity of Pānini's style is due to a principle called anuvrtti, and translated as 'dittoing' or 'rule ellipsis'. When (components of) rules that have been stated to apply at a given stage apply again at a later stage, this is often not stated explicitly, but must be understood from the context. Sometimes there are several such ellipses one inside another, and the listener or the reader of Astādhyāyf must be able to 'carry them along', as he moves forwards. In the following, I shall expose the structure of the rules 1.4.56 -62, which deal with the definition of particles. The actual form of the rules is given on the left; on the right is how they are to be understood: A (The topic is) A  except in   (is A) except in  D D (is A except in C) E in F (D is) E in F also G (D is in F) also G also H also H (is G in F) The intended anuvrtti relations may sometimes be difficult to reconstruct. But this is not Pānini's fault. He indicated at least part of the anuvrtti relations by a technical use of therising-and-falling(= svāritα) accent (1.3.11), and this, like the other accents and the nasalizations, was lost when his grammar was finally written down. To give a more complex example discussed also by Joshi & Roodbergen (1980:59), let us consider the anuvrtti in the rules 2.3.28-36, which deal with the occurrence of certain case endings, as conditioned either by meanings to be expressed or by the lexico-grammatical context. Case endings will be mentioned here explicitly, as in the original formulation. When new alternatives are men­ tioned, optionality becomes explicit. Lack of mention means either anuvrtti (= ' + ') or blocking of anuvrtti (= '-'). Whether the anuvrtti applies only to one case ending or to several alternative case endings simultaneously, is not indicated here

20

UNIVERSAL HISTORY OF LINGUISTICS

by any special symbolism, but is shown directly by the constellation of endings, pluses and minuses.

For instance, in the rule 2.3.35 the endings of accusative, instrumental, and ablative are equally possible. The downward arrows indicate discontinuous anuvrtti ('frog's leap'), which is rightly felt to be problematical. (To be sure, the 'leap' of ablative from 29 to 32 is justified insofar as the topic is the ablative ending, with the standard meaning 'movement away', and this meaning is exceptionally, and parenthetically, expressed by the genitive in 30 and the accusative in 31.) Although anuvrtti varies according to the context, there is at least the following recurrent pattern: Asserting the general case; mentioning an alternative to it; mentioning an exception to the general case; mentioning an exception to the exception. Or more explicitly: 1. 2. 3. 4.

A A or B, if  not-A, ifD not (not-A, if D), if D and E

How this semantic content is formalized, and abbreviated, may again depend on the context. In the next sub-subsection, when I discuss the order of rules, there will be reason to return to the notion of anuvrtti, and in particular to its justification. It might seem impossible to shorten Pänini's rules any further. Yet the Buddhist Candragomin (c. 400 A.D.) succeeded in this to some extent. His

INDIA

21

innovations may seem trifling to us, but at least they bear witness to the utmost seriousness with which Indian grammarians endorsed the ideal of formal brevity/ simplicity. This attitude is captured perfectly in the following 'interpretation rule' (= paribhāsa), the last of the 122 collected and explained by Nāgeśa: "Gram­ marians rejoice over the saving of even the length of half a short vowel as much as over the birth of a son" (Kielhorn 1960 [1874]:526). One does not find this paribhāsa in Pänini's grammar — because, among other things, it is much too long! — but it does express the descriptive ideal that inspires the entire Pāninian tra­ dition.8 Is it just because of the 'exotic' nature of Pänini's rules that they are apt to produce in the listener/reader an artistic experience which is seldom, if ever, produced by grammatical works in the Western tradition? I do not think so. Pänini's style has an aphoristic trenchancy, a feature proper to spoken language, that just cannot be found either in the rather rambling presentation of traditional Western linguistics or in the figures and formulas of modern linguistics. Thus I agree with Buiskool (1939:25): ...the constant application of such means as anuvrtti and the like, lends a certain logical elegance to Pänini's style, owing to the author's artistry. Along with an excessive veneration for word-memory, only a great love for his subject could enable the author to create his remarkable, nay fascinating technical language. C) Order of Rules In the sub-subsections A) and B) we have seen that factors which make Pänini's text so difficult can be fully justified. The use of determinatives, conventional labels, and anuvrtti serves the end of achieving a maximally general, concise, and elegant description. With the order of rules, however, the situation is different. One major obstacle to understanding Pānini is constituted by the fact that his rules often seem to be arranged in an arbitrary order. And unlike in the previous cases, it has not been possible until today to explain this difficulty away. The general structure of Astādhyāyfis reasonably clear. Chapter I deals with the 'metarules' (i.e. definitions and interpretation rules), various aspects of verb morphology, and the 'semantic roles'. Chapter II deals with the different types of nominal compounds and with the meanings expressed by case endings. Chapters III, IV and V deal with the general notion of affix, i.e. verb morphology and noun morphology (including the noun formation with the aid of feminine affixes as well

UNIVERSAL HISTORY OF LINGUISTICS

22

as 'primary' and 'secondary' nominal affixes). The remaining chapters VI, VII, and VIII deal with phonology, including (combinatory) accent and sandhi. However, there are several deviations from this general structure, i.e. groups of rules which seem to show up at wrong places. Yet it is not possible that Pänini would just have made a series of mistakes here. If he so wished, he was able to put the rules in a strictly logical order, as is shown by the Tripādi, or the three last quarter chapters of his grammar (cf. p. 26). Therefore it might be that the tradition has mixed up the rules at some point. Bhattoji-dîksita has followed up this suggestion and rearranged Pānini's rules in a new order which many have indeed found more illuminating. Yet it is far from clear whether Bhattoji-dîksita's efforts represent genuine progress because, for one thing, his new version of Pānini's grammar cannot be understood without previous knowledge of the original version (cf. Scharfe 1977:174-175). Therefore it is also possible that the present order of rules is the one intended by Pänini. One and the same word may belong to several topics, and therefore 'saving' this word in conformity with anuvrtti may force the grammarian to change the topic in a seemingly illogical way. It is thus possible that when the principles of logical order and of anuvrtti-type economy came into conflict, Pänini deliberately chose the latter. To conclude this subsection on the difficulty of Pänini's style, I shall illustrate the topic by quoting and translating the first five rules of AstādhyāyT. The meaning, and the justification, of the translation will become clearer as we proceed. 1.1.1. 1.1.2. 1.1.3. 1.1.4. 1.1.5. 1. 2. 3. 4. 5.

vrddhirādaic ( = vrddhih āT-aiQ adeñgunah ( = -eÑ gunah) iko gunavrddhr ( = iKah gunavrddhf) na dhātulopa ārdhadhātuke khiti ca ( = K-N-iti cd)

The sounds ā, αi, au are called 'vrddhi'. The sounds a, ey  are called 'guna'. The guna and the vrddhi replace the sounds i, u, r, l. Not before a 'half-stem'9 affix conditioning the deletion of part of the root. Nor before an affix having the determinative  or N.

INDIA

23

2.3.3 The Types of Rules Employed by Punini Before immersing ourselves in the workings of Pānini's grammar, it is good to stop and explain the types of rules he employs. This is only the more advisable, because grasping the nature of the rules eo ipso enables one to grasp the general 'move­ ment' of the grammar. The main division is between metagrammatical rules and grammatical (or 'operational') rules. A) Metagrammatical Rules Metagrammatical rules (a term without a Sanskrit equivalent) are 'second-order' rules, i.e. rules dealing with 'first-order' rules. They either define unitsfiguringin first-order rules or explain how the latter should be understood. Accordingly, metagrammatical rules are divided into definitions and interpretation rules. i) Definitions A Pāninian definition (= samjfiā-sutrd) may be either stipulative or extensional in character. A stipulative definition consists in replacing an expression, or definiendum (= samjñiri), by another term, or definiens (= samjñä), which is thought to be more unambiguous. Stipulative definitions may further be subdivided de­ pending on whether the definiens is a technical term or a conventional label. In the present context 'technical term' means a word which originally belongs to ordinary language, but is secondarily given a new, technical meaning. (Thus, in the present context 'technical term' is itself a technical term.) The meaning of 'conventional label' has already been explained in 2.3.2-A. I shall first give examples of stipulative definitions containing technical terms: 1.1.60: 1.1.9:

1.2.29:

adarśanam lopah = 'Not-being-seen [i.e. being de­ leted] [is called] 'lopa'.' tulyāsyaprayatnamsavarnam = '[Sounds that are produced with] a similar effort in the mouth [are called] 'savarna'.' uccairudāttah = '[Vowels produced with a] high [accent are called] 'udātta'.'

24

UNIVERSAL HISTORY OF LINGUISTICS

We have already encountered stipulative definitions containing conventional labels, namely the rules 1.1.61 and 1.1.64, which introduce the conventional labels LU and 27, respectively. An extensional definition consists in assigning a name to a category whose members have been, or will be, enumerated explicitly. Again, the name may be either a technical term or a conventional label. The technical term contained in one principal type of extensional definition is called adhikāra (='heading'). For instance, the rule 3.1.1. says simply pratyayah (= 'affix'). This heading is valid until the end of chapter V, which means that, in conformity with anuvrtti, the units discussed in the chapters III, IV, and V qualify as affixes. This entire sequence contains in turn subheadings such as krt, i.e. 'primary nominal suffix' (3.1.93), and taddhita, i.e. 'secondary nominal suffix' (4.1.76), with the understanding that units discussed subsequently fall under these two categories. A slightly different type of extensional definition is represented by the so-called dhātu-pātha, i.e. a list which contains all the roots of Sanskrit (= a little less than 2,000 in all), and which is appended to Pānini's grammar. It is in reference to this list that the rule 1.3.1 bhüvādayodhātavahintroduces the term 'root': 'The units thefirstof which [in the list] is bhū- are called 'roots'.' Finally, there are the extensional definitions involving conventional labels, like the rule 3.4.77 lasya ('Instead of L'), which is then followed by a list of the basic verb endings. To give another example, CLI is a general marker of the aorist and thus represents the affixes siC, Ksa, aÑ, CaÑ, and CzW (cf. 3.1.43-59). With hindsight, it is easy to see that pratyähära symbols too constitute in fact extensional definitions; thus aC and aL represent, respectively, all vowels and all sounds. The preceding discussion can be summed up as follows: technical term

conventional label

stipulative definition

lopa

LU

extensional definition

pratyaya

L

INDIA

25

To make matters more complicated, it is possible for one and the same definition to combine stipulative and extensional aspects. Consider again the first rule of the grammar, or 1.1.1 vrddhirudaic. On the one hand, it stipulates the name 'vrddhi' for āT-aiC. On the other, this latter symbol represents the set of ā, ai, and au. Before concluding this sub-sub-subsection, it is time to admit that 'defi­ nition' is in fact a cover term for two directly opposite operations. This can be made more explicit as follows: I) giving a name A to some X (where A is shorter than X) definition
. This rule says that à is lengthened if it is separated from the determinatives Ñ or N by one sound only. (This lengthening is here the 'operation involving something preceding' mentioned in 1.1.57.) Now the result of 6.4.48 might seem to meet the condition of 7.2.116. This is not so, however, because the zero substitute retains the function of its substituendum {a2 having been deleted by lopa, and not by LU). Therefore a1 is separated from N by two 'sounds', i.e. th and 0, and 7.2.116 does not apply. One gets kathi, not *kāthi. For my present purpose, the important thing is that when one applies 1.1.57 to block 7.2.116, one has to remember the preceding stage in the derivation, namely that 0 replaced à because MC followed. At the same time, however, the above example illustrates well Pänini's degree of sophistication. Finally, I shall give an example of cases where Pänini's rules seem to anticipate the result of an operation which is yet to be performed. This type of situation seems to be common in nominal compounding: 2.1.22, 30, 32-33: trttyā [= A] kartrkarane [= X] krtyair [= B] adhikārthavacane[= Y] tatpurusah [= ]: '[A noun in the] in­ strumental [= A] [employed] in [the sense of] agent or instrument [= X] [forms] a determinative compound [= C] with [words having] the affix of obligation [= B] [to express] the meaning 'extreme' [= Y]. The gist of this group of rules can be expressed as follows: The meaning Y of a compound noun  (= A-B), which is the result of the compounding operation, is a condition on this very operation, i.e. of combining A and  (of which A has the meaning X). Here the operation can be performed only if one already anticipates its result or performs it mentally, as it were, before performing it in fact. It would be wrong to infer from these prima facie violations of the principle of axiomatics that Pänini has committed a mistake. First of all, some inference rules of so-called natural deduction also require either remembering or anticipating. Such inference rules as reductio ad absurdum or conditional proof make use of temporary premises which must be demoted to the role of non-premises once they have discharged their duty. So, while operating with them, one must remember that they are not genuine premises. Or consider the inference rules of existential instantiation {—EI) and universal generalization (= UG):

INDIA

43

El is possible only if one anticipates that Fa' will in due time be replaced by '3xFx'. UG is possible only if one remembers that 'Fa' was derived from 'VxFx'. In sum, modern formal logic too accepts the use of remembering and anticipating.16 Moreover, it is a philosophical commonplace that a human being cannot live in the present moment, but rather reaches both forwards and backwards in time. In fact, these attitudes of 'protention' (= anticipating) and 'retention' (= remembering) have been considered as the defining attributes of a human being. In light of this, it should be to the credit of a descriptive technique that it at least to some extent reproduces these fundamental human attitudes. By contrast, axiomatics turns out to be an 'inhuman', machine-like ideal of description. Pänini's grammar exhibits both aspects. So much is clear, in any case, that Joshi & Roodbergen's (1975:xvii) view of Astādhyāyt as a 'machine' needs qualification. In CUT, esp. 284-286,1 distinguished between two different approaches to language description: the axiomatic approach aims solely at a maximally simple and general description of a language, whereas the psychologistic approach tries to capture the 'messiness' (i.e. the lack of simplicity and generality) that is de facto characteristic of the human psychology of language. There can be no doubt at all that Pānini opts for the former alternative (as I already indicated on pp. 8-9, 261 of CLT). This is evident e.g. from the way he derives non-basic case endings from basic ones (cf. p. 38). Besides, it hardly makes sense to assume that thefixedorder of the 295 rules in the Tripādi (i.e. 8.2-4) has some kind of psychological equivalent. In conclusion I return to the idea of the grammar 'moving' from meaning to sound. It should be clearly understood that this idea holds only approximately. In particular, it is not possible to assume that all expression rules precede all combination rules, as is shown by the fact that the former may be used as con­ ditions on the latter. If taken approximately, however, the idea of 'movement' does hold good, I think. Additional support for this view might be found in exploring the

UNIVERSAL HISTORY OF LINGUISTICS

44

notion of semantic role (= kãrakà). It is interesting to note that the kārakas do not yet constitute the starting point of the (derivational) movement. The quarter chapter 1.4 contains six prima facie definitions involving kārakas, for instance: 1.4.49: 1.4.32:

kartur Tpsitatamam karma = 'What the agent wishes most to obtain is karman [= 'patient'].' karmana yam abhipraiti sa sampradânam = 'He whom one tries to connect with the patient is sampradāna [= 'recipient'].'

There has been some uncertainty in deciding the relation of, e.g., kartur Tpsitatamam and karma to each other. Which one is the definiens, and which one is the definiendum? If the former were the definiens, the kāraka-definitions would be the only intensional definitions that occur in AstādhyāyT.If, by contrast, the definiens is karma, we have the usual type of stipulative definition. Moreover, kartur Tpsitatamam would represent something 'more original' than the kārkas themselves. I find both of these suggestions plausible. Assuming, then, that kartur Tpsitatamam is the definiendum, what, pre­ cisely, is its status? Again, there is a lot of uncertainty on this issue. To put it briefly, I think kartur Tpsitatamam, karmana yam abhipraiti, and the four other such expressions represent what I shall call 'semantico-ontologieal categories'. That is, they represent the first term in the trichotomy 'reality—meaning—form' (re­ produced here as 'category—role—case'), or the real starting point of the 'move­ ment' of AstādhyāyT. This discussion will be continued in 2.3.5-A. In particular, I shall explain and justify the distinction between 'semantico-ontological' and 'ontological' tout court.

2.3.5 Pānini's Descriptive Technique: How It Works In this subsection I shall show in more detail than before how linguistic expressions are actually described in Pānini's grammar. Because 'kāraka' is the central semantic notion, I must start by elucidating it.

INDIA

45

A) The Notion of'Käraka' The kārakas have been enumerated here on p. 32. Pänini deals with them in 1.4 and 2.3. In 1.4 they are correlated with the corresponding 'semantico-ontological categories' (as argued here on p. 44), whereas in 2.3 those meanings, including the kärakas, are discussed which are expressed by the various case endings. In addition to the rules 1.4.49 and 32, given on p. 44, I shall mention two more cases of a category—kāraka correlation: 1.4.24: 1.4.42:

dhruvam ā 'pādānam = 'What remains firm when one goes away is apädäna [= 'source'].' sudhakatamam karanam = 'The most efficient means to achieve one's purpose is karana [= 'instrument'].'

There are several additions and/or exceptions to rules of this kind, for instance: 1.4.50:

1.4.36:

1.4.37:

1.4.38: 1.4.43:

1.4.25:

'What is related to the agent in the same way as the patient, although unwished, also qualifies as a patient [= karman].' 'In connection with the root sprh- [= to desire], what the agent wishes most to obtain is recipient [and not patient].' 'In connection with roots having the meaning krudh- ['to be irritated']..., the person against whom the hatred is directed is recipient. ' 'When krudh- etc. have a preverb, he is patient [and not recipient].' 'In connection with the root div- [= 'to play'], the most efficient means to achieve one's purpose is also patient [and not only instrument].' 'In connection with roots having the meaning bhf- [= 'to fear']..., the cause of the fear is source.'

It may be good to present these data more graphically. R, P, 7, and S stand, respectively, for recipient, patient, instrument, and source. The semantico-ontologi­ cal categories and their qualifications are represented by the numbers of the corresponding rules. The categories proper are connected to their standard kärakas

46

UNIVERSAL HISTORY OF LINGUISTICS

by continuous arrows, whereas the qualifications are indicated by dotted arrows. The relation between kärakas and case endings is discussed more fully immediately below. SemanticoOntology

Semantics

(= Category)

(= Role)

Form (=

Case)

Two important lessons can be inferred from this diagram. First, although case endings are never mentioned in this context, they in fact determine the identity both of kärakas and of categories. In all the examples given so far, the käraka in question is patient, recipient, source, or instrument if, and only if, it is expressed, respectively, by accusative, dative, ablative, or instrumental. That the criterion is not semantic in nature, is shown patently by the fact that the käraka connected with one and the same action (or verb meaning) varies, e.g., from instrument to patient or from recipient to patient, according as the corresponding case ending varies from instrumental to accusative or from dative to accusative, as a result either of free variation (cf. 1.4.43) or of a variation between Vn and Preverb-Vn (cf. 1.4.37 and 38). Yet it must be emphasized that the correspondence between kärakas and case endings is not necessarily one-to-one, but may also be one-to-many and many-toone (cf. the discussion of the quarter chapter 2.3 immediately below). Second, the exceptions and/or additions to the generally valid category— käraka correlations are conditioned by co-occurrence either with specific roots (combined or not with preverbs) or with specific meanings to be expressed by a small group of roots. Therefore the significance of these exceptions and/or additions

INDIA

47

does not seem to be overly great. Yet they do suffice to demonstrate the difference in principle between semantico-ontology and semantics. Now it is finally time to justify the term 'semantico-ontology'. Why can't we just say that expressions like dhruvam apûye (1.4.24) represent the ontological level? Do we really have to postulate a 'semantico-ontologicaT level distinct from ontology, on the one hand, and from semantics, on the other? The arguments for making this tripartite distinction are found in the com­ mentary tradition. I am referring to Kätyäyana's värttikas II, IV, and V on 1.4.23 and to Patañjali's corresponding bhāsyas 6, 7, 11, and 12 as well as to Kaiyata's comments recorded, together with the authors' own notes, in Joshi & Roodbergen (1975:16-23). What is at issue, is the analysis of the following sentences: grûmasya samīpād ugacchati (= 'He comes from the vicinity of the village') and vrksasya parnampatati (= 'A leaf of the tree falls'). Kätyäyana and the others point out that in the former case the source of coming is the village while in the latter the source of falling is the tree. So should not the village and the tree qualify as dhruvam apāye, in spite of the fact that in the two sentences the corresponding words are provided with the genitive ending? The commentators are unanimous in answering this question negatively. What is important, is not how things really are (or ontology, as we would say). What is important, is the speaker's intention (= vivaksâ) to view things in a certain way; and the intention is made explicit by the sentence. In the present case the speaker intends to view the village and the tree as exhibiting a connection (= sambandha) with the vicinity and the leaf, respectively, and therefore he chooses the genitive. — The all-importance of vivaksâ, as linguis­ tically expressed, is asserted also on several other occasions. Thus, 'semantico-ontology' must be distinguished from ontology tout court. But must it also be distinguished from semantics tout court? Again, the answer is 'Yes'. Since there can be between categories and kärakas both a many-to-one correspondence (cf. 1.4.49 and 50, or 1.4.32 and 37, or 1.4.42 and 25) and a oneto-many correspondence (cf. 1.4.49 and 36, or 1.4.37 and 38, or 1.4.43), cat­ egories and kärakas cannot be identical in the sense of falling together. Therefore semantico-ontology and semantics must be two different things. And even if they did fall together, they still would not be identical in the literal sense. Exemplifi­ cations of categories are things-as-viewed-and-expressed-so-and-so, i.e. referents-aslinguistically-determined. Indian linguists should, in my opinion, be credited with realizing the difference between referent and meaning (even if they mainly use the same term artha for both). In consequence, kärakas can only be semantic counter­ parts of ('semantico-ontologicaT) categories.

48

UNIVERSAL HISTORY OF LINGUISTICS

Again, I must pause to establish the implications of the preceding dis­ cussion. First, the ontological level plays no systematic role within the Päninian tradition (cf. also 2.4.2). Therefore the semantico-ontological level might be identified as the very first level from which Pānini's derivational system starts 'moving' towards concrete form, or sound. Yet many semantic entities have no clear-cut semantico-ontological equivalent. Therefore I shall, for simplicity, continue to regard semantics as the starting point of Päninian derivations. Second, this 'movement' takes place only theory-internally. Viewed theoryexternally, it is the other way around. It is the linguistic categories which have determined the identity of semantico-ontological categories (via semantic categories, or roles). Yet it must be emphasized that even if the things falling under a given semantico-ontological category have been grouped together on the basis of a linguistic criterion, they are still things, and not some linguistic entities. It is also important to see that since the Päninian tradition recognizes the existence of an ontological, non-linguistic level (if only to dismiss it subsequently), it does not advocate the view that 'language determines thought'. Third, the Päninian tradition contains no interesting version of the iconicity (or isomorphism) thesis, i.e. of the thesis that there is a structural similarity between language and reality. To be interesting, the thesis requires that the struc­ ture of language and that of reality (more precisely: of reality-as-conceived) have been identified independently of each other, before being compared with each other. By disregarding the ontological level, the Päninian tradition makes it impossible for itself to meet this requirement. Next I shall proceed to discuss the quarter chapter 2.3 and its implications. The principal purpose is to enumerate the most important (grammatical) meanings that are expressed by the seven case endings (= sUP). On the whole, the presen­ tation is oddly meandering, with many remarks that seem to be out of place. In any case, much additional light is shed on the notion of kāraka. Recipient, instrument, source, and location are straightforward: with very few and insignificant exceptions, they are expressed by dative, instrumental, ablative, and locative, respectively. Agent and patient, by contrast, exhibit genuine variation. First, they are expressed differently depending on whether the sentence is active or passive (= nom. vs. instr., and acc. vs. nom.). Second, they may both be expressed by the genitive (i.e. as genetivus subiectivus or genetivus obiectivus). Third, the gatyarthakarman (= 'patient in the sense of movement', i.e. 'goal') may be expressed either by the accusative or by the dative. These facts may be summed up as follows:

INDIA

kartr (= agent)

înstr.

nom.

49

karman (= patient)

gen.

ace.

dat.

Thus, kārakas are the primary determinants of nominal inflection. While the primary use of other case endings is to express kāraka relationships in connection with a finite verb, the primary use of the genitive is to express a 'relation' (= sambandha) between two things. It is only in connection with verbal nouns that the genitive expresses agent or patient, as in odanasya pācakah (literally 'the cooker of gruel'). In verbal inflection kärakas play an equally central role. Remember that kartr ma karman mean not just 'agent' and 'patient', but also 'doing' and 'under­ going'. One might argue either that 'agent' and 'doing', on the one hand, and 'patient' and 'undergoing', on the other, are just one single concept, or that they are two (albeit closely related) concepts. For clarity, I choose the latter alternative. In addition to expressing such grammatical meanings as person, number, and tense/ mode, the verbal endings subsumed under tiÑ must also express the type (or 'aspect') of action, i.e. doing, undergoing, or happening (= bhāva). To this extent, the notion of käraka determines the selection of verbal endings. Moreover, most finite verb forms contain an infix indicating either doing (e.g. ŚaP) or undergoing/ happening (yaK). Thus, the type of action is expressed by discontinuous morphs (e.g. SaP...tiP) whereas person, number, tense/mood, and (part of) type of action is expressed by portmanteau morphs (e.g. tiP). Also in word-formation the notion of käraka plays a crucial role. Up to now I have been dealing with what might be called 'sentence' kärakas: they are those meaning elements that are expressed by members of sUP or tiÑ, and it is only natural that in different sentences (tokens of) one and the same word may assume different käraka roles. (In the two sentences The man teaches a woman and A woman teaches the man, the man represents first an agent and then a patient.) In connection with word-formation, however, we encounter 'lexical' kärakas, i.e. kärakas expressed by derivational affixes. It is their characteristic feature that they remain the same independently of the sentence in which they happen to occur. At least the following three subtypes, formed with the aid of krt-affixes, have to be distinguished: a) The affixes which are used to form verbal adjectives express a constant (käraka) relation between the action expressed by the root and the thing expressed

50

UNIVERSAL HISTORY OF LINGUISTICS

by the head noun. For instance, in the expression snānīyam curnam (= 'bathing powder'), the affix antyaR, added to the root snā-, indicates that the action takes the thing as an instrument, as in the sentence cūrnena(instr.) snāti ('He bathes with the help of powder'). The instrument relation remains the same irrespective of the case in which the two words happen to be. For instance, in the sentence snāntyam cürnam sprśati ('He touches bathing powder'), there is a sentence-level patient relation superimposed on a lexical instrument relation. To give another example of the same type, in the expression prapatanah parvatah ('mountain from which one falls down'), the affix ana (technically, LYUT), added to the root pat-, which is itself preceded by the preverb pra-, indicates that the action takes the thing as a source, as in the sentence parvatāt (abl.) patati ('He falls down from the mountain'). b) While verbal adjectives express an action—kāraka relation, compound nouns may express in a compressed form the relation between an action and two kärakas. Typically, one of these is the agent. For instance, to the sentence kumbham karoti ('He makes a pot') there corresponds the compound noun kumbhakārah ('pot-maker'), in which the latter part is formed by adding the krt-afñx aN to the root kr- and replacing the original root vowel r by the vrddhi ā(r). The agent—action—patient relation of the sentence remains the same in the compound noun, irrespective of the case in which the latter happens to be. c) Finally, there is the case of single nouns expressing both an action and a käraka. As a rule, a krt-afñx expresses an agent (3.4.67). For instance, the affixes aka (technically, NVUL) and trC, added to the root kr-, produce the two central terms käraka and kartr, both meaning 'doer' or 'maker'. (By contrast, the latter part of kumbha-kāra,formed with the aid of aN, cannot stand alone.) Other kärakas too, however, may be expressed. For instance, added to the root ni- ('to lead'), the affix tra (technically, Stra) expresses the instrument relation and produces, with the guna of the original vowel, the word netra ('eye' etc.). B) How to Describe a Simple Sentence Containing Simple Words: an Outline As explained, the derivation starts at the semantic level, with the action as the central concept. It has to assume one of three possible aspects. There is a number of kāraka-participants connected with the action, ranging from zero to six. The grammatical forms connected with the action are channelled through L and selected from among the members of tiÑ (not to mention such infixes as ŚaP and yaK supplied by separate expression rules), whereas the grammatical forms connected with kärakas are selected from among the members of sUP. The relevant (ex-

INDIA

51

pression) rules have been mentioned on pp. 32-33. The basic facts can be summed up as follows:

parasmaipadam

yaK + ātmane-

yaK + âtmane-

(active)/ätmane-

padam (personal

padam (impersonal

padam (medial)

passive)

passive)

nominative

instrumental

accusative

nominative

kartr2 (agent)

(instrumental)

karman2 (patient)

The entities outside the rectangle are units of meaning whereas those inside are units of form. This figure makes it clear that there is no derivational relation­ ship between active and passive. Rather, the two are just different and equivalent ways of expressing the same semantic content. Kiparsky (1982:10-11) has pointed out that there is a parallel set of rules for deriving krt-nominalizations, which constitute the basis for nominal sentences, i.e. sentences with no finite verb. This means that there is a fundamental equival­ ence not just between active and passive, but also between the following sentencetypes: devadattah pacati odanam ('Devadatta is cooking gruel') and devadattah pucakah odanasya ('Devadatta [is] a cooker of gruel'). Next, let us be a little more specific. Consider the sentence bālāh aśvam grāmāya nayanti ('The children bring the horse to the village'), presented here in a form prior to the operation of sandhi rules. The structure of the sentence could be clarified as follows:

52

UNIVERSAL HISTORY OF LINGUISTICS

The action of bringing determines the nature of its kāraka-participants from which the speaker is free to select those he intends to express. The facts of seman­ tic determination are simply assumed to be known, in accordance with the principle that meaning need not be taught (cf. p. 31). Horizontal arrows indicate determi­ nation whereas vertical arrows are identical with expression rules. Apart from agent and patient, to each (sentence) kāraka there corresponds a single case ending, with the qualification that the goal (or 'patient of movement') may be expressed either by accusative or by dative. Horizontal arrows issuing from 'doing' indicate that it triggers the application of one pair of expression rules to 'agent' and 'patient', whereas 'undergoing' would have triggered the application of another. More precisely, this is achieved by means of the anabhihite-device (cf. p. 29): Kartr (i.e. 'doing') is expressed by a-nti and, additionally (i.e. qua 'agent') by the 'neutral' case, i.e. nominative. Since karman has not yet been expressed in any other way (i.e. it has not been expressed as 'undergoing' by yaK+te), it will be expressed (as 'patient') by the accusative. Syntactic determination (i.e. 'governance') is just an after-effect of previous semantic determinations. The semantic level is here so abstract as to omit the explicit mention of tense/mode, person, and number. Selecting 'undergoing', instead of 'doing', would produce the sentence aśvah bālaih grāmāya ntyate ('The horse is brought to the village by the children'). More precisely, because karman (as 'undergoing') has already been expressed by yaK+te, it (as 'patient') must not be expressed also by the accusative. Therefore its 'patient' aspect is expressed by the neutral case ending, i.e. nominative. And because kartr has not yet been expressed at all, it will be expressed by the instrumental.

INDIA

53

The above diagram makes it clear that Pānini needs no separate level for grammatical functions like 'subject' or 'object'. The reason for this must lie in the fact that his semantic roles are so 'shallow' or close to morphological forms (cf. p. 46). Although the grand design of sentence-derivation is quite clear, there are some finer points on which commentators disagree. In particular, it is uncertain which rules have to be applied in precisely which order, to produce a description that is both exhaustive and self-consistent. To give some notion of the problems involved, I shall present here part of the tangled meaning—form relations that have to be accounted for:

This diagram calls for several comments. Arrows and continuous lines represent expression rules and substitution rules, respectively. Dotted lines rep­ resent relations which are implied, but not stated directly. Kta and krtya are (groups of) affixes providing for additional means for passive-formation (3.4.70). As has already been mentioned several times, the three aspects of action are expressed by L (3.4.69), which is then replaced by tiÑ (3.4.78). But there are also rules relating doing to the active (1.3.78) or to the medial (1.3.72), and relating both undergoing and happening to the passive, i.e. medial endings (1.3.13) plus yaK (3.1.67), as well as a pair of rules (1.4.99-100) which define L as a combination of active and medial endings. From this it follows that when e.g. nayanti ('they bring') has to be derived from 'n & doing', one in a sense loses information, if one passes through L (because, to repeat, L is common both to all three action aspects). Therefore it would seem preferable to pass through 'active' (parasmaipadam), while retaining

54

UNIVERSAL HISTORY OF LINGUISTICS

the information that it belongs to L. The principle is clear, but it is less clear how it should be formalized. The description of tense presents another problem. L is divided into ten subcategories, starting with LAT and LAN. There are rules relating LAT to the present (3.2.123) and LAN to the 'past different from today' (3.2.111). However, if nayanti has to be derived from 'n& doing & present', LAT docs not, properly speaking, enter the derivation at all. As explained, tiÑ represents the set of basic endings, and there is a group of rules (3.4.79-112), containing members of L, for deriving the non-basic endings. There is a tacit understanding that what remains after all this filtering is LAT in the active and LAN in the medial. Again, the principle is clear, but it could be formalized in a number of different ways. To be sure, there is no explicit prohibition against introducing LAT, as the (formal) substitute for 'present', into the derivation, but it is simply unclear how one should then continue. When L-categories other than LAT (in the active) or LAN (in the medial) are introduced into the derivation, they too do not function as substitutes; rather, they function as conditions triggering the application of rules belonging to 3.4.79-112; for instance tiP → NaL/LIT(='in the perfect tense, the ending of the third person singular is not tiP but NaU). An additional problem is created by the rule 3.2.124 stating that LA be replaced by the participle affixes ŚatR/ŚānaC. This creates a prima facie ruleconflict, if it is assumed that LAT must be introduced into the derivation as a specification of 3.4.69, i.e. as a substituendum to be replaced by tiÑ. Now, since 3.2.124 is an exception to the general case 3.4.69, it should prevail over the latter. This seems, however, to lead into the impossible situation where 3.4.69 will never apply. This difficulty has been taken very seriously, as shown by Kätyäyana's värttika VIII and Patañjali's bhäsyas 34 and 35 on 2.3.46 (Joshi & Roodbergen 1981:42-52, with further comments). Fully admitting the ingenuity of the proposed solution (which I leave out here), I would like to point out that the problem arises only if LAT is allowed to occur as the substituendum of tiÑ. I have given above some reasons for not allowing this, and solving the rule-conflict at hand might be adduced as a further reason. This does not yet solve the problem, however. It is still the case that at the level of semantic representation there may be two instances of '(doing/undergoing &) present' of which one should ultimately lead to (a member of) tiÑ while the other should lead first to LAT and then to SatR/ŚānaC, as in the sentence bulampacaraam (= ŚatR) śrnoti (= tiÑ) ('he hears that the child is cooking'). How are these two cases to be kept apart?

INDIA

55

This should not be overly difficult. The rule 3.2.124 actually states that ŚatR/Śunac must not be coreferential with the subject (= nominative) of the sentence; so referential indices must be used to distinguish between the two agents involved. Moreover, the rule 3.4.1: dhātusambandhe pratyayah says that affixes (like SatR) express relations to (actions expressed by) roots. When spelled out, this means that e.g. in our previous sentence the LAT (> SatR) connected with pacantam does not, properly speaking, express 'present tense', but rather 'con­ temporaneity' with the main action expressed by śru-, irrespective of whether this action itself is present or past. By contrast, the LATconnected with śrnoti genuinely expresses 'present tense'. One only needs to make the corresponding distinction at the semantic level. Now I have completed my account of how a simple sentence containing simple words is to be derived, in outline. For the moment, we remain at the level of 'abstract form' (cf. the figure on p. 39). Next I have to show how complex sentences and complex words are derived. I shall conclude by showing how abstract form is converted into concrete form. C) How to Describe a (Semantically) Complex Sentence: an Outline It is only in passing that Pānini deals with complex sentences, for instance with conditional sentences in 3.3.156. He is fully aware, however, of the existence of semantic complexity. This is evident from the rules indicating that one sentence may express several actions. Thus, the absolutive suffix Ktvā is added to a root to express anteriority when the two actions have the same agent (3.4.21), as in aśvam nitvā nivartate ('Having brought the horse, he goes back'). Similarly, the infinitive suffix twnUN is added to a root to express the future (or posteriority) when the main verb expresses an action performed for the sake of another action (3.3.10), as in asvam nitwn nivartate ('He goes back to bring the horse'). Occasionally the infinitive construction may be replaced by a simpler construction where the patient-accusative of the infinitive is in the dative and the infinitive itself is lacking (2.3.14). Thus there is a structural synonymy between edhān āhartwn vrajati ('He goes for the sake of gettingfirewood')and edhebyo vrajati ('He goes for the sake of firewood'). In the preceding examples the two actions have the same agent. The agents are different in participle constructions with SatR and SānaCusually (3.2.124 -126) and in locativus absolutus constructions always (2.3.37). There are two agents also in causative constructions. When the primary kartr has 'already' been expressed by a finite verb ending (plus nominative), the secondary kartr is expressed, in accord-

56

UNIVERSAL HISTORY OF LINGUISTICS

ance with the anabhihite principle, by an instrumental ending, as in aśvam bālair (= instr.) nāyayati ('He makes the children bring the horse'). Given that the causative affix is MC, there are two rules that explicitly correlate simple or a-NiC (= 'not-MC) sentence-types and causative or NiC sentence-types. In one case (1.4.52) it is stated that what in a simple sentence is the agent (expressed by nominative) is in a causative sentence the patient (expressed by accusative), and not the secondary agent (expressed by instrumental), as the general rule requires. In the other case (1.3.67), which deals with a somewhat peculiar use of the medial endings, it is stated that what is the patient in the simple sentence is the agent in the causative sentence. In their respective translations both Böhtlingk (1887) and Renou (1966) assign a 'transformational' interpretation to the causative construction. They translate Pänini as saying that what in a simple sentence 'was' X 'becomes' Fin the causative sentence. Interestingly enough, however, Pänini's original text lends no support to such an interpretation. Here as elsewhere, Pänini's rules contain no finite verbs, so there is no question of one sentence-type 'having been' something the other 'is' not, or of something 'becoming' something else. (Incidentally, this is in agreement with the 'non-change' interpretation required by Pänini's substitution rules; cf. p. 26). In just the same way, there is nothing 'transformational' in Pänini's way to conceptualize the active—passive relation (cf. p. 51). He is content to state what this relation is, with no intention to derive one construction from the other. To sum up, the relation between simple and causative constructions may be represented as follows:

Just as there are no transformations, there are no 'deep structures' either in Pānini's syntax. In an infinitive construction of the type 'He comes to eat', for

INDIA

57

instance, it is stated that there are two actions with a single agent (cf. above). However, there is no indication that one should postulate a level of 'abstract form' which would somehow spell out this twofold occurrence of the agent. Of course, there is the level of meaning, and the identity or non-identity of the agents must be represented at this level; but this is a different matter. Besides, there is no good reason to regard semantics as 'more abstract' than syntax. The most one can justifiably say is that semantics is different from syntax (just like the function of a tool is different from its form). On the other hand, there is no alternative to saying that at subsentential levels, from compounds 'downwards', Pänini operates with 'deep' or 'abstract' forms (cf. below). Thus there is no reason to regard complex sentences as being derived from simple ones. Given that Pänini starts from semantic structures, this means that even complex semantic structures must be non-derived or 'given'. This applies equally to structures containing others and to structures combined with others. In particular, it is not right to derive complex structures by any kind of recursive rules. Now what is true of semantic (or conceptual) structures is also true of syntactic structures understood as part of an abstract linguistic system. In fact, it seems to me that the nature of recursive rules has been generally misunderstood; and Pänini's approach is apt to highlight this misunderstanding. An abstract system possesses no temporal aspects. Within it, different structures exist (in a non-local sense) 'side by side', as mere possibilities, not 'one after another'. Accordingly, the relation of simple (syntactic) structures to (increasingly) complex structures is this: S S + S

s +s +s S + S+ S+ S

It would be wrong to describe these structures in the standard recursive way S -* ...S..., because this would mean that one generates first the first line, then the second line, and then... etc. Rather, to repeat, all the lines are given simultaneously (in a non-temporal sense). In other words, the recursive rules are unwittingly committed to the speaker's (temporal) point of view. They might still be retained, but only if it is clearly understood that they are an inadequate tool for representing language as an abstract system.

58

UNIVERSAL HISTORY OF LINGUISTICS

What was just said applies equally to recursive rules qua tools for rep­ resenting mathematical structures (instead of computational processes). It is only within a deliberately constructivist framework that they could be found acceptable. Finally, it may be good to point out that expression rules, combination rules, and substitution rules, as here defined, are not open to the same kind of criticism that was just directed against recursive rules. Unlike the latter, they do not 'create something out of nothing', with the implication that first there is this, and then that. Rather, they just show how the units and the levels of the linguistic system are related to one another. Accordingly, it goes without saying that e.g. the talk about the 'movement' of the grammar must be taken figuratively, or intending to say that what 'moves' is not the grammar, but the grammarian's attention. D) How to Describe a Complex Word: an Outline Word-formation takes up about two fifths of Pānini's grammar. A noun minus its case ending is called 'prātipadika' (1.2.45). The simplest type of prātipadika contains no derivational affix, for instance aśva ('horse'). Next comes a type formed by adding a primary or krt derivational affix to a root, for instance kr+k → kā or +tra -* netra (cf. p. 50). A secondary or taddhita derivational affix may be added to prātipadikas of either the first or the second type to form prātipadikas of the third type. The number of taddhita affixes is astonishing. Finally, compound nouns constitute the fourth type of pratipadika (1.2.46). For simplicity, I shall not discuss word-formation with krt or taddhita affixes. (Similarly, I shall pass over the formation of complex verbs, dealt with e.g. in 3.1.8-21.) Instead, I shall concentrate on nominal compounding. We have already seen on p. 50 that if the superordinate member of a compound has been formed by adding a krt affix to a root, the subordinate member stands in a kāraka relation to the action expressed by the root. For instance, in kumbha-kāra ('potmaker') kumbha ('pot') is a patient vis-à-vis kr- ('to make'). The subordinate member is called upa-pada ('by-word'), and it is recognized by the fact that in the relevant group of rules (= 3.2.1-101) the name of the corresponding karaka is in the locative (3.1.92). The initial (abstract) representation of kumbhakurah is as follows:17

INDIA

59

Deleting the accusative ending and the auxiliary symbols plus assigning the vrddhi degree to the root and applying one sandhi rule produces the surface form kumbhakārah. Compound nouns of the above type are dealt with as part of the krt af­ fixation, which covers most of the chapter III. Compound nouns properly so-called are treated of elsewhere, i.e. in the two first quarter chapters and in part of the last quarter chapter of the chapter II. The principal types of nominal compounds are the following: i) indeclinable compounds (avyayibhuva) ii) determinative compounds (tatpurusa) a) dependent (vibhakti-tatpurusa)1* ß) descriptive-coreferential (karmadhärayä) iii) copulative compounds (dvandva) iv) possessive-exocentric compounds (bahuvrthi) Indeclinable and copulative compounds seem to possess the least amount of theoretical interest, and I shall not discuss them further. With a lot of ingenuity, it is possible to put together the following definition of 'compound' (samāsd) on the basis of the rules 2.1.1-4: samarthah sup[ah] saha supā samāsah = 'A word with a case ending forms a compound noun with another word with a case ending, if it is semantically connected with the latter.' When read carefully, this definition turns out to have (at least) the following implications. First, Sanskrit compounds may contain several members; but since

60

UNIVERSAL HISTORY OF LINGUISTICS

saha supā ('with a case-ending word') is in the singular, the compounding takes places two at a time, which means that compounds have a repetitive binary struc­ ture. Second, the (binary) compound as such contains two case endings (= supah saha supā)\ but since, according to 1.2.46, a compound qualifies as a prātipadika, i.e. nominal base, a third case ending has to be added at the end to produce the abstract form of an actual word; and the two other endings must be deleted by LUK, according to 2.4.71. To illustrate, the word asvarathah ('horse-chariot') has the following abstract form, with the meaning 'chariot of horses':

All endings except the last will be deleted. This is true no matter how many members the compound in question contains. In dependent determinative compounds, and sometimes in possessiveexocentric compounds, the relation between the members is asymmetrical in the sense that one is superordinate and the other subordinate. The order of the members is indicated in the following, rather complicated way: The word whose case ending is mentioned in the nominative is upasarjana, i.e. the subordinate member (1.2.43), and it precedes the superordinate member (2.2.30). Thus, in our example, aáva-am is in the genitive, and the word 'genitive' is mentioned in the nominative (2.2.8: sasthf); therefore asva-ām is the subordinate member and precedes ratha-sU. If a compound contains no root-plus-krt combination, the (covert) case ending of the upasarjana does not express a kāraka relation. The most frequent such case ending is the genitive, as in asvarathah. All other case endings too, however, are represented, e.g. śalākayā (= instr.) kānah→Ealākākānah ('[made] blind with a splinter of wood') or caurād (= abl.) bhayam→caurabhayam ('fear of a thief). These are examples of dependent determinative compounds. In all descriptivecoreferential compounds and in most possessive-exocentric compounds the two members have identical (covert) case endings. Although the general principles governing the formation of compounds of the two last-mentioned types are reasonably clear, the question of adequate for­ malization remains disputed. According to Pānini (2.2.24), the referent of a possessive-exocentric compound is anyapadârtha, i.e. the referent (artha) of a word (pada) other (anya) than the words constituting the compound. Accordingly, the

INDIA

61

two overt members of this type of compound are equally subordinate, while the superordinate member either remains covert or is expressed by a separate word. This is Pānini's way to distinguish between a possessive-exocentric compound like n[la-kamalam ('that [e.g. a pond] which has a blue lotus') and a homonymous descriptive-coreferential compound nûa-kamalam ('blue lotus'). However, in a rare instance of disagreement with Pänini and of mutual agreement, Kātyāyana and Patañjali argue that a possessive-exocentric compound is, rather, anyavibhaktyartha, i.e. it expresses a case ending (vibhakti) other than the (covert) case endings of the words constituting the compound. In our example this 'primary' ending would be that of locative: '[a pond] in which there is a blue lotus'. Roodbergen (1974:xxxvxliii) points out, however, that in Pänini's derivational system there is no place for this kind of (sentence-size) abstract form. Instead, he proposes the following type of abstract form: [(nUa-gen.) + (kamala-gen.)] 4- nom. The justification is as follows: On the one hand, the 'other referent' aspect is accounted for, because a genitive necessarily refers to something beyond its own referent; on the other, the 'other case ending' aspect is trivially accounted for. By contrast, the abstract form of the corresponding descriptive-coreferential compound is straightforward: [(nUanom.) + (kamala-nom.)] + nom. There can be no doubt about the cleverness of Roodbergen's solution, but it entails an intuitive oddity. It is the whole point of possessive-exocentric com­ pounds that possessors are referred to by their possessions; and typically possessors are expressed by the genitive. In Roodbergen's solution, however, it is the other way around. A logical solution would be something like this: {[nila-nom.) + (kamala-nom.)] + 0-gen.} + nom. Here the implicit genitive expresses 'lexical possessor' (comparable to the lexical kārakas) whereas the outermost nominative expresses, in a perfectly regular way, the syntactic kāraka exemplified by this lexical (and merely implicit) pos­ sessor. Thus there is no need for deriving compounds from underlying sentences, which would indeed go against the spirit of Pāninian derivations. The last-mentioned point becomes clearer in a different context where Kätyäyana and Patañjali discuss the precise nature of abstract forms (Joshi & Roodbergen 1969:193-201). Again, the point of departure is the word asvarathah. We have already seen what is its canonical abstract form. But whoever knows the meaning of this word, also knows that the chariot is yoked (yuktah) with the horses. Therefore it might be suggested that the derivation (in a simplified form) should in reality be like this:

UNIVERSAL HISTORY OF LINGUISTICS

62

aavānām yuktah rathah aśvayuktah rathah aśvarathah Can this type of more abstract analysis, with the consequent lengthening of derivations, be accepted? After considering arguments pro and con, both Kätyäyana and Patañjali answer in the negative. Pāninian derivations are allowed to delete only grammatical material (like the genitive ending (n)ām), but not lexical material (like yuktah). Thefinaljustification for this view given by Patañjali (p. 200) is especially instructive: If we wish to put in the description of a construction, i.e. in its abstract form, all the information that it conveys to us, this process will never end, because — due to the functioning of associations — the amount of information conveyed by any given construction is in principle infinite. Therefore it is enough to combine aÉvunum and rathah, and to let the meaning element yuktah just be implicit in this combination.19 According to Pänini's rule 2.1.11 vibhāsā ('optionally'), compounds may be expressed equally well by corresponding analytical expressions.20 Accordingly, e.g., aÉvānām rathah and aâvarathah would be synonymous, and the latter would be, strictly speaking, superfluous. Patañjali shares this view, but Kätyäyana disagrees. He does not think that any compound is, strictly speaking, synonymous with the corresponding analytical expression. There is also a formal justification for this position, insofar as the abstract forms of the two types of construction must differ from each other: the last sU in the abstract form of the compound has no equivalent in the abstract form of the analytical expression (cf. above). These questions belong to a larger post-Pāninian discussion about the nature of compounding. To start with, Patañjali sums up the defining properties of compounds vis-à-vis analytical expressions as follows: i) ii) iii) iv) v) vi) vii)

Absence of case endings, apart from the last one The impossibility to add a new word between the members of a compound without a change of meaning Fixed order of the members A single accent Ambiguity about the grammatical number of the subordinate member Ambiguity about the syntactic relation between the members The impossibility of the subordinate member to be construed with a word outside the compound

INDIA

63

There were two basic positions. According to one, represented by Kätyäyana, compounds are non-generated (nitya), whereas according to the other, represented by Patañjali (who in this respect followed Pānini), compounds are generated (kārya, lit. 'to-be-made'). More precisely, Patañjali recognized the following alternatives in the interpretation of compounds:

The position  represents the way that analytical expressions are under­ stood, and thus implies that there is no distinction between compounds and analyti­ cal expressions. Kätyäyana accepts the position D (though with some reservations). Patañjali vacillates between the positions  and E. E) How to Describe a Simple Word: a Detailed Example Pānini's technical mastery is the most striking in morphophonology. Therefore it is befitting that when I now proceed to illustrate his descriptive technique in detail, it is from this field that I shall choose my example. More precisely, I shall show how the word-form náyanti ('they lead') is to be described. The description takes the form of a derivation which makes use of all the types of rules discussed in 2.3.3. The rules justifying each derivational step are mentioned explicitly. It is because of the anuvrtti principle that so many rules have to be mentioned: One single instruc­ tion is usually put together out of several elliptically formulated rules. 1. Action +

2. Root

+

Aspect21 + (=Doing/ Undergoing/ Happening) Doing +

Tense22 + (=Past/ Present/ Future) Present +

Number + (=Singular/ Dual/ Plural) Plural +

Person (=l./2./3.)

3.p.

64 3. Root+LAT

4. Root+tiN

5. Root+SaP+tiÑ 6. n1Ñ+SaP+tiÑ

7. nīÑ+SaP+JHi

8. nīN+SaP+anti 9. níÑ+SáP+ánti

10. níÑ+SaP+ánti 11. níÑ+SaP+anti

12. néÑ+SaP+anti

13. né+a+anti

UNIVERSAL HISTORY OF LINGUISTICS

3.2.123: LAT expresses Present [and is a member of the L-category]. 3.4.69: The L-category expresses Doing/Undergoing/Happening. 3.1.1-2, 3.4.77: The members of the L-category are affixes [Def. 3.1.1—] and occur after the root. 3.4.77-78: The members of the L-category = tiÑ. 3.4.113: Full-stem-affix = tiÑ and the units with the auxiliary symbol S. 3.1.1-2, 67-68: SaP occurs after the root and before a full-stem-affix to express Doing. 1.3.1: The roots are ... ntÑ ... 1.3.72: Ñ = With a root having this determinative the medial endings express that the action will be to the benefit of the agent. 1.3.78: Doing is expressed by the active endings [which are part of the ttW-endings]. 1.4.99, 101-102, 3.4.78: Active, Plural, 3. Person = JHi. [1.3.10: The items of two equally long lists correspond to each other in order.] 7.1.3: JH → ant. [JH→at in cases mentioned separ­ ately.] 6.1.162: The last syllable of a (primary = 1.3.1, or secondary = 3.1.32) root is accented. 3.1.3: The first syllable of an affix is accented [and SaP and anti are affixes; 3.1.1-]. 3.1.4: An affix with the auxiliary symbol P and the members of sUP are not accented. 6.1.186: A personal ending after [S]a is not accented; thus 6.1.58: A word [Def. 1.4.14] has only one accent. 6.4.1, 7.3.82, 84: In a stem [Def. 1.4.13] the middle degree of a vowel, or guna [Def. 1.1.1-2], occurs before a full-stem-affix or a half-stem affix [Def. 3.4.114]. [1.2.1, 4, 1.1.3-5: The rule is cancelled if P is lacking.] 1.3.3, 8-9: In the theoretical exposition any consonant following and an S preceding a unit are auxiliary symbols [Def. 1.3.2-8], and they are deleted.

INDIA

14. náy+a+anti

15. náy+0+anti

65

6.1.72, 77-78: In continuous speech [Def. 1.4.109] the substitution [Def. 1.1.49, 56, 66-67] eC [= e/o/ai/au] → ay/av/āy/āv/ takes place before a vowel [cf. rule 1.3.10 on line 7]. 6.1.84, 94, 96-97: A non-word-final short a, plus a following guna-vowel, is replaced by a single vowel, namely the latter one.

16. náyanti The line 1 of the derivation represents the meaning alternatives that are open to the speaker, and apart from the lexical meaning, the line 2 represents the result of the semantic specification necessitated by the communicative intention. (As Patañjali puts it, arthanimittaka eva sabdah = 'speech is of course brought about by its meaning'.) At these semantic levels the order of units is arbitrary. The line 3 states the meaning and the place of LAT, which means, in more precise terms, that the line contains one expression rule and one combination rule (supplemented by an extensional definition). The line 4 contains one (definition-like) substitution rule and one definition. The line 5 combines a combination rule and an expression rule. The line 6 contains one definition and one expression rule. The line 7 contains two expression rules (supplemented by an interpretation rule). The line 8 contains one substitution rule. The lines 9-11 contain accent rules, which may be considered as a subtype of substitution rules, except that the latter rule of the line 11 is an interpretation rule, i.e. a general condition on accent rules. The lines 12-15 contain substitution rules and metagrammatical rules. It is not possible to carry out this derivation strictly axiomatically, or in a point-by-point fashion. This follows from the morphological structure of Sanskrit: On the one hand, 'Doing' is expressed discontinuously both by SaP and by JHi; on the other, JHi alone also expresses 'Number' and 'Person'. As a consequence, the meaning units are replaced by formal units in a 'global' fashion. An agglutinative structure would be amenable to a more strictly axiomatic treatment. As explained on p. 54, it would have been more logical not to mention LAT at all (i.e. to leave out the rule 3.2.123 on the line 3) and to let the present-tense meaning of náyanti be expressed implicitly. In this example, however, I have preferred to follow the standard procedure. Notice, in any case, that if there had been just L, instead of LAT, there would have been no duplication on the lines 3 and 7, because the transition from 3.4.69 to 1.3.78 is a matter of meaning-specification.

66

UNIVERSAL HISTORY OF LINGUISTICS

To give at least some notion about the efficacy of the rules employed so far, I shall present in outline the derivation of the corresponding causative form nayáyanti ('they cause to lead'): nīN+NiC+SaP+anti

nīN+NiC+SaP+anti

naiN+NíC+SaP+anti nai+í+a+anti nāy+i+a+anti nāy+é+a+anti nāy+áy+a+anti nāy+áy+0+anti nāyáyanti

3.1.26, 1-2: MC expresses the Prompter [Def. 1.4.55] of an action done by someone else, and occurs after the root. 3.1.32: The affixes in the list headed by saN [and including MC], when added to roots, constitute [secondary] roots. [Cf. line 5 above: SaP occurs after a root.] [Cf. 6.1.162 on line 9 above. The effect is secured by] 6.1.159, 163: The last [or only] syllable of an affix with  is accented. 7.2.114-115: N brings about the full degree, or vrddhi [Def. 1.1.1, 3], of the preceding stem vowel. [Cf. line 13 above] [Cf. line 14 above] [Cf. line 12 above] [Cf. line 14 above] [Cf. line 15 above]

That generalizations have been achieved, is proved by the fact that the two forms náyanti and nāyáyantiare derived by a set of identical rules (apart from the rules relating to the causative affix MÇ). The two derivations may not yet reveal the usefulness of the rule on line 15. It is needed to derive e.g. the medial form naya+e → naye ('I lead for my own benefit'). Next I shall present in outline the derivation of the imperfect form ánayan ('they led'): áT+nīÑ+SaP+LAN

á+nay+a+LAN á+nay+a+JHi á+nay+a+JH á+nay+a+ant

3.2.111: LAN expresses the Past prior to the present day. 6.4.71: LAN requires an accented -augment. 1.1.46: A unit with  occurs before [the root]. [Cf. lines 12-14 above] [Cf. line 7 above] 3.4.100 i → θ/_# in the subcategories of L ending with N [here: LAN]. [Cf. line 8 above]

INDIA

á+nay+a+an á+nay+0+an ánayan

67

8.2.23: CC→C/_# [Cf. line 15 above]

The derivation of the medial form noyante ('they lead for their own benefit') requires a couple of rules different from those required for deriving náyanti: nīN+SaP+JHa naya+JHe

nayante

[Cf. 1.3.72 on line 6] 1.4.100-102, 3.4.78: Medial, Plural, 3. Person = JHa. 1.1.64: 77 = V(C)#. 3.4.79: 77 [here: a] → e in the medial of the subcategories of L ending with T [here: LAT\. [Cf. above]

Finally, I shall present the derivation of the passive form nīyante ('they are led') in outline: nīN+yaK+JHa

nī+ya-f JHa nïyante

3.1.1-2, 66-67: After a root and before a full-stemaffix yaK occurs to express Undergoing or Happening. 1.3.12-13: Undergoing or Happening is expressed [also] by medial endings. [Cf. above: Medial, Plural, 3. Person = JHa] 1.1.5: The auxiliary symbol  retains the basic degree of the preceding vowel. [Cf. above]

The following word-forms are derived with no additional rules: Medial Imperfect anayanta ('they led for their own benefit'), Passive Imperfect anīyanta ('they were led'), Medial Causative Present nāyayante ('they cause to lead for their own benefit'), Active Causative Imperfect anāyayan('they caused to lead'), Medial Causative Imperfect anāyayanta ('they caused to lead for their own benefit'), Passive Causative Present nāyyante ('they are caused to lead'), Passive Causative Imperfect anāyyanta ('they were caused to lead'). As a kind of résumé, I shall explain the theoretical-descriptive elements involved in the abstract form of anûyayan ('they caused to lead'):

68

UNIVERSAL HISTORY OF LINGUISTICS

1) aT, required by LAN (replaced here by tiN); , when added to an affix, indicates that the affix occurs before the root. 2) In the root list (dhātupātha), which is part of the 'theoretical exposition' (upadeśa) and thus contains non-surface forms, this is how the first sound of the root nf- is represented.23 3) Ñ, when added to a root, indicates that the verb may have medial endings in the standard meaning, i.e. to express that the action will be to the benefit of the agent. 4) N, when added to an affix, indicates that the preceding vowel has the full degree. 5) C, when added to an affix, indicates that its last syllable is accented. If the affix has only one syllable, this is vacuously the 'last' one. 4) & 5) MC expresses causativity, i.e. the idea that there is someone who causes someone else to act. A root and MC constitute a secondary root. 6) A unit with Ś qualifies as a 'full-stem-affix'. 7) An affix with P is unaccented. A full-stem-affix without P is treated as if it had Ñ, which (like K) retains the basic degree of the preceding stem vowel. Thus we have ji+SaP+JHi → jayanti and bhū+SaP+JHi → bhavanti, but ksi+Sa+JHi → ksiyanti and dhū+Śa+JHi → dhuvanti. 6) & 7) ŚaP expresses 'Doing'. 8) tiN = a pratyāhāra symbol which contains in a condensed form all basic verb endings. The verb morphology of Sanskrit is exceedingly complex. In addition to the ten subcategories of L, each of which has its own set of endings, there are 'second­ ary' and 'tertiary' finite verb-forms as well as a large number of non-finite forms. Moreover, the roots are divided into ten classes, each with its own peculiarities. Covering all this data in a way that achieves a high degree of generalization, has required not just intelligence, but also nearly infinite patience. I take it that by this time I have sufficiently well documented the high theoretical level of Pānini's grammar. To me it is beyond question that at least until the year 1970 a comparable grammar was produced nowhere in the world, that is,

INDIA

69

a grammar that would have to the same extent combined extensive coverage of data with theoretical sophistication. It seems rather obvious, moreover, that it is only during the last decades, with the advent of formal or 'algebraic' modes of description, that Western linguists have become capable to fully appreciate Pänini's achievement. This is graphically demonstrated by Whitney's (1972 [1884]) and (1972 [1893]) indignant reaction to Pänini's grammar, which he considered "a miracle of ingenuity, but of perverse and wasted ingenuity", wasted because "the task Pānini attempted was beyond the power of mortal man to accomplish, especially in the form adopted by him"; thus, for Whitney, Pänini's grammar turns out in the end to be detrimental to "clearness and objective truth". Whitney's reaction is only the more significant, because in the more common-sense field of diachronic linguistics, in particular the explanation of linguistic change, he was a first-rate theoretician and ahead of his own time (cf. CLT.2W). Against the historical-comparative background of the late 19th century, it is understandable that those who defended Pänini's formalist grammar-conception against Whitney's attacks may not have done it quite convincingly (cf. the articles reprinted in Staal 1972). Still, it is a pity that Kielhorn did not deign to engage in debate with Whitney, because on the basis of his translation of and commentary upon Nageśa's Paribhāsendusekhara, it can be said with hindsight that he must have been one of the very best representatives of theoretical linguistics of his own time. It is good to add that already in Whitney's time the value of Pänini's grammar was occasionally acknowledged also outside the circle of Indologists. Thus, von der Gabelentz (1891:22) characterizes it as a 'miracle' (Wunderwerk), i.e. as "die einzige wahrhaft vollständige Grammatik, die eine Sprache aufzuweisen hat, eine der reichsten Sprachen zudem". And he later (p. 116) repeats this assess­ ment: A 'complete grammar' sets itself the task of recording and explaining all grammatical phenomena, even the most infrequent and insignificant ones, of a given language. This task has been fulfilled only once, namely in Pänini's Wunderwerk. Finally, it is interesting to note that Bloomfield, a later compatriote of Whitney, fully grasped Pänini's significance, and apparently even borrowed some of the latter's descriptive techniques (cf. Rogers 1987). Echoing von der Gabelentz, Bloomfield (1933:11) notes that "[Pänini's grammar] describes, with the minutest detail, every inflection, derivation, and composition, and every syntactic usage of its author's speech. No other language, to this day, has been so perfectly de­ scribed." This was written in the early 30's; and, as I said before, I am for my part

UNIVERSAL HISTORY OF LINGUISTICS

70

willing to extend Pānini's period of supremacy at least by some forty additional 24

years.

2.4 Linguistics after Pänini The title of this section is, in a sense, misleading. Indian linguistics carries Pānini's imprint to such an extent that there is no linguistics 'after Pänini', if by this is meant linguistics that would have somehow 'left him behind'. Thus, in what follows I shall just consider linguists who temporally followed Pänini.

2.4.1 Kātyāyana and Patañjali Pānini's grammar would be incomprehensible without the aid of commentaries, and Kätyäyana and Patañjali are the most authoritative among his commentators. Therefore I have been relying on both throughout the preceding discussion, although I have mentioned them explicitly only a couple of times. It is my purpose in the present subsection to get their respective contributions into somewhat sharper focus. The principal question that arises in this context concerns the precise amount of compliance or independence that characterizes Kātyāyana's and Patañjali's work vis-à-vis Pänini. Did they feel free to criticize him? Or did they see it as their duty to defend him at all costs? A preliminary answer to this question was given already on p. 8. It is a pervasive feature of the Indian culture that within any given tradition the sūtra literature is followed by a bhāsya literature which endeavours to explain and supplement, rather than refute, the former. Apparently the Hindu religion set here the example. There are huge differences, and even contradictions, between the respective doctrines of the Vedas and of the Upanishads (cf. p. 7) but, by common agreement, these just had to be reconciled, or papered over, as well as possible. At least prima facie, Patañjali bears witness to this general attitude. His Mahābhāsya is divided into 85 âhnikas ('day-lessons'), and although it is otherwise a running commentary on Astādhyāyr, its first āhnika offers a sort of general introduction to the study of grammar. Now, almost one half of the first ähnika (Abhyankar & Shukla 1973:7-20, 37-40) is devoted to proving that there is religious merit to be gained from studying grammar. In arguing his case, Patañjali just accepts the authority of several proverbial sayings. He seems merely to

INDIA

71

perpetuate the same attitude when, approaching the end of the āhnika, he raises the question as to the usefulness of the hundreds of determinatives employed by Pänini. He notes that the ('deep') forms equipped with determinatives must in any case be deleted, in order to reach the actual ('surface') forms; so maybe the former could be dispensed with altogether. His definitive view, however, is as follows: This is all true; but doing this will be un-Pāninian, i.e. going against the method of Pänini. So, let the whole thing be as is done by Pänini himself: there should not be made any change of, or addition to, the already existing rules of Pänini (p. 50). This surely looks like total surrender of any critical or independent attitude vis-à-vis Pänini. The truth, however, is more complex. A closer look at Patañjali's actual commentatorial practice soon reveals that he by no means considers Pänini's sūtras as sancrosanct. The same is true of Kätyäyana as well. The kāraka discussion provides good examples of how the two commen­ tators try to improve upon Pänini. The rule 1.4.24 correlates the semantico-ontological category 'what remains firm when one goes away' and the semantic role 'source' (cf. p. 45). The rule 1.4.25 then goes on to. add that 'source' also corre­ lates with 'what causes fear', and additional qualifications are given in the rules 1.4.26-29. Now, Kätyäyana points out that also objects of disgust or of abandon­ ment should qualify as 'sources' (since they are expressed by the ablative), and he proposes corresponding additions to Pänini's rules. Patañjali claims, however, that all the examples cited either by Pänini or by Kätyäyana involve a 'going-away', either physical or mental, mentioned in the rule 1.4.24. As a consequence, he regards not only Kätyäyana's additions, but also the rules 1.4.25-29 as superfluous, or implicit in the basic rule 1.4.24. Thus where Kätyäyana proposes additions, Patañjali proposes deletions. On the other hand, Patañjali wishes to make an addition to the rule 1.4.32, which introduces the notion of 'recipient' (cf. p. 44): Because recipients, expressed by datives, also occur with patientless actions, expressed by intransitive verbs, the rule should be modified as follows: 'He whom one tries to connect with the patient [karman]or with the [mere] action [kriyā] is the recipient'.25 Cases like these can be easily multiplied. Thus it would be quite wrong to say that Pānini's awe-inspiring authority made any form of criticism impossible a priori. Now, it is quite true that Kätyäyana or Patañjali never recommend any thorough-going modifications to Pānini's grammar, let alone its total abandonment. But what does this prove? It proves either that they were unwilling (though able) to

72

UNIVERSAL HISTORY OF LINGUISTICS

do so, or that they were unable to do so. Everything that I have learned about the subject inclines me to adopt the latter alternative. It has sometimes been suggested that Kätyäyana was a severe, and even unfair, critic of Pānini, whereas Patañjali set himself the task to defend the latter against the former's attacks. I do not feel competent to appraise the issue; but what I said above, at least entails that I have seen no evidence suggesting that Kätyäyana was hostile to Pänini. This seems to be in agreement with the current majority opinion (cf. Cardona 1975:250). Next I shall give an example of Kätyäyana's and Patañjali's 'cooperation', i.e. of how their joint contribution genuinely clarifies and enlarges Pänini's legacy. At the same time, a glimpse at least is provided into Kätyäyana's style. As has been indicated before, semantics is the least developed area in Pänini's grammar. Therefore it is understandable that it is here that Kätyäyana's and Patañjali's contribution is the most substantial. 33 värttikas and 213 bhāsyas are devoted to the rule 2.1.1, which is one of Pänini's very few explicitly semantic rules:26 samarthahpadavidhih: 'Co-semantic word-rule [i.e. a rule applies to two or more words only on condition that their meanings are related].' In fact, Joshi (1968), a book of 242 pages, is a commentary upon this single rule. In his vārttika XX on 2.1.1, Kätyäyana comes to discuss the term samānādhikarana. It means either 'coreference' or 'syntactic agreement' (resulting from coreference). What is at issue, is the meaning of a descriptive-coreferential compound like vīra-purusa ('hero-man'). It gives Kätyäyana an opportunity to outline the fundamentals of his semantic theory, in five värttikas. I shall first quote them, with the translations (cf. Joshi 1968:148-168): XX samānādhikaranesūpasarrúchyānamasamarthyāt: 'Because there is no semantic connection [asamarthyāt], in the case of words which syntactically agree [samänudhikaranesu] a specific rule has to be stated [upa-samkhyäriam].' XXI dravyampadurtha iti cet: '[There is no semantic connection] if the word-meaning \padartha] is [identified with] the individual object [dravya].'

INDIA

73

XXII na vā vacanapramānyāt'.'Or [a specific rule need] not [be stated], on the authority of the rule [vacanapramānyāt] [i.e. the rule 2.1.58, which explicitly mentions vTrah among the first members of possible descriptive-coreferential compounds].' XXIII luptākhyutesu ca: '[The same applies] also [to compounds with] deleted verbal forms [as in the case of prācāryah, 'forwardteacher', derived from pragata ācāryah, 'forward-gone teacher', which is implied by the rule 2.2.18].' XXIV tadarthagatervû: 'Or [compounding is achieved, not on the basis of deletion, but] because of tadarthagati [i.e. 'implicit under­ standing of meaning', arthagati meaning just 'understanding of meaning'].' Being a commentary, Kätyäyana's text is almost devoid of auxiliary symbols. Still, as the preceding excerpt shows, his style is so concise as to make the vārttikas nearly incomprehensible. The translations make full use of Patañjali's explanations, but even so, I am quite sure that they fail to disclose Kätyäyana's intentions, at least to the average reader. To produce an adequate understanding, one needs actually to tell how Patañjali and the subsequent tradition have inter­ preted the värttikas. Thus Pānini commented upon by Kätyäyana, might rightfully be called an 'enigma shrouded in mystery'. What the passage quoted above purports to say, can be explained as follows. Consider an expression like brave man. Both brave and man refer to the same thing. If one accepts the view that the meaning of a word is identical with the thing referred to, it follows that in the expression brave man there is no semantic connection. (A connection can obtain only between two different entities, not 'between' one and the same entity.) Thus in brave man we would have syntactic agreement, but no semantic connection. This implication is unacceptable, which means that the word meaning (padārtha) cannot be identical with the individual thing (dravya). What is needed, is a conception which entails that brave and man, while referring to the same thing (dravya), express different qualities (guna). Thus, in brave man, there is both a difference (bheda) and a unity (samsarga). The situation can be depicted as follows:

74

UNIVERSAL HISTORY OF LINGUISTICS

Furthermore, the notion of 'käraka', as we have it today (cf. 2.3.5-A), is a result of Kātyāyana's and Patañjali's joint efforts. It is only on the basis of their clarifications that kārakas (as well as their semantico-ontological counterparts) can be seen as produced by a communicative intention (yivaksu) when it applies to 'pure' or prelinguistic ontology. It is also only within the post-Pāninian tradition that a terminological trichotomy dravya—sādhana—kāraka was established, corresponding rather well to the trichotomy 'ontologicaT—'semantico-ontological' —'semantic', introduced here in 2.3.5-A. Pänini deals with kärakas in 32 rules. This section is introduced by the 'heading' rule 1.4.23: kārake. 15 värttikas and 38 bhäsyas are devoted to this rule by Kätyäyana and Patañjali, respectively. What now follows, is a résumé of this general discussion. The notion of 'kāraka' is further elucidated by pointing out that within a ('global') action each käraka performs its own subsidiary action. The clue for this interpretation (which is especially plausible in connection with causative construc­ tions) was given by Pänini himself, because käraka means 'agent' and is thus virtually synonymous with kartr. The latter term is, of course, reserved for the principal agent. Already at the level of prelinguistic ontology, there has to be a distinction similar to that between 'agent' and 'patient': Knowing whether it is X which (or who) initiates an action involving Y, or vice versa, is typically a matter of simple perception, and not of language. And typically there is an agreement between ontological participants and semantic roles. Under exceptional circum-

INDIA

75

stances, however, the speaker may wish to single out some subsidiary action within the global action, and thus to turn e.g. an ontological instrument into a semantic agent (expressed by the nominative). Consider the sentence devadattah kastaih sthālT odanam pacati (lit. 'Devadatta with logs in a pot gruel cooks'), which is the favorite example of the Hindu grammarians. Ontologically Devadatta initiates the action, which means that typically he is referred to by the nominative, and thus qualifies semantically as the agent. For one reason or another, however, one may wish to emphasize the contribution made either by the logs or by the pot. This is achieved by putting the corresponding words into the nominative (and deleting any reference to Devadatta). In the resulting sentence, either the logs or the pot would express the semantic agent; but the ontological status of the corresponding things would of course remain exactly the same as before. (Remember that putting gruel in the nominative would still keep it semantically as the patient, because the verb would have to be passivized.) From Kātyāyana up to the present day, commentators have experienced great difficulties in trying to formulate the above-mentioned facts adequately. It is often said, for instance, that putting the pot in the nominative shows that 'location' can be represented as 'agent'. The correct way would be to say that the ontological participant which is normally represented as (semantic) 'location' may be rep­ resented as (semantic) 'agent'. The demarcation between ontology and semantics is further discussed in Kātyāyana's and Patañjali's comments on the rules 2.3.50 and 2.3.66, which deal with the uses of the genitive. In general, the cases other than nominative and genitive express kāraka relationships. The nominative is a 'neutral' case which expresses secondarily what has 'already' been expressed (= the anabhihite prin­ ciple). The genitive expresses the 'rest' (sesa). What is this 'rest'? Normally, it is identified as 'relation' (sambandha), i.e. a relation between two things (expressed by corresponding nouns), not between an action and a thing, as in the kāraka relationships. Patañjali points out, however, that the 'rest' just results from the speaker's unwillingness to use kārakas, which could — in principle — be used: karmādinûmavivaba śesah. Typical examples of śesa are such relationships as 'owner—owned' or 'part—whole'. Rather imaginatively, Patañjali sees these two relationships as instantiating, respectively, the kāraka relationships 'agentrecipient' and 'agent—location', with a merely 'implicit' action. (It was also suggested that the whole is not the location, but rather the agent, because it 'is' in its parts.) Whether or not one accepts such 'kārakas behind Éesa\ it is clear that

76

UNIVERSAL HISTORY OF LINGUISTICS

they are situated at the prelinguistic ontological level, and not at the semantic level (and should consequently not be called 'kārakas' at all). The preceding discussion, due to Kātyāyana and Patañjali, shows that the karakas qua semantic roles are not identical with the (semantic) 'cases' of today's case grammar. Consider these well-known sentences: 1) John opened the door with the key. 2) The door was opened by John. 3) The key opened the door. 4) The door opened. According to the case grammar, the cases expressed by John, the door, and the key (viz. agent, patient, and instrument) remain the same in each sentence. According to the Pāninian tradition, however, the door is patient in (1), (2), and (3), and agent in (4), while the key is instrument in (1) and agent in (3).27 This shows that for Pānini, semantics is to a higher degree determined by (surface) form than it is for the case grammar. To put it differently, the 'semantics' of the case grammar is identical with what I have called 'prelinguistic ontology'. Kiparsky (1982) argues that the kārakas must be interpreted not as semantic units, but as 'functions' at the level of 'abstract syntax'. Thus, e.g. in the rules 1.4.49 kartur Tpsitatamam karma (cf. here p. 44) and 3.2.123 vartamāne lat (cf. here p. 33), kartur tpsitatamam and vartamāne would stand for units of meaning while karma and lat would stand for units of abstract syntax (op. cit., p. 28). Now I think this interpretation is ruled out by the very form of Pānini's rules. In the examples above, Tpsitatamam is in the nominative, indicating that the rule is a definition. By contrast, vartamune lat cannot possibly qualify as a definition. That vartamune is in the locative, indicates that it is a unit of meaning; and the only function that lat can have is that of abstract (N.B. morphological) form being assigned to, or expressing, meaning. As far as the karakas are concerned, a locative occurs only at the derivational stage following the rule 1.4.49, as in the rule 2.3.2 karmani dvitfyā (cf. here p. 33), clearly indicating that karmani is a unit of meaning comparable to vartamune (and not a unit of syntax, as Kiparsky would have it). By contrast, when lat occurs in the derivation as the 'higher' entity contrasted with a 'lower' one, it is not in the locative (as the semantic units vartamune and karmani are), but in the genitive, as in the rules 3.4.77-78 lasya... (cf. here p. 24), indicating that it is an abstract (morphological) form to be replaced by less abstract forms (cf. here p. 34).

INDIA

77

I shall conclude this subsection by discussing a contribution that might be characterized more narrowly as Patañjali's own, i.e. a contribution not directly based either on Pānini's sūtras or on Kātyāyana's vārttikas. Having enumerated the religious merits that accrue from the study of grammar (or more precisely, from the correct use of language which is based on the knowledge of grammar), Patañjali devotes the remainder of the first āhnika to explaining how grammar should be taught. First of all, he notes that "before a grammarian, speech [as a goddess] exposes her real form just as a woman in love" (Abhyankar & Shuklar 1973:16). Unfortunately, most students of linguistics are not intelligent enough to appreciate this fact. Being reluctant to do any theoretical work, they do not see much point in studying grammar, since Sanskrit is learned even without its help (p. 19). However, such views are mentioned only to be brushed aside. But how, then, should grammar be taught? It could, in principle, be taught either with the aid of correct words or with the aid of incorrect words. However, since the number of incorrect words is very much greater than that of correct words, the only practicable way is to teach the latter. Moreover, even the correct words cannot be taught by simple enumeration. The justification for this view deserves to be quoted in extenso: The enumeration of each correct word individually is not the real way for the knowledge of correct words. It is heard that Brhaspati taught the Science of Words tofridraby means of a recital of all correct words and went on doing it for one thousand celestial years, but he could not reach the end... Then how are these correct words to be taught? There should be followed the method of laying down general principles, so that with a correspondingly small effort a man would learn bigger and bigger collections of words. What would that method be? Of course, the method of laying down general rules and exceptions... What kind of general rule should there be stated? So also what kind of exception should be given?

78

UNIVERSAL HISTORY OF LINGUISTICS

There should be stated a general rule with a wide application, as for example karmanyan [3.2.1, i.e. karmani aN\ prescribing the affix aN in the sense of the agent after a root having a word denoting the patient of the action as the preceding word [as in kwnbha-kāra, cf. here p. 50]. Then a specific exception to the rule should be given like ātonupasarge kah [3.2.3, i.e. āTah anupasarge Kah] prescribing the affix Ka instead of aN after a root ending in ā and having any word except a preposition as the antecedent (Abhyankar & Shukla 1973:20-22). By the infinite number of correct words, Patañjali does not mean words that have actually occurred in speech: "It cannot be said that a word is incorrect simply on the ground of its being unused" (p. 35). Thus, a grammar does not investigate a corpus of utterances, but rather the set of all, and only, correct words (and sentences). Such a set also contains rare or unusual words, provided they satisfy the criterion of correctness: "Although such words are not found in use, they have to be explained in grammar by stating rules, just as sacrificial sessions which last for years have to be sanctioned although they are not current" (p. 35). Then Patañjali goes on to add that the science of ritual must be able to describe even such sacri­ ficial sessions as last a thousand years (although they are not held by anybody as a matter of fact), if only they are (known to be) correct. It is hard to imagine a more eloquent argument for the openness of language and against narrow-minded corpus-thinking. What makes it especially interesting, is the comparison between language and ritual, or the view that these two, being constituted by sets of social norms, equally exemplify the superordinate notion of rule-governed behavior (for discussion, cf. GTM:2A, and CLT.3.6.2). It has been shown by Renou (1972 [1942]) that in India the science of grammar and the science of ritual share to a large extent the same vocabulary, and in such a way that grammatical terms tend to have a religious or ritualistic origin. In both cases, too, there is a sūtra literature followed by a steadily growing number of commentaries. Such descriptive notions as anuvrtti, paribhāsa, and adhikāra ('heading') are also common to both types of study. However, this similarity does not seem to have penetrated any deeper. As far as the actual nature of descriptions is concerned, the science of grammar remains qualitatively different from the science of ritual: "[II] résulte qu'a disparu dans le rituel [i.e. la science du rituel] tout le côté étonnamment formaliste de la grammaire, ... et les vastes conséquences qu'en tirent les écoles grammaticales" (cp. cit., p. 448).

INDIA

79

2.4.2 Bhartrhari After the three 'sages', the most important figure in Indian linguistics is Bhartrhari (c. 500). His contribution consists in making fully explicit the ontology presupposed by the Pāninian grammatical theory, and in embedding the resulting grammar-cumontology in the wider context of Hindu cosmology. This is achieved in his book Vākyapadtya('Study of Sentences and Words'). The book consists of three parts (kānda), each of which contains several hundred stanzas (kārika), some 2,000 in all. The kārika literature is less concise and cryptic than the sūtra literature. Yet it, too, is meant to be supplemented by commentaries. The first commentary on Vākyapadtya is supposed to have been made by Bhartrhari himself. Vākyapadtyastarts from the maximally abstract, i.e. the question about the nature of the universe and its knowability, moves forward in the direction of increasing concreteness, and ends by discussing detail questions of Pānini's gram­ mar. Although the general plan is clear, a transparent presentation of the particulars is made difficult by the fact that linguistic questions are touched upon at practically each step, and often in seemingly repetitive ways. What follows, is an attempt to give a simple and systematic account of Vākyapadtya. Several finer points of Hinduism will inevitably be lost. My main sources here are Ruegg (1959), Biardeau (1964), and Subramaniya Iyer (1969). The first part of the book starts by explaining the notion of 'tradition' (āgama). The tradition is decreed to be superior to reasoning or inference as employed by philosophical systems. Reasoning is acceptable only insofar as it agrees with the tradition. The most important representatives, or carriers, of the tradition are of course the Vedas, but also Pānini's Astādhyuyt is part of the tradition. It is the purpose of the tradition to serve as a means for attaining the moksa and the Brahman (cf. here p. 7). The Vedas are the symbol (or 'dream image') of the Brahman, and grammatical theory meant originally study of the Vedas. In addition to ordinary, unreliable sources of knowledge like perception and inference, there exists a superior and reliable source of knowledge, namely 'in­ tuition' (pratibhā). The seers who were present at the creation possessed the unadultered type of intuition, which enabled them to know the Brahman in its undifferentiated totality. In order to convey this knowledge, they created the tradition. But as time went by, the quality of the seers deteriorated; more and more commentaries were needed, but these just obfuscated, rather than revealed, the truth more and more.

80

UNIVERSAL HISTORY OF LINGUISTICS

The Brahman is One; for instance, it equally contains existence, i.e. what is no longer or not yet non-existent, and non-existence, i.e. what is no longer or not yet existent. Apart from the accomplished seers, however, human beings are not able to grasp the Brahman as it really is. Rather, human ignorance (avidyā) brings it about that the Brahman is perceived as something differentiated, i.e. in terms of temporal and spatial sequence. Time (kāla) and space (dik) are the two principal 'powers' of the Brahman, or its capacities to be perceived. Time and space are further differentiated into actions (kriyā) and participants (kâraka), respectively. Participants, in turn, are things (dravyd) that assume different roles in connection with different actions. But, to repeat, the whole phenomenal world results from human limitations or 'ignorance'.28 The knowledge of the world is inseparable from the language with which this knowledge is expressed. Knowledge qua self-awareness becomes virtually identical with language:29 without words, knowledge would reflect the world like a mirror, or inertly; but with words its reflection is alive because it reflects, and knows that it does so. Since the manifestations of the Brahman are inextricably linguistic, the Brahman itself is of linguistic nature: it is 'Word-Brahman', or 'Word-Essence' (śabda-tattva). In principle, it would seem that phenomena-asverbalized must conceal the Brahman and lead farther away from it. However, this cannot be the whole truth because someone (like Bhartrhari himself) who makes an appropriate effort is able to come to know the Brahman; and this can only happen by means of its manifestations, i.e. phenomena-as-verbalized: "The real is presented by the word through the unreal"; "treading the path of the unreal, one attains reality".30 The preceding interpretation of the relation between the Brahman and its manifestations is based on Subramaniya Iyer (1969). For her part, Biardeau (1964, esp. 308-310) argues that Bhartrhari did not regard the phenomenal world as illusory in an absolute sense. As she sees it, such a view can be rightly attributed to Buddhism, on the one hand, and to Samkara (c. 800), the creator of the so-called 'non-dualistic' school within the Vedānta system, on the other. In conformity with the Pâninian tradition, Bhartrhari is interested in 'semantic ontology', rather than in ontology tout court: "What the word presents, is the object for us." Thus, a thing (dravyd) is whatever can be referred to by a noun or a pronoun. As a consequence, abstract notions too qualify as 'things'. This is Bhartrhari's way to answer the criticism that by dealing with the nature of reality, he was exceeding the limits of his own (i.e. Pāninian) tradition, and doing philosophy, rather than linguistics. As a later commentator put it, linguists analyze the 'word-referent' (Êabdārtha), not the 'thing-referent' (vastvartha).

INDIA

81

Bhartrhari's ontology could be summed up as follows: The world is constituted by space (dik) and time (kāla) which contain things (dravya). Each thing contains a universal aspect (jāti) and a particular aspect (vyakti). Things are bundles of qualities (gum) which enable them to participate in actions (kriyā). When they actually do so, they embody a restricted number of aspects (sādhana or kārk). As noted above, it was maintained that the Brahman could be grasped in an act of intuition (pratibhā). The Brahman, in turn, was of linguistic nature in the two senses of being some kind of 'word-essence' and of being manifested by words and sentences. Making a bold connection, Bhartrhari concluded that linguistic entities are similar to the Brahman in being understood by means of intuition. This claim is not as outrageous as it might seem. It is a fact that the properties of sentences are not known on the basis either of sense-perception (i.e. 'observation') or of intro­ spection; but it is also a fact that the existence of some third way of gaining knowledge, called 'linguistic intuition', has often been felt to be problematical, or even mysterious (cf. Itkonen 1981). Now, if a transcendental entity like the Brahman is assumed to exist, it obviously cannot be known by such pedestrian means as observation or introspection. It is not quite clear whether Bhartrhari intends intuition always to possess some 'deep' qualities. Sometimes he certainly does, for instance when he characterizes intuition as an invisible force produced by actions done or sentences spoken in one's previous lives.31 At this point, the transition from the Brahman to the world has been completed at two parallel (or rather identical) levels: on the one hand, one moves via time and space to states of affairs consisting of actions and participants; on the other, one moves to sentences containing verbs and nouns. In both cases, this is a movement away from the real towards the unreal. Yet for the purposes of analysis, it is permissible to treat the unreal as if it were real. In this way Bhartrhari man­ ages to keep the best of both worlds: on the one hand, he can express his religious conviction that only the Brahman, which is devoid of distinctions, is real; on the other, he can retain all the results of Pāninian linguistics with its nearly in­ numerable distinctions. What the linguistic intuition grasps is called sphota, a term which is best left untranslated. There has been a tremendous amount of work devoted to discussing Bhartrhari's sphota doctrine. The commentators agree that Bhartrhari is struggling to express a profound insight; but it is less clear what this insight consists in. The sphota seems to be identical with 'meaningful sentence-form'. The similarity between the Brahman and the sphota, as grasped by intuition, is based on the fact that both lack sequence. It is only the external, perceptible reality, including spoken utterances, which necessarily possesses sequence. Now, it is easy to understand that

82

UNIVERSAL HISTORY OF LINGUISTICS

sentence-meanings, unlike events or states of affairs, should be free of spatial and temporal sequence. For the sphota to be genuinely Brahman-like, however, the distinction between (mental representations of) form and meaning must be elimin­ ated as well. But it is less easy to understand how this can (and why it should) be achieved. Indeed, Bhartrhari admits that grasping form and meaning as an identity is something that only true seers can do. As noted above, linguistic analysis produces results which, for Bhartrhari, are to some extent 'unreal'. At such an 'unreal' level he distinguishes, for instance, between uttered sentences (= A), their mental representations (= B), mental representations of referents (= Ç), and referents themselves (= D). Their mutual relations can be displayed as follows:

It is also only at an 'unreal' analytical level that a sentence (and a sentence meaning) is divided into words (and word meanings). On this issue Bhartrhari is strongly opposed to representatives of the Mïmamsa school. For the latter, words have an autonomous existence, and sentences and sentence meanings are put together out of words and word meanings. The disagreement between Bhartrhari and Mïmamsa is graphically expressed by their respective attitudes vis-à-vis the saying padaprakrtih sarnhitu, where pada = 'word', prakrtih = 'origin', and sarnhitu = 'sentence'. The saying means 'A sentence is padaprakrtih'. For Bhartrhari, padaprakrtih is a determinative compound, meaning 'origin of words', whereas for Mïmamsa it is a possessive-exocentric compound, meaning 'that which has words as its origin'. In the same vein, Bhartrhari and Mïmamsa disagree on how they interpret elliptical sentences. For instance, a sentence like 'The door!' may mean 'Close the door!', which implies that the hearer must supply the element close. But does this mean that he supplies the corresponding word, i.e. both its form and its meaning, or that he supplies only the corresponding meaning? Mïmamsa chooses the former alternative, whereas Bhartrhari chooses the latter one. In so doing, he faithfully perpetuates the Pāninian tradition: If lexical material should not be deleted (cf. here p. 62), then in the name of symmetry it should not be added either. Although Bhartrhari does not shrink from refuting opinions which he considers erroneous, he also displays a rather high degree of syncretic attitude. This ensues from his belief that philosophy of grammar should be independent of, and

INDIA

83

therefore common to, all the six philosophical schools (cf. here p. 8). It is even possible that Bhartrhari was open to some Buddhist influence. In any case, his tolerance is evident also from his view that there may be several equally good ways of grammatically analyzing one and the same data. For instance, he was quite willing to accept that the word-form bhavati could be analyzed as just bhu+ati, in addition to the standard Pāninian analysis bhu+(S)a(P)+ti(P). The way Bhartrhari elaborates his ontology is of great interest to any theory of general metaphysics, yet I cannot go into the details here. In the more directly grammatical part of his book, by contrast, he does not seem to add much to what Kātyāyana and Patañjali have already said. Indian representatives of rhetorics developed elaborate theories of meaning which both diverged from and transcended the semantic component of Pānini's grammar. In the present context I can only mention Kunjunni Raja (1969) for further consultation. I have found no evidence for systematic cooperation between Indian linguistics and Indian logic. The Nyāya school invented logic (cf. Matilal 1977) and constantly developed it towards greater formalization (cf. Bochenski 1961:433-445). It possessed a full-fledged philosophy of language, but failed to develop any grammatical theory that could have competed with Panini's grammar. Nor did it try to relate its own results to those of the Pāninian tradition. The mutual independence of Indian linguistics and Indian logic is confirmed by Staal (1988), which contains separate sections on the two topics, without any attempt to show that there would ever have been a fusion between the two.

2.5 Conclusion: Pānini and the Riddle of 'Progress' Pānini has an important message for anyone interested in the philosophy of science. In the West we are used to the idea that, barring such external catastrophes as wars or plagues, science just cannot help making progress. Now Pānini flatly refutes this cherished idea. If the history of a science is measured by the number of the changes (in the sense of improvements) that it has undergone, then we must say that the history of linguistics in India possesses no past and no history. Rather, it is just one big continuous present which, as it were, coils around Pānini. Nothing could tell that the voices of the individual Indian linguists, from Pānini to Joshi, are separated by hundreds and thousands of years. They seem to equally participate in a debate which goes on in our presence. The existence of a historical dimension is evident merely from the fact that once we have criticized one of our interlocutors, we cannot literally expect him to respond. Rather, we ourselves must try to elicit the

84

UNIVERSAL HISTORY OF LINGUISTICS

response by a more arduous study of his text. And more often than not, we do find a response that refutes our criticism. Thus it is not just the case that Pānini is both the oldest and the best in his own tradition (as far as there is a historical record of it). It is also the case that, at least until the late 1960's, Western linguistics has not been able to equal Pānini's achievement. Considering the central role of linguistics in the Indian education, we are now in a position to refute the traditional view about the 'static' or 'stationary' character of the Indian civilization: the reason why this civilization could not move ahead is simply that its starting point was too good to be surpassed. In a similar situation, the Western civilization would have turned out to be just as static. (To be sure, I am now ignoring the role of natural science.) Pānini refutes the customary Western notion of what science is like. In the chapter 6 I try to explain what was wrong with this notion, and how it should be emended.

INDIA

85

Notes 1)

Here as elsewhere, hyphens are used as a (non-original) means of clarification.

2)

One might suspect that Nyāya has cleverly secured its right to consider language as it really is, without the fear of being accused of heresy.

3)

Experts on non-Western cultures may also voice similar opinions; cf. Goody 1977.

4)

The oldest extant texts, representing a more recent variant of Sanskrit, date from  250 B.C. Nevertheless, Pineault (1989:311), for instance, assumes that the system of writing, called brāhmi, originated during the period 700-400 B.C. Moreover, while assuming Pānini to have lived around 500 B.C. (p. 332), he finds it probable that Astudhyayt could not have been created without at least the implicit support of writing (p. 310). The reason is that, in his view, Pānini's classification of sounds (cf. p. 17 here) as well as those of his rules that contain lists (cf. p. 37 here) had to be visu­ alized. I do not agree. To me this is a typical case of taking our Western habits of thought for granted.

5)

It is true that when sentences are artificially made to grow in complexity to the point where they are no longer spontaneously understood, the grammarian really is investigating written sentences; cf. Itkonen 1976.

6)

Under this general level of agreement, of course, there are lower-level differences between speaking and writing; but this has never been denied.

7)

The meaning of this term will shortly be clarified.

8)

Pineault (1989:354) is quite right to point out, however, that brevity is not an absolute requirement, since Pānini could easily have chosen shorter technical terms (cf. p. 23 here), if he just had wished to do so.

9)

Scharfe (1977:85) points out that in the present context dhātu means 'stem', and not 'root', as usual.

10) This point can be illustrated as follows. Traditionally it was assumed that Pānini uses three synonymous terms vu, vibhāsā, and anyatarasyām to mean 'optionally'. This would, however, be a clear violation of Pānini's general principle of economy. Kiparsky (1979) has shown that, rather than being guilty of redundancy, Pānini uses the three terms to express the three distinct meanings 'preferably', 'preferably not', and 'optionally'.

86

UNIVERSAL HISTORY OF LINGUISTICS

11) This principle is independent of 1.4.2 because exceptions can be made to precede the general rule. 12) yaK does not occur in all tenses. 13) The next rule 4.1.2 enumerates the basic case endings. 14) It has been suggested to me that if the average reader is given too many details of Pānini's grammar, he might lose his interest in this book and just put it away. Assuming that there is any truth in this rather gloomy assessment, the reader might wish to stop here and resume his reading on p. 68 (having, at least, inspected the figures on pp. 51-52 and the example on pp. 63-65). I want to make it clear, how­ ever, that all these details are absolutely vital to the general argument of this book. I am claiming, in essence, that Pānini is our contemporary and (at least) our equal in every respect, with all the consequences that this entails for the notion of 'progress in linguistics'. Surely I must have the right to justify the claim I am making. 15) It may be interesting to learn how concisely this rather complex content was original­ ly expressed: parasmin acahādeśahpūrvaviddhau sthānivat. 16) Writing non-actual intermediate forms like 'Fα', instead of 'Fa', would not change this fact. 17) Kwnbha-kāra is the basic word-form whereas kumbha-kārah is the nominative singular. 18) This term was not used by Pānini. 19) Here Patañjali anticipates the criticism that was around 1970 directed against the proliferating deep structures characteristic of 'generative semantics'. 20)

In light of Kiparsky's (1979) results (cf. note 10), however, 'optionally' is not an accurate translation of vibhāsā.

21)

It should be evident that 'aspect' is not used here in its customary sense.

22)

I mention here only the three basic temporal alternatives.

23)

For simplicity, I have not mentioned this fact before.

24) This assessment is supported by Kiparsky's (1979:18) characterization of Astudhyāyf as "the most comprehensive generative grammar written so far". 25) Joshi & Roodbergen (1975:120) point out that in a non-technical use karman may also mean 'action', implying that in this context karman might mean both 'patient' and

INDIA

87

'action'. This would make Patañjali's addition unnecessary, while retaining its content. I cannot believe, however, that Pānini would have assigned two different meanings, one technical and the other non-technical, to one and the same occurrence of a term. 26)

59 vārttikas, or the maximum number, are devoted to the rule 1.2.64, in which Pā­ nini implicitly deals with the relation of nouns to their referents.

27) It could be said that Pānini considers the agentive meaning of transitive verbs as 'prototypical'. 28) This is Kantianism before Kant; cf. p. 277. 29) It may be added that Nydya considers knowledge as primarily non-linguistic. 30)

One is reminded here of Wittgenstein (1921:6.54): "der, welcher mich versteht, ... muss sozusagen die Leiter wegwerfen, nachdem er auf ihr hinaufgestiegen ist. "

31)

One is reminded here of Plato's notion of anamnesis, and of its possible application to linguistic knowledge; cf. G7M:294-301.

3. China

In China there never developed a strong and continuous tradition either of linguistic theory or of philosophy of language. Miller (1975) mentions indigeneous lexico­ graphical work beginning with the first century A.D. and rudiments of phono­ logical description imported from India in the 6th century (cf. also Scharfe 1977: 79). Having likewise mentioned "very elaborate dictionaries", Casacchia (1989:431) offers this résumé: La tradition chinoise présentait néanmoins une lacune très grave, qui devrait la suivre jusqu'au seuil de notre siècle: l'analyse des structures grammaticales en est pratiquement absente. A côté d'une immense pro­ duction d'études concernant des mots particuliers ou des groupes de mots, on ne trouve à peu près rien concernant la description organique de la langue. All this agrees with the conventional view that Chinese thinking has been more practical than theoretical in character. The true picture, however, is a little more complicated. On the one hand, language does play a fundamental, even if largely hidden role in Confucianism, which is the principal tradition of Chinese thought. On the other, for a short period of time there did flourish a 'School of Names' devoted to philosophy of language.

3.1 Confucius Under the Chou dynasty (c. 1122-256 B.C.) the so-called 'Spring and Autumn' period (722-481 B.C.) was characterized by increasing social strife and disin­ tegration. Confucius (551-479 B.C.) — originally K'ung Ch'iu — lived in this age, and it was his overriding goal to create a class of both competent and virtuous men who, acting as government officials, would bring peace and prosperity to the country. The proper way to do so seemed to be to revert to the 'habits of the

90

UNIVERSAL HISTORY OF LINGUISTICS

ancient kings', i.e. the founders of the Chou dynasty. Thus, Confucius regarded it as his duty to keep alive and transmit the Chou civilization. Confucius called his ideal man chiln tzu (translated as 'superior man' by Bodde, the translator of Fung Yu-lan 1952, and as 'gentleman' by Lau 1979). The defining characteristic of a gentleman is benevolence (jen). As Fung Yu-lan (1952:66) sees it, benevolence has two components: individual morality (yi) and the observance of 'rites' (//), i.e. social rules of conduct. More concretely, benevolence consists in practising empathy (shu: 'Do not impose on others what you yourself do not desire'; Analects XV, 24), and in being conscientious (chung) in carrying out one's (empathetic) actions towards others. — Lau (1979:26-27, 38, 49) views the relation between yi and li in much the same way, viz. as that between individual (or internal) and social (or external). But he also attributes to yi the meaning of a moral standard against which particular acts are evaluated. This very brief characterization of what, to Confucius, it is to be a 'gentle­ man' could be rounded off by the following quotation: If a man remembers what is right at the sight of profit, is ready to lay down his life in the face of danger, and does not forget sentiments he has repeated all his life even when he has been in straitened circumstances for a long time, he may be said to be a complete man (XIV, 12). Confucius may be said to exemplify altruistic upper-class morality. Good government is defined by its ability to improve the living conditions of the common people (min). But the attitude towards the people is paternalistic and, generally, conservative. Up to now my exposition has contained nothing of specifically linguistic interest. However, in a rather surprising passage Confucius reveals that the question of language lies in effect at the very heart of his social program:1 Tzu-lu said: 'If the Lord of Wei left the administration of his state to you, what would you put first?' The Master said: 'If something has to be put first, it is, perhaps, the rectification of names.' Tzu-lu said: Ts that so? What a roundabout way you take! Why bring rectification in at all?' The Master said: 'How boorish you are. Where a gentleman is ignorant, one would expect him not to offer any opinion. When names are not correct, what is said will not sound reasonable; when what is said does not sound reasonable, affairs will not culminate in success; when affairs do not

CHINA

91

culminate in success, rites and music will not flourish; when rites and music do notflourish,punishments will not fit the crimes; when punish­ ments do not fit the crimes, the common people will not know where to put hand and foot. Thus when the gentleman names something, the name is sure to be usable in speech, and when he says something, this is sure to be practicable. The thing about the gentleman is that he is anything but casual where speech is concerned (XIII, 3). The 'rectification of names' (cheng ming) plays a pervasive role in Con­ fucius' overall system. On the one hand, it is obvious that he is constantly prac­ tising conceptual analysis, or explication, as when he tries again and again to elucidate his key concept jen (variously translated as 'benevolence', 'humanheartedness', 'humanness', 'goodness', 'love', etc.). This concern with 'names' (more precisely, with word-meanings) constitutes, as it were, the semantic aspect of his system. On the other hand, as the passage quoted above reveals (albeit in a rather cryptical way), his political theory depends on there being a 'correct' relation between words and things. In reality, however, it is quite artificial to separate in this way the two aspects in Confucius' thinking because his semantics is entirely motivated by his politics. Let us now have a closer look at the doctrine of the rectification of names. What is at issue, is the relation between words and things, and — more particular­ ly — the changes taking place in these two domains. As is evident from his veneration of the early Chou civilization, Confucius viewed the social changes taking place in his own time very negatively. In his opinion, to put it tersely, the change of things had to be not just stopped, but also cancelled. It is true that when things change, the meanings of the corresponding words are bound to change too. But this takes time. Normally the change of meanings lags behind the change of things, with the consequence that words (more precisely, word-meanings) retain to some extent the earlier situation, i.e. the situation when the things had not changed yet. When this is the case, there is a contradiction between words and things, or — as Confucius would put it — the relation between words and things is not 'correct'. A contradiction, by definition, has to be eliminated. A contradiction between A and  may be eliminated either by changing A or by changing B. It goes without saying that Confucius wished to change the things so that they would, once again, correspond to the words, and not the other way around.2 This is evident from the following passage:

92

UNIVERSAL HISTORY OF LINGUISTICS

Duke Ching of Ch'i asked Confucius about government. Confucius answered: 'Let the ruler be a ruler, the subject a subject, the father a father, the son a son (XII, 11). In the sentence 'Let the ruler be a ruler' the first 'ruler' stands for a thing whereas the second 'ruler' stands for an ideal embodied in the meaning of the corresponding word. That is, the meaning indicates what a ruler (= a thing) ought to be. There is no point in saying 'Let the ruler be a ruler', unless the ruler is not what he ought to be, i.e. unless there is a non-correspondence, or contradiction, between things and words. Indeed, in Confucius' time the traditional power of the ruler was being usurped by his feudal lords so that rulers no longer were what they used to be. And to Confucius' conservative cast of mind, rulers ought to be what they used to be. As already mentioned, the meaning of the word 'ruler' still conserved a memory of what rulers used to be, and therefore, according to Confucius, it also entailed a knowledge of what rulers ought to be. The correspon­ dence between things and words could only be re-established by the ruler taking the power away from the feudal lords. In just the same way, the father—son relationship no longer was what it used to be under the ancient kings. Sons did not show the same respect towards their fathers as they did before. Again, Confucius very much regretted this change, and did his utmost to cancel it. Confucius' hostility against the change of things extended even to cases with no possible social or political implications. He deplored (VI, 25) the fact that the word for drinking vessels with a specific form had come to be used for drinking vessels in general. According to Miller (1975:1217), this "passage ... provides the necessary key for an understanding of almost all facets of the Chinese linguistic tradition". In other respects too, Confucius would not let a discrepancy arise between words and things: The Master said: 'In antiquity men were loath to speak. This was because they counted it shameful if their person failed to keep up with their words' (IV, 22). The Master said: 'The gentleman is ashamed of his word outstripping his deed' (XIV, 27). The preceding account was inspired by Fung Yu-lan (1952:60). Yet the same passage also contains a sentence showing that it is easy to be misled by the term 'rectification of names':

CHINA

93

It was because the actualities of things no longer corresponded to their names, Confucius believed, that the world was suffering from disorder, and therefore the names must be rectified. However, as the examples given above illustrate, it is not the names that must be rectified, but the things; and their 'rectification' consists in bringing them into agreement with existing word-meanings. To this extent the term 'rectification of names' is in fact a misnomer. This has been noticed also by Dawson (1981:57). By contrast, Hansen (1983:72) argues that it is indeed names, and not things, that had to be rectified. I of course disagree with this interpretation. My reasons for doing so will be further elucidated on p. 97. Although things are the primary object of rectification, there are, to be sure, three senses in which names too may be 'rectified'. First even if the change of meanings lags behind the change of things, the former too is bound to take place after the latter has been going on for a sufficient amount of time. Therefore it is possible, and perhaps inevitable, that meanings too become 'corrupted'. 'Cor­ rupted' meanings must be rectified so as to make them constitute the data, or the starting point, for the scholar's analysis of what counts as 'gentleman', 'benevol­ ence', etc. This is rectification at the atheoretical level. Second, whether or not (atheoretical) word-meanings have changed, the scholar has a job to do. As was just pointed out, he has to analyze, or explicate, such concepts as 'benevolence'; and analyzing the concept 'A' is identical with analyzing the meaning of the word A. This amounts to rectification at the theor­ etical level, and it is part and parcel of every conceptual analysis, irrespective of the particular philosophical tradition (cf. GTMAl.O). Third, it has been taken for granted up to now that Confucius, being an 'enlightened conservative', opposed changes of any kind. But such an account certainly fails to do justice to him. We must distinguish what he claimed to be from what he was in fact. He claimed to be "a transmitter and not a creator, a believer in and lover of antiquity" (VII, 1), i.e. of the early Chou civilization. But it is quite certain that the Chou aristocracy did not, as a rule, exhibit that solicitude towards the welfare of the common people which Confucius valued so highly (cf. VI, 30). He also made it clear that the worth of a person did not depend on his social rank or on his wealth, but on his personal merits, primarily on his having "attained knowledge through study" (XVI, 9). (To be sure, the common people had little opportunity to acquire merits of this kind.) Again, this was certainly not the prevailing attitude among the Chou aristocracy. So we see that, while pretending to be just a "transmitter of antiquity", Confucius was on occasion, both atheoretically and theoretically, 'rectifying names', or making conceptual revisions, with respect

94

UNIVERSAL HISTORY OF LINGUISTICS

to what had been the case in antiquity. (In this instance, then, both names and things had to be rectified.) It is more difficult to decide what was the precise amount of rectification with respect to Confucius' own time: to what extent was his altruistic doctrine originally just a minority opinion?3 — It may be good to add that, on his own account, Confucius' adherence to the morals of antiquity was not dogmatic, but the result of a free choice: "I have no preconceptions about the permissible and the impermissible" (XVIII, 8). My preceding account of the rectification of names is based on the use of such terms as 'meaning' and 'ideal'. It is the basic tenet of Hansen (1983) that the classical Chinese philosophy was 'nominalistic' in spirit and thus dispensed with any abstract or mental entities. This he takes to entail that there was no use for 'meanings', 'ideals', or 'universale'. If he is right, my interpretation collapses; but he is not right. It is just a mistake to identify, first, abstract with mental and, second, mental with semantic (i.e. with meaning.) He tries to refute (p. 88) any appeal to meanings by arguing that, instead of relying on a 'mental apparatus', the Chinese just spoke of a correct 'practice' or use of language. But it should be common knowledge that meaning can be defined in precisely these terms. Besides, when Confucius defines what it is to be a gentleman, but despairs of ever finding gentlemen, he is necessarily speaking of an ideal. There is no other way to express the situation. I also find it odd that Hansen wishes to attribute a 'conventionalist' view of language to Confucius (pp. 79-80). He tries to justify his interpretation by referring to the fact that, in Confucius' opinion, there does occur discrepancies between words and things. But the same fact can, and should, be viewed as supporting the opposite, non-conventionalist interpretation: The isomorphism between words and things is so strong and natural that deviations from it suffice to bring the world into chaos. In conclusion I wish to touch upon some residual aspects of Confucius' thinking. We have already seen that he made a distinction between what is right and what is profitable, even profitable not just to oneself, but to the common people. What is profitable has to be furthered insofar as it is possible to do so within the limits set by what is right; but in case of a conflict, it is the latter which prevails. Yet Confucius was fully aware that his ideals could not be realized in an imperfect world: The gentleman takes office in order to do his duty. As for putting the Way (too) into practice, he knows all along that it is hopeless (XVIII, 7).

CHINA

95

Confucius had no illusions about what his disciples valued most: The Master said: I suppose I should give up hope. I have yet to meet the man who is as fond of virtue as he is of beauty in women (XV, 13). Sometimes his pessimism went even this far: The Master said: T am thinking of giving up speech.' Tzu-kung said: 'If you did not speak, what would there be for us, your disciples, to transmit?' The Master said: 'What does Heaven ever say? Yet there are four seasons going round and there are the hundred things coming into being. What does Heaven ever say?' (XVII, 19). However, when a hermit urged Confucius to give up his vain efforts and to turn his back on the world, he pulled himself together: The Master was lost in thought for a while and said: 'One cannot associate with birds and beasts. Am I not a member of this human race? Who, then, is there for me to associate with? While the Way is to be found in the Empire, I will not change places with him [i.e. the hermit]' (XVIII, 6).

3.2 Mencius The so-called 'Warring States' period (403-221 B.C.) began less than a century after Confucius' death, and social disorder intensified. 'To save the world' was the announced goal of most thinkers during this time. Mencius (c. 372 -  289 B.C.) — originally Meng K'e — was the most notable among Confucius' followers. He too believed that a class of dedicated 'gentlemen' could restore order into the empire. While Confucius' notion of 'gentleman' had been a purely normative one, Mencius added to it a psychological dimension. He did so by arguing that human nature is originally good; if it later becomes corrupted, this is due to unfavorable circumstances. It is in keeping with this attitude that he, on the one hand, did not constrain individual morality (yi) too strongly by 'rites' (//), and on the other, advocated pacifism. He went beyond Confucius in emphasizing the importance of the common people: "The people are of supreme importance; the altars to the gods of earth and grain come next; last comes the ruler" (Mencius VII B, 14). His political theory, which contains an

96

UNIVERSAL HISTORY OF LINGUISTICS

unusually strong element of economics, is expounded in much more detail than Confucius'. Mencius too claimed to be following the example of the early Chou civi­ lization. In his case, however, the tactical or propagandists reasons for this attitude are rather obvious. It suffices to quote a passage where he shows his de facto independence vis-à-vis traditional institutions: Mencius said: 'A great man will not observe a rite Qi) that is contrary to the spirit of the rites Qi), nor will he perform a duty (yi) that goes against the spirit of the dutifulness (yi) (IV B, 6). This passage could be translated more simply as: "Let a rite be a rite, and a duty a duty". It is seen at once that Mencius is here tackling the same problem of the rectification of names as Confucius is in the passage XII, 11, quoted on p. 91. There is, however, an important difference. In Confucius' case the present state of things is wrong, and the past state was right. In Mencius' case however, reference to the past has become nearly irrelevant, as shown by the context (= IV B, 3): Mencius maintains that, whatever the rites may have been, and whatever the official version of the rites may be at the moment, the subject has in fact no obligation to mourn a ruler who has behaved like an enemy towards him. Mencius is aware that what have to be rectified, are not names but things: I, too, wish to follow in the footsteps of the three sages [i.e. Yü, the Duke of Chou, and Confucius] in rectifying the hearts of men, laying heresies to rest, opposing extreme action, and banishing excessive views" (III B, 9; emphasis added). It might seem, however, that there are cases where a name has to be rectified because the corresponding thing (or person) can no longer be. Thus Mencius flatly denies that a tyrannical ruler had been a ruler (I B, 8). Accordingly, the name 'ruler' is withheld from a person who to all appearances was a ruler, but who did not match the (full) meaning of this name. Instead, the name 'criminal' is applied to him. A similar argument recurs e.g. in II A, 6 and V A, 5. It is cases like this which lead Hansen (1983:72) to claim that the 'rec­ tification of names' is after all an appropriate term: what had to be changed was indeed the way that names were being used. In Hansen's opinion, the rec­ tification is meant to solve cases of moral conflict, as in the case cited above: On the one hand, rulers should be respected; on the other, crimes should be punished.

CHINA

97

It seems to me that Hansen fails to make an important distinction, namely that between rectifying the general meaning of a word, and rectifying a particular use of a word: In general, 'ruler' means what it meant in the time of the ancient kings, and there is no need to rectify it. In this particular instance, however, one needs to make an exception and to rectify the use of 'ruler', in the sense of not applying it to (what appears to be) the ruler. Therefore, I repeat that it is things, and not names, that had primarily to be rectified. But I now add that, if this (primary) rectification had not been suc­ cessful, particular uses of names had (secondarily) to be rectified. When Hansen simply asserts that what, in Confucius' view, was wrong were not things but names, he looses sight of the past-oriented rationale of Confucius' entire phil­ osophy.4 To prove my point beyond any doubt, I offer one more quotation: The master said: 'One cannot but give assent to exemplary words, but what is important is that one should rectify oneself. One cannot but be pleased with tactful words, but what is important is that one should reform oneself. I can do nothing with the man who gives assent but does not rectify himself or the man who is pleased but does not reform himself (Analects IX, 24). It is undeniable that the contrast is here between words which are exemplary and actions which are not and thus need rectification. That the term 'rectification of names' does not occur here explicitly, is irrelevant.

3,3 The School of Names Confucius was the founder of the first philosophical school in China. What is more, it is said of him that "prior to Confucius there was no one writing in a private, non-official capacity" (Fung Yu-lan 1952:170). Soon after his death Mo Tzu created his own school, and for some time it was the only serious rival of Con­ fucianism (cf. 3.5). Around the year 300 B.C., however, there existed already several distinct schools of thought propounding their own views and arguing against one another. In the history of Chinese philosophy this is called the period of the 'Hundred Schools'. It coincided with the Warring States period, and therefore it is natural that, whatever the differences of opinion between the various thinkers, they tended to define their doctrines by the stand they took vis-à-vis the prevailing social chaos. The obvious exception was the 'School of Names', whose members took

98

UNIVERSAL HISTORY OF LINGUISTICS

up questions of linguistic and logical interest. The ultimate fate of this school vividly shows that it is not just a cliché to speak of the 'practical' nature of the Chinese thought. The members of the School of Names were called 'dialecticians'. 'Dialectic' (pien) was defined at the time as "that in which one person says a thing is so, and another says it is not so. The one who is right will win". It would be wrong to say that dialecticians were insensitive to the disorder and suffering around them. Hui Shih and Kung-sun Lung were the two leading dialecticians and, like Mo Tzu before them, they both preached 'universal love' and pacifism. Yet they were chiefly concerned with philosophy of language, and this was enough to make them targets of scholarly opprobrium. Thinkers who sharply differed between themselves were united in condemning the School of Names; and in the condemnations one can sense the tone of genuine, not-just-tactical bewilderment. Here are some samples, quoted from Fung Yu-lan (1952:193-194): [The dialecticians] threw a deceiving glamour over men's minds and altered their ideas. They could overcome men's words but could not convince their minds. Herein lay their weakness. There were some who would not follow the early kings and would not acknowledge the rules of proper conduct (li) and standards of justice (yi), but liked to deal with strange theories and to indulge in curious prop­ ositions. They were subtle but could not satisfy real needs; critical but useless; worked much but with few results. Their teaching could not serve as systematic regulations for government. ... Hui Shih was blinded by phrases and did not know facts... The School of Names made minute examination of trifling points in complicated and elaborate statements, which made it impossible for others to refute their ideas. They specialized in the definition of names, but lost sight of human feelings. One gets the impression that dialecticians established a new field of study with new standards of argumentation and could not, even by the admission of their opponents, be proven wrong on their own terms. Those outside this new discipline, however, not only lacked interest in it, but even disapproved of it. In the face of this collective disapproval, it is no wonder that the texts written by dialecticians have been lost, with the exception of a partially preserved text by Kung-sun Lung. What is known, or assumed, about the doctrine of the

CHINA

99

School of Names must therefore be pieced together from accounts given by persons who opposed it and most probably did not even understand it. What remains of Hui Shin's work are ten paradoxes recorded and explained by the Taoist philosopher Chuang Tzu (or his disciples). The common theme of these paradoxes is the relativity of all viewpoints. The first paradox sets the tone: "The greatest has nothing beyond itself, and is called the Great Unit; the smallest has nothing within itself, and is called the Little Unit." It is the explanation which, properly speaking, brings out the 'paradoxical' import of this dictum: The universe (i.e. 'Heaven and Earth') and the tip of a hair are of roughly equal size. Why? Because they both are finite whereas the Great Unit is infinitely great and the Small Unit is infinitely small; and, being bounded by infinity on both sides, any two finite things turn out to be of (roughly) equal size. Three other paradoxes are just variants of the first one. The last paradox says: "Love all things equally; the universe is one." This is clarified by the fifth paradox: "All things are in one way all similar". It then goes on to add: "All things are in another way all different. " This is made more concrete by the explanation of the third paradox: "If looking at things according to their gradated differences, we say they are relatively big, then there is nothing in the universe which is not big; if we say they are relatively small, there is nothing in the universe which is not small." Hui Shih's doctrine came in fact later to be known as 'the unity of similarity and difference'. There are two paradoxes bringing out the relativity of all local viewpoints. One of them says: "I know the center of the world; it is north of Yen [= a state in North China] and south of Yuen [= a state in South China]." There is one paradox bringing out the relativity of all temporal viewpoints: "I go to the state of Yüen today and arrived there yesterday." Finally, there are paradoxes asserting that everything is at every moment undergoing change: "The sun at noon is the sun declining; the creature born is the creature dying." Dialecticians were said to "specialize in the definition of names". Hui Shih's paradoxes might in fact be said to constitute a small 'philosophical dictionary', concentrating on such key words as 'similarity', 'difference', 'size', 'place', 'time', and 'change'. Hui Shih uses formulations calculated to catch attention, but there is nothing paradoxical, in the sense of 'unsolvable' or 'self-contradictory', in his insistence on the relativity of all viewpoints. (It is true that an all-out relativism leads to self-contradiction; but Hui Shih's paradoxes, in the form in which they have been preserved, do not justify such an extreme interpretation.) It is quite clear, however, that the relativistic spirit of Hui Shih's teaching goes against the philosophy of all those who, like Confucius, wished to save the

100

UNIVERSAL HISTORY OF LINGUISTICS

world by making everybody adhere to one single code of conduct, i.e. to one common way of acting and thinking. In this respect there is a strong, although somewhat unexpected, similarity between the School of Names and one of the two principal traditions of Taoism (cf. 3.4.). Kung-sun Lung, the other noted dialectician, explored the nature of univer­ sal concepts (chih), or universale for short. His most famous pronouncement was: "A white horse is not a horse." One of the justifications for this statement goes like this: "A white horse is 'horse' together with 'white'. 'Horse' with 'white' is not 'horse'. Therefore, I say that a white horse is not a horse." In other words, Kungsun Lung is just denying that a general concept is identical with any of its subcon­ cepts; but he has chosen a paradoxical way to express this simple truth. He belabours the point further, e.g. by pointing out that while 'horse' neither includes nor excludes any colour, 'white horse' includes whiteness and excludes blackness, yellowness etc. ; therefore, a white horse is not a horse. In elucidating the notion of 'universal', Kung-sun Lung introduces the basic components of a semantic theory. A name (ming) designates a thing (wu), which is an 'actuality' (shih). (In this context wu and shih seem to be interchangeable.) However, a name also has some kind of reference to a universal (chih). The relation of universale to things and to one another is discussed at length. Kung-sun Lung wishes to prove that even in one and the same hard white stone hardness and whiteness are in reality two separate entities. He tries to justify this claim first by referring to the fact that it is perceived by different senses that a stone is hard and that it is white. He next points out that hardness and whiteness are something general; they in no way specify in advance the things in which they will be found (and thus they need not occur together). Therefore they are different not only from each other, but also from the things in which they are found. Universals, unlike things, cannot be perceived: "Hardness does not exist in the world, but it lies concealed." On the other hand, "things cannot be said to be without universals". If there were no things in the world, it would be impossible to discover the existence of universals, because they would be permanently unknow­ able. In addition, Kung-sun Lung makes a number of remarks to the effect that while things change, universals are beyond change or manipulation. His doctrine came later to be known as 'the separateness of hard and white'. I have presented here what amounts to the 'standard' interpretation of Kungsun Lung, based on Fung Yu-lan (1952:204-212). This interpretation has been at­ tacked by Hansen (1983, chap. 5), who argues that Kung-sun Lung did not invent universals or abstract entities, but remained fully within the standard, nominalistic

CHINA

101

framework of Chinese thinking. This interpretation leads Hansen to conclude (p. 151, 160) that in asserting his paradox, Kung-sun Lung committed the following oddity. Considering a combination like ox-horse (= 'oxen and horses') as paradig­ matic, he interpreted white-horse in the same way, i.e. as meaning the sum (rather than the product) of what is white and what is a horse. The paradoxical sentence remains true, but this time not because 'white-horse' is less general than 'horse', but because it is more general (= 'white and/or horse'). I find Hansen's interpretation implausible, mainly because of the in­ coherence of his supporting arguments (pp. 144-148). First, he states that Kung-sun Lung's doctrine was incomprehensible to the Chinese tradition. But if he had remained within the main-stream nominalistic framework, it should have been comprehensible. The lack of comprehension can be explained only by assuming that Kung-sun Lung did depart from the main-stream tradition, namely by in­ venting abstract entities. Second, Hansen argues that while e.g. Zeno's claim that all movement is impossible is a genuine paradox, the same is not true of the 'abstract-entity' interpretation of Kung-sun Lung's claim that a white horse is not a horse (ex­ pressed, more adequately, as white-horse is-not horse). Why? Because "his own theory entails something that those he is trying to persuade of the theory must consider false" (p. 146). But this characterization applies, of course, to Zeno's claim as well. Thus Hansen fails to show that, on the 'abstract-entity' interpret­ ation, Kung-sun Lung's paradox is "rather silly" (p. 146). On the other hand, it does appear silly, and not paradoxical at all, to assert in conformity with the nominalist interpretation that what is a horse or a white thing is different from what is just a horse. I conclude that Kung-sun Lung, most probably, did invent and use the notion of 'universal' or 'abstract entity'. The common label 'School of Names' actually hides differences of opinion, or at least of emphasis. As Fung Yu-lan (1952:214-215, 220) sees it, Hui Shih advocated a philosophy of particulars and of change, whereas Kung-sun Lung advocated a philosophy of universals and of permanence. It would seem more correct, however, to say that Hui Shih's philosophy was based on the notion of relativity. Kung-sun Lung dealt with such universals as 'horse', 'whiteness', and 'hardness'. Hui Shih too dealt with universals (even if indirectly), but his univer­ sals, unlike those of Kung-sun Lung, are necessarily such as to admit a multi­ plicity of different viewpoints: 'similarity', 'difference', 'place', and 'time'.

102

UNIVERSAL HISTORY OF LINGUISTICS

Although dialecticians seemed to be just "making minute examination of trifling points", they were aware of the larger implications that their work might conceivably have. This is indicated by the following passage: Kung-sun Lung was a dialectician of the time of the Warring States. Dissatisfied with the divergence and confusion between names and their actualities, he used his peculiar talent to discuss the alleged inseparability of whiteness. ... He wished to extend his argument so as thereby to rectify names with their actualities, and thus transform the whole world (emphasis added). In terms of how they influenced the history of Chinese thought, Confucius and Kung-sun Lung could not be farther apart. The former was deified whereas the latter was forgotten. There is one point, however, on which they both agreed: it is the rectification of names which holds the key to saving, or transforming, the world. The School of Names also teaches a lesson of more general significance. In India ('idle') speculation on the nature of language has been practised throughout the recorded history, and it has enjoyed the greatest prestige. In China, by contrast, it did not survive. The most remarkable point, however, is that it did occur in China at all. So there were no constraints that would have forced all Chinese to think in the same ('practical') way. In this context, as in so many others, the thesis of 'cultural determinism' is just false. Of course, one might still ask why the Indian culture has been more conducive to linguistic speculation than the Chinese culture. At least a partial answer is provided by referring to the role of language in the Vedic thinking (cf. p. 6 and 70).

3,4 The Taoist School Seeing that one's best efforts to save the world lead to nothing may easily give rise to disillusionment. In Confucius' time there were several educated men who rejected the existing political system, but had also given up the hope that it could ever be replaced by a better one. In consequence they had retired from the world. These hermits did not yet present their views in any systematic form (at least no such texts have been preserved), but in their general attitude they were precursors of Taoism.

CHINA

103

Too ('the Way') has always been one of the central terms of Chinese philosophy: Rival schools would each claim to have discovered [the Way] even though what each school claimed to have discovered turned out to be very different. The Way, then, is a highly emotive term and comes very close to the term 'Truth' as found in philosophical and religious writings in the West (Lau 1979:11). For Confucius, a gentleman follows the Way. Still, the Way somehow emanates from the Heaven, which remains the highest metaphysical principle with its dual role of the Destiny and of the issuer of moral imperatives (op. cit., pp. 28-30). In Taoism the Way becomes the highest principle and acquires several qualities which the traditional Heaven lacked. Yang Chu, an older contemporary of Mencius, could well be characterized as a proto-Taoist. His doctrine was later described as follows: "Completeness of living, preservation of what is genuine, and not allowing outside things to entangle one's person: these were what Yang Chu established, and were condemned by Mencius." The following passage explains why a Confucian had to oppose Yang Chu: The principle of Yang Chu is: 'Each one for himself.'Though he might have benefited the whole world by plucking out a single hair, he would not have done it {Mencius: VII A, 26). It would be wrong, however, to equate Yang Chu's position with simple egoism or hedonism. 'Completeness of living' means a harmony between one's desires. Its opposite, which is worse than death, is 'constrained life', i.e. a life of 'submission and shame'. According to this view, social and political institutions cannot but constrain one's life and should therefore be rejected. Yang Chu held that the most important thing is to maintain the integrity of one's person, and this means living in harmony with nature. Among the Hundred Schools there was a group of men, also to be characterized as proto-Taoists, who represented a nearly opposite position. For them, the most important thing was to live naturally (or 'spontaneously'), and this meant "making no discrimination in things but simply moving with them". Spontaneity of this kind excludes planning and conscious deliberation, and leads ultimately to losing the integrity of one's

104

UNIVERSAL HISTORY OF LINGUISTICS

person, i.e. to obliterating the dividing line between one's self and one's surround­ ings. Shen Tao represents here the extreme standpoint: Shen Tao discarded knowledge, abandoned self, followed the inevitable, and was indifferent to things. Such were his principles. He said: 'Knowl­ edge is not to know.' He was one who despised knowledge and would destroy it... Shifting and slippery, he changed with circumstances... Paying no attention to past or future, he stood loftily indifferent to everything... He said: 'Let us be like creatures without knowledge. That will be sufficient. There is no use for Sages. Even a clod of earth does not miss the Way'(Fung Yu-lan 1952:153). Shen Tao and his companions wished to ignore, or transcend, all dis­ tinctions, including that between right and wrong: "Selection involves exclusion; instruction involves incompleteness. The Way omits nothing." The implications of such a view are nothing short of staggering (cf. p. 106). The label 'Taoist school' (like the label 'School of Names') was coined by scholars of the Han dynasty (206 B.C. - A.D. 220). The acknowledged masters of Taoism were Lao Tzu and Chuang Tzu. (To be sure the book Lao-tzu, also known as Tao Te Ching, was written by several persons over a considerable period of time.) Although they were grouped together in this way, there were pronounced differences between their views. Tao as defined by Lao Tzu is the first principle from which all things are born. Not being a thing, it is Non-being; but it produces Being. Tao manifests itself in the phenomenal world as 'reversion', i.e. as an eternal movement between pairs of opposites: "What is in the end to be shrunk must first be stretched... The Way that goes forward seems to go back." The Taoist Sage places himself within the movement of things and goes along with it without making any resistance. This is the principle of wu wei ('non-activity'): "The world is invariably possessed by the one who does nothing." It may seem curious that Taoism, which started as a general reaction against politics, came to have a political theory of its own. Yet there is a certain logic in this. Lao Tzu would reject the Confucian virtues of benevolence and righteousness as well as all existing institutions because they, according to him, produce conse­ quences opposite to those they were intended to produce: "The more laws are promulgated, the more thieves and bandits there will be." The Taoist Sage would rule the empire through wu wei, i.e. through non-ruling. Before this state can be reached, however, the source of disorder must be eliminated. Disorder arises from the fact that people desire more than they can get; and the more they know, the

CHINA

105

more they desire; therefore disorder is suppressed by suppressing the desires, and these are suppressed by suppressing knowledge: "The Sage rules the people by emptying their minds, filling their bellies, weakening their wills, and toughening their sinews, ever making the people without knowledge and without desire. " Apart perhaps from the way Tao is characterized, Lao Tzu's teaching is rather tough-minded and practical. With his younger contemporary Chuang Tzu (c. 369 -  286), "the most brilliant mind the Chinese race ever produced" (Hughes 1942:120), Taoism acquires a mystical dimension. Chuang Tzu too believed that good government should be based on non­ government (wu wei). But his reasons for this view were opposite to those of Lao Tzu. For Chuang Tzu, the innate nature of one person is likely to differ from that of another. People are happy only if everybody follows his own innate nature. The worst thing a ruler can do is to constrain his subjects so as to make them all alike, e.g. by suppressing their (natural) desires. That everybody should be allowed to do as he pleases, entails that all viewpoints and opinions are equally good. This is what Chuang Tzu claimed in fact. He thought that the debates between different schools were simply beside the point: Whom shall I ask to settle this dispute [between you and me]? I may ask someone who agrees with you; but since he agrees with you, how can he settle it? I may ask someone who agrees with me; but since he agrees with me, how can he settle it? I may ask someone who differs from both you and me; but since he differs from both you and me, how can he settle it? I may ask someone who agrees with both you and me; but since he agrees with both you and me, how can he settle it? (Fung Yu-lan 1952:231.) From this radical relativism, which he takes to be inescapable, Chuang Tzu draws the obvious conclusion: Regardless whether the changing sounds [i.e. the words uttered by rep­ resentatives of different schools] are relative to one another or not, let us harmonize them within the Boundary of Nature... Let us forget the distinctions of right and wrong. Let us find enjoyment in the realm of infinite and remain there (ibid.). Not to discriminate 'that' and 'this' as opposites, is the very essence of Tao (p.232).

106

UNIVERSAL HISTORY OF LINGUISTICS

Since all things are one, what room is there for speech? (p. 235.) I said here on p. 91 that "a contradiction, by definition, must be elim­ inated". Now I must add that I meant contradictions occurring in scientific contexts. Deliberate acceptance of contradictions establishes a text eo ipso as non-scientific. Among the Chinese schools of thought reviewed in this chapter, Taoism is the only one which deliberately violates the standards of scientific or (in the customary sense) rational discourse. This attitude became evident already in connection with Shen Tao, and it is quite explicit in the passages quoted above. Since Chuang Tzu embraces contradiction, it would be beside the point for me to note that he is contradicting himself when he uses speech to say that there is no use for speech.6 (Notice also that in characterizing Chuang Tzu's text as non-scientific, I made no value judgment. There is surely no obligation to produce scientific texts only.) Acceptance of contradiction disqualifies one as a scientist, but does not yet qualify him as a mystic. Chuang Tzu's mysticism is evident from the fact that the distinction between life and death figures prominently among the distinctions which he, using strongly poetic language, is anxious to eliminate: When we come, it is because we have the occasion to be born. When we go, we simply follow the natural course. Those who are quiet at the proper occasion and follow the course of nature, cannot be affected by sorrow or joy. These men were considered by the ancients as people who are released from bondage. When [people] dream, they do not know that they are dreaming. In their dream, they may even interpret dreams. Only when they are awake, they begin to know that they dreamed. By and by comes the Great Awakening, and then we shall find out that life itself is a great dream. To have attained to the human form is a source of joy. But, in the infinite evolution, there are myriads of other forms that are equally good. What an incomparable bliss it is to undergo these countless transitions! (p. 236.) Chuang Tzu describes in detail the process by which one comes to grasp Tao. First one disregards all worldly matters. Then one disregards all external things. Then one disregards one's own existence. Then one becomes enlightened, gains a vision of the One, and transcends the distinction of past and present.

CHINA

107

Having transcended the distinction of past and present, he was then able to enter the realm where life and death are no more. Then, to him, the destruction of life did not mean death, nor the prolongation of life an addition to the duration of his existence. He would follow anything; he would receive anything... Being like this, one reaches a union with the universe (p. 242). Too produces all things and their distinctions, and it is again Too in which they disappear when one understands their unimportance. In this respect Too is identical with the Brahman of Hinduism. Similarly the union of the Taoist with the universe is identical with the Upanishadic union of ātman with brahman. Chuang Tzu speaks of the 'wordlessness' of Too. Distinctions are most obviously codified in words, and therefore rejecting the former means rejecting the latter as well. In particular, Chuang Tzu repudiated the 'Rectification of Names' doctrine. Thus it is interesting to learn that there are nevertheless strong affinities between Chuang Tzu and the dialectician Hui Shih. When Chuang Tzu prepares the way for his radical relativism, he makes use of the explanations which he has given to Hui Shih's paradoxes (cf. here p. 99). Apparently the two were not only rivals, but also friends of some sort. This can be inferred from the way Chuang Tzu laments Hui Shih's death: "Ever since Hui Shih perished, I have been without my material, having no one with whom I can speak. " As Chuang Tzu saw it, Hui Shih's mistake was that he just stopped short of full-fledged Taoism. He was not ready to give up words and believed that some­ thing could be achieved by debate and argument. To Chuang Tzu, this meant leaving the business unfinished: Hui Shih might be considered as the exponent of one aspect of Too... He could not content himself with the one aspect; he diffused his energy over all things but was never satisfied... Alas! Hui Shih with his talents wandered about without achieving anything; he went after things without returning (to Tad). It was like shouting to silence an echo, or racing with his own shadow. Alas! (p. 203.) Although Confucianism and Taoism were strongly opposed to each other, there are yet some points of contact between the two. Both Fung Yu-lan (1952:129-131) and Lau (1979:45-46) have found a streak of mysticism in Mencius' view of the gentleman "being in the same stream as Heaven and Earth". One might go even farther and claim that Taoism is the constant temptation of a Confucian. When Confucius considers "giving up speech", to justify this moment

108

UNIVERSAL HISTORY OF LINGUISTICS

of despondency he uses an argument (Analects XVII, 19, quoted here on p. 95), which echoes the Taoist sentiment and is repeated almost verbatim in Chuang-tzu: "The operations of Heaven and Earth proceed with the most admirable order, yet they never speak." It would seem that Taoism is there all the time, just waiting for the Confucian to give up. Taoism preaches the vanity of words; it denies everything that linguistic theory asserts. For a linguist, to know Taoism is to know one's enemy. At the same time one may realize that even the enemy can make (non-scientific, metaphor­ ical) sense.

3.5 Mo Tzu and His School Mo Tzu (c. 479 -  381) was the founder of the second school of philosophy in China. His goal was the same as Confucius' but his methods were somewhat different. The problem to be solved was the existence of human suffering: "The people have three worries: that the hungry will not be fed, that the cold will not be clothed, and that the tired will not get rest." Human suffering could be ended only by a (utilitarian) doctrine which was based not on some abstract concept of 'morality', but on the concept of 'profitableness' (li, not to be confused with the homophonic word for 'rites'). All statements and opinions had to be tested as to their profitableness, i.e. as to whether or not they directly contributed to dim­ inishing the "three great worries of the people". The way to achieve this was to practice extreme frugality. Therefore Mo Tzu did not seek his model in the aristocratic Chou civilization, but in the older and simpler civilization of the Hsia dynasty (c. 2206 -  1766 B.C.): "Anything that caused extra expenditure, but did not add to the benefit of the people, the Sage-kings would not undertake." At this point a conflict becomes apparent between Confucius' upper-class background and Mo Tzu's lower-class background. (The latter was said to recruit his disciples from the soldier class.) Confucius favored music, elaborate funeral rites and long mourning periods, whereas to Mo Tzu all this was just a waste of time and of resources. Similarly Confucius preached 'graded love', according to which one's love should decrease as it extends to people with whom one has less and less close ties, whereas Mo Tzu preached 'universal love', according to which one should love all people equally. (To be sure, he tried to justify this view by its alleged profitableness.) Mo Tzu's fervent pacifism is also in keeping with his 'universal love' doctrine.

CHINA

109

Thus Mo Tzu seems to favor egalitarianism. Yet his political theory also presupposes a chain of command beginning with an emperor who rules with strict discipline and exacts absolute loyalty. For the sake of completeness it may be mentioned that among the 'Hundred Schools' there was an 'Agricultural School', whose members represented egalitarianism in a much more radical form: "They could see no need for Sage-kings. Desiring both ruler and subject to plough together in the fields, they overthrew the order of upper and lower classes." Mencius saw it as his mission to defend Confucius against the attacks by Yang Chu and Mo Tzu (cf. Mencius III B, 9 and VII A, 26). He regarded the former's self-love and the latter's universal love as equally extreme and equally dangerous. Chuang Tzu seems to offer a more balanced assessment of Mo Tzu. He is content to point out that with his exclusive emphasis on fulfilling the basic needs, Mo Tzu would make life rather unpalatable: I fear that to teach men thus is not to love them, while to practice this oneself is certainly not to love oneself. These objections do not suffice to overthrow Mo Tzu's system. Yet ... though men mourn, he condemns mourning. Though men enjoy music, he condemns music... It is contrary to human nature and would not be tolerated. Mo Tzu himself might be able to do it, but what about the rest of the world? (Fung Yu-lan 1952: 104.) From the viewpoint of linguistic theory Mo Tzu's doctrine is of little interest. Later Moists, however, have more to offer in this respect. Feeling threatened by the 'subversive' arguments of the School of Names, they endeavoured to refute them as well as they could. And in order to do so, they had to get acquainted with philosophy of language. Thus, together with what remains of dialecticians' work, the counter-arguments offered by later Moists represent the best thinking on language in ancient China. The 'dialectical' contribution of Moists was written during the latter half of the Warring States period. It is contained in six chapters of the book Mo-tzu, which is a collective achievement of the Moist school. It is characteristic that the text of precisely these chapters has been least well preserved. Moists set the stage for their critique of the School of Names by defining some basic epistemological and semantic concepts. Knowledge is a 'meeting' between the faculty for knowing and the object to be known. Knowledge becomes clearer when the faculty for knowing 'discusses' its object. The close similarity between linguistic and mental processes is also indicated by this statement: "When by means of the spoken one's ideas gain expression, this is the dialectical activity

110

UNIVERSAL HISTORY OF LINGUISTICS

of the mind." Knowledge is divided into several subtypes, including knowledge of names (ming), of actualities (shih), and of the correspondence (ho) between the two. (In view of Hansen's (1983, Chap. 4) arguments, it might be better to translate shih as 'stuff.) Knowledge of a name enables one to possess a type of general knowledge independent of any particular things falling under the name. Moists give a rather wide meaning to the term 'dialectic'. It means an analysis which has to discriminate between similarity and difference, right and wrong, good government and bad government. The method to carry out the analysis is to "use names to express actualities, propositions to express ideas, statements to set forth causes". 'Setting forth causes' means classifying the different methods by which — and the different degrees to which — a given proposition comes to be accepted. What is at issue is the transition from (a certain kind of) experience to proposition or from one proposition to another. Thus equipped, Moists are now ready to refute their opponents. First they tackle Hui Shih's paradoxes based on the alleged 'unity of similarity and dif­ ference'. They try to show that statements which seem to disprove common sense merely arise from a misleading or careless use of words. They themselves are careful to distinguish between four different subtypes of both 'agreement' and 'difference'. In the case of 'agreement' one has to distinguish: i) identity, ii) partwhole relationship, iii) coexistence, iv) similarity. Analogous divisions are made within 'difference'. If one, for instance, infers "All things are similar, therefore all things are one", one simply confuses similarity with part—whole relationship, i.e. one confuses things (= parts) with the universe (= whole). (To be sure, such a formulation does not occur at least among those of Hui Shih's paradoxes that have been preserved.) The issue of relativity is broached with the saying: "Different classes are not comparable ... what is longer, a tree or a night?" The meaning seems to be that paradoxes arise when two incompatible viewpoints, each of which is justifiable in itself, are mixed in a single sentence. This is a valid criticism of such paradoxes as "I know the center of the world; it is north of Yen and south of Yüen" or "I go to the state of Yüen today and arrived there yesterday". Thus Moists wished to undermine relativism by arguing for the separateness of similarity and difference (within a given context). The dialectician Kung-sun Lung had argued for the 'separateness of hard and white'. As against this, Moists argued for the 'unity of hard and white'. In a combination of oxen and horses (as expressed by ox-horse) an ox is always separ­ able from a horse, i.e. they do not 'interpenetrate', in the sense of coinciding. By contrast, in a combination of hardness and whiteness, as it occurs in a hard white

CHINA

111

stone (as expressed by hard-white-stone), hardness and whiteness interpenetrate and thus are 'inseparable' (cf. Hansen 1983:155-159). This notion of 'unity' (of e.g. hard and white) is made plausible by the fact that, in common with the main­ stream tradition of Chinese philosophy, Moists could conceive of universals only as existing in particular things. (Alternatively it might be said that they were able to dispense with universals, at least in any strong ontological sense.) The Moist view concerning the rectification of names agrees, on the whole, with Kung-sun Lung's. Both things and word-meanings, if left alone, are likely to change. The latter can nevertheless be fixed and thus made permanent. The former, however, are beyond control. For some reason, the possibility of 'fixing the things' is not taken here into consideration. To Moists, a white horse is a horse, and to ride a white horse is to ride a horse. (Kung-sun Lung's point in making the contrary claim seems to have been missed here.) However, although Moists condemned the verbal tricks supposedly performed by dialecticians, they themselves comeratherclose to sophistry in this context. They have to reconcile their 'universal love' doctrine with their accept­ ance of capital punishment. They manage to do so by arguing that although a robber is a man, to kill a robber is not to kill a man. (Analogously, if A's brother is a handsome man, and A loves his brother, it does not follow that A loves a handsome man.) An extraordinary emphasis is laid here upon how events are interpreted: it is possible to kill a robber qua robber, irrespective of the fact that he also 'happens' to be a man. This means that practically all verbs with human subjects may acquire an intensionalist reading. Under exceptional circumstances it could be argued, for instance, that to ride a white horse is not to ride a horse, i.e. just any kind of horse. (This comes, in the end, quite close to Kung-sun Lung's position. It also has an affinity with Hui Shih's insistence on the equality of different subjective viewpoints.) Moists also criticized other rival schools of thought like Confucianism and Taoism. Confucius' reverence for antiquity was deemed excessive: "Yao was a wise ruler because we judge him from the present to the past. But if the ancients were placed in the present, Yao would be incapable of good rule." Thus, even if Yao was a good ruler in his own time, his skills might just be outdated in the present time. But Moists go even farther: "Why a thing becomes so; how to find it out; and how to let others know it: these need not be the same. " That is, the real state of affairs is hidden behind the two obstacles of first finding out and then letting others know; and distortions may occur at both steps. This is the inevitable triad of 'ontology—methodology—publication'. Perhaps Yao was a good ruler, but we don't know it for sure (and, to repeat, even if he was, he would be today at best

112

UNIVERSAL HISTORY OF LINGUISTICS

an indifferent ruler). Perhaps Yao was a bad ruler, but our own investigations have misled us into thinking that he was a good ruler. Perhaps somebody has found out (with reasonable certainty) that Yao was a bad ruler, but he has deliberately misled us into thinking that he was a good ruler. Et cetera. It is interesting that Moists, though admirers of the ancient Hsia civi­ lization, exhibit such an independent attitude vis-à-vis the past. This is in keeping with the fact that Mo Tzu himself was an advocate of intellectual progress: Kung Meng Tzu said: 'The Superior Man does not create but transmits.' Mo Tzu said: 'Not at all... It seems to me that what there is good in the old should be transmitted, and what there is good in the present should be created [i.e. established], so that what is good may increase all the more' (Fung Yu-lan 1952:247). Finally, Taoists are criticized for self-contradiction, or for the lack of selfreflexivity: when they make a universal statement, they invariably exempt them­ selves from its field of application. This succinct criticism is directed against Lao Tzu: "Study is advantageous. The reason is given by those who oppose it." That is, Lao Tzu discourages people from studying; but he eo ipso encourages people to study Taoism. Chuang Tzu, in turn, holds that all opinions are equally good and condemns criticism. But, as Moists are quick to point out, "to condemn criticism is to condemn one's own condemnation". Chuang Tzu's advice to give up speech gets no better treatment: "To hold that all speech is perverse, is perverseness." The explanation of this dictum deserves to be quoted in full: To hold that all speech is perverse is not permissible. If the speech of this man [who holds this doctrine] is permissible, then speech is not perverse since there it is permissible. But if this man's speech is not permissible, then it is wrong to take it as being correct [or to utter it at all] (p. 277). According to ordinary logic, Moists are of course right and Taoists are wrong. The matter is, however, made more difficult by the fact that Taoists openly profess to abandon ordinary logic (cf. here p. 105). But then it should also be clear that rather than trying to convince their audience, they are trying to impress it; and one should take their words accordingly.

CHINA

113

3.6 Hsün Tzu Confucius had two outstanding disciples, Mencius and Hsün Tzu (c. 298 -  238). They differed by temperament and therefore developed Confucianism in two rather different directions. Mencius had claimed that human nature is originally good and is corrupted only by unfavorable circumstances. Now Hsün Tzu claimed that human nature is originally evil and may be bettered only by favorable circum­ stances. If people are left alone, only chaos will ensue: "People desire and hate the same things. Their desires are many, but things are few. Since they are few, there will inevitably be strife" (X, 3). It was in order to prevent this that the ancient Sage-kings created the rules of proper conduct and the social distinctions. Because he has no trust in the human nature Hsün Tzu lays great stress on collective norms and disregards individual freedom. When norms are adhered to, the society will flourish: Water and fire possess forces but are without life: shrubs and trees have life but no knowledge; birds and beasts have knowledge but no standards of justice. Man has force, life, knowledge, and also standards of justice. Hence he is the highest being on earth... How is it that men are able to form social organizations? Because of their social distinctions... Hence they are able to order their actions according to the four seasons, control all things, and take all things profitable in the whole world (IX, 5). Hsün Tzu agrees with other Confucians that the early Chou civilization should serve as the model. But he disagrees with his predecessors in thinking that the emperor should be given absolute power. Thus he openly represents the totalitarian streak in Confucianism. The book Hsün-tzu contains a chapter devoted to the rectification of names. Its main purpose is to combat 'heretical' views — mainly those of dialecticians — by outlining an epistemological and semantic theory based on what Hsün Tzu takes to be common sense. In these matters his thinking is very close to Moists'. For Hsün Tzu, language is conventional in character: "There are no terms assuredly appropriate of themselves. There was an agreement and things were named" (XXII, 6). 7 The (isomorphic) principle 'one referent — one form' is the natural basis of language: "When things are alike, they are named alike; when different, they are named differently... Hence he who refers to different realities should never use other than different terms... Likewise he who refers to the same reality should never use other than the same term" {ibid.). It was the responsibility of the ruler to maintain this one-to-one correspondence between words and things.

114

UNIVERSAL HISTORY OF LINGUISTICS

Others were not allowed to tamper with language: "Therefore ... making unauthor­ ized distinctions, thus confusing the correct nomenclature ... was called great wickedness. It was a crime like using false credentials or false measures." Hsün Tzu distinguishes between three different types of fallacies current in the literature of his time: i) 'fallacies in the use of terms with the result of con­ fusing the terms'; ii) 'fallacies in the use of realities with the result of confusing the terms'; iii) 'fallacies in the use of terms with the result of confusing the realities'. The three fallacies are illustrated, respectively, by the following statements: "To kill a robber is not to kill a man", "Mountains and abysses are on the same level", and "Even if you do not go and see, the center pillar exists". The justification for this trichotomy is not very clear to me. Perhaps, to Hsün Tzu the first statement, accepted by Moists, is conceptually false whereas the second statement, which exemplifies Hui Shih's relativism, is factually false. Given Hsün Tzu's no-non­ sense frame of mind, however, it is certainly not the third statement, but rather the statement denied by it, which for him represents a fallacy. Hsün Tzu's conception of 'dialectic' is the same as Moists': Names are that wherewith we designate many realities. Speech is names which combine many realities in order to express one idea. In dialectic and explanation people should separate realities from names in order to make known Tao of what one should do or should not do. Designation and giving names is the use of dialectic. Dialectic and explanation is that whereby the mind delineates Tao (XXII, 9). A wise ruler is a master of dialectic. He knows that all heretical views spring from one of the three fallacies mentioned above. But it is enough that he should know it. There is no need for him to justify his thoughts or actions to his subjects. Hsün Tzu's totalitarian attitude becomes strikingly evident here: Hence the wise princes dealt with the people by authority, and guided them to Tao; they explained things in their proclamations, gave statutes in wise maxims, and restrained them by punishments. Hence their people were turned into Tao as by magic. Why should they use dialectic? It is generally claimed that Hsün Tzu 'distorted' Confucius' thought. And certainly the latter would not have approved of the former's coercive methods. Yet it might be argued that Confucius' paternalism genuinely entails the possibility of being developed into Hsün Tzu's totalitarianism.8 This is just the step from names

CHINA

115

(or rather, things) being rectified by a community of philosophers to names being rectified by the philosopher-king alone.

3.7 The Legalist School "The great political tendency of the time was a movement from feudal rule toward a government by rulers possessing absolute power; from government by customary morality (li), and by individuals, to government by law" (Fung Yu-lan 1952:312). It was the members of the Legalist School who represented this 'modern' trend against such past-oriented 'conservatives' as Confucians and Moists. Unification of the Chinese empire was the goal shared by all thinkers at the time. But this goal was very difficult to reconcile with the amount of intellectual freedom which thinkers had become accustomed to during the last two hundred years. Legalists, more than any others, were willing to sacrifice this freedom entirely. Legalist thinking had originated already in the 6th century B.C., but it culminated at the close of the Warring States period. Its main exponent was Han Fei Tzu (d. 233 B.C.). Like Hsün Tzu, whose disciple he had been, he believed that human nature was evil, and that public use of dialectic should be forbidden. From Taoism he learned that the ruler should govern his subjects like a force of nature, being beyond clemency or cruelty. Han Fei Tzu's own contribution consisted in his expounding the idea that it is by means of laws that the ruler wields his absolute, Heaven-like power. From the viewpoint of linguistic theory Han Fei Tzu's interest lies in the fact that he gives to the 'Rectification of Names' doctrine an absurdly extreme interpretation. He does so by not distinguishing between violations of laws and violations of linguistic rules: In using the method of maintaining uniformity, names are of primary importance. When names have been rectified, things will be fixed. When names have not been rectified, things undergo change... The ruler holds the names in hand, and his subjects model their actualities after the names... When the ruler wishes to prevent wickedness, he examines into the correspondence between actualities and names, words and work... Thus when subjects have made great claims, while their actual accomplish­ ment is small, they are punished. This is not punishment because of the smallness of the accomplishment, but because the accomplishment is not equal to its name. And when subjects have made small claims while the actual accomplishment is great, they are also punished. This is not because

116

UNIVERSAL HISTORY OF LINGUISTICS

no pleasure is taken in the larger accomplishment, but because it is not in accord with the name given to it (Fung Yu-lan 1952:324-325). I said that this view is 'absurd' in its extremism. To this it might be objected that we should refrain from criticizing the views of another time and/or another culture. However, a wide spectrum of conflicting views has been displayed in this chapter. If I dislike a particular view, it is not improbable that I do so for roughly the same reasons as its erstwhile opponents did. If they were allowed to voice their criticism (and what is more, if they are today allowed to have voiced their criticism), I do not see why I should be denied the same right. China was indeed united under the Legalist banner in 221 B.C. It was namely in large measure thanks to its Legalist prime minister Li Ssu that the semibarbarian state of Ch'in managed to conquer the other states. Like Han Fei Tzu, Li Ssu had studied under Hsün Tzu and later, seeing a dangerous rival in his fellowstudent, he had him killed. Together with his master, who had assumed the title of 'First Emperor', Li Ssu organized in 213 B.C. the Burning of Books, with the intention to eliminate all scholarly knowledge that was not controlled by the ruler. This was Legalism (and political Taoism) put into practice. The Ch'in dynasty lasted, however, only until 207 B.C. Already a couple of years before, Li Ssu had shared the fate of Han Fei Tzu.

3.8 The Aftermath Soon after the installation of the Han dynasty (207 B.C. - 220 A.D.) Confucianism reached the status of orthodoxy. A system of examinations was instituted to recruit the most gifted men of the whole empire for high government posts, knowledge of the Confucian classics being the criterion of admittance. This system of government examinations persisted for more than 2,000 years, and the ensuing non-hereditary class of scholar-officials (= 'mandarins') played a decisive role in maintaining the unity both of the Chinese empire and of its intellectual tradition (cf. Needham 1977:66-70). To this limited extent, then, Confucius' dream of a 'gentleman' class did become reality. Confucianism became something of a state religion, but its nature was deeply altered in the process. Therefore during the chaotic Period of Disunity (221-589) it was eclipsed first by Neo-Taoism and then by Buddhism, which had been imported from India. Buddhism prevailed during the prosperous T'ang dynasty

CHINA

117

(618-906), but was superseded during the politically weaker Sung dynasty (9601279) by Neo-Confucianism, which remained the orthodoxy until the 20th century. It may seem curious that this period of more than 2,000 years is simply characterized as 'aftermath'. Yet this term is justified. Hsün Tzu introduced the notion that the only proper form of research was the 'study of the classics', and this attitude came to be accepted universally. Since the Han times, Confucianism had four major classics, i.e. Analects, Meneius, and two treatises from the huge secondary literature. (Neo-)Taoism had Lao-tzu and Chuang-tzu. Buddhism had its own, originally Sanskrit scriptures. Thus starting with the Han dynasty the spirit of research is that of scholasticism, i.e. of commenting upon the superior achieve­ ments of an earlier age. It goes without saying that some commentaries (or some parts thereof) are achievements in their own right. Nevertheless, they are of derivative character. And what is more important for the present purposes, they do not seem to contain much that would be of linguistic interest. Already before Confucius there existed some attempts at systematic think­ ing, notably the method of divination, codified in the book / Ching, the yin—yang doctrine, and the doctrine of five elements (= wood, fire, earth, metal, water). From the Han times onward, Confucianism absorbed all these originally separate doctrines. The result was a rather fantastic cosmological theory which in many respects was directly opposite to Confucius' sober and sceptical thinking. At the same time, however, this theory illustrates perfectly the workings of the principle of isomorphism (or iconity). Therefore it may be worthwhile to discuss it briefly. The traditional method of divination was based on the dichotomy between unbroken (= 'male') and broken (= 'female') lines, as they appear on tortoise shells or are made use of in manipulating two types of divining stalks. When three lines are combined at a time, there ensue eight different combinations (or 'trigrams') with the following meanings:

Soon two trigrams were combined at a time to yield a total of 64 'hexagrams', with a consequent proliferation in meanings.

118

UNIVERSAL HISTORY OF LINGUISTICS

The yin—yang doctrine is based on the dualism between femininity and masculinity, darkness and light, cold and warmth, moisture and dryness, softness and hardness, passivity and activity, etc. It was a natural step to combine / Ching and yin—yang, by identifying the broken lines with (various manifestations of) yin and the unbroken lines with yang. Next the five elements were brought in, together with seasons, directions, musical notes, colours, numbers, etc. The result was a theory which tried to explain the history and the constitution of the entire universe by means of a huge set of one-to-one correspondences in which units function as signs of one another.9 In particular, detailed correlations were established between microcosmos (i.e. man) and macrocosmos (i.e. the 'numerical categories of Heaven'). This kind of study persisted, with varying emphases, until the very end of the period of Neo-Confucianism (cf. Fung Yu-lan 1953: chapters II, III, XI, XII, XIII and XV, especially pp. 11-16, 30-32, 91-96, 455-464, 546-551). The Neo-Confucian cosmology is just a manifold generalization of the 'Rectification of Names' doctrine, considering that the latter too is based on an isomorphic sign-relation, i.e. a one-to-one correspondence between words and things. The Neo-Confucians meant to grasp the nature of the universe, nothing less. Although they failed to do so, their edifice possesses an obvious interest from the semiotic point of view. In a nearly unique fashion it bears wittness to the power that iconicity exerts over the human mind. It is not only that men have an urge to create signs. Rather, what is interesting, is the fact that men seem to experience insuperable difficulties in realizing the merely metaphorical or playful nature of (most of) the signs they have created. Instead, they insist on taking all of them at face value. This is where iconicity becomes an obstacle to scientific thinking. A concrete example of missed opportunities is provided by the fact that if NeoConfucians just had hit upon the idea of combining broken and unbroken lines with 'true' and 'false' (rather than with 'female' and 'male'), they might have invented truth-functional propositional logic. Chu Hsi (1130-1200), the great synthesizer of Neo-Confucianism, illustrates well the scholastic spirit of this school. He repeatedly subjects its central terms and pronouncements to minute examination. He is also extremely conscious of the language he uses, as shown by the following excerpts: "You must take the words in the sense in which they are used by the respective writers" (p. 7). "But although this is what I meant, I am bound to confess that, owing to the poverty and restriction of language, the idea was not very happily expressed" (p. 12). "It is said that the word 'good' in the expression 'The nature is good' is not moral good" (p. 33). "I have thought over these words carefully, and their meaning seems to me not

CHINA

119

different from the teaching of Yo Chi" (p. 38). "For, as a matter of fact, although they are not two entities, yet there is a difference in the meaning of the two terms which we must not fail to observe" (p. 44). "I say that the words 'Nature' and 'Mind' have each a different connotation" (p. 45). "There are many such passages in Ming Tao's writings. If you take them too literally you cannot understand any of them. The important thing is letting go the letter, to grasp the meaning and not allow yourself to be tied to one way of looking at a passage" (p. 126). Such examples could be freely multiplied; they are characteristic of Chu Hsi's approach in general. When he explains Confucius' and Mencius' 'normative psychology', he elicits fine meaning distinctions in analyzing such concepts as 'sincerity', 'ingenuousness', 'truthfulness', and 'sympathy'. But just as often he seems to be making terminological stipulations which are valid only in a theoryinternal, Neo-Confucian sense. Yet he remained the authority until the coming of the Chinese Republic. I shall not dwell upon Neo-Taoism or Buddhism. The arrival of Buddhism in China was facilitated by the fact that it has obvious and systematic affinities with Taoism. For instance, the Buddhist notion of nirvana was equated with the Taoist notion of wu wei (cf. Fung Yu-lan 1953:258, 268). The fact that Buddhism was for several centuries the predominant doctrine in China shows, once again, that the conception of the 'practical' character of the Chinese mind must not be taken in an absolute sense. Any such notion of cultural determinism is disconfirmed in the Chinese context by Buddhism even more strongly than by the (less long-lived) School of Names (cf. p. 102 here).

3.9 Conclusion: Pānini and Confucius Confucius was the founder of an ethical tradition; and the actions whose goodness he was concerned with, were political in character. What is more, Confucius had no interest in theoretical ethics, or in the explication of an abstract concept of 'good action'. He was solely interested in practical ethics: He himself both explicated and exemplified the concept of 'gentleman', i.e. the one whose actions are good, and he invited his followers to imitate and, if possible, to surpass his example. A morally inferior person, however gifted he might be intellectually, could not possibly qualify as a good Confucian. This last observation shows that Confucianism is indeed not an intellectual discipline in the Western sense of the word. We are not used to thinking that

120

UNIVERSAL HISTORY OF LINGUISTICS

eventual flaws in a person's character necessarily disqualify him as a good phil­ osopher or a good scientist. In fact, such character traits as lust for power or opportunism seem to be the rule rather than the exception in the academic world. Yet it would be wrong to conclude from the practical nature of Con­ fucianism that it is not an intellectual undertaking. Like other such undertakings, it can be developed, in the first place by detecting contradictions and eliminating them. It is just the peculiarity of Confucianism, or a consequence of its practical nature, that it detects contradictions not just within the theory but also between the theory and the everyday life or the theoretician's everyday actions. (The dividing line between theory and practice is obliterated in the process, thus giving rise — precisely — to the notion of 'practical but intellectual' undertaking.) To put it a little more formally, it is not just the case that if the theory contains a statement like 'Op' (= 'One ought to see to it that p'), then it should contain no statement 'O ¬p' (cf. GTM:1Q.2). It is also the case that the theoretician(-cum-practician) endows 'p' with a real (political) content and endorses 'Op'. If the political situation is truthfully described by , _ | p \ he has detected a contradiction which he is committed to eliminate, this time not by changing the theory, but by changing, or trying to change, the situation in the world so that it becomes p. And if he fails to do so, he has detected a contradiction between his theory and his own actions. The 'actionist' framework remains the same also in case the situation is p, because then the theoretician-cum-practician is committed to preventing it from becoming It may be added that the Confucian combination of ethics and politics is not just different from, but also better than its customary Western counterpart. Any such political theory must be deemed a failure which is totally unconcerned about the process by which politicians are actually recruited, or does not devote its energies to (an attempt at) educating responsible and altruistic politicians. For my immediate purpose here, it is important to point out that within his own tradition Confucius occupies a place identical with the one that Pānini does within his. Each of them is both the oldest and the best, which does not mean, of course, that they have been entirely without predecessors. It is not clear whether we have any right to say that in both traditions progress could (and should) have taken place. What is certain, in any case, is that it did not. A scholar is likely to have an inflated view of the importance of his own discipline. I have tried to keep this in mind, when writing the present chapter. Although I have tried to gather together all the facts about the history of Chinese thinking which I find both relevant and interesting from the linguistic point of view, I have not wished to exaggerate the importance of this particular point of view.

CHINA

121

Hansen (1983) fails to be equally cautious. As he sees it, "Chinese philosophers from the beginning in effect acknowledge that their work is the study of language alone" (pp. 119-120).10 In holding this surprising view, however, Hansen is quite alone. According to the same logic, when a social scientist studies unemployment or a physicist studies atoms, they are merely studying the words unemployment and atom, respectively.

UNIVERSAL HISTORY OF LINGUISTICS

122

Notes 1)

The style of this passage makes Dawson (1981:56-57) assume that it is a later interpolation. Yet he adds that it is fully in keeping with Confucius' general doctrine.

2)

In CLT.3.2, 3.7.3 I define 'rational action' as elimination of contradiction and 'rational explanation' as viewing actions as rational. The way Confucius implements the 'rectification of names' doctrine provides a striking illustration of these two notions.

3)

On the one hand, there is reason to believe that the aristocracy of Confucius' own time was less benevolent towards the common people than the Chou aristocracy had been (cf. Creel 1953:22). On the other, criticism of oppressive regime and of hereditary priviledges was more common in Confucius' time. So there was a multi­ plicity of moral standards, and Confucius seized upon one of them and developed it further.

4)

Recall also that, as the passage VI, 25 shows, Confucius did not limit his 'rectifi­ cation' to moral or ethical issues.

5)

This extreme faith in the transformative powers of language calls to mind Rimbaud's 'linguistic alchemy'. In the prose poem "Vierge folle" he makes 'the crazy virgin', i.e. Verlaine, ask herself whether the 'infernal husband', i.e. Rimbaud, may have secrets to change the life: "Il a peut-être des secrets pour changer la vieV (emphasis in the original). As explained in "Alchimie du verbe", this transformation would have been brought about by a revolution of language: "J'inventai la couleur des voyelles! ... Je réglai la forme et le mouvement de chaque consonne, et, avec des rythmes instinctifs, je meflattaid'inventer un verbe poétique accessible, un jour ou l'autre, à tous les sens. ... J'écrivais des silences, des nuits, je notais l'inexprimable."

6)

Yet a very similar criticism of Chuang Tzu has been made along Neo-Taoist lines. Kuo Hsiang (d. 312 A.D.), "the best of all commentators on Chuang Tzu", com­ plained that Chuang Tzu only knew the world process of transformation, but did not identify with it; he only knew about the mind in the state of inertia, but his mind was not in such a state. This is why he spoke as he did, namely with too much verbosity (cf. Fung Yu-lan 1953:173-174). The implication is that he should not have spoken at all.

7)

This passage seems to have been overlooked by Miller (1975), since he claims that "no word in the Chinese language was thought to be ... in any way the result of an arbitrary agreement on the part of the society employing it" (p. 1217).

CHINA

123

8)

It may be added that Creel (1953:152) distinguishes between Hsün Tzu's 'authoritar­ ianism' and the subsequent 'totalitarianism' of the Legalist School; cf. 3.7.

9)

If Kung-sun Lung made one think of Rimbaud, Neo-Confucianism makes one think of Baudelaire's "Correspondances": "Comme de longs échos qui de loin se confondent... les parfums, les couleurs et les sons se répondent."

10) He goes on to add, however, that since the Han dynasty linguistic philosophy was "discredited".

4. Arabia

For a universal history of linguistics, the linguistic tradition in Arabia possesses considerable interest. Curiously enough, the Western linguistic theory has been almost completely unaware of the insights that Arab linguists have contributed to the field. This is owing to the fact that until recently even the Western experts, hampered by some sort of 'Indo-European' bias, have been unable to assess Arab linguistics at its true value. In this respect the situation of Arab linguistics is rather different from that of Indian linguistics. The high value of the latter has always been recognized in Western expert circles, even if this recognition has not pen­ etrated into the consciousness of the average Western linguist.1

4.1 The Cultural Context of Arab Linguistics Prior to the 7th century A.D., it is difficult to speak of 'Arab Culture' in any very definite sense of this word. It was only with the advent of Muhammad (d. 632) and of the Islamic religion that the history of the Arab culture began, properly speak­ ing. Having conquered Egypt, Syria, and Persia around 650, the Arabs became a world power. In particular, the conquest of Alexandria made them, first unwit­ tingly, the heirs to the Greco-Roman civilization. From that time onwards, 'Arab' becomes synonymous with 'Islamic'. It denotes no longer a geographic or ethnic entity, but rather a linguistic and cultural one. The irrelevance of the ancient national borders is evident from the fact that "after 750 almost all the chief luminaries of Islam were of Persian origin: the greatest grammarian, Sībawaihi (d.  793), the greatest philosopher, Ibn Sīnā [= Avicenna] (d. 1037), the greatest physician al-Rāzī (d.  925), and the greatest theologician, al-Ghazâlï (d. 1111)" (Fakhry 1970:49-50). In the present context I need not outline the entire history of the Arab culture, for the following reason. As will be shown in 4.2, Sïbawaihi's grammar determined in a more or less definitive way the nature of all subsequent Arab linguistics. Therefore, after 793, Arab linguistics underwent practically no outside influence. At most, it needs to be explained how this kind of near-inertia was

126

UNIVERSAL HISTORY OF LINGUISTICS

possible. It might be suggested that the Arab religious attitude which lays such an emphasis on the reverence for authority and dogma created conditions conducive to transferring the same reverence to non-religious domains as well. In any case, the substantive question concerns the pre-793 influences, or influences that Sïbawaihi might have undergone. From this, an important consequence follows immediately. Traditionally, it is said that significant Arab philosophy began with al-Kindï (d.  866), and was continued, most prominently, by al-Fārābī (d. 950), Ibn Sīnā, and Ibn Rushd (= Averroes) (d. 1198), all of whom, apart from professing the Islamic faith, qualify as Aristotelians. However, because this philosophical tradition could not have influenced Sïbawaihi, it is irrelevant for my purpose. Of course, this would not be so, if any of the above-mentioned philosophers had presented a theory of language that would have gone beyond what Aristotle had said already (cf. here 5.1.2). As far as I am aware, however, this is not the case. Even if the Arab Aristotelianism, based on translations from Greek via Syriac, had no impact on Arab linguistics, it is still possible that the latter was created under some sort of more general Hellenistic influence. In the Western circles it has indeed been customary to assume that all elements of the Arab science were originally borrowed from elsewhere, mainly from Greece, and to a much lesser extent from India, de Boer's (1901:17) view is quite representative: Nicht indischer Phantasie, sondern griechischen Geistes bedurfte es dazu, das Nachdenken auf die Erkenntniss des Wirklichen zu richten. ...Überall, wo es sich nicht um blosses Aufzāhlen oder zufalliges Zusammenreihen handelt, sondern nach sachlichen oder logischen Gesichtspunkten eine Anordnung des Mannigfaltigen versucht wird, darf mit Wahrscheinlichkeit auf griechischen Einfluss geschlossen werden. More recently Versteegh (1977) has taken great pains to show that since its beginning the Arab linguistic tradition was under the Hellenistic influence. I find this implausible, for a very simple reason. The major Hellenistic work on syntax is Apollonius Dyscolus' Peri suntaxeōs. I shall analyze this work rather extensively in due course (cf. 5.2.3). At this point it is sufficient to say that Sībawaihi's syntax is in decisive respects dissimilar to Apollonius'. Moreover, Aristotle's way of reducing all sentence-types to a canonical form proved to be quite influential in the West (cf. 5.3.4 and 5.4.2); but no trace of it can be found in Sïbawaihi's syntax. Now, if  and  are dissimilar, there is not much point in arguing for A's influence upon  (or vice versa). Should there have been such an influence, the most one can say is that  has successfully warded it off.

ARABIA

127

For instance, Versteegh (1977) tries to strengthen his case by pointing out that in their grammatical examples the Greeks and the Arabs used words meaning 'man', 'horse', and 'to hit'. But what does this prove? Let us suppose that this is not just a coincidence, but a case of the Arabs imitating the Greeks. Since the Arabs created, nevertheless, a grammatical theory quite different from the Greek model, we see that they were indeed able to successfully ward off the Greek influence. Thus, instead of diminishing the originality of the Arab grammarians, Versteegh's (1977) argument ends up enhancing it. The tradition that views Arab linguistics as an off-spring of Hellenistic linguistics has been vehemently attacked by Carter (1968:51-79). He gathers an impressive array of evidence to refute the 'dogma of Hellenism' (although, to be sure, he could have just pointed to the dissimilarity between Arab linguistics and Hellenistic linguistics). I of course agree with his general conclusion. So we are left with the alternative of directing our attention to some nonHellenistic influences that might have been effective prior to 793. An obvious move is to look for some sort of indigenous Arab philosophy. Most often, the existence of such a philosophy has simply been denied. Corbin (1964), however, has devoted a large-scale study to this topic, i.e. to a non-Hellenistic kind of thinking which has mainly been preserved within the Shï'ite tradition: Là où la recherche philosophique (tahqfq) fut 'chez elle' en Islam, ce fut là où l'on réfléchit sur le fait fondamental de la prophétie et de la Révélation prophétique, avec les problèmes et la situation herméneutiques que ce fait fondamental implique (p. 8). La tâche première et dernière est de comprendre le sens vrai de ce Livre [i.e. le Qorân] (p. 14). In other words, the only indigenous type of Arab philosophy is inseparable from Islamic theology, which in turn means the explication of the Koran (and contains several schools ranging from 'scholasticism' to mysticism). This kind of thinking might conceivably be of relevance for my topic; but in fact it is not. First, Sïbawaihi's grammar has no conceptual link with the Islam (cf. 4.2). Second, from the linguistic point of view, the Islamic philosophy-cum-theology possesses no intrinsic interest. To be sure, it deals with such questions as the created/temporal vs. non-created/eternal nature of the Koran qua God's word (cf. Fakhry 1970:7679), and in the present context it might have been justifiable to pay more attention to such issues, if this had been all that the Arabs have to offer in the field of linguistic theory.2

128

UNIVERSAL HISTORY OF LINGUISTICS

Must we conclude, then, that Arab linguistics was created on an entirely autonomous foundation, without any sort of extraneous model? Not necessarily. Islamic jurisprudence had to develop very early, i.e. already before 700, given that the Koran, although the basis for deciding legal issues, could offer no direct advice about how to govern a multinational empire. From the 10th century onwards, Arab linguists explicitly recognized the broad analogy between grammar and law, or more precisely between grammatical science and jurisprudence. (This analogy can be grasped also in a non-Islamic context; cf. G7M:206-208, CLT: 131-134). It is the merit of Carter (1968, Chap. 3) to have collected the evidence, consisting mainly of terminological parallelisms, that shows Sïbawaihi to have been influenced by the study of the Islamic law. It remains true, however, that analogous results can be achieved without any overt or covert reliance upon the concept-formation in jurisprudence, as shown by the comparison with modern Western grammar-concep­ tions (cf. 4.4). Thus Sïbawaihi's relation to jurisprudence is very much like Pânini's relation to the science of ritual (cf. p. 78). Versteegh (1980) admits that Arab linguistics may have been influenced by Islamic jurisprudence. Consistent with his general pro-Hellenistic attitude, however, he argues that the central notion of qiyās, i.e. the process of applying a general rule to particular cases, was borrowed from Hellenistic jurisprudence. This argument is far from compelling. Laws have been created and applied practically everywhere. It is hard to believe that the Arabs could not have achieved this without help from others. The 'golden era' of the Arab civilization lasted from  700 to  1100. The sack of Baghdad by the Mongols in 1258 was a decisive blow, and it is question­ able whether the Arab world has ever recovered from it. Still, the grammatical tradition never ceased completely.

4.2 General Characteristics of the Arab Linguistic Tradition Sïbawaihi's grammar, known simply as Al-Kitub (= 'The Book'), is the oldest extant grammar of Classical Arabic just like Pānini's AstödhyāyT'is the oldest extant grammar of Classical Sanskrit. And just like Pânini, Sïbawaihi is unanimously regarded not only as the founder of a great linguistic tradition, but also as its best representative. As noted in 2.2, Pānini could build upon a long tradition of linguistic analysis. By contrast, it is generally thought that Arab linguistics was literally created in Sïbawaihi's time, because, according to his own testimony, he had only

ARABIA

129

two genuine teachers, who, in addition, were his own contemporaries. Talmon (1982), however, contests this view, arguing that "Sïbawaihi founded his grammati­ cal system on the groundwork of a fairly advanced school of grammar" (p. 29), of a school, to be sure, which could not have been in existence for more than one hundred years. Considering the high level of Sïbawaihi's grammar, Talmon's view is certainly plausible.3 It is strengthened by constant reference, in historical records, to two grammatical schools, those of Kufa and Basra, Sïbawaihi being pictured as the one who consolidated Basra's supremacy.4 In his own time Sïbawaihi had to defend his views vigorously: "The Kitāb depicts, as it were, a running battle between Sïbawaihi and his opponents" (Carter 1968:35). In some hundred years, however, he had become an unquestioned auth­ ority, so much so that the Kitāb can be described as "unique in Arabic grammar in the conscious novelty of its approach" (p. 290). To fully appreciate this statement, one needs to know that linguistics was a rather fashionable discipline in the Arab world. A work on the most prominent linguists, written about 960, mentions 265 names, and in a similar work written some 600 years later the number has risen to 2,209 (Owens 1988:7). Sezgin (1984) provides information about 202 grammarians who lived before 1050. Pānini is credited with having established the canon of Classical Sanskrit and thus effectively 'freezing' it in the form in which it was to remain. The Koran is generally regarded as the basis of Classical Arabic. Gātje (1985:108) points out, however, that the real norm of Classical Arabic, which Muhammad tried, some­ times unsuccessfully, to imitate, was provided by pre-Islamic Bedouin poetry. In any case, Sïbawaihi played an important role in systematizing this state of language: "It is ironical that the very success with which Sïbawaihi achieved his purpose, namely to account for all the phenomena of Arabic, has resulted in the apparent fossilization of language" (Carter 1968:175). Pānini's example shows that there may indeed be a link between the 'fossilization' of language and the 'fossilization' of linguistics (although the choice of the words could be different). But Carter takes this link too much for granted, when he moves from the one to the other without any further discussion, and continues the passage quoted above: .. .once there is nothing left to explain, the science of grammar must lose its dynamism and degenerate into a mere procedural bickering. I would go further than Fleisch and say that this stagnation and loss of initiative began with the very completion of the Kitāb, and that, to borrow Fleisch's words, Arabic grammar from that time has been nothing but the 'exploitation de l'héritage grammatical' (p. 175-176).

130

UNIVERSAL HISTORY OF LINGUISTICS

It would be just false to claim that Pānini's successors could in no way develop their own 'grammatical heritage'. In the same vein, Carter's above assessment seems to be unnecessarily gloomy. Owens (1988), for instance, claims that "with Jurjânï [d. 1078] Arabic grammatical theory reached its zenith" (p. 248), although he is quick to add that "very little of the material that Jurjânï used had not already been defined and used by those before him, and few of his ideas are really original with him" (p. 250).5 This generally accepted view of the history of Arab linguistics has been contested, or so it seems, by Bohas et al (1990), who claim that "Sïbawayhi's syntactic system is, on the whole, founded on a quite different approach from that of the classical grammarians" (p. 5); and Ibn al-Sarrāj (d. 928) is mentioned as the first representative of the 'classical' tradition. However, there has been no theoretical 'break' between Sïbawaihi and his successors, as is made clear by Bohas et al (1990:48): Sïbawayhi's morpho-phonology and morpho-syntax (i.e., notably, all that concerns the more formal aspects of the distribution of case markers) are basically identical to that of the later grammarians ... [who] were more directly interested in these two formal aspects of grammatical analysis.6 As will be seen in 4.5., Sïbawaihi described (or rather, attempted to describe) linguistic facts at two levels, namely the levels of form and of action (or 'operation'). His successors retained only the level of form, very much as Sïbawaihi had conceived of it. Consequently, what occurred in Arab grammatical theory after Sïbawaihi, was not a qualitative change, but a narrowing-down (compensated by increasing systematization). All in all, the linguistic traditions in India and in Arabia are remarkably similar, with the best linguist standing right in the beginning of the recorded history in both cases. And in both cases the implications for the notion of 'scientific progress' are equally momentous. So we are fully justified to ask, together with Jahn (1969 [1900]:XV), the following rhetorical question: Ist es nicht wunderbar, dass der erste aller uns bekannten Grammatiker den Genius der Sprache am tiefsten ... erfasst hat, so dass die Spāteren ihn zwar verstāndlicher gemacht, aber eher verflacht als vertieft haben? In addition to such basic similarities between Pānini and Sïbawaihi, there are a few minor ones. Just like Pānini, Sïbawaihi is for the uninitiated very difficult to understand, or "grössenteils unverstāndlich", as Jahn (p. v) put it. As a conse­ quence, he should be approached with the aid of commentaries like that of Sïrafï

ARABIA

131

the Elder (d. 949), who may in this respect be compared to Patañjali. Moreover, both in India and in Arabia linguistic study was at the origin motivated by religious considerations, i.e. the elucidation of the Vedas and of the Koran, respectively. However, there is no religious link between Pānini and the Vedas, and even the linguistic link is rather tenuous (cf. p. 12). Similarly Sïbawaihi has no systematic interest in the Koran. To be sure, he uses examples from the Koran to illustrate his grammatical points, but he also uses examples from poetry as well as self-invented examples (cf. Carter 1968:92). Thus, although Fakhry's (1970:2) following statement sounds plausible, it is misleading, if not downright false: Grammar, rhetoric, and the allied studies were developed during the first two centuries of the Muslim era, chiefly as a means of adequately inter­ preting or justifying the linguistic usage of the Koran and the Traditions. Finally, Sïbawaihi resembles Pānini in that also his grammar was taken to be an intellectual achievement which transcended the bounds of linguistics and acquired general validity. We have already seen that Sïbawaihi may have been influenced by Islamic jurisprudence. But he in turn influenced the way in which Islamic scholars interpreted the oral traditions (= hadith) which supposedly go back to Muhammad's sayings as memorized by his entourage. It was commonly thought that "on apprend grâce au Kitāb de Sïbawaihi la réflexion et la critique" (Langhade 1985:108). After enumerating all these similarities between Pānini and Sïbawaihi, it is good to point out that there are differences as well between them. First and foremost there is the difference of style. Pānini's rules are compressed to the utmost and packed with auxiliary symbols; he does not offer a single example sentence to illustrate his rules. By contrast, Sïbawaihi writes in complicated prose sentences, uses no auxiliary symbols, and gives thousands of examples. Then there is the difference of subject matter. Pānini concentrates on morphology and pho­ nology, and is content to sketch out his syntax and semantics. Sïbawaihi devotes his greatest attention to syntax, treats of morphology and phonology too rather exten­ sively, and neglects semantics almost entirely. Additional differences will emerge, as we proceed. Yet it is clear that from the bird's-eye perspective, differences are by far outweighed by similarities.

132

UNIVERSAL HISTORY OF LINGUISTICS

4.3 Central Aspects of the Arab Syntactic Theory In the preceding section it already became evident that, despite its history of more than one thousand years, the Arab syntactic theory has remained remarkably homogeneous, adhering to the framework constructed by Sībawaihi. Therefore I shall present it as a quasi-synchronic entity, without any systematic attempt to differentiate between contributions made at different points in time.7 My main source here is Owens (1988). The Arab syntactic theory centers around a problem of grammatical analysis which, because of its ubiquitousness, might be called the 'Grammarian's dilemma'. It focusses upon the conditions that are necessarily imposed on, and restrict the usefulness of the well-known 'clear case principle'. Any grammarian is bound to start with a 'clear case', i.e. something that he regards as indubitable or paradig­ matic. From this case he then goes on to generalize to other cases. That is, the description of this one case is ('analogically') extended by him to other cases. As he proceeds with his generalization, he comes to cases which are covered by it less and less well. He then realizes (or someone else does) that if he had started with a different 'clear case', he could have handled easily what he now finds problemati­ cal. But then, most probably, his original clear case would have turned out to be problematical. The basic phenomenon which Arab grammarians were struck by, and which they found to be in need of explanation, was nominal inflection. Classical Arabic has three cases, which have traditionally been translated as 'nominative', 'accus­ ative', and 'genitive'. In the singular of masculin nouns, for instance, the variation between case endings is -u, -a, -i. It can be explained by a corresponding variation among preceding words: a preceding verb 'causes' the following noun to be in the nominative or in the accusative, whereas a preceding preposition causes the following noun to be in the genitive. Hence, the basic explanatory schema in Arab syntax looks like this:

I shall follow Owens (1988) and call A and  'governor' and 'governed', respectively. On one interpretation, 'A governs 2?' is synonymous with ' 5 is

ARABIA

133

dependent on A'. Thus, according to Owens (1988), we have to do here with an application of dependency grammar8 As noted above, the clear cases are V {N-nom N-acc} and Prep N-gen. From there, the basic schema   is stepwise generalized to other, less clear cases. First, there are the residual cases involving nominal inflexion:

non-V

{N-nom

N-acc} and non-Prep N-gen Second, there are the cases involving non-nominal, i.e. verbal inflection: Particle V-mode. Third, there are the cases involving no inflection:   - Φ . Fourth, there are the cases where only the effect, i.e. an inflected B, is present, with the consequence that a covert cause, or A, has to be postulated:

Φ (=A)  - {case mode}. Let us explain the preceding paragraph a little. Arab linguists distinguished between three word-classes: nouns (N), verbs (V), and particles (P). Nouns include also adjectives (A), pronouns (Pr), and both active and passive participles (AP, PP). Nouns are primarily governed, but they may secondarily be governors (of other nouns). Verbs are primarily governors, but they may secondarily be governed in mode (indicative, subjunctive, or jussive). Particles are divided into governors and non-governors; the former are divided into those governing nouns (i.e. prep­ ositions, or Prep) and those governing verbs; particles are never governed. Second­ ary cases are explained as results of generalizations from primary cases. Thus nouns qua governors and verbs qua governed are seen to be relevantly similar to verbs/prepositions and nouns, respectively. 'Analogical extension' (= qiyās) is the central methodological notion of Arab grammar (cf. Baalbaki 1979:18; Sezgin 1984:28-31; Bohas et al 1990:2226). There is reason to distinguish between atheoretical (or pretheoretical) and theoretical uses of qiyās. Its atheoretical use consists in simply applying inflectional paradigms to ever new words, a process which is bound to reveal the openness of language. As Ibn Jinnï (d. 1002) put it: "Ne vois-tu pas que tu n'as jamais entendu, pas plus toi qu'un autre, tous les noms du paradigme facil or mafcuH Tu en as entendu certains et as procédé par analogie pour les autres" (Langhade 1985:111). The theoretical use of qiyās, in turn, is illustrated by the process of interpreting all the various constructions enumerated above as instances of one and the same basic

134

UNIVERSAL HISTORY OF LINGUISTICS

'governor—governed' construction. In this latter sense, then, 'analogical extension' does not mean generating new forms, but seeing apparently different forms as similar. After this summary overview, I shall present the data in more detail. I shall first deal with the two basic sentence-types of Classical Arabic, i.e. verbal sentence and nominal sentence. Next, I shall deal with subsentential constructions. I shall conclude with a discussion of suprasentential and intersentential phenomena. The basic type of construction, as shown above, has two important proper­ ties: it is binary, and it exhibits the relation of governance. The simplest type of verbal sentence, one of the two basic sentence-types of Classical Arabic, has both of these properties:

The constituents (or 'positions') of a simple verbal sentence are ficl ('action', also 'verb') and fācil ('agent'), filled here by qāma and zaydun/rajulun,respectively. In (1) a verb governs a N-nom. The requirement of binariness has the consequence that the 'objects' (= mafcūl), i.e. accusative complements, of the verb, as they occur e.g. in (2) and (3), are regarded as 'optional'.

In (2) and (3) a verb governs a N-nom and either one or two N-acc's. The 'op­ tionally' of objects is justified by distinguishing between three different cases (cf. Owens 1988:174). First, if the speaker just concentrates on the idea of action, potential objects are irrelevant, and therefore no N-acc should be present. Second, if the speaker has an object in mind, but it is evident from the context, no N-acc needs to be present. Third, if the speaker has an object in mind, and it is not

ARABIA

135

evident from the context, a N-acc must be present. — Notice that in the Western tradition there has been the opposite tendency to regard V and N-acc as constituting a whole distinct from N-nom. Governance relations analogous to those in (2) are assumed also in the following sentence, although -tu and -hu are uninflected:

(4) darab-tu-hu 'I'he' 'I hit him' Here we have a generalization from inflection to non-inflection, based on the substitution of the non-inflected for the inflected. The number of 'direct' objects may be increased by means of causativization. In addition to direct objects exemplified above, seven other types of objects were posited by Arab linguists. A verbal sentence may also contain non-accusative complements: (5) marar-tu

bi

'passed'I''by' 'I passed Zayd'

zayd-in gen

The construction bi zaydin is equivalent to a direct object, as is shown by the fact that if a second co-ordinate, meaning 'and Amr', is added, it can be either in the genitive (wa camrin) or in the accusative (wa camran). Here, however, a problem in the Arab notion of governance becomes evident. Governance relations obtain only between single words (or morphemes). Now, in (5) it is clear that -tu is governed by morar-, and zaydin is governed by bi. But, although bi zaydin admit­ tedly has the status of a direct object, it cannot be connected with marartu. This follows from the fact that bi qua particle cannot be governed; and Arab linguists just could not accept the idea of many-word units (e.g. prepositional phrases like  zaydin) being governed. Accordingly, sentences like (5) receive no coherent treatment. On one interpretation, bi zaydin is like a direct object, and yet cannot be governed by the verb. On another, zaydin alone is like a direct object, bi being somehow 'part' of the verb; but this idea is left at the level of mere suggestion (for discussion, see Owens 1988:80-81, 148-149, 176-177, 232). The situation is further complicated by the fact that outside the context of governance, i.e. when

136

UNIVERSAL HISTORY OF LINGUISTICS

dealing with the notions of substitution and constituency, Arab linguists were perfectly willing to accept the existence of many-word units (cf. 4.4). In any case, the Arab notion of governance works reasonably well in connection with verbal sentences. There is a standard word order such that the direction of governance (or 'causation') goes — in the Western orthography — from left to right, with governed units being contiguous to governing ones. Exceptions to this rule are handled by resorting to pragmatic or stylistic considerations. A unit may govern more than one unit; but it can be governed only by one unit. The Arab notion of governance agrees with the common-sense view that insofar as the notion of cause can be explicated in terms of 'conditions', it is analogous to sufficient condition, and not to necessary condition (cf. CLT:28-29). If a preposition precedes, a N-gen will follow (although it may also follow a Nnom); therefore the former governs (or 'causes') the latter. It is true, that only if there is a N-gen, there is a preposition; but this does not make the former the governor (or 'cause') of the latter. In light of the examples given so far, Classical Arabic appears to be a straightforward VSO language. This view must be qualified, however, given the existence of the nominal sentence, the other basic sentence-type of Classical Arabic. The constituents (or 'positions') of a nominal sentence are mubtada' ('topic') and khabar ('comment'). The simplest type of nominal sentence is exemplified by the following: (6) zaydun rajulun Topic Comment 'Zayd is a man' In (6) the comment position could also be filled by A (tawttun, 'tall'), AP (muntaliqun, 'leaving'), PP (madrubun, 'hit'), P + N-gen (ft l-bayti, 'in the house'), or S (cf. below). Interestingly, also the following sentence is regarded as an instance of nominal sentence: (7) zaydun qāma This seems nothing but (1) with a different word order. At first glance at least, it is hard to see why, just because of a change in word order, the 'agent' and the 'action' of (1) should become, respectively, the 'topic' and the 'comment' of (7), with the consequence that (1) and (7) are taken to belong to basically different

ARABIA

137

sentence-types. In effect, Ayoub & Bohas (1983) argue that for Western scholars this notion of nominal sentence has been the biggest single obstacle to under­ standing the Arab linguistic tradition. Yet this notion can be made fully comprehen­ sible although, to be sure, there are alternative ways of analyzing the same data. Let us, first of all, make the initial situation clear. There are two paradig­ matic structures (where all the If s are in the nominative): (8) V + N (= verbal sentence) (9) N + N (= nominal sentence) Along comes a new structure (where, again, the Nis in the nominative): (10) N + V Now the question arises: Should (10) be assigned to (8) or to (9)? (Or should it, perhaps, be regarded as a sentence-type of its own?) Arab linguists have chosen to assign (10) to (9). The first point to be made is that this choice entails a preference for substitution over permutation. One can get (10) either by permuting V and N in (8) or by substituting V for the second N in (9). The choice of the latter alternative is facilitated by the relatively fixed word order of Classical Arabic (which already became evident in the discussion of governance). We shall see that in other ways too substitution plays an important role in Arab linguistics. Still, it might be argued that the transition from (9) to (10) — or vice versa — is rather abrupt. It would be smoother, if there were an intermediate sentencetype between (9), exemplified by (6), and (10), exemplified by (7). Such a sen­ tence-type does indeed exist, and it is exemplified by the following: (11) zaydun daraba-hu rajulun 'As for Zayd, a man hit him' It is a characteristic property of this sentence-type that the comment contains a 'resumptive' pronoun referring back to the topic (which, to repeat, must be in the nominative). This means in fact that the comment constitutes an integral sentence. For instance, the comment of (11) is a sentence analogous to (4). The same sentence-type is further exemplified by the following:

138

UNIVERSAL HISTORY OF LINGUISTICS

(12) zaydun 'actâ-hu 1-rajulu kitāban 'As for Zayd, the man gave him a boot (13) zaydun qāma ghulāmu-hu 'As for Zayd, his servant stood up' The resumptive (and uninflected) pronoun -hu stands for an 'object'-accusative in (11), a 'recipient'-accusative in (12), and a genitive ('possessor') in (13). The sentence-type common to (11), (12), and (13) could be represented as follows: (14) Ni + V + {fti.Nj} Topic

Comment

N{ is in the nominative, the preferential Prx stands for the accusative or the genitive, Nj may be a pronoun, and there may be other Ws in addition to Nj Now let us assume that the category-membership of (14) is decided before that of (10). Should (14) be assigned to (8) or to (9)? There are reasons supporting both alternat­ ives. Like (8), (14) contains a verb, and could thus be labelled a 'verbal sentence' (more precisely: sentence-type). Like (9), (14) begins with a noun, i.e. a topic, about which something is said; so it could also be labelled a 'nominal sentence'. Arab linguists have opted for the second alternative. They could have chosen otherwise (and those working within other traditions have done so), but this is certainly a rational choice. Labelling (10) too as a 'nominal sentence' may have seemed less rational; but now it can be made comprehensible. (14) is in an obvious sense intermediate between (9) and (10). We just saw that it is quite natural to go from (9) to (14). But there is a striking similarity between (14) and (10): Both begin with a N-nom and contain a verb. So it is quite natural to go from (14) to (10). Thus (9), (14), and (10) all qualify as 'nominal sentences'. As a consequence, (7) should be translated as 'As for Zayd, he stood up'. It is not synonymous with (1). It would have been awkward to generalize directly from (9) to (10). But, starting from (9), we have reached (10) by means of two successive generalizations, or analogical extensions, i.e. (9) → (14) and (14) → (10). To better show the unity of (14) and (10), one might wish to postulate in (10) a resumptive pronoun in the

ARABIA

139

nominative; it is covert, but otherwise comparable to the one occurring overtly in (14). This suggestion was indeed made by the Arab grammarians (cf. Levin 1985:120-121): In (7) the word qāma contains and governs a covert pronoun in the nominative; therefore it cannot govern zaydun; therefore zaydun must be the topic. According to Ayoub & Bohas (1983:38) the same treatment was extended even to (9), to show its unity with (14) and (10): In (6) the word rajulun is followed by a covert huwa ( = 'he'). — The preceding discussion elucidates the 'Grammarian's Dilemma', and one way of solving it by means of qiyās. There still remains one sentence-type to be discussed: (15)

zaydan daraba-hu rajulun 'A man hit Zayd'

This is a hybrid construction. Like in the topic—comment construction of (11), there is a resumptive pronoun. But unlike in (11), the pre-verbal noun is in the accusative, and therefore cannot qualify as a topic. The Arab solution was to postulate a covert verb which would explain the pre-verbal N-acc, thus:

Now I can give an overview of the sentence-types containing (as we would say) a finite verb, a subject, and an object:

A), D), and E) are exemplified by (2), (15), and (11), respectively. F) is just an extension of (7). B) and C) are permutations of A), based on the principle 'important item first'. T stands for 'topic'. Square brackets indicate a covert entity. A), B), C), and D) are verbal sentences, whereas E) and F) are nominal sentences. Arab linguists would not have accepted viewing E) and F) as somehow

140

UNIVERSAL HISTORY OF LINGUISTICS

'derived' from A). Similarly D), though a verbal sentence, is not 'derived' from A). As for A), B), and C), Arab linguists were unanimous in regarding A) as primary ('ast) and the others as secondary (farc), or as 'derived' from A) (cf. Owens 1988:217-218). From the structural point of view, Arab linguists found the notions of verbal sentence and nominal sentence unproblematic. They neatly illustrate the basic binary sentence-structure ('isnād):

From the point of view of governance, however, the notion of nominal sentence is problematic (as indicated by the dotted line between topic and com­ ment). The basic type is (9). But since (9) contains only nouns, and nouns are primarily non-governors, it follows that (9) is prima facie a sentence-type without any governor. Several solutions were offered, for instance, that the topic (some­ how) governs the comment, or that they (somehow) govern each other. The generally accepted solution is more complex. It starts from construc­ tions of the type 'X + N + N', where X is a member of a small class of verbs or particles, and N + Nis a structure which, without X, would be a nominal sentence. Depending on the identity of X, there are the constructions 'X + N-nom + N-acc', 'X 4- N-acc N-acc', and 'X 4- N-acc + N-nom'; for instance:

It is commonly said that words likekā'enter' a nominal sentence and change it, so that what was the topic becomes a mere noun (ism). These 'new' words move into the sentence-initial position and govern the two nouns each in their own way. Now (simple) cases like (6) are explained on the analogy of (complex) cases like (17). It is assumed that in (6) there is a covert sentence-initial governor such that it 'causes' both nouns to be in the nominative. This is logical enough, although somewhat artificial. Difficulties arise, however, when the nature of this governor

ARABIA

141

has to be specified more narrowly. The governor is said to be "the fact of its [the topic's] being initial" (Owens 1988:55). This seems to be a category mistake. The overt governors are words, or units, like kâna. But the covert governor is a fact, namely the fact that a unit is such and such. These artificial solutions are forced upon Arab linguists by their insistent wish to achieve a maximally general solution to the Grammarian's Dilemma. Above, the verbal sentence was taken to be the clear case.9 Since it is of the tvpe a covert governor is postulated for the nominal sentence, resulting in Many grammarians, however, simply refused to seek a common explanation for the two basic sentence-types. In fact, Bohas et ah (1990:64-72) note that there was a constant tension between the 'government model', applicable to verbal sentences, and the 'predicative model', applicable to nominal sentences. Next I shall turn my attention to subsentential phenomena. The imperfective aspect (or the present tense) of the verb is inflected in indicative, subjunctive, and jussive. The two latter modes are governed straightforwardly enough by sentenceinitial particles, although in some cases the governor may be covert. Although the indicative too is an inflectional form, there is no very satisfactory account for its governance. Because verbs are primarily governors, their being governed (= inflected for mode) needs to be explained. This is done by pointing out the analogy between Vand AP. On the one hand, AP resembles V because it may govern N-nom and Nacc. On the other, AP resembles N, because it is governed (= inflected for case). AP is supposed somehow to 'confer' this quality on V, replacing case by mode in the process (Owens 1988:67). Now, AP's capacity to govern was mentioned above. AP is said to possess this capacity because of its resemblance with V (p. 203). That is, V in turn 'confers' this quality on AP. The situation might be depicted as follows:

The well-known 'argument from analogy' conforms to the following schema: If X and Y are similar in the respect W, this may justify conceptualizing X

UNIVERSAL HISTORY OF LINGUISTICS

142

and  as being similar in the respect Z. (More precisely, if X and Yare A in W, and X is  in Z, there is reason to think that also Y is  in Z.) This is how we con­ cluded, for instance, that the formal similarity between (14) and (10) justifies the view that they are also similar in exhibiting the topic-comment structure. But the case at hand violates this principle. The similarity between V and AP is identical with the fact that they both have the quality of governing N-nom and N-acc. Take this quality away from AP, and it ceases to be similar to V. Therefore V cannot 'confer' this quality on AP, on the basis of this very quality (= the similarity between V and AP), On the other hand, it can be said that AP, on the basis of its similarity to V, confers on V the quality of being governed. (Ordinary nouns too have this quality, but cannot confer it on V, because they are dissimilar to V.) In other words, the notion of 'being governed' is generalized from case to mode. Thus we must distinguish between 'similarity' and 'generalization'. The former is a symmetric notion10 whereas the latter is an asymmetric one. The defect in the preceding account was that it did not keep these two notions apart. That a circular argument has indeed been made here, is demonstrated by Anghelescu's (1985:6) following characterization: "les divers éléments ... empruntent quelque chose l'un de [sic] l'autre s'ils se ressemblent (te verbe à l'imparfait emprunte du nom la flexion et ensuite ... lui rend d'autres choses en échange)." By contrast, Baalbaki (1979:17) points out that only asymmetrical or unidirectional cases of generalization (= qiyās) are logically acceptable. Next I shall consider the 'N + N-gen' construction. Arab linguists had no doubt that the first N is the governor. This choice was made almost inevitable by the fact that, as a rule, the governor precedes the governed. Yet this particular governance relation is hard to justify, because nouns are primarily non-governors. It was suggested that (18) and (19) are (semantically) similar, and that because of this similarity the preposition li in (19) confers its inherent capacity of governing on kitābu in (18). (18) (19)

kitābu zaydin 'Zayd's book' kitābun li zaydin 'a book to Zayd'

This interpretation does not solve the problem, but only shifts it. Let us grant that kitābu in (18) governs because li in (19) governs. But what about (19) itself? How is li related to kitābun and to the rest of the sentence? Here we meet the same difficulties that we met already in connection with (5). — It may be added

ARABIA

143

that the 'N + N-gen' construction was universally regarded as equivalent to a single noun (Owens 1988:159-160). I shall single out only one more subsentential construction, namely 'N + A', which exemplifies the more inclusive 'modifier—modified' construction. Morphologically and distributionally, this construction is unproblematic. A agrees with N and, like N+N-gen, A+N is equivalent to a single noun. It is, once again, in the context of governance that problems arise: It is not clear what is the governor of A. Three solutions were offered. First, A is governed by what N is governed. Second, A is governed by the fact of its following N. Third, A is not governed at all, but merely agrees with N The generally accepted solution was a combination of the last two (cf. Owens 1988:162-164). The second solution implies the problematical view, which we already met in connection with the governance in nominal sentences, that the governor is not a unit, but a fact. Concerning the interpretation of governance, there is a vacillation here between what might be called the 'causality view' (which I take to be the original one) and the 'sequence view'. According to the former, A's governing  means that A is inherently such as to bring about an (inflectional) change in B. According to the latter, A's governing  means that the change in  somehow results from the fact that  precedes B. The latter view is symmetrical in a sense in which the former is not: If the change in  results from A's preceding B, it results equally well from B's following A. The emergence of the 'sequence view', and the consequent confusion of the two views, is comprehensible, as the governor indeed precedes the governed. It should be recalled, however, that genuine causes too precede their effects, and yet causality is not just a matter of sequence. It might seem surprising that N is not entertained as the governor of A. This has to do with the Arab view of the fixity of the governance relation: a certain particle always governs a N-gen, and a certain verb always governs the same inflections of nouns (except when some of these are occasionally deleted). By contrast, in 'N + A' constructions the N may have now this, now that inflectional ending (and the fact that the A has the same ending is deemed immaterial). So there is no temptation to lump governance and agreement together. Finally, I shall briefly consider complex sentences and relations between sentences. These two topics coincide to a large extent because a complex sentence is customarily described by showing its relation to a simple one. For instance, (6) is changed into (17) by adding kā. Moreover, (11), (12), and (13) contain complete sentences as their comments. It might seem natural to view (11), (12), and (13) as resulting from a process of topicalization which is applied to the corre­ sponding simple sentences (2), (3), and (13').

144

UNIVERSAL HISTORY OF LINGUISTICS

(13') qāma ghulāmu zaydin We have seen, however, that Arab linguists chose to regard (11), (12), and (13) as exemplifications of a basic sentence-type, i.e. nominal sentence. Relative clauses function like adjectival modifiers. When the relative pronoun is not in the nominative, a resumptive pronoun occurs in the relative clause, and — if the antecedent noun is indefinite — the relative pronoun itself is dropped:11 (20) marartu bi 1-rajuli lladfal daraba-hu zaydun 'who' 'I passed the man whom Zayd hit' (21) marartu bi rajulin 0 daraba-hu zaydun 'I passed a man whom Zayd hit' A relative clause may also occur without an antecedent, for instance: (22) 0 'alladhī daraba-hu rajulun zaydun 'The one whom a man hit is Zayd' (22) is the result of 'focalization', a process recognized and described by Arab linguists (Owens 1988:86-87, 246-247). More precisely, (22) results from applying this process to (2). Relative clauses minus relative pronouns are called 'dependent sentences'. With the aid of relative pronouns and complementizers dependent sentences are made to function as agents or objects. The unity of dependent sentences is emphasized by noting that nothing can be moved out of them or into them (p. 79). I shall illustrate intersentential relations by means of active, passive, and causative sentences. The passive sentences corresponding to (2) and (3) are these: (23) (24)

duriba zaydun 'Zayd was hit' 'uctiya zaydun kitāban 'Zayd was given a book'

ARABIA

145

In general, the relation between active and passive is described in terms of change (naqll):deletion of N-nom, change from N-acc to N-nom, and change in verb form. The principle governing this series of changes is the retention of the basic binary sentence-structure ('isnād). In a verbal sentence this is 'Verb—Agent', and if the agent is deleted, it has to be replaced by a new one, i.e. one that is chosen from among the objects. With verbs requiring two objects, there is a certain latitude in choosing the agent-to-be. According to a minority opinion, represented by Sïbawaihi among others, active and passive are treated as equals, without any attempt to derive one from the other.12 For those espousing the notion of passivization, it was seen as a process opposite to causativization: the former decreases the number of objects whereas the latter increases it (Owens 1988:175185, 223-225). Compare (2) with (23), on the one hand, and with (25), on the other: (25)

'a-draba rajulun camran zaydan 'A man made Amr hit Zayd'

The facts may be summed up as follows:

4.4 Metatheoretical Implications Carter (1968:179-187) claimed that Sïbawaihi's descriptive technique in syntax is practically identical with Bloomfield-type 'immediate constituent' analysis (= 1 analysis). Bobzin (1983:102) suggested that Sïbawaihi and his successors should rather be seen as advocating a grammar conception similar to modern dependency grammar; and he himself undertook to show, more particularly, that Sïbawaihi operated with a notion equivalent to the 'valency of the verb'. Now it is the principal claim of Owens (1988) that the Arab syntactic theory is indeed a variant of dependency grammar. As can be seen from the discussion in the preceding section, I think he proves his point quite convincingly. Although Carter (1968) is anxious to emphasize the 1 aspect of Sïbawaihi's syntax, he cannot help implicitly admitting the existence of the dependency aspect as well. This can be seen from the

146

UNIVERSAL HISTORY OF LINGUISTICS

fact that he describes Sïbawaihi's 'formalism' as creating "a situation in which the effect of a word upon the one immediately following becomes peculiarly important to the analyst" (p. 188). This of course agrees perfectly with Owens' governance, or dependency, view of Sïbawaihi's syntax. It may be that at the time of writing his 1968 book Carter just was not yet acquainted with dependency grammar. Still, it is perfectly right to say that in addition to its preponderant depen­ dency aspect, Arab syntax also has an I aspect. This is evident from the role that substitution plays in it. We have already met this notion several times, either explicitly or implicitly. It is explicitly recognized that the Comment position (or 'slot') can be filled by units of varying size and category-membership, mainly N, P + Nf and S. Similarly the Agent and Object positions can be filled either by N's or by dependent sentences. (It must be pointed out that in the Arab context 'agent' and 'object' are formal notions; cf. below.) Moreover, N + A and N + N-gen are substitutable for simple N's. This could have led to an explicit postulation of NP, but it did not. Thus both the dependency view and the 1 view are present in Arab syntax (though not to an equal extent). I have tried to do justice to both aspects with my account of camal (p. 132), which combines construction (= I) and governance. Owens does not accept this interpretation. He points out that there are constructions (e.g. P + V-mode) where the governor and the governed do not form a constituent (pp. 76-81); and even when they do, a dependency analysis may make the I notions (e.g. NP) superfluous (pp. 148-149). However, among the cases adduced by Carter (1968:210) there are several, including the basic sentence-structure 'isnud itself, where the dependency view and the I view occur together: A governs B, and AB constitutes a superordinate unit. So maybe this should be considered as the paradigmatic case, with the clear understanding that deviations from it are bound to exist. It must be added, however, that the two viewpoints are not entirely re­ conciled in the Arab system, as revealed by our discussion of (5): The I view accepts the existence of prepositional phrases, whereas the dependency view does not, to its own disadvantage. Although I accept Owens' dependency interpretation of Arab syntax, I think his treatment of substitution/constituency calls for comments. What Ifindobjection­ able, is — to put it briefly — that he identifies position with function (p. 34), and separates position from substitution (p. 76). Now, it is a fact that primary instances of such 'functional' notions as 'agent' and 'object' are defined by the following conditions: i) they occur in the same sentence with a verb; ii) they have a character­ istic form (= N-nom and N-acc, respectively); iii) they occupy characteristic positions vis-à-vis the verb (= VSO); iv) the verb agrees with the agent, not with

ARABIA

147

the object. Position is only one among these four conditions that a function (= 'agent' or 'object') has to meet, and the least important one at that. So there is no reason to identify function with position (pace Bloomfield's definition quoted both by Carter and by Owens: "the positions in which a form occurs are its functions"). Secondary instances of agents and objects (e.g. dependent sentences) are in turn defined by the fact that they can be substituted for the primary instances in their respective positions. So there is no reason to separate position from substi­ tution. It is indeed hard to see what could be meant in this context by "independent [elsewhere: a priori] syntactic positions [i.e. Agent position and Object position] at which various substitutions take place". To be sure, a position can be defined independently of which units, if any, occur in it. (For instance, nothing occurs in the position between P and N-gen.) But here we are talking about Agent and Object positions. And it just does not make sense to talk about such positions unless some units, substitutable for one another, do occur in them and thus qualify as either agents or objects. After discussing the relation of Arab syntax to dependency grammar and 1 analysis, let us take up its relation to transformational grammar (= TG). Their analysis of nominal sentences leads Ayoub & Bohas (1983) to claim that Arab linguists operated with a notion of 'représentation abstraite' and were thus effec­ tively transformational grammarians avant la lettre. The argument employed by Ayoub & Bohas is, however, rather unconvincing. Because 'asl means neither 'temporally anterior form' nor 'semantic paraphrase', they conclude that it must mean 'abstract representation'. However, they omit to mention the most obvious alternative, namely that 'asl means 'most frequent', 'standard', or 'primary' (cf. here p. 140). When the VSO word order is characterized as 'asl vis-à-vis VOS or OVS (which arefarc), this just means that VSO is the standard word order. It does not mean that VOS and OVS must be derived from an 'abstract' VSO. Rather VSO, VOS and OVS are regarded as entities situated at the same level of abstractness (or concreteness). This criticism of Ayoub & Bohas (1983) agrees with the one made by Owens (1988:220 and 305, n. 268). In his Chapter 8 he adduces a great number of examples, mainly from (morpho-)phonology and morphology, to show that the y asl—farc distinction equals the 'primary/frequent vs. secondary /infrequent' distinc­ tion, and not the 'abstract vs. concrete' distinction. His own interpretation is, more narrowly, 'unmarked vs. marked'. This may look like another anachronism. But given the nebulous state of the markedness theory, this may be an accurate inter­ pretation, after all.

148

UNIVERSAL HISTORY OF LINGUISTICS

The 'transformational' relations that Arab linguists endorsed hold between actual or 'surface' sentences, e.g. (6) → (17), (2) → (22), (2) → (23), (2) → (25), and — as we just saw — pairs of sentences exemplifying VSO→VOS and VSO → OVS. Therefore Arab syntax was not 'transformational' in any modern sense. However, the authors of the Port-Royal grammar are generally regarded as early representatives of TG, although they dealt only with actual sentences, analyzing one complex sentence as a set of simple sentences (cf. here p. 267). In this sense Arab syntacticians too practised TG. Moreover, the use of 'covert governors' might be taken to indicate that in a few cases Arab linguists did after all accept some abstract level of representation. They themselves regarded e.g. (15) as resulting from a process of deletion, or ellipsis (hadhf), applied to (16). Now (16) is indeed 'ab­ stract' in the sense that it never occurs as such. In some other constructions too an 'understood' or covert element was postulated (cf. the examples given in Baalbaki 1979 and Bohas et al. 1990:60-63). However, covert particles triggering a change in V-mode would rather seem to resemble zero morphemes and thus not to presup­ pose a separate (abstract) level of representation. — These remarks should not obscure the fact that the descriptive technique most extensively used by Arab linguists is a variant of dependency grammar. Sïbawaihi's grammar-conception clearly differs from Pānini's. It may be recalled that Pānini's grammar is a derivational system which makes linguistic entities move from semantics to phonology via syntax and morphology. Sïbawaihi, by contrast, treats syntax and morphology-cum-phonology as separate levels, without trying in any clear sense to 'move downwards' from one to the other. In addition, his grammar lacks any systematic semantic component. For instance, his 'functional' notions of 'agent' and 'object' are defined purely formally (cf. p. 134). It will be recalled that for Pānini 'agent' and 'patient' are semantic roles (kāraka) surrounding the semantic notion of 'action' (kriyā). These three are formally manifested as N-nom, N-acc, and V, respectively, with the consequence that governance relations (insofar as he operates with them at all) are for Pānini secondary phenomena, or reflections of semantic relations. The difference between Pānini and Sïbawaihi could be illustrated as follows (with due allowance to the word order differences in Sanskrit and Classical Arabic):13

ARABIA

Pānini

149

Sībawaihi

Meaning Agent Patient Act

Form

N-nom N-acc

V

V N-nom N-acc

This raises an intriguing question: Is it possible for Sïbawaihi and his successors to consistently maintain this formalist attitude and ignore the existence of the semantic dimension? It goes without saying that every linguistic tradition is bound to recognize the distinction between form and meaning, and in this respect the Arabs were no exception. They were convinced that one form could express only one meaning, and vice versa, and this principle of 'form—meaning biuniqueness' determined to a very large extent the nature of their morphological analysis (cf. Versteegh 1985). Yet the Arab grammar contains no explicitly formulated semantic component. Rather, there seems to have been at the sentence level a deliberate effort to separate form and meaning, and to concentrate on the former. This effort was not, however, entirely successful. The very parallelism between form and meaning, schematically indicated in the above figure, makes it difficult consistently to keep the two apart. This difficulty was enhanced by the fact that Arab linguists used a single set of terms to refer both to formal units and to their semantic counterparts: ficl—fācil— mafcūl, translated here as 'verb/act', 'agent', and 'object', respectively. As noted on p. 48, any interesting version of the language—reality isomor­ phism presupposes that there is a non-circular way to identify units of the reality-asconceptualized, i.e. that they can be identified without the help of linguistic units. This condition is not met by the Pāninian tradition, which, on the whole, projects the linguistic distinctions onto the world. Nor is it met by the Arab tradition, which simply identifies language with thought (or Arabic with correct thought), and is thus incapable of distinguishing between reality-as-conceptualized and reality-as-ver­ balized (cf. Corbin 1964:202; Gātje 1985:112).

4.5 Sïbawaihi's Conception of 'Discourse-Based' Linguistics Arab linguistics is famous for its concern with linguistic form. Yet, drawing on largely sporadic or indirect evidence, Belguedj (1973) argues that the earliest

150

UNIVERSAL HISTORY OF LINGUISTICS

descriptions within the Arab linguistic tradition were speaker-oriented. The object of analysis was constituted by difficult passages in the Koran or in poetry, and therefore it is natural that the elucidation of the speaker's (or the writer's) intentions gained a prominent role. When the focus shifted onto spoken language, the perspec­ tive widened to include the hearer as well. Carter (1968:242) notes that Khalil, one of Sībawaihi's two teachers, transmitted to his pupil the concern not just with the speaker, but also with the hearer. Sībawaihi discovered the formal structure of Classical Arabic, in a way that was to remain more or less definitive. But at the same time he also retained his predecessors' interest in the speaker—hearer interaction. He conceptualized this situation in the following way (cf. Carter 1968:163, 205): On the one hand there is the behavior of the speaker; on the other, there is the 'behavior' of the words spoken by the speaker; and the two types of behavior proceed in parallel. Fārisi (d. 987) uses the following metaphors to clarify the situation (Owens 1988:40, 64). At the level of language, we have a sentence like our (1), i.e. qöma zaydun. Within this sentence there is a relation of governance (i.e. causation) such that the verb, being similar to an axe, acts upon the Agent-word zayd-, which is similar to a piece of wood, and produces the nominative ending as its effect, just like the axe produces a change in the piece of wood. At the more inclusive level of linguistic actions, however, it is the speaker who is similar to an axe and who, using the verb as his proximate means (i.e. something like the sharpness of the blade), produces the nominative ending. Fārisī adds that it is only the latter metaphor which depicts the genuine type of governance (i.e. causation). Anghelescu (1985:7-8) and Bohas et al (1990:58-59) also distinguish between these two prima facie instances of causation. I find the preceding account perfectly correct, as long as the two levels are kept clearly separate. Causation obtains between spatio-temporal events or states. A language is a set (or system) of norms. Norms are not spatio-temporal entities. Neither are sentences or words, i.e. what norms apply to. Therefore at the level of language there can be at most some sort of 'quasi-causality': it ought to be that within a sentence qöma is followed by zayd-un. By contrast, there is genuine causality at the level of the speaker. He has internalized the (social) norm concern­ ing the nominative ending, and this (psychological) internalization, combined with the occurrence of a relevant speech intention, causes him to add the nominative ending after zayd-, when qöma precedes. On the other hand, this genuine psycho­ logical causation is, of course, different from (genuine) physical causation because, if one so wishes, one can violate the norm (i.e. annul the causally binding force of

ARABIA

151

its internalization), and add whatever one pleases after zayd- (for more discussion, cf. CLT:81-85, 270-275). The general 'actionist' framework of Sïbawaihi's syntax has been brought out by Carter (1968) and Bohas et al (1990, esp. pp. 38-42). According to this view, syntax must be seen as discourse, or as a result of 'discursive strategies'. I discern two aspects in this overall approach. One of them, although overlooked by the majority of Arabists, is of the highest value, whereas the other must be re­ jected. What I find valuable is Sībawaihi's attempt to elicit those intentions or operations that underlie the syntactic structure. This means that any utterance turns out to be "the global result of specific strategies, involving, at each level, a choice between several possible solutions" (Bohas et al. 1990:41). This kind of approach, even if it remains somewhat underdeveloped, has obvious modern counterparts.14 What I find less valuable, are Sïbawaihi's programmatic statements to the effect that syntactic description should be directly based on actual speech data. What makes this topic so intriguing is the fact that during the 20th century views identical with Sïbawaihi's have frequently been expressed in Western discussions of linguistic methodology (or philosophy of linguistics). What is at issue, is the notion that grammatical description should be based on concrete utterances (roughly: Saussure's parole), without any additional level of abstract sentences, or of linguis­ tic norms (roughly: Saussure's langue). This way of thinking has acquired new momentum with the rise of 'suprasentential linguistics' (= discourse analysis, conversation analysis, etc.) in the 70's. There are in fact several claims involved here, and it is good to sort them out: (A) Spoken language is the primary subject matter of linguistics, and it is qualitatively different from written language. Traditional linguistics (including I analysis, dependency grammar, and TG) has concentrated on the latter, and should therefore be abandoned. (B) There is no need to postulate abstract sentences (whose correctness or otherwise is determined by linguistic norms) in addition to concrete utterances. Traditional linguistics has concentrated on the former, and should therefore be abandoned. (C) Language as a real-life phenomenon consists in verbal exchanges that exceed the limits of one sentence. Traditional linguistics has been confined to describing sentence-length units, and should therefore be abandoned. I have already refuted the claim (A) (p. 14), and need not discuss it further. The claims (B) and (C) remain to be discussed. The fact that they were already made by Sïbawaihi should dispel the myth of their 'novelty'. At the same time Sïbawaihi reveals himself as our contemporary, not just as a representative of dependency grammar, but also as a (would-be) discourse analyst.

152

UNIVERSAL HISTORY OF LINGUISTICS

After these general remarks I shall go into the details. Sībawaihi argues that, to use terms familiar to the modern reader, his object of description is the concrete utterance, or kalām, not the abstract sentence, or what came later to be called jumla. No concept of jumla is needed to describe kalām. Since kalām is what the speaker actually says — in addition to being his act of saying it — it has no fixed length. It may consist of one word or of an entire discourse. The former type of utterance results from ellipsis. However, even though an elliptical utterance does not satisfy the (artificial) criteria of sentencehood, it is a legitimate object of grammatical description (Carter 1968:198-204). The roles of the speaker and the hearer are built-in into the concept of 'utterance' by means of the corresponding normative concept, which is mustaqim ('acceptable'). It is the speaker's responsi­ bility to make sure that his utterance will be understood by the hearer. Only when he has succeeded in this, his utterance qualifies as acceptable (pp. 218-225). Acceptability is thus a context-dependent notion, because of two formally identical utterances one may be acceptable and the other unacceptable, depending on whether or not the hearer has understood it. It is in keeping with this that meaning in general is taken to be a context-dependent notion, with the consequence that the problem of intrinsic (or literal) meaning simply does not arise (p. 203, 212). Carter finds nothing but praise for this 'non-abstract' conception of gram­ matical data. Subsequent grammarians developed a concept oí jumla, i.e. 'sentence' or, more explicitly, 'structurally complete sentence', but for Carter this was a mistake: By far the greatest advantage in this connection is the fact that the absence of a fixed concept of the sentence makes it impossible to object to any item which, though used in speech, does not appear to contain the required number of constituents to form a sentence (p. 202). Another advantage which kcààm possesses overjumla, and therein lies the superiority of the one system over the other, is that kalâm denotes an act of the speaker, while jumla denotes nothing more than a figment of the grammarian's imagination (p. 205). I take it that representatives of modern discourse (or conversation) analysis have no trouble in recognizing their own claims in the above passages. Now I shall proceed to refute Sïbawaihi's position.16 All I have to do is to show that he is forced to contradict himself. At every step he relies, and cannot help relying, on that very concept of 'structurally complete sentence' which he has supposedly gotten rid of. What I say here applies with equal force to the back-

ARABIA

153

ground assumptions of modern discourse (or conversation) analysis. (Remember that I do not oppose discourse analysis as such, but only the view that it makes traditional linguistics obsolete.) When Carter has to spell out Sïbawaihi's program in more detail, he must admit that "Sïbawaihi is saying ... that there is a point where the sentence contains enough constituents to be formally complete" (p. 228; emphasis added). Similarly, he admits that in ellipsis, or elision, "elided words still count towards the structural completeness of the sentence" (p. 229; emphasis added). Again, when he speaks of explaining "elisions and deviations from the normal speech-pattern" (p. 240; emphasis added), he presupposes precisely that concept which he thinks he has abandoned. The central concept here is ellipsis, because it is undeniably a pervasive characteristic of actual speech that it exhibits all kinds of ellipsis. Now 'ellipsis' is synonymous with 'deletion'. It is a simple conceptual point that 'deletion' is a three-place relation: X deletes Y from Z. X is the speaker, and 7 is a word. Now what is Z? (Remember, there must be a Z as a matter of conceptual necessity.) It can only be a correct sentence (where 'correct' is interchangeable with 'complete', 'normal', 'standard', 'grammatical', etc.); that is, it can only be the abstract concept of 'correct sentence'. The same is true of 'deviations' as well. A 'devi­ ation' in speech can only be a deviation from a linguistic norm, and a 'deviant' sentence can only be deviant as compared with a non-deviant, i.e. correct, sen­ tence. It is astonishing how difficult it is for those committed to discourse analysis (or more generally, to a 'concrete' or 'physicalist' conception of language) to grasp these elementary truths. Carter tries to avert these conclusions by claiming that in connection with jumla, unlike in connection with kalām, "all the possible reasons for elision would have to be specified in advance" (p. 229). But there is most certainly no such obligation. Correct sentences are potential entities existing at the level of (abstract) language. Whether or not they are uttered, and if so, whether they are uttered in a complete or in an elliptical form, is a question having nothing to do with gram­ matical theory, i.e. description of language in the strict sense. Specifying the reasons why this or that elliptical sentence has been uttered, is a task for a different branch of linguistics, i.e. empirical pragmatics or (empirical) discourse analysis. Owing to the general unpredictability of human actions, the reasons for different instances of ellipsis can only be specified post hoc. Nevertheless, there is a way to make Carter's position understandable (although it remains unacceptable). Arab linguists used the term 'ellipsis' in two different meanings (cf. Owens 1988:186-187). On the one hand, it is a process

154

UNIVERSAL HISTORY OF LINGUISTICS

which applies to actual sentences and deletes words which the hearer himself can supply because he understands them from the (linguistic or non-linguistic) context. All our examples, except (16), are (utterances of) actual sentences and can be made 'elliptical' in the present sense by deleting this or that word.17 ('Actual sentence' is a shorthand for 'sentence which may be uttered as such'. Strictly speaking, 'actual sentence' is a self-contradiction, because sentences are potential entities whereas utterances are actual entities.) On the other hand, 'ellipsis' is also a process which applies to a theoretical 'sentence' like (16) and produces an actual sentence like (15). If Carter wished to restrict jumla only to such structures as (16), he would have a point. He himself makes it quite clear, however, that jwmla is used in a much wider sense, i.e. in the sense of our 'sentence'. So my criticism stays. Theoretical or abstract sentences like (16) are postulated on the basis of taqdtr, which Baalbaki (1979:8) defines as a process of "supplying or inserting parts of utterances 'missing' through elision". As the single quotes indicate, it is only from a theoretical point of view that such parts of utterances appear to be 'missing'; and they are created by means of taqdtr, only to be deleted again when actual sentences are to be produced. In other words, taqdlr is one way to practise qiyās or analogical extension (cf. here p. 133): Postulating an underlying verb makes (16) conform to the general VSO pattern, the appearances notwithstanding. — Incidentally, 'ellipsis' is used in these two meanings also in the Western tra­ dition. The first meaning is of course the usual one, but the second occurs too, as when it is said that surface structure is elliptical vis-à-vis deep structure (cf. p. 133). Just compare (15) with (16). Let us continue. The normative concept correlated with kalām was said to be mustaqim ('acceptable'). If kalām were indeed the sole object of analysis, no other normative concept would be needed. It is revealing that Sïbawaihi also operates with a normative concept hasan, which Carter translates more narrowly as 'grammatically correct' and more widely as 'good Arabic' (pp. 225-228). An utterance is evaluated from two points of view: whether or not it is acceptable, and whether or not it is grammatically correct (or good Arabic). But this means that the entity whose correctness is evaluated is in reality not the utterance {kalām), but the sentence (jwnla) which the utterance exemplifies. Hasan is context-independent in a sense in which mustaqim is not: "mustaqim is seen ... to be independent of structural correctness" (p. 233). The point is not just thai jwnla has to be accepted in addition to kalûm. The point is, rather, that in spite of insistent claims to the contrary, Sïbawaihi's real object of description is not kalām, but jwnla. According to his 'official' program, linguistic description is about linguistic actions (= utterances), as they really

ARABIA

155

happen (p. 199, 205). On the other hand, however, Sïbawaihi's "grammar ignores such linguistic acts as ... have no characteristic grammatical consequences" (p. 209); and "Sîbawaihi is not interested in mistakes ... where the simple negli­ gence of the speaker results in an all-correct sentence [sic!] but one which is not what he intended to say" (p. 234). This is a clear contradiction. It can be removed only by giving up the 'myth of concreteness'. Utterances are really not so import­ ant. They are important only insofar as they relate to matters of grammatical correctness, or have grammatical consequences. 'Sentence' (jumla) remains the primary concept. This is evident already from the extent to which Sîbawaihi uses self-invented examples. It is just wrong to equate (utterances instantiating) selfinvented sentences with 'actual speech data'.18 Finally, Jurjānī among the later grammarians follows Sîbawaihi in empha­ sizing the study of language in its context of use. And he too denies the existence of context-independent meanings (Owens 1988:249). But he is forced to contradict himself, because when he proceeds in his semantic analysis, he sets up a basic distinction between literal and metaphorical meanings; and 'literal' is synonymous with 'context-independent'. More recent attempts to do away with literal meanings are similarly doomed to failure (cf. CLT: 166-169). In all the cases that I have discussed in what precedes, there is the same pattern. You perceive neither norms nor sentences, and therefore you think they do not exist. So you forbid the use of the corresponding terms. You do perceive utterances, and therefore you like to think that all you are doing when you are describing language is to describe utterances. However, when you try actually to account for what you are doing, you cannot help relying, perhaps in a terminologi­ cal disguise, on precisely those 'forbidden' concepts of norm and sentence which were supposed to be superfluous. It is fascinating to observe how this pattern keeps obstinately recurring throughout the history of linguistics. — The results of the preceding discussion corroborate my own position on the issue, as expressed in GTM:7.O. In conclusion, I wish to emphasize that, apart from his contributions to grammatical theory, Sîbawaihi provides many valuable insights concerning the use of language. It only needs to be seen that he is not speaking so much about actual use as about the principles governing possible use. As noted above, he views sentences as being uttered in a speech situation which contains both the speaker and the hearer, or more precisely the speaker's responsibilities and the hearer's expec­ tations (Carter 1968:239-242). Questions and commands have an obvious reference to the hearer; indeed, it would be rather absurd to view them as 'monological' speech acts, and to view the corresponding sentences as part of a 'monological' or

156

UNIVERSAL HISTORY OF LINGUISTICS

'non-functional' linguistic capacity. Now, Sïbawaihi manages to show that also statements, and the corresponding sentences, have a reference to the hearer, namely by conceptualizing them as answers to questions which the hearer could have asked (pp. 245-246). Or, in Sïbawaihi's own words: This is how we speak, even if the listener does not ask out loud, because what you say follows the extent of the questions he might pose if he were to ask you. Sïbawaihi pushes his 'interactional' view of language so far as to regard the structure of nominal sentence as motivated by communicative considerations. When the speaker has mentioned the topic, which is known to the hearer, the latter's expectations have been aroused, and then the speaker simply must add the com­ ment. To quote: And that is, as far as the person spoken to expects and anticipates, one of those cases where the speaker has no choice but to mention the predicate to the listener, because if you start with a noun you have only done so for the sake of what follows, and if you make an initial term you are obliged to mention something after the initial term, otherwise what you say will be corrupt and not allowed (p. 249). Sïbawaihi also seems to be developing a notion of multi-level, social knowledge when, discussing ellipsis, he says that "people only suppress [i.e. delete a word] when they know that you know who they [knowingly] mean" (p. 252). It does seem curious that Sïbawaihi combines his formal syntax with such pragmatic or discourse-analytical considerations. One might expect any 'inter­ actional' conception of language to contain a well-developed semantic component; but Sïbawaihi does not live up to this expectation. The reason for this state of affairs might be that in his thinking there are two strands which he is not quite able to harmonize, but of which, on the other hand, he cannot bring himself to discard either the one or the other. His successors were in this respect more straight­ forward.

ARABIA

157

4.6 Residual Issues I have chosen to concentrate here upon the syntactic component of the Arab grammar. The issues that I shall now briefly take up are, of course, 'residual' only from my present point of view. First I shall say a few words about morphology. My presentation is based on Versteegh (1985). With some exceptions, the Arab grammarians adhered to the biunique 'one form — one meaning' principle (cf. also Owens 1988:245). This means that they could accept the existence neither of portmanteau morphs nor of discontinuous morphs. When confronted with a prima facie portmanteau morph, they resorted to taqdtr (cf. p. 154) and postulated zero morphs. Thus, instead of accepting (I), they postulated (II):

This strategy was made easier by the fact that when the vowels indicating the three cases (cf. p. 132) were sometimes deleted, it was customary to assume that they still existed at some abstract level. Discontinuous morphs were in turn dispensed with by regarding one part of a prima facie discontinuous morph as a result of some phonetic process. Thus, instead accepting (I), (II) was postulated:

Notice that in the present context the form—meaning biuniqueness is considered from the viewpoint of simultaneity. It can also be considered from the viewpoint of alternativity, as when speaking of allomorphy (one meaning, several forms) or morphological ambiguity (several meanings, one form). In their phonology, the Arabs constantly operated with underlying forms which were manifested variously, depending on the phonetic context. They felt no necessity to establish an absolutely autonomous phonological level. Rather, phono-

158

UNIVERSAL HISTORY OF LINGUISTICS

logical rules could have access to morphological information. For instance, Bohas (1985) formulates a phonological rule which refers to the presence of a pronominal affix. A systematic distinction was made between nahw ('grammar', particularly 'syntax') and balāgha ('rhetorics'). The analysis of nahw assumed the existence of meaning throughout, but paid no systematic attention to it. The analysis of balâgha, by contrast, concentrated on meaning variation resulting from sentence-level variation in form. Baalbaki (1983) illustrates this difference by discussing Sïbawaihi's and Juijānï's respective approaches. For instance, Sībawaihi considered finite-verb constructions and nominalizations as essentially synonymous, whereas Jurjānī noted meaning differences between the two. Moreover, Jurjānī devoted an exceptional amount of attention to word-order variation and pointed out that it was employed to convey many different types of information. In the Arab tradition there never developed any genuine cooperation between linguistics and logic. (It will be recalled that the situation was much the same in India; cf. p. 83.) Although the grammarians and the logicians conceded that to some extent they dealt with one and the same subject matter, they insisted on viewing it from two different points of view. According to Langhade (1985: 110-112), the grammarians saw themselves as dealing with spoken language, exemplifications of which (mainly, the Koran) were memorized and transmitted by oral tradition, whereas the logicians, being dependent on translations from Greek, thought of themselves as dealing with written language. Perhaps more to the point, there was thought to be a difference in the levels of abstraction between linguistics and logic. As explained by the philosopher al-Fârâbï (cf. Langhade 1983:133136), the grammarians and the logicians investigate, respectively, what is common to one nation, and what is common to all nations. To this 'particular vs. universal' distinction at the level of thought categories, a comparable distinction is added within the hierarchy of the sciences. Linguistics is just one science among others, whereas logic as the discipline about finding the truth and avoiding the falsity is common to all sciences. It is clear that al-Fārābīs views merely reflect those of Aristotle. Therefore it is only the more interesting to note that in the West Aristotle did influence the linguistic analysis of sentence structure (cf. 5.3.4). And more generally, for some time there was a close cooperation between linguistics and logic in the West (cf. 5.3.5). Such a development just failed to take place in Arabia (as well as in India). In conclusion I take up once again the issue of cultural determinism. Outwardly at least, the Arab culture is remarkable homogeneous. Practically all Arab thinkers adhere to the Islamic faith, and although there are different traditions

ARABIA

159

within Islam, what unites them is much more important than what separates them. Yet the question arises whether this homogeneity is more apparent than real, i.e. whether there is a double standard such that for practical reasons one professes to believe something that one does not believe in fact. That this question may have to be answered affirmatively in at least some cases, is indicated by the fact that the history of Islamic science does contain a couple of famous dissidents. Al-Rāzī (d. c. 925), "the greatest physician of IslamH, rejected revelation and prophecy, which are of course central to the Islamic religion, criticized Aristotle, who was being established as the supreme authority within the Hellenistic school of Arab philosophy, and adopted views on the universe and God which are very similar to Plato's (cf. Fakhry 1970:115-124). 20 He also advocated the transmigration of the soul and its eventual release from the 'wheel of birth', views which are consonant with the Pythagorean-Platonic doctrine, but may also have been borrowed from Hinduism (ibid., pp. 46-47). So we see that, just as in India or in China, there has been in Arabia no necessity for everybody to think in the same way. Al-Rāzī's example shows that this kind of nonconformism may have been more widely spread than it appears.

4.7 Conclusion: Pānini, Confucius, and STbawaihi By this time, a 'non-Western' pattern of history has emerged in connection with the three traditions I have been investigating. In each case, the tradition has been determined by the one who stands at its beginning. To be sure, we must differen­ tiate here between recorded and unrecorded history. Only Confucius can be said to mark something like an absolute beginning. Both Pānini and Sībawaihi had prede­ cessors whose role can no longer be ascertained with exactitude. Still, one can safely infer that it was precisely Pānini's and Sībawaihi's pre-eminence which eclipsed their predecessors. And most importantly, since then the two of them have gone unchallenged for a very long time. Traditions can, and must, be considered both from an internal and an external point of view. It is possible, after all, that somebody dominates a tradition for wrong reasons, i.e. because of undeserved authority. With Pānini this is not the case: also by Western standards his authority is fully deserved. What about STbawaihi? It seems clear to me that his grammatical theory remains inferior to Pānini's. Still, Sïbawaihi's syntax is based on a few consistently formalist generaliz­ ations which are executed very systematically, and I find it doubtful whether any Western pre-1900 descriptions in syntax achieve the same theoretical level. In

160

UNIVERSAL HISTORY OF LINGUISTICS

particular, the value of Western syntactic descriptions is often diminished by an inability to keep syntactic and semantic issues separate.

ARABIA

161

Notes 1)

Professors Carter and Owens have been kind enough to send me extensive comments on the first draft of this chapter. Interestingly enough, their views are often in conflict. Unable to produce some sort of 'synthesis', I have found myself siding with Owens, mainly because his emphasis on linguistic theory seems congenial to me.

2)

It was assumed that there are several levels of reality, most obviously the microcosmos and the macrocosmos, which symbolize one another; and alchemists at­ tempted to present a cosmology which would comprise the totality of human knowl­ edge, very much like Neo-Conrucians had tried to do; cf. Corbin (1964:16-17, 186188).

3)

Sezgin (1984) too claims that Arab linguistics had become a mature discipline already before Sībawaihi's time. However, he is unique in holding that "Sībawaih hat weniger seine eigenen Meinungen mitgeteilt, als hauptsāchlich diejenigen seiner Vorgānger [i.e. al-Khalil and Yūnus] aus ihren Werken zusammengestellt" (p. 26).

4)

Today, to be sure, the opposition between the two schools is generally thought to be an after-the-fact fabrication.

5)

Both Carter and Owens are today more inclined to stress the multifariousness of the Arab tradition.

6)

Bohas et ál. mix up the temporal relations when they add that, in acting this way, the later grammarians "failed to perceive what was most original in the Kitāb". But from their point of view, of course, everything in the Kitāb was equally original. They just concentrated on a certain part of it. The remainder looks 'original' today only because it was ignored at the time.

7)

Admittedly, this is an oversimplification. For instance, some of the notions that will be mentioned below do not occur explicitly in Sībawaihi's work (although it is generally agreed that they do so implicitly). The precise relation between Sibawaihi and his successors is right now a question intensively discussed by experts on Arab linguistics. I am not competent to participate in this discussion.

8)

In evaluating this claim, I have benefited from comments made by Dr. Anneli Pajunen. Carter strongly doubts the applicability of such Western notions as 'govern­ ance'. Apart from Owens (1988), however, they are accepted e.g. by Bobzin (1983) and Bohas et al. (1990) (cf. below). Besides, the notion of 'governance' is certainly in agreement with the (Arab) notion of 'linguistic causation' (again, cf. below).

162 9)

UNIVERSAL HISTORY OF LINGUISTICS

In his comments Owens cites evidence for this view from later grammarians.

10) However, there are some interesting expections to this general rule. For instance, it is a fact that Mexico is perceived as more similar to the U.S. than the U.S. is to Mexico. 11) To Carter, the resumptive pronoun hu is the "real relative pronoun". 12) In this respect Sïbawaihi's notion of the passive is similar to Pānini's; cf. p. 51. 13) Carter denies the legitimacy of comparisons of this kind. He wishes to know "why on earth we would want to compare systems which never had anything to do with each other". There are alternative ways to answer his query. The short answer consists in pointing out the analogy with the theory of universal grammar, which is in the habit of comparing languages 'which never had anything to do with each other', i.e. which belong to language groups with distinct genealogies. Part of the long answer is given in 6.2. 14) One thinks e.g. of Halliday's 'systemic grammar'. 15) Today Carter would like to tone down these statements somewhat. 16) Carter dislikes the idea of Sībawaihi being criticized. In his opinion, we should "try to understand rather than to judge". Owens too thinks that "it is better to steer clear of value judgments about the efficacy or lack of it in Sībawaihi's and others' linguis­ tic practices". Now, as I pointed out already in my Introduction, I think value judgments are a necessary part of science, from the viewpoint both of the scientist and of the historian (or philosopher) of science. This is evident from the simple fact that a scientist always tries to be better than his rivals, and that the observer judges him to be either better or worse than his rivals. When a historian singles out one grammarian from among two thousand, he must have a reason for doing so, and the reason can only be that he thinks the one grammarian deserves to be singled out, i.e. that he is good enough to be singled out. Hiding these facts serves no purpose. Similarly, if we are obligated always to understand and never to judge, we end up with a Chuang Tzu-type absolute relativism, which, because of its self-contradictory character, is a logical impossibility. Again, it serves no purpose to pretend to be doing something that it is logically impossible to do. 17) To be quite precise, (18) and (19) are phrases, rather than sentences. 18) Nothing of what I have said above is meant to deny that empirical discourse analysis is a viable notion. I have merely shown that it necessarily presupposes the traditional notion of grammatical theory. It should also be clear that my critical remarks concerning some of Sïbawaihi's methodological statements in no way diminish the value of his syntactic theory.

ARABIA

163

19) This insight constitutes the basis for today's 'question-theoretical' approach to language. 20) Incidentally, he criticized grammarians for denying the value of any science except their own (Sezgin 1984:229).

5. Europe

In Europe we encounter a linguistic tradition markedly different from the traditions that I have discussed in the three previous chapters. Grammatical theory is not necessarily limited to one point of view only, and other linguistic subdisciplines slowly emerge in addition to grammatical theory. During the recorded history, moreover, there occur within linguistics periods of incontestable progress (counter­ balanced, to be sure, by periods of decline). Confronted with this wealth of data, one is in the danger of being over­ whelmed. Therefore it is sensible to adopt a perspective which brings some order into the data. My own solution is the following. First, I make the discussion in this chapter comparable to that in the previous chapters by concentrating on (synchronic) grammatical theory. Secondly, apart from their more specific contri­ butions, I shall consider the representatives of Western grammatical theory under the aspect of the position that they assume vis-à-vis the trichotomy 'languagemind—reality'. For this, I must set up the following continuum. reality—mind—language > mind—language > language (Sometimes the existence of mind is ignored, and language is confronted directly with reality, which gives rise to the dichotomy 'reality—language'.) I shall indicate the place that individual linguists occupy on this scale. In doing so, I must often distinguish the place they occupy in fact from the place they merely think they occupy. The theories which contain mind as one component are to the correspond­ ing extent psychological (or psycholinguistic) theories. It has been a standard assumption within the Western tradition that mind is universal in character, and that language is in turn just a reflection of mind. It follows that the history of Western linguistics has to a very large extent been a history of the notion of universal grammar. As in connection with the word 'Arabia', 'Europe' means here a cultural, rather than geographic entity. Thus North America and Australia are part of 'Europe'. It is more difficult to determne to what extent other parts of the world have become 'Europeanized'.

166

UNIVERSAL HISTORY OF LINGUISTICS

Contrary to a 'post-modern' myth, I shall argue that there are no breaks or 'ruptures' within the history of Western linguistics. If I distinguish certain subwholes within it, this is made purely for convenience. The difference between Greek philosophy of language ( = 5 . 1 ) and Greek linguistics (= 5.2) is just a matter of emphasis. The difference between Greek linguistics and Western medieval linguistics (= 5.3) is due to external factors: for several hundreds of years the tradition of linguistic research broke down, but when it was finally resumed, this was done entirely along the lines of Greek linguistics (and of Aristotelian philos­ ophy). The difference between medieval linguistics and Western grammatical theory between 1500-1900 (= 5.4) did exist, but only in imagination: contrary to what they thought, the post-medieval grammarians continued to rely, essentially, on a descriptive framework that had been established in the Middle Ages. Finally, even though the 20th-century grammatical theory (= 5.5), having accomplished genuine progress, has to some extent distanced itself from its past, many of its basic concepts can still be traced back to antiquity, as I shall now proceed to show.

5.1 Linguistic Thinking in Greek Philosophy The European tradition of systematic thought began in Greece, first as general cosmological speculation. Then there was gradual specialization and emergence of topics around which autonomous disciplines were later to be established. Thaïes (. 625-545 B.C.) assumed that water is the origin (arkhē) of the universe: everything is born out of water and returns into it. Anaximander (c. 610-545) raised speculation to a higher level of abstraction: the origin is not a stuff like water, but rather a principle, i.e. the 'unlimited' (apeiron), which then divides itself into opposite qualities like warm and cold or dry and moist; change is produced by mutual influence of the opposites. Pythagoras (c. 570-490) and his followers increased the level of abstraction and added the quantitative dimension: for them, the origin of the universe is constituted by harmonious relationships between numbers, and the visible world is just a reflection of such relationships. Cosmological speculation had to reconcile the eternal existence of the universe with the fact that it is changing all the time. The opposition between change and immutability was epitomized by Heraclitus and Parmenides, who will be discussed below. It was a common assumption that nature is permeated by spirit, and this spiritual aspect was taken to be something divine; often it was simply called 'God'.

EUROPE

167

(Democritus and the other Atomists were here the exception.) The philosopher, who spent all his time pondering upon nature, seemed to partake of its divinity. It is a striking fact that Plato and Aristotle were among the most religious of the Greek philosophers. The former believed that God had created the universe, "desiring that all things should be as like Himself as they could be". The latter believed that God was the 'Unmoved Mover' who upheld the movement of the heavenly bodies by being its purpose (i.e. 'final cause'); besides, God's sole activity, i.e. contemplation, had an openly acknowledged similarity with philosophi­ cal activity. The reason why Plato's and Aristotle's texts were preserved during the first one thousand years of the Christian era, when so many other texts perished, must be sought in part in the religious convictions which underpin their teachings. This by no means implies that Plato and Aristotle did not deserve to be preserved. It only implies that others would have deserved it too.

5.1.1 Plato The oldest extant treatise on language is Plato's (d. 349 B.C.) dialogue Cratylus (originally Kratulos). Because of its special position, it deserves to be analyzed in some detail. Plato's other contributions to the theory of language will also be discussed. The Cratylus endeavours to answer a question concerning the nature of language: Is language based on phusis or on nomos! Or, to retain the original formulation: Does language existphusei (= dat.) or nomōil Before embarking upon an analysis of the dialogue, I must first explain the meanings of these two terms. The phusis vs. nomos controversy had its roots in pre-Socratic thinking, mainly in the Sophistic movement (cf. Steinthal 1971 [1890]:54-74; Rankin 1983: 79-91). What was at issue, originally, was the relation between human nature (phusis) and the laws (nomoî) of society. A whole spectrum of different opinions were represented by individual Sophists. Human nature was generally assumed to be unjust and violent. Given this assumption, two opposite positions could still be adopted vis-à-vis the laws: either they are the laws of the ruler, invented to oppress the people; or, on the contrary, they are the laws of the people, invented to protect them from the ruler. A position intermediate between these two extremes came slowly to be accepted: people could survive only in society, and laws were required for society to come into being. Thus, laws protected people equally against human nature and against the various threats posed by physical nature. This was the origin

168

UNIVERSAL HISTORY OF LINGUISTICS

of the 'social contract' doctrine (which was formulated nearly simultaneously by Hsün Tzu in China; cf. p. 113). It was obvious that rather dissimilar laws had been instituted in different city-states (polis), not to speak of laws that were valid in barbarian societies. Acknowledging this fact was in agreement with the relativist streak in Sophistic thinking, represented most pregnantly by Protagoras and his dictum 'Man is the measure of all things'. In his legal thinking as elsewhere, Plato was a definite opponent of all forms of relativism (cf. Friedrich 1963:13-19). When formulating his legal theory, he did not try to derive it from existing laws. Rather, he tried to formulate a system of laws which approximated an ideal or universal legal system as closely as possible, the universality of the latter being presumably guaranteed by the fact that it was based on a universal ethical foundation, i.e. on the (Platonic) 'ideas' or 'forms' (eidos) of virtue and justice. Thus, to use terms that were to become current later, Plato embraced 'natural law' (ius naturale, Naturrecht) and rejected 'positive law'. This is the general background to the phusis vs. nomos debate as it is carried out in the Cratylus. The meaning of nomos is rather clear. As shown by the synonymously employed words sunthēkē and homologíā, it just means 'agreement' or 'convention'. The meaning of phusis is more difficult. So much is evident that claiming language to exist phusei does not amount to claiming that language is a 'natural' phenomenon, e.g. in the sense that linguistic utterances would be compar­ able to such natural sounds as laughing or sneezing. (To be sure, such a view was also advanced, namely by Epicurus and, following him, by Lucretius in his De Re­ nan Natura.) On the other hand, it is also evident that phusis stands for something universal and absolute, thus contrasting with nomos qua representative of what is particular and relative. It needs to be realized, however, that here too Plato gives his characteristic twist to the notion of universality, as shown by the close proxim­ ity of phusis and eidos (cf. above). Accordingly, claiming language to exist phusei amounts to claiming language to be such as to correctly reflect the nature or idea (phusis/eidos) of its referent. This interpretation is directly confirmed by the fact that the phusei view is explicitly reformulated as a claim to the effect that there obtains a 'correctness of words' (orthotēs onomatön). To sum up, the dialogue proposes to answer the following question: Is language arbitrary (nomōï), or is it non-arbitrary (phusei), namely in such a way that its own nature has been deter­ mined by the nature of its referent? The dialogue contains three participants: Hermogenes, who is a supporter of the nomōi view, Cratylus, who is a supporter of the phusei view, and Socrates, who, as usual, functions as Plato's mouthpiece. As I see it, the Cratylus consists of the following four parts: a) Socrates gives an abstract justification to the phusei

EUROPE

169

view (383A-391C). b) He tries to substantiate it by offering 'etymologies' of (supposedly) complex words (391D-421B). c) Justifications are sought for simple words and single sounds (421C-428E). d) Socrates reverses himself and starts to argue for the nomōi view by revealing the weaknesses inherent in the phusei view (428B-440E). Now I shall proceed to discuss these four parts in order. a) Socrates identifies a (social) convention for using a word with an arbi­ trary (individual) decision to use a word; and the latter is in turn identified with a full-fledged Protagoras-type relativism, which is easily shown to be self-contra­ dictory. In this way the nomöi view is at once, but wrongly, made to appear wholly implausible. Speaking is defined as a subtype of action; and action is defined as the use of an instrument (organon). Thus speaking too is defined as instrumental action. More precisely, "a name is an instrument of teaching and of distinguishing natures, as the shuttle is for distinguishing the threads of the web" (388C). In both cases there is an absolute, universal standard (eidos) of the instrument which is the same in Greece and in barbarian countries. An instrument meeting the standard is 'fitted by its nature' (phusis) for its purpose. A carpenter produces the shuttle, and a weaver uses it. Similarly a 'namegiver' (onomathetēs) produces the words, and a teacher (didaskalos, dialektikos) uses them. Interestingly enough, the principal term for 'name-giver' is nomothetēs (lit. 'law-giver'). Here Plato almost inadvertently achieves a neat synthesis between the two conflicting positions: nomos may be based on phusis; linguistic conventions may be based on the correctness (or adequacy) of words. What does the phusei character of language consist in? What does the 'correctness' of words mean? In the ideal case, a word or 'name' (onoma) is an imitation of its referent (mimèma touprugmatos, 430B). However, the relation of language 'imitating' reality extends from letters and syllables (424B) to entire sentences (424E, 43IB). Moreover, it is interesting to note that elsewhere, e.g. in the Republic 382B, language (more precisely, falsehood in language) is claimed to be also an imitation of mental phenomena {mimèma tou en tèi psukhēi pathēmatos). b) When Socrates has to justify the phusei view in concrete detail, he first turns to 'etymologizing'. He tries to prove that words in their present state are 'correct', by showing that they can be shown to result from complex expressions which reveal their 'true' or 'original' meanings. Such exertions take up most of the dialogue. From among the hundreds of examples that he adduces, I mention only one. What is the etymology of onoma itself?

170

UNIVERSAL HISTORY OF LINGUISTICS

The word onoma seems to be a compressed sentence, signifying on hou [tunkhanei] zētēma [on] ('being for which there is a search'), as is still more obvious in onomaston ('notable'), which states in so many words that real existence is that for which there is a seeking (on hou masmd) (421E). Taken at face value, all this is of course nonsense. Now, on a few occasions Socrates clearly indicates that he is just joking. But this does not help much, because as a joke his 'etymologizing' is much too extended, and therefore disas­ trous. Out of misguided respect, most commentators fail to make this point ex­ plicitly enough. Steinthal might seem to be here an honorable exception: Nun ist aber dieser Dialog ein sehr wundersames Werk, eine, wie es zunāchst scheint, durchaus fratzenhafte Caricatur, die uns mit so ver­ zerrtem Gesicht anblickt, dass man nicht weiss, ob es lacht oder weint oder ruhig ist; sein Auge schielt... (p. 42). In the sequel, however, Steinthal does his utmost to justify Plato on all points, including his vain attempts to create a science of etymology. c) Socrates points out that his way of 'etymologizing', even if correct in itself, only postpones deciding the issue. Even assuming words to be phusei in the sense of being composed of simpler words, the real question is then whether the latter too are phusei. And if they are, they obviously cannot be so in the same way as complex words. In what sense, then, are simple or 'first' words (ta pröta onomata) phusei! After considering, and rejecting, onomatopoiesis, Plato consist­ ently enough turns to sound symbolism. With obvious diffidence, he suggests, e.g., that the sounds r, t, l, i and  express, respectively, movement (or rapidity), binding (or rest), 'gliding', subtlety, and roundness. d) When Socrates proceeds to criticize the phusei view, he does so from several points of view. First, he establishes that it is possible to misuse language, i.e. either to utter false sentences or to use words incorrectly. The criticism seems to be directed against Parmenides, who had claimed that non-existence can be neither thought nor expressed (cf. Stenius 1975:89-90). Since it is, however, poss­ ible to speak about what does not exist, language is no reliable guide to reality. Things cannot be learned through words; on the contrary, the name-givers them­ selves had to know things first, before they could ('adequately') name them. What Socrates is criticizing here, is an interesting constellation of opinions. Parmenides and Heraclitus were the champions, respectively, of non-change and change. In general, Plato was sympathetic to the former and critical of the latter.

EUROPE

171

Cratylus in turn was Heraclitus' disciple. In the present context, however, he clearly defends views congenial to Parmenides. If there is a strong congruence, even to the point of coincidence, between what exists, what is thought, and what is said, it is easier to understand that words should be 'adequate' to their referents. On the other hand, under the circumstances of Heraclitean universal change words are likely to lose whatever 'adequacy' they may possess. (In fact, it is argued in the Theaetetus 183B that no words are applicable under the conditions of universal change.) The phusei view has generally been attributed to the Heraclitean doctrine, precisely on the evidence of the Cratylus. It is quite probable, however, that on this point Cratylus is defending his own idiosyncratic mixture of Heraclitus and Parmenides, rather than some 'pure' Heraclitean doctrine. The phusei view is essentially weakened if words can be used incorrectly or, what comes to the same, if there are incorrect words in addition to correct ones. For instance, the assumption of sound symbolism cannot be entirely right because the word sklēron expresses hardness but contains the sound / which should express its opposite. Socrates makes Cratylus admit that he nevertheless understands the meaning of sklëron. Since this understanding cannot be based on correctness, it must be based on something like 'custom' (ethos). And it is difficult to deny that 'custom' is here practically synonymous with 'convention' (sunthëkë). Still, Socrates is not prepared to give up the phusei view entirely. His final opinion seems to be that perhaps there is something like correctness of names; but even if there is not, at least there should be. * In this ambiguous statement one can see the origin of the 'ideal language' doctrine which was later to lead to the creation of logical semantics: I quite agree with you that words should as far as possible resemble things... For I believe that if we could always, or almost always, use likenesses, which are perfectly appropriate, this would be the most perfect state of language, as the opposite is the most imperfect (435C). The dialogue concludes with a critique of Heraclitus. It is possible that language expresses motion, rather than rest. But even if this is so, it only shows that the name-givers were mistaken, for the true nature of reality is such as to be in rest, rather than in motion. The opposite view, namely that everything, including the concept of knowledge itself, is continuously changing, leads to the paradoxical conclusion that nothing can ever be known (or expressed; cf. again Theaetetus 183B). Thus, some fixed standard of knowledge is required. Here the critique of Heraclitus overlaps with the critique of Protagoras given in the beginning of the

172

UNIVERSAL HISTORY OF LINGUISTICS

dialogue. The assumption of universal change leads to the same unacceptable conclusion as the assumption of the relative or subjective nature of all knowledge. To sum up, I quote Steinthal's characterization of the phusei view. This remarkable passage, which anticipates the correspondence and picture theories of the 20th century, was written already before 1863: Es komme darauf an, meint Sokrates, zuerst die einfachen Elemente, auf die sich alle wirklichen Dinge zurückführen lassen, aufzufinden und ihnen die Laut-Elemente der Sprache so gegenüberzustellen, wie immer eins der letzteren einem der ersteren entspricht; und sodann nach der Weise, wie in der Welt jene Elemente sich zusammenfügen, so auch die Laute zusam­ menzufassen und in den Sylben, Wörtern und Sātzen das lautliche Abbild des Wesens aller Dinge zu erkennen (1971 [1890]: 103; emphasis added). It must be emphasized, however, that on SteinthaTs interpretation Plato eventually rejects the phusei view. Here I disagree with Steinthal. As fas as the nature of language is concerned, Plato combines a picture theory with a use theory: language is (or should be) an instrument which is used in such a way as to give a one-to-one picture of reality. Moreover, the mental dimen­ sion is held to possess a structure similar to language: "when the mind is thinking, it is simply talking to itself, asking questions and answering them, and saying yes or no. ...and judgment [is] a statement pronounced, not aloud to someone else, but silently to oneself' (Theaetetus 189-190E; similarly Sophist 263). Thus Plato pos­ tulates a general 'imitational' similarity between language, thinking, and reality. However, he stops short of the Parmenides-type view according to which there is virtual identity between the three, as there would necessarily be if it were indeed impossible to say or think what does not exist. For Plato, mind is something internal which is made publicly observable by speech: "[The first meaning of logos is] giving overt expression to one's thought [dianoia] by means of vocal sound with names and verbs" (Theaetetus 206D). Mind is primary with respect to language (since the latter expresses the former). Language is used to make distinctions between aspects of reality (cf. above, Cratylus 388C). It follows that mind must already have made those same distinc­ tions. Plato's terminology for the 'language—mind—reality' triad is the following:

EUROPE

173

AU of Plato's views that I have considered here have proved to be extremely influential in the Western tradition of language theory, and they are alive still today. To be sure, the form which Plato gave to these ideas was far from defini­ tive, and therefore within this field he cannot of course be compared to such figures as Pānini or Sībawaihi. To round off the picture, it must be added that Plato also formulated the rudiments of sentence analysis. In the dialogue Sophist the personage of 'Eleatic Stranger', who acts as Plato's mouthpiece, is at pains to analyze the nature of a false statement (a problem touched upon also in the Cratylus), and in so doing he is forced to pay attention to what we would characterize as syntactic questions. He notes that (correct) statements are not just strings of words like *Walks runs sleeps or *Lion stag horse. Rather, when one is uttering a (correct) statement of "the simplest and shortest possible kind", like A man understands, he is "putting a thing together with an action by means of a name and a verb" (suntheis prāgma prāxei dV onomatos kai rhëmatos). As a consequence, "just as some things fit together, some do not, so with the signs of speech; some do not fit, but those that do fit make a statement" (262D-E). As for the notion of falsehood, Plato is able to solve Parmenides' problem, by distinguishing between two different meanings of 'non-existence' (or 'nonbeing'). If Theaetetus is sitting and I say Theaetetus is flying, I am in one sense speaking about what does not exist. However, I am not speaking about non-exist­ ence, or 'emptiness', in some absolute sense. Rather, I am speaking about some­ thing which does not exist in the way suggested by my statement, i.e. something which does exist (viz. Theaetetus), but only differently from the way suggested by my statement (263B). Since I am interested in the universalist aspects of my subject matter, I conclude this section by pointing out a few broad analogies between Greek thinkers mentioned here and Chinese thinkers discussed in Chapter 3. Heraclitus' 'dialecti­ cal' view that change is always and everywhere governed by a movement between opposites is identical with Lao Tzu's (cf. p. 104). Heraclitus also reveals a relativistic spirit somewhat akin to Chuang Tzu's: they note in identical terms that what humans find beautiful or useful, animals find ugly or useless, and vice versa. Yet, as for expounding full-blown relativism, the genuine Greek counterpart to Chuang Tzu is of course Protagoras (cf. p. 105). But Chuang Tzu also argued, in a manner strongly reminiscent of Hindu philosophers, that both the differences of opinion and the distinctions in the world ultimately vanish in, or are absorbed by, one transcen­ dent entity (be it called 'Tao' or 'Brahman'). Now this sounds just like Parmenides.

174

UNIVERSAL HISTORY OF LINGUISTICS

He taught that all change is unintelligible and therefore illusory (as shown in more detail by Zeno of Elea's paradoxes, e.g. the flying and yet immobile arrow). The same is true more generally of all the phenomenal variety that our sense organs seem to apprehend. The only thing that exists is One, as grasped by an extra­ ordinary intellectual intuition, and nothing exists beyond it.

5.1.2 Aristotle It seems fair to say that Aristotle (d. 322 B.C.) had no systematic interest in the analysis of natural language. However, he created the science of logic — practically from scratch —, and since his logic is only semi-formalized, he could not help using natural language to express it, and thus had to take a stand vis-à-vis some basic linguistic issues. His real object of interest was the theory of syllogism; and since a syllogism is composed of (three) sentences, and (as he saw it) a sentence is composed of (two) terms, he had first to outline a theory of terms (which he did in the Katēgoriai) and a theory of sentence (which he did in the Peri hermeneiās), before he could expound his theory of syllogism (which he did in the first part of Analutika, generally known as Analytica Priora)} Aristotle had an enormous influence on the intellectual history of Europe. In particular, his logic (together with Euclid's Elements) served as the exemplar of systematic thinking until the 19th century. Because of this fact, his somewhat offhand remarks on natural language acquired a tremendous importance, and largely determined the nature of linguistic theorizing in Europe until the eve of the 20th century. The first two or three pages of De Interpretation (i.e. 16a-17a, 5) are especially significant, the rest of the treatise being subordinated to logical consider­ ations. The extraordinary fate of this short passage has been well documented by Arens (1984). Aristotle's point of departure is the trichotomy already established by Plato: language—mind—reality. But he further specifies that mental experiences (pathēmata) are images or copies (homoiōmata) of things (prāgmata) whereas spoken language (phōnal) contains merely signs (sumbola, sēmeia) of mental experiences, just as written language (grammata) in turn is merely a sign of spoken language. Accordingly, things and mental experiences are common to mankind, whereas spoken and written languages vary. It follows that the relation of signifi­ cation between mind and reality, being based on similarity, must be in some sense natural. By contrast, Aristotle makes a point of noting, on three occasions, that

EUROPE

175

language is conventional, or kata sunthēkēn. This expression is contrasted with language existing phusei and language being an instrument (organon). This is clearly a comment on the Cratylus. When Coseriu (1975:71-83) claims that kata sunthēkēn means for Aristotle something quite different from what it meant for Plato, I just cannot follow his argument. Like Plato before him, Aristotle accepts the principle that 'what is in speech corresponds to what is in the mind' (ta en tēiphonēi akolouthei wis en tēi dianoiāi) (De Int. 23a, 30). More precisely, a spoken affirmation or denial is merely the expression of its mental, and primary, counterpart. But now Aristotle is faced with a problem which was to be solved only by the scholastic grammarians: If mind is a universal language that corresponds picture-like to reality, and if language (as spoken in different communities) in turn corresponds to mind, then language just cannot be in every respect arbitrary, or kata sunthēkēn. Rather, behind the apparent variability of the existing languages there must be something like a common core which 'supports' the relation of correspondence between mind and language. For instance if mind reflects the ontological distinction between thing and action, and if language corresponds to mind (here: to the mental distinction between thingconcept and action—concept), then all languages must possess the (universal) linguistic distinction between noun and verb, irrespective of the (arbitrary) material forms of nouns and verbs in particular languages. In this respect, then, Aristotle represents a step backwards vis-à-vis Plato, in spite of his greater explicitness. For Aristotle, language, mind and reality coincide in general. But he is also careful to point out that there are particular cases where this coincidence is lacking. For instance, words and things cannot literally correspond to one another, because the latter by far outnumber the former.3 Similarly, there are concepts which either have no corresponding words or which are expressed by an entire sentence. But it is just cases like these which demonstrate that it makes sense, in the first place, to speak of three different realms of phenomena and their general correspondence. Hence it would be wrong to argue, with Baratin & Desbordes (1981:18-19), that noting such occasional instances of discrepancy within the 'language—mindreality' relation makes Aristotle's approach fundamentally different from Plato's. The preceding discussion can be summed up as follows:

176

UNIVERSAL HISTORY OF LINGUISTICS

It will be recalled that for Plato both ethics and language exist phusei. Here Aristotle represents a different position. For him, language (i.e. the combination of the sound and 'what is in the sound') is arbitrary or kata sunthëkën, but ethics exists phusei: "For there really is, as everyone to some extent divines, a natural justice and injustice that is binding on all men..." (Rhetorica 1373b, 5). It must be realized, however, that phusei means two somewhat different things when applied to language and when applied to ethics. In the former case it means a rational correspondence between existing languages and reality, but in the latter case it does not mean some sort of correspondence between existing laws and justice. Rather, it amounts to the claim that there does exist a notion of justice which is common to mankind and thus rooted in human nature. There is a problem concerning the notion of meaning: what is its place within the Aristotelian framework? There can be no doubt that Aristotle has generally been understood as representing the psychologistic theory of meaning: meanings are just those mental concepts and judgements which are expressed by words and sentences (more precisely: by strings of sounds identifiable as wordforms and sentence-forms). Yet it is also possible to interpret Aristotle differently. Much depends on how we translate the expression ta en tēi phōnēi (lit. 'what is in the sound') as it occurs in the statement 'ta en tëi phönëi are signs of mental experiences'. It might seem natural to equate ta en tëi phönëi with phönai ('sounds'), but Steinthal (1971 [1890]: 191) categorically rejects this translation: "Wir würden sagen, der Laut ist Symbol der Vorstellung; aber so sagt Aristoteles hier nicht; denn ta en tëi phönëi kann nur heissen: die Bedeutung der Laute, die also verschieden ist von dem Gedanken der Seele". The second important point contained in the first pages oî De Interpretations is how Aristotle analyzes the structure of a simple statement. When presenting his analysis, I shall also make use of terminological distinctions and examples occur­ ring elsewhere in his work.

EUROPE

177

The superordinate notion is 'sentence' (logos). Its most important sub­ category is 'statement' (apophansis), i.e. the sentence-type which admits of truth or falsity. Statements are in turn subdivided into affirmations (kataphasis) and denials (apophasis). Each statement consists in a combination (sumplokë) of mental concepts (noēmata). (Actually, it would be better to speak of noëmata being 'related' to, rather than 'combined' with, each other.) The combination is sub­ divided into 'composition' (sunthesis), as in affirmations, and 'division' (diairesis), as in denials. That is, in an affirmation like 'The rose is red' the roseness and the redness are put together whereas in a denial like 'The rose is not red' they are separated from each other. (Of course, it can also be said that in a denial the roseness and the non-redness are put together.) Finally, a statement is true or false depending on whether or not the things are as it claims they are. This is Aristotle's 'official' doctrine (cf. Nuchelmans 1973:24-25). Yet it is sometimes obscured by terminological equivocations. It may occur that sunthesis is used instead of sumplokë. It may also happen that truth and falsity are envisaged as, respectively, 'composition' and 'division' of statements and states of affairs. A statement consists of one onoma and one rhēma. The notion of onoma is rather straightforward. It is exemplified by word-forms like anthröpos ('man') and hugieia ('health'), which are in the nominative. In its primary use, onoma clearly stands for '(grammatical) subject', but it also has 'noun', and 'word' as its second­ ary meanings. Rhēma is more complicated. Because it is defined by a number of features, there is some disagreement as to what Aristotle really meant by rhëma. I shall now enumerate its characteristic properties as I understand them. i) The rhëma is exemplified by word-forms like hugiainei ('enjoys-health') and badizei ('walks'). Corresponding negated and/or non-present forms are not primary instances of rhëma. ii) A rhēma both signifies (sēmainei) something and consignifies (prossēmaineï) something. For instance, hugiainei signifies 'health' (hugieia) and consignifies the present tense. iii) A rhëma always indicates that something is being said about something else. iv) A typical rhëma contains the word einai ('to be', the copula). When this is apparently not the case, as with the word-form badizei ('walks'), the word-form must nevertheless be analyzed as containing the copula, i.e. as badizön esti ('is walking'; more perspicuously e.g. in the French [ marche = [il] est marchant). What is left, once the copula has been removed from the rhëma, is an onoma (here: badizön).

178

UNIVERSAL HISTORY OF LINGUISTICS

v) The copula which is either explicitly or implicitly contained in a rhēma consignifies a combination (swnplokē)performed by the mind. Thus it is the rhēma which, thanks to its copula, makes the statement into what it is. vi) When (what might otherwise appear to be) a rhēma is uttered alone, it is an onoma. The crucial passages for iv) are to be found in somewhat later parts of De Interpretatione, i.e. in 20a, 1 and 21b, 5 (and also in Metaphysica 1017a, 35). An argument for v) occurs also in De Anima 430a, 25 and 430b, 5: "Where the alternative of true or false applies, there we always find a putting together of objects of thought in a quasi-unity. ... In each and every case that which unifies is the mind." The points ii) and v) seem to indicate quite clearly that the copula has no signification (i.e. lexical meaning), but only a consignification (i.e. grammatical meaning). It is said in iii) that the rhëma is a sign (sēmeion) of predication. The point vi) is obscure. Perhaps it means to say that a verb uttered alone is just a word. According to a more elaborate interpretation, the onoma means here the infinitive. The preceding interpretation may be summed up as follows:

EUROPE

179

It is not difficult to see why Aristotle wishes to reduce all statements to the canonical form 'A is B'. This is precisely the form which the statements figuring in a syllogism must exemplify, with the understanding that 'A' may be qualified by 'every' or 'some', and that the statement itself may be negated. Now, it is true that, when dealing e.g. with the sentence-type 'Every A is B', Aristotle also uses the following expressions: 'B is predicated of every A' (to  kata pantos tou A katēgoreitai), ' belongs to every A' (to  panti tōi A huparkhei), and 'A is entirely contained in B' (to A en holōi esti tōi B). It seems clear, however, that these expressions belong to the metalinguistic context where the nature of the syllogism is being discussed. When Aristotle has to show what kinds of statements actually occur in syllogisms, the 'A is B' format seems to prevail. Thus, what we have here is a procedure of converting natural-language sentences into the corresponding 'logical forms', i.e. forms required by a given type of (formal) logic. It goes without saying that Aristotle's logic is not the relational predicate logic of today, but rather a logic of classes or a logic with oneplace predicates. But otherwise his procedure differs in nothing from current attempts to assign 'logical forms' to sentences. What, precisely, is the relation of the logical level to the levels of language, mind, and reality? It seems fair to say that Aristotle never gave a satisfactory answer to this question. As can be seen from the diagram on p. 178, '' is defined not just by its role within a statement, but also by its referring to a 'sub­ stance'. A 'primary' substance (e.g. a particular man) consists of matter and form ('form' being identical with, e.g., the concept, or species, 'man'), whereas 'second­ ary' substances are exemplified by species and genera (e.g. 'man' and 'animal', respectively). An 'onoma2', in turn, may refer e.g. to species, genera, and such qualities as 'white' or 'walks'. Thus we see that the two types of 'onoma' (and, by implication, the notion of 'rhēma') are defined ontologically. An ontological point of view is implicit also in the expressions 'B belongs to A' and 'A is contained in B', which make syllogistics appear as a logic of class inclusion or exclusion. By contrast, the expression 'B is predicated of A' bears witness to an intralinguistic approach. Again, the 'official' definition of a statement as resulting from a sumplokë noëmatön seems to make logic a matter of psychology. To further complicate the issue, however, Aristotle does not mean to say that logic is simply identical with the mental processes of inference as they actually occur in individual persons. This becomes overwhelmingly clear from his extensive discussion, in Book II of the Anatytica Priora, of the various logical errors that

180

UNIVERSAL HISTORY OF LINGUISTICS

people are likely to commit. He seems to imply, without saying so explicitly, that logic is a general norm which everyone can be made to grasp, if only given proper coaching, and which thus can — ideally — come to possess deterministic psycho­ logical force (cf. in particular Ana, Pr. 66b, 5-67b, 25). If Aristotle does not distinguish more carefully between the different levels, the reason must be that he views them as generally isomorphic. Thus our discussion of Aristotelian logic lends further support to the thesis that, for Aristotle, there is a general (isomorphic) correspondence between language, mind, and reality, logic being somehow the fourth term in this relation of multiple correspondence. By way of a summary, I reproduce here Aristotle's well-known 'square of oppositions'. An additional reason for doing so is that within modern semiotics this square has been reinvented, and renamed the 'semiotic square'. Today it serves as the theoretical foundation of countless semiotic studies, which just goes to show, once again, that what in the human sciences appears as progress is more often than not just ignorance of the past.

If the laws of negation exhibited by this square are applied to the deontic domain (where 'Op' means 'one ought to see to it that p'), we get the following:

When translated into French, this produces the celebrated 'semiotic' fourway distinction (cf. Greimas 1983:82-90).

EUROPE

181

By now I have presented what I consider the most important in Aristotle's linguistic thinking. In the remainder of this section I shall more briefly discuss a few residual issues. In the Poética Aristotle develops a rudimentary classification of sounds and of word-classes (= 'parts-of-speech'). Its intrinsic interest is rather slight, and in any case it was soon superseded. Of much greater, though less obvious, interest are the many semantic analyses that Aristotle offers in the various parts of his oeuvre. One must in particular mention the 'Philosophical Lexicon' contained in Book V of the Metaphysica, where the meanings of more than thirty key terms such as begin­ ning, element, potency, sameness, otherness, difference etc. are analyzed. In many cases it may be asked whether any genuine progress in the definition of these terms has been achieved since Aristotle's time. Because of Aristotle's overpowering role, many of his ideas which had no direct connection with linguistic theory, nevertheless influenced Western linguistics in the following centuries. His notion of causation is a case in point. His 'official' doctrine is based on what might be called the 'artisan analogy' (cf. Physica 194b, 25-30, and Metaphysica 1013a, 25-30). For instance, when someone builds a house, the entire process may be analyzed into four 'causes': the materials which the builder uses (= 'material cause'); the idea or plan which the builder eventually realizes (= 'formal cause'); the builder himself, or rather his actions which put the house together (= 'efficient cause'); the purpose which the house is meant to serve, or 'that for the sake of which' the house is there (= 'final cause'). For Aristotle, this is the primary instance of causation, which he then generalizes to non-human nature: "Now surely as in intelligent action, so in nature" (Physica 199a, 10). But it quickly turns out that the artisan analogy does not work so well outside the sphere of intelligent action. As Aristotle sees it, the ontologically primary entities are particular things or 'substances', which are composed of form and matter. The exact relation between these two components may, however, vary in particular cases: there is a continuum leading from almost unformed matter to fully formed matter. Typically, this continuum is given a diachronic interpret­ ation: in biological growth the starting point (e.g. a seed or an embryo) represents what is unformed, whereas the end point (i.e. the tree or the grown-up man)

182

UNIVERSAL HISTORY OF LINGUISTICS

represents what is formed. But now the distinction between final cause and formal cause is seen to vanish: "... in things which come to be and are by nature ... the form must be the cause in the sense of 'that for the sake of which' " (Physica 199a, 30). Thus, there is an unresolved tension between the artisan analogy and the 'embryo analogy'. Yet what is common to both, is the firm belief that all pro­ cesses, human and non-human alike, are of teleological character (as we would say). It is interesting to note that on this issue Aristotle explicitly rejects the view of earlier philosophers who had argued either implicitly (like Empedocles) or explicitly (like Democritus) that the entire universe is governed by efficient (= nonteleological), deterministic causality. From today's perspective, Aristotle's physics represents rather simple-minded anthropomorphism (even if he makes it quite clear that nature has no conscious goals). His theory of (intelligent) action, by contrast, has not been superseded yet. In thefieldof psychology Aristotle remained the principal authority until the 17th century. His concept of mind was much too passive to be acceptable today. Yet although his view of how concepts develop and function was wrong, he was very muchrightin his analysis of what is the content of concepts, as representatives of 'non-empirical psychology' within the artificial-ingelligence movement have been forced to realize (cf. CLT: 311-312). Moreover, Aristotle said several pertinent things about the mind—body problem. To begin with, he rejects any mind—body dualism, because assuming mind and body to be separate, the first question concerns the nature of their interaction; but this question could not be answered by Aristotle's contemporaries (cf. De Anima 407b, 15-20), nor — we can add — has it been answered until the present day. In giving his own answer, Aristotle makes use of his notion of 'substance', on the one hand, and of a crucial analogy, on the other. First of all, body and mind are as inseparable as matter and form, i.e. the matter of a thing and the thing of which it is the matter. Second, the totality of mind and body is comparable to the totality of an instrument. That is, an instrument too consists of matter and form, the latter being identical with the use of the instrument. (Today it is more customary to analyze an instrument into 'form' and 'function'.) Third, both mind and use are divided into what is potential, and what is actual. In the case of mind, the two are represented by sleeping and waking, whereas in the case of using an instrument like axe, the two are represented by the 'power' of the instrument, i.e. its ability to cut, and its cutting (De Anima 412a, 1-413 a, 5).

EUROPE

183

The foregoing solution to the mind—body problem possesses considerable interest for linguistic theory, because it reveals the underlying unity of the funda­ mental psychological and linguistic questions. The instrument analogy applies, of course, not just to body and mind, but also to form and meaning; just recall the slogan 'Meaning is use' (cf. CLT: 154-156). Notice, however, that here mind is not claimed to be identical with, but merely parallel to, meaning. In the Politica Aristotle develops a comprehensive political theory which contains thorough discussions of rival conceptions as well as 'case studies' of Greek city-states; and it is well known that he defined man as zōon politikon, i.e. political (or social) animal. In view of this, it is somewhat surprising that his theory of language is formulated in terms of individual psychology. This legacy of psychologism meant that Western linguists and philosophers of language had to struggle for a long time to grasp the importance of the social dimension; and still today they expérience great difficulties in defining what is meant by 'social'. The following explanation might be suggested for this prima facie contradiction in Aristotle's thinking. Although he had an extremely wide range of interests, they were not of equal value. The highest good is the 'contemplative life'; in fact, God's sole activity is defined as 'contemplation'. In such a context the role of language easily dwindles into being just a verbalization of the thoughts of a solitary mind.

5.1.3 The Stoics In the post-Classical Greece the most influential philosophical school was Stoicism, created by Zeno of Citium (d. 261 B.C.) and perfected by Chrysippus (d. 208 B.C.). Unlike Plato and Aristotle, the Stoics developed a full-fledged grammatical theory and philosophy of language. Unfortunately, practically all of their writings have been lost, and their original doctrine must be pieced together from comments made by their later adherents and adversaries. Given this predicament, it is interesting to note that in the last decades there has been a real Renaissance of studies on Stoicism. It seems safe to say that this awakening of interest is to a large extent due to Lukasiewicz's (1935) rehabilitation of Stoic logic, continued by Mates (1953). After more than 2,000 years of general neglect and even ridicule, it has been discovered that the Stoics developed a logical theory closely resembling modern prepositional logic, and also constructed methods of formal proof. Unable to appreciate these results, the rival schools disparaged Stoic logic for its alleged triviality; and the loss of the original writings, com­ pounded by the notational errors in secondary sources, made the reconstruction of

184

UNIVERSAL HISTORY OF LINGUISTICS

Stoic logic later nearly impossible (for an overview, see Kneale & Kneale 1962: 158-176). Understandably enough, on many issues there is disagreement as to what the Stoics really said. I have chosen the interpretation that looks the most plausible to me. The Stoic grammar survives in the work of later grammarians, so I will not discuss its details here. Rather, I shall concentrate on the fundamentals of the Stoic theory of language. The Stoic philosophy constitutes a comprehensive and systematic whole. It is subdivided into three main areas: 'physics' (including psychology), ethics, and 'logic'. 'Logic' in turn consists of rhetoric and dialectic. Finally, dialectic contains two parts, according as it treats of form or of meaning. The terms for 'form' and 'meaning' are, respectively, sēmainon ('signifying') and sēmainomenon ('signified') or lekton ('what is, or can be, said'). The first thing to note about this classification is that semantics is dis­ tinguished from psychology: meanings are not mental entities (like impressions or cognitions). The Stoic metaphysics is based on a dichotomy between corporeal and incorporeal entities, with a strong predilection for what is corporeal. This gives the Stoic metaphysics its 'nominalist' character.4 Still, although the number of incor­ poreal entities or abstractions is kept to a minimum, it is impossible to dispense with them altogether. Thus, four broad classes of abstractions are allowed to exist (or rather, 'subsist'): void, place, time, and lekta (= the plural of lekton). Lekta include not only meanings, but also the four categories of corporeal entities, which — qua categories and as opposed to their members — must be admitted to be abstractions: substance, quality, disposition, and relative disposition (or more simply, relation) (cf. Rist 1969: Chap. 9). Unlike the standard Western type of ontology, the Stoic ontology is not based on the distinction between concrete thing and abstract quality (or relation). Rather, qualities and relations are taken to be corporeal entities and thus to exist at the same concrete and observable level as things are. The Stoics were the first to propose an explicitly non-psychologistic theory of meaning. It proved to be difficult, however, to explain the nature of meaningsas-lekta in more detail, apart from their being distinct from mental entities. This is why the psychologistic meaning-conception ultimately prevailed, and it has con­ tinued to be the majority opinion in the West until the present day. The Stoic view of the 'language—mind—reality' trichotomy may be summed up as follows:

EUROPE

185

There are several things to be noted about this diagram. The Stoics were keen to point out various cases of discrepancy between meaning and form. For instance, privation, negation, and opposition are conceptually three different things, and should be expressed correspondingly (as in 'colourful vs. colourless', 'equal vs. unequal', 'white vs. ©black©'). In practice, however, the different means of ex­ pression get thoroughly mixed up. Or to give a more obvious example, it occurs in Classical Greek that masculine notions are expressed by feminine forms, and vice versa. Such cases support the Stoic view that the governing principle of natural language is not regularity, called analogíā, but rather the lack of it, called anōmalíā (for more discussion, cf. pp. 196-200). The 'anomaly' view apparently resulted from comparing natural language to the ideal of 'one meaning — one form'. There are indications that the Stoics actually tried to change natural language so as to make it agree with this ideal. Alternatively this may be interpreted as an attempt to create an ideal logical language (cf. Blank 1982:35). As we have seen, Plato and Aristotle think that mind is internal language which precedes, and is expressed by, oral, external language. The Stoics too assume a distinction between internal language and external language, but view the relation between the two in more complex terms. Meanings and thoughts seem to have the same contents, but to differ as to their ontological status ( = incorporeal vs. corporeal). However, thoughts do not simply precede language. Rather, they are created out of sense-impressions with the help of language (cf. Long 1974: 124-125). In the above diagram this interaction between mind and language is indicated by the two-headed arrow. Once thoughts have been created, they consti­ tute the internal language. It is the absence of internal, rather than external, language which, according to the Stoics, separates the animals from the humans. Although the Stoics recognize the importance of language in the develop­ ment of thoughts, they agree with Plato and Aristotle in taking mind, or its results, to be of universal character. It is worth pointing out that this view did not go uncontested in Greek philosophy: Epicurus argued that different nations create different ways of thinking (cf. Steinthal 1971 [1890]: 325-326).

186

UNIVERSAL HISTORY OF LINGUISTICS

Finally, their nominalist metaphysics prevents the Stoics from assuming that reality is segmented in genera and species as straightforwardly and unproblematically as is claimed by Plato and Aristotle. Rather, according to Graeser (1978: 79-80), the Stoics regarded the external world as a 'moving continuum' and realized that "it is our mind that divides and articulates reality, arriving at logical constituents even where the physical constituents are inseparable". In this context, as in some others (cf. Long 1974:145-147), Heraclitus' influence is evident. For the present purpose, the most important implication is the following. Both Aristotle and the Stoics assumed that (in addition to its other functions) mind refers, language-like, to reality; but for Aristotle mind merely reproduces reality whereas for the Stoics mind both reproduces and creates reality. Remember, however, that the element of creation is supposed to be universal, and not dependent on individual or national variation. Next I shall delve more deeply into the difficult notion of lekton. In the linguistic context the study of lekton equals the study of semantics. Within the Western tradition, the Stoics have been nearly unique in their systematic concen­ tration on sentence-level semantics. First of all, a lekton is either complete (autoteles) or incomplete (ellipes). The subcategories of the complete lekton are statement (axiōma), question, inquiry, request, prayer, etc. (Because several of the complete lekta are expressed ident­ ically, this classification was later criticized; but it is acceptable on semantic grounds, especially from the point of view of an 'anomalist' wishing to emphasize the non-correspondence between meaning and form.) In what follows I shall concentrate on statements. They are either simple or complex. Simple statements are classified according to two parameters: on the one hand, there are three degrees of (in)definiteness; on the other, there are one type of affirmation and three types of negation (corresponding to 'contradictory', 'contrary', and 'privative'). Complex statements result from combining two simple statements with the aid of a conjunc­ tion. Let us consider the internal structure of a statement which is simple, affirmative, and — technically speaking — 'imperfectly definite'. This category is exemplified by Piatön badizei ('Plato is walking') and by Piatön Diöna paideuei ('Plato teaches Dion'). In the former example badizei is an incomplete lekton, also called katëgorëma (= 'predicate'); to be more precise, it qualifies as a complete predicate. Piatön in turn is called ptōsis (usually meaning 'grammatical case'), more precisely ptōsis orthës ('nominative'). In the latter example Diōna paideuei is a complete predicate; paideuei is an incomplete predicate, also called elation ë katëgorëma ('less than predicate'); Diöna is called ptōsis plagia ('oblique case').

EUROPE

187

All the information about statements that was given in the last two para­ graphs can be summarized with the aid of four rules of sentence-formation. I have selected them from Egli's (1987) more comprehensive list of rules, which consti­ tutes an ambitious reconstruction of the Stoic syntax-cum-semantics. The rules are reproduced here with slight modifications: i) ii) iii) iv)

An oblique case and an incomplete predicate form a complete predicate. A nominative and a complete predicate form a statement. A negation sign and a statement form a statement. Two statements combined by a conjunction form a statement.

The rules iii) and iv) are recursive. There are additional (non-recursive) rules for converting a statement into a ptōsis (i.e. accusative-cum-infinitive con­ struction) and a predicate into & ptōsis (i.e. infinitive). It must be emphasized that these rules are meant to operate at the semantic level. Thus what we have here is something like 'syntax of meanings'. There is a parallel 'syntax of forms'. For instance, the formal counterparts of katëgorêma and ptosis are rhēma and onoma. The classification of lekta clearly implies that they are sentence-size units. Here an inconsistency becomes apparent, however, because also ontological categories qua categories qualify as lekta (cf. above).5 Still, because it is the predicate, and not the ptōsis, which is called an incomplete lekton, the Stoics clearly adopted the view that at the semantic level the meaning equivalent of the finite verb is more important than, or 'governs', the meaning equivalents of the nouns. We have seen that Aristotle had his own reasons for adopting the same position, as did Pānini too. A statement {axiōma) is true or false, which means that only a lekton (though not every lekton) can be the bearer of truth. Now it seems odd to say that if we have a truth like 'The grass is green', what is true is the meaning of this sentence. One would like to say that what is true is this (meaningful) sentence-asstated. To be sure, this sentence-as-stated is true, in whatever language it is uttered, as long as the meaning remains the same. This indicates that the actual physical form is irrelevant. Yet this does not have to mean that the only alternative left is to accept meanings as truth-bearers. It is possible to choose an abstract '(meaningful) sentence-as-stated' for this job (the basis of abstraction being the similarity of the norms governing form—meaning correlations in different languages). I do not mean to say that the Stoics actually proposed this solution. But the ambiguity of the

188

UNIVERSAL HISTORY OF LINGUISTICS

notion of lekton, which means both 'meaning' and 'abstraction', certainly makes it possible to adopt this solution within the Stoic framework. The same ambiguity recurs in another context, namely when we try to determine the status of ptōsis. Although it does not qualify even as an incomplete lekton, it undeniably belongs to the level of lekta, because it is a necessary part of statements (as well as other speech acts). Therefore it is only logical that for the Stoics the analysis of the grammatical cases is part of the analysis of sēmainomenon /lekton. Besides, if it were considered as part of the analysis of sēmainon, then — it was argued — e.g. all the formal variants of the genitive singular ought to qualify as separate case endings (cf. Nuchelmans 1973:73). Now this argument can be interpreted in two different ways, due to the ambiguity of lekton: either the Stoics invented the notion of morpheme as an abstract form, i.e. an abstraction based on several (concrete) forms; or they were talking about the meaning common to a set of forms (e.g. 'possession', in the case of genitive endings). One last point about (complete) lekta. Apparently the Stoics analyzed the speech acts into two components: an abstract core, called prāgma, and an attitude or action directed towards it (called enklisis). For instance, an axiōma was divided into a prāgma indicating the semantic content and an act of apophainesthai ('to state') (cf. Nuchelmans 1973:64, 77-78). Today the same distinction is made by using the terms 'proposition' and 'illocutionary force'. It is probable that the Stoics themselves considered ethics as the most important part of their overall system. They agreed with Plato and Aristotle in holding the notions of justice and injustice to be phusei, i.e. binding on all men. In defending this position, they were responding to the Sceptics who, led by Pyrrho (d. 270 B.C.) and following in the footsteps of the Sophists, argued that nothing can be known with certainty, i.e. neither what things-in-themselves are like nor whether actions are morally good or bad. Seeing the implications of an absolute ethical relativism (or scepticism), it is easy to agree on this issue with the Stoics. However, they argued that language too exists phusei. This view is correct insofar as it just means to say that the linguistic capacity is part of human nature. As we have already seen, the Stoics indeed claimed that what distinguishes humans from animals is their possession of language, in the sense of not just being able to emit sounds, but of having articu­ late, sentence-like thoughts. However, when they had to justify the phusei view in detail, they could not think of anything better than repeating the arguments Plato had given in the Cratylus. But this means that they had to contradict their own 'anomaly' view of language: If there is an isomorphic correspondence between language and reality (and, by implication, between form and meaning), as claimed

EUROPE

189

by the phusei view, then this correspondence cannot be lacking, as claimed by the 'anomaly' view. There is a dualistic streak both in Plato's and in Aristotle's thinking: either there is the distinction between the eternal and immutable 'forms' and the temporal and mutable reality; or there is the distinction between the 'superlunary' and 'sublunary' regions of the universe, such that in the former the movement is circular and eternal whereas in the latter it is rectilinear and short-lived. As against this, the Stoic cosmology is remarkably homogeneous. There is a single principle, or logos, governing the entire universe, and both inanimate and animate beings are subordinated to the same laws of causation. This cosmological unity finds ex­ pression in a wide-ranging isomorphism between the (human) microcosmos and the (physical) macrocosmos: Other philosophers, such as Plato in Timaeus (30b), had described the universe as a zōon; but none had applied the analogy as extensively as the Stoics. In their cosmology, the universe, like a living being, was born, lived an allotted span of time, and died; like a living being it reproduced itself. The universe, like a living body, was animated by a vital heat or vital breath that passed through its entire extent; again, as with an ani­ mated body, all its parts were interconnected (Lapidge 1978:163). This shows an unmistakable similarity with the Brahminic cosmology, on the one hand (cf. p. 7), and the neo-Confucian cosmology, on the other (cf. p. 117).

5.1.4 The Relation between Mind and Reality in Greek Philosophy Rorty (1980) has given an influential account of the history of philosophy in which he defends the thesis of cultural determinism. According to him, the concept of mind in a strong sense was invented only by Descartes, i.e. not before the 17th century, and should now be abandoned in turn. For the ancient philosophers, as represented mainly by Plato and Aristotle, the mind was supposedly passive or mirror-like insofar as it was assumed to reflect reality directly; all ancient philos­ ophers were alike in being unable to grant an autonomous status to such mental reflections, or to question their trustworthiness (op. cit., pp. 45-61). There are several reasons for rejecting Rorty's account. Apart from doing less than full justice to Aristotle's psychology, it fails to mention that the Stoics had

190

UNIVERSAL HISTORY OF LINGUISTICS

a more complex view of the relation between mind and reality (cf. above). More importantly, Democritus' role as an early representative of empiricist epistemology is ignored. Yet, long before Locke, he made very clearly the distinction between primary and secondary perceptual qualities. The former, including perceptions of situatedness, size, order, and weight, corresponded to the nature, or phusis, of things, i.e. to the various configurations of atoms. The latter, including perceptions of taste, warmth, and colour, derived from the perceiver, and thus existed nomōi. Steinthal (1971 [1890]:45) gives the following résumé, in direct refutation of Rorty (1980): Vor Demokrit hatte man nur das Object im Auge, und die Subjectivitāt des Denkenden blieb ganz unbeachtet. ... Wie Vorstellungen, Erkenntnisse von den Dingen entstehen,fragteman nicht. Demokrit lenkte die Aufmerksam­ keit gerade hierauf und gab so die erste Anregung zur Psychologie und Er­ kenntnislehre. We have also seen that the Sceptics held a doctrine incompatible with Rorty's thesis of cultural determinism. According to Pyrrho, "sense-perception reveals 'what appears' to the percipient; but 'what appears' cannot be used as sound evidence from which to infer 'what is'" (Long 1974:82). "Since sence-perception, in the view of the Pyrrhonists, provides no grounds for judgments about things-inthemselves, we should not make such judgments" (p. 84). Now it does not make sense to claim, with Rorty, that all Greek philosophers regarded perceptions as a transparent guide to things, if Pyrrho and his disciples regarded perceptions as absolutely no guide to things. To be sure, Carneades, one of the 'Academic' Sceptics, developed a more informative approach to the 'mind—reality' problem. While admitting that no single perception can be absolutely trustworthy, he argued that the relative trustworthiness of perceptions increases in proportion to their mutual coherence (pp. 95-99). For this reason, he "is closer to the spirit of modern British philosophy than perhaps any other ancient thinker" (p. 106). Perhaps the most puzzling aspect of Rorty's account is his apparent aware­ ness that he may have got his facts wrong. In footnote 21 on p. 51 he mentions Michael Frede's suggestion that acquaintance with Hellenistic philosophy might reveal more continuities in the history of philosophy than any 'invention of the mind' hypothesis can allow of. Rather than following up this lead, however, Rorty prefers to maintain (and in fact, to 'radicalize') the standard "Gilson—Randall historiographical tradition", which emphasizes the discontinuities between what went before and after the year 1600. This view, at least in its radicalized form, amounts to a version of cultural determinism: the Weltanschauung of one period

EUROPE

191

changes completely vis-à-vis that of the earlier period, with the consequence that the latter is henceforth almost certain to be misunderstood. It is easy to understand why cultural determinism should exert such a strong attraction. On the one hand, it produces spectacular results: people tend to be thrilled by the idea that there are completely different, mutually impenetrable ways of thinking (either in succession or in juxtaposition). To put it briefly, disconti­ nuities, or 'sudden breaks', in history are interesting whereas continuities are boring. On the other hand, such spectacular results are achieved with little effort, namely by capitalizing on a few 'key ideas' and denying the real complexity of historical facts. The price one has to pay is lack of truth and also loss of insight. In Rorty's case this is particularly evident. He recommends abandoning the Cartesian mind—body dualism in favor of a Wittgensteinian solution. His rigid periodization of history, however, prevents him from seeing that Aristotle anticipated Wittgen­ stein. Recall the analogy between the distinctions between mind and body, on the one hand, and between meaning-as-use and form, on the other (cf. the end of 5.2). There is a simple way to get rid of the myth of cultural determinism: just think of Heraclitus and Parmenides. For one, everything was in flux; for the other, nothing was. Certainly they also had something in common. But so have Aristotle and Wittgenstein, not to speak of Pānini's and Sībawaihi's similarity with modern Western linguistics.

5.2 The Alexandrian School In the Hellenistic world Athens lost its leading role and was eclipsed by such cultural centers as Pergamon in Asia Minor and Alexandria in Egypt (of which the latter maintained its position also during the Roman Empire). Greek was the common language of the Hellenistic culture, and Homer's epics were the canonical text composed in Greek. Around 300 B.C. Homer's language was becoming opaque even to educated people; and his text was preserved in manuscripts which, apart from containing numerous errors, diverged from one another more or less widely. So there arose a need for experts who could, on the one hand, restore Homer's text into its original form and, on the other, make it comprehensible to the educated lay audience. This was the beginning of Alexandrian philology. The term 'Alexandrian School' is not meant to imply any strong doctrinal unity among its representatives. Rather, it is a convenient way of indicating that during some four hundred years the most important linguistic research within the 'Western' tradition was largely carried out in Alexandria. The most prominent

192

UNIVERSAL HISTORY OF LINGUISTICS

members of this 'school' are Aristarchus (d. 144 B.C.), who was a textual critic and a philologist in the modern sense of this word, Dionysius Thrax (d. c. 100 B.C.), who is generally credited with having written the oldest extant treatise on phonology and morphology in the West, and Apollonius Dyscolus (d.  150 A.D.), who wrote the oldest extant treatise on syntax in the West. A philological concern with Homeric texts developed gradually into a genuine grammatical theory. During this entire process, the Stoic grammar was available, and it clearly influenced both Dionysius and Apollonius. The relation between the Stoics and the Alexandrians could perhaps be depicted as follows:

The Alexandrian School left its imprint also on the Latin grammatical tradition, as witnessed by Varro and, even more strongly, by Priscian.

5.2.1 Dionysius Thrax There is general agreement concerning the gradual emergence of the Alexandrian grammar. Steinthal (1971 [1891]:82-111) argues that Aristarchus had no explicit notions of morphology or of syntax; rather, while making his emendations in Homeric manuscripts, he was just relying on his linguistic intuition and common sense. The same view has recently been reasserted by Taylor (1987:11-13). As far as the recorded history is concerned, Greek phonology and morphology were treated systematically for the first time in a short work known as Tekhnē grammatikē.Traditionally, it has been attributed to Dionysius Thrax, who was one of Aristarchus' disciples. This tradition has been contested more recently, for reasons that do not seem entirely cogent to me (cf. the end of this subsection). The Tekhnë grammatikë is, in the modern format, a text of 13 pages, divided into 20 sections (cf. Kemp 1987). Thefirstsection announces six desiderata for grammatical study, of which only the first, i.e. 'skill in reading aloud', and the fifth, i.e. 'detailed account of regular patterns' (= analogiai), are taken into account in the subsequent text.6

EUROPE

193

The sections 6-10 contain a classification of 'letters' (i.e. sounds) and syllables. The section 11 contains succinct definitions of 'word' and 'sentence', and enumerates eight word-classes (merē tou logou, lit. 'parts of the sentence'): noun, verb, participle, article, pronoun, preposition, adverb, conjunction. Each of the remaining sections is devoted in the order just given to one word-class (with the exception that conjugation, which is one of the eight 'attributes' of the verb, is treated separately in one section). The longest and most important sections are devoted to the noun and the verb. The attributes of both are enumerated and normally also exemplified, five for the noun and eight for the verb. The attributes of the noun are gender, 'species', 'shape', number, and case. There are three genders: masculine, feminine, and neuter. There are two 'species': primitive (i.e. non-derived) and derivative, of which the latter contains seven subcategories (including the comparative and superlative forms of the adjective). There are three 'shapes': simple, compound, and doubly compound, of which the second contains four subcategories. There are three numbers: singular, dual, and plural. There are five cases: nominative, genitive, dative, accusative,7 and voca­ tive. — A list of 24 subcategories of the noun is given in addition, including the positive forms of the adjective and the interrogative pronouns and adjectives. In general the criteria for subcategorization are semantic and somewhat arbitrary. — It is worth pointing out that as yet no distinction is made between inflection (= gender, number, case) and derivation/compounding (= 'species', 'shape'). The attributes of the verb are mood, genus, species, shape, number, person, tense, conjugation. There are five moods: indicative, imperative, optative, subjunc­ tive, infinitive. There are three genera: active, passive, and medial (of which the third does not yet possess its current meaning). There are two species: primitive and derivative. There are three shapes: simple, compound, and doubly compound. There are three numbers: singular, dual, and plural. There are three persons: first, second, and third. There are three tenses: present, past, and future, of which the second is subdivided into imperfect, perfect, pluperfect, and aorist. The conju­ gations are distinguished from one another on rather superficial grounds. The participle is considered a word-class of its own, having attributes both of the noun and of the verb. 'Article' contains, because of a formal similarity, both the definite article and the relative pronoun (ho and hos in masc., nom., sing.). The adverbs are presented in a list containing twenty-six semantic subcategories. The eight subcategories of 'conjunction' are still in use today.

194

UNIVERSAL HISTORY OF LINGUISTICS

As can be seen from what precedes, the classification of the word-classes is based primarily on formal criteria. These are much the same as those employed by the Stoics in their analysis of sēmainon. Yet the rationale of the Stoic grammar was quite different from that of the Alexandrian grammar. The Stoics were interested in sëmainon, or linguistic form, only insofar as it was the inevitable (and far from perfect) vehicle of sēmainomenon/lekton, or linguistic meaning. The Alexandrians, by contrast, were interested in linguistic form for its own sake, even if they freely complemented their formal classifications with semantic criteria. For instance, the noun and the verb were not defined exclusively by means of their respective attributes. It was added that the noun signifies something corporeal or incorporeal, and that the verb signifies activity or passivity. It is immediately evident that Dionysius Thrax achieved lasting results: his morphological categories and subcategories are to a large extent identical with those we use today. It is no less evident, however, that his approach is rudimentary: the distinctions made are not justified in any way, and there is no attempt at theoretical description. The best way to bring out the non-theoretical nature of the Tekhnē grammatikē is to compare it with the Astādhyāyī. As I noted before, the authenticity of the Tekhnë as a work written by Dionysius Thrax has been seriously questioned. This debate is summarized in Taylor (1987:8-11). There are two main reasons for postponing the date of the Tekhnë by a couple of hundred years. First, the inconsistency of the first section with the rest seems to indicate that what we have here is in fact a later compilation. Second, some people find it hard to believe that a work of such high quality could have been produced at such an early stage in the Western history of linguistics. At least the second argument appears to me wholly unconvincing. I would rather like to use an opposite argument. It is hard to believe that a work like the Tekhnë could not have been produced, once the Stoic grammar had been around for more than one hundred years. So I am willing to ascribe the Tekhnë to Dionysius, while admitting that it may not have been preserved in its integral form.

5.2.2 An Interlude: Varro It may seem surprising that the Roman scholar Marcus Terentius Varro (116-27 B.C.) is discussed here in a chapter devoted to the Alexandrian School. Yet this is justified insofar as, within linguistics, Varro's principal sources of inspiration were the Alexandrians, on the one hand, and the Stoics, on the other, with greater emphasis on the former. Nevertheless, it seems appropriate to characterize Varro

EUROPE

195

as an 'interlude'. Among the Roman grammarians he is unique in treating his Greek predecessors as his equals, and having the right to do so. After him, the grammati­ cal tradition of Rome (and of Constantinople, for that matter) consisted merely in perpetuating the Alexandrian heritage. As a polymath, Varro was something like the Roman equivalent of Aristotle. He wrote no less than 76 works, covering practically all branches of the contemporary science and literature. All of his works have been lost, except a treatise on agriculture and, in an incomplete form, the books V-X of his De Lingua Latina, which originally contained 25 books. The books V-VII deal with etymology while the books VIII-X deal with (synchronic) morphology from the perspective of the 'analogy vs. anomaly' controversy. (Whether or not such a controversy ever took place, except in Varro's imagination, will be discussed at the end of this subsection.) It is especially regrettable that Varro's books on syntax have been lost. According to Varro's definition, etymology is about how words have been or are being created (V, 1-2). In both cases it is a matter of 'imposing words upon things'. It may not be possible to explain the etymology of such 'primitive' words as ago ('I drive') or homo ('man'). However, the vast majority of the words result from applying a process called declinatio to the primitive words. Suppose there are 1,000 primitive words. Then from each such word we get some 500 new words (or word-forms) by means of declinatio. Let us further suppose that there are ten prefixes which can be freely added to the existing words. Now the total number of the words is 1,000 x 500 x 10 = 5 millions, or practically infinite (innumerabilis). (One is reminded here of Patañjali's rather similar argument; cf. p. 78). Even if the etymologist cannot explain the origin of the primitive words, he can explain the origin of 4,999,000 words, which is quite satisfactory (VI, 36-39; VII, 4). In this context 'etymology' covers both explaining the history of words and presenting language as a synchronic (inflectional) system. In light of the preceding considerations Varro reaches the following general conclusion concerning the nature of words: The origins of words are therefore two in number, and no more: impositio and declinatio; the one is as it were the spring, the other the brook. Men have wished that imposed nouns should be as few as possible, that they might be able to learn them more quickly; but derivative nouns they have wished to be as numerous as possible, that all might more easily say those nouns which they needed to use (VIII, 5).

196

UNIVERSAL HISTORY OF LINGUISTICS

It may be noted in passing that Varro is proposing here the agglutinative structure as the ideal structure of language, considering that his claims are meant to be universally valid: The declinatio has been introduced not only into Latin speech, but into the speech of all men, because it is useful and necessary (VIII, 3). In addition to impositio and declinatio, which deal with words, there is a third principle that must be accounted for: "how the words, when suitably combined with one another, produce a sentence" (VIII, 1). This was discussed in Varro's non-extant books on syntax. Given the value he sets in the openness of language at the word-level, it is safe to assume that he emphasized this point even more strongly in his syntax. Next Varro introduces the 'analogy vs. anomaly' controversy (VIII, 23). It is clear that impositio, i.e. name-giving, depends on the will (voluntas) of the namegiver; to this extent it is arbitrary or 'anomalous' (X, 53). On closer inspection, however, declinatio too turns out to possess an element of arbitrariness. Therefore it is divided into declinatio voluntaria ('derivation'), which is the realm of anomaly, and declinatio naturalis ('inflection'), which is the realm of analogy (VIII, 23). The justification for this distinction is obvious. Inflectional paradigms are, on the whole, quite regular, whereas derivation is bound to contain 'anomalous' gaps. Consider the following examples:8

Varro pictures Aristarchus and Chrysippus (with his disciple Crates) as the champions of analogy and anomaly, respectively (IX, 1). Analogy is based on ratio ('theory'), whereas anomaly is based on voluntas, the results of which, once they

EUROPE

197

have been accepted, become part of consuetudo ('usage'). Thus, consuetudo is bound to contain historical accidents which have no theoretical justification. Getting a clear view of what these definitions involve, is hampered by the fact that derivational morphology is not only identified as declinatio voluntaria, but is also taken to be an instance of impositio, namely an impositio that is performed in the present (e.g. IX, 34). Thus it is opposed to the impositio which was per­ formed by the 'original' name-givers, and which produced the primitive words. All these facts may now be summed up in the following diagram:

This overall schema contains the following potential sources of misunder­ standing: I-B) and II-A) are identical; I-A) and I-B) should have been dis­ tinguished more explicitly. On this issue I disagree with Taylor (1975), who simply identifies impositio with declinatio voluntaria (e.g. pp. 25-26). This intepretation cannot be accepted, because Varro explicitly mentions the impositio which was performed by the 'first' name-givers, and which produced such simple 'primitive' words as homo (VIII, 7). Moreover, it would not make much sense to set up impositio against declinatio ("like the spring and the brook"), if the former is in fact a subtype of the latter. This is acknowledged by Taylor himself when he asserts (p. 41) that declinatio voluntaria is a ('generative') process similar to declinatio naturalis. The only thing these processes (= 'the brook') can be contrasted with is their starting point (= 'the spring'), i.e. the impositio of simple primitive words like ago and homo. To be sure, Varro's terminology remains ambiguous in an infelicitous way. He should

198

UNIVERSAL HISTORY OF LINGUISTICS

have been more explicit in distinguishing the ancient impositio from the contempor­ ary one. Varro assumes the role of a judge who has to decide the 'analogy vs. anomaly' issue. He marshals extensive evidence for and against both points of view. His final verdict is that they both belong to the nature of language: But those who give us advice in the matter of speaking, some saying to follow usage and others saying to follow theory, are not so much at variance, because usage and regularity are more closely connected with each other than those advisers think. ... Therefore ... neither Anomaly nor Analogy is to be cast aside, unless man is not of soul because he is of body and of soul (IX, 2-3). When this conclusion is spelled out, it turns out to be a repetition of the earlier claim that analogy and anomaly are exemplified, respectively, by declinatio naturalis, which inflects words in tense or case, and declinatio voluntaria, which e.g. derives Romani ('inhabitants of Rome') from Roma, but Parmenses ('inhabi­ tants of Parma') from Parma. The opposition between the two principles is further softened by admitting that, on the one hand, nature plays a role not only in analogy, but also in anomaly, and, on the other, human will plays a role not only in anomaly, but also in analogy (VI, 3; X, 51). Against this general background Varro sets up a tripartite hierarchy of errors, according as they are committed by nations, individuals, or poets (IX, 4-6). The language of every nation should follow the principle of analogy. If it fails to do so, however, the resulting 'error' must be accepted as just a matter of consuetude. Even in such a case, every individual person must follow the general usage. If he fails to do so, his error is unacceptable. Poets, however, are exempted from this stipulation because they are entitled to make use of their proverbial poetic license. I give two further examples of the many insights that Varro's text contains. He has a functional view of language: "speech has been for utility's sake"; there­ fore, just as with other types of instruments, it is impractical to have formal variation which serves no function (IX, 47-48). The unfalsifiable character of sentences referring to norms seems to be anticipated in a passage which deserves to be quoted in the original: Si quis aliter putat dici oportere atque oportet, non scientiam tollit orationis sed suam inscientiam denudat (IX, 112). ('If someone thinks that people ought to speak differently from the way they in fact ought to speak,

EUROPE

199

he does not falsify the knowledge of language, but merely reveals his own lack of knowledge.') In his grammatical analysis Varro establishes a four-way classification of the word-classes based on the presence vs. absence of case and/or tense (VI, 36):

The words which inflect in case and/or tense are called 'fruitful', whereas the rest is called 'barren' (VIII, 9). While trying to do justice to the 'analogous' aspect of language, Varro represents an approach that can only be characterized as 'structuralist': The identity of a unit is determined by its relations to the neighboring units. This is true both paradigmatically and syntagmatically (as we would say): On the one hand, the nature of the last sounds in the nominative singular forms crux and Phryx becomes evident only when, e.g., the corresponding plural forms are taken into consider­ ation: cruces vs. Phryges (IX, 44). (Analogously in the verb-forms dolo and dolos vs. colo and colis; IX, 108.) On the other hand, word-forms like nemus and lepus look similar, until suitable demonstrative adjectives have been added: hoc nemus ( = neuter) vs. hie lepus ( = masc.) (IX, 94). The two cases are further explained with the aid of vivid metaphors: Taking such related units as cruces and Phryges into consideration is like bringing a lamp into the darkness and revealing the real character of things (here: crux and Phryx). Again, nemus and lepus are like two magnetic stones which, though looking similar, influence particles of steel (here: demonstrative adjectives) differently, once these have been placed nearby. Varro's craving for systematicity is also evident in that he presents inflec­ tional paradigms in the form of a matrix consisting of horizontal and vertical lines: the six cases are placed on the former whereas the three genders (and the two numbers) are placed on the latter (X, 21-23, 43-44). Only part of the paradigm is given explicitly:

200

UNIVERSAL HISTORY OF LINGUISTICS

The systematicity of the paradigm is further illustrated by means of an arithmetical analogy: 1 10 100

2 20 200

4 40 400

On the one hand, the relation between albus and alba is the same as that between albo and albae, just like 1/10 = 2/20. On the other hand, the relation between albus and albo is the same as that between alba and albae, just like 1/2 = 10/20. "Aristarchus and the grammarians of his school" are mentioned as representatives of this kind of proportional analogy in linguistic analysis. In conclusion, I take up the question about the genuinness of the 'analogy vs. anomaly' controversy. Traditionally it has been taken for granted that this was indeed one of the great points of contention in ancient linguistics, fully on a par with the 'phusei vs. nomōV controversy. More recently, however, the traditional account has been called into question. The doubts about its validity are based primarily on the fact that Varro himself is the only independent source who explicitly mentions the existence of this kind of controversy. The relevant facts have been summarized by Taylor (1987:6-8). I have two comments to make. First, if Varro invented the whole contro­ versy, it is hard to understand how he was able to do so without losing face. His readers would have known that he was lying, and would have rebuffed him accordingly. Second, even if the 'analogy vs. anomaly' controversy did not exist in reality, it should have existed. It captures perfectly the opposition between two personality types which are still widely represented in today's linguistics: one who is intent on making systematic (and mathematically expressed) generalizations, regardless of how many facts he has to sacrifice; the other who enjoys describing individual recalcitrant facts and lets the system (supposing there is any) take care of itself.

EUROPE

201

5.2.3 Apollonius Dyscolus The oldest extant syntax in the Western tradition, viz. Peri suntaxeōs by Apollonius Dyscolus, dates from the second century A.D. Apollonius wrote several works, of which the monographs on pronouns, adverbs, and conjunctions have also survived. He lived his whole life in Alexandria, but his son Herodianus, who was also a grammarian, was active in Rome. Apollonius' syntax has had a curious fate. It was the unattainable model for Priscian's syntax, which in turn served as the model for medieval grammarians; and in the period 1500-1800 Western grammatical theory was little else than unac­ knowledged imitation and continuation of what had been achieved in the Middle Ages. Yet Apollonius' syntax has remained almost completely unknown to the community of linguists, as is shown by the fact that before 1980 practically no books or articles on it had been published. Apparently it has been easy to follow Steinthal's (1971 [1891]) example and to discuss only Apollonius' morphology.9 The situation has changed since Householder's (1981) translation of, and commen­ tary on, Peri suntaxeös. Nevertheless, I still consider this as the one work most in need of (positive) reappraisal within the entire history of Western linguistics. Peri suntaxeös contains four books which purportedly deal with definite articles, relative pronouns, pronouns, solecisms (i.e. syntactic errors), verbs, participles, prepositions, and adverbs. Yet the real point of interest is how words combine, ultimately to form a sentence. (The typical Western habit to approach syntax via word-classes can thus be traced back to Apollonius.) The manner of presentation is, on the whole, not very systematic. While a huge number of examples (some 1,500, according to Householder's counting) is analyzed, the analyses are frequently interrupted by theoretical reflections or critical remarks on rival grammarians. Therefore, giving a satisfactory overview is not easy. The best I can do is to single out a few (partly overlapping) themes that seem the most worthwhile to me. A) The Attitude towards the Data For modern linguistics, synchronic grammatical description pertains to the linguistic usage of a group of speakers. Apollonius views this problem in more complex terms insofar as, in addition to usage, he recognizes two other relevant factors, namely literary tradition and grammatical theory itself. It is the combination of all three which constitutes the data of grammatical description. Perhaps the clearest statement to this effect is the following:

202

UNIVERSAL HISTORY OF LINGUISTICS

I rely not merely on poetical citations..., but on common everyday usage, the practice of the best prose-writers, and, most of all, on the force of theory which must be applied even about constructions which are not in the slightest doubt (II, 49). The following remark may serve as an illustration: "A nominative [*emautos, 'I myself] corresponding to this genitive [emautou] is impossible to find, either in grammatical theory or in actual use" (II, 138). The reference to theory (which is made here very much in the spirit of the 'analogist' standpoint) may seem incorrect either in a weaker or in a stronger sense: weaker, if it means to say that only such data can be accepted as has a ready explanation; stronger, if it means to say that data must be changed so as to fit theory. While there is some vacillation between descriptivism and prescriptivism in Apollonius' programmatic statements, it is the former which prevails nearly always in practice. It may be added that Raster (1987:54) regards exactly the same three factors as having inspired the work of the Roman grammarians, namely usus, auctoritas, and ratio. The reference to theory did not go unchallenged. Sextus Empiricus (. 200 A.D.) pointed out that to speak Greak well means simply to follow the common usage. If theory (or 'analogy') agrees with usage, it is irrelevant from the speaker's point of view; if it conflicts with usage, it is incorrect (Adversus Grammaticos, 218-220). Sextus weakens his case, however, when he goes on to argue that any theory of language is impossible, because language contains an infinite number of words, and because there can be no science of the infinite (cf. also op. cit., 81-83). Varro knew how to answer this query. It is simply Apollonius' own linguistic intuition which provides him the access to 'common everyday usage', as can be seen from the fact that a large number of his examples (some 400, according to Householder's counting) are selfinvented. (We have observed Sībawaihi following the same practice.) Moreover, the freedom with which Apollonius manipulates his self-invented examples is quite remarkable. For instance, when he discusses the relative importance of the wordclasses, he first invents a simple sentence containing instances of each word-class (except the conjunction):

EUROPE

203

He then deletes the words one by one and observes how this affects the correctness of the sentence. He comes to the conclusion that the noun and the verb are obligatory because only they cannot be freely deleted (I, 14). This is the clearest instance of theory overruling common usage (and common sense). Tra­ ditionally, the noun was considered as the most important word-class (cf. p. 213); therefore nouns had to be obligatory, inspite of the fact that e.g. the above sentence is perfectly correct without anthrōpos (and even without ho autos). Yet the meth­ odological principle itself, i.e. deleting units to establish the distinction between obligatory and optional constituents, has remained valid until the present day. There are many other examples of the deletion test, with the proviso that sometimes deletion produces a change in meaning, not incorrectness (e.g. in I, 110). There are also instances of a permutation test: ho anthröpos ho agathos ('the good man') is correct but *ho agathos ho anthröpos is not (I, 135). Furthermore, there are many cases where the form of a word is changed in one way or another, and the resulting incorrectness is noted. It is clear that the notions of 'governance' and 'agreement' presuppose an implicit test of this kind. B) Grammatical Explanation When, in the beginning of Book III, Apollonius starts discussing the notion of 'solecism', he sets himself the goal of explanatoriness: So we'd better stop and explain what the actual cause of the ungram­ maticality is, not by mere citation of examples as some linguists do, pointing out the ungrammaticality without explaining the cause. But if you don't grasp the cause, it's an excercise in futility to cite examples (III, 6). This topic is pursued, with many digressions, up to III, 54; but even afterwards it is returned to on several occasions. The central explanatory notion is akolouthiā (alternatively katallēlotēs), which will be translated here as 'concord­ ance'. The incorrectness of constructions is explained by the fact that they violate concordance. By implication, correctness is explained as non-violation of concord­ ance. First of all, concordance is defined as a purely intralinguistic phenomenon. Violations of it have nothing to do with errors concerning how words are made to refer to things. Rather, they are errors which can be recognized as such just by hearing the sentence (III, 9).

204

UNIVERSAL HISTORY OF LINGUISTICS

Apollonius attains his notion of concordance by reinterpreting all nominal and verbal attributes enumerated in 5.5.1 as corresponding parameters of (in)cor­ rectness. He explains the functioning of concordance in a somewhat circuitous way: When words have been subclassified ... according to their proper applicational positions, they act by means of their grammatical correctness as restraints on words which ... [might] have come into a position where they don't belong (III, 22). ...no word-class can be ungrammatical in respect to a category which it fails to distinguish (III, 51). It is better to illustrate this principle with a couple of examples. Consider the relation of the personal pronoun egō ( T ) to its parameters of (in)correctness:

This figure shows the correct use of egō. Crossing any other box within a parameter, when using egō, would result in a violation of concordance. However, it would be impossible to use egö incorrectly on the parameter of gender (or more outlandishly, on that of tense). Thus incorrectness is kept within 'reasonable limits'. To give another example, consider the behavior of the active, medial, and passive forms of a transitive verb on the parameter constituted by the presence vs. absence of an accusative complement (i.e. direct object):

EUROPE

205

Active and passive are opposite cases: what makes one correct, makes the other incorrect, and vice versa. By contrast, the medial form just cannot be incorrect on this parameter (III, 30 and 159). Thus, it represents a case of neutral­ ization (just like egō does with respect to gender). Violation and non-violation of concordance as well as neutralization are illustrated abundantly. It is stated, for instance, that when a finite verb is 'trans­ formed' into an infinitive or a participle, its parameters will change accordingly. In general, concordance may be identified either as governance or as agreement (as we would say): it is a matter of something 'taking' or 'requiring' something else. However, concordance also includes cases of semantic compatilibity, as when sentences like *Yesterday I am writing are ruled out. As long as incorrectness is identified with solecism, correctness concentrates on one word ( = X) 'taking' another ( = Y); it is a matter of "the proper word [ = Y] driving out an improper one [ = Y']", as Householder puts it (p. 162). Thus solecism is a relational notion. (Incorrectness as an inherent property of a single word is known as 'barbarism'.) However, incorrectness is also dealt with at a level which is more global than the one represented by solecism, as when Apollonius announces general constraints on permissible sentence-structures. Such remarks are interspersed in his general discussion; they could be taken to define a further subtype of concordance. I give here a few typical examples: ...two verbs cannot form a single construction unless they are conjoined [as shown by the correctness and incorrectness, respectively, of the sentences] philosophos esti Diōn kai dialegetai ('Dion is a philosopher and is conversing') and *philosophos esti Diōn dialegetai (*'Dion is a philos­ opher is conversing') (I, 109).

UNIVERSAL HISTORY OF LINGUISTICS

206

The verb that co-occurs with the two [possessor and possessed] cannot simultaneously agree with both, since they are different persons... (II, 139). It's impossible for a single infinitive to have two indicatives governing it, rather than just one (III, 77). Up to now, I have expounded what seems to me Apollonius' systematic use of the term 'explanation'. However, he also offers 'explanations' of a more unsystematic or local character. I shall give just one example of this kind. In II, 69 Apollonius mentions what I shall call 'rule X': objects of prepositions are never enclitic pronouns. He then adds: The important thing is to determine whether this follows logically from some more basic principle. The fact of the rule is no problem, since it's clear, but the theory is not obvious to everyone. Next Apollonius observes what I shall call 'rule T: oblique forms of pronouns which precede the verb are non-enclitic or accented, as in emoi dos ('Give me...'), whereas those which follow the verb are enclitic, as in dos moi ('Give me...'). Finally, there is 'rule Z': when prepositional phrases occur with pronouns, they must take pre-verbal position (II, 72). Now X follows from Y and Z: pronouns qua objects of prepositions are non-enclitic, because they precede the verb, and all pronouns preceding the verb are non-enclitic. There are several prima facie counter-examples to rules Y and Z, but these are brushed aside as results of reordering (huperbaton) applied to the basic structures (Prep) Non-enclitic Pronoun + V and V + Enclitic Pronoun. A similar type of reordering may apply to con­ ditional sentences: the natural order can only be 'if S 1 , S 2 ', and yet sometimes sentences with the inverse order occur in speech. The term 'explanation' can be given many different meanings. However, if we wish to retain a meaning which is even moderately close to the everyday usage, we must admit that all cases discussed above qualify as explanations. Notice, in particular, that we have here intralinguistic explanation within a single language. C) Abstract Analysis As a first approximation, grammatical analysis has to state the formal properties of and cooccurrence restrictions upon 'real' units, i.e. such units as are likely to occur

EUROPE

207

in speech. Explanation, as defined above, deals for the most part with phenomena of this type. Yet it is also possible to analyze the data in a less obvious, or more abstract, way: either by relating various real structures to one another or by postulating a level of description 'behind' real units and structures. In this sub-sub­ section I shall consider instances of this type of analysis in Peri suntaxeōs. Practi­ cally the same topic is covered by Householder (1981). On at least three occasions Apollonius analyzes the speech acts of stating, wishing, and commanding as codified in the corresponding types of sentences. In thefirstpassage (= I, 50-55) he starts by discussing the infinitive and characterizes it as "the noun implicit in the verb" or "the noun shared by the paraphrases of the moods (enklisis)". The paraphrases look like this: peripatei Truphōn ('T. is walking') → hōrisato peripatein Truphōna ('He declared that T. was walking') peripatoië Truphön ('May T. walk!') → ēuksato peripatein Truphöna ('He wished for T. to walk') peripateitö Truphön ('Let T. walk!') → prosetakse peripatein Truphöna ('He ordered T. to walk') The paraphrases consist in transforming the finite verb into an infinitive and in "adding the verb which is implicit in the mood" (I, 51). In these examples the relation between the original sentence and the paraphrase is like that between citation and comment. The same topic is touched upon in III, 25, where it is stipulated that the 'transformation' (metalëpsis) may be taken to operate in both directions: grapheis ('You are writing') ↔ horizomai se graphein ('I declare you to be writing') peripatoiës ('May you walk!') ↔ eukhomai se peripatein ('I wish for you to walk') graphe ('Write!') ↔ graphein soi prostassō ('I order you to write')

208

UNIVERSAL HISTORY OF LINGUISTICS

In these examples the relation of paraphrase is more direct than above. Now the necessarily 'ego-centered' nature of every speech act is accounted for by making the corresponding performative verb explicit and putting it in the first person singular. The same set of examples recur in III, 61. But this time the analysis is placed into a larger context, as follows: Every derivative in general can be analyzed into its base (prōtotupon) plus a word synonymous with its derivational suffix (paragōgē) ... thus Hektorides ('son of Hector') is analyzed into Hektoros huios (idem). The form gorgoteros ('more terrible') is made up of gorgos ('terrible') with the addition of māllon('more') so that it is analyzed into gorgos māllon... It would be a waste of time to pile up examples; the point is clear from these. In cases like these, where X is analyzed into Y + Z, it is said that 'Z is implicit in X' Now, 'being implicit' must be distinguished from 'being deleted by ellipsis'. For instance, māllon is implicit in gorgoteros, but gorgoteros is not elliptical, because adding māllon would produce a pleonasm, like 'more better' (II, 50). On the whole, Apollonius recognizes the same verbal attributes as Dionysius Thrax. As an exponent of abstract analysis, however, he postulates a ('deep') level of description where the attributes are explicitly represented: "in finite verbs there are very many meanings implicit" (II, 51; cf. also III, 96). For instance, an indicative contains the meaning of assertion, which is made explicit by the corre­ sponding paraphrase. When one reads what Apollonius says in different contexts, it becomes evident that he is able to distinguish between assertion (or statement), affirmation (qua subtype of assertion), and indicative mood. On some occasions, however, one or the other of these distinctions gets blurred, which makes it doubtful that Apollonius possesses a wholly consistent theory of the verb. 10 Relative clauses give Apollonius the prime occasion to relate different sentence-structures to one another. This is, in brief, his notion of relative clause: The relative pronoun [originally 'postpositive article'] requires a verb, and provides linkage through its reference to the antecedent noun, and thus does not make a simple sentence because of the construction with two verbs, one with the noun and one with the pronoun itself (I, 144).

EUROPE

209

This double nature of the relative pronoun is emphasized also on a later occasion: Though it refers back to its antecedent, it also stands in the beginning of its clause, construed with a following verb... (II, 85). As a consequence, the following examples (given in I, 144-145) turn out to be 'synonymous' with or 'equivalent' to one another: a) paregeneto ho grammatikos hos ('who') dieleksato ('The grammarian arrived who conversed') b)

ho grammatikos paregeneto kai ('and') houtos ('he') dieleksato ('The grammarian arrived and he conversed')

c)

ho grammatikos paregeneto kai dieleksato ('The grammarian arrived and conversed') The fourth synonymous structure results, by 'transformation' (katalëpsis), from a coordinated sentence like c) above (III, 29): d) egraphon kai ëniömen (T was writing and having trouble') → e) graphön ëniömen ('Writing, I was having trouble') The deletion exemplified by transforming b) into c) is possible only if the deleted element, which may be a verb or a noun, is shared by both sentences (II, 60-63):11 Truphön peripatei kai Apollönios ('T. and A. are walking') Truphōn peripatei kai dialegetai ('T. is walking and conversing') There is an additional rule to the effect that if in a 'both—and' construction the element immediately following both is either a noun or a verb, the deleted element must be, respectively, either a verb or a noun: kai Truphön peripatei kai Apollönios *kai Truphön peripatei kai dialegetai kai peripatei Truphön kai dialegetai *kai peripatei Truphön kai Apollönios

210

UNIVERSAL HISTORY OF LINGUISTICS

There is also a transformational relationship going from a finite verb to a noun via a participle (III, 190): gumnazei touton ('He trains this man') → gumnazōntouton ('training this man') → gumnastes toutou ('trainer of this man') Sentences are transformed into accusative-cum-infinitive constructions. With transitive verbs there arises a potential ambiguity because both the agent and the patient are now in the accusative. The ambiguity is resolved, however, on the grounds that the agent must precede, given that the action 'proceeds' from the agent to the patient (III, 86): Theōnhūbrise Diōna ('T. hurt D.') → legousi Theöna hübrisai Diöna ('They say that T. hurt D.') In all the examples discussed so far, the relations of paraphrase or trans­ formation hold between 'actual sentences' (for this term, cf. p. 154). However, there are also cases where one term of the relation is non-actual and where, accordingly, the transformation (rather than 'paraphrase') is necessary, and necess­ arily unidirectional. The starting points of such transformations are provided here with a star, to indicate that they are not actual sentences: *eme hūbrisa ('I hurt me') → emauton hūbrisa ('I hurt myself) (III, 5) *ton emou doulon epaisa ('I struck the slave of me') → ton emautou doulon epaisa ('I struck the slave of myself) ↔ ton emon doulon epaisa ('I struck my slave') (II, 106). A change from abstract to concrete is endorsed also in the following statement: "The basic form underlying the subjunctive is the indicative, and the conjunctions which govern it change it into its characteristic (= subjunctive) phonological form" (III, 134). As the final example of abstract analysis, I mention Apollonius' attempt to find common meanings behind accusatives, genitives, and datives governed by

EUROPE

211

verbs (III, 158-190). As he sees it, the accusative expresses the patient of (physical or mental) action, the dative expresses the beneficiary of an action, and the genitive expresses possession or domination. But since he must admit the existence of several exceptions, this analysis can hardly be considered entirely successful. D) Deciding the Category-Membership Apollonius reports several cases where he disagrees with other grammarians about the category to which a given unit should be assigned. Trypho, who seems to have been his principal rival, argues, for instance, that the form su ('you') is a vocative because it occurs in sentences like su graphe! ('You write!'). The problem as such is trivial (since Trypho is obviously wrong), but it is instructive to see how Apollonius solves it by applying the 'clear case' principle, i.e. by comparing su to analogous forms whose category-membership is not in doubt (III, 35-40). I re­ produce here only the most important arguments. a) In both of these constructions all three words are in the same case: soi (dat.) onti (dat.) agathōi (dat.) ('to you who are good') and se (acc.) onta (acc.) agathon (acc.) ('you who are good'). Therefore, in the construction su (?) ōn (nom.) agathos (nom.) the form su too must be in the nominative (and not in the vocative). b) From the constructions egō (nom.) eimi ('I am') and ekeinos (nom.) estin ('He is') it can be inferred that in the construction su (?) ei ('You are') the form su too must be in the nominative. c) Since the construction *Aristarkhe (voc.) ei ('Aristarchus! you are') is incorrect, also the construction *su (voc.) ei ('You! are') is incorrect, i.e. su must be a nominative form. d) It is a general truth that coordinated case-forms can be summarized as follows: emoi (dat.) kai soi (dat.) kai Dionūsiōi (dat.) elalësen Truphōn ('T. spoke to me and to you and to Dionysius') = hēmīn (dat.) elalësen Truphön ('T. spoke to us'). Therefore su must be a nominative form, since it is both coordinated with nominatives and included in a nominative: egō (nom.) kai su (?) kai Truphön (nom.) paragīnometha ('I and you and T. arrive') = hēmeis (nom.) paragTnometha ('We arrive'). e) What is true of coordination, is also true of disjunction. Compare the two sentences 'ē emou (gen.) akouei Truphönēsou (gen.) ē Theönos (gen.) ('T. listens either to me or to you or to Theon') and 'ē egō (nom.) aperkhomai ë su (?) ë Dionûsios (nom.) ('Either I leave, or you or D.'). The upshot is the same as before: su must be a nominative form.

212

UNIVERSAL HISTORY OF LINGUISTICS

On several occasions it has to be decided to which word-class certain ambiguous or non-standard forms belong. For instance, emou may, depending on the context, be either a personal or a possessive pronoun; or it may not be immedi­ ately obvious that impersonal verbs like dei ('it is necessary') are genuine verbs. Here I mention only one further example. Since prepositions may combine with nouns or verbs to form compound words, a distinction has to be made between a single (compound) word and a many-word construction (in practice, a prepositional phrase). The following criteria are suggested: i) A word has one accent, but a phrase has two accents (IV, 1). ii) In a phrase, unlike in a word, the preposition and the noun may be separated by an article (IV, 13). iii) If a preposition precedes a verb or a nominative, we have a word; otherwise we have a phrase (IV, 17-18). iv) A compound word has only one ending, whereas a (noun) phrase may contain several endings (IV, 33). Among the word-classes the preposition is unique in having the abovementioned ambiguity: All other parts of speech have a simple construction pattern that they take, e.g. adverbs are construed with verbs, even though there may be other words in between, articles go with nominais, [subject] nouns with their verbs, and the verbs, in turn, are in relation to [complement] nouns and pronouns, which may substitute anywhere for nouns (IV, 9). As Householder points out (p. 228), a basic sentence structure may be inferred from this characterization, as follows:

E) The Problem of Isomorphism The 'language—mind—reality' isomorphism figures in Apollonius' work less promi­ nently than in that of his more philosophically minded predecessors. Yet it is obvious that he takes the existence of some such isomorphism for granted. While trying to justify the traditional hierarchy among the word-classes, he makes the following succinct statement concerning the basic correspondence between ontological and linguistic distinctions:

EUROPE

213

The noun necessarily precedes the verb, since influencing and being influenced are properties of physical things, and things are what nouns apply to, and to things belong the special features of verbs, namely doing and experiencing (I, 16). Thus things are ontologically primary entities, and they are expressed by nouns. Whatever happens between things, i.e. acting and being acted upon, is ontologically secondary, and is expressed by verbs. ("The rest of the parts of speech are used in construction with verbs and nouns"; I, 36.) Things exemplify either primary qualities or secondary qualities like 'white' and 'good'; and primary qualities are in turn divided into general qualities like 'man' and particular qualities like 'Plato' (II, 22-23). Pronouns qua substitutes for nouns express things, but things without qualities, i.e. "pure being" (I, 138). Yet it is also claimed that personal pronouns are used like proper names, i.e. to express particular qualities (II, 41). Outside a real speech situation the personal pronouns qua deictic ( = 'pointing') words have no definite meaning (II, 42). Ontologically speaking, activity is something that 'passes over' from the agent to the patient (III, 148), which means that "the active aspect of the event precedes the passive one". Therefore it is natural that in speech the word for the agent should precede the word for the patient (III, 86-87). This, as far as I can see, sums up what Apollonius has to say about the way that the structure of reality is reflected in linguistic structure (with the qualification that the reference to the speech situation, in connection with deictic words, already transcends the limits of 'pure' ontology). Reality is of course always mediated by mind, which means that it is reality-as-conceptualized. Yet it is not pointless to consider mind as a level of its own, when speaking of isomorphism: there can be, and are, mental structures without any obvious ontological correlates (although, to repeat, the opposite is not true). As far as I can see, Apollonius gives only one clear example of how 'nonontological' mental structures are reflected in linguistic structures: [One cannot] claim that conditional conjunctions are not sentence-initial on the ground that we say phös estin, ei hémera estin ('It is light, if it is day'). For it is only in the surface order that ei hémera estin stands second, not in the explicit semantic structure. For the mind must first accept the thought 'if it is day', and then it can accept 'it is light') (II, 77). In other words, conditional relationships, in particular those of contrafactual character, are something that mind imposes on reality, rather than vice versa. Yet

214

UNIVERSAL HISTORY OF LINGUISTICS

it is meaningful to say that such relationships are reproduced by, or reflected in, language. Finally, I come to discuss the 'form—meaning' distinction in Apollonius. But first I must make a terminological remark. The word 'isomorphism' may be used either in an explanatory or in a non-explanatory sense. When reality(-asconceptualized) or mind is said to be isomorphic with language, what is meant is that the former explains the latter (or, on the Whorfian interpretation, the other way around). If one accepts the psychologistic meaning-conception, one can perfectly well speak of form—meaning isomorphism in the strong, explanatory sense. If, however, one rejects the psychologistic meaning-conception, as I do, the situation is more complicated. On the non-psychologistic interpretation, form and meaning constitute linguistic units just like form—function pairs constitute instruments (cf. p. 182). As a consequence, there is a conceptual connection between form and meaning, i.e. a connection too tight to be explanatory. It is still possible to speak here of 'isomorphism' in a weak sense in which it amounts just to a one-to-one relationship between forms and meanings. Yet neither term of this relationship explains the other. Rather, it is the relationship itself which is in need of expla­ nation. Apollonius' view of the form—meaning distinction seems thoroughly un­ objectionable. In his book On Adverbs (119, 1) he states the general principle that every word consists of two elements, the meaning (ennoia, also sēmainomenon or dēloumenon) and the sound structure (skhēma tēs phōnēs). In Apollonius' time there were conflicting views about the relative importance of form and meaning. In Peri suntaxeōs it is stated unequivocally that it is meaning, rather than form, which keeps inflectional paradigms together. For instance, the common meaning element in egō ('I'), nōi ('we two'), and hēmeis ('we') makes them members of the same paradigm, in spite of their formal differences (I, 79). The same argument recurs in II, 33, concerning words which have, for whatever reason, changed their wordclass: "The phonological forms of words do not weigh as much in classification as what is meant by them." In sum, Apollonius seems to treat formal and semantic aspects in a very balanced way. He lays down the principle that "agreement in the form and in the meaning is a necessary feature of every compound", as in mīsogunēs ('womanhater'), where "both mīsein ('to hate') and gunē ('woman') are present both in form and in meaning". By contrast, although skūteus ('shoemaker') contains the two

EUROPE

215

semantic elements of 'leather' and 'to cut', it is not a compound because its form contains only the element corresponding to 'leather', i.e. skūtos, but lacks the element corresponding to 'cut', i.e. temnein. The criteria of compoundhood are met by skūto-tomos ('leather-cutter', 'shoemaker') (II, 161). Similarly, when Apollonius discusses the definite article, he defines it first semantically, and then formally, i.e. distributionally. It is first stated that the definite article may have one of the following meanings: i) 'par excellence', ii) 'unique possession', iii) 'already mentioned', iv) 'whoever' (I, 43-44). Then types of nouns are distinguished, according as they can cooccur with the definite article (I, 46-56) or must cooccur with it (I, 57-68) or cannot cooccur with it (I, 70-93). After these clarifications it may still be asked whether Apollonius' syntactic theory has some sort of semantic foundation, or whether it is simply based on formal variation. This is the question raised by Blank (1982:28-39). As he sees it, Apollonius is a consistent follower of the Stoic grammar and thus conceives of his syntax as a syntax of lekta. This interpretation is supported by the following passage, which occurs quite at the beginning of Peri suntaxeōs: For the meaning (noēton) which subsists in each word is, in a sense, the minimal unit of the sentence, and just as the minimal units of sound compose syllables when properly linked, so, in turn, the structural com­ bining (suntaxis) of meanings will produce [semantic] 'syllables' by combining words. Just as the word is made of syllables, so the complete sentence is made by grammatical collocation of meanings (I, 2). Blank concludes that Apollonius' 'concordance' is meant to obtain between semantic, and not formal, units. First of all, I would like to point out that in many cases, particularly in connection with agreement, it is impossible to decide whether (in)correctness is about form or (grammatical) meaning. A sentence like *egö ei agathos (T are good') is grammatically incorrect, but its incorrectness is a matter both of form and of (grammatical) meaning. There are cases, however, where the consonance of form and meaning falls apart, and they are crucial from the present point of view. For instance, it may happen that the case ending required by a verb has no semantic motivation. Apollonius' discussion of the verbs tremein and pheugein — as in the sentences tremó se (acc.) ('I am frightened of you') and pheugō se ('I flee away from you') — is a case in point. He notes that "they govern the accusative although there is no action [to pass over from agent to patient]". Therefore he suggests that "the constructions under consideration now are in their full underlying form like this: tremō dia se ('I am frightened because of you') and pheugō dia se ('I run

216

UNIVERSAL HISTORY OF LINGUISTICS

away because of you')" (III, 166). The important thing is that Apollonius dis­ tinguishes here governance as a formal notion from the meanings expressed by the words between which governance obtains. Therefore I do not think it advisable to claim that Apollonius is concerned exclusively with 'syntax of meanings'. Rather, in his 'syntactic' theory he pays an equal amount of attention both to the formal and to the semantic properties of sentences. This is just as it should be. I do agree, however, with Blank's (1982) larger claim, which is to deny that there existed in Apollonius' time a sharp conflict between a 'philosophical' and a 'philological' approach to language study. At least there is no trace of such a conflict in Apollonius' work. Rather, it represents a nice synthesis between 'dataorientation' and 'theory-orientation', as these words are currently used. Indeed, nobody who has read and understood the preceding account of Peri suntaxeös can deny its strong similarities with modern linguistics. It is only the more regrettable that they have remained unknown. Within the linguistic community this state of affairs has produced a serious distortion, or lack, of historical perspective.

5.2.4 The Aftermath: Priscian The study of grammar continued until the end of the Roman Empire. Yet none of the later Latin grammarians shows the same originality as Varro. Their work has been characterized by Kaster (1987:150) as follows: Belonging for the most part to the third through sixth centuries, the handbooks collected in [Keil's Grammatici Latini] appear intent, one after the other, on imposing a deadly uniformity on their material, as the grammarians repeat the same patterns of analysis, teach the same lessons — often in precisely the same words, repeated from older sources — and use for illustration the same examples that had been repeated for gener­ ations. Kaster goes on to expose the social factors which conspired to make the fossilization of the grammatical tradition possible. It still remains unexplained why it became fossilized in fact, and — beyond that — why it declined.12 Thus I do not find the situation comparable to what happened (or failed to happen) within the tradition of Arab linguistics. Even if there was no qualitative progress, Sībawaihi remained fully understood, and on some minor points he was improved upon or at

EUROPE

217

least complemented. In the Western tradition, by contrast, Apollonius was trivialized, and many of his arguments were ignored, which may be taken as a sign that the rationale for his grammar was no longer understood. The most important works produced during this period were Ars Gram­ matica by Donatus (c. 400) and Institutiones Grammaticae by Priscian, who lived in Constantinople around the year 500. The former book was intended for peda­ gogical use, whereas the latter represents the highest 'theoretical' level that gram­ matical analysis could attain in late antiquity. Institutiones Grammaticae contains 18 books. Book I deals with sounds and letters. Book II deals with syllables, gives general definitions of the word-classes, and broaches the analysis of the noun, which is continued until Book VIII; the end of Book V as well as Books VI and VII are devoted to a discussion of the cases. Books VIII, IX, and X deal with the verb, while Book XI deals with the (present) participle, which is still regarded as a word-class of its own. Books XII and XIII deal with the pronoun. Books XIV, XV, and XVI deal, respectively, with prep­ ositions, adverbs (plus interjections), and conjunctions. Books XVII and XVIII deal with syntax (constructio), understood as a "systematic combination of [members of] word-classes with one another" (ordinatiopartium orationis inter se). The first 16 books constitute the so-called Priscianus Maior while the rest constitutes the Priscianus Minor. The structure of the two books on syntax is rather incoherent. After a few general remarks on the nature of syntax and on the word-classes (1-21), Book XVII treats at length of interrogatives (22-51) and pronouns (52-208); there is consider­ able overlap with the contents of Books XII and XIII. Although Book XVIII intends to deal with the syntax (rather than morphology) of the noun (1-39) and of the verb (40-156), much material from earlier books is repeated. The remainder of the Book XVIII (157-307) contains pairwise citations from Latin and Greek authors without any discernible overall strategy. There may be a natural explanation for the disorganized character of Priscianus Minor. In the Preface of Institutiones Priscian admits that he let his book be published too hastily. He had no choice however, since he had noticed that rival grammarians were promulgating ideas stolen from him and were thus "taking all the glory". I do not think there is much point in going into Institutiones in any detail. Priscian never goes beyond Apollonius, and very often lags behind. It is to his credit that he does nothing to hide this fact. On the contrary, in the Preface he frankly admits that he is just trying to translate into Latin the teachings of Apollonius and of his son Herodianus. He repeats his intellectual debt on several

218

UNIVERSAL HISTORY OF LINGUISTICS

occasions, as when he characterizes Apollonius as maximus auctor artis grammaticae (XI, 1) and declares (XIII, 1) that he has tried to follow Apollonius' example in everything (Apollonius cuius auctoritatem in omnibus sequendam putavi). Priscian's dependence on Apollonius can be made more concrete by noting that e.g. Book XVII is just a collection of passages from Peri suntaxeōs translated into Latin. For instance, Apollonius' example ho autos anthröpos olisthēsas sèmeron katepesen (cf. here p. 203) is translated verbatim as Idem homo lapsus heu hodie concidit (XVII, 12), with the proviso that the definite article (= ho), which has no equivalent in Latin, is 'replaced' by the interjection (= heu). And sometimes Priscian gives such a word-for-word translation of the original text that he no longer grasps its meaning. Thus it comes about that when Apollonius says (I, 18) that nouns are primary, Priscian ends up saying, in this particular context (XVII, 14-15), that verbs are primary. The reason for this muddle can be shown as follows:

In Greek, onoma, being the word(-class) for things, is taken to be ontologically primary, and therefore it means not just 'noun', but also 'word' in general. In Latin, nomen is the word(-class) for things, and nevertheless verbum, which is the word(-class) for acting and being acted upon, means not just 'verb', but also 'word' in general. It is this translational non-equivalence which Priscian fails to notice. I have characterized here Priscian as the 'aftermath' of the Alexandrian School. But this characterization could be extended much farther. Given the relation between Apollonius and Priscian, on the one hand, and between Priscian and his medieval successors, on the other, "what the grammarians of medieval western Europe inherited in the way of syntactic concepts was thoroughly Apollonian (and hence thoroughly Alexandrian)" (Percival 1987:66). Thus there was a clear line of influence from the Stoics to Apollonius, and then unbroken continuity from Apollonius to the Middle Ages (cf. 5.3), and beyond (cf. 5.4).

EUROPE

219

5.3 Linguistics in the Middle Ages After Apollonius, linguistic research declined; after Priscian, it stopped. Because of turbulent external circumstances, all that could be done during 500-800 was to maintain tolerable knowledge of Latin by the use of school-grammars. In some areas, nevertheless, knowledge of Latin deteriorated to the point that the under­ standing and copying of manuscripts became impossible. To this extent, at least, the designation 'Dark Ages' is fitting. It is interesting to reflect on what kind of conception we would have about antiquity today, if the manuscript tradition had broken down completely. That this did not happen, was to a large extent due to work carried out in monasteries in Ireland, which country, being situated on the periphery of Europe, was spared from many of those tribulations that befell the central Europe (cf. Bolgar 1954:92-95). It is true that part of the 'classical heri­ tage' was saved by the Arabs, and was to become available in the sequel. During 500-800 Donatus was the principal authority on grammar. After 800, i.e. as a result of the Carolingian Renaissance, a more theoretical point of view started gaining ground, which meant that Donatus was superseded by Priscian (cf. Law 1987:197-200). He maintained his position largely intact until the end of the Middle Ages. The medieval grammarians consciously combined the grammatical tradition of antiquity with Aristotle's philosophy. In the latter respect they could rely on such competent commentaries as those by Ammonius and Boethius, who were Priscian's contemporaries. Thus, around the year 500 philosophy of language was practised at a higher level than linguistics. By 'medieval linguistics' I shall mean here in the first place the Modistic doctrine, which flourished during 1250-1350. (This is why e.g. Peter Abaelard will be considered merely as one of the 'precursors'.) I shall also discuss the closely related 'logical linguistics' of the time. Although these achievements were certainly the most representative, they do not exhaust the contributions that were made to linguistic theory in the Middle Ages. One must also mention the phonological analysis of Icelandic by the anonymous 12th-century 'First Grammarian' (cf. Haugen 1972) and the beginnings of historical linguistics outlined by Dante Alighieri (cf. Mazzocco 1987).

220

UNIVERSAL HISTORY OF LINGUISTICS

5.3.1 Precursors: Ammonius, Boethius, Abaelard We have seen that philology arose as commentary upon literary texts. Soon also non-literary, e.g. grammatical and philosophical, texts were commented upon, and it was taken for granted that this was the proper way of doing grammatical or philosophical research. It is difficult not to see in this attitude an avowal that one is living in a 'silvery age' doomed to nourish itself from the results of the preceding 'golden age'. One of the earliest Aristotle-commentators was Porphyry (c. 234-305). His commentary on De Interpretation was available both to Ammonius (c. 450-520), who lived in Alexandria, and to Boethius (480-524), who lived in Rome, when they wrote their own commentaries. These in turn were known to all medieval philos­ opher-grammarians, as is shown by frequent references. This is why I shall discuss Ammonius and Boethius together. My discussion is based on the passages translated in Arens (1984), pp. 58-123 for Ammonius and pp. 159-204 for Boethius. It should be noted that these rather extensive passages are devoted to the two or three first pages of De Interpretations Following Porphyry, Boethius gives an overview of the meaning concep­ tions that were current in antiquity. What do words signify?13 At least the follow­ ing answers were given: a) things; b) 'incorporeal natures', subdivided into b1) Platonic forms and b2 Stoic lekta; c) psychological entities, subdivided into c1) perceptions, c2) imaginations, and c3) notions. According to Boethius, Aristotle chose the last alternative: words signify notions or concepts (i.e. 'impressions of the soul'), which arise — via imaginations — from perceptions (i.e. 'impressions of the body'). For Boethius, there is a 'natural order' such that things precede notions (which are 'formed by nature' and are hence common to Greeks, Romans, and barbarians), and notions precede, and are the 'causes' of, words (which are 'formed by institution' and are hence dissimilar in different nations). Yet there may be notions without corresponding things (e.g. 'centaur'). Within this framework words signify 'primarily' notions and 'secondarily' things. Notions too are word-like in that they 'signify' things: "Oral speech designates the mental speech which goes on in silent thinking, and this mental speech primarily conceives and designates the things" (Arens 1984:166). Ammonius gives an identical characterization, and adds that "we need not think of any other medium between concept [or notion] and thing" (p. 73). This is how the rejection of lekta is justified.

EUROPE

221

The social and institutional nature of language is grasped clearly enough. For Ammonius, it is the purpose of concepts "to enable people to communicate with one another and live as fellow-citizens" (p. 74). Boethius further clarifies this point, by comparing language to other institutions: Just as for a coin it is not enough to be copper with somefigureimprinted on it to be called a coin, but it must be the value of something, so the nouns and verbs are not only vocal forms, but are instituted for the signification of notions (p. 171). At the same time, this is a good way of bringing out the inseparability of form and function, which is characteristic of instruments. Ammonius restricts the domain of institutionality by setting up an impli­ cation 'if institutional, then human': A convention concerning the vocal forms is the work of human beings alone: neither in the world above them nor in that below them is there any conventions, for those above do not need it, and those below are not capable of any mutual agreement (p. 86). This argument was to become famous in subsequent centuries. It was taken to mean that angels conveyed their thoughts by means of telepathy and could therefore dispense with the use of sounds; animals, by contrast, had no thoughts to convey even though they made a lot of sounds. Ammonius returns to this topic to emphasize the difference between humans and animals: "Man ... has the use of ... sentences composed of [nouns and verbs] for communication as a priviledge over against brute animals; therefore he alone of all mortal beings has a self-moving soul capable of artistic work..." (p. 119-120). In a later reformulation this became the 'free will vs. machine' distinction. Ammonius discerns two 'powers' within the soul: cognitive and vital (or 'appetitive'). The former contains thinking, belief, imagination, and perception, whereas the latter contains will, choosing, inclination, and desire. The sentencetypes (eidē logou) are subclassified accordingly, the majority issuing from the vital or appetitive power of the soul (pp. 59-61). In this taxonomy there is a certain vacillaction between sentences as formally distinct units and speech acts (performed by uttering sentences) as conceptually distinct units (cf. Nuchelmans 1973:97-98):

222

UNIVERSAL HISTORY OF LINGUISTICS

Traditionally, Western philosophy has been criticized for neglecting the noncognitive functions of language. At least Ammonius must be exempted from this criticism. Even in the sequel there was some debate over the status of, e.g., the promise: is it a mood of its own? 'Speech-act theoreticians' answered yes, but 'grammarians' insisted that it is nothing but an indicative in the future tense (cf. Nuchelmans 1973:130). As far as the notion of a sentence, and in particular of the finite verb in it, is concerned, Ammonius and Boethius repeat Aristotle's position, as it has been presented here on p. 178: Every finite verb is used for predication and contains an implicit copula by means of which the human will puts the two semantically full words together and thus forms a sentence; as such, the copula has no (lexical) meaning. When explaining Aristotle's dictum that 'the noun is without time', Ammonius carefully distinguishes between lexical meaning and grammatical meaning (translated as significatio and consignificatio in Latin): That 'without time' does not mean that the noun never signifies time, but that it does not consignify it, has been clearly stated by Porphyry, because nouns [i.e. adjectives] like sēmerinos ('of today') signify time, but indicate it exclusively, not additionally like the verbs (p. 87). Boethius repeats the same distinction: "Nouns signify time, whereas the verb primarily signifies action and undergoing and carries with them an indication of time" (p. 189). Ammonius later defines 'consignifying' generally as 'secondarily indicating something in addition to the primary signification' (p. 105).

EUROPE

223

After Boethius, works of Aristotle were not translated into Latin until the second quarter of the 12th century. Already somewhat earlier there had been attempts to revive the commentary tradition, but the next significant commentary on De Interpretatione was written by Peter Abaelard (1079-1142). Abaelard is an interesting figure. Although I must confine my discussion here to his contributions to linguistic theory, it is worth pointing out that, according to Kneale & Kneale (1962:204), he was so far the best logician of the Christian era, and one who believed that it was actually possible to go beyond Aristotelian logic, instead of just repeating it. Nuchelmans (1973:276) points out that Abaelard's teachings had been forgotten by the end of the 14th century, which explains why erroneous opinions he had already disposed of could keep cropping up (cf. 5.3.6). Abaelard distinguishes the viewpoints of the speaker and the hearer as follows (Arens 1984:253): Speaker thing → notion → word

Hearer word → notion → thing

Accordingly, a word manifests the speaker's notion and produces the hearer's (similar) notion; it differs from an animal sound in that the latter is not uttered with an intention (p. 232). For a sign, whether linguistic or non-linguistic, to be significative, it is required in addition that there should exist a corresponding institution (p. 267). Hence, significance is guaranteed by the combination of intention and institution. This is a remarkable answer to the fundamental question: How is communication possible at all? It is questionable whether a better answer has been given since then (cf. CLT: 140-142, 166-169). Abaelard has a straightforward way of proving the universal character of notions: "If a Roman and a Greek see a horse at the same time, it never happens that one should esteem it to be a man, and the other a horse" (p. 248). It was customarily thought that perceptions (and ultimately, notions or concepts) are based on a likeness with the things perceived. Abaelard argues, however, that in the presence of things themselves the assumption of mental thing-like 'images' is superfluous: ...as long as the thing that is touched by a sense is present, we do not need an image, we perceive the reality of the thing in sensation and thought without any interceding image, the likeness having been conceived instead of the thing. When we have the thing itself, we do not need a likeness (p. 240).

224

UNIVERSAL HISTORY OF LINGUISTICS

The difference between Abaelard and the view he is contesting might be clarified by this figure (where A = person,  = mental image, and  = thing):

It is very much on the same grounds that the existence of 'sense-data' has been rejected in modern epistemology. Abaelard distinguishes between a simple notion (intellectus simplex) like 'man' and a composite notion (i. compositus) like 'rational mortal being', but he emphasizes that also the latter is one. Such composite notions as 'the running man' or 'the white man' do not differ qua notions from the mental equivalents of the sentences 'The man runs' or 'The man is white': in both cases it is a matter of running or whiteness 'inhering' in the man. The difference lies, rather, in the fact that only in the (mental) sentences the composite notion is also asserted (cf. Nuchelmans 1973:143-144, Arens 1984:254-255).

A sentence may also contain a sentence-like composite notion: "When we say Verum est Socratem sedere ('It is true that Socrates is sitting'), the subject sentence Socratem sedere ('that S. is sitting') is the same notion as 'Socrates sedet' ('S. is sitting'), yet it has not the declaratory form of affirmation or negation so that it could be called a proposition" (p. 256). Already Ammonius had noted that there is an ascending trichotomy 'wordsentence—syllogism' corresponding to Aristotle's three books Categoriae, De Interpretationen and Analytica Priora. Abaelard makes this idea more explicit by postulating the following mental operations: i) perception of things and actions/ qualities; ii) combination of perceptions to form a simple mental sentence; iii) com­ bination of simple mental sentences to form a complex (here: conditional) mental sentence (p. 259). This may be clarified as follows:

EUROPE

225

Notice that Abaelard's third mental operation is not yet the one that under­ lies syllogistic inferences. (A fourth operation is postulated to perform this func­ tion.) Psychologistic theories are often one-sided insofar as they identify wordmeanings with potentially conscious notions (or 'mental images'), but make things too easy for themselves by ignoring the word-classes other than noun, verb, and adjective. Abaelard is here an exception because he squarely faces the question about the notional status of prepositions and conjunctions. He does not think that there is any independent notion (intellectus) corresponding to a preposition like in. Rather, when it occurs together with a noun, as in the expression in domo ('in the house'), it signifies "the characteristic (proprietas) of the house that consists in its containing something". A conjunction like si ('if), in turn, is not underlain by any notion at all, but rather by a "mental act that enters the composition of a [complex] notion". Moreover, the conjunction ergo signifies the above-mentioned 'fourth operation', namely that the conclusion is understood to follow from the premises (cf. Nuchelmans 1973:141-142, Arens 1984:272). The traditional reified type of psychology is giving way here to a more operational conception. In his analysis of the simple sentence, Abaelard points out along the traditional lines that the finite verb has a "double function, that of a predicate and of a copula" (Arens 1984:295). Making the implicit copula explicit turns all sentences into the canonical form 'A is B' (cf. Kneale & Kneale 1962:206-207, and here p. 179). As far as the lexical meaning of verbs is concerned, Abaelard corrects the Aristotelian tradition somewhat: It is just not true that all verbs mean either acting or undergoing, witness such verbs as 'to sit', 'to live', 'to dawn', etc. (Arens 1984:283). Abaelard also presents the same 'performative analysis' of the different sentence-types as Apollonius, adding the interrogative sentence and noting explicitly that this is the way to reduce all sentence-types to statements (Nuchelmans 1973: 148-149, Arens 1984:256). Abaelard is very clear that even in the case of non-empirical sentences (as we would say) truth and falsity are decided on extra-mental grounds. For instance, a sentence like Si est homo, est animal ('If one is human, one is animate') is true, and the relation of consequence expressed by it is valid, independently of whether

226

UNIVERSAL HISTORY OF LINGUISTICS

the notion of 'human' is actually followed by the notion of 'animate' in the mind of those who either utter or hear the sentence (cf. Kneale & Kneale 1962:205-206 and Nuchelmans 1973:149-150). This is still a valid refutation of psychologism, or a demonstration that when we speak of 'conceptual truths' or 'truths based on meaning', the words 'concept' and 'meaning' must not be taken in the psychologi­ cal sense (cf. GTM: 133-134). A contemporary treatise on logic called Ars Meliduna gives a complementary argument against psychologism: The truth-bearers cannot be psychological occurrences because this would mean that truths cease to exist when people fall asleep. Therefore truth-bearers must be similar to lekta, i.e. they can be neither psychological nor physical, but must have a mode of existence of their own (suum esse per se) (cf. Pinborg 1972:57 and here p. 187). Even though single words may refer to things, a true sentence does not refer to a thing, but rather to 'some kind of relation between things' {quasi quidam rerum modus habendi se) (Kneale & Kneale 1962:206). We shall return to this topic in the beginning of 5.3.3.

5.3.2 Philosophy of Linguistics The scholastic grammarians were called the 'Modistae' because their central theoretical concept was modus. My discussion of the Modistic doctrine is based on the following texts: Boethius the Dane (= Boethius de Dacia) (1980), Siger de Cortrai (= Sigerus de Cortraco) (1977), Pseudo-Albertus Magnus (1977), and Thomas of Erfurt (1972). It is probable that both Boethius and the man behind the designation 'Pseudo-Albertus' died about 1280. Thomas was active around 1300, and Siger died in 1341. Boethius and Pseudo-Albertus were original thinkers, whereas Thomas and Siger just summed up the Modistic doctrine. The Modistae had the ambitious goal of constructing a universal grammar (grammatica universalis). Preparatory to doing so, however, they presented a wide spectrum of general views on the nature of linguistics. It seems advisable to discuss this aspect of their work first. Boethius the Dane starts by defining the relation of grammar to its neigh­ bouring disciplines. It is the task of the philosopher (philosophus, metaphysicus) to investigate the general properties of things, or the 'modes of being' (modi essendi), as well as the 'first principles' of each particular science. The natural scientist (naturalis) investigates things (including sounds, for instance), rather than their general properties. The grammarian (grammaticus) investigates word-classes (partes

EUROPE

227

orationis), or the 'modes of signifying' (modi significandi)(pp. 1-28). Since he has no interest in extralinguistic matters, he must treat e.g. the sentences 'The men are animate' and 'The men are stones' on an equal footing; for him it is enough that they are grammatically correct (p. 58). A practitioner of universal grammar, however, must of necessity consider the relation between language and extralinguis­ tic reality. Therefore he must be both a grammarian and a philosopher. Siger de Cortrai complements this picture somewhat. He too assigns to the philosopher the task of investigating the most general aspects of extralinguistic reality. He notes, however, that the natural scientist has to investigate not just observable things, but also (psychological) notions (intellectus), including the functioning of the mental language. Moreover, Siger introduces the logician (logi­ cus), whose specialty is the investigation of the correspondence between notions and things (adequatio intellectuum et rerwn). Thus, the logician is interested in truth, whereas the grammarian is interested in correctness (congruitas). Insofar as meanings are something mental, the grammarian too has an interest in the study of mind (pp. 46-48, 87-88). These facts may be summed up as follows:

What kind of discipline is grammar, as so defined? Since Aristotle, it was customary to think that the subject matter of a science had to be something immut­ able; there could be no science about things in flux. Boethius the Dane makes this requirement more specific by noting that immutability need not be taken in an absolute sense. Conditional immutability is quite enough, as when the effect is always present if the cause is present (p. 30). A genuine science was taken to be one that explains its subject matter by revealing its causes (causae, principia). This was called the 'demonstrative mode'. There was also another, inferior way of doing science, namely the 'narrative mode', which consisted in just describing the subject matter. It was possible to treat the same subject matter in both ways. At its best, grammar is a demonstrative science, namely when the correct syntactic construc­ tions are explained as resulting from the way members of the different word-classes

228

UNIVERSAL HISTORY OF LINGUISTICS

fit together. Briefly put, the modes of signifying are the 'causes' of correctness (p. 32). This method of explanation is the same as Apollonius'. Although grammar is a demonstrative science, its causes are not ultimate or primary, like the causes investigated by the philosopher (p. 54). That is, the modes of signifying qua the 'proximate' cause of correctness have been caused by the modes of being qua the 'distant' cause of correctness (cf. 5.3.3), but these lie outside the grammarian's field of competence (p. 74). We have seen, however, that universal grammar subsumes both grammar and philosophy. This method of expla­ nation goes beyond Apollonius'. It is precisely on this issue that Boethius criticizes Priscian. In so doing, however, he seems to confuse the two roles of grammarian and of universal grammarian. On the one hand, Priscian is criticized for applying the narrative mode instead of the demonstrative one, and hence for giving an opinio instead of a scientia: "constructiones multas dicit, quarum tamen causas non assignat" ('he mentions many constructions whose causes he yet fails to identify'; p. 40). On the other hand, Priscian is criticized for exactly the opposite reason, namely for dealing with such ontological categories as "acting and undergoing, substance and accident, which the grammarian qua grammarian does not consider" (p. 62). It is true, however, that Priscian's incursions into the ontological realm are somewhat half­ hearted. Maybe Boethius just wishes to delimit more clearly the areas of immanent description of language (or 'explanation by proximate causes') and of ontologicopsychological explanation of language (or 'explanation by distant causes'). In addition to the distinction between demonstrative science and narrative science, there was the distinction between theoretical science (scientia speculativa) and practical science. For Pseudo-Albertus Magnus it was obvious that grammar is a theoretical science, simply because "its goal is to know" (p. 4). Thus, grammar does not exist, at least primarily, for utility (bonum), as practical sciences do. Thomas Aquinas (1224-1274) formulates the same distinction as that between 'contemplative intellect' and 'practical intellect'. In the former case things 'measure' (or test) the intellect and determine whether a notion is true or false. In the latter case the intellect 'measures' (or tests) things, more precisely artefacts, by determining whether they are 'true' or 'false', i.e. whether or not they satisfy their criteria (Arens 1984:409-410). The same distinction is made in identical terms by Siger de Cortrai (p. 86). Now Thomas Aquinas' instrumentalist language-conception would seem to call into question the view that grammar is not a practical science:

EUROPE

229

Aristotle adds that the sentence signifies arbitrarily, i.e. by the institution of human reason and will, ... just as all artefacts are caused by human will and reason... Reason can also use the sentence and its parts like instru­ ments though they do not signify by nature ... the sentence ... is the instrument of reason ... every instrument must be defined by its purpose, which is the use of the instrument; the use of the sentence is to signify a mental conception (Arens 1984:430). In other words, since sentences (or rather, utterances) are literally produced by men, it seems that grammar is, after all, a practical science. Pseudo-Albertus Magnus has, however, a cogent counter-argument. The grammarian does not inves­ tigate those psychological occurrences which are prompted by human will and culminate in utterances being produced. What the grammarian investigates and explains is, rather, the correctness of sentences. If the modes of signifying are proportional to one another in a sentence, the sentence is necessarily correct. That is, whether it is correct or incorrect, does not depend on our will (Non est in voluntate nostra quod sit congrua vel incongrua; p. 4). To the Modistae it was self-evident that every language contains the same set of principles which, taken together, constitute the universal grammar. Boethius the Dane puts it as follows: The entire grammar which is in one language is similar to that which is in another, and they both are of the same type (species), differing only with regard to variants of form (figurationes vocum), which are grammatical accidents (p. 28). He adds that even if the universal grammar abstracts away from every individual language taken separately, it does not abstract away from all languages taken together (p. 36). For Boethius, speaking is one of the 'natural operations'; it is nature itself which has created the 'instrument' that produces speech. From this he infers that the differences between, e.g., Greek and Latin are 'unnatural', or results of teaching (doctrina). If such differences are eliminated, there remains only one language which supposedly matches the universal grammar in an intimate way and can therefore be learned without any skill (ars) (p. 56). It might be tempting to equate here 'natural' with 'innate'. This would be a mistake, however, because the modes of signifying (which constitute the universal grammar) are secondary in the sense that they have been produced by the 'modes of understanding' (modi intelligendi), as these apply to the modes of being (cf. below). Consequently enough,

230

UNIVERSAL HISTORY OF LINGUISTICS

Pseudo-Albertus Magnus states categorically: "Notitia modorum significandi intellectui non est innata" ("The knowledge of the modes of signifying is not innate in the mind"; p. 38). The universal grammar is based on the isomorphism of language with mind and reality. This will be the topic of 5.3.3. Here I touch upon only the most general aspects of this isomorphic relation. First of all, it might seem to contain some redundancy, because —- to put it bluntly — one and the same thing is said three times, as in the triad 'thing—thing-concept—noun'. Yet, as Thomas of Erfurt points out, it is impossible to dispense with any of the three terms of this relation, for the simple reason that to be (esse), to understand (inteiligere) and to signify (significare) are three different (though isomorphic) 'principles' (p. 142).14 This presupposes, of course, that it is senseless to deny the existence of external reality. As Boethius the Dane puts it, "the modes of understanding are taken from (accepti a) the very modes of being — otherwise notions would be just figments of things, if nothing outside the mind corresponded to them" (p. 24). (It is safe to assume that Boethius has been thinking of this passage from Aristotle's Metaphysica 1010b, 30: "...but that the substrata which cause the sensation should not exist even apart from sensation is impossible".) The modes of signifying have also been 'taken from' the modes of being — via the modes of understanding — and refer back to them (p. 58). These two opposite relations deserve to be depicted more graphically:

Both Boethius (p. 72) and Thomas (p. 140) note that reality precedes mind, and mind precedes language. Boethius adds rather perceptively that although reality is for us necessarily reality-as-conceptualized, rather than reality tout court, it is nevertheless meaningful to distinguish between reality and its conceptualization. (Licet ergo res non possit intelligi praeter  modwn intelligendi, distinguit tarnen intellectus inter ipsam rem et ipsum modum intelligendi; p. 70, cf. here p. 213). The need to do so is evident in connection with concepts without corre­ sponding things (cf. below). Not just reality and mind, but also mind and language are two separable levels, as is shown by frequent cases of discrepancy between mental or 'internal' language and oral or 'external' language (cf. below). All the basic ingredients of isomorphism are present in the medieval concept of 'linguistic sign', as formulated by Pseudo-Albertus Magnus (p. 20). He dis-

EUROPE

231

tinguishes words and referents as material entities, on the one hand, and their psychological counterparts, on the other:

Pierre d'Ailly (1350-1425), for instance, expresses the same sign-conception in identical terms, and adds fully in Aristotle's spirit that  and  'naturally signify' A and D, respectively (natural signification being based on similarity), whereas A/B 'conventionally signify' CID (conventional signification being based on lack of similarity) (cf. Kneale & Kneale 1962:230-231). It has been wrongly assumed that this concept of linguistic sign is an invention of modern structural linguistics. (In fact, it was already formulated by Bhartrhari; cf. 2.4.2). Notice that there is no mention of a non-psychological level of meaning. Isomorphism is a rule which admits of several exceptions. Lack of iso­ morphism between reality and mind/language appears in the formation of abstract concepts: something (like the quality of whiteness) which is not a thing is reified or thought of as if it were a thing, and it is verbalized as an (abstract) noun. Accord­ ingly, Pseudo-Albertus argues that a quality (accidens) like 'whiteness, or 'moving' may be taken in two ways: either as it is ontologically, i.e. dependent on a thing (substantia), or in its own 'essence', i.e. extracted from the thing (p. 46). When it is taken as dependent, it is verbalized as albus ('white') or movetur ('is moving'); when it is taken in itself, it is verbalized as albedo ('whiteness') or motus ('move­ ment') (cf. also p. 42). In fact, a duality can be detected already in the use of an adjective like albus. Thus Anselm of Canterbury (1033-1109) claimed that while homo both 'names' (appellat) and signifies things (i.e. men), albus 'names' (white) things and signifies a quality (i.e. whiteness) qua dependent (Pinborg 1972:43-44). Siger de Cortrai confirms Pseudo-Albertus's account (pp. 68-70). In the case of abstract concepts we have a mismatch between ontological categories and psychologico-linguistic categories. In the case of either privative or fictive concepts there is no match: nothing ontologically corresponds to such concepts (or words) as 'emptiness' or 'centaur'. Pseudo-Albertus explains such

232

UNIVERSAL HISTORY OF LINGUISTICS

concepts as being derived from something that does possess ontological roots: understanding the lack of X presupposes knowing the thing X; and understanding a fiction presupposes knowing the things out of which the fiction has been put together (p. 38). Thomas of Erfurt notes, in addition, that in the case of negations, privations, and fictions the distinction between the modes of being and the modes of understanding disappears: here to be understood is to be (eorum intelligi est eorum esse) (p. 140, also 154, 198, 208). Finally, there is a great number of cases where isomorphism is claimed to break down between mind and language. This claim is uniformly based on the assumption, made already by the Stoics, that the language of mind is perfectly logical whereas oral language is less than logical. This assumption in turn gives rise to the critique of (everyday) language. This topic will be discussed in more detail in the end of 5.3.4 and in the beginning of 5.3.5.

5.3.3 Isomorphism as the Basis for Universal Grammar The medieval philosopher-grammarians exhibit a nearly obsessive concern with referential semantics. Among the questions they raise are the following: Do words signify extramental entities or mental (= psychological) entities? If the former, are there any extramental entities in addition to things? Does a word have different significations in the sentence and taken separately? If so, which signification is the primary one? Is there a difference between meaning and referent? Does a sentence signify a thing or a combination of extramental entities or a combination of psychological entities or a combination of extramental and psychological entities? All possible answers, and all possible combinations thereof, are given to these questions. Pinborg (1972) offers an in-depth overview of this entire field. It would take too long to review here the non-Modistic conceptions of semantics. As a representative example, however, I shall mention Thomas Aquinas' analysis of the sentence Homo est albus ('The man is white'), as I understand it on the basis of Pinborg's (pp. 87-89) presentation. Thomas distinguishes between reality and mind in such a way that at the former level the sentence speaks of one thing only (i.e. the man) whereas at the latter level it speaks about two entities (i.e. the man and his whiteness). This duality finds the following expression at the linguistic level: on the one hand, there is one single sentence; on the other hand, the sentence contains two words (plus the copula as the sign of synthesis). Or more elaborately:

EUROPE

233

The ambiguity of supposition parallels that of substantia, which means both 'substrate' (materia prima) and 'thing' (materia formata). In fact, Siger de Cortrai mentions (p. 70) the ontological hierarchy 'substance1 → substance2 → form', in which what precedes 'sustains' what follows. It is good to add that in Thomas Aquinas' analysis mental and linguistic non-identity is not just a matter of nonidentical notions or words, but rather of non-identical uses of notions or words, as in Homo est homo ('A man is a man'). Here the first occurrence of homo stands for a thing while the second stands for a quality. (Notice, incidentally, that the above diagram is rather similar to the one explicating Kātyāyana's and Patañjali's seman­ tics; cf. p. 74.) The Modistae have a different view of what reality is like. It is based on the assumption that reality contains other entities in addition to things, in particular actions and qualities. Therefore things, actions, and qualities are all equally good ontological categories, even if actions and qualities are necessarily dependent on things. Thus reality contains several modes of being, the most important of which are those corresponding to thing and action. Here there is some terminological variation. The way in which things exist is called modus substantiae, m. habitus, or m. permanentis while the way in which actions exist is called modus fluxus, m. successionis, or m. fieri. I shall follow Thomas of Erfurt and designate the differ­ ence between thing and action as that between modus entis ('the mode of that which is') and modus esse ('the mode to be'). Qua ontological categories, the modes of being are identified as the (basic) 'properties of things'. As was noted in 5.3.2, reality precedes thought. Once in existence, thought applies to reality and conceptualizes it. This is expressed by saying that the modes of understanding reproduce the modes of being. More precisely, a distinction is

234

UNIVERSAL HISTORY OF LINGUISTICS

made between 'active' and 'passive' modes of understanding such that the former precede the latter. Here there is room for two different interpretations: either active and passive modes capture the distinction between psychological processes and their results; or the active modes stand for mind (i.e. processes and results), while the passive modes stand for reality-as-conceptualized (res ut intellectae). Following Aristotle, it is claimed that both the modes of being and the modes of understanding are universal in character. The latter are taken to constitute a mental language which "signifies, conceives of, and apprehends the properties of things", as Thomas of Erfurt puts it (p. 140). Just as mind follows and signifies reality, language follows and signifies mind. Hence, language signifies reality via mind. The differences between the modes of being are reflected in language as the corresponding differences between the 'modes of signifying', i.e. the word-classes: For instance, nouns, verbs, and adjectives correspond to things, actions, and qualities, respectively. The ontological justification of the word-classes is supposedly universal; the modes of signifying derived from it are called the 'essential' ones (cf. below), and they constitute the core of the universal grammar. The other modes of signifying (comprehending most of the word-class attributes enumerated by Dionysius Thrax) qualify as merely 'accidental', and they may or may not be present in a given language. Although the word-classes are universal, their material expression may of course vary from language to language. Also at the linguistic level a distinction is made between active and passive modes, with the same potential ambiguity of interpretation as in the case of the modes of understanding. The assumption of linguistic universality was all the time implicit in Aristotle's overall conception. If mind contains universally valid distinctions, and language expresses mind, then it has to express those distinctions, even though the actual forms employed to do so are arbitrary, or kata sunthēkēn. This is what is meant by saying that language is 'arbitrary'. The 'arbitrariness' of language does not mean that its relation to mind could be entirely unsystematic or chaotic. Otherwise it would not make sense to say that language corresponds to mind (cf. here p. 175). Now I shall show how the ontological justification works in practice, concentrating first on the four most important word-classes:

EUROPE

235

Although this is a perspicuous way of presentation, it gives less information than a cross-classification would give. For instance, the mode of determinate essence applies also to verbs and participles, and the mode of self-sufficience applies also to pronouns. Moreover, the mode of distance applies to verbs only when taken separately. In the sentence verbs have 'accidentally' the mode of proximity (since they combine with nouns), which means that they are indis­ tinguishable from participles, except that for the latter the mode of proximity is 'essential'. Finally, adverbs are like adjectives, except that they have modus esse. A word which belongs to some word-class is called pars orationis. It exemplifies the concept of concrete or actual word. There is also a concept of abstract word, called dictio. An abstract word is common to or underlies several concrete words. For instance, there is a common lexical meaning behind such words as doleo ('I suffer'), dolor ('pain'), dolens ('suffering'), and dolorosa ('painful'), and it justifies the postulation of an abstract word. Some of these facts may be summed up as follows:

236

UNIVERSAL HISTORY OF LINGUISTICS

Significatio is the lexical meaning common to several actual words belong­ ing to different word-classes whereas consignificatio contains both the ontological character of each word-class (i.e. the 'essential' modes of signifying) and the set of grammatical meanings possessed by all members of a given word-class (i.e. the 'accidental' modes of signifying). The concept of dictio remains problematical: a dictio is supposed to consist of significatio and vox, but here vox must exceptionally stand for some sort of abstract form, since it is a form common, e.g., to doleo, dolor, dolens, and dolorosa. The modes of being, understanding, and signifying are characteristic of simple entities, but these necessarily combine to form complex entities. Thus Boethius the Dane sets side by side (isomorphic) 'constructions' of things, notions, and words (p. 74). A sentence like Homo est albus does not refer to a single extramental entity, as Thomas Aquinas would have it, but to two such entities, namely a thing (substantia) and a quality (accidens) which 'inheres' in or 'depends' on the thing. Thing, quality, and inherence are signified, respectively, by homo, albus, and est. The Modistic doctrine is a grandiose attempt to construct a theory which would account for everything that there is: language, mind, and reality. This must be clearly understood, before one turns to examine its defects. At the moment I by­ pass the fact that the data-basis employed by the Modistae was much too narrow. There are also pitfalls of a more structural kind that threaten this type of endeav­ our. First, the three supposedly isomorphic levels may get confused. Second, there is the danger of circularity: an ontological distinction is first postulated exclusively on the basis of a linguistic distinction and is then made use of to 'explain' the latter. The Modistae were occasionally guilty of both types of error. To begin with, the mode of (in)determinate essence is also called the 'mode of (in)determinate conception' (modus (in)determinatae apprehensionis), without any recognition that this entails a shift from the modes of being to the modes of understanding. Similarly demonstrative pronouns are characterized as resulting from the 'mode of certainty and presence', without any indication that, rather than belonging to extramental reality, this mode combines a psychological state with a specific type of speech situation. The verb is also said to be motivated by an

EUROPE

237

additional mode, namely modus enuntiabilis de alio (the 'mode of that which is said about something else'), which certainly does not belong to the modes of being. There is also terminological inconsistency in the formulation of modes of signify­ ing: on some occasions they just repeat the 'preceding' modes of being, as when it is said that 'nouns and pronouns signify by (significant per) the modus entis'; but on other occasions an attempt is made to introduce special terms for the modes of signifying. In many cases supposedly ontological distinctions seem merely to reflect linguistic distinctions. In the diagram on p. 235 the modes of 'closeness' and 'proximity' are clearly synonymous; different terms are used merely to 'ontologically' distinguish adjectives from participles. It is even more difficult to find an ontological justification for distinguishing verbs from participles. While Thomas of Erfurt straightforwardly assumes a single ontological counterpart for each case ending (pp. 186-194), Boethius the Dane frankly admits that in connection with the case endings there is a discrepancy, i.e. a many—one correspondence, between reality and language (pp. 150-151). The ontological correlates of prepositions and conjunctions are also rather ad hoc. Furthermore, the notion of dictio is apt to create ontological difficulties. Both Boethius the Dane and Siger de Cortrai assume that a 'thing' (res) underlies and precedes its own properties (proprietates rei), i.e. the modes of being corre­ sponding, e.g., to the words doleo, dolor, dolens, and dolorosa (cf. above); and since a 'thing', while preceding the modes of being, produces its own psychological and linguistic correlates, it is inferred that these too precede the modes of being. However, the view that mind and language can in this way precede the structure of reality, would seriously undermine the entire Modistic edifice. I wish to emphasize, in conclusion, that none of the mistakes that I have mentioned above is conceptually necessary. Once they have been removed, the rational kernel of the Modistic doctrine remains, waiting to be developed further.

5.3.4 Grammatical Description When speaking of the ontological determination of language, the Modistae concen­ trate on the word-classes, which has sometimes been taken to mean that they have no theory of the sentence. This is a serious mistake. Being thoroughly acquainted with De Interpretatione, every Modista insists that 'truth is in the combination' (veritas est in compositione), i.e. the synthesis of notions or words which produces a (mental or oral) sentence. And they all accept Aristotle's view that this synthesis,

238

UNIVERSAL HISTORY OF LINGUISTICS

prompted by the human will, is achieved by means of the explicit or implicit copula, which in itself has no lexical meaning, but is, rather, the sign of the sentence-synthesis and, at the same time, of predication. Therefore the (finite) verb is unique among words in possessing the sentence-creating capacity. Every Modista, and more generally every medieval philosopher, also accepts the Aristotelian trichotomy of 'word—sentence—syllogism', or terminus— propositio—syllogismus,as Thomas of Erfurt puts it (p. 148). Radulphus Brito (d. 1320) gives a more elaborate taxonomy in which these three entities of increasing complexity are presented as resulting from three corresponding 'operations of the intellect', whose functioning is analyzed at the level of 'first intentions' (roughly, perceptions) and 'second intentions' (= universal concepts), on the one hand, and both in the passive mode and in the active mode, on the other (cf. Pinborg 1972: 197-199). Accordingly, words and word-classes are seen as subordinated to more inclusive wholes. Albertus Magnus (c. 1190-1280) voices the common opinion, when he notes that "when we defined the noun and the verb, we defined them as elements of the complete sentence" (Arens 1984:371). In what follows, I shall consider the syntactic theory of the Modistae. There are some difficulties connected with this undertaking. First, syntactic considerations are often mixed up with ontologico-semantic ones. Secondly, syntactic phenomena are described by several sets of terms, and it is often difficult to decide whether terminological differences imply any differences of opinion. Siger de Cortrai offers a good starting point. He observes the existence of parallel relations of dependence in ontology and in syntax: Adiectiva dependent in esse a suis substantivis sicut accidentia a substantiis (p. 10). That is, adjectives (which signify qualities) depend on their nouns (which signify things) just like qualities depend on things. The situation seems to be much the same with nouns and verbs. First, there is the question of mutual order. A property (proprium) is supposed to follow immediately that which it is a property of; being (esse) is a property of that which is (ens); therefore the verb, being the verbalization of the modus esse, must immediately follow the noun, which is the verbalization of the modus entis. This canonical word order (which is often contradicted in practice) is also motivated by the traditional view that since the action 'emanates' from, and hence 'follows' the agent, verbs must follow nouns in the nominative. It is a natural corollary that nouns in the oblique cases (qua expression of the patient) must in turn follow verbs. Second, there is the question of dependence. Becoming (fieri) is dependent on things, and every verb signifies by the mode of becoming; therefore every verb

EUROPE

239

signifies by the mode of dependence and requires a subject noun {supposition) which 'sustains its dependence' (dependentiam eius sustentat;p. 16).15 These two points of view are combined: The dependence of the verb on its nominative and on its oblique is reformulated as dependence on 'what precedes' (ex parte ante) and on 'what follows' (ex parte post), respectively. The key term in the above passage is 'to require' (requirere). This term can be given, and is in fact given, two opposite interpretations. It is a matter either of a superior commanding an inferior or of an inferior begging a superior. Accord­ ingly, there obtains the following terminological ambiguity:

Ontologically speaking, the verb depends on the noun (cf. above). Yet the verb also requires the noun to take a certain case ending. If this is conceived of as the verb governing the noun, then there is the apparent contradiction of the verb being both a 'dependent' and a 'governor'. This contradiction can be solved in two ways: either by keeping ontology and syntax separate and taking verbs to be ontologically dependents and syntactically governors; or by making syntax more consonant with ontology by interpreting requirere as petere. The fact is, however, that the Modistae were misled by their own terminology. Siger de Cortrai makes a good attempt at defining the concept of govern­ ance. Everywhere the more valuable governs the less valuable, and this principle applies also to language: Orationem complet quae proprie regit omnes alias partes et a nulla alia regitur, videlicet verbum ('The sentence is made complete by the word(-class) which governs all other words and is governed by none, namely the verb'; p. 51). Pseudo-Albertus too conceives of the verb as both depending on and governing its nouns (p. 114, 124). Siger is aware that this position may seem self-contradictory; in any case it conflicts with the traditional view that nouns are more important than verbs. He tries to solve this conflict by distinguishing between the (abstract) word-level and the (concrete) sentence-level: Although the noun, taken as such, is more valuable and more self-suf­ ficient [than the verb], the verb is more valuable as far as the entire construction is concerned, because it is more important for the essence

240

UNIVERSAL HISTORY OF LINGUISTICS

(forma) of the whole, considering that the essence of the sentence is synthesis (compositio) (pp. 52-53). This distinction was formulated, more generally, as that between partes orationis in habitu and partes orationis in actu (cf. Pinborg 1967:90). Today one might wish to say that it is the distinction between words as part of langue and words as part of parole, except that this would still not justify the view that the roles of nouns and verbs are reversed when one moves from one level onto the other (cf. here p. 299). It was also well-known that when a word occurs as part of a sentence, its meaning is more specific than it is in abstracto: significatwn eius specificatur per ea, quae sibi adiunguntur (Boethius the Dane, p. 218). To go back to Siger, in his opinion the verb is dependent on, or 'less important' than, the noun both ontologically and outside the sentence, which seems to imply that the ontological point of view is taken to prevail in the latter type of context. What Siger actually means by regimen and dependentia, must be inferred from his examples:

The relations of dependence and governance are seen to coincide, except in the 'N + A' construction. In our interpretation of Sībawaihi's syntax, if A governs B, then necessarily  depends on A. Here, in the most cases, if A governs B, it also depends on B. More terminological confusion is created by the fact that, according to Siger, a preposition does not govern, but rather 'serves' (deservit) its noun (p. 63). Notice also that this interpretation of the 'N 4- N-gen' construction entails the illogical feature that in a sentence like Filius Socratis currit ('S.'s son is running') the subject of the sentence is a dependent. Thomas of Erfurt's Grammatica Speculativa has for some reason been taken to contain the 'definitive' version of Modistic syntax (pp. 272-320). In many respects it is rather disappointing. The term regimen has been entirely replaced by dependentia (and its derivatives), and dependents are said to 'ask' (petere) or 'require' (exigere), rather than 'govern' their terminants (p. 282). By now depen-

EUROPE

241

dence has become a preponderantly syntactic notion: yet the same units as before qualify as 'dependents' and 'terminants'. This leads to oddities. For instance, in the construction cappa Socratis it is possible to think that Socrates qua possessor is ontologically more important than the cap qua possession. Syntactically, however, a similar line of argument hardly makes sense. To be more precise, every sentence is analyzed into a set of binary 'con­ structions', each of which consists of one terminant and one dependent. The sentences have a canonical word order. A dependent following its terminant constitutes an 'intransitive' construction whereas a dependent preceding its ter­ minant constitutes a 'transitive' construction. The terminant is also called principium constructionis ('the cause of the construction'). Pinborg (1977:xxx) and Covington (1984:48-51) complain that no general definition of 'dependence' has been given by the Modistae. According to Robins (1979:83), the Modistic notion of dependence "coincides ... with no single syntactic relation". Pinborg suggests, as the best approximation, the idea of dependents functioning as predications on terminants. Covington prefers an ontological inter­ pretation: a dependent has in general no thing-referent of its own, but refers (indirectly) to the thing-referent of its terminant. Apart from covering only some of the 'constructions', this interpretation entails that in Modistic syntax there is no clear distinction between formal and ontologico-semantic considerations. This is most probably true. By way of résumé, I present here the Modistic (dependency) analysis of Thomas' example sentence (p. 292) Summa doctrina Tullii Rhetoris instruit subtiliter artem Rhetoricam ('The excellent doctrine of Tullius the Rhetorician teaches subtly the rhetorical art'). The words of this sentence will be represented by corresponding letters. Notice that

reads here 'X depends on F .

It may be added that Pseudo-Albertus extends the notion of governance (regimen) so as to cover also relations between sentences (p. 122). Similarly Siger argues that in a sentence like Magistro legente pueri proficiunt ('While the teacher is reading, the boys make progress') the 'ablativus absolutas' construction magistro legente is governed by the rest of the sentence. (This is one more case where governance and dependence do not coincide.) It follows — as Siger explicitly notes

242

UNIVERSAL HISTORY OF LINGUISTICS

— that not only (members of) word-classes but also sentences, whether implicit or explicit, possess modes of signifying (pp. 54-55). According to Thomas of Erfurt, the central notion of syntax (= diasynthetica) is 'construction'. And since to know a phenomenon means to know its causes, the constructions too are known by knowing their causes, i.e. the four Aristotelian causes (pp. 292-294; cf. here p. 181). The material cause of a con­ struction are simply the two words (constructibilia) which constitute it. The formal cause is the union (unió) of the two words. There are two types of efficient cause: the internal one are those modes of signifying which differentiate between terminant and dependent, whereas the external one is the mental operation which achieves the union. The final cause is the expression of a mental concept (expressio mentis conceptus). Later on (pp. 312-314) this is characterized as the 'immediate' purpose of a construction while its 'remote' purpose is to "generate perfect sense in the mind of the hearer". The notions of congruitas and perfectio are also discussed at length. The former, being a syntactic notion, requires the compatibility of grammatical rather than lexical meanings. Thus both cappa nigra ('black cap') and cappa categorica ('categorical cap') qualify as 'congruous', although the former is 'proper' while the latter is 'improper'. Moreover, 'congruity' means similitudo when the dependent adopts the properties of the terminant, and proportio when the properties of the dependent fit in with those of the terminant (pp. 308-310). 'Perfection' is a notion which applies only to sentences. For a sentence to qualify as 'perfect', it is re­ quired, above all, that it contain both a supposition ('subject') and an apposition ('finite verb') (p. 314). As far as the more specific character of syntactic description is concerned, the Modistae continue the tradition of 'abstract analysis' (cf. 5.2.3-C). A signifi­ cant number of Apollonius' analyses were taken over and applied to Latin by Priscian. They recur in Modistic literature, confirming Siger's claim that "we must follow Priscian and Donatus" (Priscianus autem et Donatus quorum opiniones in grammaticis prosequi debemus; p. 62). I shall give here only one example, taken from Pseudo-Albertus (p. 120-122). In the sentence Sedentem atribulare est impossibile ('It is impossible for one who is sitting to be walking'), the accusative-cum-infinitive construction Sedentem atribulare is the subject (supposition). But it is also a sentence (oratio), i.e. a sentence equivalent to Sedens ambulat ('The one who is sitting is walking'). (Radulphus Brito argues that since the embedded sentence is the subject, it must be in the nominative; cf. Covington 1984:101). Moreover, sedens (lit. 'sitting') itself contains a sentence, since it is equivalent to qui est sedens ('[The one] who is

EUROPE

243

sitting'). As a consequence, the original sentence contains in fact three sentences, which may be presented in a modern notation as follows:

An actual sentence like Sedentem ambulare est impossibile is called con­ structio vocis ('vocal construction'; Boethius the Dane, p. 212) or constructio secundum sensum ('perceptible construction'; Thomas of Erfurt, p. 306), whereas its (abstract) analysis is called constructio apud (or secundum) intellectum ('mental construction'). This is of course the distinction between external/oral language and internal/mental language, which goes back to Plato. We have seen that the Modistae in general reformulated it as a distinction between modes of signifying and modes of understanding. In spite of the general isomorphism, there are cases where language fails to reflect reality and mind. Boethius the Dane notes that just as there is ambiguity (aequivocatio) resulting from the fact that one word (vox) designates many things, there is ambiguity in that one mode of signifying designates several modes of being (and of understanding) (p. 112). Case endings are a good example of this (pp. 150-152). When there is a contradiction between modes of understanding and modes of signifying, the former must be given priority. For instance, desidero legere ('I wish to read') is a transitive construction fully on a par with desidero lectionem (T wish something to read'), and it is just an accident of linguistic form that an infinitive cannot have the accusative ending (p. 212). (It may be added that for Apollonius a sentence like boulomai peripatein, i.e. 'I want to walk', was really equivalent to *boulomai emauton peripatein, i.e. T want myself to walk'; Peri suntaxeōs III, 161.) In Boethius' claim one can see the beginnings of a critique of natural language, a position that was already a characteristic feature of Stoicism (cf.

244

UNIVERSAL HISTORY OF LINGUISTICS

p. 185). It was further developed in the logical linguistics of the Middle Ages (cf. 5.3.5), and became a real fashion in the 17th century (cf. 5.4.2). Modistic syntax suffers from being limited, in general, to analyzing simple sentence-structures. The reason is obvious: the ontological motivation of complex sentences is much more problematical; what kind of state of affairs corresponds e.g. to a negated conditional sentence? Yet the notion of mental language enables the Modistae to maintain the tradition of abstract analysis and hence to consider sentences also without regard to their ontology (cf. the discussion of Pseudo­ Albertus on p. 243). Although Modistic syntax is in principle independent of contemporary logical linguistics, it is precisely the notion of mental language which provides the link between the two.

5.3.5 Logical Linguistics Medieval logicians practised a type of linguistic analysis in which a complex sentence was reduced to a set of simple sentences, i.e. sentences à la 'A is B', where A is a noun and  is a noun, an adjective, or a verb (= participle). Nouns, adjectives, and verbs were called categoremata. The complex sentences to be analyzed were such as to contain either implicit or explicit syncategoremata, or grammatical words like 'all', 'some', 'not', 'only', 'except', 'necessarily', 'if, 'to begin', etc. Depending on the type of complex sentence, the analysis was called by various names. The most common name was expositio. The analysandum or explicandum was called exponibile while the analysans or explicans was called (the set of) exponentes. When speaking of expositio, Pinborg (1972) does not hesitate to use the term 'transformational analysis', identifying exponibile and exponentes as 'surface structure' and 'deep structure', respectively. In his opinion, this type of analysis was not performed by grammarians, but only by logicians. This is not quite correct, however. Although logicians concentrated more on sentences containing syncategorematic expressions, there is no hard-and-fast division between logical analysis and grammatical analysis, as will be shown below. Besides, Modistae themselves regarded their grammatical description as an instance of expositio. Boethius the Dane, for instance, says that a participle like legem ('reading') merely 'gives to understand' (dat intelligere) an acting substance, whereas the corre­ sponding relative clause qui legit ('who is reading') 'analyzes' (exponif) it (p. 184).

EUROPE

245

Kretzmann (1982) gives many representative examples of expositio. In what follows I shall concentrate on the use that William Ockham (c. 1285-1347) makes of it. In the first three chapters of Summa Logicae he defines the notion of propositio mentalis, citing Aristotle, Boethius, and St. Augustine on the very first page to show his basic agreement with the tradition. Mental language is simpler than spoken language, since "there is no reason to postulate irrelevant elements among mental terms" (p. 52 in the 1974 edition). First, synonymy is dispensed with. Second, "although all of the grammatical features of mental names belong to spoken names, the reverse is not true" (p. 53). Thus case and number belong also to mental language, whereas gender and declension do not. Only such 'grammatical features' are ascribed to mental language as can affect the truth-value of sentences. The mental sentence is primary with respect to the spoken one. Mental terms signify naturally whereas spoken (and written) terms signify conventionally. While it has always been evident that mental terms signify things, Ockham is careful to point out that spoken terms signify these same things, even if 'second­ arily', i.e. secondarily as compared with the signifying function of mental terms. It is emphatically not the case that spoken words would 'primarily' signify mental terms. Ockham's position is, rather, that spoken terms signify things in 'sub­ ordination' to mental terms (p. 50). The latter may also be called 'concepts', 'intentions', 'likenesses', or 'intellects' (p. 74). As a logician, Ockham is interested only in the analysis of the mental language. It too contains both categorematic and syncategorematic terms. The latter are defined as follows: ...a syncategorematic term does not, properly speaking, signify anything; however, when it is combined with a categorematic expression it makes that categorematic expression signify something or supposit for [i.e. refer to] something in a determinate manner,... (p. 55). Because mental language contains syncategorematic words, and because expositio consists in analyzing these away, it follows that Ockham distinguishes between three levels of language: First, spoken language; second, mental language qua exponibile; third, mental language qua exponentes. As a first step, however, Ockham wishes to reanalyze most of the categorematic words. This is required by his austerely nominalist ontology, which admits only of individual (and concrete) things and qualities, ordinarily referred to by such expressions as 'this man' and 'this white (spot)'. As a consequence, all

246

UNIVERSAL HISTORY OF LINGUISTICS

abstract words (except, curiously enough, quality-based abstractions like 'white­ ness') must be explained away. For instance, a word like humanitas ('humanity') is analyzed as an implicit sentence containing concrete categorematic words and syncategorematic words: Necesse omnes homines... ('Necessarily all humans...'). Next, all concrete categorematic words not exemplifying the categories of thing (= 'substance') and quality, e.g. words expressing quantities and relations, are reanalyzed in terms of thing-words and quality-words (for details, cf. Loux 1974a). At this stage of the Ockhamian analysis, all sentences are represented in a form in which they contain only thing-words and/or quality-words as well as syncategorematic words. The next task is to get rid of as many syncategorematic words as possible. More precisely, Ockham intends to arrive at sentences which contain only conjunctive and disjunctive sentential connectives as well as identitystatements like 'Ai = Bj' or 'Ai ≠ Bj', where 'Ai' and 'Bj' refer to individual things and/or qualities. It is questionable, to say the least, whether this kind of reductive analysis can be carried out in practice, considering the complexity of typical naturallanguage sentences. It is good to point out, however, that Ockham's analysis has at least some merits which are too seldom recognized. Traditionally, it has been thought that in a sentence like 'This man is white' the subject stands for something particular whereas the predicate stands for something general or 'universal' (i.e. the quality of whiteness). Ockham correctly notes that this conception is unjustified. If we think of a situation where we might truthfully assert the above sentence, it is clear that the predicate in no sense expresses something more general (or abstract) than the subject: we equally perceive the man and his being white. (But, contrary to Ockham, this can be taken to mean that we need two general concepts, in addition to their particular exemplifications; cf. Itkonen 1970:113-118). The use of identity-sign may in turn be explained as follows. Ockham assumes that there is a finite number of, e.g., men and occurrences of whiteness, so to say, each of which has its own name 'A 1 , 'A2', etc. and 'B 1 , 'B2', etc., and when the sentence 'This man is white', is asserted truthfully, this is taken to amount to the assertion that the particular man pointed to by the speaker coincides with a particular occurrence of whiteness. According to Loux's (1974b:28-31) interpretation, Ockham reanalyzes the sentence-types in Aristotle's square of oppositions (cf. p. 180) truth-functionally as follows: A) [(A1=B1) V ( 1 = 2 ) V... (A1=Bn)] & [(A2=B1) V...] &... E) [(A1≠B1) & (A1≠B2) &... (A1≠Bn)] & [(A2≠B1) &...] &...

EUROPE

247

I) [(A 1 =B 1 ) V ( 1 = 2 ) V... (A1=Bn)] V [(2=1) V...] V... ) [(A1≠B1) & (1≠2) &... (A1≠Bn)] V [(A2≠B1) &...] V... Thus, the type A), i.e. 'Every A is ', is analyzed as saying that this A coincides with this  or that  or etc., and that A coincides with this  or that  etc., and etc. In other words the type A) is reanalyzed as a (finite) conjunction of disjunctions of identity-statements. The type E is in turn reanalyzed as a conjunction of (conjunctions of) negated identity-statements. And so on. It may be added that Freddoso (1980:18) prefers to reconstruct Ockham's analysis truth-conditionally, and hence less formally: A)

'Every A is B' is true iff there is something to which 'A' refers, and there is nothing to which 'A' refers and 'B' does not refer.

E)

'No A is B' is true iff there is nothing to which both 'A' and 'B' refer.

I)

'Some A is ' is true iff there is something to which both 'A' and 'B' refer.

O)

'Some A is not B' is true iff there is nothing to which 'A' refers, or there is something to which 'A' refers and 'B' does not refer.

Since 'Every A is B' entails, along the Aristotelian-scholastic lines, that there is some A, the sentence is false if there is nothing to which any 'Ai' refers. It follows, paradoxically, that under the same circumstances 'Some A is not B', i.e. the negation of 'Every A is B', is true. The same applies, of course, to 'No A is B', but this is intuitively less paradoxical. While Part I of Summa Logicae is devoted to the theory of terms, Part II is devoted to the theory of propositions. After explaining "what is required for the truth of propositions" with 'every' or 'some', Ockham proceeds to analyze pasttense and future-tense propositions and modal propositions. They are reduced to elementary, i.e. non-modal present-tense, propositions, which have the canonical form 'A is B', interpreted as an identity-statement. All this is done in the chapters 1-10 of Part II. 'Exponible' propositions in the strict sense, dealt with in the chapters 11-20, are also reduced to conjunctions or disjunctions of elementary

248

UNIVERSAL HISTORY OF LINGUISTICS

propositions. The chapters 21-29 deal with 'conversions', i.e. pairs of propositions having the form 'A is B' and 'B is A'. Finally the chapters 30-37 deal with 'hypothetical' (i.e. complex) propositions, including conditional, conjunctive, disjunctive, causal, and temporal propositions. Ockham applies his reductive analysis both to lexical and to grammatical terms. He is skeptical about there being any mechanical rules for analyzing the meaning of such lexical terms as 'blind' or 'foolish' (p. 121). In connection with grammatical terms, however, he is moving on firmer ground. I shall give here only two examples. 'Socrates, insofar as he is a man, has a colour' is called a 'reduplicative' proposition. It is characterized by Ockham as follows: If the reduplication is made to express concomitance [and not cause], then for the truth of the proposition four propositions which, as it were, ex­ pound it are required. One is a proposition in which the principal predi­ cate is truly predicated of the principal subject. The second is a prop­ osition in which what falls under the reduplication is predicated of the principal subject. The third is a proposition in which the principal predi­ cate is predicated of that which falls under the reduplication, taken univer­ sally. The fourth will be a true conditional proposition in which one argues from that which falls under the reduplication to the principal predicate, ... (p. 126; emphasis added). These rules are purely mechanical, or can be made such. When applied to the original sentence, they produce the following four sentences: 'Socrates has a colour', 'Socrates is a man', 'Every man has a colour', and 'If a man exists, then something which has a colour exists'. It is perfectly accurate to say that the combination of these four sentences represents the logical form of the original sentence. As a final example, consider Ockham's analysis of 'categorical propositions which contain the pronoun who', in the chapter 15. He starts by noting that "any proposition, being categorical in form, which contains the relative pronoun who should be given several exponents, since each such proposition is equivalent to a conjunctive proposition". He goes on to distinguish between several cases, de­ pending on how the sentence is quantified. For instance, 'A man who is white is running' is equivalent to 'A man is white and he is running' or 'A man is running and he is white'. Universal quantification, however, produces an ambiguity:

EUROPE

249

Thus Ockham distinguishes between the restrictive and non-restrictive uses of relative pronouns, as we would say. The truth-conditional character of his analysis is again evident from his remark that one or the other of the two pairs of propositions is required for the truth of the original proposition. On the other hand, it is clear that Ockham's analysis qualifies equally well as logical and as gram­ matical. In other words, there is no difference of principle between his analysis of relative clauses and Pseudo-Albertus's analysis of the sentence Sedentem ambulare est impossibile (cf. p. 243). Indeed, it is my thesis that within linguistics the collective achievement of the Middle Ages must be seen as being constituted by the Modistic grammar, on the one hand, and by Ockham-type logical linguistics, on the other. There is a division or labor such that the Modistae, being interested in ontological motivation, con­ centrate more on the analysis of simple sentences, whereas Ockham, being interested in the logically perfect mental language, concentrates more on (either implicitly or explicitly) complex sentences. The two approaches complement each other rather well. This is only the more remarkable, given their deep philosophical disagreement, which will be the topic of the next subsection.

5.3.6 The Demise of the Modistic Doctrine By Thomas of Erfurt's time the Modistic grammar had reached a stage of stag­ nation. To the non-initiated it seemed to result in nothing but a proliferation of artificial distinctions and definitions. Unable to renew itself, it became an easy target for criticism. Main-stream scholastic philosophy made several ontological commitments. It assumed that extralinguistic reality was divided into different types of entities which exemplified corresponding abstract (non-psychological) concepts. Moreover, in the realm of mind a distinction was made between subjective and objective, for instance between acts of understanding and that which was understood. While many found this kind of ontology unnecessarily rich, William Ockham was its most forceful critic. In his own nominalist philosophy he tried to get along with as few

250

UNIVERSAL HISTORY OF LINGUISTICS

types of entities as possible. In his youth he held the extreme view that ostensible differences between extralinguistic (types of) entities were just a matter of different words being used. Later he added the psychological dimension and distinguished between oral and mental words, with their respectively conventional and natural manners of signifying (cf. p. 245); the two classes of words were reduced to acts of speaking and thinking. He emphasized, moreover, that no differences existed in the external world independently of language and mind. Hence, his world-view contained only the undifferentiated 'stuff of extralinguistic and extramental reality, plus verbal and psychological occurrences. The criticism that was directed against the Modistae was of basically nominalist inspiration. In other words, it was more philosophical than linguistic in character. The principal critic was Johannes Aurifaber, who lived in Erfurt around the year 1330. His Determinatio de Modis Significandi, republished in Pinborg (1967:215-232), contains several arguments against the Modistic position. Some of them concern the correct interpretation of Aristotle's philosophy and can be omitted here. The most important arguments are the following: i) The 'modes of signifying' are unnecessary because they were not needed in antiquity. They are unnecessary also because the logicians can dispense with them. ii) Words are arbitrary (among other things, because they do not reflect all aspects of their referents). Being arbitrary, they have no systematic causes. There­ fore, the science of X being an investigation about the causes of X, there can be no science of language. In particular, universal grammar is impossible. iii) The Modistae treat mind and language alike, by making modes of signifying correspond to modes of understanding. This is a mistake, however, because mind is determined by reality whereas language is arbitrary (as argued immediately above). iv) If meanings existed, a meaningful word (or sound) would be more complex than a meaningless one. But it is not. Therefore there are no (lexical or grammatical) meanings, apart from acts of understanding. At least the following responses can be made to the preceding objections: Ad (i): The reference to antiquity is irrelevant. Perhaps modes of signifying were used implicitly; or if they were not, only the worse for ancient linguistics. The reference to logic is irrelevant for the same reasons. Besides, logic qua study of (necessary) truth may diverge from grammar qua study of correctness. Ad (ii): Words are not arbitrary, except in the trivial sense that their 'material clothing' varies from language to language. It is the basic insight of the Modistae, summarized in their notion of universal grammar, that there is a common

EUROPE

251

set of distinctions behind the variation exhibited by different languages; and such distinctions have been caused by extralinguistic distinctions. (Besides, there is nothing arbitrary in that words reflect only some aspects of their referents. It is not even clear what it would mean to reflect all aspects of a referent.) Ad (iii): It is perfectly right to treat mind and language alike, because language too is determined by extralinguistic reality (as argued immediately above). Ad (iv): Sounds-cum-meanings are tautologically more complex entities than mere sounds. But this greater complexity is of course not evident at the level of sounds. These responses suffice to refute the nominalist critique of the Modistic doctrine. But it can be shown, in addition, that nominalism itself suffers from serious defects. First, it is inconsistent to claim, on the one hand, that mental words, being determined by reality, signify naturally, and to maintain, on the other hand, that there are not distinctions in extramental reality as such. If there are universal mental distinctions which have been caused by something extramental, they cannot have been caused by some undifferentiated 'stuff. Secondly, reducing meanings to individual-psychological occurrences leads to psychologism, which had already been refuted by Abaelard and Ars Meliduna (cf. p. 226), and was refuted again by William Ockham's contemporary Walter Burley (cf. Kneale & Kneale 1962:266-271, Pinborg 1972:144-146). The inadequacy of nominalism is also evident from the fact that, when made explicit, it invalidates its own foundations. Originally, nominalism is the expression of a 'hard-nosed' attitude which intends to do away with unnecessary abstractions and to concentrate on that which is real. Perceptions are secure whereas their extramental causes are less so. It follows that the importance of extramental reality is diminished. The next logical step is to 'bracket' it entirely. This was done by Nicholas of Cusa (1401-1464), who came, paradoxically, to combine Ockhamism with Neo-Platonism.16 This transition from ultra-empiricism to de facto idealism was repeated in the 18th-century English philosophy. In sum, the arguments against the Modistae were invalid. This was quite clear to the defendants of the Modistic doctrine: for instance, the anonymous author of Sophisma 'Tantum Unum Est' (Pinborg 1967:233-268), which is a response to Johannes Aurifaber, presents most of the counter-arguments that have been given above. This was to no avail. The Zeitgeist had turned against scholasticism, and therefore the Modistae qua scholastic grammarians could not hope to get a fair hearing. In the attack against scholasticism an unlikely alliance was formed by Ockhamist nominalism and nascent Humanism. The latter prevailed, which brought

252

UNIVERSAL HISTORY OF LINGUISTICS

about a decline not only in grammatical but also in logical studies. The fate of medieval linguistics, understood as the combination of Modistic grammar with Ockham-type logical analysis, teaches an important lesson: it is perfectly possible that during the history of a human science people reject a theory, not because they have cogent arguments against it, but because they are just fed up with it. It is more difficult to imagine the same thing happening within a natural science with practical applications. When I first embarked upon the study of medieval linguistics, it was my intention to ferret out its connections with the catholic religion. I was surprised that I could find none. With hindsight this is understandable, given the close adherence to Priscian and to Aristotle evinced by the Modistae. It is a different matter that the language-conception of medieval mysticism is thoroughly imbued with religious elements (cf. Seppänen 1985: Chap. 4).

5.4 Grammatical Theory 1500-1900 During the last 25 years there has been a lot of work done on post-medieval grammatical theory in Western Europe. It is my considered opinion that the achievements of this period have been overrated, because those who specialize in the linguistics of the 'Modern Age' are likely to have little knowledge of, and no interest in, either ancient or medieval linguistics. It is true that also 'pre-modern' (in the sense of 'pre-Renaissance') linguistics has recently been studied to an increasing extent. It is my impression, however, that its permanent value has not been underscored forcefully enough. In the first three sections of this chapter I have tried to put the record straight, by showing that such scholars as Chrysippus, Apollonius, Boethius the Dane, and William Ockham are not just of an antiquarian interest for us. We do not have to lower our standards of scientificity when we approach them. In many cases they made those discoveries that have generally been thought to testify to the 'spectacular success' of synchronic grammatical theory in the 20th century. I do not deny that genuine progress has been achieved in the 20th century. In the present section I intend to show, however, that no new results were attained in the post-medieval grammatical theory before the 20th century; and some old results were forgotten. What was really new during 1500-1900, was the gradual discovery of diachronic linguistics. But it lies beyond the confines of this book.

EUROPE

253

5.4.1 The 16th Century Even during the Middle Ages, Aristotle's authority did by no means remain unchallenged. An all-out critique of Aristotelianism is given in the Universal Treatise by Nicholas of Autrecourt (c. 1300-1350), yet it does not add up to a coherent alternative. Humanism opposed Aristotelianism because it had come to be regarded as the official doctrine of the catholic church. In compensation, Plato and Cicero were rediscovered. This return to original Greek and Roman models served the interests of the middle-class which was entering the political arena. Theology, metaphysics, and logic, which had been the principal concerns of scholastic thinking, came to be replaced by ethics, politics, and rhetoric. At the same time, there was a tendency to return to original, pre-scholastic Christianity embodied by St. Augustine (cf. Vasoli 1988). It was a common complaint against scholasticism that it used 'barbarous Latin'. Scholastic logic was castigated equally because it seemed to have no practical value and because it made use of seemingly ungrammatical examples; the interest in the expositio analysis, for instance, disappeared completely (cf. Asworth 1988:164-165). A large-scale reform of logic was attempted, in the name of some sort of 'Humanist' logic. At face value, this new logic looks like a nearly total failure, when compared with scholastic logic. It should be pointed out, however, that Humanists like Lorenzo Valla (1407-1457), Rudolph Agrícola (1443-1485) or Petrus Ramus (1515-1572) did not try to compete with their scholastic prede­ cessors. They were not interested in validity or in deductive inference. Rather, they were interested in non-deductive inference, for the purpose of attaining new, probable knowledge (cf. Jardine 1988). Therefore it is not surprising that, in terms of reliability and formalization, their results stay far behind those of the scholastics. Another aspect of Humanist logic was the attempt to develop arguments that would convince the audience, which reveals the connection with rhetoric. On this issue, however, there was no absolute break with the preceding tradition because the medieval method of disputatio, exemplified e.g. in Aurifaber's Determinatio and in the response to it (cf. p. 251), had a similar goal. Deductive logic as practised by Humanists was an impoverished version of Aristotelian syllogistics.17 While there was increasing interest in neo-Platonism, Aristotelianism (both in a late-scholastic and in a 'revivalist' version) continued to be the dominant philosophy throughout the 16th century. It is interesting to note, however, that ancient scepticism was also rediscovered. It was given an up-to-date formulation by Fransisco Sanches in his 1576 book Quid Nihil Scitur ('That Nothing is Known'), and was popularized by his cousin Michel de Montaigne in Essais (1580-1588). As

254

UNIVERSAL HISTORY OF LINGUISTICS

a result, "in the post-Renaissance philosophical world the theory of knowledge became a central part of philosophical discussions" (Popkin 1988:684). Humanist grammatical theory, typified by Philipp Melanchton (1497-1560), is subordinated to pedagogical goals. In the definition of the word-classes the emphasis is put on semantic criteria. The neglect of the formal aspect leads to obliterating the distinction between lexical and grammatical meanings (Padley 1976: 38). The attitude towards the preceding tradition is internally inconsistent. On the one hand, a total break is being professed; on the other, what is actually achieved is just a diluted version of scholastic grammar. It is possible that there really was no conscious (but unacknowledged) imitation on the part of Humanist grammarians. Just like their scholastic predecessors, they sought inspiration in Quintilian, Donatus, and Priscian (plus Varro, who had been rediscovered in the meantime). Therefore it was inevitable that their results turned out to be indistinguishable from a segment of scholastic grammarians' results. This is true of even such a polemi­ cally anti-scholastic Humanist as Juan Luis Vives (1492-1540) (cf. Brekle 1985: 112-114). It is generally thought that the most significant post-medieval and pre-1600 grammarians were Thomas Linacre (1460-1524), Julius Caesar Scaliger (14841558), Petrus Ramus (or Pierre de la Ramée), and Francisais Sanctius (or Francisco Sánchez de la Brozas) (1523-1600). On this issue I have seen no reason to deviate from the tradition. It is possible to distinguish between three main divisions in the grammars written during the period under scrutiny: definition of the word-classes; (as a semiautonomous part thereof) theory of the noun, in particular of the case-system, and theory of the verb; syntax. It seems clear enough that syntax is the most interesting part of Linacre's book De Emendata Structura Latini Sermonis (1524) (cf. Padley 1976:53-55, Breva-Claramonte 1983:57-62, Padley 1985:233-239). Linacre postulates two types of constructions, viz. 'proper' (iusta) and 'figurative' (figura­ ta). Proper constructions represent the mental level (sensus animi) whereas figurat­ ive constructions represent the level of actual speech. The most important type of figurative construction is ellipsis. This repeats the age-old idea that mental language is perfect whereas oral language is defective, mainly in the sense of being trunc­ ated. According to Linacre, ellipsis occurs in every area of speech, and he proceeds to illustrate copiously this claim. Here are some representative examples: Impersonal passives like curritur (meaning something like 'It is being run') are expanded so as to match the structure of personal passives: cursus curritur ('The running is being run'). Similarly, impersonal actives like pluit ('It is raining') are supplied with a 'hidden' nominative and thus expanded into the standard

EUROPE

255

'subject & predicate' form, as in Pluvia/Deus pluit ('The rain/God is raining'). Moreover, active intransitive forms like vivo ('I live') are reinterpreted as, or derived from, 'non-elliptical' transitive forms like vivo vitam ('I live a life'). Semantically congruous nouns are added to adjectives, and present participles of esse are added to ablativus-absolutus constructions like Prisciano auctore, resulting in Prisciano enti auctore ('Priscian being the author'). Finally, coordinated sen­ tences in the plural ('A and  were consuls') are reanalyzed as a pair of sentences in the singular ('A was a consul and  was a consul'), except in case of joint coordination ('A and  are similar', 'A and  constitute C ) . — None of these analyses was invented by Linacre, but he has the merit of having brought them together. In his analysis of enallage Linacre is willing to accept that a semantic equivalence may hold between members of different word-classes on the one hand, and between single words and two-word constructions, on the other. This view is not too different from the medieval notion of dictio (cf. p. 235). In sum, "it is Linacre's De Emendala Structura that... first brings back into Renaissance grammar certain tenets of the modistic grammarians", in particular "the late-medieval distinction ... between verbum oris ['word of the mouth'] and verbum mentis ['word of the mind']" (Padley 1985:238). It is Scaliger's De Causis Linguae Latinae (1540), however, which reintro­ duces on a large scale the Fragestellung of medieval grammatical theory, because his is "the first grammar since the Middle Ages to use the metalanguage of Aristotelian philosophy" (Padley 1985:242; for more discussion, cf. Chevalier 1968:176-210, Stéfanini 1976, Padley 1976:58-77, Breva-Claramonte 1983:6372). To be sure, Scaliger purports to go directly back to Aristotle, skipping the medieval commentary tradition. Yet it is quite clear that he is heavily influenced by Modistic grammar (and by Linacre). He upholds the language—mind—reality iso­ morphism in entirely traditional terms: mind is like a mirror of reality (Est enim quasi rerum speculum intellectus noster), and language expresses mind.18 As a consequence, nouns designate what is permanent while verbs designate what is in flux. At the same time, the distinction between lexical meaning (significatio) and grammatical meaning (consignificatio) is carefully maintained: a word like dies ('day') signifies flux, but consignifies permanence, i.e. presents flux as something thing-like. A finite verb is analyzed as copula plus present participle; it is not part of the definition of the verb to express either action or undergoing. Like in Thomas of Erfurt's Grammatica Speculativa, the principal aspects of language are defined in terms of Aristotle's four causes. More interestingly, these are also made use of

256

UNIVERSAL HISTORY OF LINGUISTICS

in the description of the case system. The five cases of Latin (apart from the vocative) are given basic meanings corresponding to the relations that are to be found, either explicitly or implicitly, in every action: In omni actione est id quod agit (nom.), id quod fit (acc.), id quod factum recipit (dat.), privatio (ex + abl.), et finis cuius causa sit (gen.?). Although these relations are constant, they may occasionally be expressed in different ways. For instance, in a sentence like 'He builds a house' the accusative expresses both what is made (= 'formal cause') and out of which it is made (= 'material cause'). Here Scaliger seems to be operating with something approaching semantic roles or Pāninian kārakas. Given that Ramus made his reputation as the arch-enemy of Aristotelianism, it is ironic that, when formulating his grammatical theory, he repeatedly appeals to Aristotle's authority. His theoretical work in this field is Scholae Grammaticae (1557), but he wrote also grammars of Latin and French (cf. Chevalier 1968:247307, Padley 1976:77-96, Breva-Claramonte 1983:72-81, Padley 1985:18-46). Throughout his career Ramus was obsessed with developing a 'method' which would have been an alternative to the prevailing Aristotelianism, but it never became quite clear what his supposed originality consisted in (cf. Ong 1958:171195). In the presentation of his method, he advocated three transitions, viz. from simple to complex, from general to particular, and from clear to obscure.19 The actual description of the data was based on the following three 'laws' (taken from Aristotle's Analytica Posteriora): i) Scientific statements must be necessarily true, ii) Each scientific discipline must have its own homogeneous data-basis, iii) Differ­ ent levels of abstraction must be kept apart: what has been said about a concept need not be repeated in relation to any of its subconcepts. In the grammatical context the second law is of central importance. Ramus uses it to separate the study of lexical meaning from the study of meaning ex­ pressed by systematic formal variation, i.e. grammatical meaning (called adsignificatio, rather than consignificatio). The Modistae had already distinguished between semantic meaningfulness and syntactic correctness, and had assigned the two topics to philosophers and grammarians, respectively (cf. p. 227). Now Ramus makes this point the basis of his grammar-conception. The word-classes must be defined exclusively on the basis of endings of word-forms (vocum ipsarum fines), which means that on the semantic side the grammarian is allowed to consider only such 'adsignifications' as number, gender, degree, case, person, and tense. Accord­ ingly, words are divided into those with number and those without number, and the former are subdivided into those with case and gender (= nouns) and those with persons and tense (= verbs). It is understandable that Ramus has sometimes been regarded as a kind of proto-structuralist.

EUROPE

257

Ramus' paradoxical attitude vis-à-vis Aristotelianism is also evident from his claim that Aristotle was the 'inventor of grammar', because he realized that verbs contain the 'force' (vis) of the sentence whereas nouns are just its material: to speak is different from that which is spoken about (alterum est quod loquimur, alterum de quo loquimur). In his own time, Ramus was unique in his careful demarcation between linguistic form and the two types of meaning. Therefore it is amazing to see Foucault (1970 [1966]:35) assert, taking Ramus to be a typical Renaissance grammarian, that "language is not what it is because it has a meaning; its represen­ tative content ... has no role to play here". To repeat, for a typical 16th-century grammarian meaning, or 'representative content', was much more important than linguistic form. Ramus was just consistent enough to view meaning and form as objects for two different types of study. It may be added that it was Ramus who brought into general use the habit of representing encyclopedic knowledge by means of spatial models, or diagrams (cf. Ong 1958:197-204). Being based on the principle of proceeding from general to particular through repeated dichotomizations, however, such taxonomies are just an application of the Aristotelian genus-species apparatus. Breva-Claramonte (1983) takes great pains to prove Sanctius' originality. Yet it is undeniable that he was heavily influenced by the three grammarians discussed above. The title of his 1587 book Minerva seu de Causis Linguae Latinae is an explicit reference to Scaliger. Indeed, Sanctius endorses the latter's view of the language—mind—reality isomorphism. Although they both are ostensibly con­ cerned with the description of Latin, their view of language as expression of universally valid (because ontologically motivated) concepts entails that they are really exploring properties of a universal grammar (cf. Chevalier 1968:343-357, Lakoff 1976:362-373, Padley 1976:97-110, Clerico 1977, Breva-Claramonte 1983:209-227). In the first chapter of Minerva Sanctius, following Ramus' example, identifies grammatical description with the analysis of linguistic form, and he repeats Ramus' formal definitions of the word-classes. In the sequel, however, the emphasis is placed on semantic considerations. The noun (including the pronoun and the participle) and the verb are the principal word-classes. A third word-class, called particulum, is exemplified by prepositions and adverbs in connection, respectively, with nouns and verbs. The six cases of Latin express as many basic roles (or relations). Following Priscian V, 77 the ablative is taken always to require a preposition; when none is present, it has

258

UNIVERSAL HISTORY OF LINGUISTICS

supposedly been deleted by 'ellipsis'. (Ramus had already postulated analogous prepositions-to-be-deleted in Greek when causes or instruments are expressed by mere genitives or datives.) Sometimes prepositions are also added in a non-standard way to accusatives, or datives are replaced by prepositional constructions, in order to clarify the functions of case endings. Because the universally valid number of cases is six, an ablative formally indistinguishable from the dative is postulated in Greek. All these postulated elements are located at the level of logico-mental structure (ratio) rather than of speech (sermo). This traditional dichotomy plays a central role in Sanctius' syntax. Its most important part is the theory of ellipsis, which, along with most of the examples, has been taken from Linacre. For instance, all analyses given here on p. 255 are reproduced by Sanctius. However, he provides his non-elliptical ratio with univer­ salist underpinnings which are lacking in Linacre's De Structura. To this extent, then, Breva-Claramonte's positive appraisal of Sanctius is justified. It is obvious that Sanctius makes use of his notion of 'ellipsis-expansion' to achieve generaliz­ ations concerning sentence-structure. There is some analogy here to Sïbawaihi's descriptive practice. There are two types of justification for adding an element  to a con­ struction X, or showing that X is elliptical. Either the full construction XY occurs in the text of some venerable author; or the presence of Y is required by some sort of 'grammatical necessity'. It might be viewed as a weakness of Sanctius' overall theory that these two types of ellipsis are treated basically on an equal footing. Unlike Linacre, Sanctius does not treat enallage at all. As Padley (1985: 274-275) has noted, this points to a certain rigidity in Sanctius' system. If there can be no semantic equivalence between members of different word-classes or more generally between different constructions, and if, instead, each construction has one, and only one, fixed non-elliptical counterpart, then there is little reason to characterize the non-elliptical level as 'logical'. To be sure, it still qualifies as 'systematic', in the sense that non-elliptical structures are more regular than elliptical ones (see, again, the examples given on p. 255). Although Cratylus was rediscovered in the beginning of the 16th century, Scaliger and Ramus side with Aristotle in taking the relation of words and things to be arbitrary. Sanctius holds a compromise view, claiming that in the original language there was a natural affinity between words and things which has dis­ appeared from contemporary language, resulting in an arbitrary relation. With the growing importance of neo-Platonism the phusei view, understood as a material similarity between words and things, came to be widely accepted in the 17th and

EUROPE

259

18th centuries. So we see that Foucault (1970 [1966]) has mixed up his facts once again when he asserts that in the history of linguistics the 16th century is the age of natural similarity between language and reality whereas the 17th and 18th centuries are the age of arbitrary representation.

5.4.2 The 17th Century Much of the criticism directed against Aristotle in the 16th and 17th centuries was unsubstantiated because those who wished to refute him were not really able to do so. The criticism was the most justified in the new astronomy and mechanics: the progress made in these fields has been taken to warrant calling the period in question the 'Age of Scientific Revolution'. The most important figure of this revolution was Galileo Galilei (1564-1642), while Francis Bacon (1561-1626) was its most effective propagandist. In Book I of his Novum Organum (so called because the collection of Aristotle's logical and methodological treatises was known as Organon) Bacon defines at length the opposition between the old, Aristotelian science and the new science. Contrary to the Aristotelian and the main-stream scholastic view, he maintains that "the human understanding is like a false mirror" of the reality (XLI). To make matters worse, "words [which] are imposed according to the apprehension of the vulgar ... force and overrule the understanding" (XLIII). There is thus a fundamental opposition between ordinary perceptual knowledge, couched in ordinary language, and the true 'facts of nature'. The latter can only be revealed "by a course of experiment orderly conducted and well built up" (LXXXII). This distrust of ordinary perception and ordinary language remains a distinctive feature of the 17th century thought. It was a common complaint against traditional philosophy that, rather than looking directly at reality, it used analysis of language as a means of gaining knowledge about reality. To this it could be retorted, of course, that already Plato had warned expressis verbis against substituting analysis of words for analysis of things (cf. here p. 170). But this would be a quibble, because what was the real target of criticism was acceptance of authority and lack of experimental method. Within the human sciences, the champion of Galileo-type research is Thomas Hobbes (1588-1679). In his Leviathan, he bases his science of society on the axiom that "Life it selfe is but Motion" (1968 [1651]: 130); and he proceeds to derive the existing state of society from the two basic and opposite motions of 'Appetite' and 'Aversion', i.e. seeking of power and fear of death (cf. CLT:

260

UNIVERSAL HISTORY OF LINGUISTICS

298-300). Although more impressed by the chemist Boyle, John Locke (16321704) too in his Essay concerning Human Understanding endorses a 'dynamic' view of human beings, against the background of a machine-like physical reality (1975 [1700], Book IV, Chap. VI, 11). He distinguishes between two basic human 'actions', i.e. 'Thinking' and 'Motion', of which the latter is caused by the act of willing, itself caused by some 'Uneasiness' or 'Desire' (Book II, Chap. XXI, 1-32). In his philosophy of language Locke follows and develops William Ockham's nominalism. He repeatedly denies that we can ever know the 'real essences' of things. All we can know are 'nominal essences', i.e. the way we have come to name things. There are no distinctions (and no 'natural kinds') in extralinguistic reality; the distinctions are merely those of our own language. Just like Ockham, however, Locke is unable to carry out the nominalist program. He first notes that the names are not given in an entirely arbitrary fashion, but "according to the complex Ideas in us" (Book III, Chap. VI, 8). He then goes on to add that the naming process, based as it is on complex ideas, is ultimately grounded on the fact that "Nature, in the constant production of particular Beings, makes them not always new and various, but very much alike ... in many sensible Qualities, and probably too, in their internal frame and Constitution" (36-37). But this just means that linguistic and mental distinctions are, after all, underlain by extralinguistic and extramental ones; so nominalism collapses. It may be added that Locke is more consistent than Ockham in his view that our ideas (which are involved in the language-like operations of mind) are not universal, but "various and different in several Men" (26). Locke follows Aristotle in claiming that words are signs of 'ideas' (Book III, Chap. II, 2-3). This could charitably be taken as an infelicitous way of saying what Ockham had said, namely that words signify things via ideas. René Descartes (1596-1650) was a pioneer in the mathematization of physics. He also considered the human body as an intricate machine which, qua 'extended and unthinking being', is entirely separate from the human mind qua 'unextended and thinking being'. This 'mind—body dualism' is so extreme that the mind is claimed capable to exist without the body (1974 [1641]: 165). Compared with Aristotle's integrated mind—body conception (cf. here p. 182), this is a big step backwards. To see this, all one has to do is try to think carefully what it would be really like to be a mind which, lacking a body, has never had any sense-im­ pressions. Besides, the religious motivation for entertaining the possibility of disembodied and non-dying minds is rather obvious.

EUROPE

261

The popular wisdom has celebrated the 'Cartesian doubt' as a method for placing scientific knowledge on a secure foundation, free of any extraneous authority. This is a complete misunderstanding. Having reached the stage where only the knowledge of his own self-consciousness is indubitable (and having overlooked the certainty with which he knows the meanings of the words employed to express this insight; cf. G7M:144), Descartes faces the task of generalizing this newly-won certainty to the knowledge about the external world (including his own body). To achieve this, he finds no other way than to insist that "God is no deceiver" (1974 [1641]: 144, 158, 167): God would not have led him into thinking that the external world exists, if it did not exist in fact. This is of course a worth­ less stratagem, which means that Descartes remains enclosed within his own solipsistic self-consciousness (cf. CLT: 225-227). It would be interesting to learn in detail how and why this declaration of religious faith has come to be regarded as ushering in the age of modern, critical thought. In the 17th century there was some controversy on whether 'innate ideas' exist, with Descartes supporting this claim and Locke denying it. I shall disregard the issue here, for three reasons. First, it had no impact on linguistic theory. Second, Descartes and Locke tried to answer by philosophical analysis a question which can only be answered by empirical research. Third, and as a corollary of the second point, 'innateness' was — incongruously — identified with self-evidence, or with what was understood 'clearly and distinctly'. For Descartes, God was the supreme innate idea because he personally was able to understand it 'clearly and distinctly' (1974 [1641]: 140-143). Locke was intelligent enough to notice that the self-consciousness of a single man is not a sufficient guarantee for something to qualify as innate. Rather, the whole mankind must be taken into consideration; and there seem to be persons, and even nations, for whom the idea of God is far from 'clear and distinct' (Book I, Chap. 4, 8-10). Within linguistics, the tradition of universal grammar was continued by Tommaso Campanella (1568-1639) and Juan Caramuel y Lobkowitz (1606-82), both of whom deliberately tried to revive the Modistic doctrine (cf. Padley 1976: 160-184). While their discussion is based on Latin, Benedetto Buonmattei's Delia Lingua Toscana (1642) applies a descriptive method with universalist aspirations to a vernacular language. The most famous 17th-century work within this tradition is Grammaire générale et raisonnée (1660), commonly known as 'Port-Royal gram­ mar', written by Antoine Arnauld and Claude Lancelot. If I am going to discuss this book at some length, this is not because of its intrinsic merits. Inflated claims about the importance of the Port-Royal grammar have been made, though for quite

262

UNIVERSAL HISTORY OF LINGUISTICS

different reasons, by Foucault (1966), Chomsky (1966), and Chevalier (1968). If such claims are taken seriously, the history of Western linguistics will be com­ pletely misunderstood. The Port-Royal grammar represents a confluence of grammar and logic. On the grammatical side it must be supplemented by Nouvelle Méthode pour apprendre la langue latine (3rd ed. 1654) written by Lancelot. On the logical side it must be supplemented by La Logique ou l'Art de Penser (1662) written by Arnauld and Pierre Nicole. Nevertheless, the main emphasis will be placed here on the PortRoyal grammar. The general framework is presented in Chapter I of Part Two. There are three basic operations of mind: conceiving, judging, and reasoning. What we conceive, is also called 'idea' or 'object of our thought' (l'objet de notre pensée). Judging consists in combining two objects of thought to make an affirmation. A judgement is called 'proposition'; what is affirmed is called 'attribute' while that about which it is affirmed is called 'subject'; the subject is combined with the attribute by means of the copula. Often the copula is merely implicit, which calls for a reanalysis: A sentence like Pierre vit ('P. lives') must be reanalyzed as Pierre est vivant ('P. is living') (Chap. XIII). The combination expressed by the copula (whether explicit or implicit) is identified as a 'manner of our thought' (la manière de notre pensée). Words signify thoughts; more precisely, the word-classes other than the verb and the conjunction signify objects of thoughts while these two signify manners of thoughts. (In an alternative formulation, it is men who signify their thoughts by means of words.) Thoughts precede, and are 'made known by', words. There are two basic ontological categories: the thing (chose) or 'substance', signified by the noun, and the manner of things (la manière des choses) or 'acci­ dent', signified by the adjective (Chap. II). Because adjectives function as attri­ butes, and because a finite verb is reanalyzed as 'copula plus attribute' (Chap. XIII), it would seem to signify not just the act of affirmation, but also an accident. The technical expression 'what we conceive' (ce que nous concevons) entails an ambiguity which has often been overlooked: it may mean either 'conception of a thing' (i.e. 'idea') or 'thing as conceived'. The term 'object of thought' contains the same ambiguity between ideas and things. Thus it is not true that, on the PortRoyal view, ideas are the only referents of words. Of course, this ambiguity should not have been allowed to arise in the first place. The preceding account can be summed up as follows:

EUROPE

263

The framework of the Port-Royal grammar is thoroughly traditional. The three 'operations of mind' are identical with those three mental operations which, accepted by every medieval logician and grammarian, go back to Aristotle (cf. p. 224, 238). The notion of sentence, with the (explicit or implicit) copula expressing the combination of the two terms, is taken from Aristotle (cf. p. 179). So was the notion of accidents being dependent on substances (cf. p. 238). A careful comparison of Aristotle's account with Port-Royal's reveals that the latter contains nothing that is not already contained in the former. The reverse is not true, however, because the distinction between signification and consignification has been lost. The authenticity of this loss is not attenuated by the fact, pointed out by Padley (1976:245) and Dominicy (1984:177, n. 12), that in one passage the Port-Royal grammar uses the term manière de signifier in a way reminiscent of the Modistic modus significandi. Moreover, there is a gratuitous confusion between the Port-Royal definitions of sentence and verb: on the one hand, "to judge is to affirm" ('proposition' being another name for judgment), and "the judgment is the principal manner of our thought". On the other hand, "the verb signifies the manner of our thought", and "it is the principal function of the verb to signify the affirmation". It is therefore true both of the sentence and of the verb that it is (or signifies) both the affirmation and the manner of our thought. Finally, there is a minor difference vis-à-vis Aristotle: while the latter considers affirmation and negation as equal subtypes of 'assertion' (cf. p. 177), Arnauld and Lancelot hold that affirmation, being the primary act, is secondarily converted into negation by means of negative words. Given this traditional background, it is nothing short of staggering to read the following passage: Ainsi, Aristote s'étant arrêté à la troisième des significations ajouté à celle qui est essentielle au verbe [i.e. celle de  affirmation], l' défini vox

264

UNIVERSAL HISTORY OF LINGUISTICS

significans cum tempore, un mot qui signifie avec temps (Arnauld & Lancelot 1968 [1660]:111). Ignoring two thousand years of grammatical tradition, the authors of the Port-Royal grammar calmly assert that they have invented the centrality of the copula, its capacity to combine the two terms of the sentence, its force of affir­ mation, the reanalysis of a finite verb as 'copula plus participle', etc. This can only be a sign of great ignorance or of great dishonesty. For my part, I cannot believe that Arnauld and Lancelot could have been ignorant of the whole preceding tradition. (For instance, Thomas of Erfurt's Grammatica Speculativa had been published in Paris in 1605). One mistake breeds another. Foucault (1966: Chap. IV, 3) seizes upon the Port-Royal grammarians' new and revolutionary analysis of the verb and makes it the basis for his grandiose claim that language and being, in short the entire Western episteme, changed abruptly and totally in the Classical Age: We must not stop, as Aristotle did, at the fact that the verb signifies [sic!] tenses (there are many other words, adverbs, adjectives, nouns, that can carry temporal significations). Nor must we stop, as Scaliger did, at the fact that it expresses actions and passions, whereas nouns denote things — and permanent things (for there is precisely the very noun 'action' to be considered). ... What we must do before all else is to reveal, in all clarity, the essential function of the verb: the verb affirms,... The entire species of the verb may be reduced to the single verb that signifies to be. ... In the Classical period, language in its raw state ... vanished from sight, but language itself entered into new relations with being, ones more difficult to grasp... (Foucault 1974 [1966]:93-94; the first emphasis added). At this point I must interrupt Foucault's leaps of phantasy, just to note that nothing of what he says justifies the view of the Port-Royal grammar as in­ augurating a new era in the study of language. His view of Aristotle is completely wrong, and he does not know Scaliger any better than he does Aristotle (cf. here p. 255). Besides, the advocates of the Port-Royal grammar may be surprised to hear that, as Foucault (1974 [1966]:83) sees it, "it would be ... nonsensical to attempt to interpret [the 'general grammar' of the Classical Age] as a sort of prefiguration of a linguistics". The Port-Royal grammar has been claimed to be the first to recognize the primacy of sentences vis-à-vis words. This is absolutely false. For Aristotle, the words which make up a sentence are subordinated to it, and his notion of sentence

EUROPE

265

(qua object of the second 'operation of the intellect') was the one accepted through­ out the Middle Ages (cf. p. 224). The primacy of the sentence was declared openly, e.g. by Albertus Magnus (cf. p. 238). It is true that the Modistae tend to discuss syntax by means of discussing word-classes. But the Port-Royal grammarians operate in just the same way: The second part of their grammar contains 25 chapters, and apart from the first, introductory chapter, the chapters 2-24 deal with word-classes, while only the last chapter, comprehending the total of five pages, deals with syntax. Yet Arnauld and Lancelot do manage, in a truly medieval fashion, to raise syntactic questions also in chapters ostensibly devoted to wordclasses. Sometimes it has been assumed that the revolutionary aspect of the PortRoyal grammar consists in its virtual identification of mental and linguistic pro­ cesses, shown by the fact that judgments are just called 'propositions'. This 'revolution', however, goes back to Plato; I just repeat a quotation from Theaetetus which says exactly that which is said by Arnauld and Lancelot: "Judgment [is] a statement pronounced, not aloud to someone else, but silently to oneself' (cf. here p. 172). Again, it may be suggested that what is new and revolutionary in the PortRoyal grammar, is the proximity, and even coincidence, of grammar and logic. But the same is true of Aristotle already (cf. here p. 179). Moreover, the Port-Royal logic is famous for having conjoined logic and psychology. Exactly the same thing, however, was done by Ockham, who identified mental language with logically perfect language (cf. here p. 245). It is true that, in addition to traditional syllogistics, the Port-Royal logic contains an attempt to develop 'une logique des idées', or an intensional logic interpreted psychologically, which is something of a novelty in the history of Western logic. Kneale and Kneale (1962:315-320), though somewhat skeptical, admit that the Port-Royal logic "nevertheless deserves its high reputation". Finally, it has been claimed that the revolutionary aspect of the Port-Royal grammar consists in its notion of linguistic sign, which anticipates de Saussure's corresponding ('binary') notion. The relevant definition is given in the Port-Royal logic (rather than grammar): "le signe enferme deux idées, l'une de la chose qui représente, l'autre de la chose représentée" (Book I, Chap. IV). It is self-evident, however, that this notion (which should be characterized 'quaternary', rather than 'binary') is identical with the one given by Pseudo-Albertus Magnus and Pierre d'Ailly (cf. here p. 231). Up to now I have examined the general framework of the Port-Royal grammar, and I have shown that, apart from possible influences emanating from the

266

UNIVERSAL HISTORY OF LINGUISTICS

attempt to construct an intensional logic, it contains nothing that would not have been said before. Next I shall proceed to examine in more detail the grammatical descriptions given by Arnauld and Lancelot. Let us see whether we find here any justification for considering the Port-Royal grammar as a landmark in the history of Western linguistics. Just like in Apollonius, a noun may express either a particular idea (Socrate) or a general idea (homme) (cf. p. 213). Just like in Anselm, an adjective like white signifies directly (though confusedly) that which is white, and indirectly (though distinctly) the quality of being white (cf. p. 231). Only Latin and Greek are claimed to possess a case system, which is a curious remark made by a supposedly 'general' grammar. Yet, since pronouns in many languages do have cases, these must be dealt with. The accusative is used in connection with verbs signifying an action which 'passes' from the agent to the patient; in French this is expressed not by a case ending but by the 'natural order' 'agent—action—patient', noted already by Apollonius (cf. p. 213). Like in Priscian (V, 78-79), the ablative is taken to be a prepositional case without any signification of its own. On the analogy of Latin, Greek is assumed to possess an ablative indistinguishable from the dative. The French indefinite and definite articles are taken to exemplify a Latin-type case system: nominative = U crime si horrible mérite la mort; accusative: // a commis un crime; ablative: Il est puni pour un crime; dative: // a eu recours à m crime; genitive: Il est coupable d'un crime. The notion of 'manner of thought' is clarified further in Ch. XVI and XXIII in connection with verbal modes and conjunctions. First, not just a simple affir­ mation, but also 'conditional' and 'modified' affirmations, occurring in sub­ ordinated sentences and expressed by the subjunctive, qualify as 'manners of thought'. Secondly, not just affirmations but also such mental movements as wish, permission, and will, variously expressed in different languages, qualify as 'man­ ners of thought'. Third, the same is true of the acts of asking, negating and joining sentences together. The common denominator in all these cases is said to be the fact that in the external world there is no (ontological) entity that would correspond to the linguistic expression. This is the traditional justification for distinguishing syncategorematic from categorematic terms (cf. p. 245), even if in other respects this distinction does not fall together with that between (expressions of) objects vs. manners of thought. Finally, the syntax properly so-called, discussed in Ch. XXIV, is divided into agreement (construction de convenance) and governance (construction de régime), which are claimed to be universal and language-specific respectively. This view of 'syntax' echoes Thomas of Erfurt's (cf. p. 242).

EUROPE

267

In all this there is nothing novel. The authors seem to realize this to some extent, because they themselves claim that, apart from their analysis of the verb, they have made only two original discoveries, one concerning relative clauses (Ch. IX) and the other concerning infinitives (Ch. XVII). We have seen that the claim about the analysis of the verb is false. Let us see whether the two other claims fare any better. What is at issue, is the fact that one sentence can contain several (implicit) sentences. For instance, the (complex) sentence Dieu invisible a créé le monde visible contains the following judgments/sentences: Dieu est invisible, Dieu a créé le monde, and Le monde est visible. The same judgments/sentences are contained, more obviously, in the (complex) sentence Dieu, qui est invisible, a créé le monde, qui est visible. It follows that these two 'complex sentences' (defined in Logique II, 5) are equivalent not just to each other, but also to the ('composite') sentence Dieu est invisible, et il a crée le monde, et celui-ci est visible; in other words, the relative pronoun may be equivalent to the combination of and plus a personal/ demonstrative pronoun (as noted in the beginning of the second part of Ch. IX in Grammaire). Moreover, not just an adjective but also a (present) participle com­ bined with a noun is equivalent to a relative-clause construction, as in canis currens ('a running dog') = canis qui currit ('a dog which is running'). Now, Arnauld and Lancelot claim to be the first to have discovered that a sentence containing the relative pronoun may be part of the subject or of the attribute of another sentence {la proposition dans laquelle il entre ... peut faire partie du sujet ou de l'attribut d'une autre proposition). It is impossible to see what is supposed to be new about this approach. The idea that sentences may contain other (implicit) sentences is as old as Western syntax. Already Apollonius noted explicitly the equivalences between relative-clause constructions, co-ordinated sentences, and participle constructions; in particular, he carefully shows how a relative pronoun joins together two sentences with two finite verbs (cf. p. 208). Ever since, it has been a commonplace that a noun plus an adjective/participle equals a sentence minus its assertoric force (cf. p. 224 and 242). It may be added that in this respect the medieval tradition has been more inventive than the Port-Royal grammar: Siger de Cortrai (p. 13), for instance, takes also 'N + N-gen' constructions (e.g. the cap of Socrates) to be equivalent to sentences (like Socrates possesses a cap). The third and last claim of originality is closely connected to the previous one: Arnauld and Lancelot think they have discovered that just like the relative pronoun, the infinitive too can join two sentences together. For instance, Scio malum esse fugiendum ('I know that evil is to be shunned', literally 'I know evil to

268

UNIVERSAL HISTORY OF LINGUISTICS

be to-be-shunned') contains two sentences of which only the first is affirmed. Again, Arnauld and Lancelot exhibit an appalling ignorance of the Western grammatical tradition. It has always been known that (accusative-cum-)infinitive constructions are semantically equivalent to sentences (cf. p. 187, 210, 224, 242). Notice, incidentally, that it is rather odd to say that in a construction A +  (= I wish + to walk) it is  which joins A and  together. It is much more natural to say that the joining-together, performed by the speaker, is made possible by the nature of A, i.e. the finite verb which either permits or requires the (accusativecum-)infinitive construction. It is good to add that Logique II, 6-7 contains a more sophisticated treat­ ment than Grammaire, Ch. IX, especially because a distinction is made there between explicative and restrictive relative clauses. It is said that in a sentence like The people who are pious are charitable the mind joins together the ideas of 'people' and 'pious' without, however, actually affirming that there are any pious people; rather, the affirmation in question is merely 'virtual'. In whatever way this should be understood in detail, it is not true, contrary to Pariente (1985:84, 86), that the distinction between the two types of relative clauses was discovered by Arnauld and Nicole (cf. here p. 249). To what extent is Chomsky's (1966) view of the Port-Royal grammar as a forerunner of transformational theory justified? In the Port-Royal grammar there is no reference to innate ideas (cf. Dominicy 1984:13) or to creative use of language (cf. Pariente 1985:45) or to recursivity. If there is a 'depth vs. surface' distinction (a fact much more in evidence in Lancelot's Nouvelle Méthode; cf. Lakoff 1976), it goes back to Peri suntaxeōs. The notion of universal grammar goes back to medieval linguistics (cf. 5.3.3.). Padley (1985:283-324) offers an overview of the controversy surrounding Chomsky (1966). To what extent is Chevalier's (1968) view of the revolutionary character of the Port-Royal grammar justified? This question is more difficult to answer because the reasons for this view never become clear, in the first place. Formulations like the following are of no help: "[It is the principal merit of the Port-Royal grammar] d'avoir défini le rôle de la proposition, définition inconnue dans la grammaire jusqu'alors, et par là, d'avoir fondé une syntaxe autonome" (p. 490). Yet Chevalier himself insists that linguistic form was subordinated to psychologico-logical considerations; so in what sense was the syntax autonomous? Besides, the notion of sentence had been defined in identical terms since Aristotle. Or consider this: "Les formes peuvent être entièrement muables, les schémes internes ne le sont pas. C'est là la révolution" (p. 499). But this 'revolution' was inaugurated by Aristotle and constituted the principal claim of Modistic grammar. There are many more such

EUROPE

269

remarks but they all suffer from the same weakness, namely ignorance of ancient and medieval linguistics. It may be added that Joly (1977) is one of the very few historians of linguistics writing in French whose attitude towards the Port-Royal grammar is not one of uncritical admiration. It is clear that the most original contribution made within the Port-Royal research program obtains in the field of the 'psychologization of logic'. Therefore it is understandable that both Dominicy (1984) and Pariente (1985) devote the greatest amount of space to reconstructing this logic. The basic relation is an idea 'containing' (enfermant) another. For instance, the idea of 'green apple' contains the idea of 'apple' (whereas the class of apples contains the class of green apples). Although 'idea' and 'class' behave inversely, there is potential ambiguity in that a (complex) idea is considered as a class of (simple) ideas. A sentence like The earth is round is understood as affirming that the idea of the earth contains the idea of roundness. Perhaps significantly, Dominicy (1984) and Pariente (1985) interpret even this simple relation differently, the former in terms of inclusion (p. 55, 164) and the latter in terms of identity (p. 67, 268). Both display great ingenuity in attempting to translate Aristotelian logic into a logic of ideas. It may be asked, however, where is the line between the actual contribution of the Port-Royal logic and its modern interpretation. For instance, Dominicy's claim (pp. 41-44, 85-87) that Arnauld and Nicole were really doing possible-worlds semantics à la Montague seems to be a clear instance of the latter. This interpretation also has the drawback that, contrary to the received view, it requires depsychologizing the Port-Royal logic. The study of universal grammar went hand in hand with the planning of a universal language. The motivation for both endeavours was the same, namely distrust of ordinary language, forcefully expressed not just by Bacon, Hobbes, and Locke, but also by Descartes and his school. In some cases it is not clear what the status of the ideal, logically perfect language was supposed to be, i.e. whether it was taken to be a real psychological phenomenon, lying hidden beneath the irregularities of existing languages, or whether there was a distinct recognition that it had to be constructed expressly, to get rid of irregularities embodied equally by the existing languages and by whatever psychological mechanism they have in common. Therefore it is understandable that sometimes an artificial universal language was identified with the universal grammar, or rather with the one language which, being common to all existing languages, would be such as to be adequately described by the universal grammar. These two lines of thinking, and their partial confluence, constitute the main topic of Padley (1976), Salmon (1979), and Padley (1985).

270

UNIVERSAL HISTORY OF LINGUISTICS

The leading figure in the universal-language movement was John Wilkins (1614-1672). It was his purpose to construct an artificial language which would describe the whole body of existing knowledge and would, moreover, do so by means of a "vocabulary which would show its meaning by its form" (Salmon 1979:132). Primitive expressions would correspond to primitive notions (or 'things'), and the increasing complexity of notions would be reflected as the corresponding increase in the complexity of expression (cf. Aarsleff 1982:260262). This general program had several immediate predecessors, but ultimately it goes back to Cratylus (cf. p. 172). Wilkins was a member of the Royal Society, and he professed an allegiance to Bacon's new conception of science. Therefore it is somewhat surprising that he categorized the world along thoroughly Aristotelian lines. In this respect his philosophical vocabulary does not differ from Ramus' taxonomies (cf. p. 257). Moreover the 'philosophical grammar' which he had to construct as a supplement to his universal language means a deliberate return to Modistic grammar (cf. Padley 1976:198-208, Salmon 1979:102-123). With hindsight it is easy to see that Wilkins attempted the impossible; his project was soon forgotten. Still, the idea of a universal language lingered on. Scholars like Wilkins countenanced the notion that there should be a transparent, one-to-one relation of representation between words and things. This is the basis for Foucault's (1966) view about what Western episteme was like in the Classical Age. It is incomprehensible, however, why he thinks he can let the Wilkins-type thinking represent the Western Classical thought in its entirety. First, the model to be imitated in the human sciences was the Galilean and Newtonian description of motion, rather than the static Aristotelian taxonomy as implemented by Wilkins (cf. p. 259). Second, in assuming an immediate and unproblematical access to things, Wilkins was far from representing some sort of consensus in the 17th-century thought. In spite of their considerable differences on other issues, Descartes and Locke agreed that knowledge of one's own sensations and thoughts is much more secure than knowledge of the external world. Indeed, it is Rorty's (1980) claim that Classical thought is characterized by its lack of immediate and uproblematical access to things (cf. p. 189). In other words, Foucault (1966) and Rorty (1980) make exactly opposite claims. Of course, they are both wrong, because in their urge to simplify, they remain blind to the real complexity of intellectual history. One can admit this complexity, without being overwhelmed by historical details.

EUROPE

271

5.4.3 The 18th Century G.W. Leibniz (1646-1716) is in several respects the central figure in the late 17thcentury and early 18th-century thought. Towards language he had an attitude which could be characterized as 'many-layered'. First, taking logic to be a scientia generalis, he is the first to formalize it, but unlike Aristotle in his (only semiformalized) syllogistics, he prefers an intensionalist interpretation (cf. Kauppi 1960). Second, he is also interested in a Wilkins-type universal language, or a characteristica universalis, in which "die Zeichen ... die Struktur der Ideen [oder Begriffe] genau abbilden... Da die Begriffe die Dinge abbilden, besteht ein Iso­ morphismus zwischen Dingen, Begriffen und Zeichen" (Heinekamp 1976:530531). Yet understandably enough, this grandiose plan was not carried out beyond a few programmatic statements. Third, at a more practicable level, Leibniz takes an existing language, i.e. Latin, and tries to purify it of its irregularities so as to transform it into a lingua philosophica, or a language which would be adequately described by a grammatica rationalis/universalis. By eliminating the semantically irrelevant distinctions like those between the grammatical genders, declensions, and conjugations as well as those between nouns and adjectives etc., all sentences will be reduced to the canonical type 'A is B' (cf. Heinekamp 1976:533-535, Lenders 1976:587-589). The similarities with Ockham's lingua mentalis are self-evident (cf. p. 245). Fourth, it is possible to piece together a general theory of signs out of Leibniz's fragmentary remarks. This is the task that Dascal (1978) has set himself. To a large extent, notions that were needed in connection with the three preceding points recur in this semiotic context. Later, semiotics became something of a fashion among the Enlightenment philosophers (cf. Auroux 1979). Fifth, Leibniz concerned himself also more directly with the analysis of language. To be sure, those of his remarks that may have a bearing on synchronic grammatical theory are too general to permit any exact assessment, or any compari­ son with grammar-conceptions of the 20th century (cf. Dascal 1987:125-144). Within linguistics, most of his energies were devoted to etymological research and to what might be called a prefiguration of comparative linguistics. In this respect he illustrates well the growing interest in history, which was characteristic of the 18th century. Leibniz was, of course, Newton's rival as an inventor of differential calculus qua description of motion. David Hume (1711-1776) turns out to have been a parallel figure in an interesting way. On the one hand, he deliberately

272

UNIVERSAL HISTORY OF LINGUISTICS

imitated Newton in this attempt to construct a 'science of human nature', the laws of association taking the place of the laws of motion (Hume 1962 [1740]: 353-353). In its general design his research program owes much to Hobbes (cf. CLT:298301). On the other hand, around 1750 Hume turned away from philosophy (and speculative psychology) and devoted his energies to writing a history of England. Contemporaneously with Hume, there existed a genuine 'School of Scottish Historians', with Adam Smith (1723-1790) as its leader. Anticipating the basic insights of Marxist historiography, it constructed a purportedly universal theory about how social and ideological structures of 'civil societies' evolve, via welldefined stages, in response to changes in economic relations (cf. Skinner 1979: 29-32). Already earlier Giambattista Vico (1668-1744) had reshaped the philosophy of history, by identifying verum ('what is [known to be] true') and factum ('what is done'), and thus postulating a qualitative difference between people's 'internal' knowledge of human actions and their 'external' knowledge of physical events. Vico's insight suffers, however, from the fact that he fails to make the additional distinction between normative and non-normative actions, of which the latter are known much less securely than the former (cf. GTM: 195-196). Furthermore, even if actions are made by men and can therefore be known by them in a way that physical events cannot be, the same is not true of the 'laws of history' to which, according to Vico, human actions conform. Therefore it is not clear how Vico can, on his own terms, claim to know these laws as well (cf. Berlin 1976:112-114). The notion of history was also central to the French Enlightenment. More precisely, in this context the history of mankind was considered from the perspec­ tive of progress, as a gradual ascent towards more and more rational stages of culture. This normative attitude has been much criticized, but it should be remem­ bered that Montesquieu, for instance, far from countenancing a schematic, univer­ sally valid conception of history, paid attention to special developments occasioned by differences in geography and climate. It is just simple-minded, and wrong, to think that a normative attitude automatically excluded any genuine 'sense of historicity'. E.B. de Condillac (1715-1780) reformulated the controversy between Descartes and Locke in terms of hypothetical or fictional history. What he intended to prove, more precisely, is that abstract knowledge (as it exists today) can develop without any help of innate ideas, and even of Locke's 'reflection', out of simple sensations (as they presumably existed at the beginning of human history). What makes this account interesting from the present point of view is the role played by language: thought and language are assumed to develop in ('dialectical') interaction, one making it possible for the other to reach the next highest level on the scale of

EUROPE

273

intellectual evolution. In this way the question of the origin of language moved to the center of attention, and the nature of universal grammar, which had earlier been discussed in terms of atemporal structure, was now discussed on the scale of time (cf. Aarsleff 1982:166). For instance, when Adam Smith gave his account of how the word-classes had presumably developed, he at the same time characterized what he took to be the real or universal nature of each word-class (cf. Smith 1983 [1761]:203-226). To be sure, it was by no means a novel idea to explain the nature of linguistic units by showing how, and why, they must presumably have been created. Apollonius makes occasionally use of this descriptive device, for instance when he tries to explain, in II, 44-45, why the first and second person pronouns were 'introduced'. The Port-Royal grammar uses fictional history more frequently, among other things to give its own explanation, in Ch. VIII, of the problem treated by Apollonius. Condillac's views had a strong impact on the debate about the natural order of thoughts and words. Ever since Apollonius, SVO had been taken to be the natural word order because it reflects the temporal order 'agent—action—patient' exhibited by the state of affairs spoken about (cf. p. 213). The authors of the PortRoyal grammar took the word order to reflect the order of thoughts, rather than the order of entities in the external world: first, thinking pre-exists language and is expressed by it; second, thinking has a fixed, 'natural' order, and word-order in turn is 'natural' to the extent that it reflects the order of thoughts. To be sure, Lancelot and Arnauld seem to have held different views as to what, precisely, the 'natural' word order consists in: the former emphasizes the illegitimacy of discon­ tinuous constituents while the latter condemns the constructions where the subject (qua topic, not qua agent-word) follows the predicate (cf. Ogle 1980). All this changes with Condillac. As he sees it, thought does not pre-exist language, but is rather created through language. More precisely, prelinguistic thinking is simultaneous and resembles a painting, whereas articulate thought is identical with verbalized thought, and therefore shares the linear character of language. This is why Condillac labels languages as méthodes analytiques. Because there is no pre-existent logical order, there is no universally valid 'natural' order of words either (cf. Ricken 1978:96-97). It may be added that, according to Scaglione (1980), Buffier was the first to formulate, in 1732, the opposition between simul­ taneity of thought and linearity of speech.20 Diderot accepts Condillac's con­ clusions, but in his Lettre sur les sourds et muets (1751) he adds the interesting remark that it is only the sign languages of the deaf which make an investigation of prelinguistic thinking possible. It is hard to understand why it took over two hundred years until this line of inquiry was seriously taken up (cf. Furth 1966).

274

UNIVERSAL HISTORY OF LINGUISTICS

The most important grammarians of the time were C.C. Du Marsais (1676 -1756) and Nicolas Beauzée (1717-1789). As far as the relation between language and thought was concerned, they adopted a rather conservative position. On the one hand, Du Marsais claims, fully in the spirit of the Port-Royal grammar, that thought pre-exists, and is independent of, language; on the other, he concedes that if thought is to be conveyed, it mustfirstbe 'decomposed' by language (cf. Ricken 1978:134-135). Beauzée (1974 [1767]: II, 466-469) reaffirms, essentially, the same position. His account contains, however, a couple of confusions which it is good to clear up. In conformity with the mainstream philosophy of the 18th century, Beauzée asserts that language speaks of thought (la pensée), not of extramental entities. In its primary state thought is 'indivisible' (which seems to be a concession to Condillac). It is then analyzed by logic and put in what Beauzée calls ordre analytique, and it is this (natural) order which particular languages must, in principle, reproduce in their word order. (This accords with Beauzée's view, expressed in his Preface, xxiii, that universal grammar is an exposition of a logique naturelle.) The 'analytical order' is universally valid and innate. A contradiction arises, however, when Beauzée wishes to justify the (innate) word order by appealing to the traditional arguments: the object follows the verb because it takes time before an action reaches its patient, and the adjective follows the noun because prius est esse quam sic esse ('being precedes being so-and-so'). First, these 'justifications' refer to extramental entities, not thoughts. Second, this entails, absurdly, that the extramental and extralinguistic temporal order 'agentaction—patient' comes to be reflected in the linguistic temporal order 'subjectverb—object' via 'indivisible' thought with no order. Third, and most important, if something is innate, it must not be justified, e.g. by deriving it from an interaction between perception and reality. This justification for their universal grammar was open to the Modistae precisely because of their non-innatist stance. It is probable that Du Marsais made a more original contribution to the development of grammatical theory than Beauzée. Yet I can concentrate on the latter's Grammaire générale (1767), precisely because he is so heavily dependent on his predecessor. 'Agreement' and 'governance' had traditionally been among the central concepts of syntactic theory, but now they are made, under the names of concord­ ance and régime, the very basis of sentence analysis. Their 'logical' (or conceptual) counterparts are the relations of 'identity' and 'determination'. As Bartlett (1974: 25-28) explains it, there is identity, manifested as agreement, on the one hand between the noun and its adjective, as in la loi naturelle ('the natural law'), and on

EUROPE

275

the other, between the subject and its verb, as in le soleil luit ('the sun shines'), derived — in the old Aristotelian tradition — from le soleil est luisant ('the sun is shining'). There is determination, manifested as governance, in the remaining constructions. Prepositions are marks of determination, as in roi de France ('king of France'). Where there is determination/governance without any (overt) prep­ osition, as in aimer Dieu ('to love God'), a covert preposition must be postulated at the logical level, resulting in this case in aimer à l'égard de Dieu ('to love with respect to God'). The transition from the logical level to the syntactical one can take place only by means of 'ellipsis' (as employed already by Linacre and Sanctius). These facts may be summed up as follows:

In addition to ellipsis, Beauzée recognizes the existence of pleonasm. It was Du Marsais, however, who had integrated, on a much vaster scale, the traditional rhetorical figures into grammatical theory (cf. Soublin 1976). 'Identity' adds two entities together whereas 'determination' relates two entities to each other by means of a ('logical') preposition which may or may not be overt. For instance, non-restrictive and restrictive relative clauses are interpreted as instances of identity and determination, respectively (Beauzée 1974 [1767]: II, 29-32). Complément is used as a superordinate term both for logical relations (identity/determination) and syntactic ones (agreement/governance). Finally, Beauzée accepts Du Marsais' reinterpretation of the role of the (explicit or implicit) copula: it no longer performs the function of combining the subject and the attribute. Rather, together with a participle or an adjective, it now constitutes the attribute (II, 7-10). The idea seems to be that there is no need to use an overt sign to express the fact that the subject and the attribute have been combined. Similarly, no such sign is used to express the fact that la loi and

276

UNIVERSAL HISTORY OF LINGUISTICS

naturelle have been combined to produce the expression la loi naturelle. When there is a relation of determination, on the other hand, the (logical) preposition could be taken to be precisely such a sign. Universal grammar was practised also in England, James Harris (17091780) being its most influential representative. In designing his theory of language, he made a conscious effort to go back to Aristotle and, more generally, to the classical grammatical tradition; it is especially noteworthy that in addition to Priscian, he had first-hand knowledge of Apollonius. His system of the wordclasses, as described by Joly (1976), is more perspicuous than that of his French counterparts, as can be seen from the following figure:

At the turn of the century the 'tradition of Condillac' was continued and expanded into a 'revolutionary' philosophy of culture by the so-called Idéologues (cf. Busse & Trabant 1986). They do not seem to have contributed very much to grammatical theory (cf. Baum 1975), and therefore I shall omit them in the present context.

5.4.4 The 19th Century Throughout the history of Western thought the problem has been continuously raised how, or even whether, it is possible for a human being (qua subject) to

EUROPE

277

acquire knowledge about the external world (qua object). We have already seen this in the case of the Stoics (p. 186), Democritus (p. 190), Pyrrho (p. 190), Ockham (p. 250), and Locke (p. 260). Scepticism about the possibility of crossing the 'subject—object divide' resulted in emphasizing the priviledged status, or even certainty, of self-knowledge, as in the case of Descartes (p. 261). Immanuel Kant (1724-1804) tried to solve this age-old problem by refining the distinction between how things are and how they appear to be: on the one hand, it is a contingent fact that human beings are equipped with such and such a mental apparatus; on the other, it is a necessary fact that, given this apparatus, they perceive the external world (= 'things-in-themselves') as they do. It is obvious at once that Kant is merely repeating Bhartrhari's view that space and time, instead of belonging to the external world as it really is, are just attributes of human 'ignorance' (cf. p. 80), with only the difference that Bhartrhari is more consistent in regarding the unknown reality as One: if we really do not know what reality is like, we have no right to speak of things-in-themselves in plural; rather, reality is just 'something', an idea better conveyed by One than by Many. Indeed, Arthur Schopenhauer (1788-1860) later added this correction to Kant's doctrine. Until the beginning of the 19th century philosophers, including Kant, had envisaged the subject—object relation as that of a solitary person facing the external world. Sometimes this person was simply taken as representative of human beings in general. But it was just as possible to regard other persons as part of the external world, which then gave rise to a solipsistic world-view. All this changed with G.W.F. Hegel (1770-1831). He was probably the first to clearly realize that humans in the literal sense of this word are primarily social beings: the communality of many people with one another, rather than having to be precariously inferred, constitutes the very framework within which the individuality of different persons must be (secondarily) established. Because of the difficulty of Hegel's style, it is better to quote on this issue one of his (non-German) commen­ tators: All these [mental] acts, we see, involve at least the appeal to many selves, to society, to other spirits. We have no life alone. There is no merely inner self. ... My inner self is already an outer, a revealed, an expressed self. The only mind then is a world of many related minds. It is of the essence of the consciousness to find its inner reality by losing itself in outer, but spiritual relationships. Who am I then at this moment? I am just this knot of relationships to other moments and other people" (Royce 1892:209-210).

278

UNIVERSAL HISTORY OF LINGUISTICS

The primacy of the social as well as the dependence of consciousness on public behavior were graphically expressed by Marx and Engels in their Deutsche Ideo­ logie: Es zeigt sich hier, dass die Individuen allerdings einander machen, phy­ sisch und geistig, aber nicht sich machen (p. 37). ... Die Weise, in der die Menschen ihre Lebensmittel produzieren, ...ist vielmehr schon eine bestimmte Art der Tätigkeit dieser Individuen, eine bestimmte Art, ihr Leben zu äussern, eine bestimmte Lebensweise derselben. Wie die Indivi­ duen ihr Leben äussern, so sind sie (p. 21). The subject—object problem also receives a rather elegant solution: "Mein Verhältnis zu meiner Umgebung ist mein Bewusstsein" (p. 30). Not content to deal with consciousness, Schopenhauer was among the first to realize the importance of unconscious motives, thus anticipating Freud. In investigating the way that reality is conceptualized, Ludwig Feuerbach pointed out that in addition to abstract intellect and to social relationships, a significant role is also played by the human body. 21 History had become a concern for philosophers, thanks to efforts made by Vico and the Enlightenment philosophers. Yet it can be claimed with good justice that it was only in Hegel's work that the historical point of view was made the basis of an entire philosophical system. This is most obviously the case in his Phäno­ menologie des Geistes. This remarkable book is divided into three parts which can, in accordance with Lukács (1973 [1948]:728), be seen as dealing with the subjec­ tive spirit, the objective spirit, and the absolute spirit. In each part there is a gradual ascent from the least to the most perfect which remains structurally the same, the only difference being that the entity which performs the ascent is conceived differently in each case: it is first the individual, then the society, and finally the conceptual system. In other words, three historical processes are set here in parallel, namely the life-history of an individual person, the world history qua history of societies, and the quasi-logical development from lower to higher systems of thought. The driving force behind the ascent is (the experience of) contradiction, more precisely a contradiction between what is the case and what ought to be the 22

case. Hegel's view of history is as controversial as it is fascinating. To me its greatest shortcoming consists in the fact that Hegel does not seem to be able to make up his mind as to whether the type of historiography he is practising is of ideal (i.e. normative) or of factual (i.e. descriptive) character. This irresolution results in a great deal of arbitrariness. On the one hand, everything which conforms

EUROPE

279

to the 'norm' (i.e. what serves the self-realization of the absolute spirit as de­ ciphered by Hegel) is pronounced to have been necessary, whereas everything else qualifies as merely contingent. On the other, periods like the early Middle Ages which quite obviously violate the norm of intellectual and moral progress are misrepresented so as to make them comply with the norm just the same. It seems to be Hegel's selective, if not haughty, attitude towards historical facts which sanctions, ultimately, much of today's 'philosophical historiography'. Represented by Foucault and Rorty among others, it is characterized by chopping up history into discrete periods and then reducing each period to a single attribute. From the viewpoint of linguistics too, Hegel is a fitting symbol for the 19th century because in this period the emphasis shifted from universal grammar to historical linguistics. I refrain from following this development in any detail, in keeping with the general design of my book (cf. p. 165). Instead, I shall concen­ trate on the more hidden side of the 19th-century linguistics, by showing how the ultimately Aristotelian notion of (universal) grammatical theory lived on, even in the thinking of principal representatives of the 'new' historical linguistics. At the same time the myth will be dispelled that the start of the 19th century signalled any kind of sudden or absolute break in the history of linguistics. Rasmus Rask (1787-1832) and Franz Bopp (1791-1867) are generally credited with having laid the foundations either of scientific historical linguistics or of scientific linguistics tout court. The latter assumption is too obviously wrong to need any refuting, but the former assumption too is baseless. Measured by any standards, Sámuel Gyarmathi (1751-1830), following in the footsteps of his fellow Hungarian P. Sajnovics (1733-1785), established comparative linguistics fully in the sense of Rask and Bopp, the only difference being that the languages treated by him belong to the Finno-Ugric and not to the Indo-European family (cf. Gulya 1974: 266-272). The title of Gyarmathi's book indicates its scope and intent: Affinitas Linguae Hungaricae cum Unguis Fennicae Originis grammatice demonstrata (1983 [1799]). The structure of the book bears witness to the haste in which it had to be written: The first part compares Hungarian with Finnish and Lapp, without paying any attention to the differences between the latter two languages; the second part compares Hungarian with Estonian, and the third adds a few remarks on seven less well known Finno-Ugric languages. What is of more direct interest in the present context, Gyarmathi also wrote a large-scale synchronic grammar of Hungarian which Hanzeli (1983:xiv) characterizes as "thefirstcomplete grammaire raisonnée of a language other than the major languages of Western Europe". It is also interesting to note that Rask and Bopp had much more in common with the 18th-century thought than has generally been acknowledged. Hjelmslev

280

UNIVERSAL HISTORY OF LINGUISTICS

(1966 [1951]: 186) points out that Rask was interested, above all, in constructing a grammaire générale, which, to be sure, had to be based on cross-linguistic data, and not on philosophical speculation. Hjelmslev argues repeatedly that Rask did not yet distinguish between typological and genetic affinities among languages. He even goes so far as to claim that to Rask change of language (as opposed to change of speech) is inconceivable: one system does not change into another, but is rather replaced by it (p. 191). Percival (1974:308) confirms this view: "Rather than assume a gradual transformation of structure, Rask posited that when a language changes, its grammatical structure is abandoned and then replaced by a different one." This attitude could be explained by the fact that "Rask had difficulties in realizing the full importance and true character of what was later (with a misleading term) called 'phonetic laws'" (Diderichsen 1974:299). Rask is an heir to the 17th and 18th centuries in the sense that he remains committed to the ideal of form—meaning isomorphism: "Rask recherche partout la conformité entre le contenu et l'expression. Il la maintient à titre de principe, tout en admettant qu'elle n'est pas partout réalisée au même degré" (Hjelmslev 1966 [1951]: 188). The lack of isomorphism in existing languages serves as an incentive for Rask to engage in planning an artificial language, called pasilalie, in which the isomorphism would be perfect. Thus Rask maintains the notion of universal gram­ mar/language in both of its senses (cf. p. 269). Traditional motifs are equally evident in Bopp's thinking. It is quite amazing to see that he fully accepts Aristotle's sentence-analysis, with the either explicit or implicit copula still performing the essential sentence-constituting function: A verb, in the most restricted meaning of the term, is that part of speech by which a subject is connected with its attribute. According to this definition it would appear that there can exist only one verb, namely, the substantive verb, in Latin esse; in English, to be. The Latin verb, dat, expresses the proposition, 'he gives', or 'he is giving': the letter t, indi­ cating the third person, is the subject, da expresses the attribute of giving, and the grammatical copula is understood. In the verb potest ['he is able'], the latter is expressed, and potest unites in itself the three essential parts of speech, t being the subject, es the copula,andpot the attribute (Bopp 1974 [1820]:22-23). Ironically, it is this logical analysis which guides Bopp in his historicoempirical investigations. Everywhere he wishes to see traces of the Indo-European copula in one of its two variants (Sanskrit as-, bhū-, Latin es-, fu-).

EUROPE

281

Bopp's outlook on language is the same as Rask's. First, consider the notion of change: "Bopp compares forms of language which he takes to be static. Histori­ cal genesis, glottogony, and the diachronic process do not interest him as such" (Verburg 1966 [1950]:236). Second, consider the notion of isomorphism: "Bopp's view is: words represent thoughts. The world of thought is composed of parts, of mathematical elements, so to say, (Grundbegriffe, accessorische Verhältnisbegriffe), and the expression of this world of thoughts runs parallel in its structure with that of the world expressed; it does so necessarily" (ibid.). As a consequence, Bopp is forced to countenance agglutinative structure as the only intelligible type of grammatical structure. Any other type creates insuperable difficulties for him. It must be added that, even after Rask and Bopp, the ideal of meaningform isomorphism continued to dominate historical-comparative linguistics. Kiparsky (1974:334-335) illustrates this by reproducing Schleicher's attempt, in the mid-19th century, to reconstruct the Proto-Indo-European medial endings for the 1. and 2. person plural. Three assumptions constitute the basis for this attempt: the active endings for sg. 1. and 2. p. are originally pronominal forms ma and tva; the plural endings are formed, in conformity with their meanings, by combining the singular endings: ma-tva ('we') and tva-tva ('you'); the meaning of the medial forms under consideration is something like 'We ... ourselves' and 'You ... yourselves'. As a consequence, the requisite endings are claimed to be ma-tva-matva and tva-tva-tva-tva. (But notice that these should really be only dual endings; the plural should require at least three repetitions.) — Such extreme adherence to the principle of isomorphism should not obscure the fact that this principle does play a central role in the explanation of linguistic change (cf. LT:208-210, Itkonen 1984, Nyman 1987). In sum, the diachronic linguistics of the 19th century was underlain by considerations stemming from the tradition of universal grammar. But this tradition was also perpetuated in a direct form. In his book Organism der Sprache (1827, 2nd edition 1841) Karl Ferdinand Becker endeavours to construct an allgemeine Grammatik in the spirit of Aristotle and the Stoics; and he repeats verbatim the Modistic view that "der Unterschied der einzelnen Sprachen liegt in den besonderen Formen des Ausdrucks für die allgemeinen Formen der Anschauung und des Denkens" (p. xviii). Following the advice of Adolf Trendelenburg, a contemporary logician, Becker intends to eliminate the distinction between grammar and logic: for him, the structure common to all languages is the structure of human thought, and uncovering the latter amounts eo ipso to uncovering the principles of logic. More precisely, this universal structure is semantic, rather than syntactic or formal, and

282

UNIVERSAL HISTORY OF LINGUISTICS

Becker correctly points out that, knowingly or unknowingly, linguists have always relied on it. The following remarks against the autonomy of descriptions of particular languages continue to apply today with equal force: ..., vielmehr hat die Grammatik von je her neben der Besonderheit die Einheit aller Sprachen anerkannt. Denn was ist es anders, als die Aner­ kennung dieser Einheit, wenn die Grammatik in fast allen Sprachen gleiche Wortformen, Kasus, Modus, Präpositionen, Konjunktionen u.s.w. mit denselben Namen unterschieden hat? Nicht die Gleichkeit der Formen, sondern die Gleichkeit der Bedeutung sprach sie damit aus (Becker 1841: xix). Even if Becker's plan for universal grammar is commendable, its realization is less so. He starts from an opposition between thing (Sein) and action (Thätigkeit) which superficially resembles the Modistic distinction between the modes of permanence and flux. Becker, however, opposes the Aristotelian tradition in taking 'action' to be the primary concept, and for this interpretation he seeks support in comparative linguistics: He takes it to be an established fact that the roots of 'all languages' are of verbal character; as a consequence, nouns qua words referring to things must have been derived from verbs (p. 70). The difference between action/ verb and thing/noun is interpreted both ontologically and historically: action is general and primary whereas thing is particular and secondary. Thinking is charac­ terized by two movements: taking the particular into the general, and reducing the general to the particular. These movements find their linguistic expressions, on the one hand, in the predicative relation (the particular and the general being rep­ resented by the subject and the predicate) and on the other, in the attributive relation ( = specifying the subject) and the 'objective' relation ( = specifying the predicate either by an object or by an adverb) (pp. 158-162). Becker also em­ phasizes that, while verbs are primary with respect to nouns, movement in space is the common core of all verb meanings: non-sensible entities can be expressed only by a sort of 'translation' process, i.e. by using words originally referring to sensible entities (pp. 72-81). From the above it is evident that Becker's notion of 'logic' is a somewhat idiosyncratic one. Nevertheless, Steinthal (1972 [1881]:44-72) feels compelled to fight Becker's identification of grammar and logic as part of a larger tradition which goes back to Aristotle via Condillac, the Port-Royal grammar, Scaliger, Thomas of Erfurt, and the Stoics. Steinthal himself identifies 'logic' and '(laws of) thought', but he denies that language and thought are identical. The very fact that the former is said to express the latter proves that they are two different things; and

EUROPE

283

if A expresses , it by no means follows that the forms of A are the forms of B; all that can legitimately be inferred is that the forms of A are the forms of expressing  (. 56). After mentioning the often illogical character of natural language, Steinthal has to admit that an attenuated form of 'philosophical' (or 'logical', 'general', 'ideal') grammar was also held by Wilhelm von Humboldt, whom he otherwise considers as his model (pp. 63-65). According to this view, philosophical grammar occupies a sort of middle position between logic proper and linguistic grammar insofar as it contains those forms of thought (i.e. logical principles) which are, or rather, happen to be, manifested in all natural languages. Steinthal, however, finds even this notion of philosophical or universal grammar unacceptable. To prove his point, he repeats the traditional view that grammarians and logicians just concern themselves with dissimilar types of problems: Semantically anomalous sentences like This square table is round disturb logicians, not grammarians, but the situation is reversed in case of grammatically incorrect sentences like This table are round (p. 70). This is no definitive solution, however, because the same division of labor between grammarians and logicians was also accepted by the medieval representa­ tives of philosophical grammar (cf. p. 227). Trendelenburg, who was already mentioned in connection with Becker, advocated a continuation of Leibniz's program to construct a characteristica univer­ salis or universal language; his own term for it was Begriffsschrift. As Haaparanta (1988) points out, Gottlob Frege (1848-1925) seems to have responded to Trende­ lenburg's prompting, as shown by the fact that he chose Begriffsschrifi as the title of the book in which, in 1879, he introduced modern formal logic. Yet Frege went in a direction very different from Becker's. Instead of identifying formal logic and linguistic grammar, he draws a line between the two, by adopting the mathematical 'argument—function' terminology instead of the grammatical 'subject—predicate' terminology (cf. Kneale & Kneale 1962:483-485). It is quite interesting to note, however, that it was possible to reach Frege's results by means of a semi-grammatical analysis. This was done by Charles Sanders Peirce (1839-1914), as argued by Ferriani (1987). Peirce knew well the work of the Modistae, and he respected it. Yet he criticized it, and the subsequent tradition, for having mistaken 'Arian' grammar for universal grammar. No genuinely universal grammar can afford to neglect the comparative study of typologically divergent languages. Peirce's sentence-analysis, which he in the traditional fashion bases on word-classes, contains three elements: Nouns and adjectives are dispens­ able, because they can be replaced by verbs; the copula is dispensable (as already demonstrated by a few 'Arian' languages); pronouns like this represent things in the

284

UNIVERSAL HISTORY OF LINGUISTICS

most direct way. As a consequence, there emerge sentence-structures like Lab structurally analogous to the Fregean L(a,b), except that Peirce has no inhibitions against applying the terms 'subject' and 'predicate' to a and b, and to L, respect­ ively (cf. Kneale & Kneale 1962:430-432). — In pointing out the dispensability of the copula, Peirce exposes a weakness, or at least an oversight, in the traditional Aristotelian conception. If the subject and the attribute have to be combined for a sentence to arise, there is no logical reason why this act of combining should also be expressed linguistically. Given that the subject, the copula, and the attribute are equally words, i.e. thing-like entities, it would seem that one more act is needed to put them together. And if this act too must be linguistically expressed, an infinite regress ensues. Therefore it must be enough that the act of combining just 'shows itself, instead of being overtly expressed. (As noted on p. 275, Du Marsais and Beauzée already had an inkling of this fact.) Frege also made a lasting contribution to the philosophy of logic, a contri­ bution which can be generalized to the philosophy of linguistics with only minor modifications. In the latter half of the 19th century it was the majority opinion that logic had to be founded on psychology. Frege fought this psychologistic view energetically, repeating and refining arguments already employed by Abaelard (cf. p. 225). On the other hand, Frege opposed no less strongly the formalist conception according to which formal logic consists in nothing but manipulation of inherently meaningless marks. Perhaps inevitably, his own positive answer is less clear. He emphasizes repeatedly that logic deals with 'laws of thought' (Denkgesetze), in the sense of prescribing how people ought to think, instead of describing how they think in fact (1893:XV-XVI). This normative dimension is independent from factual opinions in the same way as being-true (Wahrsein) is from being-held-to-betrue (Fiirwahrgehaltenwerden). Frege's normative conception of logic acquires a Platonistic cast when he adds that the laws of logic are valid eternally: "[Sie sind] Grenzsteine in einem ewigen Grunde gefestigt, von unserem Denken überflutbar zwar, doch nicht verrückbar". Frege explicitly denies the similarity between logic and language. Statements describing how "people speak in the grammatically correct way" are about empiri­ cal regularities comparable to those exhibited by the functioning of digestion or by the changes in fashion. Unlike logical laws, "grammatical laws express the nature of human thinking and change with it". On this issue Frege needs to be corrected. The distinction between psychological and normative is just as mandatory in language as it is in logic. Once this is realized, the notion of a non-human norm which people may or may not grasp in principle, but just happen to grasp in practice, loses much of whatever initial plausibility it may have had. Grammatical

EUROPE

285

theory and formal logic turn out to have the same methodological status, based on the ontological similarity of their respective subject matters (cf. GTM.7.0, 10.0). Having flirted with psychologism in his earlier work, Edmund Husserl (1859-1938) opposed it in his Logische Untersuchungen (1900-1901, 2nd edition 1913). Yet his overall account is less convincing than Frege's, for two reasons. First, he mistakenly conceives of normativity as in some sense 'containing' a psychological component, which then makes him feel justified to criticize the Fregean position for implicit psychologism (cf. Husserl 1913, I:53-58). Second, he remains centrally concerned with an analysis of logical experience, thus committing the psychologistic fallacy himself, by his own later admission (cf. Juntunen 1986: 66-69). As Frege sees it, formal logic speaks about the same things as natural language, but much better, the relation between the two being comparable to that between the microscope and the naked eye. Husserl, however, assumes a much more intimate connection between logic and language, as can be seen from the fact that he, in addition to exploring the notion of philosophical-universal grammar in some depth, intends to base logic on it. When Husserl (1913) approaches the notion of reine Grammatik, he announces, more specifically, that he intends it as a 'syntax of meanings' (Formen­ lehre der Bedeutungen) (p. 295); and the principal division will be made between categorematic (or independent) and syncategorematic (or dependent) meanings, not just forms (p. 305). It is obvious at once that what Husserl is practising is identical with the Stoicist analysis of lekta. Like Steinthal, he next distinguishes between semantically anomalous and grammatically incorrect sentences, and assigns their study to logicians and to grammarians, respectively (pp. 318-321). The use which he makes of this distinction, however, is quite different from Steinthal's. Semantic anomaly (Widersinn) is illustrated with such standard examples as This algebraic number is green, and contrasted with such semantically normal sentences as This tree is green. Grammatical incorrectness (Unsinn), by contrast, is not illustrated with such minor violations as those against rules of agreement, as in Steinthal's case, but rather with violations against the 'categorematic vs. syncategorematic' distinction, like This tree is and. Husserl makes two points in this connection: First, the 'grammatical' incorrectness is not of formal, but of semantic nature: it consists in a wrong combination of meaning categories. Second, both semantic normality and semantic anomaly (qua concern for logicians) presuppose grammatical (= categorial-semantic) correctness, which means that ('pure') grammar is the foun­ dation of logic.

286

UNIVERSAL HISTORY OF LINGUISTICS

The latter claim may seem preposterous, but if taken in the sense intended by Husserl, it is difficult to deny. I also find the former claim absolutely correct. Husserl elaborates on it, pointing out (just like Becker had done) that any universal grammar, employed to describe either German or Chinese, is necessarily based on such semantic (rather than formal) notions as 'plural', 'modality', 'conditional', 'negation' etc. (pp. 338-339). Here Husserl brings out quite well the absurdity of all attempts to make the form of natural language into the primary object of linguistic study. Of course, it is possible to pretend that linguistic forms are not, first and foremost, bearers of linguistic meaning, but this is a different matter. Within the universal grammar Husserl distinguishes between an aprioristic part and an empirical one, the former being identical with his reine Grammatik. It contains, first, primitive meanings (nominal, adjectival, syncategorematic) as well as rules for combining them into a simple sentence-meaning. Second, there are rules for constructing more complex meanings in such a way that they still belong to the basic meaning categories; for instance: N and N = N, A and A = A, AN = N, S and S = S, S or S = S, if S, S = S (pp. 330-331). Moreover, nominalizations of adjectives or of sentences exemplify cases where meanings are 'transformed' while their 'kernel' remains unchanged (pp. 324-325). As a result, it is possible to produce in a regular way an infinite number of meaning forms: ... wir können offenbar, willkürlich variierend, innerhalb der gesetzlichen Zulässigkeit verschiedene der Verknüpfungsformen kombinatorisch zu den Konstruktionen benutzen und so in gesetzlicher Weise Unendlichkeiten von komplizierten Formen erzeugt denken (Husserl 1913, II:332). There is an interesting sequel to Husserl's occupation with philosophicaluniversal grammar. In his youth Martin Heidegger (1889-1976) was under Husserl's influence and took up topics dictated by the latter's interests. In his dissertation, published in 1916, Heidegger investigated the Grammatica Speculativa (whose author was at the time taken to be Duns Scotus rather than Thomas of Erfurt) and presented it as a full-fledged realization of Husserl's program of philosophical-universal grammar. This interpretation may be accepted in some general sense: the most important modi significandi are indeed identical with Husserl's 'aprioristic' semantic categories. However, Heidegger's concomitant attempt to picture Thomas of Erfurt as a phenomenologist avant la lettre is less successful. Later Heidegger's language-conception was to diverge sharply from Husserl's. These developments have been exhaustively reported by Kusch (1989). In this subsection I have up to now been describing the 'hidden' history of the theory of universal grammar in the 19th century, as it took place in the interior

EUROPE

287

either of diachronic linguistics or of philosophical/logical linguistics. Towards the end of the century, however, there were also deliberate attempts to construct an explicit and empirical theory of universal grammar. It was in this spirit that Steinthal & Misteli (1972 [1893]) laid the foundations of modern linguistic ty­ pology, and that von der Gabelentz (1840-1893) wrote his encyclopedic work Die Sprachwissenschaft (1891), in which he tried, as the subtitle announces, to present 'the tasks, the methods, and the achievements' of linguistics. From the wealth of information provided by von der Gabelentz, it is possible to piece together the outlines of a theory that may quite neatly be placed within the tradition of universal grammar that has been discussed in this chapter. Yet unlike his predecessors, von der Gabelentz, himself an expert on Chinese, bases his views on an impressive amount of cross-linguistic data, i.e. data stemming not just from Indo-European languages, but also from several Finno-Ugrian, Asian, African, and American-Indian languages. In his terminology, he rejects 'general or philosophical grammars', while accepting their goals, and opts for 'general linguis­ tics', which he understands as 'part of anthropology in the wider sense of this word'. Since languages resemble such cultural phenomena as religions and customs, linguistics cannot be a natural science, even if it has to rely on the 'inductive method' (pp. 11-15). The linguistic system, also called the 'form of the language', is learned by the child on the basis of 'unconscious abstraction out of multiple experience': "Die Wirkung dieser unbewussten Abstraction nennen wir Analogie". This process, which presupposes the existence of surprisingly abstract concepts in the child's mind, must be explained by the psychologist, not by the linguist. Yet even a linguist is able to observe that there is considerable variation in how rapidly or slowly individual children learn their first language (pp. 63-66). It would be a defeatist attitude to view the linguistic capacity as innate: "Mit der Frage nach den angeborenen Ideen brauchen wir uns hier nicht zu beschäftigen. Eine Idee für angeboren erklären, heisst erklären, dass sie unklärbar sei" (p. 365). If one tries to explain the linguistic structure, the basic ontological divisions offer the obvious place to start: "der Mensch brauchte nur die Welt zu betrachten, um des Unterschiedes zwischen Ding, Eigenschaft und Thätigkeit inne zu werden" (p. 366). It was a natural urge to express these conceptual (or 'logical') distinctions by corresponding linguistic distinctions: "die logischen Kategorien drängten dazu, sich in grammatische Classen zu gestalten, in Redetheile zu son­ dern" (p. 367). However, these ontological-grammatical correspondences are not exactly the same in all languages. Rather, some languages exemplify a 'predomi­ nantly nominal conception', whereas others exemplify a 'predominantly verbal

288

UNIVERSAL HISTORY OF LINGUISTICS

conception'. That is to say, languages like Mayan or Nahuatl tend to express not just actions, but also qualities and even things by 'verbs', whereas languages like Sanskrit or Malay tend to express not just qualities and things, but also actions by 'adjectives' and 'nouns' (pp. 367-369). From today's perspective, the most problematic aspect of von der Gabelentz' work is his attempt to carry out a Sprachwürderung, i.e. a ranking of languages according to their higher or lower 'spiritual value' (pp. 371-457). His own attitude towards this kind of undertaking seems to be ambivalent. He notes (p. 376) that it is not possible to find linguistic features that uniquely characterize the languages of civilized nations. But then he immediately reverses this judgment: "Und doch bleibt es dabei: soweit die Gesittung oder Roheit der Völker von ihrer geistigen Beanlagung abhängig ist, muss jenen der Werth der Sprachen ent­ sprechen." And he proceeds to search for data that might substantiate this thesis. As is to be expected, the results are less than conclusive. It seems befitting to conclude this subsection by taking up the theory of synchronic linguistics more explicitly, and I shall do so by examining the account that the leading neogrammarian Hermann Paul (1846-1921) gave of it in his Prinzipien der Sprachgeschichte (1880, 9th ed. 1975). This is in keeping with my wish to focus on the neglected side of the 19th century: it is not commonly thought that Paul had much to say about the synchronic study of language. (Koerner's article from 1973 may be mentioned as an exception.) In implicit reference to Rask and Bopp, Paul notes (p. 23) that historical linguistics has often consisted merely in comparing states of language with one another. This approach is unsatisfactory because it loses sight of linguistic change. Change is important because it is only in it that the causal nexus obtains (p. 24, also p. 11, note 1). Science is equivalent to explanation (p. 21), and explaining means, precisely, revealing the causal nexus. Therefore there is no scientific study of language over and above historical (i.e. diachronic) study (p. 20). (This verdict is immediately qualified, however; cf. below). 'Historical grammar' was preceded by 'descriptive grammar', by which Paul clearly means the 16th and 17th century grammars treated here in 5.4.1 and 5.4.2, as can be inferred from his brief reference to their theory of ellipsis (p. 313). To Paul, descriptive grammar is non-scientific because of its non-causal character. It studies abstractions, and between abstractions there can be no causal nexus. This assessment of descriptive grammar (today called 'autonomous linguis­ tics') is factually correct, but there is no reason to accept the value judgment it contains.

EUROPE

289

It is of decisive importance to realize that in addition to historical grammar and descriptive grammar, Paul also postulates, and describes, a third type of grammar (esp. pp. 29-31), which is meant to be the psychological study of a given synchronic state of language (Sprachzustand); it may appropriately be called 'psychological grammar'. The phenomena investigated by it are situated below the level of consciousness; Steinthal is applauded for having repeatedly emphasized this fact (p. 25). Unconscious phenomena can be investigated only by trying to infer their nature from their observable effects (p. 29). Therefore the results achieved by psychological grammar must not be confused with those achieved by descriptive grammar, which relies on the method of 'grammatical reflection' (p. 27, 31). (This is a clear statement of the fundamental difference between psycholinguistics and autonomous linguistics.) Even if the ultimate purpose of psychological grammar qua study of synchronic states of language is to produce data for historical grammar, which purports to explain the change from one state to another, it nevertheless qualifies as a scientific grammar (p. 31, note 1), because it is clearly interested in answering (part of) the causal question. Psychological grammar conceives of its primary object as a 'system of groups of unconscious concepts' (Organismus von unbewussten Vorstellungsgruppen). Concepts pertain to lexical and grammatical meanings, on the one hand, and to sounds, word-forms, suffixes, and syntactic relationships, on the other (pp. 26-27). The groups are held together by similarity of form and meaning, as in the case of inflexional paradigms (also pp. 189-190). The general notion of grammar is non-discrete because the associative force which, on the one hand, holds groups together and, on the other, relates them to one another is assumed to be a matter of degree. Moreover, this whole system is constantly in transition: the relations of force are shifting, new connections are being created, and old ones are being discarded. The psycholinguistic system is not just something that has been memorized. Rather, it is underlain by a combinatory activity which produces new expressions and sentences on the analogy of old ones. In language-acquisition the rules for generating an infinite number of new sentences are abstracted from encountered utterances insofar as these contain a 'common element': Es war ein Grundirrtum der älteren Sprachwissenschaft, dass sie alles Gesprochene, so lange es von dem bestehenden Usus nicht abweicht, als etwas bloss gedächtnismässig Reproduziertes behandelt hat, ... Zwar hat schon W. v. Humboldt nachdrücklich betont, dass das Sprechen ein immerwährendes Schaffen ist. ... Es ist eine nicht zu bezweifelnde Tat­ sache, dass eine Menge Wortformen und syntaktische Verbindungen, die

290

UNIVERSAL HISTORY OF LINGUISTICS

niemals von aussen in die Seele eingeführt sind, mit Hilfe der Proportio­ nengruppen nicht bloss erzeugt werden können, sondern auch immerfort zuversichtlich erzeugt werden, ... Ohne weiteres wird zugegeben werden müssen, dass die wenigsten Sätze, die wir aussprechen, als solche auswen­ dig gelernt sind, dass vielmehr die meisten erst im Augenblicke zusam­ mengesetzt werden. ... Bei dem natürlichen Erlernen der Muttersprache... hören [wir] nach und nach eine Anzahl von Sätzen, die auf dieselbe Art zusammengesetzt sind und sich deshalb zu einer Gruppe zusammenschliessen, ... und so wird die Regel unbewusst aus den Mustern abstrahiert" (Paul 1975:109-111). Although it is convenient to let a given state of language be represented by one psycholinguistic system, Paul emphasizes that in point of fact each person has his own system, and systems may vary from one person to the next (p. 27). These individual systems (which are connected with one another by sound, or the physical aspect of language) are the bearers both of linguistic states and of linguistic changes (p. 28). What changes is really not (social) language but the set of individual (psychological) systems. Each generation has to create its own language anew, which means that language learning is the most important cause of linguistic change (p. 34). As an advocate of 'methodological individualism', Paul argues strongly against Völkerpsychologie ('psychology of nations') as represented by Steinthal and Wundt (pp. 8-16). I shall finally say a few words about Paul's sentence-analysis (pp. 121150). Here too he maintains the distinction between psychological and grammatical categories. Being the primary entities, the former are expressed by the latter, and syntactic change is claimed to be typically caused by a discrepancy between the two levels. For Paul, the simplest type of sentence consists in a combination of two notions {Vorstellungen), namely (psychological) subject and (psychological) predicate. These two are defined, respectively, as that (= A) to which the speaker directs the hearer's attention, and as that which the speaker makes the hearer think about A (pp. 124-125). As a consequence, psychological subjects and predicates are in principle independent of the corresponding grammatical categories. There is a general tendency to prepose the linguistic expression of the psychological subject and to accentuate the linguistic expression of the psychological predicate (pp. 126-127). Even to one-word sentences there corresponds at the psychological level a combination of two notions, one of which must be inferred from the speech situation (pp. 129-133). (This works well in ellipsis, but less well in case of impersonal verbs.) Negative sentences do not really exemplify combination, but

EUROPE

291

separation of what was supposed to be combined (p. 133). (The Aristotelian flavor of this analysis is obvious.) Statements, directives (Aufforderungen), and questions are on an equal footing. Next Paul shows how to make this simple picture more complex. His basic idea is that the (psychological) 'subject—predicate' relationship repeats itself in the 'determined—determination' (Bestimmtes—Bestimmung) relationship: determination is just a 'degraded' predicate. If we assume, for simplicity, a coincidence of psychological and grammatical categories, it may be noted that primary subjects are determined by adjectives, nouns, and prepositional phrases, but also by so-called apo koinou predicates (as in the Middle High German example do spranc von dem gesidele her Hagene also sprach, lit. 'then jumped from the chair Sir Hagen thus spoke'). Primary predicates are in turn determined by objects and adverbial phrases. In each case a determination of level n may in the sequel become a determined of level n + 1. Finally, subordinated clauses too enter into the relationship of determination. As a simple example, consider the sentence Karl fährt morgen nach Berlin (pp. 282-283). If uttered in a neutral context, it consists of four elements, as follows:

Karl and fährt are the primary subject and the primary predicate, but fährt is also the secondary subject (= determined) with the two secondary predicates (= determinations) morgen and nach Berlin. Depending on the speech situation, however, the sentence may consist of only two elements in the sense that each of the four elements may be the psychological predicate (= 'new information') while the rest constitutes the psychological subject (= 'old information'). By making the distinction between psychological grammar and descriptive grammar, Paul places himself in the long tradition which, starting from Plato, operates with the two levels of mental language and oral language. In the preceding tradition the relation between the two had been conceived of as a very close one, the mental (or 'logical') language being just a purified version of the oral one, i.e. a version elicited just by means of logical reflection. Paul, however, breaks new ground by taking the claim to psychological reality seriously and by insisting, as a consequence, that grammatical or logical reflection is not sufficient in the investi­ gation of unconscious phenomena. This is why he is ready to accept a greater

292

UNIVERSAL HISTORY OF LINGUISTICS

divergence between psychological and grammatical than any of his predecessors was, as far as I can see.

5.5 The 20th Century In the present century there has been a huge increase in the amount of linguistic research and, parallel to it, clear progress in the linguistic theory-construction. Only ignorance of historical facts could make one think, however, that there has been anything like a 'total break' vis-à-vis the linguistics of the preceding centuries. I cannot hope to survey the facts of the 20th century grammatical theory in any detail. Therefore I shall be content to delineate two main research programs that are discernible in our century, namely 'autonomous linguistics' and 'causal (or functional) linguistics'. As before, the general 'intellectual climate' of the century will be described first, by focusing on a few seminal philosophers.

5.5.1 Some Philosophical Background In his program of 'transcendental phenomenology', which was launched in Logische Untersuchungen, Husserl continued and elaborated the Cartesian tradition in Western philosophy. It was his overriding purpose to reach the rock bottom of apodictic (i.e. absolutely certain) knowledge, and he thought he found it in the knowledge of 'ideal essences' (which might perhaps be popularly translated as 'concepts as such'). Essences are 'constituted' in various psychological acts, all of which share the same basic intentional or directional structure: a given essence (e.g. '5' or 'house' or whatever) is what various acts are directed to, and what is common to all of them; an essence is purified of whatever contingent features may adhere to thought-contents exemplifying it. Acts of a particular type (e.g. per­ ceiving or expecting) must themselves be seen in turn as exemplifying a corre­ sponding essence. What is left, is supposedly the intentional structure of a 'pure' or 'transcendental' consciousness. The existence of any world outside it is 'bracketed', i.e. neither affirmed nor denied, but simply left out of consideration. It is this program which Husserl tried to implement during the first two decades of the 20th century. Gradually he had to admit, however, that what is thought does not exist in a 'transcendental' vacuum, but is rather grounded in a given cultural context, or what he came to call Lebenswelt ('life-world'). This insight was expressed in his Krisis book, written around the mid-thirties and

EUROPE

293

published in 1954. The book criticizes the modern natural sciences as well as logic and mathematics for unwarranted self-sufficience: they consistently overlook the fact that they spring from a certain type of culturally determined, pretheoretical thinking. At the same time, Husserl makes the questionable attempt to show, by 'bracketing' the entire life-world, that transcendental phenomenology, and it alone, is free from any socio-historical presuppositions. The whole of Husserl's philosophy can be construed as a reaction against a momentous shift in scientific thinking that had occurred somewhat earlier, namely the abandonment of the 'quest for certainty': "... during the course of the nine­ teenth century, the view of science as aiming at certainty gave way among most scientists to a more modest program of producing theories that were plausible, probable, or well tested" (Laudan 1984:83). Against this background, Husserl could be viewed as stubbornly insisting on the unique position of philosophy. We have noted above that Heidegger started out as a disciple of Husserl. In his Sein und Zeit (1927), however, he presents a conception of philosophy which is distinctly different from Husserl's. The latter had been preoccupied with the question of apodictic knowledge, which goes back to the distinction between subject and object, or man and world: is it possible for man to acquire indubitable knowl­ edge about the world external to him? In answering this question negatively, Husserl put himself on the side of man, or ('transcendental') consciousness, thus perpetuating the Cartesian tradition. Now Heidegger abandons the entire framework withing which it makes sense, in the first place, to ask the question which was answered negatively by Husserl. It is a fact that we 'always already' live in the world, and it is this fact which needs to be explicated. Thus, instead of opposing man and the world to each other, Heidegger starts from the all-encompassing notion of 'man-in-the-world'. This change of perspective has much to recommend itself, and it is likely that Husserl's notion of 'life-world' was influenced by it. Accepting Heidegger's general stance does not force one into accepting his more specific, and at times outlandish, views about what the human existence is like. In his later work Heidegger becomes increasingly hostile towards rational argument and advises his readers to replace science and philosophy by poetry. — It may be added that the idea of starting from the totality, rather than constructing it bit by bit, by no means originated with Heidegger. It can be found fully ex­ pressed both in Hegel and in Marx (cf. here p. 278). Besides, there is no need to regard these alternatives as exclusive of each other. Maybe the 'man vs. world' dichotomy is a theoretical development of the unitary 'man-in-the-world' notion; but there is no reason why this development (besides others) should not be pursued, if only to probe its limits.

294

UNIVERSAL HISTORY OF LINGUISTICS

It has been customary to regard Husserl and Heidegger as the main rep­ resentatives of the 'Continental' philosophical tradition in the 20th century. The other tradition, namely that of Anglo-American 'analytic philosophy', is often thought to have been launched by Ludwig Wittgenstein in the earlier phase of his career, but it really goes back to Frege, through the intermediary of Bertrand Russell. Interestingly enough, Wittgenstein came rather near to the Continental way of thinking in his later phase. Wittgenstein (1921) constructs a formal language, with Frege and Russell as his acknowledged predecessors (3.325); the requirement of unambiguous ex­ pression, voiced already by the Stoics, shows at the same time the connection with the ideal-language movement. The ontology of the world which the formal language talks about is that of 'logical atomism' as outlined by Russell. What separates Wittgenstein from Russell, is his analysis of the picturing relation {abbildende Relation, Projektion) between sentences and states of affairs: "Der Konfiguration der einfachen Zeichen entspricht die Konfiguration der Gegenstände in der Sach­ lage" (3.21). Once units of (logical) thought are interposed between 'names' (i.e. simple signs) and things (cf. 4.01: "Der Satz ist ein Modell der Wirklichkeit, so wie wir sie uns denken"; also 3.2.), Wittgenstein turns out to represent the language—mind—reality isomorphism in its classical form. Indeed, Steinthal's characterization of Plato's phusei program captures perfectly the spirit of Wittgen­ stein's Tractatus period (cf. here p. 172). More conspicuously than his predecessors, Wittgenstein dwells on the fact that a sentence is a state of affairs with exactly the samerightas that which it talks about. This, however, creates a problem: In the extralinguistic state of affairs where, e.g., a boy is hitting a dog, there are two things and one (two-place) relation, whereas in the corresponding sentence there are, prima facie, three things (i.e. the words boy, hit, dog) and one (three-place) relation. How, then, is iso­ morphism to be achieved? One solution to this problem, hinted at in 3.1432, and explicitly presented by Stenius (1960), is to 'embed' hit in the relation between boy and dog. Stenius also justified in the present context the use of the term 'iso­ morphism' in its technical sense (for discussion, and application to linguistic semantics, cf. Itkonen 1970). In elaborating his 'picture theory' Wittgenstein has to pay special attention to the extralinguistic correlates of linguistic entities. Already Plato knew that there is no single extralinguistic state of affairs corresponding, point-by-point, to a false sentence (p. 173). In the Middle Ages it was agreed that syncategorematic terms do not refer to anything (p. 245). Now Wittgenstein argues in a similar vein that there is nothing extralinguistic corresponding to the negation sign '¬' (4.0621). It

EUROPE

295

follows that the same must be true of the signs which are or can be defined by means of negation, i.e. the connectives and the quantifiers (5.42, 5.441). Only names and predicates have ontological or extralinguistic correlates of some sort. All other signs express (mental) operations performed on (representations of) simple states of affairs. Tractatus represents a static and solipsistic conception of (ideal) language. From the end of the '20s onwards Wittgenstein adopted an increasingly critical attitude towards this conception. His critique is expressed most fully in Philosophi­ cal Investigations (1953), in which the emphasis is on the use-based and social nature of (natural) language. When the isomorphism between extralinguistic and linguistic states of affairs is the center of attention, one easily loses sight of the fact that the latter are used in a characteristic fashion, i.e. normally to assert something. The same use is implicit in such isomorphic representations par excellence as geographical maps. This insight, that every picture may be used and understood in many different ways, is exploited by Wittgenstein in a fully original way. Ever since Plato and Aristotle it had been thought that the existence of mental images (or 'pictures') intervening between words and things explains how the former become attached to the latter. But Wittgenstein points out that this, as such, explains nothing because, logically speaking, mental images need to be supplemented by 'instructions of use'. What these should be like, is not known; at least they cannot be anything like images, because then they would need additional instructions; and so on. The most that can be said with certainty is that human beings are in fact so constituted as to be able to attach words to things. — It may be added that there is no contradiction between Wittgenstein's earlier 'picture theory' and his later 'use theory'. Rather, the former may be regarded as remaining embedded in the latter. In this way Wittgenstein's career combines the two central notions of Plato's theory of language, namely phusei and organon. The social aspect of Wittgenstein's theory of language is most evident in his claim that there can be no so-called private languages. These are languages con­ sisting of rules which only one person, i.e. their inventor, can follow in principle. The main argument against private languages is based on the fact that the speaker, being dependent on his own memory of what the rules consist in, has no way of distinguishing between what is correct and what merely seems, at a given moment, to him to be correct (for more discussion, cf. GTM: 109-121). Wittgenstein's argument has often been criticized by claiming that there not only can be, but in fact are private languages. It is invariably the case, however, that the term 'private language' is then used incorrectly, i.e. not in the way

296

UNIVERSAL HISTORY OF LINGUISTICS

intended by Wittgenstein. For instance, the following entities do not qualify as private languages: unconscious mental grammars, pidgins, or 'home-sign systems' invented by deaf children of hearing parents (cf. LT:239-243, and Itkonen 1984). Moreover, contrary to Pateman (1987:127), the private-language argument cannot be invalidated by reference to universal scepticism. Other people's judgments provide independent and thus genuine checks on the (in)correctness of my actions whereas my own judgements provide no independent checks, that is, no checks tout court. (To be sure, they work perfectly well most of the time.) Now, it is perhaps imaginable that everybody might be mistaken or confused at a given moment. But it must be admitted at least that genuine checks are better than no checks (cf. Itkonen 1988a). Wittgenstein's theory of language is just a part, even if a central part, of his later philosophy. Its basic goal could be said to be the overcoming of the Cartesian tradition. Mental phenomena are shown to be inseparable from public actions, which in turn cannot be separated from the types of contexts in which they charac­ teristically occur (for an extended discussion, cf. GTM:4.0). A simple example, which also brings out Wittgenstein's affinity with Heidegger, may clarify the issue: As a phenomenon of human life, somebody smiling in a given situation is the primary fact. It is only as the result of an artificial analysis that the smile is first taken out of its context and then divided into the physical, geometrically measurable 'outside' and the mental 'inside'. More than anyone else, Wittgenstein is responsible for the 'linguistic' character of the 20th-century philosophy, either in its formal (or 'analytic') or in its non-formal (or 'ordinary-language') version. The same interest in language, or more generally in signification, is evident in the growth of semiotics, beginning in the early '70s. As welcome as this development is, there are at least two very real pitfalls threatening to devalue its output: first, overdrawing the analogy between language and other social domains, and thus seeing norms for form—meaning correlation where they do not exist in fact; second, mistaking effect for cause, and thus seeing signs as initiating, rather than expressing, social behavior. The latter mistake has been blatantly committed by the so-called poststructuralist philosophy (cf. Itkonen 1988c). Poststructuralism is ultimately motivated by the view that language determines thought in its entirety, a view that is flatly contradicted by contemporary psychological research (cf. Weiskrantz 1988). It has generally been the case that philosophy gives illumination to linguis­ tics, rather than vice versa, but in the present century the situation may have changed somewhat. Analytical philosophy displays a considerable lack of selfreflection: while it analyzes 'lower-level' conceptual systems, it rarely, if ever,

EUROPE

297

analyzes itself. In this respect it might be helped by linguistic metatheory, within which the notion of 'intuitional science', equally applicable to grammatical theory, analytical philosophy, and philosophical logic, has been developed in recent years (cf. GTM: 10.0-11.0, Katz 1981, CLT.3.7.3).

5.5.2 Autonomous Linguistics Paul (1975 [1880]:33) had already claimed that his entire theory of linguistic methodology concentrates on the question concerning the relation between the social and the individual aspects of language: "Es ergibt sich demnach, dass sich die ganze Prinzipienlehre der Sprachgeschichte um die Frage konzentriert: wie verhält sich der [soziale] Sprachusus zur individuellen Sprechtätigkeit?" In keeping with his methodological individualism, Paul regards the individual aspect as the primary one. A truthful description of a given state of language should take the psychologi­ cal grammars of all speakers into consideration. Unfortunately, this is not a practical possibility, and therefore one has to be content with describing the language of only a few speakers, or even of a single speaker (p. 29). Saussure (1962 [1916]) addresses the same question, but he answers it in a very different way. The social aspect (langue) is primary vis-à-vis the individual aspect (parole) and constitutes therefore the genuine subject matter of linguistics. There will be no scientific study of language until this truth has been grasped (pp. 24-25, 31). At the same time, it is enough to investigate the consciousness of individual speakers (i.e. their linguistic intuition), and conceivably of a single speaker, in order to elicit the properties of langue (p. 128). It is the particular merit of this assessment that it sanctions the practice of grammatical description which has been (and will be) followed always and everywhere. For instance, it is evident from the account given here in Chapters 2 and 4 that Pānini and Sïbawaihi investi­ gated the langue of their respective communities by means of their respective linguistic intuitions. Saussure uses the term langage to designate the superordinate concept which subsumes both langue and parole. Langage in turn issues from the language capacity (faculté de langage), which is 'natural', rather than social (like langue) or mental (like parole). Moreover, since parole is characterized as an 'act' (of speaking and understanding), it must contain the psycho-physiological perform­ ances, on the one hand, and the physical utterances, on the other. What seems to be lacking, is the equivalent of Paul's psychological grammar, or rather a clear distinction between it and langue qua social institution.

298

UNIVERSAL HISTORY OF LINGUISTICS

This last point becomes clearer when we examine Saussure's notion of 'linguistic sign'. It is characterized as a mental entity (entité psychique) consisting of a 'concept' and an 'acoustic image', also called 'signified' and 'signifier', respectively (pp. 98-99). Now langue is said to be 'constituted' by linguistic signs (p. 107). But this is a contradiction because langue is social whereas linguistic sign is individual-psychological. The same contradiction is apparent in the passages (p. 30, 32) where it is claimed that langue is situated 'in the brain'. In other words, what we have here is the contrast between the psychologistic conception of language and the social-normative one. Saussure was well on his way to overcome the former, but did not quite succeed. Within both components of the linguistic sign a further distinction is introduced, namely that between 'substance' and 'form' (pp. 155-158). The substance of the signifier is constituted by the whole spectrum of vocal sounds while the substance of the signified is something like 'thought minus conceptual distinctions'. By 'form' is meant the way in which each language structures the two substances. The notion of 'form' is further clarified with the aid of the notion of 'value' (pp. 158-166). The value, or identity, of 'concepts', on the one hand, and of 'acoustic images' (of words and individual sounds) on the other, is entirely determined by how they differ from other, comparable units: "dans la langue, il n'y a que des différences (p. 166). This calls for comments. First, the notion of 'linguistic sign' was anticipated by Bhartrhari and by Pseudo-Albertus Magnus (cf. p. 82 and p. 231), except that Saussure's version exhibits a gratuitous asymmetry unknown to his predecessors. If the substance of the signifier is something non-mental, the same ought to be true of the substance of the signified as well: it ought to be the extralinguistic reality 'before' a given language has structured it in such and such a way. By identifying the substance of the signified with 'amorphous thought' (which, unlike vocal sounds, is not an ontologically real entity), Saussure is committed to the views that there is no thinking over and above thinking expressed by oral languages, and that each (oral) language creates its own way of thinking. Both of these views are obviously false (see e.g. Weiskrantz 1988). Second, it is a very old idea that an entity A is determined by comparing it with  and C, i.e. with what it is not. Spinoza expressed this idea with his dictum Omnis determinatio est negado, and Hegel (1970 [1807]:95) formulated it as follows: Nämlich wenn die vielen bestimmten Eigenschaften ... sich durchaus nur auf sich selbst bezögen, so wären sie keine bestimmten; denn sie sind dies

EUROPE

299

nur, insofern sie sich unterscheiden und sich auf andere als entgegen­ gesetzte beziehen. If this is all that Saussure has in mind, then his thesis is correct, but not very exciting. But when he adds that the differences in langue are 'without any positive terms', he goes too far. Suppose I try to define A, B, and  'purely negatively', i.e. by what they are not. The result, apart from being extremely unwieldy, still contains two positive terms, and circularly:  and  are involved in the definitions of  and C:

Linguistic signs are situated simultaneously on two dimensions: associative and syntagmatic (pp. 170-175). It has become customary to replace 'associative' by 'paradigmatic', but this does not quite capture Saussure's meaning. He really intends to bring under this head everything that a given sign makes one think of, because of the similarity of either form or of meaning. Therefore his associative relations are in fact indistinguishable from Paul's Verbindungen zwischen Vorstel­ lungsgruppen, which, as we have noted, may be of varying strength (cf. p. 289). There is some uncertainty concerning the correct interpretation of the syntagmatic dimension. It is generally thought that Saussure assigned syntax to parole; accordingly, idiomatic expressions would be the largest syntagmatic entities belonging to langue. This view, however, is not supported by textual evidence. Saussure says quite clearly that sentences are formed on the basis of 'general and latent types' which belong to langue. The details of his description agree with Paul's account of 'syntactic analogy' (cf. here p. 289): ... il faut attribuer à la langue, non à la parole, tous les types de syntagmes construits sur des formes régulières... [par ex.] un mot comme indécorable... Il en est exactement de même des phrases [= sentences] et des groupes de mots établis sur des patrons réguliers; des combinaisons comme la terre tourne, que vous dit-il?.etc., répondent à des types généraux, qui ont à leur tour leur support dans la langue... (p. 173).

300

UNIVERSAL HISTORY OF LINGUISTICS

Notre mémoire tient en réserve tous les types de syntagmes plus ou moins complexes, de quelque espèce ou étendue qu'ils puissent être, et au moment de les employer, nous faisons intervenir les groupes associatifs pour fixer notre choix. ... Ce principe s'applique aux syntagmes et aux phrases [= sentences] de tous les types, même les plus complexes. Au moment où nous prononçons la phrase: 'que vous dit-il', nous faisons varier un élément dans un type syntagmatique latent, par example 'que te dit-il' — 'que nous dit-il', etc., ... (p. 179). Engler (1975:865-866, 870-871) confirms that in his last years Saussure came to regard syntax as part of langue. It is Saussure's (1962 [1916]) central thesis that every synchronic state of language constitutes a system, and must be treated as such. By contrast, he re­ garded linguistic changes as affecting single units only, which means that diachronic linguistics could not be a systematic discipline. In this he was obviously mistaken. Saussure's conception of synchronic linguistics can be summarized as the set of the following dichotomies: langue—parole, expression—content, form—sub­ stance, syntagmatic—paradigmatic. Hjelmslev (1963 [1943]) integrates these dichot­ omies into a coherent, closely knit theory, and adds three 'functions' between 'for­ mally' defined units: 'interdependence' (= if A, then B, and vice versa), 'determi­ nation' (= if A, then  or not-B, and if B, then A), and 'constellation' (= if A, then  or not-B, and vice versa). It was Hjelmslev's explicit purpose to base a universally valid descriptive technique, or universal grammar, on these three functions. Yet his approach, called 'glossematics', would seem to give rise to a rather impoverished view of language (cf. Itkonen 1968). Therefore it is interesting to note that Sampson (1980: Chap. 7) regards Lamb's 'stratificational grammar' as an implementation of Hjelmslev's central ideas. The Hjelmslev-type 'functions' also play an important role within the school of 'dependency grammar', which, created by Tesnière (1959), enjoys a great vogue on the Continent (cf. Korhonen 1977). Tesnière leaves his basic notion undefined. Yet, as shown by his examples, it is meant to apply to two paradigmatic cases: On the one hand, in an A + N construction the adjective 'depends on' the noun. On the other, in a sentence like John gave the book to Mary, the nouns John and the book as well as the prepositional phrase to Mary 'depend on' the verb gave. As far as I can see, this notion of dependence must be explicated as follows. In the first construction N is obligatory and A is optional. In the second construction all elements are equally obligatory. What makes the others 'depend on' the verb is the fact that in the formula 'If X is given, then Y, Z etc. are given', much more specific information results from replacing X by gave than by any other unit. The A + N

EUROPE

301

construction meets the criteria of Hjelmslev's 'determination': if a noun is present, an adjective may or may not be present, but if an adjective is present, a noun must be present. (Notice that 'not-B' means here the absence of, rather than an entity different from, B). The gave-construction would also seem to qualify as a 'determi­ nation': if (e.g.) to Mary is present, NV N (and in particular, gave) may or may not be present, but if gave is present, N N to N must be present. (Notice that 'not-B' means here the set of entities different from B, rather than the absence of B.) This, however, gives the wrong result insofar as gave turns out to 'depend on' to Mary. Therefore I prefer the characterization I first gave. Saussure and Hjelmslev received their training in the field of Indo-European linguistics. In the United States linguistic theorizing was influenced by the presence of Indian languages which exhibit a remarkable structural diversity. Therefore it is understandable that North American linguistics tended to emphasize the inadequacy of the traditional grammatical categories which had remained practically the same ever since Dionysius Thrax. This attitude (exhibited to some extent already by von der Gabelentz) is very much in display in Sapir (1962 [1921]). On more careful examination, however, it turns out that Sapir abandons neither the great Western tradition nor its universalist aspirations. Sapir's view of the correspondence between language and reality could have been taken directly from the Modistae: "We must have objects, actions, qualities to talk about, and these must have their corresponding symbols in independent words or in radical elements" (p. 93). The notion of sentence is the Aristotelian one: "[The sentence] combines a subject of discourse with a statement in regard to this subject. Subject and 'predicate' may be combined in a single word...; each may be expressed independently... ; each or either may be so qualified as to lead to complex propositions of many sorts" (p. 35). While 'subject' and 'predicate' are functional notions, the corresponding formal notions are 'noun' and 'verb' having, in con­ formity with the isomorphism thesis, 'thing' and 'action' as their extralinguistic correlates: The subject of discourse is a noun. As the most common subject of discourse is either a person or a thing, the noun clusters about concrete objects of that order. As the thing predicated of a subject is generally an activity ... the form which has been set aside for the business of predicat­ ing, in other words, the verb, clusters about concepts of activity. No language wholly fails to distinguish noun and verb,... (p. 119). The indispensability of subject and predicate is established by means of a 'reduction analysis' identical with the one already practised by Apollonius (p. 36;

302

UNIVERSAL HISTORY OF LINGUISTICS

cf. here p. 203). Like Paul and Saussure before him, Sapir argues that new sentences are created on the basis of abstract sentence-types and in analogy with previously created sentences: ... underlying the finished sentence is a living sentence type, of fixed formal characteristics. These fixed types or actual sentence-groundworks may be freely overlaid by such additional material as the speaker or writer cares to put on, but they are themselves rigidly 'given' by tradition as are the radical and grammatical elements abstracted from the finished word. New words may be consciously created from these fundamental elements on the analogy of old ones, but hardly new types of words. In the same way new sentences are being constantly created, but always on strictly traditional lines (p. 37; emphasis added; cf. also p. 85). The characteristics discussed so far are claimed to belong to all languages. Under this level of universal validity Sapir establishes a wide spectrum of linguistic diversity. But again the very notions which are employed to fix the diversity are meant to be universally valid. On the formal side Sapir speaks of 'grammatical processes' (Chap. IV), and on the semantic side he speaks of 'grammatical con­ cepts' (Chap. V). The former include word order, composition, affixation, internal modification, reduplication, and accentual differences (p. 61). The latter are first divided into two groups, namely 'material content' and 'relation'; an additional subdivision is then made within each, namely between 'basic' and 'derivational' concepts having material content, on the one hand, and between 'concrete' and 'pure' relational concepts, on the other (pp. 101-102). Basic concepts correspond, roughly, to lexical meanings while all the others exemplify grammatical meanings in the direction of increasing abstractness. Cross-linguistically, there is great freedom as to how 'concepts' are expressed by 'processes' (pp. 59-61). Yet it is universally true that in each language "analogous concepts and relations are most conveniently symbolized in analogous forms" (p. 38). Sapir pays a great deal of attention to the intuitive acceptability of gram­ matical description, and even speaks (p. 33) of the 'psychological reality' of descriptive units. Yet he makes it quite clear that as a scientific discipline, linguis­ tics remains distinct from psychology: We can profitably discuss the intention, the form, and the history of speech as we discuss the nature of any other phase of human culture — say art or religion — as an institutional or cultural entity, leaving the

EUROPE

303

organic and psychological mechanisms back of it as something to be taken for granted (p. 11). As Sapir's interests went far beyond the bounds of linguistics, he came to develop a full-fledged philosophy of human sciences. It is summarized in GTM: 61-68. Jespersen's (1924) approach to grammatical description bears an interesting resemblance to Sapir's insofar as he too distinguishes between the 'meaning-toform' and 'form-to-meaning' directions. Instead of such terms as 'grammatical concepts' and 'grammatical processes', however, he uses the rather ill-chosen terms 'syntax' and 'morphology'. To these two levels he adds a third one, namely that of 'notional categories', which are claimed to represent "the outside world, as re­ flected in the human mind" (p. 54). We seem to be here in the presence of the familiar trichotomy 'ontology—semantics/meaning—syntax/form'. For instance, in discussing English past-tense forms (pp. 56-57), Jespersen assigns to them the common meaning (or 'function') 'preterit' and, depending on the case at hand, such notions as 'past time', 'unreality in present time' (if we knew), 'future time' (it is time you went to bed), or 'all times' (men were deceivers ever). This is like arguing, in conformity with the Pāninian tradition, that every nominative in an active sentence has the semantic role 'agent', even if its ontological referent may or may not be an agent. On several occasions Jespersen discusses the possibility of universal grammar. He is clearly fascinated by the idea, but he is also afraid of repeating the mistakes of his predecessors, i.e. of making claims unsupported by cross-linguistic data. Yet he does believe, in common with the Aristotelian tradition, that the notional categories are universal (p. 55), and as a consequence, it is his conviction that, by carefully moving from notion via function to form, it is possible to discover the "great principles underlying the grammars of all languages" (p. 344). In his view of language acquisition and its end result Jespersen agrees with von der Gabelentz, Paul, Saussure, and Sapir: ... free expressions ... have to be created in each case anew by the speaker... The sentence he thus creates may, or may not, be different in some one or more respects from anything he has ever heard or uttered before; that is of no importance to our inquiry. What is essential is that in pronouncing it he conforms to a certain pattern. No matter what words he inserts, he builds up the sentence in the same way, and even without any special grammatical training we feel that the two sentences

304

UNIVERSAL HISTORY OF LINGUISTICS

John gave Mary the apple, My oncle lent the joiner five shillings are analogous, that is, they are made after the same pattern. ... Now, how do such [sentence] types come into existence in the mind of a speaker? ...from innumerable sentences heard and understood [the child] will abstract some notion of their structure which is definite enough to guide him in framing sentences of his own,... (p. 19; emphasis added). Bloomfield (1933) represents a type of North American linguistics rather different from Sapir's. In his self-consciously methodological statements, his behaviorism made him reject the study of meaning,23 but did not otherwise hamper his descriptive practice. Although an advocate of studying linguistic corpora, he himself gives only self-invented examples in his chapter on syntax. Besides, he clearly recognizes the open-ended nature of language; for him this fact is 'obvious' and in no need of special emphasis: "...it is obvious that most speech forms are regular, in the sense that the speaker who knows the constituents and the grammatical pattern, can utter them without ever having heard them; moreover, the observer cannot hope to list them, since the possibilities of combination are practically infinite. ... A grammatical pattern (sentence-type, construction, or substitution) is often called an analogy (Bloomfield 1933:275; the first emphasis added). As regards the possibility of universal grammar, Bloomfield adopts a cautious attitude, noting that the "only useful generalizations about language are inductive generalizations" (p. 20). In this he is quite correct. It is much more sensible to base claims about putative universals on several languages than on a single language (e.g., English). Making non-inductive generalizations was the basic mistake committed by the Modistae. In his morphology Bloomfield operates with 'underlying forms' (p. 218) and with descriptive processes applied 'in a certain order' (p. 222). This technique became more pronounced in his later writings. As noted here on p. 69, it is possible that he was in this respect influenced by Pānini. Harris continues the Bloomfieldean emphasis on linguistic form, more precisely on "regularities in the distributional relations among certain features of speech" (1951:5). He notes explicitly that although, in his opinion, the grammarian must start from a corpus, his description necessarily goes beyond it:

EUROPE

305

The interest in our analysis of the corpus derives primarily from the fact that it can serve as a predictive sample of the language (p. 244; emphasis added). When a linguist offers his results as a system representing the language as a whole, he is predicting that the elements set up for his corpus will satisfy all other bits of talking in that language (p. 17; emphasis added). It is in this, and only in this, context that the following often-quoted statement must be understood: "The over-all purpose of work in descriptive linguistics is to obtain a compact one-one representation of the stock of utterances in the corpus" (p. 366). Notice also that since the Harris-type description makes predictions, it must be testable, and liable to disconfirmation, just like any other scientific theory. The main theoretical notion employed by Harris (1946) is 'frame of substitution', and the considerations which lead to the selection of one frame of substitution over others refer to the over-all simplicity of the total description: The criterion which decides for -ing, and against un-, as the relevant environment in determining substitution classes [for verbs] is therefor a criterion of usefulness throughout the grammar, a configurational consider­ ation (p. 164, n.6). In his early work Chomsky showed that methods developed for the study of formal languages were applicable to natural languages as well. It seemed possible to situate any kind of language in an overarching hierarchy according to the type of grammar suitable for it. Natural-language grammars were conceived of as axiomatic systems, phrase-structure rules and transformation rules functioning as rules of inference meant to generate all and only correct sentences of a given language starting from the axiomatic S-symbol. Although in Chomsky's later work the emphasis on grammatical rules has been replaced by that on constraints and principles that are meant to jointly entail the (in)correctness of all and only ( i n ­ correct sentences, establishing higher standards of formal description remains his lasting achievement. In his dissertation Chomsky is still a follower of Bloomfield and Harris. He defends the former's antimentalist program against suggestions that the criteria of scientific significance should be relaxed so as to admit non-physical entities as well: The fact that a certain general criterion of significance has been abandoned does not mean that the bars are down and that 'ideas' and 'meanings'

306

UNIVERSAL HISTORY OF LINGUISTICS

become proper terms for linguistics. If this rejection of an old criterion ... is followed by a new analysis of 'significance', then if this is at all ad­ equate, it seems to me that it will rule out mentalism for what were essentially Bloomfield's reasons, i.e., its obscurity and inherent untestability (1975 [1955]:86; cf. here n. 9, 23). It is Chomsky's goal to achieve a synthesis of Bloomfield's antimentalism and Harris's distributionalism at the level of syntax: The notions that enter into linguistic theory are those concerned with the physical properties of utterances, the formal arrangement of parts of utterances, ...and finally, formal properties of systems of representation and of grammars. ... We will refer to linguistic analysis carried out in these terms as 'distributional analysis' (p. 127), this term [being] borrowed from Harris (p. 63, n.1). Chomsky (1957:55) also echoes Harris in his claim that "the only ultimate criterion in evaluation is the simplicity of the whole system" (cf. here p. 305). So we see that, contrary to a wide-spread misconception, in its first explicit formu­ lation Chomsky's linguistic theory did not embody any significant break with the preceding tradition. On this issue the level of historiographical writing has been very low. In Chomsky (1965) the study of linguistic form, now supplied with a fullfledged mentalistic interpretation, is seen as a study of innate universal grammar. The assumption of linguistic universals is not based on data, but rather on concep­ tual arguments (cf. below). In Chomsky (1980) the innate linguistic faculty (or 'organ') is conceived of in ultimately biological terms, and its autonomy vis-à-vis other comparable faculties is justified by a general appeal to the 'modularity' of mind. In the present context it is of no importance that Chomsky's theory of syntax has undergone several modifications. What is interesting, is the fact that while he has all the time continued to analyze the syntax of English by means of self-invented sentences which his own linguistic intuition deems correct, his interpretation of, and justification for, what he is doing has changed completely: he first practises antimentalist (or physicalist) distributional analysis, then mentalist syntax as part of general psychology, and finally study of universal grammar as one aspect of biological endowment. It is not possible, however, to move from one science to another, and from this to a third one, while doing exactly the same thing. Therefore Chomsky is a 'psychologist' or a 'biologist' only with quotes. In reality

EUROPE

307

he remains within the tradition of autonomous linguistics. If mind is taken to rest on a neurophysiological foundation, and language is declared to be a module of mind, and syntax is declared to be the central module of language, then it might seem that one has acquired the right to study English syntax and call it either psychology or neurophysiology. This chain of reasoning is not cogent, however. Alternative, clearly dissimilar theories of syntax are available, and they have an equal claim to psychological reality. At present the question as to the relation between grammatical theory and psychology remains wide open. More precisely, the notion of analogy, buttressed by recent work on auditory phonetics and neuro­ physiology,24 offers a viable alternative to modularity (cf. 5.5.3). Innate universal grammar is invoked by Chomsky to explain the 'logical problem of language acquisition', or the 'fact' that children acquire their first language so rapidly and on the basis of such a limited and degenerate data. But this is far from being a fact, because the central notions are either undefined or fal­ lacious. Compared with what isfirst-languageacquisition supposed to be rapid? With second-language acquisition by adults? But this is irrelevant. With visual perception, e.g. with learning to distinguish different shapes? But then language acquisition is not rapid. Moreover, the data is not limited because children hear thousands and thousands of utterances. Nor is it degenerate, as has been shown by sociolinguistic studies. (Besides, there is characteristic behaviour accompanying slips of the tongue, on the part both of speakers and of hearers, which may be taken to imply that incorrect utterances come labelled as incorrect.) An even stronger argument in favor of innate universal grammar has been advanced, namely language acquisition on the basis of no data. This argument, however, suffers from subjectivity. It is repeatedly claimed that it is 'hard to imagine' or 'not plausible' that children should be able to make such and such generalizations or to perceive such and such analogical relationships. But the majority of linguists have found such processes quite easy to imagine. What this suggests, is that the logical problem of language-acquisition must be replaced by the corresponding empirical problem. Instead of just imagining what children might or might not do, one must observe what they do in fact. And observational accounts like Bates (1979) clearly indicate that children abstract general (analogical) prin­ ciples from particular instances of use-in-context. Chomsky (1975 [1955]:85) was still optimistic enough to believe that "linguistic theory may actually provide a literal discovery procedure". He soon gave up this hope, however, and has since then regarded the lack of discovery pro­ cedures as entailing the vacuity of "such ideas as analogy, induction, association, reliable procedures, good reasons, or justification in any generally useful sense"

308

UNIVERSAL HISTORY OF LINGUISTICS

(1986:12). What this amounts to, is just recognition of the fact that so far no algorithm has been devised for deriving grammars from corpora. But the impli­ cations of this fact are not those envisaged by Chomsky. If it is sufficient to describe language acquisition, without any algorithm, as a process of non-analogical data triggering specific innate capacity, then no algorithm is needed to substantiate the alternative view of analogical data triggering non-specific innate capacity.25 On the other hand, the lack of an algorithm is predictable, if language acquisition is taken to be a creative, or at least flexible process which is, within the limits set by biological maturation, impelled by unconscious rationality (cf.